Abstract
Classical sexual selection theory has been used to explain some of the very common sexual dimorphisms in the animal kingdom. It is well known that sexually selected traits experience strong selection pressures from processes such as sperm competition and female choice, and therefore evolve at accelerated rates. To date most studies of sexual selection theory have been ecology-based focusing on phenotypes. It is currently unclear how and if sexual selection operates on the molecular level. In this study I used gene expression data for chicken to determine what types of selectional processes operate in genes which experience sex-biased expression in both gonad and brain tissue. Mean d
N/d
Sas well as the M-K test were used to determine the types of selection acting upon these genes. I found increased signatures of selection for female-biased genes during
embryonic development, coinciding with female meiosis. Conversely male-biased genes
showed higher selective signatures during adulthood, during which male spermatogenesis
occurs. Positive selection on genes implicit in the reproductive tract of chicken could not
be determined due to the lack of a comprehensive polymorphism database for chicken. I
argue that sex-biased genes in chicken gonad experience accelerated rates of evolution
due to sexual selection. No obvious distinction in evolutionary rates between male and
female-biased genes expressed in the brain could be detected.
Introduction
Sexual dimorphisms are abundant in the animal kingdom, ranging from the most visually evident, morphological and behavioural, to more inconspicuous, physiological and gametic. Classically, sexual dimorphisms are explained by sexual selection theory whereby certain traits related to reproduction are selected for during mating. Sexually selected traits such as tail length, colour, and size are under strong directional selection and therefore experience faster rates of evolution than non-sexually selected traits (Zhang et al. 2004). Most tests to date of the role that sexual selection plays on the evolution of morphological traits have been on the level of the phenotype, and involved mating assays. It is however important to understand the underlying genetic changes which produce morphological variation during the evolution of a species (Carroll 2000), particularly the genetic mechanism underlying sexually selected traits. Thus, an interesting question that arises is whether or not sexual selection can be detected at the molecular level.
A surprising fact of sexual dimorphisms is that sexual differences persist only on the phenotypic level whereas at the coding level, males and females are nearly identical (Ellegren & Parsch 2007). Thus differences which do occur do so because of cis-
regulatory changes affecting the expression of genes (Carroll 2000). That is, the same set of genes is regulated differentially such that expression levels between males and females vary leading to distinct sexual dimorphisms. Candidate genes which could shed light on the evolution of sexual dimorphisms are those genes which exhibit sex-biased expression (Connallon & Knowles 2005). Genes with sex-biased expression, i.e. produce sexually dimorphic expression profiles, are commonly referred to as sex-biased genes (Ellegren &
Parsch 2007). Furthermore, these dimorphic expression patterns are frequently due to up-
regulation of sex-biased genes one sex, and, to a lesser extent down-regulation in the
other sex (Connallon & Knowles 2005 and Zhang et al. 2004). Therefore, studying
changes in sex-biased genes could shed light into how sexual dimorphisms persist in
nature.
Sex-biased genes often result when a single gene develops antagonistic selection
pressures in males and females (Ellegren & Parsch 2007). This occurs because males and females often differ in fitness optima for a particular trait. For example, testosterone expression optima are different between males and females since it is essential for male reproduction and detrimental to female reproduction (Moller et al. 2005). Traits, such as testosterone levels, which are beneficial to one sex and detrimental to the other sex, are termed sexually antagonistic (Rice & Chippindale 2002), and the conflict which arises between the sexes is resolved through sex-biased gene expression (Connallon & Knowles 2005). Typically, sex-biased gene expression results from sexual antagonism and the accumulation adaptive mutations which occur predominantly in the male lineages, implying that male traits evolve at a faster rate than females (Connallon & Knowles 2005, and Zhang et al. 2004) This process is further complicated by the location and number of tissues a particular gene is expressed in.
Expression in multiple tissues, a process we refer to here as pleiotropy, has been shown to constrain the resolution of sexual antagonism in chicken and mouse (Mank et al.
2007), as well as Drosophila (Zhang et al. 2007 and Larracuente et al. 2008). It has been suggested that as a gene is expressed in more tissues, negative selection becomes stronger and the gene thus becomes constrained (Khaitovich et al. 2008). This arises because each tissue has different selection pressures, which leads to constraints on the possible changes which can occur in gene expression in order for that gene to be functional in a particular tissue, this in turn constrains divergence of that gene. By this mechanism sex-biased gene expression is expected to be highest in genes expressed in fewer tissues because they will not be constrained and can thus develop differential expression.
