Heterozygote advantage and pleiotropy
contribute to intraspecic color trait
Chiara De Pasqual,1,2 ,3 Kaisa Suisto,1Jimi Kirvesoja,1Swanne Gordon,4Tarmo Ketola,1
and Johanna Mappes1,2
1Department of Biological and Environmental Science, University of Jyväskylä, Jyväskylä 40014, Finland
2Organismal and Evolutionary Biology Research Program, University of Helsinki, Helsinki 00014, Finland
4Department of Ecology and Evolutionary Biology, Cornell University, Ithaca, New York 14853
Received March 31, 2022
Accepted June 29, 2022
The persistence of intrapopulation phenotypic variation typically requires some form of balancing selection because drift and
directional selection eventually erode genetic variation. Heterozygote advantage remains a classic explanation for the maintenance
of genetic variation in the face of selection. However, examples of heterozygote advantage, other than those associated with
disease resistance, are rather uncommon. Across most of its distribution, males of the aposematic moth Arctia plantaginis have
two hindwing phenotypes determined by a heritable one locus-two allele polymorphism (genotypes: WW/Wy =white morph,
yy =yellow morph). Using genotyped moths, we show that the presence of one or two copies of the yellow allele affects several
life-history traits. Reproductive output of both males and females and female mating success are negatively affected by two copies
of the yellow allele. Females carrying one yellow allele (i.e., Wy) have higher fertility, hatching success, and offspring survival than
either homozygote, thus leading to strong heterozygote advantage. Our results indicate strong female contribution especially at
the postcopulatory stage in maintaining the color polymorphism. The interplay between heterozygote advantage, yellow allele
pleiotropic effect, and morph-specic predation pressure may exert balancing selection on the color locus, suggesting that color
polymorphism may be maintained through complex interactions between natural and sexual selection.
KEY WORDS: Color locus, heterozygote advantage, intraspecic trait variation, life-history traits, pleiotropy, wood tiger moth.
The origin and maintenance of polymorphism—the co-
occurrence of more than two distinct morphs—within natural
populations constitute a long-standing conundrum in evolution-
ary biology (Ford 1945; Huxley 1955; White 2017). Drift alone
can erode phenotypic variation from populations in a few hun-
dred generations (Nevo et al. 1997). If traits are under selection,
polymorphism is even more puzzling. Theory predicts that traits
contributing to the fitness of individuals should be under strong
natural and stabilizing selection and drive the more fit morph
to fixation (Endler, 1988; Cardé and Baker, 1984). Still, color
polymorphic populations are widespread in nature (e.g., Sinervo
and Lively 1996; Pryke and Griffith, 2007; Maan and Cummings,
2008; Hegna et al., 2015). Traits (i.e., coloration) may be shaped
by complex evolutionary processes through multiple and nonmu-
tually exclusive selective pressures (Gray and McKinnon, 2007),
which drive and maintain phenotypic variation and genetic diver-
sity in nature (Fisher 1930; Ford 1945).
Coloration, for example, plays an important role in a variety
of ecological and physiological processes (Endler and Mappes,
2017; Cuthill et al., 2017), from camouflage (Duarte et al., 2017),
to warning coloration (Mappes et al., 2005) and sexual selection
(Maan and Cummings, 2008). Thus, color polymorphism may be
the result of natural selection (Gray and McKinnon 2007), sexual
selection (Wellenreuther et al., 2014), their combination (Maan
and Cummings, 2008), and/or pleiotropic effects (i.e., when a
single locus affects two or more phenotypic traits) because color
© 2022 The Authors. Evolution published by Wiley Periodicals LLC on behalf of The Society for the Study of Evolution.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original
work is properly cited.
C. DE PASQUAL ET AL.
morphs are often genetically correlated with other traits (McK-
innon and Pierotti 2010). Alternative color morphs often dif-
fer in features other than color (McKinnon and Pierotti 2010).
For example, variable morph-specific behavioral strategies, such
as territoriality (Sinervo and Lively 1996), aggressiveness and
dominance (Pryke and Griffith 2007), or alternative reproductive
strategies, may exist (Sinervo and Lively 1996; Tuttle 2003).
Complex phenotypes can be controlled by simple ge-
netic mechanisms (i.e., one or few genes). In Drosophila
melanogaster, a gene responsible for cuticle pigmentation, yel-
low, has pleiotropic effects on other traits in males. The lack of
function of the yellow gene disrupts body pigmentation expres-
sion, male courtship behavior, and mating success (Bastock 1956;
Wilson et al., 1976; Massey et al., 2019) caused by a morpholog-
ical and structural change on the leg section used by the male
to grasp the female (i.e., sex comb) (Massey et al., 2019). In the
case of the common wall lizard (Podarcis muralis), the simple ge-
netic basis of the color polymorphism leads to pleiotropic effects
in numerous traits (Andrade et al., 2019), including morphology
(Sacchi et al., 2007), behavior (Abalos et al., 2016), physiology
(Galeotti et al., 2010), immunology (Calsbeek et al., 2010), and
reproduction (Galeotti et al., 2013).
Intraspecific color polymorphism maintenance typically re-
quires some form of balancing selection, achieved through color
morph fluctuations resulting from negative frequency-dependent
selection (FDS) (Wellenreuther et al., 2014) or independent of
the relative abundance of a morph (Pryke and Griffith, 2007;
Hedrick et al., 2016). Negative FDS mediated by sexual selec-
tion can maintain multiple color morphs in natural populations,
for example, through alternative male reproductive strategies in
the side-blotched lizards (Uta stansburiana) (Sinervo and Lively
1996), through rare morph advantage in guppies (Poecilia retic-
ulata) (Hughes et al., 2013), or through FD sexually antagonistic
selection in blue-tailed damselflies (Ischnura elegans) (Svensson
and Abbott, 2005; Svensson et al., 2005). In populations with sta-
ble morph frequencies, nonrandom mating, in concert with other
selective forces, can prevent the loss of color morphs through
within-morph mating (i.e., assortative mating) (Pryke and Griffith
2007) or can promote morph maintenance through disassortative
mating that maintains high heterozygosity and genetic variation
within a population (Hedrick et al., 2016; Maisonneuve et al.,
The presence of two different alleles at a locus (i.e., het-
erozygosity) provides a basis for phenotypic variation within
populations, for example, by expressing alternative color morphs.
If heterozygote individuals have a fitness advantage over the
homozygote ones, the persistence of phenotypic polymorphism
and genetic variability can be aided through heterozygote ad-
vantage (Fisher 1922, 1930; Hedrick 2012). Despite decades of
research, the majority of studies have focused on heterozygote
advantage as a phenomenon of disease resistance, especially
in humans (e.g., the sickle cell anemia, Allison 1954; AIDS,
Carrington et al. 1999), in the environment (e.g., pesticide
resistance, Greaves et al. 1977; infection resistance, Frelinger,
1972), or to maximize fecundity in livestock (Gemmell and Slate
2006). Recently, due to the advantages of the modern molecular
biological methods, there is an increasing number of studies
focusing on the role of heterozygote advantage in color poly-
morphic wild populations (Krüger et al., 2001; Coulson et al.,
2011; Hedrick et al., 2014; Llaurens et al., 2017; Strickland
et al., 2021). Heterozygote advantage is not an easy task to study
in wild populations. The challenges lie in gathering life-history
traits of the different genotypes and, sometimes, the lack of
knowledge of the genetic basis of the polymorphic trait.
Compelling examples of the fitness advantage of heterozy-
gote individuals are phenotypic variability of sexually selected
traits (Coulson et al., 2011; Krüger et al., 2001; Johnston et al.,
2013; Hedrick et al., 2014; Maisonneuve et al., 2021), concur-
rently with other selective forces. In the common buzzard (Buteo
buteo), the plumage color polymorphism is maintained through
heterozygote advantage, which counterbalances maladaptive as-
sortative mate choice due to maternal sexual imprinting (Krüger
et al., 2001). The color coat of wolves in Yellowstone National
Park represents another well-known example, whose stable color
polymorphism maintenance is due to heterozygote advantage
(Coulson et al., 2011; Hedrick et al., 2014) coupled with weak
selection (Hedrick et al., 2014) and a strong contribution of disas-
sortative mating (Hedrick et al., 2016). Complex polymorphisms
can thus be maintained by the interplay of multiple selective pres-
sures, of which heterozygote advantage is one vastly understud-
ied mechanism, and which altogether may determine phenotype-
specific advantages culminating in the coexistence of multiple
The wood tiger moth (Arctia plantaginis)representsa
compelling study species to investigate how different selec-
tive pressures can act on a single color locus and maintain
within-population trait variation. In this system, male hindwing
coloration is determined by a simple genetic basis (Suomalainen
1938; Nokelainen et al., 2022b; Brien et al., 2022): a one locus-
two allele polymorphism (dominant W allele and recessive y
allele), which translates into white (genotype: WW, Wy) and
yellow (genotype: yy) males. Because this is an aposematic
moth species, the color trait is not only used for intraspecific
communication (i.e., sexual selection) but also to advertise their
unpalatability to predators (i.e., interspecific communication).
