1 NEUTRAL AND ADAPTIVE GENETIC STRUCTURE OF THE SOUTH AMERICAN SPECIES OF NOTHOFAGUS SUBGENUS LOPHOZONIA NATURAL HISTORY, CONSERVATION, AND TREE IMPROVEMENT IMPLICATIONS By RODRIGO VERGARA A DISSERTATION PRESENTED TO THE GR ADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2011
2 2011 Rodrigo Vergara
3 To my wife Tracy for her constant support and my kids Sofia and Nico for their unconditional smiles and love.
4 ACKNOWLEDGMENTS I would like to offer my most sincere gratitude to the people and institutions that have made this project possible. First, I thank my advisor Pam Soltis for her guidance encouragement, and e ndless belief that one day I would finish this endeavor. I also thank my committee members Matt Gitzendanner, Dudley Huber, Doug Soltis, and Tim White for their advice throughout the development of my project and dissertation. Especial acknowledgment to my sources of funding including the Organization of American States (OAS/LASPAU) grant, the Florida Museum of Natural History fellowship, the Department of Biology scholarships, and the financial support from the Soltis Lab for laboratory supplies and sequen cing. Thanks to my fellow students from the Soltis Lab for their friendship, willingness to share ideas and knowledge. T hank s to many members of the Instituto de Silvicultura at Universidad Austral de Chile for their valuable support in identifying sample sites in the field and providing the m e a ns for sampling process ing Thanks go to the genetics crew at the Instituto Forestal in Concepcion for offering me the progeny provenance trials information and contacts, to Corporacion Nacional Forestal for granting me permission to collect samples in protected areas and for assisting me in the field, to Ag ricola y Forestal Taquihue for allowing me access to the trials planted in their land and providing me with shelter and field assistance and to the people from th e Genomics Core Facility in the Department of Biology at East Carolina University, for their kindness in providing me with hardward and softwere support in the last stages of my research I also thank many people and friends that helped in different stage s of this research: Diego Alarcon, Cesar Sepulveda, and Aristides Leiva helped in site
5 identification and accessibility; Felipe Schultz, Alvaro Aguilar, Oscar Reyes, Hector Reyes, and Victor Vera assisted in the field; Emilio Cuq, Ivo Cuq, Linsay Fernandez Salvador, and Claudia Paez, helped in sample processing and laboratory measurements Special thanks go to my wife Tracy Van Holt for her u nconditional support love, and invaluable help with general logistics, field assistance, and in the elaboration of m aps.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 11 LIST OF ABBREVIATIONS ................................ ................................ ........................... 13 ABSTRACT ................................ ................................ ................................ ................... 16 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 18 2 POPULATION GENETIC STRUCTURE AND GENETIC DIVERSITY OF THE SOUTH AMERICAN SPECIES OF NOTHOFAGUS SUBGENUS LOPHOZONIA NATURAL HISTORY INFERRED FROM MULTI LOCUS NUCLEAR MICROSATELLITE DATA ................................ ................................ .... 22 Introductory Remarks ................................ ................................ .............................. 22 Materials and Methods ................................ ................................ ............................ 27 Study Site, Sample Collection, and Storage ................................ ..................... 27 DNA Extraction ................................ ................................ ................................ 28 PCR Amplifications and Genotyping ................................ ................................ 29 Data Analysis ................................ ................................ ................................ ... 29 Detection of scoring errors ................................ ................................ ......... 29 Linkage disequilibrium analysis ................................ ................................ .. 30 Genetic diversity within populations ................................ ........................... 30 Genetic structure ................................ ................................ ........................ 31 Hybridization analysis ................................ ................................ ................ 33 Results ................................ ................................ ................................ .................... 33 Detection of Scoring Errors and Linkage Disequilibrium ................................ .. 33 Genetic Div ersity within Populations ................................ ................................ 34 Genetic Structure Inferred from Bayesian Clustering ................................ ....... 36 Genetic Structure Inferred from AMOVA and Pa irwise Genetic Distances ....... 37 Hybridization Analysis ................................ ................................ ...................... 39 Discussion ................................ ................................ ................................ .............. 40 Genetic Variation within Populations at Nuclear Microsatellite Loci ................. 40 Differences in within Species Genetic Variation ................................ ............... 42 Ou tcrossing and Inbreeding ................................ ................................ ............. 43 Structure and Isolation by Distance (IBD) ................................ ......................... 44 Structure in Nothofagus obliqua ................................ ................................ 45 Structure in Nothofagus alpina ................................ ................................ ... 46
7 Structure in Nothofagus glauca ................................ ................................ .. 47 Contribution to Genetic Variation from Hybridization ................................ ........ 48 Natural History and the Glacial Refugia Hypotheses ................................ ........ 50 Concluding Remarks ................................ ................................ ........................ 53 3 MORPHOLOGICAL GENETIC VARIATION WITHIN AND AMONG PROVENANCES OF NOTHOFAGUS OBLIQUA AND N. ALPINA GROWING IN CHILE. ADAPTIVE DIFFERENCES INFERRED FROM A PROVENANCE PROGENY TRIAL ................................ ................................ ................................ ... 66 Introductory Remarks ................................ ................................ .............................. 66 Materials and Methods ................................ ................................ ............................ 72 Study Area, Seed Collection, and Trials General Description .......................... 72 Experimental Design, Measurements, and Data Editing ................................ .. 73 Analysis of Variance ................................ ................................ ......................... 74 Analysis of variables from all five blocks ................................ .................... 75 Analysis of variables from block one ................................ .......................... 76 Gen etic Parameter Estimation ................................ ................................ .......... 77 Estimates for variables from all five blocks ................................ ................ 77 Estimates for variables from block one ................................ ...................... 79 Among Provenance Differentiation ................................ ................................ ... 79 Canonical Correlation, Discriminant, and Cluster Analysis ............................... 80 Analysis at the provenance level ................................ ................................ 80 Analysis at the family level ................................ ................................ ......... 81 Results ................................ ................................ ................................ .................... 81 Best Models for the Analysis of Variance ................................ ......................... 81 Genetic Parameter Estimates ................................ ................................ ........... 83 Among Provenan ce Differentiation ................................ ................................ ... 84 Patterns of Variation Related to Single Environmental Variables ..................... 85 Multivariate Analysis ................................ ................................ ......................... 86 Multivariate association between morphological and environmental traits ................................ ................................ ................................ ........ 86 Group membership of provenances ................................ ........................... 87 Genetic similarities among provenances through morphological traits ...... 87 Provenance membership of families ................................ .......................... 88 Discussion ................................ ................................ ................................ .............. 89 Within Provenance Heritabilities ................................ ................................ ....... 89 Genetic Correlations among Traits ................................ ................................ ... 90 Among Provenance Differentiation ................................ ................................ ... 92 Growth traits ................................ ................................ ............................... 92 Other adaptive traits ................................ ................................ ................... 94 Non adaptive traits ................................ ................................ ..................... 94 Geographic and Environmental Patterns of Variation ................................ ....... 95 Si ngle trait analysis for N. obliqua ................................ .............................. 95 Single trait analysis for N. alpina ................................ .............................. 100 Multi trait analysis for N. obliqua ................................ .............................. 100 Multi trait analysis for N. alpina ................................ ................................ 102
8 Concluding Remarks ................................ ................................ ...................... 103 4 SYSTEMATICS, CONSERVATION GENETICS, AND BREEDING ZONES DEFINITION BASED ON NEUTRAL AND ADAPTIVE PATTERNS OF GENETIC VARIATION IN NOTHOFAGUS OBLIQUA N. ALPINA AND N. GLAUCA ................................ ................................ ................................ ............... 124 Introductory Remarks ................................ ................................ ............................ 124 Material and Methods ................................ ................................ ........................... 127 Sources of Data Used for the Analyses ................................ .......................... 127 Analysis of Genetic Similarities among Species ................................ ............. 127 Bayesian analysis ................................ ................................ .................... 127 Pairwise genetic distances ................................ ................................ ....... 128 Methods of Ranking Populations for Conservation ................................ ......... 129 Rankings based on allelic richness ................................ .......................... 129 Rankings based on dendrograms ................................ ............................ 129 Results ................................ ................................ ................................ .................. 130 Among Species Genetic Similarity Inferred from Bayesian Clustering ........... 130 Among Species Genetic Similarities Inferred from Pairwise Genetic Distances ................................ ................................ ................................ .... 131 on to Total Allelic Richness ................................ ................................ ................................ ..... 132 Ranking for Conservation Using Population Genetic Distinctness ................. 132 Discussion ................................ ................................ ................................ ............ 133 Genetic Similarities among Species ................................ ............................... 133 Identification of Conservation Priorities ................................ .......................... 136 Evolutionarily significant units ................................ ................................ .. 136 Conservation priorities based on population conservation values ............ 137 Methodological considerations ................................ ................................ 139 General recommendations ................................ ................................ ....... 139 Definition of Breeding Zones for Tree Improvement Strategies ...................... 142 5 CONCLUSIONS ................................ ................................ ................................ ... 154 LIST OF REFERENCES ................................ ................................ ............................. 157 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 169
9 LIST OF TABLES Table page 2 1 Genetic variability and differentiation assessed with nuclear genetic markers in other similar studies ................................ ................................ ........................ 54 2 2 List of sampled populations of Nothofagus obliqua N. alpina and N. glauca ... 55 2 3 Selected microsatellite loci used in Nothofagus obliqua N. alpina and N. glauca ................................ ................................ ................................ ................. 56 2 4 Within population diversity indices for Nothofagus obliqua N. alpina and N. glauca ................................ ................................ ................................ ................. 57 2 5 Global AMOVA s howing the partition of genetic variation among and within populations for Nothofagus obliqua N. alpina and N. glauca ............................ 59 2 6 Membership proportions averaged over all populations of Nothofagu s obliqua N. alpina and N. glauca individuals i n the inferred clusters .................. 59 3 1 Provenances of Nothofagus obliqua and N. alpina analyzed in my study. I made all measurements in the progeny pro ve nance trials (PPT) .................... 105 3 2 Genetic parameters for Nothofagus obliqua and N. alpina obtained from measurements in all five blocks of four year old progeny provenance trials. ... 107 3 3 Genetic parameters for Nothofagus obliqua and N. alpina obtained from measurements in block one of four year old progeny provenance t rials ......... 107 3 4 Narrow sense genetic correlations (r A ) among morphological traits in Nothofagus obliqua and N. alpina obtained from measurements ..................... 107 3 5 Provenance least square means (LSM) for all morphological traits in Nothofagus obliqua and N. alpina ................................ ................................ .... 108 3 6 Among provenance signi ficant differences obtained using provenances least square means and treating the effect of pr ovenance as fixed in the models .... 110 3 7 Standardized canonical coefficients for the two canonical variables morpho1 and morpho2 and for env1 and env2 in Nothofagus obliqua ........................... 110 3 8 Standardized canonical coefficients for the two canonical variables morpho1 and morpho2 and for env1 and env2 in Nothofagus alpina ............................. 111 3 9 Stepwise discriminant analysis of environmentally homogeneous groups of provenances using provenance LSM from nine morphological variables ......... 111
10 3 10 Posterior probability of group membership from miscl assified provenances of the canonical discriminant analysis of enviro nmentally homogeneous groups 112 3 11 Cross validation summary and error rate estimates of the canonical discriminant func tion of environmentally homogeneous groups in N. obliqua .. 112 3 12 Cross validation summary and error rate estimates of the canonical discriminant function of environmentally homogeneous groups in N. alpina .... 113 4 1 Rankings for conservation value in populations of N. obliqua N. alpina and N. glauca analyzed in my study using dat a from three different data sets ........ 145 4 2 Top ranked populations in conservation value for Nothofagus obliqua N. alpina and N. glauca ................................ ................................ ....................... 147
11 LIST OF FIGURES Figure page 2 1 Range of distribution for a) Nothofagus obliqua b) N. alpina and c) N. glauca in grey, showing the location of the sampled populations used in my study. ...... 60 2 2 Pearson Correlations ( r ) between latitude and mean squared allele size differences ( MSD ) for Nothofagus obliqua and N. alpina ................................ ... 61 2 3 Log posterior probabilities ( LnP[D] K values (E vanno et al., 2005) against K (number of popu lation clusters) ................................ .......................... 62 2 4 Results of S TRUCTURE 2.3.2 analysis (Pritchard et al., 2000) showing from K =2 to the most likely K in a) N. obliqua b) N. alpina and c) N. glauca ............ 63 2 5 Correlations between pairwise Slatkin linerized R ST and pairwise geographic distances (Mantel tests) to evaluate isolation by distance ( IBD ) ......................... 64 2 6 Unrooted neighbo r joining trees obtained in P HYLIP 3.69 using R ST values for all pairs of sampled populations ................................ ................................ ......... 65 3 1 Range of distribution for a) Nothofagus obliqua and b) N. alpina in grey, showing the location of the sampled provenances used in my study ............... 114 3 2 Grouping of provenances of a) Nothofagus obliqua and b) N. alpina according to a non hierarchical cluster analysis done using MAT and MAP ... 115 3 3 Significant differences among geographic origins of Nothofagus obliqua provenances obtained using contrasts ................................ ............................. 116 3 4 Significant differences among groups of provenances of similar environmental characteristics in N. obliqua obtained using contrasts. .............. 117 3 5 Significant differences among geographic origins of Nothofagus alpina provenances obtained using contrasts. ................................ ............................ 118 3 6 Pearson correlations ( r ) among morphological and environme ntal variables for Nothofagus obliqua .. ................................ ................................ ................... 119 3 7 Pearson correlations ( r ) among morphological and environmental variables for Nothofagus alpina ................................ ................................ ...................... 120 3 8 Canonical correlations ( r ) of a) the first canoni cal variables morpho1 vs. env1 and b) the second canoni cal variables morpho2 vs. env2 for N. obliqua ......... 121 3 9 Canonical correlations ( r ) of a) the first canoni cal variables morpho1 vs. env1 and b) the second canoni cal variables morpho2 vs. env2 for N. alpina ........... 122
12 3 10 variance dendrograms indica ting closeness in adaptive traits among provenances in a) Nothofagus obliqua and b) N.alpina ............... 123 4 1 Range of distribution for a) Nothofagus obliqua b) N. alpina and c) N. glauca in grey sho wing the location of populations sampled ................................ ........ 148 4 2 Log posterior probabilities ( LnP[D] K values against K (number of population clusters) obtain using S TRUCTURE ................................ .................. 149 4 3 Results of S TRUCTURE 2.3.2 an alysis combining nMSAT data from 20 populations of Nothofagus obliqua 12 of N. alpina and eight of N. glauca ...... 150 4 4 Dendrograms showing the genetic similarities among Nothofagus obliqua N.alpina and N.glauca inferred from nMSAT data. ................................ .......... 151 4 5 Percentage contribution to total diversity ( P T ) obtained from den d r ograms, contribution to the total allelic richness ( C T ), and the respective rankings ........ 152 4 6 Percentage contribution to the total diversity ( P T s distinctness obtained from den d r ograms and its respective ranking ................ 153
13 LIST OF ABBREVIATION S proportion of genetic variance contain ed among populations relative to the total genetic variance (for quantitative data) IS allele size based inbreeding coefficient A average number of alleles per locus ADAP morphological data from daptive traits measured in common gardens A M OVA A nalysis of M olecular V ariance BIC Baye sian Information C riterion bp b ase pair CCC c ubic clustering criterion CG p coefficient for provenance genetic gain cpDNA maternally inherited chloroplast DNA C T contribution of a population to the total allelic richness DA d iscri minant analysis DENS leaf density (g/dm 2 ) DIAM stem diameter at the root collar (mm) DMAT absolute differences in mean annual temperature between the trial site and the provenance origin ER error rate ESU evolutionar il y significant unit F IS inbreeding coef ficient FORK stem forking (%) F ST proportion of genetic variance contained among populations relative to the total genetic variance (based on the infinite allele model)
14 G W Garza Williamson index to detect evidence of recent bottlenecks h 2 within provenanc e individual narrow sense heritability h 2 f family narrow sense heritability H E expected heterozygosity H O observed heterozygosity HT total height (cm) HVAB heterogeneous variance among blocks HWE Hardy Weinberg equilibrium IAM i nfinite allele model IBD i so lation by distance K number of assumed clusters LAREA leaf area (cm 2 ) LD l inkage disequilibrium LGM l ast glacial maximum LnP(D) log posterior probability LSM l east square means LVEIN number of lateral veins on the leaf blade MAP mean annual precipitation M AT mean annual temperature MC Markov chain MCMC Markov chain Monte Carlo MSD mean squared allele size difference among individuals within populations N e m effective rate of migration (number of individuals per generation) NJ neighbor joining. Ref ering to a technique to build similarity tree s nMSAT nuclear microsatellite
15 PCR polymerase chain reaction PPT progeny provenance trial P T percentage contribution of a population to the total diversity measured in a dendrogram R 2 H volume index (L) r A narrow sense gen etic correlations among traits RCBD randomized complete block design REML restricted maximum likelihood analysis R ST proportion of genetic variance contained among populations relative to the total genetic variance (based on the stepwise mutation model) SD stomatal density (stomata/mm 2 ) SI shape complexity index SL stomatal length (m) SMM stepwise mutation model STP single tree plot STR stem straightness SURV survival (%)
16 Abstract of 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 NEUTRAL AND ADAPTIVE GENETIC STRUCTURE OF THE SOUTH AMERICAN SPECIES OF NOTHOFAGUS SUBGENUS LOPHOZONIA NATURAL HISTORY, CON SERVATION, AND TREE IMPROVEME NT IMPLICATIONS By Rodrigo Vergara December 2011 Chair: Pamela S. Soltis Major: Botany The combined analysis of neutral and adaptive genetic variability is important to evaluate the genetic structure of forest trees and to help developing strategies of conservation and tree gen etic improvement I investigate the genetic diversity o f the South American species of Nothofagus subgenus Lophosonia N. obliqua N. alpina and N. glauca emphasizing the ir i ntra and inter population al variability I analyzed their genetic diversity by 1) measuring neutral variability at nuclear microsatellite DNA loc i and 2) measur ing morphological traits on progeny provenance trials (i.e. common gardens) For neutral markers I found relatively high genetic diversity levels ( H E =0. 50 0.62, and 0.66 for N. glauca N. alpina and N. obliqua respectively ) and low but significant genetic structure ( R ST =9, 11, and 15% for N. glauca N. obliqua and N. alpina respectively ) I n N. obliqua this limited structure was spatially organized in t hree latitudinal groups I also detected what appears to be introgression of N. alpina genes into N. obliqua in the northern populations For growth traits (morphological ) I found higher genetic structure ( =0.57 0.89 ) than for neutral markers and in N. obliqua these estimates were substantially higher ( =0.85 0.89 ) than in N. alpina ( =0.57 0.63 )
17 H eritabilities in growth traits were higher for N. obliqua ( h 2 =0.14 0.30) than for N. alpina ( h 2 =0.06 0.14 ) There was a trend of faster growth in N. obliqu a populations adapted to warmer and rainier environments and a tendency of presenting less dense stomata in populations adapted to colder environments. My neut ral nuclear marker results s up p ort the multiple refugia hypothesis suggesting several centers of genetic diversity Moreover, these results indicate that N. obliqua and N. alpina are more genetically similar to each other than to N. glauca and N. obliqua does not have sufficiently differentiated sub groups that could represent new taxa T he morpholog ical traits demonstrate that natural selecti on play s an important role in generating adaptive variation in N. obliqua indicating that this species may have a better chance to adapt to future climatic changes and should respond better to artificial selecti on than N. alpina Finally, combin ing both data types I propose conservation priorities among and within species and bree ding zones within species for tree improvement stra tegies.
18 CHAPTER 1 INTRODUCTION The genus Nothofagus (Nothofagaceae ; southern bee ches ) is an important element of the Southern Hemisphere temperate flora. Approximately 35 species are disjunctly distributed among South America, Australia, New Zealand, New Caledonia, and New Guinea (Manos 19 97). Of these species, nine grow in the tempe rate forest of Chile and Argentina between Santiago de Chile (33 00'S) and Tierra del Fuego (5600'S) The Chilean forest is one of the few temperate forests in the world that harbor s a high proportion of endemic tree species (Armesto et al 19 95). Nothof agus is the backbone of these forest ecosystems where most forest types are dominated by at least one of these species, making the genus ecologically important. The natural history of these forests ( i.e. centers of origin, colonization routes, and current distributional patterns) has been mainly driven by repeated glaciations during the Pleistocene particularly during the last glacial event, which occurred between 22,000 and 10,000 years ago ( Villagran et al. 19 95). Nevertheless, in the Holocene, after th e last ice age, anthropogenic influences became considerable in modifying the landscape for the species of Nothofagus The species of subgenus Lophosonia roble ( N othofagus obliqua (Mirb.) Oerst.), rauli ( N. alpina (Poepp. et Endl.) Oerst.), and hualo ( N. glauca (Phil.) Krasser.), are significant because of their high quality wood, and there is evidence that their populations have steadily degraded because of their commercia l value. Recently, and after the most desirable phenotypes were ex terminated, harves ting for firewood, clear
19 cutting, and subsequent plantations of exotic pine and eucalyptus species have systematically di minished the already degraded forests (Vergara and Bohle 2 000). In the last 30 years, Chilean policies have subsidized plantations, an d as a result more information is available about exotic species than about native Chilean species, preventing native fores try programs from prospering. A new era in forestry now promotes a sustainable harvest of native species, yet relatively little is kn own about the genetic diversity of populations -an essential component of sustainable forest management. Genetic diversity is important in the forests because it in theory, helps the species to be resilient from disturbances such as fire, pests, and clima te change ( Zobel and Talbert 19 84 ). Genetic diversity is one measure of biodiversity that allows a detailed description of the health of the population -distinct from traditional measures of biodiversity that measur e the health of the ecosystem. Genetic d iversity can be divided in to two major components: 1) neutral variability, which is influenced by mutations, gene flow, and genetic drift, but not by selection and therefore does not have adaptive significance for the populations; and 2) adaptive variabili ty, which is influenced by mutations, gene flow, genetic drift, and selection ( McKay and Latta 2 002). In conservation genetics, genetic variation within and between species is usually analyzed using neutral typically molecular, markers such as various DN A markers (or historically, isozymes) The use of these techniques is relatively inexpensive and fast ( McKay and Latta 2 002 ; van Tienderen et al 2 002 ). Unfortunately, variability obtained from the analysis of neutral markers has been shown to be uncorre lated with adaptive variability obtained from the analysis of morphological markers (Reed and Frankham 2 001 ; McKay and Latta 2 002 ). Adaptive genetic variation is a very important
20 component for conservation, because whereas neutral variation determines th e underlying potential for longer term evolutionary changes, adaptive variation determines the evolutionary potential to respond to more immediate changes (McKay and Latta 2 002 ) The combined analysis of both measures of genetic variability is a very powe rful approach to understand breeding systems, gene flow, selection, genetic drift, and their interactions. Thus, the combined analysis of neutral and adaptive markers is a reliable way to analyze the natural history and genetic structure of a spec ies, and the genetic similaritie s with other related species. This information, in turn, is useful for developing strategies of conservation and genetic improvement of forest trees. For two of the three South A merican species belonging to Nothofagus subgenus Lophoz onia ( N. obliqua and N. alpina ), there are several studies in genecological variation among populations ( Donoso 19 87 ), population genetics using isozymes and molecular markers ( Marchelli et al 19 98 ; Pineda 2 000 ; Marchelli and Gallo 2 001 ), and common g arden experime nts under greenhouse conditions ( Ipinza et al 2 000 ). These studies are a valuable starting point for conservat ion genetics in these species. Nevertheless, we need ed an integrate d analysis of the subgenu s that would include N. glauca possib le interspecific hybridization processes, a more intensive sampling in natural populations, and the analysis of morphological traits measured in common garden experiments. This research examine s both the neutral and adaptive patterns of genetic variation o f the species of subgenus Lophozonia across the temperate forests of southern Chile
21 clarifying their natural history and genetic similarity among them prioritizing sites for conservation of the gene pool, and providing guidelines for breeding strategies. My specific objectives are to: 1) analyze the genetic variability and genetic structure of populations including inter and intra population al variation using molecular markers ( Chapter 2) 2) clarify the natural history of the species in relation to their centers of origin, glacial ref ug ia and genetic bottleneck s (Chapter 2), 3) evaluate the patterns of adaptive genetic varia bility including inter and intrapopulation variation using morphological markers (Chapter 3), 4) identify the underlying environme ntal factors that influence the patterns of adaptive genetic variation (Chapter 3), 5 ) analyze the genetic similarity among the species (Chapter 4) 6 ) identify conservation priorities for the species, ranking the populations to be protected (Chapter 4) a nd 7 ) propose breeding zones for the species tree improvement strateg ies from the point of view of t h eir genetic struc ture (Chapter 4)
2 2 CHAPTER 2 POPULATION GENETIC STRUCTURE AND GENETIC DIVERSITY OF THE SOUTH AMERICAN SPECIES OF NOTHOFAGUS SUBGENUS LOP HOZONIA NATURAL HISTORY INFERRED FROM MULTI LOCUS NUCLEAR MICROSATELLITE DATA Intro duct o ry Remarks The current genetic structure and diversity of natural plant populations can be seen as the resulting product of the interaction of biology, geography and c limatic change (Hewitt 2 000). Important biological factors are the breeding system, life form, seed dispersal, and pollination mechanism of the species ( Hamrick 1 982 ; Hamrick and Godt 1 996), which together with the geographical range influence the level of isolation by distance (Wright 1 943) and the effectiveness of geographical barriers against colonization and pollen flow. In addition, the dramatic climatic changes that occurred during the ice ages in the Quaternary had major effects on the distributi on of plant genetic diversity especially in the boreal and temperate regions of the Northern Hemisphere (Soltis et al. 1 997 ; Hewitt 2 000), but also in austral and temperate regions in the Southern Hemisphere (Ogden 1 989; Premoli et al. 2 003; Marchelli and Gallo 2 004 ; Azpilicueta et al. 2 009; Worth et al. 2 009; Mathiasen and Premoli 2 010) The west coast of southern South America in Chile and western Argentina between 3300'S and 4130'S is crossed longitudinally by two mountain ranges separated by t he Central Valley. The Coastal Range has altitudes between 2 000 to 500 m a.s.l., and the Andes, altitudes between 6 ,0 00 to 3 ,0 00 m a.s.l., both becoming overall lower from north to south. This area is the habitat of the Mediterranean Forests between 3300 'S and 3630'S, and the Temperate Rainforests south of 3730'S, w ith an ecotonal zone in between : the Transitional Forests (Donoso, 1982; Veblen and Schlegel, 1982)
23 Nothofagus obliqua (Mirb.) Oerst., N. alpina ( Poepp. et Endl.) Oerst. (= N. nervosa ), and N. glauca (Phil.) Krasser. are symp atric South American endemics. They belong to subgenus Lophozonia along with two species from Australia ( N. cunninghamii (Hook.) Oerst. and N. moorei (Muell.) Krasser.), and one from New Zealand ( N. menziessii (Hook.) Oer st.) (Manos 1 997). This group of deciduous species grows in both Mediterranean and Temperate Rainforest regions in Chile and adjacent areas in Argentina occupying different elevations from the Central Valley to the Coasta l and Andean mountain ranges (O rma zabal and Benoit 1 987). Of the three species, N. obliqu a has the most extended geographical distribution covering nearly 1, 000 km in longitude, N. alpina has an intermediate extension with 700 km, and N. glauca has a narrower distribution covering approxi mately 400 km ( Figure 2 1). These species ar e tall long lived trees easily reaching 30 m in height and 300 years of age These monoecious species have anemophilous pollination and an outcrossing, highly self incompatible breeding system (Riveros et al. 1 9 95, Gallo et al. 1 997, Ipinza and Espejo 2 000). An important difference in reproductive biology among them is seed dispersal, which is predomina nt ly carried by gravity in N. glauca and by a combination of wind and gravity in N. obliqua and N. alpina (Do noso 1 993). The current habitat of these species was largely affec ted by repeated glaciations during the Quaternary influencing the distributional pattern of forests. During the last glacial maximum (LGM 20 ,0 00 yr B.P.), a large proportion of the curre nt forestland was covered by glaciers (Villagran et al. 1 995). A dditionally, periglacial effects in the surrounding areas in the central valley changed climatic patterns and vegetation
24 composition that have been studied mainly through palynological resear ch ( Villagran et al. 1 995) In the Central V alley of central Chile (33 36 S), the current climate is characterized by warm and dry summers, and consequently, sclerophyllous forests. However, with more precipitation at higher elevations, it is possi ble to find populations of Nothofagus spp. that form relict forests in the northern limit of its distributions, such as the N. obliqua island forests in the highlands of the Coastal Range between 33 and 34 S ( Donoso 1 993; Villagran 2 001). During t he LGM glaciers re ached altitudes as low as 1 200 3, 000 m lower than today, creating climatic conditi ons that favored the colonization of valleys by Nothofagus (Heusser 1 990), and probably, eradicated Nothofagus from the mountains. The central valley the n could have been one large panmitic glacial refugium or several isolated ones. The end of the ice age (10 ,0 00 yr B.P.) gradually brought drier and warmer conditions to the area, pushing Nothofagus forests to the ir current di s tribution s in the mountains. S omething similar occurred in the middle portion of the ranges (36 39 S). Currently, on the top of the Nahuelbuta Mountains (Coastal Range), there are isolated forests, which have their main distribution in the Andes ( e.g. N. alpina ). This distributi on can be interpreted as the remains of glacial populations that were growing in the central valley during the ice age (Villagran 2 001), taking advantage of a colder and wetter climate than today. At present, this area, however, is wet enough to allow oth er Nothofagus species ( e.g N. obliqua ) to live in the central valley. According to pollen records, r e colonization of Nahuelbuta by Nothofagus spp. started about 6 ,0 00 yr B.P. in the Holocene (Villagran 2 001).
