The Pseudoautosomal Regions of the U/V Sex Chromosomes of the Brown Alga Ectocarpus Exhibit Unusual Features.

The recombining regions of sex chromosomes (pseudoautosomal regions, PARs) are predicted to exhibit unusual features due to their being genetically linked to the nonrecombining, sex-determining region. This phenomenon is expected to occur in both diploid (XY, ZW) and haploid (UV) sexual systems, with slightly different consequences for UV sexual systems because of the absence of masking during the haploid phase (when sex is expressed) and because there is no homozygous sex in these systems. Despite a considerable amount of theoretical work on PAR genetics and evolution, these genomic regions have remained poorly characterized empirically. We show here that although the PARs of the U/V sex chromosomes of the brown alga Ectocarpus recombine at a similar rate to autosomal regions of the genome, they exhibit many genomic features typical of nonrecombining regions. The PARs were enriched in clusters of genes that are preferentially, and often exclusively, expressed during the sporophyte generation of the life cycle, and many of these genes appear to have evolved since the Ectocarpales diverged from other brown algal lineages. A modeling-based approach was used to investigate possible evolutionary mechanisms underlying this enrichment in sporophyte-biased genes. Our results are consistent with the evolution of the PAR in haploid systems being influenced by differential selection pressures in males and females acting on alleles that are advantageous during the sporophyte generation of the life cycle.

Introduction

Sex chromosomes have commonly been found to possess strikingly distinctive features compared with autosomes, for example in terms of the content and density of genes and repeat sequences. These characteristics are thought to be a consequence of suppression of recombination between the sex chromosomes (X and Y or Z and W in diploid systems, or U and V in haploid systems; reviewed in Otto et al. [2011]). A broadly established model of sex chromosome evolution predicts gradual expansion of the region of suppressed recombination, driven by selection for linkage between the sex-determining region (SDR) and loci at which selection differs between males and females (Charlesworth et al. 2005; Immler and Otto 2015). Expansion of the SDR reduces the recombining portion of the sex chromosome, the so-called pseudoautosomal region (PAR). However, the recombining region is usually not lost completely and it is thought that most species retain a PAR because homologous recombination in this region plays a critical role in chromosomal pairing and segregation during meiosis (Rouyer et al. 1986; Shi et al. 2001). Moreover, there are situations where sexually antagonistic (SA) forces may be too weak to drive a marked expansion of the SDR, and an extensive PAR may be preserved. This may be expected to occur, for example, in organisms with a low level of phenotypic sexual dimorphism (e.g., Ahmed et al. 2014) or where SA selection has been resolved by alternative processes such as the evolution of sex-biased gene expression (Vicoso et al. 2013).

The evolutionary fate of PAR genes is expected to differ from that of either autosomal or fully sex-linked genes. In particular, sex differences in allele frequencies should be maintained more easily in the PAR, either due to SA polymorphisms (which are maintained under a wider range of conditions than on autosomes), or to other forms of selection, such as heterozygous advantage (Otto et al. 2011). These effects are expected to be strongest very near the SDR, and to decay as the genetic distance from the SDR increases (the rate of decay being inversely proportional to the strength of selection maintaining polymorphism; Charlesworth et al. 2014; Kirkpatrick and Guerrero 2014).

There has been little empirical work on PARs. Analyzes of the structure and genetic behavior of the PAR have mainly focused on organisms that have old, well-differentiated sex chromosomes such as humans and other mammals (Flaquer et al. 2008; Raudsepp et al. 2012), and more recently birds (Smeds et al. 2014). These PARs have been shown to exhibit several unusual features compared with autosomes, including higher levels of recombination (Soriano et al. 1987; Lien et al. 2000; Kondo et al. 2001), greater abundance of repetitive DNA (Hinch et al. 2014; Smeds et al. 2014), and differences in GC content (Montoya-Burgos et al. 2003). These studies focused on organisms with diploid sexual chromosome systems (XY and ZW), whereas in a large number of taxa including many red, brown, and green algae, land plants and fungi, sex is determined during the haploid phase of the life cycle (UV systems; Bachtrog et al. 2011). Many of the theoretical predictions made for diploid sexual systems are also relevant to UV sexual systems, for example concerning the evolution of recombination suppression and the maintenance of sex differences in allele frequencies in the PAR (Immler and Otto 2015). Some effects, such as the potential of sex differences in selection to drive gene differentiation in the PAR, are expected to be stronger in UV systems because the U and V chromosomes occur only in females and males, respectively (in contrast with the X, for example, which can occur in males and females). At present however, few empirical data are available for haploid sexual systems to test these various predictions.