It is well known that genes involved in sex related processes experience faster rates of change in their coding sequences than genes which are not related to sex (Zhang et al.
2004). Thus sexual selection theory may be an ideal construct with which we can use to
explain the evolution of sex-biased genes. Two classical components of sexual selection
which can affect the evolution of phenotypes are female choice and sperm competition.
other potential mates, which directly skews the allele frequency in the next generation (Wigby & Chapman 2004). Furthermore, the environment created by a female when she mates multiply encourages competition between sperm for fertilization success;
promoting evolution of adaptive traits (Wigby & Chapman 2004).
Several types of tests have been adopted in order to distinguish forces influencing the allele frequency spectrum, including the d
N/d
Stest (Miyata & Yasunaga 1980),
McDonald-Kreitman test (McDonald & Kreitman 1991), Tajima’s D (Tajima 1989), and Fay and Wu’s H ( Fay & Wu 2000). Two common forms of natural selection are positive and purifying selection; several studies have shown that these types of selection affect gene expression (Eyre-Walker 2006, Shapiro et al. 2007, Larracuente et al. 2008, Singh et al. 2007, Khaitovich et al. 2008). The d
N/d
Stest is probably the most informative of these tests to predict the type of selection, as the other aforementioned population genetic tests measure only departures from neutrality (Yang & Bielawski 2000).
The magnitude of sexual dimorphisms in birds makes this an interesting clade with which to study the evolution of sex-biased gene expression. Furthermore, females are the
heterogametic sex which provides an interesting contrast to studies on mammals and Drosophila which are male heterogametic. Additionally it has been noted that female heterogamety is under different selective forces than male heterogamety (Axelsson et al.
2004). In this paper I will try to determine whether or not the evolution of sex-biased genes can be explained by classical sexual selection theory. I use chicken microarray data to study the evolution of sex-biased gene expression in a female heterogametic system.
Microarray data allows for the unique opportunity to determine which ‘evolutionary
processes shape the genome’ (Connallon & Knowles 2005). I calculated divergence
estimates in genes expressed in brain and gonad. This along with analysis of chicken
polymorphisms should allow me to determine which processes, either positive selection
or genetic drift, influence the pattern of evolution for sex-biased genes in chicken.
Materials and Methods
15 male and 15 female white leghorn chickens were euthanized at each of embryonic days (ED) 10, 15, 19 and reproductive adulthood. Rinn et al. (2005) found that sex-biased gene expression occurs in mouse embryos at day 10.5 after fertilization, and sex
differences persist into adulthood. Therefore expression profiles during development have the potential to change intensity throughout development, and as such a study incorporating changes throughout development could be interesting. Animals were immediately dissected and whole brain samples were taken from all time points, and gonad samples from the latter three time points. Gene expression profiles were obtained using the Affymetrix microarray genechip platform. Complete details of the sample preparation and microarray methods can be found in Mank & Ellegren (2009).
The degree of sex-bias was calculated as the log
2(male/female). Fold-change cutoffs were used to define the type of bias for each gene; male, female or unbiased. Fold-change refers to the degree of difference in expression of a gene between the sexes. Because gonadal tissue is highly sexually dimorphic, we used higher fold change levels to
compensate for the increased degree of sex-bias seen in this tissue compared to brain, as previously implemented (Mank et al. 2007). Brain fold-change values were 125%, 133%
and 150%, while 200%, 250% and 300% were used for gonad. Genes which did not meet these cutoff values were classified as unbiased.
Divergence analysis
Divergence in protein sequence between two species is commonly measured as the ratio
(ω) of nonsynonymous (d
N) and synonymous (d
S) mutations (Yang & Bielawski 2000),
where a nonsynonymous mutation has the effect of changing the amino acid, while a
synonymous mutation, a product of the redundancy of the genetic code, is silent. d
Nand
d
Sare defined as the number of mutations divided by the total possible mutational sites of
the same type. In this type of analysis the nature of selection driving protein evolution
can be deduced. The following inferences can be made (Yang & Bielawski 2000): ω >1
is an indication of positive selection since there are more selectively advantageous
mutations are being purged due to purifying selection there are fewer nonsynonymous than synonymous mutations and ω < 1. In the neutral case (ω = 1), nonsynonymous mutations are fixed at the same rate as synonymous mutations.