25 Finally, t he area south of 40 S underwent different periglacial effects. The proximity of the glaciers during the last ice age allowed only the most cold resistant and hygrophilous forest elements to survive in discontinuous populations in lowland sites in the Central Valley and in the Coastal Ra nge (Villagran 2 001). Vegetation was dominated by non arboreal taxa mixed with these cold resistant and hygrophilous forest elements resembling a parkland with varying degrees of openness (Villagran et al. 1 995, Moreno et al. 1 999). Gradually, after 14 200 B.P., more mesic taxa started arriving in the area. Thermophyllous forest taxa ( e.g N. obliqua N. alpina ) might have expanded slowly to this area in the Holocene (after 10 000 yr B.P.) when climatic conditions started to be warm and moist enough to s upport that vegetation (Moreno et al. 1 999), which suggests that the refugia for those taxa were probably localized north of 40 S, in the Coastal Range and the Central Valley. The hypothesis of glacial refugia for N. obliqua and N. alpina localized nor th of 40 S in places i ncluding the valleys near Nahuelbuta (Villagran 2 001), or Rucaancu (39 Villagran 1 991 ) are supported by the extant pollen records. However, there is the possibility tha t these species survived in low numbers in multiple scattered refugia associated with favorable microclimates at higher altitudes and latitudes without leaving any trace in the pollen record ( Markgraf et al. 1 996 ). This last idea is directly supported by genetic studies in N. obliqua (Azpilicueta et al. 2 009) and N. alpina (Marchelli et al. 1 998; Carrasco et al. 2 002; Marchelli and Gallo 2 006), and indirectly supported by genetic studies in other tree species in the area (Al l nutt et al. 1 999; Bekessy et al. 2 002; Premoli et al. 2 003;
26 Nunez Avila and Armesto 2 006 ) and southward (Pastorino et al. 2 009; Mathiasen and Premoli 2 010 ). In accordance with Nothofagus anemophilous pollination, genecological studies in N. obliqua ( Donoso 1 979a ) and N. alpina ( Donoso 1 987) and populati on genetics studies conducted using nuclear markers in N. alpina (Pineda 2 000 ; Carrasco and Eaton 2 002; Carrasco et al. 2 009), show a north to south clinal pattern of variation produced by large scale pollen flow and a climatic cline between the norther n and southern populations. Th is pollen flow also facilitates interbreeding within the subgenus, generating natural hybrids i n speci fic environmental conditions ( i.e. N. alpina x N. obliqua ( Donoso et al. 1 990 ; Gallo et al. 1 997; Marchelli and Gallo 2 00 1) and N. obliqua x N. glauca (= N. leonii Espinosa) ( Donoso 1 979b )) Th us, anywhere th ese species grow in sympatry, there is a potential for hybridization and introgression among them Likewise, th is extensive pollen flow has been shown to maintain relat ively high levels of genetic variability and low population differen t iation in N. alpina and in most Nothofagus species (Table 2 1). The goal of my study is to evaluate, through nuclear microsatellite markers, the levels of genetic diversity and structure of the three South American species of subgenus Lophozonia ( Nothofagus ) in order to infer how these genetic parameters are influenced by climatic change, geography and the species biology. Nuclear microsatellites are a powerful tool widely used to estimate neutral genetic variation. They are single locus, co dominant, highly variable molecular markers, allowing for fine scale analysis of local and regional patterns of genetic diversity ( Selkoe and Toonen 2 006). Nuclear markers are biparentally inherited, t racing both p ollen flow and seed
27 dispersal. Thus, in th is paper I will compare results with studies using other biparentally inherited markers in N. alpina : allozymes (Pineda 2 000; Carrasco and Eaton 2 002), and RAPDs (Carrasco et al. 2 009). My microsate llite data will also complement recent studies in N. obliqua (Azpilicueta et al. 2 009) and N. alpina (Marchelli and Gallo 2 006 ) based on chloroplast DNA which is maternally inherited and trac es exclusively colonization by seed dispersal. I used seven mi crosatellite loci to obtain within and among population parameters of selectively neutral genetic variation in these species growing in Chile. I expect to observe, in general, relatively high levels of genetic diversity and low structure among the populat ions of the species. However, I hypothesize that population s of the three species growing among the Mediterranean Forests between 3300'S and 3630'S would have reduced genetic diversity and higher leve ls of different iation due to their isolation since the end of the last glacial period ( Heusser 1 990; Villagran 2 001 ). Also, I expect to see lower genetic diversity in the narrow range N. glauca than in the some w hat more widespread N. obliqua and N. alpina Finally, I do not expect to see a north to south pa ttern of diminishing variation due to founder effects in the colonization after the LGM as in the patterns often seen in plants in the Northern Hemisphere ( Soltis et al. 1 997 ; Taberlet et al. 1 998; Hewitt 2 000; P etit et al. 2 003; Soltis et al. 2 006). Instead, I expect different centers of variation along the current distribution of the species in line with the multiple refugia hypothesis in South America ( Markgraf et al. 1 996 ) Materials and M ethods Study S ite S ampl e Collection, and S torage I sampled populations across most of the species range of distribution in Chile, including the latitudinal and altitudinal observed variation ( Figure 2 1). I collected
28 samples of fresh tissue from 16 individuals in each of 20 populations of N othofagus obliqua 12 p opulations of N. alpina and 8 populations of N. glauca (Table 2 2) Within this sample, I included two newly described populations that were outside the known range of distribution before 2001, one for N. alpina (Sepulveda and Stoll 2 003) and one for N. glauca (Le Quesne and Sandoval 2 001 ), and regarded the populations described as N. macrocarpa by Vazquez and Rodriguez (1999) (Table 2 2) as N. obliqua following Donoso (1979a) I did not sample populations from Argentina. For N. obliqua and N. alpina I obtained samples mostly from progeny provenance trials, where I collected approximately 5g of fresh leaf tissue for each tree in January 2004. These trials were established in the spring of 2000 by the FONDEF D96/1052 UACH INFOR project in Fundo Arquilhue Valdivia province, Los Rios Region, Chile (4014'S, 7203'W, 304 m a.s.l.), consisting of 31 and 14 N. obliqua and N. alpina populations respectively. I sampled populations where there were typically 10 mother trees ( i.e. open pollinated families) rep res ented with five planted individuals per family therefore six out of 16 individuals are expected to be half siblings to another individual in each sample. For N. glauca and some populations of N. obliqua and N. alpina I collected dormant buds in July 200 3 from randomly selected trees separated, when possible, by at least 20 meters in natural populations (Table 2 2 ). All samples were dried and stored in silica gel, and transported to the Laboratory of Molecular Systematics and Evolutionary Genetics at the Florida Museum of Natural History, University of Florida, US A, for DNA extraction and genotyping. DNA E xtraction After pulverizin g buds and leaf tissue in a bead mill, I extracted total genomic DNA using a modified CTAB protocol for silica dried tissue ( Do yle and Doyle 1 987) I re
29 suspended the DNA pellets in 100 L TE buffer, quantified DNA concentration using a NanoDrop ND 1000 Spectrophotometer, diluted samples to have concentrations in a range of 50 to 200 ng/ L and stored samples at 20C until use. PCR A mplifications a nd G enotyping After DNA extraction I used high resolution microsatellite markers to assess genetic variability. I screened a total of 22 microsatellite loci : a set of 14 loci developed by Jones et al. (2004) f o r N. cunninghamii three developed for N. glauca and N. obliqua ( Azpi licueta et al. 2 004), and five for N. alpina (Marchelli et al. 2 008). I prepared and r a n the polymerase chain reactions ( PCR ) following Jones et al. (2004) and adding a labeled M13 tail primer using 384 well p lates and a Bio Rad Thermo Cycler. After preliminary amplifications, I selected seven loci that were polymorphic and amplified consistently across species and populations at a standardized annealing temperature of 52 C and 2.0 mM MgCl 2 (Table 2 3). I genot yped the samples for all selected primers using the automated sequencer ABI 3730xl DNA Analy z er (Applied Biosystems Inc. ) at the ICBR facility, University of Florida. I used four different fluorescent dyes to label the M13 primer to combine four loci in e ach run by pooling them together in each well of a 96 well plate. After the runs I performed fragment analysis and scoring of alleles using GeneMapper 4.0 (Applied Biosystems, Inc.), repeating unsuccessful amplifications once and treating second round fa ilures as missing data. Data A nalysis Detection of scoring errors Three types of errors stuttering, large allele dropouts, and null alleles are common in microsatellite data and can create scoring bias (Dewoody et al. 2 006). I
30 attempted to mitigate the effects of scoring errors by identifying and correcting them using M ICRO C HECKER 2.2.3 ( van Oosterhout et al. 2 004 ). I analyzed each locus/population combination using the Bonferroni (Dunn Sidak) adjusted 95% confidence interval in the Monte Carlo simulat ions t o detect d eviations from expected allele distributions. In the cases in which there was evidence of null alleles, I adjusted the allele and genotype frequencies following the procedure suggested in M ICRO C HECKER I used this adjusted database for all further analyses. Link a ge disequilibrium analysis I base d my population genetics analyses assuming that microsatellite loci are random ly distributed and independent of each other. I employed G ENE P OP 4.0 (Rousset 2 008) to perform a linkage disequilibrium (LD) analysis using the log likelihood ratio statistic ( G test) and a Markov chain algorithm with 2 0 batches and 5 ,0 00 iterations per batch on each populati on to detect linkage between pairs of loci. Also, I double checked the results in A RLEQUIN 3.11 (Ex coffier et al. 2 005) with their l ikelihood ratio LD procedure ( 1, 000 permutations) and applied the standard Bonferroni technique to obtain the proper significance for multiple comparisons (Rice 1 989). Genetic diversity within populations For each species I calc ulated intra population diversity indices using A RLEQUIN 3.11 (Excoffier et al. 2 005) and G ENE P OP 4.0 (Rousset 2 008) including number of alleles per locus ( A ), observed ( H O ) and expected ( H E ) heterozygosity, inbreeding coefficient ( F IS ), the Gar za Williamson ( G W ) index to detect evidence of recent bottlenecks, mean squared allele size difference among individuals within populations ( MSD ), and allele size based inbreeding coefficient ( IS ). I looked for deviations from Hardy Weinberg equilibrium (HWE) with HWE exact tests, using the Markov chain (MC)
31 algorithm (chain length: 1,000,000, dememorization steps: 100, 000) in A RLEQUIN 3.11 (Excoffier et al. 2 005), and applied the standard Bonferroni technique (Rice 1 989) to correct for multiple loci. F inally, I used SAS software 9.2 (SAS 2 008) to test if there were differences in diversity indices ( i.e. A H E MSD and G W ) across latitude and elevation obtaining Pearson correlation coefficients ( r ), and among species, locations ( i.e. Coast Andes ), la titudinal groups ( i.e. North, South), glacial refugia hypothesis ( i.e. s i.e. natural population, provenance trial), running non parametric Kruskal Wallis tests. Additionally, I tested for differences in inbreeding coefficients ( F IS and IS ) between sources of plant material. I also applied the standard Bonferroni technique (Rice 1 989) in all these analyses. Genetic st r ucture To investigate the patterns of within species genetic structure, I first carried out an individual based approach assuming no specific mutation model using Bayesian clustering in the program S TRUCTURE 2.3.2 (Pritchard et al. 2 000) Following Falush et al. (2003), I used the admixture model with correlated allele freque ncies. I analyzed the three species separately with 320, 192, 128 individuals and 20, 12, 8 sampled populations for N. obliqua N. alpina and N. glauca respectively, without including population information in the analyses. I ran S TRUCTURE at multiple K values ( K =number of assumed clusters in the data); for N. obliqua K =1 to 22, N. alpina K =1 to 14, and N. glauca K =1 to 10. For each species I performed 10 separated runs at each K I used a burn in period of 100, 000 an d 200, 000 Markov chain Monte Carlo (MCMC) iterations for each run obtaining posterior probabilities ( Ln P[D] ) to detect the most likely K I chose K
32 at the highest Ln P[D] and K (Evanno et al. 2 005) ; a criteria used to facilitate the selection of K when Ln P[D] turn s asymptoti c. Finally, I employed D ISTRUCT 1.1 (Rosenberg 2 004) to visualize and edit the S TRUCTURE outputs After I had a better understanding about defined population clusters in each species, I employed locus by locus analysis of molecular variance ( AMOVA ) in A RL EQUIN 3.11 (Excoffier et al. 2 005) to obtain the partition of genetic variation within and among popula tions and, when applicable, within and among clust ers of populations as obtained in S TRUCTURE I performed AMOVA assuming both the stepwise mutation mod el (SMM) obtaining R ST (Slatkin 1 995) and the infinite allele model (IAM) obtaining F statistics (Wright 1 965) as comparison. I tested the significance of the AMOVA outputs with 1, 000 permutations. From the R ST values I estimated the effective m igration rate ( N e m ) following the infinite island population model (Wright 1 931). I also obtained pairwise R ST and pairwise F ST matrices as measures of genetic different iation among populations with 100 permutat ions to obtain significance using standard B onferroni corrections o n all pairwise differences (Rice 1 989). Also, I utilized Mantel tests ( 1, 000 permutations) to compare pairwise R ST and F ST values to evaluate the influence of the chosen mutation model, and to compare Slatkin linearized R ST and g eog raphic distances as an isolation by distance ( IBD ) analysis following Frantz et al. (2009) recommendation to measure IBD when interpreting S TRUCTURE outputs Finally, I input ted the pairwise R ST matri ces into P HYLIP 3.6 9 ( Felsenstein 1 989 ) to obtain unro oted neighbor joining (NJ) trees using N EIGHBOR to infer population s imilarities within species. I visualized the resulting trees with T REE V IEW 1.6.6 (Page 1 996). Before obtaining the pairwise R ST matri ces I performed statistical tests for
33 unequal contri bution of loci using A NIMAL F ARM 1.0 (Landry et al. 2 002) in each species, to exclude from the analysis any locus with too much contribution to the SMM based distance coefficients (Landry et al. 2 002). Hybridization analysis To try to determine the extent of hybridization among the three studied species, I looked for evidence of admixture using S TRUCTURE 2.3.2 (Pritchard et al. 2 000). I followed the same procedures described above, but in this case I analyzed the populations of all three species together, totali ng 640 individuals and 40 populations. After finding the optimal K (number of clusters in the sample), I evaluated the correspondence between the clusters and the morphological identity of the species to examine the admixture proportions found among species. Results Detection of Scoring E rrors and Linkage D isequilibrium Out of all locus/population combinations, I identified 35% of the homozygote excesses to be due to null alleles in Nothofagus obliqua 8% in N. alpina and 9% in N. glauca I w as able to correct approximately half of the cases in M ICRO C HECKER (van Oosterhout et al. 2 004) I regarded the uncorrectable cases as missing values Remarkably, locus NnBIO111 in N. obliqua had putative null alleles in all 20 populations and most of them cou ld not be corrected using M ICRO C HECKER T herefore I excluded locus NnBIO111 from all further analyses in this species (Table 2 3). Combining N. obliqua N. alpina and N. glauca I found a total of 16 cases of significant linkage disequilibrium (LD, =0. 05) over 487 possible combinations (3.3%) and none of them was consistent across populations. The most consistent event was the LD between ncutas06 and ncutas13 but it was only significant in three out of 40
34 populations indicating that it was probably a relationship due to chance alone. All the other cases of LD were significant in only one or two populations. Genet ic Diversity within Populations While I d id not measure polymorphism in my study because it is usually close to 100% when using microsatellit e loci selected specifically for their polymorphism there was one monomorphic locus ( ncutas04 ) in all N. obliqua and N. alpina populations, but not in N. glauca (Table 2 3). In contrast, other measures of genetic diversity, i.e. A H E and MSD were alway s lower in N. glauca than in N. alpina and lower in N. alpina than in N. obliqua (Table 2 4). Pairwise non parametric comparisons among the species yielded highly significant results between N. obliqua and N. glauca for the three parameters ( p <0.01), sign ificant results between N. obliqua and N. alpina for A and MSD ( p <0.05), and between N. alpina and N. glauca for H E ( p <0.05). Even though only populations La Campana (1) and Hueyusca (32) in N. obliqua and N. alpina respectively, showed significant deviat ions from HWE ( =0.05) consistently across all loci, there is a general tendency in N. obliqua and N. glauca to have positive inbreeding coefficients F IS and IS across populations in clear contrast with N. alpina (Table 2 4). Finally, the Garza William son index ( G W ) was very similar across species and across populations within species, ranging overall from 0.494 to 0.847 ( Table 2 4). There was a general trend in measures of diversity for N. alpina to increase with latitude. However, t he only statistica l ly significan t corre lations were latitude vs. MSD in N. obliqua ( r =0.630, p =0.011) and latitude vs. MSD in N. alpina ( r =0.689, p =0.076), but only when population Nahuelbuta (25) of N. alpina was excluded as an outlier ( Figure 2
35 2). In contrast, there was no significant correlation or general trend between diversity measures and elevation. Also, in comparisons between Coast and Andes there was a general trend of higher levels of genetic variation in the Coast for all species, but when I tested the differenc es using non parametric comparisons, I did not find any statistically significant relationships. I hypothesized that samples obtained from provenance trials would show inbreeding and lower levels of genetic diversity in comparison with samples from natura l populations because in the trials, six out of the16 sampled individuals were half siblings. However, non parametric Kruskal Wallis tests did not show any significance or trends showing higher inbreeding or lower genetic diversity in sa mples obtained from the trials either in N. obliqua or in N. alpina (all p values > 0.5). To avoid the confounding effect that source of plant material could have in comparing north and south groups of populations, I combined N. obliqua and N. alpina to contrast the levels o f genetic diversity between these two latitudinal groups; north, including all populations from the Mediterranean Forests, north of the parallel 3630'S, and south, including all the rest ( Figure 2 1, Table 2 2). I did not take into account N. glauca popul ations in this analysis because seven of eight populations are from the Mediterranean Forests, and inclusion would have confounded the latitude and species effects. In this analysis, there was a general trend of less genetic diversity in the North, but onl y MSD was significantly lower there ( p =0.048). Finally, and also combining N. obliqua and N. alpina populations, the tests to compare glacial refugia hypotheses were all not significant and without trends, favoring the multiple refugia hypothesis.
36 G enetic Structure I nferred from Bayesian C lustering The Bayesian analysis performed in each species separately yielded log posterior probabilities ( LnP[D] ) for each K (number of assumed clusters). The maximum LnP[D] was at K =3 for N. obliqua at K =7 for N. alpina and at K =2 for N. glauca ( Figure 2 3). While all three LnP[D] curves had clear peaks and none of them turned asymptotic with the increase of K I K (Evanno et al. 2 005) t o compare both criteria and obtain an idea of the signal strength at the optimal K In N. obliqua and N. glauca K agrees with the optimal K found by LnP[D] but in N. obliqua the signal is relatively stronger in comp arison with a weak signal in N. glauca In N. alpina K does not agree with the K found by LnP[D] Here, the optimal K is 2 instead of 7 ( Figure 2 3b). For N. obliqua ( Figure 2 4a ) with optimal K =3, there is a clear differentia tion between the northern populations (1 to 5) that coincide s with the Mediterranean Forests, and everything else. Among the rest I found a transition group (7 to 12) appearing different from the southern populations (13 to 20) that fall on the Temperate R ainforests. Only Ninhue (6), which belongs geographically to the northern group, is clustering with the south. It is also evident that there is significant gene flow among groups inferred from the levels of admixture seen in the S TRUCTURE diagram ( Figure 2 4a). For N. alpina ( Figure 2 4b) with optimal K =2 or K =7 depending on the criteria, there is no geographic differentiation into groups. Moreover, the membership charts from K =2 to K =7 show a lack of identity in the populations. The only two populations th at stay together regardless of the number of clusters are Siete Tazas (21) and Bullileo Alto (23). Both populations are located in the Andes, in the northern tip of N. alpina
37 distribution, relatively close together and isolated from other populations ( Figu re 2 1). Other populations that individually show an identity (membership to a cluster greater than 50%) are Tregualemu (22), Santa Barbara (26), Jauja (27), and Pichipillahuen (28). For N. glauca ( Figure 2 4c) with optimal K gradual from north to south. Most of the populations are a relatively even combination of the two clusters, except the northernmost population, Loncha (33), and the southernmost population, Quilleco (40), which have a clear and opposite membership. These two populations are also very isolated from other populations of the species ( Figure 2 1). In the three species, the number of optimal clusters K was far from reaching the total number of sampled populations, indicating extensive gene flow, either historic ally or ongoing, among them. Genetic Structure I nferred from AMOVA and P airwise Genetic D istances My within species SMM based AMOVA analyses indicated that the genetic variation within populations was always greater than 80%. The R ST values were 0.11, 0.16 and 0.087 for N. obliqua N. alpina and N. glauca respectively representing low to moderate but statistically highly significant ( p <0.00 001) levels of genetic different iation among populations ( Table 2 5) Using the R ST values, I estimated the effectiv e rates of migration ( N e m ) in 2.0, 1.3, and 2.6 individuals per generation respectively. For comparison, I also ran an IAM based AMOVA and obtained very similar among population variation values for N. obliqua ( F ST =0.12), N. alpina ( F ST =0.14), and N. glauc a ( F ST =0.074). In N. obliqua S TRUCTURE also found that the species could be divided in to three spatially defined and homogeneous groups of populations. As a result I ran an SMM based AMOVA including that extra group level. In this analysis, the
38 variation among groups ( R CT ) was 0.114 ( p <0.00001) and the variation among populations within groups ( R SC ) was only 0.037 ( p <0.00001) also indicating homogeneity within groups with substantially more intra group ( N e m =6.5) than inter group migration ( N e m =1.9) Cons equently, the total variation among populations ( R ST ) was 0.146 ( p <0.00001), higher than the value obtained in the previous AMOVA for N. obliqua ( R ST =0.11), because I excluded Ninhue (6), which did not have a well d efined membership to any group. In a M ant el test, the pairwise R ST matrices obtained separately for each species correlated moderately to the pairwise F ST matrices indicating a connection between the two values, b ut also important differences. r ) obtained in the M antel tests were 0.69, 0.49, and 0.64 for N. obliqua N. alpina and N. glauca respectively. For each species, I also conducted another set of Mantel tests between the pairwise Slatkin linearized R ST (Slatkin, 1995) and geographic distance matrice s to test for IBD In N. obliqua ( r =0.73) I found a strong and highly significant ( p =<0.0001) relationship ( Figure 2 5a), in N. alpina ( r = 0.06, p =0.604) I found no relationship ( Figure 2 5b), and in N. glauca ( r =0.33, p =0.096) I found a positive but weak relationship between genetic and geographic distances ( Figure 2 5c) After confirming that there were not significant inequalities in the contribution of loci to the SMM based distance coefficients I obtained unrooted NJ trees from the pairwise R ST matric es for each species ( Fig ure 2 6). In N. obliqua ( Figure 2 6a), there are two d efined branches in the NJ tree: one composed only of populations from the Temperate Rainforests in the south, including Llancacura (19) and Purranque (20), and another composed o f populations from the central and transitional part of the
39 distribution, including Reserva uble (8) and Ralco (10). All the populations from the Mediterranean Forest in the north from La Campana (1) to Ninhue (6) except Bullileo Alto (5) did not cluster closely together with any other population. They are genetically closer to the central and south branches according to the geographic distances. Interestingly, two populations from the south, Galletue (13) and Choshuenco (18), are also part of this group o f somewhat isolated populations. For N. alpina ( Figure 2 6b) there are not defined branches in the tree and the genetic similaritie s among populations do not agree in general with their geographic distributions. However, similarly to N. obliqua the two po pulations from the Andes in the northern area of the species distribution appear genetically isolated from the rest and from each other. N othofagus glauca ( Figure 2 6c) exhibits a greater association between the geographic and genetic distances, with three main groups of populations in the NJ tree. The extreme populations Loncha (33) in the north and Quilleco (40) in the south are genetically separated from each other and from the rest, and geographically isolated, forming one branch each. The rest of the p opulations fall in a third branch with variable levels of isolation. Hybridization A nalysis By running S TRUCTURE and including all populations from the t h ree species together, I obtained the optimal number of clusters, K =3, with a high correspondence (>90% ) to the morphological identity of the species on all N. alpina and N. glauca populations, and on populations (6) to (20) in N. obliqua Northern populations (1) to (5) in N. obliqua had higher admixture proportions, but their identity to the right cluster was still greater than 50% (data not shown). Under this scenario, in N. obliqua there was an
40 average admixture proportion of 10.3% coming from N. alpina and 1.9% coming from N. glauca far higher than the admixture going to N. alpina or N. glauca ( Table 2 6). It is clear however, that the admixture in N. obliqua is not distributed evenly among all populations, and most of it is in the northern populations. Due to this uneven distribution in admixture proportions, I also inspected data using K =4, which I knew placed these northern populations in a different cluster, allowing me to have a separate estimation of admixture in it. Surprisingly, most of the interspecies admixture in N. obliqua went away in this new analysis reducing the overall admixture propo rtion in N. obliqua to 3.4% coming from N. alpina and 1.4% coming from N. glauca without changing much the values going to N. alpina and N. glauca ( Table 2 6). In addition, the distribution of the admixture in N. obliqua is more even, with identities alw ays greater that 90% except in La Campana (1) and Loncha (2) with 86% each. Discussion Genetic Variation within P opulations at Nuclear Microsatellite L oci The high levels of nuclear microsatellite genetic variation I obtained in my study were not surprisin g and agreed in general with other studies of related, large sized, long lived, outcrossing, wind pollinated trees. For nuclear microsatellite loci, I found one report of genetic diversity ( H E =0.474) and number of alleles per locus ( A =3.8 and 4.3) from Not hofagus alpina in Argentina (Milleron et al. 2 010) and several from related genera like Fagus and Quercus from the Northern Hemisphere that ranged from 0.597 to 0.864 for H E and from 4.9 to 14.9 for A ( Table 2 1). The values found in my study ( H E =0.50 0. 66, A =4.5 6.2) are within t hese ranges. However, they are high er than the values reported by Milleron et al. (2010) for N. alpina and fall in the lower end of the range known for Fagus and Quercus I attribute this to the span of the geographical
41 distribut ions of the species in my study which are more narrowly distributed than those species from the Northern He misphere N othofagus alpina from Chile has lower RAPD diversity than other long lived forest species, a result attributed to intense exploitation an d substitution during the last century (Carrasco et al. 2 009). Also, Taylor et al. (2005) found in Australian N. moorei values of ISSR diversity similar to the ones found by Carrasco et al. (2009); however, N. moorei is an isolated species with a very nar row distribution. Interestingly, several reports studying allozyme variation in N. alpina found levels of variation similar to (Marchelli and Gallo 2 001 ; 2004) and even higher than (Pineda 2 000; Carrasco and Eaton 2 002) other related genera like Castane a Quercus and Fagus The evidence obtained from my study indicates that despite a century of exploitation and substitution, the dramatically reduced fragments of forest probably still contain an important fraction of their neutral genetic variability, agr eeing with the findings of Craft and Ashley (2007) in Quercus macrocarpa but not with what Carrasco et al. (2009) hypothesize for N. alpina or Torres Diaz et al. (2007) for N. alessandrii Also, according to the se relatively high levels of variation foun d in my study, populations from these three species appear to have overcome the severe climatic changes that occurred during the last 20, 000 years. Hamrick (2004) explains the resilience of forest trees to climatic change and habitat fragmentation, indicat ing that characteristics such as individual longevity, high neutral within population genetic variation, and extensive pollen flow may make them especi ally resistant to the loss of genetic diversity and further extinction in chang ing environmen ts
42 Differen ces in within S pecies G e netic V ariation N othifagus obliqua N. alpina and N. glauca have relatively high levels of genetic variation, as shown by A H E and MSD Yet, there are statistically significant differences among these species, and those differenc es correspond with their geographic ranges. N othofagus glauca with a narrow distribution, has the lowest values of genetic diversity in all three parameters, followed by N. alpina with a regional distribution. Finally, N. obliqua also with a regional dis tribution but somewhat larger than that of N. alpina is the most genetically diverse species. These findings agree with Hamrick and Godt (1996), who found that geographic range is one of the predictors of genetic diversity. In this case, it appears that h aving a wider distribution helps to create and maintain genetic variability, or conversely, having genetic variability helps to improve the chances to colonize heterogeneous environments. Even though it may be misleading to compare this range influenced tr end of genetic variation obtained for the Chilean species of subgenus Lophozonia with other genera (Gitzendanner and Soltis 2 000) the trends in levels of genetic diversity ( H E ) in relation to geographic range are remarkable. Using the studies with nuclea r microsatellite markers for Fagus and Quercus listed in Table 2 1, I can find larger genetic diversity values ( H E =0.68 0.87) for widespread species such as F. sylvatica and intermediate values ( H E =0.60 0.66) for regional species like Q garryana similar to values for N. alpina ( H E =0.62) and N. obliqua ( H E =0.66). Finally, narrowly distributed N. glauca appears in my study with a lower H E =0.50. It is obvious that these few examples from Fagaceae are far from showing a generalized trend in the group, and th at I would need nuclear microsatellite data for many more species of Nothofagus to confirm the trends.