Previous studies have indeed shown that multiple selective
pressures act on the male coloration. The two male morphs are
differently protected against predators (Nokelainen et al., 2014;
Rojas et al., 2017; Winters et al., 2021), with yellow males gener-
ally having higher survival (Nokelainen et al., 2012; Rojas et al.,
BALANCING SELECTION ON THE COLOR LOCUS
Figure 1. Examples of wood tiger moth hindwing coloration in Finland. Female’s (top row) hindwing coloration varies from light to dark
red, whereas male’s (bottom row) coloration is either white (left and middle photo) or yellow (right photo). White males can either be
WW or Wy for the dominant W color allele, whereas yellow males are homozygote for the y recessive allele. Photos: Chiara De Pasqual.
2017). In addition, male morph mating advantage is dependent
on the morph frequency (Gordon et al., 2015) and males that
origin from “mixed-morph lines” have higher mating success
compared to the moths that originated from more monomorphic
lines (Gordon et al., 2018), which suggests that heterozygote
advantage may also contribute to the color polymorphism in this
species. Here, we test the hypothesis that heterozygote advantage
is contributing to male hindwing color polymorphism in the
wood tiger moth. By using genotyped lines of moths reared in a
greenhouse and life-history traits collected across 19 generations
(i.e., 7 years), we subjected the three color genotypes (WW,
Wy, and yy) to multiple tests. We test whether heterozygote
individuals have (1) higher mating success, either through higher
probability of copulating (copulation observations) or lower
probability of unsuccessful matings; (2) higher reproductive
output by testing fecundity, fertility, and hatching success; and
(3) higher longevity by testing the adults’ life span.
Material and Methods
The wood tiger moth (Arctia plantaginis) (formerly Parasemia
plantaginis; Rönkä et al., 2016) is a polymorphic and aposematic
moth species. The male hindwing coloration is determined by a
simple genetic mechanism where a one locus-two allele (W and
y allele) polymorphism translates into white (WW or Wy geno-
type) or yellow (yy genotype) male morphs (Suomalainen 1938;
Nokelainen et al., 2022b) (Fig. 1). Females do not phenotypi-
cally express the male color alleles as their hindwing coloration
varies continuously from yellow to red but pass the color alleles
to their offspring (Nokelainen et al., 2022b) (Fig. 1). The wood
tiger moth is a capital breeder; it does not feed at the adult stage,
making the larval diet very important for both their development
and the adult stage (e.g., sperm quality, egg numbers) (Tammaru
and Haukioja 1996). Adults only live for 1 or 2 weeks after their
emergence and spend their adulthood looking for suitable mates.
Females lay on average 250 eggs within a few days from the
copulation event. Larvae hatch after about 7 days (Chargé et al.,
2016), and start feeding on a variety of weedy plants (e.g., Plan-
tago sp., Taraxacum sp., and Rumex sp.).
In Finland, the wood tiger moth has one generation per year
and the flight season happens between mid-June and mid-July,
depending on the latitude. It is both a diurnal and crepuscular
species as it flies during daytime hours (Rojas et al., 2015), but
shows mate searching flying activity between ∼5:00 p.m. and
10:00 p.m. with mating activity that can extend into the night
(Nokelainen et al., 2012; Gordon et al. 2015) and a mating peak
in laboratory around sunset (∼10:00 p.m. to 11:00 p.m.) (pers.
obs.). Under laboratory conditions, it can produce up to three gen-
erations per year.
MOTHS REARING AND STOCK MAINTENANCE
The laboratory stock was established in 2013 at the Depart-
ment of Biological and Environmental Science, University of
Jyväskylä (Finland) and new individuals were introduced yearly
to the stock to maintain the genetic variability. During the
stock maintenance, individual females were offered one ran-
domly selected male to ensure offspring’s paternity. Greenhouse
EVOLUTION 2022 3
C. DE PASQUAL ET AL.
temperature roughly followed the outdoor temperature (20–
25°C) and natural light. Individuals were paired in 13 ×7×9cm
(h×w×l) transparent plastic boxes with mesh on the lid. Each
box was provided with a small piece of moistened paper, where
the moths could drink, and to offer a substrate for later oviposi-
tion. Three genotype lines have been established in the stock for
experimental purposes, each one composed by numerous fam-
ilies. To avoid high inbreeding coefficient that could affect the
moths’ survival and the experimental results, controlled matings
are performed in each generation to ensure the most variable
genotype-family combination. The life-history traits (fecundity,
fertility, hatching success, offspring survival, and mating success)
analyzed in this article come from 19 generations (i.e., 7 years) of
data collection. Because mating pairs for the stock maintenance
are not individually observed for successful copulation events,
we followed a subset of these matings to determine whether
heterozygote individuals have higher probability of copulating.
These same individuals were then used to test for the individual’s
longevity. We introduce here the terminology used in the follow-
ing sections; at the precopulatory stage, we use “copulation prob-
ability” to define the likelihood of the paired individuals to cop-
ulate; at the postcopulatory stage, we use “reproductive output”
when referring to fecundity, fertility, and hatching success, and
“mating success” to refer to the likelihood of reproductive fail-
ure. Finally, throughout this work, when referring to “genotype,”
we refer to the sire or dam’s genotype.
PRECOPULATORY STAGE: COPULATION PROBABILITY
AND MATING DELAY
We followed a total of 292 pairs, of which 180 were white (87
WW and 61 Wy genotypes) and 112 were yellow (yy) males.
Among females, 73 were WW, 53 Wy, and 89 yy (see the Sup-
porting Information for the complete crossing scheme). Each
male was paired with a single female. Pairs were set at 4:00 p.m.
and observed until midnight, approximately 1 hour after sunset
when moths were not active anymore. All moths were 1–7 days
old. We considered a mating to be successful if the mating pair
was successfully formed within the 8 hours of observation. Oth-
erwise, we considered it as not successful. We recorded the cop-
ulation success of each pair and the time it took to start mating
(henceforth “mating delay”).
POSTCOPULATORY STAGE: REPRODUCTIVE OUTPUT
To test the reproductive output of the different genotypes, we
compared the fecundity (number of eggs), fertility (number of
hatched larvae), and hatching success over 19 generations (i.e.,
7 years) of life-history trait data collected during routine mainte-
nance of the common garden stock population. Because individ-
uals had been reared in the greenhouse for several generations,
we controlled for the effect of inbreeding coefficient by adding it
as fixed effect and tested its potential interactive effect with the
genotype in the analyses of reproductive traits (see the Supporting
Information for inbreeding coefficient calculation and Table S1).
For each mating pair, the number of laid eggs was counted 4 days
after the female had laid her first egg, and larvae were counted
14 days after the first one had hatched. The hatching success was
calculated as the total number of larvae that hatched divided by
the total number of eggs the female had laid. Larvae were divided
to groups of 30, 14 days after hatching. This counting gives us an
indication of the genotype’s survival. We also tested for genotype
differences in oviposition day and hatching day (i.e., the number
of days it took for each individual to, respectively, lay the first
egg or for the first larvae to hatch). A total of 2714 genotyped in-
dividuals were used for these analyses, of which 1566 were sires
(111 WW, 522 Wy, and 933 yy) and 1148 were dams (150 WW,
351 Wy, and 647 yy).
POSTCOPULATORY STAGE: MATING SUCCESS
Because the life-history trait data collected during stock mainte-
nance mainly take into account successful matings and thus rep-
resent fitness after selection, it is important to separately analyze
those who failed either to mate or produce viable offspring. Be-
cause the lack of offspring also translates in the lack of full known
genotype, we classified individuals either as having a W (either
WW or Wy genotype) or a y (i.e., yy genotype) allele. We identi-
fied three stages of failure: no eggs laid (i.e., no eggs), eggs were
laid but no larvae hatched (i.e., egg hatching), and larvae hatched
but none reached adulthood (i.e., adult eclosion). A total of 1059
matings (out of 2357 set) were considered unsuccessful (44.9%)
with 1568 individuals and 561 pairs included in the analyses.
