Genetic Architecture of Hybrid Male Sterility in Drosophila: Analysis of Intraspecies Variation for Interspecies Isolation
Markow TA (2008) Genetic Architecture of Hybrid Male Sterility in Drosophila: Analysis of Intraspecies Variation for Interspecies
Isolation. PLoS ONE 3(8): e3076. doi:10.1371/journal.pone.0003076
Genetic Architecture of Hybrid Male Sterility in Drosophila : Analysis of Intraspecies Variation for Interspecies Isolation
Laura K. Reed 0
Brooke A. LaFlamme 0
Therese A. Markow 0
Pawel Michalak, University of Texas Arlington, United States of America
0 1 Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, Arizona, United States of America, 2 Department of Genetics, North Carolina State University , Raleigh , North Carolina, United States of America, 3 Department of Molecular Biology and Genetics, Cornell University , Ithaca , New York, United States of America, 4 Division of Biological Sciences, University of California San Diego , La Jolla, California , United States of America
Background: The genetic basis of postzygotic isolation is a central puzzle in evolutionary biology. Evolutionary forces causing hybrid sterility or inviability act on the responsible genes while they still are polymorphic, thus we have to study these traits as they arise, before isolation is complete. Methodology/Principal Findings: Isofemale strains of D. mojavensis vary significantly in their production of sterile F1 sons when females are crossed to D. arizonae males. We took advantage of the intraspecific polymorphism, in a novel design, to perform quantitative trait locus (QTL) mapping analyses directly on F1 hybrid male sterility itself. We found that the genetic architecture of the polymorphism for hybrid male sterility (HMS) in the F1 is complex, involving multiple QTL, epistasis, and cytoplasmic effects. Conclusions/Significance: The role of extensive intraspecific polymorphism, multiple QTL, and epistatic interactions in HMS in this young species pair shows that HMS is arising as a complex trait in this system. Directional selection alone would be unlikely to maintain polymorphism at multiple loci, thus we hypothesize that directional selection is unlikely to be the only evolutionary force influencing postzygotic isolation.
Funding: National Science Foundation support for this work was provided by Interdisciplinary Research Training Group on Plant-Insect Interactions
(DBI9602249) to the University of Arizona and the Integrative Graduate Education and Research Training Grant in Evolutionary, Computational, and Molecular
Approaches to Genome Structure and Function (DGE-0114420) to the University of Arizona, Doctoral Dissertation Improvement Grant (DEB 0414184) to LKR, and
grants (DEB 95-10645 and DEB 0075312) to TAM. In addition, the Undergraduate Biological Research Program (UBRP) at the University of Arizona supported BAL.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
During the modern synthesis, Dobzhansky and Mayr proposed
the Biological Species Concept that defines species by their ability
to exchange genetic material within their group while being
prevented from exchanging genetic material between groups [1,2].
Reproductive isolating mechanisms subsequently have been the
focus of extensive research such that a great deal now is
understood about mechanisms of both pre and postzygotic
isolation. Capturing the process of speciation early enough to
determine the initial genetic causes of reproductive isolation,
however, is challenging. Consequently a gap still remains in our
understanding of the mechanisms underlying the very first stages
of postzygotic isolation.
Postzygotic isolation exists when a hybrid, or its progeny,
experience a reduction in fitness relative to the parental types.
Whether the genetic variation needed for postzygotic isolation is
segregating within species in the form of epistatic variation [3,4] or
instead arises as de novo mutations in allopatric populations 
remains controversial. Thus, there is need for further empirical
study to determine the roles of segregating and de novo
polymorphism in speciation, and the underlying architecture of
such variation .
Three general patterns of postzygotic isolation are well
supported. First, Haldanes rule  states that decreases in hybrid
fitness are more common in the heterogametic sex; second, hybrid
sterility often appears earlier in species divergence than hybrid
inviability [7,8]; and third, heterogametic isolation generally
appears earlier in speciation than homogametic isolation [8,9].
Taken together, these patterns suggest that the earliest
manifestation of postzygotic isolation likely will be hybrid sterility in the
heterogametic sex. These patterns hold across diverse taxa in
which the male is the heterogametic sex, such as mammals and
Drosophila and interestingly, also for systems where the female is the
heterogametic sex, such as birds and Lepidoptera . Various
theories exist about the basis for Haldanes rule . The
explanations with the most consistent empirical support revolve
around the X (Z)-chromosome , leading to the expectation
that the X-chromosome plays a substantial genetic role in the
occurrence of postzygotic isolation in any given system.
The problem of how a trait that decreases the fitness of hybrids
could evolve between two populations sharing a common ancestral
genome without either population passing through a state of
reduced fitness is solved by the Dobzhansky-Muller (DM) model of
reproductive isolation [1,1517]. For example, if derived alleles of
two loci arise independently in two populations then, upon
meeting in a hybrid, they may be incompatible, leading to
decreased fitness . Several known interacting genomic regions
have been implicated in hybrid incompatibilities . In
addition, the incompatible alleles of the Lhr and Hmr genes in the
Drosophila melanogaster D. simulans pair have been demonstrated
functionally to cause hybrid male lethality .
Expanding the Dobzhansky-Muller model, Orr  proposed
that the number of two locus incompatibilities between two
populations will increase as a function of the square of their
divergence times. If more complex genetic scenarios are
considered, the accumulation of incompatibilities is even faster.
Those genetic incompatibilities that initially defined the two
species therefore, can rapidly become obscured by subsequent
genetic differentiation. Efforts to elucidate the origin of isolation
can be facilitated by identifying the architecture of postzygotic
isolating mechanisms in incipient species, where the number of
postzygotic incompatibilities are still few and, ideally, not yet fixed
within the populations
A major lingering question in speciation genetics concerns the
actual genetic architecture of postzygotic isolation as it first
emerges. Does it have a simple or a complex genetic basis at the
earliest stages of the speciation process? A burdensome
disadvantage to genetic studies of postzygotic isolation is that the phenotype
of interest, sterility or inviability, precludes direct genetic crossing
schemes. Many successful studies of species pairs, in which the
factors underlying isolation are effectively fixed, utilized the
common occurrence of asymmetry  and/or Haldanes Rule.
