Imp 1

RESEARCH ARTICLE
Haploid selection, sex ratio bias, and

transitions between sex-determining systems
Michael Francis Scott1☯*, Matthew Miles Osmond2☯, Sarah Perin Otto2
1 UCL Genetics Institute, Department of Genetics, Evolution and Environment, University College London,
London, United Kingdom, 2 Department of Zoology, University of British Columbia, Vancouver, British
Columbia, Canada
☯ These authors contributed equally to this work.

* [email protected]
a1111111111
a1111111111
a1111111111 Abstract
a1111111111
a1111111111 Sex determination is remarkably dynamic; many taxa display shifts in the location of sex-
determining loci or the evolution of entirely new sex-determining systems. Predominant the-
ories for why we observe such transitions generally conclude that novel sex-determining
systems are favoured by selection if they equalise the sex ratio or increase linkage with a
OPEN ACCESS locus that experiences different selection in males versus females. We use population
Citation: Scott MF, Osmond MM, Otto SP (2018) genetic models to extend these theories in two ways: (1) We consider the dynamics of loci
Haploid selection, sex ratio bias, and transitions very tightly linked to the ancestral sex-determining loci, e.g., within the nonrecombining
between sex-determining systems. PLoS Biol region of the ancestral sex chromosomes. Variation at such loci can favour the spread of
16(6): e2005609. https://doi.org/10.1371/journal.
new sex-determining systems in which the heterogametic sex changes (XY to ZW or ZW to
pbio.2005609
XY) and the new sex-determining region is less closely linked (or even unlinked) to the locus
Academic Editor: Hanna Kokko, University of
under selection. (2) We consider selection upon haploid genotypes either during gametic
Zurich, Switzerland
competition (e.g., pollen competition) or meiosis (i.e., nonmendelian segregation), which
Received: February 7, 2018
can cause the zygotic sex ratio to become biased. Haploid selection can drive transitions
Accepted: June 6, 2018 between sex-determining systems without requiring selection to act differently in diploid
Published: June 25, 2018 males versus females. With haploid selection, we find that transitions between male and
Copyright: © 2018 Scott et al. This is an open female heterogamety can evolve so that linkage with the sex-determining locus is either
access article distributed under the terms of the strengthened or weakened. Furthermore, we find that sex ratio biases may increase or
Creative Commons Attribution License, which decrease with the spread of new sex chromosomes, which implies that transitions between
permits unrestricted use, distribution, and
sex-determining systems cannot be simply predicted by selection to equalise the sex ratio.
reproduction in any medium, provided the original
author and source are credited. In fact, under many conditions, we find that transitions in sex determination are favoured
equally strongly in cases in which the sex ratio bias increases or decreases. Overall, our
Data Availability Statement: All relevant data are
within the paper and its Supporting Information models predict that sex determination systems should be highly dynamic, particularly when
files. haploid selection is present, consistent with the evolutionary lability of this trait in many taxa.
Funding: Natural Sciences and Engineering
Research Council of Canada (grant number CGS-D
6564 to MMO; RGPIN-2016-03711 to SPO). The
funder had no role in study design, data collection
and analysis, decision to publish, or preparation of Author summary
the manuscript. Systems of sex determination are strikingly diverse and labile in many clades. This poses
Competing interests: The authors have declared the question: what drives transitions between sex-determining systems? Here, we use
that no competing interests exist.
PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 1 / 25

Transitions between sex-determining systems
Abbreviations: ESD, environmental sex

determination; GSD, genetic sex determination. models to derive conditions under which new sex-determining systems spread. Prevailing
views suggest that new sex-determining systems are favoured when they equalise the sex
ratio and/or when they are more closely linked to genes that experience differential selec-
tion in males and females. Our models include selection upon haploid genotypes (meiotic
drive or gametic competition), which biases the sex ratio and occurs differently in male
and female gametes. Surprisingly, we find the two forces (selection to equalise the sex
ratio and the benefits of hitchhiking alongside driven alleles that distort the sex ratio) will
often be equally strong, and thus neither is sufficient to explain the spread of new sex-
determining systems in every case. We also find that new sex-determining alleles can
spread despite being less closely linked to selected loci, as long as initial linkage is tight or
haploid selection is present. Our models therefore predict that loci in previously unex-
pected genomic locations and/or experiencing various types of selection (including hap-
loid selection) can now be implicated as drivers of transitions between sex-determining
systems.
Introduction
Animals and angiosperms exhibit extremely diverse sex-determining systems (reviewed in [1–
5]). Among species with genetic sex determination (GSD), some taxa have heterogametic
males (XY) and homogametic females (XX), including mammals and most dioecious plants
[6], whereas other taxa have homogametic males (ZZ) and heterogametic females (ZW),
including Lepidoptera and birds. Within several taxa, the chromosome that harbours the mas-
ter sex-determining locus changes, due either to translocation of the master sex-determining
locus or to the evolution of a new master locus. During these transitions, the heterogametic sex
can remain the same (hereafter ‘cis-GSD transitions’) as in salmonids [7, 8], Diptera [9], and
Oryzias [10]. Alternatively, species can switch between male and female heterogamety
(XY$ZW, hereafter ‘trans-GSD transitions’), as in snakes [11], lizards [12], eight of 26 teleost
fish families [13], true fruit flies (tephritids, [9]), amphibians [14], the angiosperm genus Silene
[15], the angiosperm family Salicaceae [16, 17], and Coleoptera and Hemiptera (plate 2 [3]).
Indeed, in some cases, both male and female heterogametic sex-determining systems can be
found in the same species, as reported in houseflies [18], midges [19], frogs [20], cichlid fish
[21], tilapia [22], sea bass [23], and lab strains of zebrafish [24, 25]. In addition, multiple transi-
tions have occurred between genetic and environmental sex determination (ESD) systems
(GSD$ESD), e.g., in reptiles and fishes [5, 12, 13, 26–29]. In sum, accumulating evidence
indicates that transitions between sex-determining systems are common [4].
It has been suggested that sex ratio selection is a particularly dominant force in the evolu-
tion of sex determination (e.g., [1]; [3]). Classic ‘fisherian’ sex ratio selection favours a 1:1
zygotic sex ratio when assuming that males and females are equally costly to produce [30, 31].
This follows from the fact that, for an autosomal locus, half of the genetic material is inherited
from a male and half from a female [32]. Thus, if the sex ratio is biased, an individual of the
rarer sex will, on average, contribute more genetic material to the next generation. Selection
therefore typically favours mutants that increase investment in the rarer sex, including new sex
determination systems.
The evolution of sex determination is also thought to be strongly influenced by differences
in selection between the sexes [3, 33, 34]. For example, loci experiencing sexual antagonism
have been shown to favour the spread of new genetic sex-determining alleles that are closely
linked [35–37]. Linkage allows a stronger favourable association to build up between a male-

beneficial allele and a neo-Y allele, for example. Such associations can favour cis-GSD transi-
tions [35], trans-GSD transitions [36], and new partially masculinising or partially feminising
alleles in a population with ESD [37]. By similar logic, however, existing sexually antagonistic
alleles associated with the current sex-determining locus are expected to hinder the spread of a
new sex-determining system [35, 36].
One novel feature of the models developed here is that we explicitly consider the mainte-
nance of genetic variation around the ancestral sex-determining locus (e.g., within the nonre-
combining region of a sex chromosome). Counterintuitively, when linkage is tight between
the sex-determining locus and a selected locus, an allele good for females can be at higher fre-
quency on the ancestral Y than on the ancestral X under a variety of forms of selection. In
addition, selection on ancestral X chromosomes in males can prevent the X from becoming
optimally specialised for female-beneficial alleles. These factors, in turn, can favour a new ZW
sex-determining locus that has weaker linkage with loci under selection, which was not
revealed by previous theory [36]. A similar argument applies to ZW$XY transitions. Thus,
we show that selected loci in very tight linkage with the ancestral GSD locus can favour trans-
GSD transitions, during which linkage associations are actually weakened.
Most significantly, we include haploid selection (gametic competition or meiotic drive) in
models describing cis-GSD, trans-GSD, and GSD-to-ESD transitions. This poses an apparent
evolutionary problem. On one hand, haploid selection is typically sex limited in that it usually
occurs among gametes produced by one sex only [38–41]. Therefore, one might expect new
sex-determining systems to benefit from close linkage with haploid-selected loci, as found for
loci that experience diploid sex differences in selection [35–37]. On the other hand, associa-
tions between sex-determining loci and haploid-selected loci generate biased zygotic sex ratios,
which should generally hinder the spread of new sex-determining systems.
Two previous studies have considered the spread of GSD with sex-limited meiotic drive
[42, 43] under a limited number of possible genetic architectures and diploid-selective regimes.
Ubeda and colleagues [43] considered ancestral ESD (with no sex ratio bias) and numerically
showed that new GSD alleles can spread if they arise in linkage with meiotic drive loci. For
example, a masculinising allele spreads in association with an allele that is favoured during
male meiosis, causing sex ratios to become male biased. This suggests that the benefits of asso-
ciating with driving alleles can overwhelm selection to balance the sex ratio. However,
Kozielska and colleagues [42] considered an ancestral GSD system that is perfectly linked to a
meiotic driver and therefore exhibiting an ancestral sex ratio bias. They found that a new,
completely unlinked GSD system can spread if it generates the rarer sex, creating a balanced
sex ratio. This suggests that fisherian sex ratio selection can overwhelm the benefits of being
associated with driving alleles. It is thus currently unclear when haploid selection favours
increased versus decreased linkage between haploid-selected loci and a new sex-determining
locus. In addition, because the sex ratio is determined by linkage between haploid-selected loci
and the sex-determining locus, it is also unclear when fisherian sex ratio selection is the most
important driver of transitions between sex-determining systems.
Here, we analytically find the conditions under which new GSD or ESD systems spread in
ancestral GSD systems with any degree of linkage between the loci involved and arbitrary
forms of haploid and diploid selection. Doing so, we reconcile and generalise the results of
Kozielska and colleagues [42] and Ubeda and colleagues [43] by deriving conditions for the
spread of new GSD systems that alter linkage with haploid-selected loci. Our result is qualita-
tively distinct from those for diploid selection alone [35, 36] and suggests that haploid selection
is more likely to promote transitions between sex determination systems. We also show that
transitions involving haploid selection cannot be simply explained by invoking sex ratio
selection. In particular, under a wide range of conditions, we show that transitions in sex-

determining systems are favoured equally strongly in situations in which sex ratio biases
increase or decrease (and in situations in which sex ratio biases are ancestrally present or
absent). Finally, we show that ESD may not evolve, even if the sex ratio is initially biased by
haploid selection, which is not predicted by previous theories for transitions to ESD [1, 31, 32].
Together, our results suggest that both selection to equalise the sex ratio and the benefits of
associating with haploid-selected alleles can drive transitions between sex-determining sys-
tems, leading to stronger or weaker sex linkage and increased or decreased sex ratio bias.
Model
We consider transitions between ancestral and novel sex-determining systems using a three-
locus model, each locus having two alleles (Fig 1). A full description of our model, including
recursion equations, is given in S1 Text. Locus X is the ancestral sex-determining region, with
alleles X and Y (or Z and W). Locus A is a locus under selection, with alleles A and a. Locus M
is a novel sex-determining region, at which the null allele (M) is initially fixed in the population
such that sex of zygotes is determined by the genotype at the ancestral sex-determining region,
X; XX genotypes become females, and XY become males (or ZW become females, and ZZ
Fig 1. Outline of model features. Panel A: Recombination rate parameters between the ancestral sex-determining
locus (X, here assumed to have alleles X and Y), a locus under selection (A, with alleles A and a), and a new sex-
determining locus (M, with alleles M and m). Panel B: Haploid selection is often sex limited, occurring during haploid
production or competition in one sex (shown here in males by dashed circles). If X or Y alleles are linked with alleles
that experience haploid selection in males (r < 1/2), then zygotic sex ratios can become biased because either X- or Y-
bearing male gametes/gametophytes will be more abundant after haploid selection. Similarly, zygotic sex ratio biases
can arise if haploid-selected alleles are linked with new sex-determining alleles (R < 1/2). However, the zygotic sex
ratio is not biased by male haploid selection in ZW sex-determining systems. Panel C: During cis-GSD transitions (XY
to XY or ZW to ZW), a neo-Y allele (m) spreads until all males bear the neo-Y, and the ancestral Y allele is lost. Panel
D: During trans-GSD transitions (XY to ZW or ZW to XY), a neo-W allele (m) spreads until all females bear the neo-
W, and the ancestral X allele is lost. Neo-W alleles allow Y-associated alleles into females, which may impede or aid
their spread. GSD, genetic sex determination.
https://doi.org/10.1371/journal.pbio.2005609.g001

