Genetics of Tomato Spotted Wilt Virus Resistance in Peanut (Arachis hypogaea L.)

baldessari_j.pdf

CHAPTER 3
ARTIFICIAL INOCULATION STUDIES INT THE TSWV-PEANUT PATHOSYSTEM

Introduction

Tomato spotted wilt virus (Bunyaviridae: Tospovirus) is a species of quasi-spherical

enveloped particles containing three single-stranded RNA (Kormelink et al., 1992). It is a

cosmopolitan pathogen of 1090 plant species in 85 families (Parrella et al., 2003). It is vectored

in a propagative and circulative manner exclusively by thrips of the genera Thrips and

Frankliniella (Ullman et al., 2002). Wherever TSWV incidence has increased enough to cause

economic losses, it has remained a chronic problem in many economically important crops. This

is the case in the Southeastern USA, where the peanut (Arachis hypogaea L.) crop has suffered

intermittent heavy losses since 1993. The initial increasing trend in TSWV incidence has now

been reversed (Brown et al., 2007) by a combination of factors including planting date, stand

density, row pattern, insecticide use and resistant cultivars (Culbreath et al., 2003). The

resistance level of the cultivar is the most important tool in the management of this disease

(Brown et al., 2007). So far most resistance assessments have been conducted in the field and

relied on natural epidemics. This means that a valid characterization of each genotype has been

resource-consuming involving several seasons and locations. As an alternative approach,

artificial inoculation methods have been developed to reduce the time needed to assess a

genotype targeting its commercial release as a cultivar and also to diminish the overall resource

requirements .

Several artificial inoculation methods have been described for the TSWV-peanut

pathosystem (Halliwell and Philley, 1974; Clemente et al., 1990; Pereira, 1993; Hoffman et al.,

1998). According to published results the most consistent one was developed by Mandal and

coworkers (2001). In this method important factors are the antioxidants in the extraction buffer












Table 2-4. Heritability (S.E.) estimates for stunting and foliar symptoms caused by TSWV on
peanut populations from fiye crosses at different assessment dates in five tests in
Floirida. Estimates were calculated using univariate Animal Models.
Test Variable 30 DAP 60 DAP 120 DAP
Citra 2005 stunting 0.59 (0.05) 0.59 (0.05) 0.71 (0.04)
foliar symptoms 0.65 (0.04) 0.43 (0.08) 0.28 (0.11)


Citra 2006


Marianna 2006


Marianna 2007


stunting
foliar symptoms

stunting
foliar symptoms

stunting
foliar symptoms


0.02 (0.01)
0.02 (0.01)

0.03 (0.01)
0.04 (0.01)

0.07 (0.01)
0.08 (0.02)

0.05 (0.02)
0.06 (0.02)


0.02 (0.01)
0.03 (0.01)

0.09 (0.02)
0.12 (0.03)

0.26 (0.03)
0.28 (0.03)

0.28(0.03)
0.31(0.03)


a
a

b
b

0.01 (0.01)
0.01 (0.01)


Quincy 2007 stunting
foliar symptoms
a: incidence too low to allow analysis.


b
b
b: date not assessed.


Table 2-5. Phenotypic and genetic correlation (S.E.) estimates between stunting and foliar
symptoms caused by TSWV on peanut populations from fiye crosses at different
assessment dates in five tests in Florida. Estimates were calculated using univariate
Animal Models.


30 DAP
0.80 (0.02) / 0.88 (0.03)


60 DAP
0.82 (0.02) / 0.99 (0.01)


120 DAP
0.83 / 0.99 e


Citra 2005

Citra 2006


0.91 (0.1) /0.95 (0.03)

0.84 (0.01) /0.95 (0.02)

0.92 (0.01) /0.99 (0.0.1)


Marianna 2006

Marianna 2007


0.93 (0.01) /0.95 (0.03)


0.90 / 0.99 e


Quincy 2007 c 0.91 / 0.99 e 0.92 (0.01) / 0.99 (0.01)
a: incidence too low to allow analysis. b: Singularity in datamatrix c: date not assessed. d:
parameters didn't converge. e: Standard Error not available because REML estimate was bounded.





















AP-3 / NemaTAM


NemaTAM / AP-3

















012345
SCmre


6
r:
a,
rr
e
c 0.19-
ar
C
m
a,
u
0.10-


012345
SCmre


v0n 034


.2-NemaTAM / DP-1 02-DP-1 / NC94002




S0.19- 0.17-




0.10- .9





0.00 0.00
0 12 34 5 0 12 34 5
score score




Figure 2-8. Frequency distributions for TSWV-induced stunting scores among F2 populations from four peanut crosses field tested at

Marianna, Florida in 2007.









and the type and amount of abrasives used in the rubbing. Oxidation, which reduces the life of

TSWV outside the cell, seems to be a common remark among the bibliographical sources and

thus different concentrations of several reducing chemicals have been evaluated (Halliwell and

Philley, 1974; Clemente et al., 1990; Pereira, 1993; Mandal et al., 2001). With reference to

abrasives, Clemente et al. (1990) found no differences in the rates of TSWV artificial inoculation

using various grit sizes of Carborundum whereas Mandal and coworkers (2001) reported that the

type of added abrasive in the inoculum was very important. Some other factors have been cited

in the literature as influencing the outcome on inoculation experiments. Ng et al. (2004) reported

that the concentration of virus that effectively reached the target tissues was very important in

the transmission efficiency of Lettuce infectious yellows closterovirus to lettuce. Even with

standardized conditions, different factors can produce varied actual damage after inoculation,

which could lead to an inconsistent number of viral particles entering the leaf (Pereira, 1993).

Consequently a better assessment of the actual damage inflicted while rubbing could be

extremely useful in ruling out this factor as a cause of variability.

A fact that has hampered the resistance assessment in the TSWV-peanut pathosystem is

that external symptoms do not always reflect the concentration of TSWV in the plant (Resende et

al., 2000; Mandal et al., 2001; Lyerly et al., 2007). Additionally, asymptomatic TSWV infections

in peanut have been reported (Culbreath et al., 1992). Nonetheless, the frequency with which

plants within a cultivar express symptoms early after inoculation has been suggested as a good

indicator of viral titer in the tissue (Kresta et al., 1995) and also of resistance (Culbreath at al.,

2003).

While an artificial inoculation method should accelerate the screening process leading to

the detection of resistant genotypes, it should also reduce the environmental variation of plant










Test 2: Determining the importance of amount of rubbing on infection rate

Leaves were rubbed four, six or eight times using a cotton swab. Once prepared the

inoculum was used immediately. Ten plants were inoculated per treatment using the same

inoculum batch and these 30 plants (10x3) were considered a block within the RCBD. There

were four blocks totaling 120 plants. As controls, eight plants were rubbed with buffer only plus

abrasives and eight more plants were left untreated.

Test 3: Evaluating the influence of inoculum concentration on infection rate

Two inocula, each using a different tissue:buffer ratio, were compared (1:10 and 1:20).

After obtaining the usual 1:10 inoculum, half of it was allocated in another mortar and a similar

volume of buffer was added. Once prepared both inocula were used immediately. Different

swabs were used for each level of the dilution factor. Ten plants were inoculated per treatment

and these 20 plants (10x2) were considered a block within a RCBD. There were four blocks

totaling 80 plants. As controls, eight plants were rubbed with buffer only plus abrasives and eight

more plants were left untreated.

Imposing Treatments

In Test 1 plants were inoculated by rubbing them four times with a cotton swab. The time

to inoculate each plant was about ten seconds so for every treatment the real inoculation time

between the first and the last plant was almost two minutes. In the Test 2 plants were inoculated

according to the layout L1-L2-L3-L3-L2-L1-L 1-L2-L3 -L3 -L2-L 1 (L means "Factor Level")and

so on. In Test 3 plants were inoculated according to the layout L1-L2-L2-L1-L1 -L2-L2-L1 and

so on.





-o F2 NemaTAM / DP-1
-m-- F3 NemaTAM / DP-1
F2 DP-1 / NC94002
F3 DP-1 / NC94002
-a- NC94002
-*- NemaTAM
--DP-1


30DAP


60DAP


120DAP


Assessment date








Figure 2-6. Observed stunting severity among different populations from five peanut crosses, at three different dates at Marianna,
Florida in 2007.









Coinciding with published research, resistance differences among genotypes were most

noticeable under the severe epidemics in Marianna (Culbreath et al., 1997; Tillman et al., 2007).

The observed resistance grouping was in good agreement with the information provided by the

present version of the Peanut Disease Risk Index (Brown et al, 2007).

Murakami et al. (2006) reported that planting dates diverged in spotted wilt incidence as

the season progressed under a mild epidemic but were similar under a severe one. However, in

the present study rather the opposite was true.

Under a mild epidemic in Citra, differences in spotted wilt intensity between PDs were

easily observed early in the season (PD2 vs. PD4 at AD70), but they tended to diminish as the

season advanced. Meanwhile, in Marianna the effect of PD was not important until late in the

season (132 AD) where all the PDs differed in the intensity DIRs. Curiously, the spotted wilt

intensity spiked at the third PD which coincides with the extension recommendation for

plantings with reduced spotted wilt risk (Brown et al., 2007). This further underscores the

seasonal unpredictability of spotted wilt epidemics and the need to plant resistant varieties. For

example, Tillman and coworkers (2007) reported that June plantings were less conducive to

severe spotted wilt in most of the genotypes tested in the Florida Panhandle. In the present study,

however, early May planting seemed best for reduction of spotted wilt damage.

Differing from results reported by Hurt et al. (2005), in the present study the interactions

among genotypes, location and planting dates were not important. This was so even when all the

three factors were j ointly analyzed, which highlights the advantages of modeling variance and

error structures (Gilmour et al., 2006).

Phenotypic and genotypic correlations are of the same sign and similar magnitude most of

the time (Lynch and Walsh, 1998). This seemed to be the case in the present study. The fact that













Table 2-3. REML variance estimates for stunting and foliar symptoms caused by TSWV in populations derived from five peanut
crosses tested at Citra, Florida in 2005 and 2006, Marianna, Florida in 2006 and 2007 and Quincy, Florida in 2007.


Stunting


Foliar symptoms


Citra 2005


G52Block ~2Plot

a ~0.040

a ~0.040

a ~0.042


O2A

2.539

2.539

2.827



c

0.021

0.036



b

0.087

0.305



0.013

0.235

0.698


O2NA O2e

1.699

a 1.699


(2Block 2SPlot

a ~0.155

a ~0.198

a ~0.278


O2A

4.096

2.315

1.048



c

0.030

0.065



b

0.133

0.358



0.017

0.320

0.790


(2NA

a

a

a


(2e

2.050

2.889

2.455



c

1.423

2.110



b

3.403

2.298



1.012

3.578

1.783


Assessment date

30 DAP

60 DAP

120 DAP


4

Citra 200


1.11



c

1.07

1.53



B

2.77

2.64



0.94

2.95

1.73


30 DAP

tj 60 DAP

120 DAP



30 DAP

60 DAP

120 DAP



30 DAP

60 DAP

120 DAP


c

0.001

0



b

0

0.004



0

0.015

0.095


c

0.017

0.054



b

0.106

0.382



0.014

0.102

0.250


0 0.002

9 0.000

Marianna 2006



4 0

10.012

Marianna 2007

6 0.001

5 0.034

9 0.056


c

0.018

0.083



b

0.149

0.358



0.011

0.140

0.211










found in the literature even though the actual damage inflicted to the leaf directly influences the

number of virions that reach the target parenchymatous tissue (Ng et al., 2004). Damage also

impacts the degree and speed of development of necrosis caused by abrasion (Hoffman et al.,

1998). No necrosis developed after inoculation in the abrasion area so apparently the number of

rubbings applied was rather mild. Peanut is known for its strong load of waxes in its leaves

(Samdur et al., 2003) so it' s possible that the abrasion didn't always provide enough injury to

serve as entry points for the virions.

As in Test 1, the factor Inoculum Batches was statistically significant. It is evident that

even pooling tissue with similar symptoms was not enough to avoid important variability in the

infectious capacity of the inoculum batches.

Test 3

The percentage of infected plants in this test was greater than in the previous two tests and

approached the levels reported by other authors (Hoffman et al., 1998; Mandal et al., 2001).

Surprisingly, no symptomatic plant was observed. Lyerly et al. (2002) reported that there were

instances where peanut plants infected with TSWV exhibited no visible symptoms and some

plants even recovered from initial infection and appeared normal. Krista et al. (1995) found that

peanut leaves with very low ELISA titer exhibited extremely mild symptoms. However, the

ELISA titers obtained in this test are in the usual range obtained while using similar conditions

(S. Mullis, pers. comm.) and overall are similar to those obtained in Tests 1 and 2 described in

this chapter (data not shown). Mandal et al. (2001) observed a delay in symptom expression

associated with plant age. Since age is associated with plant size, it could be possible that the

older plants used in the present test were displaying this kind of delay, when compared to the

smaller ones used in test 1 and 2.






















III Marianna '07 F2

0 Marianna '07 F3

5 Quincy '07 F2

o Quincy '07 F3


AP-3 /
NemaTA M


NemaTAM / NemaTAM / NemaTAM /
AP-3 DP-1 NC94002


DP-1/
NC94002


Figure 2-16. Percentage of individuals and (number ofF3 families) displaying BLUPs for TSWV-induced stunting above their best
parent, in each of five peanut crosses tested at Marianna and Quincy, Florida in 2007.










Table 1-10. Type B genetic correlations and [standard errors] for transformed spotted wilt
disease intensity ratings at four planting dates assessed in different times in Citra and
Marianna, Florida in 2005.
Coefficients obtained from Univariate Analysis
Location Assessment Date Genetic Correlation
Citra 70 0.91[0.17]
90 0.83 [0.32]
112 0.95 [0.11]
132 1[0.01]
Marianna 90 0.91[0.08]
112 0.97[0.03]
132 0.87[0.09]
Coefficients obtained from Bivariate Analysis
Planting Assessment Date
Date 90 112 132
2 1[0.09]
3 0.71[0.22] 1[0.13] 1[0.25]
4 0.73[0.18]
Coefficients obtained from Multivariate Analysis
Assessment Date
112
Cell Citra-PD4 Marianna-PD3 Marianna-PD4
Citra-PD3 0 1 [0.11]b 1 [0.13]b 1 [0.09]b
Citra-PD4 0.91 [0.12] 0.73 [0.18]
Marianna-PD3 0.94 [0.08]
AD132
Cell Citra-PD3 Marianna-PD2 Marianna-PD3
Citra-PD2 1 (0.24)b 1 [0.08]b 1 [0.07]b
Citra-PD3 1 [0.20]b 1 [0.20]b
Marianna-PD2 1 [0.04]b
Correlations were calculated among planting dates (across locations), between locations (across
planting dates) and among cells (location by planting date combination) by applying different mixed
models. b Variance components and genetic correlations were kept in the theoretical range by
constraining the covariance matrix (Gilmour et al. 2006). PD: Planting Date.


Table 1-11. Entry-mean repeatability estimates and their [standard errors] for transformed
spotted wilt intensity ratings at three planting dates assessed three times in Citra and
Marianna, Florida in 2005. Values for each location are separated by a slash, Citra
being on the left and Marianna on the right.
Assessment Date (days after pIlanting~)
Planting Date 90 112 132
2 0.81 [0.11] /0.94 [0.04]
3 0.82 [0.10] /0.94 [0.04] 0.76 [0.14] /0.86 [0.08] 0.57 [0.26] / 0.94 [0.04]
4 0.90 [0.06] / 0.98 [0.01]










Culbreath, A. K., J.W. Todd, D.W. Gorbet, F.M. Shokes, and H.R. Pappu. 1997. Field response
of new peanut cultivar UF 91 108 to Tomato spotted wilt virus. Plant Dis. 81(12): 1410-1415.

Culbreath AK, J.W. Todd, D.W. Gorbet, S.L. Brown, J.A. Baldwin, H.R. Pappu, and F.M.
Shokes. 2000. Reaction of peanut cultivars to spotted wilt. Peanut Sci. 27(1):3 5-39

Culbreath, A.K., J.W. Todd, and S.L. Brown. 2003. Epidemiology and management of Tomato
spotted wilt virus in peanut. Annu. Rev. Phytopathol. 41:53-75.

Culbreath, A.K., D.W. Gorbet, N. Martinez-Ochoa, C.C. Holbrook, J.W. Todd, T.G. Isleib, and
B. Tillman. 2005. High levels of field resistance to Tomato spotted wilt virus in peanut
breeding lines derived from hypogaea and hirsuta botanical varieties. Peanut Science
32(1):20-24.

Demsky, J.W., and D.R. Reddy. 2004. Diseases caused by viruses. In: Compendium of peanut
diseases, 2nd Ed. N. Kokalis-Burelle, D. M. Porter, R. Rodriguez-Kabana, D. H. Smith, and
P. Subrahmanyam. APS Press. Saint Paul, MN.

Develey-Riviere, MP, and E. Galiana. 2007. Resistance to pathogens and host developmental
stage: a multifaceted relationship within the plant kingdom. New Phytologist 175(3):405-
416.

Eisen, E. J. and A. M. Saxton. 1983. Genotype by environment interactions and genetic
correlations involving two environmental factors. Theoretical Applied Genetics 67(1):75-86.

Falconer, D.S., and T.F.C. Mackay. 1996. Introduction to Quantitative Genetics. 4th Edn.
Longman, Essex, England

Finne, M.A., O.A. Rognli, and I. Schjelderup. 2000. Genetic variation in a Norwegian
germplasm collection of white clover (Trifolium repens L.) 2. Genotypic variation,
heritability and phenotypic stability. Euphytica 112(1):45-56

German, T.L., D.E. Ullman, and J.W. Moyer. 1992. Tospoviruses: Diagnosis, molecular biology,
phylogeny and vector relationships. Annual Rev. Phytopathol. 30:315-348

Gilmour, A.R., B.J. Gogel, B.R. Cullis, and R. Thompson. 2006. ASReml User Guide Release
2.0. VSN International Ltd, Hemel Hempstead, HP1 1ES, UK

Gilmour, A.R., B.R. Cullis, S.J. Welham, B.J. Gogel, and R. Thompson. 2004. An efficient
computing strategy for prediction in mixed linear models. Comput. Stat. and Data Analysis
44: 571-586.

Gorbet, D.W. 1999. University of Florida peanut breeding program. Proceedings Soil and Crop
Science Society of Florida 58:2-7.

Gorbet, D.W., 2003. New University of Florida peanut varieties for 2003. UF/IFAS Agric. Exp.
Stn. Marianna NFREC Research Report 03-2.










LIST OF REFERENCES


Agresti, A. 1996. An Introduction to Categorical Data Analysis. Wiley-Interscience. New York,
NY.

Banks, B.D., I.L. Mao, and J.P. Walter. 1985. Robustness of the restricted maximum likelihood
estimator derived under normality as applied to data with skewed distributions. J. Dairy Sci.
68: 1785-1792.

Betran, F.J., S. Bhatnagar, T. Isakeit, G. Odvody, and K. Mayfield. 2006. Aflatoxin
accumulation and associated traits in QPM maize inbreds and their testcrosses. Euphytica
152(2):247-257.

Branch, W.D. 1996. Registration of 'Georgia Green' peanut. Crop Sci. 36(3):806.

Branch, W.D. 2002. Registration of 'Georgia-01R' Peanut. Crop Sci. 42(6): 1750-1751

Branch, W.D. 2003. Registration of 'Georgia-02C' Peanut. Crop Sci. 43(5): 1883-1884.

Branch, W.D., T.B. Brenneman, and A.K. Culbreath. 2003. Tomato spotted wilt virus resistance
among high and normal O/L ratio peanut cultivars with and without irrigation. Crop Prot.
22:141-145.

Brown, S.L., J.W. Todd, A.K. Culbreath, J. Beasley, B. Kemerait, E. Prostko, T. Brenneman, N.
Smith, D. Gorbet, B. Tillman, R. Weeks, A. Hagan, W. Faircloth, D. Rowland and R.
Pittman 2007. Minimizing Spotted Wilt of Peanut including the 2007 Version of the Tomato
Spotted Wilt Risk Index. http ://www.tomatospottedwiltinfo. org/peanut/ri skindex.htm. (last
accessed January 2007)

Bruening, G. 2006. Resistance to infection. In: Natural Resistance Mechanisms of Plants to
Viruses. Chpt. 10. Ed. G. Loebenstein and J.P. Carr, 2006. Ed. Springer. Dordrecht,
Germany.

Canady, M.A., M.R. Stevens, M.S. Barineau, and J.W. Scott. 2001. Tomato spotted wilt virus
(TSWV) resistance in tomato derived from Lycopersicon chilense Dun. LA 1938. Euphytica
117(1):19-25

Clemente, T.E., A.K. Weissinger, and M.K. Beute. 1990. Mechanical inoculation of Tomato
spotted wilt virus on peanut. Proc. Am. Peanut Res. Ed. Soc. 22:27 (Stone Mountain, GA)

Connover, J.W. 1998. Practical Nonparametric Statistics. 3rd. ed. John Wiley & Sons, NY.

Culbreath, A.K., J.W. Todd, and J.W. Demski. 1992. Comparison of hidden and apparent spotted
wilt epidemics in peanut. Proc. Am. Peanut Res. Ed. Soc. 24:39 (Norfolk, VA).

Culbreath, A.K., J.W. Todd, D.W. Gorbet, W.D. Branch, R.K. Sprenkel, F.M. Shokes, and J.W.
Demski. 1996. Disease progress of Tomato spotted wilt virus in selected peanut cultivars and
advanced breeding lines. Plant Dis. 80(1):70-73









Planting Date Effect

i) Univariate analysis for planting dates

The importance of planting date as a factor determining the DIR was variable according to

the location. In Citra it ranged from highly significant (p=0.006) to slightly non-significant

(p=0.054) depending on the genotypes and planting dates included in each analysis (Table 1-5).

Earlier AD (70 & 90) showed greater significant differences than later ones (112 & 132),

although different PDs were compared in each analysis. At 70AD and 132AD, the later PD

exhibited higher predicted DIR (Table 1-6), while at 90AD the opposite was true.

In Marianna, planting dates were statistically different only at 132AD, with the PDs

ranking 3>4>2 in DIR.

The ratios of planting date by genotype interaction variance to the biggest error variance

were most of the time close to zero (Table 1-5). The only exception was Marianna at 132AD

where the ratio was 0.8, yet small compared to the genotypic variance. In comparison, Genotypes

seemed far more important as a variability source with the ratios of genotypic variance to the

biggest error variance ranging from 0.4 to 1.7 at Citra and from 2.4 to 4.4 at Marianna. In order

to compare ADs, when this ratio was calculated only from PD 2 and 3, its value went from 1.5 at

AD90 to 3.3 at AD 112 (data not shown), suggesting increasing variability over time.

In Citra, the variance component for the PD by block interaction varied from small, 1/10 of

the smallest error variance, at AD 112 to negligible at the other three ADs (data not shown).

Meanwhile, in Marianna it ranged from rather small, 1/4 of the smallest error variance, to

important (1.5 times the smallest error variance).

The ratio of error variances among PDs was quite similar among ADs at Citra, ranging

from 1 to 1.4 while at Marianna they varied widely from 1.5 to 7.6.




















































9825
3.21 (1.53)
3.66 (1.59)
0 (0.07)
0 (0.02)
0.04 (0.34)


Test Plots in each generation Design and block Sowing
Parents F1 F2 F3b BC number d date
CR w/variable
5/30
Citra '05 13" 1 2-5 NA NA replications
Citra '06 4-8 1-2 1-9 2 NA RCB, 2 blocks 5/24
Marianna '06 6-12 1 6d 2-3 NA RCB, 3 blocks 4/28
Marianna '07 6-12 1-2 9 2 0-2 RCB, 3 blocks 4/24
Quincy '07 6-12 0-2 9 2 0-2 RCB, 3 blocks 4/25
Plot number in each generation varied depending on seed availability. b All F2 3 familieS changed
from year to year and some changed from test to test within years. Only AP-3 grown. d F2 DP-1 /
NC94002 not grown


a


Table 2-2. Mean (S.D.) score for each spotted wilt symptom at 30, 60 and 120 days after
planting, at each of five field tests in which five peanut populations were evaluated in
Florida.


30 Days After Planting
1073 N/A 10713 10669 N/A
0.66 (1.58) N/A 0.17 (0.79) 0.32 (0.99) N/A
1.15 (1.92) N/A 0.17 (0.69) 0.33 (1.02) N/A
0.08 (0.54) N/A 0 (0) 0.01 (0.14) N/A
0.08 (0.51) N/A 0 (0.02) 0 (0.07) N/A
0 (0) N/A 0 (0) 0 (0) N/A
60 Days After Planting
1073 3867 10646 10538 10032
0.69 (1.63) 0.36 (1.06) 0.99 (1.73) 1.27 (1.81) 1.24 (1.91)
1.24 (1.98) 0.46 (1.22) 1.24 (1.92) 0.15 (2.01) 1.43 (2.13)
0.09 (0.58) 0.01 (0.15) 0.01 (0.16) 0.02 (0.2) 0.09 (0.47)
0.09 (0.54) 0 (0.03) 0.01 (0.13) 0 (0.1) 0 (0.08)
0 (0) 0.01 (0.16) 0.00 (0.04) 0.10 (0.5) 0.06 (0.39)


Table 2-1. Sowing date, replication number and design of tests assessing performance against
spotted wilt in five peanut crosses in Florida.


2005
Citra


2006


2007


Citra


Marianna Marianna


Quincy


Stunting
Spots and Mosaic
Tip Death
Leaf Necrosi s
Yellowing


Stunting
Spots and Mosaic
Tip Death
Leaf Necrosi s
Yellowing


Stunting
Spots and Mosaic
Tip Death
Leaf Necrosi s
Yellowing
N/A: Not Available


120 Days After Planting
3809 10246 10028
0.56 (1.29) 2.34 (1.85) 3.27 (1.69)
0.75 (1.51) 3.13 (1.76) 3.63 (1.7)
0.01 (0.15) 0.01 (0.11) 0.03 (0.27)
0 (0.03) 0 (0.03) 0 (0.02)
0.01 (0.12) 0.04 (0.12) 0.05 (0.39)


1071
0.68 (1.48)
1.23 (1.77)
0.07 (0.49)
0.03 (0.28)
0 (0)









Inoculum Preparation

Infected leaf tissue from greenhouse grown peanut plants (cv. Georgia Green) were

collected and pre-chilled in a refrigerator and ground (1:10 [wt/vol] tissue:buffer unless

otherwise stated) with freshly prepared ice-cold 0.01 M potassium phosphate buffer, pH 7.0,

containing 0.2% sodium sulfite and 0.01 M 2-mercaptoethanol using a chilled pestle and mortar

as described by Mandal et al. (2001). Debris was removed by squeezing the ground extract

through a pad of nonabsorbent cotton. To this homogenate, Celite 545 (Fisher Scientific, Fair

Lawn, NJ) and Carborundum 320 grit (Fisher Scientific) were each added to a final

concentration of 1% each. The inoculum was kept on ice until the inoculation process was

completed.

Sap Inoculation

Test plants were dusted with Carborundum on the youngest fully expanded leaf 12-14 days

after sowing. Two leaflets (one basal and one apical) were inoculated by rubbing them four times

(unless otherwise stated) with a cotton swab (Johnson & Johnson, Skillman, NJ) dipped in the

inoculum. After inoculation the plants were sprayed with distilled water and placed in a growth

room at 25/190C, 50% RH, 12-h light period and 15 klx of light intensity and were irrigated

every two days using distilled water.

Description of Tests

Test 1: Effect of elapsed time from preparation to inoculation on infection frequency

Three time lapses from the inoculum preparation (zero, ten and twenty minutes) were

compared. At each time, 10 plants were inoculated using the same inoculum batch and these 30

plants (10x3) were considered a block within a randomized complete block design. There were

three blocks totaling 90 inoculated plants. Additionally as controls, six plants were rubbed with

buffer only plus abrasives and six more plants were left untreated.










Koll, S. and C. Biitner. 2000. Cell-to-cell movement of plant viruses through plasmodesmata: a
review. Arch. Phytopath. Pflanz. 33(2):99-110.

Kormelink, R., P. de Haan, C. Meurs, D. Peters, and R., Goldbach. 1992. The nucleotide
sequence of the M RNA segment of Tomato spotted wilt virus, a bunyavirus with two
ambisense RAN segments. J. Gen. Virol. 73:2795-2804.

Kresta, K.K., F.L. Mitchell, and J.W. Smith, Jr. 1995. Survey by ELISA of thrips
(Thysanoptera: Thripidae) vectored Tomato spotted wilt virus distribution in foliage and
flowers of field-infected peanut. Peanut Sci. 22:141-149.

Kucharek, T., L. Brown, F. Johnson, and J. Funderburk. 2000. Tomato spotted wilt virus of
agronomic, vegetable, and ornamental crops. Florida Coop. Ext. Service, Circ-914, 13 pp.

Lu, P., D.A. Huber, and T.L. White. 2001. Comparison of multivariate and univariate methods
for the estimation of Type B genetic correlations. Silvae Genetica 50(1): 13-22.

Lyerly, J.H., H.T. Stalker, J.W. Moyer, and K. Hoffman. 2002. Evaluation of Arachis species for
resistance to Tomato spotted wilt virus. Peanut Science 29, 79-84.

Lynch, M., and B. Walsh. 1998. Genetics and Analysis of Quantitative Traits. Sinauer
Associates, Inc., Sunderland, MA.

Mandal, B., H.R. Pappu, and A.K. Culbreath. 2001. Factors affecting mechanical transmission of
Tomato spotted wilt virus to peanut (Arachis hypogaea L.). Plant Dis. 85:1259-1263.

Mandal, B, H.R. Pappu, A.S. Csinos, and A.K. Culbreath. 2006. Response of peanut, pepper,
tobacco, and tomato cultivars to two biologically distinct isolates of Tomato spotted wilt
virus. Plant Dis. 90(9): 1150-1155.

McKeown, S.P., J.W. Todd, A.K. Culbreath, D.W. Gorbet, and J.R. Weeks. 2001. Planting date
effects on tomato spotted wilt in resistant and susceptible peanut cultivars. Phytopathology
91:S60.

Mitchell, F.L. 1996. Implementation of the IPM planting window for management of Tomato
spotted wilt virus and avoidance of peanut yellowing death. Final Compliance Report, Texas
Pest Management Association, Biologically Intensive Integrated Pest Management Grant
Program. http://stephenville .tamu. edu/fmitchel/ento/tswy3 .pdf.

Moury B., K.G. Selassie, G. Marchoux, A.M. Daubeze, and A. Palloix. 1998. High temperature
effects on hypersensitive resistance to Tomato spotted wilt tospovirus (TSWV) in pepper
(Capsicum chinense Jacq.). Eur. J. Plant Pathol. 104(3):489-498.

Mrode, R. 2005. Linear Models for the Prediction of Animal Breeding Values. 2nd. Edn. CABI
Publishing, Trowbridge, UK.










Gorbet, D.W. 2006. Registration of 'Carver' Peanut. Crop Sci. 46(6):2713-2714.

Gorbet, D.W. 2007a. Registration of 'ANorden' Peanut. J. Plant Registr. 1(2):123-124.

Gorbet, D.W. 2007b. Registration of 'AP-3' Peanut. J. Plant Registr. 1(2): 126-127.

Gorbet, D.W. and D.A. Knauft. 2000. Registration of 'SunOleic 97R' peanut. Crop Sci.
40(4):1190-1191.

Gorbet, D.W. and F.M. Shokes. 2002. Registration of 'C-99R' Peanut. Crop Sci. 42(6):2207.

Groves, R.L., J.F. Walgenbach, J.W. Moyer, and G.G. Kennedy. 2003. Seasonal Dispersal
Patterns of Frankliniella fusca (Thysanoptera: Thripidae) and Tomato spotted wilt virus
occurrence in Central and Eastern North Carolina. J. Econ. Entomol. 96(1): 1-11.

Hall, M.D., and D.A. Van Sanford. 2003. Diallel analysis of Fusarium head blight resistance in
soft red winter wheat. Crop Sci. 43(6):1663-1670

Hallauer, A.R. and J.B. Miranda, Fo. 1988. Quantitative Genetics in Maize Breeding. 2nd Ed.
Iowa State Univ. Press, Ames, IA, USA.

Halliwell, R.S., and G. Philley. 1974. Spotted wilt of peanut in Texas. Plant Dis. Rep. 58(1):23-
25.

Henderson, C.R. 1976. A simple method for computing the inverse of a numerator relationship
matrix used in prediction of breeding values. Biometrics 32(1): 69-83.

Hoffman, K., S.M. Geske, and J.W. Moyer. 1998. Pathogenesis of Tomato spotted wilt virus in
peanut plants dually infected with peanut mottle virus. Plant Dis. 82:610-614.

Holland, J.B. 2006. Estimating genotypic correlations and their standard errors using
multivariate restricted maximum likelihood estimation with SAS Proc MIXED. Crop Sci
46(2):642-654.

Holland, J.B., W.E. Nyquist, and C.T. Cervantes-Martinez. 2003. Estimating and interpreting
heritability for plant breeding: an update. Plant Breed. Rev. 22:9-111.

Holland, J. B., D.V. Uhr, D. Jeffers, and M.M. Goodman. 1998. Inheritance of resistance to
southern corn rust in tropical-by-corn-belt maize populations. Theor. Appl. Genet. 96(3):
232-241.

Huber, D.A., T.L. White, and G.R. Hodge. 1994. Variance component estimation techniques
compared for two mating designs with forest genetic architecture through computer
simulation. Theor. Appl. Genet. 88: 236-242.

Hurt, C.A., R.L. Brandenburg, D.L. Jordan, G.G. Kennedy, and J.E. Bailey. 2005. Management
of spotted wilt vectored by Frankliniella fusca (Thysanoptera: Thripidae) in Virginia market-
type peanut. J. Econ. Entomol. 98(5):1435-1440.













AP-3 + DP-1 NC94002 ~tNemaTAM


100

90

80

en 70

S60







20




30 DAP 60 DAP 120 DAP


Figure 2-5. Spotted wilt incidence in four peanut genotypes at three assessment dates at Marianna, Florida in 2007.









The high values of both Type B genetic correlation coefficients and repeatability estimates

suggested a strong genetic determination of the observed genotypic differences in spotted wilt

intensity ratings. This emphasizes the importance of resistant cultivars in the management of

spotted wilt.









NC94002, AP-3, Georgia-02C, DP-1 and C-99R while two genotypes showed intermediate

values (Georgia Green and ANorden).

Location Effect

The univariate analysis first tried (described under point "i' of Material and Methods), is

similar to the usual ANOVA plus the specification of different error terms for each location. It

provided variance components of the interaction Location by PD that were rather important most

of the time (data not shown) and were due to heterogeneous variances among planting dates.

Consequently the decision was made to model a variance for each combination of location and

planting date as described under point "ii" of Material and Methods. With this model, the

location effect was found significant in all the PD and AD (Table 1-3) with Marianna exhibiting

higher predicted values (Table 1-4). In the most extreme case, PD3 at AD90, the Marianna

predicted value for transformed DIR was five times larger.

The ratio of error variance for Marianna to the error variance for Citra was mostly above

one (Table 1-3), suggesting higher data variability at Marianna. An extreme value of 4.5 for this

ratio was observed at PD 3, AD 112.

In Marianna, the ratio of the block variance component to the error variance ranged from 0

to 0.7. At Citra, the DIR were extremely close to zero (data not shown), denoting the rather

minor importance of block as a variability source.