While the genetic dissections of incompatibility phenotypes by
backcrosses and introgression [10,21,24,26,27] have contributed
significantly to our current understanding of the genetics of
postzygotic isolation, it is not without limitations. Since the
crossing schemes employed potentially disrupted the very
coadapted gene complexes that might be critical to the F1 phenotype
we cannot be certain that the genetic architecture for the
incompatibility observed in the F1 hybrid has been fully
characterized. Evidence of substantial within species epistasis
contributing to postzygotic isolation has been demonstrated in
Tribolium by joint-scaling analysis [28,29] showing that co-adapted
gene complexes may indeed be important in the architecture of
In the present study, we circumvented the limitations of the
hybrid phenotype by exploiting within-species polymorphism for
between-species postzygotic isolation to determine the genetic
architecture of segregating variation for F1 hybrid male sterility
directly. We used Drosophila mojavensis and D. arizonae, a recently
diverged species pair (0.661.2 my [30,31]) and a model system for
the study of speciation . Crosses between D. mojavensis females
and D. arizonae males exhibit widespread between-population
[33,34] and within-population variation for the HMS phenotype
suggesting polymorphism at multiple loci within D. mojavensis .
The polymorphism for HMS shows that this trait is not yet fixed in
D. mojavensis. Since intraspecific evolution is the process by which
gene frequencies change, only a trait exhibiting genetic variation is
capable of evolving. Thus HMS in this system is still capable of
experiencing the forces of evolution (selection and/or drift) and
can give insight into how HMS evolves. In a novel design, we have
utilized the within-population variation for production of F1 HMS
to determine the genomic regions and epistatic interactions that
shape the architecture of postzygotic isolation within the F1 hybrid
males. Below we address the following questions: What is the
nature and number of polymorphic loci within D. mojavensis
contributing to the HMS phenotype? And what can that tell us, in
conjunction with other information about this species pair, about
the evolution of HMS?
Materials and Methods
Lines used for mapping
Genetic lines were derived from lines used in the HMS
polymorphism study by Reed and Markow . Isofemale D.
mojavensis lines were collected from Santa Catalina Island (CI) in
April 2001. The Low (CI-10) line exhibiting low levels of HMS
was inbred through full-sib mating for six generations and the
High (CI-12) line exhibiting high levels of HMS was inbred for
eight generations. The lines had minimum inbreeding coefficients
of 0.734 and 0.826 respectively . They showed significant
differences from each other for HMS (presence of motile sperm)
observed when females from those lines were crossed to D. arizonae
males from an inbred line (minimum inbreeding coefficient of
0.859) collected at Peralta Canyon Trail Head, east of Phoenix,
Arizona (AZ), in April 1997. All lines show normal male fertility
when the males are genetically pure D. mojavensis or D. arizonae
. The sterility phenotype is only manifested in the species
hybrid state. Homozygosity of the High line, used in the D.
mojavensis genome sequencing project, had been confirmed by
sequencing five marker loci in 10 females (L. Matzkin, unpublished)
and an additional seven loci in 10 individuals by Agencourt
Biosciences, and none showed sequence variation. The two D.
mojavensis QTL lines were homokaryotypic, lacking inversion
polymorphism. The lines used showed extreme contrasting HMS
(sperm motility) phenotypes in heterospecific crosses (Table 1) as
well as strong viability.
Identifying Markers and Genotyping
We designed fluorescently-tagged PCR primers for flanking
sequences of microsatellite loci identified by Ross et al. .
Additional candidate microsatellite sequences were drawn from
Staten et al. . Potential loci were genotyped in a panel of four
individuals from each D. mojavensis line to confirm that the lines
contained different alleles for the locus. We genotyped 132
potential loci but only found 25 that distinguished the High and
1Cross is Genotype of the D. mojavensis mother.
2Motility is defined as a male having one or more motile sperm.
Low lines (primer sequences not already described in Staten et al.
 shown in Table S1). Polymorphic marker discovery was
inhibited by the general deficit of neutral microsatellite variation in
the wild population from which the mapping lines were derived
. We had markers on the X-chromosome and the four major
autosomes but failed to find a marker on the 6th dot chromosome.
Markers were assigned to chromosome using one or more of the
following methods. 1) BLASTing  the flanking sequence
against the D. melanogaster genome sequence to identify which
Muller element and thereby to what chromosome in D. mojavensis it
belonged; 2) linkage mapping to allozyme markers assigned to
chromosome by Zouros [40,41]; 3) looking for close linkage
between assigned markers and unassigned makers. Informative
loci were multiplexed and PCR amplified using HotStarTaq DNA
Polymerase (Qiagen Inc.). Genotyping was performed by the
Genomic Analysis and Technology Core (GATC) at the
University of Arizona using an ABI 3730 DNA Analyzer.
Genotypes were read using Genotyper ver. 1.1 from Applied
The crossing design used was an innovative modification of a
standard F2 line cross (Figure 1). Five virgin F2 females, at six days
post-eclosion, were paired with five sexually mature virgin male D.
arizonae (AZ) for 48 hours. After 48 hours, the females were placed
individually into vials, allowed to oviposit for four days, then
removed and frozen for later genotyping. Virgin hybrid males
were collected from those vials producing viable adults. Hybrid
males were allowed to reach sexual maturity at 912 days
posteclosion before phenotyping. Use of the F2 mothers genotypes
allowed for a largest possible sample of recombinant genotypes
(given limited resources for genotyping) while the use of the sons
phenotypes allowed for a good estimate of their mothers average
phenotype, maximizing the quality and quantity of the data. A
reciprocal design was also performed; instead of using Low females
crossed to High males at the first step (F2 cross) High females were
crossed to Low males (RF2 cross). A large number of attempted
matings between F2 and RF2 females and D. arizonae males failed
to produce offspring as was expected considering the significant
prezygotic isolation between the species. Seven percent of the F2
(224 of 3214) and the 12% of the RF2 (278 of 2339) mothers
produced larvae after being mated to D. arizonae males.
Mating and/or oviposition took place on standard
opuntiabanana food (http://flyfood.arl.arizona.edu/opuntia.php3#s2.1)
while virgins were maintained on standard cornmeal food (http://
flyfood.arl.arizona.edu/cornmeal.php3). Cultures were
maintained at 24+/21 degree Celsius and under a strict 12h: 12h
light: dark cycle.
Male Drosophila store mature sperm in a seminal vesicle, a
structure at the base of each testis. To assess sperm motility, males
were anesthetized with ether just prior to dissection. Testes and
seminal vesicles were dissected from the male in sperm buffer
(0.05 M Tris, 1.1% NaCL, 0.1% Glucose, 0.01% L-arginine,
0.01% L-lysine, pH 8.7). Reproductive organs were transferred to
a fresh slide with an 11 microliter droplet of sperm buffer,
stretched with forceps to be linear, and then placed under a
coverslip. Pressing the coverslip down on the sample forced all
sperm out of the seminal vesicle, allowing motility of the sperm to
be observed under dark-field microscopy. The entire area where
potentially motile sperm could be found was visually scanned and
each seminal vesicle assigned an integer score from 0 to 6 on a
sperm motility scale.