become males). To evaluate the evolution of new sex-determining systems, we consider the
spread of a novel sex-determining allele (m) at the M locus.
Here, we assume that the M locus is ‘epistatically dominant’ over the X locus such that
zygotes with at least one m allele develop as females with probability k and as males with prob-
ability 1 − k, regardless of the X locus genotype. With k = 0, the m allele is a masculiniser (a
neo-Y allele), and with k = 1, the m allele is a feminiser (a neo-W allele). With intermediate k,
we can interpret m as an ESD allele, such that zygotes develop as females in a proportion (k) of
the environments they experience. The assumption that derived sex-determining loci are epis-
tatically dominant is motivated by empirical systems in which multiple sex-determining alleles
segregate (i.e., X, Y, Z, and W alleles present), such as cichlid fish [21], platyfish (Xiphophorus
maculatus [44]), houseflies (Musca domestica [45]), western clawed frogs (Xenopus tropicalis
[46]), and Rana rugosa [20]. Nevertheless, our supplementary analysis file (S1 File) allows
other dominance relationships between loci to be specified (see also [35] supplementary mate-
rial for a numerical analysis).
We consider two forms of selection upon haploid genotypes, ‘gametic competition’ and
‘meiotic drive’. During gametic competition, we assume that a representative sample of all
gametes/gametophytes (hereafter ‘gametes’) compete with others of the same sex for fertilisa-
tion, which implies a polygamous mating system. Relative fitnesses in sex ◯ 2 {♀,♂} during
gametic competition are given by w◯ ◯
A and w a (see Table 1). On the other hand, meiotic drive
in our model only affects the segregation of gametes produced by heterozygotes. Specifically,
gametes produced by Aa heterozygotes of sex ◯ bear allele A with probability α◯. We note
that competition between sperm produced by a single male (e.g., in a monogamous mating
system) would be appropriately modelled as male meiotic drive, as only the frequency of gam-
etes produced by heterozygotes would be affected. However, we do not consider scenarios in
which there is competition among gametes produced by a small number of males/females
(e.g., [47]).
In each generation, we census the genotype frequencies in male and female gametes before
gametic competition. After gametic competition, conjugation between male and female gam-
etes occurs at random. The resulting zygotes develop as males or females, depending on their
genotypes at the X and M loci. Diploid males and females then experience viability and/or
individual-based fertility selection, with relative fitnesses w◯ ◯ ◯
AA , wAa , and waa . We do not con-
sider fertility selection that depends on the mating partner, e.g., sexual selection with variation
in choosiness. The next generation of gametes is produced by meiosis, during which recombi-
nation and sex-specific meiotic drive can occur. Recombination (i.e., an odd number of cross-
overs) occurs between loci X and A with probability r, between loci A and M with probability
Table 1. Relative fitness of different genotypes in sex, ◯ 2 {♀,♂}.

Genotype Relative fitness during gametic competition
A w◯
A ¼ 1þt
◯
a w◯
a ¼ 1
Genotype Relative fitness during diploid selection

AA w◯
AA ¼ 1 þ s
◯
Aa w◯ ◯ ◯
Aa ¼ 1 þ h s
aa w◯
aa ¼ 1
Genotype Transmission during meiosis in Aa heterozygotes

A a◯ ¼ 1=2 þ a◯
D =2
a 1 a◯ ¼ 1=2 a◯
D =2
https://doi.org/10.1371/journal.pbio.2005609.t001

R, and between loci X and M with probability ρ. Any linear order of the loci can be modelled
with appropriate choices of r, R, and ρ (see Fig 1A and S1 Table). Our model is entirely deter-
ministic and hence ignores chance fluctuations in allele frequencies due to genetic drift.
The model outlined above describes both ancestral XY and ZW sex-determining systems.
Without loss of generality, we refer to the ancestrally heterogametic sex as male and the
ancestrally homogametic sex as female. That is, we primarily describe an ancestral XY sex-
determining system, but our model is equally applicable to an ancestral ZW sex-determining
system (relabelling the ancestrally heterogametic sex as female and the ancestrally homoga-
metic sex as male and switching the labels of males and females throughout). We use a super-
script to specify the ancestral sex-determining system described, e.g., (XY) for ancestral XY sex-
determination.
In the ancestral population, it is convenient to follow the frequency of the A allele among
female gametes (eggs), p♀X , and among X-bearing, p♂X , or among Y-bearing, p♂Y , male gametes
(sperm/pollen). We also track the fraction of male gametes that are Y-bearing, qY, which may
deviate from 1/2 because of meiotic drive in males. We consider only equilibrium frequencies
of alleles, p^◯ ^Y , when determining the invasion of new sex-
i , and Y-bearing male gametes, q
determining factors. We use ξ to measure the sex ratio (fraction male) among zygotes, which is
determined by the allele frequencies and haploid selection coefficients (S2 Table).
Results
We begin by describing the general conditions under which new GSD alleles can spread within
a population, without explicitly specifying ancestral allele frequencies. These general condi-
tions then allow us to consider several special cases of interest in subsequent sections, in which
equilibrium ancestral allele frequencies are explicitly calculated. Finally, we consider the spread
of alleles that cause sex to be determined environmentally.
Generic invasion by a neo-Y or neo-W

The evolution of a new sex-determining system requires that a rare mutant allele, m, at the
novel sex-determining locus, M, increases in frequency when rare. Specifically, m invades
when lðXY
m
Þ
> 1, in which lðXY
m
Þ
is the leading eigenvalue of the system of eight equations
describing m-bearing gamete frequencies, Eqs S1.1. This system simplifies substantially for an
epistatically dominant neo-Y (k = 0) or neo-W (k = 1); see S3 Text for details.
Invasion by a neo-Y or a neo-W primarily depends on the ‘haplotypic growth rates’
(denoted by LðXY Þ
mi ) of the neo-sex determination allele m on background i 2 {A,a}, without
accounting for loss due to recombination (R = 0); see Table 2. If both haplotypic growth rates
are greater than 1 (LðXY Þ ðXY Þ
mA , Lma > 1), then the new sex-determining allele invades regardless of
the rate of recombination between the new sex-determining locus and the selected locus (R).
Conversely, if both haplotypic growth rates are less than 1 (LðXY Þ ðXY Þ
mA , Lma < 1), then invasion can
never occur. Finally, if only one haplotypic growth rate is greater than 1, the new sex-deter-
mining allele can always invade when arising at a locus that is tightly linked to the selected
locus (R 0). Furthermore, it can be shown that the leading eigenvalue declines with recombi-
nation rate, R, and invasion requires that R is sufficiently small such that
wðXY Þ ðXY Þ
ma =ðLma 1Þ þ wðXY Þ ðXY Þ
mA =ðLmA 1Þ < 1: ð1Þ
Here, wðXY
mi
Þ
> 0 is the rate at which mutant haplotypes on background i 2 {A,a} recombine
onto the other A locus background in heterozygotes (which is proportional to R; see Table 2).
This is a ‘dissociative force’ that breaks down linkage disequilibrium.

Table 2. Parameters determining invasion of mutant neo-Y and neo-W alleles into an ancestrally XY system.
m is a neo-Y (k = 0)
1
LðXY Þ
^♀ ♀ ♂ ♂
Y 0 A ¼ ð2zÞ ½p X wA wA wAA þ ð1 p^♀X Þw♀a w♂A w♂Aa ð1 þ a♂4 Þ=ðw
♀H w
♂H w
♂D Þ
ðXY Þ 1
L Y0 a ¼ ð2zÞ ½ð1 p^ Þw w w þ p^♀X w♀A w♂a w♂Aa ð1
♀
X
♀
a
♂
a
♂
aa a♂4 Þ=ðw
♀H w
♂H w
♂D Þ
1
wðXY Þ
Y 0 A ¼ Rð2zÞ ½ð1 p^♀X Þw♀a w♂A w♂Aa ð1 þ a♂4 Þ=ðw
♀H w
♂H w
♂D Þ
ðXY Þ 1
wY0 a ¼ Rð2zÞ ½p^♀X w♀A w♂a w♂Aa ð1 a♂4 Þ=ðw
♀H w
♂H w
♂D Þ
m is a neo-W (k = 1)
1
LðXY Þ
W 0 A ¼ ½2ð1 zÞ ½p♂ w♂A w♀A w♀AA þ ð1 p♂ Þw♂A w♀A w♀Aa ð1 þ a♀D Þ=ðw
♀H w
♂H w
♀D Þ
1
LðXY Þ
W 0 a ¼ ½2ð1 zÞ ½ð1 p♂ Þw♂a w♀a w♀aa þ p♂ w♂A w♀a w♀Aa ð1 a♀D Þ=ðw
♀H w
♂H w
♀D Þ
ðXY Þ 1
wW0 A ¼ R½2ð1 zÞ ½ð1 ♂ ♂
a
♀
A
♀
p Þw w w ð1 þ a Þ=ðw
Aa w
w ♀
♀D Þ
D
♀
H
♂
H
1
wðXY Þ
W 0 a ¼ R½2ð1 zÞ ½p♂ w♂A w♀a w♀Aa ð1 a♀D Þ=ðw
♀H w
♂H w
♀D Þ
p♂ ¼ ð1 q^Y Þp^♂X þ q^Y p^♂Y is the average frequency of the A allele among X- and Y-bearing male gametes. q^Y is the
◯
frequency of Y-bearing male gametes. ξ is the zygotic sex ratio (fraction male). w D is the mean fitness of diploids of
◯
sex ◯ 2 {♀,♂}. w H is the mean fitness of haploids from sex ◯ (see S2 Table). R is the rate of recombination between
◯
the neo-sex determiner and the selected locus. Selection terms (w◯
i , aD ) are described in Table 1.
Condition 1 may or may not be satisfied for the full range of locations of the new sex-deter-
mining locus, including R = 1/2 (e.g., on an autosome), depending on the nature of selection.
Interpreting this condition, if we assume that only the mA haplotype would increase in fre-
quency when R = 0 (i.e., LðXY Þ ðXY Þ
ma < 1 < LmA ), then the first term on the left-hand side of Eq (1)
is negative, and invasion requires that the growth rate of mA haplotypes (LðXY
mA
Þ
1 > 0) and
ðXY Þ
the rate at which they are produced by recombination (wma ) are sufficiently large relative to
the rate of decline of ma haplotypes (1 LðXY Þ
ma > 0) and the rate at which m and A are dissoci-
ated by recombination (wðXY Þ

