<%BANNER%>

Interactions of Temperature with the Dynamics of Aedes Aegypti (l.) Development in Household Vessels

Permanent Link: http://ufdc.ufl.edu/UFE0041509/00001

Material Information

Title: Interactions of Temperature with the Dynamics of Aedes Aegypti (l.) Development in Household Vessels
Physical Description: 1 online resource (154 p.)
Language: english
Creator: Padmanabha, Harish
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: aegypti, behavior, cell, chikungunya, climate, dengue, ecology, energetics, human, metabolism, model, resources, size, temperature, vector
Entomology and Nematology -- Dissertations, Academic -- UF
Genre: Entomology and Nematology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The mosquito Aedes aegypti (L.) transmits dengue and chikungunya viruses among human hosts in cities at temperatures ranging 20-30degC, where it thrives in water that humans store in vessels. Two processes are known to drive the dynamics of A. aegypti production in these habitats: resource limitation and human behavior. Although temperature affects mosquito size and maturation rate, there is little understanding of how temperature may affect the capacity of mosquitoes to develop in food limited, domestic environments. In this dissertation we use modeling, experimental and field approaches to explore the general hypothesis that the effects of human and resource-mediated processes on the size, development rate and pupation success of A. aegypti may be modified across temperature gradients or cities that vary in temperature. We experimentally show that due to increased energy demands, larvae at higher temperature may experience tradeoffs between development rate, size and starvation resistance. Moreover, our model of growth and energy storage in mosquitoes indicates that these tradeoffs can both explain the commonly observed effects of temperature on mosquito size and result in interactive effects of food and temperature on development rate. A food-temperature developmental interaction was morphologically corroborated by analyzing wing and epidermal cell size in experimentally reared A. aegypti. Our data suggest that in resource poor habitats commonly inhabited by urban A. aegypti, development rate may be lower at 24-26degC in comparison to 20-22 or 28-30degC, thereby challenging the fundamental assumptions of widely used temperature-driven models of A. aegypti dynamics. Our field study in dengue endemic areas of Colombia indicates that the outcome of this food-temperature interaction depends on specific household behaviors, including the frequency of using, emptying and placing lids on domestic water. Moreover, each of these acts in the context of particular socio-cultural perceptions, altitude, urban architecture and water supply. Nonetheless, our data suggest that climate-induced decreases in water security predicted throughout Colombia may indirectly increase A. aegypti abundance through changes in human behavior. These results demonstrate the underlying interactions between human and mosquito responses to a changing environment and enhancing our capacity to predict A. aegypti production and design locally adapted intervention strategies.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Harish Padmanabha.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Lounibos, Leon P.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2012-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041509:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041509/00001

Material Information

Title: Interactions of Temperature with the Dynamics of Aedes Aegypti (l.) Development in Household Vessels
Physical Description: 1 online resource (154 p.)
Language: english
Creator: Padmanabha, Harish
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: aegypti, behavior, cell, chikungunya, climate, dengue, ecology, energetics, human, metabolism, model, resources, size, temperature, vector
Entomology and Nematology -- Dissertations, Academic -- UF
Genre: Entomology and Nematology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The mosquito Aedes aegypti (L.) transmits dengue and chikungunya viruses among human hosts in cities at temperatures ranging 20-30degC, where it thrives in water that humans store in vessels. Two processes are known to drive the dynamics of A. aegypti production in these habitats: resource limitation and human behavior. Although temperature affects mosquito size and maturation rate, there is little understanding of how temperature may affect the capacity of mosquitoes to develop in food limited, domestic environments. In this dissertation we use modeling, experimental and field approaches to explore the general hypothesis that the effects of human and resource-mediated processes on the size, development rate and pupation success of A. aegypti may be modified across temperature gradients or cities that vary in temperature. We experimentally show that due to increased energy demands, larvae at higher temperature may experience tradeoffs between development rate, size and starvation resistance. Moreover, our model of growth and energy storage in mosquitoes indicates that these tradeoffs can both explain the commonly observed effects of temperature on mosquito size and result in interactive effects of food and temperature on development rate. A food-temperature developmental interaction was morphologically corroborated by analyzing wing and epidermal cell size in experimentally reared A. aegypti. Our data suggest that in resource poor habitats commonly inhabited by urban A. aegypti, development rate may be lower at 24-26degC in comparison to 20-22 or 28-30degC, thereby challenging the fundamental assumptions of widely used temperature-driven models of A. aegypti dynamics. Our field study in dengue endemic areas of Colombia indicates that the outcome of this food-temperature interaction depends on specific household behaviors, including the frequency of using, emptying and placing lids on domestic water. Moreover, each of these acts in the context of particular socio-cultural perceptions, altitude, urban architecture and water supply. Nonetheless, our data suggest that climate-induced decreases in water security predicted throughout Colombia may indirectly increase A. aegypti abundance through changes in human behavior. These results demonstrate the underlying interactions between human and mosquito responses to a changing environment and enhancing our capacity to predict A. aegypti production and design locally adapted intervention strategies.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Harish Padmanabha.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Lounibos, Leon P.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2012-08-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041509:00001


This item has the following downloads:


Full Text

PAGE 1

INTERACT IONS OF TEMPERATURE WITH THE DYNAMICS OF Aedes aegypti (L.) DEVELOPMENT IN HOUSEHOLD VESSELS By HARISH PADMANABHA A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010 1

PAGE 2

2010 Harish Padm anabha 2

PAGE 3

I dedicate this work to R ichard Levins for stimulating my interest in understanding complexity. 3

PAGE 4

ACKNOWLEDGMENTS I thank Phil Lounibos, Cynthia Lord, Ben Bolk er and Jorge Rey for their contributions and dedication to teaching; my family and in particular Natalia, for their support of this endeavor; Fabio Correa, Edison Soto, Naoya Nish imura, Richard Escher, Camilo Rubio, Lucia Suarez, Marcelo Torres and Jorge de la Salas for their contributions; Salua Osorio in the National Institute of Health of Colombia for administra tive support; health departments of Barranquilla, Bucarmanga and Armenia for logistical support in the field work. This dissertation was largely funded by the Global Environmental Facility/Wor ld Bank Integrated Na tional Adaptation Pilot to Climate Change. 4

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS ..............................................................................................................4 LIST OF TABLES ..........................................................................................................................7 LIST OF FIGURES ........................................................................................................................8 ABSTRACT .................................................................................................................................. 10 CHAPTER 1 INTRODUCTION ..............................................................................................................12 1.1 General Eco-Physiological Aspects of Developm ent and Growth in A. aegypti ..........13 1.2 Water Storage Behavior and A. aegypti Larval Developm ent......................................15 1.3 Food Limitation and A. aegypti Developm ent..............................................................17 1.4 Temperature Effects on A. aegypti and Ectotherm Development.................................19 1.5 Hypotheses and Research Questions............................................................................22 2 ECOLOGICAL LINKS BE T WEEN WATER STORAGE BEHAVIORS AND Aedes aegypti PRODUCTION: IMPLICATIONS FOR DENGUE VECTOR CONTROL IN VARIABLE CLIMATES...................................................................................................2 4 2.1 Introduction........................................................................................................ ........... 24 2.2 Materials and Methods.................................................................................................. 26 2.2.1 Study Area........................................................................................................... 26 2.2.2 Container Selection..............................................................................................27 2.2.3 Physical Description of Houses and Study Vessels.............................................28 2.2.4 Data Collection.................................................................................................. .. 29 2.2.5 Data analysis.................................................................................................... .... 30 2.3 Results................................................................................................................... 32 2.3.1 Study Containers Water Storage Practices and A. aegypti Production ...............32 2.3.2 Qualitative Assessment of Behavior and Motivations .........................................33 2.3.3 Quantitative Assess m ent of City-Specific Behaviors..........................................34 2.3.4 Effects of Water Storage Behaviors and City on A. aegypti Production .............34 2.4 Discussion..................................................................................................................... 36 3 CELL SIZE AND NUMBER RELATIONSHIPS IN Aedes aegypti (L.) W INGS SHOW DIFFERENTIAL AND INTERAC TIVE EFFECT S OF TEMPERATURE AND FEEDING RATE ON MOSQUITO MORPHOLOGY ...................................................50 3.1 Introduction........................................................................................................ ........... 50 3.2 Materials and Methods.................................................................................................. 53 3.2.1 Experimental Rearing of Aedes aegypti ............................................................... 53 3.2.2 Wing Photography and Image Analysis..............................................................55 3.2.3 Data Analysis.................................................................................................... ... 56 5

PAGE 6

3.3 Results.......................................................................................................................57 3.4 Discussion..................................................................................................................... 60 4 ENERGY STORAGE STRATEGIES EXP LAIN DEVELOPING DISEASE VECTORS RESPONSE TO RISING TEMPERATURE ......................................................................73 4.1 Introduction........................................................................................................ ........... 73 4.2 Methods and Results................................................................................................. .... 75 4.2.1 Laboratory Experiments on Effects of Te mperatur e on Development................75 4.2.2 Model of Growth and Energy Storage in Aedes aegypti (L.) ..............................83 4.2.3 Parameter Estimation and Model Fitting Using Experim ental Data...................87 4.2.4 Sensitivity Analyses............................................................................................. 98 4.2.5 Model Predictions..............................................................................................10 3 4.3 Discussion................................................................................................................... 108 5 CONCLUSIONS ...............................................................................................................137 5.1 Interactions between Container Environment and A. aegypti Growth and Development........................................................................................................138 5.2 Human Ecological In teractions and the Dynam ics of A. aegypti Production.............141 LIST OF REFERENCES ............................................................................................................146 BIOGRAPHICAL SKETCH ......................................................................................................154 6

PAGE 7

LIST OF TABLES Table Page ........................................................................................42 2-1 Water storage containers (infested with A. aegypti la rvae) studied, follow-up interval missing data and pupation rates 2-2 Emptying, usage, lidding and A. aegypti production in study vessels ..............................42 2-3 City-wide differences in patterns of use, em ptying and lidding of vessels ......................43 2-4 Patterns of use, emptying and rate of A. aegypti production across survey periods .........43 2-5 Univariate associations between wate r storage behaviors and city and m ean A. aegypti pupal production in infested vessels in order of l og-likelihood (LL) ...............................44 2-6 Multivariate models of mean rate of A. aegypti pupal production, using generalized negative binom ial regression for overall data and within cities ......................................45 1 2-7 Overall and city-speci fic best fitting models of household behaviors and A. aegypti pupal production .........................................................................................................................46 3-1 Random effects models of wing ar ea and m ean cell size in female A. aegypti in high and low food experiments .................................................................................................66 3-2 Random effects models of wing area and m ean cell size in male A. aegypti in high and low food experiments ........................................................................................................66 3-3 Random effects models of Z in female and male A. aegypti in high and low food experim ents .......................................................................................................................67 diff 3-4 Summary table of significance of main and in teractive effects of food and tem perature treatments in random effects m odel of wing size, cell size and Z ................................67 diff 4-1 Parameters of model and used to describe experim ental data ........................................115 4-2 Joint narrow range ML parameter space for maximum growth efficiency (g), coefficient of energy storage (c ) and exponent of metabolism ( m ).................................................116 s 7

PAGE 8

LIST OF FIGURES Figure Page 2-1 Study cities within Colombia. ...........................................................................................47 2-2 Water storage vessels in study areas.. ...............................................................................48 2-3 Effect of emptying in terval on the rate of A. aegypti pupal production. ..........................49 3-1 Images used to determine wing size and epidermal cells s ize in A. Aegypti .. ..................68 3-2 Mean time to pupation among experimental treatm ents of food and temperature. ..........69 3-3 Wing measurements in female A. aegypti across food and temperature treatm ents. ........70 3-4 Wing measurements in male A. aegypti across f ood and temperature treatments.. ..........71 3-5 Significant interactive effect s of food and tem perature on z .diff ......................................72 4-1 Minimum feeding days required to pupa te upon transfer to distilled water across temperatures.. .............................................................................................117 4-2 Weight trajectory at 28 C.. .............................................................................................118 4-3 Mean W ( 95% CI) among temperatures.L4..................................................................118 4-4 Starvation survival in unfed newly larvae and in 1-day fed larvae. ................................119 4-5 Starvation survival, as measur ed by the W eibull scale parameter ( ) of starved A. aegypti among temperature and feeding treatments.. ....................................120 4-6 Joint space of coefficient ( c ) and exponent (m ) of m etabolism that reproduces mean starvation resistance of unfed larvae from 20 to 30 C.m......................121 4-7 Estimates of temperature dependent param eters based on Experiments A-D: ...............122 4-8 Maximum likelihood model fits to observed data.. ........................................................123 4-9 Sensitivity of weight after 1 to 4 days feeding to v ariation in model parameters with respect to 28 C ML fit.. ........................................................................125 4-10 Sensitivity of starvation survival after 1 to 4 days feeding to v ariation in model parameters with respect to 28 C fit.. ...............................................................126 8

PAGE 9

4-11 Sensitivity of energy reserves (E) after 1 to 4 days feeding to variation in model parameters with respect to 28 C fit.. ...............................................................127 4-12 Sensitivity of proportion of energy reserves (E/W) after 1 to 4 days feeding to variation in model parameters with respect to 28 C fit.. ...............................128 4-13 Profiles of simulated versus e xperimental survival for variation in growth efficiency (r) and energy allocation (c ), by temperature and feeding age.s....................129 4-14 Simulated ML trajectories of mass and energy s tores across temperature. ....................133 4-15 Simulated energy stores as a fraction of body weight across temperature and feeding age using M L paramete r scenarios for each temperature.. ..........................134 4-16 Food limitation and time to attain 100% pupation.. ........................................................135 4-17 Effects of temperature on cumulative pupation.under food li m ited conditions (3g/6h). ........................................................................................................136 5-1 Schematic of eco-social system of domestic Aedes aegypti (L.) production. .................145 9

PAGE 10

Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy INTERACTIONS OF TEMPERATURE WITH THE DYNAMICS OF Aedes aegypti DEVELOPMENT IN HOUSEHOLD VESSELS By Harish Padmanabha August 2010 Chair: Leon Phillip Lounibos Major: Entomology and Nematology The mosquito Aedes aegypti (L.) transmits dengue and chikungunya viruses among human hosts in cities at temperatures ranging 20-30 C, where it thrives in water that humans store in vessels. Two processes ar e known to drive the dynamics of A. aegypti production in these habitats: resource limitation and human be havior. Although temperature affects mosquito size and maturation rate, there is little understanding of how temp erature may affect the capacity of mosquitoes to develop in f ood limited, domestic environments. In this dissertation we use modeling, experimental and field ap proaches to explore the general hypothesis that the effects of human and resource-mediated processes on the size, development rate and pupation success of A. aegypti may be modified across temper ature gradients or cities that vary in temperature. We experimentally show that due to increased en ergy demands, larvae at higher temperature may experience tradeoffs between development rate, size and starvation resistance. Moreover, our model of growth and energy storage in mosquitoes indicates that these tr adeoffs can both explain the commonly observed effects of temperature on mosquito size and result in interactive effects of food and temperature on development rate. A food-temperature developmental interaction was morphologically corroborated by analyzing wing and epidermal cell size in experimentally 10

PAGE 11

reared A. aegypti Our data suggest that in resource poor habitats commonly inhabited by urban A. aegypti developm ent rate may be lower at 2426C in comparison to 20-22 or 28-30C, thereby challenging the fundamental assumptions of widely used temperature-driven models of A. aegypti dynamics. Our field study in dengue endemi c areas of Colombia indicates that the outcome of this food-temperature interaction de pends on specific household behaviors, including the frequency of using, emptying and placing lids on domestic wate r. Moreover, each of these acts in the context of particular socio-cultural perceptions, altitude, urban architecture and water supply. Nonetheless, our data suggest that climat e-induced decreases in wa ter security predicted throughout Colombia may indirectly increase A. aeypti abundance through changes in human behavior. These results demonstrate the underlyi ng interactions between human and mosquito responses to a changing environment a nd enhancing our capacity to predict A. aegypti production and design locally adapted intervention strategies. 11

PAGE 12

CHAPTER 1 INTRODUCTION The mosquito Aedes aegypti (L.) is a major global vector of the two most prevalent arboviral diseases in the world, de ngue fever (DF) and chikungunya fever, and continues to play a role in the transmission of urban yellow fever. This mosquito is uniquely adapted to environments where humans reside, both in aquati c stages that inhabit anthropogenic vessels and as adults that feed primarily on human blood. Human transportation networks and massive urbanization following the failed eradication ca mpaign in the 1950-60s have contributed to burgeoning A. aegypti populations and persistent DF infection in most neotropical cities (Gubler, 2002). Colombia, for example, reports an average of over 40,000 DF cases each year (National Institute of Health of Colombia, 2008), hyperendemic circulation of all four DF serotypes, and currently has the highest dengue hemorrhagic feve r (DHF) incidence rate in South America (Pan American Health Organization, 2008). Given th e costs and ineffectiv eness of insecticide spraying, reducing the rate of adult pr oduction in the container habitats of A. aegypti immatures remains the best prevention strategy for mitigatin g DF, for which no vaccines or treatments are available (Farrar et al., 2007). In Colombia close to 80% of the population liv es in plateaus and foothills of the Andes mountains, where endemic DF circulates in diverse climates with mean temperatures ranging from 21 to 30 C. As in many of the worlds neighbor hoods where DF is endemic, domestic water storage vessels are the primary aquatic habitat of A. aegypti in Colombia (Padmanabha et al., unpublished data; Romero-Vivas et al., 2006). Most field studies of this habitat associate two processes with spatio-temporal variation in A. aegypti production and adult size: larval food limitation (Strickman and Kittayapong, 2003; Arrivi llaga and Barrera, 2004; Morrison et al., 2004; Barrera et al., 2006b), and hu man water storage behaviors su ch as water use (Hammond et 12

PAGE 13

al., 2007), vessel em ptying (Subra, 1983), cleani ng (Chan et al., 1998) and covering with lids (Morrison et al., 2004). These processes may act independently or interactively by inducing starvation, extending larval development time and/ or eliminating aquatic stages before they emerge as adults. Although temperature is also a well documented determinant of development rate and size of mosquitoes, there is surprisingl y little understanding of how it may interact with the processes that regulate A. aegypti production in dengue endemic environments. Knowledge of how the effects of food limitation and human behavior are mediated by ambient temperature can lead to more effective climateand habi tat-specific intervention strategies to reduce A. aegypti production. Moreover, there is a need to develop understan ding of how mosquitoes will respond to a warming mean temperatures at different disease endemic altitudes. This study uses a combination of field surveys, experimental manipulations, morphological analyses and simulation modeling to investigate how water storage practices and larval feeding affect larval development in A. aegypti In particular, we ex plore how the effects of each is modified across a temperature gradient, focusing on the range of 20-30 C in which dengue is endemic in Colombia. 1.1 General Eco-Physiological Aspects of Development and Growth in A. aegypti Insect development is regulated by the endocri ne system as it responds to nutrient uptake and environmental conditions. The release of the hormone ecdysone triggers molting between immature stages and juvenile hormone (JH) titer controls the type of molt e.g., from larva to pupa or pupa to adult (Nation, 2008). Feeding and starvation experime nts with fourth instar (L4) A. aegypti indicate that cessation of JH secretion occurs after a minimal L4 feeding time and attainment of a critical mass, which is higher in females than in males (Chambers and Klowden, 1990; Timmermann and Briegel, 1999; Telang et al., 2007 ). If larvae are starved such that they 13

PAGE 14

do not attain this m ass, JH levels do not fall su fficiently to induce meta morphosis (Nishiura et al., 2007). In Drosophila melanogaster, the critical weight does not correlate with food availability, although this has not been directly studied in A. aegypti Increasing temperature increases A. aegypti development rate (Rueda et al., 1990) and decreases critical mass and final body size (Chambers and Klowden, 1990). A large body of evidence suggests that the level of stored energy is a key factor regulating A. aegypti development. Wigglesworth (1942) obse rved microscopically that starved L4 A. aegypti died when lipid droplets disappeared fr om the body (from Gilp in and McClelland, 1979). Depletion of lipids, but not overall mass, in starved A. aegypti has since been repeatedly confirmed (Gilpin and McClelland, 1979; Timmerma n and Briegel, 1999; Telang et al., 2007). In general studies employing biochemical assays of developing larvae have implicated energy reserves in diverse forms, including lipids and glycogen stores, in the mechanism that triggers pupation. (Gilpin and McClelland, 1979; Chambe rs and Klowden, 1990; Telang et al., 2007) Lipids also appear to play a role in defining the characterist ic sigmoidal weight trajectory in larval A. aegypti, with roughly 80% of growth occurring in L4 (Telang et al., 2007; Nishiura et al., 2007). When reared upon high prot ein laboratory diets, lipid c ontent as a percentage of body mass has been shown to remain relatively steady or slightly decrease as the L4 stage progresses (Timmerman and Briegel, 1999; Nishiura et al ., 2007; Telang et al., 2007). Moreover the lipid incorporation rate has been s hown to decrease during the L4 stage, particularly during the interval between attainme nt of critical mass and maximum larval size, when almost 50% of final size is attained (Timmerman and Briegel, 1999; Nishiura et al., 2007). This is supported by findings of Dye (1982), showing that experimental data on the time to maturation was best reproduced by a mathematical model when a te rm representing decreasing food assimilation 14

PAGE 15

efficiency with larval ag e was incorporated. T ogether these findings sugg est that energy storage may directly or indirectly regulate survival, pupation and growth in A. aegypti larvae. 1.2 Water Storage Behavior and A. aegypti Larval Development Water storage vessels are unquestionably an important larval habitat of domestic A. aegypti and are often the most abunda nt and productive urban contai ners in a particular area (e.g., Southwood et al., 1972, Reuben et al., 1978; Barre ra et al., 1993; Vu et al., 2005; Bisset et al., 2006; Romero-Vivas et al., 2006; Burkot et al., 2007; Hammond et al., 2007; Koenraadt et al., 2008). Larvae in water storage containers experience a number of unique processes related to human water storage practices that do not occur in other container habitats of mosquitoes, such as discarded tires, that passively receive rainfall The first is that they are purposely filled and drained by humans and experience a much larger fluctuations in wa ter level at daily or weekly timescales (Koenraadt et al., 2008). This provides a much larger hatching stimulus for eggs, increasing the frequency of larval infestation across vessels and pot entially reducing intraspecific resource competition by decreasing the size of larval cohorts that hatch synchronously. The second is that the processes of water extraction, emptying and refilling generate a highly unstable and often nutrient poor environment for larval populations (Subra and Mouchet, 1984; Arrivillaga and Barrera, 2004) that is frequently subject to stochastic collapses when containers are emptied (Subra, 1983). Frequent emptying may eliminate aquatic stages before they emerge as adults, thereby reducing the rate of A. aegypti production (Subra, 1983). Other water storage behaviors, such as water use (Hammond et al., 2007), clea ning (Chan et al., 1998) and covering with lids (Morrison et al., 2004) may reduce mosquito production by limiting nutrient input, oviposition or egg retention on cont ainer walls. However, not all of these are 15

PAGE 16

am enable for mosquito control in particular water storage contexts. For example, frequent water usage may undermine the effects of lid placement and result in higher mos quito infestation rates because lid placement may be less consistent in vessels where water is frequently extracted (Phuanukoonnon et al., 2005; Hammond et al., 2007). Residents may be unwilling to empty vessels if water is stored primarily to compensa te for unpredictable interru ptions in piped water (Barrera et al., 1993). These water storage dynamics may contribute to the recurrent finding th at in residential, urban areas A. aegypti production is highly variab le at small spatio-tem poral scales. Spatially, virtually all studies of pup ae find that the majority of pupae are generated by only a few sampling units (days, containers or houses) whereas the majority of sampling units are not productive (Subra, 1983; Focks et al., 1995; Bisset et al., 2006; Koenraadt et al., 2008). Getis et al. (2003) found that, while pupae clus ter in containers within spec ific houses, la rval infested vessels in Peru was randomly distributed acro ss houses, suggesting that the determinants of dengue vector production reside within the household level. Simila rly, Subra (1983) corroborated that in a Kenyan village, househol d behaviors were responsible fo r the large daily variation in pupal production. While container maintenance may cause the majority of water storage vessels in a particular community to be devoid of organic matter and/or A. aegypti immatures, variation in household-container interactions may result in a few vessels that prod uce large numbers of pupae (Subra, 1983). Clearly a deeper understanding of human-cont ainer interactions is essential for determining sustainable community-based interv entions to reduce mosquitoes. Assessment of relationships between human behavior and vect or production through th e widely used crosssectional pupal/demographic surveys (Focks a nd Chadee, 1997) is problematic, however. When 16

PAGE 17

residen ts of communities with a history of dengue transmission and A. aegypti control programs are visited by vector control pe rsonnel, their self-perceptions, desi re to please the interviewer, and the difficulties in quantifying sometimes unconscious interacti ons with containers all limit the accuracy of self-reported behaviors in surveys. (Scrimshaw, 1990; Kasprzyk, 2005) Moreover, surveys with predetermined variables may not reveal the subjective experience of individuals in a particular social and physical context that influen ces their behavior (Golafshani, 2003). In order to understand the ecol ogical coupling of the dynamics of A. aegypti production and human behavior, longitudinal studies of hous eholds with direct, repeated observation of human behavior, such as those carried out in a Kenyan village in the 1970s (Subra, 1983; Subra and Mouchet, 1984), need to be carried out in modern, dengue endemic urban areas. 1.3 Food Limitation and A. aegypti Development There is ample eviden ce indicating that A. aegypti is frequently limited by the resources available in larval habitats. This has been observed in field populations of A. aegypti by comparing body size/mass indicators of fiel d-collected pupae with those of laboratory populations reared across food gradients (TunLin et al., 2000; Strickman and Kittayapong, 2003; Barrera et al., 2006a). Domestic A. aegypti larvae experience starvation in nature (Arrivillaga and Barrera, 2004) and have been show n to differ from its subspecies, the sylvatic A. aegypti formosus, in the increased ability of larvae to withstand starvation (Gilpin and McClelland, 1979). Other studies have correlated larval food proxy variables such as exposure to leaf drip, rainwater and human food particles to A. aegypti pupal production in field containers (Subra and Mouchet, 1984; Morrison et al., 200 4; Barrera et al., 2006b; Hammond et al., 2007), indicating that larval food resources may limit adult population size. While food limitation is common in natural A. aegypti habitats, the extent to which it is driven by the depletion of res ource by competing larvae is un clear. Broadly speaking, food 17

PAGE 18

lim itation arises whenever larv ae incorporate into biomass le ss food than their physiological capacity to process ingested food. Food limitatio n may arise without resource competition in various situations. If water volum e is high and resources are presen t but diluted or aggregated to the extent that larvae spend most of their time and energy searching for food, then food limitation may be density-independent (Legros et al., 2009). For example, Subra and Mouchet (1984) found that addition of larval food dram atically increased pupal production in natural habitats, whereas addition of larvae did not affect production. Rash ed and Mulla (1989) demonstrated that A. aegypti larvae ingest inert particles more than other mosquito species. This suggests that in habitats w ith abundant inert particles, A. aegypti larvae may saturate their guts to the point where food consumption is driven by the rate of digestion and excretion. Therefore if larvae do not readily distinguish between nutri tive and non-nutritive pa rticles and the matter ingested is largely inert indige stible particles, f ood limitation, as measured by starvation or body size, will arise independen tly of larval density Few field studies have directly demonstrat ed density-dependent resource competition. In containers found in an enclosed Buddhist monastery in Thailand, S outhwood et al. (1972) enumerated all life stages of A. aegypti monthly over a 1-year period. The authors used life table methods to infer that density-dependent mortality from egg to the sec ond instar regulated the population dynamics. Barrera et al. (2006a) found that larval density was positively correlated with pupal output, but negatively correlated with size, indicating that competition was strong enough to reduce larval growth in the interval between attaining critical a nd maximum mass, but not strong enough to induce mortality. Where it doe s exist in the urban Neotropics, competition affecting A. aegypti is usually intraspecific, notwithstan ding occasional co-occurrence of this 18