LONGEVITY OF GENOTYPES
To follow individual longevity but avoid multiple matings, we
removed the male from the mating box at about 1:00 p.m. the day
after the mating and kept them in separated jars to follow their
All analyses were performed in Rstudio (version 1.4.1717) (R
Core Team 2013). The effect of individual full-allele combina-
tions (i.e., genotype) was tested both at the pre- and postcopula-
tory stage. Because several traits showed a general disadvantage
of the yy genotype at the postcopulatory stage, we tested the ef-
fect of the y allele at the pair level. We classified the pairs either
based on the number of y alleles in the pair (henceforth “number
of y allele,” from 0 when both individuals are WW, to 4 when
both are yy) or based on individuals that either had one W allele
or both yy alleles (henceforth “pair type”). This allowed to test,
respectively, for the effect of the y allele regardless of, or consid-
ering, the sex of the moth (see Table S2 for the sample size and
BALANCING SELECTION ON THE COLOR LOCUS
Figure 2. Summary of the effect of the genotype and alleles on the traits included in the analyzes. “No difference” refers to no effect of
the genotype/alleles on the denoted trait. Full-colored moths/dots indicate that the corresponding genotype/alleles plays a role on the
denoted trait, and the size of the moth indicates the higher or lower trait output in genotype comparison. Grayscale moths/dots indicate
no signicant effect of such genotype/allele; however, the size hints at the higher or lower trend on such trait.
spelled-out pair classification, and Fig. 2 for a summary of the
Moth weight and age
Because mate choice and mating success can be affected by
size and age, we tested whether genotype differences existed
among the moths used at the precopulatory (weight and age) and
postcopulatory stage (weight only) by fitting linear models with
either “weight” or “age” as response variables and male or female
genotype as fixed effects using the “lm” function (“stats” package
version 4.1.1). We compared the mean weight and age between
genotypes with F-tests implemented with the “aov” function
(“stats” package) and performed pairwise post hoc comparisons
by estimated least-square means using the “lsmeans” function
(Tukey HSD adjustment; “lsmeans” package version 2.30-0).
Precopulatory stage: Copulation probability and
Copulation event was recorded as a binary variable: 1 if the pair
formed, 0 otherwise. At the individual level, we first tested for
differences in the copulation probability by setting two General-
ized Linear Models (GLMs) (one for males and one for females)
with “copulation probability” as response variable, modeled with
binomial distribution, and genotype, weight, their interaction, and
age as fixed factors. We included the interaction between geno-
type and weight because of significant differences in weight be-
tween genotypes (reported later). We tested the overall effects of
the variables with Chi-square test implemented with the “anova”
We then analyzed the mating delay. Across years, the trials
were performed by using moths reared in the three different
generations, thus carried out in slightly different seasonal time.
Because they mate preferentially 1–2 hours before the sunset
(pers. obs.) and the sunset time is ∼9:30 p.m. in the first and
third generations, and ∼11:00 p.m. during the second generation,
we first tested whether the mating delay (response variable)
was significantly affected by the generation time (fixed factor)
and controlled for the effect of the year (random effect) with
a Cox Proportional Hazard Model (henceforth “Cox model”)
(function “coxph,” “survival” package, version 3.2-11). Be-
cause the mating delay was significantly affected by generation
(χ2=143.14, df =2, P≤2.2 ×10–16) with the second genera-
tion (mean ±SE =334 ±20 min) leading to higher mating delay
compared to the first (mean ±SE =262 ±11 min; estimated
marginal means =–0.726 ±0.111, z=–6.552, P≤0.0001)
and the third (mean ±SE =246 ±11 min) generations (es-
timated marginal means =–0.905 ±0.216, z-ratio =–3.67,
P-value =0.0001), we standardized the mating delay to make it
comparable for later analyses by centering the mean (mean =0
and SD =1). We tested the effect of genotype, weight, age, and
generation (fixed effects) on mating delay (response variable)
with two Cox models: one for males and one for females. The
EVOLUTION 2022 5
C. DE PASQUAL ET AL.
male model included the interactions “genotype x generation,”
“genotype x weight,” and age as fixed effects, whereas the
female model included the interaction genotype by weight, gen-
eration, weight, and age. We did not fit “genotype ×generation”
interactions because we did not test WW females in the second
generation. We then tested for the effect of “number of y allele”
and “pair type” on copulation probability (response variable 1)
and mating delay (response variable 2) by fitting two GLMs
with binomial responses (for response variable 1) and two Cox
models (for response variable 2).
Postcopulatory stage: Reproductive output
To test for differences in the number of eggs, larvae, and hatch-
ing success (response variables), we fit four Generalized Linear
Mixed Models (GLMMs) per response variable: two with Pois-
son distribution and two with negative binomial distribution, of
which two accounted for zero inflated distribution (“glmmTMB”
function from “glmmTMB” package version 1.1.3). We included
genotype, weight, inbreeding coefficient, and two interactions
(“genotype ×weight” and “genotype ×inbreeding coefficient”)
as fixed effects and family as random effect to control for
the effect of relatedness. We standardized both the weight and
the inbreeding coefficient variables (by centering the mean and
SD =1) to include them in the interaction with a discrete vari-
able (the genotype). The model with the lowest AIC value was se-
lected as the best (Table S8, Panel a). For all three response vari-
ables, we used type III analyses of variance to test for the effect of
the interactions on the response variable, and if the effects were
not significant (P>0.05), we removed the interactions from the
final model. Finally, we performed genotype pairwise compar-
isons based on estimated marginal means (“emmeans” function
of the “emmeans” package, version 1.7.2).
In addition, by considering pairs with only one WW and one
yy individual (which ensures Wy offspring), we tested whether
the heterozygote advantage could come from the dam or sire’s
side. We thus tested whether fecundity, fertility, and hatching suc-
cess (three response variables) differed between pairs (fixed fac-
tor) by fitting two GLMs per response variable, one with a Pois-
son and the other with a negative binomial distribution. We chose
the models with negative binomial distribution due to their lower
AIC (Table S3).
Finally, we tested for the effect of genotype, “number of y
allele,” and “pair type” (fixed factors) on the number of days
both to lay eggs (response variable “oviposition day”) and for
the eggs to hatch (response variable “hatching day”) by setting
two GLMMs per response variable, one with Poisson and one
with negative binomial distribution, genotype as fixed factor and
generation as random factor, and four GLMs with the same re-
sponse variables and distribution, but either “number y allele” or
“pair type” as fixed effects. GLMs with Poisson distribution were
chosen because of their lower AICs (Table S4).
Postcopulatory stage: Mating success
To test for differences in mating success between the W and y
allele (fixed factor) and “pair type” (fixed factor), we fit four
GLMMs, one to test for the allele effect regardless of sex, two
models considering moth sex (one for males and one for females),
and the final one for the pair effect. In all the four models, we
determined the probability of successfully mating (response vari-
able) by the count of successful over the unsuccessful matings
through the “cbind” function and modeled with binomial distri-
bution. We set generation as a random effect and used the func-
tion “weights” to specify the total number of matings that were
set per generation. We tested the effect of the “pair type” using
pairwise comparisons based on estimated marginal means (Tukey
Longevity of genotypes
To test whether longevity differed between genotypes, we fit two
Cox models, one for males and one for females, with individuals’
life span (days) as response variable, and genotype as fixed factor.
MOTHS WEIGHT AND AGE
Genotype did not affect male (F(2;1199) =2.567, P=0.077)
or female (F(2;938) =0.246, P=0.782) weight of the indi-
viduals used in the postcopulatory analyses, but did for those
used at the precopulatory stage. In both sexes, WW individ-
uals were significantly heavier than yy individuals (estimated
marginal means; males =14.10 ±3.39, t=4.162, P=0.0001;
females =16.40 ±6.18, t=2.652, P=0.0233), and WW
females were also heavier compared to Wy females (estimated
marginal means =23.69 ±7.03, t=3.371, P=0.0026). Age
did not differ between male (F(2,257) =0.898, P=0.409) or fe-
male genotypes (F(2,212) =1.357, P=0.26).
PRECOPULATORY STAGE: COPULATION PROBABILITY
AND MATING DELAY
Although we found genotype-specific differences in weight, the
copulation probability in either sex was not affected by their
interaction (males: genotype ×weight =χ2(2,250) =5.5438,
P=0.0625; females: genotype ×weight =χ2(2,205) =2.4382,
P=0.2955). Copulation probability was not affected by male
or female genotype, male weight, or male and female age (Ta-
ble S5, Panel a). Interestingly, the heavier the female, the lower
the copulation probability (GLM; Estimate =–0.4776 ±0.1538,
z=–3.106, P=0.0019). The mating delay was significantly
affected by the generation, suggesting that environmental cues
BALANCING SELECTION ON THE COLOR LOCUS
(e.g., the sunset/light) may influence the mating behavior (Ta-
ble S5, Panel b). Males took significantly longer in the second
generation compared to the first and third (coxph; second vs.
first; exp(coef) =2.1212 ±0.2739, z=2.745, P=0.0060;
second vs. third; exp(coef) =2.7969 ±0.2884, z=3.567,
P=0.0004), whereas females took significantly longer only
compared to the third generation (coxph; second vs. third;
exp(coef) =3.0448 ±0.3285, z=3.389, P=0.0007). Be-
sides the effect of the environmental cues, no other traits played
a significant effect on the mating delay. These include the lack
of interaction between male genotype and generation (LR test;
χ2=0.8738, df =2, P=0.6460), the lack of genotype-specific
effect of weight (LR test; male genotype ×weight: χ2=3.1702,
df =2, P=0.2049; female genotype ×weight: χ2=1.5839,
df =2, P=0.4530), and lack of significant effect of genotype,
weight or age, both in males and females (Table S5, Panel b).