mA ).
The haplotypic growth rates and dissociative forces are listed in Table 2 for a neo-Y and
neo-W invading an ancestrally XY system. From this table and the arguments above, we draw
four main points about the generic invasion of neo-Y and neo-W mutations (without specify-
ing the ancestral equilibrium): (1) Fisherian sex ratio selection will favour the spread of a neo-
W and inhibit the spread of a neo-Y if the ancestral zygotic sex ratio is biased towards males
(i.e., the first factor of the LðXY Þ
mi is greater than 1 for a neo-W and less than 1 for a neo-Y when
ξ > 1/2). Thus, neo-sex-determining alleles that specify the rarer sex are favoured by fisherian
sex ratio selection. (2) In addition, the new sex-determining allele has associations with alleles
favoured by either haploid or diploid selection (fitness terms in square brackets). Importantly,
invasion by a neo-Y (neo-W) does not directly depend on the fitness of female (male) diploids.
This is because a dominant neo-Y (neo-W) is always found in males (females), and therefore
the frequency of the neo-Y (neo-W), m, only changes in males (females), Fig 1C and 1D. (3)
Haploid selection thus plays two roles, generating fisherian selection to equalise the ancestral
sex ratio (through ξ) and generating selection for the neo-Y/neo-W through associations with
haploid-selected loci, which can distort the sex ratio. Each role influences the invasion dynam-
ics of a new sex-determining allele, allowing the sex ratio to become more or less biased during
a transition (as previously found in two special cases; [42, 43]). (4) Finally, Table 2 shows that
the genetic contexts that arise during cis- and trans-GSD transitions are qualitatively different.
This is because, in an ancestrally XY system, a gamete with the neo-Y always pairs with a
female gamete containing an X, Fig 1C. By contrast, a gamete with a neo-W can pair with an
X- or Y-bearing male gamete, Fig 1D. Consequently, neo-W-bearing females obtain a different

frequency of A alleles from mating compared to ancestral (MM) females (p♂ versus p^♂X , respec-
tively). This can inhibit or favour the spread of a neo-W.
In order to explicitly determine the conditions under which a new sex-determining allele
spreads, we next calculate the equilibrium frequency of the A allele (i.e., p^♀X , p^♂X , and p^♂Y ) and
Y-bearing male gametes (^ q Y ) in the ancestral population. Because only the A locus experi-
ences selection directly, any deterministic evolution requires that there be a polymorphism at
the A locus. Polymorphisms can be maintained by mutation-selection balance or occur tran-
siently during the spread of beneficial alleles. Here, however, we focus on polymorphisms
maintained by selection for longer periods. Such polymorphisms can be maintained by het-
erozygote advantage, sexually antagonistic selection, ploidally antagonistic selection, or a
combination [48]. We analytically calculate equilibrium frequencies using two alternative
simplifying assumptions: (1) the A locus is tightly linked to the nonrecombining region
around the ancestral sex-determining locus (r 0), or (2) selection is weak relative to recom-
bination (s◯, t◯, a◯ D r). The ancestral equilibria and their stability conditions are given in
S2 Text.
Tight linkage with the ancestral sex-determining locus (r 0)

When there is complete linkage between the ancestral sex-determining locus and the selected
locus A (r = 0), either the A allele or the a allele must be fixed in gametes containing a Y allele
(S2 Text). Because the labelling of alleles is arbitrary, we will assume that the a locus is fixed in
gametes with a Y (p^♂Y ¼ 0), without loss of generality. If there are two alleles maintained at the
A locus, the A allele can be fixed (p^♀X ¼ p^♂X ¼ 1) or segregating at an intermediate frequency
(0 < p^♀X ;p^♂X < 1) in gametes with an X.
We find that a neo-Y allele can never invade an ancestral XY system that already has tight
linkage with the locus under selection (lðXY Y0
Þ
1 when r = 0; for details, see S1 File). In essence,
through tight linkage with the A locus, the ancestral Y becomes strongly specialised on the
allele that has the highest fitness across male haploid and diploid phases. It is thus not possible
for a neo-Y to create males that have higher fitness than the ancestral Y, and cis-GSD transi-
tions are never favoured.
Neo-W alleles, on the other hand, can invade an ancestral XY system (the full invasion con-
ditions are given in S3 Text; Eqs S3.1 and S3.2). Invasion occurs when neo-W females can have
higher fitness than the XX females in the ancestral population. Neo-W invasion is possible
under all forms of selection that can maintain a polymorphism (sexually antagonistic selection,
overdominance, ploidally antagonistic selection, or some combination, e.g., S2, S3 and S8
Figs). Thus,
Conclusion 1: Selection on loci in or near the nonrecombining region around the ancestral
sex-determining locus (r 0) prevents cis-GSD transitions (XY $ XY, ZW $ ZW) but
can spur trans-GSD transitions (XY $ ZW).
To clarify conditions under which trans-GSD transitions can occur, we focus here on cases
in which there is no haploid selection (and hence no sex ratio bias) and discuss the additional
effect of haploid selection in S3 Text. Broadly, it is possible for neo-W females to have higher
fitness than XX females for two reasons. Firstly, because the ancestral X experiences selection
in both males and females, the X may be unable to specialise strongly on an allele favoured in
females. Secondly, an allele can be associated with the Y and yet favoured in females. In turn, a
neo-W can spread because (a) it is only found in females and is therefore unleashed from
counterselection in males (corresponding to LðXY Þ
W 0 A > 1), and/or (b) it allows Y-associated
alleles into females (corresponding to LðXY Þ

W 0 a > 1).

Fig 2. When the ancestral XY locus is tightly linked to a locus under selection (r 0), one or both neo-W haplotypes can spread (no haploid selection). We vary
the fitness of male homozygotes relative to heterozygotes (w◯ Aa ¼ 1) and only consider stable equilibria at which both A locus alleles are maintained and the a allele is
initially fixed on the Y (nonhatched region). Here, selection in females can favour the A allele (panel A, w♀aa ¼ 0:85, w♀AA ¼ 1:05), favour the a allele (panel B,
w♀aa ¼ 1:05, w♀AA ¼ 0:85), or be overdominant (panel C, w♀aa ¼ w♀AA ¼ 0:6). If either haplotypic growth rate (LðXY Þ ðXY Þ
W 0 A or LW 0 a ) is greater than 1, then a rare neo-W allele
can spread for at least some values of R > r (grey regions). The parameter values marked with an asterisk correspond to the fitnesses used in Fig 3C. S3 Fig shows the
dynamics arising with the parameters marked with a dagger.
We first give an example in which neo-W-A haplotypes can spread because the neo-W is
unleashed from counterselection in males (case [a], in which LðXY Þ
W 0 A > 1). When A is female
beneficial and a is male beneficial, the A allele can be fixed (p^♀X ¼ p^♂X ¼ 1) or polymorphic
(0 < p^♀X ;p^♂X < 1) on the X. In this case, polymorphism on the ancestral X indicates suboptimal
specialisation for females fitness, which occurs because the A allele is counterselected in males
(requires that w♂Aa be sufficiently small relative to w♂aa ). Neo-Ws, however, spend no time in
males and can build stronger associations with the female-beneficial A allele, allowing them to
spread (see grey region in Fig 2A).
We next give an example in which neo-W-a haplotypes can spread because they bring in
female-beneficial alleles associated with the Y (case [b], in which LðXY Þ
W 0 a > 1). When there is
overdominance in males, X-A Y-a males have high fitness, and the A allele is favoured by selec-
tion on the X background in males. Therefore, the A allele can be polymorphic or even fixed
on the X background, despite selection favouring the a allele in females (e.g., see nonhatched
region in Fig 2B and [49, 50]). In such cases, neo-W-a haplotypes can spread because they cre-
ate more Aa and aa females when pairing with an X-bearing gamete from males and because
they bring more of the Y-a haplotype into females, in whom it has higher fitness (Fig 1D).
In some cases, both neo-W-A and neo-W-a haplotypes can spread. For example, when AA
individuals have low fitness in females, yet the A is polymorphic or fixed on the X background
due to overdominance in males (Fig 2B and 2C), both neo-W-A and neo-W-a haplotypes pro-
duce fewer unfit AA females. This is true for the neo-W-A haplotype because it can pair with a
Y-a haplotype and still be female. Whenever both haplotypic growth rates are greater than 1,
invasion by a neo-W is expected regardless of its linkage with the selected locus (i.e., for any
R); see S1 and S2 Figs for examples. As a consequence, evolution can favour a new sex determi-
nation system on a different chromosome (R = 1/2), despite the fact that this unlinks the sex-
determining locus from the selected locus.

When only one neo-W haplotype has a growth rate greater than 1 (see Fig 2), a neo-W allele
can invade as long as Eq (1) is satisfied, which may require that the recombination rate, R, is
small enough. Nevertheless, because we assume here that r is small, these results indicate that a
more loosely linked sex-determining region (r < R) can spread. For example, tightly sex-linked
loci that experience sexually antagonistic selection can drive trans-GSD transitions in which
the new sex-determining locus is less closely linked (R > r, Fig 3), but the analysis in S1 File
indicates that a new unlinked sex-determining allele (R = 1/2) cannot invade when selection is
purely sexually antagonistic (directional selection in each sex and no haploid selection).
Assuming selection is weak relative to recombination, van Doorn and Kirkpatrick [36]
showed that invasion by a neo-W allele occurs under the same conditions as its fixation in
females. An equivalent analysis is not possible when recombination rates are low. However,
numerical simulations demonstrate that, with tight sex linkage, neo-Y or neo-W alleles do not
necessarily reach fixation in males or females, respectively, which can lead to the stable mainte-
nance of a mixed sex-determining system, in which X, Y, and neo-W alleles all segregate (e.g.,
S9B and S9C Fig).
From the arguments above, we reach
Conclusion 2: With tight linkage between a selected locus and the ancestral sex-determin-
ing locus (r 0), trans-GSD transitions (XY $ ZW) can be favoured by selection even if
they weaken sex-linkage (r < R), potentially shifting sex determination to a different chro-
mosome (R = 1/2). Such transitions can also lead to the maintenance of multifactorial sex
determination systems.
With haploid selection, Conclusions 1 and 2 continue to apply (S3 Text). The parameters
for which neo-W-A and neo-W-a haplotypes spread under various forms of haploid selection
are plotted in S4, S5, S6 and S7 Figs. In particular, we note that adding haploid selection allows
shifts in sex determination to a different chromosome (R = 1/2) even when selection is sexually
antagonistic, with directional selection in each diploid sex, e.g., S3 Fig. Furthermore, haploid
selection allows variation to be maintained by ploidally antagonistic selection, under which
trans-GSD transitions may also be favoured, S8 Fig. Some cases of XY ! ZW transitions in
which r = 0, R = 1/2, and selection is ploidally antagonistic (meiotic drive in males opposed by
Fig 3. Transitions between XY and ZW systems can occur even when the new sex-determining locus is less tightly linked to a locus under sexually antagonistic
selection (no haploid selection). In panel A, linkage is initially tight relative to selection, and a neo-W can invade (above dashed line) even when it is less tightly linked
with the selected locus (r < R; unshaded region around ). In panel B, linkage is loose enough relative to selection that the analytical results assuming weak selection
hold, and a neo-W allele can only invade when it arises at a locus more tightly linked with the selected locus (R < r; shaded region). In panel C, we vary the
recombination rate between the neo-W and the selected locus (R) for a fixed recombination rate between the ancestral sex-determining locus and the selected locus
(r = 0.005). Coloured markers show recombination rates for which the temporal dynamics of invasion are plotted in the inset (frequency of females carrying a neo-W),
demonstrating that neo-W alleles can reach fixation if they are more (black) or less (red) closely linked to a locus experiencing sexually antagonistic selection. A very
loosely linked neo-W does not spread in this case (blue and green lines overlap and go to 0 in inset). Fitness parameters are w♀AA ¼ 1:05, w♂aa ¼ 1:2, w♀aa ¼ w♂AA ¼ 0:85,
w◯Aa ¼ 1. SDR, sex-determining region.

diploid selection) were studied by Kozielska and colleagues [42], who found that sex ratio
biases are reduced during these transitions. However, such transitions are not always driven
by selection to reduce sex ratio bias. For example, with XY sex determination and haploid
selection in females, sex ratios are not ancestrally biased, yet a neo-W can invade (S8 Fig). We
further discuss how the spread of neo-sex-determining alleles is influenced by associations
with haploid-selected loci in the next section.
Loose linkage with the ancestral sex-determining region