In this study we estimated values of d
Nand d
Sfor chicken using a phylogenetic approach.
This was accomplished by first identifying 1:1 orthologs for significantly expressed chicken genes based on the ENSEMBL chicken predicted coding regions with reciprocal BLAST to the zebra finch (Taeniopygia guttata) genome draft sequence. For all resulting 1:1 orthologs, PAML (Yang 1997) was used to estimate the number of synonymous and nonsynonymous sites and changes between species. The resulting data were then filtered;
all genes which were mutationally saturated (d
S>2) were removed from further analysis to reduce to problems with mutational saturation. Transcripts with coding alignments
<100bp or which had premature stop codons were also eliminated. Finally, due to the confounding effects of the accelerated rate of evolution for Z-linked genes (Mank et al.
2007), in combination with the overall male-bias of the chromosome due to lack of complete dosage compensation (Ellegren et al. 2007), I focused all further analysis solely on the autosomes.
For each class of genes, average d
Nand d
Swere calculated by dividing the total sum of substitutions by the sum of the number of possible sites for each class of genes (male- biased, female-biased, and unbiased). Mean values of d
N/d
Swere then calculated from these values. Measurements of d
N/d
Swere calculated in this manner because this
methodology weighs divergence estimates by the alignment length and at the same time allows for the inclusion of loci with d
Sequal to zero (Mank et al. 2007).
Significant differences in divergence estimates among expression classes were assessed
with permutation tests of 1000 replicates, comparing sex-biased categories to the
unbiased estimates.
Tissue of maximal expression
Genes expressed in the brain or gonad are not necessarily limited to that tissue, as genes may have many functions throughout the body, and may therefore be expressed in many tissues and organs. In order to ensure my estimates of divergence reflect either gonad or brain related evolutionary paths, I determined the tissue that each gene was maximally expressed in.
Chicken expressed sequence tags (EST) were collected for chicken from the UniGene library (ftp://ncbi.nlm.nih.gov/repository/unigene/). The tissue of maximal expression for each gene was calculated by
where TPM
iis the number of transcripts per million per tissue, library
iis the total number of unique ESTs per library. A complete description of these methods can be found in Mank et al. (2007).
Estimates of d
N, d
Sand d
N/d
Swere calculated for genes maximally expressed in either brain or gonad. Permutations tests (1000 replicates) were once again carried out to compare the sex-biased categories to the unbiased estimates.
Gene ontology
Some classes of genes experience stronger selection pressures than others; therefore I
tested whether or not different functional classes of genes may have introduced biases in
our results. Gene function was determined from the Gene Ontology (GO) Consortium
(Gene Ontology Consortium 2000). ONTOLOGIZER 2.0 (Robinson et al. 2004) was
used to identify if any GO terms were overrepresented in each expression class of the
analysis, using term-for-term comparison and the westfall-young-step-down correction.
Polymorphism analysis
As a compliment to the divergence test above, a polymorphism analysis was also performed. This test, called the McDonald-Kreitman test (MK test), compares the numbers of synonymous and nonsynonymous substitutions which occur either as polymorphisms or are fixed within the species. The MK test can distinguish between relaxed constraint acting upon sex-biased genes resulting in neutral accumulation of substitutions or positive selection of adaptive substitutions in sex-biased genes (Zhang et al. 2004). If the ratio of nonsynonymous to synonymous polymorphisms is greater than that for fixed differences i.e. d
N/d
S> p
N/p
S, than you can conclude that positive selection has occurred (McDonald & Kreitman 1991, Eyre-Walker 2006, Axelsson & Ellegren 2009). This type of test is more effective than tests of divergence in disentangling purifying and positive selection (Shapiro et al. 2007). Instead of comparing
polymorphism estimates on a gene-by-gene basis, here, a variation of the MK test was carried out whereby pooled polymorphism estimates for each sex-bias category were used.