We have recently shown that the UV sex chromosomes of the brown alga Ectocarpus have a small nonrecombining SDR, despite being at least 70 My old, and that this region is bordered by two relatively large PARs (Ahmed et al. 2014). Here, we show that the PARs of these chromosomes recombine at a similar rate to autosomal regions of the genome and yet exhibit many features typical of nonrecombining regions. The PARs are enriched in physically linked clusters of genes that are preferentially, and often exclusively, expressed during the sporophyte generation of the life cycle and many of these genes appear to have evolved since the Ectocarpales diverged from other brown algal lineages. A model is presented that provides a possible mechanism for the accumulation of these sporophyte-biased genes on the PARs.

Results

The PAR of the Ectocarpus Sex Chromosome Exhibits Unusual Structural Features

The PARs of the Ectocarpus sex chromosome (linkage group 30, LG30) represent about 2 Mb of sequence on each side of the 1 Mb SDR. We have previously noted that the PARs exhibit a number of structural differences compared with the autosomes. For instance, values for gene density, mean intron length, and percentage of GC content are intermediate between those of the autosomes and the SDR (Ahmed et al. 2014).

Several studies (Burt 2002; Jensen-Seaman et al. 2004) suggest that chromosome size should be taken into account when comparative analyzes of chromosome structure are carried out. In Ectocarpus, transposable element (TE) content tends to be negatively correlated with linkage group physical size (Spearman’s correlation test ρ = −0.113, P = 0.598) whereas gene density and GC percentage increase with chromosome size (Spearman’s correlation test ρ = 0.303, P = 0.151 and ρ = 0.284, P = 0.178, respectively). Consequently, to analyze in detail the unusual structural features of the Ectocarpus PARs, we compared the sex chromosome not only to the autosomal regions as a whole (all chromosomes apart from the sex chromosome) but also with one specific chromosome, linkage group 4 (LG04), which is of similar size (5.028 Mb) to the sex chromosome. For this comparison, all genes on LG30 and LG04 were manually curated to produce high-quality annotations for both chromosomes. Comparison of these two genomic regions showed that the PARs contained more TE sequences and lower gene density than LG04, and that GC content and the size of coding regions were significantly lower for the PAR, compared with LG04 (fig. 1A–D). Moreover, PAR genes tended to have longer introns, and fewer and smaller exons on average than genes on LG04 (fig. 1E–H). All of these differences were also detected when the PARs were compared with the autosomes (fig. 1A–D; supplementary fig. S1, Supplementary Material online), confirming that the PARs are unusual. Moreover, the features that distinguish the PARs from the autosomes were not confined to the regions that were close to the SDR. The PARs exhibited some structural heterogeneity along their length, with, for example, a significant negative correlation between TE content and gene content (Pearson’s correlation test r = −0.606, P < 0.01), but we found no evidence that the features that distinguish the PARs from the autosomes (gene structure, GC content, etc.) were more marked in the vicinity of the SDR (supplementary table S1, Supplementary Material online). These unusual structural features are therefore characteristic of the entire PARs.