PAGE 19

species with Culex quinquefasciatus in large containers with high or ganic content, such as storm drains and sewers (Legros et al, 2009). Using the results of Southwood et al. (1972), Gilpin and McClelland (1979) constructed a detailed model of A. aegypti production that successfully reprodu ced effects of variation in foodlarval ratio on pupation success in labo ratory experiments. They modeled A. aegypti growth using an exponential function in which the fraction of larvae that cross the critical mass threshold increased with the food:larvae ratio until the syst em become saturated with food at a maximum pupation rate. In further experiments on inte rference and resource competition within and between larval stages, Dye (1982, 1984) largel y corroborated the results of Gilpin and McClelland. The competition-driven formulati on of larval growth used by Gilpin and McClelland served as the basis for subsequent models inte grating temperature effects on A. aegypti production (Focks et al., 1993; Jetten and Focks, 1997; Williams et al., 2008; Kearney et al., 2009). This formulation, however, assumes that effects of food limitation are temperature independent and therefore cannot be applie d to study food-temperature interactions. 1.4 Temperature Effects on A. aegypti and Ectotherm Development In A. aegypti the effects of water temperature on th e development rate of aquatic stages and on adult size are well documented (Rueda et al., 1990; Tun-Lin et al, 2000). Temperature effects on development rate have been describe d by an enzyme kinetics model which assumes that development rate is controlled by a single ra te-limiting enzyme that is denatured reversibly at high and low temperatures (F ocks et al., 1993). This mode l has been shown to explain temperature effects on development rate in a wide range of ectotherms (Sharpe and DeMichele, 1977; Schoolfield et al., 1981), and is consitent with field and laboratory studies show that A. aegypti development rate increases linearly with te mperature in the 15-35C range, with a sharp increase in mortality below 20 C and above 30 C (Rueda et al., 1990, Tun-Lin et al., 2000). 19

PAGE 20

Studies also show that adult si ze increases m onotonically as temp erature decreases in the range of 15 to 35C (Rueda et al., 1990, Tun-Li n et al., 2000). The critic al mass for pupation also varies inversely with temperat ure (Chambers and Klowden, 1990). The enzyme-kinetics model forms the basis of temperature effects on vector development in the weather driven model, CIMSIM (Focks et al., 1993; Focks et al, 1995; Jetten and Focks, 1997). While the model predicts a m onotonic increase in productivity in A. aegypti between 20 and 30C, field studies do not reveal a consistent pattern. In two studies of productivity across individual containers in Puerto Rico (Barrera et al., 2006b) an d Australia (Tun-Lin et al., 2000), temperature in the 22-30C range was negatively associated with productivity, and both studies found that resource competition was the principa l regulatory factor. Conve rsely, Favier et al. (2006) found that the number of pupae per positive container was correlated with seasonal temperature in an area of Southern Brazil wi th a large (~ 6C) seasonal fluctuation in temperature. The pattern followed by A. aegypti of faster development to a smaller final size with increasing temperature, dubbed the temperature-size rule (TSR) in ectothems, is one of the most universal observations in biology (Atkinson a nd Sibly, 1997); yet the und erlying developmental processes that generate this phenomenon in ectot herms remain controversial, and the issue has not been directly addressed in mosquitoes. For example, a mechan istic explanation is lacking for the widespread observation that while increased food and lower temperature both increase final ectotherm size, they have opposite effects on maturation time. VanderHave and deJong (1996), for example, postulated that the underlying m echanisms controlling cell division (mitosis) are different from those controlling differentiation (maturation rate), and thus environmental effects on body size may be reflected in the relative diffe rences in cell size and number, According to 20

PAGE 21

their m odel, which also assumes that temperature increases the activity of a single-rate limiting enzyme (Sharpe and deMichele, 1977), the TS R should apply if the process of cellular differentiation is more sensitive to temperature than cellular growth. This model predicts that size increases at lower temperatur es should be generated by increases in cell size rather than cell number. Indeed, in a review of 12 studies on allo metric growth in insects, 10 of which were carried out on D. melanogaster it was concluded that, temperature generally affects body size through larger effects on cell size (Arendt, 2007). There is no reason to believe that these results apply in mosquitoes. However, they suggest that investigation of the allo metry of body vis--vis cell size in mosquitoes may shed light on differential impacts of temperature and food intake. An important question in vector borne dis ease is how mosquitoes will meet increased energy demands in a warming climate (Laffert y, 2009). However the assumption that size is determined exclusively through a single temperatur e-dependent enzyme, as used in prior models of A. aegypti immature dynamics (Focks et al, 1993), neglects potential interactions of temperature with energy acquisition and metabolism. An alternative group of models examines the TSR through the dynamic effects of temperature on catabolism and anabolism (Bertalanffy, 1960; Berrigan and Charnov, 1994; Perrin, 1995). Bertalanffy (1960) proposed a simple equation to outline the growth rates of organisms: n mwwdtdw /, where w is the mass of the organism, and are the coefficients of anabolism and catabolism and m and n are the exponents that desc ribe the weight dependence of anabolism and catabolism. Studies generally find that these exponents range between 0.5 and 1, indicating that anabolic and catabolic activity increases less than linearly with heightened weight. By assuming that the exponents m and n are constants, Perrin ( 1995) showed that the TSR can only arise if catabolic 21

PAGE 22

coef ficient is more sensitive than the anabolic coefficient to increasing temperature and cited empirical support in studies of fish. Moreover, by assuming that energy and mass are directly proportional, Angilletta and Dunham (2003) showed that, for the Bertalanffy-Perin explanation of the TSR to hold, growth efficiency must decrea se with increasing temperature. Recently, Karl and Fischer (2008), using detailed laboratory experiments, showed that both food intake and assimilation efficiency are greater at lo wer temperatures in the copper butterfly, Lycaena tityrus. However, in a review of 97 studies across a rang e of taxa, it was found that growth efficiency was overwhelmingly positively correlated with temperature (Angilletta and Dunham, 2003). The authors conclude that empirical studies more strongly support the hypothesis that temperature effects emerge in the norm of development th at is, temperature modifies the trajectory of resource acquisition and us age through the exponents m and n in the Bertalanffy equation (Angilletta and Dunham, 2003). By affecting the exponents in addition to the coefficients of growth, temperature may cause the anabolic-catabo lic balance to vary between the early and later stages of development. In A. aegypti this conclusion is supported by experimental work of, Rashed and Mulla (1989), who found that the food consumption rate per A. aegypti larva increased by 50% between 18C and 31C, suggesting that the Perin paradigm does not apply. Overall, these studies suggest that the way in which ectotherms allocate resources is a crucial aspect of the mechanisms determining how size and development rate change with increasing temperature. (Bochdanovits and De Jong, 2003; Kozlowski et al., 2004) 1.5 Hypotheses and Research Questions In dengue endemic areas of Colombia, climat e warming may have complex interactions with larval ecology, as temperature directly affects the same principal outcomes as food intake and human behavior: survival, development time and growth. Because of its adaptation to low resource habitats, domestic A. aegypti in particular may exhibit unique interactions between 22

PAGE 23

tem perature, food consumption and growth. These bi ological interactions o ccur in a sociological context of household behaviors towards domestic container habitats, which in turn, may vary across climates. Given that A. aegypti populations thrive between 20-30 C, the temperaturespecific response to increasing en ergy demands is likely to determine how this species responds to climate change. Since accumulation of energy stores are likely involved in the physiological process leading to A. aegypti maturation, which occurs sooner at increased temperature, it is unlikely that smaller body size at higher temperat ures comes at the cost of energy storage. Ultimately, the impacts of climate change on dengue control will depend on how A. aegypti responds to the increased ener getic demands of higher temper ature (Lafferty, 2009), and how this response modifies the effect s of water storage practices on A. aegypti production in specific ecological and social contexts. In chapters 2-4 we explore the genera l hypothesis that temperature impacts on A. aegypti development are mediated by interactions with hu man behavior and feeding rate over the course of development. In particular, we address ea ch of the following questions using either experimental temperature manipulations or a natural altitude gradient in Colombia: 1. How do household behavior s affect the rate of A. aegypti production and how are these relationships modified by the motivations for st oring water in the context of an altitudegenerated temperature gradient? 2. How do simultaneous manipulations of feeding rate and temperature differentially impact the size and number of epidermal cells in A. aegypti wings? 3. How do 2-degree deviations in temperatur e from 20 to 30C impact tradeoffs between development rate, growth and starvation resistance in A. aegypti ? 23

PAGE 24

CHAPTER 2 ECOL OGICAL LINKS BETWEEN WATER STORAGE BEHAVIORS AND Aedes aegypti PRODUCTION: IMPLICATIONS FOR DENGUE VECTOR CONTROL IN VARIABLE CLIMATES 2.1 Introduction Throughout the world domestic vessels used fo r water storage are frequently the most abundant and productive habitats of immature stages of Aedes aegypti (L.), the principal vector of dengue fever (DF) and urban yellow feve r (Southwood et al., 1972; Reuben et al., 1978; Barrera et al., 1993). This mosquito is uniquely influenced by human water storage, which, in turn, may affect indices of dengue vector abundance (Morrison et al., 2004; Barrera et al., 2006a; Hammond et al., 2007). Accordingly, understanding how specific patterns of water usage affect A. aegypti larval ecology may be a key step in identif ying specific behaviors to target in dengue prevention interventions. (Elder and Lloyd, 2007) Modifying human behavior for dengue control requires knowledge of local customs of water storage and how the resulting behaviors affect the processes that generate adult A. aegypti While behaviors such as water use (Hamm ond et al., 2007), vessel emptying (Subra, 1983), cleaning (Chan et al., 1998) and covering with li ds (Morrison et al., 2004) may reduce mosquito production, their implementation may not always be practical in particular water storage contexts. For example, frequent water usage ma y undermine the effects of lid placement and result in higher mosquito infestation ra tes in lidded vessels (Phuanukoonnon et al., 2005). Residents may be unwilling to empty vessels if water is stored primarily to compensate for unpredictable interru ptions in piped water (Barrera et al., 1993). Among the unique processes that preimaginal mosquitoes experience in water storage vessels is a higher probability that aquatic stages are washed away by containe r emptying (Subra, 1983). Emptying, however, will only be effective in reducing A. aegypti production if it occurs before immature stages complete 24

PAGE 25

developm ent to adult. Thus, from an ecological perspective, emptying an d associated behaviors may interact with conditions affecting development time, including temperature and food availability (Subra and Mouchet, 1984; Arrivillaga and Barrera 2004; Barrera et al., 2006b). Traditional entomological surveys are not suitable for correlating wa ter storage practices with A. aegypti production. Respondents' self-perceptions, their desire to pl ease vector control personnel, and the difficulties in quantifying tempor ally variable actions towards containers all limit the accuracy of self-reported behaviors in surveys. (Scrimshaw, 1990; Kasprzyk, 2005) Moreover, surveys with predeter mined variables may not reveal the subjective experience of individuals in a particular social and physical context that influences their behavior (Golafshani, 2003). From a biological perspectiv e, one-time pupal counts may al so provide biased estimates of productivity across heterogeneous climates be cause higher temperature shortens the pupal stage, and pupae observed in warmer areas comple te larval development fa ster than counterparts in colder areas (Christophers, 1960). Repeated household visits can address these difficulties (Subra, 1983), by averaging pupation across a fixe d interval and characterizing human behavior through a combination of direct, repeated observa tion and open ended interviews (Trotter et al., 2001) In Colombia, dengue viruses persist at altitudes from sea level to 1600m, where A. aegypti immature stages experience average temperatures ranging from 30-20 C in household containers. This temperature range is associated with larval developmental completion between 9-4 days under optimal feeding conditions (Rue da et al., 1990) and may cause the impact of human behaviors on pupal production to vary across ci ties at different altit udes. In this study we tested this hypothesis using a l ongitudinal (7 to 15 day) study of 235 households with larval infested water storage vessels in three Colomb ian cities 5, 950 and 1550 meters above sea level 25

PAGE 26

(m sl). In particular, we characterized motivatio ns for household water usage and the effects of water storage practices on the rate of A. aegypti production in order to suggest the behaviors most likely to reduce A. aegypti production in water storag e vessels in each city. 2.2 Materials and Methods 2.2.1 Study Area This study was conducted in two DF endemic neighborhoods in each of the cities Armenia (1550 msl), Bucaramanga (950 msl) an d Barranquilla (5 msl). All six neighborhoods had a socio-economic ranking of 2 (on a 1 to 6 scale of each municipal planning department), characterized by low-income, planned housing un its. In 2007 all six neighborhoods had a stable piped water supply with occasional interruptions Routine vector control in 2007 varied across cities: in Armenia there were no known interventions; in Bucaramanga, adulticides were applied in homes within 100 m of dengue cases and larv icides were applied to storm drains; in Barranquilla there was a city-wide mass communication progra m to promote cleaning, scrubbing and lid placement in domestic vessels. Data collection in each city was conducted in January, June and October 2007. Ten-year mean ambient temperatures ( C) and relative humidity (%) measur ed in airport weather stations for each of these respective months are: Armenia 20.3 (79.7%), 20.3 (81.4%), 20.0 (82.7%); Bucaramanga 23.3 (81.4%), 23.3 (84.9%), 22.9 (85.3%); Barranquilla 26.8 (78.4%), 28.3 (81.4%), 27.6 (85.4%) (Institute of Environmenta l Studies of Colombia, 2009). It should be noted, however, that the weather stations for Ar menia and Bucaramanga differ in altitude as compared to the study neighborhoods, which causes temperature to differ between airports and our study areas (see below). Mean cumulative monthly rainfall (mm) for January, June and October is the following: Armenia 144.0, 183.7, 278.8; Bucaramanga 78.3, 99.4, 93.6; Barranquilla 1.7, 80.2, 152.8 (Institute of Environm ental Studies of Colombia, 2009). Seasonal 26

PAGE 27

clim ate patterns are not consistent across cities although January is the driest of the three survey periods in all three. Based on their high rates of dengue persistence (2004 to 2006) neighborhoods were selected as part of a separate study of human ecology and A. aegypti production (not reported here). This study consisted of tri-annual surveys of 7159 households (35-50% of each neighborhood) and their water-holding containers, carried out in th e same months as the present study. Preliminary 2007 data from this study show ed that water storage vessels accounted for over 90% of all A. aegypti pupae in each city, with Breteau Indices ranging 18.3-28.1 infested vessels/100 premises inspected in Arme nia, 11.5-24.4 in Bucaramanga and 6.7-13.2 in Barranquilla. 2.2.2 Container Selection For the present study we selected actively-fille d water storage vessels found infested with A. aegypti larvae or pupae in the triannual surveys described above and carried out visits every other day over 10-15 days after detecting the vessel. In each survey period we sought authorization to conduct the study in all premises with a larval or pupal infested vessels within a subset of 240-360 houses in Bucaramanga, 360-480 houses in Armenia and 600-720 houses in Barranquilla. Sampling effort was determined based on available manpower in each city. Houses with more than one infested actively filled water storage vessel (found rarely and only in Barranquilla) were excluded from the study in or der avoid the added complication of assessing potential variation in behavior towards multiple vessels, and no houses were repeatedly followed in different study periods. Residents were informed that particip ation involved receiving a visit every two days in order to count all mosquito pupae and to qu ery their container use in the previous 48 hours. The minimum inclusion criterion was at leas t three visits over a minimum 27

PAGE 28

seven-day interval. Containers not surveyed on two consecutive visits because residents were absent at the tim e of our visit we re eliminated from the study. 2.2.3 Physical Description of Houses and Study Vessels In houses of all six neighborhoods water storage vessels were located in a concrete wash area in the patio an enclosed semi-roof covered ar ea towards the rear of the house. In Barranquilla, patios range from 5 up to 30 m2. This sharply contrasts with Bucaramanga and Armenia, where patios range 1-5 m2. All vessels studied in Bucar manga and Armenia were at least partially roof-covered and re ceived little or no sunlight a nd rainfall. In Barranquilla, the warmest city, 46% of study vessels were classified as without roof c over but received limited direct sunlight due to thei r proximity to the houses. Ninety-two percent of study vessels in Armenia and Bucarmanga were lavaderos, or washbasins and faucet with a capacity of 100-150 L, an opening of approximately 800 cm2 and an attached scrubbing board, elevated approxi mately 1m above the floor (Figure 2A). Lavaderos have a removable drain to easily empty water. In Barranquilla homes were not constructed with a built in washbasin, so residents obtained or constr ucted water storage vessels. Combined with the larger patio size, this resulted in larg er variability in Barranquilla vessel types, including 500 L cement tanks, 200 L metallic and plastic drums, 20-50 L plastic buckets /jugs and occasional > 1000 L cement albercas (pools) (Figure 2B). However, because we exclusively studied actively filled water storage vessels, observed water volume varied on a daily basis and was almost entirely dependent on household use patter ns, given high shade cover and thus low evapotranspiration. 28

PAGE 29

2.2.4 Data Collection 2.2.4.1 Water temperature Digital tem perature loggers (Embedded Data Systems) were placed in water storage vessels in each neighborhood over 10-15 days during the June 2007 surveys. Loggers were placed in households where we felt there was mi nimal risk of loss or theft; there was no indication of any systematic differences in water temperatures between these and our study households. Eight loggers were placed in Bucaramanga, and 12 in Barranquilla and Armenia, respectively. ANOVA and pairwise T-tests were employed to comp are mean temperatures from hourly recordings in each contai ner across cities. Repr esentative of conditions in study vessels, all Bucaramanga and Armenia loggers were placed in roof-covered lavaderos whereas 50% of the Barranquilla loggers were plac ed in plastic and metal drums pa rtially exposed to sunlight and the other 50% in large cement basins. All ve ssels contained at least 20 L of water, 2.2.4.2 Entomological surveys All pupae were collected upon discovery to avoid the possibility of re-counting on successive visits, and all mosquito immatures we re eliminated at the end of the study. Pupae were returned to the public health laboratory of each municipal health department and allowed to emerge to adult in order to confirm species identification. For each vessel the mean two-day rate of production was defined as the total number of A. aegypti pupae collected over the study period divided by the number of observatio ns (visits in which entomol ogical survey was carried out). 2.2.4.3 Human behavioral assessments Behaviors were characterized through semi-s tructured interviews conducted in each visit with a household member that interacted frequently with vessels. On each visit we carried out an informal interview of residents, allowing them to express the reasons for using stored water and the nature of the use. Studies show that such inductive data collection reduces self-reporting 29

PAGE 30

biases associated with structured survey with pre-determ ined variables that subjects may subjectively interpret (G olafshani, 2003). Ethnographic narratives (Caprara et al., 2009) were employed in order to qualitativel y describe the motivations underl ying water storage behaviors in each city. Three behaviors were consis tently recorded for each c ontainer-household: whether the water was used (self-reported) or emptied (sel f-reported triangulated with observation) in the prior 48 h, and whether the vessel was lidded (obs ervation triangulated with self-reporting). 2.2.5 Data analysis Given the stochastic nature of human behaviors and short follow-up time (7-15 days) per vessel, emptying interval and usage were grouped into discrete levels, using biologically and sociologically relevant cutoffs. Wa ter usage (extraction for domestic use) was classified into the following categories of decreasing use: (1) every visit, (2) interm ittently or (3) never. Increasing water emptying interval was classified as: (1 ) < 7 days between occu rrences, corresponding roughly to average larval development times a bove 24C under optimal food conditions (Rueda et al., 1990); (2) 7-15 days be tween occurrences and (3) no em ptying recorded in the study. Usage of a container lid (cove ring entire vessel opening) was dichotomized. Spearman rank correlations and contingency table ( 2) analyses were used to corre late behaviors in each city. 2.2.5.1 Explanatory variables entered in model Because of the potentially complex causal pathways between behavior and physical variables such as container loca tion, volume, debris, microclimate and others (see discussion) we limited independent variables to usage (categori cal), emptying (categorical), lid placement, survey period, city and the interac tion terms Barranquilla*emptying7-15days and Armenia*emptying 7-15days. Based on prelim inary data inspection, we included these interactions in order to test fo r a non-linear effect of emptying on production in the vicinity of 30

PAGE 31

the tim e to pupation, as minimum larval development period (under optimal laboratory conditions unlikely to be met in the field) is 8 and 4 days at the temperatures of Armenia (21.9C) and Barranquilla (28.3 C), respectively (Rueda et al., 1990). 2.2.5.2 Model building and selection Based on the distribution of the dependent variable, mean pupae/two-day interval/vessel, we used negative binomial regression in ST ATA 8.0 (StataCorp. 2001. Statistical Software: Release 8.0. College Station, TX: Stata Corporation). A saturated multivariate model was constructed using the explanatory variables listed above Potential variation in the aggregation parameter ( ) [as defined in variance equation for th e negative binomial model (Eq. 2-1)] across city (Table 2-1) and behavioral (Table 2-2) strata, was accounted for by employing the generalized negative bi nomial regression ( gnbreg ) command. This procedure tested whether specification of across one of these groups improved th e saturated model by parameterizing the equation (2-2). All variables were entered into Equation 2-2 as dichot omous (dummy variables for each category in ordered variables). The al pha parameterization that most improved the overall saturated model was then maintained in deriving best-fitting models. 2 2xxs. (2-1) Xaa10ln (2-2) We selected the best fitting model using th e Akaike Information Criterion (AIC), defined as, LLkAIC 22 (2-3) where k is the number of free parameters and LL is the log-likelihood (Hobbs and Hilborn, 2006). The minimum AIC model was selected among the 20 combinations of main effects and 31

PAGE 32

interaction term s specified above with the highe st likelihoods. Uncertainty in model selection was evaluated using AIC weights (wr) defined as, R i r reewr1 2 2/ (2-4) where r is the difference between the AIC value of model r and the lowest AIC model, and R is the total number of models considered. The index wr can be interpreted as an indictor of the odds that model r would emerge as the best model among the attempted models given a different dataset (Hobbs and Hilborn, 2006). 2.2.5.3 City-specific models City-specific models were constructed using the same process and alpha parameterization as the overall model. Potential non linear-effects of emptying interval in th e vicinity of larval development period were explored by inserti ng emptying < 7 days and emptying 7-15days as dummy variables for comparison against non-emp tied vessels, instead of the categorical approach used in the overall and univariate analyses. 2.3 Results 2.3.1 Study Containers, Water Storage Practices and A. aegypti Production Overall, 235 water storage vessels infested with A. aegypti larvae or pupae were included in the study. Mean production was similar in Armenia and Barranquilla, but lower in Bucaramanga (Table 2-1). A similar proportion of vessels across cities had missing data due to absence of residents. Mean water temperatures coincided with altitude and were significantly different between cities (F(2) = 212.3 p<0.001). There were also si gnificant pairwise differences between the two colder cities, even after a Bonferonni adjustme nt of the significance level to /n, or 0.0025 (tbucaramanga-armenia = 6.84 p<0.001). Variation in manpow er, infestation levels, resident cooperation and follow-up success all contributed to the observed differences in sampling effort 32

PAGE 33

across cities. Overall, 5.3 observations were made per container, corresponding to a mean follow up interval of roughly 11 days (Table 2-1). Water in 83.4% of the vessels was extracted for domestic use at least once during the follow-up period and 80.4% lacked a lid (Table 2-2). Forty-two per cent of the vessels were never emptied and these produced 67.2% of the total A. aegypti pupae recovered; 92% of total pupae were produced in vessels emptied at least every 7 days. Producti on rate was similar across water use groups and was lower in lidded vessels (Table 2-2). Follow-up interval was similar across behavioral strata. 2.3.2 Qualitative Assessment of Behavior and Motivations We observed strong relationships between water storage behavior and container structure. In Armenia and Bucaramanga, residents stored water primarily for the convenience of using the easily accessible washbasin instead of tap water to wash clothes and, secondarily, for cleaning floors, flushing toilets, watering plants, or in ca se of interruptions in piped service. Potential interruption in piped supply was the major factor influencing water storage in Barranquilla, however no clear interactions betw een interruptions in piped serv ice and water storage practices were observed. In Barranquilla elderly residents, who lived most of their lives without piped water, were often those that had the custom of st oring water just in case Residents specifically mentioned an interruption in piped service as th e reason for water use on only 11 occasions in 10 different houses (6 in Barranquilla and 4 in Buca ramanga). In all cities the large majority of vessels followed contained exclusively tap wa ter, and no vessel followed was exclusively rainwater. Container structure influenced lid placem ent. Metallic and plastic drums (200 L), observed only in Barranquilla, often had tailor-made lids, as did 1000 L cement/plastic tanks (Figure 2-2B). In some instances residents in Barranquilla improvised lids on 1000 L permanent 33

PAGE 34

basins with household m aterials such as wooden boards, plastic bags, cardboard or roof tiles. Lavaderos such as Figure 2-2A were seldom lidded. 2.3.3 Quantitative Assessment of City-Specific Behaviors Container lid placement was less frequent in Ar menia than in the other two cities (Table 2-3). On average water was used less freque ntly in Barranquilla than in Armenia and Bucaramanga, whereas the emptying interval in Barranquilla was almost double that of the Andean cities (Table 2-3). Twenty seven per cent of vessels in Barra nquilla were never used during the observation period, as compared to 6.9% and 9.3% in Armenia and Bucaramanga, respectively. Water usage and emptying were sign ificantly correlated (p<0.05) in all 3 cities (Table 2-3), although the correlation was much higher in Barranquilla (Table 2-3). In all three cities, containe rs surveyed in the dry seas on (January 2007) had a notably higher mean interval between emptying events (Tab le 2-4). In Barranquilla, this also coincided with decreased usage, whereas there was no appa rent pattern of water usage across surveys in Bucaramanga and Armenia. Production rate, partic ularly in Armenia and Barranquilla, was also higher in the dry season survey (Table 2-4). 2.3.4 Effects of Water Storage Behaviors and City on A. aegypti Production The mean rate of mosquito production acr oss vessels was well approximated by a negative binomial distribution ( = 3.35, 2 (1) = 6505, p<0.01, rejecting the null hypothesis that =0 in Equation 2-2). Among univariate models, in creasing emptying interval best described the variation, followed by lid placement and survey period (Table 2-5). Vessels in Bucarmanga were associoated with lower pupation as compared to the other two cities (negative coefficient in Table 2-5). Decreasing use fre uquency, Armenia and Barranquilla had higher AICs than the intercept only model (Table 25). In univariate models, each of the three cities was dummy 34

PAGE 35

coded as a d ichotomous variable in order to co mpare the effect of each individually against the data from the other two. The log-likelihood (LL) of the standard nbreg saturated model with all behaviors, interaction terms and city effects was -694.3 (A IC=1404.6) as compared to -690.9 (AIC=1399.8) when gnbreg was employed using the equation ln( ) = 0.86 + 0.67(emptying < 7days), indicating that the data were best describe d when the model took into account increased aggregation of production in containe rs emptied < 7days. Specification of for other behaviors, categorical emptying interval, or cities generated smaller or no improvements in AIC as compared to the model with unspecified alpha; therefore all subsequent multivariate models (Tables 2-6 and 2-7) employe d this parameterization of Less frequent emptying was associated with increased production. Barranquilla*7-15day emptying had a positive and Armenia*7-15day a nega tive effect, with the latter having a 0.07 probability of being selected in the be st-fitting model based on AIC weights (wr) (Table 2-6), whereas the former was the stronge st predictor in the best fitti ng model. In the Barranquilla only model, emptying < 7days had a strong negativ e effect on production (Table 2-7), whereas emptying 7-15 days did not improve the description of the data (Table 2-6). In Bucaramanga, inclusion of both emptying intervals improved the prediction of production ra te as compared to non-emptied vessels (Table 2-6), but emptying < 7 days had a 3-fold greater effect in reducing production than emptying 7-15 days (Table 2-7). In Armenia em ptying < 7 and 7-15 days had a similar effect in reducing producti on (Table 2-7). In Barranquilla production rate is similar in vessels not emptied and emptied 7-15 days (Fig ure 2-3), whereas in Armenia, production is similar in vessels emptied <7 days and emptie d 7-15 days. In Bucaramanga, production appears to linearly increase with emptying category (Figure 2-3). 35

PAGE 36

Water usage frequency did not significantly improve the model fits of the overall data (Table 2-6). Decreased usage was slightly associ ated with reduced production in the best-fitting models for Armenia and Barranquilla (Table 2-7), but the same models w ithout the usage term had large wr values (0.29 and 0.35, respectively) i ndicating a relatively hi gh probability that water usage would not appear in the best-fitting model in an other data set (Table 2-6). Lid placement was associated with reduced production overall, but only improved model fits for Barranquilla (Tables 2-6 and 2-7). Includ ing survey period improved model fits for the overall data set, with wr = 0.70 as compared to wr = 0.10 when this variable was omitted (Table 2-6). Survey period remained in the best-fi tting models of Barranquilla and Bucaramanga, but with opposite effects (Table 2-7). 2.4 Discussion Human behaviors generate unique habitats fo r mosquito development in domestic water storage containers. Since recognizing important behaviors is a key step in designing communitybased A. aegypti control strategies (Elder et al, 1998; Elder and Lloyd, 2007), there is a need to understand how water storage prac tices interact with human a nd mosquito ecology. For dengue prevention policy, the results of this study suggest that in defining community-based interventions, vector control programs should include assessment of local ecological factors affecting larval development rate, prevalent wate r storage behaviors, and how the latter interacts with environmental features such as container structure and seasonality. Our results are consistent with a non-lin ear relationship between pupal production and household water emptying interval, whereby the impact of emptying A. aegypti habitats every 715days on reducing production may decrease in citi es with increasing mean temperature within the 20-30 C range. In Barranquilla (29.3 C) the effect of emptying on production was strongly 36