The ratio of genotypic variance to the biggest location error variance ranged from 2.5 to

20.2. In every case, the error variance for Marianna was used as the denominator in the ratio. It

was clear that genotypic variance was the main source of data variability.











Table 1-3. Location effect and variance ratios for bivariate analysis performed on transformed
spotted wilt disease intensity ratings at three planting dates assessed in different times
at Citra and Marianna, Florida in 2005.
P-value>F
Planting Assressment ~2,Mo, 2 bl~~ 2bc 2 2 2 2 2
Date date fr 0e/ c 0b C 0b MI>e
location
2 132 0.004 1.4 147712.9 0.7 5.1
3 90 <0.001 3.3 10.5 0.0 4.9
112 0.004 4.5 1137.2 0.7 2.5
132 <0.001 1.3 10.5 0.0 5.6
4 112 <0.001 0.7 0.7 0.2 20.2
G2eM: Marianna error variance. G2eC : Citra error variance. (T2bM: Marianna block variance. o2bC : Citra
block variance. o21g>: biggest location*genotype variance. (T2e>: biggest location error variance.


Table 1-4. Transformed spotted wilt disease intensity rating predicted values, standard errors
(SE) and standard errors of mean differences (SED) for planting dates assessed at
different times at Citra and Marianna, Florida in 2005.
Planting Assessment Predicted
Location SE SED
Date date Value
Citra 0.3881 0.0391
2 132 0.0786
Marianna 0.7858 0.1011
Citra 0.1879 0.0339
90 0.0875
Marianna 0.9461 0.1045
3 112 Citra 0.3349 0.0391 019
Marianna 1.0253 0.1400
Citra 0.4247 0.0331
132 0.0698
Marianna 1.0097 0.0923
Citra 0.4337 0.0552
4 112 0.0899
Marianna 0.9532 0.1108










v) Bivariate analysis for different ADs. This was designed to address the

correlation among DIR from different ADs for every available Location by PD combination (see

Table 1-1). Additionally, the effect of PD on each DIR in the pair being analyzed, was

determined.

A detailed explanation of each analysis follows:

i) Univariate analysis for location

This analysis was performed for each planting date separately to test the difference in the

DIR between locations. For example, for PD 2, the only score available at both locations was the

one obtained at 132 AD. Thus, this data subset was analyzed and from that the differences

among Locations were tested.

The arcsine(square root(DIR) values observed in a PD at a determined AD were modeled




y = pU + X Z + Zvu, + Zvuy, + Z~b lb + e (Eq. 1-2)


where y is a vector containing the arcsine(square root(DIR) values, ZIis the p x 1 vector of


a constant and fixed effect Location, Xl is an n x p design matrix of full column rank which

associates observations with the appropriate combination of the fixed effect. The u vectors (ue,

up and ulb) arT q X 1 Vectors for the random terms variety, Location by variety and Location by

block while Z matrices (Z,, Zv, and Z~b) arT H X q design matrices for those random effects

mentioned above. The random effects and error are assumed to be independent Gaussian

variables with zero means and variance structures var(ui) = di lbi (where bl is the length of ut; i =


1....3) and var(e) = d In .









Variance components for heritability estimation today are mostly obtained through Mixed

Linear Model approaches. They offer the flexibility of analyzing various types of unbalanced

data coming from non-traditional mating designs with good precision (Holland et al., 2003).

Using this approach, random effects such as breeding values can also be obtained (Lynch

and Walsh, 1998). Breeding value is the sum of the additive effects of an individual's genes

(Lynch and Walsh. 1998).

Recently, the use of mixed models coupled with REML has allowed the accurate

estimation of additive variance and consequently of the breeding values. A common linear mixed

model used mostly by animal breeders is the Animal Model (Mrode, 2005). It utilizes all genetic

relationships among the individuals being analyzed, in order to obtain a more accurate estimation

of the additive variance than traditional methods (Henderson, 1976; Lynch and Walsh, 1998).

Breeding value estimates (BLUPs), obtained through the Animal Model are accurate because

they take into consideration not only the performance of the individual but also that of its

relatives (Mrode, 2005).

Breeding values are used to choose individuals in a population that are superior for a trait

and that will provide a better progeny. They can also be an integral part of a selective index to

choose individuals based on several traits (Mrode, 2005).

To gain insight into the genetics of resistance to TSWV in peanut, the obj ectives of this

study were: 1) to provide heritability estimates from crosses involving different sources of

resistance while assessing their potential to generate superior progenies; and 2) to explore the

relationship among different symptoms of infection by TSWV.

Material and Methods

To study the inheritance of resistance to TSWV in peanut, three resistant genotypes (AP-3,

DP-1 and NC94002) and a susceptible genotype (NemaTAM) were mated.







































To Monica and Victoria.
































0.5


O


2.5


2


1.5


1


E F2 0 F3


AP-3 / NemaTAM NemaTAM /AP-3 NemaTAM / DP-1 NemaTAM /
NC94002


DP-1 /NC94002


Figure 2-15. Variability of best linear unbiased predictors for TSWV-induced stunting in the F2 and F3 generations of five peanut
crosses tested at Quincy, Florida in 2007.










LIST OF TABLES


Table page

1-1 Layout of data collection for planting date studies of peanut in Marianna and Citra,
Florida in 2005. ............. ...............38.....

1-2 Means and standard errors for tomato spotted wilt disease intensity ratings at
different planting dates at Citra and Marianna, Florida in 2005 ................. ................ ..38

1-3 Location effect and variance ratios for bivariate analysis performed on transformed
spotted wilt disease intensity ratings at three planting dates assessed in different
times at Citra and Marianna, Florida in 2005. ............. ...............39.....

1-4 Transformed spotted wilt disease intensity rating predicted values, standard errors
(SE) and standard errors of mean differences (SED) for planting dates assessed at
different times at Citra and Marianna, Florida in 2005. ............. .....................3

1-5 Planting date effect and variance ratios for univariate analysis performed on
transformed spotted wilt disease intensity ratings at four planting dates assessed in
different times at Citra and Marianna, Florida in 2005. ............. .....................4

1-6 Transformed spotted wilt disease intensity rating predicted values, standard errors
(SE) and standard errors of mean differences (SED) for at four planting dates
assessed in different times at Citra and Marianna, Florida in 2005 ................. ........._.....40

1-7 Location by planting date combination (cell) effect and variance ratios for
multivariate analysis performed on transformed spotted wilt disease intensity ratings
at different planting dates assessed at different times at Citra and Marianna, Florida
in 2005. ............. ...............40.....

1-8 Transformed spotted wilt disease intensity rating predicted values, standard errors
(SE) and standard errors of mean differences (SED) for location by planting date
combinations (cells) assessed at different times in Citra and Marianna, Florida in
2005. Each assessment date was analyzed separately. ............. ...............41.....

1-9 Phenotypic and genetic correlations among transformed spotted wilt disease intensity
ratings at four planting dates assessed in different times at Citra and Marianna,
Florida in 2005. Values in brackets are standard errors............... ...............41.

1-10 Type B genetic correlations and [standard errors] for transformed spotted wilt
disease intensity ratings at four planting dates assessed in different times in Citra and
Marianna, Florida in 2005. a............ ...............42.....

1-11 Entry-mean repeatability estimates and their [standard errors] for transformed
spotted wilt intensity ratings at three planting dates assessed three times in Citra and
Marianna, Florida in 2005. Values for each location are separated by a slash, Citra
being on the left and Marianna on the right. .............. ...............42....









Neither "Inoculum Dilution" (p=0.4136) nor "Inoculum Batch" (p=0.9734) were found

significant in determining the "ELISA Status" after the inoculation. The infectivity among

batches was much less variable than in the previous two tests (Table 3-3).

Discussion

Test 1

Very low infection levels were attained compared with data from literature using this

method (Mandal et al., 2001). Although low infection rates are not uncommon, no single factor

has been detected as the cause (N. Martinez-Ochoa, pers. comm.).

The marked impact of plant age at inoculation on infection success has been demonstrated

by several authors. Mandal and co-workers (2001) and Hoffman et al. (1998) obtained high

percentages of symptomatic plants (75% and 90% respectively) for plants at 14 DAP. In spite of

using the technique of Mandal et al. (2001), the percentages obtained in this experiment were

smaller and similar to those reported by Pereira (1993). According to Noordam (1973), Branch et

al. (2003) and S. Mullis (pers. comm.) well-irrigated non-stressed plants are more prone to

become infected or to develop symptoms. As the small volume of substrate in which the plants

were raised in the present work tended to dry very easily, the plants were subj ected to short but

frequent periods of water stress which could have contributed to the lower-than-expected number

of infected plants. The irrigation problem was solved in the subsequent tests by increasing the

applied water volume.

Similar to Culbreath et al. (1992), almost half of the ELISA positive plants in this study

were asymptomatic. Hoffmann et al. (1998) reported that symptomatic TSWV infected leaves

were readily detected by ELISA although asymptomatic leaves of infected plants did not always

give positive ELISA readings. Similarly, four plants visually scored in our experiment as

infected were not detected as such by ELISA. Kresta et al. (1995) in peanut and Canady et al.









the average breeding value of the F2 and F3 pOpulations were usually intermediate to their

parents' breeding values. Both facts seem consistent with an additive mode of action (Falconer &

MacKay, 1996).

As additive variance increases with selfing (Nyquist, 1991), individual BLUP tend to

increase so more individuals in each distributional extreme can be found in F3 than in F2.

Although further selfing continues to increase the additive variance, resource limitations always

force some type of selection in early generations (Simmonds, 1979).

The low individual heritability but good reliability of family BLUP suggests that taking

into consideration the family performance for spotted wilt resistance when selecting individuals

among F2:3 familieS, as is frequently practiced in peanut breeding programs in Southeastern

USA, is a safe breeding strategy (Falconer and Mackay, 1996; Hallauer and Miranda, 1988).

When selecting in populations derived from the resistant parents used here, inclusion of spotted

wilt resistance a part of a selective individual multi-trait index is acceptable. However, it should

be given a moderate weight because of its modest heritability.

It seems clear that RxR crosses would provide better populations to select for spotted wilt

resistance and those having NC94002 as a parent would display the best response to selection.

Conclusion

The use of the Animal Model provided accurate and precise heritability estimates. They

ranged from 0.01 to 0.71, but were most frequently in the low to medium range. The estimates

increased as the epidemics progressed.

The almost exclusive spotted wilt symptoms detected in each test were stunting and foliar

symptoms (spots or mosaics). Tip death, leaf necrosis and yellowing chlorosiss) were rare. There

was high phenotypic and genotypic correlation among stunting and foliar symptoms suggesting

either pleiotropy or a very strong coupling linkage among their genetic determinants.










Simpson, C.E., J.L. Starr, G.T. Church, M.D. Burow, and A.H. Paterson. 2003. Registration of
'NemaTAM' Peanut. Crop Sci. 43(4):1561.

Soler S., M.J. Diez, and F. Nuez. 1998. Effect of temperature regime and growth stage
interaction on pattern of virus presence in TSWV-resistant accessions of Capsicum chinense.
Plant Dis. 82: 1199-1204.

Tillman, B.L., D.W. Gorbet, and P.C. Andersen. 2007. Influence of planting date on yield and
spotted wilt of runner market type peanut. Peanut Science 34(2):79-84.

Ullman, D.E., R. Meideros, L.R. Campbell, A.E. Whitfield, J.L. Sherwood, and T.L. German.
2002. Thrips as vectors of Tospoviruses. Adv. Bot. Res. 36: 1 13-140.

Venuprasad, R., H.R. Lafitte, and G.N. Atlin. 2007. Response to direct selection for grain yield
under drought stress in rice. Crop Sci. 47(1):285-293.

Westfall, P.H. 1987. A comparison of variance component estimates for arbitrary underlying
distributions. J. Amer. Stat. Assoc. 82(399): 866-874.

Yamada, Y. 1962. Genotype by environment interaction and genetic correlation of the same trait
under different environments. Jap. J. Genet. 37: 498-509.

Yang, R.C., N.K. Dhir and F.C. Yeh. 1998. Intraclass correlation of polychotomous responses of
Lodgepole pine to infection of Western gall rust: a simulation study. Silvae Gen. 47(2-
3):108-115.

Zobel, B. J. and J. T. Talbert. 1984. Applied Forest Tree Improvement. John Wiley & Sons, New
York.










2-13 Generation-mean best linear unbiased predictors for TSWV-induced stunting in
populations from five peanut crosses tested at Marianna, Florida in 2007 (all
reliabilities were above 0.9). .............. ...............85....

2-15 Variability of best linear unbiased predictors for TSWV-induced stunting in the F2
and F3 generations of five peanut crosses tested at Quincy, Florida in 2007. .................. .87

2-16 Percentage of individuals and (number ofF3 families) displaying BLUPs for TSWV-
induced stunting above their best parent, in each of five peanut crosses tested at
Marianna and Quincy, Florida in 2007. ............. ...............88.....










The logits of the unknown binomial probabilities (i.e., the logarithms of the odds) are

modeled as a linear function of the Xi. The unknown parameters pjare usually estimated by

maximum likelihood.

The full model containing both elapsed time and inoculum batch as factors was used and in

the event of a factor being found non-significant, it was removed from the model (Agresti, 1996).

Fisher' s Exact Test was used to detect association between "Appearance of systemic symptoms"

and "ELISA status".

Results

Test 1

The overall percentage of inoculated plants showing visual symptoms of systemic infection

was 18% (most of them with mild severity, Table 3-1) while no localized symptoms were

observed in the inoculated leaves. According to the ELISA test, 43% of the plants were

systemically infected. Four out of 16 plants showing systemic symptoms failed to be detected by

ELISA. Nonetheless, the association between systemic symptoms and ELISA status was

statistically significant (p=0.0289) according to Fisher' s Exact Test.

Neither the elapsed time from inoculum preparation nor inoculum batch were significant

(p=0.09 and 0.13 respectively) in the logistic regression for systemic symptoms. In the case of

the serological status, both factors were found significant (p=0.02 and 0.05).

The obtained Maximum Likelihood Estimates can be seen in Table 3-2

The Prediction Equations obtained for the logit of the probability of a Positive ELISA

result under each treatment were then:

For Time O' and Batch 3, logit(AZ)=-0.3336+0 Time+0 Batch

For Time O' and Batch 1, logit(AZ)=-0.3336+0 Time+(-0.7978 Batch)












Table 3-1. Effect of elapsed time after preparation of Tomato spotted wilt virus inoculum on the
number of peanut plants declared infected by visual examination or serological
means.
Treatment Symptomatic plants a ELISA-positive plants
O' batch 1 1 2
10' batch 1 6 5
20' batch 1 1 1
O' batch 2 4 3
10' batch 2 2 9
20' batch 2 0 5
O' batch 3 2 9
10' batch 3 0 4
20' batch 3 0 1
a Number of plants for each treatment=10


Table 3-3. Effect of inoculum dilution on the number of peanut plants declared infected by
ELISA.
Treatment ELISA-positive plants a
10:1 -batch 1 8
20:1 -batch 1 8
10:1 batch 2 7
20:1 batch 2 8
10:1 batch 3 10
20:1 batch 3 6
10:1 batch 4 8
20:1 batch 4 8
a Number of plants for each treatment=10


Table 3-2. Maximum Likelihood Estimates for time and inoculum batch effects on artificial
inoculation of Georgia Green peanut.
Parameter Estimate S.E.
Intercept -0.3336 0.2354
Time (10') 0.768 0.3274
Time (20') -0.9494 0.3504
Batch (1) -0.7978 0.3442
Batch (2) 0.6253 0.3274










Table 2-3. Continued


Stunting


Foliar symptoms


Quincy 2007
O2Block O2Plot


G2Block 02Plot


G2A
b


02NA ~2e


G2NA


Assessment date
30 DAP


60 DAP

120 DAP


0.003 0.197 0.197 0

0.008 0.243 0.640 0


3.227

1.396


0.004

0.019


0.248

0.274


0.268

0.738


3.983

1.394


o2Block: Block variance; o2Plot: Plot variance; o2A: Additive variance; G2NA: Non-additive variance. a: term not included in the
model. b: date not assessed. c: incidence too low to allow analysis























Maximum
Average
O Minimum


Figure 2-11. Breeding values' reliability for TSWV-induced stunting among individuals in segregating populations derived from fiye
peanut crosses evaluated in Hyve Hield tests.










present in the tested genotypes performed consistently when faced with the viral populations

present at both locations, which could arise from similarities in the viral consensus sequence

between locations (i.e. they are essentially the same) or the fact that both consensus sequences

induce similar rankings on the tested genotypes in much the same way Mandal et al. (2006)

reported for Georgia isolates.

Genotypic variances were the most important random variation source in this study. By

having a wide range of resistances in the tested genotypes the genotypic variance is expected to

be increased (Betran et al., 2006), thus increasing the repeatability.

Another cause of high genetic variances relative to the error term was probably the correct

modeling of the experimental data (Gilmour et al., 2006), which was accomplished throughout

the present study.

Additionally, genetic causes (biochemical pathways) could predominantly have established

the performance of the genotypes (Lynch and Walsh, 1998). This last explanation seemed to be

supported by the high genetic correlation coefficients obtained here.

Conclusions

The modeling of correct variance and covariance structures of the tests provided a good

estimation of variances and covariances which in turn allowed determination that location was a

significant factor in establishing the spotted wilt ratings observed among genotypes. Meanwhile,

planting date was only a significant factor under light epidemics or late in the season under

heavy epidemics.

The high correlation among assessment dates indicated that the relative performance of

genotypes can be perceived early in the season and the genotypic differences tend to persist until

harvest time.









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

GENETICS OF TOMATO SPOTTED WILT VIRUS RESISTANCE INT PEANUT
(Arachis hypogaea L.)

By

Jorge J. Baldessari

August 2008

Chair: Barry L. Tillman
Major: Agronomy

Tomato spotted wilt virus (TSWV, Bunyaviridae: Tospovirus) is a maj or peanut pathogen

in the USA. Its management involves, among other factors, the use of resistant cultivars and

recommended planting dates.

Ten genotypes with varied degrees of resistance were field tested in two locations and four

planting dates with the following obj ectives: 1) to ascertain the importance of planting dates and

location as determining factors of spotted wilt epidemic intensity, 2) t evaluate the consistency in

the performance of an array of genotypes with contrasting spotted wilt resistance assessed at

different times, and 3) to provide an estimation of how much genotypic consistency can be

ascribed to genetic causes. Results indicated that location was a significant factor in determining

the spotted wilt damage, while planting date was significant only under a light epidemic or late

in the season under a heavy epidemic. The high correlation between assessment dates implied

that genotypic performance was perceived early and differences persisted until harvest. High

Type B genetic correlation and repeatability suggested a strong genetic determination of

resistance.


















-a Citra '05 -a- Citra '06 - Marianna '06 ~-- Marianna '07 -* Quincy '07

3.5






2.









0.5 .

0 .

30DAP 60DAP 120DAP

Assessment dates


Figure 2-3. Stunting severity progression in five peanut crosses, in five field tests assessed at three dates in three Florida locations.









The vector of observations y is assumed to be univariate normal with mean E(y)= XP and

vrac-vaarl~~-~valiance Var~y)= ZB Vg Z'B + Zeaf Z' + Zi Aaj Z' + ZNAZ'NA ~e

The REML method was used to estimate additive and non-additive variance components

by using the ASREML software (Gilmour et al., 2006). Individual predicted breeding values

(BLUPs) were obtained from the solutions for the individual effects in the above mentioned

Animal Model. Mean F2 and F2:3 BLUPs (breeding values) were obtained by averaging the

BLUPs of all the individuals in the corresponding population. BLUP Reliabilities (a measure of

their accuracy) for parents were obtained by using the formula:


r = PEDICION(Eq. 2-2)


whr PREDICTION 'S the variance~ of the~ predicted breeding value andu uji h adtv

variance estimation from the corresponding model.

Mean F2 and F2:3 BLUPs were obtained by averaging the squared prediction standard

errors for the BLUPs of all the individuals in the corresponding population and dividing them by

the square of n (number of squared BLUPs used in the average).

Narrow-sense heritability was estimated on individual values, as


h2 2
h ,2 (Eq. 2-3)


where A is the additive variance, NA is the non-additive variance, "Pis the plot


variance and a2 is the residual variance. The variance components were directly provided from

the model fitting. The heritability estimates are probably upwardly biased because they may

contain genotype-by-environment interaction variation not accounted for, because calculations

were performed for each single environment (Nyquist, 1991).










variety and PD by block while Z matrices (Z,, Z,, and Zpb) arT H X q design matrices for those
random terms mentioned above.

The random effects and error were assumed to be independent Gaussian variables with

zero means and variance structures:

var(ug) = gsllo, var(uPg) dPg 20, Var(Upb) = pb,13 and var(ey) = dI60*

Thus, the random terms were assumed to have a variance resulting from the direct product

of three G correlation structures (one for each random term) while the error variance was

described by specifying a unique unstructured correlation R structure.

In addition to the previous assumptions, the bivariate analysis also involves the following

ones:


cov(uADa )AU OADaADb J10 COV(MPgADa ) PgADb = A~A b20 COV(UpbAa~ )pbADb = pbwb 3


and cov(eADa) = ADb ADaADb 60 Thus random effects and errors are correlated between variables

(DIRs from different ADs).

The bivariate model for the Citra data subset can be written as


y = (I2 0 X) T+ (I2 0 Zg) Ug + (I2 0 Zb) Ub + e (Eq. 1-7)


where A=y,,a>,) A : u=UA,) s~a) RA bAM = bA~ ,b and e = (eADa> ADb .~ In turn,


A,,a = the vector containing the arcsine(square root(DIR) for AD "a"; while T, is the 1 x p vector


of a constant, X, is an p x n design matrix of full column rank which associates observations with

the constant, the u vectors (ugADa and ubADa ) are 1 x q vectors for the random terms variety and

block while Z matrices (Z, and Zb) arT H X q design matrices for those random terms mentioned

above.









Both tested factors (inoculum dilution and inoculum batch) were statistically non-

significant, probably due to a similar number of viable virions reaching the target tissues

between both treatments. The smallest dilution used here (10:1) is certainly weaker than that

reported by some researchers (Hoffman et al., 1998; Pereira, 1993; Mandal et al., 2001). This

suggests that even a 20: 1 buffer:tissue ratio can provide similar infection rates than higher ratios

allowing a more efficient use of tissue donor plants. This factor can be especially important

while preparing batches for a large number of plants from limited amounts of infected tissue

from donor plants.

Conclusions

Following the protocol suggested by Mandal et al. (2001), this study determined that both

number of rubbings and inoculum dilution had no effect on the outcome of artificial inoculations.

The elapsed time from inoculum preparation showed an unexpected trend as the infectivity did

not fall with time as suggested by some authors (Halliwell and Philley, 1974; Clemente et al.,

1990; Mandal et al., 2001). A new test using more replications per treatment could provide

further insight on this issue.

Inoculum batch was an important factor, probably highlighting the fact that viral titer is

highly variable even when using infected tissue with similar characteristics (age, plant position).

This stresses the importance of standardizing the inoculation process.

In two of the three tests the difference between the percentages of symptomatic and

ELISA-detected plants was similar to that reported by Culbreath et al. (1992) but this was not the

case in Test 3. Other factors could exist that influence the visual symptom development that was

not controlled in this test.

The overall low infection rates obtained in comparison with other reports using similar

techniques clearly suggest that additional work is necessary to detect which factors caused the










The random effects and error were assumed to be independent Gaussian variables with

zero means and variance structures:

var(ug) = C gl,0. var Hi)= a 3 and var(eJ)- = ; Ioo.

Thus, the random terms were assumed to have a variance resulting from the direct product

of two G structures, an unstructured correlation for Genotypes and a general correlation for

blocks while the error variance was described by specifying a unique unstructured correlation R

structure.

In addition to the previous assumptions, the bivariate analysis also involves the following


ones:s~ cov2tem\D 4Da4Db~,ly COV(Zlh )21, m CT^4a4Dbl and cov(e,,4Da, =DhADadDh 60

Thus random effects and errors are correlated between variables (DIRs from different ADs).

The bivariate analysis also provided an estimate of "true" Type A genetic correlation (two

traits measured on the same experimental unit) through bivariate REML estimation among DIR

obtained at different AD (Gilmour et al., 2006).

Calculation of Genetic Correlations

Traditional (Type A) genetic correlation was calculated among DIRs obtained on the same

plot at different ADs. Using the genotypic variance and covariance component estimates

obtained from the corresponding linear mixed model described under section v) Bivariate

analysis for different ADs. The genotypic correlation between DIRs from ADj and ADk, WaS

estimated as


COV k
r 2 2 (Eq. 1-8)









Evaluation of Inoculated Plants and Analysis of Data

The measured variables in each experiment were as follows:

Appearance ofsystemic symptoms: as viral lesions on inoculated leaves were not observed,

the plants were considered as "infected" when chlorotic spots followed by mosaic rings and

necrotic spots developed in the newly emerging leaves (systemic symptoms). In their absence the

plants were considered "healthy".

Serological detection of TSWV by ELISA : optical density (OD) values greater than the

average value plus 3 times the standard deviation (cut-off value) of the two negative control

wells, belonging to healthy plants of C1 1-2-3 9 peanut line, were considered positive for the

presence of TSWV. Due to contamination of the negative controls in Test 1, a cut-off value was

set taking into account the usual values obtained for this type of control. As the highest value

ever obtained for this negative control has been 0.006 in several previous ELISA (data not

shown) it was considered reasonably conservative to use 0.06 as a cut-off value.

TSWV infection was confirmed by alkaline phosphatase labeled DAS-ELISA according to

manufacturer' s instructions (Agdia Inc., Elkhart, Indiana). Absorbance was measured at 405 nm

with an automated microplate reader (Model 680, Bio-RAD, Hercules, CA,USA). Two

replications were made on each sample, and averages were used for evaluation.

Recording of symptomatic plants and ELISA were done 3 weeks post-inoculation. Two

apical leaflets in the youngest leaf plus one apical leaflet on the youngest fully expanded leaf and

young secondary roots were used for ELISA. If new leaves were observed as symptomatic they

were used for ELISA instead of using random leaves. Since the tested tissue was not weighted

and the obtained macerate volume was variable, the ELISA values were used only to categorize

the plants as infected or not.










Standard errors of the correlations were calculated using the Taylor Series Expansion

method (Gilmour et al., 2006).

Calculation of Repeatability

An entry-mean repeatability of performance in each environment was calculated as an

intra-class correlation using the corresponding variance components from the linear mixed

models described under section ii) Bivariate analysis through G structure specification, according

to the formula (Holland et al., 1998):


R= -


r (Eq. 1-11)

2
where, o, Je2 and "r" are the genotypic variance, the error variance and the number of

replications at that location, respectively.



Results

There was a clear difference in the intensity of the epidemic between locations. By 132

DAP the highest DIR in Citra was less than half the smallest DIR recorded at Marianna (Table 1-

2).

The epidemic progression was slow but steady at Citra while at Marianna it was fast and

abrupt, reaching "final" intensity as early as 90 DAP (Fig. 1-1). The slight reduction in intensity

observed in the figure was caused by the harvest of susceptible genotypes prior to the assessment

at 132 DAP, reducing the overall intensity DIRs for that AD. Among genotypes, a clearer

separation of resistance groups was observed at Marianna, under a heavy epidemic compared to

Citra (Figs. 1-2 & 1-3). The groups' conformation was as expected, the susceptible genotypes

comprising F-43 5HOL, NemaTAM and SunOleic 97R, the resistant genotypes comprising




















4.



3 .5. - -




2 .5 - : : : - -- -


2 - - - -- -




















Fiue2-3 eerto-ea et ierunisdprdcor o SW -nuedsain nppuain fo iv ent rse


tete at Maiana Floid in 207 allrelibiltie wer abve .9)










Standard error of heritability was calculated using the Taylor Series Expansion method

(Gilmour et al., 2006).

Genetic and phenotypic correlation coefficients were calculated as

cov
r, = a"Y (Eq. 2-4)



weecova- is the additive covariance between trait X (stunting) and Y (spots); a2 and

Ya are the additive variances for traits X and Y respectively.


cov, + cov, + cov + cov
axy Py x
phb 2 2 (Eq.2-5)
phx ph,


where o hx x Ax x x being & hx the phenotypic variance, is the

2 2 2
additive variance, ONAx is the non-additive genetic variance, Gpx is the plot variance and "ex

is the residual variance for the trait X. Similar variances apply also for trait Y.

Standard errors of correlations were calculated using the Taylor Series Expansion method

(Gilmour et al., 2006).

Results

The predominant spotted wilt symptoms in every test were stunting and foliar symptoms.

Tip death, leaf necrosis and yellowing incidences never reached more than 5% in any test (Fig 2-

1). Thus, no analysis was performed on them.

Epidemics varied considerably in their intensity among tests, ranging from light at Citra

2005 and 2006 to severe at Marianna and Quincy in 2007 (Table 2-2).









































Citra '06


Figure 2-1. Incidence of spotted wilt symptoms at 120 days after planting in populations from five peanut crosses tested in five Florida
environments.


m


O Stunting O Spots 5 Tip Death Leaf Necrosis O Yellowing


III


I


III


I I I I


100

90

80

70

60

50

40

30

20

10
-


n


Citra '05


Marianna '06 Marianna '07


Quincy '07










5. Always droopy, yellow leaves, reddish slightly dehydrated stems.

In all variables a score of 0 was assigned if no symptoms were apparent. The plants were

scored at three points in time: 30, 60 and 120 days after planting (DAP), except at Citra in 2006

and Quincy in 2007 where no scoring was done 30 DAP.

Traditional mating designs used to estimate genetic variance components are applicable

only when parental components are unrelated (Hallauer and Miranda, 1988). By using REML to

estimate genetic variance components in a mixed model approach, it is possible to account for

the relationship among individuals. A mixed model approach using a single trait Animal Model

(Mrode, 2005) was employed to estimate genetic variance components from populations derived

from the crosses. Since F2:3 familieS made up most of the data and they mostly changed among

environments, the analyses were performed by year and location according to the following

model :

y = Xj3 + ZBOB + Zpup + Ziui + ZNANNA + e (Eq. 2-1)

where y is the vector of observation for each individual; ZB and uB are the incidence matrix

andu vector of randomll block effects B~NID(0, VBj), with 2 <; B < 3. ZP andU uP are the~ inidenceI




inidenceI~, mlatrlix andu vectorv of randomll additive effects I~(0 -Y\ VA j). ZNA andU uNA are the~




errors, which are NID(0, a, ).

The additive matrix A (Henderson, 1976) has dimensions 37013x37013 with diagonal

elements equal to 1 and off-diagonal elements equal to two times coancestry coefficient (2rxy)

between the individuals in the study. The diagonal non-additive matrix NA has dimensions

37013x37013 and was expected to account for each variation not accounted for the A matrix.









Genotypic correlations were high at both locations, ranging from 0.85 to 1 (Table 1-9) with

rather small standard errors. Many coefficients were fixed at the maximum theoretical value of 1

by constraining the covariance matrix.

There was a moderate and significant correlation among the phenotypic and genotypic

correlation coefficients (0.76, p=0.04) by Spearman's rank correlation but due to the reduced

number of data pairs, this significance should be taken cautiously.

A high within location consistency of genotypic performance for DIR at different PDs and

ADs was observed. Type B genetic correlation coefficients were high, ranging from 0.83 to 1

with modest standard errors, with Citra displaying larger errors (Table 1-10). There was also

consistency of genotypic performance across locations at the same PD and AD. In three out of

five cases, the correlation coefficient was 1 while in the remaining two cases it was above 0.7

(Table 1-10).

When calculated from the combination of locations x planting date (cell), the correlation

coefficient between transformed DIR was very high, with all but one value above 0.9 (Table 1-

10). All the coefficients for AD 132 had a value of one, suggesting that the performance of the

evaluated genotypes in a certain environment (combination of location x planting date) was

predictable based upon its performance in another environment.

Irrespective of the type of genetic correlation calculated, the general results showed that

the genotypic performances in the tested environments were highly correlated between locations

and among ADs and PDs. Additionally, the modeling of variance structures allowed better

estimations of these correlations through elimination of scale effects and error shrinkage.

Repeatability

Repeatability values for the transformed DIR at different PDs and AD were high (Table 1-

1 1) and their precision good (small standard errors). All but one of the values was above 0.75





--- Citra05 --- Citra06 ---- Marianna06 -a- Marianna07 -- -- Quincy07


100

90

80

70


50

40

30

20

10


30 DAP


60 DAP


120 DAP


Assessment date






Figure 2-2. Percentage of plants displaying spotted wilt symptoms in five peanut crosses in five field tests assessed at three dates in
five Florida environments.









As the genetic correlation between stunting and foliar symptoms was very high, the

analysis of breeding values (BLUPs) is presented only for stunting.

Breeding Values (BLUPs)

The BLUPs obtained from each test showed a clear distinction among the susceptible

(NemaTAM) and the resistant parents (AP-3, DP-1 and NC94002) (Fig. 2-10). The only

exception was Citra 2005, where the only resistant parent grown was AP-3. In most tests

NC94002 displayed the best (smallest) breeding value for stunting in every test it was in,

whereas DP-1 was better than AP-3 in 3 out of 4 tests. A similar pattern was also observed for

BLUPs for foliar symptoms (data not shown). There was, however, an important variation in the

BLUPs of each parent among tests. This is somewhat expected because, although most of the

individuals in each test were genetically related, they were not identical.

The reliability of the breeding values (which provides a measure of their accuracy) was

intermediate for all tests, except for the ones in 2006 (Fig. 2-11) in which the average reliability

for individuals in segregating generations was zero. The parents in each test had intermediate to

high reliability because of their great number of relatives included in each test, as each parent

was grown alongside with its Fl, F2, F2:3 and even backcross individuals.

Both tests in 2007 showed the most damaging epidemics and the best breeding value

reliabilities. Consequently, their comparison follows. The rank correlation among the generation-

mean BLUPs at both 2007 tests was highly significant (r=0.93, p<0.0001) suggesting a very

similar breeding value of each generation irrespective of the location in that year.

Generation BLUPs

When average BLUPs were calculated for the different generations in both 2007 tests, the

best ones belonged to the RxR (resistant with resistant) cross (Figs. 2-12 and 2-13). Among the























Citra ---*--- Marianna


VI

RI
1~ 0.5
O
cn
r
~ 0.4
r
a,
cn
co 0.3
a,
cn

r 0.2
co
a,
r
0.1



0


--------------------------------~----..


90 DAP


112 DAP


132 DAP


Figure 1-1.


Change of spotted wilt disease intensity ratings over time at Citra and Marianna,

Florida in 2005 (averaged over all planting dates and genotypes)









{exp[-0.3336+(0.7680 Time)+(0 Batch)}/{1+ exp[-0.3336+(0.7680 Time)+(0 Batch)}=

0.61

For Time 10' and Batch 1,

{exp[-0.3336+(0.7680 Time)+(-0.7978 Batch)}/{1+ exp[-0.3336+(0.7680 Time)+(-0.7978

Batch)}= 0.41

For Time 10' and Batch 2,

{exp[-0.3336+(0.7680 Time)+(0.6253 Batch)}/{ 1+ exp[-0.3336+(0.7680 Time)+( 0.6253

Batch)}= 0.74

For Time 20' and Batch 3,

{exp[-0.3336+(-0.9494 Time)+(0 Batch)}/{1+ exp[-0.3336+(-0.9494 Time)+(0 Batch)}=

0.22

For Time 20' and Batch 1,

{exp[-0.3336+(-0.9494 Time)+(-0.7978 Batch)}/{ 1+ exp[-0.3336+(-0.9494 Time)+(-

0.7978 Batch)}= 0. 11

For Time 20' and Batch 2,

{exp[-0.3336+(-0.9494 Time)+(0.6253 Batch)}/{1+ exp[-0.3336+(-0.9494 Time)+(0.6253

Batch)}= 0.34

As can be seen from the parameter estimates for each factor level, the probability of

obtaining a positive ELISA increased by using the second inoculum batch or by using the

inoculum 10 minutes after its preparation while that probability decreased by using the first

inoculum batch or by using the inoculum 20 minutes after its preparation.