43 Outcrossing and I nbreeding Nothofagus species have an outcrossing, highly self incompatible breeding system (Riveros et al. 1 995 ; Gallo et al. 1 997; Ip inza and Espejo 2 000). However, the incidence of at least some inbreeding in natural populations of Nothofagus ( e.g. Premoli 1 997; Pineda 2 000; Carrasco and Eaton 2 002) and other Fagaceae ( e.g Muir et al. 2 004; Lee et al. 2 006; Buiteveld et al. 2 00 7; Marsico et al. 2 009) is the ru le. This is probably because forest stands are composed of groups of related trees that allow breeding between cousins, parents and offspring, half sibs, and even full sibs. This may explain why samples obtained from prove nance trials (some of them half sibs) had neither lower genetic variation nor higher inbreeding coefficients than the natural populations. In my study, after correcting for null alleles, I found only two statistically significant departures from Hardy Wein berg equilibrium (HWE): for populations La Campana (1) in N. obliqua and Hueyusca (32) in N. alpina ( Table 2 4). But here, it is important to keep in mind that the algorithm used to detect and correct null alleles could have masked some of the true inbreed ing. Despite the lack of significance in inbreeding coefficients, F IS and IS on most of the populations, the tendency to consistently have F IS and IS greater than zero still holds for N. obliqua and N. glauca This consistency suggests that with a large r sample size a moderate and statistically significant amount of inbreeding could be detected in the two species. In N. alpina there is no trend among populations, th us without counting Hueyusca (32), the F IS and IS values are probably due to chance alon e The general compliance of N. alpina to HWE could mean that this species has a more efficient way
44 to avoid inbreeding, but so far all the evidence in the literature states the opposite (Pineda 2 000; Carrasco and Eaton 2 002; Marchelly and Gallo 2 004). With regard to the inbred populations, it is reasonable that La Campana (1) at the northern limit of N. obliqua and Hueyusca (32) at the southern limit of N. alpina ( Figure 2 1) showed evidence of inbreeding, because of their isolation and relatively small population sizes. However, it is not apparent why only those populations would have inbreeding and not other similar small populations in the area like Loncha (2) and Bellavista (3) in the north or Llancacura (31) in the south. Further research will be n ec essary to examine these issues. Structure and Isolation by D istance (IBD) According to simulations by Frantz et al. (2009), it appears that strong IBD would have an effect on the clustering algorithm in S TRUCTURE mainly by making the posterior probabili clusters. Fortunately, even though I have strong evidence of IBD in N. obliqua and some evidence in N. glauca when running S TRUCTURE my posterior probabilities were not asymptotic and had a peak that was also coincident with the optimal number of clusters given by K In addition, Frantz et al. (2009) claim that clusters obtained by S TRUCTURE in these conditions of strong IBD have too much overlap I n my case, particularly with N. g lauca this is true. However, I believe that overlap is precisely a product of what is really going on in the distribution of variation, and I prefer not to consider possible clusters where the variation is actually clinal Another consideration about the methodology employed to infer among population structure is the difference in using SMM based or IAM based approaches. In my case,
45 R ST and F ST values obtained from AMOVA were very similar, but when comparing population pairwise R ST with pairwise F ST values in each species, the Mantel tests delivered only moderate correlations (0.5 to 0.7), indicating that the choice of the mutation model has a large influence on the results and further conclusions from the analysis, and they cannot be used interchangeably. Structure in Nothofagus obliqua The percentage of among population neutral variation found for N. obliqua (11%) is moderate in comparison to N. alpina and N. gla u c a in my study, and to other Nothofagus and Fagaceae ( Table 2 1). This indicates that gene flo w ( N e m =2.0) is able to maintain the genetic homogeneity of the species quite well. Additionally, IBD plays an important role ( Figure 2 5a), suggesting that a good part of the inter population variability is clinal and not due to abrupt differences between populations. I did, however, find a relatively clear subdivision into three homogeneous groups of populations ( Figure 2 4a). There is gene flow among these groups ( N e m =1.9), but it is more restricted than the gene flow among populations within groups ( N e m = 6.5). These groups are distributed latitudinally, each with members from very different environments ( Table 2 2), indicating a constant gene interchange among elevations and locations that apparently do not challenge local adaptation. The subdivision into three groups also agrees in general with the divide in two forest types and an ecotonal zone proposed by Donoso (1982) and Veblen and Schlegel (1982) ( Figure 2 1), suggesting that around the southern limit of the Mediterranean Forests at latitude 3630'S, there is a geographic barrier probably the uble River valley, slowing colonization and pollen flow and creating a divide between the northern and transitional groups. The divide between the transitional and southern
46 groups approximately at latitude 3830 'S does not match any apparent geographic barrier to gene flow, but it coincides surprisingly well with a major change in climate (Fuenzalida 1 965). The NJ tree ( Figure 2 6a) agrees fairly well with the two divides, but suggests that populations in the no rthern group have a higher level of genetic isolation from each other than the populations in the other groups. This makes sense because the northern populations are more geographically isolated from each other than those from the rest of the country. All this implies that the organization of the among population genetic variability in N. obliqua follows a clinal pattern of variation similar to that described by Donoso et al. (2009) of N. obliqua using chloroplast DNA (cpDNA) haplotypes recognized a different pattern of variation showing that populations in the Coastal Range are very distinctive from each other and also different than those from the Central Valley or the Andes However, the pattern of variation in the Central Valley and in the Andes in Chile is similar to that found in my study, following a north to south line and finding a divide approximately at the same latitude as my south transition boundary (3830'S). Azp ilicueta et al. (2009) did not find a divide at the uble River valley at latitude 3630'S, but did at 3500'S, with the isolated populations on the highlands of the north of the distribution forming their own cluster. Structure in Nothofagus alpina Among population neutral variation in N. alpina was the highest in my study (16%, N e m =1.3), and it is among the highest values reported for Nothofagus and Fagaceae ( Table 2 1). This value might be larger than in N. obliqua because the N. alpina coastal populatio ns are more isolated from each other and from the Andean populations.
47 Furthermore, the level of among population variation combined with the absence of IBD in this species ( Figure 2 5b) indicates the presence of selective barriers to gene flow, and the abs ence of apparent geographical groups ( Figure 2 4b) supports this idea. The clinal pattern found in N. obliqua is not repeated for N. alpina Figure 2 6b also shows a lack of spatial structure, with high genetic similarities between very distant populations like Nahuelbuta (25) and Curarrehue (29) or between Tregualemu (22), Jauja (27) and Llancacura (31). My findings do not agree in general with the genetic clusters obtained for the Chilean populations of N. alpina using allozymes (Pineda 2 000 ; Carrasco an d Eaton 2 002) and RAPDs ( Carrasco et al. 2 009) These studies show an overall correspondence between genetic and spatial patterns of variation, although in all of them there are cases of populations clustering with geographic distant groups. Marchelli an d Gallo (2006) used cpDNA haplotypes to describe the patterns of variation in N. alpina finding no divides among populations on the Chilean side of the Andes and isolated haplotypes in two populations sampled in the Coastal Range, agreeing with a spatiall y undefined structure. Structure in Nothofagus glauca In the narrowly distributed N. glauca among population neutral variation (8.7%) is lower than those for N. obliqua and N. alpina but still moderate in comparison to other Nothofagus and Fagaceae ( Table 2 1) indicating moderate to high levels of gene flow among populations ( N e m =2.6). There is a weak tendency for the populations to be isolated by distance ( Figure 2 5c), but the two populations driving this tendency the northernmost population Loncha (33) and the southernmost population Quilleco (40) are also very isolated from the N. glauca core range ( Figure 2 1). Within the six
48 populations making this core there is no IBD and there is apparent east/west gene flow as suggested by the genetic similarity be tween Coastal Tregualemu (37) and Andean Alico (39) and the genetic dissimilarity between Bullileo (38) and Alico (39), two Andean populations geographically close but on different sides of a group of mountains (Fig ures 2 1 and 2 6c). The results of the S T RUCTURE analysis also support extensive gene flow among populations and the greater isolation of Loncha (33) and Quilleco (40), with no apparent geographical groups ( Figure 2 4c). Unfortunately, there is virtually nothing in the literature about the patter ns of genetic or morphological variation in this species to support or dispute my results. Contribution to Genetic Variation from Hybridization A potential source of genetic variation and structure in the studied populations is hybridization within the sub genus. N atural F 1 hybrids have been found i n speci fic environmental conditions i.e. N. alpina x N. obliqua ( Donoso et al. 1 990 ; Gallo et al. 1 997; Marchelli and Gallo 2 001) N. obliqua x N. glauca (= N. leonii Espinosa) ( Donoso 1 979b), and populations with different levels of introgression mainly towards N. obliqua The level and sources of admixture I obtained in S TRUCTURE ( Table 2 6) agree with the hybridization and introgression patterns proposed previously within the subgenus, suggesting that they reflect the real events involving the sampled populations. Given that I sampled only populations with an unequivocal morphological identity for each species, evidence of admixture indicates that some hybridization and introgression must have occurred long ago However, the signs of admixture between N. alpina and N. glauca are probably an artifact because they appear in low frequency
49 (less than 1%), many cases occur in populations that never were in sympatry, and there are no reports of hybridization betwee n these two species. I examined two different outcomes from the S TRUCTURE analysis to discuss admixture among the three species. When using the optimal number of clusters ( K =3), with each species belonging to its own cluster, I found high levels of admixtu re going from N. alpina towards N. obliqua in the highlands of the northern part of the distribution in the Mediterranean Forests between 3300'S and 3630'S. This result was very surprising because currently in this area the species do not seem to interbr eed and the populations with more admixture, i.e. La Campana (1), Loncha (2), and Bellavista (3), are isolated from any N. alpina populations. Interestingly, according to Vazquez and Rodriguez (1999), the N. obliqua populations from this area, that they be lieve should be regarded as N. macrocarpa are morphologically more similar to N. alpina than to N. obliqua from the south. If these results are real, they will be strong evidence of past hybridization between N. obliqua and N. alpina with marked introgres sion towards N. obliqua that probably occurred around 10 000 years B.P. in the glacial refugia located in the valleys between 3300'S and 3600'S (Heusser 1 990). Thus, the role of past hybridization between N. obliqua and N. alpina would be a major force in shaping the patterns of genetic structure in N. obliqua However, I also examined the suboptimal K =4. With it, S TRUCTURE splits the N. obliqua populations precisely into one cluster composed of the northern populations and the other cluster containing t he rest. This removes most of the admixture coming from N al p ina to N. obliqua ( Table 2 6), in conflict with the ancient hybridization/introgressi on hypothesis described above. Even though K =3 is undoubtedly the optimal number of
50 clusters under K (Evanno et al. 2 005) and could be giving us important clues to infer LGM hybridization events responsible for shaping the genetic diversity in these species, assuming K =4 better explains the hybridization processes from the last couple of thou sand years. Na tural History and the Glacial Refugia H ypotheses The effects of the last glacial maximum (LGM) on the distribution patterns and genetic variation of N. obliqua and N. alpina have been revealed by palynological and cpDNA studies. Pollen profil es obtained at different latitudes and elevations in Chile support the glacial refugia hypothesis localizing refugia in low elevations north of 40 S and agree with the general tendency seen in the Northern Hemisphere of lower latitude refugia ( Soltis e t al. 1 997 ; 2006; Hewitt 2 000 ). On the other hand, the distributions of haplotypes obtained through cpDNA analysis in both species agree with the multiple glacial refugia hypothesis suggested for virtually all tree species in southern South America ( Alln utt et al., 1999; Azpilicueta et al., 2009; Bekessy et al., 2 002; Carrasco et al., 2002; Nunez Avila and Armesto, 2006 ; Marchelli and Gallo, 2006; Marchelli et al., 1998; Mathiasen and Premoli, 2010 ; Pastorino et al., 2009; Premoli et al., 2003 ) The multi ple refugia hypothesis including high latitude refugia is also postulated as an alternative hypothesis for some species in North America ( Soltis et al. 1 997 ; 2006) and remarkably proposed by Magri et al. (2006) in Europe for Fagus sylvatica a species tha t was formerly described as an example for the low latitude hypothesis ( Demesure et al. 1 996 ). My results agree with the multiple refugia hypothesis and therefore with the existence of several centers of genetic diversity across the geographical ranges of the species. Also, the lack of consistent signs in my data showing bottlenecks or repeated
51 founder events indicates that these refugia harbored enough amounts of variation and pollen flow has been an important agent in keeping populations genetically di verse and connected from the time of the first re colonization events after the LGM. N othofagus obliqua ( Azpilicueta et al. 2 009) and N. alpina (Marchelli and Gallo 2 006) show evidence of several refugia in the Coastal R do not seem to have participated in the postglacial migrations and remained confined within their areas of origin. My results indicate that genes did move from the Andes to the Coastal Range refugia and vice versa through pollen flow, generating great among p opulation admixture in N. obliqua and a somewhat more restricted admixture in N. alpina Nevertheless, there is a slight but statistically nonsignificant tendency for populations in the coastal area to be more genetically diverse than those in the Andes ba sed on my microsatellite data. In the work of Azpilicueta et al. (2009) the Chilean populations of N. obliqua had two relatively widespread haplotypes currently present in the Central Valley and the They seem to have been gene rated in two different with palynological records (Villagran 1 991), or further nort h in the Mediterranean My microsatellite analysis matches the division between also suggesting the existence of one or more glacial refugia in this area. Farther north, my data do not seem to support the results of Azpilicueta et al. (2009). First, I found that
52 according to the levels of genetic variabilit y ( Table 2 4), a more probable area as a the Mediterranean Forests and north of the Nahuelbuta mountains. Second, I differentiated another group of populations north of 36 Forests that according to my microsatellite data is slightly less genetically diverse t han the populations southward. This group of currently isolated populations probably derived from different glacial refugia located in the Cen tral Valley, Coastal Range, and Andean valleys. In N. alpina Marchelli and Gallo (2006) did not find clear cpDNA haplotypic divides in Chilean populations, and their data suggest a northward postglacial colonization of the Andes from only a single refugiu according to Carrasco et al. My results partially support this trend with slight evidence for a diminishing diversity to the north ( Table 2 4, Figure 2 2), and lack of spatially explicit groups of populations ( Figure 2 4). However, there is a marked increase in genetic diversity in populations Recinto (24) and Nahuelbuta (25), in the same area where I proposed the refugium for N. obliqua thus making this area a potenti al refugium for N. alpina as well. The population genetics of N. glauca which mostl y occurrs in intermediate elevations in both mountain ranges w ere probably only mildly influenced by the climatic changes since the last glaciation. My data indicate that the limited genetic structure of this species is due to two very isolated populations ( Figure 2 1): the northernmost population Loncha (33), from the highlands of the Coastal Range, and the southernmost population Quilleco (40), from 100 km south of the ma in geographical range of the
53 species (Le Quesne and Sandoval 2 001). The large and heavy N. glauca seeds make it difficult to believe that Quilleco (40) is a 100 km dispersal event. More probably, there were small N. glauca stands in the Andean foothills a nd Central Valley connecting all the populations of the species during the LGM about 20, 000 yr BP when moist conditions at low elevations were ideal to maintain Nothofagus forest. Current drier conditions could have made most of the populations gradually d isa ppear in the area, except Quille co (40). Moreover, this population has been isolated from other hypothetical N. glauca populations in the area by the great alluvial cone of Lake Laja deposi ted in the Laja valley about 10, 000 yr BP (Thiele et al. 1 998). Conclu ding Remarks My results indicate that the levels of neutral genetic variation exhibited by N. obliqua N. alpina and N. glauca are high and correspond to values of other tree species with similar natural history and range of distribution. There is little but significant genetic structure explained by isolation by distance within population continuums, geographic isolation of certain populations and some hypothesized geographic barriers to gene flow. Also, the tendency of these species to form stands with different levels of kinship did not refle ct on significant inbreeding indicating in general a tendency to avoid selfing On the o ther hand, the evidence of ancient and current hybridization a s an important contributor to the intra specific genetic di versity w ithin the subgenus indicates that it may be a major source of variability and differen t iation for the northern populations of N. obliqua Finally, I observed that the patter n s of genetic variation in these three species support the multiple refugi a hypothesis for N. obliqua and N. alpina and suggest several centers of genetic diversity across the areas of distribution for N. obliqua N. alpina and N. glauca
54 Table 2 1 Genetic variability and differentiation assessed with nuclear genetic marker s in other similar studies Species Range Marker K N Loci A H E F ST Reference From genus Nothofagus N. alpin a reg Allozymes 22 19 7 2.9 0.484 9.6 Pineda 2000 (= N. nervosa ) Allozymes 18 36 10 3.0 0.289 5.1 Carrasco and Eaton, 2002 RAPDs 22 27 33 0.150 12.4 Carrasco et al. 2009 Allozymes 11 112 8 2.3 0.173 3.8 Marchelli and Gallo 2001 Allozymes 20 115 8 3.4 0.180 5.2 Marchelli and Gallo 2004 Allozymes 2 71 6 1.9 0.126 Milleron et al. 2010 Allozymes 2 30 6 1.6 0.163 Millero n et al. 2010 Microsats 2 71 3 4.3 0.474 Milleron et al. 2010 Microsats 2 30 3 3.8 0.474 Milleron et al. 2010 N. alessandrii narr Allozymes 7 27 7 1.8 0.182 25.7 Torres Diaz et al. 2007 N. nitida narr Allozymes 4 42 15 1.3 0.045 4.7 Premol i 1997 N. betuloides reg Allozymes 4 28 15 1.5 0.116 12.0 Premoli 1997 N. dombeyi reg Allozymes 5 34 15 1.6 0.093 7.4 Premoli 1997 N. pumilio reg Allozymes 41 29 7 1.4 0.070 20.0 Mathiasen and Premoli 2010 N. Antarctica reg Allozymes 12 48 2 2.7 0. 185 11.0 Pastorino et al. 2009 N. truncate reg Allozymes 30 57 5 1.3 0.051 4.9 Haase 1992 N. menziesii reg Allozymes 5 52 15 1.5 0.116 Haase 1993 N. moorei narr ISSRs 20 7 42 1.8 0.168 10.4 Taylor et al. 2005 From other related genera Fa gus sylvatica wide Microsats 10 130 4 14.9 0.829 5.8 Buiteveld et al. 2007 F. japonica reg Microsats 16 34 13 8.6 0.659 2.3 Hiraoka and Tomaru 2009 Quercus glauca wide Microsats 10 19 4 6.5 0.741 4.2 Lee et al. 2006 Q. macrocarpa wide Microsats 14 3 4 5 11.2 0.864 2.7 Craft and Ashley 2007 Q. petraea wide Microsats 5 60 6 8.6 0.755 20.1 Bruschi et al. 2003 Microsats 7 30 13 7.0 0.797 0.8 Muir et al. 2004 Q. semiserrata wide Microsats 10 39 8 8.2 0.679 12.0 Pakkad et al. 2008 Q. garryana reg Microsats 22 15 7 4.9 0.597 4.9 Marsico et al. 2009 Range (Hamrick and Godt 1 996): wide=widespread, reg=regional, narr=narrow; K number of populations; N average sample size per population; Loci, number of loci effectively analyzed; A mean number of alleles per locus; H E Average expected heterozygosity; F ST F ST (Wright 1 965) or its equivalent G ST (Nei 1 973) in percentage.
55 Table 2 2. List of sampled populations of N othofagus obliqua N. alpina and N. glauca # Population Location Lat itude (S) Longitude (W) Elevation (m a.s.l.) Source of plant material Tissue collected N. obliqu a 1 La Campana Â¢ Coast 3258' 7107' 1,380 Natural population Buds 2 Loncha Â¢ Coast 3409' 7057' 870 Natural population Buds 3 Bellavista Â¢ Andes 34 46' 7044' 950 Natural population Buds 4 Siete Tazas Alto Â¢ Andes 3528' 7059' 1,460 Natural population Buds 5 Bullileo Alto Andes 3620' 7123' 1,150 Natural population Buds 6 Ninhue Coast 3623' 7225' 150 Provenance trial Leaves 7 Recinto Andes 36 51' 7137' 710 Provenance trial Leaves 8 Reserva uble Andes 3706' 7115' 1,890 Provenance trial Leaves 9 Santa Barbara Andes 3740' 7159' 430 Provenance trial Leaves 10 Ralco Andes 3751' 7133' 810 Provenance trial Leaves 11 Victoria Central Valley 3814' 7218' 480 Provenance trial Leaves 12 Pichipillahuen Coast 3817' 7304' 450 Provenance trial Leaves 13 Galletue Andes 3840' 7118' 1,450 Provenance trial Leaves 14 Cunco Andes 3852' 7200' 410 Provenance trial Leaves 15 Curarrehue Andes 392 5' 7135' 910 Provenance trial Leaves 16 Cruces Coast 3932' 7305' 160 Provenance trial Leaves 17 Malalhue Central Valley 3933' 7233' 140 Provenance trial Leaves 18 Choshuenco Andes 3950' 7205' 220 Provenance trial Leaves 19 Llancacura Coast 4016 7318' 60 Provenance trial Leaves 20 Purranque Coast 4055' 7311' 90 Provenance trial Leaves N. alpina 21 Siete Tazas Andes 3529' 7053' 760 Provenance trial Leaves 22 Tregualemu Coast 3603' 7240' 510 Natural population Buds 23 Bullileo Al to Andes 3618' 7124' 1 650 Provenance trial Leaves 24 Recinto Andes 3650' 7136' 810 Provenance trial Leaves 25 Nahuelbuta Coast 3731' 7252' 940 Provenance trial Leaves 26 Santa Barbara Andes 3744' 7154' 430 Provenance trial Leaves 27 Jauja Ande s 3802' 7202' 670 Provenance trial Leaves 28 Pichipillahuen Coast 3818' 7303' 420 Provenance trial Leaves 29 Curarrehue Andes 3926' 7135' 840 Provenance trial Leaves 30 Releco Andes 3951' 7155' 1 160 Provenance trial Leaves 31 Llancacura Coast 4017' 7321' 380 Provenance trial Leaves 32 Hueyusca Coast 4059' 7330' 410 Provenance trial Leaves N. glauca 33 Loncha Coast 3410' 7101' 1 110 Natural population Buds 34 Alto Huelon Coast 3506' 7204' 200 Natural population Buds 35 Siete T azas Andes 3526' 7103' 830 Natural population Buds 36 Los Ruiles Coast 3550' 7230' 220 Natural population Buds 37 Tregualemu Coast 3559' 7240' 510 Natural population Buds 38 Bullileo Andes 3618' 7124' 710 Natural population Buds 39 Alico Andes 3635' 7128' 620 Natural population Buds 40 Quilleco Andes 3728' 7158' 340 Natural population Buds The coordinates and elevation of the source populations sampled for the provenance trials is only a rough approximation. There are not records of the e xact position of the mother trees harvested to plant the trials. Â§ Elevations were obtained entering geographic coordinates in a digital elevation model based on Shuttle Radar Topography Mission (SRTM) Finished 3 arc second (90 m) raster elevation data set. Â¢ Populations described as N. macrocarpa by Vazquez and Rodriguez (1999).
56 Table 2 3 Selected microsatellite loci used in N othofagus obliqua N. a lpin a and N. glauca Locus Primer sequences Repeat of reported allele Size Total alleles per locus N. obliqua N. alpina N. glauca ncutas 04 F: CTCCCGTGAGAAGGTTTGAAT R: AATGGGCATATGGTTATTGTGATAG (CA) 11 337 357 1 1 10 ncutas 06 F: TTTCCCTCCATGAATACTTG R: AATGGCTTGATATTGTTACC (CT) 14 357 409 27 9 8 ncutas 08 F: TTGAATGGCTTGACTTGTAA R: GATGGGTGAGAAT TTTGACT (AC) 12 217 233 7 7 5 ncutas 12 F: GCATCATCCCATCCTAAGTTAT R: CTGAACACTGGCATCTTTAATG (CA) 16 216 250 19 9 6 ncutas 13 F: TAACCCACCACTCTTGCCGAAGT R: GGAACGGCCTCCACATCTCA (CT) 16 304 358 21 25 12 ncutas 22 F: GATGGGGTTATCATAGGTGTCGT R: TCAGCGAGAATTCCT TTGATGTA (CT) 14 (AC) 11 292 326 18 11 14 NnBIO 111 Â§ F: TATGTGAACGCGTCTGCTTC R: CGCTCTTCAGACCAGAAAGG (GT) 2 A(GT) 10 134 162 10 5 Size range in my study (bp). Developed by Jones et al. (2004). Â§ Developed by Marchelli et al. (2008).
57 Table 2 4 Within po p ulation diversity indices for N othofagus obliqua N. alpina and N. glauca # Population N Loci A H O H E F IS MSD IS G W N. obliqu a 1 La Campana 15.8 4 6.5 0.493 0.779 0.376 56.9 0.371 0.517 2 Loncha 14.8 5 7.0 0.681 0.714 0.044 78.7 0.072 0.63 3 3 Bellavista 15.2 5 6.2 0.589 0.688 0.147 48.9 0.090 0.565 4 Siete Tazas Alto 15.4 5 5.0 0.553 0.597 0.077 34.7 0.125 0.721 5 Bullileo Alto 14.4 5 5.4 0.579 0.620 0.070 33.0 0.028 0.651 6 Ninhue 15.8 5 6.8 0.609 0.711 0.147 73.4 0.088 0.684 7 Recin to 15.6 5 6.8 0.632 0.689 0.086 90.2 0.042 0.628 8 Reserva uble 16.0 4 7.0 0.609 0.683 0.111 68.7 0.054 0.732 9 Santa Barbara 15.8 5 5.8 0.560 0.617 0.095 68.0 0.028 0.537 10 Ralco 14.6 5 4.8 0.438 0.534 0.192 56.6 0.043 0.494 11 Victoria 14.4 5 5.6 0.546 0.568 0.045 71.6 0.092 0.532 12 Pichipillahuen 15.5 4 5.8 0.598 0.621 0.038 39.6 0.042 0.714 13 Galletue 14.6 5 6.8 0.683 0.770 0.112 131.3 0.293 0.523 14 Cunco 15.6 5 7.4 0.659 0.680 0.031 92.0 0.396 0.847 15 Curarrehue 15.2 5 7.0 0.679 0.697 0.022 90.0 0.035 0.638 16 Cruces 15.0 5 7.0 0.655 0.685 0.046 96.5 0.189 0.749 17 Malalhue 15.8 5 6.8 0.573 0.631 0.094 105.3 0.254 0.721 18 Choshuenco 15.0 5 6.6 0.645 0.648 0.003 111.5 0.294 0.611 19 Llancacura 15.2 5 6.2 0.614 0.659 0.070 82.9 0.18 8 0.570 20 Purranque 15.4 5 6.8 0.610 0.658 0.080 129.5 0.192 0.684 Average 15.3 4.9 6.2 0.600 0.662 0.094 78.0 0.124 0.625 N. alpina 21 Siete Tazas 15.8 6 3.8 0.485 0.419 0.165 27.6 0.187 0.598 22 Tregualemu 16.0 6 4.3 0.688 0.649 0.062 30.4 0.007 0.579 23 Bullileo Alto 15.8 6 4.7 0.494 0.541 0.088 37.5 0.072 0.528 24 Recinto 16.0 6 6.8 0.719 0.717 0.003 44.0 0.190 0.731 25 Nahuelbuta 15.8 6 7.3 0.746 0.761 0.018 115.2 0.016 0.701 26 Santa Barbara 15.8 6 5.3 0.666 0.555 0.206 44.9 0.260 0.693 27 Jauja 16.0 6 4.8 0.563 0.545 0.033 53.6 0.094 0.526 28 Pichipillahuen 16.0 6 5.7 0.625 0.572 0.097 61.9 0.524 0.774 29 Curarrehue 16.0 6 5.8 0.760 0.665 0.148 52.1 0.358 0.645 30 Releco 16.0 6 5.7 0.604 0.621 0.027 51.0 0.165 0.5 77 31 Llancacura 15.3 6 6.2 0.690 0.719 0.036 61.2 0.051 0.688 32 Hueyusca 14.5 6 5.2 0.396 0.635 0.382 42.9 0.349 0.668 Average 15.8 6.0 5.5 0.620 0.617 0.014 51.8 0.081 0.642
58 Table 2 4. Continued. # Population N Loci A H O H E F IS MSD IS G W N. glauca 33 Loncha 14.9 7 4.7 0.448 0.543 0.181 51.2 0.380 0.648 34 Alto Huelon 16.0 6 5.2 0.552 0.548 0.007 55.3 0.123 0.700 35 Siete Tazas 15.0 7 4.0 0.394 0.497 0.215 31.1 0.296 0.713 36 Los Ruiles 15.9 7 4.1 0.460 0.480 0.043 24.8 0.017 0.769 37 Tregualemu 14.8 6 4.2 0.476 0.522 0.094 29.9 0.042 0.714 38 Bullileo 15.9 7 4.3 0.448 0.455 0.018 32.0 0.149 0.834 39 Alico 15.1 7 5.7 0.441 0.535 0.184 43.9 0.131 0.641 40 Quilleco 15.7 7 4.1 0.436 0.431 0.011 52.8 0.101 0.573 Average 15. 4 6.8 4.5 0.457 0.502 0.089 40.1 0.151 0.699 Note: N average sample size; Loci, number of loci effectively analyzed; A mean number of alleles per locus; H O Average observed heterozygosity; H E Average expected heterozygosity; F IS inbreeding coefficien t (Wright, 1965); MSD mean squared allele size difference among individuals within populations (Goldstein et al., 1995); IS allele size based inbreeding coefficient (Rousset, 1996); G W Garza Williamson index (Excoffier et al., 2005). *Significant devi ation from Hardy Weinberg equilibrium ( =0.05) after standard Bonferroni correction (Rice, 1989).
59 Table 2 5 Global analysis of molecular variance ( AMOVA ) showing the partition of genetic variation among and within populations for N othofagus obliqua N alpina and N. glauca Results are a weighted average over usable loci. I performed the analyses under the stepwise mutation model (SMM) using R ST like sum of squared size differences with 1 000 permutations Source of variation d.f. Sum of squares Vari ance components Percentage of variation R ST Â§ N. obliqu a Among populations 19 2 449.8 3.274 11.0 0.110 Within populations 606 16 108.3 26.519 89.0 Total 625 18 558.1 29.793 N. alpin a Among populations 11 3 009.8 7.430 16.0 0.160 Within po pulations 366 14 290.3 38.910 84.0 Total 377 17 300.1 46.340 N. glauca Among populations 7 834.4 2.909 8.7 0.087 Within populations 241 7 264.9 30.367 91.3 Total 248 8 099.3 33.276 Average degrees of freedom across loci. Â§ All R ST are highl y significant ( p < 0.00001). Table 2 6 Membership proportions averaged over all populations of Nothofagus obliqua N. alpina and N. glauca individuals in the inferred clusters obtained using Structure Species Clusters (for K = 3) Clusters (for K = 4) 1 2 3 1 Â§ 2 Â§ 1+2 3 4 N. obliqua 0.878 0.103 0.019 0.311 0.641 0.952 0.034 0.014 N. alpina 0.022 0.972 0.006 0.037 0.013 0.050 0.944 0.006 N. glauca 0.016 0.009 0.975 0.011 0.015 0.026 0.007 0.967 Â§ Clusters 1 and 2 represent northern and southern N. obliqua populations respectively.
60 Fig ure 2 1 Range of distribut ion for a) N othofagus obliqua b) N. alpina and c) N. glauca in grey, showing the location of the sampled populations used in my study Last Glacial Maximum (LGM) extent of the ice she et adapted from Hollin and Schilling (1981).
61 Fig ure 2 2 Pearson Correlations ( r ) between latitude and mean squared allele size differences ( MSD ) for N othofagus obliqua and N. alpina There is no correlation between MSD and N. glauca The straight line s show the tendency of the relationships and p is the p value after Bonferroni correction. In N. alpina population Nahuelbuta (25) was regarded as an outlier and excluded from the analysis.
62 Figure 2 3 Log posterior probabilities ( LnP[D] ) a K values (Evanno et al. 2 005) against K (number of population clusters). I obtain all values using S TRUCTURE 2.3.2 (Pritchard et al. 2 000) for a) twenty two potential clusters in N othofagus obliqua b) fourteen in N. alpina and c) ten in N. glauca I chose the most likely K in each species at the highest LnP[D] K values. I identified three, seven, and two clusters in N obliqua N. alpina and N. glauca respectively.
63 F igure 2 4 Results of S TRUCTURE 2.3.2 analysis (Pritchard et al. 2 000) showing from K =2 to the most likely K in a) N othofagus obliqua b) N. alpina and c) N. glauca Clusters of populations are represented by colors. Populations are defined by vertical lines and ordered north to south within each species. Within individuals, the proportion of each color indicates membership to the given cluster. I employed D ISTRUCT 1.1 (Rosenberg 2 004) to visualize and edit the S TRUCTURE outputs. K =2 K =3 c ) a) b ) K =3 K =4 K =5 K =6 K =7 K =2
64 Figure 2 5 Correlations between pairwise Slatkin li nerized R ST (Slatkin 1 995) and pairwise geographic distances (Mantel tests) to evaluate isolation by distance ( IBD ) in a) N othofagus obliqua b) N. alpina and c) N. glauca I obtained Pearson correlation coefficients ( r ) and p values ( p ) for each test.