PAGE 37

modified upward in the 7-15day interval, whereas in Arm enia there was a weak modification towards lower production in this interval (Table 2-6). Moreover in Armenia, the effects of emptying < 7 and 7-15 days as compared to non -emptied vessels were approximately the same, whereas in Barranquilla 7-15day emptied vessels had similar produc tivity to those that were never emptied; Bucaramanga (23.9 C) demonstrated significant in creases in productivity across the 3 emptying levels (Figure 2-3, Table 2-7). Thus, the impact of human emptying on vector production appears to be dependent on city-specific ecological fact ors such as temperature that may affect development rate, such that small changes in emptying interval may have a large impact on A. aegypti output. In Australia human adaptati on to increased drought has been shown to drive A. aegypti infestation more than the direct temp erature effects (Beebe et al., 2009). Our results suggest that emptying frequency was lower in the dry season survey which would indicate that climate change-induced drought s in Colombia may increase A. aegypti production via human adaptation. However, st udy households were not selected randomly nor revisited in different seasons. Future studies should focu s on human perceptions of climate variability, seasonality and long term climate change, and how these affect container interactions. We suggest that temperature-driven models of A. aegypti production (Focks et al., 1993, Jetten and Focks, 1997) could be improved by incorporation of human beha vioral dynamics. We found that container struct ure and location in the househ old environment shapes the underlying motivations for water st orage and behavior towards vessels. The cement washbasin, present in virtually every house in Armenia a nd Bucaramanga, allowed residents to handwash clothes using stored, instead of tap, water. Accordingly, over 90% of residents in Armenia and Bucaramanga used stored water in comparison to only 63% in Barranquilla. Households that frequently used vessels were more apt to regular ly empty water, but frequent use may make lid 37

PAGE 38

usage m ore cumbersome (Phuanukoonnon et al., 2005). In contrast, emptying was less frequent in households that store water not for regular use. After controlling for emptying frequency, however, increased usage of wa sh basins was not associated with production or may have slightly increased it in Armenia and Barra nquilla perhaps through increased egg hatching stimuli through frequent fluctuations in water level. In Barranquilla, the lack of a built-in storage vessel compelled residents to actively obtain ve ssels to store water, largely in case of interruptions in piped service, thereby generating a stronger correlation between emptying and use (Table 2-3). Moreover, greater distance betw een stored water and washing area reduced the use of stored water for cloth washing. We suggest that le ss frequent use in Barranquilla, combined with the structure of storage drum s and possibly community education programs, allowed lid placement to be more frequent and e ffective in reducing production as compared to the other two cities. (Table 2-3 and Table 2-7). For vector control, understanding intera ctions between larval ecology and human behavior is important for negotia ting intervention strategies with communities so as to maximize the sustainability of commun ity-wide behavior modification (Elder and Lloyd, 2007). For example, in Armenia emptying at least every tw o weeks is sufficient to reduce vector output, whereas a more intense promotional campaign would be required in Bucaramanga as the decrease achieved by emptying once per week is more than 3-fold that obtained upon emptying once every two weeks (Table 2-7). In Barranquilla promotion of emptying once a week may be useful in frequently used ve ssels, but other strategies, such as lid placement, should be considered for households that do not regularly use stored water. In the dry season, particularly in Barranquilla, people may be more reluctant to empty vessels suggesting a role for seasonal larviciding programs. We concl ude that in Colombia, vector control programs could be 38

PAGE 39

optim ized by specifying the emptyi ng frequency required to reduce A. aegypti production in a particular altitude or season and promoting a lternative behaviors when the underlying motivation for water storage, as influenced by local containe r structure or reliability of water service, does not permit fulfillment of this objective. We found that container emptyi ng drives variation in the bo th mean and aggregation of A. aegypti production across water storage vessels. Fi fty-eight percent of the total production was concentrated in 7% of the vessels, of which only on1y one was emptied more than once every seven days. This result bears a remarkable resemblance to findings of Subra (1983) who followed 53 vessels in a small Kenyan village over 63 days. Moreover, using a generalized model, we show that aggregation of pupal pr oduction was higher in vessels emptied < 7days, most likely because of a large number of freque ntly emptied vessels that never produced pupae. A. aegypti pupae are usually highly aggregated in cross sectional surveys, and our results indicate that the dynamics of emptying may generate this consistent pattern in which the majority of larval infested vessels in urban areas produce few or no pupae (Focks a nd Chadee, 1997; Getis et al., 2003; Morrison et al., 2004; Barrera et al., 2006a). In addition to flushing out immatures, emptying may reduce larval food, a well docum ented regulator of pupal production in water storage vessels (Subra and Mouc het, 1984, Arrivillaga and Barre ra, 2004, Barrera et al., 2006b). Human perceptions and socio-cu ltural factors interact with the physical environment to define human motivations. These motivations dete rmine human-container interactions, including the frequency and intensity of water usage and replenishment, the time water goes unperturbed without emptying or cleaning, loca tion, lid placement, etc. These processes, in turn, determine key A. aegypti habitat features such as the quality of vessel construc tion, location (shade, sun, rainfall), exposure and accumulation of detritu s, and the dynamics of water volume which 39

PAGE 40

ultim ately shape the proximal ecological processes that regulate larval development including density dependent and independent intraspecific resource competition, microclimate, oviposition, egg accumulation, egg hatching frequency, etc. In this study we focused on directly associating human behaviors with A. aegypti pupal output and chose to exclude intermediate variable s such as daily volume, container size, location and structure from the model selection processes. Because of the complex and multiple causal pathways through which human behavior, contai ner structure and proxi mal physical variables may interact, and given limited sample size, we chose to exclude physical variables. For example, in Barranquilla, with a relatively stable water supply and no built-in washbasin, a frail person may chose to store water in easily mane uverable and sealable 20 L buckets, instead of a large 200 L plastic tank, given the more difficult maintenance and that water is only used in emergency situations. Other residents may choose a large 1000 L tank for eas ier access to water, but due to intermittent usage, it may be placed in an unobtrusive location, increasing exposure to sunlight, rainfall and nutrient en trance. Alternatively, for a renter, such a vessel may condition decreased usage because of the distance betw een the vessel and the washing area, thereby reversing causality between container structure and water storag e behavior. In certain cases unrecorded physical variables, could have modified the impacts of human behavior for example, container texture and contour may affect how efficiently larvae or food particles are removed when permanent basins are emptied by removing the drain. Indeed, while vessels in Armenia and Bucarmanga were virtually identical in size, material and use patterns (Figure 2-2, Table 2-3), Bucaramanga washbasins tended to have smoother finishing, which may have contributed to the observed lower production (Table 2-1). Mo reover, our inductive behavior analysis, employed to minimize the chances that residents would modify be haviors in response to 40

PAGE 41

our surveys, im peded quantificati on of other sociological variables such as interruptions in piped water supply, container cleaning/scr ubbing or socio-economic factors. Nonetheless this study suggests key variables that may interact with human behavior including container structure, contour, location, human demogra phics, temperature and seasonal variation in water service We show that A. aegypti production and human behavior are a coupled human-ecological system (Wilcox and Co lwell, 2005; Liu et al., 2007); research on dengue prevention in variable climates must focus the shared depende ncy of households and A. aegypti immatures on domestic vessels and their water. 41

PAGE 42

Table 2-1.Water storage c ontainers (infested with A. aegypti larvae) studied, follow-up interval, m issing data and pupation rates City Vessels studied (mean no. observations/ vessel) Mean 2-day A. aegypti pupal count ( *) Containers with missing data** (% total) Mean water temperature (C) (range) Armenia 77 (5.6) 13.9 (4.6) 22 (26%) 21.9 (21.0-23.3) Bucaramanga 47 (5.8) 7.2 (3.2) 16 (30%) 23.9 (22.8-25.5) Barranquilla 111 (4.9) 12.0 (2.8) 30 (25%) 29.0 (27.2-31.0) OVERALL 235 (5.3) 11.6 (3.5) 68 (26%) Note: Aggregation parameter ( ), defined in the negative binomial formulation: 2 2xxs where =0 indicates that a Poisson model is ad equate for describing the distribution of production across vessels ** Residents were absent on one visi t and container was not surveyed Table 2-2. Emptying, usage, lidding and A. aegypti production in study vessels No. vessels (avg. no obs/vessel) Mean A. aegypti pupae/ 2-day interval ( )* Percent total vessels Percent total A. aegypti pupae <7 66 (5.3) 3.4 (4.9) 28.1% 8.0% 7-15 70 (5.9) 8.8 (2.5) 29.8% 24.8% Max. days between emptying Not emptied 99 (4.9)19.2 (2.6)42.1% 67.2% 2days 92 (5.3)11.7 (5.0)39.1% 40.1% 3-15 days 104 (5.5) 10.6 (3.1) 44.3% 41.0% Max. days between water usage Not used 39 (4.8)14.3 (2.4)16.6% 18.9% No 189 (5.2) 13.4 (3.6) 80.4% 91.4% Lid placement Yes 46 (5.7) 4.3 (1.6)19.6% 8.6% Note: Aggregation parameter ( ), defined in the negative binomial formulation: 2 2xxs where =0 indicates a random produc tion rate across vessels. 42

PAGE 43

Table 2-3. C ity-wide differences in pattern s of use, emptying and lidding of vessels Armenia Bucaramanga Barranquilla Containers with lids (% tota l) 6 (8.0%)11 (23%) 29 (26%) Avg. water emptying interval (SD) 8.5 (9.8) 10.2 (8.8) 13.1 (10.4) Avg. water usage interval (SD) 2.5 (6.6) 2.6 (5.6) 4.7 (5.4) Containers never used during study (% total) 5 (6.9%)4 (9.3%) 30 (27%) Spearmans Rho, d.f., correlating usage vs. emptying interval (p-value) 0.31, 47 (0.005) 0.32, 77 (0.029) 0.61, 111 (<0.001) Contingency table 2 (2) Usage (always, intermittent, never) vs. lid placement (pvalue) 3.4 (0.33) 6.7 (0.08) 2.9 (0.40) Contingency table 2 (2) Emptying (<7days, 715days, never) vs. lid placement (p-value) 5.0 (0.08) 0.49 (0.78) 1.5 (0.47) Table 2-4. Patterns of use, emptying and rate of A. aegypti production across survey periods City Survey period in 2007 (no. vessels followed) Container use interval (s.e.) Container emptying interval (s.e.) Mean 2-day rate of A. aegypti production (s.e.) January (24) 2.4 (6.6) 12.6 (9.6) 23.1 (48.2) July (18) 2.4 (8.5) 8.0 (12.7) 13.3 (15.5) Armenia Oct-Nov (35) 2.6 (6.0)7.20 (9.5) 7.8 (26.1) January (13) 2.7 (5.1) 28.4 (14.8) 8.6 (15.5) July (21) 2.3 (6.6) 7.9 (7.3) 5.6 (15.0) Bucaramanga Oct-Nov (13) 3.1 (5.3)8.8 (11.9) 8.3 (6.9) January (23) 5.8 (4.7) 39.4 (17.0) 28.9 (46.0) July (51) 4.8 (5.9) 10.0 (9.6) 6.5 (9.7) Barranquilla Oct-Nov (37) 4.1 (5.4)13.2 (11.2) 9.1 (28.2) 43

PAGE 44

Table 2-5.U nivariate associations between water storage behaviors and city and mean A. aegypti pupal production in infested vessels in order of log-likelihood (LL) Variable Coefficient (95% CI) LL AIC Emptying interval 0.84 (0.54,1.15) 0 0 Lid placement -1.19 (-1.8,-0.62) -10.16 20.33 Survey period -0.45 (-0.73, -0.18) -10.51 21.03 Bucaramanga -0.58 (-1.17, 0.015) -14.17 28.35 Armenia 0.27 (-0.23, 0.77) -15.23 30.47 Use category 0.056 (-0.15, 0.26) -15.65 31.43 Barranquilla 0.061 (-0.41, 0.53) -15.76 31.53 Intercept only 2.45 (2.22, 2.69) -15.79 29.59 44

PAGE 45

Table 2-6. Multiv ariate models of mean rate of A. aegypti pupal production, using generalized negative binomial regression1 for overall data and within cities Dataset Explanatory variables (sign of effect) LL AIC (wr) Emptying interval (+), decreasing use (+), lid (-), emptying7-15 *Barranquilla (+), empty7-15*Armenia (-), Armenia (+), Barranquilla (+), survey period -690.91 1399.84 (7x10-5) Emptying interval (+), lid (-), emptying7-15*Barranquilla (+), Armenia (+), survey period -691.52 1395.03 (0.70) Emptying interval (+), lid (-), emptying7-15*Barranquilla (+), Armenia (+) -692.99 1395.97 (0.10) Total Emptying interval (+), lid (-), emptying7-15 days*Barranquilla (+), empty7-15days*Armenia (-), Armenia (+), survey period -691.10 1396.20 (0.07) Emptying<7days (-), emptying7-15days (-), decreasing use (), lid (-), survey period -216.98 445.96 (0.001) Emptying<7days (-), emptying7-15days (-), decreasing use () -217.44 442.89 (0.48) Emptying<7days (-), emptying7-15days (-) -218.57 443.13 (0.29) Armenia Decreasing use (+), emptying7-15days (-) -218.76 443.53 (0.13) Emptying<7days (-), emptying7-15days (-), decreasing use (+), lid (-), survey period -116.54 245.09 (0.001) Emptying<7days (-), emptying7-15days (-), survey period -116.91 241.82 (0.93) Emptying<7days (-), emptying7-15days (-) -118.77 243.53 (0.03) Bucaramanga Emptying<7days (-) -119.78 243.55 (0.03) Emptying<7days (-), emptying7-15days (-), decreasing use (+), lid (-), survey period2 -343.33 698.65 (0.02) Emptying<7days (-), lid (-), survey period, decreasing use (-) -343.52 697.05 (0.62) Barranquilla Emptying<7days (-), survey period, lid (-) -344.66 697.33 (0.35) Note: For dataset, the saturated model is given first, followed by the 3 models with the lowest AIC (bold indicates best-fitting models) out of the 20 highest likelihood models considered for each data set. 1Using the gnbreg command in STATA 8.0 in which the aggregation of production across vessels was parameterized according to ) 7 ( ln10days emptyingaa In Barranquilla the saturated model was the 3rd best fitting model. 45

PAGE 46

Table 2-7. O verall and city-specific best fitting models of household behaviors and A. aegypti pupal production Model (Chi-2 comparison to intercept-only model) Explanatory variables Coefficient (95% CI) Emptying interval (linear trend) 0.82 (0.51, 1.14) Lid placement -0.65 (-1.25,-0.05) Barranquilla*emptying 7-16 days 0.93 (0.22, 1.64) Armenia 0.47 (-0.05, 0.98) Overall ( 2 (5)=39.5, p<0.001) Survey period -0.26 (-0.56, 0.03) Emptying interval < 7 days -1.26 (-2.60, 0.11) Emptying interval 7-16 days -1.32 (-2.30,-0.35) Armenia ( 2 (3)=13.85, p=0.003) Decreasing use -0.60 (-1.30, 0.08) Emptying interval < 7 days -4.80 (-6.7, -3.00) Emptying interval 7-16 days -1.55 (-2.9, -0.21) Bucaramanga ( 2 (3)=14.82, p=0.002) Survey period 0.78 (-0.06, 1.60) Emptying interval < 7 days -1.65 (-2.5, -0.83) Lid -0.98 (-1.6, -0.28) Decreasing use -0.26 (-0.60, 0.08) Barranquilla ( 2 (4)=27.1, p<0.001) Survey period -0.53 (-0.93,-0.12) 46

PAGE 47

Figure 2-1. Study cities within Colombia. 47

PAGE 48

A B Figure 2-2. Water storage vessels in study areas. A) Typical la vaderos of Ar menia (left) and Bucaramanga (right). B) Diverse wate r storage vessels of Barranquilla. 48

PAGE 49

` Figure 2-3. Effect of emptyi ng interval on the rate of A. aegypti pupal production (mean, standard error) in larval infested vessels in three dengue endemic cities with different temperatures. 49

PAGE 50

CHAPTER 3 CELL SIZE AND NUMBER RELATIONSHIPS IN Aedes aegypti (L.) WINGS SHOW DIFFERENTIAL AND INTERACTIVE EFFECTS OF TEMPERATURE AND FEEDING RATE ON MOSQUITO MORPHOLOLOGY 3.1 Introduction Temperature and food availability in the habitats of developing ectotherms are fundamental determinants of variatio n in adult size. In the mosquito Aedes aegypti (L.), ecological processes affecting adult size may also affect variation in th e capacity of populations to transmit diseases such as dengue and chi kungunya. This is because adult size has been correlated with maturation rate (Rueda et al., 1990; TunLin et al., 2000), fecundity (Steinwascher, 1982; Nakasathit and Scott, 1998; Ponlawat and Harringt on, 2007), biting habits (Klowden et al., 1988; Nasci, 1991), dispersal (Maciel de Freitas et al., 2007), survival (Reiskind and Lounibos, 2009), and infecti ousness (Alto et al., 2008; West brook et al., 2009). In these studies size variation was experimentally generated across gradients of resource limitation or temperature; however in adult mos quitoes collected in field settings there is currently no way of distinguishing between these two effects on size. Thermal and nutritional constraints affect attainment of tw o developmental milestones in mosquitoes: critical weight (minimum weight required for pupation) and asymptotic or final weight, both of which are lower in males than in females. The interval to cessation of growth (ICG) is the period between attainment of the cr itical and final weights, during which up to 50% of total growth may occur under optimal feed ing conditions (Davidow itz and Nijhout, 2004, Nishiura et al., 2007; Telang et al., 2007). In creasing temperature [w ithin thermal optima] reduces both critical and final weights in A. aegypti (Chambers and Klowden, 1990; Rueda et al., 1990); this generates a crossing of growth trajectories such that la rvae reared in warm conditions reach the critical weight sooner th an larvae in cooler conditions, but the latter eventually reach a 50

PAGE 51

larger asym ptotic size. In contra st, food variation genera tes nested growth trajectories such that larvae reared with more food reach the critical we ight faster and grow to a larger final weight. Thus, variation in food intake a nd temperature induce di verging growth trajec tories specifically during the ICG. Because not all body components grow at the same rate over the course of development it may be that the differential impa cts of food and temperature on growth trajectory in mosquitoes may be observed in morphologic indicators of allometry. Body size in multicellular organi sms can be viewed as having two components: the size of cells and their number. In in sects, body size is determined by th e size of the surface epidermis that secretes the exoskeleton; thus, epidemeral cells are the key building block of overall size (French et al., 1998). In Dipterans wing cell size is conveniently estimated based on the density of trichomes (wing hairs), each of which represents a single cell. Arendt (2007) reviewed a total of 14 studies comparing cell size and number relationships in Drosophila melanogaster all of which analyzed size and number of epiderma l wing cells. In the only study that we found comparing temperature effects on cell size in mu ltiple organs, Azevedo et al. (2002) found that temperature affected epidermal cell sizes consistently in fruit fly wings, feet and eyes. In general, studies on D. melanogaster indicate that colder temperature affects female size through cell size rather than cell number, whereas in males co lder temperature increases both cell size and number. (Azevedo et al., 2002; French et al., 2002; de Moed et al, 1997; Arendt, 2007). To our knowlege no studies have looked at the effects of cell size and number on the function or shape of wings. In the only study to investigate the combined effects of temperature and food in D. melanogaster, de Moed et al. (1997) found differential a nd interactive effects of temperature and food on cell size and number. Because organism s in natural environments simultaneously experience resource and temperature variation th ere is a need for more research on their 51

PAGE 52

com bined effects on allometric growth (Arendt, 2007). While the de Moed et al. (1997) study suggests that cell size and numb er relationships may reflect the mechanistically diverging impacts of food and temperature on development, ther e is no reason to expect the relationships in observed in D. melanigaster to hold true in mosquitoes. In A. aegypti up to 80% of growth may occur in th e fourth instar (Telang et al., 2007), including the thoracic proliferation of imaginal discs, which give rise to the wing epidermis (Christophers, 1960; Nishiura, 2002). Therefore, wing morphology is likely to be highly sensitive to the environmental conditions that c ontribute to size varia tion in field collected mosquitoes. A. aegypti winglength has been shown to incr ease with lower temperature (Tun-Lin et al, 2000) and heightened resources, induced ei ther by low larval dens ities or high food input (Bedhomme et al, 2003; Jirakanjanakit et al ., 2007). Bedhomme et al (2003) found that heightened feeding time increa sed the winglength of female A. aegypti more so than males. Jirakanjanakit et al (2007) al so found significant allometric effects of food and density treatments on the geometry of wing veins; moreove r, this study showed th at allometric effects were similar across sexes in mid-range food/density treatments but not at extremely high and low density conditions. While variation in both reso urce and temperature conditions has been shown in field A. aegypti habitats (Arrivillega and Barrera, 1996; Strickman et al., 2003; Barrera et al., 2006b; Tun-Lin et al., 2000), no studies to our know ledge have investigated their differential or interactive effects on adult wing morphology. Here, we investigated cell size and cell number in wings of A. aegypti reared under temperature and food conditions designed to si mulate those encountered in dengue endemic areas of different altitudes in Colombia. In particular, we determined the independent and interactive effects of food a nd temperature treatments on diff erences between standardized 52

PAGE 53

indices of cell size and num ber, highlighting their contribution to overall wing size for both males and females. 3.2 Materials and Methods 3.2.1 Experimental Rearing of Aedes aegypti This study was carried out in Universidad Nacional de Colombia, Bogot, Colombia (elevation: 2640m, latitude 4 N, mean temp: 13.0 C), in rearing condition s designed to simulate typical domestic habitats at different dengue-endemic elevations in Colombia. Twenty newly hatched A. aegypti larvae (F2) from Barranquilla, Colo mbia (altitude 5m, latitude 10 N, mean temperature 26-29 C), were raised on different levels of standardized household detritus (see below) in 20 L buckets (filled with at least 19. 5 L of water), each with a 5W aquarium heater (Aquarios S.A., Bogot, Colombia) submerged at the bottom. Buckets were placed in one of two temperature treatments: cold buc kets were exposed to Bogot indoor temperatures, and warm buckets were placed in a non-insulated room (3x4m) maintained at 25 C by an electric space heater. These treatments were standardized prior to experiments so that the cold group water would fluctuate between 21-23 C and the warm group between 27-29C. Water temperatures were measured hourly in 1 bucket within each te mperature group, using an I-button temperature logger (Embedded Data Systems, Lawrenceburg, KY) submerged at the bottom of each vessel and daily in each bucket (because of only a few loggers available). In order to prepare a standard ized household detritus for f ood treatments, we distributed plastic bags in 30 homes located in a dengue e ndemic Colombian city and asked residents to sweep the area immediately surrounding their wa ter storage vessel and deposit the contents obtained over a 24 hour period. Bags were obt ained from 25 houses. After removing big rocks and synthetic substances, a 0.8 kg mixture resulted, composed (by weight) of 52% 53

PAGE 54

dirt/sed iment/small rocks, 11% leaves and stems and 37% diverse organic matter (fibrous material/hairs, insects, human/pet food, seeds). We ground this mixture into sediment in order to visually minimize biases in food application across containers and over time. Contents were dried and a manual grinder was used to convert > 95% of the mixture into particles capable of passing through netting with pores smaller than 0.25mm2. The remaining larger particles were discarded. Upon sifting, the resulting mixture cons isted of a high density granulose bottom layer and a fibrous top layer. Food applied to each ve ssel included both layers and was measured on a balance with approximately mg error. In order to simulate a range of food conditions that A. aegypti may experience in household vessels, two consecutive experiments were carried out in each temperature treatment. In Experiment 1, 5 groups of 3 replicate vessels received one of the follo wing treatments every 3 days beginning on the day when the 20 newly ha tched larvae were added (day 0): 50, 100, 200, 400 and 800 mg of sediment (hereby referred to by the respective daily app lication rates: 16.7, 33.3, 66.7, 133.3, 266.7 mg/day) Based on the p upation results of Experiment 1 (hereby called high food experiment), we sought to increase food scarcity in Experiment 2 (hereby called low food experiment). However we observed that 40 mg was the minimum amount of food applied that could contain matter from both of the layers described above. Accordingly, in the low food experiment, we varied food input frequency instead of amount. Gr oups of 4 replicate vessels per temperature were each assigned one of the followi ng daily probabilities of addition of 40 g of sediment: 0.1, 0.25, 0.5, 0.75. A random number generator determined the days in which each replicate received a food treatment. We used th is method instead of food application at constant intervals because constant food input under f ood limiting conditions may stall development in the L4 stage (Gilpin and McClelland, 1979). Food application rates ( s.e.) for each group were 54

PAGE 55

the following: 2.9 (.5), 11.4 (.4), 20.0 ( 20.4) and 27.8 (.8) m g/day. Because these rates are calculated over the time of the last pupa tion event and are not ne cessarily indicative of the food experienced by most of the resulting mo squitoes, we hereby use the following expected values to define food applica tion rates in the low food experi ment: 4, 10, 20 and 30 mg/day. Experiments 1 and 2 gave a total of 62 vessels. F ourth instar larvae (L4) and pupae were counted daily until all larvae died or pupated. All pupae were removed and allowed to complete development at 25 C, as we are unaware of evidence linki ng temperature in the pupal stage with wing development. 3.2.2 Wing Photography and Image Analysis The right wing of each surviving mosquito with undamaged wings was mounted on a cover slip, with the dorsal side facing up, using a 2:1 mixture of water with transparent paper glue. Digital cameras and Optika Vision Pro soft ware were employed in order to produce two images of each wing: 1) a 2048 x 1536 pixel im age of the entire wing using a dissecting microscope (2.5x) (Figure 3-1A) and 2) a 480 x 640 pixel image of a 0.0108 mm2 area in the third posterior wing cell between the 1st and 2nd anal veins using a compound microscope (40x) (Figure 3-1B). Image analyses were conducted using Image J software (National Institutes of Health). The software automatically calculated the area enclosed by a trace of the perimeter of the wing using the image in Figure 3-1A. It also performed an automated hair count by dichotomizing the color spectrum in Figure 3-1B and counting the number of black points. Because wing veins, split hairs, or discontinuities in the background color due to mounting imperfections were all potential sources of error, count s were conducted on a smaller 0.0032 mm2 square area within Figure 3-1B. For each image we located the square that maximized the visualization of hairs and 55

PAGE 56

m inimized blurriness and debris. Im ages with noticeable errors in dichotomization of hairs were discarded. In the absence of evidence on hetero geneities in epidermal cell size across mosquito wings, we considered that our procedure achieved an appropria te balance between including a representative number of hairs in each take (~45-85) and maximizing image quality. Mean cell size was estimated by the reciprocal of hair counts, and an index of wing cell number was defined as wing area divided by the estimate of mean cell size. 3.2.3 Data Analysis Due to the variation in food input schedules and contrasting deve lopmental strategies across sexes, data were analyzed separately for males and females and in the high and low food experiments. We first determined the effects of food input rate, temperature treatment and the food-temperature interaction on the two variables measured independently for each observation, wing size and cell size. Subsequently, we analyz ed whether treatments had a larger impact on cell size or number. Because of the large difference in scaling of wing and cell size, we used Z-scores to standardize wing area (Zwing), mean cell size (Zsize) and cell number index (Znumber). Z-scores are defined as Z=( xi x )/ where x is the measured value for each mosquito i, i x is the mean and is the standard deviation of each of the three measur ements outcomes. Z measures the magnitude and direction of the deviation of each observation fr om the overall mean, positive if above the mean, negative if below, generating a standard normal di stribution in the overall data with mean of 0 and standard deviation of 1. We determined wh ether treatments had a larger impact on cell size or cell number through the difference in z-scores, defined for each mosquito i as: i number, i size, i diff,Z Z Z (3-1) 56

