Trade-offs and heritability of a key sex pheromone component in Heliothis subflexa moths.

Trade-offs and heritability of a key sex pheromone component in

Heliothis subflexa moths.

Daily Supervisor: Elise Fruitet

Assessor: Dr. Emily Burdfield-Steel Examinator: Prof. dr. Astrid T. Groot Bachelor student: Max Boot

Student ID: 11779233 University of Amsterdam

Abstract

Life history traits are vital for the survival, growth and reproduction of organisms, the resources needed for this can only be allocated once resulting in trade-offs. Heliothis subflexa moths are known for producing multicomponent sex pheromones, the acetates in this sex pheromone have a dual function, which act as both an attractant for conspecific males and as a repellent for allospecific males. Both geographic and temporal variation have been found in the sex pheromone blend of female H. subflexa. Artificial selection for over 9 generations has resulted in two acetate selection lines, for high and low acetate production. This research determined the heritability of the total amount of produced acetates in both selection lines and looked at potential trade-offs between acetate production and two life history traits: developmental time and adult’s survival rate. The heritability of the total amount of produced acetates was almost identical for both selection lines, at around 40%. This means that 40% of the phenotypic variation is caused by the genetic variation in both lines. Overall the high line showed a decline in the developmental time over the course of four generations, suggesting that no trade-off is present between acetate production and developmental time. The high selection line seemed to have the highest survival rate on both treatments which did not have a food limitation. This suggests that there could be a potential trade-off present between acetate production and adult survival rate when food is limited and that the trade-off is absent or hidden when food is unlimited.

Keywords

Introduction

Life history traits, theories and strategies

All living organisms have one thing in common, they need to grow and survive in order to reproduce. These traits combined are referred to as life history traits (Stephen C. Stearns, 1992) which influence the fitness of the organism (Stephen C. Stearns, 1976). Multiple life history traits are clustered together in order to maximize reproduction through optimal allocation of limited resources (Olderbak et al., 2014). Adaptation of life history strategies to reoccurring environmental conditions over evolutionary time is induced by natural and sexual selection, while developmental experiences adapt life history strategies to environmental conditions encountered during the organisms life (Ellis et al., 2009). It has been found that variability for these life history strategies exists between and within species (Ellis et al., 2009; Promislow & Harvey, 1990) all across the animal kingdoms (Harvey & Clutton‐Brock, 1985; Jervis et al., 2001; Mesquita et al., 2016; Molleman et al., 2011; Xu et al., 2020) and even in plants (Adler et al., 2014; Ellner, 1987; Salguero-Gómez et al., 2016).

Trade-offs in sexual traits

Trade-offs arise when a decision either consciously or subconsciously needs to be made by an individual. This decision can cause problems, mainly in the allocation of resources for specific processes such as survival or reproduction (S. C. Stearns, 1989). Sexual traits are used to attract mates by both males and females in order to increase reproductive success, however in most cases a fitness cost is associated with these traits (Fitzpatrick et al., 1995; Peters et al., 2004; Sheldon & Verhulst, 1996; Zahavi, 1975). Preferences for the sexual traits of potential mates are heritable and have evolved over time (Bakker & Pomiankowski, 1995; Iwasa & Pomiankowski, 1994). The main driver of the evolution of these preferences is sexual selection (Jones & Ratterman, 2009; Smith, 1991). While natural selection focuses on the adaptation of organisms to their environment hence increasing their survival as first described by Charles Darwin. Sexual selection on the other hand focuses on improving the reproductive success of an individual by so called non-survival adaptations. Thus it can promote traits which may seem useless from a survival point of view but are in fact useful for the attraction of mates and competing with members of the same sex.

Chemical signals

The physical appearances of animals are not the only thing under the influence of sexual selection: it can also affect olfactory signals. These signals are found in multiple species (Ache & Young, 2005) and especially in insects (Steiger & Stökl, 2014; Tegoni et al., 2004; Vereecken et al., 2007). Insects use olfactory signals, in the form of pheromones which can cause physiological or behaviour changes and act as attractants or repellents for other individuals (Cardé & Baker, 1984; Tegoni et al., 2004). Pheromones can travel over long distances and play an important role in mate choice (Harari & Steinitz, 2013; Hildebrand, 1995; Johansson & Jones, 2007). These species-specific functions of the pheromones are also the basis for prezygotic isolation between species (Cardé & Baker, 1984). The sex pheromones produced by the females consist of multiple components, of which the composition can differ between and within species (Byers, 2006). Therefore even seemingly minor changes in the pheromone blend can have detrimental effects on the efficiency of the pheromone (Hildebrand, 1995) which could lead to less or no mate attraction and even the attraction of allospecific mates (Steiger & Stökl, 2014). A wide variety of these pheromones are found in insects, but the order of Lepidoptera shows potentially the highest diversity (Ando et al., 2004). Moths in particular show a large variety in sex pheromones (A. Groot et al., 2007; Astrid T. Groot et al., 2006; N. Vickers & Baker, 1997; N. J. Vickers, 2002).

Trade-offs between pheromone production and life history traits

This large variety in pheromones comes with a cost either associated with the production of the pheromone itself or indirect on other life history traits, which can for example lead to a trade-off between pheromone production and reproduction in insect species (Steiger et al., 2012). An example can be found in Lobesia botrana moths, the produced pheromone is a condition dependent trait, the production of which is costly resulting in the reduction of signalling behaviour, fecundity and survival of the female (Harari et al., 2011). The produced pheromones are used by the males to determine the size of the females (Harari et al., 2011), the male preference for a specific pheromone blend could have resulted from the differences in the amount of pheromone released or the ratio of the pheromone components (Collins & Cardé, 1985; Jaffe et al., 2007). A different example of a trade-off found in moths is the cost of pheromone production and fecundity, the pheromone production of

females can be reduced after mating (Foster, 2009). After Heliothis virescens females have mated, their haemolymph trehalose (sugar) concentration rapidly decreases (Foster, 2009). Individuals that were given trehalose showed an increase in pheromone titre, which could facilitate further mating and increased fecundity (Foster, 2009). This sugar feeding could also increase the amount of produced juvenile hormone, which causes increased fecundity by faster egg maturation (Ramaswamy et al., 1997). Another type of trade-off is also found in H. virescens:females that were injected with Serratia entomophila a type of bacteria, had a significant higher ratio of 16:Ald to Z11-16:Ald. The latter being a major sex pheromone component, which is essential for mate attraction (Astrid T. Groot et al., 2014). This change in pheromone composition resulted in less male attraction compared to healthy females, which suggests a trade-off between reproduction and immunity investment (Barthel et al., 2015).

Trade-offs within the pheromone blend

Trade-offs are not only found between pheromone production and other life history traits, but also within the pheromone blend itself. This occurs when two compounds share the same precursor, when one of the two compounds increases in amount the other one will decrease. This has been found for a few compounds in Heliothis subflexa moths, which share the same precursor (A. T. Groot et al., 2013). One of these compounds, known as acetates, share the same precursor with the aldehyde 16:Ald (Jurenka, 2003). A strong negative phenotypic correlation has been found between the relative amounts of 16:Ald and the two acetates Z9-16:OAc and Z11-16:OAc. This suggests that when more acetates are produced, less 16:Ald is produced and vice versa (Astrid T. Groot, Estock, et al., 2009). The competition for the same precursor 16:CoA could explain the negative correlation between the compounds (Astrid T. Groot, Estock, et al., 2009) and could be seen as a potential trade-off. Therefore production of high amounts of acetates could have a cost, namely the production of smaller amounts of critical components needed for conspecific mate attraction (N. J. Vickers, 2002).