65 Fig ure 2 6 Unrooted neighbo r joining trees obtained in P HYLIP 3.69 (Felsenstein 1 989) using R ST values for all p airs of sampled populations in a) N othofagus obliqua b) N.alpina and c) N.glauca
66 CHAPTER 3 MORPHOLOGICAL GENETI C VARIATION WITHIN A ND A MONG PROVENANCES OF NOTHOFAGUS OBLIQUA AND N. ALPINA GROWING IN CHILE. AD APTIVE DIFFERENCES INFERRED FROM A PROVENANCE PROGENY TRIAL Introduct o ry Remarks Nothofagus obliqua (Mirb.) Oerst. and N. alpina (Poepp. et Endl.) Oerst. (= N. nervosa ) are two sympat ric South American endemics belonging to subgenus Lophozonia (Manos, 1997). These deciduous hardwood tree species are of remarkable ecological and economic importance in Chile because of their role as habitat for many plant and animal species, fast growth, and high quality wood (Donoso, 1993). N othofagus obliqua and N. alpina forests cover as a whole about 1 000 km from north to south between Valparaiso (3300'S) and Puerto Montt (4130'S) in Chile and adjacent areas in Argentina ( Figure 3 1), growing at di fferent altitudes from the Central Valley to the mountains (Ormazabal and Benoit, 1987). These species grow i n a range of climatic conditions depending on latitude and elevation. In the northern part of their distribution in central Chile, habitat of the M editerranean Forests long summers. In these conditions, N. obliqua and N. alpina grow typically in the mountains at elevations between 400 and 2 ,0 00 m a.s.l., and between 500 and 1 ,5 00 m a.s. l. respectively, where there is enough precipitation for the species to survive ( Donoso, 1982). Conversely, in the southern part of their distribution, corresponding to the Temperate Rainforests (south of 3730'S), the climate becomes in general, colder a nd rainier with precipitations distributed year around and only a couple of months relatively dry over the summer. Here, N. obliqua grow s predominately from sea level to 400 m a.s.l., and N. alpina generally f r o m 700 to 900 m a.s.l. (Veblen and Schlegel,
67 1 982). Despite the two distinctive environmental conditions described above corresponding to the main northern and southern forest types, the variation in temperature and precipitations follow both latitudinal and altitudinal clines. According to Di Castri and Hajek (1976), in this area the mean annual temperature at sea level diminishes gradually north to south from 15 C to 10 C, and the mean annual precipitation ranges from 300 to 1 ,5 00 mm dependi ng on elevation in the north increas ing progressively southw ard where it ranges from 1 ,0 00 mm in the valleys to 3 ,0 00 mm or more in the west slopes of the mountains. These climatic gradient s are the most conspicuous reason why it is possible to observe the phenotypic differences among provenances found in N. obliqu a (Donoso, 1979; Ipinza et al. 2000) and N. alpina (Donoso, 1987; Medina, 2001 ), which show clinal variation following those climatic gradient s. Although the traits studied by these authors (Donoso, 1979 ; 1987; Ipinza et al. 2000; Medina, 2001) are relat ed to flowers and fruits and therefore less likely to be affected by the environment ( Clausen et al., 1940 ) these patterns of variation could correspond to either plasticity or genetic adaptation, because the experimental design of those studies does not allow separat ion between the patterns of phenotypic diversity from those of genetic diversity. Genetic diversity is one measure of biodiversity that allows a detailed description of a G enetic diversity is important in the forests becau se i n theory it helps the species to be resilient to disturbances such as fire, pests, and climate change (White et al. 2007; Zobel and Talbert, 1984). Genetic diversity can be divided in two major components: 1) Neutral variability, which is influenced by mutations, gene flow, and genetic drift, but not affected by selection and therefore does not have
68 adaptive significance for the populations; and 2) A daptive variability, which is additionally and most importantly influenced by selection (McKay and Lat ta, 2002; White et al. 2007). Genetic variation within and between species is often analyzed using neutral markers such as isozymes or molecular markers. The use of these techniques is relatively inexpensive and fast (van Tienderen et al. 2002). Unfortun ately, variability obtained from the analysis of neutral markers has been shown to be uncorrelated with adaptive variability obtained from the analysis of morphological traits (Reed and Frankham, 2001; McKay and Latta, 2002). Adaptive genetic variation is a very important component in populations, because whereas neutral variation determines the underlying potential for longer term evolutionary changes, adaptive variation determines the evolutionary potential to respond to more immediate changes (McKay and Latta, 2002). T he main goal of my study is to evaluate the patterns of adaptive genetic variation in N. obliqua and N. alpina using quantitative traits related to adaptability to specific environmental conditions, growth, and other morphological variables I focus o n identifying the und erlying factors that make up the adaptive genetic structure in these two species. I could not find any other study that examined the partitioning of genetic variances among provenances versus the variance among families with in provenances in Nothofagus However, this has been done in several other forest tree species. A review examining some provenance progeny tests in Castanea sativa (Tchatchoua and Aravanopoulos, 2010a ; 2010b), Swietenia macrophylla (Wightman et al. 2008), Pinus palustris Taxus brevifolia and Populus trichocarpa (White et al. 2007), Gmelina
69 arborea (Hodge and Dvorak, 2004), Araucaria angustifolia (Sebbenn et al. 2003), Pinus taeda (Sierra Lucero et al. 2002), and Picea glauca (Li et al. 1993), shows t hat for growth traits such as growth in height, stem diameter, and volume, the proportion of among provenance variance as compared to the variance among famil ies within provenances, is considerable usually ranging between 40% and 80%. In other traits like stem straightness, stem forking, and disease resistance, the proportion of among provenance variance is usually much lower than for growth traits, ranging from 0% to 50% indicating that for these variables in many cases almost all of the variation can be found among famil ies within a single provenance. The amount of among provenance adaptive genetic variation found in tree species also depend s on the ecological range where provenances grow. There is a tendency of finding less among provenance variation in species or groups of provenances growing in regions that are ecologically homogeneous. For example, Xie (2008) reported the among provenance variation in growth traits for Alnus rubra finding that there was not a statistically significant difference among provenances growing within each of two ecologically homogeneous regions in the north and south of British Columbia, Canada, indicating that virtually all the adaptive genetic variation within both regions was allocated among families within provenances. De spite the fact there is virtually no progeny provenance studies analyzing the partitioning of genetic variances in Nothofagus there are several common garden studies that analyze the significance of among provenance variation and environmental trends in m orphological traits. I n common garden studies t here seems to be a tendency to find more among provenance variation in species with wider ecological ranges. In the
70 genus Nothofagus those ranges are more or less associated to the different subgen era Thus the species of subgenus Nothofagus have the wider ecological range, the species of subgenus Fuscospora the narrower ecological range, and the species of subgenus Lophozonia which is the subgenus of N. obliqua and N. alpina are somewhere in the middle. Am ong the species of subgenus Nothofagus research in N. pumilio showed highly significant among provenance genetic differences between two elevations in shoot length, leaf area, and leaf phenology (Premoli et al. 2007), as well as in stomatal density (Prem oli and Brewer, 2007), with the higher elevation having shorter shoots, smaller leaves, late phenology, and less stomatal density. In subgenus Lophozonia a series of three papers in N. cunninghamii indicate that genetic differences in leaf area among elev ations were never significant, and that differences were significant in leaf thickness and stomatal density with thicker leaves and less stomata per unit area at higher elevations, but only when the range in elevation was wide enough (Hovenden, 2001; Hoven den and Vander Schoor, 2 003; 2006). In N. menziesii Sun and Sweet (1996) found that provenances growing at higher elevations were genetically more frost tolerant ; Ledgard and Norton (1988) found that the growth period was significantly shorter, ending soo ner, and therefore plants were growing slower in provenance s exposed to lower temperatures; and Wilcox and Ledgard (1983) found significant genetic variation among provenances in growth rate and leaf size. In N. obliqua Puntieri et al. (2006) found no sig nificant among provenance differences in growth rate, however the ecological range of the analyzed provenances was very narrow. Finally, in species of subgenus Fuscospora there are fewer significant among provenance differences
71 detected in adaptive traits In N. alessandrii Santelices et al. (2009) found no significant differences in growth rate. In N. fusca Wilcox and Ledgard (1983) found no significant differences in growth rate or leaf size, and Ledgard and Norton (1988) found the same for growth period but some differences in growth rate. Finally, N. solandri has among provenance differences in frost tolerance (Sun and Sweet, 1996), growth, and leaf size (Ledgard and Norton, 1988; Wilcox and Ledgard, 1983), but not in growth period (Ledgard and Norton 1988). My study is an attempt to genetically evaluate the patterns of variation in several morphological traits for N. obliqua and N. alpina The specific objectives of the study are: a) to partition the species mo rphological genetic variation among pro venances and among famil ies within provenances b) to find relationships between spatial patterns of adaptive and climatic variability across the study area, c) to estimate the level of genetic control or heritability on the variables measured in the trial s, and d) to evaluate the relative performance of provenances in growth traits I hypothesize that there will be higher levels of among provenance variation in N. obliqua than in N. alpina because of the wider ecological range in N. obliqua and in growth traits more than in form traits, the latter of which seem to have little adaptive value. Also, I hypothesize that the patterns of among provenance variation will follow a latitudinal cline related to the climatic variation patterns in the study area and fo und by other authors (Donoso, 1979 ; 1987; Ipinza et al. 2000; Medina, 2001). Finally, I expect to find faster growth rate s in provenances that originate in milder climates similar to the conditions of the progeny provenance trials.
72 In the summer of 2004 I measured one N. obliqua and one N. alpina progeny provenance trial, both established in the spring of 2000 by the FONDEF D96/1052 UACH INFOR project in Fundo Arquilhue, Valdivia Province, Los Rios Region, Chile. The trials contain one to 11 open pollinate d families from each of 31 and 14 provenances for N. obliqua and N. alpina respectively (Table 3 1). I measured surviva l, total height, diameter, stem straightness and stem forking on every four year old tree in the trials. In a sub sample, I measured lea f morphometrics related to leaf size, leaf shape, and stomatal features. Materials and M ethods Study Area, S eed C ollection, and T rials G eneral D escription I measured the offspring of open pollinated mother trees belonging to provenances systematically samp led by the project FONDEF D96/1052 (Universidad Austral de Chile Instituto Forestal) across most of the species range of distribution in Chile ( Figure 3 1) includ ing the populations described as N othofagus macrocarpa by Vazquez and Rodriguez (1999) (Table 3 1) following Donoso (1979). In the summer of 1999, at each provenance, seeds from 10 mother trees were harvested. Trees were selected maintaining a minimum distance of 100 m between them to maximize genetic variation. After a pre germination treatment, seeds were planted and grown in 80 mL containers under greenhouse conditions for approximately seven months (Medina, 2001; Rodriguez and Medina, 2000). In the spring of 2000, the project simultaneously planted one N. obliqua and one N. alpina progeny prove nance trial in Fundo Arquilhue, Province, Los Rios Region, Chile (4014'S, 7203'W, 304 m a.s.l.). The tr ia l s were planted in a 4x2 m grid ( i.e. density of 1, 250 trees / ha), on a site with deep fertile
73 volcanic soils, a mean annual temperature of 9.8 C, and 3 ,5 00 mm in mean annual precipitation. Due to restrictions in plant availability, the trials finally included between one and 11 open pollinated families from 31 N. obliqua and 14 N. alpina provenances (Table 3 1). Experimental Design, Measurements, and D ata E diting At the field location, both trials were planted in a randomized complete block design (RCBD) using five complete blocks. Families from all provenances wer e placed at random in each block using single tree plots (STPs). The block sizes were 1 984 and 1 ,0 08 m 2 for N. obliqua and N. alpina respectively, and there were two rows of plants surrounding the trials to avoid uneven competition due to border effects. In the summer of 2004, when trees were four years of age I measured survival ( SURV ), total height ( HT in cm), stem diameter at the root collar ( DIAM in mm), the presence (1) or absence (0) of stem forking ( FORK ), and stem straightness using a r elative sca le from very contorted (0) to very straight (4) ( STR ) on every tree in the trials. I used HT and DIAM to calculate a volume index assuming a cone ( R 2 H in L): R 2 H = (( DIAM / 20) 2 x HT ) / 1000 (3 1) Simultaneously, I obtained a sub sample of five preformed a nd well exposed leaves from each tree on block one at each trial. I harvested, transported, and stored the samples in a 4 C ref rigerator in the Instituto de Silvicultura, Universidad Austral de Chile. I first took surface impressions using clear nail polis h from the abaxial leaf surface between veins of one leaf from each tree. I mounted the impressions on microscopy slides to measure stomatal length ( SL in m) in 10 stomata and stomatal density ( SD in stomata/mm 2 ) in three microscopy fields using x400 and x160 magnification respectively (modified from Steubing et al. (2002)). Right after taking the
74 surface impressions, I scanned all the leaves creating digital images to later measure leaf perimeter ( LPER in cm) and leaf area ( LAREA in cm 2 ) using the P ATCH A NALYST 3.1 extension for A RC GIS 9.1 (Rempel and Carr 2003), and number of lateral veins on the leaf blade ( LVEIN ) counted by eye. Employing P ATCH A NALYST I also calculated a s hape complexity index ( SI ) using LPER and LAREA : SI = LPER / (2* LAREA ) (3 2) w here SI =1 in a circle and is progressively larger than one when the shapes are more irregular. Finally, and right after scanning I dried all the leaves samples in an oven at 80C overnight, and obtained dry weight ( DW in g) for each 5 leaf sample (mod ified from Steubing et al. (2002)). With this, I calculated leaf density ( DENS in g/dm 2 ) using: DENS = ( DW / LAREA )*100 (3 3) I built two data sets with the variables subject to analysis. Data set 1 included SURV growth traits ( HT DIAM and R 2 H ), and for m traits ( FORK and STR ) from the five blocks, and data set 2 included variables LAREA LVEIN DENS SI SL and SD from block one alone. Using SAS software 9.2 (SAS 2008) I ra n P ROC U NIVARIATE to detect suspected outliers by flagging observations that we re at least three standard deviations away from the variable mean. Also, I plot ted the relationship between the highly correlated HT and DIAM to detect out of range data pairs. I re measured the suspicious observations or eliminated them in the cases wher e re measurement was not possible. Analysis of V ariance Trials designed using RCBD with STPs are optimal in obtaining genetic structure and parameter estimates (White et al. 2007). However, the variation in micro site conditions within a field test become s a problem when the block size is too large and
75 the number of replicated blocks is small. In my STP RCBD trials, the block sizes were around 0.2 and 0.1 ha for N. obliqua and N. alpina respectively, and there were only five replicates in each of them. In general, in standard forest sites the general rule is that the block size should be less than 0.1 ha (Matheson, 1989), indicating that the block size for N. obliqua was too large, and for N. alpina was at the limit. To solve this problem I added the positi on of each tree in the trials using rows and columns as random effects nested within blocks in the models as a way to account for the within block spatial variation (White et al. 2007). Analysis of variables from all five blocks For each species/variable combination, I compared nine different models using restricted maximum likelihood analysis (REML) in P ROC M IXED in SAS 9.2 (SAS 2008). I used the Bayesian Information Criterion (BIC) (Schwarz, 1978) to find the optimal model maximizing likelihood and acco unting for overfitting. I started with the simplest model with no effects, and then progressively I added the fixed effect of the provenance, and the random effects of block, provenance by block interaction, family nested within provenance, row nested with in blocks, and column nested within blocks. I additionally ra n all the models that included the block effect allowing heterogeneous variance among blocks. Here I show the full mo del including all the effects: y ijklm = + i + j + ( ) ij + (i)k + (j)l + (j)m + ijklm (3 4) where y ijklm = phenotype of the k th family from the i th provenance in the l th row and m th column within the j th block, = overall mean,
76 i = fixed effect of the i th provenance, j = random effect of the j th block, ( ) ij = random i nteraction of the i th provenance and the j th block, (i)k = random effect of the k th family within the i th provenance, (j)l = random effect of the l th row within the j th block, (j)m = random effect of the m th column within the j th block, and ijklm = r ando m residual error After finding the best model for each variable, I ran the analyses treating all ef fects but provenance as random to estimate the provenances least square means (LSM) and to obtain significance of the among provenances differences I a lso ran the models treating provenance as a random effect to obtain the partition of variation among provenance and among families within provenances. In P ROC M IXED I used the Kenward Roger method to test the significance of fixed effects, and asymptotic t tests to evaluate random effects. Analysis of variables from block one I used the same approach as above to analyze these variables, but eliminating the family and block effects from the models. Thus, for each species/variable combination, I compared five different models. I started with the simplest model with no effects, and then progressively I added the fixed effect of the provenance, and the random effects of row and column. Here, the full model is : y ijkl = + i + j + k + ijkl (3 5) where
77 y ijkl = phenotype of the l th individual from the i th provenance in the j th row and k th column, = overall mean, i = fixed effect of the i th provenance, j = random effect of the j th row, k = random effect of the k th column, and ijkl = random residual erro r Additionally I ra n bivariate analyses in ASR EML 2.00a (Gilmour et al. 2006) of all block one variables with HT and STR which I measured in the whole trial. In this way I took advantage of the genetic correlations among bivariate pairs using the patter ns of spatial variation in the whole trial rather than only in block one to obtain the variance components. Genetic Parameter E stimation Estimates for variables from all five blocks I obtained variance components from all the sources of variation in the tr ials to estimate within provenance individual ( h 2 ) and family ( h 2 f ) narrow sense heritability for each trait pooling together all of the family variances across provenances. Also for each trait, I estimated a coefficient for provenance genetic gain ( CG p ): h 2 = r 1 2 f(p) / ( 2 f(p) + 2 e ) (3 6) h 2 f = 2 f(p) / [ 2 f(p) + ( 2 e / b)] (3 7) CG p = 2 p / [ 2 p + ( 2 f(p) / c) + ( 2 e / (bc))] (3 8) where 2 p = variance among provenances,
78 2 f(p) = variance among families within provenance, 2 e = error variance, r = coefficient of relationship, b = harmonic mean of the number of individuals per family, and c = harmonic mean of the number of families per provenance. I adjusted the coefficient of relationship for half sib families ( r = 0.25) following Squillace (1974). The selfing rate estimates for Nothofagus varied from 2% to 6% in a contro lled pollination study (Riveros et al. 1995) and in an allozyme study (Gallo et al. 1997). Therefore, considering a selfing rate of 5%, and assuming about 10 local unrelated polle n donors, and less than 10 non local pollen donors, the adjusted r was 0.29, significantly smaller than the conservative r = 0.35 used for Nothofagus by Ipinza et al. (2000). In my analysis I used r = 0.29 to obtain additive variances and calculate h 2 and compared these values with the ones obtained using r = 0.35. To obtain the partition of variation within and among provenances, I estimated the proportion of the total morphological genetic variation that is due to differences among provenances using: = 2 p / ( 2 p + 2 f(p) ) (3 9) where = proportion of among provenance variation, 2 p = variance among provenances, and 2 f(p) = variance among families within provenance.
79 Estimates for variables from block one For this analysis, I only obtained variance c omponents to estimate the coefficient for provenance genetic gain ( CG p ). Here, I could not obtain the partition of variation among provenance and among families within provenances. I compared results obtained from both the univariate analysis r a n on S AS an d the bivariate analysis r a n on ASR EML Among P rovenance D ifferentiation When I treated the effects of provenance as fixed effects, I could estimate least square means (LSM) for each variable / provenance combination and obtain the significance of the dif ferences among provenances. Using LSM, I ra n multiple comparison tests among groups of provenances applying the standard Bonferroni correction for multiple comparisons (Rice 1989) at each variable. I grouped the provenances in two ways. First, according t o their geographic origin I found three groups; provenances from the Mediterranean Forests the T ransition al Forests Temperate Rainforests (3800' S). And second, according to a non hi erarchical cl uster analysis performed using the provenances mean annual temperature ( MAT ) and mean annual precipitation ( MAP ) ( Table 3 1). I also obtained correlations among family and provenance LSM for single morphological variables and MAT MAP and the absolute dif ferences in MAT between the trial site and the provenance origin ( DMAT ). Due to the low sample size to test the significance of the correlations using provenance LSM (n=31 and n=14 for N. obliqua and N. alpina respectively) I only tested the dependent vari ables HT DENS SL and SD which were non multicolinear and most adaptively meaningful.
80 Canonical C orrelation, D iscriminant, and C luster A nalysis All multivariate analysis procedures used in this Section: P ROC C ANCOR P ROC S TEPDISC P ROC D ISCRIM P ROC V AR CLUS P ROC F ASTCLUS and P ROC C LUSTER are part of SAS 9.2 (SAS, 2008) and were chosen following McGarigal et al. (2000). Analysis at the provenance level I first employed canonical correlation in P ROC C ANCOR to see how the combination of the environmental variables MAT and MAP were correlated to the morphological variables using squared canonical correlations and significance values. In addition, the standardized canonical coefficients identified which morphological and environmental variables contributed the most to each canonical variable. Also, as a way to define groups of provenances integrating the environmental variables MAT and MAP I performed cluster analyses using standardized values (z scores) of both of these variables. I used P ROC F ASTCLUS to ru n non hierarchical cluster analyses to obtain a stable solution for the number of clusters or groups. I compared the pseudo F statistic (Pseudo F) and the cubic clustering criterion (CCC) simultaneously among solutions to choose the optimal number of group s After finding the optimal number of provenance groups, I ra n stepwise discriminant analysis (DA) in P ROC S TEPDISC to correlate group membership of provenances with the provenance LSM of the non multicolinear morphological variables allowing variables to enter and t o stay in the model if p I conducted a hierarchical cluster analysis on the morphological variables using correlation matrices in P ROC V ARCLUS I input the significant variables identified wi th P ROC S TEPDISC in P ROC D ISCRIM to obtain the posterior probability of membership in groups and to run a cross validation assessing the homogeneity of group dispersions.
81 Finally, to obtain dendrograms indicating closeness in adaptive traits among provenan ces I first performed non hierarchical cluster analyses in P ROC F ASTCLUS combining all non multicolinear morphological traits to obtain a stable solution for the membership in main clusters. I compared the pseudo F statistic (Pseudo F) and the cubic cluste ring criterion (CCC) simultaneously among solutions to choose the optimal number of groups using standardized values of the provenance LSM (z scores). Then, I performed hierarchical cluster analyses in P ROC C LUSTER testing several clustering methods ( i.e. minimum variance). I chose the dendrogram from the method that had the best match in main clusters with the grouping obtained from the more stable non hierarchical cluster analyses. An alysis at the family level Similar to the analysis at the provenance level I ra n stepwise DA in P ROC S TEPDISC to correlate provenance membership of families with the family LSM of the non multicolinear morphological variables and then I used P ROC D ISCRIM to obtain the posterior probability of membership in provenances and to run a cross validation assessing the homogeneity of provenance dispersions. Results Best M odels for the A nalysis of V ariance Using the Bayesian Information Criterion (BIC) (Schwarz, 1 978) I chose the best of nine models for each species/variable combination. For the variables measured in all five blocks, I choose three models. For STR in Nothofagus obliqua and HT DIAM and R 2 H in N. obliqua and N. alpina the optimal model was: y ijklm = + i + j + (i)k + (j)l + (j)m + ijklm (allowing HVAB) (3 10)
82 For FORK in N. obliqua and STR in N. alpina : y ijk = + i + j + (i)k + ijk (allowing HVAB) (3 11) Finally, for FORK in N. alpina : y ij = + i + j + ij (3 12) where y ijklm = pheno type of the k th family from the i th provenance in the l th row and m th column within the j th block, = overall mean, i = fixed effect of the i th provenance, j = random effect of the j th block, (i)k = random effect of the k th family within the i th prov enance, (j)l = random effect of the l th row within the j th block, (j)m = random effect of the m th column within the j th block, ijklm = random residual error, and HVAB = heterogeneous variance among blocks. Likewise, for the variables measured only in block one, I choose three models. For LVEIN SI SL and SD in N. obliqua and LVEIN DENS and SD in N. alpina the optimal model was: y ikl = + i + k + ikl (3 13) For LAREA in N. obliqua and SL in N. alpina : y ijl = + i + j + ijl (3 14) Finally, f or DENS in N. obliqua and LAREA and SI in N. alpina :
83 y il = + i + il (3 15) where y ijkl = phenotype of the l th individual from the i th provenance in the j th row and k th column, = overall mean, i = fixed effect of the i th provenance, j = random eff ect of the j th row, k = random effect of the k th column, and ijkl = random residual error Genetic P arameter E stimates With the variance components obtained from the analysis of the chosen models I obtained the genetic parameter values shown in Tables 3 2 and 3 3. Within provenance individual ( h 2 ) and family ( h 2 f ) narrow sense heritabilities were substantial for all variables in N. obliqua and for HT and STR in N. alpina In N. alpina the heritabilities for DIAM and R 2 H were low and FORK had no heritabil ity. Also, values were consistently higher in N. obliqua than in N. alpina for all variables ( Table 3 2). The values of the coefficient for provenance genetic gain ( CG p ) were in general much higher than the values for heritabilities and higher in N. obliqu a than in N. alpina except for FORK STR LAREA and SI (Tables 3 2 and 3 3). The values were all greater than 0.5 except for FORK in N. obliqua ( =0) indicating that the proportion of the among provenance variation was usually greater than the proporti on of the among family variation ( Table 3 2).
84 The narrow sense genetic correlations among traits (r A ) are shown in Table 3 4. There was a wide range of r A values among variables, but the correlations of trait pairs HT / SL STR / SI STR / SL in N. obliqua and HT / SL STR / SL STR / SD in N. alpina were greater than 0.75 and are regarded as strong. Among Provenance D ifferent iation Using the provenance LSM for all traits ( Table 3 5) I tested the statistical significance of the differen ces among provenances. Different iation among provenances in N. obliqua was not significant ( >0.05) for FORK and SL significant ( <0.05) for SI and SD and highly significant ( <0.01) for all other variables including the growth traits ( HT DIAM R 2 H ). Conversely, in N. alpina the diff erences among provenances were highly significant only for HT and STR significant for DIAM R 2 H LAREA and SI and not significant for all other variables ( Table 3 6). These significant differences imply that at least one provenance was different than th e average value of all provenances. To avoid performing too many multiple comparisons among all provenances in each species, which would seriously hinder the power of the tests, I used contrast analysis to compare groups of provenances defined by latitudin al differences in type of forests (Mediterranean Forests, Transition al Forests and Temperate Rainforests) and defined by a non hierarchical cluster analysis based on environmental variables ( MAT and MAP ) that found three groups of provenances for N. obliq ua and two for N. alpina ( Figure 3 2). Results in Figure 3 3 show that for N. obliqua HT DIAM and R 2 H were significantly lower ( p <0.0006) in the Mediterranean Forests than in the other two zones. Also SL had the same pattern, but the values were higher i n the Mediterranean Forests ( p =0.045). The variables STR and LAREA and SI
85 did not follow the same tendency, however the values were still lower in the Mediterranean Forests than in the Temperate Rainforests ( p <0.0006, p =0.018, and p =0.003 respectively). I also found for N. obliqua that HT DIAM and R 2 H were significantly higher ( p <0.0006) in Warm and Rainy provenances than in the other two groups ( Figure 3 4). In the case of SL the Warm and Rainy provenances were only significantly lower in value than the Warm and Dry group ( p =0.026), and in SD the values were only significantly lower than the Cold and Rainy group ( p =0.006). For N. alpina I did not find any geographic or environmental pa ttern of variation in any of the putatively adaptive traits like the g rowth traits, LAREA DENS SL or SD The only significant differences were found between the Mediterranean Forests and the Temperate Rainforests in SI ( Figure 3 5). The values of SI were lower in the Temperat e Rainforests ( p =0.013 ). Patterns of Variation Related to S ingle Environmental V ariables I obtained Pearson correlations ( r ) and their significance using Bonferroni corrections of provenance and family least square means (LSM) for HT DENS SL and SD with MAT MAP and DMAT to detect possible trends o f adaptation to the environmental variables. In N. obliqua I found a trend of increased HT with an increment in MAP using the provenance LSM ( Figure 3 6a, r =0.45, p =0.132, n=31). Although r was not significant at this level, the correlation using family LS M was consistent with that result becoming highly significant due to the increase in the number of observations (Fig, 6b, r =0.26, p <0.0001, n=247). Something similar occurred with the correlation between SD and MAT ( Figure 3 6c and d) but in this case the increase in the number of observation (from n=31 to n=247) did not improve the significance of the correlation
86 (from p =0.091 to p =0.062). Finally, the relationships of HT with MAT and DMAT only showed a trend when using family LSM ( Figure 3 6e and f, n=247 ). HT tended to significantly increase with MAT ( r =0.23, p =0.002) and decrease with DMAT ( r =0.25, p <0.001). In N. alpina I only found a trend of decreased DENS with an increment in MAP using the provenance LSM ( Figure 3 7a, n=14). Although I obtained a hig h correlation value ( r =0.69), it was only slightly significant ( p =0.083) and that significance did not improve when using the family LSM ( Figure 3 7b, r =0.27, p =0.062, n=104). Multivariate A nalysis Multivariate association between morphological and environ mental traits In the canonical correlation analysis, the standardized canonical coefficients indicate the relative contribution of each trait to the canonical variables. For N. obliqua the canonical variable morpho1 was dominated by HT However, SD and, i n a lower degree, SI were also important to correlate with env1, in which MAT was somewhat more important than MAP ( Table 3 7 ). In the other hand, in morpho2 HT STR and FORK were prevalent to correlate with env2, in which MAP clearly dominated over MAT ( Table 3 7 ). I obtained a high canonical correlation ( r ) for morpho1 vs. env1 ( r =0.86) and a moderate one for morpho2 vs. env2 ( r =0.61) ( Figure 3 8). For N. alpina morpho1 was controlled by SD and SL but also STR LVEIN and DENS played a role to correlat e with env1, in which MAP was a little more important than MAT ( Table 3 8 ). The second canonical variable, morpho2 was clearly dominated by LVEIN and also DENS and SL are important to correlate with env2, in which MAT dominates over MAP ( Table 3 8 ). The c anonical correlations for morpho1 vs. env1 and for morpho2 vs. env2 were very high with r =0.96 and r =0.91 respectively ( Figure 3 9).