PAGE 57

According to this definition, if a particular treatm ent cause s a disproportionate increase in cell size compared to cell number, then Zsize,i will be large compared to Znumber,i, and Zdiff,i will be a large positive number. In the opposite case, that a treatment results in an inordinate increase in cell number compared to cell size, Zdiff,i will be negative. Therefore, predictors positively associated with Zdiff will have a larger impact on cell size and a negative association indicates a larger impact on cell number. No association with Zdiff means no allometric effect. Because mosquitoes were produced from 62 different buckets, each exposed to random differences in our heterogeneous field-collected food, in addition to a highly unbalanced design with many containers yielding only 1 or 2 individuals per sex us ed in wing analyses, we used multiple random-effects, maximum likelihood (ML) regression using the procedure xtreg in STATA 8.0 Statistical Software (StataCorp., Co llege Station, TX). Container was the group variable assigned random effects. We quantified the combined effects of temperature (warm group = 27-28C versus cool group = 21-22C), the Z-score of feed ing rate (hereby referred to has food), and the food temperature interacti on (hereby referred to as warm*food) on the dependent variables Zwing, Zsize and Zdiff. Wald tests ( = 0.05) were employed to determine the significance of regression coeffi cients. A separate model was c onstructed for each sex and food input experiment for a total of four models for each dependent variable. By evaluating significant associations with Zwing,, Zcell and Zdiff conclusions were drawn as to whether treatments effects on overall wing size were due to a la rger effect on cell size or number. 3.3 Results Mean daily temperatures across containers were 22.2 C (range across vessels: 21.7-22.5) and 28.2C (range across vessels 27.9-28.6) in cool and warm treatments, respectively, which were underestimates, as temperature in each cont ainer was generally measured in the cooler 57

PAGE 58

morning hours. In the containers that were m onitored hourly, average daily minimum and maximum temperature over the two expe rimental trials were 22.0 and 22.9 C (mean 22.5 C) in the cold treatment and 27.9 and 29.1 C (mean 28.5 C) in the warm treatment. A total of 128 and 216 mosquitoes were analy zed in the low and high food experiments, respectively, as compared to 231 and 312 pupae that were harvested in each. This discrepancy was due to a combination of relatively high pupal mortality in food limited treatments, newly emerged mosquitoes that drowne d or whose right wing was otherw ise folded or damaged, and inadequate wing mounting. Mean time to pupation decreased with increasing food input and was lower in the warm group in all feeding regimes (Figure 3-2). Time to pupation in the highest food treatments in the low food experiment was highe r than that of the lowest feeding groups of the high food experiment (Figure 3-2). Given that mean food addition rate was comparable among these groups, the data suggest that the rand om feeding method may have had an effect on lowering development rate. Total wing area of females was significantly larg er in those mosquitoes reared in the cold temperature as opposed to the warm in both expe riments (Table3-1, Figure 3-3). Increasing food addition rate was associated with larger female wing area in th e high food experiment (Figure 33B) but not in the low food experiment (Figure 33, Table 3-2). The lower temperature treatment was also significantly associated with larger cell size in both experi ments (Table3-2) although the effects were less pronounced than for wing si ze (Figure 3-3). In the high food experiment increased feeding rate was not associated with cell size. In the low food experiment increased feeding rate was associated with larger epidermal cells in the warm treatment but not in the cold (Figure 3-3), as evident in the significant warm*f ood interaction term in the model of female cell size for this experiment (Table 3-1). In the tw o lowest food groups (4 and 10 mg/day in Figure 358

PAGE 59

3) cell size was notice ably larger in the cold treatment; this tre nd switched at 30mg/day, with increasing mean cell size in the warm treatment (Figure 3-3). As with females, the lower temperature treatment significantly increased male wing size in both experiments, whereas increased food add ition was associated with larger wings only in the high food experiment and not associated with cell size in either e xperiment (Figure 3-4, Table 3-2). Unlike females the cold treatme nt increased cell size only in the low food experiment. In the high food experiment, there was a significant food*war m interactive effect, such that in low food conditions males reared in the cold treatment had larger cells, but in the higher feeding groups mean cell size was larger in the warm treatment (Table 3-2, Figure 3-4). There was a significant warm*f ood interactive effect on Zdiff in females in the low food experiment, indicating that the relative effects of food on cell number and size differed between temperature treatments (Table 3-3). In particular the positive coefficient indicates that in the warm group increasing food input rate is associated with a larger increase in cell si ze than in cell number whereas in the cold group higher food induces a larger increase in cell number. This is evident in the crossing of the regression lines in Figure 3-5a with mosquitoes below the y=0 plane having a larger cell number relative to cell size. Neither of the treatments or their interaction are significantly associated with Zdiff in males in the low food experiment or females in the high food experiment (Table 3-3). In males in the high food experiment, food input was associated with a larger increase in cell number th an in cell size, as is evidenced by the larger density of dots (mosquitoes) above the y=0 plane in the low food input rates as compared to the high food input rates (Figure 3-5b); however, this effect varied among temperature groups as evidenced in the signific ant, positive warm*food interaction (Tab le 3-3). The larger impact of 59

PAGE 60

food on cell num ber relative to cell size was mo re pronounced in the cold treatment (Figure 35b). Table 3-4 qualitatively summarizes the signifi cant model results for wing size, cell size and Zdiff, indicating whether treatment effects on overall size were brought about by a larger effect on cell size or cell number Mosquitoes reared in the cold treatment had larger wing and cell size (i.e. a significantly nega tive effect of warm), with th e exception of males in the high food experiment (Table 3-4). However, the lack of a significant association with Zdiff indicates that the effects of the cold treatment on wing size were not generated through a significantly greater effect on cell size relative to cell number or vice vers a. Food addition rate increased wing size in both sexes only in the high food experiment. In males this was due to a larger increase in cell number than in cell size (signifi cant negative association with Zdiff), whereas no allometric effect was detected in females. Interestingly, the warm*food interaction term was not associated with wing size but did have a pos itive effect on cell size and Zdiff for males in the high food experiment and females in the low food experiment Given that the cold treatment, but not food, increased female wingsize and cell size in the low food experiment, th e warm*food interactive effect on Zdiff suggests that the cold treatment increas ed wing size through a larger increase in cell size at low feeding rates but through cell number as high feedi ng rates. For males in the high food experiment, both food and temperature increased wing size, but neither had a significant effect on cell size (Table 3-2, Figure 3-4). Thus the interactive effect on Zdiff mean that the relative contribution of cell size and number to the observed wing size variation was specific to both temperature group and food app lication rate (Figure 3-5) 3.4 Discussion Although both thermal and resource regulati on in aquatic stages are key processes shaping the size, abundance and distribution of mo squitoes, surprisingly little is known about the 60

PAGE 61

underlying m echanisms through which they affect growth and development. Here, for the first time that we know of in a mosquito, we demons trate that independent, interactive and sexually dimorphic impacts of food and te mperature in the larval environment can be observed in adult morphology. We show that while both lower te mperature and heightened feeding increase wing size, they differ in the magnitude of their impacts on epidermal cell size and cell number in A. aegypti, reared under conditions simulating thos e observed in dengue endemic areas in Colombia. Moreover, our results indicate that temperature and resource availability have interactive effects on mosquito de velopment, such that the effect of each changes with the level of the other. Numerous prior studies show that wingsize in A. aegypti increases with lower rearing temperatures and increased food provision (R ueda et al, 1990; Tun-Lin et al, 2000; Jirakanjanakit et al., 2007), but to our knowle dge no studies have i nvestigated food and temperature effects on cell size. In D. melanogaster, the taxonomically closest organism on which this data exists, studies consistently show that lower temperature increases cell size, with lesser or no impact on cell number (Arendt, 2007). In addition, DeMoed et al (1997) found that increased food has larger effects on D. melanogaster cell number than on cell size. In the experiments presented here A. aegypti raised at 22C had both larg er wings and larger epidermal cells than those raised at 28C. However, our data indicate that the effect of the cold treatment on cell size was approximately similar to its effects on cell number, as evidenced in the consistently non-significant impact on Zdiff in both experiments and sexe s (Tables 3-3 and 3-4). Food application increased wing size only in the high food experiment but was not associated with cell size in either sex. Moreover, in males, increa sed food had a significantly larger impact on cell number than on cell size, particularly in the cold treatment (Table 3-3, Figure 3-5), in accordance 61

PAGE 62

with DeMoed et al (1997). These results dem onstrate m orphologically di stinct effects of food and temperature and suggest that they differ in their impacts on cell growth and division. Also consistent with DeMoed et al (1997) we found significant interactive effects of food and temperature on cell size, such that heightened food input increased cell size in the warmer group, but not the colder group. This interactive effect caused a positive association with Zdiff, indicating that food had a larger effect on cell size than cell number in the warm treatment, but had a larger effect on cell number in the cold treatment. For females, this occurred in the low food experiment and for males it occurred in the high food experiment. However, unlike D. melanogaster (DeMoed et al, 1997) this interaction was not pres ent in overall wing size. The cold treatment, but not feeding rate, increased wing size in females in the low food experiment. We suggest that in extremely food limited hab itats (4 and 10 mg/day in our experimental conditions), lower temperature increases female size through larger increa ses in cell size than cell number, whereas at higher food levels both ce ll number and cell size contribute. In males in the high food experiment, both increased food a nd lower temperature associated with larger wings, whereas neither had an independent effect on cell size (Table 3-2, Figure 3-4). Therefore, the warm*food interaction indicates that the relative contribution of cell size and number to environmental variation in male wing size is sp ecific to the combination of temperature and resource conditions (Table 3-4). Morphologically, the size of insects is the product of the number a nd size of epidermal cells composing the exoskeleton, whereas physiologi cally, size is a product of the growth rate and the duration of growing stag es (Davidozitz and Nijhout, 2004). In mosquitoes food resources increase growth rate; by contrast lower temp erature prolongs growing time while increased temperature promotes increased food assimilatio n rate (Rashed and Mulla, 1989). Sex-specific 62

PAGE 63

developm ent strategies have also been shown to interact with environmental conditions. For example increased resources have been shown to favor heightened ener gy storage and decreased time to maturation in male A. aegypti whereas in females increased resources have greater effects on structural growth and size and sma ller effects on development time (Timmermann and Biegel, 1999; Bedhomme et al, 2003). We can sp eculate as to how these sexually dimorphic ecological relationships fit with the results pres ented here. For example, females in the cold treatment in the 4 and 10mg/day groups (low food experiment) would have had exceptionally low food uptake and energy storage; thus cold te mperature may have caused a disproportionate increase in cell size over number as a consequence of the need to maximize size by prolonging development. By contrast males in the warm treatment in the 133.3 and 266.7 mg/day groups (high food experiment) would have had exceptionally high food assimilation rates; therefore a disproportionate increase in ce ll size in the wing epidermis may reflect heightened cellular energy storage [in other tissues] within a very short window of devel opment. Assessing the validity of these assertions however, would require unders tanding how the Barranquilla A. aegypti strain used in our experiments ma y have influenced the results. In field surveys of A. aegypti populations throughout the worl d larval nutrient indicators consistently demonstrate la rge variation across urban A. aegypti habitats (Subra and Mouchet, 1984; Tun-Lin et al., 2000; Strickman et al., 2003; Arrivillaga and Barrera, 2004; Barrera et al., 2006b). In a study of geometric relationships in A. aegypti wing veins, the allometric effects of resource availability on wing geometry changed at extremely low and high larval densities/food inputs, and these treatments were di scarded in the statistical analysis (Jirakanjanak it et al., 2007). However, given that food limitation in larval stages is considered a critical process in determining the dynamics of adult A. aegypti populations (Southwood et al., 1972), 63

PAGE 64

understanding the effects of ex trem e resource conditions may be highly relevant for field A. aegypti Moreover, as heterogeneity in the food particles available to A. aegypti affects larval nutrition (Rashed and Mulla, 1989), heterogenou s detritus and microrganisms that larvae encounter in field habitats are like ly to bare little resemblance to standardized laboratory diets. Thus, we chose an experimental design which maxi mized the similarity of our experiments with the wide range of resource conditions experienced by A. aegypti in the field, and minimized the potentially large stochastic effects of administ ering small quantities of heterogeneous detritus food. The use of random food input in the low food experiment introduced a further source of variance, which combined with th e variation in frequency instead of quantity of food input, made it impossible to combine the analysis in order to better isolate the inde pendent effects of food abundance. Nonetheless, the design employed was able to generate extremely nutritionally stressed mosquitoes as evidenced in the mean development times on the order of 20-25 days (males and females combined) (Figure 3-2). Desp ite small sample size in the female low food experiment, the data indicates th at the wide range of food abunda nce was essential in detecting interactive effects of heightened feeding rate and temperature. Although alternative designs of food application and the usage of replicate incubators for temp erature treatments may have generated clearer results, the resemblance of our experiment to field conditions may increase the prospects for field validation. The food-temperature interactions that we demonstrate in both sexes may represent a first step in developing the means to assess larval habitat quality based on adult characteristics in this important disease vector. As a first approximation to field testing our results, adult A. aegypti could be collected across a range of climates an d habitats. We predict that as nutritional stress experienced in habitats decreases, the differen ce between standardized indices of cell size and 64

PAGE 65

num ber should increase at 28C a nd decrease at 22C. Furthermor e, variation in indices should occur at the low end of the resource spectrum in females and at the high end in males. Considering the limitations of our experimental design, field validation of these predictions could represent a significant advance in unders tanding the regulatory pr ocesses in natural A. aegypti populations. Moreover, further experiment al studies that allow comparison of our findings to other arthropod vectors and A. aegypti strains from different climates can provide insights on how natural selection in the face of warming climates may impact vector borne disease. 65

PAGE 66

Table 3-1. R andom effects models of wing area and mean cell size in female A. aegypti in high and low food experiments High food experiment n=104 Low food experiment n=43 Dependent variable Treatment Coefficient (95% CI) P Coefficient (95% CI) P Wing area (z-score) -1.27 (-1.56, -0.99) <0.01 -1.40 (-1.85, -0.95) <0.01 Warm 0.50 Food (0.25, 0.75) <0.01 -0.07 (-0.25, 0.39) 0.68 Food*warm -0.15 -0.21 0.53 0.18 (-0.60, 0.31) (-0.52, 0.09) -0.62 Warm -0.46 (-0.81, -0.1 1) 0.01 (-1.16, -0.07) 0.03 Food 0.15 (-0.17, 0.48) 0.35 -0.29 (-0.69, 0.10) 0.14 Mean cell size (zscore) Food*warm -0.07 (-0.46, 0.32) 0.65 0.74 (0.10, 1.21) 0.02 Note: *Using xtreg procedure in STATA 8.0. Table 3-2. Random effects models of wing area and mean cell size in male A. aegypti in high and low food experiments High food experiment n=112 Low food experiment n=85 Dependent variable Treatment coefficient (95% CI) P coefficient (95% CI) P Warm -1.30 (-1.60, -1.01) <0.01 -1,33 (-1.73, -0.93) <0.01 Food 0.38 (0.19 0.57) <0.01 0.15 (-0.14, 0.44) 0.31 Wing area (z-score) Food*warm -0.14 -0.15 0.27 (-0.55,0.27) (-0.40, 0.11) 0.50 Warm -0.66 (-0.99, -0.34) <0.01 -0.31 0.14 (-0.71, 0.10) Food -0.27 (-0.55, 0.01) 0.06 -0.03 (-0.22, 0.29) 0.79 Mean cell size (zscore) 0.54 Food*warm (0.14, 0,94) 0.01 -0.04 (-0.37, 0.33) 0.83 Note: *Using xtreg procedure in STATA 8.0. 66

PAGE 67

Table 3-3. R andom effects models of Zdiff in female and male A. aegypti in high and low food experiments High food experiment Low food experiment Sex Treatment coefficient (95% CI) p coefficient (95% CI) p 0.53 (-0.07, 1.12) 0.08 0.51 (-0.45, 1.47) 0.30 Warm -0.31 Food (-0.86, 0.24) 0.27 -0.55 (-1.23, 0.14) 0.12 Female 1.26 0.14 0.67 Food*warm (0.3, 2.23) (-0.51, 0.79) 0.01 Warm 0.58 0.13 (-0.16, 1.32) 0.12 (-0.46, 0.71) 0.68 Food -0.83 (-1.33, -0.33) <0.01 -0.05 ((-0.51, 0.41) 0.83 Male Food*warm 1.08 <0.01 (0.38, 1.78) 0.08 (-0.52, 0.67) 0.80 *Using xtreg procedure in STATA 8.0. Table 3-4. Summary table of significance of main and interactiv e effects of food and temperature treatments in random effects model of wing size, cell size and Zdiff High food experiment Low food experiment Sex Treatment Zwing Zcell Zdiff Zwing Zcell Zdiff Warm NS NS NS Food + NS NS NS NS Males Warm*Food NS + + NS NS NS Warm NS NS Food + NS NS NS NS NS Fe males Warm*Food NS NS NS NS + + Note: NS means non-significant at =0.05 67

PAGE 68

A 0.25m m B 16m Figure 3-1. Images used to determin e wing size and epidermal cells size in A. Aegypti A) Image of A. aegypti wing used to measure wing area. Squa re indicates region in which hair counts were conducted. B) Image in 3rd posterior wing cell used count hairs in order to estimate cell size. 68

PAGE 69

A 0 2 4 6 8 10 12 14 16 1816.733.366.7133.3266.7Mean food application rate (mg/day)Mean time to pupation (days ) cold warm B 0 5 10 15 20 25 30 35 4102030Mean food application rate (mg/day)Mean time to pupation (days ) cold warm Figure 3-2. Mean time to pupation among experiment al treatments of food and temperature. A) High food experiment (n=312). B) Low food experiment (n=231). Figure comprises both males and females and all pupae excluded from the wing analyses due to mortality, damaged wings, or inadequate mounting. 69

PAGE 70

A B 1 1.25 1.5 1.75 2 2.25 2.5 4102030Mean food input rate (mg/day)Wing area (mm2) cold warm 1 1.25 1.5 1.75 2 2.25 2.5 16.733.366.7133.3266.7Mean food input rate (mg/day)Wing area (mm2) cold warm C D 30 35 40 45 50 55 60 65 4102030Mean food input rate (mg/day)Mean cell area (m2) cold warm 30 35 40 45 50 55 60 65 16.733.366.7133.3266.7Mean food input rate (mg/day)Mean cell area (m2) cold warm Figure 3-3. Wing measurements in female A. aegypti across food and temperature treatments. A) Wing area low f ood experiment. B) Wing area hi gh food experiment. C) Average epidermal cell area in low food experiment D) Average epidermal cell area in high food experiment. 70

PAGE 71

A B 0.5 0.7 0.9 1.1 1.3 1.5 4102030Mean food input rate (mg/day)Wing area (mm2) cold warm 0.5 0.7 0.9 1.1 1.3 1.5 16.733.366.7133.3266.7Mean food input rate (mg/day)Wing area (mm2) cold warm C D 30 35 40 45 50 55 60 4102030Mean food input rate (mg/day)Mean cell area (m2) cold warm 30 35 40 45 50 55 60 16.733.366.7133.3266.7Mean food input rate (mg/day)Mean cell area (m2) cold warm Figure 3-4. Wing measurements in male A. aegypti across food and temperature treatments. A) Wing area low food experiment. B) Av erage epidermal cell area low food experiment. C) Wing area high food expe riment. D) Average epidermal cell area high food experiment. 71

PAGE 72

A B Figure 3-5. Significant interactive e ffects of food and temperature on zdiff (labeled as difference in y-axis). A) Females in low food expe riment. B) Males in high food experiment. Black lines are the linear regr ession of feeding rate in e ach temperature. Dots are individual mosquitoes; black lines depict regression of food effects within each temperature group. Zfood is the z-transformed food application level calculated separately for the high and low food experiments. 72

PAGE 73

CHAPTER 4 ENERGY STORAGE STRATEGIES EXPLAIN DEVELOPING DISEASE VECTORS RESPONSE TO RISING TEMPERATURE 4.1 Introduction Higher temperature increases the devel opment rates of mosquitoes by speeding up biochemical reactions. Nonetheless, heighten ed reaction rates require more energy, and surprisingly little is known about how mosquitoes, and ectotherms in general, compensate energetically in order to matu re faster in warmer conditions (Lafferty, 2009). It has been suggested that since organisms must allocate finite resources towards competing biological needs, temperature may affect the fitness tradeo ffs between developing faster, meeting increased energy demand and maximizing size (Kozlowski et al., 2004). Such tradeoffs provide an attractive basis for the appearance of the te mperature-size rule (TSR), the observation of decreased ectotherm size at higher temperatures. Indeed, despite being recognized as one of the most consistent rules in bi ology the TSR has eluded genera l, taxonomically widespread explanations (Atkinson and Sibly, 1997; Angilleta and Dunham, 2003). For holometabolic ectotherms such as mosquitoes, temperature may affect the optimization of resource allocation between energy and growth in immature stages. Mosquitoes will be favored by avoiding depletion of energy reserves, which are involved in determining the timing of of pupation (Gilpin and McClelland, 1979; Chambers and Klowden, 1990; Telang et al., 2007). Larval energy reserves ha ve also been associated with adult flight ability (Nayar and Van Handel, 1971) and egg production (Zhou et al., 2004). Therefore, if mosquitoes are compelled to devote more resources towards energy storage instead of structural growth in order to compensate for increased metabolic activity at higher temperatures, final size may be reduced, thereby explaining the TSR. However, a temp erature-induced reducti on in body size is also likely to have fitness costs as larger mosquito es have increased fecund ity (Steinwascher, 1982; 73

PAGE 74

Nakasathit and Scott, 1998; Ponlawat and Harr ington, 2007), bite m ore persistently (Nasci, 1991), and survive longer (Reiski nd and Lounibos, 2009). Clearly th e effects of temperature on the growth-energy storage tradeoff demand attent ion; however, this issue has been scarcely addressed in part because histor ically, life-history m odels and empirical studies have assumed that energy storage and weight gain are directly proportional to one another in a temperature independent manner (Bertalanffy, 1960; Strong and Daborn, 1980; Berrigan and Charnov, 1994; Perrin, 1995). However, stored energy is but only one component of total mass, and if temperature were to affect the tradeoff between growth and energy storage, then it would produce changes in the proportion of total mass dedicated to energy storage. Such an energy-growth tradeoff may be particularly important for mosquitoes that thrive in food limited habitats. Under f ood limitation energy stores may ta ke on the additional role of bolstering short-term starvation resistance (Gilpin a nd McClelland, 1979), whereas structural growth may increase energy expenditure (Bertala nffy, 1960) and therefore susceptibility to food limitation. This takes on particular relevance for Aedes aegypti (L.), the major global vector of dengue fever, chikungunya and urban yellow fever. The principal larval habitats of this mosquito, domestic urban contai ners, are notoriously resource poor (Subra and Mouchet, 1984; Arrivillaga and Barrera, 2004; Stri ckman et al., 2003), and numerous field studies correlate the dynamics of adult production (Subra, 1983; Barr era et al., 2006b) and abundance (Southwood et al., 1972) with food limitation in the larval stage. Therefore, increased energy expenditure at higher temperatures may aff ect the ability of larval A. aegypti to obtain the resources necessary to mature in common urban habitats; this may add an additional pressure on developing larvae of this species to energetically compensate at higher temperature. 74

PAGE 75

Here we m odel the potential tradeoffs between metabolism, energy storage and growth in A. aegypti larvae. In order to determine how rising temperature impacts these tradeoffs, we parameterized the model at 2 C intervals between 20 and 30 C, using laboratory experiments on weight gain, maturation time and starvation resistance. We use th e model to determine the most likely mechanisms of resource allocati on that give rise to the TSR in A. aegypti and their potential impact on the tradeoffs between size, starvation resistance and development rate as temperature increases. Using maximum likelihood m odel fits we make predictions of the impacts of rising temperature on energy re serves through the course of development and the rate of A. aegypti production in habitats with variable resource conditions. 4.2 Methods and Results 4.2.1 Laboratory Experiments on Effects of Temperature on Development 4.2.1.1 Experimental conditions Temperature impacts on development were in vestigated by measuring development rate, growth and starvation resistance in larval Aedes aegypti (F3) from Barranquilla, Colombia (altitude 5 m, latitude 10 N, mean monthly temperatures 26-29 C), using incubators calibrated to 20, 22, 24, 26, 28 and 30C. At each temperature experi ments were carried out in one incubator in which water temperature was measur ed hourly in a 5 ml cup, using an I-button temperature logger (Embedded Data Systems, Lawr enceburg, KY). In order to simulate natural conditions, the food used was a standardized house hold detritus collected in the vicinity of 30 water storage vessels in a dengue endemic city in Colombia. No a priori information was available on the quality of this food, but based on the observed development rates (see below), it appeared comparable to standard laboratory food s used for mosquito rearing (Rueda et al, 1990). 75

PAGE 76

In all feeding regim es, 100 mg of sediment was pre-incubated for 24 h in 0.9 L of water (in 1L feeding cups) in order to allow more time for microorganism development as the sediment had been freeze dried. Ten newly hatched first instar larvae (L1) were added to each feeding cup, and another 100 mg of food was added after 48 h of feeding was completed. These feeding conditions had relatively small water volume in comparison to typical A. aegypti field habitats (Padmanabha et al, 2010) and very high food to larval ratio (albeit of an unknown food quality) in comparison to experimental studies of resource competition (Gilpin and McClelland, 1979; Jirakanjanakit et al, 2007); these conditions were chosen sp ecifically in order to minimize resource competition and the time spent searching for food. In order to measure the minimum feed ing time required to commit to pupation (Experiment A) and starvation resistance (Experiment D) each larvae in four replicate vessels (n = 40) was placed in a 5 mL filled cup with dis tilled water and monitored daily until death or pupation. At each temperature this was carried ou t daily beginning with newly hatched larvae (that were never fed) on Day 0, until the day (specific to each temperature) in which pupae were observed in the feeding cups. As a supplement to Experiment A, feeding was continued in four replicate feeding vessels at each temperature until the day in which 50% of the larvae had pupated, which we define as the median age of pupation ( Apup). All observations on pupation and mortality were made on a daily basis, and no larval mortality in the feeding cups was observed. In Experiment B we used the same feeding and temperature conditions to measure the dry weight trajectory at 28C in larvae sacrificed da ily after feeding from 0 to 4 days and incubation for 24 h in a drying oven. Because the error in our microgram balance was comparable to the weight of early instar larvae, 0-3 day fed larv ae were weighed in ten different groups. Group size was 10 for unfed larvae and 5, 4 and 2 for 1, 2 and 3-day fed larvae, respectively; this 76

PAGE 77

correspond ed to five, four and two replicate feed ing cups for each respect ive feeding time. After four days feeding dry weights were measured in individual larvae/pupae in three replicate cups (n=30). This latter procedure was carried out at all six temperatures in order to determine WL4 the weight of late L4 or ear ly pupae (Experiment C). Since Apup was four days at 28C the same 30 larvae were used in experiments B and C for this temperature. No larval mortality was observed in either of these experiments. Expe riments B and C were conducted after Experiments A and D had terminated. 4.2.1.2 Temperature effects on mimimum feed ing time required for commitment to pupation and median time to pupation (Experiment A ) Feeding time required to pupate after transfer to distilled water decreased monotonically with increased temperature (F igure 4-1). All larvae that di d not pupate incurred starvation mortality. Over 60% of larvae committed to pupation after 3 days feeding at 30 C, whereas at 20C larvae needed to feed for 7 days in order to reach this pupation success (Figure 4-1). For larvae that commited to pupation, mean time (days SD) to pupation after transfer to distilled water was for 30C: 1.25 ( 0.44), 28C: 1.5 ( 0.66), 26C: 1.8 ( 0.70), 24C: 2.6 ( 1.4), 22C: 2.9 ( 1.1) and 20C: 5.4 ( 2.4). These data indicate that transfer to distilled water stimulated larvae to initiate the physiological pr ocess of pupation, which proceeded faster with increasing temperature. At 24 a nd 20C we observed outliers that initiated pupation after 5 days in starvation, but subsequently died before emerging as adults. At 20, 22, 24, 26, 28 and 30C our Apup measurement was 9, 8, 7, 6, 5 and 5 days, respectively. The respective cumulative pupation percent at Apup for each temperature was 30C: 92.5%, 28C: 60%, 26C: 100%, 24C: 82.5%, 22C: 50%, 20C: 52.5%. In these cups we did not record pupation percentage on a daily basis, only on the day in which we counted at least 20 pupae among the four feeding cups used to determine Apup, at which point Experiment A was 77