Heritability of the pheromone blends

Before a trait can evolve it needs to meet multiple criteria. Firstly, there needs to be variation in the given trait among the population. Some of these variations may result in a fitness advantage for the individual over other individuals. This variation needs to be heritable in order to be given to the next generation. Overall the heritability of specific pheromone blends have been found to be relatively high in most insects (Johansson & Jones, 2007). One of the moth species in which a high phenotypic variability in the sex pheromone composition has been found is Heltiohis virescens (Astrid T. Groot et al., 2014). This moth species is a New World generalist herbivore, which feeds on over 37 plant species from 14 different families (Sheck & Gould, 1993). The phenotypic variation of this species was in the large part caused by changes in the relative amount of the major sex pheromone component Z11-16:Ald and its saturated counterpart 16:Ald (Astrid T. Groot et al., 2014). Two artificial selection lines that selected for the ratio of 16:Ald to Z11-16:Ald were created, resulting in a high and a low line (Astrid T. Groot et al., 2014). Offspring of the six mothers with the highest ratio were used for the next generation of selection and this process was continued for 8 generations, this was done also done for the low line (Astrid T. Groot et al., 2014). These selection lines showed an asymmetrical effect, females selected for a relative high amount of 16:Ald to Z11-16:Ald showed a significant increase in the ratio of 16:Ald to Z11-16:Ald (Astrid T. Groot et al., 2014). While the low selection line only showed an effect of selection in the first 4 generations, suggesting that the lower limit of this ratio was reached earlier (Astrid T. Groot et al., 2014). This variation has a genetic basis, which was found by the high heritability of this trait and its reaction to selection (Astrid T. Groot et al., 2014). Furthermore, this variation is likely caused by one or more closely linked genes (Astrid T. Groot et al., 2014). H. virescens is not the only species with high phenotypic variation in sex pheromone blend, as it has also been shown in the closely related species Heltiohis subflexa (Astrid T. Groot, Inglis, et al., 2009).

Heliothis subflexa

The existence of variation, in both species, is surprising because the sex pheromone blend is assumed to be under strong stabilizing selection, as females producing unusual pheromone blends would have more difficulty in finding a mate (Bengtsson & Löfstedt, 2007; Astrid T. Groot et al., 2006). Therefore it is believed that interference from closely related species could have caused the evolution of this signal diversity (Astrid T. Groot et al., 2006). In H. subflexa, coexistence with the closely related species H. virescens is believed to counter this intraspecific stabilizing selection by intense interspecific selection, resulting in a greater divergence between the sex

pheromones of the two species (Astrid T. Groot et al., 2006). Recent research has also found a major QTL that impacts variation in pheromone blends between and within species (A. T. Groot et al., 2013). This means that the same genes could play a role in premating isolation at both inter as intraspecific level (A. T. Groot et al., 2013). The known phenotypic variation and the discovered QTL make the Heltiohtis subflexa an ideal species to use to work with to determine any sex pheromone related heritability and trade-offs.

Heliothis subflexa sex pheromone

H. subflexa is a new-world specialist herbivore moth species that exclusively feeds on plants in the genus Physalis (Laster, 1972). The females of this species produce sex pheromones de novo consisting of multiple components, which can be divided into major and critical secondary sex pheromone components. Heliothis sublexa as well as two closely related species Heliothis virescens and Helicoverpa zea share the same major sex pheromone component Z11–16:Ald (Astrid T. Groot, Inglis, et al., 2009). However H. subflexa has different critical secondary sex pheromone components compared to H. virescens and H. zea, which are Z9–16:Ald and Z11–16:OH (Heath et al., 1990; Nojima et al., 2018; Teal et al., 1981; N. J. Vickers, 2002). The latter component has also been shown to repel both H. zea (Quero & Baker, 1999) and H. virescens (Vetter & Baker, 1983). However H. virescens also produces this compound which at low concentrations attracts conspecific males (Astrid T. Groot et al., 2018) while at high concentrations acts as a repellent (Ramaswamy et al., 1985). Furthermore, the following three compounds are only made by H. subflexa: Z7–16:OAc, Z9–16:OAc and Z11–16:OAc, which are commonly referred to as acetates. The component that has a known dual function of these three is Z11-16:OAc, which acts both as an attractant for H. subflexa and as a repellent for both H. virescens (A. Groot et al., 2007; Astrid T. Groot et al., 2006; N. Vickers & Baker, 1997; N. J. Vickers, 2002) and H. zea (Fadamiro & Baker, 1997; Quero & Baker, 1999). Without this the H. virescens males could hybridize with H. subflexa females resulting in sterile male offspring, which is not beneficial for both species (Laster, 1972). Not only do these pheromones consist of multiple different components, the amount in which they are produced can also differ substantially.

Variation in Heliothis subflexa sex pheromones

Recent studies have found that H. subflexa pheromone blends show geographic and temporal variation within and between areas (Astrid T. Groot, Inglis, et al., 2009). The amount of produced acetates differed per region depending on the presence of H. virescens (Astrid T. Groot, Inglis, et al., 2009). Female H. subflexa moths would produce more acetates in regions with a high number of H. virescens and the opposite happened when almost no H. virescens were present (Astrid T. Groot, Inglis, et al., 2009). The temporal variation in the pheromone blend coincided with changes in pheromone lure traps, the number of caught H. virescens differed per year and the females changed their acetate production accordingly (Astrid T. Groot, Inglis, et al., 2009). This variation was thought to have come by different olfactory cues in the environment, which led to a change in sex pheromone composition (Astrid T. Groot, Inglis, et al., 2009). Research has shown that the amount of Z11‐16:OAc is significantly higher in the sex pheromone of adults, that had been exposed to sex pheromones of H. virescens in an earlier life stage (A. T. Groot et al., 2010). This adaptive phenotypic plasticity in the moth sex pheromone, is referred to as the “experience” hypothesis (A. T. Groot et al., 2010). The results from that research also suggest that behavioural differentiation may precede genetic divergence, which could result in more assortative mating when the amount of acetates are increased (A. T. Groot et al., 2010; Astrid T. Groot et al., 2006). Other findings have found that also the male response on the sex pheromone can show geographic variation (A. Groot et al., 2007) this has also been found in other moth species (Cork et al., 1992; El-Sayed et al., 2003; Gemeno et al., 2000; Hansson et al., 1990; Klun, 1975). Depending on the environmental conditions this plasticity could be lost due to specialization resulting in genetic fixation of the specific gene (Price et al., 2003). This phenotypic plasticity however can also be costly, because the female needs to perceive surrounding pheromones and use this feedback in gene feedback pathways (Price et al., 2003). This potential cost could lead to similar pheromone trade-offs as mentioned earlier, which could potentially also be present in H. subflexa, however these trade-offs are also highly depend on the environmental factors. The effect of environmental conditions can also be variable in this species, resulting in different pheromone compositions per area. Therefore still a lot of research needs to be done on the heritability of the sex pheromone blend, especially the acetylated compounds and potential trade-offs that arise due to the variation in the composition of the sex pheromone blend.