Here, we assume that selection is weak (s◯, t◯, a◯ D of order ε, in which ε is some number
much less than 1) and thus implicitly assume that all recombination rates (r, R, and ρ) are large
relative to selection. To leading order in selection,
1 ðr RÞ
lðXY
Y0
Þ
¼ 1 þ pð1 pÞSA 2 þ Oðε3 Þ ð2Þ
4 rR
and
lðXY Þ ðXY Þ
W 0 ¼ lY 0 þ ½ð2a♂D 2a♀D þ t ♂ t ♀ Þðp^♂Y p^♂X Þ=2 þ Oðε3 Þ; ð3Þ
in which p is the frequency of A, to leading order (Eq S2.3), and SA ¼ ðs ♂ þ a♂D þ t ♂ Þ ðs ♀ þ
a♀D þ t ♀ Þ describes sex differences in selection for the A versus a allele across diploid
selection, meiosis, and gametic competition. The diploid selection term,
s ◯ ¼ ½ps◯ þ ð1 pÞh◯ s◯ ½ph◯ s◯ þ ð1 pÞ, is the difference in fitness between A and a
alleles in diploids of sex ◯ 2 {♀,♂}. The difference in A allele frequency among Y-bearing
sperm versus X-bearing sperm is, at equilibrium, p^♂Y p^♂X ¼ pð1 pÞSA ð1 2rÞ=ð2rÞ.
Eq (2) demonstrates that, under weak selection, a neo-Y allele will invade an XY system
ðXY Þ
(lY 0 > 1) if and only if it is more closely linked to the selected locus than the ancestral sex-
determining locus (i.e., if R < r). This echoes our results above, in which a neo-Y could never
invade if r 0. It is also consistent with the results of [35], who considered diploid selection
only and also found that cis-GSD transitions can only occur when the new sex-determining
locus is more closely linked to a locus under sexually antagonistic selection.
Conclusion 3A: New sex-determining alleles causing cis-GSD transitions (XY $ XY or
ZW $ ZW) are favoured if they arise more closely linked with a locus that experiences
(haploid and/or diploid) selection than the ancestral sex-determining locus (R < r).
Similarly, in the absence of haploid selection (t ◯ ¼ a◯ D ¼ 0), Eq (3) indicates that trans-
GSD transitions can occur if and only if the new sex-determining locus is more closely linked
to a locus under selection, R < r, as found by [36]. With haploid selection, a neo-W is also usu-
ally favoured when it is more closely linked to the selected locus than the ancestral sex-deter-
mining region is (R < r, e.g., Figs 3B and 4); this is true unless the last term in Eq (3) is
negative and dominant over the first, which requires relatively restrictive combinations of
selection and recombination parameters. For example, with haploid selection, a neo-W will
always be favoured if it arises in linkage with a selected locus (R < 1/2) that is ancestrally auto-
somal (r = 1/2, leading to p^♂Y p^♂X ¼ 0).
Conclusion 3B: New sex-determining alleles causing trans-GSD transitions (XY $ ZW)
are usually favoured if they arise more closely linked with a locus that experiences (haploid
and/or diploid) selection than the ancestral sex-determining locus (R < r).
However, with haploid selection and some ancestral sex linkage (r < 1/2; allowing allele fre-
quency differences on the X and Y), the term in square brackets in Eq (3) can be positive. This
leads to

Fig 4. Ploidally antagonistic selection allows a less tightly linked neo-W allele to invade. In panel A, male drive
(a♂D ¼ 1=20, t ◯ ¼ a♀D ¼ 0) opposes selection in diploids (no sex differences: s◯ = 1/10, h◯ = 7/10). In panel B,
gametic competition in males (t♂ = −1/10, t ♀ ¼ a◯ ♂ ♀
D ¼ 0) opposes selection in diploids (sex differences: s = 3/20, s =
◯
1/20, h = 7/10). In either case, the new sex-determining allele can invade (above dashed line) regardless of R, even
when linkage to the selected locus is reduced (white regions). SDR, sex-determining region.
Conclusion 3C: With haploid selection, new sex-determining alleles causing trans-GSD
transitions (XY $ ZW) can spread even if they arise further from a locus that experiences
selection than the ancestral sex-determining locus (r < R).
To clarify the parameter space under which neo-W alleles spread despite looser linkage
with the selected locus (R > r), we focus on cases in which dominance coefficients are equal in
the two sexes, h♀ = h♂, and haploid selection only occurs in one sex (e.g., during male meiosis
only). Table 3 then gives the conditions required for unlinked (R = 1/2) neo-W invasion when
there is some ancestral sex linkage (r < 1/2; e.g., the selected locus is on the ancestral sex chro-
mosome, and the novel sex-determining locus arises on an autosome). These special cases
indicate that neo-W invasion occurs for a large fraction of the parameter space, even though
the neo-W uncouples the sex-determining locus from a locus under selection. Fig 4 then dem-
onstrates that under these conditions, neo-W alleles can spread when they are more loosely or
more closely linked to the locus that experiences haploid selection, i.e., Conclusions 3B and 3C
(compare with Fig 3A for diploid sexually antagonistic selection alone).
We can also compare transitions among different GSD systems, as these are associated
with different effects on the sex ratio, which can increase, decrease, or remain equal. For
example, if there is meiotic drive in males only (a♂D 6¼ 0, a♀D ¼ 0), without gametic competition
(t♀ = t♂ = 0), the zygotic sex ratio is initially biased only when the ancestral sex-determining
system is XY (Figs 1B and 5A) and not ZW (Figs 1B and 5B). If fisherian sex ratio selection
dominated, we would thus expect a difference in the potential for XY-to-ZW and ZW-to-XY
transitions. However, invasion by a neo-W allele into an XY system and invasion by a neo-Y
Table 3. Invasion conditions for a neo-W allele at an unlinked locus (R = 1/2) into an ancestral XY system with
linkage (r < 1/2) and a single form of haploid selection.
Scenario Assumptions neo-W spreads (lðXY Þ
W 0 > 1) if
male drive only h♂ = h♀, t ♀ ¼ t ♂ ¼ a♀D ¼ 0 s♀ s♂ > 0

female drive only ♂ ♀ ♀ ♂
h =h ,t ¼t ¼a ¼0 ♂
D
s♀ s♂ > 0
male gametic competition only ♂ ♀ ♀
h =h ,t ¼a ¼a ¼0♀
D
♂
D
s♀(s♂ − s♀) > 0
female gametic competition only ♂ ♀
h = h , t ♂ ¼ a♀D ¼ a♂D ¼ 0 s♂(s♀ − s♂) > 0

Fig 5. Fisherian sex ratio selection alone is not a good predictor of turnover between sex-determining systems. In this figure, selection is ploidally
antagonistic, with haploid selection favouring the a allele during male meiosis. In panel A, male meiotic drive in an ancestral XY system causes a male bias (see
Fig 1B), allowing a neo-W to invade and replace the ancestral sex-determining system (inset shows the frequency of females carrying a neo-W), which balances
the zygotic sex ratio. In panel B, male drive in an ancestral ZW system has no effect on the zygotic sex ratio (50:50 at generation 0), yet a neo-Y can invade and
replace the ancestral sex-determining system (inset shows the frequency of males carrying a neo-Y). Parameters: s♀ = s♂ = 0.2, h♀ = h♂ = 0.7, t ♀ ¼ t ♂ ¼ a♀D ¼ 0,
a♂D ¼ 0:1, r = 0.02.
allele into a ZW system occur under the same conditions (lðXY Y0

Þ
¼ lðZWÞ
W0 and lðXY Þ ðZWÞ
W 0 ¼ lY 0 , at
least to order ε ), implying that
2
Conclusion 4: When selection is weak relative to recombination, trans-GSD transitions in

the presence of haploid selection are favoured as often and as strongly, whether they erase
ancestral sex ratio bias (benefiting from fisherian sex ratio selection) or generate sex ratio
bias (benefiting from associations with selected alleles).
For example, in Fig 5A, neo-W alleles invade an ancestral XY system in which females are
initially rare, equalising the sex ratio (as occurs in [42]). However, Fig 5B shows that a neo-Y
can invade the resulting ZW system under the same conditions. When R < 1/2, the invading
neo-Y becomes associated with the male meiotic drive allele, and the zygotic sex ratio evolves
to become male biased (as occurs in [43], beginning from ESD). In this case, the neo-Y spreads
because it is often found in males and can, if it carries the driven allele a, benefit from haploid
selection in males (Fig 5B).
While equalising the sex ratio and benefiting from associations with selected alleles are two
primary reasons why haploid selection spurs sex chromosome transitions, more complex situ-
ations also arise. For example, with R = 1/2 in Fig 5B (green curve), the neo-Y allele spreads
despite the fact that it cannot benefit from drive because free recombination moves it ran-
domly between driven and nondriven backgrounds. Nevertheless, the unlinked neo-Y can
spread because males bearing it more often carry the nondriven allele A and have higher aver-
age diploid fitness compared to ZZ males, which bear a high frequency of the driven allele a
from their mothers.
Transitions to environmentally determined sex

We next consider the case in which the new sex-determining allele, m, causes sex to be deter-
mined probabilistically or by heterogeneous environmental conditions (ESD). In particular,

we assume individuals carrying allele m develop as females with probability k 2 (0,1). In our
deterministic model, this means the fraction of females in the subpopulation containing m is
exactly k, even when m is rare (i.e., ESD does not introduce any additional variance in sex
determination). We also assume that the environmental conditions that determine sex do not
differentially affect the fitness of males versus females. Such correlations can favour environ-
mental sex-determining systems by allowing each sex to be produced in the environment in
which it has highest fitness; in the absence of these correlations, previous theory would predict
that ESD is favoured when it produces more equal sex ratios than the ancestral system (see
reviews by [1, 31, 32]).
The characteristic polynomial determining the leading eigenvalue (Eqs S1.1) does not factor
for ESD (0 < k < 1) as it does for a neo-Y (k = 0) or neo-W (k = 1) allele. We therefore focus
on weak selection here, in which case the leading eigenvalue is
2
ð1 2kÞ r R
lðXY Þ
ESD0 ¼ 1þ pð1 pÞSA 2
4 rR ð4Þ
kðp^♂Y p^♂X Þ
þ ½kð2a♂D 2a♀D þ t ♂ ♀
t Þ 2ð1 3
kÞSA þ Oðε Þ:
2
This reduces to lðXY

Y0
Þ
when k = 0, and lðXY Þ
W 0 when k = 1. Of particular interest are ESD muta-
tions that cause half of their carriers to develop as females and half as males (k = 1/2), creating
equal sex ratios. The spread of such mutations is determined by
ðXY Þ
1 ðlY 0 jR¼1=2 1Þ þ ðlðXY Þ
W 0 jR¼1=2 1Þ
lðXY Þ
ESD0 ¼ 1 þ þ Oðε3 Þ; ð5Þ
2 2
in which lðXY Þ ðXY Þ

Y 0 jR¼1=2 and lW 0 jR¼1=2 represent lY 0
ðXY Þ
and lðXY Þ
W 0 when evaluated at R = 1/2 (Eqs 2 and
3). That is, ESD with k = 1/2 behaves as if the M and A loci were unlinked, regardless of the
actual value of R. This is because sex is randomised each generation in individuals bearing the
m allele, preventing associations from building up between it and alleles at locus A. Eq (5)
shows that the ESD mutation gets half of the fitness of a feminising mutation (neo-W) and half
of the fitness of a masculinising mutation (neo-Y) but only has an effect one-half of the time
(the other half of the time it produces the same sex as the ancestral system would have). As dis-
cussed above, lðXY Þ
Y 0 jR¼1=2 is necessarily less than or equal to 1 when selection is weak (Conclusion
3A), but lðXY Þ

W 0 jR¼1=2 can be greater than 1 if there is haploid selection (see Conclusion 3C). That
is, with haploid selection, an allele causing ESD can invade an ancestrally XY system because it
generates females that are either rare or have high fitness, in the same manner as a neo-W
(likewise, ESD invades a ZW system for the same reasons that a neo-Y can).
Significantly, Eq (5) is the same whether ESD is invading an ancestrally XY or ZW system
(because lðXY
Y0
Þ
¼ lðZWÞ
W0 and lðXY Þ ðZWÞ
W 0 ¼ lY 0 ). Thus, focusing solely on fisherian selection to equal-
ise the sex ratio does not fully explain GSD-to-ESD transitions. For example, when the ances-
tral sex-determining system is XY, the sex ratio is biased by male haploid selection. When the
ancestral sex-determining system is ZW, the sex ratio is not biased. Nevertheless, ESD is
equally likely to invade both XY (through lðXY Þ
W 0 ) and ZW (through lY 0
ðZWÞ
) systems, equalising
the zygotic sex ratio in the former case but potentially transiently biasing it in the latter. In
addition, we note that ESD may not invade, even if the sex ratio is initially biased (e.g., with
drive in males only, r < 1/2, h♀ = h♂, and s♀ s♂ < 0, then lðXY Þ
W 0 < 1, see Table 3). We conclude
that, as with neo-W and neo-Y mutations,
Conclusion 5: Transitions from GSD to ESD are not straightforwardly predicted by selec-
tion to balance the zygotic sex ratio when haploid selection is present.