PAGE 78

discontinued. Other sources of uncertainty in our Apup estimate include the course 24h time scale of observation, unknown sex ratios (time to pupa tion in 50% of the population may overestimate and underestimate pupation rate in males and female s, respectively) and the fact that pupation percent was not recorded separately for each feeding cup. Nonetheless our Apup estimate was similar to Rueda et al (1990), who recorded me dian time to pupation as 4.9 and 9.3 days at 30 and 20C, respectively. This simila rity is noteworthy, given importa nt differences in the protocol of this study, including the use a laboratory colony of A. aegypti 4-6 h interval between observations, a high protein homogeneous f ood source and a fixed 1:1 sex ratio. 4.2.1.3 Dry weight trajectory at 28C (Experiment B ) Dry weight through the course of larval development at 28C showed a roughly sigmoidal trajectory, with exponential growth occurring between 2 and 3 days feeding and considerably slowing between 3 and 4 days feed ing (Figure 4-2). Larval instars after each feeding day were the following at 28C: one day: L2, two day: L3 three and four days: L4. Over 75% of total weight was gained in the second ha lf of development. This corresponds roughly to the finer scale observations of Te lang et al (2007) in which 80% of total larval growth occurred in the L4 stage. The little growth from L1-L3 stag es indicates that a sigmoidal function is a more accurate description of the data than a less than linear, exponential function. 4.2.2.4 Temperature effects on WL4 (Experiment C ) At 20, 22, 24, 26, 28 and 30C, WL4 was measured at 8, 7, 6, 5, 4 and 4 days, respectively. At each temperature this was a day prior to Apup. We chose this measurement time, rather than weight at pupation fo r a number of reasons including: (1) pupation was not defined in our model fitting procedure (see below), (2) the large time interval (~ 24 h) between laboratory observations in comparison to the model time step (6 h) and (3) our model was not sex or stage specific. For simplicity in model fitting, these factors favored weight measurement in larvae at a 78

PAGE 79

fixed age after the inflextion point in the ta il end of the weight traj ectory (Figure 4-2), rather than at a particular developmen tal stage. At 20 and 22C WL4 was measured 2-days longer than the minimum feeding time required for at least 50% of larvae to commit to pupation, and one day longer at 24, 26 and 30C (Figure 4-1). Given that only larvae in the middle of the L4 stage commit to pupation (Telang et al, 2007; Nishiu ra et al, 2007), we ar e confident that our WL4 estimation reflects the weight of late L4 that ha ve surpassed the inflectio n point in the sigmoidal weight gain trajectory (Figure 42) at these temperatures. Sim ilarly Figure 4-2 indicates that weight increase after four days feeding at 28C is likely to be small. Mean WL4 decreased monotonically with increas ed temperature (Figure 4-3). This coincides with other studies that show a lower a dult size/weight at higher rearing temperatures in A. aegypti (Rueda et al., 1990; Tun-Li n et al., 2000). From 20-28C the effect of temperature, although monotonic with respect to the means, was small and of variable magnitude among 2C intervals, with large overlaps in the 95% CIs among all treatments (Figure 4-3). At 30C, however, there was a noticeable reduction in WL4 As with Apup these observations may be highly influenced by unknown differences in sex ratios among temperatures. We note that pupae were included in the average WL4 measurement in the following percentages of the total number of individuals (n=30) measured: 20C: 3.3%, 22C: 6.7%, 24C: 47%, 26C: 40%, 28C: 3.3%, 30C:10%. With the exception of 24 and 26C, pupa l weight was lower than the group means, suggesting that most of the pupae were males. However, differences in the proportion of pupae included in the WL4 measurement could be caused either by sex-ratios or slight non-linearities in the effects of temperature on developmental rate that may be accentuated due to our course, 24 h interval between observations. Th is is supported by the independe nt observation in Experiment 79

PAGE 80

A of a high fraction that pupated at 2 4 (82. 5%) and 26C (100%) pupation at each respective Apup. 4.2.1.4 Temperature effects on starvation resista nce through the course of development (Experiment D ) Starvation survival in the 40 larvae in each temperature-feeding treatment was characterized by obtaining maximum likelihood (ML) parameter estimates of a Weibull distribution of time to failure. The We ibull hazard rate is defined as: 1 ,, ,,,)/)(/(),|(ATk ATATAT ATATtk kth (4-1) where ),|(,, ATATkth is the hazard of death of larvae in each time period (t) subsequent to transfer to distilled water, given the estimat ed parameters of the Weibull distribution ( kT,A, T,A), kT,A is the shape parameter and T,A is the scale parameter, both of which were estimated via ML analysis of observed survival in the 40 larvae starved in each temperature (subscript T) and feeding time (subscript A) treatment. Under a constant shape parameter ( k ), increases in the scale parameter ( ) will push the distribution towards the right and flatten h(t) whereas decreasing raises the distribution and pushes it to the left. Graphically, is found at the time with the maximum hazard of death and is th erefore referred to as the charac teristic life. By contrast, the shape parameter describes the trajectory of th e hazard over time (decr easing to zero if <1, increasing to infinity if >1), which in these expe riments is expected to increase as starved larvae have increasingly fewer energy reserves to ma intain basal metabolism (k>1). The maximum feeding day from each temperature whose surviv al data was employed for model fitting had at least 10 larvae that experienced starvation mortality without pupating. This was 3, 4, 4, 4, 5 and 6 for 30, 28, 26, 24, 22 and 20C respectively. 80

PAGE 81

Since the accum ulation of su fficient energy reserves determines both pupation and starvation resistance (Gilpin and McClelland, 1979; Telang et al., 2007), we assumed that larvae that pupated in distilled larvae had higher energy reserves upon removal of food than those that did not pupate. Therefore ML estimates of the We ibull survival (scale parameter) assumed that all larvae that pupated after being placed in dist illed water are right censored at the starvation time of the longest surviving larv a that did not pupate This assumption overlooks other factors, such as overall weight or genetic factors that ma y also contribute to starvation resistance. By contrast, interval censoring would have assumed th at larvae that pupated had the same hazard as those that did not, which is clear ly contrary to the evidence linking energy re serves and pupation. Alternatively, mean starvation re sistance estimates exclude pupae, which are more likely to be males, and would have considerably reduced sample size in many cas es. Although we know of know studies that examine sexual dimorphism in starvation resistance in mosquito larvae, because males accumulate lipids more efficiently and weigh less than females (Timmermann and Briegal, 1999), they may have heightened starvation resistance. Thus, while none of these alternatives are ideal we c onsidered that treating pupae as right censored was the best representation of starvation resistance in the original 40 larvae cohort. In unfed larvae, starvation resistance, as measured by the Weibull scale parameter ( ), decreased as temperature rose above 22C; however, it dropped off steeply at 20C, which may approach the cold toleran ce of the tropical, sea level A. aegypti strain used (Figure 4-4). Figure 4-4 also depicts changes in re lative starvation resistance of unfed vis--vis 1-day fed larvae across temperatures. At 20 and 22C starvation resistance was similar or slightly higher in unfed as compared to 1-day fed larvae. At 24 and 26C, survival was lower after one day of feeding 81

PAGE 82

with no overlap in the 95% CIs of at 24C and only a slight overlap at 2 6C (Fi gure 4-4). At 28 and particularly 30C, however, survival was clearly highe r in the 1-day fed group (Figure 4-4). In general, the data support the logical assump tion of an increased risk of mortality as the starvation period increases and la rval energy reserves are deplet ed, with k>2 for virtually all temperature-feeding treatments. The exception to th is pattern occurred in 1-day fed larvae at 20 and 22C, which had a bimodal hazard, with high numbers of larvae dying both at the beginning and at the end of the starvation period. This generated the large confidence intervals for the scale parameter in these treatments (Figures 4-4 and 45) and is also reflected in the shape parameter estimate in 1-day fed larvae at 20C (1.3, 95% CI: 1.0, 1.6) and 22C (0.78, 95% CI: 0.61, 0.98), much lower in comparison to all other feeding a nd temperature regimes. For example, the shape parameter (95% CI) for starvation resistance in 1-day fed larvae in the other temperatures was 3.1 (2.5, 3.9) at 24C, 3.3 (2.6, 4.1) at 26C, 5.5 (4 .2, 7.0) at 28C and 6.1 (4.9, 7.7) at 30C. A rise in starvation resistance was observed over the course of deve lopment, consistent among temperature treatments (Figure 4-5). In 1-day fed larvae starva tion resistance was approximately constant across temp eratures. In the 2-day fed group was 7.6 (6.9-8.4) at 24C and 8.5 (7.9-9.1) at 26C, in comparison to 10.1 (9.5-10.7) at 22C, 9.6 (9.1-10.1) at 28C and 11.0 (10.1-12.0) at 30C (Figure 4-5). In the 3 and 4-day fed groups, starvation resistance was clearly higher in the th ree warmest temperatures (Figure 45). It should be noted that right censoring for pupae occurred in the Weibull surviv al estimate (Figure 4-5) on the final feeding day at each temperature except for 24C. At this temperature no pupae were observed in the group starved after 4-days feeding, but in th e 5-day fed group 38 of 40 larvae (95%) pupated after transfer to distilled water (Figure 4-1). 82

PAGE 83

4.2.2 Model of Growth and Energy Storage in Aedes aegypti (L.) We developed a model of A. aegypti larval development that simulated growth and energy storage under our experimental conditions in order to achieve following objectives: 1) use the model to explore potential trad eoffs between development rate, size, starvation resistance and energy reserves; 2) use the experimental data to determine the temperat ure-dependence of model parameters and the sensitivity of the model fit to parameter variation; 3) use the fitted model to predict temperature impacts on energy reserves and pupation rate under a continuum of food scenarios. Through a set of discrete time equations, th e model tracks larval weight and energy storage until pupation in i ndividual larvae. Food assimilated into biomass may be partitioned into structural growth, stored energy or basal meta bolic needs (Kozlowski et al., 2004). Metabolic costs, which depend on weight, are paid first (Kozlowski et al., 2004), and thus excess resources are allocated between gr owth and energy storage, both of wh ich contribute to weight gain. As evidenced in prior studies, energy stores determ ine both starvation survival and the timing of pupation (Wigglesworth, 1942, Gilpin and McCl elland, 1969; Telang et al, 2007). Based on the hypothesized tradeoff between energy storage an d size, we also model the rate of food assimilation as a function of energy stores. Theref ore a change in energetic requirements and/or food assimilation efficiency may affect growth rate directly or indi rectly by modifying how larvae optimize the allocation of resources betw een structural growth and energy storage. We estimated and constrained parameters for the model using the results of Experiments A-C, simulated growth and starvation in individual la rvae, and then assessed the fit of predicted starvation resistance to temperat ure and age specific data from Experiment D. Given larval development times on the order of 5-9 days (R ueda et al, 1990) we considered 6 h as a biologically reasonable time step for tracking A. aegypti growth and energy reserves (Gilpin and 83

PAGE 84

McClelland, 1979). Simulated growth and survival were sum med over 24 h intervals in order to compare model outputs with daily experimental observations. 4.2.2.1 Nutrient assimilation into biomass For simplicity, larvae do not incur metabol ic costs for searching or handling food, corresponding roughly to our experimental c onditions (see above). In mosquitoes and A. aegypti in particular, our own data (Fi gure 4-2) and numerous other studies indicate a sigmoidal growth trajectory, with a burst in growth at approximately the molt to fourth instar (Dye, 1984; Nishiura et al., 2007; Telang et al., 2007). Equation 4-2 describes the rate of nutrient assimilation when nutrients are not limiting. In accordance with the sigmoidal weight trajectory, nutrients assimilated [hereby referred to as food assimilated] ( FA,i,t) for each larva i of feeding age A per unit time are assumed to take on a Gaussian function over the course of development when larvae are assimilating food at their physiological capacity. ]2/)[( ,,22thtEE titiAegWF. (4-2) Under unlimited food conditions, su ch as those assumed in our feeding experiments used to parameterize the model, the food available to la rvae is greater than the right side of Equation 4-2. In later simulations we use the model to predict the impacts of temperature in food limiting conditions, in which food assimilated may be less than this quantity. The proportion of total body weight ( Wi,t) that can be assimilated into biomass when food assimilation is at a maximum is te rmed maximum growth efficiency ( g ). The key assumption that links growth and energy stores over the cour se of development is that the location of the hump in assimilation rate is triggered by a threshold level of energy stores ( Eth). Therefore increasing energy stores before reaching Eth will tend to increase food assimilated, whereas 84

PAGE 85

energy sto rage after Eth (during the latter part of the L4 stage) will decrease food assimilated. This assumption is supported by Timmermann an d Briegel (1999), whose experimental work showed decreasing lipid assimilati on rate in the last instar of A. aegypti, which according to Equation 4-2, would increase grow th in the fourth instar as Ei,t would remain in the vicinity of Eth. Moreover, Telang et al (2007) showed that larv al weight upon molting to the last instar is approximately 20% of the final weight attained under optimal feeding conditions; this suggests that a sigmoidal function is a be tter description of weight trajecto ry than an exponential or linear to plateu-type shape. The spread parameter ( ) determines the trajectory of increase and decrease of the food assimilation rate as Ei,t approaches and surpasses Eth. Weight Equation 4-3 describes weight ( g) dynamics of food assimilated into body mass and the loss due to consumption of en ergy reserves in order to meet basal metabolic needs. m timtiAttiWcFWW )(, ,, 1,. (4-3) The parameter cm is the coefficient of metabolic e xpenditure as a proportion of larval weight ( Wi,t) scaled by m the exponent describing weight dependence of metabolism or metabolic allometry (Bokma, 2004). These paramete rs are roughly analogous to those describing catabolism in the original Bertalanffy (1960) mode l. However, as customary in models optimal resource allocation (Kozlowski et al, 2004), we ight is lost only under conditions when food assimilated does not meet the energetic demands of basal metabolic functions (Eq. 4-5). 4.2.2.2 Stored energy dynamics Allocation of resources towards stored energy, represented in units of mass ( g), is described in Equation 4-4 when food assimila ted is greater than the minimum metabolic requirement. Equation 4-5 describes the expenditu re of stored energy when food assimilated 85

PAGE 86

does not cover m etabolic requirements. Larval deat h from starvation occurs in the time interval immediately after energy reserves are depleted to zero. sm timtiAsttiWcFcEE ])([, ,, 1, if ))((, ,, m timtiAWcF (4-4) m timtiAttiWcFEE )(, ,, 1,if ))((, ,, m timtiAWcF (4-5) The efficiency of energy storage ( cs) represents the proportion scaled by the exponent s, of food assimilated above the minimum metabolic requirements that a larva allocates towards energy storage. Although not formal ly tracked in the model, this formulation assumes that all food assimilated that is not allocat ed towards energy stores is devot ed to structural growth, such that total weight ( Wi,t) is the sum of energy reserves ( Ei,t) and structural mass ( Si,t),: tititiSEW,,, (4-6) We model energy storage as a stochastic pr ocess with independent trials for each 0.1 g of food assimilated. In mosquitoes energy reserv es may take the form of carbohydrates, lipids or amino acids, each of which may vary with diffe rent food sources (Wigglesworth, 1942; Nayar and VanHandel, 1971; Timmermann and Br iegel, 1999). For simplicity, given the heterogeneous, field collected detritus used in th e experiments, we assume that each particle of food assimilated has the same probability of being converted into energy reserves. The probability that each piece of assimilated food (Eq. 4-2) is allocated towards energy stores in (Eq. 4-4) is therefore assumed to take on a binomial distribution with a probability p equal to the coefficient of energy storage ( cs) and a number of trials n equal to the scaled amount of food assimilated above maintenance costs within every 6 h interval. Given our assum ption that all food assimilated has the same pot ential to be converted into energy reserves, sm mtiAWcF ) (,, 86

PAGE 87

the valu e of the parameter cs represents the optimal proportion of excess food assimilated at a particular weight that a larva allocates toward s energy storage instead of structural growth. 4.2.2.3 Pupation Pupation is modeled in latter simulations in which we use the model to predict development rate under an array of feeding c onditions. Stored energy is assumed to trigger pupation (Gilpin and McClella nd, 1979; Klowden and Cham bers, 1990; Timmermann and Briegel, 1999; Nishiura et al ., 2007; Telang et al., 2007), such that larvae pupate when stored energy reaches Epup. Unlike other studies of A. aegypti (Chambers and Klowden, 1990) we do not assume a threshold, or criti cal weight required for pupation, as this is unlikely to be directly involved in the physiological mechanism of pupati on and may also vary with food availability (Gilpin and McClelland, 1979; Nishiu ra et al, 2007). Alt hough critical weight has been shown to increase with lower temperature (Chambers and Klowden, 1990) the relationship between Epup and temperature has not been studied to our knowledge. Therefore, we did not make any temperature assumptions with regards to Epup, per se although, based on the data on commitment to pupation (Experiment A, Figure 4-1), we did assume that the rate of accumulation of energy reserves increased with temperature (see model fitting section below). An additional criterion for pupation is a temperature-dependent minimum development age Apup (Experiment A). Larvae that reach Epup will only pupate if they have also completed Apup, whereas larvae that reach Epup after Apup pupate at the beginning of the next time interval. 4.2.3 Parameter Estimation and Model Fitting Using Experimental Data 4.2.3.1 Summary of model fitting strategy and assumptions of temperature effects on parameters The model fitting process consisted of specifying and constraining parameters in order to match Experiments A, B, C and starvation resistance of unfed larvae, a nd then using ML analysis to 87

PAGE 88

assess the fit of the m odel to Experiment D, sepa rately for each temperature. Therefore, model fits to Experiment D are all predicated on the observations that: (1) risi ng temperature induces commitment to pupation with a shorter feeding time (Experiment A, Figur e 4-1), (2) at 28C, larval weight gain in optimal feeding conditions follows a sigmoi dal trajectory (Experiment B, Figure 4-2), (3) the TSR applies in A. aegypti with a monotonically decreasing WL4 as temperature rises from 20-30C (E xperiment C, Figure 4-3) and (4 ) starvation resistance of unfed larvae decreases as temperature rises from 22 to 30C (Figure 4-4). Because pupation requires the accumulation of sufficient energy reserves and starvation mortality occurs when energy reserves are depleted, the model also assumes, based on observations (1) and (4) above, that energy reserves rise faster in feeding larvae and are spent faster in starved larvae as temperature rises. Accordingly our parameter estimates, sens itivity analyses and model predictions suggest how the observed temperature effects on A. aegypti development are brought about and the implications for temperature effects across a sp ectrum of resource conditions. Implicitly, we assume that neither age nor food limitation affects growth and metabolic parameters. Mathamatical explanations of the TSR s uggest that temperature may affect the coefficients and/or exponents of catabolism and anabolism, which are usually modeled as the processes that define the rate of growth (Bertalanffy, 1960; Berrigan and Charnov, 1994; Perrin, 1995 Angilleta and Dunham, 2003; Angilletta et al 2004; Kozlowski et al., 2004). Similarly we also assume that temperature may affect the coefficients ( cm and cs) and/or exponents ( m and s ) of catabolic and anabolic activ ity, with the key difference that in our model these processes determine the dynamics of stored energy, which only partially cont ributes to total mass and also affects food assimilation rate. We also assume that temperature may affect maximum food assimilation efficiency (g ), an assumption well supported in ectotherms (Angilleta and Dunham, 88

PAGE 89

2003) and in A. aegypti in particular (Rashed an d Mulla, 1989). In our m odel g determines the peek of the Gaussian food assimilation rate wher eas the other other two parameters involved in the growth trajectory, Eth and determine how the feeding rate va ries with energy reserves. For simplicity in the absence of evidence to the contrary, we assume that both Eth and are temperature independent. These assumptions re garding the temperature dependence of model parameters suppose that temperature may directly affect weight gain by affecting the overall efficiency of food assimilation or indirectly by af fecting the rate of consumption or build-up of energy reserves; however, the assumed temperature independence of Eth and means that temperature will not affect how f ood assimilation responds to a particular level of stored energy. We divide the parameters (Table 4-1) in to four categories: initial conditions (Wo, Eo), assumed temperature independent parameters ( Eth), temperature dependent parameters specified to match experimental data (cm, s), and temperature dependent pa rameters evaluated through ML model fits to observed starvation survival and 28 C weight trajectory after successive feeding days ( g m cs). Table 1 summarizes the method of estimation of each parameter. 4.2.3.2 Initial conditions (Wo, Eo) Initial weight ( Wo) was assumed to be the same across larvae and set to the mean of the group means measured in newly hatched larvae (unfed larvae in Experiment C). Since no information was available on the energy stores present in newly hatched mosquito larvae ( Eo), we arbitrarily assigned an average Eo equal to 40% of Wo. Genetic variability and varia tion in egg resources are tw o potential sources in which Wo or Eo may vary among larvae. However since we ight in newly hatched larvae was not individually measured (due to the limits of our weighing device), the only information we had that would allow us to infer variation in initial conditions was in the starvation survival of newly hatched larvae. Because starvation resistance involves depletion of energy reserves, we 89

PAGE 90

incorporated this variation into the assignati on of initial energy re serves for each larva ( Eo). Using Equation 4-5, we chose cm so that stored energy of a st arved, newly hatched L1 would reach zero at a time corresponding to observed mean starvation resistance at each temperature. Weight dependence of metabolism, represented in the exponent m was set at 2/3 the value assumed in prior A. aegypti models (Gilpin and McClelland, 1979). [However, this assumption was later relaxed when the model fit was compared to experimental data at all six temperatures (see below).] Using T,A and kT,A, estimated based on starvation of unfed larvae in Experiment D (Figure 4-4, Table 4-1) we generated a dist ribution of starvation su rvival time upon hatching among 100 larvae in each of the si x temperatures. Having assumed m and fit cm this enabled us to use Equation 4-4 to iteratively generate an Eo value for each simulated larva. Histograms of the 600 Eo fits were then employed to construct a temperature-independent distribution of initial energy reserves. From this di stribution we randomly assigned Eo to each larva in all subsequent simulations. 4.2.3.3 Temperature dependent parameters chosen to match data ( cm, s ) As described above, the coefficient of metabolism ( cm) was chosen using Equation 4-4 to match mean starvation survival of the 40 newly hatched larvae observed at each temperature (unfed larvae in experiment D), for a given value of m This approach, instead of choosing cm for each individual larva, ensured that cm and Eo were not coupled as an artifact of the fitting process. Figure 4-6 gives the joint cm versus m parameter space that reproduced starvation resistance of newly hatched larvae at each temp erature. Starvation resistance of newly hatched larvae is much more sensitive to the coefficient th an the exponent of metabolism, such that at the most common ranges of the exponent m evoked by ecologists (0.5-1) (Bokma, 2004; Irlich et al., 2009) cm must take on a narrow range between 0.02 and 0.08 at all temperatures. The 90

PAGE 91

interdepend ence of cm and m given by the slope of the curves in Figure 4-6, roughly decreases with lower temperature and higher values of m For a given m value, the relative value of cm among temperatures corresponds inversely to the relative starvation survival of unfed larvae among temperatures (Figure 4-4); th at is, the fit value of cm increases as temperature rises fr om 22-30C (Figure 4-6). At 20C cm lies between the 26 and 28C estimates (Figure 4-6). Our approach to fitting cm was based on the assumption that the differential survival of unfed larvae among temperatures reflected differences in the energetic costs of metabolic maintenance. However, increased metabolic requirements is an unlikely cause of higher mortality of unfed larvae at 20C as co mpared to 22-26C (Figure 4-4). Genetic factors related to the cold tolerance of the A. aegypti strain used are a more likely cause, but not accounted for in the m odel. We randomly assigned a value for WL4 to each simulated larva, based on histograms generated from the 30 larvae wei ghed in each temperature (Experi ment C). The simulation chose a value of s for each larva so that it w ould comply with its assigned WL4 This exponent determines whether energy allocation levels off ( s < 1), is proportional to ( s = 1), or increases ( s > 1) with more food assimilated in the latter portion of the L4 st age; we considered it biologically unreasonable for energy storage to decrease with incr eased weight, and thus model fitting was constrained to non-negative s values. 4.2.3.4 Temperature independent parameters specified ( Eth, ) Because we were unable to find studies that describe feeding efficiency or the trajectory of energy reserves in ea rly larval instars, we arbitrarily specified Eth as approximately ten times Eo, or 15 g. We subsequently assumed a value that produced a mean larval weight trajectory that was roughly consistent (vis ually) with observed group mean weight of larvae fed for 1, 2 and 3 days at 28 C (Experiment C). For this analysis the metabolism exponent (m ) was varied 91

PAGE 92

from 0.5 to 0.75 (Bokma, 2004), as opposed to assu ming the 0.66 value that was used to estimate variation in Eo. We made this modification since weight trajectory is temperature dependent, and a number of studies have suggested temper ature dependence in the catabolism exponent (Angilleta and Dunham, 2003; Kozlowski et al, 2004). Growth efficiency ( g ) was varied from 0.3 to 0.8 and the energy storage coefficient ( cs) from 0.1 to 0.4; the metabolism coefficient cm and energy storage exponent ( s ) were chosen to match mean st arvation survival of unfed larvae and WL4 at 28C, respectively, given the value of m g and cs in each trial. This analysis produced similar weight trajectory to observed data when the value was in approximately the 17-28 g range. This large spread ( ) in the Gaussian curve relative to the mean ( Eth) indicates a gradually increasing and decreasing food assimila tion rate as energy re serves approach and surpass Eth. By contrast small values of would have produced a low food assimilation rate at the beginning of development and a steep rise and drop in the vicinity of Eth. Using this information we assumed = 23 and Eth = 15 g for all temperatures th roughout the rest of the process of model fitting and generation of predictions. 4.2.3.5 Estimation of temperature-d ependent parameters (m, cs, g) through ML comparison of model to experimental data We obtained ML estimates for the metabolic exponent ( m ), maximum growth efficiency ( g ) and the coefficient of energy storage ( cs) by comparing starvation resistance in larvae fed successively for one to six days (Experiment D) to that of larvae simulated using Equations 4-2 to 4-5. At 28C this comparison was constraine d to parameter spaces that also maximized the likelihood of the observed weight trajectory (E xperiment B). At all other temperatures the parameter space was constrained to ensure a fast er rate of increase in energy reserves with increasing temperature, consiste nt with our data on commitme nt to pupation (Experiment A). 92

PAGE 93

In order to fit the m odel to experimental da ta on starvation resistance at each feeding age we used the following temp erature-specific function: (4-7) max11 ,, ,),|(lnA F n i ATATi THkttpr LLwhere LLH,T is the log-likelihood that death hazard of simulated larvae follows the Weibull probability density function estimated from observed data at temperature T ti is time to death from starvation of each simulated larva, assumed to occur at the beginning of the time interval after energy reserves fall to zero, n is the number of larvae in each temperature-feeding age, F is feeding age the days reared from hatching in feeding cups prior to starvation and Amax is the maximum feeding age that was starved at each temperature. This formulation summed the Weibull probabilities of the pr edicted time of death of each simulated larva, based on the observed T,A and kT,A. ML fits maximized the mean LLH,T among 100 simulations (n = 40 larvae in each) for each parameter combination. In order to ensure that the model fits comp lied with the observed weight trajectory, ML estimates of g m and cs at 28C involved maximizing the sum LLH,28 + LLW,28, where the latter term refers to the likelihood that simulated larv ae comply with the observed weight trajectory from 1 to 3 days feeding (weight of the 4-day fed group was used to fit s ) (Experiment B). Since weight was measured at the group level for thes e treatments, we determined the likelihood that the observed mean and variance in weight among groups described the distribution of group weights in a pool of simulated larvae that were randomly resamp led and placed into groups of the same size used for weighing. This is su mmarized in the following log-likelihood function: 3 1 40 1 2 ,, 28,),| (lndi dgdgi Wxxpr LL, (4-8) 93

PAGE 94

where is the total weight of group i consisting of n resam pled larvae (n=5 for 1-day fed, 4 for 2-day fed and 2 for 3-day fed larvae), ix g,d is the observed mean weight among groups of larvae after d feeding-days and 2 g,d is the observed variance. At all other temperatures (20, 22, 24, 26 and 30C) the estimation of g m and cs (through maximization of LLH,T) was constrained to ensure compliance with commitment to pupation (Experiment A). This was achieved by first fitting the model at 28 C and then assuming that that the minimum percentile that pupated at a higher temperature had on average (among 100 simulations) more energy reserves than the maximum percentile that did not pupate at a lower temperature (Figure 4-1). Energy st orage constraints for 30, 26 and 24 C were defined based on the ML 28 C scenario, at which parameters were i ndependently estimated based on both weight trajectory and starvation survival For example, in Experiment A among larvae starved after four days feeding, 92.5% pupated at 30C, 72.5 % pupated at 28C, 50% pupated at 26C and none pupated at 24C (Figure 4-1). Therefore, at 26C the joint g m and cs parameter space was constrained such that the average value of the 50th percentile in energy reserves after four days feeding was lower than the average 22.5th percentile ML value at 28 C; at 24C we considered only the parameter space in which the average valu e of the highest energy reserves after four days feeding was lower than the 22.5th percentile value at 28 C, and at 30C the 7.5th energy percentile was constrained to be higher than the 22.5th percentile value at 28C. Because of the large gap between time to commitment to pupation between high and low temperatures, constraints at 22 C were defined based on ML 24 C estimates and 20 C constraints were based on ML 22 C estimates. Pupation was not allowed in the model fits to Experiment D, as Epup was fit a posteriori based on the predicted energy trajector ies (see model prediction section). 94