Aim and expectations

Heliothis subflexa have been artificially selected on the total amount of produced acetates for more than a year to create two specific acetate producing lines, referred to as the low and the high line. Combining the selection line information with the pheromone composition data allows performing a pedigree analysis on the heritability of the variation in pheromone compounds, most importantly the acetates. The outcomes of this can be used to further expand our knowledge on the evolution of these pheromone blends and to what the degree the genetic variation in the population causes the variation in the produced pheromone blends. Expected is that the heritability of the total amount of produced acetates in the high line is higher than the low line, which is most likely caused by the stronger response to selection of the high line, while the low acetate selection line stays relatively stable and does not have such a strong response to selection. This is shown in supplementary figure 1, that depicts the evolution of the total acetate level per line. Furthermore, the potential effect of acetate production on the developmental time will be determined. This can shed more light on the potential costs associated with sex pheromone production by the female. Hypothesised is that there will be a trade-off between developmental time and acetate production. This is expected to be caused by the potential high cost of a high acetate production which will lead to a longer developmental time compared to organisms which have a lower acetate production. Thus if the female produces high amounts of acetates then the developmental time is expected to be longer. Lastly, whether acetate production has an effect on the adult survival rate under diet and food stress will also be determined. Exposing individuals to these two stressors, could actually expose potential offs that are hidden under optimal lab conditions. Therefore, it is hypothesised that there will be a trade-off between acetate production and adult survival rate. This is also expected to be caused by the potential high cost of acetate production, therefore the adult survival rate of high acetate producing females will be much lower compared to low acetate producing females under both stresses.

Methods

Organisms and selection lines:

The pedigree and developmental data:

Heliothis subflexa moths, a new world herbivore specialist, were field collected from North Carolina in 2006, which were reared into adulthood in the laboratory (Astrid T. Groot et al., 2014). Each mating consisted of a single-pair cross, which got its own unique number in order to keep track of the families. The unique mating number further referred to as family ID, is a mating specific number which is given to all the offspring of that mating. This family ID will later be used to retrieve both the parental and grandparental data of each female as well as the information on the analysed pheromone blend. The date of egg collection and the date of female emergence from the pupae were also recorded and were later used to calculate the developmental time. The pairs were kept in mating cups which was closed off with a piece of gauze on which the females laid their eggs. When the eggs turned black or neonates had hatched the gauze was collected from the mating cup and placed on Petri-dishes with pinto bean diet. The females that were still alive were injected with PBAN to boost pheromone production and their gland was extracted. Pheromone samples were analysed using GC-ICD in the 2 coming weeks after the eggs were collected. In this order, only the offspring of females that passed the threshold of their line were kept and set-up in separate cup with pinto bean diet to avoid cannibalism. Both selection lines started with the offspring of the females that produced more than 22% of acetate (the 3 Acetates combined) for the high line and less than 16% of acetate for the low line. These values were based on the data collected on the original population (called then back-up) and represent the 1st _{and 3}rd _{quartile. The selection pressure was} increased at generation 3 (new threshold < 22% and >14%) and again at generation 5 (<24% and >12%). The evolution of both section lines as well as the non-selected back-up line can be found in supplementary figure 1. The diet and temperature stress experiment:

Heliothis subflexa neonates were collected from the mating cups of generation 10 for the high selection line and 9 for the low selection line. The neonates were directly placed in separate cups containing different types of diet to prevent cannibalism. The diet consisted of a mixture of agar and nutrients. The ‘normal’ or 100% diet contained a mixture of 76 gram Agar and 576 gram nutrients, while the ‘reduced’ diet also referred to as 25% diet contained 76 gram Agar and 144 gram nutrients. The nutrients consisted out of a mixture of sucrose, 50% soy flour, stabilized wheat germ, Wesson salt mix, USDA vitamin premix, fibre, sorbic acid, methyl paraben and

ascorbic acid. Overall more neonates were put on the 100% diet because these were also used for the temperature stress experiment. Once separated, the neonates were placed in climate chambers with a 14 hours light and 10 hours day rhythm (lights go out at 12:00) and kept under either 25° degrees Celsius for at least two weeks. After this period half of the neonates on normal diet (100%) were kept under 21° C and the rest remained under 25°C. This resulted in three different set ups per line determined by diet type and temperature. Neonates that acted as the control were kept on 100% diet and at 25°C. Neonates that were exposed to diet stress treatment were kept at 25°C and on 25% diet, the temperature stress treatment consisted of individuals that were kept at 21° C and on 100% diet this was done for both the high and low selection lines.

After a month of being exposed to these treatments the number of larvae, pupae and adultsthat were alive were recorded for seven days. At the end the total number of neonates that survived into adulthood as well as the number of neonates that died were calculated to determine the survival rate of both lines for each treatment.

Data

acquisition

All data was acquired from Groot lab located at the University of Amsterdam in the Netherlands. The parental and developmental data were collected during one year, while the data of the diet and temperature stress experiments were collected in April and May of 2020. The following subsections give a brief summary of how the data was edited to make it suitable for statistical analysis, a full explanation about the datasets and alterations can be found in the data repository.

Pedigree for the selection lines:

The parental data of each phenotyped female and the grandparental data of the father were collected. The family number/ID that was given to each unique mating, was used to determine the origin of each individual. This led to the creation of two separate files, one for the high and one for the low selection lines. A separate file containing only the sample ID, family ID and total amount of acetates of each female was also made. These two files were made for each line and were needed to create the pedigrees. However, in order for the pedigree analysis to be performed each individual needs to have unique ID, the family ID was not suitable for this. The reason for this was that the family ID occurred multiple times and therefore were not really unique, a script in R was used to convert the family IDs into unique IDs. These new IDs were created by adding a M or F depending on the sex and a number after the family ID. For family IDs that occurred multiple times the number was increased for each instance. For example family ID 9654 for the mother would become 9654_F_1 if this number occurred another time the second entry would become 9654_F_2, this way all individuals got an unique number. This procedure was carried out for both selection lines and allowed for the creation of one pedigree data frame for each line. The total amount of produced acetates together with the gas chromatograph sample id number, family number, the parental and grandparental data of that individual were combined into one data frame as well. Each mother ID got the same type of labelling as in the pedigree data, this was done by looking at the gas chromatograph sample ID number which is unique for each female. By combining both the gas chromatograph ID with the Mother ID you will get the same ID as in the pedigree, which allows the pedigree analysis to be performed.

Pedigree analysis, phenotype choice

For the pedigree analysis the last four generations of both selection lines were used to calculate the heritability of the total amount of produced acetates. One of the reasons for this was the data availability, the data needed to be digitalized in order for me to analyse it. Especially the first few generations had a lot of erroneous data, which was the reason to start with the later generations that had the most accurate data. Furthermore, the process of figuring out how to convert the family IDs into unique IDs was also very time consuming resulting in overall less time to add the other generations. The four generations that were used to determine the heritability, would have been the generations that had been longest under the selection pressure. So if a difference in the heritability of the total amount of acetates would be present between the two selection lines, it could potentially be found in these generations. The total amount of produced acetates was used as the response variable instead of looking at each of the three acetates separate or at the other components. The reason for this was that when the two selection lines were created, there was selected for the total amounts of acetates

Developmental time:

The developmental time in days was calculated as the difference between the date on which the eggs were collected and the date of hatching. The same data that was used to construct the pedigree was used to calculate this, however only data from generation 5 till 8 of both selection lines could be used as the date of eggs collection wasn’t recorded before. As with the pedigree analysis, the family IDs were used to link the date of egg collection to the right individuals. This was done by using the family ID of which the data of egg collection was recorded and looking it up in the parental records. For example if the data of egg collection was recorded for the family H6231, then the family ID 6231 is looked up in the parental record for the high line. Only the female offspring of this mating were of interest because they produce the acetates, this could result in a family ID occurring multiple times for the female. These females with the same family ID only originated from the same mating and were not the same individual. For each female the data of hatching was recorded, which was subtracted from the acquired date of egg collection. This resulted in the developmental time in days which was calculated for each female. Diet:

The number of neonates that matured into adulthood and that died before this were calculated for each family for both the low and high line. These were used to calculate the survival rate for each family per treatment, the adult survival rate was calculated using only the total number of adults that survived. Thus the number ofadults that were alive were recorded on each of the seven days, these were combined together and subtracted from the original number of neonates that were setup. The reason for this was because the number of dead larvae were most often not recorded while the number of adults that survived were.