Unlike mammalian systems that have short sperm, Drosophila
sperm are extremely long, forming a tangled wiggling knot, in
which the movement of any individual sperm cannot be tracked.
The zero-to-six scale is a reasonable means to approximate the
amount of motility exhibited by Drosophila sperm, and is a
refinement of the presence/absence of motility scheme used in
many previous studies of sperm motility in Drosophila [33,42,43]. A
quantitative measurement of sperm motility more suitable for
quantitative genetic analysis results from the zero-to-six scale. A
score of 0 meant there was no motility observed, 1 was one or
two motile sperm, 2 being several motile sperm, 3 being
several motile sperm in two or more areas of the field, 4 is three
to four areas of moderate motility, 5 being several large areas of
high motility, and 6 being wild-type where most of the field
showed high levels of sperm motility (as is observed in non-hybrid
males of these species). A males phenotype was the average score
of his two seminal vesicles. We phenotyped up to ten sons for each
mother (mean family size of 3.4) and found no trends of decreasing
motility scores with time (first verses second testis scored). The
genotype of the male being scored was not known during
phenotyping. A subset of samples were also scored and rescored
in a randomized order while blind to sample identity. Replicate
scores were highly correlated (R2 = 0.79, p,0.0001), validating
this method for estimating Drosophila sperm motility suggesting its
utility for future studies. Sperm motility is one phenotypic
component of sterility. We have shown in previous work that
motile sperm are required though not always sufficient for males to
be able to reproduce in this system due to additional genetic
polymorphism for other phenotypic components of sterility .
Studies of other phenotypic components of sterility would be
interesting though not required to assess the importance of natural
genetic variation for HMS as estimated by sperm motility in this
F2 or RF2 females that successfully produced hybrid male
offspring were genotyped for the 25 microsatellite markers
described above. Maximum likelihood marker (Figure S1) order
was calculated for each chromosome and the Haldane mapping
function was used to assign linkage distances between markers in
centiMorgans (cM) using Mapmaker 3.0 .
Autosomal genotypes were pooled from the F2 and RF2 crosses
for linkage map calculation while they are separated for the
Xchromosome due to differences in X-chromosome inheritance
patterns in the two crosses. Marker density was an average of 27.6
QTL mapping was conducted using Windows QTL
Cartographer 2.5 [45,46]. We conducted composite interval mapping
[47,48] on the F2, RF2, and combined data sets using a common
linkage maps for the autosomes and their respective linkage maps
for the X-chromosome. Composite interval mapping (CIM) uses
cofactors from other regions to control for genome wide influences
when regression analysis at any given genomic location is being
tested [47,48]. For our study a multiple regression analysis at each
2 cM segment of each chromosome was performed using five
randomly selected markers from elsewhere in the genome as
cofactors to control for genetic effects in other genomic regions
(Model 6 in QTLCartographer). A 10 cM region around the test
location was excluded for selection of the five control markers. A
likelihood ratio (LR) test was performed for each position
comparing the null hypothesis that no QTL is present at that
position to the alternative that there is a QTL present. To control
for any violations of normality, multiple testing, or the possible
disruption of the asymptotic distribution of the chi-squared test
statistic by idiosyncrasies of the study (due to factors such as
multiple QTL per chromosome or distorted segregation ratios),
QTL peak significance was assessed by 1000 permutations .
Two datasets were used in mapping. In the one called sons, sons
of each mother were analyzed with their mothers genotype and
their individual motility phenotype to account for the within
mother variation in her sons motility scores. Analyses were also
conducted on the second dataset using the mothers Best Linear
Unbiased Prediction (BLUP) [50,51], which considers both the
mean and variance in offspring scores, as a summary of her
breeding quality for hybrid male sperm motility calculated in
the SAS/STAT Mixed Procedure (SAS 9.1.3, SAS Institute Inc.
Cary, NC 20002004). The BLUP data is well suited to determine
the basis of additive (heritable) variation while the sons data set is
better for detecting genetic loci experiencing non-additive effects.
We assessed epistatic interactions using methods modified from
Moehring and Mackay  and Morgan and Mackay . We
tested for interaction effects between all pairs of markers using a
repeated measures model of the form:
Y ~ m z F z C z M1 z M2 z M11M2 z E
where F is the effect of mother, C is the effect of the direction of
the cross (F2 vs. RF2), M1 is the effect of genotype at marker 1 and
M2 is the effect of genotype at marker 2. All 300 possible two-way
interactions were tested and significance of the interaction term
(M1*M2) was assessed by Bonferroni correction and False
Discovery Rate (FDR). All calculations were implemented using
the SAS Mixed procedure (SAS 9.1.3, SAS Institute Inc., Cary,
Maternal effects were identified as a significant cross by main
effect interaction. Hybrid male phenotype (n = 304) was regressed
against the interaction between marker genotypes as the four main
effect loci and the direction of the cross (F2 vs. RF2) in the multiple
regression model of the form:
Y ~ m z C z M1 z M11C z E
Where C in the direction of the cross and M1 is the genotype at the
marker associated with one of the main effect QTL.
Inbred D. mojavensis lines showing High and Low phenotypic
levels of HMS when crossed to D. arizonae were used in a standard
F2 (and reciprocal) intercross QTL mapping design (Figure 1).
Sperm motility in the parental, F1, RF1, F2, and RF2 lines are
presented in Table 1. High and Low lines showed significant
differences in the proportion of inter-species hybrid sons with
motile sperm (91% and 46% respectively), and motility clearly was
dominant in all hybrids (Table 1). Sperm motility is necessary but
not always sufficient for fertility , thus motility is used as a
correlate to absolute fertility. Sperm motility scores were normally
distributed (Shapiro-Wilks W = 0.99, p = 0.38), with a mean of
Female F2s (149 F2 and 155 RF2), of the species D. mojavensis,
were crossed to male D. arizonae for QTL analysis. Sperm motility
was scored, from 0 to 6, in the interspecific hybrid males. 304 F2
mothers were genotyped at 25 microsatellite markers across the
genome and used to generate the linkage map. Phenotypes were
mapped relative to genotype in two ways. One was to use each
sons motility score, hereafter referred to as sons, as a sample of the
mothers phenotype and the other was to map a mothers Best
Linear Unbiased Prediction (BLUP) for her sons motility scores.