F

Structural characteristics of the PAR compared with the SDR, LG04, and autosomes. (A) Percentage of TE calculated per supercontig; (B) gene density per supercontig; (C) percentage of GC per gene; (D) percentage of GC3 per coding sequence (CDS); (E) gene size; (F) CDS size per gene; (G) total intron length per gene; (H) number of exons per gene. Statistical differences were tested using pairwise Mann–Whitney U test. Letters shared in common between the groups indicate no significant difference.

Recombination along the Sex Chromosome

The structural analysis described above strongly indicated that the Ectocarpus PARs exhibit features resembling those of the nonrecombining SDR. Recombination is completely suppressed within the SDR of the Ectocarpus sex chromosome (Ahmed et al. 2014) but analysis of molecular marker segregation has confirmed that the PARs recombine during meiosis (Heesch et al. 2010). In order to build a more comprehensive recombination map of the Ectocarpus sex chromosome, an expanded segregating population of 280 individuals was genotyped with 23 LG30 markers. The average recombination rate for the PARs was 40 cM/Mb whereas the average recombination rate for autosomes was 23 cM/Mb. Comparisons of average rates between adjacent markers indicated that this difference was not significant (Mann–Whitney U test, P = 0.28). However, recombination events were unevenly distributed along the sex chromosome (fig. 2). Specifically, two regions of high recombination (one of them recombining at about ten times the genome average) were found on each side of the SDR. Recombination between markers within these peaks was significantly higher than the background recombination rate on the sex chromosome (Mann–Whitney U test, P = 0.0038). When markers within these recombination peaks were excluded, the PARs had an average recombination rate of 15.3 cM/Mb, which was still not significantly different from the genome average (Mann–Whitney U test, P = 0.388). Globally, we found no significant correlation between recombination rate and TE or gene content (Pearson’s correlation test, P > 0.05) along the length of the PARs, although there was a tendency for regions that exhibited higher recombination rates to have higher gene density and lower TE density (fig. 2).

F

Recombination frequency and distribution of TEs and gene density in the sex chromosome of Ectocarpus. The x axis indicates the physical position along the sex chromosome. Upper panel: y axis indicates the recombination rate (cM/Mb). Error bars represent 95% confidence intervals (CI). The recombination frequency around the SDR is unusually high. The average recombination between two adjacent markers on the PAR is 40.3 cM/Mb (18.8–66.9, 95% CI), compared with 217.2 and 95.0 cM/Mb for the peaks at the borders of the SDR. The red dashed line represents the average recombination frequency over the entire Ectocarpus genome. The black and red lines on the x axis indicate boundaries between supercontigs (sctgs) and the midpoints of supercontigs, respectively. Gray background rectangle above the upper graph indicates the distribution of generation-biased genes along the sex chromosome. Orange: sporophyte-biased genes; green: gametophyte biased genes. Horizontal bars and asterisks represent clusters of sporophyte-biased genes. See also figure 3. Gene and TE density along the Ectocarpus sex chromosome on the lower panel are represented by the solid red and dashed blue lines, respectively. Analysis of gene and TE density was performed by calculating the percentage of bases on each supercontig that are part of a gene or a TE, respectively. Vertical gray dashed lines indicate the boundaries between the PARs and the SDR. Note that the existing genetic map only allowed 70% of the genome sequence to be assigned to linkage groups (Heesch et al. 2010) and therefore we cannot exclude the possibility that missing scaffolds have led to an underestimation of the Mb/cM ratio in some regions of the sex chromosome.

F

PAR gene expression during different life cycle stages. (A) Average gene expression (log2RPKM) of all autosomes, LG04 (a linkage group of similar size to the sex chromosome) and PAR genes in male and female gametophytes (immature and fertile), and sporophytes. Letters shared between groups indicate no significant difference (Mann–Whitney U test, P < 6.0 × 10−5). (B) Heatmap showing the expression levels of PAR genes during different life cycle stages relative to their position on the sex chromosome (the SDR is excluded). Clusters of sporophyte-biased genes (also represented in fig. 2) are highlighted by asterisks. GA: gametophyte; SP: sporophyte (C) Enrichment of sporophyte-biased genes on the PAR compared with autosomes (χ2 test with Bonferroni correction, ***Padj = 6.03 × 10−5). (D) Expression of sporophyte-biased genes on autosomes, LG04 and PAR measured during the sporophyte (pink) and gametophyte (green) generations. Mean gene expression levels (log2RPKM) at several degrees of generation-bias (from FC > 1 to FC > 10) are shown. Error bars represent standard errors of the mean.