Discussion
New sex determination systems are typically expected to spread when they equalise the sex
ratio and/or when they increase linkage with loci that experience sex differences in selection
[33, 34]. In accordance with the latter mechanism, we find that sex differences in selection at
the haploid stage can favour cis- or trans-GSD transitions that tighten sex linkage (Conclusion
3A and 3B). Contrary to this expectation, however, we find that trans-GSD transitions can be
favoured that loosen linkage with the sex-determining locus, either when linkage is initially
tight (Conclusions 1 and 2, Figs 2 and 3) or when there is haploid selection (Conclusion 3C,
Figs 4 and 5). Furthermore, we show that the spread of new sex determination systems is not
dominated by selection to balance the sex ratio (Conclusions 4 and 5, Fig 5).
On the one hand, sex ratio biases caused by haploid selection can facilitate trans-GSD transi-
tions or transitions from GSD to ESD [42]. For instance, alleles favoured by haploid selection
in males often become associated with the Y allele, which leads to an ancestral male-biased
zygotic sex ratio. This male bias increases the potential for a neo-W or ESD allele to invade
(Table 2), equalising the sex ratio (e.g., see Fig 5B; for related examples, see [42]). On the other
hand, sex ratio selection can be overwhelmed by additional selective effects, preventing a neo-
W or ESD allele from invading, even if it would balance the sex ratio (e.g., when selection also
acts in opposite directions in male and female diploids, Table 3). Indeed, transitions between
sex-determining systems can generate stronger sex ratio biases (e.g., Fig 5A and step 1 in [43]).
In one of our key results, we find that with weak selection, there is no difference in conditions
allowing XY-to-ZW and ZW-to-XY transitions (Conclusion 4), even when haploid selection
always acts in the same sex (e.g., males). That is, the sex ratio bias created by male haploid selec-
tion facilitates the spread of a neo-W allele into an XY system to the same degree that male hap-
loid selection drives the spread of a neo-Y into a ZW system with a 1:1 sex ratio (Fig 5).
Because both fisherian selection to equalise the sex ratio and the benefits of hitchhiking
with driven alleles can facilitate transitions among sex chromosome systems, we predict that
haploid selection should increase the lability of sex determination systems. Even in animal and
plant species that have much larger and more conspicuous diploid phases than haploid phases,
many loci have been shown to experience haploid selection through gamete competition and/
or meiotic drive [38–41, 51–56], which can generate biased sex ratios [57–64]. In animals, a
relatively small proportion of all genes are thought to be expressed and selected during compe-
tition in animal sperm [39, 65, 66]. Nevertheless, expression in the gamete is not required for
haploid selection if the fitness of a gamete depends on its ability to condense DNA [67]. Fur-
thermore, expression during gamete production often underlies systems of meiotic drive [68–
70], which may be a common form of haploid selection in animals [71]. Recent studies have
demonstrated that sperm competition, even within a single ejaculate, can alter haploid allele
frequencies and increase offspring fitness [72, 73]. In plants, competition among gametophytes
may be particularly important. It is estimated that 60%–70% of all genes are expressed in the
male gametophyte, and these genes exhibit stronger signatures of selection than randomly cho-
sen genes [74–76]. Furthermore, artificial selection pressures applied to male gametophytes
are known to cause a response to selection (e.g., [77–80]).
Linking haploid expression with the evolution of sex-determination, a recent transcriptome
analysis in Rumex shows that pollen-biased expression (relative to expression in flower buds
or leaves) is enhanced among XY-linked genes, compared to autosomal genes or compared to
hemizygous genes that are only linked to the X [81]. In addition, Y-linked genes are overex-
pressed relative to X-linked genes in pollen (but not in flower buds or leaves). This suggests
that the spread of neo-Y chromosomes in this clade could have been favoured through linkage
with haploid-selected genes rather than those under sexually antagonistic selection.

Frequent turnovers driven by haploid selection may help to explain the relative rarity of
heteromorphic sex chromosomes in plants. If haploid selection is strong, but selective dif-
ferences between male and female diploids are weak, we specifically predict that trans-GSD
transitions are favoured more strongly than cis-GSD transitions, with transitions to ESD
intermediate (e.g., with js ♂ s ♀ j << ja♂D a♀D þ t ♂ t ♀ j, we have lðXY Þ ðXY Þ
W 0 > lY 0 ; Eq 3).
Among the relatively few dioecious clades in which multiple species have well-characterised
sex chromosomes [6], trans-GSD transitions have been inferred in Silene otites [15] and
in Salicaceae [16, 17]. Assuming that transitions from dioecy to hermaphroditism (equal
parental investment in male and female gametes) are favoured in a similar manner to the
ESD examined here (equal probability of zygotes developing as males or females), our
results suggest that competition among haploid pollen could drive transitions between
dioecy and hermaphroditism, which are frequent in plants [82, 83]. To further examine
this link, future theory could also include inbreeding, which is an important consideration
during transitions between dioecy and hermaphroditism [84]. Future empirical studies
could look for evidence of haploid selection acting on former sex chromosomes in her-
maphroditic species (e.g., a study such as [81] on ancestral, rather than derived, sex
chromosomes).
New sex-determining alleles have previously been shown to spread when they arise in link-
age with loci that experience sex differences in selection because beneficial associations build
up between alleles that determine sex and alleles that are favoured in that sex [35–37, 43]. In
support of this hypothesis, researchers have identified genes on recently derived sex chromo-
somes that might be under sexually antagonistic selection [21, 85–87]. However, we show that,
if selected loci are tightly linked to the ancestral sex-determining locus, they can drive trans-
GSD transitions that reduce sex-linkage (Conclusions 1 and 2), thus widening the range of
genomic locations where selection could be driving observed trans-GSD transitions. In addi-
tion, we find that polymorphic sex-determining systems (X, Y, and neo-W alleles all segregat-
ing) can be maintained when a selected locus is tightly linked to the ancestral sex-determining
system (e.g., S9B and S9C Fig), which is not possible with loose linkage [36]. This pair of con-
clusions applies in cases with or without haploid selection.
Our tight linkage result—in particular, the prediction that invasion can lead to polymor-
phic sex determination—is consistent with empirical data from species in which new femi-
nising mutations are found segregating with ancestral XY loci. For example, in the platyfish
(X. maculatus), X,Y, and W alleles segregate at one locus (or two closely linked loci) near
potentially sexually antagonistic genes for pigmentation and sexual maturity [44, 88–90].
Furthermore, several rodent species maintain feminising alleles along with the ancestral X
and Y sex determination alleles (reviewed in [91]). In nine Akadon rodent species, it appears
that male-determining sry expression is suppressed by an autosomal feminising allele (a neo-
W allele), creating XY females [92, 93]. XY females have increased fitness relative to XX
females [94]. However, it is not yet clear whether loci linked to the feminising factor or
the ancestral Y cause this effect. Most convincingly, in Mus minutoides, females can have
XX, XX , or X Y genotypes [95]. Previous theory would predict that the dominant X chro-
mosome (potentially an autosome that has fused with the sex chromosome) harbours female-
beneficial alleles, driving its spread. However, XX and XX females have similar fitness,
whereas X Y female fitness is enhanced [96–98]. Although Y-linkage of female-beneficial
alleles is counterintuitive, our model suggests that it can be stably maintained when linkage
is initially tight between the sex-determining region and the selected locus, subsequently
favouring new feminising mutations, which would be a parsimonious explanation for the
spread of feminising alleles in this case.

Our models assume that sex-determining alleles do not experience direct selection except
via their associations with sex and selected alleles. However, in some cases, there may be signif-
icant degeneration around the sex-limited allele (Y or W) in the ancestral sex-determining
region because recessive deleterious mutations and/or deletions accumulate in the surround-
ing nonrecombining regions [99–102]. During trans-GSD transitions, but not cis-GSD transi-
tions, any recessive deleterious alleles linked to the Y or W are revealed to selection in YY or
WW individuals [4]. This phenomenon was studied by van Doorn and Kirkpatrick [36], who
found that degeneration can prevent fixation of a neo-W or a neo-Y allele, leading to a mixed
sex-determining system in which the ancestral and new sex-determining loci are both segre-
gating. However, they noted that very rare recombination events around the ancestral sex-
determining locus can allow the completion of trans-GSD transitions. Degeneration around
the Y or W could explain why trans-GSD transitions are not observed to be much more com-
mon than cis-GSD transitions despite the fact that our models demonstrate that they are
favoured under a wider range of conditions, especially with haploid selection. For example,
there are a dozen sex chromosome configurations among dipteran species but only one transi-
tion between male and female heterogamety [9], but Y degeneration or absence is also very
common among Diptera [9].
In this study, we have only considered new sex-determining alleles of large effect. How-
ever, we expect similar selective forces to act on masculinising and feminising alleles of
weaker effect. For example, small-effect masculinising and feminising alleles within a thresh-
old model of sex determination can be favoured when linked to loci that experience sexually
antagonistic selection [37]. These results echo those for large-effect neo-Y and neo-W alleles
[35, 36]. It should be noted, however, that the dynamics of sex-determining alleles with very
weak effect will be influenced by genetic drift, which itself has been shown to bias transitions
towards epistatically dominant sex-determining systems when there is no direct selection
[103].
Conclusion
We have shown that tight sex linkage and haploid selection can drive previously unexpected
transitions between sex-determining systems. In particular, both can select for new sex-deter-
mining loci that are more loosely linked to loci under selection (Conclusions 2 and 3C). In
addition, haploid selection can cause transitions in GSD analogous to those caused by purely
sexually antagonistic selection, eliminating the need for differences in selection between male
and female diploids (Conclusion 3A, 3B, and 3C). We conclude that haploid selection should
be considered as a pivotal factor driving transitions between sex-determining systems. Fur-
ther, transitions involving haploid selection can eliminate or generate sex ratio biases; to
leading order, selection to balance the sex ratio and the benefits of hitchhiking with haploid-
selected alleles, leading to a biased sex ratio, are of equal magnitude (Conclusions 4 and 5).
Overall, our results suggest several novel scenarios under which new sex-determining sys-
tems are favoured, which could help to explain why the evolution of sex-determining systems
is so dynamic.
Supporting information
S1 File. Supplementary Mathematica file. This file can be used to rederive our results and
generate figures.
(NB)