Although there was no precopulatory selection at the indi-
vidual level, a closer look at the copulation probability and mat-
ing delay suggests that the allele combination may play an indi-
rect role in these traits, at least for some genotypes. The number
of y alleles in the mating pair significantly affected the copula-
tion probability (χ2=12.996, df =4, P=0.0113), where pairs
with zero y alleles had a higher copulation probability in general,
and significantly higher than pairs with one, two, and three y al-
leles (Table S6). We found, however, no significant effect of the
pair type (χ2=3.6337, df =3, P=0.3038) on the copulation
probability, suggesting a general effect of the allele combinations
on the mating success rather than sex-specific contribution. For
the mating delay, we found somewhat the opposite pattern, as
it did not differ according to the number of y alleles (LR test;
χ2=2.26, df =4, P=0.6872) but the “femyy +maleW-allele”
pair type mated significantly faster than all the other pair types
(Table S7). Genotype-specific advantages might be relative to the
mating partner and thus can arise at the pair level.
POSTCOPULATORY STAGE: REPRODUCTIVE OUTPUT
For the six final models selected, the lowest AICs were given
by the zero inflated with negative binomial models (Table S8,
Panel a). Neither interactions (“genotype ×weight” and “geno-
type ×inbreeding coefficient”) were significant and were ex-
cluded from the final models (Table S8, Panels b–d). This sug-
gests that weight and inbreeding coefficient did not affect the
reproductive output in a genotype-specific manner, despite, for
example, genotype-specific differences in the inbreeding coef-
ficient. Although the genotype did not explain the mean differ-
ences in fecundity, fertility, and hatching success (Table S9 [Pan-
els b and d], Table S10 [Panel b], Table S11 [Panel b], and Table
1 [Panels a and c]), it had a strong effect on the probability of
reproductive failure. This suggests that genotypes differ in their
likelihood of reproductive failure rather than the number of eggs,
larvae, or proportion of eggs hatched.
Genotype, female weight, and inbreeding coefficient had a
significant effect on the fecundity trait (Table S9, Panel a). yy
males had significantly fewer eggs (mean ±SE =149.2 ±3.7)
compared to WW (mean ±SE =167.9 ±8.3) and Wy
(mean ±SE =171.6 ±4.9) males (Table S9, Panel c;
Fig. 3a). yy females laid a significantly lower number of
eggs (mean ±SE =161 ±4.2) compared to Wy females
(mean ±SE =202.3 ±5) but not compared to WW females
(mean ±SE =166.2 ±7.8) (Table S9, Panels d and e; Fig. 2d).
The yy genotype disadvantage was due to both a lower egg count
and a higher probability of failing to have eggs at all, both in
males and females (Table S9, Panels c and e). Weight had a sig-
nificant effect in females (Table S9, Panel a) with the heavier
the female, the higher the number of eggs laid (Table S9, Panel
e), whereas no significant effect of the weight was detected for
males (Table S9, Panel a). Weight had a significant effect on the
count (number) of eggs laid but did not affect the probability of
zero count (Table S9, Panel e). No interaction between inbreeding
coefficient and genotype was detected but the inbreeding coeffi-
cient had a significant effect on the number of eggs laid (Table
S9, Panel a), with the higher its value, the lower the egg count
(Table S9, Panels c and e). Interestingly, this did not affect the
probability of having zero eggs (Table S9, Panels c and e).
Wy females had a significantly lower probability of egg
hatching (i.e., having larvae) failure (Table 1, Panel b; Fig. 3e).
This was not repeated in males, as yy males had lower prob-
ability of having larvae than WW and Wy males (Table S10,
Panel b; Fig. 3b). The significant differences were in the proba-
bilities of failure (zeroes) and not in the number (count) of lar-
vae. Therefore, the female Wy advantage is due to the signif-
icantly lower probability in failing to have larvae at all com-
pared to the other two genotypes (Table 1, Panel b). The ef-
fect of female weight on fertility was significant (Table S10,
Panel a) with the heavier the female, the higher the number
of larvae that hatched (Table 1, Panel b). This was not seen
in males (Table S10, Panel c). The inbreeding coefficient had
a significant effect on the fertility trait (Table S10 [Panel c]
and Table 1 [Panel b]), where the higher the inbreeding coef-
ficient, the lower the number (count) of larvae in males only
(Table S10, Panel c) but not in females (Table 1, Panel b).
In addition, the higher the inbreeding coefficient, the higher the
probability of zero larva both in males and in females (Table S10
[Panel c] and Table 1 [Panel b]).
The hatching success was significantly affected by the in-
dividual genotype (Table S11 [Panel c] and Table 1 [Panel
d]), with Wy females having a higher likelihood of hatch-
ing success compared to the other two genotypes (WW
mean ±SE =0.46 ±0.03, Wy mean ±SE =0.67 ±0.02, yy
EVOLUTION 2022 7
C. DE PASQUAL ET AL.
Figure 3. The graph illustrates differences in the fecundity, fertility, and hatching success between genotypes, in males (top row) and
females (bottom row). Statistically signicant differences are marked with asterisks.
mean ±SE =0.47 ±0.02; Table 1, Panel d; Fig. 3f). In males,
the yy genotype had a lower likelihood of hatching success than
the other two genotypes (WW mean ±SE =0.46 ±0.04, Wy
mean ±SE =0.53 ±0.02, yy mean ±SE =0.40 ±0.01;
Table S11, Panel c, Fig. 3c). We found, therefore, strong fe-
male heterozygote advantage in fertility and hatching success ex-
pressed in their higher likelihood of having larvae and higher
likelihood of hatching success. Weight had a significant effect
in males but not in females (Table S11, Panel a). Interestingly,
the heavier the male, the lower the probability of hatching suc-
cess (Table S11,Panel c). The inbreeding coefficient significantly
affected males and females (Table S11 [Panel c] and Table 1
[Panel d]) with lower probability of hatching success as its value
Finally, the Wy advantage does not seem to be due to ei-
ther maternal or paternal effect. The number of eggs (glm.nb; es-
timate =0.031 ±0.127, z=0.24, P=0.81), larvae (glm.nb;
estimate =–0.066 ±0.175, z=–0.377, P=0.706), or the
hatching success (glm.nb; estimate =0.164 ±0.288, z=0.589,
P=0.556) did not differ between pairs where either the dam or
the sire was WW and the other yy. This suggests that the higher
Wy fitness is due to the allele combination (W and y) per se,
rather than being determined by the dam or sire’s side. We found
no differences based on the individual genotype or due to the ef-
fect of the pair for the oviposition day and hatching day (Table
S4) suggesting no particular effect of the color locus on these
POSTCOPULATORY STAGE: MATING SUCCESS
With 78% of the unsuccessful matings having eggs and larvae,
the mating failure is more likely to take place at the postcopu-
latory rather than precopulatory stage. The most sensitive stage
seems to be the egg-hatching stage (62%), which was signifi-
cantly higher than matings that had no eggs (23%; χ2=17.89,
df =1, P=2.335 ×10–5) and than matings that had no adult
eclosing (15%; χ2=28.69, df =1, P=8.502 ×10–8). No dif-
ferences were found between the no-egg and adult-eclosing stage
(χ2=1.68, df =1, P=0.19). There was no effect of either
the sire or the dam at the different stage levels (no-eggs stage,
χ2=0.56, df =1, P=0.46; egg-hatching stage, χ2=0.072,
df =1, P=0.79; adult-eclosing stage, χ2=0.13, df =1,
P=0.72), suggesting no sex-specific cause of failure. Y-allele
individuals (i.e., yy genotype) had a significantly higher proba-
bility of failing to have offspring than W allele individuals (W
vs. y; estimate =–0.075 ±0.007, z=–10.475, P≤2×10–16).
These results were likely influenced by females, as y allele fe-
males failed significantly more than W allele females (W vs. y
females; estimate =–0.272 ±0.020, z=–13.38, P≤2×10–16),
whereas y allele males had significantly higher probability of suc-
ceeding in having offspring compared to W allele males (W vs.
y males; 0.044 ±0.011, z=4.014, P=5.97 ×10–5). The gen-
eration effect accounted in average for 15% of the variation in
the probability of failing (16% in females and 13% in males).
At the pair level, “female yy +male yy” and “female yy +
male W-allele” pair types had the lowest probability of having
BALANCING SELECTION ON THE COLOR LOCUS
Tab l e 1 . Panels (a) and (c) report female genotype pairwise comparisons for the fertility and hatching success traits. Panels (b) and (d)
report the GLMM output for, respectively, the fertility and hatching success trait in females.