Test 2

The OD cut-off value for this test was set at 0.004. Thirty nine percent of the plants were

declared positive by ELISA while only 13% displayed visual symptoms. Despite this difference













Individual BLUPs ........._._.. ..... .___ ...............60.....

Family BLUPs................ ...............61.
Discussion ........._._.. ..... ___ ...............61.....
Conclusion ........._._ ........_. ...............67.....


3 ARTIFICIAL INOCULATION STUDIES IN THE TSWV-PEANUT
PATHOSY STEM .............. ...............89....


Introducti on ................. ...............89.................
Materials and Methods .............. ...............91....
Plant Culture ................. ...............91.................

Inoculum Preparation .............. ...............92....

Sap Inoculation ................. ...............92.................
Description of Tests .................. ........... ... ..... .. ...... .............9
Test 1: Effect of elapsed time from preparation to inoculation on infection

frequency ................ .......... .. ... ... .... ........9
Test 2: Determining the importance of amount of rubbing on infection rate ..........93
Test 3: Evaluating the influence of inoculum concentration on infection rate ........93

Imposing Treatments ................. ...... ...... ... ............ .............9
Evaluation of Inoculated Plants and Analysi s of Data ................ ................ ...._.94
Re sults........._... _...... ._ ._ ...............96....
Test 1 .............. ...............96....
Test 2 .............. ...............98....
Test 3 .............. ...............99....
Discussion ........._... ...... ._ ._ ...............100...
Test 1 .............. ...............100....
Test 2 .............. ...............101....
Test 3 .............. ...............102....
Conclusions............... ..............10


LIST OF REFERENCES ........._... ...... ._._ ...............106...


BIOGRAPHICAL SKETCH ........._... ......___ ...............112....









means. The random terms were assumed to have different variance structures: diagonal G

structure for blocks (no correlations among blocks from different cells) and unstructured

correlation G structure for the genotypes (there is covariance among genotypes between cells)

while the error variance was described by specifying an Identity R structure for each Cell.

v) Bivariate analysis for different assessment dates

For multivariate linear mixed methods, measurements from different environments are

treated as different variable with different variance and covariance structures which are

simultaneously estimated using the REML approach (Schaeffer & Wilton, 1978). Consequently,

the main weakness of univariate methods (i.e. heterogeneous variances) is properly addressed

(Lu, 2001).

A bivariate analysis (a special case of multivariate) was used here to estimate the genetic

correlation between DIR by jointly analyzing two AD's from the same location.

In Marianna, DIRs at each AD were taken in at least two different PDs (see Table 1-1).

This also allowed testing the effects of PDs on both AD DIRs being analyzed bivariately. In the

case of Citra, each combination of AD DIRs being analyzed was taken only in one PD so no

testing for PD effect was possible.

The bivariate model for the Marianna data subset can be written as


y = (I, 0 Xp> Zp + (Iz 0 Zg) Ug + (Iz 0 Zpg) Upg + (Iz 0 Zpb) upb + e (Eq. 1-6)


where y =(yA,,,a>, A =b )gA ADb) U= AN )PgADb) U(pbA )pbADb ) and

e = (eAlua Aob In turn A'a = the vector containing: the arcsine(square root(DIR) for AD "a";

while T, is the 1 x p vector of a constant and fixed effect PD, X, is an p x n design matrix of full

column rank which associates observations with the appropriate combination of the fixed effect,

the u vectors (ugA~a "pgAN and upb ) are 1 x q vectors for the random terms variety, PD by









The source of resistance for each resistant parent was believed to be unique (D. Gorbet,

pers. comm.). AP-3 and DP-1 are related through a common ancestor, their grandparent

Florunner, which is extremely susceptible to TSWV (Culbreath et al., 1997; Tillman et al.,

2007).

AP-3 is a runner-type cultivar whose parents display no noticeable spotted wilt resistance

(Gorbet, 2007). DP-1 is also a runner-type cultivar and traces its resistance back to its

grandparent PI 203396, which has produced numerous lines with good resistance (D. Gorbet,

pers. comm.). This PI is a typical member of the hypogaea botanical variety. NC94002, traces its

resistance back to its parent PI 57663 8, which is an accession belonging to the hirsuta botanical

variety (D. Gorbet and T. Isleib, pers. comm.).

The resistant and susceptible parents were mated in the following combinations AP-3/

NemaTAM, NemaTAM/AP-3, NemaTAM/DP-1, NemaTAM/NC94002 and DP-1/NC94002.

The resulting Fl, Backcross, F2 and F3 pOpulations, together with their parents were field tested.

For each cross, 25 F 2:3 (F2-derived in F3) familieS with enough seed were randomly selected and

included in each test in at least two replications.

Field tests were conducted at the University of Florida Plant Science Research and

Education Unit in Citra, Florida on a Candler Sand (Hyperthermic, uncoated Typic

Quartzipsamments) during 2005 and 2006, at the North Florida Research and Education Center

near Marianna, Florida on a Chipola loamy sand (Loamy, kaolinitic, thermic Arenic

Kanhapludults) during 2006 and 2007 and at the North Florida Research and Education Center in

Quincy, Florida on an Orangeburg fine sandy loam (fine loamy, siliceous, thermic, Typic

Paleudults) during 2007. The tests details are shown in table 2-1.









The Animal Model utilizes all relationships among individuals in a test by using a

numerator relationship matrix (Mrode, 2005) thus accounting for most of the additive variance

which results in more accurate estimations of variance components and breeding values. The fact

that most of the individuals on each test were related to some extent, even when belonging to

different crosses, provided better estimates of the additive covariances than the use of traditional

heritability estimation methods (Henderson, 1976; Lynch and Walsh, 1998). Although similar

kinds of populations were used each year, the variance components were quite variable among

tests. This is sometimes the case, even when repeating tests few days apart (Chapter 1 of this

dissertation), doing them in a laboratory setting (Rapp & Juntila, 2001) or in the field (Finne et

al., 2000).

Spotted wilt epidemics are highly variable among locations and even from year to year at a

single location (Culbreath et al., 2003). The range of intensity of the epidemic observed among

the tests was certainly wide. Apparently the geographical location was more important than the

year, in accordance with the results described in the Chapter 1 of this dissertation.

Heritability estimates are influenced by the relative amount of total variation due to genetic

causes (Lynch and Walsh, 1998). The wide range of epidemic intensity probably accounted for a

large part of the variability in heritability estimates. Inaccurate heritability estimates are often

obtained when the frequency of a category in a polychotomous variable is very high across the

whole population of individuals being tested (Yang et al., 1998). In the present study most of the

individuals showed a reasonable dispersion in the score frequencies. The use of individual values

is known to provide better heritability estimates than plot means under most conditions when a

REML approach is used (Huber et al., 1994). Consequently, even when the heritability estimates

obtained here were quite variable, they are expected to be accurate. Heritability estimates in the










2. Noticeable shortening of internodes, plant size about 60% of a healthy one.

3. Marked shortening of intemodes, plant size about 40% of a healthy one, leaflets showing
signs of poor expansion.

4. Very marked shortening of internodes, plant size about 30% of a healthy one, leaflets
poorly unfolded.

5. Extreme shortening of intemodes, plant size about a 20% of a healthy one, plant shows
unfolded leaves in crowded limb tips. No leaf has been unfolded recently.

Foliar symptoms were related to damage in photosynthetic pigments as shown by the presence of

spots or mosaic patterns. Their incidence and severity were jointly assessed according to the

following ordered scale:

1. A few hardly noticeable (faint) foliar symptoms in a few leaves.
2. Noticeable (easily observable) foliar symptoms in a few branches.
3. Marked (very evident) foliar symptoms in most branches.
4. Very marked (covering most leaves) foliar symptoms in most branches.
5. Very marked foliar symptoms in all branches.

Tip death was assessed according to the symptom incidence on the plant, as follows:

1. One stem tip dead
2. Up to 25% of the tips dead
3. Up to 50% of the tips dead
4. Up to 75% of the tips dead
5. All tips dead

Leaf necrosis was assessed where necrotic lesions tend to coalesce forming patches of dead

tissue on the leaf. Its degree was scored as follows:

1. A few leaves with less than 1/10 of their surface necrotic.
2. Noticeable necrosis in a few branches.
3. Extended necrosis (up to 50% of the leaves affected) in most branches.
4. Very extended necrosis (up to 75% of the leaves affected) in most branches
5. Most leaves necrotic in all branches.

Yellowing was assessed as follows:

1. Leaves turn slightly yellow, no or very little leaf folding.
2. Leaves noticeably yellow, some leaf folding especially in the afternoon.
3. Leaves very yellow, noticeable leaf folding especially in the afternoon.
4. Grayish yellow leaves drooping in the afternoon, some reddish hue appears in the stems.











Table 1-5. Planting date effect and variance ratios for univariate analysis performed on
transformed spotted wilt disease intensity ratings at four planting dates assessed in
different times at Citra and Marianna, Florida in 2005.
P-value>F
Location Assessment date for planting GZe> 2e< 2pldSI.g 2e> 2 e>
date
Citra 70 0.006 1.1 0.1 1.1
90 0.023 1.0 0.1 0.4
112 0.054 1.1 0.1 1.7
132 0.031 1.4 0.0 0.9
Marianna 90 0.092 1.5 0.4 4.4
112 0.096 7.6 0.1 2.4
132 0.045 3.5 0.8 3.8
G2e>: biggest error variance. (r2e<: Smallest error variance. a2pld~g : planting date by genotype
interaction variance. G2g : genotypic vanance.


Table 1-6. Transformed spotted wilt disease intensity rating predicted values, standard errors
(SE) and standard errors of mean differences (SED) for at four planting dates
assessed in different times at Citra and Marianna, Florida in 2005.

Location Assessment date Planting Date Predicted Value SE SED

Citra 2 0.2199 0.0304
70 0.0223
4 0.2996 0.0300
1 0.2425 0.0298
90 0.0304
3 0.1502 0.0301
2 0.3692 0.0355
132 0.0249
3 0.4247 0.0370
Marianna 2 0.6755 0.1067
132 3 0.9298 0.1057 0.0899
4 0.8387 0.1091



Table 1-7. Location by planting date combination (cell) effect and variance ratios for
multivariate analysis performed on transformed spotted wilt disease intensity ratings
at different planting dates assessed at different times at Citra and Marianna, Florida in
2005.
Asesmn -vleF o 2 2 2 2 2 2 2 2 2
dasesete CeVall>fo 2 > e 0 b> 0 b 0b> ec 0 cg< 0 e< 0cg> 0e>
112 <0.001 6.9 285267592 0.7 1.6 20.2
132 <0.001 1.6 36619522 0.8 0.7 7.1
G2e> : biggest error variance. G2e<: Smallest error variance. G2b> : biggest block variance. (T2b< : Smallest
block variance. G2cg : cell by genotype interaction variance. a2g : genotypic variance.














S4/19/2005 m 5/4/2005 m 5/19/2005


6/6/2005


o C


cP39e ~e~p~S\oL 91e
"~~,`"~" arjGssL\oO\'d~c


ozC
o""


Figure 1-2. Spotted wilt disease intensity ratings at 112 days after planting for ten genotypes
planted at different dates at Marianna, Florida in 2005.









GENETICS OF TOMATO SPOTTED WILT VIRUS RESISTANCE INT PEANUT
(Arachis hypogaea L.)


















By

JORGE JAVIER BALDES SARI


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2008










1998). If the latter were the case, the genetic determinants of both types of symptoms would

have to be tightly linked, as thousands of F3 individuals were assessed and some recombinants

should have occurred thus reducing the correlation. Similar incidences among genotypes at 30

DAP suggested that thrips feeding preference was not an issue, in agreement with previous

findings (Culbreath et al., 1996; 1997).

In spite of the fact that the parents are inbred there was still a wide range of severity in the

symptoms suggesting either different inoculation times or differential progression of the disease

among plants within a genotype. Taking into account that at least a week is required to develop

symptoms (Hoffman et al., 1998) it seems clear that a potential period of three weeks of

inoculum exposure could cause a wide range of symptom severities very early in the season.

Incomplete penetrance of resistance to TSWV has been reported in tomato when

inoculated by thrips (Rosello et al., 2001). This could also explain the variable symptom severity

observed in the present study, particularly in the resistant parents. Whichever the case, the score

distribution didn't suggest the use of traditional Mendelian segregation analysis (Lynch and

Walsh, 1998) so a quantitative approach to analyze the symptoms scores as polychotomous

variables was necessary.

The heritability values varied noticeably among tests, which is usually the case when

calculating estimates even from similar populations (Nyquist, 1991; Lynch and Walsh, 1998).

However, the small standard errors suggested the estimates were rather precise.

In the present study three unrelated resistant genotypes, which represent the different

sources of resistance to spotted wilt known to date (D. Gorbet, pers. comm.) were used. By

crossing them to a susceptible parent, segregating populations with wide variability in spotted

wilt resistance were obtained.








I


m 5/10/2005 6/2/2005


0.6
u> 0.5
rr0.4
C

0.
0.


II


Figure 1-3. Spotted wilt disease intensity ratings at 112 days after planting for ten genotypes
planted at different dates at Citra, Florida in 2005.


,1 Ill









Location by Planting date (cell) effect

v) Multivariate analysis through G structure specification

The analysis using common combinations of planting dates and assessment dates between

both locations (referred herein as "cells") showed that the cell effect was highly significant at

both ADs (Table 1-7), with Citra showing smaller predicted values than Marianna (Table 1-8).

Cells of different location were always significanlty different. At 1 12 DAP, within-location

cells were only different for citra. At 132 DAP, within-location cells were only different for

Marianna.

Cells showed different variability with Marianna' s cells displaying the most variability

(data not shown). The ratio of the greatest error variance to the smallest (Marianna at PD4 : Citra

at PD 3) was 8.8 for 112AD while it was 1.2 (Marianna at PD 3 : Citra at PD 3) for 132AD (data

not shown).

The block variance had a negligible value at AD 112 for PD 3 and at AD132 for both PDs

at Citra (data not shown). In Marianna the block variance had some importance, but not in Citra.

The ratio of block variance for a cell to its error variance at 1 12AD was highest for PD 3 at

Marianna, while for 132AD the highest ratio was observed for PD 2 in Marianna.

The ratio of genotypic variance within a location to the error variance for that location

ranged from 0.7 to 20.2 with 5 out of 8 values above 2 (data not shown), showing that the

genotypic variance was the largest variability source for spotted wilt intensity DIRs.

Phenotypic and Genetic Correlations

The DIRs a plot received at different assessment dates were strongly correlated.

Phenotypic correlations were medium to high (Table 1-9). In Citra the values ranged wider and

below (0.43-0.89) those in Marianna (0.85-1).









reaction through the standardization of several factors. This should allow a better estimation of

the genotype' s true reaction to the virus. Consequently, the obj ective of this series of studies was

to determine the relative importance of age of inoculum, virus concentration in the inoculum, and

amount of rubbing during inoculation on the frequency of infection. A secondary objective was

to determine if there was an association between ELISA values and symptom expression. To

address these obj ectives three studies were conducted during the spring of 2005.

Materials and Methods

Plant Culture

Georgia Green, a widely grown runner market-type cultivar was used in all tests. This

cultivar displays some Hield resistance to TSWV (Culbreath et al., 1996) but it is susceptible

under artificial inoculation (Mandal et al., 2001).

One seed was planted in each 164 ml plastic container (Cone-tainer C10, Stuewe & Sons,

Corvallis, Oregon) containing all purpose professional growing mix consisting of Canadian

sphagnum peat moss 75 to 85%, perlite 15 to 20%, and vermiculite 5 to 10% (Berger Peat Moss,

Saint-Modeste, Quebec, Canada) and irrigated every other day with distilled water. No fertilizer

was added to the mix. Test plants were grown until inoculation was performed in a chamber

made of a shelf with fluorescent lights (Gro-Lux, Osram Sylvania, Danvers, Massachusetts) all

surrounded by a transparent plastic sheet. Conditions inside this chamber were 12-h light period

(12 klx intensity) and 230C min. and 340C max. Those seedlings with uniform size and vigor

were used for inoculation 12-14 days after planting. The average seedling height was variable

among tests ranging from fiye to seven cm, whereas the average number of fully expanded

leaves ranged from two to Hyve.










Murakami, M., M. Gallo-Meagher, D.W. Gorbet, and R.L. Meagher. 2006. Utilizing
immunoassays to determine systemic Tomato spotted wilt virus infection for elucidating field
resistance in peanut. Crop Protection 25(3):23 5-243.

Nagata, T., L. S. Boiteux, N. lizuka, and A.N. Dusi. 1993. Identification of phenotypic variation
of tospovirus isolates in Brazil based of tospovirus analysis and differential host response.
Fitopat. Brasil. 18: 425-430.

Ng, J.C.K., T. Tian, B.W. Falk. 2004. Quantitative parameters determining whitefly (Bemisia
tabaci) transmission of Lettuce infectious yellows virus and an engineered defective RNA.
Journal of Gen. Vir. 85(9):2697-2707

Northfield, T.D. 2005. Thrips competition and spatiotemporal dynamics on reproductive hosts.
Ph.D. Diss. University of Florida, Gainesville, FL.

Nyquist, W.E. 1991. Estimation of heritability and prediction of selection response in plant-
populations. Critical Rev. in Plant Sci. 10(3):235-322.

Parrella, G., P. Gognalons, K. Gebre-Selassie, C. Vovlas, and G. Marchoux. 2003. An update of
the host range of Tomato spotted wilt virus. J. Plant pathol. 85 (4): 227-264.

Pereira, M.J. 1993. Tomato spotted wilt virus in peanut (Arachis hypogaea L.): screening
technique and assessment of genetic resistance levels. M. S. Thesis. University of Florida,
Gainesville, FL.

Prins, M., M.M.H. Storms, R. Kormelink, P. De Haan and R. Goldbach. 1997. Transgenic
tobacco plants expressing the putative movement protein of Tomato spotted wilt tospovirus
exhibit aberrations in growth and appearance. Transg. Res. 6(2):245-251

Rapp, K, and O. Junttila, 2001. Heritability estimates of winter hardiness in white clover based
on field and laboratory experiments. Acta Agricult. Scan. Sect. B-Soil and Plant Sci. 50(3-
4):143-148

Resende, L.V., W.R. Maluf, A.D. Figueira, and J.T.V. Resende. 2000. Correlations between
symptoms and DAS-ELISA values in two sources of resistance against Tomato spotted wilt
virus. Brazilian J. Microbiol. 31(2):135-139

Rosello, S., B. Ricarte, M.J. Diez, and F. Nuez. 2001. Resistance to Tomato spotted wilt virus
introgressed from Lycopersicon peruvianum in line UPV 1 may be allelic to Sw-5 and can be
used to enhance the resistance of hybrids cultivars. Euphytica 1 19(2):3 57-3 67.

SAS Institute. 2000. Statistical Analysis Software for Windows. Version 8.1. SAS Institute,
Cary, NC.

Searle, S.R., G. Casella, and C.E. McCulloch. 1992. Variance Components. John Wiley & Sons,
New York.

Simmonds, N.W. 1979. Principles of Crop Improvement. Longman, London, UK.









Table 1-1. Layout of data collection for planting date studies of peanut in Marianna and Citra,
Florida in 2005.
Planting Assessment Dates (in Days After Planting)
Location
date 70 DAP 90 DAP 112 DAP 132 DAP
Citra la X
2 X X
3 X X Xb
4 X X
Marianna 1 X
2 X X Xc
3 X X Xc
4 Xd Xcd
The X indicates the cells in which spotted wilt damage was assessed. a Genotypes F43 5HO,
NemaTAM and NC94002 were not planted. b F435HO was already dug. 0 F435HO
and SunOleic 97R were already dug. d Replication one was discarded.


Table 1-2. Means and standard errors for tomato spotted wilt disease intensity ratings at different
planting dates at Citra and Marianna, Florida in 2005.
Location AD 70 AD 90 AD 112 AD132
Planting Mean S.E. Mean S.E. Mean S.E. Mean S.E.
date (n) (n) (n) (n)
Citra la 0.066 0.010
(2 1)
2 0.057 0.009 0.157 0.018
(30) (30)
3 0.048 0.008 0.123 0.016 0.179 0.019
(30) (30) (27) b
4 0.097 0.014 0.191 0.026
(30) (30)
Marianna 1 0.545 0.054
(30)
2 0.514 0.054 0.480 0.056 0.442 0.049
(30) (30) (24) 0
3 0.640 0.670 0.673
(30) 0.051 (30) 0.050 (24) c 0.047
4 0.632 0.548
(20)d 0.063 (14)cd 0.070
aGenotypes F435HO, NemaTAM and NC94002 were not planted.b F435HO was already dug.
F435HO and SunOleic 97R were already dug. d Replication one was discarded.









Florida on a Chipola loamy sand (Loamy, kaolinitic, thermic Arenic Kanhapludults), during the

summer of 2005.

Eight cultivars and two breeding lines with variable maturity, belonging to three market

groups (spanish, runner, virginia) and having varied response to TSWV were tested. Based on

previous research the genotypes were considered to have three different reactions to TSWV:

susceptible, moderately resistant (intermediate) and resistant. The susceptible group included F-

43 5HO (Gorbet, pers. comm.), NemaTAM (Simpson et al., 2003) and SunOleic 97R (Gorbet &

Knauft, 2000). Georgia Green (Branch, 1996) and ANorden (Gorbet, 2007a) constituted the

intermediate group while C-99R (Gorbet and Shokes 2002), NC94002 (Gorbet, pers. comm.),

DP-1 (Gorbet, 2003), AP-3 (Gorbet, 2007b) and Georgia-02C (Branch 2003) formed the

resistant group.

The planting window at both locations was divided so that a similar number of days

elapsed between successive planting dates. At Citra the planting dates were 03/29/2005,

04/19/2005, 05/10/2005 and 06/2/2005 while at Marianna they were 4/19/05, 5/04/05, 5/19/05

and 6/06/05.

The experimental design was a 2 x 4 x 10 (location x planting date x cultivar) factorial

which was planted in a randomized complete block, split-split-plot design, with location being

the whole-plot, planting date the sub-plot and cultivar the sub-sub-plot. Plots consisted of two

4.5 m long rows spaced 0.91 m apart. The planting density was 18 seeds/m. The first planting

date at Citra didn't include F43 5HO, NemaTAM or NC94002.

After sowing, plots were maintained according to commercial peanut production practices

for the region with fertilizer, herbicide, fungicide, insecticide and irrigation applied as

recommended by the University of Florida extension guidelines.












TABLE OF CONTENTS


page

ACKNOWLEDGMENTS .............. ...............4.....


LIST OF TABLES ............_...... .__ ...............7....


LIST OF FIGURES .............. ...............9.....


AB S TRAC T ............._. .......... ..............._ 1 1..


CHAPTER


1 EFFECT OF PLANTING DATE ON SPOTTED WILT EXPRESSION IN
DIFFERENTIALLY SUSCEPTIBLE PEANUT (ARACHIS HYPOGAEA L.)
CULT IVARS ............ ......__ ...............13....


Introducti on ............ ......_ ...............13....
M material and M ethod s ............ ......_ ...............15...
Field Trials............... ...............15.
Trait M easurements ............ ..... ._ ...............17...
Statistical Analysis .............. ...............17....
Variables and factors ............ _. ..... ...............18...

Anaylses performed............... ...............1
Linear mixed analyses ........._...... ................ 19...___. ....
Calculation of Genetic Correlations ...._._._.. ..... ..__... ....__._ ...........2
Calculation of Repeatability ........._...... ...............28...__........
Re sults.........._.. ... ... .._ ...............28.....
Location Effect .........._.... ...............29..__..........
Planting Date Effect ................. ......... ..............3
Location by Planting date (cell) effect .............. ...............31....
Phenotypic and Genetic Correlations ................ ...............31.......___....
Repeatability ........._... ...... ..... ...............32....
Discussion ........._... ...... ..... ...............33....
Conclusions............... ..............3


2 HERITABILITY AND BREEDING VALUES OF TSWV RESISTANCE INT
POPULATIONS FROM PEANUT (ARACHIS HYPOGAEA L.) CROSSES ................... ..46


Introducti on ........._...... ...............46..__..........
Material and Methods ........._...... ...............49...__........
Re sults........._...... .. ..._... .....__. .............5
Observed Variance Components .............. ...............57....
H eritability...................... ..............5
Phenotypic and Genetic Correlations ................ ...............58.......___....
Breeding Values (BLUPs) ............ ............ ...............59...
Generation BLUPs .............. ...............59....









BIOGRAPHICAL SKETCH

Jorge Javier Baldessari was born in San Francisco, Cordoba Province, Argentina, on April

3rd, 1967. After moving to Cordoba City, he graduated from high school and he enrolled in the

Universidad Nacional de Cordoba (UNC), graduating in 1992 with a degree in agricultural

engineering. He then received a graduate research fellowship from the UNC Science Secretariat

to work on chickpea breeding. In 1994 he took a position as a peanut breeder in the Manfredi

Experimental Station of the Instituto Nacional de Tecnologia Agropecuaria. While working as a

breeder, he started his M.Sc. studies, receiving in 2000 a M. Sc. degree in plant breeding from the

Facultad de Ciencias Agrarias of the Universidad Nacional de Rosario.

He started his doctoral studies in August 2004. Upon graduation he will return to his

breeding duties at Manfredi Experimental Station.









Heritability is a genetic parameter of paramount importance for efficient plant breeding but

no estimates have been published for resistance to TSWV in peanuts. To provide such estimates

and assess resistance sources, Hyve populations from three resistant and a susceptible parent were

Hield tested in five environments in Florida, USA. Approximately 36,300 total plants were

individually assessed three times for Hyve spotted wilt symptoms using a six level scale. Each

environment was individually analyzed using an Animal Model containing block, plot, additive

and non-additive terms. High phenotypic (0.80-0.93) and genetic (0.88-0.99) correlation

estimates between stunting and spots/mosaic were obtained. Individual-basis heritability

estimates showed a wide range (0.01-0.71) although values most frequently were in the low-

medium range. This suggests individual selection for resistance to spotted wilt should not be

applied in early generations within the tested populations. The resistant parents produced

populations with similar breeding values when crossed to the susceptible parent, while the

population from a cross between resistant parents exhibited the best breeding values for

resistance to spotted wilt.

A published inoculation method was used to study if inoculum age, viral concentration,

and extent of rubbing during inoculation affected the frequency of infection. Results showed that

neither number of rubbings nor inoculum concentration were important factors. Inoculum

showed better infectivity 10 minutes after preparation than at zero or twenty minutes after

preparation. Inoculum batch was an important factor; highlighting the fact that viral titer is

highly variable even when collected from similar plant tissues. The overall low infection rates

suggest that additional work is necessary for mechanical inoculation to be a reliable research

tool .









breeding lines available. Most of them trace their resistance back to two unrelated sources, PI

203395/6 (both PIs come from the same original accession) and PI 57663 8. The former, typical

hypogaea botanical variety members, are the spotted wilt resistance source of Georgia Green

(Branch 1996), Georgia 01R (2002), DP-1 (Gorbet and Tillman, 2008, in press) and many others.

Meanwhile, PI 57663 8 (a hirsuta botanical variety member) is the source of the resistance of

several breeding lines that had shown remarkable field resistance to spotted wilt (Culbreath et al.,

2005, D. Gorbet, pers. comm.).

The existence of a different mechanism of resistance between these two PIs has been

hypothesized (Culbreath et al., 2005). Nonetheless, it has been shown that resistance is unrelated

to vector non-preference or reproduction (Culbreath et al., 1996, 1997, 2000).

A third source of TSWV resistance has been observed in sisterline cultivars AP-3 (Gorbet,

2007) and Carver (Gorbet, 2006). In this case, the origin of the resistance is uncertain as neither

of the parents of these two cultivars is resistant (D. Gorbet, pers. comm.)

In tomato and pepper, maj or genes for resistance to TSWV have been described, having

different modes of action and penetrance (Rosello et al., 2001, Soler et al., 1998 Moury et al.,

1998).

In species where inherited resistance is conditional or ambiguous, with no clear or

consistent phenotypes, traditional genetic analyses become difficult (Lynch and Walsh, 1998;

Bruening, 2006) and the heritability becomes the most important information for the plant

breeder (Simmonds, 1979). The heritability (in its narrow sense) expresses which proportion of

the phenotypic variability can be transmitted from parent to offspring (Falconer & MacKay,

1996) and it is the main determinant of the expected response to selection (Hallauer & Miranda,

1988).









These scale effects cause the interaction term in the ANOVA to become significant (Eisen

and Saxton, 1983). A way to cope with this biological fact is to move away from the usual

approach of classical linear models towards a mixed linear models approach which provides the

flexibility of modeling data means, variances and covariances by specifying the correct structure

of variances and errors (Gilmour et al., 2006).

Genetic correlations among traits indicate the degree of change in one trait as a result of a

change in another trait (Zobel and Talbert, 1984). Estimates of type B genetic correlations are

also used as quantitative measures of genotype by environment interactions (Lu et al., 2001).

Type B is the genetic correlation of the same trait measured on the same individual at different

environments (Yamada, 1962).

Several methods can be used to estimate Type B genetic correlation. The simplest ones are

called generically "univariate" because by using univariate linear models they calculate genetic

correlations according different procedures. They are easy to calculate but can be biased if data

are severely unbalanced or variances are very different (Lu et al., 2001).

With the increase in computational power of computers, statistical software using restricted

maximum likelihood (REML) techniques has become widely available and provides a means to

calculate Type B genetic correlations referred generically as "multivariate analysis". These

methods can estimate genetic variances and covariances simultaneously using a REML approach

(Holland, 2006). For these methods, the traits being correlated (the same trait in different

environments) are handled with attention to their variance-covariance structure, thus solving the

main limitation of the univariate methods. The REML approach is better for handling unbalanced

data for the purpose of variance component estimation (Searle et al., 1992). Multivariate methods










Breeding values for the different generations of the five crosses tested seemed to suggest

additivity as the main mode of action in the determination of the resistance to spotted wilt. The

resistant parents produced populations with similar breeding values when crossed to the

susceptible parent. The population from a cross between resistant parents exhibited the best

breeding values for resistance to spotted wilt.

Based on the calculated heritability estimates, pedigree selection within the populations

used in this study should not put too much weight on individual selection in early generations

based on resistance to TSWV. More emphasis on including resistance as a part of a multi-trait

individual selective index with a corresponding moderate weighting seems recommendable.

Additionally, familial performance can provide surrogate estimations of an individual's real

resistance.









SxR (susceptible with resistant) crosses, those involving the most resistant parent (NC94002)

were the best at Marianna.

Reciprocal crosses between AP-3 and NemaTAM had very similar average F2 BLUPs in

both locations, implying no maternal effect.

The average BLUPs for all F3 pOpulations were similar to the ones for F2 pOpulations at

both locations. This suggests that the 25 F2 plants sampled in 2006 in each F2 pOpulation were

able to capture most of the variability present in that population.

The average BLUP for F2 and F3 pOpulations were usually intermediate to their parents,

suggesting additivity. The only exception was the F3 pOpulation for the RxR cross at Quincy

2007, which had a higher (worse) BLUP than its most susceptible parent (DP-1). However, this

could have been caused by the fact that the sampling of the F2 plants in the previous year didn't

reflect correctly the true genetic composition for that population.

Individual BLUPs

The percentage of individuals with BLUPs better than their best parent varied widely

depending on the cross and the test. In general and as expected (because of increasing additive

variance due to selfing), the comparison between F2 and F2:3 pOpulations showed the latter

having both higher individual BLUP variability (Figs. 2-14 and 2-15) and higher percentage of

individuals with BLUP superior to their best parent (Fig. 2-16).

Comparing both tests grown in 2007, it can be seen that in Quincy 2007, 30% and 48% of

F2 and F3 individuals (respectively) had better breeding values than NC94002 in the cross

between this line and DP-1 (Fig. 2-16). Surprisingly, the same cross in Marianna 2007 only

showed 6% and 3% (F2 and F3) Of individuals better than NC94002 and it failed to produce an

individual with breeding value better than NC94002 in the other three tests (data not shown). The

remaining cross involving NC94002 as a parent also showed a difference in the percentage of











2~2 2
S1 I M o

while the error variance was described by specifying an Identity R structure for each

Location.

This analysis also provided an estimate of Type B genetic correlation (bivariate REML

estimation) among DIR from different locations at each AD by modeling the G structure of the

interaction between locations and genotypes. The correlation among genotypes was obtained at

the genotypic rather than the phenotypic level because through the specification of a correlation

model applied to the G structure for the Genotype factor it' s possible to obtain a true estimation

of the genotypic correlation (Gilmour et al., 2006).

Estimates of Repeatability were also obtained at each Location.

iii) Univariate analysis for planting dates

This analysis was performed for each location separately to test the difference in the DIR

among PDs. For example, for Marianna, the DIRs obtained at AD 90 which were available only

for PDs 2 & 3, (Table 1-1) were analyzed and the difference between those PDs was tested.

The arcsine(square root(DIR) observed in a Location at an AD were modeled as


y = pU + X, T, + Zvu, + Z,,u,v + Zpb pb + e (Eq. 1-4)


where y is a vector containing the arcsine(square root(DIR), T, is the p x 1 vector of a


constant and fixed effect PD, X is an n x p design matrix of full column rank which associates

observations with the appropriate fixed effect level. The u vectors (u,, u,v and upb) arT q X 1

vectors for the random terms variety, PD by variety and PD by block effects while Z matrices (Z,

,Z~v and Zpb) arT H X q design matrices for those random effects mentioned above. The random










CO2
where COVJk is the estimated genotypic covariance between DIRs j and k and gJis

the estimated genotypic variance for the score obtained at ADj.

For the same trait, for example, a score obtained at ADj but assessed in different

experimental units (i.e. different PD or Location), a Type B correlation, model II (Yamada,

1962) was calculated by using the corresponding ratio of variances, depending on the nature of

the performed analysis (univariate or bivariate). If calculated from variance components obtained

through the univariate analyses described under sections i) Univariate analysis for Location or

iii) Univariate analysis for PDs, the Type B genetic correlation was estimated as


r, = (Eq. 1-9)



where "gis the estimated genotypic variance for the score obtained at that Location or PD


and "' is the corresponding interaction term between genotype and either Location or PD.

When calculated from bivariate or multivariate analyses, as shown under sections ii)

Bivariate analysis through G structure specification and iv) Multivariate analysis through G

structure specification, the Type B genetic correlation was estimated as

COV
gclc2
"In 2 2~ (Eq. 1-10)
Rel Rc2


where COVanclc is the estimated genotypic covariance between DIRs from locations one


and two and ~gc, is the estimated genotypic variance for the score obtained at Location number


































O 2008 Jorge J. Baldessari









CHAPTER 1
EFFECT OF PLANTINTG DATE ON SPOTTED WILT EXPRESSION INT DIFFERENTIALLY
SUSCEPTIBLE PEANUT (ARACHIS HYPOGAEA L.) CULTIVARS

Introduction

Peanut spotted wilt caused by Tomato spotted wilt virus (TSWV,

Bunyaviridae: Tospovirus) can cause significant losses in the Southeastern USA. Epidemics of

TSWV are highly variable among locations and even from year to year at a single location

(Culbreath et al., 2003).