87 Group membership of provenances I tested whether a combination of morphological traits would have the power to discriminate group membership of provenances using the groups based on environmental variables ( Figure 3 2). The stepwise discriminant analysis indicated that only two morphological variables ( HT and SD ) in N. obliqua and only one ( DENS ) in N alpina were useful to dis criminate groups of provenances ( Table 3 9 ). Using these variables I obtained posterior probabilities of group membership indicating that overall 21 out of 31 (68%) N. obliqua provenances and 11 out of 14 (79%) N. alpina provenances were classified correct ly in the groups predicted by the environmental variables. In N. obliqua group Warm and Rainy did not have any misclassified provenances, group Cold and Rainy had two, and group Warm and Dry had eight misclassifications ( Table 3 10 ). In N. alpina group C old and Rainy had one misclassification, and group Warm and Dry had two ( Table 3 10 ). After a cross validation, the canonical discriminant functions were able to estimate the correct group for 50, 36, and 82% of the N. obliqua provenances in the Cold and R ainy Warm and Dry and Warm and Rainy group respectively ( Table 3 11 ). In N. alpina the cross validation showed that 80 and 78% of the provenances were correctly classified into the Cold and Rainy and Warm and Dry groups respectively ( Table 3 1 2 ). Geneti c similarities among provenances through morphological traits I obtained dendrograms indicating closeness in adaptive traits among provenances by matching more stable results of non hierarchical cluster analyses with a series of hierarchical cluster analys es obtained using different clustering methods. I got variance method in both species ( Figure 3 10).
88 In N. obliqua ( Figure 3 10a), there were two well defined clusters in the dendrogram. One small cluster composed of three provenances from the Mediterranean Forests that also belong to the Warm and Dry group, and Galletue (17), which belongs to the Temperate Rainforests and to the Cold and Rainy group. The second cluster is large and contains all the other provenances. However, this cluster is subsequently composed of three branches. One that contains the rest of the Mediterranean Forests provenances except Vilches (03) and most of the Transition al Forests provenances except Reserva uble (10) and Lago Lanalhue (13) but without a climatic identity, and the other two that include most of the provenances from the Temperate Rainforests and differentiate from each other because one contains predominately Warm and Dry provenances and the other contai ns predominately Warm and Rainy provenances. In N. alpina ( Figure 3 10b) there were three main clusters. However, they did not correspond to either the type of forests or the environment groups based on MAT and MAP Provenance membership of families Using discriminant analysis I tested if a combination of morphological traits would have the power to discriminate provenance membership of families. Results indicated that, in general, the posterior probability of membership in provenances was very low and prov enance dispersions very high. The error rates ( ER ) obtained with cross validation averaged 0.87 in N. obliqua and 0.90 in N. alpina Only provenances Reserva uble (10, ER =0.29) from N. obliqua and Recinto (35, ER =0.40) from N. alpina had error rates lower than 0.60.
89 Discussion Within Provenance H eritabilities The individual tree heritabilities ( h 2 ) I obtained for N othofagus obliqua in growth traits ( HT DIAM R 2 H ) and form traits ( FORK and STR ) ( Table 3 2) were well within the overall range of 0.1 to 0.3 s tated for forest tree species (White et al. 2007) and, in general, close to the average values obtained by Cornelius (1994) for height (0.28), diameter (0.23), volume (0.21), and stem form (0.28) from a meta analysis with 67 forest tree heritability repor ts. Although the values obtained in my research represent single test locations and therefore are upwardly biased (White et al. 2007), the heritabilities in N. obliqua are substantial and indicate that there is an important portion of the phenotypic varia tion within provenances that is genetically controlled and therefore there is a potential for both differentiation of populations through natural selection and genetic gains through artificial selection. Ipinza et al. (2000) measured the stem height of fiv e month old seedlings obtained from the same mother trees I analyze here Th e s e author s reported a significantly higher heritability in height ( h 2 =0.50). However, these plants were growing in the much controlled envir onment of a greenhouse emphasizing the fact that environmental homogeneity is one of the chief factors that control heritability (Cornelius, 1994). The lower h 2 values I obtained in N. alpina for the same variables indicate that the response of this species to either natural or artificial selec tion will be poorer than in N. obliqua Medina (2001) also found higher h 2 values in the offspring of the same mother trees growing in a greenhouse for N. alpina When the plants were seven months o ld this author estimated h 2 =0.52 and h 2 =0.40 for stem hei ght and stem diameter respectively, confirming the trend found in N. obliqua As for field trials, with higher
90 environmental variation, there are estimates of h 2 =0.07 for stem diameter and stem height in a 1 year old progeny test of N. alpina (Gutirrez an d Ipinza, 2000), which are comparable with the values obtained in the current study Comparing h 2 estimates among traits, there is a general trend for height and form to have higher values than for diameter and volume (Cornelius, 1994; White et al. 2007), probably because growth in diameter and, consequently, in volume have an extra environmental source of variation when the tr ia l density is uneven. I observed this trend in N. alpina but not in N. obliqua where the h 2 estimates were considerably higher in DIAM and R 2 H than in HT I do not have an explanation for these unexpected results provided that the N. obliqua and N. alpina trails were planted at the same density and overall survival was only slightly better in N. obliqua (97%) than in N. alpina (94%). The h 2 estimates obtained using the coefficients of relationship r =0.29 defined in my study are slightly higher than the estimates obtained using the more conservative r =0.35 defined by Ipinza et al. (2000), but do not change the trends of my results. How ever, it would be wise to use the conservative estimates when projecting future genetic gains of individual selections in a genetic improvement program. For that matter, the family heritability ( h 2 f ) is also important for its use in selection indices becau se it is typically larger than h 2 and therefore can improve selection efficiency. The h 2 f estimates in N. obliqua and N. alpina are larger than the h 2 estimates however, the differences are not that significant to take advantage of them in a selection inde x. The reason for this is that the trials have only five replicates to obtain h 2 f Genetic C orrelations among T raits Obtaining genetic correlations among traits ( r A ) was not a goal in my study but as I ra n the bivariate analysis in ASR EML (Gilmour et al. 2006) to obtain better estimates of
91 the variance components in the block one variables, I obtained values for r A ( Table 3 4). Genetic correlations are a measure of pleiotropy. Thus, traits with a strong r A are influenced by a similar set of genes and there fore changes in the allelic frequency of those genes will influence both traits (White et al. 2007). For example, the strong r A estimates I obtained between HT and SL would imply that any time t here is a selective pressure to increase height growth in N. o bliqua or N. alpina there would be a decrease or an increase in stomatal length respectively. Having strong trait trait genetic correlations like these could have interesting ecological and economical implications. In my example, assuming that measures of SL do not vary with age, I could potentially use SL for early selection of HT in one year old N. obliqua or N. alpina seedlings growing in a greenhouse with increased environment control. Likewise, I could use SI for early selection of STR in N. obliqua o r SL for early selection of STR in N. alpina However, we have to keep in mind that genetic correlations are very difficult to estimate with precision. Precise r A estimates come from large progeny tests with lots of families and repeated in many locations. In my case, even though I analyzed many families, I did not analyze multiple locations. Therefore, these results must be taken with caution and genetic correlations must be confirmed in further analys e s. Moreover, in some cases genetic correlations are ve ry incongruent between N. obliqua and N. alpina It is not clear to me why the underlying genetic mechanisms to determine, for example, stomatal length or stomatal density would be so different in two sympatric and closely related species.
92 Among P rovenance D ifferent iation Growth traits In my study I found important levels of among provenance differentiation ( i.e. structure) in N. obliqua for growth traits ( HT DIAM and R 2 H ). Growth traits in N. obliqua present estimates that are larger than the typical v alues obtained for forest trees which range from =0.4 to = 0. 8 ( Hodge and Dvorak, 2004; Li et al. 1993 ; S ebbenn et al. 2003; Sierra Lucero et al. 2002; Tchatchoua and Aravanopoulos, 2010a ; 2010b ; White et al. 2007; Wightman et al. 2008), indicating adaptation to specific environments in this species. Additionally, the large CG p values I obtained in N. obliqua indicate that a large fraction of the among provenance phenotypic variability is genetically controlled. CG p is equivalent to h 2 in the sense t hat it is a measure for evaluating the potential of a species trait to respond to natural or artificial selection at the provenance level. If we consider growth as a measure of adaptation, the large CG p estimates (>0.8) indicate that the potential for futu re adaptation is high and that this species would be, in theory, well prepared to withstand ongoing climatic changes. Also, for breeding programs, growth traits are obviously important and my data make it clear that the genetic gain due to selection of spe cific provenances is highly probable. A limitation of my study is that I obtained data from only one site without the possibility of testing provenance x environment interactions. Therefore, results are only applicable to sites similar to the one the proge ny provenance trials were planted. Keeping this in mind, the ranking of the 25% best prov enances in growth traits are Lago Lanalhue (13), Curanilahue (11), Curarrehue (22), Malalhue (24) Llifen (27), Lastarria (21), Choshuenco (25), and Futrono (26).
93 In N alpina the estimates for growth traits are lower than in N. obliqua but they are within the typical values for forest trees indicating moderately structured populations. Also, the differences among provenances are significant, but p values are not as impressive ( p <0.02) Thus, the potential for future adaptation is lower in N. alpina and the chances to withstand climatic changes might be also lower. I t is important to keep in mind that when analyzing progenies it is impossible to separate within family from environmental variation and therefore the estimates are upwardly biased (White et al. 2007), and unfortunately, the extent of this bias is unknown. Nevertheless, because all the estimates in my study and in most of the literature are obtained from provenance progeny tests, values are comparable among species and traits. In my study the between species differences in structure make sense because N. obliqua has a wider ecological range than N. alpina (Donoso, 1982; Veblen and Schlegel, 19982) and th ere is a tendency to find less structure is species growing in ecologically homogeneous regions (Xie, 2008). The lower CG p values in N. alpina are an indication that the phenotypic variance among provenances in this species is, in a greater way, due to pla sticity and therefore the expected genetic gain due to selection of specific provenances would be less promising. Thus, using a provenance ranking with this species may not be useful Still, just to show the best provenances, the ranking with the 25% best provenances in growth traits includes Santa Barbara (37), Pichipillahuen (39), Nahuelbuta (36), and Releco (43).
94 Other adaptive traits For other putative adaptive traits like LAREA DENS SL and SD I could not obtain values of structure. However, the stat istical significance of the differences among provenances gives us a general idea of variation. In N. obliqua there is highly significant among provenance variation for LAREA and DENS ( p <0.005), and significant for SD ( p =0.039). This variation appears to be adaptively important in the two former traits given that their genetic control at the provenance level was moderately high ( CG p > 0.4). On the contrary, CG p was low for SD and for the non significant SL In N. alpina LAREA was also significantly differe nt among provenances ( p =0.021) and CG p was moderately high. However, DENS SL and SD were all non significant and had low genetic control at the provenance level confirming that N. alpina has lower potential for future adaptation than N. obliqua Non adapt ive traits For form traits like stem straightness and forking, which do not have a known adaptive value, the estimates in the literature are logically low and range from 0 to 0.4. However, I obtained inconsistent estimates in my study. In N. obliqua the lack of structure, genetic control at the provenance level, and significance of FORK agree with the literature. However, N. obliqua had moderate structure for STR with relatively high CG p and highly significant differences among provenances. Also in N. a lpina the high estimates in both FORK and STR had different meaning because while FORK was not significant and had low CG p STR was highly significant and had relatively high CG p Ruling adaptation out, these estimates for STR in both species might be d ue to random genetic drift in chloroplast genes that cannot be easily equilibrated by
95 chloroplast gene flow from seeds is these species (Azpilicueta et al. 2009; Marchelli and Gallo, 2006). Another plausib le reason is that this different iation is a by pro duct of selective pressure in other genetically correlated traits with putative adaptive value like LAREA and SL in N. obliqua or DENS SL and SD in N. alpina Regardless of the reason, my data indicate that STR has a good potential for obtaining genetic gain due to selection of specific provenances in both species. The case of FORK in N. alpina is different since all of its genetic variation is allocated among provenances with a complete lack of among family variability, so this value is probably a matt er of chance alone. Other apparently non adaptive traits like LVEIN in N. obliqua and SI in N. alpina had moderately high CG p values. The s ame reasons as apply for STR may explain these estimates. Geographic and E nvironmental P atterns of V ariation Single t rait analysis for N. obliqua Growth traits ( HT DIAM and R 2 H ) in N. obliqua present a cl ear and significant geographic pattern that is somewhat related also to environmental variation. The better performance in growth traits observed, in average, for prov enances from the Temperate Rainforests (colder and rainier) and the Transition al Forest in comparison with provenances from the Mediterranean Forests (warmer and dryer) indicates that the latter may have an adaptive disadvantage growing in the Temperate Ra inforest environment where the trial w as planted. Nevertheless, these geographic areas that coincide with the north, central and south parts of the species distribution have within area climatic variation in temperature and precipitations mainly related to elevation. This is why I also grouped populations according to the climatic parameters MAT and MAP
96 The highly and significant ly better performance in growth traits observed, in average, for provenances from the Warm and Rainy group in comparison to the W arm and Dry and Cold and Rainy groups also show that provenances adapted to environments similar to the one where the trial was planted had an advantage. Interestingly, the ranking with the 25% best provenances show that for this species, six out of the to p eight are from the Temperate Rainforest and grow geographically very close to the study site. And the other two (Lago Lanalhue (13) and Curanilahue (11)), remarkably top one and top two, are coastal provenances from the Transition al Forests that grow ver y close to each other. Now, regarding environmental conditions, the latter two share the Warm and Rainy group with most of the top eight. Only Curarrehue (22) and Futrono (26) are from the Cold and Rainy group, but are very rainy and not too cold. Converse ly, the bottom eight provenances tend to grow far from the study site, mostly in the Mediterranean Forest, and also tend to belong to the Warm and Dry group confirming the climatic tendency. In the bottom eight, Reserva uble (10) and Galletue (17) are two interesting exceptions belonging to the Cold and Rainy group, but they differ from top eight provenances 22 and 26 in that the former grow in environments much colder than the study site. In general, there are two possible reasons for this pattern of vari ation. The first is genetic adaptation to slow growth to withstand adverse conditions in their site of origin like extremely cold or dry environments, restrictive soil characterist ics, or wind. The second is mal adaptation to new conditions that are not nec essarily extreme, but that would prevent the development of the full potential of a plant. Unfortunately, I could not test this because of the lack of reciprocal gardens in my study. Nevertheless, there is a
97 study that measured in situ growth rates showing that N. obliqua provenances belonging to the bottom eight i n my study do not necessarily have poor growth rates when growing in their sites of origin (D onoso et al. 1993), indicating that at least in some cases the reason for slow growth in provenances i s mal adaptation to the new site. Still, adaptation to grow slowly could be playing a role in, for example, very cold provenances as it has been reported in N. pumilio (Premoli et al. 2007), N. menziesii (Ledgard and Norton, 1988), and N. solandry (Ledgar d and Norton, 1988; Wilcox and Ledgard, 1983). The significant increase in HT due to the increase in MAP on the provenance confirms the importance of genetic adaptation to the environment to increase the chances of a superior performance. I can say this be highest MAP value, so the relationship between HT and MAP also represents a significant increase in HT when MAP from the provenance is more similar to that of the DM AT (the differ ence in MAT and the provenance) have a significant and positive influence in HT Because of the clinal nature of the variation in MAP and MAT throughout the study site, it is not surprising that the performance in growth traits als o follows a general north to south cline. To attempt reducing the noise due to elevation and physiography I selected a set of provenances located in the Andes and at average elevation for the latitude. The provenances were Bellavista (2), Vilches (3), Reci nto (9), Santa Barbara (12), Cunco (19) and Llifen (27). For this set of provenances, the clinal variation in growth traits was very clear agreeing with the general clinal trends in other traits like seed or flower size in
98 N. obliqua (Donoso, 1979; Ipinza et al. 2000 ), N. alpina (Donoso, 1987; Medina, 2001), and N. dombeyi (Chultz, 2005) in Chile. Despite being app arently non adaptive trait s STR and SI showed highly significant geographic patterns similar to g rowth traits with better STR in provenances fr om the Temperate Rainforests than from the other zones, and higher SI (more shape complexity) in the Temperate Rainforest than in the Mediterranean Forests. However, these patterns are not significantly related to environmental variation. The ranking in ST R with the top eight provenances show that all of them are from the Temperate Rainforest s but do not grow very close to the study site. On the other hand, six of the bottom eight provenances grow in the Mediterranean Forest s and the other two in the Trans ition al Forests To compare the latitudinal cline in STR and SI with the one obtained in HT I selected the same six provenances located in the Andes on average elevation s at each latitude I obtained a clear latitudinal cline in STR but not in SI These p atterns of variation rules out the random drift explanation for the among provenance variation of the s e trait s and suggest that genetic correlations between these and other traits with adaptive value could be the reason. Indeed, my findings show high r A va lues between STR and SI between STR and SL and a moderate r A value between STR and LAREA. LAREA DENS SD and SL are all putative adaptive traits with known variation patterns in some Nothofagus species. In general, when the elevation range is wide enou gh genotypes in high elevation provenances exposed to very cold environments may have smaller leaves (Premoli et al. 2007), thicker leaves (Hovenden and Vander Schoor, 2003 ; 2006), less dense stomata (Hovenden and Vander Schoor, 2003 ; 2006;
99 Premoli and Br ewer, 2007), and possibly smaller stomata. In my study results show a clear and highly significant lower SD in the Cold and Rainy provenances compared with the other two groups and a slightly significant positive correlation between SD and MAT suggesting that N. obliqua genotypes growing in colder environments are genetically adapted to cope with hydric stress due to low temperatures reducing the stomatal density. Interestingly, hydric stress due to drought seems not to have any impact on stomatal density in this species. The stomatal size measured as SL was significantly larger in the Mediterranean Forest and also in the Warm and Dry provenances, but this trend goes against the expectations and seems to be driven by only three provenances from the northern most area of distribution (Tiltil (1), Bellavista (2), and Los Ruiles (4)). Besides, the correlation between SL and MAP is very low and not significant. All of this suggests that this trait is probably not an adaptation to drought and may be the product o f the high genetic correlation between SL and HT or it may be related to random genetic drift in chloroplast genes. As Premoli et al. (2007) stated for N. pumilio genotypes in very cold environments had reduced leaf size. Instead, my findings indicate tha t the average LAREA was significantly larger in the overall colder Temperate Rainforests, and that in addition, there was a not significant difference among environmental groups or correlation with MAT Finally I had the notion that denser leaves could be an adaptation to resist drought or extreme cold, but DENS appeared to be non adaptive in N. obliqua Even though it had highly significant among provenance variation, it did not follow any trend related to spatial distributions, temperature, or rainfall.
100 Single trait analysis for N. alpina In addition to its lower among provena nce different iation and structure, I obtained very little evidence that the traits I measured in my study had adaptive significance for N. alpina T here was no evidence of a pattern of variation related to spatial distribution or environmental variability in any but two traits. For SI as in N. obliqua I am not sure of the reason why shape complexity would vary geographically. Even more, given that the trend is the opposite in both s pecies and is not significantly related to environmental conditions. Therefore, this variation pattern might be also due to random genetic drift in chloroplast genes The only adaptively meaningful relationship in this species was the slightly significant correlation between DENS and MAP indicating that, unlike N. obliqua leaf density could have a role in controlling water losses within N. alpina plants. This correlation, however, was not enough to be detected as a significant difference between Cold and R ainy and Warm and Dry groups of provenances. Multi trait analysis for N. obliqua The high canonical correlations relating morphological and environmental variables in N. obliqua indicates that there is a strong effect of MAT and MAP on the morphology of th e provenances and the most important variables for the canonical models agree in general with the single trait analyses. The results here stress the importance of the environmental conditions where genotypes are adapted, and suggest that those adaptations influence the characteristics in growth and anatomy that those same genotypes would have when growing in a different environment. The canonical correlations confirm important positive effects of warmer and rainier environments on growth and stomatal densit y (env1 vs. morpho1), but also highlight negative effects of
101 precipitation on stem characteristics (env2 vs. morp ho2), which were not seen in other analysis. The discriminant analysis on group membership emphasized also the importance of HT and SD as trait s discriminating group of provenances predicted by environmental variables. However, error rates obtained in the cross validation analysis are very high indicating that there must be other environmental variables that correlate better with the morphologica l traits I studied. Six out of 10 misclassified provenances were originally assigned to the Warm and Dry group and grow at low elevations in the coast or the Central Valley of the Temperate Rainforests with precipitations ranging between 1 300 and 1 700 mm I think that probably the cluster analysis used to create the groups used a threshold in precipitations that was too high. Also, the amount of precipitations in the dry season could have being a better predictor considering that in the Temperate Rainfore sts summers are considerably shorter than in the Mediterranean Forest s and somewhat shorter than in the Transition al Forests Unfortunately, summer precipitation data was not available for the provenance sites at the moment of my study. In the case of the discriminant analysis on provenance membership, results indicate that it is not possible to predict by any means the membership of single families on provenances. This indeed makes sense because of the large family variance in every trait, and because the sample points defined as provenances are arbitrary and in many cases two or more of them could be part of a morphologically homogeneous population. The only exception to this was provenance Reserva uble (10) that grows at the highest elevation among the s ites of my study The families on this provenance have a uniq ue combination of low growth, low fork percentage, and high leaf density.
102 Finally, when I tried grouping provenances using only morphology, the resulting dendrogram related very well with a combi nation of geography and environmental characteristics confirming structure and the influence of adaptation on at least some traits in this species. The provenances that looked more misplaced within the dendrogram were Galletue (17) placed in a Mediterranea n/ Warn and Dry branch, and Reserva uble (10) placed in a Rainforest branch. Interestingly, these are the two provenances growing in extreme environments at high elevations that make them to have unique adaptive characteristics difficult to assign to any p articular cluster. Multi trait analysis for N. alpina The high canonical correlations relating morphological and environmental variables in N. alpina are very surprising because I found little evidence of relationships among the species single traits and M AT and MAP Therefore, the high values obtained here are probably due to overfitting considering that the sample size was only n=14 and I combine d eight variables in the models Still, the canonical correlations suggest positive effects of warmer and raini er environments on stomatal density and stomatal size (env1 vs. morpho1), and contradictorily suggest negative effects of temperature on stomatal size (en v2 vs. morpho2). Even though the s e effects were not seen in other analysis, the relationship between w armer and rainier environments and stomatal density matches the one found in N. obliqua and strengthen this idea. The discriminant analysis on group membership highlight ed the importance of DENS as the only trait discriminating group of provenances predict ed by environmental variables. DENS also was the only variable in the single trait analyses that had any significant relationship with the environment and here it did a good job discriminating groups. Error rates obtained in the cross validation analysis w ere low and only three out
103 of 14 provenances were misclassified. Like in N. obliqua two out of the three misclassified provenances were originally assigned to the Warm and Dry group and grow with precipitations around 2 000 mm. This confirms my thoughts d iscussed for N. obliqua about the problem with finding the correct threshold in precipitations and the possibility that the amount of precipitations in the dry season could have being a better predictor. In the case of the discriminant analysis on provenan ce membership, results are also similar to the ones found in N. obliqua I cannot predict membership of single families on provenances. The exception here was provenance Recinto (35) that, unlike Reserva uble ( 10), does not grow under any particular condi tions respect to the other provenances The families from this provenance have a unique combination of low growth, high leaf area, and high leaf shape complexity. Finally, when I tried grouping provenances using only morphology, the resulting dendrogram di d not relate at all with any geographic or environmental characteristics confirming in general the low influence of adaptation on most of the traits in this species. Conclu ding Remarks My study shows very dissimilar characteristics between N. obliqua and N alpina regarding the patterns of adaptive genetic variation and the effect of environmental factors on those patterns. On one hand, N obliqua is a species that presents substantial individual heritabilities in growth and form traits large proportion of genetic variation due to among provenance differentiation in growth traits, large coefficients for provenance genetic gains in growth and other adaptive traits and several significant associations between morphological and geographic and environmental pat terns of variation. Conversely, N. alpina has lower individual heritabilities in all traits, only moderate
104 proportion of genetic variation due to among provenance differentiation in growth traits, lower coefficients for provenance genetic gains, and little association between morphological and geographic and environmental patterns of variation These differences between the two species are in agreement with N. obliqua range and imply that in this species there is a higher potential for se lection of provenances and individual trees than in N. alpina It seems likely that N. obliqua would have a better chance to adapt to rapid climatic changes through natur al selection and that breeders would obtain larger geneti c gains by selecting the pro venances and individuals with desirable traits in N. obliqua than in N. alpina In N. obliqua the association of morphological features like growth, stem form, leaf shape, and stomatal density with geogr aphic and environmental variation indicates that pro venances are adapted to particular conditions of temperature and precipitations. For example, t ree s of this species tend to gr o w faster when they co me from provenances that had similar environmental conditions i.e. warm and rainy, in relation to the place they were planted. This probably happens because provenances growing in very cold or very dry environments may be genetically adapted to slow growth to endure harsh conditions or because they are maladapted to the new conditions perhaps due to, among oth ers new diseases or the lack of a specific mycorrhizal association W e would only have a more complete picture by testing a similar set of genotypes in colder and also in drier environments.
105 Table 3 1 Provenances of N othofagus obliqua and N. alpina an alyzed in my study I made all measurements in the progeny provenance trials (PPT) locate in Fundo Ar quilhue, Los Rios Region, Chile # Provenance N F Â£ Location Latitude (S) Longitude (W) Elevation (m a.s.l.) MAT (C) MAP (mm) N. obliqua 1 Tilt il Â¢ 3 Coast 3305' 7058' 980 11.8 500 2 Bellavista Â¢ 1 Andes 3413' 7044' 970 10.7 1,300 3 Vilches Â¢ 6 Andes 3535' 7104' 1 310 8.0 1,500 4 Los Ruiles 3 Coast 3550' 7234' 230 12.0 1,000 5 Quirihue 10 Coast 3618' 7232' 260 11.9 1,000 6 Bullileo Al to 3 Andes 3622' 7122' 1 650 6.5 2,000 7 Ninhue 10 Coast 3623' 7223' 150 14.0 800 8 Cayumanqui 5 Coast 3642' 7229' 230 11.9 1,200 9 Recinto 9 Andes 3651' 7140' 710 11.2 2,200 10 Reserva uble 8 Andes 3656' 7113' 1 890 5.3 3,000 11 Curanilahu e 6 Coast 3728' 7321' 140 12.5 2,000 12 Santa Barbara 10 Andes 3740' 7158' 430 10.5 1,800 13 Lago Lanalhue 7 Coast 3751' 7321' 230 12.1 2,000 14 Ralco 10 Andes 3753' 7135' 810 9.5 4,000 15 Victoria 10 Central Valley 3812' 7210' 480 11.4 1,500 16 Pichipillahuen 9 Coast 3820' 7303' 450 11.0 1,700 17 Galletue 6 Andes 3837' 7116' 1 450 5.8 2,000 18 Quepe 10 Central Valley 3852' 7230' 120 12.4 1,700 19 Cunco 10 Andes 3852' 7151' 410 11.0 2,700 20 Lago Colico 10 Andes 3901' 7201' 430 10.9 2,500 21 Lastarria 9 Coast 3923' 7242' 180 12.3 2,200 22 Curarrehue 7 Andes 3923' 7132' 910 7.5 3,200 23 Cruces 10 Coast 3931' 7304' 160 11.7 2,300 24 Malalhue 9 Central Valley 3931' 7232' 140 11.8 2,500 25 Choshuenco 8 Andes 3951' 7206 220 10.4 3,000 26 Futrono 10 Andes 4005' 7220' 690 8.1 2,500 27 Llifen 8 Andes 4011' 7215' 290 10.1 2,900 28 Llancacura 10 Coast 4018' 7326' 60 11.5 1,300 29 Rio Negro 10 Coast 4047' 7316' 110 11.2 1,200 30 Rupanco 10 Central Valley 4049' 7 254' 130 11.1 1,600 31 Purranque 10 Coast 4052' 7314' 90 11.2 1,500
106 Table 3 1 Continued # Provenance N F Â£ Location Latitude (S) Longitude (W) Elevation (m a.s.l.) MAT (C) MAP (mm) N. alpina 32 Siete Tazas 10 Andes 3525' 7102' 760 10.7 1,400 33 Vilches 5 Andes 3535' 7104' 1 200 8.5 1,500 34 Bullileo Alto 10 Andes 3622' 7122' 1 650 6.5 2,000 35 Recinto 10 Andes 3651' 7140' 810 10.7 2,200 36 Nahuelbuta 6 Coast 3740' 7302' 940 8.7 2,000 37 Santa Barbara 11 Andes 3740' 71 58' 430 10.5 1,800 38 Jauja 10 Andes 3806' 7158' 670 9.3 1,900 39 Pichipillahuen 10 Coast 3820' 7303' 420 11.2 1,700 40 Malalcahuello 3 Andes 3828' 7138' 970 8.2 3,500 41 Melipeuco 2 Andes 3850' 7140' 550 10.3 3,200 42 Curarrehue 10 Andes 392 3' 7132' 840 7.9 3,200 43 Releco 10 Andes 3943' 7208' 1 160 6.2 3,000 44 Llancacura 9 Coast 4017' 7321' 380 9.9 1,500 45 Hueyusca 8 Coast 4055' 7335' 410 9.5 1,700 PP T trials (Arquilhue) Andes 4014' 7203' 310 9.8 3,500 Â£ NF=Number of familie s. The coordinates and elevation of the source provenances sampled for the provenance trials is only an approximation. There are not records of the exact position of the mother trees harvested to plant the trials. Â§ Elevations were obtained entering geogra phic coordinates in a digital elevation model based on Shuttle Radar Topography Mission (SRTM) Finished 3 arc second (90 m) raster elevation data set. Â¢ Provenances described as N. macrocarpa by Vazquez and Rodriguez (1999). MAT (mean annual temperature) an d MAP (mean annual precipitation) were obtained (1985)) and adjusted according to Luebert and Pliscoff (2006) and Amigo and Ramirez (1998).