(a) Pairwise comparisons based on estimated marginal means; Tukey HSD adjustment
Contrast Estimate SE df t P
Wy-WW 0.074 0.082 925 0.894 0.644
Wy- yy 0.035 0.052 925 0.674 0.779
WW-yy –0.039 0.078 925 –0.495 0.874
(b) Zero inflated; Intercept =Wy genotype
Estimate SE zP
Intercept 5.054 0.040 125.31 <2×10–16
WW genotype –0.073 0.082 –0.89 0.371
yy genotype –0.035 0.052 –0.67 0.500
Weight 0.217 0.026 8.47 <2×10–16
Inbreeding coefficient –0.010 0.027 –0.39 0.699
Zero inflated model
Intercept –2.456 0.229 –10.721 <2×10–16
WW genotype 1.225 0.310 3.946 7.95 ×10–5
yy genotype 1.566 0.249 6.293 3.12 ×10–10
Weight –0.026 0.087 –0.300 0.764
Inbreeding coefficient 0.432 0.092 4.682 2.84 ×10–6
(c) Pairwise comparisons based on estimated marginal means; Tukey HSD adjustment
Contrast Estimate SE df t P
Wy-WW 0.089 0.068 834 1.324 0.382
Wy- yy 0.041 0.044 834 0.933 0.619
WW-yy –0.048 0.064 834 –0.756 0.730
(d) Zero inflated; Intercept =Wy genotype
Estimate Std. Error zP
Intercept 4.284 0.034 126.06 <2×10–16
WW genotype –0.089 0.068 –1.32 0.185
yy genotype –0.041 0.044 –0.93 0.351
Weight 0.0004 0.021 0.02 0.986
Inbreeding coefficient 0.005 0.023 0.23 0.817
Zero inflated model
Intercept –2.350 0.230 –10.215 <2×10–16
WW genotype 1.210 0.312 3.876 0.0001
yy genotype 1.559 0.251 6.216 5.12 ×10–10
Weight –0.005 0.089 –0.060 0.952
Inbreeding coefficient 0.382 0.095 4.003 6.25 ×10–5
offspring (estimated marginal means =0.0527 ±0.0426,
z=1.239, P=0.6023), whereas the probability of failing sig-
nificantly differed between all the other pair comparisons (P-
values <0.05) (Table S12; Fig. 4).
LONGEVITY OF GENOTYPES
All males’ genotypes had similar life spans after mating once (LR
test; males; χ2=0.297, df =2, P=0.862), whereas WW females
lived significantly longer than the other two female genotypes
(coxph; WW vs. Wy; exp(coeff) =2.2614 ±0.3653, z=2.234,
P=0.0255; WW vs. yy; exp(coeff) =1.99 ±0.3192, z=2.156,
P=0.0311). For a summary of these results, see Figure 2.
We investigated the effect of color alleles and genotypes from
pre- to postcopulatory stage in maintaining warning color poly-
morphism within wood tiger moth populations. Carrying one or
EVOLUTION 2022 9
C. DE PASQUAL ET AL.
Figure 4. Coefcient estimates of the probability of pair type’s mating success. Except for the nonsignicant difference between “female
yy +male yy” and “female yy +male W-allele,” all the other pairwise comparisons were signicantly different (see Table S12 for a
two copies of the yellow allele affected the reproductive fitness
in a stage-specific way, from higher likelihood of reproductive
output when females carry one copy of the allele (i.e., heterozy-
gote advantage), to lower likelihood of reproductive output suc-
cess and lower mating success when individuals carry two copies.
Thus, the yellow allele might have a pleiotropic effect on sev-
eral life-history traits that can contribute to the maintenance of
polymorphism in male coloration. Although we found little con-
tribution of male genotype across the reproductive sequence, fe-
male genotype had a significant effect, especially for reproduc-
tive success, and likely therefore contributes to the persistence of
polymorphism in male coloration. Although all the genotypes, re-
gardless of the sex, had an equal copulation probability and mat-
ing delay, Wy females had higher reproductive output (fertility
and hatching success) and thus higher offspring survival. Pairs
with yy females had shorter mating delay and were more likely
to fail in having any offspring. The presence of the yellow allele
affected the fitness both at the individual and pair level, such as a
lower reproductive output in males, and across different steps of
the reproductive process for females. Our results thus show the
role of genotype-dependent female reproductive success in main-
taining male hindwing coloration. Overall, these results suggest
that the color locus is pleiotropic with a number of life-history
traits, allowing for the maintenance of within-species phenotypic
WEAK EFFECT OF PRECOPULATORY SELECTION
At the precopulatory stage, 43% of paired individuals did not
copulate suggesting some form of female or male rejection. The
lack of copulation probability and mating delay differences be-
tween genotypes suggests that precopulatory selection may be a
weak selective force on the genotypes and, at this stage, neither
males nor females can avoid mating with partners with lower
fitness prospects. These results are in accordance with a previ-
ous study with a similar mating experiment setup that showed
equal mating probability between white and yellow phenotypes
(Chargé et al., 2016). The hypothesis that sexual selection is more
likely to take place after the copulation event, rather than result-
ing from precopulatory selection, may be further supported by
the low (23%) percentage of failed matings that did not have
10 EVOLUTION 2022
BALANCING SELECTION ON THE COLOR LOCUS
eggs, a proxy for the lack of copulation event. However, we can-
not exclude that the lack of differences in the copulation prob-
ability may have been masked by a trade-off between securing
at least one mating (and therefore some offspring) and exerting
mate choice (see Kokko and Mappes 2005).
Females of different species have been shown to ex-
ert stronger sexual selection when presented with a choice
(Dougherty and Shuker, 2015). Virgin females, due to the un-
certainty of finding a second mate and the risk of dying un-
mated, are expected to be less choosy and may accept to mat-
ing randomly if they fear no further male will be encountered
(Kokko and Mappes, 2005; Dougherty and Shuker, 2015). In
addition, individuals may get choosier in later matings (Kokko
and Mappes 2005; Gao et al., 2020), which might explain the
lack of differences in copulation probability. This explanation
may also be supported by the lack of differences in the mat-
ing delay; if any choice were to be made based on some trait, it
might have been translated into a different mating delay. Instead,
the mating delay was higher in the second generation because
this species is mostly sexually active around sunset (pers. obs.),
which is about 2 hours later than the first and third generations
(∼11:30 p.m. vs. ∼9:30 p.m.). Other studies on the species have
shown that differences in male copulation probability, and par-
ticularly the white male advantage, may be condition dependent
(stress-induced condition; Nokelainen et al., 2012), due to the ef-
fect of white mixed-lineage advantage (more heterozygous indi-
viduals; Gordon et al., 2018), or context dependent, in which the
most common morph has higher mating success (Gordon et al.,
2015). Mating differences, or lack thereof, in the wood tiger moth
may be, therefore, determined by the ecological context or be
based on a different trait (e.g., the sex pheromone).
HETEROZYGOTE ADVANTAGE FOR THE
MAINTENANCE OF COLOR POLYMORPHISM
At the postcopulatory stage, we found a significant effect of the
genotype on fecundity, fertility, and hatching success. In particu-
lar, heterozygote (Wy) females had higher likelihood of fertility,
offspring survival, and hatching success than the other two geno-
types, suggesting that male hindwing coloration is maintained
by a rather strong heterozygote advantage effect. The Wy advan-
tage does not seem to be due to either dam or sire’s effect (i.e.,
Ww ×yy pairs do not show differences in their reproductive out-
put) or due to differences in oviposition or hatching strategies,
suggesting that the heterozygote advantage is a consequence of
the W and y allele combination. Wy females had, indeed, a sig-
nificantly lower probability of zero fertility, which translated into
higher hatching success than both the homozygotes. Our results
add to a few other known cases of heterozygote advantage (Buteo
buteo, Krüger et al., 2001; wolves, Hedrick et al., 2014; Helico-
nius numata, Jay et al., 2021). The advantage of the dominant (W)
allele in our species does not appear to change for fitness-related
measures supported by the general advantage of Wy (and WW
genotype) and over the general disadvantage of the yy genotype
throughout the reproductive output, a pattern somewhat oppo-
site to the wolf of the Yellowstone National Park (Coulson et al.,
2011; Hedrick et al., 2014). In contrast, the heterozygosity advan-
tage in the wood tiger moth may be context dependent: in mating
probability either due to female choice or intrasexual competition
(Gordon et al., 2018), in the reproductive output (this study), or as
defense against predators (Winters et al., 2021), which suggests
the importance of considering both natural and sexual selective
PLEIOTROPIC EFFECT OF THE YELLOW ALLELE
The presence of one or two copies of the yellow allele affected
several steps of the reproductive sequence, from copulatory prob-
ability, to mating delay, reproductive output, and mating success,
especially in females. Females carrying one yellow allele (i.e.,
Wy) had higher reproductive output than the other two genotypes
(i.e., heterozygote advantage). The W and y allele combination
might therefore lead to a genetic compatibility advantage that
give rise to increased offspring survival, and higher likelihood of
hatched eggs (i.e., hatching success). Bearing two copies of the
yellow allele affected other traits of the reproductive sequence,
such as copulation probability, mating delay, and mating success
of pairs with yy females. About 55% of the pairs with yy females
copulated (against, e.g., 80% of WW ×WW pairs), whereas pairs
with yy females and white (i.e., WW or Wy) males copulated
faster than the other pair types. WW and Wy males have higher
reproductive output than yy males, which might be a reason why
yy females are more willing to accept white males compared to
the yy males. Females carrying two copies of the yellow allele
had a higher likelihood of reproductive failure regardless of the
male they mated with. Mating with a yy female may thus be par-
ticularly costly to males.