Resistant cultivars are the single most important factor in the management of this disease

(Brown et al., 2007). Although lack of genotype by environment (GxE) interaction under a wide

range of conditions has been reported in the literature (McKeown et al., 2001), this is not the

norm. Tillman et al. (2007) and Murakami et al. (2006) reported significant genotype x year and

genotype x planting date interactions for final spotted wilt ratings. Culbreath et al. (1997) also

found genotype by year interaction for some locations but not others for final spotted wilt

intensity ratings. Similarly, in a two year study, Culbreath et al. (2005) found that genotypes

interacted with locations, although similar trends were observed across locations and years. As

more factors enter the equation, multiple interactions can sometimes occur (Hurt et al., 2005).

GxE interaction can be caused by genotype cross-over, a change in the relative ranking

across environments, due to the interaction among pathogenicity factors and resistance genes

(Develey-Riviere & Galiana, 2007). However, no genotype by isolate interaction has been

reported so far in the TSWV-peanut pathosystem (Mandal et al., 2006). GxE interaction can also

have statistical causes. Heterocedasticity (unequal variances) is a violation of the ANOVA

assumptions commonly found in biological experiments (Eisen and Saxton, 1983). This is caused

by the relationship between variance and mean which is usually referred to as "scale effects"

(Falconer and MacKay, 1996).









Locations with heavier epidemics, like Marianna, showed greater variability between years

than locations with lighter epidemics (Citra).

Geographical area seemed important as Citra (North-Central Florida) and Marianna

(Florida Panhandle) in 2006 were widely different in their epidemics whereas Marianna and

Quincy, only 50 miles apart, displayed similar epidemics in 2007.

The epidemic progression was also very different among tests. As opposing examples, at

Citra 2005 the epidemic changed very little after the first assessment date both in incidence (Fig

2-2) and severity (Fig. 2-3) while at Marianna 2007 there was a steady increase in both.

In general, the epidemic progression was rather different among genotypes, both in

incidence and severity. Susceptible genotypes (like NemaTAM) exhibited a faster rate of

increase in the percentage of symptomatic plants, especially in tests with the more severe

epidemics (Fig. 2-4, 2-5). Symptom severity progression was also different among genotypes

(Fig. 2-6). Most of the genotypes showed a similar average score for both stunting and foliar

symptoms (data not shown). In most of the genotypes a range of severity from zero to five was

observed at 30 DAP for both symptoms although the frequencies among genotypes were

different.

The scores for stunting or foliar symptoms showed unimodal and multimodal distributions

with varying degree of dispersion according to the environment and genotype. Even

homogeneous homozygous genotypes like the parents (Fig. 2-7) showed dispersion. As no

obvious dominant/susceptible threshold was observed among classes (Fig. 2-8), the study of

inheritance of resistance as a quantitative trait was pursued.












Table 1-8. Transformed spotted wilt disease intensity rating predicted values, standard errors
(SE) and standard errors of mean differences (SED) for location by planting date
combinations (cells) assessed at different times in Citra and Marianna, Florida in
2005. Each assessment date was analyzed separately.
Cell Assessment date Predicted Value SE SED
1 (Citra, P.D.3) 112 0.3349 0.0394
2 (Citra, P.D.4) 0.4337 0.0552
0.0981
3 (Marianna, P.D.3) 1.0253 0.1397
4 (Marianna, P.D.4) 0.9532 0.1108
1 (Citra, P.D.2) 132 0.3881 0.0401
2 (Citra, P.D.3) 0.4358 0.0325
0.0681
3 (Marianna, P.D.2) 0.7863 0.1014
4 (Marianna, P.D.3) 1.0599 0.0955
P.D. Planting Date.


Table 1-9. Phenotypic and genetic correlations among transformed spotted wilt disease intensity
ratings at four planting dates assessed in different times at Citra and Marianna,
Florida in 2005. Values in brackets are standard errors.
Phenotypic Correlation [SE] Genetic Correlation [SE]
Assessment Assessment
Location 112 132 112 132
Date Date
70 0.83 [0.08] 0.84 [0.07] 70 0.95 [0.06] 1 [0.08] a
Citra 90 0.43 [0.19] 0.75 [0.15] 90 .085 [0.22] 1 [0.21] a
112 0.71 [0.08] 112 0.99 [0.01]
90 0.85 [0.06] 0.87 [0.06] 90 1 [0.02] a 1 [0.03] a
Marianna
112 0.89 [0.09] 112 1 [0.01] a
a Genetic correlations were kept in the theoretical range by constraining the covariance matrix (Gilmour
et al. 2006)









In all the three tests, treatments were established by the combination of inoculum batch and

the level of the factor being tested (elapsed time, # of rubbings or inoculum dilution). Each

treatment was represented by 10 inoculated plants.

Both binary variables (appearance of systemic symptoms and TSWV ELISA detection)

were analyzed by Multiple Logistic Regression with a binomial distribution and logit link

function using SAS (SAS Institute, 2000). The applied model in Test I was

logit~-tc)=+ P: Time +P3 Batch (Eq. 3-1)

where n: is the probability of the plant being symptomatic or being ELISA positive

depending on the response variable being analyzed. The parameter Pi refers to the effect of the

"i" level of a factor (say Time) on the log odds that the dependent variable equals one of the two

possible outcomes, say "infected", controlling the levels of Batch (Agresti, 1996). "Time"

denotes the amount of elapsed time from inoculum preparation (0', 10' and 20') while "Batch"

denotes the inoculum batch used.

The adjusted models for Test 2 were:

logit~tc)=+ P3 Rubbings + P4 Batch; (Eq. 3-2)

and

logit~tc)=+ P4 Batch; (Eq. 3-3)

where "rubbings" denotes the number of rubbings applied during inoculation, being the

rest of the terms as described in Test 1.

The applied model for Test 3 was

logit~tc)=+ P,2 Dilution +p,4 Batch (Eq. 3-4)

where "Dilution" denotes the inoculum dilutions tested, being the rest of the terms as

described in Test 1.









Among the methods used for heritability estimation, the most used is the variance

component method because of its adaptability to different situations (Nyquist, 1991). Heritability

estimates depend not only on the genetic factors in the populations being analyzed but also on

the environment in which they are tested (Falconer & MacKay, 1996). In most situations, better

discrimination among genotypes is feasible by testing the populations in certain types of

environments (Hall and Van Sanford, 2003, Venuprasad et al., 2007). These environments can be

laboratory settings (Rapp & Juntila, 2001); locations (Finne et al., 2000) or even planting dates

(Chapter 1 of this dissertation).

The expression of disease resistance is usually scored as a polychotomous variable,

commonly called "ordered scale" (Connover, 1998). For this kind of trait, the unit of analysis

can be the individual observations in the native scale (Huber et al., 1994) or plot means

combined with a transformation such as arcsine or logistic (Holland et al., 1998). The use of

individual values is known to provide better heritability estimates than plot means under most

conditions when a REML approach is used in the linear mixed model context (Huber et al.,

1994). Although the typical analytical approach is to adjust a threshold model, this usually

yields similar results to REML variance estimation under the native scale under a wide range of

conditions (Banks et al., 1985; Westfall 1987). Inaccurate heritability estimates are sometimes

obtained when the frequency of a category in a polychotomous variable is very high across the

whole population of individuals being tested (Yang et al., 1998). Under these conditions, the

threshold model is only superior to estimates obtained from REML on the native scale if

incidences are extreme and heritabilities are low to medium (Lopes et al., 2000).









CHAPTER 2
HERITABILITY AND BREEDING VALUES OF TSWV RESISTANCE INT POPULATIONS
FROM PEANUT (ARACHIS HYPOGAEA L.) CROSSES

Introduction

Tomato spotted wilt virus (Bunyaviridae: Tospovirus TSWV) is a worldwide problem in

both greenhouse and field crops (German et al., 1992). Since the first report of TSWV affecting

peanut in Texas in the early '70s, it has become a maj or limiting factor to peanut production in

the US (Culbreath et al., 2003).

Spotted wilt disease symptoms develop at least a week after inoculation (Hoffman et al.,

1998) and can be seen quite early in the season (Chapter 1 of this dissertation). Typical

aboveground symptoms include concentric ringspots, mosaic patterns, stunting, varied degrees of

apical and leaf necroses and general chlorosis, also known as yellowing (Mitchell, 1996,

Culbreath et al., 2003, Demsky & Reddy, 2004,). The presence of some of these symptoms, like

stunting and yellowing, has been linked to the action of specific viral proteins (Prins et al 1997,

Koll & Biitner, 2000). However, the nature and severity of symptoms depends on the consensus

(predominant) RNA sequence in the virus population (Nagata et al., 1993; Mandal et al., 2006).

Generally, epidemic patterns vary significantly across locations, particularly if the

locations are distant and contain dissimilar agroecosystems (Culbreath, 2003; Groves et al.,

2003) with different crop species that react dissimilarly to TSWV (Kucharek et al., 2000) or

different weeds that influence thrips dynamics (Northfield, 2005). Whatever the environment, the

rate of progress of spotted wilt epidemics has been shown to be cultivar dependent in peanut

(Culbreath et al., 1997; Murakami et al., 2006; Chapter 1 of this dissertation). Resistant cultivars

are the most important factor in the management of the disease (Brown et al., 2007).

Spotted wilt resistance is an important goal for peanut breeders in Southeastern USA

(Gorbet, 1999, Tillman et al., 2007). At present there are several resistant peanut cultivars and










2-1 Sowing date, replication number and design of tests assessing performance against
spotted wilt in five peanut crosses in Florida. ......___ .... ... .__ ......_ .........6

2-2 Mean (S.D.) score for each spotted wilt symptom at 30, 60 and 120 days after
planting, at each of five field tests in which five peanut populations were evaluated
in Florida ................. ...............69.................

2-3 REML variance estimates for stunting and foliar symptoms caused by TSWV in
populations derived from five peanut crosses tested at Citra, Florida in 2005 and
2006, Marianna, Florida in 2006 and 2007 and Quincy, Florida in 2007..........................70

2-4 Heritability (S.E.) estimates for stunting and foliar symptoms caused by TSWV on
peanut populations from five crosses at different assessment dates in five tests in
Floirida. Estimates were calculated using univariate Animal Models. .........._.... .............72

2-5 Phenotypic and genetic correlation (S.E.) estimates between stunting and foliar
symptoms caused by TSWV on peanut populations from five crosses at different
assessment dates in five tests in Florida. Estimates were calculated using univariate
Animal Model s............... ...............72.

3-1 Effect of elapsed time after preparation of Tomato spotted wilt virus inoculum on the
number of peanut plants declared infected by visual examination or serological
m eans. ............. ...............105....

3-2 Maximum Likelihood Estimates for time and inoculum batch effects on artificial
inoculation of Georgia Green peanut ................. ...............105........... ...

3-3 Effect of inoculum dilution on the number of peanut plants declared infected by
ELI SA. ............ .................105..









present study were usually in the medium to low range, with the values increasing as the season

progressed. This was expected as the difference in symptom severity in genotypes with different

resistance tends to increase with time (Culbreath et at., 1997; Murakami et al., 2006) unless the

epidemic reaches a final intensity very early in the season (Chapter 1 of this dissertation). If the

additive variance and consequently the heritability tend to increase toward harvest time, selection

for resistant genotypes would be more effective when conducted closer to harvest (Hallauer &

Miranda, 1988). Heritability estimates for a trait is at the core of any individual multi-trait

selection index. Its magnitude determines the importance (weight) that' s assigned to the trait

while selecting individuals based on that index (Hallauer and Miranda, 1988).

In the University of Florida Peanut Breeding Program (UFPBP), performance of

segregating populations against TSWV has been assessed based on a "holistic" score assigned to

plots or plants in which all spotted wilt symptoms are considered (Gorbet, 1999, B. Tillman,

pers. comm.). Simmons (1979) pointed out that every breeder has in his/her mind a multi-trait

selective index but it is usually not put into writing. Taking into account the high genetic

correlation between the most frequent spotted wilt symptoms and the "workable" value of

heritability estimates here reported, it seems reasonable that the inclusion of spotted wilt

resistance as a part of this "unwritten" selective index could be the cause of the observable

improvement in the overall level of spotted wilt resistance in the breeding populations observed

in the UFPBP compared to older, but good performing genotypes from the pre-TSWV era like

Florunner, Sunrunner or SunOleic 97R (Culbreath et al., 2005; Tillman et al., 2007).

The variable reliability observed among and within tests could only be explained by a

variable importance of the environment, as the genetic structure and the type of genetic

relationships among individuals were quite similar in each test. When using full sib records (as is





Figure 2-12. Generation-mean best linear unbiased predictors for TSWV-induced starting in populations from five peanut crosses
tested at Quincy, Florida in 2007 (all reliabilities were above 0.9).









observed outcomes in the present work while using the method described by Mandal et al.


(2001).














AP-3 + DP-1 NC94002 NemaTAM



100

90

80

70

S60





2 0




30




30 DAP 60 DAP 120 DAP


Figure 2-4. Spotted wilt incidence in four peanut genotypes at three assessment dates at Marianna, Florida in 2006.










The random terms were assumed to have a unique G structure while the error variance was

described by specifying an Identity R structure for each Location.

This analysis also provided an estimate of Type B genetic correlation (Yamada II) among

DIR at certain AD from different Locations.

ii) Bivariate analysis through G structure specification

This analysis tested the differences in DIR among locations at certain AD and PD. The

data were modeled as


y = X Z + Z~,u, + Z~b lb + e (Eq. 1-3)


where y is a vector containing the arcsine(square root(DIR) for the corresponding AD, ZIis


the p x 1 vector of a constant and fixed effect Location, Xl is an n x p design matrix of full

column rank which associates observations with the appropriate combination of fixed effect. The

u vectors (ul, and ulb) arT q X 1 Vectors for the random terms Location by variety and Location by

block effects while Z matrices (Z~,, and Z~b) arT H X q design matrices for those random effects

mentioned above. The random effects and error are assumed to be independent Gaussian

variables with zero means. The random terms were assumed to have different variance

structures: diagonal G structure for blocks (no correlations among blocks from different

locations)


(0 0
va~block = 0 a 0O



and a correlation G structure for the genotypes (there is covariance among genotypes

between locations):









For Time O' and Batch 2, logit(AZ)=-0.3336+0 Time+(0.6253 Batch)

For Time 10' and Batch 3, logit(A2)=-0.3336+(0.7680 Time)+0 Batch

For Time 10' and Batch 1, logit(i2)=-0.3336+(0.7680 Time)+(-0.7978 Batch)

For Time 10' and Batch 2, logit(A2)=-0.3336+(0.7680 Time)+(0.6253 Batch)

For Time 20' and Batch 3, logit(A2)=-0.3336+(-0.9494 Time)+0 Batch

For Time 20' and Batch 1, logit(A2)=-0.3336+(-0.9494 Time)+(-0.7978 Batch)

For Time 20' and Batch 2, logit(i2)=-0.3336+(-0.9494 Time)+(0.6253 Batch)

These prediction equations can be expressed as the predicted probability of a plant being

infected. For example, using the third inoculum batch twenty minutes after it was prepared, the

predicted probability of a plant to be ELISA positive would be:

{exp[-0.3336+(-0.9494 Time)+(0.6253 Batch)}/{1+ exp[-0.3336+(-0.9494 Time)+(0.6253

Batch)}= 0.34

Similarly, the predicted probabilities for the other treatments were:

For Time O' and Batch 3,

{exp[-0.3336+(0 Time)+(0 Batch)}/{1+ exp[-0.3336+(0 Time)+(0 Batch)}= 0.42

For Time O' and Batch 1,

{exp[-0.3336+(0 Time)+(-0.7978 Batch)}/{1+ exp[-0.3336+(0 Time)+(-0.7978 Batch)}=

0.24

For Time O' and Batch 2,

{exp[-0.3336+(0 Time)+(0.6253 Batch)}/{1+ exp[-0.3336+(0 Time)+(0.6253 Batch)}=

0.57

For Time 10' and Batch 3,









showing that the theoretical upper limit for the heritability of the intensity was quite high. In

general, the repeatabilities at Citra were smaller than at Marianna.

Discussion

Some locations tend to have stronger spotted wilt epidemics than others (Culbreath et al.,

2003). Marianna has usually displayed severe epidemics since the late '90s (Tillman et al., 2007;

Culbreath et al., 2005) while spotted wilt is a lesser problem at Citra (Gorbet, pers. comm.). The

epidemics observed in 2005 followed a similar trend, with spotted wilt being consistently more

prevalent in every PD and AD in Marianna. This was reflected not only in the final DIR but also

on the fact that they peaked earlier in the season in every planting date.

The epidemic seemed to almost reach its peak at Marianna at around 90 days after planting

as there was little disease progression afterwards. This rapid establishment of near maximal DIR

at such an early assessment date is uncommon in the literature. Even the worse epidemics

reported so far have reached final intensities at later assessment dates (Murakami et al., 2006,

Culbreath et al., 1997). This was especially important because early TSWV infection typically

leads to more damage inflicted by the virus (Culbreath et al., 2003). In contrast to Marianna, the

progression of the epidemic at Citra showed a typical steady increase similar to mild epidemics

reported elsewhere (Murakami et al., 2006, Culbreath et al., 1997).

In the present study, the level of spotted wilt intensity recorded for well known genotypes

resembled those of severe epidemics described by some authors (Culbreath et al., 2005, Tillman

et al., 2007). These authors reported intensity ratings above 0.8 for susceptible genotypes such as

SunOleic 97R and variable ratings for the moderately susceptible Georgia Green. These values

are in agreement with ours, suggesting that final intensity ratings were in the expected range for

a wide array of genotypes of varying resistance levels.





















R = reliability of BLUP ii 9


,=0.41

R-0.7 7,R=0.94


R=0 .63 R=0 .89
R=0.44
~R=0.40


R= .6 R =0.44 R=09 =.8
R.=0.92 R=0.89

R=0 .61

R=0.44


- 4- DP-1 AP3 A NC94002 -g- NemaTAM


Citra '05


Citra '06


Marianna '06


Marianna '07


Quincy '07


Figure 2-10. Best linear unbiased predictors (relative to test average) for TSWV-induced stunting in parents of five peanut crosses
assessed in five field tests in Florida.









in Marianna 2006 and Quincy 2007 where the estimated average values for both spotted wilt

symptoms were smaller than the observed ones. However, in Citra 2006 and Marianna 2007 the

agreement between observed and estimated averages was very good.

Heritability

Heritability estimates varied widely among tests (Table 2-4). The highest heritability was

observed at Citra 2005 where only F2 pOpulations and AP-3 were grown. In the 2006 tests, the

heritability was extremely low whereas in the 2007 tests the estimates were close to 0.3.

Heritability among assessment dates within a test tended to increase with time, but the

extent of increase depended on the test. The only exception to this trend was the heritability of

foliar symptoms at Citra 2005 which diminished as the season progressed.

Heritability estimates were quite precise as suggested by the small standard errors.

Phenotypic and Genetic Correlations

In some assessment dates the calculation of correlations was hampered by estimation

problems such as lack of parameter convergence in the REML process or singularity in the data

matrix. In other cases the REML estimate was calculated by bounding it within its theoretical

space so no standard error was available.

Correlations between stunting and foliar symptoms were high in value and with small

standard errors in every test and assessment date (Table 2-5). Phenotypic correlations ranged

from 0.80 to 0.93 with the test at Citra 2005 showing the lowest values in each assessment date.

Correlation coefficients from tests in 2007 were very similar. Genotypic correlations were higher

than the phenotypic ones, ranging from 0.88 to 0.99. In the second and third assessment dates,

three tests (Citra 2005, and Marianna and Quincy 2007) showed almost perfect genetic

correlation.









can also apply constraints to estimates of genetic variances and covariances so that estimates of

genetic correlation remains within the theoretical parameter space (Lu et al., 2001).

Type B genetic correlations range between 0 and 1. Its value indicates the correlation of

genotypic performance across sites, thus assessing GxE interaction. High values indicate reduced

GxE interaction and high genotypic determination for the trait studied (Lu et al., 2001, Lynch

and Walsh, 1998).

An additional rough estimate regarding the importance of genotype in the observed

performance across environments is the repeatability, which is the intraclass correlation among

observations of the same trait at different moments. It provides the theoretical upper limit for the

heritability (Falconer & Mackay, 1996).

For spotted wilt resistance a better distinction of the genotypic performance has usually

been observed under strong epidemics (Culbreath et al., 2003). A better separation of a wide

array of responses allows more accurate estimations of repeatability (Betran et al., 2006).

In order to better understand the nature of the GxE interaction frequently reported in the

TSWV-pathosystem, the present study pursued the following obj ectives: 1) To ascertain the

importance of planting date and location as determining factors of spotted wilt epidemic

intensity, 2) To evaluate the consistency in the performance of an array of genotypes with

contrasting spotted wilt resistance assessed at different times, and 3) To provide an estimation of

how much of the genotypic consistency can be ascribed to genetic causes.

Material and Methods

Field Trials

Two field tests were conducted at the University of Florida Plant Science Research and

Education Unit in Citra, Florida on a Candler Sand (Hyperthermic, uncoated Typic

Quartzipsamments) and at the North Florida Research and Education Center near Marianna,





m Obs. Stunting Estim. Stunting


O Obs. Foliar symptoms a Estim. Foliar symptoms


Quincy '07


Marianna '06


Marianna '07


Citra '06


Figure 2-9. Observed vs. estimated overall severity in stunting and foliar symptoms caused by TSWV on Hyve peanut crosses in five
Hield tests in Florida.


Citra '05










and genotype) were only performed for two different ADs and not all the PDs. Simpler analyses

were also performed according to data availability. In most of the analyses, the effects of fixed

factors (location, PD or their interaction) were evaluated. When statistical significance was

below 5%, predicted marginal means and overall standard errors were obtained as suggested by

Gilmour et al. (2004). Additionally, different types of genetic correlation were obtained for DIR

from different ADs in every analysis.

Linear mixed analyses

The data analysis began by building simple linear mixed models (including genotypes plus

either Location or PD) to test for differences in levels of the fixed effects and then progressed

towards more complex models containing location, planting date and genotype at the same time.

The general progression in complexity followed this order:

i) Univariate analysis for location. It studied, for each PD and AD, the effect of

Locations.

ii) Bivariate analysis through G structure specification. It applied a bivariate

model containing location and genotype with specification of G structures by means of which the

correlation between DIR at each location was estimated. This analysis was performed for each

PD separately.

iii) Univariate analysis for PDs. It studied, for each location and AD, the effect of

PDs.

iv) Multivariate analysis through G structure specification. This applied a

multivariate approach. The DIR from each Loc*PD combination (cell) was regarded as a

variable through the specification of the corresponding G structure. Consequently, a correlation

among the DIR was obtained.










LIST OF FIGURES


FiMr page

1-1 Change of spotted wilt disease intensity ratings over time at Citra and Marianna,
Florida in 2005 (averaged over all planting dates and genotypes) .............. ..................43

1-2 Spotted wilt disease intensity ratings at 112 days after planting for ten genotypes
planted at different dates at Marianna, Florida in 2005. .................. ................4

1-3 Spotted wilt disease intensity ratings at 112 days after planting for ten genotypes
planted at different dates at Citra, Florida in 2005. ............. ...............45.....

2-1 Incidence of spotted wilt symptoms at 120 days after planting in populations from
five peanut crosses tested in five Florida environments. ............. .....................7

2-2 Percentage of plants displaying spotted wilt symptoms in five peanut crosses in five
field tests assessed at three dates in five Florida environments ................. ................ ...74

2-4 Spotted wilt incidence in four peanut genotypes at three assessment dates at
Marianna, Florida in 2006............... ...............76..

2-5 Spotted wilt incidence in four peanut genotypes at three assessment dates at
Marianna, Florida in 2007............... ...............77..

2-6 Observed stunting severity among different populations from five peanut crosses, at
three different dates at Marianna, Florida in 2007............... ...............78..

2-7 Frequency distributions for TSWV-induced stunting scores among parents of five
peanut crosses field tested at Marianna, Florida in 2007 ................. .......................79

2-8 Frequency distributions for TSWV-induced stunting scores among F2 populations
from four peanut crosses field tested at Marianna, Florida in 2007. ............. .................80

2-9 Observed vs. estimated overall severity in stunting and foliar symptoms caused by
TSWV on five peanut crosses in five field tests in Florida. ............. .....................8

2-10 Best linear unbiased predictors (relative to test average) for TSWV-induced stunting
in parents of five peanut crosses assessed in five field tests in Florida. .................. ..........82

2-11 Breeding values' reliability for TSWV-induced stunting among individuals in
segregating populations derived from five peanut crosses evaluated in five field tests....83

2-12 Generation-mean best linear unbiased predictors for TSWV-induced stunting in
populations from five peanut crosses tested at Quincy, Florida in 2007 (all
reliabilities were above 0.9). .............. ...............84....









individuals with better (smaller) BLUP than NC94002 observed between tests in 2007. In this

case, however, the difference between tests grown in 2007 was less drastic than in the cross DP-1

x NC94002. While the crosses between AP-3 and NemaTAM showed small percentages of good

individuals in Marianna, they failed to deliver better BLUPs than AP-3 in Quincy. Meanwhile,

the cross between NemaTAM and DP-1 didn't produce any individual with better breeding value

than DP-1.

Family BLUPs

In four out of five crosses there were individuals with better BLUPs than their best parent.

However, the proportion of families with average BLUPs for stunting better than their best

parent was rather small (Fig. 2-16). Most crosses produced one or two superior families in

Marianna 2007. The only exception to this general trend was the cross between two resistant

parents (DP-1/NC94002), which produced 13 out of 25 F2:3 familieS with better (smaller)

average BLUPs than the most resistant parent in Quincy 2007. However, the same cross only

produced one superior family in Marianna 2007.

Discussion

Locations have generally different epidemic patterns, particularly if they are significantly

distant and contain different agroecosystems (Culbreath et al., 2003; Groves et al., 2003).

Among the locations tested in this study, Citra typically does not have serious spotted wilt

epidemics even under agronomic practices that increase the likelihood of strong epidemics

(Tillman, pers. comm., Chapter 1 of this dissertation). One likely explanation is that the

agroecosystem of Marion County (where Citra is located) is dominated by warm-season grass

pastures while crops that are TSWV host species are a small minority (2002 Census of

Agri culture).

PAGE 1

1 GENETICS OF TOMATO SPOTTED WI LT VIRUS RESISTANCE IN PEANUT ( Arachis hypogaea L.) By JORGE JAVIER BALDESSARI A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008

PAGE 2

2 2008 Jorge J. Baldessari

PAGE 3

3 To Monica and Victoria.

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank Drs. Barry Tillm an, Da vid Wofford, Daniel Gorbet, Jane Polston, Albert Culbreath, Paul Lyrene and Kevin Ke nworthy for serving as dedicated committee members. My special appreciation goes to th e chairperson, Dr. Tillman, whose patience and support made this doctoral experience a very educational one. I am also very indebted to Dr. Dudley Hube r whom out of the goodness of his heart helped me solve some genetic and statistical conundrums. Heartfelt thanks go to Eric Ostmark, Jus tin McKinney, Mark Gomillion, Angus Branch, Serafin Aguirre and John Allen fo r their invaluable technical a ssistance. My deep appreciation goes to the whole Peanut crew at the NFREC at Marianna for th eir help during those long days under the hot summer sun. Fanchao Lis assist ance during the field ra ting is also greatly appreciated. Big thanks go to Eric and Bl anca Ostmark for their friendship and support. Finally, my wife Monica and daughter Victor ia deserve special mention. Their continual encouragement and unconditional love helped me through the path leading to this dissertation.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........7 LIST OF FIGURES.........................................................................................................................9 ABSTRACT...................................................................................................................................11 CHAP TER 1 EFFECT OF PLANTING DATE ON SPOTTED WILT EXPRESSION IN DIFFERENTIALLY SUSCEPTIBLE PE ANUT (ARACHIS HYPOGAEA L.) CULTIVARS ..........................................................................................................................13 Introduction................................................................................................................... ..........13 Material and Methods.............................................................................................................15 Field Trials................................................................................................................... ....15 Trait Measurements......................................................................................................... 17 Statistical Analysis.......................................................................................................... 17 Variables and factors................................................................................................18 Anaylses performed..................................................................................................18 Linear mixed analyses..............................................................................................19 Calculation of Genetic Correlations......................................................................... 26 Calculation of Repeatability..................................................................................... 28 Results.....................................................................................................................................28 Location Effect................................................................................................................29 Planting Date Effect........................................................................................................30 Location by Planting date (cell) effect............................................................................ 31 Phenotypic and Genetic Correlations.............................................................................. 31 Repeatability....................................................................................................................32 Discussion...............................................................................................................................33 Conclusions.............................................................................................................................36 2 HERITABILITY AND BREEDING VAL UES OF TSWV RESISTANCE IN POPULATI ONS FROM PEANUT (A RACHIS HYPOGAEA L.) CROSSES..................... 46 Introduction................................................................................................................... ..........46 Material and Methods.............................................................................................................49 Results.....................................................................................................................................55 Observed Variance Components.....................................................................................57 Heritability................................................................................................................... ....58 Phenotypic and Genetic Correlations.............................................................................. 58 Breeding Values (BLUPs)...............................................................................................59 Generation BLUPs................................................................................................... 59

PAGE 6

6 Individual BLUPs.....................................................................................................60 Family BLUPs..........................................................................................................61 Discussion...............................................................................................................................61 Conclusion..............................................................................................................................67 3 ARTIFICIAL INOCULATION STUDIES IN THE TSWV-PEANUT PATHOSYSTEM ...................................................................................................................89 Introduction................................................................................................................... ..........89 Materials and Methods...........................................................................................................91 Plant Culture....................................................................................................................91 Inoculum Preparation...................................................................................................... 92 Sap Inoculation................................................................................................................92 Description of Tests......................................................................................................... 92 Test 1: Effect of elapsed time from preparation to inoculation on infection frequency ..............................................................................................................92 Test 2: Determining the importance of a mount of rubbing on infection rate..........93 Test 3: Evaluating the influence of inoc ulum concentration on infection rate........ 93 Imposing Treatments....................................................................................................... 93 Evaluation of Inoculated Plan ts and A nalysis of Data.................................................... 94 Results.....................................................................................................................................96 Test 1...............................................................................................................................96 Test 2...............................................................................................................................98 Test 3...............................................................................................................................99 Discussion.............................................................................................................................100 Test 1.............................................................................................................................100 Test 2.............................................................................................................................101 Test 3.............................................................................................................................102 Conclusions...........................................................................................................................103 LIST OF REFERENCES.............................................................................................................106 BIOGRAPHICAL SKETCH.......................................................................................................112

PAGE 7

7 LIST OF TABLES Table page 1-1 Layout of data collection for planting date studies of peanut in Marianna and Citra, Florida in 2005. ............................................................................................................... ...38 1-2 Means and standard errors for tomato spotted wilt dis ease intensity ratings at different planting dates at Citra and Marianna, Florida in 2005........................................ 38 1-3 Location effect and variance ratios for bi variate analysis perform ed on transformed spotted wilt disease intensity ratings at three planting dates a ssessed in different times at Citra and Marianna, Florida in 2005.................................................................... 39 1-4 Transformed spotted wilt disease intensity rating predicted values, standard errors (SE) and standard errors of m ean differences (SED) for planting dates assessed at different times at Citra a nd Marianna, Florida in 2005..................................................... 39 1-5 Planting date effect and variance ratios for univariate analysis perform ed on transformed spotted wilt disease intensity ratings at four planti ng dates assessed in different times at Citra a nd Marianna, Florida in 2005..................................................... 40 1-6 Transformed spotted wilt disease intensity rating predicted values, standard errors (SE) and standard errors of m ean diff erences (SED) for at four planting dates assessed in different times at Ci tra and Marianna, Florida in 2005...................................40 1-7 Location by planting date combination (cell) effect and variance ratios fo r multivariate analysis performed on transformed spotted wilt disease intensity ratings at different planting dates a ssessed at different times at Citra and Marianna, Florida in 2005...............................................................................................................................40 1-8 Transformed spotted wilt disease intensity rating predicted values, standard errors (SE) and standard errors of m ean differences (SED) for location by planting date combinations (cells) assessed at different times in Citra and Marianna, Florida in 2005. Each assessment date was analyzed separately....................................................... 41 1-9 Phenotypic and genetic correlations among transfor med spotted wilt disease intensity ratings at four planting dates assessed in different times at Citra and Marianna, Florida in 2005. Values in br ackets are standard errors..................................................... 41 1-10 Type B genetic correlations and [standard errors] for transformed spotted wilt disease intensity ratings at f our planting dates assessed in different tim es in Citra and Marianna, Florida in 2005. a...............................................................................................42 1-11 Entry-mean repeatability estimates a nd their [standard errors] for transform ed spotted wilt intensity ratings at three planti ng dates assessed three times in Citra and Marianna, Florida in 2005. Values for each location are separated by a slash, Citra being on the left and Marianna on the right....................................................................... 42

PAGE 8

8 2-1 Sowing date, replication number and de sign of tests assessing perform ance against spotted wilt in five pean ut crosses in Florida.....................................................................69 2-2 Mean (S.D.) score for each spotte d wilt sym ptom at 30, 60 and 120 days after planting, at each of five field tests in wh ich five peanut populations were evaluated in Florida............................................................................................................................69 2-3 REML variance estimates for stunting and foliar symptoms caused by TSWV in populations derived from five peanut crosses tested at Citra, Florida in 2005 and 2006, Marianna, Florida in 2006 and 2007 and Quincy, Florida in 2007.......................... 70 2-4 Heritability (S.E.) estimates for stun ting and foliar sym ptoms caused by TSWV on peanut populations from five crosses at di fferent assessment dates in five tests in Floirida. Estimates were calculated using univariate Animal Models............................... 72 2-5 Phenotypic and genetic correlation (S.E .) estim ates between stunting and foliar symptoms caused by TSWV on peanut populations from five crosses at different assessment dates in five tests in Florida. Estimates were calcul ated using univariate Animal Models...................................................................................................................72 3-1 Effect of elapsed time afte r preparation of Tom ato spo tted wilt virus inoculum on the number of peanut plants declared infect ed by visual examination or serological means...............................................................................................................................105 3-2 Maximum Likelihood Estimates for time and inoculum batch effects on artificial inoculation of Geor gia Green peanut. .............................................................................. 105 3-3 Effect of inoculum dilution on the number of peanut plants declared infected by ELISA. .............................................................................................................................105

PAGE 9

9 LIST OF FIGURES Figure page 1-1 Change of spotted wilt disease intensity rating s over time at Citra and Marianna, Florida in 2005 (averaged over a ll planting dates and genotypes)....................................43 1-2 Spotted wilt disease intensity ratings at 112 days after planting for ten genotypes planted at different dates at Marianna, Florida in 2005. .................................................... 44 1-3 Spotted wilt disease intensity ratings at 112 days after planting for ten genotypes planted at different dates at Citra, Florida in 2005. ........................................................... 45 2-1 Incidence of spotted wilt symptoms at 120 days after planting in populations from five peanut crosses tested in five Florida environments....................................................73 2-2 Percentage of plants displa ying spotted wilt symptom s in five peanut crosses in five field tests assessed at three dates in five Florida environments......................................... 74 2-4 Spotted wilt incidence in four peanut genotypes at three assessm ent dates at Marianna, Florida in 2006..................................................................................................76 2-5 Spotted wilt incidence in four peanut genotypes at three assessm ent dates at Marianna, Florida in 2007..................................................................................................77 2-6 Observed stunting severity among different populations from five peanut crosses, at three different dates at Ma rianna, Florida in 2007.............................................................78 2-7 Frequency distributions for TSWV-induced stunting scores among parents of five peanut crosses field tested at Marianna, Florida in 2007. .................................................. 79 2-8 Frequency distributions for TSWV-indu ced stunting scores among F2 populations from four peanut crosses field test ed at Marianna, Florida in 2007.................................. 80 2-9 Observed vs. estimated overall severity in stunting and foliar sym ptoms caused by TSWV on five peanut crosses in five field tests in Florida............................................... 81 2-10 Best linear unbiased predic to rs (relative to test aver age) for TSWV-induced stunting in parents of five peanut crosses asse ssed in five field tests in Florida.............................82 2-11 Breeding values reliability for TSWV -induced stunting am ong individuals in segregating populations derived fr om five peanut crosses eval uated in five field tests....83 2-12 Generation-mean best linear unbiased predictors for TSWV-induced stunting in populations from five peanut crosses tested at Quincy, Florida in 2007 (all reliabilities were above 0.9)............................................................................................... 84

PAGE 10

10 2-13 Generation-mean best linear unbiased predictors for TSWV-induced stunting in populations from five peanut crosses te sted at Marianna, Florida in 2007 (all reliabilities were above 0.9)............................................................................................... 85 2-15 Variability of best linea r unbiased predictors for TSW V-induced stunting in the F2 and F3 generations of five peanut crosses tested at Quincy, Florida in 2007.................... 87 2-16 Percentage of individuals and (number of F3 families) displaying BLUPs for TSWVinduced stunting above their be st parent, in each of five peanut crosses tested at Marianna and Quincy, Florida in 2007.............................................................................. 88

PAGE 11

11 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy GENETICS OF TOMATO SPOTTED WI LT VIRUS RESISTANCE IN PEANUT (Arachis hypogaea L.) By Jorge J. Baldessari August 2008 Chair: Barry L. Tillman Major: Agronomy Tomato spotted wilt virus (TSWV, Bunyaviridae:Tospovirus) is a major peanut pathogen in the USA. Its management involves, among other factors, the use of resistant cultivars and recommended planting dates. Ten genotypes with varied degrees of resistance were field tested in two locations and four planting dates with the following objectives: 1) to ascertain the importance of planting dates and location as determining factors of spotted wilt epidemic intensity, 2) t evaluate the consistency in the performance of an array of genotypes with contra sting spotted wilt resistance assessed at different times, and 3) to provide an estima tion of how much genotypic consistency can be ascribed to genetic causes. Results indicated that location was a significant factor in determining the spotted wilt damage, while planting date was significant only under a light epidemic or late in the season under a heavy epidemic. The hi gh correlation between assessment dates implied that genotypic performance was perceived early and differences persisted until harvest. High Type B genetic correlation and repeatability suggested a strong genetic determination of resistance.