Sons better captures non-additive effects, while BLUP highlights
the additive effects in the architecture.
There were two striking outcomes of the main effects analysis;
first that the F2 and RF2 mapping populations did not match more
closely, and second, no main effects were detected on the
Xchromosome. Main effect QTL were detected, however, on the
2nd, 3rd, and 5th chromosomes (Table 2, Figure 2). Note that
estimated effect sizes (percent of phenotypic variation explained) of
all detected QTL are likely to be overestimates due to the Beavis
effect [54,55], thus, there are likely to be additional genetic factors
contributing to observed differences between lines that were not
detected in this study. Also the actual genotype of the hybrid males
at a given marker can only fall into two possible genotypic classes,
High or Low for the D. mojavensis half of their genome: they will be
monomorphic with respect to the D. arizonae half of their genome.
Thus, the variance due to non-additive effects identified in the
QTL mapping corresponds to intra-genome epistasis rather then
intra-locus dominance. Finally, remember that these two lines are
only a sample from a larger population. There may be additional
polymorphic factors on the X-chromosome or autosomes
segregating in the population as a whole that were not captured
by this biparental cross. One QTL with estimated effect size of
10.922.1% occurred on the 3rd chromosome at about 60 cM and
was observed in both datasets (sons and BLUP) and across all
population samples (F2, RF2, and combined). The RF2 population
sample did not reveal any other significant QTL.
QTL were also identified on the 2nd chromosome. Analysis of
the sons dataset in the F2 and combined populations found two and
one QTL on the 2nd chromosome respectively (Table 2, Figure 2).
The combined population sons QTL at 58 cM has a large
Position (cM2) LR3
additive effect of 2.23 motility units and estimated effect size is
55.3%. The two QTL found in the F2 population sons dataset (at
32 and 66 cM) also had large (and opposite) non-additive effects of
22.17 and 2.16 motility units, respectively. The RF2 sons dataset
had an elevated likelihood ratio on the 2nd chromosome but it did
not reach significance. The BLUP datasets exposed a significant
QTL on 2nd chromosome for the combined population (60 cM)
and the RF2 population (50 cM) with a non-additive effect of 0.92
to 1.10 motility units, with estimated effect size of 44.3 to 59.3%.
The lack of a significant 2nd chromosome QTL in the F2 BLUP
dataset can be explained by the apparent equal and opposite
nonadditive effects of the two QTL found in the sons dataset.
Chromosome five also contained a QTL, but only in the F2
population sons dataset. It was at 54.5 cM with a relatively large
additive effect of 20.80 and moderate non-additive effect of
20.43 motility units with an estimated effect size of only 9%.
We tested for biases due to variation in family size by rerunning
all analyses of both the BLUP and sons datasets on families limited to
three randomly selected sons. All QTLs replicated with controlled
family size (data not shown) so we are confident that the mapping
results are not due to biases in family size. We also looked for any
biases introduced by segregation bias due to prezygotic isolation
between the species or hybrid male over dominance in viability. We
found that 15 of the 25 markers were not in the expected 1:2:1
(autosomes) or 1:1 (x-chromosome) ratios having excess
heterozygotes, but there was no association between violation of these ratios
and the location of the QTL (X2 = 0.45, p = 0.50). Thus, there is
evidence of genetic variation for prezygotic isolation with a possible
heterozygote advantage that may be worth mapping in future
studies, but that variation did not introduce bias into the mapping
for HMS. The linkage map, though, is likely only accurate for the
population that was genotyped and cannot be applied to the
F2females that failed to mother hybrid sons, because two or more
segregation distortion loci within a given marker interval can bias
estimates of recombination rate .
The lack of a segregating main effect on the X-chromosome was
surprising, as this chromosome has been generally shown to play a
major role in the evolution of hybrid male sterility . There
may still be factors on the X-chromosome (or autosomes)
segregating in the ancestral population as a whole whose effects
were not captured by the two lines used in this study, and there
also may be fixed factors on the X-chromosome that we cannot
detect with this design. Additionally, there were several major
discrepancies between the F2 and RF2 populations. Specifically,
four main effect QTL were identified in the F2 population; two
had largely additive effects on the 3rd and 5th chromosomes and
two had largely non-additive effects on the 2nd chromosome. In
the RF2 population, however, only a single, largely additive QTL
was found on the 3rd chromosome. The combined population data
set shows one 2nd chromosome QTL and the 3rd chromosome
QTL. These unexpected results motivated further analyses of the
underlying architectural complexity of HMS.
There are several possible explanations for the discrepancy
between F2 and RF2 populations. First, we may have failed to
detect QTL due to simple sampling variance between the F2 and
RF2 populations. Second, the different compositions of the
Xchromosomes in the reciprocal populations could have lead to an
X-by-autosome interaction. Finally, an interaction between the
nature of the cross (e.g. cytoplasm) and the QTLs could account for
(or underlie) the observed differences between reciprocal crosses.
We address each of these below.
Sampling Variance: To test whether the inevitable force of
sampling variance was driving the differences observed between
the F2 and RF2 populations, we subsampled from the datasets to
test for consistency of the signal. From each population 100
datasets were created by randomly selecting 80% of the data
points from the original dataset. Complete mapping analyses,
including permutations for significance thresholds, were then
performed on the subsampled datasets. Despite the reduced power
of the smaller datasets, the QTLs found on the full dataset
remained detectable in between 27% and 70% of the subsampled
datasets (Figure S2), while spurious QTL were only detected in a
maximum of 24% (e.g. see left portion of chromosome 5 in the F2
dataset). Subsampling indicates that QTL identified in the
complete dataset remain detectable in a large portion of the
subsets despite variation in sampling and spurious QTL (those not
detected in the complete dataset but in the subsets) rarely arise due
to sampling variance. The consistency between our subsampled
datasets and our complete dataset suggests that our findings are
fairly robust to variation in sampling and that the differences
between the F2 and RF2 are likely biologically real, though some of
the differences could still be due in part to sampling variance.
Inheritance of the X-chromosome: If an important X-linked genetic
difference between the parental lines interacts with the autosomes,
the cross direction will affect the frequency of that factor in the
mapping population. This, in turn, would influence the detection
and nature of autosomal QTL that interact with the X. The
Xlinked factor could primarily influence the penetrance or
expression of the genetic variation at the autosomal factor and
thus have a significant epistatic effect without having a main effect.