We thus suggest that genotypic differences in life-history
traits are likely due to pleiotropic effects of the yellow allele,
especially because these effects are expressed in females with-
out the yellow phenotype. The pleiotropic effect of the yellow
allele extends to males as well. The male yellow coloration con-
fers better protection against predators (Nokelainen et al., 2012,
2014; Rojas et al., 2017), but there are trade-offs with the mat-
ing probability (Nokelainen et al., 2012), the reproductive output
(Gordon et al., 2018; this study), and their ability to disperse
(Gordon et al., unpublished). Recent examples of the pleiotropic
effect of color loci on life-history traits have been found in the
warningly colored seed bug (Lygaeus simulans) (Balfour et al.,
2018) and Heliconius numata (Jay et al., 2021).
The male coloration in the wood tiger moth is likely reg-
ulated by a yellow-family gene (Brien et al., 2022), which is
EVOLUTION 2022 11
C. DE PASQUAL ET AL.
conserved across insects (Ferguson et al., 2011) and has well-
known functions in the melanin production pathway (Wittkopp
et al., 2002). Yellow geneshavealsobeenshowntohave
pleiotropic effects on life-history and behavioral traits (Bastock,
1956; Massey et al., 2019; Connahs et al., 2021). Loss of the
yellow gene function in D. melanogaster results in reduced mat-
ing success due to changes in the courtship behavior (Bastock,
1956) and to structural changes in the sex combs used to grasp
the female (Massey et al., 2019). The yellow gene has the oppo-
site effect in Bicyclus anynana where its expression needs to be
suppressed for the males to properly express courtship behavior
(Connahs et al., 2021). Thus, we suspect that the yellow locus
is influencing life-history traits as well as wing coloration also in
the wood tiger moth, although the exact genetic mechanism is yet
GENERAL IMPLICATIONS FOR THE MAINTENANCE
OF THE COLOR POLYMORPHISM
Across its distribution, the wood tiger moth shows a striking
level of phenotypic diversity, both across and within populations
(Hegna et al., 2015). From our results, the overall advantage
of individuals bearing at least one dominant W allele, and the
pleiotropic effect of the yellow allele, could theoretically explain
populations that are naturally W male biased, such as the Finnish
population. However, the 2:1 (white:yellow) ratio expected by the
dominant W allele advantage is hardly found in natural popula-
tions, even in the light of the higher likelihood of y-bearing in-
dividuals to show disadvantage along the reproductive sequence.
This suggests that other mechanisms and selective forces are at
play. The extensive literature on this study system shows indeed
that male morphs experience a multitude of morph-specific se-
lective pressures, from predation (Nokelainen et al., 2012, 2014;
Rojas et al., 2017, Winters et al., 2021) linked also to light envi-
ronment (Nokelainen et al., 2022a), to immune response (Noke-
lainen et al., 2013), and density-dependent effects (Gordon et al.,
2015). This likely affects the expected ratio of white and yel-
low morphs in natural populations. Future quantifications of the
genotype frequencies of natural populations will shed more light
on the mechanisms maintaining both alleles.
PRE- AND POSTCOPULATORY EFFECT OF TRAITS
BEYOND COLOR GENOTYPE
Mate choice and mating probability may also be based on size
or age. However, no effect of age in either sex or male weight
was found to affect the copulation probability and mating de-
lay despite white males being heavier than yellow males (and in
general, WW individuals being heavier than the other two geno-
types). The lack of age and weight effect could be due to the lack
of mate choice or adaptation to lab conditions. Although female
weight did not play a role in mating delay, it was interesting to
notice that the heavier the female, the lower her copulation proba-
bility. A similar result was found in azure damselflies males (Co-
enagrion puella) in which lighter males have higher mating suc-
cess (Banks and Thompson, 1985). Banks and Thompson (1985)
put forward the hypothesis that heavier males may be less ac-
tive due to their bigger size, thus less likely to find a female. As
in our experiment, mating trials were carried out in a confined
space and females were more easily spotted by males than in a
natural scenario, the lower copulation probability of heavier fe-
males may be due to between-females behavioral differences. For
instance, heavier wood tiger moth females may be more prone to
actively reject males than lighter females. Although this hypoth-
esis should be properly tested, it has been already shown that in
Lepidoptera male harassment can be costly to females (Merrill
et al., 2018) and females actively reject males to the point they
can override male preference (Chouteau et al., 2017). This be-
havioral hypothesis is also in line with the lowest yy female mate
acceptance toward males carrying the W allele that lacks the dele-
terious elements associated with the y allele when expressed in
homozygote yy males (at least for the reproductive success).
At the reproductive output stage, female, and not male,
weight played a significant role in fecundity and fertility, de-
spite larger males produce bigger spermatophores (Chargé et al.,
2016). Not surprisingly, heavier females laid more eggs. This is
in accordance with Santostefano et al. (2018) and it is expected
because this species is a capital breeder and females are born
with all the eggs that can be potentially fertilized (Tammaru and
Haukioja, 1996). Heavier females also had higher fertility. There-
fore, female weight may be a trait that males could select for.
It is interesting to notice that heavy males had lower hatching
success. A previous study on the wood tiger moth (Santostefano
et al., 2018) showed that mating with heavier males led to a lower
number of eggs laid. Because heavier males produce bigger sper-
matophores (Chargé et al., 2016), a negative correlation between
male weight and hatching success may be the result of trade-
offs; being heavy and therefore having invested more resources
into mass development may trade-off with the quality of the sper-
matophore, or heavier males may spend more energy than lighter
males in finding and/or courting a female, therefore lowering the
resources available for spermatophore production.
Altogether these results suggest that wood tiger moth male col-
oration is maintained through stage-specific color allele and
genotype advantages across the reproductive sequence, from cop-
ulation probability to offspring survival. Although individuals
do not seem to avoid mating with partners with lower fitness
prospects, the strong female heterozygote advantage in fertility,
12 EVOLUTION 2022
BALANCING SELECTION ON THE COLOR LOCUS
hatching success, and offspring survival offers a powerful mech-
anism for both alleles to be maintained within the population.
Male hindwing coloration seems also to be maintained through
the pleiotropic effect of the yellow allele, which affects specific
traits of the reproductive sequence, from shortening the mating
delay, to being correlated with higher reproductive failure and
in general, with the reproductive output. In nature, populations
are typically exposed to complex ecological interactions, multi-
ple mechanisms, and selective forces. Such multiple mechanisms
concurrently interact and allow for life-history trait variability
maintenance through pleiotropy (this study, Mérot et al., 2020)
and thus maintain complex color polymorphisms even in the sit-
uation when selection is positively frequency dependent (Gordon
et al., 2015; Chouteau et al., 2016).
CDP collected the precopulatory stage data, analyzed the data, and wrote
this article. KS and JK created the genotype lines, reared numerous gen-
erations of moths, and collected the life-history traits data. SG initi-
ated data analyses. TK helped with the inbreeding coefficient analyses
and contributed to data analysis. JM conceptualized and coordinated the
work. SG, TK, and JM contributed to the writing of the article. All au-
thors read and approved the final version of the manuscript.
This article is the result of the work of numerous greenhouse workers
and students who, in the past decade, have patiently mated thousands
of moths and collected very valuable long-term data. We also thank T.
Tuomaala and T. Salmi for assistance with the mating experiment, J.
Valkonen for assistance with stats, and F. Guillaume, M. Brien, E. Koch,
and two anonymous reviewers for comments on a previous version of
the manuscript. This work was supported by the Academy of Finland
(project no. 345091 to JM).
CONFLICT OF INTEREST
The authors declare no conflict of interest.
The data used in this study are accessible from the following repository:
https://doi.org/10.5061/dryad.g1jwstqth. The data will be released after
Abalos, J., Pérez i de Lanuza, G., Carazo, P. & Font, E. (2016) The role of
male coloration in the outcome of staged contests in the European com-
mon wall lizard (Podarcis muralis). Behaviour, 153, 607–631.
Allison A.C. (1954) Protection afforded by sickle-cell trait against subtertian
malarial infection. Br.Med.J., 1, 290–294.
Andrade, P., Pinho, C., Pérez i de Lanuza, G., Afonso, S., Brejcha, J., Rubin,
C.J., Wallerman, O., Pereira, P., Sabatino, S.J. & Bellati, A. (2019) Reg-
ulatory changes in pterin and carotenoid genes underlie balanced color
polymorphisms in the wall lizard. Proc. Natl. Acad. Sci. USA, 116,
Balfour, V.L., Aumont, C., Dougherty, L.R. & Shuker, D.M. (2018) The fit-
ness effects of a pale mutant in the aposematic seed bug Lygaeus sim-
ulans indicate pleiotropy between warning coloration and life history.