PAGE 12

12 Heritability is a genetic parameter of param ount importance for efficient plant breeding but no estimates have been published for resistance to TSWV in peanuts. To provide such estimates and assess resistance sources, five populations from three resistant and a susceptible parent were field tested in five environments in Florid a, USA. Approximately 36,300 total plants were individually assessed three times for five spotte d wilt symptoms using a six level scale. Each environment was individually analyzed using an Animal Model containing block, plot, additive and non-additive terms. High phenotypic (0.80-0 .93) and genetic (0.88-0.99) correlation estimates between stunting and spots/mosaic were obtained. Individua l-basis heritability estimates showed a wide range (0.01-0.71) altho ugh values most frequently were in the lowmedium range. This suggests individual select ion for resistance to spotted wilt should not be applied in early generations within the tested populations. The resi stant parents produced populations with similar breeding values when crossed to the susceptible parent, while the population from a cross between resistant pare nts exhibited the best breeding values for resistance to spotted wilt. A published inoculation method was used to st udy if inoculum age, viral concentration, and extent of rubbing during inocul ation affected the frequency of infection. Results showed that neither number of rubbings nor inoculum concen tration were important factors. Inoculum showed better infectivity 10 mi nutes after preparation than at zero or twenty minutes after preparation. Inoculum batch was an important factor; highlighting the fact that viral titer is highly variable even when collected from sim ilar plant tissues. The overall low infection rates suggest that additional work is necessary for mech anical inoculation to be a reliable research tool.

PAGE 13

13 CHAPTER 1 EFFECT OF PLANTING DATE ON SPOTTED W ILT EXPRESSION IN DIFFERENTIALLY SUSCEPTIBLE PEANUT (ARACHIS HYPOGAEA L.) CULTIVARS Introduction Peanut spotted wilt c aused by Tomato spotted wilt virus (TSWV, Bunyaviridae:Tospovirus ) can cause significant losses in the Southeastern USA. Epidemics of TSWV are highly variable among lo cations and even from year to year at a single location (Culbreath et al., 2003). Resistant cultivars are the single most important factor in the manageme nt of this disease (Brown et al., 2007). Although lack of genotype by environment (GxE) interaction under a wide range of conditions has been repo rted in the literature (McKeown et al., 2001), this is not the norm. Tillman et al. (2007) and Murakami et al (2006) reported signific ant genotype x year and genotype x planting date interactions for final spotted wilt ratings. Culbreath et al. (1997) also found genotype by year interaction for some locations but not others for final spotted wilt intensity ratings. Similarly, in a two year study, Culbreath et al (2005) found that genotypes interacted with locati ons, although similar trends were obser ved across locations and years. As more factors enter the equation, multiple interactions can some times occur (Hurt et al., 2005). GxE interaction can be caused by genotype cross-over, a change in the relative ranking across environments, due to the interaction am ong pathogenicity factor s and resistance genes (Develey-Riviere & Galiana, 2007). However, no genotype by isolate in teraction has been reported so far in the TSWV-peanut pathosystem (Mandal et al., 2006). GxE interaction can also have statistical causes. Heterocedasticity (unequ al variances) is a violation of the ANOVA assumptions commonly found in biol ogical experiments (Eisen and Saxton, 1983). This is caused by the relationship between variance and mean whic h is usually referred to as scale effects (Falconer and MacKay, 1996).

PAGE 14

14 These scale effects cause the interaction term in the ANOVA to become significant (Eisen and Saxton, 1983). A way to cope wi th this biological fact is to move away from the usual approach of classical linear models towards a mixed linear models approach which provides the flexibility of modeling data means, variances an d covariances by specifying the correct structure of variances and errors (Gilmour et al., 2006). Genetic correlations among traits indicate the degree of change in one trait as a result of a change in another trait (Zobel and Talbert, 1984) Estimates of type B genetic correlations are also used as quantitative meas ures of genotype by environmen t interactions (Lu et al., 2001). Type B is the genetic correlation of the same tr ait measured on the same individual at different environments (Yamada, 1962). Several methods can be used to estimate Type B genetic correlation. The simplest ones are called generically univariate because by using uni variate linear models they calculate genetic correlations according different procedures. They are easy to calculate but can be biased if data are severely unbalanced or variances are very different (Lu et al., 2001). With the increase in computational power of com puters, statistical soft ware using restricted maximum likelihood (REML) techniques has become widely available and provides a means to calculate Type B genetic correlations referred generically as multivariate analysis. These methods can estimate genetic variances and covariances simultaneously using a REML approach (Holland, 2006). For these methods, the traits being correlated (the same trait in different environments) are handled with attention to thei r variance-covariance structure, thus solving the main limitation of the univariate methods. The REML approach is bette r for handling unbalanced data for the purpose of variance component estimation (Searle et al., 1992). Multivariate methods

PAGE 15

15 can also apply constraints to estimates of genetic variances and covariances so that estimates of genetic correlation remains within the theo retical parameter space (Lu et al., 2001). Type B genetic correlations range between 0 and 1. Its value indicat es the correlation of genotypic performance across sites, thus assessing GxE interaction. High values indicate reduced GxE interaction and high genotypic determinatio n for the trait studied (Lu et al., 2001, Lynch and Walsh, 1998). An additional rough estimate regarding the importance of genotype in the observed performance across environments is the repeatab ility, which is the intr aclass correlation among observations of the same trait at different moment s. It provides the theoretical upper limit for the heritability (Falconer & Mackay, 1996). For spotted wilt resistance a better distinction of the genotypic performance has usually been observed under strong epidemics (Culbreath et al., 2003). A better separation of a wide array of responses allows more accurate estima tions of repeatability (Betrn et al., 2006). In order to better understand the nature of the GxE interactio n frequently reported in the TSWV-pathosystem, the present study pursued the following objectives: 1) To ascertain the importance of planting date and location as determining factors of spotted wilt epidemic intensity, 2) To evaluate the consistency in the performance of an array of genotypes with contrasting spotted wilt resistance assessed at different times, and 3) To provide an estimation of how much of the genotypic consistency can be ascribed to genetic causes. Material and Methods Field Trials Two field tests were conducted at the Univer sity of Florida Plant Scien ce Research and Education Unit in Citra, Florida on a Ca ndler Sand (Hyperthermic, uncoated Typic Quartzipsamments) and at the North Florida Re search and Education Center near Marianna,

PAGE 16

16 Florida on a Chipola loamy sand (Loamy, kaolinitic thermic Arenic Kanhapludults), during the summer of 2005. Eight cultivars and two breedi ng lines with variable maturity, belonging to three market groups (spanish, runner, virginia ) and having varied response to TSWV were tested. Based on previous research the genotypes were considered to have three different reactions to TSWV: susceptible, moderately resistant (intermediate) and resistant. The su sceptible group included F435HO (Gorbet, pers. comm.), NemaTAM (Simpson et al., 2003) and SunOleic 97R (Gorbet & Knauft, 2000). Georgia Green (Branch, 1996) an d ANorden (Gorbet, 2007a) constituted the intermediate group while C-99R (Gorbet and Shokes 2002), NC94002 (Gorbet, pers. comm.), DP-1 (Gorbet, 2003), AP-3 (Gorbet, 2007b) and Georgia-02C (Branch 2003) formed the resistant group. The planting window at both locations was di vided so that a similar number of days elapsed between successive planting dates. At Citra the planting dates were 03/29/2005, 04/19/2005, 05/10/2005 and 06/2/2005 wh ile at Marianna they we re 4/19/05, 5/04/05, 5/19/05 and 6/06/05. The experimental design was a 2 x 4 x 10 (loc ation x planting date x cultivar) factorial which was planted in a randomized complete bloc k, split-split-plot desi gn, with location being the whole-plot, planting da te the sub-plot and cult ivar the sub-sub-plot. Plots consisted of two 4.5 m long rows spaced 0.91 m apart. The planting density was 18 seeds/m. The first planting date at Citra didnt includ e F435HO, NemaTAM or NC94002. After sowing, plots were maintained according to commercial peanut production practices for the region with fertilizer, herbicide, f ungicide, insecticide and irrigation applied as recommended by the University of Florida extension guidelines.

PAGE 17

17 Trait Measurements Natural TS WV-infected thrips populations were relied upon to cause spotted wilt epidemics. Although asymptomatic infections of TS WV in peanut have been reported (Culbreath et al., 1992) all discussion about s potted wilt incidence in the present study refers to symptomatic plants. A disease intensity rating (DIR ) representing a combination of incidence and severity was calculated by counting the number of foci of plants severely affected by spotted wilt for each plot divided by the number of potential foci (Culbreath et al., 1997). A focus represented 0.31 m or less of linear row with plants severely stunted, killed, or show ing intense chlorosis due to TSWV. Strong reduction (<50% than healthy) in height or width of the peanut row was required for a row portion to be considered severely aff ected with regard to st unting and subsequently declared as a focus. If a portion showed less than severe reductions, it was counted as one half of a focus. The DIR were measured at the four assessment dates (ADs) observed in Table 1-1, expressed as days after planting (DAP), along with other details. Samples were taken from symptomatic plants and tested by DAS-ELISA (SRA 30400/0096 kit, AGDIA, Elkhart, IN) to confirm TSWV infection. Statistical Analysis All the analyses carri ed out in this study em ployed linear mixed models which were performed using ASREML (Gilmour et al, 2006). In short, and following the notation presented by the authors, the software uses the following linear mixed model. If y denotes an n x 1 vector of observations the linear mixed model (lato sensu) can be expressed as

PAGE 18

18y = X + Zu + e (Eq. 1-1) where is the p x 1 vector of fixed effects, X is an n x p design matrix of full column rank which associates observations with the appropriate combination of fixed effects, u is the q x 1 vector of random effects, Z is the n x q desi gn matrix which associates observations with the appropriate combination of random effects, a nd e is the n x 1 vector of residual errors. The model (Eq. 1-1) is called a linear mixed mo del or linear mixed effects model. It is assumed: )(0 0)( 0 0 ~ R G N e u where the matrices G and R are functions of parameters and respectively. The parameter is a variance parameter which is usually referred to as the scale parameter. Both G and R matrices can assume different structures depending on th e inherent structure of the data. Its necessary, in consequence, to find the right matrix structur e so that a correct and meaningful analysis of th e data can be performed. Variables and factors The variable under study in all the analyses was the DIR. The ratings were transform ed to arcsine(square root(DIR)) to improve normality of the original variable. The factors included in the analyses were Genotype, Location and Planting Date (PD). Both Location and PD were considered fixed ef fects while the Genotype was considered random (regarded as a sample of the popul ation of genotypes that could be grown in Southeastern USA). Anaylses performed The fact that not all PD were evalu ated on all the AD caused imbalance in the design (Table 1-1). Consequently, anal yses involving the three factors at the same time (location, PD

PAGE 19

19 and genotype) were only performe d for two different ADs and not all the PDs. Simpler analyses were also performed according to data availabilit y. In most of the analyses, the effects of fixed factors (location, PD or their interaction) were evaluated. Wh en statistical significance was below 5%, predicted marginal means and overall st andard errors were obtained as suggested by Gilmour et al. (2004). Additionally, different types of genetic correlation were obtained for DIR from different ADs in every analysis. Linear mixed analyses The data analysis began by building sim ple lin ear mixed models (including genotypes plus either Location or PD) to test for differences in levels of the fixed e ffects and then progressed towards more complex models containing location, planting date and genot ype at the same time. The general progression in comp lexity followed this order: i) Univariate analysis for location It studied, for each PD and AD, the effect of Locations. ii) Bivariate analysis through G structure specification. It applied a bivariate model containing location and genot ype with specification of G st ructures by means of which the correlation between DIR at each location was estim ated. This analysis was performed for each PD separately. iii) Univariate analysis for PDs. It studied, for each location and AD, the effect of PDs. iv) Multivariate analysis through G structure specification. This applied a multivariate approach. The DIR from each Loc* PD combination (cell) was regarded as a variable through the spec ification of the corresponding G stru cture. Consequently, a correlation among the DIR was obtained.

PAGE 20

20 v) Bivariate analysis for different ADs. This was designed to address the correlation among DIR from different ADs for every available Location by PD combination (see Table 1-1). Additionally, the e ffect of PD on each DIR in the pair being analyzed, was determined. A detailed explanation of each analysis follows: i) Univariate analysis for location This analy sis was performed for each planting da te separately to test the difference in the DIR between locations. For example, for PD 2, the only score available at both locations was the one obtained at 132 AD. Thus, this data subset was analyzed a nd from that the differences among Locations were tested. The arcsine(square root(DIR) values observed in a PD at a determined AD were modeled as y = + X ll + Zvuv + Zlvulv + Zlbulb + e (Eq. 1-2) where y is a vector containing the arcsine(square root(DIR) values, l is the p x 1 vector of a constant and fixed effect Lo cation, Xl is an n x p design ma trix of full column rank which associates observations with the appropriate co mbination of the fixed effect. The u vectors (uv ulv and ulb) are q x 1 vectors for the random terms variety, Location by variety and Location by block while Z matrices (Zv Zlv and Zlb) are n x q design matrices for those random effects mentioned above. The random effects and error are assumed to be independent Gaussian variables with zero means and variance structures var(ui) = 2i Ibi (where bi is the length of ui; i = 1.3 ) and var(e ) = 2 In .

PAGE 21

21 The random terms were assumed to have a uni que G structure while the error variance was described by specifying an Identity R structure for each Location. This analysis also provided an estimate of Type B genetic correlation (Yamada II) among DIR at certain AD from different Locations. ii) Bivariate analysis through G structure specification This analy sis tested the differences in DI R among locations at certain AD and PD. The data were modeled as y = Xll + Zlgulg + Zlbulb + e (Eq. 1-3) where y is a vector containing the arcsin e(square root(DIR) fo r the corresponding AD, l is the p x 1 vector of a constant and fixed effect Location, Xl is an n x p design matrix of full column rank which associates observations with th e appropriate combination of fixed effect. The u vectors ( ulg and ulb) are q x 1 vectors for the random terms Location by variety and Location by block effects while Z matrices (Zlg and Zlb) are n x q design matrices for those random effects mentioned above. The random effects and error are assumed to be independent Gaussian variables with zero means. The random term s were assumed to have different variance structures: diagonal G structure for blocks (no correlations among blocks from different locations) 2 3 2 2 2 100 0 0 00 varb b b block and a correlation G structure for the genotypes (there is covariance among genotypes between locations):

PAGE 22

22 22 1 2 1 2 1 22 1varM C genotype while the error variance was described by specifying an Identity R structure for each Location. This analysis also provided an estimate of Type B genetic correlation (bivariate REML estimation) among DIR from different locations at each AD by modeling the G structure of the interaction between locations and genotypes. The correlation among ge notypes was obtained at the genotypic rather than the phenotypic level because through the specification of a correlation model applied to the G structure for the Genotype f actor its possible to ob tain a true estimation of the genotypic correlati on (Gilmour et al., 2006). Estimates of Repeatability were also obtained at each Location. iii) Univariate analysis for planting dates This analy sis was performed for each location se parately to test the difference in the DIR among PDs. For example, for Marianna, the DIRs obtained at AD 90 which were available only for PDs 2 & 3, (Table 1-1) were analyzed and the difference between those PDs was tested. The arcsine(square root(DIR) observed in a Location at an AD were modeled as y = + X pp + Zvuv + Zpvupv + Zpbupb + e (Eq. 1-4) where y is a vector containing the arcsine(square root(DIR), p is the p x 1 vector of a constant and fixed effect PD, X is an n x p design matrix of full column rank which associates observations with the appropriate fixed effect level. The u vectors (uv upv and upb) are q x 1 vectors for the random terms variety, PD by variety and PD by block effects while Z matrices (Zv Zpv and Zpb) are n x q design matrices for those random effects mentioned above. The random

PAGE 23

23 effects and error are assumed to be independent Gaussian variables with zero means and variance structures var(ui) = 2i Ibi (where bi is the length of ui; i = 1.3 ) and var(e ) = 2 In The G structure was assumed unique and an Identity R structure was specified for each PD. This analysis also allowed the estimation of Type B genetic correlation (Yamada II) among DIR from different PDs at each AD. iv) Multivariate analysis through G structure specification This analysis was perform ed analyzing both lo cations at the same time. The PDs analyzed were those in which both locations have DIRs for the same AD. Thus, for AD112, PDs 3 & 4 for both locations were analyzed while for AD132 PDs 2 & 3 were the ones tested. For this analysis, the data structure was modi fied by creating a new factor (cell) which was the combination of the levels of the factors Location and PD. Consequently, when analyzing AD112 the four cells were Citr a-PD3, Citra-PD4, Marianna-PD3 and Marianna-PD4 while in analyzing AD132 the cells we re Citra-PD2, Citra-PD3, Mari anna-PD2 and Marianna-PD3. The arcsine(square root(DIR) observed in a Lo cation at each of two planting dates at an AD were modeled as y = Xcc + Zcgucg + Zcbucb + e (Eq. 1-5) where y is a vector containing the arcsin e(square root(DIR) fo r the corresponding AD, c is the p x 1 vector of a constant and fixed effect Cell, Xc is an n x p design matrix of full column rank which associates observations with the appropriate combination of fixed effect. The u vectors ( ucg and ucb) are q x 1 vectors for the random te rms Cell by variety and Cell by block while Z matrices (Zcg and Zcb) are n x q design matrices for those random effects mentioned above. The random effects and error are assumed to be independent Gaussian variables with zero

PAGE 24

24 means. The random terms were assumed to have different variance structures: diagonal G structure for blocks (no correlations among blocks from diff erent cells) a nd unstructured correlation G structure for the ge notypes (there is covariance among genotypes between cells) while the error variance was described by speci fying an Identity R structure for each Cell. v) Bivariate analysis fo r diffe rent assessment dates For multivariate linear mixed methods, measurements from different environments are treated as different variable with different variance and covariance structures which are simultaneously estimated using the REML appr oach (Schaeffer & Wilton, 1978). Consequently, the main weakness of univariate methods (i.e. heterogeneous variances) is properly addressed (Lu, 2001). A bivariate analysis (a special case of multivar iate) was used here to estimate the genetic correlation between DIR by jointly analyz ing two ADs from the same location. In Marianna, DIRs at each AD were taken in at least two different PDs (see Table 1-1). This also allowed testing the e ffects of PDs on both AD DIRs bei ng analyzed bivariately. In the case of Citra, each combination of AD DIRs being analyzed was taken only in one PD so no testing for PD effect was possible. The bivariate model for the Marianna data subset can be written as y = ( I2 Xp)p + ( I2 Zg) ug + ( I2 Zpg) upg + ( I2 Zpb) upb + e (Eq. 1-6) where '''),(ADb ADayyy ; '''),(ADb ADagguuu ; '''),(ADb ADapg pguuu ; '''),(ADb ADapb pbuuu and '''),(ADb ADaeee In turn ADay = the vector containing the arcsine(square root(DIR) for AD a; while p is the 1 x p vector of a constant and fixed effect PD, Xp is an p x n design matrix of full column rank which associates observations with th e appropriate combinati on of the fixed effect, the u vectors ('ADagu,'ADapgu and 'ADapbu) are 1 x q vectors for the random terms variety, PD by

PAGE 25

25 variety and PD by block while Z matrices (Zg Zpg and Zpb) are n x q design matrices for those random terms mentioned above. The random effects and error were assumed to be independent Gaussian variables with zero means and variance structures: var(ug)= 2g I10 var(upg)= 2pg I20 var(upb)= 2pbj I3 and var(ej)= 2 I60. Thus, the random terms were assumed to have a variance resulting from the direct product of three G correlation structures (one for each random term) while the error variance was described by specifying a unique unstr uctured correlation R structure. In addition to the previous assumptions, the bi variate analysis also involves the following ones: 10 ')cov(I uuADaADb ADb ADag gg 20 ')cov(I uuADaADb ADb ADapg pg pg 3 ')cov(I uuADaADb ADb ADapb pb pb and 60 ')cov( I eeADaADb ADb ADa Thus random effects and errors are correlated be tween variables (DIRs from different ADs). The bivariate model for the Citra data subset can be written as y = ( I2 X) + ( I2 Zg) ug + ( I2 Zb) ub + e (Eq. 1-7) where '''),(ADb ADayyy ; '''),(ADb ADagguuu ; '''),(ADb ADabbuuu and '''),(ADb ADaeee In turn, ADay = the vector containing the arcsine(square root(DIR) for AD a; while p is the 1 x p vector of a constant, Xp is an p x n design matrix of full column rank which associates observations with the constant, the u vectors ('ADagu and 'ADabu) are 1 x q vectors for the random terms variety and block while Z matrices ( Zg and Zb) are n x q design matrices for those random terms mentioned above.

PAGE 26

26 The random effects and error were assumed to be independent Gaussian variables with zero means and variance structures: var(ug)= 2g I10 var(ub)= 2b I3 and var(ej)= 2j I60. Thus, the random terms were assumed to have a variance resulting from the direct product of two G structures, an unstructured correla tion for Genotypes and a general correlation for blocks while the error variance was described by specifying a unique unstructured correlation R structure. In addition to the previous assumptions, the bi variate analysis also involves the following ones: 10 ')cov( I uuADaADb ADb ADag gg 3 ')cov( I uuADaADb ADb ADab bb and 60 ')cov( I eeADaADb ADb ADa Thus random effects and errors are correlated between variables (DIRs from different ADs). The bivariate analysis also provided an estima te of true Type A ge netic correlation (two traits measured on the same experimental un it) through bivariate REML estimation among DIR obtained at different AD (Gilmour et al., 2006). Calculation of Genetic Correlations Traditional (Type A) genetic correlation was calculated amon g DIRs obtained on the same plot at different ADs. Using the genotypic variance and c ovariance component estimates obtained from the corresponding linear mixed model described under section v) Bivariate analysis for different ADs. The genot ypic correlation betw een DIRs from ADj and ADk was estimated as 22covkj jkgg g gr (Eq. 1-8)

PAGE 27

27 where jkgcov is the estimated genotypic covariance between DIRs j and k and 2jg is the estimated genotypic variance for the score obtained at ADj. For the same trait, for example, a score obtained at ADj but assessed in different experimental units (i.e. differe nt PD or Location), a Type B correlation, model II (Yamada, 1962) was calculated by using the corresponding ratio of variances depending on the nature of the performed analysis (univariat e or bivariate). If calculated from variance components obtained through the univariate analyses de scribed under sections i) Univar iate analysis for Location or iii) Univariate analysis for PDs, the Type B genetic correlation was estimated as 22 2 gfg g BIIr (Eq. 1-9) where 2 g is the estimated genotypic variance for the score obtained at that Location or PD and 2 gf is the corresponding inter action term between genotype and either Location or PD. When calculated from bivariate or multivaria te analyses, as shown under sections ii) Bivariate analysis through G structure specif ication and iv) Multivariate analysis through G structure specification, the Type B genetic correlation was estimated as 222 1 21covc c cc IIgg g Br (Eq. 1-10) where 21covccg is the estimated genotypic covariance between DIRs from locations one and two and 21 cg is the estimated genotypic variance fo r the score obtained at Location number 1.

PAGE 28

28 Standard errors of the correlations were ca lculated using the Taylor Series Expansion method (Gilmour et al., 2006). Calculation of Repeatability An entry-mean repeatability of performance in each environment was calculated as an intra-class correlation using the corresponding va riance components from the linear mixed models described under section ii ) Bivariate analysis through G structure specification, according to the formula (Hol land et al., 1998): r Re g g 2 2 2 (Eq. 1-11) where, 2 g 2 e and r are the genotypic variance, the error variance and the number of replications at that location, respectively. Results There was a clear difference in the intensity of the epidemic between locations. By 132 DAP the highest DIR in Citra was less than half th e smallest DIR recorded at Marianna (Table 12). The epidemic progression was slow but steady at Citra while at Marianna it was fast and abrupt, reaching final intensity as early as 90 DAP (Fig. 1-1). Th e slight reduction in intensity observed in the figure was caused by the harvest of susceptible genotypes prior to the assessment at 132 DAP, reducing the overall intensity DIRs for that AD. Among genotypes, a clearer separation of resistance groups wa s observed at Marianna, under a heavy epidemic compared to Citra (Figs. 1-2 & 1-3). The groups conformatio n was as expected, th e susceptible genotypes comprising F-435HOL, NemaTAM and SunOleic 97R, the resistant ge notypes comprising

PAGE 29

29 NC94002, AP-3, Georgia-02C, DP-1 and C-99R while two genotypes showed intermediate values (Georgia Green and ANorden). Location Effect The univariate analysis first tried (described under point i of Mate rial and Methods), is similar to the usual ANOVA plus the specification of different error terms for each location. It provided variance components of th e interaction Location by PD that were rather important most of the time (data not shown) and were due to heterogeneous variances among planting dates. Consequently the decision was made to model a variance for each combination of location and planting date as described under point ii of Material and Methods. With this model, the location effect was found significan t in all the PD and AD (Table 1-3) with Marianna exhibiting higher predicted values (Table 1-4). In the most extreme case, PD3 at AD90, the Marianna predicted value for transformed DIR was five times larger. The ratio of error variance for Marianna to th e error variance for Citra was mostly above one (Table 1-3), suggesting higher data variability at Marianna. An extreme value of 4.5 for this ratio was observed at PD 3, AD112. In Marianna, the ratio of the block variance component to the error variance ranged from 0 to 0.7. At Citra, the DIR were extremely clos e to zero (data not shown), denoting the rather minor importance of block as a variability source. The ratio of genotypic variance to the biggest location error variance ranged from 2.5 to 20.2. In every case, the error vari ance for Marianna was used as the denominator in the ratio. It was clear that genotypic variance was the main source of data variability.

PAGE 30

30Planting Date Effect i) Univariate analysis for planting dates The importance of planting date as a factor determining the DIR was variable according to the location. In Citra it ranged from highly si gnificant (p=0.006) to slightly non-significant (p=0.054) depending on the genotypes and planting dates included in each analysis (Table 1-5). Earlier AD (70 & 90) showed greater significan t differences than later ones (112 & 132), although different PDs were compared in each analysis. At 70AD and 132AD, the later PD exhibited higher predicted DIR (Table 16), while at 90AD the opposite was true. In Marianna, planting dates were statistical ly different only at 132AD, with the PDs ranking 3>4>2 in DIR. The ratios of planting date by genotype intera ction variance to the biggest error variance were most of the time close to zero (Table 1-5). The only exception was Marianna at 132AD where the ratio was 0.8, yet small compared to the genotypic variance. In comparison, Genotypes seemed far more important as a variability sour ce with the ratios of genotypic variance to the biggest error variance ranging from 0.4 to 1.7 at Citra and from 2. 4 to 4.4 at Marianna. In order to compare ADs, when this ratio was calculated only from PD 2 and 3, its value went from 1.5 at AD90 to 3.3 at AD112 (data not shown), s uggesting increasing variability over time. In Citra, the variance component for the PD by block intera ction varied from small, 1/10 of the smallest error variance, at AD112 to neglig ible at the other thr ee ADs (data not shown). Meanwhile, in Marianna it ranged from rather small, 1/4 of the smallest error variance, to important (1.5 times the smallest error variance). The ratio of error variances among PDs was quite similar among ADs at Citra, ranging from 1 to 1.4 while at Marianna they varied widely from 1.5 to 7.6.

PAGE 31

31Location by Planting date (cell) effect v) Multivariate analysis thro ugh G structure s pecification The analysis using common combinations of planting dates and assessment dates between both locations (referred herein as cells) showed that the cell effect was highly significant at both ADs (Table 1-7), with Citra showing smaller predicted values than Marianna (Table 1-8). Cells of different location were always significanlty different. At 112 DAP, within-location cells were only different for citra. At 132 DAP within-location cells were only different for Marianna. Cells showed different variability with Mariannas cells displaying the most variability (data not shown). The ratio of th e greatest error variance to the sm allest (Marianna at PD4 : Citra at PD 3) was 8.8 for 112AD while it was 1.2 (Maria nna at PD 3 : Citra at PD 3) for 132AD (data not shown). The block variance had a negligible value at AD112 for PD 3 and at AD132 for both PDs at Citra (data not shown). In Marianna the block variance had some importance, but not in Citra. The ratio of block variance for a cell to its er ror variance at 112AD was highest for PD 3 at Marianna, while for 132AD the highest ratio was observed for PD 2 in Marianna. The ratio of genotypic variance within a loca tion to the error vari ance for that location ranged from 0.7 to 20.2 with 5 out of 8 values above 2 (data not show n), showing that the genotypic variance was the largest variability source for spotted wilt intensity DIRs. Phenotypic and Genetic Correlations The DIRs a plot received at different a ssessment dates were strongly correlated. Phenotypic correlations were medium to high (Table 1-9). In Citra the values ranged wider and below (0.43-0.89) those in Marianna (0.85-1).

PAGE 32

32 Genotypic correlations were high at both locations, ranging from 0.85 to 1 (Table 1-9) with rather small standard errors. Many coefficients we re fixed at the maximum theoretical value of 1 by constraining the covariance matrix. There was a moderate and significant co rrelation among the phenotypic and genotypic correlation coefficients (0.76, p=0.04) by Spear mans rank correlation but due to the reduced number of data pairs, this signi ficance should be taken cautiously. A high within location consistency of genotypi c performance for DIR at different PDs and ADs was observed. Type B genetic correlation coefficients were high, ranging from 0.83 to 1 with modest standard errors, with Citra displayi ng larger errors (Table 1-10). There was also consistency of genotypic performance across locations at the same PD and AD. In three out of five cases, the correlation coefficient was 1 wh ile in the remaining two cases it was above 0.7 (Table 1-10). When calculated from the combination of locatio ns x planting date (cell), the correlation coefficient between transformed DIR was very high, with all but one value above 0.9 (Table 110). All the coefficients for AD 132 had a value of one, suggesting that the perfor mance of the evaluated genotypes in a certain environment (combination of location x planting date) was predictable based upon its performa nce in another environment. Irrespective of the type of genetic correlation calculated, the general results showed that the genotypic performances in the tested enviro nments were highly correlated between locations and among ADs and PDs. Additionally, the mode ling of variance structures allowed better estimations of these correlations through elimin ation of scale effect s and error shrinkage. Repeatability Repeatability values for the transformed DIR at different PDs and AD were high (Table 111) and their precision good (small standard errors). All but one of the values was above 0.75

PAGE 33

33 showing that the theoretical upper limit for the heritability of the intensity was quite high. In general, the repeatabilities at Citr a were smaller than at Marianna. Discussion Some locations tend to have st ronger spotted wilt epidemics than others (Culbreath et al., 2003). Marianna has usually displaye d severe epidemics since the la te s (Tillm an et al., 2007; Culbreath et al., 2005) while spotted wilt is a lesser problem at C itra (Gorbet, pers. comm.). The epidemics observed in 2005 followed a similar tre nd, with spotted wilt be ing consistently more prevalent in every PD and AD in Marianna. This was reflected not only in the final DIR but also on the fact that they peaked earlier in the season in every planting date. The epidemic seemed to almost reach its peak at Marianna at around 90 days after planting as there was little disease progression afterwards This rapid establishment of near maximal DIR at such an early assessment date is uncommon in the literature. Even the worse epidemics reported so far have reached fina l intensities at later assessmen t dates (Murakami et al., 2006, Culbreath et al., 1997). This wa s especially important because early TSWV infection typically leads to more damage inflicted by the virus (Culbreath et al., 2003). In contrast to Marianna, the progression of the epidemic at Citra showed a ty pical steady increase similar to mild epidemics reported elsewhere (Murakami et al., 2006, Culbreath et al., 1997). In the present study, the level of spotted wi lt intensity recorded for well known genotypes resembled those of severe epidemics described by some authors (Culbreath et al., 2005, Tillman et al., 2007). These authors report ed intensity ratings above 0.8 fo r susceptible genotypes such as SunOleic 97R and variable ratings for the mode rately susceptible Georgia Green. These values are in agreement with ours, suggesting that final intensity ratings were in the expected range for a wide array of genotypes of varying resistance levels.