To test for X-by-autosome interactions, as well as for other
epistasis, we looked for significant pairwise interactions between all
markers across the combined dataset (n = 304). We found seven
interactions with a FDR significance of 0.05 or less, two of which
were between the X-chromosome and the autosomes (Table 3,
Figure 2). Epistatic interactions between the X-chromosome and
the autosomes therefore could be contributing to some of the
differences observed between the F2 and RF2 populations. We also
Markers and their corresponding chromosomes with interaction effects with a False Discovery Rate (FDR) of 0.05 or less. Interactions significant after a Bonferroni
correction are in bold and interactions significant at an FDR of 0.05 are italicized.
found 5 inter-autosomal epistatic interactions. Notice that markers
on the 4th chromosome were involved in five of the seven
interactions, thus the 4th chromosome appears to be playing an
important epistatic role in the HMS in this system worthy of
additional study. Only one of nine markers involved in these
pairwise interactions (m4_9_1 on the 5th chromosome) resides in
or near a main effect QTL. Two additional markers (a1_2_1 on
chromosome 3 and m1_10_11 on chromosome 4) did not appear
in the list of significant marker by marker interactions, but
exhibited significant deviations from the additive expectation after
a Bonferroni correction, indicating that there are at least two other
unidentified epistatic interactions. Taken together, not only does
the X-chromosome appear to play an epistatic role in shaping the
genetic architecture of HMS in this system, but so does
withinspecies epistasis throughout the rest of the genome.
Maternal effects: Cytoplasmic factors may also underlie
interactions between the direction of the cross and the QTLs. We tested
for marker by cross interactions for the four markers showing main
effects and found a significant interaction effect of marker
dmoj2210 on the 2nd chromosome with cross (p = 0.0142), after
controlling for the main effect of the 3rd chromosome (Figure 3).
This marker did not show evidence of interacting with the
Xchromosome in the test for epistasis above. Chromosome 2 is
where the large and opposing non-additive QTL were found in the
F2 population but not in the RF2 populations. The significant cross
by marker interaction on the 2nd chromosome is a likely
contributor to the differences seen in the mapping for the two
crosses. The cross effect is not necessarily independent of the
epistatic effect of the X chromosome found above. But since
dmoj2210 did not show any evidence of interacting with the
Xchromosome, it is likely that the cytoplasm plays a role in HMS.
Thus, both the X-chromosome and the cytoplasmic environment
should be further explored in future studies of HMS.
The first question we proposed to answer in this study was: what
is the basic architecture of HMS in the early stages of its evolution?
Is HMS caused by only a few factors of large effect or does it have
a more complex basis, involving multiple loci and epistatic
interactions? An earlier study  showed evidence of multiple
polymorphic factors for HMS within D. mojavensis and additional
data show genetic polymorphisms for HMS segregating in the
sister species D. arizonae (Reed, unpublished). Our present QTL
study reveals that HMS, even at the earliest stages of speciation,
can be surprisingly complex. We found within-population
variation not only for several QTL of main effect, but also for
significant within-species epistatic interactions. Similar complexity
has been seen in other systems, such as with the Lhr-Hmr gene pair,
where the incompatibility requires an epistatic interaction with the
hybrid genetic background to be expressed . There is likely to
be more segregating genetic variation in this species that was not
captured by these two particular mapping lines, and fixed genetic
factors that could not be detected by this design; potentially both
of which further contribute to the complexity of this trait. While
previous investigations on more diverged species pairs  have
revealed a multi-factored genetic basis for post-zygotic isolation,
these studies were unable to determine whether the polygenicity
had arisen early or was a byproduct of long divergence time. We
show that HMS has a polygenic and epistatic basis even early in
the evolutionary process, before the ability to produce HMS is
fixed within a species. These findings are consistent with the
complex genetic basis of early postzygotic isolation found in
In addition to the unexpected complexity of HMS in this
incipient species pair, we were surprised to discover no main-effect
QTL segregating on the X-chromosome, though we did find an
epistatic effect. While we may have lacked the statistical power to
detect X-linked main effects, main effects were detected on other
chromosomes indicating that sample size alone is not the full
explanation. There may still be fixed X-linked effects that could
not be detected by this design, and fixed effects on the X are more
likely due its smaller effective population size. Demuth and Wade
did demonstrate the role of epistatic variation on the
Xchromosome on Haldanes Rule both theoretically  and
empirically [28,29]. We also found such a role for the
Xchromosome in our epistasis tests.
Finally, our data show a striking difference between reciprocal
crosses, suggesting a potential role for maternal effects and
genome-wide epistatic variation on HMS. Cytoplasmic and
epistatic effects have been demonstrated in other systems as well
[24,28,29,59]. Zouros and colleagues have mapped the genetic
basis of HMS resulting from the reciprocal cross (female D. arizonae
mated to male D. mojavensis) to the third, fourth and
Ychromosomes [40,41,6062]. Whether or not the third
chromosome effect in the cross using D. arizonae mothers is a function of
the same factor or factors as revealed in our study remains
The second question we proposed to answer was: what can the
genetic architecture of HMS tell us, in conjunction with other
information, about its evolution? Substantial within-species natural
polymorphism for between species postzygotic isolation has been
characterized in many independent taxa thus far, including at least
nine Drosophila species groups, [33,34,43,6370], Tribolium beetles
[71,72], Chorthippus parallelus grasshoppers , Mus musculus ,
Xiphophorus fish , Crepis hawksberg weed , Gossypium cotton
 and Mimulus monkey flowers . Thus, intraspecific
polymorphism for postzygotic isolation appears to be a common
phenomenon. Shuker et al.  argue that witnessing substantial
within-species polymorphism for hybrid incompatibilities means
that loci contributing to the hybrid phenotype are likely to be
neutral or nearly-neutral in the parental species genetic
background. Finding that a trait is polymorphic at multiple loci,
supports the role of drift and/or balancing selection in the genetics
of the trait. If directional selection was occurring on the genes
related to the trait, no matter what the within- species function of
the genes, they would be polymorphic for only the briefest of times
. The probability of catching multiple polymorphic loci under
directional selection would be minute. Considering that there is
evidence of balancing selection on male reproductive proteins in
several systems  it is not unreasonable to hypothesize that
it might also be playing a role in HMS.
A handful of genes causing hybrid incompatibilities have been
identified in other species and all but one (a gene transposition
) show evidence of being under positive selection (OdsH in D.
mauritiana, ; Xmrk-2 in Xiphophorus, ; Nup96, Hmr, Lhr in the
D. melanogaster/D. simulans species pair [24,87,88]), leading to the
impression that most postzygotic speciation genes experience
directional selection. Given our findings with D. mojavensis and in
other systems discussed above, however, we believe it is premature
to discount the role of drift and balancing selection in speciation.