PAGE 95

At all tem peratures we conducted first a likeli hood analysis of the fu ll range of the joint m cs and g space, varying each at 0.05 intervals between 0.05 and 1 (i.e. 8000 parameter combinations). A narrow range, joint ML parame ter space was defined by taking the two values for each parameter that yielded the highest ov erall likelihood among the 8000 combinations, and adding 0.05 to the higher value and subtracting 0. 05 from the lower value. This yielded a joint parameter space of 0.15 for all three parameters at all temperatures, ex cept for the metabolism exponent ( m ) at 20C and growth efficiency ( g ) at 24C, both of which had an ML narrow range space of 0.2 (Table 4-2). The definition of this narrow range was not constrained by the assumption of energy reserves with respect to Experiment A, as described above. We subsequently carried out a likelihood an alysis over the narrow range ML parameter space (Table 4-2), varying each of the three pa rameters at 0.01 intervals (i.e. 4096 parameter combinations at 22, 26-30C and 5376 combinations at 20 and 24C). This was repeated 100 times for each temperature, constraining the analysis only to ML parameter combinations that were compliant with the energy reserves assu mption regarding pupation in Experiment A. The estimates of g m and cs used to generate model predictions was considered the mean of the 100 ML values generated for this narrow range anal ysis and 95% CIs were set at the 2.5 and 97.5 percentiles (Figure 4-7A). The metabolic coefficient cm was assumed constant among all simulated larvae and was fit at each temper ature to reproduce obser ved mean starvation resistance in unfed larvae given the ML estimate of m. The energy storage exponent s was fit to match the randomly assigned WL4 of each simulated larva, given the ML estimates of g m and cs and the fit of cm. Mean and SD of the fit value of s among individual larv ae are reported in Figure 4-7B. The values of Eo, and Eth were assumed as descri bed above (Table 4-1). 95

PAGE 96

Log-like lihood estimates (LLH,T + LLW,28 for 28C and LLH,T for all other temperatures) were the sum of 40 larvae fed until Amax at each temperature, averaged over 100 simulations for each parameter combinations (Equations 4-7 and 4-8). Averaged maximum likelihood estimates were the following: 20C: -553.0 ( Amax = 6), 22C: -472.9 ( Amax = 5), 24C: -332.0 ( Amax = 4 days), 26C: -329.5 ( Amax = 4), 28C: -555.1 ( Amax = 4 plus 1-3 days feeding for weight) and 30C: -219.7 ( Amax = 3). Likelihoods were in creasingly negative in t hose temperatures with more data included (Bolker, 2008). For example LLH,28C + LLW,28C summed 280 larvae as compared to only 90 larvae for LLH,30C. Maximized log-likelihoods, standardized per larva, were as follows: 20C: -2.30, 22C: -2.36, 24C: -2.07, 26C: -2.05, 28C: -1.98 and 30C: -1.83. This indicates a slightly less precise model fit at 20 and 22C. ML estimates of growth efficiency ( g ) demonstrate an approximately linear increasing relationship with temperature (Figure 4-7A). The metabolic coefficient ( cm) decreases from 20 to 22 C (due to the low observed surviv al of newly hatched larvae at 20 C), and increases monotonically with temperature at 22-30 C, with an increasing slope as temperature rises (Figure 4-7A). Energy storage coefficient ( cs) is roughly constant from 20 to 26 C, but shows a sharp increasing trend from 26-30 C (Figure 4-7A). The mean estimate of the energy storage exponent ( s ) (fit to individual larvae) increases until 26 C and fall dramatically at 28 C, the temperature at which cs increased (Figure 4-7B). Estimates of s are clearly higher at 24 and 26 C as compared to the other temperatures. From 22-30C, the exponent of metabolism (m ) fluctuates from 0.5 to 0.58, at the lower end of the 0. 5-0.75 range commonly cite d by ecologists. At 20 C, however, the value of m is much lower and inconsistent with the literature (B okma, 2004). This is an artifact of the joint dependence of m and cm to starvation resistance of unfed larvae (Figure 4-6), which was unexpectedly low at 20C (Fig ure 4-4), resulting in a large fit value for cm in 96

PAGE 97

com parison to 22C (Figures 4-6 and 4-7A ). Notably, across all temperatures m is significantly smaller than s, indicating that while energy storage varies close to li nearly with increasing food assimilated, as overall weight increases, basal metabolic demands are proportionally lower (Eq. 4-4 and Eq. 4-5). Figure 4-8 depicts the comparison of simulate d and observed mean weight trajectory at 28C and starvation resistance at all temperatures, using parameter values shown in Figure 4-7. In order to directly compare simulated and observed means in starvation treatments with pupation, simulated means in Figure 4-8 were calculated without the longest surviving larvae in each temperature-food category, corresponding to th e number of observed larvae that pupated in starvation in Experiment A (i.e. percent of la rvae truncated in each treatment corresponds to pupation percentage in Figure 41). Figure 4-8 shows that usi ng ML parameter estimates, the model on the whole adequately reproduced weight trajectory at 28C (Experiment B) and starvation resistance at each temp erature (Experiment D), with simulated mean survivals located well within observed standard devi ations and vice versa. Slight parameter changes (not shown) within the narrow ML ranges (Table 4-2) permitted a closer fit between observed and simulated means at all temperatures, but these shifts gene rated consistently lower likelihoods, attesting to the differences between mean survival and the Weibull hazard function. Deviations between simulated and observed data are particularly evident at 20 C. In concert with increased confidence intervals for estimates of both m and cs (Figure 4-7), this further calls into question the biological validity at 20C of our assumpti on equating starvation resi stance of unfed larvae with basal metabolic parameters cm and m. 97

PAGE 98

4.2.4 Sensitivity Analyses 4.2.4.1 Effects of parameters on weight, starvation survival and energy reserves In order to characterize tradeoffs between gr owth, starvation survival, the mass of energy reserves and the proportion of weight dedicated to energy reserves (E/W) in feeding larvae, we determined the independent effect of each para meter on each of these four outcomes. This analysis focused on the behavior of the determin istic model in which stochastic energy storage was eliminated; thus, cs represented the proportion of all food assimilated into biomass in each 6hr time step that went towards energy storage as opposed to structural growth (Equation 4-6). We also excluded individual variation in Eo and WL4, simulating one larva representing mean conditions. All temperature dependent parameters were independently varied to % of their 28 C ML estimate in order to assess their effect on model behavior be yond the range of ML parameter estimates, and all model fitting constraints to ensure consistency with Experiments AC were also removed. The values of Eo, Eth and assumed in Table 4-1 were maintained. Since individual variation was removed in this an alysis, we used as the baseline value of s the mean value given in Figure 4-7B. Overall, maximum growth efficiency ( g ) (green curve in Figures 4-9 to 4-12), had a strong positive effect on weight ga in (Figure 4-9) but had little impact on starvation resistance (Figure 4-10). In contrast, the paramete rs describing basal metabolic activity ( cm and m red shades) had a large negative impact on starvation resistance, with little impact on weight gain. Increases in energy storage parameters ( cs and s, blue shades) clearly cause a tradeoff between growth and starvation resistan ce, by favoring starvation resist ance throughout development but inducing a sharp decline in size after 3 and 4 days feeding (Figure 4-9C and 4-9D, Figure 4-10). Interestingly, weight after 4 days feeding shows more sensitivity to energy storage parameters (negative relationship) than growth parameters (positive relationship) (Figure 4-9D). This 98

PAGE 99

indica tes that in the late fourth instar, the indirect effect on weight gain of increased energy storage on reducing food assimilated outweigh the direct effects of heightened growth efficiency (Equation 4-2). These results indica te that relative to the 28C parameter estimates, reduced size at higher temperatures cannot be explained the effects of temperature on the metabolic parameters ( cm and m ) in A. aegypti as postulated by Perrin et al (1994). Moreover, since higher temperature increases the maximu m food assimilation efficiency ( g ) (Figure 4-7A), reduced size at higher temperatures in the vicinity of ML parameter estimates can only be attained through increased allocation of resources towards energy stores. At lowe r temperatures reduced energy storage parameters will increase food assimilation in L4 by maintaining energy reserves closer to Eth (the peek of the food assimilatio n trajectory); moreover, since both cs and s are less than one, a larger fraction of the heightened food assimilated will be devoted to structural growth, as the proportion of food devoted to energy storage decreases. Figures 4-11 and 4-12 show the impact of parameter variation on absolute energy reserves ( E ) and percent energy reserves ( E/W ). Although neither was expe rimentally evaluated, the former is directly linked to commitment to pupation (Gilpin and McClelland, 1979; Chambers and Klowden, 1990; Telang et al., 2007), whereas the latter is a likely indicator of the ability of the larva to withstand a reduction in resources at a particular stage of development. Metabolic parameters (red shades) clearly have litt le effect on energy reserves or percent energy reserves in non-starvation c onditions (Figures 4-11 and 4-12). In contrast, increased maximum growth efficiency generates cl ear tradeoffs for larvae between the quantity and proportion of energy reserves. After 1-3 days feeding g has the strongest positive effect of all parameters on absolute energy reserves (Figure 4-11), but due to a stronger im pact on weight, has paradoxically the strongest negative impact on proportion energy reserves (Figure 4-12 ). This indicates that 99

PAGE 100

tem perature driven increases in maximum growth efficiency (Figure 4-7, Table 4-2) will likely decrease time to maturation, but may reduce th e ability of larvae to withstand food limited conditions. Extremely low values of g have highly irregular impact s on percent energy reserves (Figure 4-12) due to non-linear and differential impacts on weight (Figure 4-9) and total energy reserves (Figure 4-11). In the vicinity of 28C estimates, increasing g has a similar effect on total energy reserves as increasing cs after 1-day feeding (Figure 4-11A), but clearly has a la rger capacity to increase energy reserves after 2 days feeding and beyond (F igure 4-10B-D). The reason for this pattern is that g has a strong positive effect on food assimilated whereas cs and s decrease food assimilated in the latter portion of development. Figures 4-11 and 4-12 also exhibit a change in the sensitivity of percent energy reserves to the coefficient ( cs) vis--vis exponent ( s ) of energy storage as development progres ses. After one day feeding cs improves both total and percent energy reserves, whereas s has little effect (Figures 4-11A and 4-12A). After two days feeding and beyond cs and s have relatively similar effects on percen t energy reserves in the vicinity of their 28 C values (Figures 4-11B-D and 4-12B-D). Together with the estimated temperature de pendence of the parameters (Figure 4-7), these results indicate that te mperature-induced increases in development rate and metabolism have energetic costs that may in fluence larval capacity to mature in food-limited environments, assuming that the parameters do not vary with food conditions. By in creasing energy storage with heightened temperature, larvae may energetically compensate for reduced energy reserves (due to an increased g ) and starvation resistance (due to an increased cm), but at the expense of final size, as predicted by the TSR. Such compen satory energy storage may occur either at the initial or final stages of larval development, through cs or s, respectively. 100

PAGE 101

4.2.4.2 Sensitivity of observed starvation resistan ce to the efficiencies of growth and energy storage Figures 4-9 to 4-12 show that in the model system, in th e face of increasing maximum growth efficiency ( g ) (Figure 4-7), attaining smaller fina l size at higher temperatures requires heightened energy storage. Our next step was to evaluate the sensitivity of the observed starvation resistance to simultaneous changes in growth and ener gy storage. We explored the joint parameter space of maximum growth efficiency (g ) and energy storage efficiency (cs) that reproduced observed mean starvation survival in each feeding-temperature regime. For this analysis groups of 4000 larvae were simulated across the parameter space of cs and g (step of 0.15) including the ML estimates of each at all six temperatures. We employed the stochastic model with individual vari ations in initial lipids and final weight as pr eviously described, in which the exponent of energy storage ( s) was individually fit to match asymptotic weight for each csg combination. As previously described cm was chosen to match mean starvation resistance in unfed larvae at each temperature, given the value of m (see below). The curves in Figure 4-13A-F depict the line bisection between the areas of the csg space that generated a mean starvation survival greater and less than the observed mean. For example, at 26 C (Figure 4-13D), mean starvation resistan ce of 4-day fed larvae (yellow curve) was reproducible within an extremely narrow range of g (~0.4) but through nearly the entire cs range. Survival of 2 and 3-day fed larvae (green and red curves in Figure 4-13D) shows a larger interdependence and clear ne gative relationship between cs and g indicating a compensatory effect of the efficiency of growth vis--vi s energy storage during the middle stages of development. Survival of 1-day fed larvae (blu e curve in Figure 4-13D) is reproducible only when cs is the vicinity of 0.13, whereas it can be reproduced across a wide range of g values. The energy storage exponent, s, clearly impacts the interd ependency of the data on cs and g When 101

PAGE 102

the system is insensitive to either of these parameters (at the beginning of development for g and the end of development for cs) they are both associated with s, although the association is much greater with cs than g (i.e. in Figure 4-13D the small black numbers decline more rapidly going down the yellow curve than from right to left on the blue curve). This indicates that the insensitivity of the system to either parameter, particularly cs, is due to the potential for compensatory effects in the exponent s. These effects of feedin g age on the differential sensitivities of the system to g and cs hold at all temperatures (Figures 4-13A to 4-13F). At each temperature m was set at a value in which th e curves converge at a single parameter region (see legend in Figure 4-13). The convergence regions do not correspond exactly to ML estimates for each parameter (Fig ure 4-7), because the curves depict the region where simulated and observed mean survival coincide, whereas ML parameter estimates were based on the probability that simulate d larvae reproduced ML estimates of T,A and kT,A, based on the observed data. Discrepancies between model fits to the mean survival and the Weibull distribution may be caused by skew ness and/or outliers in the obser ved or simulated data, which are accentuated by the increasing death hazard ( k > 1) over time in foodless conditions. Differences in the distribution of randomly assigned WL4, affecting the fit value of s, may have also contributed to discrepancies. Nonetheless, the 1-day fed curves hover at a value relatively similar to the temperaturespecific ML estimate for cs, whereas the vertical (final day fe eding) curves persist at values similar to the ML estimates for g The increase in the level of cs for survival of 1-day fed larvae (blue line) is particularly noteworthy at 28 and 30 C. These results are consistent with Figure 410, indicating that energy st orage through the coefficient cs impacts starvation resistance at the 102

PAGE 103

beginning of larval development, whereas the exponent s can improve starvation resistance at the end of development. Figure 4-13 also suggests that at higher temperatures there is a heightened sensitivity to maximum growth efficiency at earlier feeding ages. At 28 and 30 C, the slope of the 1-day (blue) shows a slight increase (negative) as compared to the lower temperatures and that of the 2-day fed curves (green) show a marked negative incr ease (Figures 4-13E and 4-13F). The slope of 2day fed larvae at 30 C, in particular, is comparable to th at of 3-day fed larva at 24 (Figure 413C) and 26 C (Figure 4-13D). This indi cates the potential for incr eased interdependency of energy storage and growth efficiency in early instars at 28 and 30 C. Taken together with the experimentally observed increase in starvation resistance in the 1-day fed vis--vis unfed groups at 28 and 30 C (Experiment D, Figure 4-4), and the negative effect of heightened growth efficiency on percent energy reserves in the model (Figure 4-12), thes e results suggest the sensitivity of larval development to changes in specific parameters may vary with age; in particular, we suggest that above 26 C, larvae are compelled to store energy in earlier, as opposed to exclusively in later developmental stag es; this may be a compensatory mechanism for high growth efficiency. 4.2.5 Model Predictions 4.2.5.1 Weight, stored energy and reserve proporti on as a function of temperature in unlimited food conditions Using parameter estimates of g, m, cs and cm (Figure 4-7) we projected accumulated energy stores and weight across a ll 6 temperatures. As previously described, energy storage was a stochastic process and Eo and WL4 were randomly assigned to each simulated larva, with the latter used to individually [re]fit s. Mean s among temperatures corresponded well to the values in Figure 4-7 (20C: 0.73 0.12, 22C: 0.83 0.13, 24C: 0.95 0.15, 26C: 1.0 0.14, 28C: 103

PAGE 104

0.81 0.11, 30C: 0.79 0.12), as expected sinc e all of the other parameters were identical. Figure 4-14 shows means and standard deviations of weight and energy reserves through the course of development as obtained from a simula tion of 4000 larvae after 1-5 days of feeding on excess food. With warming temperature, energy re serves rise slower than weight in the early instars (Figure 4-14A and 4-14B), approximate ly proportional to wei ght in the middle of development (Figure 4-14C), and faster than weight in later inst ars (Figures 4-14D and 4-14E). Thus, the model predicts that the observed TSR in Experiment C (Figure 4-3) was generated due to the negative feedback of rising energy reserves on feeding rate in the last larval instar, in which most growth occurs (Gilp in and McClelland, 1979; Nishiura et al., 2007; Telang et al., 2007). In the latter feeding days growth rate slows substantia lly in the higher temperatures, generating the predicted leveling off of th e temperature-weight curve in 4-14D-E. Concurrent with this pattern is a changing temperature dependency across development. In early stage larvae both weight and energy rese rves increase at a higher rate in the warmer temperatures from 26-30 C, but as development progresses the slope of the temperature curves is higher from 20-26 C. This causes the gradual switch in the concavity of energy reserve and weight curves as development pr ogresses (Figures 4-14A to 4-14E). The sharp increase in energy reserves from 26-30 C after 1-day feeding (Figures 4-14A and 4-14B) parallels ML estimates for cs (Figure 4-7A) and the heightened sensitivity of energy reserves to cs at the beginning of development (Figure 4-11C). In contrast the he ightened slope of weight and energy reserves from 20-26 C in later developmental stages (Figures 4-14D and 4-14E) is lik ely a result of the increased storage exponent s at these temperatures (Figure 4-7B). Using the same simulation conditions, we predicted proportion energy stores ( E/W ) across temperature and feeding ages. Th is simulation was carried out until Apup at each 104

PAGE 105

tem perature in order to predict how E/W in newly pupated mosquitoes may vary with temperature. Figure 4-15 shows a U-shaped patt ern across temperatures after 1-3 feeding days, with lower values at 24 and 26 C as compared to the warmer and colder extremes. This is consistent with the larger increase in weight than energy reserves from 20-26 C and compensation of energy reserves from 26-30 C (Figures 4-14A and 4-14B). Taken together with Figures 4-12 and 4-13, these re sults indicate that increasing maximum growth efficiency ( g ) leads to lower percent energy reserves unless it is compensated for in the early stages of development through cs. The U-shape emerges because cs does not rise until 28 C. After 4-days feeding, however, 24 and 26 C have larger percent ener gy reserves than 20 and 22 C, presumably due to a higher energy storage exponent s. Also noteworthy in Figure 4-15 is that percent energy reserves at Apup increase roughly monotonically with increased temperature. This is consistent with the larger impact of temperat ure on energy reserves than on weight in the latter feeding ages (Figure 4-14). 4.2.5.2 Combined effects of temperature and food limitation on development rate and cumulative pupation Using ML parameter estimates a value for the threshold energy reserves required for pupation ( Epup) was fit based on Experiment A, the propor tion of larvae that pupated subsequent to transfer into distil led water in each temperature-feeding ag e. This fit was carried out using 100 simulations of 40 larvae each, using the parameter values shown in Figure 4-7, with stochastic energy storage and individual variation in Eo and WL4 ; thus, s was again [re]fit to individual larvae in each simulation, yielding very similar average s values as in Figure 4-7. We assumed that Epup corresponded to the average predicted energy re serves in the lowest percentile of larvae that pupated upon transfer to distilled water (Fig ure 4-1) Potential se x ratio biases, arising because males mature before females, were addressed by adjusting Epup to the mean energy 105

PAGE 106

stores in th e larva with the nth highest energy level in the feeding treatment in which at least 70% of larvae pupated after st arvation, where n is the number of la rvae that pupated after transfer to distilled water. This corresponde d to 4 days feeding at 28 and 30C, 5 days at 24 and 26C, 6 days at 22C and 7 days at 20C. The fit value for Epup was as follows for each temperature: 20 C: 41.36 0.16, 22 C: 51.04 0.19, 24 C: 54.70 0.18, 26 C: 53.82 0.26, 28 C: 65.29 0.15, 30 C 67.18 0.13. This pattern of increased Epup at higher temperatur es (not withstanding the slightly lower value at 26 as compared to 24C) is an expected outcome of our model fitting assumption of increased accumulation of energy re serves at higher temperatures, at a rate proportional to the data on commitme nt to pupation (Figure 4-1). After fitting Epup we simulated growth and pupation in 1000 larvae, varying FA,i,t from 1 to70 g / 6h and thereby including a range in which FAi,t fell below the right hand side of Equation 4-2 (i.e. no longer in excess). As with prior simulations in which FA,i,t was equal to Equation 4-2, a constant amount of food was added at the beginning of each 6 h time step, all of which was presumed to be assimilated into bi omass. This assumption presumes that the food available to a larva is equiva lent to the food assimilated ( FA,i,t). Larvae were simulated in isolation in order to avoid making assumptions about resource competition. Stochastic energy storage and individual variation in Eo and WL4 were assumed as previously described. Values of Eo, Wo, and Eth were the same as those assumed in Table 1 and prior simulations. The only source of mortality assumed for each larva was depletion of energy reserves below zero. We varied the timing of ons et of food limitation as follows: (1) immediately upon hatching, (2) after completing 2-days of feeding in excess ( FA,i,t equal to Equation 4-2) (3) early L4 (feeding in excess until median ener gy stores among the 1000 si mulated larvae reached 106

PAGE 107

Eth) and (4) middle L4 (feeding in excess until median energy stores matched the midpoint between Eth and Epup). Figure 4-16 shows the effect of food level on the time required to attain 100% pupation allowing a maximum of 60 days to pupate. Figure 4-17 shows the cumulative pupation rate when the rate of food input is 3g/6hrs, which co rresponded to the minimu m food level in which pupation occurred within 40 days at all six temperatures, when food limitation onset immediately upon hatching. Both diagrams indicate that larvae in this model system are more sensitive to food limitation at 24 and 26 C, with consistently lower de velopment rates in food limited scenarios. While this U-shaped relationship between temperature and development rate is more pronounced in early ontogeny, it is clearly maintained even when food limited conditions initiate when larvae are close to reaching Epup (Figures 4-16D and 4-17D). The interaction between the effects of food and temperature on development rate are evident in the cro ssing of the curves in Figure 4-16. At higher food levels development rate increases monotonically with temperature with curves parallel to the x-axis (F igure 4-16), indicating that larvae reach Epup before Apup, At lower food levels, development rate is roughly similar at 20, 22, 28, and 30 C (Figures 4-16 and 4-17). Within these temperatures larval development rate is highest at 30 C, although the differences are slight an d not consistent when food limitation begins at different developmental stages (Figures 4-16 and 4-17). Thes e data indicate that a reduction in percent energy reserves at 24 and 26 C (Figure 4-15) may reduce the capacity of larvae to mature in food limited habitats. Together with Figures 4-13 and 4-14, these data suggest that if larvae do not energetically compensate for increased food a ssimilation early in development, rising temperatures enhances the effects of food limitation. 107

PAGE 108

4.3 Discussion Although heightened temperatur e increases the development rate of ectotherms, its ultimate effects on the abundance and distribution of species depend on how organisms compensate for increased energetic demands (Laffe rty, 2009). Here we test ed the hypothesis that the observation of reduced size at higher temperat ures in ectotherms, the Temperature Size Rule, may be a consequence of this compensatory re sponse. We modeled ectotherm development as a function of a negative feedback between grow th and energy storage and compared it to laboratory experiments on the mosquito Aedes aegypti, carried out at 2C interval between 2030C. Upon fitting the model to experimental obs ervations of heightened development rate, lower final weight and reduced st arvation resistance of unfed larvae with rising temperature, the model successfully reproduced i ndependent data on starvation resi stance through the course of development at all six temperatures. Sensitivity analyses on the key temperature dependent parameters indicated a simple, novel and potenti ally general, mechanistic explanation for temperature effects on ectotherm growth and deve lopment: that heightened temperature produces a net increase in the rate of accumulation of energy reserves, which reduces the rate of conversion of food into biomass in the final larv al instar. Moreover, ou r findings indicate that the developmental stage at which compensatory energy storage occurs may qualitatively change with small temperature increments. These resu lts suggest a potentia l interaction between temperature and the resources available to mosquito larvae that could cause local habitat features to modify the effects of temperature on the dynamics of mosquito production. Most explanations of the TSR in ectotherms c ite differential impacts of temperature on the rate or allometric scali ng of growth and metabolism (Berrigan and Charnov, 1994; Perrin, 1995; Angilleta et al., 2004). Implic it in these explanations is the assumption that weight and energy stores are interchangeable and directly proportional to one another (Strong and Daborn, 108

PAGE 109

1980; Angilleta and Dunham 2003). Since these expl anations assume that weight determines maturation rate and final size, tradeoffs betw een these two developmental outcomes generating the TSR can only arise through the differential imp acts of temperature on the coefficients or exponents in the growth equation (Bertalannffy, 1960). However, at least in mosquitoes, such a formulation ignores evidence that implic ates energy reserves, and not weight per se in the causal mechanism associated with pupa tion (Gilpin and McClelland, 1979; Chambers and Klowden, 1990; Telang et al., 2007). Here we model independently the dynamics of energy storage and weight gain in developing A. aegypti larvae. Each of these proce sses were linked to each other through the assumption that stori ng energy favors increasing growth rate up to a period of exponential growth in the early fourth instar, after which furt her energy surplus reduces food conversion into biomass (Equation 2). In this m odel mosquitoes that store more energy commit to pupation sooner and additionally reduce food intake in the latter portion of the last larval instar. This mechanism reduces development time in larvae that store more energy, thereby shortening the interval to cessa tion of growth (ICG), the peri od between commitment to pupation and attainment of final weight (Davidowitz and Nijhout, 200 4, Nishiura et al., 2007). Besides the generality of the TSR itself, ther e are a number of pieces of information from empirical studies that su pport the basic premise of our model and the tradeoffs we attribute to rising temperature. First, despite the arbitrar y assumption of 40% initial energy reserves, our results indicate that temperature induced depleti on of energy reserves has little direct impact on observed body size variation, as ev idenced by the low sensitivity of weight to variation in the metabolic coefficient ( cm) and exponent (m ) (Figure 4-9). This is likel y due to the low values of cm relative to weight and is supported by Telang et al. (2007), who found that fourth instar A. aegypti did not have significantly reduced weight after starvation for 36 hours. Secondly, the 109

PAGE 110

sigm oidal weight trajectory gene rated by our use of a Gaussian function to model growth rate with decreasing food assimilation in the ICG, is well supported in laboratory studies of A. aegypti, in addition to those presen ted here (Dye, 1984; Telang et al., 2007). Finally, in A. aegypti and in most ectotherms, growth efficiency tends to increase with temperature (Rashed and Mulla, 1990; Angilleta and Dunham, 2003), coinciding with our ML estimates that show a linear increase in maximum f ood assimilation efficiency ( g ) from 20 to 30 C. Using our model, we demonstrate that increasing g is the key process that permits greater accumulation of energy reserves and thereby faster commitment to pupation at higher temperat ures (Figures 4-1); however, this may generate a deficit in energy reserves as a percent of body weight in the absence of compensatory energy storage in the early instars (Figure 4-12 ). Moreover, the model predicts that in early development rising temperature increases weight gain more than energy reserves, whereas the reverse is true at later de velopmental stages (Figur e 4-14). We suggest that our proposed mechanism of a negative feedback between energy storage and food assimilation in the last larval instar may provide a general explanation for the temperature-size rule in holometabolic ectotherms. While asymptotic size decreases with temp erature, asymptotic energy reserves as a proportion of body weight ( E/W ) is predicted to increase with temperature (Figures 4-14 and 415). Sensitivity analyses in concert with temperat ure-specific parameter estimates indicated that while temperature-induced increas es in growth efficiency ( g ) reduce percent energy reserves (Figure 4-12) by increasing weight more than reserves, temperature effects on energy storage parameters ( cs and s) more than compensate by reducing wei ght and increasing reserves (Figures 4-9 and 4-11). This overcompensation may allow A. aegypti larvae in higher temperatures to overcome the effects increased metabolic coefficient ( cm) on faster depletion of energy reserves 110