PAGE 34

34 Coinciding with published research, resistan ce differences among genotypes were most noticeable under the severe epidemics in Marianna (Culbreath et al., 1997; Tillman et al., 2007). The observed resistance grouping was in good agr eement with the information provided by the present version of the Peanut Dis ease Risk Index (Brown et al, 2007). Murakami et al. (2006) reported that planting dates diverged in spotted wilt incidence as the season progressed under a mild epidemic but were similar under a severe one. However, in the present study rather the opposite was true. Under a mild epidemic in Citra, difference s in spotted wilt intensity between PDs were easily observed early in the season (PD2 vs. PD4 at AD70), but they tended to diminish as the season advanced. Meanwhile, in Marianna the effect of PD was not important until late in the season (132 AD) where all the PDs differed in th e intensity DIRs. Curiously, the spotted wilt intensity spiked at the third PD which coincides with the extension recommendation for plantings with reduced spotted wilt risk (Bro wn et al., 2007). This further underscores the seasonal unpredictability of spotted wilt epidemics and the need to plant resistant varieties. For example, Tillman and coworkers (2007) reporte d that June plantings were less conducive to severe spotted wilt in most of the genotypes tested in the Florida Panhandle. In the present study, however, early May planting seemed best for reduction of spotted wilt damage. Differing from results reported by Hurt et al. (2005), in the pr esent study the interactions among genotypes, location and planting dates were not important. This was so even when all the three factors were jointly analyzed, which high lights the advantages of modeling variance and error structures (Gilmour et al., 2006). Phenotypic and genotypic correla tions are of the same sign and similar magnitude most of the time (Lynch and Walsh, 1998). This seemed to be the case in the present study. The fact that

PAGE 35

35 genetic correlations were higher than the phenot ypic ones can be explained because the variance and covariance components used in the calculat ions come from a well specified mixed model analysis, thus only containing genotypic effects from which experiment al noise was removed (i.e. scale effects as the season progresses). Consequently, they are a much more precise estimation of the genetic cause of similarity among the spotted wilt ratings that a genotype will receive at different assessme nt dates (Holland et al., 2003). The high genetic correlation among assessment dates indicated a high consistency in the rankings obtained by the genotypes across the asse ssment dates. This finding coincides with those reported by Culbreath et al. (1997) and Murakami at al. (2006), who reported reduced cross-over of genotypic rankings at different assessment dates. Consequently, the differences among genotypes in their performance against TS WV should be perceived no matter the moment they are compared. Despite the marked differences between the disease intensity between locations, there was a marked consistency of phenotypic performance across locations, PDs and their combinations, as indicated by the high Type B ge netic correlation coefficients. This suggests a scenario similar to that pres ented by McKeown et al. (2001) and different to the ones described frequently in published research where GxE interaction seemed to be the norm (Culbreath et al., 1997, 2005; Tillman et al ., 2007). This apparent discrepancy could be caused by the use of ANOVA in those references, coupled with scale effects. The fact that Culbreath et al. (2005) reported that even though GxE interacti on was present, the genotypic trends were consistent across locations and years seems to back up this explanation. The high Type B genetic correlation suggests th at there was no cross interaction among genotypes and environments. This finding could sugge st that the wide array of resistance genes

PAGE 36

36 present in the tested genotypes performed cons istently when faced w ith the viral populations present at both locations, which could arise from similarities in the viral consensus sequence between locations (i.e. they are essentially the same) or the fact that both consensus sequences induce similar rankings on the tested genotypes in much the same way Mandal et al. (2006) reported for Georgia isolates. Genotypic variances were the most important random variation source in this study. By having a wide range of resistances in the tested genotypes the ge notypic variance is expected to be increased (Betrn et al., 2006), th us increasing the repeatability. Another cause of high genetic variances relative to the error term was probably the correct modeling of the experimental data (Gilmour et al., 2006), which was accomplished throughout the present study. Additionally, genetic causes (biochemical path ways) could predominantly have established the performance of the genotypes (Lynch and Walsh, 1998). This last explanation seemed to be supported by the high genetic correla tion coefficients obtained here. Conclusions The modeling of correct variance and covarian ce structures of the tests provided a good estimation of variances and covariances which in turn allowed determination that location was a significant factor in es tablishing the spotted wilt ratings observed among genotypes. Meanwhile, planting date was only a significan t factor under light epidemics or late in the season under heavy epidemics. The high correlation among assessment dates i ndicated that the rela tive performance of genotypes can be perceived early in the season a nd the genotypic differences tend to persist until harvest time.

PAGE 37

37 The high values of both Type B genetic correlati on coefficients and repeatability estimates suggested a strong genetic determination of the observed genotypic differences in spotted wilt intensity ratings. This emphasizes the importance of resistant cultivars in the management of spotted wilt.

PAGE 38

38 Table 1-1. Layout of data collect ion for planting date studies of peanut in Marianna and Citra, Florida in 2005. Assessment Dates (in Days After Planting) Location Planting date 70 DAP 90 DAP 112 DAP 132 DAP Citra 1a X 2 X X 3 X X Xb 4 X X Marianna 1 X 2 X X Xc 3 X X Xc 4 Xd Xcd The X indicates the cells in which spotted wilt damage was assessed. a Genotypes F435HO, NemaTAM and NC94002 were not planted. b F435HO was already dug. c F435HO and SunOleic 97R were already dug. d Replication one was discarded. Table 1-2. Means and standard errors for tomato spotted wilt disease intens ity ratings at different planting dates at Citra and Marianna, Florida in 2005. AD 70 AD 90 AD 112 AD132 Location Planting date Mean (n) S.E. Mean (n) S.E. Mean (n) S.E. Mean (n) S.E. 1a 0.066 (21) 0.010 2 0.057 (30) 0.009 0.157 (30) 0.018 3 0.048 (30) 0.008 0.123 (30) 0.016 0.179 (27) b 0.019 Citra 4 0.097 (30) 0.014 0.191 (30) 0.026 1 0.545 (30) 0.054 2 0.514 (30) 0.054 0.480 (30) 0.056 0.442 (24) c 0.049 3 0.640 (30) 0.051 0.670 (30) 0.050 0.673 (24) c 0.047 Marianna 4 0.632 (20)d 0.063 0.548 (14)cd 0.070 a Genotypes F435HO, NemaTAM and NC94002 were not planted. b F435HO was already dug. c F435HO and SunOleic 97R were already dug. d Replication one was discarded.

PAGE 39

39 Table 1-3. Location effect and variance ratios fo r bivariate analysis performed on transformed spotted wilt disease intensity ratings at thr ee planting dates assesse d in different times at Citra and Marianna, Florida in 2005. Planting Date Assessment date P-value>F for location 2 eM/ 2 eC2 bM/ 2 bC2 bM/ 2 eM2 lg>/ 2 e> 2 132 0.004 1.4 147712.9 0.7 5.1 3 90 <0.001 3.3 10.5 0.0 4.9 112 0.004 4.5 1137.2 0.7 2.5 132 <0.001 1.3 10.5 0.0 5.6 4 112 <0.001 0.7 0.7 0.2 20.2 2 eM: Marianna error variance. 2 eC : Citra error variance. 2 bM: Marianna block variance. 2 bC : Citra block variance. 2 lg>: biggest location*genotype variance. 2 e>: biggest location error variance. Table 1-4. Transformed spotted wilt disease inte nsity rating predicted values, standard errors (SE) and standard errors of mean differ ences (SED) for planting dates assessed at different times at Citra a nd Marianna, Florida in 2005. Planting Date Assessment date Location Predicted Value SE SED Citra 0.3881 0.0391 2 132 Marianna 0.7858 0.1011 0.0786 Citra 0.1879 0.0339 90 Marianna 0.9461 0.1045 0.0875 Citra 0.3349 0.0391 112 Marianna 1.0253 0.1400 0.1193 Citra 0.4247 0.0331 3 132 Marianna 1.0097 0.0923 0.0698 Citra 0.4337 0.0552 4 112 Marianna 0.9532 0.1108 0.0899

PAGE 40

40 Table 1-5. Planting date effect and variance ratios for univariate an alysis performed on transformed spotted wilt disease intensity ra tings at four planti ng dates assessed in different times at Citra a nd Marianna, Florida in 2005. Location Assessment date P-value>F for planting date 2 e>/ 2 e<2 pld.g/ 2 e>2 g/ 2 e> Citra 70 0.006 1.1 0.1 1.1 90 0.023 1.0 0.1 0.4 112 0.054 1.1 0.1 1.7 132 0.031 1.4 0.0 0.9 Marianna 90 0.092 1.5 0.4 4.4 112 0.096 7.6 0.1 2.4 132 0.045 3.5 0.8 3.8 2 e> : biggest error variance. 2 e< : smallest error variance. 2 pld.g : planting date by genotype interaction variance. 2 g : genotypic variance. Table 1-6. Transformed spotted wilt disease inte nsity rating predicted values, standard errors (SE) and standard errors of mean diff erences (SED) for at four planting dates assessed in different times at Ci tra and Marianna, Florida in 2005. Location Assessment date Planting Date Predicted Value SE SED 2 0.2199 0.0304 70 4 0.2996 0.0300 0.0223 1 0.2425 0.0298 90 3 0.1502 0.0301 0.0304 2 0.3692 0.0355 Citra 132 3 0.4247 0.0370 0.0249 2 0.6755 0.1067 3 0.9298 0.1057 Marianna 132 4 0.8387 0.1091 0.0899 Table 1-7. Location by planting date combinati on (cell) effect and variance ratios for multivariate analysis performed on transformed spotted wilt disease intensity ratings at different planting dates asse ssed at different times at Citra and Marianna, Florida in 2005. Assessment date P-value>F for Cell 2 e>/ 2 e<2 b>/ 2 b<2 b>/ 2 ec2 cg/ 2 e> 112 <0.001 6.9 2852675920.7 1.6 20.2 132 <0.001 1.6 366195220.8 0.7 7.1 2 e> : biggest error variance. 2 e< : smallest error variance. 2 b> : biggest block variance. 2 b< : smallest block variance. 2 cg : cell by genotype interaction variance. 2 g : genotypic variance.

PAGE 41

41 Table 1-8. Transformed spotted wilt disease inte nsity rating predicted values, standard errors (SE) and standard errors of mean differ ences (SED) for location by planting date combinations (cells) assessed at different times in Citra and Marianna, Florida in 2005. Each assessment date was analyzed separately. Cell Assessment date Predicted Value SE SED 1 (Citra, P.D.3) 0.3349 0.0394 2 (Citra, P.D.4) 0.4337 0.0552 3 (Marianna, P.D.3) 1.0253 0.1397 4 (Marianna, P.D.4) 112 0.9532 0.1108 0.0981 1 (Citra, P.D.2) 0.3881 0.0401 2 (Citra, P.D.3) 0.4358 0.0325 3 (Marianna, P.D.2) 0.7863 0.1014 4 (Marianna, P.D.3) 132 1.0599 0.0955 0.0681 P.D. Planting Date. Table 1-9. Phenotypic and genetic correlations among transformed spotted wilt disease intensity ratings at four planting dates assessed in different times at Citra and Marianna, Florida in 2005. Values in brackets are standard errors. Phenotypic Correlation [SE] Genetic Correlation [SE] Location Assessment Date 112 132 Assessment Date 112 132 70 0.83 [0.08] 0.84 [0.07] 70 0.95 [0.06] 1 [0.08] a 90 0.43 [0.19] 0.75 [0.15] 90 .085 [0.22] 1 [0.21] a Citra 112 0.71 [0.08] 112 0.99 [0.01] 90 0.85 [0.06] 0.87 [0.06] 90 1 [0.02] a 1 [0.03] a Marianna 112 0.89 [0.09] 112 1 [0.01] a a Genetic correlations were kept in the theoretical range by constraining the covariance matrix (Gilmour et al. 2006)

PAGE 42

42 Table 1-10. Type B genetic correlations and [standard errors] for transformed spotted wilt disease intensity ratings at f our planting dates assessed in different times in Citra and Marianna, Florida in 2005. a Coefficients obtained from Univariate Analysis Location Assessment Date Genetic Correlation 70 0.91[0.17] 90 0.83[0.32] 112 0.95[0.11] Citra 132 1[0.01] 90 0.91[0.08] 112 0.97[0.03] Marianna 132 0.87[0.09] Coefficients obtained from Bivariate Analysis Assessment Date Planting Date 90 112 132 2 1[0.09] 3 0.71[0.22] 1[0.13] 1[0.25] 4 0.73[0.18] Coefficients obtained from Multivariate Analysis Assessment Date 112 Cell Citra-PD4 Marianna-PD3 Marianna-PD4 Citra-PD3 c 1 [0.11]b 1 [0.13]b 1 [0.09]b Citra-PD4 0.91 [0.12] 0.73 [0.18] Marianna-PD3 0.94 [0.08] AD132 Cell Citra-PD3 Marianna-PD2 Marianna-PD3 Citra-PD2 1 (0.24)b 1 [0.08]b 1 [0.07]b Citra-PD3 1 [0.20]b 1 [0.20]b Marianna-PD2 1 [0.04]b a Correlations were calculated among planting dat es (across locations), between locations (across planting dates) and among cells (location by plantin g date combination) by applying different mixed models. b Variance components and genetic correlations were kept in the theoretical range by constraining the covariance matrix (Gilmour et al. 2006). c PD: Planting Date. Table 1-11. Entry-mean repeatability estimates and their [standard errors] for transformed spotted wilt intensity ratings at three planti ng dates assessed three times in Citra and Marianna, Florida in 2005. Values for each location are separated by a slash, Citra being on the left and Marianna on the right. Assessment Date (days after planting) Planting Date 90 112 132 2 0.81 [0.11] / 0.94 [0.04] 3 0.82 [0.10] / 0.94 [0.04] 0.76 [0.14] / 0.86 [0.08] 0.57 [0.26] / 0.94 [0.04] 4 0.90 [0.06] / 0.98 [0.01]

PAGE 43

43 0 0.1 0.2 0.3 0.4 0.5 0.6 0.790 DAP 112 DAP 132 DAPMean Disease Intensity Ratin g Citra Marianna Figure 1-1. Change of spotted wi lt disease intensity ratings over time at Citra and Marianna, Florida in 2005 (averaged over a ll planting dates and genotypes)

PAGE 44

44 0 0.2 0.4 0.6 0.8 1N C 9 4 0 0 2 A P 3 G e o r g i a 0 2 C D P 1 C 9 9 R A N o r d e n G e o r g ia G r e e n N e m a T A M F 4 3 5 H O L S u n O l e ic 9 7 RDisease Intensity Ratin g 4/19/2005 5/4/2005 5/19/2005 6/6/2005 Figure 1-2. Spotted wilt diseas e intensity ratings at 112 days after planting for ten genotypes planted at different dates at Marianna, Florida in 2005.

PAGE 45

45 0 0.1 0.2 0.3 0.4 0.5 0.6N C 9 4 0 0 2 A P 3 G e o r g i a 0 2 C C 9 9 R D P 1 G e o r g i a G r e e n A N o r d e n N e m a T A M S u n O l e i c 9 7 R F 4 3 5 H O LDisease Intensity Ratin g 5/10/2005 6/2/2005 Figure 1-3. Spotted wilt diseas e intensity ratings at 112 days after planting for ten genotypes planted at different dates at Citra, Florida in 2005.

PAGE 46

46 CHAPTER 2 HERITABILITY AND BREEDING VALUES OF TSWV RESISTANCE IN POPULATIONS FROM PE ANUT (ARACHIS HYPOGAEA L.) CROSSES Introduction Tomato spotted wilt virus ( Bunyaviridae: Tospovirus TSWV) is a worldwide problem in both greenhouse and field crops (German et al., 1992) Since the first report of TSWV affecting peanut in Texas in the early 70s, it has become a major limiting f actor to peanut production in the US (Culbreath et al., 2003). Spotted wilt disease symptoms develop at leas t a week after inoculat ion (Hoffman et al., 1998) and can be seen quite early in the seas on (Chapter 1 of this dissertation). Typical aboveground symptoms include concen tric ringspots, mosaic patterns, stunting, varied degrees of apical and leaf necroses and general chlorosis, also know n as yellowing (Mitchell, 1996, Culbreath et al., 2003, Demsky & Reddy, 2004,). The presence of some of these symptoms, like stunting and yellowing, has been linked to the acti on of specific viral proteins (Prins et al 1997, Koll & Btner, 2000). However, the nature and se verity of symptoms depends on the consensus (predominant) RNA sequence in the virus popula tion (Nagata et al., 1993; Mandal et al., 2006). Generally, epidemic patterns vary signifi cantly across locations, particularly if the locations are distant and contai n dissimilar agroecosystems (Culbreath, 2003; Groves et al., 2003) with different crop species that react dissimila rly to TSWV (Kucharek et al., 2000) or different weeds that influence thrips dynamics (N orthfield, 2005). Whatever the environment, the rate of progress of spotted wilt epidemics has b een shown to be cultivar dependent in peanut (Culbreath et al., 1997; Murakami et al., 2006; Chapter 1 of this di ssertation). Resistant cultivars are the most important factor in the manage ment of the disease (Brown et al., 2007). Spotted wilt resistance is an important goal for peanut breeders in Southeastern USA (Gorbet, 1999, Tillman et al., 2007). At present there are several resistant peanut cultivars and

PAGE 47

47 breeding lines available. Most of them trace their resistance back to two unrelated sources, PI 203395/6 (both PIs come from the same original accession) and PI 576638. The former, typical hypogaea botanical variety members, are the spotte d wilt resistance source of Georgia Green (Branch 1996), Georgia 01R (2002), DP-1 (Gorbet and Tillman, 2008, in press) and many others. Meanwhile, PI 576638 (a hirsuta botanical variety member) is the source of the resistance of several breeding lines that had shown remarkable field resistance to spotted wilt (Culbreath et al., 2005, D. Gorbet, pers. comm.). The existence of a different mechanism of resistance between these two PIs has been hypothesized (Culbreath et al., 2005). Nonetheless, it has been s hown that resistance is unrelated to vector non-preference or reproduction (Culbrea th et al., 1996, 1997, 2000). A third source of TSWV resistance has been obs erved in sisterline cultivars AP-3 (Gorbet, 2007) and Carver (Gorbet, 2006). In this case, the or igin of the resistance is uncertain as neither of the parents of these two cultivars is resistant (D. Gorbet, pers. comm.) In tomato and pepper, major genes for resistance to TSWV have been described, having different modes of action and penetrance (Rosel lo et al., 2001, Soler et al., 1998 Moury et al., 1998). In species where inherited resistance is conditional or ambiguous, with no clear or consistent phenotypes, traditi onal genetic analyses become di fficult (Lynch and Walsh, 1998; Bruening, 2006) and the heritability becomes the most important information for the plant breeder (Simmonds, 1979). The heritability (in its narrow sense) expresses which proportion of the phenotypic variability can be transmitted from parent to offspring (Falconer & MacKay, 1996) and it is the main determinant of the expected response to select ion (Hallauer & Miranda, 1988).

PAGE 48

48 Among the methods used for heritability es timation, the most us ed is the variance component method because of its adaptability to different situations (N yquist, 1991). Heritability estimates depend not only on the genetic factors in the populations being analyzed but also on the environment in which they are tested (Falco ner & MacKay, 1996). In mo st situations, better discrimination among genotypes is feasible by te sting the populations in certain types of environments (Hall and Van Sanford, 2003, Venupr asad et al., 2007). These environments can be laboratory settings (Rapp & Junt ila, 2001); locations (Finne et al., 2000) or even planting dates (Chapter 1 of this dissertation). The expression of disease resistance is us ually scored as a polychotomous variable, commonly called ordered scale (Connover, 1998). Fo r this kind of trait, the unit of analysis can be the individual observati ons in the native scale (Hube r et al., 1994) or plot means combined with a transformation such as arcsine or logistic (Holland et al., 1998). The use of individual values is known to provide better heritability estim ates than plot means under most conditions when a REML approach is used in the linear mixed model context (Huber et al., 1994). Although the typical analytic al approach is to adjust a threshold model, this usually yields similar results to REML variance estimation under the native scale under a wide range of conditions (Banks et al., 1985; Westfall 1987). In accurate heritability estimates are sometimes obtained when the frequency of a category in a polychotomous variable is very high across the whole population of individuals being tested (Yang et al., 1998). Under these conditions, the threshold model is only superior to estimates obtained from REML on the native scale if incidences are extreme and heritabilities are low to medium (Lopes et al., 2000).

PAGE 49

49 Variance components for herita bility estimation today are mostly obtained through Mixed Linear Model approaches. They offer the flexib ility of analyzing various types of unbalanced data coming from non-traditional mating desi gns with good precision (Holland et al., 2003). Using this approach, random effects such as breeding values can al so be obtained (Lynch and Walsh, 1998). Breeding value is the sum of th e additive effects of an individuals genes (Lynch and Walsh. 1998). Recently, the use of mixed models coupled with REML has a llowed the accurate estimation of additive variance and consequently of the breeding values. A common linear mixed model used mostly by animal breeders is the Anim al Model (Mrode, 2005). It utilizes all genetic relationships among the individuals being analyzed in order to obtain a more accurate estimation of the additive variance than traditional me thods (Henderson, 1976; Lynch and Walsh, 1998). Breeding value estimates (BLUPs ), obtained through the Animal Model are accurate because they take into consideration not only the perfor mance of the individual but also that of its relatives (Mrode, 2005). Breeding values are used to choose individuals in a population that are superior for a trait and that will provide a better progeny. They can also be an integral part of a selective index to choose individuals based on se veral traits (Mrode, 2005). To gain insight into the genetics of resistan ce to TSWV in peanut, the objectives of this study were: 1) to provide herita bility estimates from crosses involving different sources of resistance while assessing their potential to gene rate superior progenies; and 2) to explore the relationship among different symp toms of infection by TSWV. Material and Methods To study the inheritance of resi stance to TSWV in peanut, th ree resistant genotypes (AP-3, DP-1 and NC94002) and a susceptible genotype (NemaTAM) were mated.

PAGE 50

50 The source of resistance for each resistant pa rent was believed to be unique (D. Gorbet, pers. comm.). AP-3 and DP-1 are related th rough a common ancestor, their grandparent Florunner, which is extremely susceptible to TSWV (Culbreath et al., 1997; Tillman et al., 2007). AP-3 is a runner-type cultivar whose parents display no noticeable spotted wilt resistance (Gorbet, 2007). DP-1 is also a runner-type cul tivar and traces its resistance back to its grandparent PI 203396, which has produced numer ous lines with good re sistance (D. Gorbet, pers. comm.). This PI is a typical member of the hypogaea botanical variety. NC94002, traces its resistance back to its parent PI 576638, which is an accession belonging to the hirsuta botanical variety (D. Gorbet and T. Isleib, pers. comm.). The resistant and susceptible parents were mated in the following combinations AP-3/ NemaTAM, NemaTAM/AP-3, NemaTAM/DP-1, NemaTAM/NC94002 and DP-1/NC94002. The resulting F1, Backcross, F2 and F3 populations, together with thei r parents were field tested. For each cross, 25 F 2:3 (F2-derived in F3) families with enough seed were randomly selected and included in each test in at least two replications. Field tests were conducted at the University of Florida Plant Science Research and Education Unit in Citra, Florida on a Candler Sand (Hyperthermic, uncoated Typic Quartzipsamments) during 2005 and 2006, at the No rth Florida Research and Education Center near Marianna, Florida on a Chipola loamy sand (Loamy, kaolinitic, thermic Arenic Kanhapludults) during 2006 and 2007 and at the North Florida Research and Education Center in Quincy, Florida on an Orangeburg fine sandy lo am (fine loamy, siliceous, thermic, Typic Paleudults) during 2007. The tests details are shown in table 2-1.

PAGE 51

51 The different crosses were randomized in each block, and the different generations of each cross were randomized within the cross subplots. The number of replications for each generation was variable within (by cross) and among tests, as can be seen in table 2-1. At Citra, plots were one row, 0.9 m wide and 4.5 m long, planted with 6 seeds/m in 2005 and 3 seeds/m in 2006. At Marianna and Quincy plots were two rows, 1.8 m wide and 4.5 m long, and planted with 3 seeds/m. Tests were subjected to cultivation practices such as early planting, low plant population, no phorate (insecticide) applicati on and single row planting pattern that are known to favor the development of spotted wilt epidemics (Culbreath et al., 2003). Natural TSWV-infected thrips populations were relied upon to cau se spotted wilt epidemics. Samples were taken from symptomatic plants and tested by DAS-ELISA (SRA 30400/0096 kit, AGDIA, Elkhart, IN) to confirm TSWV infection. With the exception of phorate at planting, plots were maintained according commercial peanut production practices for the region with fertilizer, herbicide, fungicide, insecticide and irrigation applied as recommended by the Univers ity of Florida Extension Service guidelines. Every other plant was marked with a flag and each individual was asse ssed for five typical foliar symptoms of spotted wilt, which are menti oned in the literature (Mitchell, 1996, Culbreath et al., 2003; Demsky & Reddy, 2004): stunting, foliar symptoms (spots and/or mosaic), tip death, leaf necrosis and yellowing (chloros is). In total, about 36,300 indi vidual plants were evaluated. The following ad-hoc scales were used to score each symptom. Stunting was defined as a reducti on in plant size in at least on e dimension (height, width). The degree of stunting was then assigned to a degree in the following ordered scale: 1. Light shortening of internodes, rendering th e plant size about 80% of a normal plant (or reducing one of the plant's dimensi ons to about 80% of a normal one).

PAGE 52

52 2. Noticeable shortening of internodes, plant size about 60% of a healthy one. 3. Marked shortening of internodes, plant size about 40% of a healt hy one, leaflets showing signs of poor expansion. 4. Very marked shortening of internodes, pl ant size about 30% of a healthy one, leaflets poorly unfolded. 5. Extreme shortening of internodes, plant si ze about a 20% of a healthy one, plant shows unfolded leaves in crowded limb tips. No leaf has been unfolded recently. Foliar symptoms were related to damage in photos ynthetic pigments as shown by the presence of spots or mosaic patterns. Thei r incidence and severity were jointly assessed according to the following ordered scale: 1. A few hardly noticeable (faint) foliar symptoms in a few leaves. 2. Noticeable (easily observable) foliar symptoms in a few branches. 3. Marked (very evident) foliar symptoms in most braches. 4. Very marked (covering most leaves) foliar symptoms in most branches. 5. Very marked foliar symptoms in all braches. Tip death was assessed according to the symptom incidence on the plant, as follows: 1. One stem tip dead 2. Up to 25% of the tips dead 3. Up to 50% of the tips dead 4. Up to 75% of the tips dead 5. All tips dead Leaf necrosis was assessed where necrotic le sions tend to coalesce forming patches of dead tissue on the leaf. Its degree was scored as follows: 1. A few leaves with less than 1/10 of their surface necrotic. 2. Noticeable necrosis in a few branches. 3. Extended necrosis (up to 50% of the leaves affected) in most braches. 4. Very extended necrosis (up to 75% of the leaves affected) in most branches 5. Most leaves necr otic in all branches. Yellowing was assessed as follows: 1. Leaves turn slight ly yellow, no or very little leaf folding. 2. Leaves noticeably yellow, some leaf folding especially in the afternoon. 3. Leaves very yellow, noticeable leaf folding especially in the afternoon. 4. Grayish yellow leaves drooping in the afternoon, some reddish hue appears in the stems.

PAGE 53

53 5. Always droopy, yellow leaves, re ddish slightly dehydrated stems. In all variables a score of 0 was assigned if no symptoms were appa rent. The plants were scored at three points in time: 30, 60 and 120 da ys after planting (DAP), except at Citra in 2006 and Quincy in 2007 where no scoring was done 30 DAP. Traditional mating designs used to estimate genetic variance components are applicable only when parental components are unrelated (H allauer and Miranda, 1988). By using REML to estimate genetic variance components in a mixe d model approach, it is possible to account for the relationship among individuals. A mixed model approach using a single trait Animal Model (Mrode, 2005) was employed to estimate genetic variance components from populations derived from the crosses. Since F2:3 families made up most of the da ta and they mostly changed among environments, the analyses were performed by year and location according to the following model: y = X + ZBuB + ZPuP + ZIuI + ZNAuNA + e (Eq. 2-1) where y is the vector of observation for each individual; ZB and uB are the incidence matrix and vector of random bl ock effects B~NID(0, 2 B), with 2 B 3. ZP and uP are the incidence matrix and vector of random plot effects P~NID(0, 2 P), with 40 P 429. ZI and uI are the incidence matrix and vector of random additive effects I~N((0, 2 AA). ZNA and uNA are the incidence matrix and vector of random non-additive ge netic effects NA~ NID(0, 2 NA). e are the errors, which are NID(0, 2 e). The additive matrix A (Henderson, 1976) has dimens ions 37013x37013 with diagonal elements equal to 1 and off-diagonal elements equal to two times coancestry coefficient (2rxy) between the individuals in the study. The di agonal non-additive matrix NA has dimensions 37013x37013 and was expected to account for each variation not accounted for the A matrix.

PAGE 54

54 The vector of observations y is assumed to be univariate normal with mean E(y)= X and variance-covariance Var(y)= ZB2 BZB + ZP2 PZP + ZI2 AAZI + ZNAZNA + 2 eI. The REML method was used to estimate additive and non-additive variance components by using the ASREML software (Gilmour et al., 2006). Individual pred icted breeding values (BLUPs) were obtained from the solutions for th e individual effects in the above mentioned Animal Model. Mean F2 and F2:3 BLUPs (breeding values) were obtained by averaging the BLUPs of all the individuals in the corresponding population. BLUP Reliabilities (a measure of their accuracy) for parents were obtained by using the formula: 2 21A PREDICTIONr (Eq. 2-2) where 2 PREDICTION is the variance of the pr edicted breeding value and 2 A is the additive variance estimation from the corresponding model. Mean F2 and F2:3 BLUPs were obtained by averaging the squared prediction standard errors for the BLUPs of all th e individuals in the correspondi ng population and dividing them by the square of n (number of squared BLUPs used in the average). Narrow-sense heritability was es timated on individual values, as 2222 2 2 ePNAA A ih (Eq. 2-3) where 2 A is the additive variance, 2NA is the non-additive variance, 2 P is the plot variance and 2e is the residual variance. The variance components were directly provided from the model fitting. The heritability estimates are probably upwardly biased because they may contain genotype-by-environment interaction va riation not accounted for, because calculations were performed for each single environment (Nyquist, 1991).

PAGE 55

55 Standard error of heritability was calculated using the Taylor Series Expansion method (Gilmour et al., 2006). Genetic and phenotypic correlation coefficients were calculated as 22covYX XYaa a gr (Eq. 2-4) where XYacov is the additive covariance between trait X (stunting) and Y (spots); 2Xa and 2Ya are the additive variances for traits X and Y respectively. 22covcov covcovyx xy xy xy xyphph e P NA a phr (Eq. 2-5) where 22 22 2x x x x xeP NA a ph being 2xph the phenotypic variance, 2xa is the additive variance, 2xNA is the non-additive genetic variance, 2xP is the plot variance and 2xe is the residual variance for the trait X. Similar variances apply also for trait Y. Standard errors of correlations were calcula ted using the Taylor Series Expansion method (Gilmour et al., 2006). Results The predominant spotted wilt symptoms in every test were stunting and foliar symptoms. Tip death, leaf necrosis and yello wing incidences never reached more than 5% in any test (Fig 21). Thus, no analysis was performed on them. Epidemics varied considerably in their intensity among tests, ranging from light at Citra 2005 and 2006 to severe at Marianna and Quincy in 2007 (Table 2-2).

PAGE 56

56 Locations with heavier epidemics, like Maria nna, showed greater variability between years than locations with lighter epidemics (Citra). Geographical area seemed important as Citr a (North-Central Florida) and Marianna (Florida Panhandle) in 2006 we re widely different in their epidemics whereas Marianna and Quincy, only 50 miles apart, displayed similar epidemics in 2007. The epidemic progression was also very di fferent among tests. As opposing examples, at Citra 2005 the epidemic changed very little after the firs t assessment date both in incidence (Fig 2-2) and severity (Fig. 2-3) while at Mari anna 2007 there was a steady increase in both. In general, the epidemic progression was rather different among genotypes, both in incidence and severity. Susceptible genotypes (l ike NemaTAM) exhibite d a faster rate of increase in the percentage of symptomatic plan ts, especially in tests with the more severe epidemics (Fig. 2-4, 2-5). Symptom severity progression was also different among genotypes (Fig. 2-6). Most of the genotypes showed a similar average sc ore for both stunting and foliar symptoms (data not shown). In most of the genot ypes a range of severity from zero to five was observed at 30 DAP for both symptoms a lthough the frequencies among genotypes were different. The scores for stunting or foliar symptoms showed unimodal and multimodal distributions with varying degree of dispersion according to the envir onment and genotype. Even homogeneous homozygous genotypes like the pare nts (Fig. 2-7) showed dispersion. As no obvious dominant/susceptible th reshold was observed among classes (Fig. 2-8), the study of inheritance of resistance as a quantitative trait was pursued.

PAGE 57

57Observed Variance Components Mixed models were fitted for each assessm ent date for each test, provided there was enough variation in the data. In the case of Marianna 2006, low spotted wilt incidence (5% stunted plants) would not allow data analysis for the first assessment date. In both tests at Citra and in Ma rianna 2006, singularity in the data matrix forced fit to a model without a non-additive variance. The relative importance of each variance component in the model observed in each test was similar for both variable s analyzed (Table 2-3). The most important variance component in most cases was the residual variance. The second most important variance source was the a dditive variance, which ranged from slightly superior to the residual varian ce (in Citra 2005) to a rather sma ll relative size in Citra 2006 and the first two assessment date in Marianna 2006. Non-additive variance, where calculated, was never different from zero. Block variance was neglig ible most of the time. Plot variance was in some cases rather small, as in Citra 2005 whereas in other cases it was more important than the additive variance, as in Marianna 2006. The proportion of variation accounted for by th e additive and plot sources tended to increase with passing time, with the former increasing more noticeably. Changes in the relative importance of the variance estimates in each year varied between locations. While at Marianna plot and additive variances were rather small in both years, at Citra, their relative importance changed with the year. This was probably due to the different generations tested in both years (only F2s and AP-3 in Citra ). The adjustment provided by the model caused a change in the average stunting and foliar symptoms in some tests (Fig. 2-9), suggesting that the performance of relatives provided an important adjustment on the predic ted performance of individuals. This was especially noticeable

PAGE 58

58 in Marianna 2006 and Quincy 2007 where the estim ated average values for both spotted wilt symptoms were smaller than the observed ones. However, in Citra 2006 and Marianna 2007 the agreement between observed and estimated averages was very good. Heritability Heritability estimates varied widely among tests (Table 2-4). The highest heritability was observed at Citra 2005 where only F2 populations and AP-3 were grown. In the 2006 tests, the heritability was extremely low whereas in the 2007 tests the estimates were close to 0.3. Heritability among assessment dates within a test tended to increas e with time, but the extent of increase depended on the test. The only exception to this trend was the heritability of foliar symptoms at Citra 2005 which diminished as the season progressed. Heritability estimates were quite precise as suggested by the sm all standard errors. Phenotypic and Genetic Correlations In some assessment dates the calculation of correlations was hampered by estimation problems such as lack of parameter convergence in the REML process or singularity in the data matrix. In other cases the REML estimate wa s calculated by bounding it w ithin its theoretical space so no standard error was available. Correlations between stunting and foliar symp toms were high in value and with small standard errors in every test and assessment date (Table 2-5). Phe notypic correlations ranged from 0.80 to 0.93 with the test at Citra 2005 show ing the lowest values in each assessment date. Correlation coefficients from tests in 2007 were very similar. Genotypic correlations were higher than the phenotypic ones, ranging from 0.88 to 0. 99. In the second and third assessment dates, three tests (Citra 2005, and Marianna and Qu incy 2007) showed almost perfect genetic correlation.