S2 File. Supplementary Mathematica file in CDF form. This file can be used to rederive our
results and generate figures with a free online viewer (www.wolfram.com/cdf-player/). CDF,
computable document format.
(CDF)
S3 File. Supplementary Mathematica file in PDF form. This file can be used to see how we
have derived our results and have generated figures with any PDF viewer. PDF, portable docu-
ment format.
(PDF)
S1 Table. Substitutions for different loci orders, assuming no interference.
(PDF)
S2 Table. Mean fitnesses and zygotic sex ratio in the resident population (M fixed, XY sex
determination).
(PDF)
S1 Text. Recursion equations and complete model description.
(PDF)
S2 Text. Equilibria and stability conditions when M allele is fixed.
(PDF)
S3 Text. Invasion conditions for the m allele.
(PDF)
S1 Fig. With overdominance, loci near the ancestral sex-determining locus (r 0) can
favour neo-W alleles that are less tightly linked (R > r). In panels A and B, the a allele
is favoured in females (w♀aa ¼ 1:05, w◯ ♀
Aa ¼ 1, wAA ¼ 0:85), and selection in males is overdom-
♂ ♂
inant (waa ¼ wAA ¼ 0:75). In panels C and D, selection in males and females is overdomi-
nant (w♀aa ¼ w♀AA ¼ 0:6, w♂aa ¼ 0:5, w♂AA ¼ 0:7, w◯Aa ¼ 1). There is no haploid selection
t ◯ ¼ a◯D ¼ 0. These parameters are marked by daggers in Fig 2B and 2C, which show that
neo-W invasion is expected for any R (LðXY Þ ðXY Þ

W 0 A , LW 0 a > 1) if the a allele is nearly fixed on the Y
(black lines in this figure; not stable for r 0). Equilibria in which the A allele is more com-
mon among Y-bearing male gametes can also be stable and allow neo-W invasion for these
parameters (blue lines).
(TIF)
S2 Fig. Following invasion by a neo-W allele, there can be a complete transition to a new
sex-determining system, maintenance of both ancestral XY and neo-ZW sex-determining
systems, or loss of the new sex-determining allele. Here, we plot the frequency of the neo-W
allele among female gametes. Panels A, C, and D show cases in which a steady state is reached
with the neo-W at a frequency below 0.5, in which case ancestral X and Y alleles also both seg-
regate. In all cases, we assume that the a allele is initially more common than the A allele on
the Y background (Y-a is fixed when r = 0). When r > 0 (panels B and D), Y-A haplotypes cre-
ated by recombination can become more common than Y-a haplotypes as the neo-W spreads.
In B, this leads to loss of the neo-W, and the system goes to an equilibrium with X-a and Y-A
haplotypes fixed (equilibrium A0 ), such that all females have the high fitness genotype aa, and
all males are Aa. For the parameters in B, neo-W alleles have negative invasion fitness when
the Y-A haplotype is ancestrally more common than Y-a (compare blue to black curves in
S1A Fig and S2B Fig near the ancestral sex-determining locus). In contrast, the neo-W is not
lost in panel D, as it is favoured regardless of whether Y-A or Y-a haplotypes predominate

(again, compare blue to black curves in S1C and S1D Fig).

(TIF)
S3 Fig. When there is sexually antagonistic selection and haploid selection, a neo-W allele
may invade for any R. Panel A shows that the invasion fitness of a neo-W is positive, even
when r < R (unshaded region). In panel B, we vary the recombination rate between the neo-W
and the selected locus (R) for a fixed recombination rate between the ancestral sex-determin-
ing locus and the selected locus (r = 0.005). Coloured markers show recombination rates for
which the temporal dynamics of neo-W invasion are plotted in panel C (black R = 0.001, red
R = 0.02, blue R = 0.1, green R = 0.5). The diploid selection parameters used in this plot are the
same as in Fig 3. There is also meiotic drive in males favouring a (a♂D ¼ 0:08); this full set of
parameters is marked by an asterisk in S4A Fig. When R = 0.5 (green curve), the neo-W does
not reach fixation, and X, Y, Z, and W alleles are all maintained in the population; see S9C Fig.
(TIF)
S4 Fig. Parameters for which neo-W-A and neo-W-a haplotypes spread when there is male
meiotic drive at a locus that is tightly linked to the ancestral XY locus (r 0). This figure is
equivalent to Fig 2 but with meiotic drive in males. In panels A–C, meiotic drive in males
favours the a allele (a♂D ¼ 0:16), creating male-biased sex ratios and generally increasing
LðXY Þ ðXY Þ ðXY Þ ðXY Þ
W 0 A and LW 0 a . By contrast, LW 0 A and LW 0 a tend to be reduced when meiotic drive in males
favours the A allele (a♂D ¼ 0:16), panels D–F.
(TIF)
S5 Fig. Parameters for which neo-W-A and neo-W-a haplotypes spread when there is male
gametic competition at a locus that is tightly linked to the ancestral XY locus (r 0). This
figure is equivalent to Fig 2 but with gametic competition in males. The a allele is favoured
during male gametic competition in panels A–C (w♂a ¼ 1:16, w♂A ¼ 1), which creates male-
biased sex ratios and increases LðXY Þ ðXY Þ ðXY Þ ðXY Þ
W 0 A and LW 0 a . By contrast, LW 0 A and LW 0 a tend to be reduced
when the A allele is favoured during male gametic competition, panels D–F. Compared to the
meiotic drive parameters in S4 Fig, the effect of these male gametic competition parameters on
the sex ratio is smaller. For example, in S4A–S4C Fig, the ancestral sex ratio is α♂ = 0.58 at
equilibrium (B), and in panels A–C of this plot, the ancestral sex ratio is w♂a =ðw♂A þ w♂a Þ ¼
0:537 at equilibrium (B).
(TIF)
S6 Fig. Parameters for which neo-W-A and neo-W-a haplotypes spread when there is
female meiotic drive at a locus that is tightly linked to the ancestral XY locus (r 0). This
figure is equivalent to Fig 2 but with meiotic drive in females. The a allele is favoured by mei-
otic drive in females in panels A–C (a♀D ¼ 0:16), which increases LðXY Þ ðXY Þ
W 0 a and decreases LW 0 A .
Female meiotic drive in favour of the A allele (panels D–F, a♂D ¼ 0:16) has the opposite
effect.
(TIF)
S7 Fig. Parameters for which neo-W-A and neo-W-a haplotypes spread when there is
female gametic competition at a locus that is tightly linked to the ancestral XY locus
(r 0). This figure is equivalent to Fig 2 but with gametic competition in females. The a allele
is favoured during female gametic competition in females in panels A–C (w♀a ¼ 1:16, w♀A ¼ 1),
which increases LðXY Þ ðXY Þ
W 0 a and decreases LW 0 A . The A allele is favoured during gametic competition
in panels D–F (w♀a ¼ 1, w♀A ¼ 1:16), giving the opposite effect on LðXY Þ ðXY Þ
W 0 a and LW 0 A .
(TIF)

S8 Fig. Ploidally antagonistic selection can drive the spread of neo-W alleles. A–D show
when each of the neo-W haplotypes invades an internally stable equilibrium with a fixed on
the Y (found by setting r = 0). The y-axis shows directional selection in diploids of both sexes,
s♀ = s♂, and the x-axes show sex-limited drive, a◯ ◯
D , or haploid competition, t . The top-left
and bottom-right quadrants therefore imply ploidally antagonistic selection (and these are the
only places where neo-W haplotypes can invade). Dominance is equal in both sexes, h♀ = h♂ =
3/4. E–F show the temporal dynamics of neo-W frequency in females, with parameters given
by the asterisks in the corresponding A–D plot, with r = 1/200, for four different R. Black
R = 1/1000, Red R = 2/100, Blue R = 1/10, Green R = 1/2.
(TIF)
S9 Fig. Pseudofixation of neo-W or maintenance of multiple sex-determining alleles. The
curves show the frequencies of the neo-W (red), ancestral Y (blue), and A allele (black) among
female gametes (solid curves) and among male gametes (dashed curves). In panel A, there is a
complete transition from XY sex determination (XX-ZZ females and XY-ZZ males, labeling
allele M as Z and the new allele m as W) to ZW sex determination (YY-ZW females and YY-ZZ
males). In panels B and C, a polymorphism is maintained at both the ancestral XY locus and the
new ZW locus, such that there are males with genotypes XY-ZZ and YY-ZZ and females with
genotypes XX-ZZ, XX-ZW, XY-ZW, and YY-ZW. In panel A, selection is ploidally antagonistic
with drive in males (parameters as in the green curve in Fig 5B). In panel B, there is overdomi-
nance in both sexes and no haploid selection (parameters as in the green curve in S2C Fig). In
panel C, there is sexually antagonistic selection in diploids with drive in males (parameters as in
the green curve in S4C Fig). In all cases, the initial equilibrium frequency has a near fixation on
the Y.
(TIF)
Acknowledgments
We thank Georgy Sandler and Stephen Wright for sharing their results with us, and we thank
Bret Payseur for helpful comments on a previous version of this manuscript.
Author Contributions
Conceptualization: Michael Francis Scott.
Formal analysis: Michael Francis Scott, Matthew Miles Osmond, Sarah Perin Otto.
Investigation: Michael Francis Scott, Matthew Miles Osmond.
Supervision: Sarah Perin Otto.
Writing – original draft: Michael Francis Scott, Matthew Miles Osmond.
Writing – review & editing: Michael Francis Scott, Matthew Miles Osmond, Sarah Perin
Otto.
References
1. Bull JJ. Evolution of sex determining mechanisms. Menlo Park, CA: The Benjamin Cummings Pub-
lishing Company; 1983.
2. Charlesworth D, Mank JE. The birds and the bees and the flowers and the trees: lessons from genetic
mapping of sex determination in plants and animals. Genetics. 2010; 186(1):9–31. https://doi.org/10.
1534/genetics.110.117697 PMID: 20855574
3. Beukeboom LW, Perrin N. The evolution of sex determination. Oxford, UK: Oxford University Press;
2014.