107 Table 3 2 Genetic parameters fo r N othofagus obliqua and N. alpina obtained from measurements in all five blocks of four year old progeny provenance trials located in Fundo Ar quilhue, Los Rios Region, Chile Trait N. obliqua N. alpina h 2 0.29 h 2 0.35 h 2 f CG p h 2 0.29 h 2 0.35 h 2 f CG p HT 0.17 0.14 0.20 0.89 0.89 0.14 0.12 0.16 0.54 0.57 D IAM 0.30 0.25 0.32 0.88 0.82 0.07 0.06 0.09 0.46 0.63 R 2 H 0.22 0.18 0.24 0.85 0.82 0.08 0.06 0.09 0.45 0.62 FORK 0.15 0.13 0.18 0.00 0.00 0.00 0.00 0.00 0.20 1.00 STR 0.21 0.17 0.23 0.62 0.57 0 .10 0.08 0.11 0.64 0.74 h 2 0.29 and h 2 0.35 are the within provenance individual tree heritabilities using coefficients of relationship ( r ) = 0.29 and 0.35 respectively. h 2 f is the family heritability, CG p is the coefficient for provenance genetic gain, and is the proportion of the total morphological genetic variation that is due t o differences among provenances. Table 3 3 Genetic parameters for N othofagus obliqua and N. alpina obtained from measurements in block one of four year old progeny provenanc e trials located in Fundo Ar quilhue, Los Rios Region, Chile Trait N. obliqua N. alpina CG p CG pASR CG p CG pASR LAREA 0.49 0.51 0.36 0.50 LVEIN 0.47 0.48 0.26 0.40 DENS 0.43 0.46 0.28 0.24 SI 0.23 0.20 0.42 0.44 SL 0.06 0.33 0.02 0.25 S D 0.21 0.21 0.00 0.10 CG p and CG pASR are the coefficient for provenance genetic gain obtained from P ROC M IXED in SAS (SAS, 2008) and from the bivariate analysis in ASR EML (Gilmour et al. 2006) respectively. Table 3 4 Narrow sense genetic correlatio ns (r A ) among morphological traits in N othofagus obliqua and N. alpina obtained from measurements in four year old progeny provenance trials located in Fundo Ar quilhue, Los Rios Region, Chile LAREA LVEIN DENS SI SL SD N. obliqua HT 0.092 0.196 0.251 0.319 0.977 0.025 STR 0.447 0.344 0.046 0.978 0.845 0.202 N. alpina HT 0.452 0.107 0.321 0.457 0.781 0.407 STR 0.066 0.250 0.529 0.491 0.955 0.816
108 Table 3 5 Provenance least square means (LSM) for all morphological tr aits in N othofagus obliqua and N. alpina # Provenance SURV (%) HT (cm) DIAM (mm) R 2 H (L) FORK (%) STR (0 to 4) LAREA (cm 2 ) LVEIN (#) DENS (g/dm 2 ) SI SL (m) SD (#/mm 2 ) N. obliqua 1 Tiltil 80 200 26 0.565 25 0.9 11.1 8.9 3.98 1.33 26.2 306 2 Bellavista 80 197 31 0.467 24 0.3 5.9 7.8 3.88 1.21 26.8 236 3 Vilches 87 302 48 1.906 41 0.8 10.8 9.2 3.60 1.33 23.7 204 4 Los Ruiles 93 308 43 1.499 54 0.9 7.5 7.8 3.96 1.33 27.5 217 5 Quirihue 92 331 50 2.349 44 0.9 9.2 9.4 4.29 1.28 24.3 222 6 Bullileo Alto 93 319 51 2.523 65 0.6 11.6 9.1 4.38 1.33 23.7 178 7 Ninhue 90 327 50 2.469 47 1.0 10.3 8.8 4.13 1.33 24.3 239 8 Cayumanqui 92 347 53 2.817 54 0.8 10.8 9.0 4.18 1.34 23.6 235 9 Recinto 96 346 53 2.780 39 1.1 9.0 8.6 3.97 1.35 24.6 245 10 Reserva uble 98 270 40 1.329 36 1.3 11.5 9.0 4.38 1.37 23.9 218 11 Curanilahue 97 399 67 5.072 63 1.0 7.4 9.0 4.23 1.39 23.0 234 12 Santa Barbara 98 330 52 2.521 40 1.0 10.1 9.6 4.00 1.36 23.9 234 13 Lago Lanalhue 94 397 70 5.256 56 1.1 10.0 9.1 3.79 1 .35 22.9 246 14 Ralco 94 344 54 2.878 50 0.7 11.1 9.2 4.15 1.32 23.8 232 15 Victoria 100 321 45 1.914 38 1.2 9.7 8.6 3.83 1.36 24.3 206 16 Pichipillahuen 98 373 57 3.495 39 1.6 11.6 8.5 4.48 1.40 24.5 256 17 Galletue 97 265 41 1.483 48 1.0 11.0 8.2 4.4 0 1.55 25.2 213 18 Quepe 96 366 57 3.386 31 1.4 8.5 8.6 3.52 1.34 22.3 233 19 Cunco 98 351 52 2.614 47 1.1 11.5 9.2 3.96 1.35 23.4 244 20 Lago Colico 98 376 58 3.596 49 1.2 11.0 8.8 4.17 1.38 22.5 215 21 Lastarria 100 388 60 3.814 37 1.3 11.3 8.8 3.82 1.39 23.6 236 22 Curarrehue 100 392 62 4.117 50 1.1 15.1 8.7 3.78 1.37 23.2 211 23 Cruces 98 371 58 3.513 43 1.3 11.7 9.2 3.68 1.39 24.2 211 24 Malalhue 98 387 59 4.079 39 1.5 10.3 8.5 3.89 1.38 23.1 215 25 Choshuenco 98 387 58 3.805 30 1.7 12.2 9.6 4. 32 1.35 22.7 246 26 Futrono 100 385 59 3.780 39 1.5 11.8 9.5 3.72 1.39 23.4 205 27 Llifen 100 388 61 3.975 45 1.5 11.9 8.4 4.11 1.34 24.9 212 28 Llancacura 98 358 54 2.885 43 1.2 13.9 9.2 3.63 1.44 24.1 216 29 Rio Negro 100 331 49 2.211 31 1.5 11.4 8.5 4.13 1.37 24.3 228 30 Rupanco 96 359 51 2.771 41 1.6 12.6 9.1 4.14 1.38 23.1 240 31 Purranque 100 332 47 2.033 37 1.6 12.1 8.3 4.01 1.43 23.9 219 Species average 97 340 52 2.835 43 1.2 10.8 8.8 4.02 1.36 24.0 227
109 Table 3 5. C ontinued # Provenanc e SURV (%) HT (cm) DIAM (mm) R 2 H (L) FORK (%) STR (0 to 4) LAREA (cm 2 ) LVEIN (#) DENS (g/dm 2 ) SI SL (m) SD (#/mm 2 ) N. alpin a 32 Siete Tazas 96 324 49 2.451 46 2.9 14.3 15.0 4.78 1.32 26.4 191 33 Vilches 100 303 50 2.350 56 2.0 17.0 16.8 4.80 1.32 23.8 164 34 Bullileo Alto 86 343 51 2.907 45 2.9 14.3 14.6 4.81 1.31 27.6 170 35 Recinto 96 286 46 1.844 23 2.6 19.0 16.6 4.26 1.33 25.5 205 36 Nahuelbuta 83 318 56 3.039 56 2.1 16.1 14.8 4.92 1.26 23.8 195 37 Santa Barbara 97 364 56 3.574 4 6 2.7 16.6 15.8 4.02 1.27 24.7 203 38 Jauja 94 335 52 2.712 48 2.6 15.9 16.1 4.41 1.28 25.3 190 39 Pichipillahuen 92 343 56 3.111 52 2.4 14.4 16.4 4.81 1.27 27.2 175 40 Malalcahuello 93 268 46 1.662 21 2.9 15.1 15.4 3.70 1.24 25.2 213 41 Melipeuco 60 2 83 41 1.728 50 2.3 24.6 19.0 4.29 1.29 23.8 205 42 Curarrehue 94 322 51 2.598 36 2.4 15.4 15.8 4.19 1.29 25.9 183 43 Releco 98 330 53 2.908 49 2.6 16.9 15.9 4.05 1.28 26.3 179 44 Llancacura 98 299 45 1.865 45 2.4 14.1 15.4 4.39 1.29 24.6 177 45 Hueyusc a 97 311 49 2.371 42 2.4 10.6 14.5 4.45 1.25 26.0 184 Species average 94 316 50 2.509 44 2.5 16.0 15.9 4.42 1.28 25.4 188 Note : SURV =survival, HT =total height, DIAM =stem diameter at the root collar, R 2 H =volume index, FORK =stem forking, STR = stem straigh tness LAREA = leaf area, LVEIN =number of lateral veins on the leaf blade, DENS =leaf density, SI =shape index, SL = stomatal length, SD = stomatal density.
110 Table 3 6 Among provenance significant differences obtained for Nothofagus obliqua and N. alpina us ing provenances least square means and treating the effect of provenance as fixed in the models. Only the p values of statistically significant differences in morphological traits are shown ( =0.05) Species HT DIAM R 2 H STR LAREA LVEIN DENS SI SD N. obliq ua <0.0001 <0.0001 <0.0001 <0.0001 0.0004 0.0004 0.0030 0.0413 0.0389 N. alpina 0.0034 0.0136 0.0180 0.0001 0.0209 0.0275 Table 3 7 Standardized canonical coefficients for the two canonical variables morpho1 and morpho2 that describe the association between the morphological and canonical variables, and for env1 and env2 that describe the association between the environmental and canonical variables in N othofagus obliqua V ariable s Canonical variables Morphological morpho1 morpho2 HT 1.1765 1.0503 STR 0.1694 1.0309 FORK 0.1861 0.8470 DENS 0.0835 0.5572 SI 0.3238 0.4773 LVEIN 0.1532 0.0726 SD 0.5232 0.4767 SL 0.0469 0.2412 Environmental env1 env2 MAT 1.1077 0.1950 MAP 0.6804 0.8956
111 Table 3 8 Stan dardized canonical coefficients for the two canonical variables morpho1 and morpho2 that describe the association between the morphological and canonical variables, and for env1 and env2 that describe the association between the environmental and canonical variables in N othofagus alpina V ariable Canonical variables Morphological morpho1 morpho2 HT 0.2472 1.0051 STR 0.4603 0.8260 FORK 0.1171 1.1936 DENS 0.3664 1.5281 SI 0.0838 0.4607 LVEIN 0.4097 2.1560 SD 0.7864 0.2948 SL 0.7102 1. 4560 Environmental env1 env2 MAT 0.724 0.8315 MAP 1.0591 0.3065 Table 3 9 Stepwise discriminant analysis of environmentally homogeneous groups of provenances using provenance LSM from nine non multico linear morphological variables. Only total height ( HT ) and stomatal density ( SD ) in N othofagus obliqua and leaf density ( DENS ) in N. alpina were useful to discriminate groups of provenances Step Variable Partial R 2 p value ASCC N. obliqua 1 HT 0.272 0.0117 0.136 2 SD 0.341 0.0036 0.298 N. alpina 1 DENS 0.202 0.1065 0.202 average squared canonical correlation.
112 Table 3 10 Posterior p robabi lity of group membership f rom mis classified pr o venances of the canonical discriminant analysis of envir onmentally homogeneous groups in N oth ofagus obliqua and N. a lpina provenances ( Fig ure 3 2) # Provenance From group To group Cold and Rainy Warm and Dry Warm and Rainy N. obliqua 4 Los Ruiles Warm and Dry Cold and Rainy 0.62 0.38 0.00 8 Cayumanqui Warm and Dry Warm and Rainy 0.04 0.3 6 0.60 15 Victoria Warm and Dry Cold and Rainy 0.79 0.21 0.00 16 Pichipillahuen Warm and Dry Warm and Rainy 0.00 0.18 0.82 18 Quepe Warm and Dry Warm and Rainy 0.03 0.15 0.82 22 Curarrehue Cold and Rainy Warm and Rainy 0.24 0.17 0.59 26 Futrono Cold a nd Rainy Warm and Rainy 0.39 0.19 0.42 28 Llancacura Warm and Dry Warm and Rainy 0.26 0.24 0.50 30 Rupanco Warm and Dry Warm and Rainy 0.01 0.20 0.78 31 Purranque Warm and Dry Cold and Rainy 0.47 0.43 0.09 N. alpina 34 Bullileo Alto Cold and Rain y Warm and Dry 0.22 0.78 35 Recinto Warm and Dry Cold and Rainy 0.58 0.42 37 Santa Barbara Warm and Dry Cold and Rainy 0.73 0.27 Table 3 11 Cross validation summary and error rate estimates of the canonical discriminant function of environment ally homogeneous groups in N othofagus obliqua provenances ( Figure 3 2a). The diagonal shows number of provenances and percent correctly classified. Off diagonal numbers are the misclassified provenances. The discriminant analysis was done using total heigh t ( HT ) and stomatal density ( SD ). Predicted g roup Actual g roup Cold and R ainy Warm and Dry Warm and Rainy Total Cold and R ainy 3 1 2 6 50% 17% 33% 100% Warm and Dry 4 5 5 14 29% 36% 36% 100% Warm and Rainy 0 2 9 11 0% 18% 82 % 100% Total 7 8 16 31 23% 26% 52% 100% Error rates 50% 64% 18% 44% Priors 33% 33% 33%
113 Table 3 12 Cross validation summary and error rate estimates of the canonical discriminant function of environmentally homogeneous groups i n N othofagus alpina provenances ( Figure 3 2b). The diagonal shows number of provenances and percent correctly classified. Off diagonal numbers are the misclassified provenances. The discriminant analysis was done using leaf density ( DENS ). Predicted g roup Actual g roup Cold and R ainy Warm and D ry Total Cold and R ainy 4 1 5 80% 20% 100% Warm and Dry 2 7 9 22% 78% 100% Total 6 8 14 43% 57% 100% Error rates 20% 22% 21% Priors 50% 50%
114 Fig ure 3 1 Range of distribu tion for a) N othofagus obliqua and b) N. alpina in grey, showing the location of the sampled provenances used in my study and the progeny provenance trial.
115 Fig ure 3 2 Grouping of provenances of a) N othofagus obliqua and b) N. alpina according to a n on hierarchical cluster analysis done using environmental variables MAT (mean annual temperature) and MAP (mean annual precipitation). Close to each point is the number ID (#) for the provenances ( Table 3 1). Dashed lines delimit groups of environmentally homogeneous provenances.
116 Fig ure 3 3 Significant differences among geographic origins of N othofagus obliqua provenances obtained using contrasts in total height ( HT ), stem diameter at the root collar ( DIAM ), volume index ( R 2 H ), stem straightness ( STR ), leaf area ( LAREA ), leaf shape index ( SI ), and stomatal length ( SL ). Dashed horizontal lines represent the mean values for each geographic origin. Different letters in the top left corner of each geographic origin indicate statistically significant differen ces ( =0.05). The p values after Bonferroni correction for multiple comparisons were p <0.0001 for HT DIAM R 2 H and STR p =0.018 for LAREA p =0.003 for SI and p =0.045 for SL
117 Fig ure 3 4 Significant differences among groups of provenances of similar e nvironmental characteristics in N othofagus obliqua obtained using contrasts in total height ( HT ), stem diameter at the root collar ( DIAM ), volume index ( R 2 H ), stomatal length ( SL ), and stomatal density ( SD ). Dashed horizontal lines represent the mean value s for each provenance group. Different letters in the top left corner of each group indicate statistically significant differences ( =0.05). p values after Bonferroni correction for multiple comparisons were p <0.0001 for HT DIAM and R 2 H p =0.026 for SL and p =0.006 for SD
118 Fig ure 3 5 Significant differences among geographic origins of N othofagus alpina provenances obtained using contrasts in leaf shape index ( SI ). Dashed horizontal lines represent the mean values for each geographic origin. Different letters in the top left corner of each geographic origin indicate statistically significant differences ( =0.05). The p value after Bonferroni correction for multiple comparisons was p =0.013 for the difference between Mediterranean Forest and Temperate Rai nforest.
119 Fig ure 3 6 Pearson correlations ( r ) among morphological and environmental variables for N othofagus obliqua The straight lines show the tendency of the relationships and p are p values obtained after Bonferroni correction for multiple comparis ons. HT = total height, SD = stomatal density, MAP = mean annual precipitation, MAT = mean annual temperature, and DMAT = absolute difference in MAT between the trial site and the provenance origin. Scatter plots a) and c) are at the provenance level (n=31 ). Scatter plots b), d), e), and f) are at the family level (n=247).
120 Fig ure 3 7 Pearson correlations ( r ) among morphological and environmental variables for N othofagus alpina The straight lines show the tendency of the relationships and p are p values obtained after Bonferroni correction for multiple comparisons. DENS = leaf density, and MAP = mean annual precipitation. Scatter plot a) is at the provenance level (n=14) and scatter plot b) is at the family level (n=104).
121 Fig ure 3 8 Canonical correl ations ( r ) of a) the first canonical variables morpho1 vs. env1 ( r =0.86) and b) the second canonical variables morpho2 vs. env2 ( r =0.61) obtained for N othofagus obliqua
122 Fig ure 3 9 Canonical correlations ( r ) of a) the first canonical variables morpho1 vs. env1 ( r =0.96) and b) the second canonical variables morpho2 vs. env2 ( r =0.91) obtained for N othofagus alpina
123 Fig ure 3 10 variance dendrograms indicating closeness in adaptive traits among provenances in a) N othofagus obliqua and b) N.alpina The numbers below the dendrograms are semi partial R 2 values, which are a measure of distance between branche s.
124 CHAPTER 4 SYSTEMATICS, CONSERV ATION GENETICS, AND BREEDING ZONES DEFIN ITION BASED ON NEUTRAL AND ADAPTIVE PATTERNS OF GENETIC VA RIATION IN NOTHOFAGUS OBLIQUA N. ALPINA AND N. GLAUCA Introduct o ry Remarks A sound approach to making decisions about the conservation and sustainable use of a group of closely related species should include resolving the systematic relationship among th ose species, assessing the potential and extent of hybridization, and estimating their within species variation structure considering both neutral and adaptive genetic variation. The genetic variation within and among species is often analyzed using neutra l molecular markers such as isozymes or an assortment of DNA markers. The use of these techniques is relatively inexpensive and fast (McKay and Latta 2002 ; van Tienderen et al. 2002). Unfortunately, variability obtained from the analysis of neutral marke rs has been shown to be uncorrelated with adaptive variability obtained from the analysis of morphological markers (Reed and Frankham 2 001 ; McKay and Latta 2 002). Adaptive genetic variation is a very important component for conservation, because whereas neutral variation determines the underlying potential for longer term evolutionary changes, adaptive variation determines the evolutionary potential to respond to more immediate changes (McKay and Latta 2 002). Thus, the combined analysis of neutral and ad aptive markers is important for the development of conservation and genetic improvement strategies of forest trees. Nothofagus obliqua (Mirb.) Oerst., N. alpina (Poepp. et Endl.) Oerst. (= N. nervosa ), and N. glauca (Phil.) Krasser. make up the South Ameri can clade of subgenus Lophozonia (Manos, 1997). These three species are sympatric, endemic, and
125 closely related, but the systematic relationships within the clade are not clear. There is a combined consensus tree recovered by Manos (1997) showing a closer relationship between N. alpina and N. glauca leaving N. obliqua as the sister to their branch. However, the moderate support for that topology (bootstrap=77%), the current geographic range of the species (Figure 4 1 ), and the evidence that hybridization o ccurs between N. obliqua and both N. alpina and N. glauca but not between the latter two ( Donoso, 19 79; Donoso et al. 1990 ) suggests that N. alpina and N. glauca might not be the most genetically similar species with in the clade. In addition, Vazquez and Rodriguez (1999) proposed a new species for the subgenus ( N. macrocarpa (A.DC.) F.M. Vazquez & R.A. Rodr., formerly N. obliqua var. macrocarpa A.DC.) and recognized three subspecies of N. obliqua (ssp. obliqua (Mirb.) Oerst., spp. valdiviana (Phil.) F.M. Vazquez & R.A. Rodr., and ssp. andina F.M. Vazquez & R.A. Rodr.), adding a new element to the discussion. The conservation status of these species overall is not critical. Nothofagus obliqua and N. alpina have been classified as out of danger or near threa tened and N. glauca as vulnerable by Benoit (1989) and Gonzalez (1998a; 1998b). However, threats to particular populations mainly in the northern area of distribution 363 but also in other localities across their range, are serious and may po tentially lead to local extinction. Besides, the populations of these species are poorly represented in the Chilean National Wild Lands System (SNASPE). Overall, less than 6%, 8%, and 2% of N. obliqua N. alpina and N. glauca respectively, are within pro tected areas (Ormazabal and Benoit, 1987).
126 The species of Lophozonia in South America are greatly valued because of their high quality wood and fast growth. Therefore, it is important to incorporate the use of their wood in the conservation equation. T here is evidence that their populations have steadily degraded because of their commercial value and the most desirable phenotypes are gone (Vergara and Bohle 2 000) Thus, br eeding programs leading to genetic improvement of these tree species are needed to a ttempt the recover ing of genotypes that would maximize growth and wood quality and also general fitness to the current and future environment s. A tree improvement strategy should take in to account, among many elements, the genetic structure and variability of populations to assure the conservation of within population variability as well as regional genetic identities The tree improvement strategy proposed for the populations of N. obliqua and N. alpina growing in Chile (Ipinza and Gutierrez, 2000) uses a s general guidance regions of provenance defined by Vergara ( 2000). However, this strategy only uses the regions of provenance to prioritize the regions where the selection and genetic tests should be done. It does not explicitly use the regions of proven ance to organize the genetic material in breeding zones mainly because of the lack, at the time, of neutral or adaptive genetic information needed to characterize such regions. The objectives of my study are to : 1) elucidate the systematics of the South Am erican clade of subgenus Lophozonia including their inter and infra specific genetic similarities by using high resolution nuclear microsat ellite data for populations of the three species (Chapter 2) ; 2) identify conservation priorities among the studied p opulations and generate rankings that will take into account uniqueness and complementarity of populations within each species by combining nuclear microsatellite
127 data (Chapter 2), morphological data obtained from common gardens (Chapter 3), and maternally inherited chloroplast DNA data (Azpilicueta et al. 2009; Marchelli and Gallo, 2006) ; a nd 3) using the latter three sets of information, evaluate the relevance of defining breeding zones in the t ree improvement strategy for the species belonging to Nothof agus subgenus Lophozonia Material and Methods Sources of Data Used for the Analyses The data I used in the se analyses came from three sources that obtained genetic information from the same seed lots in most of the population s (Table 4 1) and in some case s, the same individuals. The first source included data from seven high resolution nuclear microsatellite loci (the nMSAT data set) obtained for 20 populations of N othofagus obliqua 12 populations of N. alpina and 8 populations of N. glauca (Chapter 2). The second source consisted of morphological data from 14 adaptive traits measured in common gardens ( the ADAP data set ) with a randomized complete block design, representing open pollinated families from 31 N. obliqua and 14 N. alpina po pulation s (Chapter 3). T he third source comprised PCR RFLP fragment data from maternally inherited chloroplast DNA regions (the cpDNA data set) obtained from 27 N. obliqua (Azpilicueta et al., 2009) and 26 N. alpina (Marchelli and Gallo, 2006) populations. All three data so urces covered most of the species ranges of distribution in Chile (Figure 4 1) Analysis of Genetic Similarities among Species Bayesian analysis To examine the geneti c similarities among species, I used an individual based approach assuming no specific mut ation model using Bayesian clustering in S TRUCTURE
128 2.3.2 (Pritchard et al. 2 000). Using the nMSAT data set, I analyzed all the populations of the three species together with a total of 640 individuals from 40 populations I used the admixture model with c orrelated allele frequencies and without including population information in the analyses as recommended by Falush et al. (2003). I ran S TRUCTURE at multiple K values ( K =number of assumed clusters in the data) from K =1 to K =8. At each K I carried out 10 s eparate runs of 200,000 generations each with a burn in of 100 000 I obtained posterior probabilities ( LnP[D] ) and computed K (Evanno et al. 2 005) to select the optimal K when LnP[D] turns asymptotic. I analyzed this data set with the seven original nMSAT loci and also by eliminating locus ncutas04 because it was only polymorphic in N. glauca (Chapter 2) and could bi as the results. I employed D ISTRUCT 1. 1 (Rosenberg 2 004) to visualize and edit the S TRUCTURE outputs. Pairwise genetic distances Another way t o examine the gen etic similarity among species is by obtaining pairwise genetic distances among the populations o f all three species simultaneously. Again u sing the nMSAT data set and assuming a stepwise mutation model ( SMM, Slatkin 1 995) I ran A RLEQUIN 3.11 (Excoffier et al. 2 005) to obtain a pairwise R ST matrix with 100 permutations to obtain significance, usin g standard Bonferroni corrections on all pairwise differences (Rice 1 989). As in the Bayesian analysis, I ran A RLEQUIN with and without locus ncutas04 to compare results. Pairwise R ST are measures of genetic differentiation among populations and were used to obtain unrooted neighbor joining (NJ) trees in P HYLIP 3.69 (Felsenstein 1 989) employing the component N EIGHBOR to infer species similarities within the clade I visualized the resulting trees with T REE V IEW 1.6.6 (Page 1 996 ). Additionally, S TRUCTURE 2.3.2 (Pritchard et al. 2 000) also delivers N EIGHBOR /P HYLIP unrooted neighbo r joining trees
129 from allele frequency divergence values assuming an infinite allele model ( IAM ) for K pairs of S TRUCTURE clusters. I u sed these trees to compare results among all methods. Methods of Ranking Populations for Conservation Rankings based on allelic richness Allelic richness is a diversity parameter of high priority in conservation genetics and is most useful when applied to highly variable markers like microsatellite D NA (Petit et al. 1998). I used the approach of Petit et al. (1998) to obtain the relative conservation value for each popula tion within each species based o n the marginal allelic richness contributed b y the populations to the total allelic richness of the species Thus, the populations w ere ranked according to their contribution to the total allelic richness ( C T ) using: C T ( k ) = ( r T r T \ k ) / ( r T 1) (4 1) where C T ( k ) = contribution of the k th population to the total allelic richness, r T = estimator of the total allelic richness, and r T \ k = estimator of the total allelic richness when the k th population is excluded. I obtained allelic richness estimations from the nMSAT and cpDNA data sets. To input allelic richness values in the equation, I standardize d them using rarefaction to correct for uneven sample size (Kalinowski, 2004) using HP R ARE 1.1 (Kalinowski, 2005). Rankings based on dendrograms Other ranking options based on conservation value are given by systematics with the assessment of taxonomic di stinctness (Vane Wright et al. 1991 ; Faith, 1994) The various systematics approaches developed to obtain conservation values are intended
130 to give greater conservation weight to taxa that are unique and although they were designed to define conservation p riorities among different species and groups of species, I applied them to ra nk populations within species based on the nMSAT and ADAP dendro grams obtained in Chapter 2 and Chapter 3 respectively I utilized the root weighting procedure developed by Vane Wright et al. (1991) including May ( May, 1990) to prioritize populations for conservation based on topology and without directly considering genetic distances among populations. For each populat ion in a dendro gram I traced all th e nodes to the den d r ogram counting all the branches at each node. The populations were then ranked obtaining the percentage contribution f or each population to the total diversity ( P T ) with: P T ( k ) = 100[( b T / b k b T / b k )] (4 2) where P T ( k ) = percentage contribution of the k th population to the total diversity, b T = total branch count adding all populations, and b k = total branch count for the k th population. Results Among Species Genetic Similarity Infer red from Bayesian Clustering The analysis in S TRUCTURE grouping all forty populations from the three species together produced log posterior probabilities ( LnP[D] ) for K =1 to K =8. Figure 4 2 shows the curves of LnP[D] K (Evanno et al 2005) defining a peak to find the optimal number of clusters ( K K value s in K =3 obtained for both analyses th at with all seven loci and th at with the six polymorphic loci indicated a
131 strong signal to iden tify three clusters coinciding with the spec ies boundaries defined in my study (Figure 4 3 a and b ). I also show the results for K =2 and K =4 because, although suboptimal, they give an idea of the among species similarities ( K =2) and of putative within species different iation ( K =4) in this South Ameri can clade. In this case, results are quite different between the two analyses. For K =2 and using the seven loci (Figure 4 3a), N othofagus obliqua clearly clustered with N. alpina leaving N. glauca in its own cluster. However, using only the six polymorphic loci (Figure 4 3b), N. obliqua clustered with N. glauca instead. Also, for K =4 and using the seven loci (Figure 4 3a), N. obliqua was subdivided in to two groups with populations from the Mediterranean Forest ( 36 2 ) separated from the Transition al Forests and Termperate R ainforests populations south of 36 2 (Figure 4 1). When using only the six polymorphic loci (Figure 4 3b) the limit between the two groups in N. obliqua change d and now most of the populations from the Mediterranean and Trans itional Forests 38 2 ) cluster toge ther and apart from the Temperate Rainforests populations (Figure 4 1). Am ong Species Genetic Similarities Inferred from Pairwise Genetic Distances I obtained two unrooted NJ t rees using pairwise R ST matrices fo r all pairs of populations combined o ne using all seven available loci (Figure 4 4a) and another using the six polymorphic loci (Figure 4 4b). Both tre es show three clearly defined groups matching the t hree species analyzed Almost all populat ions fell we ll within each group with the sole exception of Loncha (5) from N. obliqua which seems to be apart from every population with a slight tendency to the N. glauca branch These two trees do not show a trend of similarity among species or of geographic withi n species differen t iation. Coastal populations Pichipillahuen (26) and Cruces (36) from N. obliqua
132 cluster very close to each other and apart from all the other populations in the species. However, this similarity does not make sense geographically (Figure 4 1). As another way to observe among species similaritie s a nd possible geographic within species differen t iation I obtained four unrooted NJ trees assuming the infinite allele model (IAM) among K =3 and K =4 pairs of population clusters f rom the S TRUCTURE analyses. For each K I obtained one tree using all seven available loci (Figure 4 4c) and another using the six polymorphic loci (Figure 4 4d). Results sh ow a greater similatiry of N. obliqua with N. alpina rather than with N. glauca and little evid ence of within species different iation in N. obliqua across all four dendrograms. Richness I obtained the contribution to total allelic richness ( C T ) for nuclear markers (nMSAT data set) in 20 N. obliqua 12 N. alpina and 8 N. glauca populations (Figure 4 5) and for chloroplast markers (cpDNA data set) in 27 N. obliqua and 26 N. alpina popu lations. O f these, 11 N. obliqua and 8 N. alpina populations are common to both data sets or are di fferent samples of the same population (Table 4 1). Table 4 1 shows the within species conservation rankings (rnk C T ) for the contribution to C T in both data sets and highlights approximately the top 25% populations for each species/marker combination. O f the 19 po pulations common to both data sets eight belonged to the top 25% in just one marker and only Nahuelbuta (54) and Pichipillahuen (57) in N. alpina were ranked top 25% for both sets of markers. Ranking for Conservation Using Population Genetic Dist inctness Using dendrograms based on among population genetic similaritie s I calculated the percentage contribution to total diversity ( P T ) of each population based on
133 population distinctness for nuclear markers (nMSAT data set) in 20 N. obliqua 12 N. alp ina and 8 N. glauca populations (Figure 4 5) and for adaptive markers (ADAP data set) in 31 N. obliqua and 14 N. alpina populations (Figure 4 6) In this case 18 N. obliqua and 11 N. alpina populations are common for both data sets or are different samp les of the same population (Table 4 1). Figures 4 5 and 4 6 show the P T values and their respective within species conservation rankings in which it is important to note that the approximately top 25% ranked populations account for about 50 to 60% of the t otal diversity in each species/marker combination. Also Table 4 1 shows the within species conservation rankings (rnk P T ) based on P T values in both data sets and highlights approximately the top 25% populations for each s pecies/marker combination. O f the 29 populations common to both markers, eight belonged to the top 25% in just one marker and only La Campana/Tiltil (1/2), Sierras de Bellavista/Bellavista (7/8), and Galletue (27) in N. obliqua were ranked top 25% for both markers. Discussion Genetic Simi larities among Species The two maj or questions about the systematic relationship s in the South American clade of subgenus Lophozonia relate to the topology among the three species and to the subdivision of N othofagus obliqua in two sister species and other subspecies. Manos (1997) with his combined consensus tree obtained analyzing ITS sequences, plastid sequences, and morphology placed N. alpina and N. glauca in a clade and N. obliqua as the ir sister species. My results analyzing seven highly variable n uclear microsatellite loci did not agre e with this topology and favored gre a ter similarity between N. obliqua and N. alpina with N. glauca more genetically distant in both the S TRUCTURE analysi s (Figure 4 3a) and the neighbo r joining (NJ) trees obtained