Statistical analysis:

All the data was analysed using the statistical program R version 3.6.3 (2020-02-29). Pedigree analysis:

To estimate the heritability of the acetate production in female Heliothis subflexa moths, phenotypic data from grandparental and parental lines were used. Bayesian Animal models were fitted in MCMCglmm (Hadfield, 2019) using an Inverse-Gamma distribution as prior. The behaviour, autocorrelation, effective size and convergence were checked following the tutorial made by Pierre de Villemereuil (De Villemereuil, 2012). Then a single response model was run for both lines with the total amount of acetate as response variable and ran 1 million iterations, discarding the first 100,000 samples as burn-in. The mean and 95% Honest Posterior Density (HPD) interval of the heritability of this trait were calculated.

Developmental time:

Developmental time (coded as the time in days between date of egg collection and date of egg hatching) was analysed as a Gamma variable using a Generalized Linear Model with the dplyr package (Wickham et al., 2020). Generation (5,6,7,8), Line type (Low vs high) and the interaction were included as fixed effects, no random effects were added. To determine which model had the best fit, the full model was compared with ‘lesser’ models using an ANOVA with the likelihood-ratio tests (LRT) function.

Diet and Temperature stress:

Adults survival (coded as a two-column vector, number of adults alive and number of neonates that failed to survive into adulthood) was analysed as a binomial variable using a Generalized Linear Mixed Model with the lme4 package (Bates et al., 2014). Diet type (100 vs 25), Line type (Low vs High) and the interaction were included as fixed effects. Temperature was added as a random effect. The fixed effects were tested using an ANOVA with the likelihood-ratio tests (LRT) function to determine whether including these parameters would give the model a higher accuracy.

Model choice

The generalized linear model used to determine what influenced the developmental time in days for both lines, was chosen out of 4 different models. These models were compared with each other and the model with the best fit was chosen over the other models (table 3). In the end the full model turned out to have the best fit and more biological relevance compared to the other models. Especially the biological relevance was important,

because the main goal of this experiment was to determine whether line, generation or the interaction of the two had an effect on the developmental time. When these three are in the same model, you can exactly see which variable influences the developmental time. If only generation would be added to the model, you would not have known if only generation or the selection lines had an effect on the developmental time. In this case line is very interesting because without we would not know whether acetate production (high or low) has an effect on the developmental time.

For the diet and temperature stress experiment the full model was chosen over the model with only diet as fixed effect and temperature as random effect, even though that model did not have a significantly better fit but did have the lowest AIC value of all three constructed models (table 5). I decided to use the full model over the simplified model because the full model had more biological relevance. If only diet type was added to the model, you would not have known whether only diet type or the selection lines had an effect on the adult survival rate. If the selection lines were not added to the model, we would not know whether acetate production had an effect on the adult survival rate. In addition the inclusion of the interaction between diet and line would show whether the survival rate of the two lines differ per diet type, which was the aim of this experiment.

Results

Pedigree analysis:

The heritability of the acetate production was determined for both selection lines using a Bayesian animal model that was run for 1.000.000 iterations. This model showed that the heritability for both selection lines is almost identical to each other, with the low acetate selection line having a slightly higher acetate production (figure 1). The heritability was calculated using the combined data of the 6th _{till the 9}th _{generations of both selection lines.} The mean, median and the 95% Honest Posterior Density of the heritability of both lines are shown in table 1. The mean heritability of the variance in the acetate production were 0.391 for the high line and 0.399 for the low line. The Bayesian Animal models gave very large 95% Honest Posterior Densities between 2.344e-06 and 0.652 , 3.330e-06 and 0.692 respectively.

Figure 1. Interquartile range of the heritability of the total amount of acetates for both selection lines. The heritability was calculated using Bayesian animal model, run for 1.000.000 iterations with the first 100.000 as burn-in.

The pedigree constructed from generations 6 till 9 of both lines was used as input for this model. The high selection line is shown in red and the low selection line in blue. The mean heritability of both lines are indicated with a diamond and are respectively 0.39 for the high line and 0.40 for the low line. The sample sizes for both lines are equal with n= 9000 for both lines.

Table 1. Mean, median and Honest Posterior Density (PHD) of the heritability of both selection lines.

Mean Median Lower bound.

HPD

Upper bound. HPD

High line 0.3911345 0.3991683 2.344603e-06 0.6524492 Low line 0.399239 0.3998848 3.330892e-06 0.6926318 Developmental time:

The effect of both generation and line on the developmental time of female H. subflexa moths were analysed using a Generalized Linear Model. This generalized linear model showed that the generations had a significant effect (p = 0.00675) on the developmental time (figure 2). While the developmental time of the low line compared to the high line showed no significant difference (p < 0.05)(figure 3), however this effect was almost significant. Lastly, the interaction between the line and generation also had a significant effect (p = 0.01897) on the developmental time (Figure 4). These p-values are summarised in table 2 along with the estimated regression parameters, standard deviations and t-values that were generated by the Generalized Linear Model. The results of the model comparison are shown in table 3.

The means and medians of each generation tend to be relatively the same and only tend to be higher when more outliners are present. An overall trend can be seen in the slow decline of the mean developmental time which in the fifth generation was almost 50 days and in the eighth generation declined to 45 days (figure 2). The same trend can be seen when for each generation both lines are plotted next to each other (figure 4). In this figure only the high selection line tend to show this trend, at the fifth and sixth generations both lines have roughly the same means after which the difference between those two becomes greater (figure 4). The mean and median developmental time in days for the high line start to decline, but the low line stays relatively the same (figure 4). The potential trend seen in the high line however is not significant, but suggests that a correlation between a lower developmental time and high acetate levels is present (figure 4). Also the mean and median of the high line in the eighth generation are more apart from each other, with the median being much lower than the mean value of 45 days. Furthermore, both lines did not show a significant effect on the developmental time with the low line having a slightly higher mean and median developmental time (figure 3).

Figure 2. Generation effect on developmental time of both lines combined. The Generalized Linear Model showed that generations had a hyper significant effect on the developmental time (P = 0.00128). The interquartile range of generations 5 to 8 are plotted against the developmental time in days of each individual. The means of the generations are depicted with a diamond in each corresponding boxplot. The values of the mean developmental time for each generation are respectively: Gen5: 47.55 , Gen6: 46.39 , Gen7: 46.48 , Gen8: 45.76 days. The sample sizes for each generation are respectively for generation 5 till 8: n= 715 , n= 2262 , n= 2898 and n= 1040.

Figure 3. Selection line effect on the female developmental time. The Generalized Linear Model showed that line had a non-significant effect on the developmental time (P = 0.05106). The interquartile ranges of the two selection lines are plotted against the developmental time in days of all four generations combined. The high selection line is depicted in red while the low selection line is depicted in blue. The means are depicted with diamonds with corresponding colouring, the mean values are respectively 46.06 for the high line and 46.99 for the low line. The sample sizes for each line are respectively n= 3635 for the high line and n=3280 for the low line.