SNPs (single nucleotide polymorphisms) were obtained for chicken from ChickVD (http://chicken.genomics.org.cn/), which houses variation data for broiler, silkie and layer chickens sequenced to ~25% coverage (Wang et al. 2005). Using SNPs which map to the loci used in the divergence analysis, numbers of nonsynonymous and synonymous polymorphisms were calculated. This allows for direct comparison of divergence and polymorphism data.
Estimates of p
N, p
Sand p
N/p
Swere calculated by dividing the sum of the number of mutations divided by the sum of mutational sites per each expression class. As mentioned above, this provides a more realistic calculation of these values. Significance testing for the difference between sex-biased and unbiased classes was evaluated through
permutations tests with 1000 replicates.
Results
Due to the sheer volume of analyses, only a sub-set of the results are presented and discussed in the body of the paper. Unless otherwise stated all results pertain to fold change cutoff 200% or 125%. Full results for all fold change levels and time points can be found in the appendix of the paper. The absence of these results from the discussion, should in no way affect the interpretation of this paper.
Divergence analysis
There were significant differences between estimates of divergence for chicken gonad
sex biased genes throughout development (Figure 1). Adult chicken male biased genes
show the highest divergence (d
N/d
S=0.120, P<0.05). However, during embryonic
development, it is the female biased genes which are more highly diverged. At ED 19
there is a large, significant difference in female divergence rates compared with unbiased
genes (d
N/d
S= 0.146, P<0.05). This pattern was not apparent in sex biased brain genes.
A
Adult gonad fc>2
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
unbiased mb fb
dN/dS
*
B
Embryo day 19 Gonad fc>2
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
unbiased mb fb
dN/dS
*
C
Em bryo day 15 fc>2 and 2.5
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Unbiased mb fb
dN/dS
Figure 1. Comparison of average divergence estimates for chicken sex biased gonad genes during development (A=Adult, B=ED 19, and C=ED 15). Mb = male bias, fb = female biased. Significant differences of sex biased genes from unbiased genes are marked by * (permutation test with 1000 replicates, P<0.05).
There was no significant difference between sex-biased categories across all time points
for brain expressed genes (See appendix for actual values). Generally, there were not
many brain genes which were sex-biased across all time points. This results in low
statistical power of the data. Thus, permutations tests were not conducted when there
were less than 50 sex-biased genes, as results are non-interpretable. For those classes
where there were enough sex-biased genes, there was no significant differences between
divergence values for the sex biased classes (P>0.05).
A pattern in the expression data for brain (Figure 2), although insignificant, did immerge, which is different from that of the gonad expressional divergence analysis. In embryonic chickens female-biased genes show the greatest divergence and this pattern persists into adulthood. Thus there appears to be no change in the expressional differences of brain genes through development in chicken.
A
Adult Brain fc>1.25
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
unbiased mb fb
B
Em bryo day 19 Brain fc>1.25
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
unbiased mb fb
dN/dS
C
Em bryo day 15 Brain fc>1.25
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
unbiased mb fb
dN/dS
D
Em bryo day 10 brain fc>1.25
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
unbiased mb f b
dN/dS
Figure 2. Comparison of average divergence estimates for chicken sex-biased brain genes during development (A=Adult, B=ED 19, C=ED 15, and D=Ed 10). Mb = male bias, fb = female biased.
Significant differences of sex biased genes from unbiased genes are marked by * (permutation test with
1000 replicates, P<0.05).
Tissue of maximal expression
Mean divergence estimates of sex-biased gene classes were calculated for those genes which have primary expression in gonad only (Figure 3). The resulting dataset was a subset of the autosomal data used in the above analysis. Mean divergence estimates for tissue specific sex biased genes differ from the autosomal estimates of d
N/d
S. Adult estimates of d
N/d
Sdid not vary between the autosomal and tissue specific analyses; with female biased classes being more diverged. However, embryonic tissue specific mean d
N/d
Sprofiles did change from having highest female divergence to highest male divergence. However, there was a severe deficit of sex-biased genes in both ED19 and ED15, thus permutation tests were not carried out due to lack of statistical support.
A
Tissue specific adult gonad fc>2
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
ub mb fb
dN/dS
B
Tissue specific ED19 gonad fc>2
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
ub mb fb
dN/dS
C
Tissue specific ED15 gonad fc2 and 2.5
0.095 0.1 0.105 0.11 0.115 0.12 0.125
ub mb f b
dN/dS