134 a ssuming the infinite allele model (IAM) (Figure 4 4c). One locus was polymorphic only in N. glauca with N. obliqua and N. alpina sharing a common and unique allele perhaps arising from a generalized failure in the P CR reactions for that locus in the latte r two species These results could therefore indicate a closer but artifactual, relationship between N. obliqua and N. alpina than between either and N. glauca However, the analysis excluding that locus did not change the results in the IAM/NJ trees (Fig ure 4 4d) and although it changed the results in the S TRUCTURE analysis, it grouped N. obliqua and N. glauca together (Figure 4 3b) still disagreeing with the topology obtained by Manos (1997). Thus, genetic similarities of microsatellite loci clearly co nflict with the results of phylogenetic analysis. With regard to variation patterns within N. obliqua a study based on morphological features (leaf and cupule characteristics) and directed to herbarium specimens preserved worldwide (Vazquez and Rodriguez 1999) argues that the northern populations of N. obliqua formerly N. obliqua var. macrocarpa should be regarded as a new species they name N. macrocarpa Additionally, they propose three subspecies of N. obliqua based on differences in their cupulae and leaf pubescence. The genetic evidence from microsatellite loci does not support the definition of a new species in the northern part of the distribution of N. obliqua and furthermore does not seem to match the classification of three subspecies for N. obl iqua Both S TRUCTURE analyses show with strong signal, that the optimal number of clusters K =3 (Figure 4 2) correspond s to the three original species in th e clade (Figure 4 3). When K =4, N. obliqua split in two subgroups ; however, the groups composition s differed depending on the number of loci used. With all seven loci the Mediterranean Forest group extends
135 to matching fairly well the geographic distribution of the former N. obliqua var. macrocarpa which ranges and above 800 m a.s.l. (Ormazabal and Benoit, 1987) but contains Bullileo Alto (14), a population never included within var. macrocarpa Conversely, when using only the six polymorphic loci, t he northern group extended farther south, to (Mediterranean and Transitional Forests) and excluded Ninhue (16 ) and Recinto (17), which fit better with the Temperate Rainforest group. A S TRUCTURE analysis r a n only for N. obliqua populations (Chapter 2) found an optimal K =3 for the species, with geographically well defin ed groups and evidence of gene flow among groups. Finally, both unrooted NJ trees obtained using pairwise R ST values (Figure 4 4a and b) show that the populations defined as N. macrocarpa by Vazquez and Rodriguez (1999) do not form a monophyletic group apa rt from nor within N. obliqua Despite the conspicuous morphological differences among N. obliqua populations used by Vazquez and Rodriguez (1999) to propose N. macrocarpa as a new species and three subspecies within N. obliqua the genetic evidence from m icrosatellite loci d o not support this taxo nomic arrangement. My data seem to agree more closely with the notion of a north to south cline as proposed for N. obliqua by Donoso (1979a) where the northern populations are probably just an extreme of t he spec ies cline. However, from a genetic point of view and considering gene flow, it is possible to observe two breaks at and suggesting t he presence of three distinct groups within the species. These groups do not seem to match the proposed su bspecies for N. obliqua bu t these proposed subspecies are not geographically explicit and the morphological features used by Vazquez and Rodriguez (1999) were not recorded in
136 my study A powerful test in the future would be to obtain these morphological fe atures from populations described as each of the subspecies and compare t he characteristics with those from the same populations growing in common garden s and thus to discriminate ada ptation from phenotypic plasticity. The genetic differences observed in t his study for N. obliqua may represent an early stage of speciation that could yield different species in the future if the isolation among groups become larger and permanent However, chances are that this could take millions of years provided that the la st dated speciation event in the subgenus took place between 9 and 21 mya (Knapp et al., 2005). Identification of Conservation Priorities Evolutionarily significant units One way to define important populations for conservation is via the concept of evolut ionar ily significant units (ESU s ) identified considering historical population structure rather than current adaptation (Moritz, 1994). Thus, topologies defined employing neutral markers like microsatellites are a reasonable approach to attempt to define E SU s in this study. ESU s are used to make decisions in c onservation, where priorities go to clades of populations that are reciprocally monophyletic and therefore are independently evolving (Moritz, 1994). The discussion about the s ystematic relationships a mong the s pecies and possible subspecies in the South American Lophozonia clade suggests that N. alpina and N. glauca contain only one ESU each. On the other hand, N. obliqua is a more structured species with higher levels of genetic variation arranged geo graphically. Indeed, there are three distinctive groups of populations detected in N. obliqua (Chapter 2): t he populations within the Mediterranean Forest ( 3 6 2 ), the populations growing
137 ), and the popul ations growing within the Temperate Rainforests ( 41 ). These groups do not seem to be sufficiently different to represent ESUs ; n evertheless, conservation geneticists could use them as guidance for conservation efforts making sure, for example, that all three groups a re represented in a reserve network. Conservation priorities based on population conservation values In addition to these general units of conservation, each analyzed population has one to four values for conservation obtained from three different data sets. Table 4 1 shows the rankings where low numbers mean high conservation value and therefore high conservation priority. Each of the four rankings emphasizes a different aspect of genetic variation. While neutral data sets nMSAT and cpDNA represent the underlying potential for long term evolutionary changes, the adaptive data set ADAPT emphasizes the evolutionary potential to respond to more immediate changes like global warming. For this reason I believe that conservation prioritie s should first take in to account the adaptive variation in each species and afterwards include the neutral variation data. To help preserve the specific short term a daptive capacity I arbitrarily chose to target approximately the top 25% of the N. obliqua and N. alpina populations in the P T ranking from the ADAPT data set (Table 4 1). For N. obliqua the top populations wer e well distributed across branches in the dendro gram (Figure 4 6 a ) and across the species range (Figure 4 1 a ) indicating that an impor tant part of the adaptive variation is due to variation of environmental conditions. For N. alpina the top populations for conservation clustered in two small branches of the dendro gram (Figure 4 6b) characterized by slower growth and shorter stomata. For thermore, t he southern part of
138 (Figure 4 1b) Here, the method failed to captur e part of the adapt ive variability of the species excluding all populations th at ar e fast growing and have long stomata Once the short t erm adaptive capacity is protected conservation geneticists can complement the conservation efforts by protecting populations with high potential for long term evolutionary changes. For that, I used the same criteria as for the ADAPT data set and chose the top 25% of the N. obliqua N. alpina and N. glauca populations in the C T and P T rankings from the nMSAT data set and of the N. obliqua and N. alpina populations in the C T ranking from the cpDNA data set (T able 4 1). Table 4 2 shows the proposed top populations in conservation value indicating the overall ranking and the partial ranking values for all four measurements. For N. obliqua of the 18 populations ranked in the top 25% for at least on e conservati on value I selected nine that were top ranked in adaptive value and complemented the list with five populations of high neutral value plus Quila Quina (41), which was the only population from the East Andes in Argentina (Figure 4 1a) and which had two rar e cpDNA haplotypes (Table 4 2). Likewise, from the 12 top ranked populations in N. alpina I selected four of high adaptive value and complemented the list with four populations of high neutral value plus Tregualemu (51), a very isolated population (Figure 4 1b) with potentially high adaptive value. The selections included population s from all main branches in the dendrograms capturing most of the adaptive variability (Figures 4 5 and 4 6) as well as 93% and 100% of the cpDNA haplotypes in N. obliqua and N alpina respectively. In the case of N. glauca there are no adaptive data or data from cpDNA to build an overall ranking using the same approach Thus, I used the nMSAT data set alone to detect the top conservation priorities adding Los
139 Ruiles (81) to have a better representation in all maj or branch es of the nMSAT dendr ogram (Figure 4 5c) and across the species geographic range (Figure 4 1c). Methodological considerations I based my overall conservation rankings chiefly o n what I c onsider adaptive cons ervation value. It was intended to capture a set of unique adaptive traits that could enhance the capacity of each species to adapt to rapid environmental changes. I included several morphological traits that were apparently non adaptive, i.e. stem forking stem straightness leaf shape index, and number of lateral veins on the leaf blade (Chapter 3), following the rationale that if there are significant genetic differences in these traits among populations, those differences could have an unknown adaptive or preadaptive value. Also, the method I used to ra nk populations based o n adaptive traits dendrograms appears uneven because it favors populations in small branches with unique trait combinations disregarding sets of traits that are common to several pop ula tions (Figure 4 6). However, assuming that the sampling was systematic across the ranges of distribution, the method takes in to account that if a set of traits is shared by a large group of populations, the probability of extinction of that trait combin ation decreases due to metapopulation dynamics and therefore the urge ncy of conservation is reduced Still the probability that some un represented areas like the East Andes in Argentina contain unique trait combinations is high. In those cases, I made sur e that at least one population was selected as a priority for conservation (Table 4 2). General recommendations Based on the conservation status of N. obliqua N. alpina and N. glauca in their natural range s (Benoit, 1989; Gonzalez, 1998a; 1998b), their g enetic similarities and
140 the genetic conservation value for each population as obtained here N. glauca should have t op priority in conservation. N othofagus glauca is the most distinctive species in the clade (Figure 4 4c and d), it is a vulnerable species across its range and it has only 2% representation in protected areas (Ormazabal and Benoit, 1987), the lowest within the clade. Among the populations with high conservation value, Loncha (78) and Los Ruiles (81) are at least partially protected. Therefo re the t op conservation priorit y should be Qu illeco (85) and Alico (84). The protec t i on of Quilleco (85) is critical because it is a small and isolated population growing close to a village and surrounded by agricultural land and forest plantations (Le Qu esne and Sandoval, 2001) and has a high probability of local extinction. A second priority should be to complement the protected areas of Loncha (78) and Los Ruiles (81) with the protection of other stands in the surrounding areas It is important to aim f or the conservation of several stands adjacent to the sampled point of a population when they are available. The conservation of single, isolated stands tend s to be fragile because of the possibility of unexpected catastrophes like wild fires or diseases. In N. obliqua several populations with high conservation value are growing within n ational parks or reserves. O f the 15 populations selected, Galletue (27), La Campana (1), Los Ruiles (12), Reserva uble (20), Loncha (5), and Quila Quina (41) are growing within protected areas. There is, however, a group of selected populations mainly from the Mediterranean Forests that are locally en dangered and deserve special att ention and the top priority. These are Bullileo Alto (14), Bellavista (8), Curanilahue (21), Rupanco (46), Alto Colorado (6), Cruces (36), and Lampa (3). Particularly threatened and isol ated is Alto Colorado (6). S econd ary priority should go to Choshuenco (38) and
141 Curarrehue (33) because they are not seriously threatened and are surrounded by pro tected areas in which the species is present T he third priority would be to complement the protected areas of La Campana (1), Los Ruiles (12), Res erva uble (20), and Loncha (5) with the protection of other stands in the surrounding area s. Populations Gal letue ( 27) and Quila Quina (41) appear to be adequately protected I proposed for N. obliqua three g roups of populations that, although not ESU s, are distinct worth y of separate protect ion and therefore, should be well represented in the proposed conserv ation priorities. The groups growing in the Mediterranean Forests and the Temperate Rainforests have good representation, but the group corresponding to the Transitional Forests is under represented with only two populations that are priorit y for conservat ion. Here I suggest the in clusion of two additional populations from areas that were not sampled in this study such as near the city of Concepcion and the eastern slopes of the Nahuelbuta mountain range. Finally, in N. alpina there also are populations w ith high conservation value that are placed within national parks or reserves: the top ranked Nahuelbuta (54), and also Vilches (50) and Hua Hum (71). Thus, the t op priority for conservation fall s on the vulnerable populations Recinto (53), Melipeuco (59), Pichipillahuen (57), and Bullileo Alto (52). The second priority goes to Neltume (65) and Tregualem u (51). Neltume (65) is not seriously threatened and is close to many other protected populations and Tregualemu (51), despite being an endangered, very r a re, and isolated population, is not genetically unique base on the loci and traits scored in this study As a third priority, onl y Nahuelbuta (54) requires conservation of surrounding st ands due to its
142 vulnerability. Populations Vilches (50) and Hua Hum ( 71) are probably already adequately protected. Definition of Breeding Zones for Tree Improvement Strategies One critical step in the elaboration of a tree improvement strategy especially for native species is the definition of breeding zones E ach breedi ng zone will have its own improvement program influencing the costs and flexibility of the tree improvement operations (White et al., 2007). There are two main reasons for having separa te breeding zones. The first is to mantain the genetic identity of loca l populations to conserve their original allele combinations intact, and the second reason is to maintain genotypes adapted to general climatically homogeneous regions in their zone to avoid maladaptation. The genetic evidence gathered in this s tudy indica tes that there is no reason to subdivide the populations of either N. glauca or N. alpina in to different breeding zones ; each of these species should be regarded as i t s own unique zone. N othofagus glauca grow mainly in a climatically homogeneous region cha racterized by the presence of the Medi terranean Forests (Figure 4 1) with extensive pollen flow and therefore not much genetic structure (Chapter 2). On the other hand, N. alpina spans three different climatic regions and presents higher among population g enetic variation (Chapter 2). However, this heterogeneity does not prevent pollen flow among populations, does not follow a defined geographic pattern and therefore does not make sense for defining breeding zones. Furthermore the adaptive variability mea sured for N. alpina indicates that specific combinations of adaptive traits are not distributed following maj or geographic areas (Chapter 3).
143 The case of N. obliqua is different because the genetic evidence indicates that there are three groups of populati ons that could be regarded as different breeding zones. Like N. alpina N. obliqua spans three different climatic regions and presents among population levels of genetic variation in between those of N. alpina and N. glauca (Chapter 2). The di stinction is that this different iation shows a geographic pattern that matches fairly well the north to south climatic variation and follows the general definition of Mediterranean Forests, Transitional Forest, and Temperate Rainforests Also, the AMOVA in Chapter 2 re veals that the extensive gene flow among populations is not the rule when measured among groups supporting the idea of using three different breeding zones in a tree improvement strategy. Nevertheless, defining two breeding zones may be more appropriate f or N. obliqua Observing the among group differences in adaptive traits, the Transitional Forests group always grouped with either the Mediterranean Forests or the Temperate Rainforests and was never significant ly different from both of them at the same ti me. In all growth traits and in stomatal length the Transitional Forests group is similar to the Temperate Rainforest group indicating similar overall adaptation to southern climates (Chapter 3). Thus, my recommendation is to define two breeding zones wi th separate tree improvement programs in N. obliqua : b growing within the Mediterranean Forests boundaries ( 3 6 2 ), and breeding 41 ). The c urre nt tree improvement strategy developed for N. obliqua and N. alpina (Ipinza and Gutierrez, 2000) do not propose within species breeding zones. However, the priorities stated in the strategy and the implementation of the breeding program for
144 N. obliqua do n ot include selections outside the Temperate Ra inforests and therefore do not conflict with the zoning stated in my study. The breeding zones definition in N. obliqua does not coincide with the maj or macroclimates identified for central Chile ( Fuenzalida, 1 965) but coincides well with the boundaries of the regions of provenance proposed by Vergara (2000 ). Thus, breeding zone North is comprise d of regions of provenance 1 C, 2 C, and 8 A, while breedin g zone South contains all other regions of provenance The translocation of seeds among provenances and among regions of provenances within each bree ding zone should be do ne with caution. P ollen flow from translocated trees into a new population should not be a problem because it would be similar to the extensive pollen flow that seems to occur naturally within each breeding zone, and therefore the effects on genetic identity should be limited. However, seed production from the translocated trees will certainly alter the chloroplast composition in populations of u nique haplotypes like th ose found for N. obliqua ( Azpilicueta et al., 2009 ) and N. alpina ( Marchelli and Gallo, 2006 ) in ce rtain coastal populations. Gene flow via seeds in Nothofagus is much slower than pollen flow and in isolated populations is an extrem ely rare event. As a result, the effects of translocating plants with a different ha plotype composition will result in a serious long term threat to genetic identity and eventually to genetic variability. For this reason, it is important to widen the chlor oplast haplotype characterization in this group to include other isolated populations and more individuals per population in N. obliqua and N. alpina and also to include N. glauca Once there is information about the chloroplast composition of enough popu lations, it will be necessary to limit trans locations only to individuals with matching haplotypes.
145 Table 4 1 Rankings for conservation value in populations of N othofagus obliqua N. alpina and N. glauca analyzed in my study using dat a from three diffe rent data sets # Population Location nMSAT cpDNA ADAP rnk C T rnk P T rnk C T rnk P T N. obliqua 1 La Campana a Coast 2 1 2 Tiltil a Coast 9 2 3 Lampa a Coast 6 4 Alhue a Coast 3 5 Loncha a Coast 10 1 6 Alto Colorado Coast 2 7 Sierras de Bellavista a Andes 5 4 8 Bellavista a Andes 15 3 9 Altos de Lircay a Andes 20 10 Siete Tazas Alto a Andes 19 4 11 Vilches a Andes 26 11 12 Los Ruiles Coast 1 4 13 Quirihue Coast 17 14 Bullileo Alto Andes 20 15 2 15 Embalse Bullileo Andes 16 16 Ninhue Coast 10 7 11 17 Cayumanqui Coast 10 17 18 Recinto Andes 4 15 8 11 19 Lagunas Epulafquen East Andes 27 20 Reserva uble Andes 3 11 7 21 Curanilahue Coast 4 22 Santa Barbara Andes 13 8 21 17 23 Lago Lanalhue Coast 26 24 Ralco Andes 17 11 17 25 Victoria Central Valley 17 19 21 11 26 Pichipillahuen Coast 16 11 11 27 Galletue Andes 1 5 1 28 Quepe Cen tral Valley 23 7 29 Cunco Andes 8 8 26 30 Lago Colico Andes 17 31 Pulmari East Andes 13 32 Lastarria Coast 26 33 Curarrehue Andes 4 19 17 34 Cabecera Oeste Quillen East Andes 14 35 Pilolil East Andes 15 36 Cruces Coast 12 11 4 17 37 Malalhue Central Valley 8 10 18 26 38 Choshuenco Andes 6 6 7 39 Futrono Andes 18 11 40 Pio Protto East Andes 24 41 Quila Quina East Andes 12 42 Hua Hum East Andes 25 43 Lli fen Andes 17 44 Llancacura Coast 13 15 6 17 45 Rio Negro Coast 11 26 46 Rupanco Central Valley 17 7 47 Purranque Coast 6 19 26
146 Table 4 1 C ontinued # Population Location nMSAT cpDNA ADAP rnk C T rnk P T rnk C T rnk P T N. alpina 48 Los Quenes Andes 7 49 Siete Tazas Andes 11 1 8 50 Vilches Andes 15 2 51 Tregualemu Coast 6 10 52 Bullileo Alto Andes 11 1 8 53 Recinto Andes 2 8 15 2 54 Nahuelbuta Coast 1 4 2 2 55 Santa Barbara Andes 10 4 5 8 56 Jauja Andes 9 11 7 13 57 Pichipillahuen ( Chol Chol ) Coast 3 3 3 6 58 Malalcahuello Andes 7 2 59 Melipeuco Andes 11 1 60 Curarrehue Andes 7 4 13 12 61 Quilalelfu Andes 7 62 Lanin East Andes 5 63 Paimun East Andes 18 64 Puerto Canoas East Andes 19 65 Neltume Andes 1 66 Releco Andes 7 7 11 13 67 Playa Bonita East Andes 19 68 Yuco Quilaquina East Andes 25 69 Quilanlahue Chidiak East Andes 26 7 0 Bandurrias East Andes 19 71 Hua Hum East Andes 4 72 Lago Queni East Andes 19 73 Arquilhue Andes 17 74 Llancacura / Las Trancas Coast 4 11 13 11 75 Espejo Chico East Andes 19 76 Peninsula Rauli East Andes 19 77 Hueyusca Coast 4 9 6 N. glauca 78 Loncha Coast 1 1 79 Alto Huelon Coast 3 6 80 Siete Tazas Andes 7 7 81 Los Ruiles Coast 6 3 82 Tregualemu Coast 5 3 83 Bullileo Andes 8 7 84 Alico Andes 2 3 85 Quilleco Andes 3 1 a populations described as N. macrocarpa by Vazquez and Rodriguez (1999) Note: nMSAT=nuclear microsatellite loci (Chapter 2) cpDNA= PCR RFLP fragments from chloroplast DNA ( Azpilicueta et al. 2009; Marchelli and Gallo 2006) ADAP= morphological data from quantitative traits (Chapter 3) rnk C T =within species conservation ranking based on the contribution of a population to the total allelic richness (Petit et al., 1998), rnk P T =within species conservation ranking based on populations distinctness based on dendro grams (May, 1 990; Vane Wright et al., 1991). P opulation p airs 1/2, 4/5, 7/8, and 9/10 are actually different samples of the same population
147 Table 4 2. Top ranked populations in conservation value for N othofagu s obliqua N. alpina and N. glauca first consireding adaptive variation and complemented with neutral variation information. # Population Location Overall ra nking nMSAT cpDNA ADAP rnk C T rnk P T rnk C T rnk P T N. obliqua 27 Galletue Ande s 1 1 5 1 1/2 La Campana/Tiltil Coast 2 2 1 9 2 14 Bullileo Alto Andes 3 20 15 2 7/8 S. Bellavista/Bellavista Andes 4 15 3 5 4 12 Los Ruiles Coast 5 1 4 21 Curanilahue Coast 6 4 20 Reserva uble Andes 7 3 11 7 38 Choshu enco Andes 8 6 6 7 46 Rupanco Central Valley 9 17 7 4/5 Alhue/Loncha Coast 10 10 1 3 6 Alto Colorado Coast 11 2 36 Cruces Coast 12 12 11 4 17 3 Lampa Coast 13 6 33 Curarrehue Andes 14 4 19 17 41 Quila Quina East A ndes 15 12 N. alpina 54 Nahuelbuta Coast 1 1 4 2 2 53 Recinto Andes 2 2 8 15 2 59 Melipeuco Andes 3 11 1 50 Vilches Andes 4 15 2 57 Pichipillahuen (Chol Chol) Coast 5 3 3 3 6 52 Bullileo Alto Andes 6 11 1 8 65 Neltume Andes 7 1 71 Hua Hum East Andes 8 4 51 Tregualemu Coast 9 6 10 N. glauca 78 Loncha Coast 1 1 1 85 Quilleco Andes 2 3 1 84 Alico Andes 3 2 3 81 Los Ruiles Coast 4 6 3 Note: nMSAT=nuclear microsatellite loci (Chapter 2), cpDNA= PCR RFLP fragments from chloroplast DNA (Azpilicueta et al. 2009; Marchelli and Gallo, 2006) ADAP= morphological data from quantitative traits (Chapter 3) rnk C T =within species conservation ranking based on the contr ibution of a population to the total allelic richness (Petit et al., 1998), rnk P T =within species conservation ranking based on populations distinctness based on dendro grams (May, 1 990; Vane Wright et al., 1991). Samples representing the same population ar e merged in the same row.
148 Fig ure 4 1. Range of distribution for a) N othofagus obliqua b) N. alpina and c) N. glauca in grey sho wing the location of populations sampled by at least one of the three data sources used in my study Last Glacial Maximum ( LGM) extent of the ice sheet adapted from Hollin and Schilling (1981).
149 Figure 4 2. Log posterior probabilities ( LnP[D] K values (Evanno et al. 2 005) against K (number of population clusters) obtain using S TRUCTURE 2.3.2 (Pritchard et al. 2 000) f or eight potential clusters combining nMSAT data from 20 populations of N othofagus obliqua 12 of N. alpina and eight of N. glauca T he most likely K K va lue was three clusters in both a) the analysis usin g all seven available loci and b) the analysis using only the six loci that were polymorphic across all species.
150 Figure 4 3 Results of S TRUCTURE 2.3.2 analysis (Pritchard et al. 2000) combining nMSAT data from 20 populations of N othofagus obliqua 12 of N. alpi na and eight of N. glauca Results show clustering from K =2 to K =4 Clusters of populations are represented by colors. Populations are defined by vertical lines and ordered north to south within each species. I employed D ISTRUCT 1.1 (Rosenberg 2 004) to v isualize and edit the S TRUCTURE outputs. a) Analysis u sing all seven available loci, b) analysis using only the six loci that were polymorphic across all species. =Most likely K N obliqua N. alpina N. glauca b ) K =2 K = 4 K = 3 1 5 8 10 14 16 18 20 22 24 25 26 27 29 33 3 6 37 38 44 47 49 51 52 53 54 55 56 57 60 66 74 77 78 79 80 81 82 83 84 85 1 5 8 10 14 16 18 20 22 24 25 26 27 29 33 36 37 38 44 49 51 52 53 54 55 56 57 60 66 7 4 77 78 79 80 81 82 83 84 85 N. obliqua N. alpina N. glauca K =2 K = 4 K = 3 a)
151 Figure 4 4 Dendrograms showing the genetic similarities among N othofag us obliqua N.alpina and N.glauca inferred from nMSAT data. Letters a) and b) represent unrooted neighbo r joining trees obtained using the component N EIGHBOR in P HYLIP 3.69 (Felsenstein 1 989) from R ST values obtained in A RLEQUIN 3.11 (Excoffier et al. 2 005) for all pairs of sampled populations. Letters c) and d) also represent N EIGHBOR /P HYLIP unrooted neighbo r joining trees but obtained in S TRUCTURE 2.3.2 (Pritchard et al. 2 000) from allele frequency divergence values for K =3 and K =4 pairs of S TRUCTURE clusters. M embership of each cluster is in Figure 4 3 =Most likely K All dendrograms were acquired using both all seven available loci, and using only the six loci that were polymorphic across all species. N. obliqua N. a lpina N.glauca N. obliqua N. a lpina N.glauca
152 Figure 4 5. Percentage contribution to the t otal diversity ( P T distinctness obtained from dendro grams, contribution to the total allelic richness ( C T ), a nd the respective rankings for a) Nothofagus obliqua b) N. alpina and c) N. glauca Allelic richness and den d r ograms showi ng the genetic similarities among populations were inferred from the nMSAT data set (Chapter 2).
153 Figure 4 6 Percentage contribution to the total diversity ( P T ) distinctness obtained from dendro grams and its respective ranking for a) N othofagus obliqua and b) N. alpina The den d r ograms showing the genetic similarities among populations were inferred from the ADAP data set (Chapter 3 ).
154 CHAPTER 5 C ONCLUSIONS This study highlights the relatively high levels of neutral genetic variatio n of the South American species of Nothofagus subgenus Lophozonia that are in concert with values of other tree species of similar biology T he genetic structure obtained by neutral markers was relatively low due to extensive gene flow H owever, for growth traits there was more structure with moderate values in Nothofagus alpina and high values in N obliqua reveal ing that natural selection is playing an important role in creating adaptive variation at least in the latter species, which also has a wider ec ological range than N. alpina For adaptive traits in general, N. obliqua also has higher levels of individual heritability, higher coefficients for provenance genetic gains, and more significant associations of traits with geographic and environmental pat terns of variation than N. alpina T his indicates that N. obliqua may have a better chance to adapt to future climatic changes through natural selection and breeders would obtain larger genetic gains by selecting the correct prov enances and best individua ls in this species relative to N. alpina In N. obliqua the association of growth, stem straightness leaf shape, and stomatal density values with geographic and environmental variation indicates genetic adapt ation to different temperature s and precipitat ion s including the tendency of trees growing faster when they are from provenances that ha ve similar environmental conditions than the plac e where they are planted. The patter n s of neutral genetic variation among and within species obtained in my study als o support: 1) the multiple refugia hypothesis for N. obliqua and N. alpina suggesting several centers of genetic diversity across their areas of distribution 2) a
155 topology within the clade base on overall genetic similarity, that place s N. obliqua and N alpina together in a branch and N. glauca as their sister species, 3) the lack of sufficient genetic structure in N. obliqua to support the place ment of its northern populations in a separate species (i.e N. macrocarpa ) or designations of subspecies wit hin the remaining populations, and 4 ) the occurrence of ancient and current hybridization as a co ntributor to the intra specific genetic diversity w ithin the subgenus and as a major source of variability and differentiation for the northern populations of N obliqua Finally, combining neutral and adaptive genetic variation in this study allowed me to propose conservation priorities among and within species and to define breeding zones within species useful for tree improvement strategies. According to my co mbined analysis and also based on the general conservation status of the three species in their natural range s it appears that at the species level N. glauca should have the t op priority for conservation and then as a second priority N. obliqua and N. a lpina The choice of priorities within species and among populations should always go to populations with outstanding uniqueness in their adaptive characteristics T his list of populations should then be complemented with other populations with unique alle les and genotypes obtained from neutral markers. Within the se selected populations the t op priority is unprotected, small, isolated populations The s econd priority is populations insufficiently protected larger in size, or with higher connectivity. In th is study I proposed to give high conservation priority to four fifteen, and seven areas for N. glauca N. obliqua and N. alpina respectively.
156 Regar ding the definition of breeding zones, my integrated analysis indicates that there is no reason to subdivi de the populations of N. glauca or N. alpina in to different breeding zones ; each of these species should be regarded as i t s own unique zone. For N. obliqua the genetic evidence using neutral markers indicates that there are three groups of populations that could be regarded as different breeding zones Nevertheless, observing the among group differences in adaptive traits suggests that defining two breeding zones may be more appropriate for this species : b populations growing wit As a final recommendation, t he translocation of seeds among p o pulations within each breeding zone should be done w ith caution. While pollen flow from trees translocated into a new population should not be a problem because extensive pollen flow occurs naturally within breeding zone s, seed production from the translocated trees will certainly alter the chloroplast comp osition in populations of unique haplotypes because immigration from seed dispersal is an extremely rare event in isolated populations Therefore, translocations may results in the loss of ge netic identity and variability in the long term For this reason, after a more complete chloroplast haplotype characterization it would be desirable to limit trans locations only to individuals with matching haplotypes.