Figure 4. Generation and line effect on the developmental time for each line separate. The interaction between the selection lines and generation showed a significant effect according to the Generalized Linear Model (P = 0.01897). The interquartile range of generations 5 till 8 are plotted against the developmental time in days of the individuals of both lines. Each generation has two boxplots the high lines are depicted in red and the low lines in blue. The mean developmental time for each generation of both lines are depicted in diamonds with the same colours as the boxplots. These are respectively for the high and low lines: Gen5: 47.45 and 47.63 , Gen6: 46.42 and 46.37 , Gen7: 45.71 and 47.34 , Gen8: 44.93 and 46.94. The sample sizes for each line are respectively for the high and low lines; Gen5: n= 330 and n= 385 , Gen6: n= 1164 and n= 1098 , Gen7: n= 1533 and n= 1365 , Gen8: n= 608 and n= 432.

Table 2. Estimated regression parameters, standard errors, t-values and P-values for the Gamma GLMM presented in figures 2 till 4.

Estimate Std. error t value P-value Intercept 0.0192296 0.0007737 24.853 < 2e-16

Generation 0.0003795 0.0001176 3.228 0.00128

LineL 0.0021490 0.0011003 1.953 0.05106

Generation:LineL -0.0003952 0.0001682 -2.350 0.01897

Table 3. Results of Likelihood Ratio Tests comparing the Gamma generalized linear models. With nested and complex representing the variables in the model. Nested being the ‘simple’ model and complex the ‘complex’ model, represented by the added variables.

Nested Complex AIC

Nested AIC Complex Deviance Degrees of Freedom P-value 1 Generation 6430 6425.8 0.070718 1 0.01801

Generation Generation + Line 6425.8 6419.2 0.097728 1 0.005145

Generation + Line

Generation + Line + Generation * Line

Diet:

The effects of Diet type (25% and 100%), Line (High and Low) and the interaction of Line and diet type on the survival rate of adult moths were determined using a Generalized Linear Mixed Model, with temperature as a random effect. This model showed that diet type has a significant effect on the adult survival rate (figure 5). The interaction between line and diet type seems to follow the same trend as diet type, both lines have a much lower survival rate on the 25% compared to the 100% (figure 6). However this interaction between diet type and line as well as the line effect on the survival rate were not significant (P= 0.2348 and P= 0.2097) (table 4). Furthermore, both lines had the highest survival rate on the temperature stress experiment (figure 7) and the overall lowest survival rate on the diet stress. The p-values are summarised in table 4 along with the estimated regression parameters, standard deviations and t-values that were generated by the Generalized Linear Model. The results of the model comparison are shown in table 5.

The adult survival rate was influenced by the diet the neonates were raised on (figure 5), this is also shown when both line and diet type are plotted together (figure 6). Even though both lines have their lowest survival rate on the 25% diet and thus their highest on the 100% diet, there seems to be an opposite line effect/response. While the high line seems to be doing worse on the reduced diet compared to the low line, it does have an almost 4% difference in mean survival rate compared to the low line. Both line and the interaction between line and diet type seem to have an effect on the adult survival rate, however this effect is not significant (figure 6 and table 3). The last graph which shows the interactions between diet type and temperature for each line (not included in the model) which does show a similar effect as shown in figure 6.

Temperature was not added as a fixed effect in the Generalized Linear Model, graph 7 shows the effect of the three treatments on the adult survival rate. This graph showed that both lines had the highest survival rate on the temperature stress treatment, with almost similar mean values. The control treatment on the other hand showed a much larger difference in adult survival rate between the two lines, the high line had a mean survival rate that was 9% higher (26.66%) compared to the low line (17.20%). The high line has the lowest survival rate on the diet stress treatment while the low line has a slightly higher survival rate compared to the control treatment (figure 7).

Figure 5. Effect of diet type on the survival rate of adult Heliothis subflexa moths. The Generalized Linear Mixed Model showed that diet type significantly effects the survival rate of adults (P = 0.0302). The interquartile range of the survival rate for both diet types are plotted, the 25% diet in green and the 100% diet in orange. The diamonds represent the mean survival rate of both diet types, which are respectively 16.86% for the 25% diet and 25.75% for the 100% diet. The sample sizes of the two diet types are respectively n= 427 for the 25% and n= 683 for the 100% diet.

Figure 6. Effect of diet type on the adult survival rate per selection line. The Generalized Linear Mixed Model showed that the interaction of diet type and line had no significant effect on the adult survival rate (P = 0.2097). The interquartile ranges of the survival rate for both the diet types and lines separate are shown here. The high lines are depicted in red and the low lines in blue. The means of each line are depicted by a diamond in each boxplot which are respectively for the high and low lines; 25% diet: 14.56% and 18.59% , 100% diet: 27.92% And 24.22%. The sample sizes for each line are respectively for the high and low lines; 25% diet: n= 213 and n= 214 , 100% diet: n= 330 and n= 353.

Figure 7. Effect of diet type on the adult survival rate per line. The interquartile range of the adult survival is plotted against the diet and temperature treatments per selection line. The high lines are depicted in red and the low lines in blue. The temperature and diet treatments are plotted on the x-axis, the first number represents the diet type while the second is for the temperature. Three treatments were used from left to right; Temperature stress, control and diet stress. The means of each line are depicted with a diamond and are respectively for the high and low line; temperature stress: 28.28% and 30.46% , control: 26.66% and 17.20% , diet stress: 14.56% and 18.59%. The sample sizes for each line are respectively for the high and low lines; temperature stress: n= 180 and n= 193 , control: n= 150 and n= 160 , diet stress: n= 213 and n= 214.

Table 4. Estimated regression parameters, standard errors, t-values and P-values for the Binomial GLMM presented in figures 5 and 6.

Estimate Std. error t value P-value

Intercept -1.5944 0.2487 -6.412 1.44e-10

Diet type 100% 0.5636 0.2600 2.167 0.0302

LineL 0.3232 0.2577 1.254 0.2097

Diet type 100: LineL -0.3697 0.3112 -1.188 0.2348

Table 5. Results of Likelihood Ratio Tests comparing the Binomial generalized linear mixed models. With nested and complex representing the variables in the model. Nested being the ‘simple’ model and complex the ‘complex’ model, represented by the added variables. All models have temperature added as random effect, only the variables that differ for each model are shown.

Nested Complex AIC

Nested AIC Complex Deviance Nested Deviance Complex Degrees of Freedom P-value Diet type Diet type + line 354.72 356.48 348.72 348.48 1 0.6251 Diet type +

line

Diet type + line + Diet type * line

Discussion

The heritability of the total amount of produced acetates was almost identical for both selection lines at around 0.40. This means that 40% of the variation in the total amount of produced acetates is caused by the genetic variation in the populations of both acetate selection lines (figure 1). Overall ectotherms tend to have a significantly lower heritability for morphological traits compared to endotherms (Mousseau & Roff, 1987), which could explain the relatively low heritability that has been found in both lines. This sounds surprising when looking at supplementary figure 1, which clearly shows that the high acetate selection line has a stronger response to selection. This was also the reason why the high acetate selection line was expected to have a higher heritability compared to the low acetate selection line, which did not show such strong response to selection. However, a weakly heritable trait is not less likely to be selected. Biological traits that are closely associated with fitness will generally have lower heritability’s than traits that are loosely connected with fitness (Mousseau & Roff, 1987). This could suggest that acetate production is closely related with fitness, which could support the potential trade-off between adult survival rate and acetate production. The difference in response to selection cannot be explained by this nearly identical heritability of the total amount of acetates, this suggest that a different trait or process influences the strong response to selection. It would be interesting to see whether there is a difference in body size between these two lines, if so whether there is a difference in heritability of this trait. Then the correlation between the body size and total amount of produced acetates can be determined to see whether body size does influence the total amount of acetates produced, thus the way how the moths respond to selection. These kind of correlations have already been found in other ectotherms as in developmental time and fecundity with body size (Honěk, 1993; Kusano, 1982; Roff, 1986; Symonds et al., 2012) . Similar correlations could be present in H. subflexa between acetate production and body size.