4. Bachtrog D, Mank JE, Peichel CL, Kirkpatrick M, Otto SP, Ashman TL, et al. Sex determination: why
so many ways of doing it? PLoS Biol. 2014; 12(7):e1001899. https://doi.org/10.1371/journal.pbio.
1001899 PMID: 24983465
5. Pennell MW, Mank JE, Peichel CL. Transitions in sex determination and sex chromosomes across
vertebrate species. Molecular Ecology. 2018; Early Online. https://doi.org/10.1111/mec.14540 PMID:
29451715
6. Ming R, Bendahmane A, Renner SS. Sex chromosomes in land plants. Annual Review of Plant Biol-
ogy. 2011; 62(1):485–514.
7. Li J, Phillips RB, Harwood AS, Koop BF, Davidson WS. Identification of the sex chromosomes of
brown trout (Salmo trutta) and their comparison with the corresponding chromosomes in Atlantic
salmon (Salmo salar) and rainbow trout (Oncorhynchus mykiss). Cytogenetic and Genome Research.
2011; 133(1):25–33. https://doi.org/10.1159/000323410 PMID: 21252487
8. Yano A, Nicol B, Jouanno E, Quillet E, Fostier A, Guyomard R, et al. The sexually dimorphic on the Y-
chromosome gene (sdY) is a conserved male-specific Y-chromosome sequence in many salmonids.
Evolutionary Applications. 2012; 6(3):486–496. https://doi.org/10.1111/eva.12032 PMID: 23745140
9. Vicoso B, Bachtrog D. Numerous transitions of sex chromosomes in Diptera. PLoS Biol. 2015; 13(4):
e1002078. https://doi.org/10.1371/journal.pbio.1002078 PMID: 25879221
10. Myosho T, Otake H, Masuyama H, Matsuda M, Kuroki Y, Fujiyama A, et al. Tracing the emergence of
a novel sex-determining gene in medaka, Oryzias luzonensis. Genetics. 2012; 191(1):163–170.
https://doi.org/10.1534/genetics.111.137497 PMID: 22367037
11. Gamble T, Castoe TA, Nielse SV, Banks JL, Card DC, Schield DR, et al. The discovery of XY sex chro-
mosomes in a Boa and Python. Current Biology. 2017; 27:2148–2152. https://doi.org/10.1016/j.cub.
2017.06.010 PMID: 28690112
12. Ezaz T, Sarre SD, O’Meally D. Sex chromosome evolution in lizards: independent origins and rapid
transitions. Cytogenetic and Genome Research. 2009; 127:249–260. https://doi.org/10.1159/
000300507 PMID: 20332599
13. Mank JE, Promislow DEL, Avise JC. Evolution of alternative sex-determining mechanisms in teleost
fishes. Biological Journal of the Linnean Society. 2006; 87(1):83–93.
14. Hillis DM, Green DM. Evolutionary changes of heterogametic sex in the phylogenetic history of
amphibians. Journal of Evolutionary Biology. 1990; 3(1):49–64.
15. Slancarova V, Zdanska J, Janousek B, Talianova M, Zschach C, Zluvova J, et al. Evolution of sex deter-
mination systems with heterogametic males and females in Silene. Evolution. 2013; 67(12):3669–
3677. https://doi.org/10.1111/evo.12223 PMID: 24299418
16. Pucholt P, Rönnberg-Wästljung AC, Berlin S. Single locus sex determination and female heterogam-
ety in the baskey willow (Salix viminalis L.). Heredity. 2015; 114:575–583. https://doi.org/10.1038/hdy.
2014.125 PMID: 25649501
17. Pucholt P, Wright A, Conze LL, Mank JE, Berlin S. Recent sex chromosome divergence despite
ancient dioecy in the willow Salix viminalis. Molecular Biology and Evolution. 2017; 34:1991–2001.
https://doi.org/10.1093/molbev/msx144 PMID: 28453634
18. McDonald IC, Evenson P, Nickel CA, Johnson OA. House fly genetics: isolation of a female determin-
ing factor on chromosome 4. Annals of the Entomological Society of America. 1978; 71:692–694.
19. Thompson PE. Male and female heterogamety in population of Chironomus tentans (Diptera: Chirono-
midae). The Canadian Entomologist. 1971; 103:369–372.
20. Ogata M, Hasegawa Y, Ohtani H, Mineyama M, Miura I. The ZZ/ZW sex-determining mechanism orig-
inated twice and independently during evolution of the frog, Rana rugosa. Heredity. 2007; 100(1):92–
99. https://doi.org/10.1038/sj.hdy.6801068 PMID: 18000521
21. Ser JR, Roberts RB, Kocher TD. Multiple interacting loci control sex determination in lake Malawi cich-
lid fish. Evolution. 2010; 64(2):486–501. https://doi.org/10.1111/j.1558-5646.2009.00871.x PMID:
19863587
22. Lee BY, Hulata G, Kocher TD. Two unlinked loci controlling the sex of blue tilapia (Oreochromis
aureus). Heredity. 2004; 92:543–549. https://doi.org/10.1038/sj.hdy.6800453 PMID: 15100706
23. Vandeputte M, Dupont-Nivet M, Chavanne H, Chatain B. A polygenic hypothesis for sex determination
in the European sea bass Dicentrarchus labrax. Genetics. 2007; 176:1049–1057. https://doi.org/10.
1534/genetics.107.072140 PMID: 17435246
24. Liew WC, Bartfai R, Lim Z, Sreeninvasan R, Siegfried KR, Orban L. Polygenic sex determination sys-
tem in Zebrafish. PLoS ONE. 2012; 4:e34397.
25. Wilson CA, High SK, McCluskey BM, Amores A, Yan Y, Titus TA, et al. Wild sex in zebrafish: loss of
the natural sex determinant in domesticated strains. Genetics. 2014; 198(3):1291–1308. https://doi.
org/10.1534/genetics.114.169284 PMID: 25233988

26. Conover DO, Heins SW. Adaptive variation in environmental and genetic sex determination in a fish.
Nature. 1987; 326(6112):496–498. https://doi.org/10.1038/326496a0 PMID: 3561487
27. Pokorná M, Kratochvı́l L. Phylogeny of sex-determining mechanisms in squamate reptiles: are sex
chromosomes an evolutionary trap? Zoological Journal of the Linnean Society. 2009; 156:168–183.
28. Pen I, Uller T, Feldmeyer B, Harts A, While GM, Wapstra E. Climate-driven population divergence in
sex-determining systems. Nature. 2010; 468(7322):436–438. https://doi.org/10.1038/nature09512
PMID: 20981009
29. Holleley CE, O’Meally D, Sarre SD, Marshall Graves JA, Ezaz T, Matsubara K, et al. Sex reversal trig-
gers the rapid transition from genetic to temperature-dependent sex. Nature. 2015; 523(7558):79–82.
https://doi.org/10.1038/nature14574 PMID: 26135451
30. Fisher R. The genetical theory of natural selection. London: Clarendon Press; 1930.
31. Charnov EL. The theory of sex allocation. Princeton, NJ: Princeton University Press; 1982.
32. West S. Sex allocation. Princeton, NJ: Princeton University Press; 2009.
33. Blaser O, Grossen C, Neuenschwander S, Perrin N. Sex-chromosome turnovers induced by deleteri-
ous mutation load. Evolution. 2012; 67:635–645. https://doi.org/10.1111/j.1558-5646.2012.01810.x
PMID: 23461315
34. van Doorn GS. Patterns and mechanisms of evolutionary transitions between genetic sex-determining
systems. Cold Spring Harbour Perspectives in Biology. 2014; 6:a017681.
35. van Doorn GS, Kirkpatrick M. Turnover of sex chromosomes induced by sexual conflict. Nature. 2007;
449(7164):909–912. https://doi.org/10.1038/nature06178 PMID: 17943130
36. van Doorn GS, Kirkpatrick M. Transitions between male and female heterogamety caused by sex-
antagonistic selection. Genetics. 2010; 186(2):629–645. https://doi.org/10.1534/genetics.110.118596
PMID: 20628036
37. Muralidhar P, Veller C. Sexual antagonism and the instability of environmental sex determination.
Nature Ecology and Evolution. 2018; 2:343–351. https://doi.org/10.1038/s41559-017-0427-9 PMID:
29335574
38. Mulcahy DL, Sari-Gorla M, Mulcahy GB. Pollen selection—past, present and future. Sexual Plant
Reproduction. 1996; 9(6):353–356.
39. Joseph S, Kirkpatrick M. Haploid selection in animals. Trends in Ecology & Evolution. 2004;
19(11):592–597.
40. Úbeda F, Haig D. On the evolutionary stability of Mendelian segregation. Genetics. 2005; 170(3):1345–
1357. https://doi.org/10.1534/genetics.104.036889 PMID: 15911587
41. Lindholm AK, Dyer KA, Firman RC, Fishman L, Forstmeier W, Holman L, et al. The ecology and evolu-
tionary dynamics of meiotic drive. Trends in Ecology & Evolution. 2016; 31(4):315–326.
42. Kozielska M, Weissing FJ, Beukeboom LW, Pen I. Segregation distortion and the evolution of sex-
determining mechanisms. Heredity. 2010; 104:100–112. https://doi.org/10.1038/hdy.2009.104 PMID:
19672280
43. Úbeda F, Patten MM, Wild G. On the origin of sex chromosomes from meiotic drive. Proceedings of
the Royal Society B: Biological Sciences. 2015; 282(1798):20141932. https://doi.org/10.1098/rspb.
2014.1932 PMID: 25392470
44. Kallman K. Evidence for the existence of transformer genes for sex in the teleost Xiphophorus macula-
tus. Genetics. 1968; 60:811–828. PMID: 5732420
45. Dubendorfer A, Hediger M, Burghardt G, Bopp D. Musca domestica, a window on the evolution of sex-
determining mechanisms in insects. International Journal of Developmental Biology. 2002; 46(1):75–
79. PMID: 11902690
46. Roco AS, Olmstead AW, Degitz SJ, Amano T, Zimmerman LB, Bullejos M. Coexistence of Y, W, and
Z sex chromosomes in Xenopus tropicalis. Proceedings of the National Academy of Sciences. 2015;
112(34):E4752–E4761.
47. Holman L, Price TAR, Wedell N, Kokko H. Coevolutionary dynamics of polyandry and sex-linked mei-
otic drive. Evolution. 2015; 69(3):709–720. https://doi.org/10.1111/evo.12595 PMID: 25565579
48. Immler S, Arnqvist G, Otto SP. Ploidally antagonistic selection maintains stable genetic polymorphism.
Evolution. 2012; 66(1):55–65. https://doi.org/10.1111/j.1558-5646.2011.01399.x PMID: 22220864
49. Lloyd DG, Webb C. Secondary sex characters in plants. Botanical Review. 1977; 43:177–216.
50. Otto SP. Selective maintenance of recombination between the sex chromosomes. Journal of Evolu-
tionary Biology. 2014; 27:1431–1442. https://doi.org/10.1111/jeb.12324 PMID: 24529284
51. Lalanne E, Michaelidis C, Moore JM, Gagliano W, Johnson A, Patel R, et al. Analysis of transposon
insertion mutants highlights the diversity of mechanisms underlying male progamic development in

Arabidopsis. Genetics. 2004; 167:1975–1986. https://doi.org/10.1534/genetics.104.030270 PMID:

15342534
52. Fishman L, Willis JH. A novel meiotic drive locus almost completely distorts segregation in Mimulus
(monkeyflower) hybrids. Genetics. 2005; 169:347–353. https://doi.org/10.1534/genetics.104.032789
PMID: 15466426
53. Leppälä J, Bechsgaard JS, Schierup MH, Savolainen O. Transmission ratio distortion in Arabidopsis
lyrata: effects of population divergence and the S-locus. Heredity. 2008; 100:71–78. https://doi.org/10.
1038/sj.hdy.6801066 PMID: 17940548
54. Leppälä J, Bokma F, Savolainen O. Investigating incipient speciation in Arabidopsis lyrata from pat-
terns of transmission ratio distortion. Genetics. 2013; 194:697–708. https://doi.org/10.1534/genetics.
113.152561 PMID: 23666938
55. Didion JP, Morgan AP, Clayschulte AMF, Mcmullon RC, Yadgary L, Petkov PM, et al. A multi-mega-
base copy number gain causes maternal transmission ratio distortion on mouse chromosome 2. PLoS
Genet. 2015; 11:e1004850. https://doi.org/10.1371/journal.pgen.1004850 PMID: 25679959
56. Didion JP, Morgan AP, Yadgary L, Bell TA, McMullan RC, Ortiz de Solorzano L, et al. R2d2 drives self-
ish genetic sweeps in the house mouse. Molecular Biology and Evolution. 2016; 33:1381–1395.
https://doi.org/10.1093/molbev/msw036 PMID: 26882987
57. Jaenike J. Sex chromosome meiotic drive. Annual Review of Ecology and Systematics. 2001; 32:25–
49.
58. Burt A, Trivers R. Genes in conflict: the biology of selfish genetic elements. Cambridge, MA: Belknap
Press; 2006.
59. Lloyd DG. Female-predominant sex ratios in angiosperms. Heredity. 1974; 32(1):35–44.
60. Conn JS, Blum U. Sex ratio of Rumex hastatulus: the effect of environmental factors and certation. Evo-
lution. 1981; 35(6):1108–1116. https://doi.org/10.1111/j.1558-5646.1981.tb04980.x PMID: 28563383
61. Stehlik I, Barrett S. Mechanisms governing sex-ratio variation in dioecious Rumex nivalis. Evolution.
2005; 59(4):814–825. PMID: 15926691
62. Stehlik I, Barrett SCH. Pollination intensity influences sex ratios in dioecious Rumex nivalis, a wind-pol-
linated plant. Evolution. 2006; 60(6):1207–1214. PMID: 16892971
63. Field DL, Pickup M, Barrett SCH. The influence of pollination intensity on fertilization success, progeny
sex ratio, and fitness in a wind-pollinated, dioecious plant. International Journal of Plant Sciences.
2012; 173(2):184–191.
64. Field DL, Pickup M, Barrett SCH. Comparative analyses of sex-ratio variation in dioecious flowering
plants. Evolution. 2013; 67(3):661–672. https://doi.org/10.1111/evo.12001 PMID: 23461317
65. Zheng Y, Deng X, Martin-DeLeon PA. Lack of sharing of Spam1 (Ph-20) among mouse spermatids
and transmission ratio distortion. Biology of Reproduction. 2001; 64(6):1730–1738. PMID: 11369602
66. Vibranovski MD, Chalopin DS, Lopes HF, Long M, Karr TL. Direct evidence for postmeiotic transcrip-
tion during Drosophila melanogaster spermatogenesis. Genetics. 2010; 186(1):431–433. https://doi.
org/10.1534/genetics.110.118919 PMID: 20610406
67. Immler S, Otto SP. The evolutionary consequences of selection at the haploid gametic stage. The
American Naturalist. https://doi.org/10.1086/698483 In press.
68. Tau Y, Araripe L, Kingan SB, Ke Y, Xiao H, Hartl DL. A sex-ratio meiotic drive system in Drosophila
simulans. II: an X-linked distorter. PLoS Biol. 2007; 5(11):e293. https://doi.org/10.1371/journal.pbio.
0050293 PMID: 17988173
69. Cocquet J, Ellis PJI, Mahadevaiah SK, Affara NA, Vaiman D, Burgoyne PS. A genetic basis for a post-
meiotic X versus Y chromosome intragenomic conflict in the mouse. PLoS Genet. 2012; 8(9):
e1002900. https://doi.org/10.1371/journal.pgen.1002900 PMID: 23028340
70. Helleu Q, Gérard PR, Dubruille R, Ogereau D, Prud’homme B, Loppin B, et al. Rapid evolution of a Y-
chromosome heterochromatin protein underlies sex chromosome meiotic drive. Proceedings of the
National Academy of Sciences. 2016; 113(15):4110–4115. https://doi.org/10.1073/pnas.1519332113
PMID: 26979956
71. Bachtrog D, Ellison C, Leonard C, Landeen E, Gibilisco L, Phadnis N. Rampant cryptic sex chromo-
some drive in Drosophila. bioRxiv. 2018; https://doi.org/10.1101/324368
72. Immler S, Hotzy C, Alavioon G, Petersson E, Arnqvist G. Sperm variation within a single ejaculate
affects offspring development in Atlantic salmon. Biology Letters. 2014; 10(2):20131040. https://doi.
org/10.1098/rsbl.2013.1040 PMID: 24522632
73. Alavioon G, Hotzy C, Nakhro K, Rudolf S, Scofield DG, Zajitschek S, et al. Haploid selection within a
single ejaculate increases offspring fitness. Proceedings of the National Academy of Sciences. 2017;
114:8053–8058.

74. Borg M, Brownfield L, Twell D. Male gametophyte development: a molecular perspective. Journal of
Experimental Botany. 2009; 60(5):1465–1478. https://doi.org/10.1093/jxb/ern355 PMID: 19213812
75. Arunkumar R, Josephs EB, Williamson RJ, Wright SI. Pollen-specific, but not sperm-specific, genes
show stronger purifying selection and higher rates of positive selection than sporophytic genes in Cap-
sella grandiflora. Molecular Biology and Evolution. 2013; 30(11):2475–2486. https://doi.org/10.1093/
molbev/mst149 PMID: 23997108
76. Gossmann TI, Schmid MW, Grossniklaus U, Schmid KJ. Selection-driven evolution of sex-biased
genes is consistent with sexual selection in Arabidopsis thaliana. Molecular biology and evolution.
2014; 31(3):574–583. https://doi.org/10.1093/molbev/mst226 PMID: 24273323
77. Hormaza JI, Herrero M. Male gametophytic selection as a plant breeding tool. Scientia horticulturae.
1996; 65(4):321–333.
78. Ravikumar RL, Patil BS, Salimath PM. Drought tolerance in sorghum by pollen selection using osmotic
stress. Euphytica. 2003; 133:371–376.
79. Hedhly A, Hormaza JI, Herrero M. Effect of temperature on pollen tube kinetics and dynamics in sweet
cherry, Prunus avium (Rosaceae). American journal of botany. 2004; 91(4):558–564. https://doi.org/
10.3732/ajb.91.4.558 PMID: 21653411
80. Clarke HJ, Khan TN, Siddique KHM. Pollen selection for chilling tolerance at hybridisation leads to
improved chickpea cultivars. Euphytica. 2004; 139(1):65–74.
81. Sandler G, Beaudry FEG, Barrett SCH, Wright SI. The effects of haploid selection on Y chromosome
evolution in a dioecious plant. Evolution Letters. https://doi.org/10.1002/evl3.60
82. Käfer J, Marais GAB, Pannell J. On the rarity of dioecy in flowering plants. Molecular Ecology. 2017;
26:1225–1241. https://doi.org/10.1111/mec.14020 PMID: 28101895
83. Goldberg EE, Otto SP, Vamosi JC, Mayrose I, Sabath N, Ming R. Macroevolutionary synthesis of flow-
ering plant sexual systems. Evolution. 2017; 71:898–912. https://doi.org/10.1111/evo.13181 PMID:
28085192
84. Charlesworth B, Charlesworth D. A model for the evolution of dioecy and gynodioecy. The American
Naturalist. 1978; 112(988):975–997.
85. Lindholm A, Breden F. Sex chromosomes and sexual selection in poeciliid fishes. The American Natu-
ralist. 2002; 160(S6):S214–24.
86. Tripathi N, Hoffmann M, Willing EM, Lanz C, Weigel D, Dreyer C. Genetic linkage map of the guppy,
Poecilia reticulata, and quantitative trait loci analysis of male size and colour variation. Proceedings of
the Royal Society B: Biological Sciences. 2009; 276(1665):2195–2208. https://doi.org/10.1098/rspb.
2008.1930 PMID: 19324769
87. Roberts RB, Ser JR, Kocher TD. Sexual conflict resolved by invasion of a novel sex determiner in
Lake Malawi cichlid fishes. Science. 2009; 326(5955):998–1001. https://doi.org/10.1126/science.
1174705 PMID: 19797625
88. Kallman K. Genetics and geography of sex determination in the poeciliid fish, Xiphophorus maculatus.
Zoologica. 1965; 50(13):151–190.
89. Volff JN, Schartl M. Variability of genetic sex determination in poeciliid fishes. Genetica. 2001;
111(1):101–110.
90. Schulteis C, Zhou Q, Froschauer A, Nanda I, Selz Y, Schmidt C, et al. Molecular analysis of the sex-
determining region of the platyfish Xiphophorus maculatus. Zebrafish. 2006; 3(3):299–309. https://
doi.org/10.1089/zeb.2006.3.299 PMID: 18377211
91. Fredga K. Bizarre mammalian sex-determining mechanisms. In: Short RV, Balaban E, editors.
The differences between the sexes. Cambridge, USA: Cambridge University Press; 1994. p. 419–
432.
92. Bianchi NO. Akodon sex reversed females: the never ending story. Cytogenetic and Genome
Research. 2002; 96:60–65. https://doi.org/10.1159/000063029 PMID: 12438781
93. Sánchez A, Marchal JA, Romero-Fernández I, Pinna-Senn E, Ortiz MI, Bella JL, et al. No differences
in the Sry gene between males and XY females in Akodon (Rodentia, Cricetidae). Sexual Develop-
ment. 2010; 4:155–161. https://doi.org/10.1159/000309780 PMID: 20453479
94. Hoekstra HE, Hoekstra JM. An unusual sex-determination system in South American field mice
(genus Akodon): the role of mutation, selection, and meiotic drive in maintaining XY females. Evolu-
tion. 2001; 55(1):190–197. PMID: 11263738
95. Veyrunes F, Chevret P, Catalan J, Castiglia R, Watson J, Dobigny G, et al. A novel sex
determination system in a close relative of the house mouse. Proceedings of the Royal Society
B: Biological Sciences. 2010; 277(1684):1049–1056. https://doi.org/10.1098/rspb.2009.1925
PMID: 20007182

96. Saunders PA, Perez J, Rahmoun M, Ronce O, Crochet PA, Veyrunes F. XY females do better than
the XX in the African pygmy mouse, Mus minutoides. Evolution. 2014; 68(7):2119–2127. https://doi.
org/10.1111/evo.12387 PMID: 24611447
97. Saunders PA, Franco T, Sottas C, Maurice T, Guila G, Veyrunes F. XY females do better than the XX
in the African pygmy mouse, Mus minutoides. Scientific Reports. 2016; 6:22881e.
98. Veyrunes F, Perez J. X inactivation in a mammal species with three sex chromosomes. Chromosoma.
2017; 127(2):261–267. https://doi.org/10.1007/s00412-017-0657-2 PMID: 29256059
99. Rice WR. Evolution of the Y sex chromosome in animals. BioScience. 1996; 46(5):331–343.
100. Charlesworth B, Charlesworth D. The degeneration of Y chromosomes. Philosophical Transactions of
the Royal Society B: Biological sciences. 2000; 355(1403):1563–1572.
101. Bachtrog D. A dynamic view of sex chromosome evolution. Current opinion in genetics & develop-
ment. 2006; 16(6):578–585.
102. Marais GAB, Nicolas M, Bergero R, Chambrier P, Kejnovsky E, Monéger F, et al. Evidence for degen-
eration of the Y chromosome in the dioecious plant Silene latifolia. Current Biology. 2008; 18(7):545–
549. https://doi.org/10.1016/j.cub.2008.03.023 PMID: 18394889
103. Veller C, Muralidhar P, Constable GWA, Nowak MA. Drift-induced selection between male and female
heterogamety. Genetics. 2017; 207(2):711–727. https://doi.org/10.1534/genetics.117.300151 PMID:
28821587

Imp 1

Uploaded by

Copyright:

Available Formats

Imp 1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Imp 1

Uploaded by

Copyright:

Available Formats

RESEARCH ARTICLE

Haploid selection, sex ratio bias, and

☯ These authors contributed equally to this work.

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 1 / 25

Abbreviations: ESD, environmental sex

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 2 / 25

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 3 / 25

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 4 / 25

Table 1. Relative fitness of different genotypes in sex, ◯ 2 {♀,♂}.

Genotype Relative fitness during diploid selection

Genotype Transmission during meiosis in Aa heterozygotes

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 5 / 25

Generic invasion by a neo-Y or neo-W

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 6 / 25

ated by recombination (wðXY Þ

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 7 / 25

Tight linkage with the ancestral sex-determining locus (r  0)

alleles into females (corresponding to LðXY Þ

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 8 / 25

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 9 / 25

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 10 / 25

Loose linkage with the ancestral sex-determining region

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 11 / 25

male drive only h♂ = h♀, t ♀ ¼ t ♂ ¼ a♀D ¼ 0 s♀ s♂ > 0

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 12 / 25

allele into a ZW system occur under the same conditions (lðXY Y0

Conclusion 4: When selection is weak relative to recombination, trans-GSD transitions in

Transitions to environmentally determined sex

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 13 / 25

This reduces to lðXY

in which lðXY Þ ðXY Þ

3A), but lðXY Þ

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 14 / 25

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 15 / 25

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 16 / 25

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 17 / 25

neo-W invasion is expected for any R (LðXY Þ ðXY Þ

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 18 / 25

(again, compare blue to black curves in S1C and S1D Fig).

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 19 / 25

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 20 / 25

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 21 / 25

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 22 / 25

Arabidopsis. Genetics. 2004; 167:1975–1986. https://doi.org/10.1534/genetics.104.030270 PMID:

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 23 / 25

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 24 / 25

PLOS Biology | https://doi.org/10.1371/journal.pbio.2005609 June 25, 2018 25 / 25

You might also like

Tight linkage with the ancestral sex-determining locus (r 0)