Empirically, we need to characterize more completely the genetic
architecture of incompatibility phenotypes during early stages to
find reproductive isolation genes before they are fixed and test
those loci for evidence of selection. In the present study we have
begun the empirical aspect of this process.
Figure S1 Linkage map for D. mojavensis. Distance between
markers given in centiMorgans.
Figure S2 Subsampling effect on QTL detection. Bar graphs of
count (right x-axis, out of 100) of 80%-subsampled datasets at each
mapping interval showing a likelihood ratio greater than the
significance threshold determined by permutation for each subset.
Solid lines indicate likelihood ratio (left x-axis) calculated from the
complete dataset. A. F2-sons dataset. B. RF2-sons dataset.
Subsampling indicates that QTL identified in the complete dataset
remain detectable in a large portion of the subsets and spurious
QTL (those not detected in the completely dataset) rarely arise due
to sampling variance. Thus, patterns observed in the complete
dataset appear robust to sampling variance.
Found at: doi:10.1371/journal.pone.0003076.s003 (0.16 MB
We greatly appreciated the assistance from M. Wasserman in confirming
homosequentiality of the two D. mojavensis QTL lines by inspection of
polytene chromosomes. We thank C. Ross for assistance in collecting
specimens and M. Lowrance for assistance with inbreeding. We also
appreciate advice on this study and manuscript from I. Dworkin, G.
Gibson, E. Kelleher, C. Machado, T. Mackay, L. Matzkin, T. Morgan, D.
Nielsen, D. Papaj, B. Payseur, and B. Walsh. We also are greatly indebted
to the Santa Catalina Island Conservancy for collecting permits and
logistical support in collecting specimens.
Conceived and designed the experiments: LKR TAM. Performed the
experiments: LKR BAL. Analyzed the data: LKR. Wrote the paper: LKR
1. Dobzhansky T ( 1937 ) Genetics and the origin of species . New York : Columbia University.
2. Mayr E ( 1942 ) Systematics and the Origin of Species . New York: Columbia University Press.
3. Goodnight CJ , Wade MJ ( 2000 ) The ongoing synthesis: A reply to Coyne , Barton, and Turelli . Evolution 54 : 317 - 324 .
4. Wade MJ , Goodnight CJ ( 1998 ) Perspective: The theories of Fisher and Wright in the context of metapopulations: When nature does many small experiments . Evolution 52 : 1537 - 1553 .
5. Coyne JA , Barton NH , Turelli M ( 2000 ) Is Wright's shifting balance process important in evolution? Evolution 54 : 306 - 317 .
6. Haldane JBS ( 1922 ) Sex ratio and unisexual sterility in animal hybrids . Journal of Genetics 12 : 101 - 109 .
7. Wu CI ( 1992 ) A note on Haldane's rule: hybrid inviability verses hybrid sterility . Evolution 46 : 1584 - 1587 .
8. Coyne JA , Orr HA ( 1997 ) ''Patterns of speciation in Drosophila'' revisited . Evolution 51 : 295 - 303 .
9. Coyne JA , Orr HA ( 1989 ) Patterns of Speciation in Drosophila . Evolution 43 : 362 - 381 .
10. Coyne J , Orr HA ( 2004 ) Speciation. Sunderland, MA: Sinauer Associates Inc.
11. Turelli M , Orr HA ( 1995 ) The dominance theory of Haldane's rule . Genetics 140 : 389 - 402 .
12. Orr HA , Presgraves DC ( 2000 ) Speciation by postzygotic isolation: forces, genes and molecules . Bioessays 22 : 1085 - 1094 .
13. Demuth JP , Wade MJ ( 2005 ) On the theoretical and empirical framework for studying genetic interactions within and among species . American Naturalist 165 : 524 - 536 .
14. Masly JP , Presgraves DC ( 2007 ) High-Resolution Genome-Wide Dissection of the Two Rules of Speciation in Drosophila . PLoS Biology 5 : e243 .
15. Muller HJ ( 1942 ) Isolating mechanisms, evolution and temperature . Biological Symposia 6 : 71 - 125 .
16. Muller HJ ( 1940 ) Bearing of the Drosophila work on systematics . In: Huxley J , ed. The New Systematics. Oxford: Clarendon. pp 185 - 268 .
17. Bateson W ( 1909 ) Heredity and variation in modern lights . In: Seward AC, ed. Darwin and Modern Science. Cambridge, UK: Cambridge University Press . pp 85 - 101 .
18. Orr HA ( 1995 ) The Population-Genetics of Speciation - the Evolution of Hybrid Incompatibilities . Genetics 139 : 1805 - 1813 .
19. Carvajal AR , Gandarela MR , Naveira HF ( 1996 ) A three-locus system of interspecific incompatibility underlies male inviability in hybrids between Drosophila buzzatii and D-koepferae . Genetica 98 : 1 - 19 .
20. Christie P , Macnair MR ( 1984 ) Complementary Lethal Factors in 2 NorthAmerican Populations of the Yellow Monkey Flower . Journal of Heredity 75 : 510 - 511 .
21. Sweigart AL , Fishman L , Willis JH ( 2006 ) A simple genetic incompatibility causes hybrid male sterility in mimulus . Genetics 172 : 2465 - 2479 .
22. Hutchinson JB ( 1932 ) The genetics of cotton . Part VII. ''Crumpled' ': A new dominant in Asiatic cottons produced by complementary factors . Journal of Genetics 25 : 281 - 291 .
23. Kazianis S , Gutbrod H , Nairn RS , McEntire BB , Della Coletta L , et al. ( 1998 ) Localization of a CDKN2 gene in Linkage Group V of Xiphophorus fishes defines it as a candidate for the DIFF tumor suppressor . Genes Chromosomes & Cancer 22 : 210 - 220 .
24. Brideau NJ , Flores HA , Wang J , Maheshwari S , Wang X , et al. ( 2006 ) Two Dobzhansky-Muller genes interact to cause hybrid lethality in Drosophila . Science 314 : 1292 - 1295 .
25. Turelli M , Moyle LC ( 2007 ) Asymmetric postmating isolation: Darwin's corollary to Haldane's rule . Genetics 176 : 1059 - 1088 .
26. Barbash DA , Ashburner M ( 2003 ) A novel system of fertility rescue in Drosophila hybrids reveals a link between hybrid lethality and female sterility . Genetics 163 : 217 - 226 .
27. Presgraves DC ( 2003 ) A fine-scale genetic analysis of hybrid Incompatibilities in drosophila . Genetics 163 : 955 - 972 .