PAGE 111

and reduced starvation resistan ce (Figure 4-10). Thus, because food limitation (Arrivillaga and Barrera, 2004; Barrera et al., 2006b) and stochastic food entran ce (Subra and Mouchet, 1984) are common regulators of urban A. aegypti population dynamics (Southwood et al., 1972), it may be that A. aegypti energetically overcompensates in order to reduce vulnerability to food scarcity. However, the high frequency of food limitation in A. aegypti habitats means that heightened energy reserves with temperature ar e unlikely to be observed in field A. aegypti populations. Moreover, increased temperature may increase expend iture of larval reserv es in the pupal and/or adult stages, thereby providing another adaptive bene fit of heightened percent reserves in larvae. We suggest that temperature-i nduced energetic compensation should be investigat ed in adult stages and in other mosquito species, particul arly those that readil y inhabit anthropogenic habitats with large variation in resource availability, such as A. albobictus or Anopheles gambiae spp. Our data indicate that crossing a critical temp erature may generate a qualitative change in the developmental stage in which comp ensatory energy storage occurs in A. aegypti Sensitivity analyses show that while energy storage can occu r through either the coe fficient or the exponent of energy storage, increases in the coefficient cs will favor energy reserves and starvation resistance in both the early and la tter stages of development (Figur es 4-10 to 4-12). In contrast, increased storage exponent s improves energy balance only after 2-days of optimal feeding (Figures 4-10 to 4-12). ML parame ter estimates show an increasing s and relatively constant cs from 20-26 C. In contrast cs shows a sharp linear rise in the 26-30 C range, whereas s drops at 28 and 30 C to the levels of 22 and 20 C (Table 4-2, Figure 4-7). Moreover, the model fit to starvation resistance is highly sensitive to cs at 1-day feeding, with little dependence on g or s (Figure 4-13). In the middle stages of development the impact of cs on the model fit was highly 111

PAGE 112

dependent on the values of g and s (Figure 4-13). Interestingly, at 30 C the dependency of observed starvation resistance of cs on values of g arises in earlier feeding days (Figure 4-13F), suggesting a decreased window for compensato ry energy storage at high temperature. The model and ML estimates suggest that the experimentally observed increase in starvation resistance in the 1-day fed as compared to unfed groups at 28 and 30 C (Figure 4-4) was due to an increase in energy storage at the initial stages of development (Figure 4-13), in response to high growth efficien cy (Figure 4-4). At 24 and 26C, ML estimates (relative to 20 and 22C) suggest that increased metabolism ( cm) and growth efficiency ( g ) without an early increase in energy storage (cs) generated the observed decline in starvation resistance after 1-day feeding in comparison to unfed larvae (Figure 4-4) This interpretation of ML estimates is also consistent with the reduced starvation resist ance in the 2-day fed group at 24 and 26C, in comparison to 22, 28 and 30C (Figure 4-5). Th ese results support the notion of a norm of development (Angilleta et al., 2004), whereby ectotherms modify developmental strategies in response to temperature va riation. In the case of A. aegypti we suggest the potential for a threshold temperature that generates a switch in energy compensation strategies from end of development to the beginning of development; in our laboratory system this occurred between 26 and 28 C. We speculate that such a response coul d involve an epigenet ic trigger in early ontogeny, such that larvae initially raised at 28 C and subsequently transferred to 26 C would exhibit heightened starvation resistance comp ared to larvae raised continuously at 26 C. In summary, the model suggests three different developmenta l patterns in our experimental temperature treatments: 20-22 C: low metabolic demand, low energy storage and increased growth towards the end of development; 24-26 C: increased food assimilation and metabolic demand early in ontogeny with compensatory energy storage late in ontogeny; 28112

PAGE 113

30 C: high m etabolism, food assimilation and energetic storage at the beginning of development. As a result of these patterns, percent energy reserves at the beginni ng and middle stages of development are predicted to be higher at 20, 22, 28 and 30 C as compared to 24 and 26 C (Figure 4-15). We show that in food limited envi ronments this could tran slate into a U-shaped relationship between temperature and development rate. The model predicts a substantially lower capacity to mature in food limited environm ents at mid range temperatures (24-26 C in our experimental system) as compared to the colder and warmer extremes of the thermal optima of A. aegypti (Figures 4-16 and 4-17). Interestingly, while percent energy reserves at 24 and 26 C are predicted to recover after 3 and 4 days of optimal feedi ng (Figure 4-15), when food limitation onsets in beginning or middle of the last instar, it continues to have a larger impact on maturation time at these temperatures as compared to the colder temperatures (Figures 4-16C and D and 417C and D). This is likely due to an increased effect of metabolic parameters on starvation resistance when food is limited (Figure 4-10), si nce larvae must expend en ergy reserves in order to maintain basic functions (Eq. 4-5). Understanding how disease vectors respond to temperature changes is essential for predicting their abund ance and distribution in heterogeneous environments. Our results indicate that increased temperature may cause mosquitoes to assimilate more resources in early ontogeny and less in late ontogeny. Furthermore, we sh ow that amidst increased metabolic and food assimilation rates, temperature-induced changes in the timing of compensatory energy storage can generate an interaction between the eff ects of food availability and temperature on A. aegypti development time (Figures 4-16 and 4-17). This in teraction calls into que stion the fundamental assumption of a monotonic relations hip between the rate of emer gence and warming temperature in the 20-30 C range (Sharpe and DeMichele, 1977; Sc hoolfield et al., 1981; Rueda et al., 1990) 113

PAGE 114

and the temperatu re-driven predictions of A. aegypti production models that are based on this assumption (Focks et al., 1993; Jetten and Fock s, 1997; Kearney et al., 2009; Magori et al., 2009; Williams et al., 2009). Testing our predicted reso urce-temperature interaction in experimental vessels in which water temperature, larval dens ity and food input are si multaneously varied can lead to the development of improved models in order to assess vector populations in heterogeneous environments. The prediction of an inter active effect of temperature and resources on mosquito development rate suggests that temperature-driven predictions of the rate of adult emergence in A. aegypti should be habitat specific, and at larger spatial scales may depend on the container landscape of a particular area. Mo reover, the effects of food lim itation, via larval competition and/or limited opportunity for nutrient entrance, may vary among altitudes, latitudes and seasons. For example, we can speculate that in a regi on where temperature fluctuates between 25 and 28C, seasonal variation in A. aegypti production may be much larg er in a high housing density urban community with intra-domiciliary A. aegypti container habitats exposure to few nutrient sources, as compared to a peri-urban community in which largely outdoor vessels receive increased nutrient input. Given that in many countries mosquitoes readily transmit disease amidst temperatures that vary on the or der of 4-6 degrees in the 20-30 C range, understanding how these results apply to other mo squito species and strains of A. aegypti will be an important step in determining the long term impact of climate ch ange and variation on mosquito borne disease. 114

PAGE 115

Table 4-1. P arameters of model and used to describe experimental data Equation Symbol Description Es timation and key assumptions Wo Initial weight of newly hatched larva Temperature independent, assumed constant among larvae at experimentally measured mean value, 4.1g. 4-2 ,4-3, 4-4, 4-5 WL4 Weight of late L4/early pupae Measured individually in larvae/pupae (n=30) at each temperature; randomly assigned to each simulated larva using a distribution based on measurments. Used for model f itting Eo Initial mean stored energy of newly hatched larva Temperature independent; mean set arbitrarily at 1.5g or 40% Wo; individual variation randomly generated using a distribution derived from variation in starvation re sistance of unfed larvae 4-5 Eth Amount of stored energy that triggers exponential food assimilation (beginning of L4) Temperature independent; se t arbitarily at 15g (10 times Eo). 4-2 Spread of gaussian feeding rate 4-2 Tem perature independent, set at 23g; value roughly consistent with 28 C weight trajectory given Eth assumption (above) over a range of values for g, m and cs 4-2, 4-4, 4-5 cm energy depletion coefficient Chosen to match mean survival of unfed larvae at each temperature, using equation 2b, given the values of Eo and m 4-4 s energy storage exponent Chosen to match asymptotic weight assigned to each larva, given the value of cs, m and g. 4-2, 4-4, 4-5 m energy metabolism exponent ML estimate based on starvation resistance and commitment to pupation at each temperature. 28 C ML estimate also based on weight trajectory from 1-3 days feeding. 4-4 cs energy storage coefficient Same as m 4-2 g Maximum growth efficiency Same as m 4-1, 4-7 T,A Weibull scale parameter for starvation resistance ML fit to experimental data from each temperature and feeding starvation treatment, based on Weibull hazard function 115

PAGE 116

Table 4-1. C ontinued Table 4-2. Joint narrow range ML parameter space for maximum growth efficiency (g), coefficient of energy storage (cs) and exponent of metabolism (m ) Equation Sy mbo l Description Estimation and key assumptions 4-1 kT, A Weibull shape parameter for starvation survival Same as T,A Used in food limitation scenarios Epup Amount of stored energy reserves that triggers pupation Chosen to match energy reserves of minimum percentile that pupated after transferal to distilled water at each temperature for feeding regimes in which at least 70% of starved larvae pupated 4-2, 4-3, 44, 4-5, 4-7 A Feeding age Days of continuous larval feeding upon hatching Used in food limitation scenarios Apup Median developmental time required to pupate Days of feeding after which at least 50% of larvae pupated Temperature (C) Maximum growth efficiency ( g ) Coefficient of energy storage ( cs) Metabolism exponent ( m ) 20 0.2-0.350.1-0.250.3-0.5 22 0.25-0.40.1-0.250.45-0.6 24 0.3-0.5 0.1-0.25 0.5-0.65 26 0.35-0.50.1-0.250.45-0.6 28 0.45-0.60.2-0.350.45-0.6 30 0.5-0.650.25-0.40.45-0.6 Note: Each parameter was varied at 0.05 interv als through the 0-1 range. Ranges represent the two values for each parameter that generated the highest likelihood, with 0.05 added to each end of the range. At 28 C ML fits are based on the summ ation of the likelihood of the observed hazard function for starvation resistance (LLH,T) and weight trajectory ( LLW,28). At all other temperatures, estimates are based soley on LLH,T with no other cons traints outside the fits of s and cm to WL4 and starvation resistance of unfed larvae, respectively. Mean Eo=1.5g, =23g, Eth=15g (Table 4-1). 116

PAGE 117

0 10 20 30 40 50 60 70 80 90 100 345678Feeding days completed upon starvationPercent pupation 20C 22C 24C 26C 28C 30C Figure 4-1. Minimum feeding days required to pupate upon transf er to distilled water across temperatures. Larvae experience starvation mortality in those treatments with less than 100% pupation. Curves at each temperature terminate on the day in which pupae first appear before tran sfer to starving conditions. 117

PAGE 118

0 100 200 300 400 500 600 700 01234Feeding days completed after egg hatchingDry weight (g) Figure 4-2. Weight trajectory at 28 C. Error bars show standard error of dry weight across groups of larvae multiplied by group size (n) in each feeding group as follows: newly hatched (n=10), 1-day fed (n=5) and 2-day fed (n=4) and 3-day fed (n=2); 4-day fed larvae were weighed individually (n=30). 300 350 400 450 500 550 600 650 202224262830Temperature (C)Weight (g) Figure 4-3. Mean WL4 ( 95% CI) among temperatures. Measured after 4, 5, 6, 7 and 8 days feeding at 30, 28, 26, 24, 22, 20C respectively. 118

PAGE 119

0 1 2 3 4 5 6 7 8 9 202224262830 Temperature (C)Survival (days) unfed newly hatched 1-day fed Figure 4-4. Starvation survival in unfed newly larvae and in 1-day fed larvae. Starvation survival as measured by the scale parameter ( ) of the Weibull distribution (%CI) in unfed newly larvae (in black) and in 1-day fed larvae (in grey). 119

PAGE 120

0 10 20 30 40 50 60 123456Days fed before starvationStarvation survival ( ) Figure 4-5. Starvation surv ival, as measured by the Weibull scale parameter ( ) of starved A. aegypti among temperature and feeding treatment s. Color coding is as follows: dark blue: 20C, light blue: 22C, green: 24C, yellow: 26C, orange: 28C, red: 30C. Data from one day fed larvae are repeated from Figure 4-4. In all treatments at least 10 larvae experienced starvation mortality without pupating. Maximum feeding time complying with this criterion was 3 days at 30C, 4 days at 28, 26, and 24C, 5 days at 22C and 6 days at 20C. 120

PAGE 121

0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.140.01 0.07 0.13 0. 1 9 0.25 0.31 0. 3 7 0.43 0.49 0.55 0.61 0. 6 7 0.73 0. 7 9 0.85 0.91 0. 9 7Metabolism exponent(m)Metabolism coefficient (c_m ) 20C 22C 24C 26C 28C 30C Figure 4-6. Joint space of coefficient ( cm) and exponent (m ) of metabolism that reproduces mean starvation resistan ce of unfed larvae from 20 to 30 C. 121

PAGE 122

A 0.34 (0.32,0.35) 0.40 (0.39,0.41) 0.44 (0.43,0.46) 0.59 (0.57,0.61) 0.23 (0.22,0.24) 0.52 (0.50,0.53) 0.31 (0.28,0.33) 0.23 (0.21,0.26) 0.15 (0.14,0.16) 0.15 (0.15,0.16) 0.16 (0.13,0.18) 0.18 (0.18,0.19)0.046 0.041 0.035 0.033 0.069 0.0550.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 202224262830Temperature (C)g, c_s 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07c_m g (95%CI) c_s (95%CI) c_m Figure 4-7. Estimates of temperature dependent parameters based on Experiments A-D: A) Coefficients of growth (maximum growth efficiency, g ), energy storage ( cs) and metabolism (cm). B) Exponents of metabolism (m ) and energy storage (s ). 95% CIs of parameters estimated through ML analysis ( g cs, and m as described in Table 4-2 are based on 100 repetitions of the ML an alysis (see text) for each temperature. Mean Eo = 1.5 g, = 23 g, Eth = 15 g (Table 4-1). The energy storage exponent ( s) is chosen to match the WL4 assigned to each larva; mean and SD among larvae are given in the figure. The co efficient of metabolism ( cm) is fit to mean starvation resistance of unfed larvae given the ML value of m ; it is assigned the same value to each larva. 122

PAGE 123

B 0.52 (0.51,0.52) 0.50 (0.49,0.50) 0.50 (0.50,0.51) 0.58 (0.57,0.59) 0.54 (0.53,0.55) 0.36 (0.34,0.38) 0.800.02 0.810.02 1.010.02 0.960.02 0.810.02 0.730.02 0.0 0.2 0.4 0.6 0.8 1.0 202224262830Temperature (C)(m, s) m (95%CI) s (SD) Figure 4-7. Continued A 0 100 200 300 400 500 600 700 1234Feeding Age (days)Dry mass ( g) Observed (SD) Simulated Figure 4-8. Maxim um likelihood model fits to ob served data. A) Model fi t to observed weight trajectory after 1-3 days feeding at 28 C. Simulated 4-day fed weight is fit to the observed distribution among larvae (see text) Error bars show standard error of dry weight across groups of larvae for 0 (n=10), 1 (n=5) and 2 (n=4) and 3 (n=2) days fed groups. For 4-day fed larvae bars are st andard errors across individuals (n=30). B) Maximum likelihood estimates of mean starvation survival across feeding age at 20, 22, 24, 26, 28 and 30 C. Triangles and circles are observed and simulated means, respectively. Pupation is not accounted for in model fitting; for treatments in which larvae pupated after transfer to starvati on, the corresponding nu mber of simulated larvae with the highest energy reserves af ter feeding are removed prior to calculation of means. Dotted error bars are observed SDs, solid error bars are SDs among 4000 simulated larvae. 123

PAGE 124

B 0 5 10 15 20 25 30 35 201 202 203 204 205 206 221 222 223 224 225 241 242 243 244 261 262 263 264 281 282 283 284 301 302 303Temperature (C) Feeding daysMean starvation resistanc e Simulated (SD) Observed (SD) Figure 4-8. Continued 124

PAGE 125

A B 0 20 40 60 80 100 120 050100150200Percent variation from 28C fitWeight (g) c_m m s c_s g 0 100 200 300 400 500 600 050100150200Percent variation from 28C fitWeight (g) c_m m s c_s g C D 0 100 200 300 400 500 600 700 050100150200Percent variation from 28C fitWeight (g) c_m m s c_s g 0 200 400 600 800 1000 1200 1400 1600 050100150200Percent variation from 28C fitWeight (g) c_m m s c_s g Figure 4-9. Sensitivity of weight after 1 to 4 days feeding to variation in model parameters with respect to 28 C ML fit. A) 1-day fed, B) 2-day fed, C) 3-day fed, D) 4-day fed. Maximum 1500g weight allowed. Shades of red are energy metabolism parameters ( cm and m ), shades of blue ( cs and s) are energy storage parameters and green is maximum growth efficiency ( g ). Each temperature depende ndent parameter is varied individually, maintaining all others at mean value given in Figure 4-7. Eo=1.5g, =23g, Eth=15g (Table 1). No individual variation in Eo, s or stochastic energy storage; one larva simulated. 125

PAGE 126

A B 0 5 10 15 20 25 30 35 40 050100150200Percent variation from 28C fitStarvation survival (days ) c_m m s c_s g 0 5 10 15 20 25 30 35 40 050100150200Percent variation from 28C fitStarvation survival (days ) c_m m s c_s g C D 0 5 10 15 20 25 30 35 40 050100150200Percent variation from 28C fitStarvation survival (days ) c_m m s c_s g 0 5 10 15 20 25 30 35 40 050100150200Percent variation from 28C fitStarvation survival (days ) c_m m s c_s g Figure 4-10. Sensitivity of starvation survival af ter 1 to 4 days feeding to variation in model parameters with respect to 28 C fit. A) 1-day fed, B) 2-day fed, C) 3-day fed, D) 4day fed. Maximum 40 days survival allowe d. Shades of red are energy metabolism parameters ( cm and m ), shades of blue ( cs and s) are energy storage parameters and green is maximum growth efficiency ( g ). All parameter values and simulation conditions are identic al to Figure 4-9. 126

PAGE 127

A B 0 2 4 6 8 10 12 14 16 18 20 050100150200Percent variation from 28C fitEnergy reserves (g) c_m m s c_s g 0 10 20 30 40 50 60 70 80 050100150200Percent variation from 28C fitEnergy reserves (g) c_m m s c_s g C D 0 10 20 30 40 50 60 70 80 90 050100150200Percent variation from 28C fitEnergy reserves (g) c_m m s c_s g 0 10 20 30 40 50 60 70 80 90 100 050100150200Percent variation from 28C fitEnergy reserves (g) c_m m s c_s g Figure 4-11. Sensitivity of energy reserves (E) afte r 1 to 4 days feeding to variation in model parameters with respect to 28 C fit. A) 1-day fed, B) 2-day fed, C) 3-day fed, D) 4day fed. Shades of red are energy metabolism parameters ( cm and m ), shades of blue ( cs and s ) are energy storage parameters and green is maximum growth efficiency ( g ). All parameter values and simulation conditions are identical to Figure 4-9. 127

PAGE 128

A B 0% 20% 40% 60% 80% 100% 050100150200Percent variation from 28C fitPercent ennergy reserves (E/W ) c_m m s c_s g 0% 20% 40% 60% 80% 100% 050100150200Percent variation from 28C fitPercent ennergy reserves (E/W ) c_m m s c_s g C D 0% 20% 40% 60% 80% 100% 050100150200Percent variation from 28C fitPercent ennergy reserves (E/W ) c_m m s c_s g 0% 20% 40% 60% 80% 100%050100150200Percent variation from 28C fitPercent ennergy reserves (E/W ) c_m m s c_s g Figure 4-12.Sensitivity of proportion of energy rese rves (E/W) after 1 to 4 days feeding to variation in model parameters with respect to 28 C fit. A) 1-day fed, B) 2-day fed, C) 3-day fed, D) 4-day fed. Shades of red are energy metabolism parameters ( cm and m ), shades of blue ( cs and s) are energy storage parameters and green is maximum growth efficiency ( g ); All parameter values and simulation conditions are identical to Figure 4-9. 128

PAGE 129

A Figure 4-13. Profiles of simulated versus experiment al survival for variation in growth efficiency ( g ) and energy allocation ( cs), by temperature and feeding age: A) 20 C. B) 22 C. C) 24 C. D) 26 C. E) 28 C. F) 30 C. Each curve represents the joint g and cs space in which the model reproduces observed starvation resistance, with feeding treatments color-coded. + indicates survival is great er than observed for all feeding ages, indicates lower simulated survival at all feeding ages, indicates higher survival at some feeding ages and lower at others, a nd x means that asymptotic weight could not be reproduced if the energy storage exponent s took on a positive value. Numbers adjacent to lines are the average fit valu e of the energy allocation exponent s for a particular parameter combination. For all temperatures Eo=1.5g, =23g, Eth=15g (Table 4-1). Metabolism coefficient ( cm) and exponent (m ) used for 20C: cm=0.062, m =0.30; 22C: cm=0.039, m =0.47; 24C: cm=0.04, m =0.60; 26C: cm=0.049, m =0.46; 28C: cm=0.060, m =0.47; 30C: cm=0.075, m =0.47. 129

PAGE 130

B C Figure 4-13. Continued 130

PAGE 131

D E Figure 4-13. Continued 131

PAGE 132

F Figure 4-13. Continued 132

PAGE 133

A B 0 5 10 15 20 25 30 35202224262830temperature (C)weight (g)0 2 4 6 8 10 12 14energy reserves (g) weight energy reserves 0 25 50 75 100 125 150 175 200 202224262830temperature (C)weight (g)0 10 20 30 40 50energy reserves (g) weight energy reserves C D 0 100 200 300 400 500 202224262830temperature (C)weight (g)0 20 40 60 80energy reserves (g) weight energy reserves 0 100 200 300 400 500 600 700 202224262830temperature (C)weight (g)0 20 40 60 80energy reserves (g) weight energy reserves Figure 4-14.Simulated ML trajectories of mass and energy stores across temperature. A) 1 dayfed larvae. B) 2 day-fed larvae. C) 3 dayfed larvae. D) 4 day-fed larvae. E) 5-day fed larvae. Error bars represent standa rd error in a simulation of 4000 larvae. 133

PAGE 134

E 0 100 200 300 400 500 600 700 800 202224262830temperature (C)weight (g)0 20 40 60 80energy reserves (g) weight energy reserves Figure 4-14. Continued 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 123456789Larval age (days feeding)Energy reserves (E/W)20C 22C 24C 26C 28C 30C Apup Figure 4-15. Sim ulated energy stor es as a fraction of body weight across temperature and feeding age using ML parameter scenarios for each temperature. Error bars represent 95% CIs in a simulation of 4000 larvae. Apup the laboratory observed median days to pupation (Experiment A) is the last bar fo r each temperature as highlighted by the arrows. 134

PAGE 135

A B 0 10 20 30 40 50 60 5101520253035Food availibility (g/6hrs)Time to 100% pupation (days ) 20C 22C 24C 26C 28C 30C 0 10 20 30 40 50 60 5101520253035Food availibility (g/6hrs)Time to 100% pupation (days) 20C 22C 24C 26C 28C 30C C D 0 10 20 30 40 50 60 5101520253035Food availibility (g/6hrs)Time to 100% pupation (days) 20C 22C 24C 26C 28C 30C 0 10 20 30 40 50 60 5101520253035Food availibility (g/6hrs)Time to 100% pupation (days) 20C 22C 24C 26C 28C 30C Figure 4-16. Food limitation and time to attain 100% pupation. In th e first two panels larvae at all temperatures are food lim ited (A) immediately after ha tching or (B) after feeding with excess food for 2 days. In C and D, larv ae feed in excess for different times at each temperature until reaching (C) Eth or (D) the midpoint between Eth and Epup. Xaxis begins at 3g/6h in all panels, th e lowest food level in which pupation is possible within 60 days in all four developmental scenarios. 135

PAGE 136

A B 0 20 40 60 80 100 510152025303540Larval age (days)Cumulative pupation (%) 20C 22C 24C 26C 28C 30C 0 20 40 60 80 100 510152025303540Larval age (days)Cumulative pupation (%) 20C 22C 24C 26C 28C 30C C D 0 20 40 60 80 100 Day5101520253035Larval age (days)Cumulative pupation (%) 20C 22C 24C 26C 28C 30C 0 20 40 60 80 100 Day5101520253035Larval age (days)Cumulative pupation (%) 20C 22C 24C 26C 28C 30C Figure 4-17. Effects of temperature on cu mulative pupation.under food limited conditions (3g/6h) In the first two panels larvae at all temperatures are food limited (A) immediately after hatching or (B) after feed ing with excess food for 2 days. In C and D, larvae feed in excess for different times at each temperature until reaching (C) Eth or (D) the midpoint between Eth and Epup. 136

PAGE 137

CHAPTER 5 CONCLUSI ONS In Chapters 2-4 we explored how temperature variation in a ra nge observed in dengue endemic areas in Colombia interacts with a number of ecological processes that drive the production of domestic A. aegypti populations. The study system consists of socio-biological processes that interact with each other in response to abiotic conditions. Abiotic components include altitude, temperature, stored water, containers and larv al food (ignoring that microorganisms consumed by larvae are living). The two biotic processes are (1) human filling, emptying, usage and maintenance of stored water in domestic vessels and (2) food assimilation and energy storage in A. aegypti larvae that allows growth a nd metamorphosis into pupae. We have seen that the effects human-media ted and resource-mediated processes that determine the size, development rate and pupation success of A. aegypti can be modified across experimental temperature gradients or cities th at vary in temperature. Moreover, we have developed a mechanistic framework th at allows investigation of how A. aegypti responds to changing temperature and nutrien t conditions that is amenable to future scaling-up to a community of vessels. Figure 5-1 shows a schematic of the eco-soc ial system under study. It depicts how the proximal environmental cues experienced by larv ae are ultimately shaped by the interaction between climate and human-mediated habitat dynamics. Habitat resources include the abundance, location, environmental exposure and permanency of water-holding vessels, which, in turn, determine environmental determinants of larval development, such as water temperature, food availability and habitat stability. Moreover, these habita t features are themselves a product of human behavioral responses to variable envi ronments. Key features of the human environment that engender water storage behaviors include the cost of water, the stability of 137

PAGE 138

w ater service, variation in the st ructure of containers and local cultural norms, all of which may interact with individual percepti ons, age, sex, income, etc.. A key feature of this system is that climate variation may have simultaneous direct and indirect impacts on mosquito development either through its direct eff ects on water temperature or the abundance of water-holding vessels, or through its effects on human water stor age behaviors. Below we discuss how A. aegypti larvae process specific cues in the household environment and ultimatel y how the interactions between larval and human ecology determines the development rate and size of emerging A. aegypti We emphasize that human-ecological interactions are key features of the regulation of A. aegypti production among domestic vessels in dengue endemic neighborhoods of Colombia. 5.1 Interactions between the Container Environment and A. aegypti Growth and Development If, for simplicity, we overlook the potential for A. aegypti larvae to modify their environment through movement within containers, larvae are subject to a set of externally determined conditions determined by human activ ity that is ultimately responsible for the existence, location, permanency of all urban vessels. These conditions dictate a maximum time that a habitat will persist for larva to develop, as well as the temperature, nutrient and larval density conditions which determine biologically how much time a developing larva needs to pupate and emerge to adults. Temperature variation is particularly acute among our Colombian field sites, as humanassociated determinants of microclimate, incl uding container material, water volume and sun exposure act over a range of altitudes. By fitti ng a novel developmental model to experimental data, we explored how larvae are likely to inco rporate water temperature into the processes of growth and development. We showed that larv ae exposed to different environmental conditions may exhibit changes in the initial growth rate, fi nal growth rate and/or the duration of growth. 138

PAGE 139

W hile a positive linear impact of increasing temperature on a developmental parameter is likely due to faster biochemical reactions, tradeoffs between development rate, size, fecundity and/or survival may cause mosquitoes to respond by modifying a facet of development that does not directly act through the enzyme-kinetics mech anism (Angilleta and Dunham, 2004; Kozlowski et al, 2004). Our results suggest that while heightened metabolism and food assimilation rate are inevitable at higher temperatur es (within the 20-30C optima), mosquitoes can compensate for negative impacts by modifying the duration and timing of growth, through changes in how they distribute acquired resources. In particular, by supposing that increased energy stores cause larvae to reduce food assimilation in the latter stages of development, we show that mosquitoes may reduce their final feeding rate; moreover, this reduction can generate the negative association between temperature and adult size commonly observed in mosquitoes and most ectotherm species. Throughout the preceding chapters we emphasi ze that large variati on in the resources available to A. aegypti across domestic vessels in a part icular human community is well documented. We saw that the tradeoffs in te mperature-induced increases in growth and metabolic rate may be more acute when resources are absent, as temperature increases maintenance costs of larvae; this compels A. aegypti larvae to increase the accumulation of energy reserves in order to ensure pupation. We found that if increased energy allocation occurs at the beginning of development it may signi ficantly reduce the nega tive effects of food limitation on development rate and pupation suc cess. However, if temperature-associated increases in energy storage occur at the end of development, they may result in a reduction in development rate in situations of food limitation. Indeed, both model predictions based on ML fits to experimental data on starvation resistance, and simultaneous manipulation of food and 139