Ecol. Evol., 8, 12855–12866.
Banks, M.J. & Thompson, D.J. (1985) Lifetime mating success in the dam-
selfly Coenagrion puella.Anim. Behav., 33, 1175–1183.
Bastock & M. (1956) A gene mutation which changes a behavior pattern.
Evolution, 10, 421–439.
Brien, M.N., Orteu, A., Yen, E.C., Galarza, J.A., Kirvesoja, J., Pakkanen, H.,
Wakamatsu, K., Jiggins, C.D. & Mappes, J. (2022) Colour polymor-
phism associated with a gene duplication in male wood tiger moths.
Cacoyianni, Z., Kovacs, I.V. & Hoffmann, A.A. (1995) Laboratory adap-
tation and inbreeding in Helicoverpa punctigera (Lepidoptera, Noctu-
idae). Australian journal of zoology, 43(1), 83–90.
Calsbeek, B., Hasselquist, D. & Clobert, J. (2010) Multivariate phenotypes
and the potential for alternative phenotypic optima in wall lizard (Po-
darcis muralis) ventral colour morphs. Journal of evolutionary biology,
Cardé, R.T. & Baker, T.C. (1984) Sexual communication with pheromones.
In Chemical ecology of insects (ed. W. J. Bell & R. T. Cardé), pp. 355–
383. Chapman & Hall, Lond.
Carrington, M., Nelson, G.W., Martin, M.P., Kissner, T., Vlahov, D., Goedert,
J.J., Kaslow, R., Bushbinder, S., Hoots, K. & O’Brien, S.J. (1999) HLA
and HIV-1: heterozygote advantage and B∗35-Cw∗04 disadvantage.
Science, 283(5408), 1748–1752.
Chargé, R., Wedell, N., Lindstedt, C., Hämäläinen, L., Övermark, E. &
Mappes, J. (2016) Variation in male fertility in a polymorphic moth,
Parasemia plantaginis.Animal Behaviour, 111, 33–40.
Chouteau, M., Arias, M. & Joron, M. (2016) Warning signals are under pos-
itive frequency-dependent selection in nature. Proceedings of the na-
tional Academy of Sciences, 113(8), 2164–2169.
Chouteau, M., Llaurens, V., Piron-Prunier, F. & Joron, M. (2017) Poly-
morphism at a mimicry supergene maintained by opposing frequency-
dependent selection pressures. Proceedings of the National Academy of
Sciences, 114(31), 8325–8329.
Connahs, H., Tan, E., Dion, E., Ter, Y., Matsuoka, Y., Bear, A. & Monteiro, A.
(2021) Developmental plasticity in male courtship in Bicyclus anynana
butterflies is driven by hormone regulation of the yellow gene. bioRxiv,
Coulson, T., MacNulty, D.R., Stahler, D.R., VonHoldt, B., Wayne, R.K.
& Smith, D.W. (2011) Modeling effects of environmental change on
wolf population dynamics, trait evolution, and life history. Science,
Cuthill, I.C., Allen, W.L., Arbuckle, K., Caspers, B., Chaplin, G., Hauber,
M.E., Hill, G.E., Jablonski, N.G., Jiggins, C.D., Kelber, A., Mappes, J.,
Marshall, J., Merrill, R., Osorio, D., Prum, R., Roberts, N.W., Roulin,
A., Rowland, H.M., Sherratt, T.N., Skelhorn, J., Speed, M.P., Stevens,
M., Stoddard, M.C., Staurt-Fox, D., Talas, L., Tibbetts, E. & Caro, T.
(2017) The biology of color. Science, 357(6350), eaan0221.
Dougherty, L.R. & Shuker, D.M. (2015) The effect of experimental design on
the measurement of mate choice: a meta-analysis. Behavioral Ecology,
Duarte, R.C., Flores, A.A. & Stevens, M. (2017) Camouflage through
colour change: mechanisms, adaptive value and ecological significance.
Philosophical Transactions of the Royal Society B: Biological Sciences,
Endler, J.A. (1988) Frequency-dependent predation, crypsis and aposematic
coloration. Philosophical Transactions of the Royal Society of London.
B, Biological Sciences, 319(1196), 505–523.
Endler, J.A. & Mappes, J. (2017) The current and future state of animal col-
oration research. Philosophical Transactions of the Royal Society B: Bi-
ological Sciences, 372(1724), 20160352.
EVOLUTION 2022 13
C. DE PASQUAL ET AL.
Ferguson, L.C., Green, J., Surridge, A. & Jiggins, C.D. (2011) Evolution of
the insect yellow gene family. Molecular biology and evolution, 28(1),
Fisher, R.A. (1922) XXI.— On the dominance ratio. Proceedings of the royal
society of Edinburgh, 42, 321–341.
———. (1930) The evolution of dominance in certain polymorphic species.
The American Naturalist, 64(694), 385–406.
Ford, E.B. (1945) Polymorphism. Biological Reviews, 20(2), 73–88.
Frelinger, J.A. (1972) The maintenance of transferrin polymorphism in pi-
geons. Proceedings of the National Academy of Sciences, 69(2), 326–
Galeotti, P., Pellitteri-Rosa, D., Sacchi, R., Gentilli, A., Pupin, F., Rubolini,
D. & Fasola, M. (2010) Sex-, morph-and size-specific susceptibility to
stress measured by haematological variables in captive common wall
lizard Podarcis muralis.Comparative Biochemistry and Physiology part
A: Molecular & Integrative Physiology, 157(4), 354–363.
Galeotti, P., Sacchi, R., Pellitteri-Rosa, D., Bellati, A., Cocca, W., Gentilli,
A., Scali, S. & Fasola, M. (2013) Colour polymorphism and alternative
breeding strategies: effects of parent’s colour morph on fitness traits in
the common wall lizard. Evolutionary Biology, 40(3), 385–394.
Gao, K., van Wijk, M., Clement, Z., Egas, M. & Groot, A.T. (2020) A life-
history perspective on sexual selection in a polygamous species. BMC
evolutionary biology, 20, 1–10.
Gemmell, N.J. & Slate, J. (2006) Heterozygote advantage for fecundity. PLoS
One, 1(1), e125.
Gordon, S.P., Kokko, H., Rojas, B., Nokelainen, O. & Mappes, J. (2015)
Colour polymorphism torn apart by opposing positive frequency-
dependent selection, yet maintained in space. Journal of Animal Ecol-
ogy, 84(6), 1555–1564.
Gordon, S.P., Burdillat, S. & Mappes, J. (2018) Phenotype-dependent mate
choice and the influence of mixed-morph lineage on the reproductive
success of a polymorphic and aposematic moth. Evolutionary Ecology,
Greaves, J.H., Redfern, R., Ayres, P.B. & Gill, J.E. (1977) Warfarin resistance:
a balanced polymorphism in the Norway rat. Genetics Research, 30(3),
Gray, S.M. & McKinnon, J.S. (2007) Linking color polymorphism mainte-
nance and speciation. Trends in ecology & evolution, 22(2), 71–79.
Hedrick, P.W. (2012) What is the evidence for heterozygote advantage selec-
tion? Trends in ecology & evolution, 27(12), 698–704.
Hedrick, P.W., Stahler, D.R. & Dekker, D. (2014) Heterozygote advantage in
a finite population: black color in wolves. Journal of Heredity, 105(4),
Hedrick, P.W., Smith, D.W. & Stahler, D.R. (2016) Negative-assortative mat-
ing for color in wolves. Evolution, 70(4), 757–766.
Hegna, R.H., Galarza, J.A. & Mappes, J. (2015) Global phylogeography and
geographical variation in warning coloration of the wood tiger moth
(Parasemia plantaginis). Journal of Biogeography, 42(8), 1469–1481.
Hughes, K.A., Houde, A.E., Price, A.C. & Rodd, F.H. (2013) Mating ad-
vantage for rare males in wild guppy populations. Nature, 503(7474),
Huxley, J. (1955) Morphism and evolution. Heredity, 9(1), 1–52.
Jay, P., Chouteau, M., Whibley, A., Bastide, H., Parrinello, H., Llaurens, V. &
Joron, M. (2021) Mutation load at a mimicry supergene sheds new light
on the evolution of inversion polymorphisms. Nature genetics, 53(3),
Johnston, S.E., Gratten, J., Berenos, C., Pilkington, J.G., Clutton-Brock, T.H.,
Pemberton, J.M. & Slate, J. (2013) Life history trade-offs at a single
locus maintain sexually selected genetic variation. Nature, 502(7469),
Kokko, H. & Mappes, J. (2005) Sexual selection when fertilization is not
guaranteed. Evolution, 59(9), 1876–1885.