PAGE 59

59 As the genetic correlation between stunti ng and foliar symptoms was very high, the analysis of breeding values (BLUPs ) is presented only for stunting. Breeding Values (BLUPs) The BLUPs obtained from each test showed a clear distinction among the susceptible (NemaTAM) and the resistant parents (AP-3, DP-1 and NC94002) (Fig. 2-10). The only exception was Citra 2005, where the only resist ant parent grown was AP-3. In most tests NC94002 displayed the best (smallest) breeding va lue for stunting in every test it was in, whereas DP-1 was better than AP-3 in 3 out of 4 tests. A similar pattern was also observed for BLUPs for foliar symptoms (data not shown). There was, however, an important variation in the BLUPs of each parent among tests. This is so mewhat expected because, although most of the individuals in each test we re genetically related, they were not identical. The reliability of the breeding values (which provides a measure of their accuracy) was intermediate for all tests, except for the ones in 2006 (Fig. 2-11) in which the average reliability for individuals in segregating generations was zero The parents in each test had intermediate to high reliability because of their great number of relatives included in each test, as each parent was grown alongside with its F1, F2, F2:3 and even backcross individuals. Both tests in 2007 showed the most damaging epidemics and the best breeding value reliabilities. Consequently, their comparison fo llows. The rank correlati on among the generationmean BLUPs at both 2007 tests was highly significant (r=0.93 p<0.0001) suggesting a very similar breeding value of each generation ir respective of the location in that year. Generation BLUPs When average BLUPs were calculated for the different generations in both 2007 tests, the best ones belonged to the RxR (resistant with resistant) cross (Figs. 2-12 and 2-13). Among the

PAGE 60

60 SxR (susceptible with resistant) crosses, those invol ving the most resistant parent (NC94002) were the best at Marianna. Reciprocal crosses between AP-3 and NemaTAM had very similar average F2 BLUPs in both locations, implying no maternal effect. The average BLUPs for all F3 populations were sim ilar to the ones for F2 populations at both locations. This suggests that the 25 F2 plants sampled in 2006 in each F2 population were able to capture most of the vari ability present in that population. The average BLUP for F2 and F3 populations were usually intermediate to their parents, suggesting additivity. The only exception was the F3 population for the RxR cross at Quincy 2007, which had a higher (worse) BL UP than its most susceptible parent (DP-1). However, this could have been caused by the fact that the sampling of the F2 plants in the previous year didnt reflect correctly the true geneti c composition for that population. Individual BLUPs The percentage of individuals with BLUPs be tter than their best parent varied widely depending on the cross and the test. In general an d as expected (because of increasing additive variance due to selfing), the comparison between F2 and F2:3 populations showed the latter having both higher individual BLUP variability (Figs. 2-14 and 2-15) and higher percentage of individuals with BLUP superior to their best parent (Fig. 2-16). Comparing both tests grown in 2007, it can be seen that in Quincy 2007, 30% and 48% of F2 and F3 individuals (respectively) had better breeding values than NC94002 in the cross between this line and DP-1 (Fig 2-16). Surprisingly, the sa me cross in Marianna 2007 only showed 6% and 3% (F2 and F3) of individuals better than NC 94002 and it failed to produce an individual with breeding value better than NC94002 in the other th ree tests (data not shown). The remaining cross involving NC94002 as a parent also showed a diffe rence in the percentage of

PAGE 61

61 individuals with better (smaller) BLUP than NC94002 observed between tests in 2007. In this case, however, the difference between tests grown in 2007 was less drastic than in the cross DP-1 x NC94002. While the crosses between AP-3 a nd NemaTAM showed small percentages of good individuals in Marianna, they fa iled to deliver better BLUPs than AP-3 in Quincy. Meanwhile, the cross between NemaTAM and DP-1 didnt produce any individual with better breeding value than DP-1. Family BLUPs In four out of five crosses th ere were individuals with better BLUPs than their best parent. However, the proportion of families with averag e BLUPs for stunting bett er than their best parent was rather small (Fig. 2-16). Most cro sses produced one or two superior families in Marianna 2007. The only exception to this gene ral trend was the cross between two resistant parents (DP-1/NC94002), which produced 13 out of 25 F2:3 families with better (smaller) average BLUPs than the most resistant parent in Quincy 2007. However, the same cross only produced one superior family in Marianna 2007. Discussion Locations have generally different epidemic patt erns, particularly if they are significantly distant and contain different agroecosystems (Culbreath et al., 2003; Groves et al., 2003). Among the locations tested in this study, Citra typically does not have serious spotted wilt epidemics even under agronomic practices that increase the likelihood of strong epidemics (Tillman, pers. comm., Chapter 1 of this dissert ation). One likely explanation is that the agroecosystem of Marion County (where Citra is located) is dominated by warm-season grass pastures while crops that are TSWV host sp ecies are a small minor ity (2002 Census of Agriculture).

PAGE 62

62 Meanwhile, both Marianna and Quincy have ro utinely severe spotted wilt epidemics (T. Momol, pers. comm., Tillman et al., 2007, Culbreath et al., 2005). The important proportion of farmland devoted to susceptible cr ops, like peanut, tobacco, toma to and vegetables (2002 Census of Agriculture) and/or the presence of weed speci es that are better hosts of thrips during early spring could explain the chronica lly strong epidemics in these tw o locations (Kucharek et al., 1990; 2002 Census of Agriculture, Northfield, 2005). In the present study, there were different ep idemic patterns among genotypes with the most resistant genotypes showing a slower progressi on, most noticeably unde r severe epidemics. Similar findings have been reported previously (Culbreath et al.,, 1997; Murakami et al.,, 2006; Chapter 1 of this dissertation). Five typical spotted wilt symptoms were se lected from the literature to be assessed (Demsky & Reddy, 2004, Culbreath et al., 2003), but only stunting and foliar symptoms reached significant incidences. Yellowing has been reported in Texas as a very important symptom of TSWV (Mitchell, 1996). Yellowing (chlorosis) and stunting in tobacco are caused by TSWV NSM protein (Prins et al., 1997) which increases the molecular exclusion limit of plasmodesmata causing traffic disruption of substances across them. Thus, it would be expected that both symptoms would be present frequently when a plant displays spotted wilt symptoms Nonetheless, in the presen t study stunting was common but chlorosis was observed only occasionally. This could be due to the fact that the predominant viral sequence present in the tests mainly cause s stunting and foliar symptoms (Nagata et al., 1993; Mandal et al., 2006) The high genetic correlation between stunting and foliar symptoms suggests either pleiotropy acting on the same di rection or gametic phase disequilibrium (Lynch and Walsh,

PAGE 63

63 1998). If the latter were the cas e, the genetic determinants of both types of symptoms would have to be tightly linked, as thousands of F3 individuals were assesse d and some recombinants should have occurred thus reducing the correla tion. Similar incidences among genotypes at 30 DAP suggested that thrips feeding preference wa s not an issue, in agreement with previous findings (Culbreath et al., 1996; 1997). In spite of the fact that the parents are inbred there was still a wide range of severity in the symptoms suggesting either different inoculation times or differential progression of the disease among plants within a genotype. Taking into account that at least a week is required to develop symptoms (Hoffman et al., 1998) it seems clear that a potential peri od of three weeks of inoculum exposure could cause a wide range of symptom severities very early in the season. Incomplete penetrance of resistance to TSWV has been reported in tomato when inoculated by thrips (Rosello et al., 2001). This could also explai n the variable symptom severity observed in the present study, par ticularly in the resistant parent s. Whichever the case, the score distribution didnt suggest the use of traditiona l Mendelian segregation analysis (Lynch and Walsh, 1998) so a quantitative approach to analyze the symptoms scores as polychotomous variables was necessary. The heritability values varied noticeably among tests, which is usually the case when calculating estimates even from similar popul ations (Nyquist, 1991; Lynch and Walsh, 1998). However, the small standard errors suggested the estimates were rather precise. In the present study three unrelated resistan t genotypes, which represent the different sources of resistance to spotte d wilt known to date (D. Gorbet, pers. comm.) were used. By crossing them to a susceptible parent, segregatin g populations with wide variability in spotted wilt resistance were obtained.

PAGE 64

64 The Animal Model utilizes all relationship s among individuals in a test by using a numerator relationship matrix (Mrode, 2005) thus accounting for most of the additive variance which results in more accurate estimations of variance components and breeding values. The fact that most of the individuals on each test were related to some extent, even when belonging to different crosses, provided better estimates of the additive covariances than the use of traditional heritability estimation methods (Henderson, 197 6; Lynch and Walsh, 1998). Although similar kinds of populations were used each year, the variance components were quite variable among tests. This is sometimes the case, even when re peating tests few days ap art (Chapter 1 of this dissertation), doing them in a laboratory setting (Rapp & Juntila, 2001) or in the field (Finne et al., 2000). Spotted wilt epidemics are highly variable among locations and even from year to year at a single location (Culbreath et al ., 2003). The range of intensity of the epidemic observed among the tests was certainly wide. Apparently the geographical location was more important than the year, in accordance with the results describe d in the Chapter 1 of this dissertation. Heritability estimates are influenced by the re lative amount of total variation due to genetic causes (Lynch and Walsh, 1998). The wide range of epidemic intensity probably accounted for a large part of the variability in heritability estima tes. Inaccurate heritability estimates are often obtained when the frequency of a category in a polychotomous variable is very high across the whole population of individuals being tested (Yang et al., 1998). In the present study most of the individuals showed a reasonable di spersion in the score frequencies. The use of individual values is known to provide better heritability estimates than plot means under most conditions when a REML approach is used (Huber et al., 1994). Consequently, even wh en the heritability estimates obtained here were quite variable, they are expected to be accurate. Herita bility estimates in the

PAGE 65

65 present study were usually in the medium to lo w range, with the values increasing as the season progressed. This was expected as the difference in symptom severity in genotypes with different resistance tends to increase with time (Culbr eath et at., 1997; Murakami et al., 2006) unless the epidemic reaches a final intensity very early in th e season (Chapter 1 of this dissertation). If the additive variance and consequently the heritability tend to increase toward harvest time, selection for resistant genotypes would be more effective when conducted closer to harvest (Hallauer & Miranda, 1988). Heritability estimates for a trait is at the core of any individual multi-trait selection index. Its magnitude determines the im portance (weight) thats assigned to the trait while selecting individual s based on that index (Ha llauer and Miranda, 1988). In the University of Florida Peanut Br eeding Program (UFPBP), performance of segregating populations against TSWV has been assessed based on a holistic score assigned to plots or plants in which all s potted wilt symptoms are consider ed (Gorbet, 1999, B. Tillman, pers. comm.). Simmons (1979) pointed out that ev ery breeder has in his/her mind a multi-trait selective index but it is usually not put into writing. Taking into account the high genetic correlation between the most frequent spotted wilt symptoms and the workable value of heritability estimates here reported, it seems reasonable that the inclusion of spotted wilt resistance as a part of this unwritten sele ctive index could be the cause of the observable improvement in the overall level of spotted wilt resist ance in the breeding populations observed in the UFPBP compared to older, but good perf orming genotypes from the pre-TSWV era like Florunner, Sunrunner or SunOleic 97R (Culbr eath et al., 2005; Tillman et al., 2007). The variable reliability observed among and within tests could only be explained by a variable importance of the environment, as th e genetic structure and the type of genetic relationships among individuals were quite similar in each test. When using full sib records (as is

PAGE 66

66 our case in the F2:3 families), reliability always increases as the number of tested sibs increases but with decreasing relative gain (Mrode, 2005). The increase in reliability from increasing the number of sibs is larger when heritability is lower. Although in this study the number of sibs tested varied among tests, this doesnt seem to account for the observed variation in the BLUP estimate reliabilities Even within each test and among F2:3 families of the same cross, the reliabilities had important variation. This seem s to point to experimental noise as the most probable cause of this fact (Mrode, 2005). As expected under the assumption of additive mode of action for spotted wilt resistance, the cross between resistant parents produced the population with best BLUPs. Culbreath et al. (2005) reported that genotypes that inherite d spotted wilt resistance from PI 203395 were less resistant than genotypes derived from another resistance source (PI 576638). They suggested the possibility that both sources could pr ovide different resistance genes. In the present study, both resistance sour ces were represented by DP-1 and NC94002, respectively. This cross provided few or no indi viduals with better bree ding value than NC94002 in four tests but it provided a hi gh percentage of superior indivi duals in Quincy 2007. This could point to different resistance mechanisms for gene s coming from these two resistance sources, as suggested by Culbreath et al (2005). The reason these differ ent mechanisms would only be appreciated in one out of five tests is unknown. Perhaps Quincy provided a better discriminating environment for such effect. This ability of some environments to be better at discriminating has also been cited for other tra its and species (Hall and Van Sa nford, 2003; Venuprasad et al., 2007). All the SxR crosses produced populations with similar breed ing values. Most populations had a rather small proportion of i ndividuals with better BLUP than their resistant pa rent. In fact,

PAGE 67

67 the average breeding value of the F2 and F3 populations were usually intermediate to their parents breeding values. Both fact s seem consistent with an addi tive mode of action (Falconer & MacKay, 1996). As additive variance increases with self ing (Nyquist, 1991), individual BLUP tend to increase so more individuals in each distributional extreme can be found in F3 than in F2. Although further selfing continues to increase the additive vari ance, resource limitations always force some type of selection in early generations (Simmonds, 1979). The low individual heritability but good reliability of family BLUP suggests that taking into consideration the family performance for s potted wilt resistance when selecting individuals among F2:3 families, as is frequently practiced in peanut breeding programs in Southeastern USA, is a safe breeding strategy (Falconer and Mackay, 1996; Halla uer and Miranda, 1988). When selecting in populations derived from the re sistant parents used here, inclusion of spotted wilt resistance a part of a selective individual multi-trait index is acceptable. However, it should be given a moderate weight becau se of its modest heritability. It seems clear that RxR crosses would provide better populations to select for spotted wilt resistance and those having NC94002 as a parent w ould display the best response to selection. Conclusion The use of the Animal Model provided accura te and precise heritability estimates. They ranged from 0.01 to 0.71, but were most frequen tly in the low to medium range. The estimates increased as the epidemics progressed. The almost exclusive spotted wilt symptoms detected in each test we re stunting and foliar symptoms (spots or mosaics). Tip death, leaf necr osis and yellowing (chlorosis) were rare. There was high phenotypic and genotypic correlation among stunting and foliar symptoms suggesting either pleiotropy or a very strong coupli ng linkage among their genetic determinants.

PAGE 68

68 Breeding values for the different generations of the five crosses tested seemed to suggest additivity as the main mode of action in the de termination of the resistance to spotted wilt. The resistant parents produced populations with si milar breeding values when crossed to the susceptible parent. The population from a cross between resistant parents exhibited the best breeding values for resistance to spotted wilt. Based on the calculated heritability estimates pedigree selection w ithin the populations used in this study should not put too much weight on individual selection in early generations based on resistance to TSWV. More emphasis on incl uding resistance as a pa rt of a multi-trait individual selective index w ith a corresponding moderate wei ghting seems recommendable. Additionally, familial performance can provide surrogate estimations of an individuals real resistance.

PAGE 69

69 Table 2-1. Sowing date, replicat ion number and design of tests assessing performance against spotted wilt in five peanut crosses in Florida. Plots in each generationa Test Parents F1 F2 F3 b BC Design and block number d Sowing date Citra 13c 1 2-5 NA NA CR w/variable replications 5/30 Citra 4-8 1-2 1-9 2 NA RCB, 2 blocks 5/24 Marianna 6-12 1 6d 2-3 NA RCB, 3 blocks 4/28 Marianna 6-12 1-2 9 2 0-2 RCB, 3 blocks 4/24 Quincy 6-12 0-2 9 2 0-2 RCB, 3 blocks 4/25 a Plot number in each generation varied depending on seed availability. b All F2:3 families changed from year to year and some changed from test to test within years. c Only AP-3 grown. d F2 DP-1 / NC94002 not grown Table 2-2. Mean (S.D.) score for each spotted wilt symptom at 30, 60 and 120 days after planting, at each of five field tests in whic h five peanut populations were evaluated in Florida. 2005 2006 2007 Citra Citra Marianna Marianna Quincy 30 Days After Planting n 1073 N/A 10713 10669 N/A Stunting 0.66 (1.58) N/A 0.17 (0.79) 0.32 (0.99) N/A Spots and Mosaic 1.15 (1.92) N/A 0.17 (0.69) 0.33 (1.02) N/A Tip Death 0.08 (0.54) N/A 0 (0) 0.01 (0.14) N/A Leaf Necrosis 0.08 (0.51) N/A 0 (0.02) 0 (0.07) N/A Yellowing 0 (0) N/A 0 (0) 0 (0) N/A 60 Days After Planting n 1073 3867 10646 10538 10032 Stunting 0.69 (1.63) 0.36 (1.06) 0.99 (1.73) 1.27 (1.81) 1.24 (1.91) Spots and Mosaic 1.24 (1.98) 0.46 (1.22) 1.24 (1.92) 0.15 (2.01) 1.43 (2.13) Tip Death 0.09 (0.58) 0.01 (0.15) 0.01 (0.16) 0.02 (0.2) 0.09 (0.47) Leaf Necrosis 0.09 (0.54) 0 (0.03) 0.01 (0.13) 0 (0.1) 0 (0.08) Yellowing 0 (0) 0.01 (0.16) 0.00 (0.04) 0.10 (0.5) 0.06 (0.39) 120 Days After Planting n 1071 3809 10246 10028 9825 Stunting 0.68 (1.48) 0.56 (1.29) 2.34 (1.85) 3.27 (1.69) 3.21 (1.53) Spots and Mosaic 1.23 (1.77) 0.75 (1.51) 3.13 (1.76) 3.63 (1.7) 3.66 (1.59) Tip Death 0.07 (0.49) 0.01 (0.15) 0.01 (0.11) 0.03 (0.27) 0 (0.07) Leaf Necrosis 0.03 (0.28) 0 (0.03) 0 (0.03) 0 (0.02) 0 (0.02) Yellowing 0 (0) 0.01 (0.12) 0.04 (0.12) 0.05 (0.39) 0.04 (0.34) N/A: Not Available

PAGE 70

70 Table 2-3. REML variance estimates for stunting and foliar symp toms caused by TSWV in populations derived from five peanut crosses tested at Citra, Florida in 2005 and 2006, Marianna, Florida in 200 6 and 2007 and Quincy, Florida in 2007. Stunting Foliar symptoms Citra 2005 Assessment date 2 Block 2 Plot 2 A2 NA2 e 2 Block 2 Plot 2 A2 NA2 e 30 DAP a 0.040 2.539 1.699 a 0.155 4.096 a 2.050 60 DAP a 0.040 2.539 a 1.699 a 0.198 2.315 a 2.889 120 DAP a 0.042 2.827 a 1.114 a 0.278 1.048 a 2.455 Citra 2006 30 DAP c c c c c C c c c c 60 DAP 0.001 0.017 0.021 a 1.070 0.002 0.018 0.030 a 1.423 120 DAP 0 0.054 0.036 a 1.539 0.000 0.083 0.065 a 2.110 Marianna 2006 30 DAP b b b b B B b b b b 60 DAP 0 0.106 0.087 a 2.774 0 0.149 0.133 a 3.403 120 DAP 0.004 0.382 0.305 a 2.641 0.012 0.358 0.358 a 2.298 Marianna 2007 30 DAP 0 0.014 0.013 a 0.946 0.001 0.011 0.017 a 1.012 60 DAP 0.015 0.102 0.235 0 2.955 0.034 0.140 0.320 0 3.578 120 DAP 0.095 0.250 0.698 0 1.739 0.056 0.211 0.790 0 1.783

PAGE 71

71Table 2-3. Continued Stunting Foliar symptoms Quincy 2007 Assessment date 2 Block 2 Plot 2 A2 NA2 e 2 Block 2 Plot 2 A2 NA2 e 30 DAP b b b b b b b b b b 60 DAP 0.003 0.197 0.197 0 3.227 0.004 0.248 0.268 0 3.983 120 DAP 0.008 0.243 0.640 0 1.396 0.019 0.274 0.738 0 1.394 2 Block: Block variance; 2 Plot: Plot variance; 2 A: Additive variance; 2 NA: Non-additive variance. a: term not included in the model. b: date not assessed. c: incidence too low to allow analysis

PAGE 72

72 Table 2-4. Heritability (S.E.) estimates for stunting and foliar symptoms caused by TSWV on peanut populations from five crosses at different assessmen t dates in five tests in Floirida. Estimates were calculate d using univariate Animal Models. Test Variable 30 DAP 60 DAP 120 DAP Citra 2005 stunting 0.59 (0.05) 0.59 (0.05) 0.71 (0.04) foliar symptoms 0.65 (0.04) 0.43 (0.08) 0.28 (0.11) Citra 2006 stunting a 0.02 (0.01) 0.02 (0.01) foliar symptoms a 0.02 (0.01) 0.03 (0.01) Marianna 2006 stunting b 0.03 (0.01) 0.09 (0.02) foliar symptoms b 0.04 (0.01) 0.12 (0.03) Marianna 2007 stunting 0.01 (0.01) 0.07 (0.01) 0.26 (0.03) foliar symptoms 0.01 (0.01) 0.08 (0.02) 0.28 (0.03) Quincy 2007 stunting b 0.05 (0.02) 0.28(0.03) foliar symptoms b 0.06 (0.02) 0.31(0.03) a: incidence too low to allow analysis. b: date not assessed. Table 2-5. Phenotypic and genetic correlation (S.E.) estimates between stunting and foliar symptoms caused by TSWV on peanut populations from five crosses at different assessment dates in five tests in Florida. Estimates were calcul ated using univariate Animal Models. 30 DAP 60 DAP 120 DAP Citra 2005 0.80 (0.02) / 0.88 (0.03) 0.82 (0.02) / 0.99 (0.01) 0.83 / 0.99 e Citra 2006 a b 0.91 (0.1) / 0.95 (0.03) Marianna 2006 c 0.93 (0.01) / 0.95 (0.03) 0.84 (0.01) / 0.95 (0.02) Marianna 2007 d 0.90 / 0.99 e 0.92 (0.01) / 0.99 (0.0.1) Quincy 2007 c 0.91 / 0.99 e 0.92 (0.01) / 0.99 (0.01) a: incidence too low to allow analysis. b: Singularity in datamatrix c: date not assessed. d: parameters didn't converge. e: Standard Error not available because REML estimate was bounded.

PAGE 73

73 0 10 20 30 40 50 60 70 80 90 100Citra '05Citra '06Marianna '06Marianna '07Quincy '07% plants with each sympto m Stunting Spots Tip Death Leaf Necrosis Yellowing Figure 2-1. Incidence of spotted wilt sympto ms at 120 days after planting in populations from five peanut crosses tested in fiv e Florida environments.

PAGE 74

74 0 10 20 30 40 50 60 70 80 90 100 30 DAP60 DAP120 DAPAssessment date% symptomatic plants Citra05 Citra06 Marianna06 Marianna07 Quincy07 Figure 2-2. Percentage of plants displaying spotted wilt symptoms in five peanut crosses in five field tests assessed at three dates in five Florida environments.

PAGE 75

75 0 0.5 1 1.5 2 2.5 3 3.5 30DAP60DAP120DAPAssessment datesMean stunting score Citra '05 Citra '06 Marianna '06 Marianna '07 Quincy '07 Figure 2-3. Stunting severity progression in five peanut crosses, in five field tests assessed at three dates in three Florida locations.

PAGE 76

76 0 10 20 30 40 50 60 70 80 90 100 30 DAP60 DAP120 DAP% symptomatic plants AP-3 DP-1 NC94002 NemaTAM Figure 2-4. Spotted wilt incide nce in four peanut genotypes at three assessment dates at Marianna, Florida in 2006.

PAGE 77

77 0 10 20 30 40 50 60 70 80 90 100 30 DAP60 DAP120 DAP% symptomatic plants AP-3 DP-1 NC94002 NemaTAM Figure 2-5. Spotted wilt incide nce in four peanut genotypes at three assessment dates at Marianna, Florida in 2007.

PAGE 78

78 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 30DAP60DAP120DAPAssessment dateStunting F2 NemaTAM / DP-1 F3 NemaTAM / DP-1 F2 DP-1 / NC94002 F3 DP-1 / NC94002 NC94002 NemaTAM DP-1 Figure 2-6. Observed stunting severity among different populations from five peanut crosses, at thr ee different dates at Marian na, Florida in 2007.

PAGE 79

79 Figure 2-7. Frequency dist ributions for TSWV-induced stunting scores among parent s of five peanut crosse s field tested at Marianna, Florida in 2007.

PAGE 80

80 Figure 2-8. Frequency dist ributions for TSWV-induced stunting scores among F2 populations from four pea nut crosses field tested at Marianna, Florida in 2007.

PAGE 81

81 0 0.5 1 1.5 2 2.5 3 3.5 4Citra '05 Citra '06 Marianna '06 Marianna '07 Quincy '07mean score Obs. Stunting Estim. Stunting Obs. Foliar symptoms Estim. Foliar symptoms Figure 2-9. Observed vs. estimated overall severity in stunting and foliar symptoms ca used by TSWV on five peanut crosses in five field tests in Florida.

PAGE 82

82 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 Citra '05Citra '06Marianna '06Marianna '07Quincy '07BLUPs DP-1 AP-3 NC94002 NemaTAMR=0.65 R=0.63 R=0.61 R=0.61 R=0.44 R=0.44 R=0.44 R=0.41 R=0.67 R=0.44 R=0.40 R=0.66 R=0.89 R=0.94 R=0.92 R=0.92 R=0.92 R=0.89 R=0.89 R=0.90 R = reliability of BLUP Figure 2-10. Best linear unbiased pr edictors (relative to test average) for TSWV-induced stunti ng in parents of five peanut cro sses assessed in five field tests in Florida.

PAGE 83

83 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Citra '05Citra '06Marianna '06Marianna '07Quincy '07Reliabilit y Maximum Average Minimum Figure 2-11. Breeding values reliability fo r TSWV-induced stunting among individuals in segregating populations derived from f ive peanut crosses evaluated in five field tests.

PAGE 84

84 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5A P 3 F 3 A P 3 / N e m a T A M F 1 A P 3 / N e m a T A M F 2 A P 3 / N e m a T A M A P 3 / N e m a T A M 2 F 2 N e m a T A M / A P 3 F 1 N e m a T A M / A P 3 F 3 N e m a T A M / A P 3 N e m a T A M N e m a T A M* 2 / D P 1 F 3 N e m a T A M / D P 1 F 2 N e m a T A M / D P 1 F 1 N e m a T A M / D P 1 N e m a T A M / D P 1 2 D P 1 F 3 D P 1 / N C 9 4 0 0 2 F 2 D P 1 / N C 9 4 0 0 2 N C 9 4 0 0 2 F 2 N e m a T A M / N C 9 4 0 0 2 F 3 N e m a T A M / N C 9 4 0 0 2 N e m a T A MStunting BLUPs Figure 2-12. Generation-mean best linear unbi ased predictors for TSWV-induced stunting in populations from five peanut crosses tested at Quincy, Florida in 2007 (a ll reliabilities were above 0.9).

PAGE 85

85 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5AP-3 F3 AP-3 / N emaTAM F1 A P -3 / Ne m a TA M F2 AP-3 / NemaTAM AP-3 / NemaTAM*2 F2 N e m a TA M / AP3 F1 N e m a TA M / AP-3 F 3 NemaT A M / AP 3 Nema T AM Ne m a T AM*2 / DP1 F3 NemaTAM / DP-1 F2 Ne m a T AM / DP-1 F1 Ne m a T AM / D P 1 Nema T AM / DP-1*2 DP1 F 3 DP-1 / NC9 4 002 F 2 DP1 / N C9 4 002 F 1 D P -1 / NC 9 40 02 NC 9 4002 F1 NemaTAM / NC9400 2 F3 Nem a T AM / NC 9 40 0 2 F2 Ne maTAM / N C 94 0 0 2 Ne m aT AMStunting BLUPs Figure 2-13. Generation-mean best linear unbi ased predictors for TSWV-induced stunting in populations from five peanut crosses tested at Marianna, Florida in 2007 (all reliabilities were above 0.9).

PAGE 86

86 0 0.5 1 1.5 2 2.5 3 AP-3 / NemaTAMNemaTAM / AP-3NemaTAM / DP-1NemaTAM / NC94002 DP-1 / NC94002Stunting BLUPs F2 F3 Figure 2-14. Variability of best linear unbiased predictors for TSWV-induced stunting in the F2 and F3 generations of five peanut crosses tested at Marianna, Florida in 2007.

PAGE 87

87 0 0.5 1 1.5 2 2.5 3 3.5AP-3 / NemaTAMNemaTAM / AP-3NemaTAM / DP-1NemaTAM / NC94002 DP-1 / NC94002Stunting BLUPs F2 F3 Figure 2-15. Variability of best linear unbiased predictors for TSWV-induced stunting in the F2 and F3 generations of five peanut crosses tested at Quincy, Florida in 2007.

PAGE 88

88 0 5 10 15 20 25 30 35 40 45 50AP-3 / NemaTAM NemaTAM / AP-3 NemaTAM / DP-1 NemaTAM / NC94002 DP-1 / NC94002% of individuals Marianna '07 F2 Marianna '07 F3 Quincy '07 F2 Quincy '07 F3(2) (1) (13) (1)(1) (1) Figure 2-16. Percentage of individuals and (number of F3 families) displaying BLUPs for TSWV-indu ced stunting above their best parent, in each of five pea nut crosses tested at Marianna and Quincy, Florida in 2007.

PAGE 89

89 CHAPTER 3 ARTIFICIAL INOCULATION STUDIES IN THE TSWVPEANUT PATHOSYSTEM Introduction Tomato spotted wilt virus ( Bunyaviridae:Tospovirus) is a species of quasi-spherical enveloped particles containing three single-stranded RNA (Kor melink et al., 1992). It is a cosmopolitan pathogen of 1090 plant species in 85 fa milies (Parrella et al., 2003). It is vectored in a propagative and circulative manner exclusively by thrips of the genera Thrips and Frankliniella (Ullman et al., 2002). Wherever TSWV in cidence has increased enough to cause economic losses, it has remained a chronic prob lem in many economically important crops. This is the case in the Southeastern USA, where the peanut ( Arachis hypogaea L.) crop has suffered intermittent heavy losses since 1993. The initial increasing tre nd in TSWV incidence has now been reversed (Brown et al., 2007) by a combin ation of factors includ ing planting date, stand density, row pattern, insecticid e use and resistant cultivars (Culbreath et al., 2003). The resistance level of the cultivar is the most important tool in the management of this disease (Brown et al., 2007). So far most resistance asse ssments have been conducted in the field and relied on natural epidemics. This means that a valid characterization of each genotype has been resource-consuming involving several seasons and locations. As an alternative approach, artificial inoculation methods have been developed to reduce the time needed to assess a genotype targeting its commercial release as a cultiv ar and also to diminish the overall resource requirements. Several artificial inocula tion methods have been described for the TSWV-peanut pathosystem (Halliwell and Philley, 1974; Clemente et al., 1990; Pereira, 1993; Hoffman et al., 1998). According to published resu lts the most consistent one was developed by Mandal and coworkers (2001). In this method important factor s are the antioxidants in the extraction buffer

PAGE 90

90 and the type and amount of abrasives used in th e rubbing. Oxidation, which reduces the life of TSWV outside the cell, seems to be a common remark among the bibliographical sources and thus different concentrations of several reduci ng chemicals have been evaluated (Halliwell and Philley, 1974; Clemente et al., 1990; Pereira, 1993; Mandal et al., 2001). With reference to abrasives, Clemente et al. (1990) found no differen ces in the rates of TSWV artificial inoculation using various grit sizes of Ca rborundum whereas Mandal and cowo rkers (2001) reported that the type of added abrasive in the inoculum was very important. Some other factors have been cited in the literature as influencing the outcome on inoculation experime nts. Ng et al. (2004) reported that the concentration of virus that effectively reached the target tissues was very important in the transmission efficiency of Lettuce infectious yellows closterovirus to lettuce. Even with standardized conditions, different factors can produce varied act ual damage after inoculation, which could lead to an inconsis tent number of viral particles entering the leaf (Pereira, 1993). Consequently a better assessmen t of the actual damage inflicted while rubbing could be extremely useful in ruling out this factor as a cause of variability. A fact that has hampered the resistance assessment in the TSWV-peanut pathosystem is that external symptoms do not always reflect the concentration of TSWV in the plant (Resende et al., 2000; Mandal et al., 2001; Lyerly et al., 2007) Additionally, asymptomatic TSWV infections in peanut have been reported (Culbreath et al., 1992). Nonetheless, th e frequency with which plants within a cultivar express symptoms early after inoculation has been suggested as a good indicator of viral titer in the tissue (Kresta et al., 1995) and also of resistance (Culbreath at al., 2003). While an artificial inocula tion method should accelerate the screening process leading to the detection of resistant genot ypes, it should also reduce the e nvironmental variation of plant

PAGE 91

91 reaction through the st andardization of several factors. This should allow a better estimation of the genotypes true reaction to the virus. Consequently, the objective of this series of studies was to determine the relative importance of age of i noculum, virus concentration in the inoculum, and amount of rubbing during inoculation on the frequency of infection. A secondary objective was to determine if there was an association betw een ELISA values and symptom expression. To address these objectives three studies were conducted during the spring of 2005. Materials and Methods Plant Culture Georgia Green, a widely grown runner market-t ype cultivar was used in all tests. This cultivar displays some field resistance to TS WV (Culbreath et al., 1996 ) but it is susceptible under artificial inoculati on (Mandal et al., 2001). One seed was planted in each 164 ml plastic container (Cone-tainer C10, Stuewe & Sons, Corvallis, Oregon) containing all purpose prof essional growing mix consisting of Canadian sphagnum peat moss 75 to 85%, perlite 15 to 20%, and vermiculite 5 to 10 % (Berger Peat Moss, Saint-Modeste, Quebec, Canada) and irrigated every other day with distilled water. No fertilizer was added to the mix. Test plants were grow n until inoculation was performed in a chamber made of a shelf with fluorescent lights (Gro-L ux, Osram Sylvania, Danvers, Massachusetts) all surrounded by a transparent plastic sheet. Conditions inside this chamber were 12-h light period (12 klx intensity) and 23C min. and 34C max. Those seedlings with uniform size and vigor were used for inoculation 12-14 days after plan ting. The average seedling height was variable among tests ranging from five to seven cm, wh ereas the average number of fully expanded leaves ranged from two to five.

PAGE 92

92 Inoculum Preparation Infected leaf tissue from greenhouse grown peanut plants (cv. Georgia Green) were collected and pre-chilled in a refrigerator and ground (1:10 [wt/vol] tissue:buffer unless otherwise stated) with freshly prepared ice-cold 0.01 M potassi um phosphate buffer, pH 7.0, containing 0.2% sodium sulfite and 0.01 M 2-mer captoethanol using a chil led pestle and mortar as described by Mandal et al. (2001). Debris was removed by squeezing the ground extract through a pad of nonabsorbent cotton. To this ho mogenate, Celite 545 (Fisher Scientific, Fair Lawn, NJ) and Carborundum 320 grit (Fisher Scientific) were each added to a final concentration of 1% each. The inoculum was kept on ice until the i noculation process was completed. Sap Inoculation Test plants were dusted with Carborundum on the youngest fully expanded leaf 12-14 days after sowing. Two leaflets (one basal and one apical) were inocul ated by rubbing them four times (unless otherwise stated) with a cotton swab (Johnson & Johnson, Skillman, NJ) dipped in the inoculum. After inoculation the plants were spra yed with distilled water and placed in a growth room at 25/19C, 50% RH, 12-h light period and 15 klx of light intensity and were irrigated every two days using distilled water. Description of Tests Test 1: Effect of elapsed ti me from preparation to inocul ation on infection frequency Three time lapses from the inoculum prepar ation (zero, ten and twenty minutes) were compared. At each time, 10 plants were inoculated using the same inoculum batch and these 30 plants (10x3) were considered a block within a randomized comp lete block design. There were three blocks totaling 90 inoculated plants. Additi onally as controls, six pl ants were rubbed with buffer only plus abrasives and six mo re plants were left untreated.