The developmental time of female H. subflexa moths was significantly affected by both generation as well as the interaction between generation and selection line (figures 2 and 4). The high selection line did show a declining trend in developmental time over the course of the four generations, ultimately reaching the lowest developmental time of the two selection lines (figure 4). If a trade-off was present between acetate production and developmental time, then the high acetate selection line would have shown an increase in the developmental time over the four generations. This result, and the absences of a potential trade-off between developmental time and (a high) acetate production were not in line with, my expectations. The shown trend could have been caused by a line and generation specific effect on the developmental time, resulting in the decrease of the developmental time of the high selection lines. However this trend could have also been caused by the difference in sample sizes of both the four generations. Both the sixth and the seventh generations had much higher sample sizes compared to the other two generations, which had a difference of more than 1000. However the sample sizes of the two selection lines did not differ that much from each other, with only the sixth and the ninth generations having much lower sample sizes compared to the other two. Adding data of generation 10 and higher with more equal sample sizes, could show whether this trend is really caused by a generation effect or is just caused by a difference in sample sizes. Another point of interest regarding the developmental time is that a relatively short time scale was used, only four generations ranging from 4 till 8. As mentioned here above, adding information of more generations could show a potentially greater trend and difference between the two selection lines.

The last experiment showed that the diet type on which the neonates were kept significantly influences the adult survival rate (figure 5), which proves that diet type does have an effect on the adult survival rate. However the fact that the high line seemed to have a higher survival rate compared to the low line on the treatments containing the 100% diets was not expected (figure 7). When diet is not limited (100%) temperature does not seem to affect the adult survival rate for high acetate producing females. When food is limited both lines have a considerably lower survival rate, but the high line has the lowest of the two. This suggests that there could be a trade-off between survival and acetate production when food is limited, however due to the setup of the experiment this cannot be specifically determined. The main reason why temperature was added as a random effect instead of a fixed effect to the Generalized Linear Mixed Model was because not all diet types were kept under the same temperatures. At first glance an apparent effect of diet type on the survival rate of the adults can be seen (figure 5). Even though diet type has a significant effect on the survival rate, this figure does give a false image. It looks like there is a large difference caused by the two diet types especially when the survival rates

for both lines are plotted alongside each other (figure 6). When temperature is added you can see that the 100% diet is overrepresented in the experiment because it contains two different temperature treatments while 25% diet only contains one treatment, which is 25°C (figure 7). Originally it was not the plan to look at both diet and temperature stress, but due to the corona crisis it was decided to also look at the effect of temperature stress on the adult survival rate. The design was relatively balanced for the diet stress treatment for which families were equally distributed among treatments to prevent a family effect on the outcome. Adding an additional treatment to the experiment caused the design to become unbalanced resulting in unequal distribution of the families over the three treatments. This was the reason why temperature was added as a random effect to account for the unequal family distribution among the treatments.

Having an unbalanced family design could have caused the families to reveal a potential hidden preference for specific temperatures. Families that were kept on lower temperatures could have had a preference for lower temperatures while the families that were kept under control conditions had the opposite preference. The differences between these preferences for the families could have resulted in the different responses to the diet stress and control treatments.

Another explanation could be that the individuals in general ‘prefer’ a suboptimal temperature, this could result in a slower metabolism and thus a longer developmental time. This longer developmental time could at the same time result in a higher adult survival rate. Previous research has shown that both the duration of the egg stage, larval stage, pupal stage (combined are the developmental time) as well as the longevity of the adults (amount of survived days) were the longest when H. subflexa was kept on 25°C or lower temperatures (G. D. Butler et al., 1979). A similar temperature effect on both the different life stages as well as the adult survival rate was found in H. virescens (G. D. Butler & Hamilton, 1976) and H. zea (George D. Butler, 1976). Interesting to see is that the H. subflexa pupae go into diapause when the temperature drops below 20°C (G. D. Butler et al., 1979), this could suggest that a trade-off could be present between survival and development when the temperatures are too low. This research also compared the duration of these life stages of H. subflexa with H. virescens and found that the egg and larval stages of H. subflexa did significantly differ from H. virescens, which one average took longer on both low (20°C) and high (30°C) temperatures (G. D. Butler et al., 1979). This could suggest that there is a trade-off between acetate production and developmental time present, when the organisms are exposed to different (extreme) temperatures.

In the adult survival rate experiment, two out of the three treatments had no food limitation, which could have resulted in a more moderate metabolic rate in which the organisms ‘can take their time’ to further develop resulting in a higher survival rate. The fact that the survival rate in the 25°C setup was much lower compared to the 21°C, could have been caused by the opposite effect. A higher temperature could result in a higher metabolic rate that causes an organism to eat much more in order to keep up with this increased metabolic rate. The individuals exposed to the diet stress treatment, could potentially not keep up with the high metabolic rate which resulted in a much lower survival rate compared to control treatment. The metabolic rate was not determined in all three treatments and therefore no solid conclusions can be made, however this could suggest that metabolic rate which is influenced by the temperature in the environment played a role in the adult survival rate. Looking back at the introduction where the geographic and temporal variation of the sex pheromone is described, this research shows that the difference in the amount of produced acetates between populations with and without H. virescens is still not clear. If a high acetate production has no cost as the results of the developmental time experiment suggest than its strange why the populations with and without H. virescens differ so much. If no cost is associated with the production of high amounts of acetates, you would expect to find similar acetate levels in these different populations, this is not the case suggesting that something other than the presence of H. virescens is causing this differentiation. Potential other environmental factors other than the presence of closely related species could influence the acetate production, like diet as found in the stress experiment. Overall the heritability was found not to be super high, while the high selection line did show a stronger response to selection. This could be explained by the fact that only the last 4 generations are used and that during the creation of both selection lines that the variation decreased resulting in these relatively low identical values. The low heritability could also mean that a lot of variation in the total amount of produced acetates caused by environmental factors even though both lines are kept under the same conditions in the laboratory. It could be that the conditions in the laboratory do influence the heritability of the total amount of

produced acetates, however due to the fact that the organisms are kept under the same conditions this would be really peculiar, this definitely something that is worth looking at more in future research.

Overall it would be interesting to redo the diet and temperature stress experiment with equal group sizes, family distributions and temperature treatments. It would also be interesting to keep individuals under more than two temperatures, now only 25°C and 21°C were used, but the effect of higher and even lower temperatures on the survival rate of the moths could also cause differences between the two selection lines. Especially, to see whether these lower and higher temperatures could reveal a potential hidden trade-off, for example between survival and growth (diapause or not). Furthermore, it would be interesting to see what the effect of a relative lower temperature on female H. subflexa would be if they were kept on 25% diet. As suggested above, this might lead to a slightly higher survival rate compared to the other 25% diet treatment. However families do need to be randomly distributed among treatments in order to rule out a family effect on the results.