PAGE 140

tem perature conditions in developi ng larvae, indicated that the developmental response of larvae to variation in resource conditions may be modi fied by temperature. In particular at 28C increased larval resources may favor cell si ze more than cell number, whereas at 22C heightened resources produce a larger increase in cell number. Assuming that the resources ava ilable to larvae in water storage vessels are similar across the Colombian landscape, a 2C seasonal or ENSO-associated temperature shift may have starkly different effects on the A. aegypti production among cities with different altitudes in the 0-1500 m range. Moreover, given evidence that spatia l variation in vector production is related to resource availability (Barrera et al, 2006), temperature cha nges may also have differential impacts on the spatial distribution of emerging mosqu itoes across altitudes. This is due to the fact that the characteristic aggregation of A. aegypti pupae in a few vessels in a community (Getis et al, 2003) may be linked to resource limitation in a majority of vesse ls (Barrera et al, 2006). Thus, if a 2C temperature shift were to alleviate reso urce limitation at certain mean temperatures (and alitutdes) but not others, there would be a differential impact on the spatial dist ribution of pupae and potentially on human exposure to dengue vectors. A third aspect of A. aegypti s response to container conditions is the interac tion of larvae with each other. Indeed, in order to scale up our model to predict A. aegypti production dynamics in a community of vessels, it will be first necessa ry to model how conspecific larvae compete for resources. Notwithstanding the potential for variation in search efficiency among larvae, our model predicts that larvae will exert the largest competitive pressure, i.e. consume the most food, in the early and middle portions of the fourth in star when larvae possess only 50 to 75% of their final weight under optimal food cond itions. Food assimilation rate decr eases in the latter half of the fourth instar with increasing energy reserv es. By contrast, other models of competition 140

PAGE 141

(Gilpin and McClelland, 1979; Focks et al., 1993 ) assume that competition is temperature independent and that larvae monotonically incr ease food consumption with weight. Once food searching is incorporated into the model, our fram ework is likely to generate novel insights into the dynamics of intra-specific resource competition in A. aegypti. 5.2 Human Ecological Interactions and the Dynamics of A. aegypti Production As container water is purposefully stored for potential human use, larvae that inhabit domestic vessels are subject to a constant risk of being wash ed away or desiccated through emptying events. Simultaneously, however, human use and replenishment of water allows eggs to be deposited and exposed to a regular hatching stimulus thr oughout the height of the vessel. This contrasts from treehole mosquito habitats or large abandoned vessels that depend only on rainfall for variation in water level; these may be susceptible to seasonal or climate change induced droughts in which large egg banks may accu mulate at certain levels with little egg hatching. By contrast, in vessels in which water is regularly ex tracted and replenished due to usage for common household activities such as clean ing and cloth washing, larvae are likely to hatch more frequently in smaller and irregular cohorts; this could potentially alleviate resource competition and increase the likelihood of contin uous presence of a larva. Accordingly, in residential areas with high levels of water storag e and usage of stored water, we may expect an increased abundance and frequency of larval in festation. For example in our study areas in Bucaramanga and Armenia where greater than 90 % of households store water, we readily find that 20 to 30% of premis es are infested with A. aegypti larvae at a given time. In Chapter 2 we saw that the key process that mediates the impact of water use on A. aegypti production is water emptying. Emptying flus hes both larvae and nutrients in containers; because larva hatch faster than nutrients accumu late with water replenishment, emptying may simultaneously increase both larval mortality and density-dependent resource competition. Thus 141

PAGE 142

frequently emptied vessels are m ore likely to de monstrate an interaction between (1) the impacts of temperature-food interaction on development ra te discussed above, and (2) the upper limit to A. aegypti development time set by average emptying frequency. This supposition ignores the non-zero probability that a larva may survive an emptying event. Nonetheless, we found that on average, more frequent emptyi ng was required to reduce pupal pr oduction in the warmer cities. Moreover water storage vessels that expe rienced emptying in Bucaramanga (24-25 C) had a mean production rate approximately 50% of those in Armenia (21-22 C) and Barranquilla (2730 C), consistent with model predictions for f ood limited vessels. Moreover, our preliminary modeling studies (not presented here), paramete rized using the associat ions between emptying frequency and pupal production rate described Chapter 2, suggest that stochastic water emptying that is independent among houses can expl ain the characteris tic aggregation in A. aegypti production. We found that human water st orage behavior is largely de termined through individual interactions with the physical and sociological environment. Fo r example residents who store water not for regular usage but rath er in case of interruptions in piped water supply are less likely to empty vessels. By contrast, emptying is more frequent in containers w hose water is regularly extracted as residents are more likely to empty when the water level reaches a minimum. All of our study neighborhoods have a stable water supply with occasional interrup tions that are seldom prolonged for more than a few hours. However, we observed a universally lower frequency of emptying and higher A. aegypti production rate in the dry seas on, when interruptions in piped service are presumably or perceived to be mo re frequent. Thus, human adaptation to climate variation, rather than the dire ct impact of climate itself, ma y drive the association between climate and vector production (B eebe et al., 2009). Elderly people tended to perceive a greater 142

PAGE 143

risk of going without water and we re more likely to store water jus t in case. This is likely an outcom e of customs developed over many years or multiple generations of living without a continuous water supply that are not easily modifi ed when efficient water delivery systems are built. Interestingly, climate change, which is predicted to reduce rainfall and increase water insecurity in all three study cities, may proceed at a similar time-scale as cultural adaptation. This would suggest that the customs of the elderly may once again take hold in younger generations. Not only cultural practices, but the structur e of the vessel itself may influence the motivations and patterns of water storage be havior. In houses with permanent washbasins, residents dont have to actively de cide to store water and obtain a vessel, but rather, they use the vessel that came built-in as part of the house. Culturally, in Colombia hand-washing of undergarments is commonplace even in household with washing machines. In order to save water and for convenience residents almost universally opt to use water stored in the attached washbasin to scrub clothes instead of taking tap water. Thus, the motivation for water storage is primarily for the convenience of using stored water to wash clothes and secondarily for other reasons, such as interruptions in water service. In areas without built in wa sh-basins, by contrast, households must purposefully obtai n a vessel and thus water storag e is less frequent; moreover stored water often goes without usage. Unli dded, unused water ofte n contributes to the accumulation of A. aegypti egg banks with less frequent hatching stimuli. In the face of increasing drought conditions, uns table water supply and decreas ed incentive to reduce water storage throughout Colombia, our da ta indicate that lid placement can be effective method of reducing A. aegypti production. We consistently observe d less debris, although not necessarily lower larval infestation rates in lidded vessels, suggesting that the major action of lids is to 143

PAGE 144

reduce the input of larval nutrients. However th e structure of per manent washbasin makes lid placement on these containers extremely cumbersome and uncommon. In short, sustainable management of A. aegypti production in changing urban environments requires understa nding variation in motivati on for household water storage behavior. Achieving such knowledge will require th e use of social science methods capable of addressing the complex interactions between in dividual perceptions a nd the socio-economic, cultural and physical features of the human envi ronment. In the preceding work we have moved forward towards developing a framework to study the joint interaction of the three separate ecological relationshi ps: human behaviorA. aegypti production, climateA. aegypti production and climate-human behavior. Understanding how this joint interaction plays out in human communities is the key to being able to apply our findings to predict the dynamics of A. aegypti production in changing environments. 144

PAGE 145

Human water storage practices and behavior towards domestic vessels Aedes aegypti growth and development Container and water dynamics Stability of water service, costs of water, cultural norms, container structure Resources, larval density, water temperature, habitat stability Proximal environmental cues that drive behavioral and physiological responses of humans and mosquito larvae Climat e Figure 5-1. Schem atic of eco-s ocial system of domestic Aedes aegypti (L.) production. 145

PAGE 146

LI ST OF REFERENCES Alto BW, Reiskind MH, Lounibos LP (2008) Size al ters susceptibility of vectors to dengue virus infection and dissemination. American Journal of Tropi cal Medicine and Hygiene 79:688-695 Angille tta MJ, Dunham AE (2003) The temperature-si ze rule in ectotherms: Simple evolutionary explanations may not be general. American Naturalist 162:332 Angilletta MJ, Steury TD, Sears MW (2004) Temperature, growth rate and body size in ectotherms: Fitting pieces of a life history puzzle. Integrative and Comparative Biology 44:498 509 Arendt J (2007) Ecological co rrelates of body size in relation to cell size and cell num ber: Patterns in flies, fish, fruits and foliage. Biological Reviews of the Cambridge Philosophical Society 82:241-256 Arrivillaga J, Barrera R (2004) Food as a limiting factor for Aedes aegypti in water-storage containers. Journal of Vector Ecology 29:11-20 Azevedo RB, French V, Partridge L (2002) Te mperature modulates epidermal cell size in Drosophila melanogaster Journal of Insect Physiology 48:231-237 Atkinson D, Sibly RM (1997) Why are or ganism s usually bigger in colder environments? Making sense of a life history puzzle. Trends in Ecology and Evolution 12: 235-239 Barrera R, Am ador M, Clark GG (2006a) Use of the pupal survey technique for measuring Aedes aegypti ( Diptera: Culicidae) productivity in Puerto Rico. American Journal of Tropical Medicine and Hygiene 74:290-302 Barrera R, Amador M, Clark GG (2006b) Ecological fact ors influencing Aedes aegypti ( Diptera: Culicidae ) productivity in artifi cial containers in Salinas, Puerto Rico. Journal of Medical Entomology 43:484-492 Barrera R, Avila J, Gonzlez-Tllez S (1993) Un reliable supply of potab le water and elevated Aedes aegypti larval indices: A causal relationship? Journal of the American Mosquito Control Association 9:189-195 Bedhomme S, Agnew P, Sidobre C, Michalakis Y (2003) Sexspecific reaction norms to intraspecific larval competition in the mosquito Aedes aegypti Journal of Evolu tionary Biology 16: 721-730 Beebe NW Cooper RD, Mottram P, Sweeney AW (2009) Austra lia's Dengue risk driven by human adaptation to climate change. PLoS Neglected Tropical Diseases 3:e429 Berrigan D Charnov EL (1994) Reaction norms for age and size at maturity in response to temperature: A puzzle for life historians. Oikos 70:474 146

PAGE 147

Bertalanffy L (1960) Principles and theory of growth. In: Fundam ental Aspects of Normal and Malignant Growth, Nowinski WW (edito r), Amsterdam: Elsevier, pp 137-259 Bisset JA, Marquetti MC, Suarez S, Rodriguez MM, Padmanabha H (2006) Application of the pupal/demographic-survey methodology in an area of Havana, Cuba, with low densities of Aedes aegypti (L.). Annals of Tropical Medicine and Parasitology 100: S45-S51 Bolker B (2008) Ecological Mode ls and Data in R. Princeton : Princeton University Press Bokma F (2004) Evidence against universal metabolic theory. Funtional Ecology 18:184-187 Burkot TR, Handzel T, Schmaedick MA, Tufa J, Roberts JM, Graves PM (2007) Productivity of natural and artificial containers for Aedes polynesiensis and Aedes aegypti in four American Samoan villages. Medical and Veterinary Entomology 21:22-29 Bochdanovits Z, De Jong G (2003) Temperature dependent larval resource allocation shaping adult body size in Drosophila melanogaster. Journal of Evolutionary Biology 16:1159-1167 Caprara A, De Oliveira-Lima JW, Correia-Pequeno A, Gondim P, Sommerfeld J (2009) Irregular water supply, household usage and dengue: A bios ocial study in the Brazilian Northeast. Cadernos da Sade Pblica 25:S125-S136 Chambers GM, Klowden MJ (1990) Correlation of nutritiona l reserves with a critical weight for pupation in larval Aedes aegypti mosquitoes. Journal of the American Mosquito Control Association 6 : 394-399 Chan AS, Sherm an C, Lozano RC, Fernndez EA Winch PJ, Leontsini E (1998) Development of an indicator to evaluate the impact, on a community-based Aedes aegypti control intervention, of improved cleaning of water-sto rage containers by householders Annals of Tropical Medicine and Parasitology 92:317-329 Christopher RS (1960) Aedes aegypti (L.) the Yellow Fever Mosquito, Its Life History Bionom ics and Structure, London: Cambridge University Press Davidowitz G, Nijhout HF (2004) Physiological basis of reaction norms: the interaction among growth rate, the duration of growth and body size. Integrative and Com parative Biology 44:443 De Moed GH, De Jong G, Scha rloo W (1997) Environmental e ffects on body size variation in Drosophila melanogaster and its cellular basis. Genetics Research 70:35-43 Dye C (1982) Intraspe cific competition amongst larval Aedes aegypti : Food exploitation or chemical interference. Ecological Entomology 7:39-46 147

PAGE 148

Dye C (1984) Com petition amongst larval Aedes aegypti : The role of interference. Ecological Entomology 9:355-357 Elder JP, Apodaca X, Parra Medina DM, Zuiga-d e-Nuncio ML (1998) Strategies for Health Education: Theoretical Models. In: Handbook of Immigrant Health, Loue S. (editor), New York, NY: Plenum Publishing Corporation, pp 567-585 Elder J, Lloyd LS (2007) Achiev ing Behavior Change for Dengue Control: Methods, Scaling-up, and Sustainability: A Working Paper. World H ealth Organization on behalf of the Special Programme for Research and Trai ning in Tropical Diseases, Scie ntific Working Group Report on Dengue, 1-5 October 2006, Geneva, Swit zerland. Paper No. 7.3. Available: http://apps.who.int/tdr/publications/td r-res earch-publications/swg-reportdengue/pdf/swg_dengue_2.pdf [accessed Jun 10, 2008] Farrar J, Focks D, Gubler D, Barrera R, Gu zman MG, Simmons C, et al. (2007) Editorial: Towards a global dengue research agenda. Tropical Medicine and International Health 12: 695699 Favier C, Degallier N, Vilarinhos PdeT, de Carvalho Mdo S, Yoshizawa MA, Knox MB (2006) Effects of climate and different management strategies on Aedes aegypti breeding sites: A longitudinal survey in Braslia (DF, Brazil). Tropical Medicine and International Health 11: 1104-1118 Focks DA, Chadee DD (1997) Pupal survey : An epidem iologically significant surveillance method for Aedes aegypti : An example using data from Trinidad. American Journal of Tropical Medicine and Hygiene 56:159-167 Focks DA, Haile DG, Daniels E, Mount GA (1993) Dynamic life table model for Aedes aegypti ( Diptera, Culcidae ) Analysis of the literat ure and model development. Journal of Medical Entomology 30:1003-1017 Focks DA, Daniels E, Haile DG, Keesling JE (1995) A simulation model of the epidemiology of urban dengue fever: literature an alysis, model development, preliminary validation, and samples of simulation results. American Journal of Tropic al Medicine and Hygiene 53:489-506 French V, Feast M, Partridge L ( 1998) Body size and cell size in Drosophila : the dev elopmental response to temperature. Journal of Insect Physiology 44:1081-1089 Getis A, Morrison AC, Gray K, Scott TW (2003) Characteristics of the spatial pattern of the dengue vector, Aedes aegypti in Iquitos, Peru. American Journal of Tr opical Medicine and Hygiene 69:494-505 Gilpin ME, McClelland GAH (1979) System analysis of the yellow fever mosquito Aedes aegypti Fortschritte fr Zoologie 25:355-388 148

PAGE 149

Golafshani N (2003) Understa nd ing reliability and validity in qualitative research. The Qualitative Report 8:597-607 Gubler DJ (2002) Epidemic dengue/dengue hemorrh agic fever as a public health, social and economic problem in the 21st century. Trends in Microbiology 10:100-103 Hammond SN, Gordon AL, Lugo Edel C, Moreno G, Kuan GM, Lpez MM, et al. (2007) Characterization of Aedes aegypti ( Diptera : Culcidae) production sites in urban Nicaragua. Journal of Medical Entomology 44:851-860 Hobbs NT, Hilborn R (2006) Altern atives to statistical hypothesis testing in ecology: A guide to self teaching. Ecological Applications 16:5 Institute of Environmental Studies of Colombia (2009). Integrated National Adaptation Pilot. Institute of environmental Studi es of Colombia. Available: http://institucional.ideam.gov.co /jsp/loader.jsf ?lServicio=Publicaciones&lTipo=publicaciones&l Funcion=loadContenidoPublicacion&id=1011 [accessed Feb 12, 2008] Irlich UM, Terblanche JS, Blackburn TM, Chown SL (2009) Insect ratetemperature relationships: Environmenta l variation and the metabolic theory of ecology. American Naturalist 174:819-835 Jetten TH, Focks DA (1997) Potential changes in the distribution of dengue transmission under climate warming. American Journal of Tropica l Medicine and Hygiene 57:285-297 Jirakanjanakit N, Leemingsawat S, Thongrungkiat S, Apiwathnasorn C, Singhaniyom S, Bellec C, Dujardin JP (2007) Influence of larval dens ity or food variation on th e geometry of the wing of Aedes ( Stegomyia) aegypti Tropical Medicine and International Health 12:1354-1360 Karl I, Fischer K (2008) Why get big in the cold ? Towards a solution to a life-history puzzle. Oecologia 155:215-225 Kasprzyk D (2005) Measurem ent error in househ old surveys: Sources and measurement. mathematical policy research. In: Household Sample Surveys in Developing and Transition Countries, Department of Economic and Social Affairs (editors), Washington, DC: United Nations Publications, pp 171-198 Kearney M Porter WP, Williams C, R itchie S, H offmann AA (2009) Integrating biophysical models and evolutionary theory to predict climatic impacts on species' ranges: the dengue mosquito Aedes aegypti in Australia Functional Ecology 23 : 528-538 Klowden MJ, Blackm er JL, Chambers GM (1988) Effects of larval nutrition on the host-seeking behavior of adult Aedes aegypti mosquitoes. Journal of the American Mosquito Control Association 4:73-75 149

PAGE 150

Koenraadt C J, Aldstadt J, Kijchalao U, Sithip rasasna R, Getis A, Jones JW, Scott TW (2008) Spatial and temporal patterns in pupal and adult production of the dengue vector Aedes aegypti in Kamphaeng Phet, Thailand. American Journal of Tropica l Medicine and Hygiene 79:230-238 Koz owski J, Czarno eski M, Danko M (2004) Can optimal resource allocation models explain Why ectotherms grow larger in cold? Integrative and Com par ative Biology 44:480 Lafferty KD (2009) the ecology of clim ate change and infectious diseases Ecology 90:888-900 Legros M, L loyd AL, Huang Y, G ould F (2009) Density-dependent intraspecific competition in the larval stage of Aedes aegypti ( Diptera : Culicidae ): Revisiting the current paradigm. Journal of Med ical Entomology 46:409-419 Liu J, Dietz T, Carpenter SR, Laberti M, Folk e C, Moran M, et al. (2007) C omplexity of Coupled Human and Natural Systems. Science 317:1513-1516 Maciel-De-Freitas R, Codeo CT, Loureno-De-O liveira R (2007) Body size-associated survival and dispersal rates of Aedes aegypti in Rio de Janeiro. Medical and Veterinary Entomology 21:284-292 Magori K, L egros M, Puente ME, Focks DA, Scott TW, Lloyd AL, Gould F ( 2009) SkeeterBuster: A stochastic, spatially explicit m odeling tool for studying Aedes aegypti population replacement and population suppression strategies. PLoS Neglected Tropical Diseases 3:e508 Morrison AC, Gray K, Getis A, Astete H, Sihui ncha M, Focks D, et al. (2004) Tem poral and geographic patterns of Aedes aegypti ( Diptera : Culicidae ) production in Iquitos, Peru Journal of Medical Entomology 41:1123-1142 Naksathit AT, Scott TW (1998) Effect of fe m ale size on fecundity and survivorship of Aedes aegypti fed only human blood versus human blood plus sugar. Journal of the American Mosquito Control Association 14:148-152 Nasci RS (1991) Influence of larval and adult nutrition on biting persistence in Aedes aegypti ( Di ptera : Culicidae ). Journal of Medical Entomology 28:522-526 Nation JL (2008) Insect Physio logy and Biochem istry, Boca Rat on, FL: Taylor and Francis Group CRC Press National Institute of Health of Colombia (2008) Informe Quincenal Epidemiologico Nacional: National Epidemiological Surveillance System (SIVIGILA). National Institute of Health of Colombia. Available: http://www.ins.gov.co/?idcategoria=1680&download=Y [accessed Dic 12, 2008] Nayar JK, Van Handel E (1971) The fuel for sust ained flight. Journal of Insect Physiology 17:471-481 150

PAGE 151

Nishiura JT (2002) Coordinated morphological changes in m idgut, imaginal discs, and respiratory trumpets during metamorphosis of Aedes aegypti ( Diptera : Culicidae) Annals of the Entom ological Society of America. 95 : 498-504 Nishiura JT, Burgos C, Aya S, Goryacheva Y, Lo W (2007) Modulation of larval nutrition affects midgut neutral lipid stor age and temporal pattern of tran scription factor expression during mosquito metamorphosis. Journal of Insect Physiology 53: 47-58 Padm anabha H, Soto E, Mosquera M, Lord CC, Lounibos LP (2010) Ecological links between water storage behaviors and Aedes aegypti production: implications fo r dengue vector control in variable climates. Ecohealth DOI: 10.1007/s10393-010-0301-6 Pan American Health Organization ( 2008) Dengue Surveillance. Available: http://www.paho.org/english/ad/dpc/cd/dengue.htm [accessed Dec 20, 2009] Perrin, N (1995) About Berrigan and Charnovs life history puzzle. Oikos 73:137 Phuanukoonnon S, Mueller I, Bryan JH (2005) Effectiveness of dengue control practices in household water containers in Northeast Thailand. Tropical Medicine and International Health 10:755-763 Ponlawat A, Harrington LC (2007) Age and body size influence male sperm capacity of the dengue vector Aedes aegypti ( Diptera : Culicidae ). Journal of Medical Entomology 44:422-426 Rashed SS, Mulla MS (1989) Factors influencing i ngestion of particulate m aterials by mosquito larvae ( Diptera : Culicidae ). Journal of Medical Entomology 26:210-216 Reiskind MH, Lounibos LP (2009) Effects of intras pecific larval competit ion on adult longevity in the mosquitoes Aedes aegypti and Aedes albopictus Medical and Veterinary Entomology 23:62-68 Reuben R, Das PK, Sa muel D, Brooks GD ( 1978) Estimation of daily emergence of Aedes aegypti ( Diptera : Cilicidae ) in Sonepat, India. Journal of Medical Entomology 16:705-714 Romero-Vivas CM, Arango-Padilla P, Falcona r AK (2006) Pupal-produ ctivity surveys to identify the key container habitats of Aedes aegypti (L.) in Barranquilla, th e principal seaport of Colombia. Annals of Tropical Medicine and Parasitology 100:S87-S95 Rueda LM, Patel KJ, Axtell RC, Stinner RE (1990) Temperature-dependent development and survival rates of Culex quinquefasciatus and Aedes aegypti (Diptera: Culicidae). Journal of Medical Entomology 27:892-898 Schoolfield RM, Sharp PJH, Manguson CE ( 1981) Nonlinear regression of biological tem perature dependent rate models ba sed on absolute reaction rate theory. Journal of Theoretical Biology 77:608-620 151

PAGE 152

Scrim shaw SCM (1990) Combining quantitative and qualitative me thods in the study of intrahousehold resource allocation. In : Intra-household Resource A llocation: Food and Nutrition Bulletin, Supp. 15, Rogers BL, Schlossman NP (editors), Tokyo: United Nations University Press, pp 186-198 Sharpe PJ, DeMichele DW (1977) Reaction kinetics of poikilotherm development. Journal of Theoretical Biology 64: 649-670 Southwood TR, Murdie G, Yasuno M, Tonn RH, R eader PM (1972) Studies on the life budget of Aedes aegypti in W at Samphaya, Bangkok, Thailand. Bulletin of the Worl d Health Organization 46:211-226 Steinwascher K ( 1982) Relationship between pupal mass and adult survivorship and fecundity for Aedes aegypti Environmental Entomology 11: 150 153 Strickm an D, Kittayapong P (2003) Dengue and its vectors in Thailand: Calculated transmission risk from total pupal counts of Aedes aegypti and association of winglength measurements with aspects of the larval habitat. American Journal of Tropical Medicine and Hygiene 68:209-217 Strong KW, Daborn GR (1980) The influence of temperature on energy budget variables, body size, and seasonal occurrence of the isopod Idotea baltica (Pallas). Canadian Journal of Zoology 58:1992-1996 Subra R (1983) The regulation of preimaginal populations of Aedes aegypti L. ( Diptera : Culicidae ) on the Kenya coast. I. Preimaginal population dynamics and the role of human behaviour. Annals of Tropical Medicine and Parasitology 77:195-201 Subra R, Mouchet J (1984) The regula tion of preimaginal populations of Aedes aegypti (L.) ( Diptera : Culicidae ) on the Kenya coast. II. Food as a main regulatory factor. Annals of Tropical Medicine and Parasitology 78:63-70 Telang A, Frame L, Brown MR (2007) Larval f eeding duration affects ec dysteroid levels and nutritional reserves regulating pupal comm itment in the yellow fever mosquito Aedes aegypti ( Diptera : Culicidae ). Journal of Experimental Biology 210: 854-864 Timmer mann SE, Briegel H (1999) La rval growth and biosynthesis of reserves in mosquitoes. Journal of Insect Physiology 45:461-470 Trotter RT, Needle RH, Goosby E, Bates C, Sing er MA (2001) Methodological model for rapid assessment, response and evaluation: The RARE program in public health. Field Methods 13:137; DOI: 10.1177/1525822X0101300202 [Online May 2001] Trpis M, Hausermann W ( 1986) Dispersal and other population parameters of Aedes aegypti in an African village and their pos sible significance in epidemiol ogy of vector-borne diseases. American Journal of Tropic al Medicine and Hygiene 35:1263-1279 152

PAGE 153

Tun-Lin W Burkot TR, Kay BH (2000) Effects of temperature and larval diet on development rates and survival of the dengue vector Aedes aegypti in north Queensland, Australia. Medical and Veterinary Entomology 14:31-37 VanderHave TM, DeJong G (1996) Adult size in ect otherm s: Temperature effects on growth and differentiation. Journal of Theoretical Biology 183: 329-340 Vu SN, Nguyen TY, Tran VP, Truong UN, Le QM, Le VL, et al. (2005) Elim ination of dengue by community programs using Mesocyclops ( Copepoda) against Aedes aegypti in central Vietnam. American Journal of Tropic al Medicine and Hygiene 72:67-73 Westbrook CJ, Reiskind MH, Pesko K, Greene K, Lounibos LP (2009) Larval environmental temperature and the susceptibility of Aedes albopictus to chikungunya virus. Vector-Borne and Zoonotic Diseases (In press). Wilcox BA, Colwell RR (2005) Emerging and reemer ging infectious diseases biocomplexity as an interdisciplinary paradigm. EcoHealth 2:244-257 Williams CR, Johnson PH, Long SA, Rapley LP Ritchie SA (2008) Rapid estimation of Aedes aegypti population size using simulati on modeling, with a novel approach to calibration and field validation. Journal of Medical Entomology 45:1173-1179 Zhou G, Pennington JE, Wells MA ( 2004) Utilization of pre-existing energy stores of female Aedes aegypti m osquitoes during the first gonotrophic cy cle. Insect Biochemistry and Molecular Biology 34:919-925 153

PAGE 154

154 BIOGRAPHICAL SKETCH Harish Padmanabha leads a World Bank funde d project on climate change and dengue ecology in Colombia, based in the Instituto Naci onal de Salud of the Colombian Ministry of Health. For 8 years he has worked in the ecol ogy and control of emerging diseases such as dengue, hantavirus, leptospira and spotted feve r rickettsia in Panama Cuba and Colombia, through appointments in institutions such as the Panamerican Health Organization (PAHO), National Institute of Health of Colombia, local health departments in Colombia, the Gorgas Institute and the Pedro Kouri Tropical Medicine Institute. He has an MSc in Population Sciences and International Health from Harvard University He lives with his wife and two children in Barranquilla, Colombia.