28. Demuth JP , Wade MJ ( 2007 ) Population differentiation in the beetle Tribolium castaneum . I. Genetic architecture . Evolution 61 : 494 - 509 .
29. Demuth JP , Wade MJ ( 2007 ) Population differentiation in the beetle Tribolium castaneum . II. Haldane's rule and incipient speciation . Evolution 61 : 694 - 699 .
30. Matzkin LM ( 2004 ) Population genetics and geographic variation of alcohol dehydrogenase (Adh) paralogs and glucose-6-phosphate dehydrogenase (G6pd) in Drosophila mojavensis . Molecular Biology and Evolution 21 : 276 - 285 .
31. Reed LK , Nyboer M , Markow TA ( 2007 ) Evolutionary relationships of Drosophila mojavensis geographic host races and their sister species Drosophila arizonae . Molecular Ecology 16 : 1007 - 1022 .
32. Markow TA , Hocutt GD ( 1998 ) Reproductive isolation in Sonoran Desert Drosophila: testing the limits of the rules . In: J . HD, S.H. B, eds. Endless Forms: species and speciation . Oxford: Oxford University Press. pp 234 - 244 .
33. Reed LK , Markow TA ( 2004 ) Early events in speciation: Polymorphism for hybrid male sterility in Drosophila . Proceedings of the National Academy of Sciences of the United States of America 101 : 9009 - 9012 .
34. Ruiz A , Heed WB , Wasserman M ( 1990 ) Evolution of the Mojavensis Cluster of Cactophilic Drosophila with Descriptions of 2 New Species . Journal of Heredity 81 : 30 - 42 .
35. Falconer DS , Mackay TFC ( 1996 ) Introduction to quantitative genetics . Harlow. England: Prentice Hall.
36. Ross CL , Dyer KA , Erez T , Miller SJ , Jaenike J , et al. ( 2003 ) Rapid divergence of microsatellite abundance among species of Drosophila . Molecular Biology and Evolution 20 : 1143 - 1157 .
37. Staten R , Schully SD , Noor MAF ( 2004 ) A microsatellite linkage map of Drosophila mojavensis . Bmc Genetics 5.
38. Ross CL , Markow TA ( 2006 ) Microsatellite variation among diverging populations of Drosophila mojavensis . Journal of Evolutionary Biology 19 : 1691 - 1700 .
39. Altschul S , Boguski M , Gish W , Wootton J ( 1994 ) Issues in searching molecular sequence databases . Nature Genetics 6 : 119 - 129 .
40. Zouros E ( 1981 ) An autosome-Y chromosome combination that causes sterility in D. mojavensis and D. arizonensis hybrids . Drosophila Information Service 56 : 167 - 168 .
41. Zouros E ( 1991 ) Searching for speciaton genes in the species pair Drosophila mojavnesis and D. arizonae. In: Young JPW, ed. Molecular Techniques in Taxonomy. Berlin: Springer-Verlag. pp 233 - 247 .
42. Orr HA , Irving S ( 2001 ) Complex epistasis and the genetic basis of hybrid sterility in the Drosophila pseudoobscura Bogota-USA hybridization . Genetics 158 : 1089 - 1100 .
43. Kopp A , Frank AK ( 2005 ) Speciation in progress? A continuum of reproductive isolation in Drosophila bipectinata . Genetica 125 : 55 - 68 .
44. Lander E , Abrahamson J , Barlow A , Daly M , Lincoln S , et al. ( 1987 ) Mapmaker a Computer Package for Constructing Genetic-Linkage Maps . Cytogenetics and Cell Genetics 46 : 642 - 642 .
45. Basten CJ , Weir BS , Zeng ZB ( 1994 ) Zmap-a QTL cartographer . In: Smith C, Gavora JS , Benkel B , Chesnais J , Fairfull W , et al., eds. Proceedings of the 5th world congress on genetics applied to livestock production:Computing strategies and software . Guelph, Ontario, Canada: Organizing committee, 5th world congress on genetics applied to livestock production . pp 65 - 66 .
46. Basten CJ , Weir BS , Zeng ZB ( 2002 ) QTL Cartographer. 1.16 ed. Raleigh, NC: Department of Statistics, North Carolina State Univeristy.
47. Zeng ZB ( 1993 ) Theoretical Basis for Separation of Multiple Linked Gene Effects in Mapping Quantitative Trait Loci . Proceedings of the National Academy of Sciences of the United States of America 90 : 10972 - 10976 .
48. Zeng ZB ( 1994 ) Precision Mapping of Quantitative Trait Loci . Genetics 136 : 1457 - 1468 .
49. Churchill GA , Doerge RW ( 1994 ) Empirical Threshold Values for Quantitative Trait Mapping . Genetics 138 : 963 - 971 .
50. Mclean RA , Sanders WL , Stroup WW ( 1991 ) A Unified Approach to Mixed Linear-Models . American Statistician 45 : 54 - 64 .
51. Lynch M , Walsh B ( 1998 ) Genetics and analysis of quantitative traits . Sunderland, MA: Sinauer Associates Inc.
52. Moehring AJ , Mackay TFC ( 2004 ) The quantitative genetic basis of male mating behavior in Drosophila melanogaster . Genetics 167 : 1249 - 1263 .
53. Morgan TJ , Mackay TFC ( 2006 ) Quantitative trait loci for thermotolerance phenotypes in Drosophila melanogaster . Heredity 96 : 232 - 242 .
54. Beavis W ( 1994 ) The power and deceit of QTL experiments: lessons from comparative QTL studies. Report of the 49th annual corn and sorghum reseach conference . Washington, DC: American Seed Trade Association . pp 252 - 268 .
55. Xu SZ ( 2003 ) Estimating polygenic effects using markers of the entire genome . Genetics 163 : 789 - 801 .
56. Fisher RA , Bailey NTJ ( 1949 ) The Estimation of Linkage with Differential Viability . Heredity 3: 215 - 228 .
57. Lorieux M , Goffinet B , Perrier X , Deleon DG , Lanaud C ( 1995 ) MaximumLikelihood Models for Mapping Genetic-Markers Showing Segregation Distortion .1. Backcross Populations. Theoretical and Applied Genetics 90 : 73 - 80 .
58. Lorieux M , Perrier X , Goffinet B , Lanaud C , Deleon DG ( 1995 ) MaximumLikelihood Models for Mapping Genetic-Markers Showing Segregation Distortion .2. F2 Populations. Theoretical and Applied Genetics 90 : 81 - 89 .