Krüger, O., Lindström, J. & Amos, W. (2001) Maladaptive mate choice main-
tained by heterozygote advantage. Evolution, 55(6), 1207–1214.
Llaurens, V., Whibley, A. & Joron, M. (2017) Genetic architecture and bal-
ancing selection: the life and death of differentiated variants. Molecular
ecology, 26(9), 2430–2448.
Maan, M.E. & Cummings, M.E. (2008) Female preferences for aposematic
signal components in a polymorphic poison frog. Evolution, 62(9),
Maisonneuve, L., Beneteau, T., Joron, M., Smadi, C. & Llaurens, V. (2021)
When do opposites attract? A model uncovering the evolution of disas-
sortative mating. The American Naturalist, 198(5), 625–641.
Mappes, J., Marples, N. & Endler, J.A. (2005) The complex business of sur-
vival by aposematism. Trends in ecology & evolution, 20(11), 598–603.
Massey, J.H., Chung, D., Siwanowicz, I., Stern, D.L. & Wittkopp, P.J. (2019)
The yellow gene influences Drosophila male mating success through
sex comb melanization. Elife, 8, e49388.
Mérot, C., Llaurens, V., Normandeau, E., Bernatchez, L. & Wellenreuther, M.
(2020) Balancing selection via life-history trade-offs maintains an in-
version polymorphism in a seaweed fly. Nature communications, 11(1),
Merrill, R.M., Neggazi, S., Morrison, C.R., Crisp, R. & McMillan, W.O.
(2018) Experimental manipulation of Heliconius warning patterns re-
duces harassment of previously mated females. bioRxiv, https://doi.org/
McKinnon, J.S. & Pierotti, M.E. (2010) Colour polymorphism and corre-
lated characters: genetic mechanisms and evolution. Molecular ecology,
Nevo, E., Kirzhner, V., Beiles, A. & Korol, A. (1997) Selection versus ran-
dom drift: long–term polymorphism persistence in small populations
(evidence and modelling). Philosophical Transactions of the Royal So-
ciety of London. Series B: Biological Sciences, 352(1351), 381–389.
Nokelainen, O., Hegna, R.H., Reudler, J.H., Lindstedt, C. & Mappes, J.
(2012) Trade-off between warning signal efficacy and mating success
in the wood tiger moth. Proceedings of the Royal Society B: Biological
Sciences, 279(1727), 257–265.
Nokelainen, O., Lindstedt, C. & Mappes, J. (2013) Environment-mediated
morph-linked immune and life-history responses in the aposematic
wood tiger moth. Journal of Animal Ecology, 82(3), 653–662.
Nokelainen, O., Valkonen, J., Lindstedt, C. & Mappes, J. (2014) Changes in
predator community structure shifts the efficacy of two warning signals
in Arctiid moths. Journal of Animal Ecology, 83(3), 598–605.
Nokelainen, O., de Moraes Rezende, F., Valkonen, J.K. & Mappes, J. (2022a)
Context-dependent coloration of prey and predator decision making in
contrasting light environments. Behav. Ecol., 33, 77–86.
Nokelainen, O., Galarza, J.A., Suisto, K., Kirvesoja, J. & Mappes, J. (2022b)
Genetic colour variation visible for predators and conspecifics is con-
cealed from humans in a polymorphic moth. Journal of Evolutionary
Biology, 35, 467–478
Pryke, S.R. & Griffith, S.C. (2007) The relative role of male vs. female mate
choice in maintaining assortative pairing among discrete colour morphs.
Journal of evolutionary biology, 20(4), 1512–1521.
R Core Team. (2013) R: a language and environment for statistical comput-
ing. R Foundation for statistical computing, Vienna.
Rojas, B., Gordon, S.P. & Mappes, J. (2015) Frequency-dependent flight ac-
tivity in the colour polymorphic wood tiger moth. Current Zoology,
Rojas, B., Burdfield-Steel, E., Pakkanen, H., Suisto, K., Maczka, M., Schulz,
S. & Mappes, J. (2017) How to fight multiple enemies: target-specific
14 EVOLUTION 2022
BALANCING SELECTION ON THE COLOR LOCUS
chemical defences in an aposematic moth. Proceedings of the Royal
Society B: Biological Sciences, 284(1863), 20171424.
Rönkä, K., Mappes, J., Kaila, L. & Wahlberg, N. (2016) Putting Par ase mi a
in its phylogenetic place: a molecular analysis of the subtribe Arctiina
(Lepidoptera). Systematic Entomology, 41(4), 844–853.
Roush, R.T. (1986) Inbreeding depression and laboratory adaptation in He-
liothis virescens (Lepidoptera: Noctuidae). Annals of the Entomological
Society of America, 79(4), 583–587.
Sacchi, R., Scali, S., Pupin, F., Gentilli, A., Galeotti, P. & Fasola, M. (2007)
Microgeographic variation of colour morph frequency and biometry of
common wall lizards. Journal of Zoology, 273(4), 389–396.
Santostefano, F., Galarza, J.A. & Mappes, J. (2018) Testing the direct
and genetic benefit hypotheses of polyandry in the wood tiger moth.
Behavioral Ecology and Sociobiology, 72(7), 1–10.
Sinervo, B. & Lively, C.M. (1996) The rock–paper–scissors game and
the evolution of alternative male strategies. Nature, 380(6571),
Strickland, L.R., Fuller, R.C., Windsor, D. & Cáceres, C.E. (2021) A poten-
tial role for overdominance in the maintenance of colour variation in the
Neotropical tortoise beetle, Chelymorpha alternans.Journal of Evolu-
tionary Biology, 34(5), 779–791.
Suomalainen, E. (1938) Die Erblichkeitsverhältnisse des männlichen Dimor-
phismus bei Parasemia plantaginis. Hereditas, 24(3), 386–390.
Svensson, E.I. & Abbott, J. (2005) Evolutionary dynamics and population
biology of a polymorphic insect. Journal of evolutionary biology, 18(6),
Svensson, E.I., Abbott, J. & Härdling, R. (2005) Female polymorphism, fre-
quency dependence, and rapid evolutionary dynamics in natural popu-
lations. The American Naturalist, 165(5), 567–576.
Tammaru, T. & Haukioja, E. (1996) Capital breeders and income breeders
among Lepidoptera: consequences to population dynamics. Oikos, 561–
Tuttle, E.M. (2003) Alternative reproductive strategies in the white-throated
sparrow: behavioral and genetic evidence. Behavioral Ecology, 14(3),
Wellenreuther, M., Svensson, E.I. & Hansson, B. (2014)Sexual selection and
genetic colour polymorphisms in animals. Molecular ecology,23(22),
White, T.E. (2017) Digest: strengthening the link between sexual selection
and color polymorphism. Evolution, 71, 1913-1914.
Wilson, R., Burnet, B., Eastwood, L. & Connolly, K. (1976) Behavioural
pleiotropy of the yellow gene in Drosophila melanogaster.Genetics Re-
search, 28(1), 75–88.
Winters, A.E., Lommi, J., Kirvesoja, J., Nokelainen, O. & Mappes, J.
(2021) Multimodal aposematic defenses through the predation se-
quence. Frontiers in Ecology and Evolution, 9, 657740.
Wittkopp, P.J., Vaccaro, K. & Carroll, S.B. (2002) Evolution of yellow gene
regulation and pigmentation in Drosophila.Current Biology, 12(18), pp.
Associate Editor: M. Kronforst
Handling Editor: T. Chapman
Additional supporting information may be found online in the Supporting Information section at the end of the article.
Table Supp. info 1. Table a) reports the model selection for testing if inbreeding coefficients differ at the genotype level, and genotype pairwise compar-
Table Supp. Info 2. Sample size of the individuals tested and all possible pair combinations used in the experiments.
Table Supp. Info 3. The table below reports the model selection for the analyzes that tested whether the heterozygote advantage was linked to the sire or
Table Supp. Info 4 The upper table below reports the model selection for the oviposition and hatching day considering both models at the individual level
(males and females) and at the pair level (number of yellow alleles in the pair and the pair type).
Table Supp. Info 5. Table a) reports the effect of the genotype, weight and age on the mating probability, for males (upper part) and females (lower part)
separately through Chi-square test.
Table Supp. Info 6. The upper part of the table reports the effect of the number of yellow allele and the pair type on the copulation probability through
Table Supp. Info 7. The upper table reports the effect of the number of yellow allele and the pair type on the mating delay through LR-test.
Table Supp. Info 8. Table a) reports the model selection for the fecundity, fertility and hatching success traits for males and females.
Table Supp. Info 9. Table a) reports the effect of genotype, weight and inbreeding coefficient on the fecundity of males and females.
Table Supp. Info 10. Table a) reports the effect of genotype, weight and inbreeding coefficient on the fertility of males and females.
Table Supp. Info 11. Table a) reports the effect of genotype, weight and inbreeding coefficient on the hatching success of males and females.
Table Supp. Info 12. The table reports the pairwise comparisons between pair type for the mating success.
EVOLUTION 2022 15