This follow up experiment could be combined with the developmental time experiment. During these experiments the developmental time of the individuals could also be recorded, this way we can also determine the effect of diet type and temperature on the developmental time. It would be interesting to determine whether there is a difference in developmental time between the two selection lines if they are kept under different temperatures and on different diet types. Lower temperatures would probably result in a relatively longer developmental time compared to higher temperatures because the metabolism will be much slower (Neven, 2000). This is mainly caused by the fact that insects are exothermic, which means that their metabolic rate is extremely dependent on environmental temperature (Neven, 2000).

Furthermore, the organisms could be exposed to different stressor as well, such as the presence of H. virescens during the mating as well as when the neonate is in a pupae, research has shown that the amount of produced acetates increases when females are exposed to H. virescens sex pheromones in early life (A. T. Groot et al., 2010). The increase of acetates could also have a cost associated with it, which could show a potential trade-off. Lastly the effect of infecting a neonate with a parasite on the developmental time and most importantly on the (adult) survival rate would also be interesting to investigate. Especially, whether the two selection lines differ from each other which again could expose a hidden trade-off. Parasite infection in H. zea has shown that infection does lengthen the developmental time and does not increase the mortality (Gaugler & Brooks, 1975) a similar effect on the developmental progression in weight has been found in Trichoplusia ni (Thompson, 1982), but the response could of course be host and parasite specific which could result in different or similar effects in H. subflexa.

As for the heritability experiment, it would be nice to add the other generations to see if using a larger time scale could show a difference between the two selection lines. Furthermore, the heritability of each acetate can be determined to see whether there is a difference in the heritability of these acetates within and between both selection lines. At last the heritability of the total amount of produced acetates as well as each acetate separate can be calculated for each generation separate. This could show whether the heritability of these traits changes over the course of 9 generations or if it relatively stayed the same. If the heritability stays roughly the same for both selection lines then the strong response to selection of the high line could be caused by environmental effects or non-additive genetic variation, which would be interesting to further investigate.

Overall it is important to keep in mind that trade-offs are almost everywhere but that they can be hard to detect. Organisms in the laboratory are normally kept under optimal conditions, which could easily hide potential trade-offs. Therefore exposing organisms to stressful treatments such as a reduced diet or extreme temperatures is a good way to shift their balance from optimal to less/non-optimal conditions, and gain information on processes that may be occurring in the wild.

Acknowledgements

I would like to thank Astrid Groot for allowing me to do my bachelor project in the Groot lab, Elise Fruitet and Emily Burdfield-steel for their help and supervision during this project especially during the corona pandemic, Thomas Blankers for helping me out with creating and performing the pedigree analysis in R, and

Jet ten Berge for digitalizing the data and therefore making it more accessible.

Supplementary figures

Supplementary figure 1. The evolution of acetate ratio for each selection line and per generation (Unpublished data, courtesy of the Groot lab). The mean proportion of acetates produced in the pheromone blend of Hs females across 9 generations from high (orange) and low (green) selection lines. The non-selected back-up line is shown in blue. Stars represent an increase of the selection pressure. Bars represent standard error, N ~ 2800. The stars indicate the changes in selection pressure for the total amount of produced acetates.

References

Ache, B. W., & Young, J. M. (2005). Olfaction: Diverse Species, Conserved Principles. Neuron, 48(3), 417–430. https://doi.org/10.1016/j.neuron.2005.10.022

Adler, P. B., Salguero-Gomez, R., Compagnoni, A., Hsu, J. S., Ray-Mukherjee, J., Mbeau-Ache, C., & Franco, M. (2014). Functional traits explain variation in plant life history strategies. Proceedings of the National Academy of Sciences, 111(2), 740–745.

https://doi.org/10.1073/pnas.1315179111

Ando, T., Inomata, S.-I., & Yamamoto, M. (2004). Lepidopteran sex pheromones. Topics in Current Chemistry, 239, 51–96. https://doi.org/10.1007/b95449

Bakker, T. C. M., & Pomiankowski, A. (1995). The genetic basis of female mate preferences. Journal of Evolutionary Biology, 8(2), 129–171. https://doi.org/10.1046/j.1420-9101.1995.8020129.x Barthel, A., Staudacher, H., Schmaltz, A., Heckel, D. G., & Groot, A. T. (2015). Sex-specific

consequences of an induced immune response on reproduction in a moth. BMC Evolutionary Biology, 15(1), 282. https://doi.org/10.1186/s12862-015-0562-3

Bates, D., Mächler, M., Bolker, B., & Walker, S. (2014). Fitting Linear Mixed-Effects Models Using lme4. ArXiv e-prints, arXiv:1406. https://doi.org/10.18637/jss.v067.i01

Bengtsson, B. O., & Löfstedt, C. (2007). Direct and indirect selection in moth pheromone evolution: Population genetical simulations of asymmetric sexual interactions. Biological Journal of the Linnean Society, 90(1), 117–123. https://doi.org/10.1111/j.1095-8312.2007.00715.x

Butler, G. D., & Hamilton, A. G. (1976). Development Time of Heliothis virescens in Relation to Constant Temperature ,. Environmental Entomology, 5(4), 759–760. https://doi.org/10.1093/ee/5.4.759 Butler, G. D., Hamilton, A. G., & Proshold, F. I. (1979). Developmental times of Heliothis virescens and

H. subflexa in Relation to Constant Temperature,. Annals of the Entomological Society of America, 72(2), 263–266. https://doi.org/10.1093/aesa/72.2.263

Butler, George D. (1976). Bollworm: Development in Relation to Temperature and Larval Food ,. Environmental Entomology, 5(3), 520–522. https://doi.org/10.1093/ee/5.3.520

Byers, J. A. (2006). Pheromone component patterns of moth evolution revealed by computer analysis of the Pherolist. Journal of Animal Ecology, 75(2), 399–407. https://doi.org/10.1111/j.1365-2656.2006.01060.x

Cardé, R. T., & Baker, T. C. (1984). Sexual Communication with Pheromones. In W. J. Bell & R. T. Cardé (Red.), Chemical Ecology of Insects (pp. 355–383). Springer US. https://doi.org/10.1007/978-1-4899-3368-3_13

Collins, R. D., & Cardé, R. T. (1985). Variation in and Heritability of Aspects of Pheromone Production in the Pink Bollworm Moth, Pectinophora gossypiella (Lepidoptera: Gelechiidae). Annals of the Entomological Society of America, 78(2), 229–234. https://doi.org/10.1093/aesa/78.2.229 Cork, A., Boo, K. S., Dunkelblum, E., Hall, D. R., Jee-Rajunga, K., Kehat, M., Kong Jie, E., Park, K. C.,

budworm,Helicoverpa assulta (Guenee) (Lepidoptera: Noctuidae): Identification and field testing. Journal of Chemical Ecology, 18(3), 403–418. https://doi.org/10.1007/BF00994240

De Villemereuil, P. (2012, oktober). (PDF) Fitting Linear Mixed-Effects Models Using lme4. https://www.researchgate.net/publication/263352347_Fitting_Linear_Mixed-Effects_Models_Using_lme4

Ellis, B. J., Figueredo, A. J., Brumbach, B. H., & Schlomer, G. L. (2009). Fundamental Dimensions of Environmental Risk: The Impact of Harsh versus Unpredictable Environments on the Evolution and Development of Life History Strategies. Human Nature, 20(2), 204–268.