59. Pontecorvo G ( 1943 ) Viability interactions between chromosomes of Drosophila melanogaster and Drosophila simulans . Journal of Genetics 45 : 51 - 66 .
60. Vigneault G , Zouros E ( 1986 ) The Genetics of Asymmetrical Male-Sterility in Drosophila-Mojavensis and Drosophila-Arizonensis Hybrids - Interactions between the Y-Chromosome and Autosomes . Evolution 40 : 1160 - 1170 .
61. Pantazidis AC , Zouros E ( 1988 ) Location of an Autosomal Factor Causing Sterility in Drosophila-Mojavensis Males Carrying the Drosophila-Arizonensis Y-Chromosome . Heredity 60 : 299 - 304 .
62. Zouros E , Lofdahl K , Martin PA ( 1988 ) Male Hybrid Sterility in Drosophila - Interactions between Autosomes and Sex-Chromosomes in Crosses of Drosophila-Mojavensis and Drosophila-Arizonensis . Evolution 42 : 1321 - 1331 .
63. Macknight RH ( 1939 ) The sex-determining mechanism of Drosophila miranda . Genetics 24 : 180 - 201 .
64. Orr HA , Irving S ( 2005 ) Segregation distortion in hybrids between the Bogota and USA subspecies of Drosophila pseudoobscura . Genetics 169 : 671 - 682 .
65. Patterson JT , Stone WS ( 1952 ) Evolution in the Genus Drosophila . New York : Macmillian.
66. Kimura M ( 1987 ) Habitat differentiation and speciaiton in the Drosophila auraria species complex (Diptera: Drosophilidae) . Kontyu 55 : 429 - 436 .
67. Ayala FJ , Tracey ML , Barr LG , Ehrenfeld JG ( 1974 ) Genetic and reproductive differentiation of the subspecies Drosophila equinoxialis caribbensis . Evolution 28 : 24 - 41 .
68. Mainland GB ( 1942 ) Genetic relationships in the Drosophila funebris group . University of Texas Publications 4228 : 74 - 112 .
69. Nolte V , Weigel D , Schlotterer C ( 2008 ) The impact of shared ancestral variation on hybrid male lethality - a 16 codon indel in the Drosophila simulans Lhr gene . Journal of Evolutionary Biology 21 : 551 - 555 .
70. Crow JF ( 1942 ) Cross fertility and isolating mechanisms in the Drosophila mulleri group . University of Texas Publications 4228 : 53 - 67 .
71. Wade MJ , Johnson NA ( 1994 ) Reproductive Isolation between 2 Species of Flour Beetles, Tribolium-Castaneum and T-Freemani - Variation within and among Geographical Populations of T-Castaneum . Heredity 72 : 155 - 162 .
72. Wade MJ , Johnson NA , Jones R , Siguel V , McNaughton M ( 1997 ) Genetic variation segregating in natural populations of Tribolium castaneum affecting traits observed in hybrids with T-freemani . Genetics 147 : 1235 - 1247 .
73. Shuker DM , Underwood K , King TM , Butlin RK ( 2005 ) Patterns of male sterility in a grasshopper hybrid zone imply accumulation of hybrid incompatibilities without selection . Proceedings of the Royal Society BBiological Sciences 272 : 2491 - 2497 .
74. Good JM , Handel MA , Nachman MW ( 2008 ) Asymmetry and polymorphism of hybrid male sterility during the early stages of speciation in house mice . Evolution 62 : 50 - 65 .
75. Nairn RS , Kazianis S , McEntire BB , DellaColetta L , Walter RB , et al. ( 1996 ) A CDKN2-like polymorphism in Xiphophorus LG V is associated with UV-Binduced melanoma formation in platyfish-swordtail hybrids . Proceedings of the National Academy of Sciences of the United States of America 93 : 13042 - 13047 .
76. Hollingshead L ( 1930 ) A lethal factor in crepis effective only in an interspecific hybrid . Genetics 15 : 114 - 140 .
77. Stephens SG ( 1950 ) The Genetics of Corky .2. Further Studies on Its Genetic Basis in Relation to the General Problem of Interspecific Isolating Mechanisms . Journal of Genetics 50 : 9 - 20 .
78. Sweigart AL , Mason AR , Willis JH ( 2007 ) Natural variation for a hybrid incompatibility between two species of Mimulus . Evolution 61 : 141 - 151 .
79. Kimura M , Ohta T ( 1969 ) Average Number of Generations until Fixation of a Mutant Gene in a Finite Population . Genetics 61 : 763 -&.
80. Gasper J , Swanson WJ ( 2006 ) Molecular population genetics of the gene encoding the human fertilization protein zonadhesin reveals rapid adaptive evolution . American Journal of Human Genetics 79 : 820 - 830 .
81. Hamm D , Mautz BS , Wolfner MF , Aquadro CF , Swanson WJ ( 2007 ) Evidence of amino acid diversity-enhancing selection within humans and among primates at the candidate sperm-receptor gene PKDREJ . American Journal of Human Genetics 81 : 44 - 52 .
82. Begun DJ , Whitley P , Todd BL , Waldrip-Dail HM , Clark AG ( 2000 ) Molecular population genetics of male accessory gland proteins in drosophila . Genetics 156 : 1879 - 1888 .
83. Levitan DR , Ferrell DL ( 2006 ) Selection on gamete recognition proteins depends on sex , density, and genotype frequency. Science 312 : 267 - 269 .
84. Masly JP , Jones CD , Noor MAF , Locke J , Orr HA ( 2006 ) Gene transposition as a cause of hybrid sterility in Drosophila . Science 313 : 1448 - 1450 .
85. Ting CT , Tsaur SC , Wu ML , Wu CI ( 1998 ) A rapidly evolving homeobox at the site of a hybrid sterility gene . Science 282 : 1501 - 1504 .
86. Schartl M , Hornung U , Gutbrod H , Volff JN , Wittbrodt J ( 1999 ) Melanoma loss-of-function mutants in xiphophorus caused by Xmrk-oncogene deletion and gene disruption by a transposable element . Genetics 153 : 1385 - 1394 .
87. Barbash DA , Siino DF , Tarone AM , Roote J ( 2003 ) A rapidly evolving MYBrelated protein causes species isolation in Drosophila . Proceedings of the National Academy of Sciences of the United States of America 100 : 5302 - 5307 .
88. Presgraves DC , Balagopalan L , Abmayr SM , Orr HA ( 2003 ) Adaptive evolution drives divergence of a hybrid inviability gene between two species of Drosophila . Nature 423 : 715 - 719 .