PAGE 93

93 Test 2: Determining the importance of amount of rubbing on infection rate Leaves were rubbed four, six or eight tim es using a cotton swab. Once prepared the inoculum was used immediately. Ten plants we re inoculated per treatment using the same inoculum batch and these 30 plants (10x3) were considered a block within the RCBD. There were four blocks totaling 120 plants. As controls eight plants were rubbed with buffer only plus abrasives and eight more pl ants were left untreated. Test 3: Evaluating the influence of inoculum concentration on infection rate Two inocula, each using a different tissue:buf fer ratio, were compared (1:10 and 1:20). After obtaining the usual 1:10 inoculum, half of it was allocated in another mortar and a similar volume of buffer was added. Once prepared both inocula were used immediately. Different swabs were used for each level of the dilution factor. Ten plants were inoculated per treatment and these 20 plants (10x2) were considered a block within a RCBD. There were four blocks totaling 80 plants. As controls, ei ght plants were rubbed with buffe r only plus abrasives and eight more plants were left untreated. Imposing Treatments In Test 1 plants were inoculated by rubbing them four times with a cotton swab. The time to inoculate each plant was about ten seconds so for every treatment the real inoculation time between the first and the last plant was almost tw o minutes. In the Test 2 plants were inoculated according to the layout L1-L2-L3-L3-L2-L1-L1-L2-L3-L3-L2-L1 (L means Factor Level)and so on. In Test 3 plants were inoculated accord ing to the layout L1-L2-L2-L1-L1-L2-L2-L1 and so on.

PAGE 94

94 Evaluation of Inoculated Plants and Analysis of Data The measured variables in each experiment were as follows: Appearance of systemic symptoms : as viral lesions on inoculated leaves were not observed, the plants were considered as infected when chlorotic spots followed by mosaic rings and necrotic spots developed in the newly emerging l eaves (systemic symptoms). In their absence the plants were considered healthy. Serological detection of TSWV by ELISA : optical density (OD) values greater than the average value plus 3 times the standard deviati on (cut-off value) of the two negative control wells, belonging to healthy plants of C11-2-39 peanut line, were considered positive for the presence of TSWV. Due to contamination of the negative controls in Test 1, a cut-off value was set taking into account the usual values obtained fo r this type of control. As the highest value ever obtained for this negative control has b een 0.006 in several previous ELISA (data not shown) it was considered reasonably conser vative to use 0.06 as a cut-off value. TSWV infection was confirmed by alkaline pho sphatase labeled DAS-ELISA according to manufacturers instruct ions (Agdia Inc., Elkhart, Indiana). Absorbance was measured at 405 nm with an automated microplate reader (Mode l 680, Bio-RAD, Hercules, CA,USA). Two replications were made on each sample, and averages were used for evaluation. Recording of symptomatic plants and ELISA were done 3 weeks post-inoculation. Two apical leaflets in the youngest l eaf plus one apical leaflet on th e youngest fully expanded leaf and young secondary roots were used for ELISA. If ne w leaves were observed as symptomatic they were used for ELISA instead of using random l eaves. Since the tested tissue was not weighted and the obtained macerate volume was variable, the ELISA values were used only to categorize the plants as infected or not.

PAGE 95

95 In all the three tests, treatments were estab lished by the combination of inoculum batch and the level of the factor being tested (elapsed time, # of rubbings or inoculum dilution). Each treatment was represented by 10 inoculated plants. Both binary variables (appearance of syst emic symptoms and TSWV ELISA detection) were analyzed by Multiple Logistic Regressi on with a binomial distribution and logit link function using SAS (SAS Institute, 2 000). The applied model in Test 1 was logit( )= + 3 1Time +3 1Batch (Eq. 3-1) where is the probability of the plant being symptomatic or being ELISA positive depending on the response variable being analyzed. The parameter i refers to the effect of the i level of a factor (say Time) on the log odds th at the dependent variable equals one of the two possible outcomes, say infected, controlli ng the levels of Batch (Agresti, 1996). Time denotes the amount of elapsed time from inoculum preparation (0, 10 and 20) while Batch denotes the inoculum batch used. The adjusted models for Test 2 were: logit( )= + 3 1Rubbings +4 1Batch; (Eq. 3-2) and logit( )= +4 1Batch; (Eq. 3-3) where rubbings denotes the number of rubbi ngs applied during inoc ulation, being the rest of the terms as described in Test 1. The applied model for Test 3 was logit( )= + 2 1Dilution +4 1Batch (Eq. 3-4) where Dilution denotes the inoculum dilutions tested, being the rest of the terms as described in Test 1.

PAGE 96

96 The logits of the unknown binomial probabilitie s (i.e., the logarithms of the odds) are modeled as a linear function of the Xi. The unknown parameters j are usually estimated by maximum likelihood. The full model containing both elapsed time and i noculum batch as factors was used and in the event of a factor being found non-significant, it was rem oved from the model (Agresti, 1996). Fishers Exact Test was used to detect associ ation between Appearance of systemic symptoms and ELISA status. Results Test 1 The overall percentage of inocul ated plants showing visual symptoms of systemic infection was 18% (most of them with mild severity, Table 3-1) while no local ized symptoms were observed in the inoculated leaves. According to the ELISA test, 43% of the plants were systemically infected. Four out of 16 plants show ing systemic symptoms failed to be detected by ELISA. Nonetheless, the association between systemic symptoms and ELISA status was statistically significant (p=0.0289) a ccording to Fishers Exact Test. Neither the elapsed time from inoculum prepar ation nor inoculum batch were significant (p=0.09 and 0.13 respectively) in the logistic regression for syst emic symptoms. In the case of the serological status, both factors were found significant (p=0.02 and 0.05). The obtained Maximum Likelihood Estimat es can be seen in Table 3-2 The Prediction Equations obtained for the logit of the probability of a Positive ELISA result under each treatment were then: For Time 0 and Batch 3, logit( )=-0.3336+0 Time+0 Batch For Time 0 and Batch 1, logit( )=-0.3336+0 Time+(-0.7978 Batch)

PAGE 97

97 For Time 0 and Batch 2, logit( )=-0.3336+0 Time+(0.6253 Batch) For Time 10 and Batch 3, logit( )=-0.3336+(0.7680 Time)+0 Batch For Time 10 and Batch 1, logit( )=-0.3336+(0.7680 Time)+(-0.7978 Batch) For Time 10 and Batch 2, logit( )=-0.3336+(0.7680 Time)+(0.6253 Batch)For Time 20 and Batch 3, logit( )=-0.3336+(-0.9494 Time)+0 Batch For Time 20 and Batch 1, logit( )=-0.3336+(-0.9494 Time)+(-0.7978 Batch) For Time 20 and Batch 2, logit( )=-0.3336+(-0.9494 Time)+(0.6253 Batch) These prediction equations can be expressed as the predicted probability of a plant being infected. For example, using the third inoculum batch twenty minutes after it was prepared, the predicted probability of a plant to be ELISA positive would be: {exp[-0.3336+(-0.9494 Time)+(0.6253 Batch)}/ {1+ exp[-0.3336+(-0.9494 Time)+(0.6253 Batch)}= 0.34 Similarly, the predicted probabilities for the other treatments were: For Time 0 and Batch 3, {exp[-0.3336+(0 Time)+(0 Batch)}/{1+ exp[-0.3336+(0 Time)+(0 Batch)}= 0.42 For Time 0 and Batch 1, {exp[-0.3336+(0 Time)+(-0.7978 Batch)}/{1+ exp[-0.3336+(0 Time)+(-0.7978 Batch)}= 0.24 For Time 0 and Batch 2, {exp[-0.3336+(0 Time)+(0.6253 Batch)}/{1+ exp[-0.3336+(0 Time)+(0.6253 Batch)}= 0.57 For Time 10 and Batch 3,

PAGE 98

98 {exp[-0.3336+(0.7680 Time)+(0 Batch)}/{1+ exp[-0.3336+(0.7680 Time)+(0 Batch)}= 0.61 For Time 10 and Batch 1, {exp[-0.3336+(0.7680 Time)+(-0.7978 Batch)}/{1+ exp[-0.3336+(0.7680 Time)+(-0.7978 Batch)}= 0.41 For Time 10 and Batch 2, {exp[-0.3336+(0.7680 Time)+(0.6253 Batch)}/{1+ exp[-0.3336+(0.7680 Time)+( 0.6253 Batch)}= 0.74 For Time 20 and Batch 3, {exp[-0.3336+(-0.9494 Time)+(0 Batch)}/{1+ exp[-0.3336+(-0.9494 Time)+(0 Batch)}= 0.22 For Time 20 and Batch 1, {exp[-0.3336+(-0.9494 Time)+(-0.7978 Batch) }/{1+ exp[-0.3336+(-0.9494 Time)+(0.7978 Batch)}= 0.11 For Time 20 and Batch 2, {exp[-0.3336+(-0.9494 Time)+(0.6253 Batch)}/ {1+ exp[-0.3336+(-0.9494 Time)+(0.6253 Batch)}= 0.34 As can be seen from the parameter estimates for each factor level, the probability of obtaining a positive ELISA increased by using the second inoculum batch or by using the inoculum 10 minutes after its pr eparation while that probability decreased by using the first inoculum batch or by using the inocul um 20 minutes after its preparation. Test 2 The OD cut-off value for this test was set at 0.004. Thirty nine percent of the plants were declared positive by ELISA while only 13% displaye d visual symptoms. Despite this difference

PAGE 99

99 between plants declared infected by visual or serological means, Systemic symptoms and ELISA status were highly signifi cantly associated (p<0.0001). As in the previous test, neither number of rubbings nor inoculum batch were found significant in determining the appearance of syst emic symptoms after the inoculations (p=0.93 and p=0.65 respectively). In the case of response variable ELISA St atus, the factor number of rubbings was nonsignificant. Consequently, a mode l containing only Batch was adjusted. Under this model, Batch was found significant (p=0.0357) as a fact or determining the ELISA Status after the artificial inoculations. The diff erence of infectivity between th e most and the least infective batches was very noticeable (53% vs 20% of infected plants af ter inoculation). The respective predicted probabilities of a plant being infected after being inocul ated with these extreme batches were: {exp[-0.4865+(0.4865 Batch)}/{1+ exp[-0.4865+(0.4865 Batch)}= 0.5, for the most infective Batch. {exp[-0.4865+(-0.8998 Batch)}/{1+ exp[-0.4865+ (-0.8998 Batch)}= 0.2, for the least infective Batch. Test 3 The plants used in this study although of simila r age to those of the pr evious tests in this project were slightly more developed, being 7 cm high and having an average of two more fully expanded leaves than those used previously (2-3 vs. 5). The cause of this was a slightly higher average temperature in the growth chamber, pr obably caused by the air conditioning being shutoff in the building during most of this period. Th e OD cut-off value for this test was set at 0.005. Seventy nine percent of the plants were de clared positive by ELISA, while no plant displayed visual symptoms.

PAGE 100

100 Neither Inoculum Dilution (p=0.4136) nor Inoculum Batch (p=0.9734) were found significant in determining the ELISA Status after the inoculation. The infectivity among batches was much less variable than in the previous two tests (Table 3-3). Discussion Test 1 Very low infection levels were attained comp ared with data from literature using this method (Mandal et al., 2001). A lthough low infection rates are not uncommon, no single factor has been detected as the cause (N. Martinez-Ochoa, pers. comm.). The marked impact of plant age at inoculati on on infection success has been demonstrated by several authors. Mandal and co-workers (2 001) and Hoffman et al. (1998) obtained high percentages of symptomatic plants (75% and 90% respectively) for plants at 14 DAP. In spite of using the technique of Mandal et al. (2001), the percentages obtain ed in this experiment were smaller and similar to those reported by Pereira (1993). According to Noordam (1973), Branch et al. (2003) and S. Mullis (pers. comm.) well-irri gated non-stressed plants are more prone to become infected or to develop symptoms. As th e small volume of substrate in which the plants were raised in the present work tended to dry very easily, the plants were subjected to short but frequent periods of water stress which could have contributed to the lower-than-expected number of infected plants. The irrigati on problem was solved in the s ubsequent tests by increasing the applied water volume. Similar to Culbreath et al. (1992), almost half of the ELISA positive plants in this study were asymptomatic. Hoffmann et al. (1998) repor ted that symptomatic TSWV infected leaves were readily detected by ELISA a lthough asymptomatic leaves of in fected plants did not always give positive ELISA readings. Similarly, four plants visually scored in our experiment as infected were not detected as such by ELISA. Kres ta et al. (1995) in peanut and Canady et al.

PAGE 101

101 (2001) in tomato also observed a similar phenom enon. A possible explan ation could be that several plant parts were used for ELISA and that this pooling of tissues displaying mild symptoms and no symptoms could have lowere d the absorbance value for the pooled sample. The effect of time between inoculum prep aration and inoculation on the frequency of symptomatic plants was important although the observed trend seemed illogical. Tomato spotted wilt virus is usually described as a sh ort living virus outside the cel l even in reducing solutions (Halliwell & Philley, 1974; Hoffman et al., 1998 ). In the present experiment, a declining percentage of infected plants was expected as the time from inoculum preparation increased. Nonetheless, the inoculum seemed to be less infective immediately af ter preparation than 10 min. later. The most plausible explanation for th is result could be a sampling induced statistical bias artifact due to the small numb er of blocks (batches) used. The inherent viral load variability among batches has been addressed by some authors, who suggested the use of samples from similar tissues and showing similar symptoms (Hoffman et al., 1998; Mandal et al., 2001). Nonetheless, Inoculum Batch re ached statistical significance in our study, indicating that even when similar tissue (age and position) was extract ed from the donor plants and pooled to prepare inoculum, viral concentration was still highly variable. Test 2 As in Test 1, the percentage of infected pl ants was far lower than that reported by Mandal and coworkers (2001), even though the inoculation pr otocol was very similar. Two thirds of the ELISA plants were asymptomatic. In contrast to Test 1, every plant displaying symptoms was confirmed by ELISA in Test 2. With reference to the number of rubbings durin g inoculation, the fact or was statistically non-significant. The range used here (4-8 rubbin gs) is typical of what has been used while inoculating (N. Ochoa, pers. comm .). Its non-significance could expl ain why this detail is not

PAGE 102

102 found in the literature even though th e actual damage inflicted to the leaf directly influences the number of virions that reach the target parenchymatous tissue (Ng et al., 2004). Damage also impacts the degree and speed of development of necrosis caused by abrasion (Hoffman et al., 1998). No necrosis developed after inoculation in the abrasion area so apparently the number of rubbings applied was rather mild. Peanut is know n for its strong load of waxes in its leaves (Samdur et al., 2003) so its possible that the abrasion didnt always provide enough injury to serve as entry points for the virions. As in Test 1, the factor Inoculum Batches wa s statistically significant. It is evident that even pooling tissue with similar symptoms was no t enough to avoid important variability in the infectious capacity of the inoculum batches. Test 3 The percentage of infected plants in this test was greater than in the previous two tests and approached the levels reported by other auth ors (Hoffman et al., 1998; Mandal et al., 2001). Surprisingly, no symptomatic plant was observed. Lyerly et al. (2002) re ported that there were instances where peanut plants infected with TSWV exhibited no visible symptoms and some plants even recovered from initial infection and appeared normal. Krista et al. (1995) found that peanut leaves with very low ELISA titer exhibited extremely mild symptoms. However, the ELISA titers obtained in this test are in the usual range obtained while using similar conditions (S. Mullis, pers. comm.) and overall are similar to those obtained in Test s 1 and 2 described in this chapter (data not shown). Mandal et al. (2001) observed a delay in symptom expression associated with plant age. Since age is associat ed with plant size, it c ould be possible that the older plants used in the present test were disp laying this kind of delay, when compared to the smaller ones used in test 1 and 2.

PAGE 103

103 Both tested factors (inoculum dilution and inoculum ba tch) were statistically nonsignificant, probably due to a similar number of viable virions reaching the target tissues between both treatments. The smallest dilution used here (10:1) is certainly weaker than that reported by some researchers (Hoffman et al., 1998; Pereira, 1993; Mandal et al., 2001). This suggests that even a 20:1 buffer:tissue ratio can provide similar infection rates than higher ratios allowing a more efficient use of tissue donor plan ts. This factor can be especially important while preparing batches for a la rge number of plants from li mited amounts of infected tissue from donor plants. Conclusions Following the protocol suggested by Mandal et al. (2001), this study determined that both number of rubbings and inoculum dilution had no effect on the outco me of artificial inoculations. The elapsed time from inoculum preparation show ed an unexpected trend as the infectivity did not fall with time as suggested by some authors (Halliwell and Philley, 1974; Clemente et al., 1990; Mandal et al., 2001). A new test using more replications per treatment could provide further insight on this issue. Inoculum batch was an important factor, probably highlighting th e fact that viral titer is highly variable even when using infected tissue with similar characteristics (age, plant position). This stresses the importance of stan dardizing the inoculation process. In two of the three tests the difference between the percentages of symptomatic and ELISA-detected plants was similar to that reported by Culbreath et al. (1992) but this was not the case in Test 3. Other factors c ould exist that influence the visual symptom development that was not controlled in this test. The overall low infection rates obtained in comparison with other reports using similar techniques clearly suggest that additional work is necessary to detect which factors caused the

PAGE 104

104 observed outcomes in the present work while using the method described by Mandal et al. (2001).

PAGE 105

105 Table 3-1. Effect of elapsed time after preparatio n of Tomato spotted wilt virus inoculum on the number of peanut plants declared infect ed by visual examination or serological means. Treatment Symptomatic plants a ELISA-positive plants 0 batch 1 1 2 10 batch 1 6 5 20 batch 1 1 1 0 batch 2 4 3 10 batch 2 2 9 20 batch 2 0 5 0 batch 3 2 9 10 batch 3 0 4 20 batch 3 0 1 a Number of plants for each treatment=10 Table 3-2. Maximum Likelihood Estimates for ti me and inoculum batch effects on artificial inoculation of Georgia Green peanut. Parameter Estimate S.E. Intercept -0.3336 0.2354 Time (10') 0.768 0.3274 Time (20') -0.9494 0.3504 Batch (1) -0.7978 0.3442 Batch (2) 0.6253 0.3274 Table 3-3. Effect of inoculum dilution on the nu mber of peanut plants declared infected by ELISA. Treatment ELISA-positive plants a 10:1 batch 1 8 20:1 batch 1 8 10:1 batch 2 7 20:1 batch 2 8 10:1 batch 3 10 20:1 batch 3 6 10:1 batch 4 8 20:1 batch 4 8 a Number of plants for each treatment=10

PAGE 106

106 LIST OF REFERENCES Agresti, A. 1996. An Introduction to Categorical Data Analysis. W iley-Interscience. New York, NY. Banks, B.D., I.L. Mao, and J.P. Walter. 1985. Robustness of the restri cted maximum likelihood estimator derived under normality as applied to da ta with skewed distributions. J. Dairy Sci. 68: 1785-1792. Betrn, F.J., S. Bhatnagar, T. Isakeit, G. Odvody, and K. Mayfield. 2006. Aflatoxin accumulation and associated traits in QPM ma ize inbreds and their testcrosses. Euphytica 152(2):247-257. Branch, W.D. 1996. Registration of Georgia Green peanut. Crop Sci. 36(3):806. Branch, W.D. 2002. Registration of Georg ia-01R Peanut. Crop Sci. 42(6): 1750-1751 Branch, W.D. 2003. Registration of Georgia -02C Peanut. Crop Sci. 43(5):1883-1884. Branch, W.D., T.B. Brenneman, and A.K. Culbreat h. 2003. Tomato spotted wilt virus resistance among high and normal O/L ratio peanut cultivar s with and without irrigation. Crop Prot. 22:141-145. Brown, S.L., J.W. Todd, A.K. Culbreath, J. Beasley, B. Kemerait, E. Pros tko, T. Brenneman, N. Smith, D. Gorbet, B. Tillman, R. Weeks, A. Hagan, W. Faircloth, D. Rowland and R. Pittman 2007. Minimizing Spotted Wilt of Pean ut including the 2007 Version of the Tomato Spotted Wilt Risk Index. http://www.tomatospottedwiltinfo.o rg/peanut/riskindex.htm (last accessed January 2007) Bruening, G. 2006. Resistance to infection. In: Natural Resistance Mechanisms of Plants to Viruses. Chpt. 10. Ed. G. Loebenstein and J.P. Carr, 2006. Ed. Springer. Dordrecht, Germany. Canady, M.A., M.R. Stevens, M.S. Barineau, an d J.W. Scott. 2001. Tomato spotted wilt virus (TSWV) resistance in tomato derived from Lycopersicon chilense Dun. LA 1938. Euphytica 117(1):19 Clemente, T.E., A.K. Weissinger, and M.K. Be ute. 1990. Mechanical inoculation of Tomato spotted wilt virus on peanut. Proc. Am. Peanut Res. Ed. Soc. 22:27 (Stone Mountain, GA) Connover, J.W. 1998. Practical Nonparametric St atistics. 3rd. ed. John Wiley & Sons, NY. Culbreath, A.K., J.W. Todd, and J.W. Demski. 1992. Comparison of hidden and apparent spotted wilt epidemics in peanut. Proc. Am. Peanut Res. Ed. Soc. 24:39 (Norfolk, VA). Culbreath, A.K., J.W. Todd, D.W. Gorbet, W.D. Branch, R.K. Sprenkel, F.M. Shokes, and J.W. Demski. 1996. Disease progress of Tomato spotted wilt virus in selected peanut cultivars and advanced breeding lines. Plant Dis. 80(1):70

PAGE 107

107 Culbreath, A. K., J.W. Todd, D.W. Gorbet, F. M. Shokes, and H.R. Pappu. 1997. Field response of new peanut cultivar UF 91108 to Tomato spotted wilt virus. Plan t Dis. 81(12):1410-1415. Culbreath AK, J.W. Todd, D.W. Gorbet, S.L. Brown, J.A. Baldwin, H.R. Pappu, and F.M. Shokes. 2000. Reaction of peanut cultivars to spotted wilt. Peanut Sci. 27(1):35 Culbreath, A.K., J.W. Todd, and S.L. Brown. 2003. Epidemiology and management of Tomato spotted wilt virus in peanut. A nnu. Rev. Phytopathol. 41:53-75. Culbreath, A.K., D.W. Gorbet, N. Martinez-Ochoa, C.C. Holbrook, J.W. Todd, T.G. Isleib, and B. Tillman. 2005. High levels of field resistan ce to Tomato spotted wilt virus in peanut breeding lines derived from hypogaea and hirsuta botanical varieties. Peanut Science 32(1):20. Demsky, J.W., and D.R. Reddy. 2004. Diseases cau sed by viruses. In: Compendium of peanut diseases, 2nd Ed. N. Kokalis-Burelle, D. M. Po rter, R. Rodrguez-Kbana, D. H. Smith, and P. Subrahmanyam. APS Press. Saint Paul, MN. Develey-Riviere, MP, and E. Galiana. 2007. Re sistance to pathogens and host developmental stage: a multifaceted relationship within the plant kingdom. New Phytologist 175(3):405416. Eisen, E. J. and A. M. Saxton. 1983. Genot ype by environment interactions and genetic correlations involving two envir onmental factors. Theoretical Applied Genetics 67(1):75-86. Falconer, D.S., and T.F.C. Mackay. 1996. Introd uction to Quantitative Genetics. 4th Edn. Longman, Essex, England Finne, M.A., O.A. Rognli, and I. Schjelderup. 2000. Genetic vari ation in a Norwegian germplasm collection of white clover (Trifo lium repens L.) 2. Genotypic variation, heritability and phenotypic stab ility. Euphytica 112(1):45-56 German, T.L., D.E. Ullman, and J.W. Moyer. 1992. Tospoviruses: Diagnosis, molecular biology, phylogeny and vector relationships. Annual Rev. Phytopathol. 30:315-348 Gilmour, A.R., B.J. Gogel, B.R. Cullis, and R. Thompson. 2006. ASReml User Guide Release 2.0. VSN International Ltd, Hemel Hempstead, HP1 1ES, UK Gilmour, A.R., B.R. Cullis, S.J. Welham, B.J. Gogel, and R. Thompson. 2004. An efficient computing strategy for prediction in mixed linear models. Comput. Stat. and Data Analysis 44: 571-586. Gorbet, D.W. 1999. University of Florida peanut breeding program. Proceedings Soil and Crop Science Society of Florida 58:2-7. Gorbet, D.W., 2003. New University of Florida p eanut varieties for 2003. UF/IFAS Agric. Exp. Stn. Marianna NFREC Research Report 03-2.

PAGE 108

108 Gorbet, D.W. 2006. Registration of Car ver Peanut. Crop Sci. 46(6):2713-2714. Gorbet, D.W. 2007a. Registration of ANorden Peanut. J. Plant Registr. 1(2):123-124. Gorbet, D.W. 2007b. Registration of AP-3 Peanut. J. Plant Registr. 1(2):126-127. Gorbet, D.W. and D.A. Knauft. 2000. Registration of SunOleic 97R peanut. Crop Sci. 40(4):1190-1191. Gorbet, D.W. and F.M. Shokes. 2002. Registra tion of C-99R Peanut. Crop Sci. 42(6):2207. Groves, R.L., J.F. Walgenbach, J.W. Moyer, and G.G. Kennedy. 2003. Seasonal Dispersal Patterns of Frankliniella fus ca (Thysanoptera:Thripidae) and Tomato spotted wilt virus occurrence in Central and Eastern North Ca rolina. J. Econ. Entomol. 96(1):1-11. Hall, M.D., and D.A. Van Sanford. 2003. Diallel an alysis of Fusarium head blight resistance in soft red winter wheat. Crop Sci. 43(6):1663 Hallauer, A.R. and J.B. Miranda, Fo. 1988. Qu antitative Genetics in Maize Breeding. 2nd Ed. Iowa State Univ. Press, Ames, IA, USA. Halliwell, R.S., and G. Philley. 1974. Spotted wilt of peanut in Texas. Plant Dis. Rep. 58(1):2325. Henderson, C.R. 1976. A simple method for compu ting the inverse of a numerator relationship matrix used in prediction of breed ing values. Biometrics 32(1): 69-83. Hoffman, K., S.M. Geske, and J.W. Moyer. 1998. Pathogenesis of Tomato spotted wilt virus in peanut plants dually infected with pe anut mottle virus. Plant Dis. 82:610-614. Holland, J.B. 2006. Estimating genotypic correla tions and their standard errors using multivariate restricted maximum likelihood es timation with SAS Proc MIXED. Crop Sci 46(2):642. Holland, J.B., W.E. Nyquist, and C.T. Cervan tes-Martinez. 2003. Estimating and interpreting heritability for plant breeding: an update. Plant Breed. Rev. 22:9-111. Holland, J. B., D.V. Uhr, D. Jeffers, and M.M. Goodman. 1998. Inheritance of resistance to southern corn rust in tropical-by-corn-bel t maize populations. Theor. Appl. Genet. 96(3): 232-241. Huber, D.A., T.L. White, and G.R. Hodge. 1994. Variance component estimation techniques compared for two mating designs with fore st genetic architectur e through computer simulation. Theor. Appl. Genet. 88: 236-242. Hurt, C.A., R.L. Brandenburg, D.L. Jordan, G.G. Kennedy, and J.E. Bailey. 2005. Management of spotted wilt vectored by Frankliniella fusca (Thysanoptera:Thripidae) in Virginia markettype peanut. J. Econ. Entomol. 98(5):1435-1440.

PAGE 109

109 Koll, S. and C. Btner. 2000. Cell-to-cell movement of plant viruses through plasmodesmata: a review. Arch. Phytopath. Pflanz. 33(2):99-110. Kormelink, R., P. de Haan, C. Meurs, D. Peters, and R., Goldbach. 1992. The nucleotide sequence of the M RNA segment of Tomato spotted wilt virus, a bunyavirus with two ambisense RAN segments. J. Gen. Virol. 73:2795-2804. Kresta, K.K., F.L. Mitchell, and J.W. Sm ith, Jr. 1995. Survey by ELISA of thrips (Thysanoptera:Thripidae) vectored Tomato s potted wilt virus distribution in foliage and flowers of field-infected p eanut. Peanut Sci. 22:141-149. Kucharek, T., L. Brown, F. Johnson, and J. Funderburk. 2000. Tomato spotted wilt virus of agronomic, vegetable, and ornamental crops. Fl orida Coop. Ext. Service, Circ-914, 13 pp. Lu, P., D.A. Huber, and T.L. White. 2001. Comp arison of multivariate and univariate methods for the estimation of Type B genetic correlations. Silvae Genetica 50(1):13-22. Lyerly, J.H., H.T. Stalker, J.W. Moyer, and K. Hoffman. 2002. Evaluation of Arachis species for resistance to Tomato spotted wilt virus. Peanut Science 29, 79. Lynch, M., and B. Walsh. 1998. Genetics and Analysis of Quantitative Traits. Sinauer Associates, Inc., Sunderland, MA. Mandal, B., H.R. Pappu, and A.K. Culbreath. 2001. Factors affecting mech anical transmission of Tomato spotted wilt virus to peanut (Arachis hypogaea L.). Plant Dis. 85:1259-1263. Mandal, B, H.R. Pappu, A.S. Csinos, and A.K. Culbreath. 2006. Response of peanut, pepper, tobacco, and tomato cultivars to two biologically distinct isol ates of Tomato spotted wilt virus. Plant Dis. 90(9):1150-1155. McKeown, S.P., J.W. Todd, A.K. Culbreath, D. W. Gorbet, and J.R. Weeks. 2001. Planting date effects on tomato spotted wilt in resistant a nd susceptible peanut cultivars. Phytopathology 91:S60. Mitchell, F.L. 1996. Implementation of the IP M planting window for management of Tomato spotted wilt virus and avoidan ce of peanut yellowing death. Final Compliance Report, Texas Pest Management Association, Biologically In tensive Integrated Pest Management Grant Program. http://stephenville.tamu.edu/fmitchel/ento/tswv3.pdf. Moury B., K.G. Selassie, G. Marchoux, A.M. Daubze, and A. Palloix. 1998. High temperature effects on hypersensitive resistance to Tomato spotted wilt tospovirus (TSWV) in pepper (Capsicum chinense Jacq.). Eur. J. Plant Pathol. 104(3):489. Mrode, R. 2005. Linear Models for the Predictio n of Animal Breeding Values. 2nd. Edn. CABI Publishing, Trowbridge, UK.

PAGE 110

110 Murakami, M., M. Gallo-Meagher, D.W. Gorbet, and R.L. Meagher. 2006. Utilizing immunoassays to determine systemic Tomato spo tted wilt virus infection for elucidating field resistance in peanut. Crop Protection 25(3):235-243. Nagata, T., L.S. Boiteux, N. Iizuka, and A.N. Dusi. 1993. Identification of phenotypic variation of tospovirus isolates in Braz il based of tospovirus analysis and differential host response. Fitopat. Brasil. 18: 425-430. Ng, J.C.K., T. Tian, B.W. Falk. 2004. Quantitativ e parameters determining whitefly (Bemisia tabaci) transmission of Lettuce infectious yell ows virus and an engineered defective RNA. Journal of Gen. Vir. 85(9):2697-2707 Northfield, T.D. 2005. Thrips competition and spatiotemporal dynamics on reproductive hosts. Ph.D. Diss. University of Florida, Gainesville, FL. Nyquist, W.E. 1991. Estimation of heritability an d prediction of selection response in plantpopulations. Critical Rev. in Plant Sci. 10(3):235-322. Parrella, G., P. Gognalons, K. Gebre-Selassie, C. Vovlas, and G. Marchoux. 2003. An update of the host range of Tomato spotted wilt viru s. J. Plant pathol. 85 (4): 227-264. Pereira, M.J. 1993. Tomato spotted wilt virus in peanut (Arachis hypogaea L.): screening technique and assessment of genetic resistance le vels. M.S. Thesis. University of Florida, Gainesville, FL. Prins, M., M.M.H. Storms, R. Kormelink, P. De Haan and R. Goldbach. 1997. Transgenic tobacco plants expressing the putative moveme nt protein of Tomato spotted wilt tospovirus exhibit aberrations in growth and appearance. Transg. Res. 6(2):245 Rapp, K, and O. Junttila, 2001. Heritability estimat es of winter hardiness in white clover based on field and laboratory experiment s. Acta Agricult. Scan. Sect. B-Soil and Plant Sci. 50(34):143-148 Resende, L.V., W.R. Maluf, A.D. Figueira, and J.T.V. Resende. 2000. Correlations between symptoms and DAS-ELISA values in two sources of resistance against Tomato spotted wilt virus. Brazilian J. Microbiol. 31(2):135-139 Rosello, S., B. Ricarte, M.J. Diez, and F. Nuez. 2001. Resistance to Tomato spotted wilt virus introgressed from Lycopersicon peruvianum in lin e UPV 1 may be allelic to Sw-5 and can be used to enhance the resistance of hyb rids cultivars. Euphytica 119(2):357. SAS Institute. 2000. Statistical Analysis Softwa re for Windows. Version 8.1. SAS Institute, Cary, NC. Searle, S.R., G. Casella, and C.E. McCulloch. 1992. Variance Components. John Wiley & Sons, New York. Simmonds, N.W. 1979. Principles of Crop Improvement. Longman, London, UK.

PAGE 111

111 Simpson, C.E., J.L. Starr, G.T. Church, M.D. Burow, and A.H. Paterson. 2003. Registration of NemaTAM Peanut. Crop Sci. 43(4):1561. Soler S., M.J. Diez, and F. Nuez. 1998. Effect of temperature regime and growth stage interaction on pattern of viru s presence in TSWV-resistant acce ssions of Capsicum chinense. Plant Dis. 82:1199. Tillman, B.L., D.W. Gorbet, and P.C. Andersen. 2007. Influence of planting date on yield and spotted wilt of runner market type peanut. Peanut Science 34(2):79. Ullman, D.E., R. Meideros, L.R. Campbell, A.E. Whitfield, J.L. Sherwood, and T.L. German. 2002. Thrips as vectors of Tospoviruses. Adv. Bot. Res. 36: 113-140. Venuprasad, R., H.R. Lafitte, and G.N. Atlin. 2007. Response to direct selection for grain yield under drought stress in rice. Crop Sci. 47(1):285. Westfall, P.H. 1987. A comparison of variance component estimates for arbitrary underlying distributions. J. Amer. Stat. Assoc. 82(399): 866-874. Yamada, Y. 1962. Genotype by environment interacti on and genetic correlation of the same trait under different environments. Jap. J. Genet. 37: 498. Yang, R.C., N.K. Dhir and F.C. Yeh. 1998. Intr aclass correlation of pol ychotomous responses of Lodgepole pine to infection of Western gall rust: a simulation study. Silvae Gen. 47(2 3):108-115. Zobel, B. J. and J. T. Talbert. 1984. Applied Forest Tree Improvement. John Wiley & Sons, New York.

PAGE 112

112 BIOGRAPHICAL SKETCH Jorge Javier Baldessari was bor n in San Francisco, Cordoba Province, A rgentina, on April 3rd, 1967. After moving to Cordoba City, he gra duated from high school a nd he enrolled in the Universidad Nacional de Cordoba (UNC), gra duating in 1992 with a degree in agricultural engineering. He then received a graduate rese arch fellowship from the UNC Science Secretariat to work on chickpea breeding. In 1994 he took a position as a peanut breeder in the Manfredi Experimental Station of the In stituto Nacional de Tecnologia Ag ropecuaria. While working as a breeder, he started his M.Sc. st udies, receiving in 2000 a M.Sc. degree in plant breeding from the Facultad de Ciencias Agrarias of th e Universidad Nacional de Rosario. He started his doctoral studies in August 2004. Upon graduation he will return to his breeding duties at Manfredi Experimental Station.