https://doi.org/10.1007/s12110-009-9063-7

Ellner, S. (1987). Alternate plant life history strategies and coexistence in randomly varying environments. Vegetatio, 69(1), 199–208. https://doi.org/10.1007/BF00038701

El-Sayed, A., Delisle, J., Lury, N., Gut, L., Judd, G., Legrand, S., Reissig, W., Roelofs, W., Unelius, C., & Trimble, R. (2003). Geographic Variation in Pheromone Chemistry, Antennal Electrophysiology, and Pheromone-Mediated Trap Catch of North American Populations of the Obliquebanded Leafroller. Environmental Entomology, 32, 470–476. https://doi.org/10.1603/0046-225X-32.3.470

Fadamiro, Henr. y, & Baker, T. C. (1997). Helicoverpa zea males (Lepidoptera: Noctuidae) respond to the intermittent fine structure of their sex pheromone plume and an antagonist in a flight tunnel. Physiological Entomology, 22(4), 316–324.

https://doi.org/10.1111/j.1365-3032.1997.tb01175.x

Fitzpatrick, S., Berglund, A., & Rosenqvist, G. (1995). Ornaments or offspring: Costs to reproductive success restrict sexual selection processes. Biological Journal of the Linnean Society, 55(3), 251–260. https://doi.org/10.1111/j.1095-8312.1995.tb01063.x

Foster, S. (2009). Sugar feeding via trehalose haemolymph concentration affects sex pheromone production in mated Heliothis virescens moths. Journal of Experimental Biology, 212(17), 2789– 2794. https://doi.org/10.1242/jeb.030676

Gaugler, R. R., & Brooks, W. M. (1975). Sublethal effects of infection by Nosema heliothidis in the corn earworm, Heliothis zea. Journal of Invertebrate Pathology, 26(1), 57–63.

Gemeno, C., Lutfallah, A. F., & Haynes, K. F. (2000). Pheromone Blend Variation and Cross-Attraction among Populations of the Black Cutworm Moth (Lepidoptera: Noctuidae). Annals of the Entomological Society of America, 93(6), 1322–1328. https://doi.org/10.1603/0013-8746(2000)093[1322:PBVACA]2.0.CO;2

Groot, A., Santangelo, R., Ricci, E., Brownie, C., Gould, F., & Schal, C. (2007). Differential Attraction of Heliothis subflexa Males to Synthetic Pheromone Lures in Eastern US and Western Mexico. Journal of chemical ecology, 33, 353–368. https://doi.org/10.1007/s10886-006-9233-6 Groot, A. T., CLAβEN, A., Staudacher, H., Schal, C., & Heckel, D. G. (2010). Phenotypic plasticity in

sexual communication signal of a noctuid moth. Journal of Evolutionary Biology, 23(12), 2731– 2738. https://doi.org/10.1111/j.1420-9101.2010.02124.x

Groot, A. T., Staudacher, H., Barthel, A., Inglis, O., Schöfl, G., Santangelo, R. G., Gebauer‐Jung, S., Vogel, H., Emerson, J., Schal, C., Heckel, D. G., & Gould, F. (2013). One quantitative trait locus for intra- and interspecific variation in a sex pheromone. Molecular Ecology, 22(4), 1065–1080. https://doi.org/10.1111/mec.12171

Groot, Astrid T., Estock, M. L., Horovitz, J. L., Hamilton, J., Santangelo, R. G., Schal, C., & Gould, F. (2009). QTL analysis of sex pheromone blend differences between two closely related moths: Insights into divergence in biosynthetic pathways. Insect Biochemistry and Molecular Biology, 39(8), 568–577. https://doi.org/10.1016/j.ibmb.2009.05.002

Groot, Astrid T., Horovitz, J. L., Hamilton, J., Santangelo, R. G., Schal, C., & Gould, F. (2006).

Experimental evidence for interspecific directional selection on moth pheromone communication. Proceedings of the National Academy of Sciences, 103(15), 5858–5863.

https://doi.org/10.1073/pnas.0508609103

Groot, Astrid T., Inglis, O., Bowdridge, S., Santangelo, R. G., Blanco, C., Jr, J. D. L., Vargas, A. T., Gould, F., & Schal, C. (2009). Geographic and Temporal Variation in Moth Chemical Communication. Evolution, 63(8), 1987–2003. https://doi.org/10.1111/j.1558-5646.2009.00702.x

Groot, Astrid T., Nojima, S., Heath, J. J., Ammagarahalli, B., van Wijk, M., Claβen, A., Santangelo, R. G., Lopez, J., & Schal, C. (2018). Alcohol Contributes to Attraction of Heliothis (= Chloridea)

virescens Males to Females. Journal of Chemical Ecology, 44(7–8), 621–630. https://doi.org/10.1007/s10886-018-0995-4

Groot, Astrid T., Schöfl, G., Inglis, O., Donnerhacke, S., Classen, A., Schmalz, A., Santangelo, R. G., Emerson, J., Gould, F., Schal, C., & Heckel, D. G. (2014). Within-population variability in a moth sex pheromone blend: Genetic basis and behavioural consequences. Proceedings of the Royal Society B: Biological Sciences, 281(1779), 20133054. https://doi.org/10.1098/rspb.2013.3054 Hadfield, J. (2019). MCMCglmm: MCMC Generalised Linear Mixed Models (2.29) [Computer software].

https://CRAN.R-project.org/package=MCMCglmm

Hansson, B. S., Tóth, M., Löfstedt, C., Szöcs, G., Subchev, M., & Löfqvist, J. (1990). Pheromone variation among eastern European and a western Asian population of the turnip mothAgrotis segetum. Journal of Chemical Ecology, 16(5), 1611–1622. https://doi.org/10.1007/BF01014094 Harari, A. R., & Steinitz, H. (2013). The evolution of female sex pheromones. Current Zoology, 59(4),

569–578. https://doi.org/10.1093/czoolo/59.4.569

Harari, A. R., Zahavi, T., & Thiéry, D. (2011). Fitness Cost of Pheromone Production in Signaling Female Moths. Evolution, 65(6), 1572–1582. https://doi.org/10.1111/j.1558-5646.2011.01252.x Harvey, P. H., & Clutton‐Brock, T. H. (1985). Life History Variation in Primates. Evolution, 39(3), 559–

581. https://doi.org/10.1111/j.1558-5646.1985.tb00395.x

Heath, R. R., Mitchell, E. R., & Cibrian Tovar, J. (1990). Effect of release rate and ratio of (Z)-11-hexadecen-1-ol from synthetic pheromone blends on trap capture ofHeliothis subflexa (Lepidoptera: Noctuidae). Journal of Chemical Ecology, 16(4), 1259–1268.

https://doi.org/10.1007/BF01021024

Hildebrand, J. G. (1995). Analysis of chemical signals by nervous systems. Proceedings of the National Academy of Sciences, 92(1), 67–74. https://doi.org/10.1073/pnas.92.1.67

Honěk, A. (1993). Intraspecific Variation in Body Size and Fecundity in Insects: A General Relationship. Oikos, 66(3), 483–492. JSTOR. https://doi.org/10.2307/3544943

Iwasa, Y., & Pomiankowski, A. (1994). The Evolution of Mate Preferences for Multiple Sexual Ornaments. Evolution, 48(3), 853–867. https://doi.org/10.1111/j.1558-5646.1994.tb01367.x