157 LIST OF REFERENCES Allnutt T R Newton A C Lara A Premoli A C Armesto J J Vergara R Gardner M ., 1999. Genetic variation in Fitzroya cupressoides (alerce), a threatened South American conifer. Mol. Ecol. 8, 975 987. Amigo, J., Ramrez, C., 1998. A bioclimatic classification of Chile: woodland communities in the temperate zone. Plant Eco l. 136, 9 26. Armesto, J.J., Leon Lobos, P., Arroyo, M.K., 1995. Los bosques templados del sur de Chile y Argentina: Una isla biogeografica. In : Armesto, J.J., Villagran, C., Arroyo, M.K. (Eds.), Ecologia de los bosques natives de Chile. Ed. Universitaria, Santiago, Chile, pp. 23 28. Azpilicueta M M Caron H Gallo L A ., 2004. SSR markers for analyzing South American Nothofagus species. Silvae Genet. 53, 240 243. Azpilicueta, M.M., Marchelli, P., Gallo, L.A. 2009. The effects of Quaternary glaciations in Patagonia as evidenced by chloroplast DNA phylogeography of Southern beech Nothofagus obliqua Tree Genet. Genomes 5, 561 571. Bekessy S A Allnutt T R Premoli A C Lara A Ennos R A Burgman M A Cortes M Newton A C ., 2002 Genetic var iation in the vulnerable and endemic Monkey Puzzle tree, detected using RAPDs. Heredity 88, 243 249. Benoit, I.L. (Ed ), 1989. Red List of Chilean Terrestrial Flora (Part One). CONAF, Santiago, Chile. Bruschi P Vendramin G G Bussotti F Grossoni P ., 2003. Morphological and molecular diversity among Italian populations of Quercus petraea (Fagaceae). Ann. B ot. 91, 707 716. Buiteveld J Vendramin G G Leonardi S Kamer K Geburek T ., 2007 Genetic diversity and differentiation in European bee ch ( Fagus sylvatica L.) stands varying in management history. For. Ecol. Manage. 247, 98 106. Carrasco B Eaton L ., 2002. Natural history and genetic structure of raul ( Nothofagus nervosa (Phil.) Dim. et Mil.). For. Genet. 9, 275 284. Carrasco B Gar nier M Eaton L Guevara R Car M ., 2009 Proximal causes of genetic variation between and within populations of raul ( Nothofagus nervosa ). Cien. Inv. Agr. 36, 229 238. Chultz, F.A., 2003. Evaluacion del comportamiento geografico en parametros sem inales y de crecimiento inicial de las plantas, en procedencias de Nothofagus dombeyi (Mirb.) Oerst. B.S. Thesis. Universidad Austral de Chile, Valdivia, Chile.
158 Clausen, J., Keck, D.D., Hiesey, W.M., 1940. Experimental Studies on the Nature of Species. I.E ffect of Varied Environments on Western North American Plants. Publication #520. Carnegie Institution of Washington. Washington, DC, US. Cornelius, J., 1994. Heritabilities and additive genetic coefficients of variation in forest trees. Can. J. For. Res. 2 4, 372 379. Craft K J Ashley M V ., 2007 Landscape genetic structure of bur oak ( Quercus macrocarpa ) savannas in Illinois. For. Ecol. Manage. 239, 13 20. Demesure B Comps B Petit J ., 1996. Chloroplast DNA phylogeography of the common beech ( Fagu s sylvatica L.) in Europe. Evolution 50, 2515 2520. Dewoody J Nason J D Hipkins V D ., 2006. Mitigating scoring errors in microsatellite data from wild populations. Mol. Ecol. Notes 6, 951 957. Di Castri, F., Hajek, E., 1976. Bioclimatologa de Chile Vicerrectora Acadmica, U. Catlica de Chile, Santiago, Chile. Donoso, C., 197 9 a Genecological differentiation in Nothofagus obliqua (Mirb.) Oerst. in Chile. For. Ecol. Manage. 2, 53 66. Donoso, C., 1979 b Nothofagus leonii Espino sa, a natural hybrid between Nothofagus obliqua (Mirb.) Orest. and Nothofagus glauca (Phil.) Krasser. N ew Z eal J. Bot. 17, 353 360. Donoso, C., 1982. Resena ecologica de los bosques mediterraneos de Chile. Bosque 4, 117 146. Donoso, C., 1987. Variacion natural en especies de Nothofagus en Chile. Bosque 8, 85 95. Donoso, C., 1993. Bosques templados de Chile y Argentina. Variacion, estructura y dinamica, first ed. Editorial Universitaria, Santiago, Chile. Donoso, C ., Morales, J., Romero, M., 1990. Natural hybridization between Nothofagus obliqua (Mirb.) Oerst. and N. alpina (Poepp. and Endl.) Oerst. in forests of southern Chile. Rev. Chil. Hist. Nat. 63, 49 60. Donoso, P., Donoso, C., Sandoval, V., 1993. Proposicin de zonas de crecimiento de renovales de roble ( Nothofagus obliqua ) y raul ( Nothofagus alpina ) en su rango de distribucin natural. Bosque 14, 37 55. Doyle J J Doyle J L ., 1987 A rapid isolation procedure for small quantities of fresh tissue. Phytoch e m. Bull. 19, 11 15.
159 Evanno G Regnaut S Goudet J ., 2005. Detecting the number of clusters of individuals using the software Structure: a simulation study Mol. Ecol. 14, 2611 2620. Excoffier L Estoup A Cornuet J M ., 2005 Bayesian analysis of an admixture model with mutations and arbitrarily linked markers. Genetics 169, 1727 1738. Excoffier L Laval G Schneider S ., 2005. Arlequin ver. 3.0: An integrated software package for population genetics data analysis. Evol. Bioinform. Online 1, 4 7 50 Faith, D.P., 1992. Conservation evaluati on and phylogenetic diversity. Biol. Conserv. 61, 1 10. Faith, D.P., 1994. Genetic diversity and taxonomi c priorities for conservation. Biol. Conserv. 68 69 74. Falush D., Stephens M., Pritchard J.K., 2003. I nference of population structure using multilocus genotype data: linked loci and correlated allele frequencies Genetics 164, 1567 1587 Felsenstein J ., 1989. PHYLIP Phylogeny Inference Package ( Version 3.2). Cladistics 5, 164 166. Frantz A C Celli na S Krier A Schley L Burke T ., 2009 Using spatial Bayesian methods to determine the genetic structure of a continuously distributed population: clusters or isolation by distance? J. App l. Ecol. 46, 493 505. Fuenzalida H ., 1965. Clima. In: Geog rafia economica de Chile. Texto refundido. Corporacion de Fomento de la Prod uccion (CORFO), Santiago, Chile, pp. 99 152. Gallo, L.A., Marchelli, P., Breitembucher, A., 1997. Morphological and allozymic evidence of natural hybridization between two southern beeches ( Nothofagus spp .) and its relation to heterozygosity and height growth. For. Genet. 4, 15 23. Gilmour, A.R., Gogel, B.J., Cullis, B.R., Thompson, R., 2006. ASReml User Guide Release 2.0. VSN International Ltd, Hemel Hempstead. Gitzendanner M A Soltis P S ., 2000 Patterns of genetic variation in rare and widespread plant congeners. Am. J. Bot. 87, 783 792. Goldstein D B Ruiz Linares A Cavalli Sforza L L Feldman M W ., 1995 An evaluation of genetic distances for use with microsatellite loci. Genetics 139, 463 471. G onza lez, M. 1998 a Nothofagus alpina In: IUCN 2010. IUCN Red List of Threatened Species. Version 2010.4. . Downloaded on 20 May 2011.
160 Gonz a lez, M. 1998 b Nothofagus glauca In: IUCN 2010. IUCN Red List of Threatened Species. Version 2010.4. . Downloaded on 20 May 2011 Gutirrez, B., Ipinza, R. 2000. Evaluacin de parmetros genticos en Nothofagu s In : Ipinza, R., Gutirrez, B., Emhart, V. (Eds. ), Dome sticacin y Mejora Gentica d e Roble y Raul. Universidad Au stral de Chile / Instituto Forestal, Valdivia, Chile pp. 371 390. Haase P ., 1992 Isozyme variability and biogeography of Nothofagus truncata (Fagaceae). New Zeal J. Bot. 30, 315 328. Haase P ., 1993 Genetic variation, ge ne flow, and the 'founder effect' in pioneer populations of Nothofagus menziesii (Fagaceae), South Island, New Zealand. J. Biogeogr. 20, 79 85. Hamrick J L ., 1982 Plant population genetics and evolution. Am. J. Bot. 69, 1685 1693. Hamrick J L ., 2004. Re sponse of forest trees to global environmental changes. For. Ecol. Manage. 197, 323 335. Hamrick J L Godt, M J W ., 1996 Effects of life history traits on genetic diversity in plant species. Philos. T. R. Soc. B 351, 1291 1298. Heusser C J ., 1990. Ice age vegetation and climate of subtropical Chile. Palaeogeogr., Palaeoclimatol. Palaeoecol. 80, 107 127. Hewitt G ., 2000. The genetic legacy of the Quaternary ice ages. Nature 405, 907 913. Hiraoka K Tomaru N ., 2009 Population genetic structure of Fa gus japonica revealed by nuclear micro satellite markers. Int. J. Plant Sci. 170, 748 758. Hodge G.R., Dvorak, W.S., 2004. The CAMCORE international provenance progeny trials of Gmelina arborea : genetic parameters and potential gain. New For. 28, 147 166. Hollin J T Schilling D H ., 1981. Late Wisconsin Weichselian mountain glaciers and small ice caps. In: Denton G H Hughes T J (Eds.), The Last Great Ice Sheets. Wiley, New York, USA pp. 179 206 Hovenden M.J., 2001. The influence of temperature a nd genotype on growth and stomatal morphology of southern beech, Nothofagus cunninghamii (Nothofagaceae). Aust. J. Bot. 49, 427 434. Hovenden, M.J., Vander Schoor, J.K., 2003. Nature vs nurture in the leaf morphology of Southern beech, Nothofagus cunningha mii (Nothofagaceae). New Phytol. 161, 585 594.
161 Hovenden, M.J., Vander Schoor, J.K., 2006. The response of leaf morphology to irradiance depends on altitude of origin in Nothofagus cunninghamii New Phytol. 169, 291 297. Ipinza R Espejo J ., 2000. Biolog ia reproductiva en Nothofagus In: Ipinza, R., Gutierrez, B., Emhart, V. (Eds.), Domesticacion y mejor a genetica de rauli y roble. Universidad Au stral de Chile / Instituto Forestal Valdivia, Chile pp. 75 93. Ipinza R., Gutierrez, B., 2000. Estrategia de mejora genetica para N. alpina and N. obliqua In: Ipinza R Gutierrez B Emhart V (Eds.), Domesticacion y mejor a genetica de rauli y roble. Universidad Au stral de Chile / Instituto Forestal, Valdivia, Chile pp. 403 417 Ipinza, R., Gutirrez, B., Medina, A. 2000. Variacin gentica del peso y forma de la semilla y crecimiento inicial de procedencias naturales de N. alpina y N. obliqua en Chile In: Ipinza, R., Gutierrez, B., Emhart, V. (Eds.), Domesticacion y mejora genetica de rauli y roble. Univ ersidad Au stral de Chile / Instituto Forestal, Valdivia, Chile pp. 323 347. Jones R C Vaillancourt R E Jordan G J ., 2004 Microsatellites for use in Nothofagus cunninghamii (Nothofagaceae ) and related species. Mol. Ecol. Notes 4, 14 16. Kalinowski, S.T., 2004. Counting alleles with rarefaction: Private alleles and hierarchical sampling desig ns. Conserv. Genet. 5, 539 543. Kalinowski, S.T., 2005. HP RARE 1.0: a computer program for performing rarefaction on measures of allelic richness. Mol. Ecol. No tes 5, 187 189. Knapp, M., Stockler, K., Havell, D., Delsuc, F., Sebastiani, F., Lockhart, P.J., 2005. Relaxed molecular clock provides evidence for long distance dispersal of Nothofagus (southern beech). Plos Biol. 3, 38 43. Landry P A Koskinen M T Primmer C R ., 2002 Deriving evolutionary relationships among populations using microsatellites and ( 2 : all loci are equal, but some are more equal than others ... Genetics 161, 1339 1347. Ledgard, N.J., Norton, D.A., 1988. Shoot growth in 2 3 year old Nothofagus seedlings. New Zeal. J. Ecol. 11, 105 108. Lee Y J Hwang S Y Ho K C Lin T P ., 2006 Source populations of Quercus glauca in the last glacial age in Taiwan revealed by nuclear mi crosatellite markers. J. Hered. 97, 261 269. Le Quesne C Sandoval L ., 2001 New southern limit for Nothofagus glauca (Phil.) Krasser. Gayana, 58, 139 142.
162 Li, P., Beaulieu, J., Corriveau, A., Bousquet, J., 1993. Genetic variation in juvenile growth and phenology of white spruce provenance progeny test. Silva e Genet. 42, 52 60. Luebert, F., Pliscoff, P., 2006. Sinopsis bioclimtica y vegetacional de Chile. Editorial Universitaria, Santiago, Chile. Magri D Vendramin G G Comps B Dupanloup I Geburek T Gomory D Latatowa M Litt T Paule L Roure J M Tantau I van der Knaap W O Petit R J de Beaulieu J L ., 2006 A new scenario for the Quaternary history of European beech populations: palaeobotanical evidence and genet ic consequences. New Phytol. 171, 199 221. Manos, P.S., 1997. Sy stematics of Nothofagus (Nothofagaceae) based on rDNA spacer sequences (ITS): Taxonomic congruence with morphology and plastid sequences. Am. J. Bot. 84, 1137 1155. Evol. 9, 373 375. Marchelli P Caron H Azpilicueta M M Gallo L A ., 2008. Primer note: A new set of highly polymorphic nuclear microsatellite markers for Nothofagus nervosa and related South Am erican species. Silvae Genet. 57, 82 85. Marchelli P Gallo L Scholz F Ziegenhagen B ., 1998 Chloroplast DNA markers reveal a geographical divide across Argentinean southern beech Nothofagus nervosa (Phil.) Dim. et Mil. di stribution area. Theor. Appl. Genet. 97, 642 646. Marchelli P Gallo L A ., 20 01 Genetic diversity and differentiation in a southern beech subjected to introg ressive hybridization. Heredity 87, 284 293. Marchelli P Gallo L A ., 2004 The combined role of glaciation and hybridization in shaping the distribution of genetic variati on in a Patagonian southern beech. J. Biogeogr. 31, 451 460. Marchelli, P., Gallo, L.A., 2006. Multiple ice age refugia in a southern beech of South America as evidenced by chloroplast DNA markers. Conserv. Genet. 7, 591 603. Markgraf V Romero E J Vi llagran C ., 1996. History and Paleoecology of South American Nothofagus forests. In: Veblen, T.T., Hill, R.S., Read, J. (Eds.), The ecology and biogeography of Nothofagus forests. Yale University Press, New Haven & London pp. 354 386. Marsico TD, Hellma nn JJ, Romero Severson J (2009) Patterns of seed dispersal and pollen flow in Quercus garryana (Fagaceae) following post glacial climatic c hanges. J. Biogeogr. 36, 929 941.
163 Matheson, A.C., 1989. Statistical methods and problems in testing large numbers of genotypes across sites. In: Gibson, G.L., Griffin, A.R., Matheson, A.C. (Eds.), Breeding Tropical Trees: Population Structure and Genetic Improvement Strategies in Clonal and Seedling Forestry. Proceedings IUFRO meeting, 28 November 3 December 1988, Pattay a, Thailand. Oxford Forestry Ins titute, Oxford, UK, pp. 93 105. Mathiasen P Premoli A C ., 2010 Out in the cold: genetic variation of Nothofagus pumilio (Nothofagaceae) provides evidence for latitudinally distinct evolutionary histories in austral Sout h America. Mol. Ecol. 19, 371 385. May, R. M. 1990 Tax onomy as destiny. Nature 347, 129 1 30. McGarigal, K., Cushman, S., Stafford, S., 2000. Multivariate statistics for wildlife and ecology research. Springer Verlag New York, Inc. McKay, J.K., Latta, R.G ., 2002. Adaptive population divergence: markers, QTL and traits. Trends Ecol. Evol. 17, 285 291. Medina, A.M., 2001. Evaluacion del comportamiento genetico y geografico en parametros seminales y de crecimiento inicial de las plantas, en procedencias de No thofagus alpina (Poep. et Endl.) Oerst.). B.S. Thesis. Universidad Austral de Chile, Valdivia, Chile. Millern M Gallo L Marchelli P ., 2008. The effect of volcanism on postglacial migration and seed dispersal. A case study in southern South A merica. Tree Genet. Genomes 4, 435 443. Moreno P I Lowell T V Jacobson G L Denton G H ., 1999 Abrupt vegetation and climate changes during the last glacial maximum and last termination in the Chilean Lake District: A case study from Canal de la Punti lla (41S). Geogr. Ann. A 81 285 311. Muir G Lowe A J Fleming C C Vogl C ., 2004 High nuclear genetic diversity, high levels of outcrossing and low differentiation among remnant populations of Quercus petraea at the margin of its range in Ireland. Ann. Bot. 93, 691 697. Nei M ., 1973 Analysis of gene diversity in sub divided populations. P. Nat l Acad. Sci. USA 70, 3321 3323. Nunez Avila M C Armesto J J ., 2006. Relict islands of the temperate rainforest tree Aextoxicon punctatum (Aextoxicaceae) in semi arid Chile: genetic diversity and b iogeographic history. Aust. J. Bot. 54, 733 743. Ogden J ., 1989 On the coenospecies concept and tree migrations during the oscillations of the Ple istocene climate. J. Roy. Soc. New Zeal 19, 249 262.
164 Ormazabal, C., Benoit, I., 1987. El estado de conservacion del genero Nothofagus en Chile. Bosque 8, 109 120. Page R D M ., 1996. TreeView: A n application to display phylogenetic trees on personal computers. Comput. Appl. Biosci. 12, 357 358. Pakkad G Ueno S Yo shimaru H ., 2008 Genetic diversity and differentiation of Quercus semiserrata Roxb. in northern Thailand revealed by nuclear and chloroplas t microsatellite markers. For. Ecol. Manage. 255, 1067 1077. Pastorino M J Marchelli P Milleron M Soliani C Gallo L A ., 2009 The effect of different glaciation patterns over the current genetic structure of the southern beech Nothofagus antarctica Genetica 136, 79 88. Petit R J Aguinagalde I de Beaulieu J L Bittkau C Brewer S Cheddadi R Ennos R Fineschi S Grivet D Lascoux M Mohanty A Mller Starck G Demesure Musch B Palm A Martn J P Rendell S Vendramin G G ., 2003 Glacial refugia: hotspots but not melting pots of genetic diversity. Science 300, 1563 1565. Petit, R.J., El Mousadik, A., Pons, O., 1998. Identifying populations for conservation on the basis of genetic markers. Conserv. Biol. 12, 844 855. Pineda G ., 2000 Variabilidad aloenzimatica de N. alpina en Chile. In: Ipinza, R., Gutierrez, B., Emhart, V. (Eds.), Domesticacion y mejora genetica de rauli y roble. Universidad Au stral de Chile / Instituto Forestal, Valdivia, Chile pp. 95 119. Premoli, A.C., 1997. Genetic variation in a geographically restricted and two widespread species of South American Nothofagus J. Biogeogr. 24, 883 892. Premoli A C Vergara R Souto C P Lara A Newton A C ., 2003. Lowland valleys shelter the ancient conifer Fitzroya cupressoides in the Central Depression of southern Chile. J. Roy. Soc. New Zeal. 33, 623 631. Premoli, A.C., Brewer, C.A., 2007. Environmental v. genetically driven variation in ecophysiological traits of Nothofagus pumilio from contrasting elevations. Aust. J. Bot. 55, 585 591. Premoli, A.C., Raffaele, E., Mathiasen, P., 2007. Morphological and ph enological differences in Nothofagus pumilio from contrasting elevations: Evidence from a common garden. Austral Ecol. 32, 515 523. Pritchard, J.K., Stephens, M., Donnelly, P., 2000. Inference of population structure using multilocus genotype data. Genetic s 155, 945 959. Puntieri, J.G., Grosfeld, J.E., Stecconi, M., Brion, C., Azpilicueta, M.M., Gallo, L., 2006. Desarrollo temprano del roble ( Nothofagus obliqua ): un analisis arquitectural de procedencias de Argentina. Bosque 27, 44 51.
165 Reed, D.H., Frankham, R., 2001. How closely correlated are molecular and quantitative measures of genetic variation? A meta analysis. Evolution 55, 1095 1103. Rempel, R.S., Carr, A.P. 2003. Patch Analyst 3.1 Extension for A RC GIS 9.1 Available from http://flash.lakeheadu.ca/~ rrempel/patch/index.html. Accessed March 2008. Rice W R ., 1989. Analyzing tables of statistical tests. Evolution 43, 223 225. Riveros, M., Parades, M.A., Rosas, M.T., Cardenas, E., Armesto, J., Arroyo, M.K., Palma, B., 1995. Reproductive biology in specie s of the genus Nothofagus Environ. Exp. Bot. 35, 519 524. Rodriguez, J., Medina, A. 2000. Viverizacin de procedencias de N. alpina y N. obliqua en contenedores. In: I pinza, R., Gutirrez, B., Emhart, V. (Eds.), Dome sticacin y Mejora Gentica de Roble y Raul. Universidad Austral de Chile / Instituto Forestal, Valdivia, Chile pp. 307 322. Romero, H., 1985. Geografa de los climas. Tomo XI. Coleccin de geografa de Chile. Editorial IGM, Santiago, Chile. Rosenberg N.A., 2004. Distruct: a program for the graphical display of population structure Mol. Ecol. Notes 4, 137 138. Rousset F ., 1996 Equilibrium values of measures of population subdivision for stepwi se mutation processes. Genetics 142, 1357 1362. Rousset F ., 2008 Genepop'007: a complete reimple mentation of the Genepop software for Windows and Linux. Mol. Ecol. Res. 8, 103 106. Santelices, R., Navarro, R.M., Drake, F., 2009. Caracterizacion del material forestal d e reproduccion de cinco procedencias de Nothofagus alessandrii Espinosa, una especie en peligro de extincion. Interciencia 34, 113 120. SAS, 2008. SAS/STAT software release 9.2, SAS Institute Inc. Cary, NC. Schwarz, G.E., 1978. Estimating the dimension of a model. Ann. Stat. 6, 461 464. Sebbenn, A.M., Pontinha, A.A.S., Giannotti, E., K ageyama, P.Y., 2003. Genetic variation in provenance progeny test of Araucaria angustifolia (bert.) O. Ktze. in Sao Paulo, Brazil. Silvae Genet. 52, 181 184. Selkoe K A Toonen R J ., 2006 Microsatellites for ecologists: a practical guide to using and e valuating microsat ellite markers. Ecol. Lett. 9, 615 629. Sepulveda C A Stoll A ., 2003 Nothofagus alpina (Poepp. et Endl.) Oerst. (Fagaceae) in the coastal area of the Maule Regio n, central Chile. Gayana Bot. 60, 132 133.
166 Sierra Lucero, V., McKeand, S .E., Huber, D.A., Rockwood, D.L., White, T.L., 2002. Performance differences and genetic parameters for four coastal provenances of loblolly pine in the Southeastern United States. For. Sci. 48, 732 742. Slatkin, M., 1995. A measure of population subdivisi on based on micr osatellite allele frequencies. Genetics 139, 457 462. Soltis D E Gitzendanner M A Strenge D D Soltis P S ., 1997 Chloroplast DNA intraspecific phylogeography of plants from the Pacific Northwest of North America. Plant Syst. Evol. 206, 353 373. Soltis D E Morris A B McLachlan J S Manos P S Soltis P S ., 2006. Comparative phylogeography of unglaciated eastern North America. Mol. Ecol. 15, 4261 4293. Squillace, A. E., 1974. Average genetic correlations among offspring fro m open pollinated forest trees. Silvae Genet. 23, 149 1 56. Steubing, L., Godoy, R., Alberdi, M., 2002. Metodos de ecologia vegetal. Editorial Universitaria, Santiago, Chile. Sun, O.J., Sweet, G.B., 1996. Comparison of frost tolerance of Nothofagus solandri var. cliffortioides (Hook. f.) Poole and Nothofagus menziesii (Hook. f.) Oerst. New Zeal. J. Bot. 34, 273 278. Taberlet P Fumagalli L Wust Saucy A G Cosson J F ., 1998 Comparative phylogeography and postglacial coloniza tion routes in Europe. Mol Ecol. 7, 453 464. Taylor K J Lowe A J Hunter R J Ridgway T Gresshoff P M Rossetto M ., 2005. Genetic diversity and regional identity in the Australian remnant Nothofagus moorei Aust. J. Bo t. 53, 437 444. Tchatchoua, D. T., Aravanopoulos, F. A. 2010a. Evaluation of selected European chestnut ( Castanea sativa ) provenances I: Inter provenance genetic variation. Acta Hort. 866, 203 213. Tchatchoua, D. T., Aravanopoulos, F. A. 2010b. Evaluation of selected European chestnut ( Castanea sativa ) provenances II: Intra provenance family variation. Acta Hort. 866, 215 224. Thiele R Moreno H Elgueta S Lahsen A Rebolledo S Petit Breuilh M ., 1998 Evolucion geologico geomorfologica cuaternaria del tramo superior del valle del rio La ja. Rev. Geol. Chile 25, 229 253. Torres Diaz C Ruiz E Gonzalez F Fuentes G Cavieres L A ., 2007. Genetic diversity in Nothofagus alessandrii (Fagac eae), an endangered endemic tree species of the coastal maulino forest of central Chile. Ann. Bo t. 100, 75 82.
167 van Oosterhout C Hutchinson W F Wills D P M Shipley P ., 2004 Micro Checker: software for identifying and correcting genotyping errors in microsatellit e data. Mol. Ecol. Notes 4, 535 538. van Tienderen, P.H., de Haan, A.A., van der Linden, C.G., Vosman B., 2002. Biodiversity assessment using markers for ecologically important traits. Trends Ecol. Evol. 17, 577 582. Vane W right, R.I., Humphries, C.J., Williams, P.H. 1991. What to protect? Systematics and the agon y of choice. Biol. C onserv. 55, 235 254. Vazquez, F.M., Rodriguez, R.A., 1999. A new subspecies and two new combinations of Nothofagus Blu me (Nothofagaceae) from Chile. Bot. J. Linn. Soc. 129, 75 83. Veblen, T., Schlegel, F., 1982. Resena ecologica de los bosques del sur de Chile. Bosque 4, 73 115. Vergara, R., 2000. Regiones de procedencia de N. alpina y N. obliqua In: Ipinza, R., Gutierrez, B., Emhart, V. (Eds.), Domesticacion y mejora genetica de rauli y roble Universidad Austral de Chile / Inst ituto Forestal, Valdivia, Chile, pp. 121 132 Vergara, R., Bohle, K., 2000. Estado actual del recurso forestal de N. alpina y N. obliqua In: : Ipinza, R., Gutierrez, B., Emhart, V. (Eds.), Domesticacion y mejora genetica de rauli y roble Universidad Austral de Chile / Inst ituto F orestal, Valdivia, Chile, pp. 40 52. Villagran C ., 1991 Historia de los bosques templados del sur de Chile durante el Tar diglacial y Postglacial. Rev. Ch il Hist. Nat. 64, 447 460. Villagran C ., 2001. A model for the history of vegetation of the Coastal Range of central southern Chile: Darwin's glacial hypothesis. Re v Chil. Hist. Nat. 74, 793 803. Villagran C Moreno P Villa R ., 1995 Antecedentes palinologicos acerca de la historia cuaternaria de los bosques chilenos. In: Armesto, J.J., Villagran C., Arroyo, M.K. (Eds.), Ecologia de los bosques nativos de Chile Editorial Universitaria, Santi ago, Chile pp. 51 70. White, T.L., Adams, W.T., Neale, D.B., 2007. Forest genetics. Wallingford, Oxfordshire, UK, CABI Pub. Wightman, K.E., Ward, S.E., Hagg ar, J.P., Santiago, B.R., Cornelius, J.P., 2008. Performance and genetic variation of big leaf mahogany ( Swietenia macrophylla King) in provenance and progeny trials in the Yucatan Peninsula of Mexico. For. Ecol. Manage. 255, 346 355. Wilcox, M.D., Ledgard N.J., 1983. Provenance variation in the New Zealand species of Nothofagus New Zeal. J. Ecol. 6, 19 31.
168 Worth J R P Jordan G J McKinnon G E Vaillancourt R E ., 2009 The major Australian cool temperate rainforest tree Nothofagus cunninghamii wi thstood Pleistocene glacial aridity within multiple regions: evidence from t he chloroplast. New Phytol. 182, 519 532. Wright S ., 1931 Evolution in Mendelian populations. Genetics 16, 97 159. Wright, S., 1943. Isolation by distance. Genetics 28, 114 138. Wright S ., 1965 The interpretation of population structure by F statistics with special regard to systems of mating. Evolution 19, 395 420. Xie, C.Y., 2008. Ten year results from red alder ( Alnus rubra Bong.) provenance progeny testing and their implicat ions for genetic improvement. New For. 36, 273 284. Zobel, B., Talbert, J., 1984. Applied Forest Tree Improvement. John Wiley and Sons, Inc., NY, USA.
169 BIOGRAPHICAL SKETCH Rodrigo Vergara was born in Santiago Chile in 1969, and was raised in Chillan abou t 400 km south of Santiago where he went to school in Colegio Padre A. Hurtado. After high school he moved south to Valdivia roughly 450 km south of Chillan, to study Forest Engineering at Universidad Austral de Chile where he graduated on 1995 with his u Nothofagus alpina (Poepp. et. Forest Genetics in the Department of Silviculture at Universidad Austral de Chile where h e participated on several projects related to tree breeding and co nservation genetics. In 2000 he returned to school to pursue h is Master s of Science in Forest Genetics in the School of Fores t Resources and Conservation at the University of F lorida where h e graduated i n 2003 with his thesis predicted breeding values using field trials with large rectangular plots of slash pine ( Pinus elliottii var. elliottii e Department of Botany currently Department of Biology, under the advisement of Dr. Pamela S. Soltis and graduated in 2011.