Genetic recombination rates along the PARs were also studied in a segregating family generated from two parental strains of another Ectocarpus species, Ectocarpus siliculosus lineage 1a (Stache-Crain et al. 1997), demonstrating that the PARs are also a recombining region in this sister species (supplementary fig S2 and table S2, Supplementary Material online).

Expression Patterns of PAR Genes during the Ectocarpus Life Cycle

The PARs contain 209 protein-coding genes. We investigated their patterns of expression, using RNA-Seq, at several stages of the life cycle of Ectocarpus, including male and female immature and fertile gametophytes, and different tissues of the sporophyte generation. The PAR genes exhibited significantly lower mean expression levels than genes on LG04 (median 5.88 RPKM (reads per kilo base pair per million) for the PARs compared with 11.16 RPKM for LG04; Mann–Whitney U test, P = 4.50 × 10−10) and than autosomal genes in general (median 9.88 RPKM for all autosomes; Mann–Whitney U test, P < 1.10 × 10−7) (fig. 3A). This difference in transcript abundance was particularly marked during the gametophyte generation, and slightly less marked during the sporophyte generation.

A heatmap representing the expression levels of the PAR genes revealed a striking pattern (fig. 3B; supplementary fig. S3, Supplementary Material online). Several clusters of genes had coordinated patterns of expression during the life cycle, including two clusters of PAR genes that were strongly upregulated during the sporophyte generation, and a cluster of genes that exhibited transcription below the detection limit (RPKM < 1), during both the gametophyte and the sporophyte generations. The sporophyte-biased gene clusters were localized in regions of the PAR that exhibited low levels of recombination (in supercontigs sctg_96 and sctg_266, fig. 2; supplementary table S3, Supplementary Material online). No other linkage group exhibited similar patterns of generation-biased gene clusters (supplementary fig. S3, Supplementary Material online).

To further analyze the relationship between genomic location and life cycle expression pattern, we carried out a genome-wide analysis to identify genes that were differentially expressed during the alternation between the sporophyte and gametophyte generations of the life cycle. About 25% of the genes in the Ectocarpus genome were significantly differentially regulated between the generations (fold change [FC] ≥ 2, false discovery rate [FDR] < 0.1), with slightly fewer sporophyte-biased genes (∼12% of the genome, 1,883 genes) than gametophyte-specific genes (∼13%, 2,083 genes). The PAR was found to be significantly enriched in genes that are upregulated during the sporophyte generation (χ2 test, Padj = 2.2 × 10−7, Bonferroni correction) (fig. 3C), while none of the autosomes exhibited a significant enrichment in sporophyte-biased genes (supplementary figs. S3 and S4, Supplementary Material online).

To examine the relationship between level of expression and degree of generation-bias, the sporophyte-biased genes on the PARs, on LG04, and on all autosomes were grouped according to fold-change in transcript abundance between the sporophyte and gametophyte generations, and the mean expression level (RPKM) of each group was plotted (fig. 3D). For LG04, and for autosomal genes in general, the degree of sporophyte-biased expression was determined by the level of expression in the gametophyte, so that their high fold difference correlated with low gametophyte expression. In contrast, all the sporophyte-biased genes on the PAR exhibited very low levels of expression in the gametophyte generation and the degree of sporophyte-biased expression (fold change) was determined both by attenuation of expression during the gametophyte generation and by the strength of expression during the sporophyte generation.

Two types of measurement can be used to describe the expression of a gene in a multicellular organism: the level of gene expression in terms of the number of transcripts present in a particular tissue, and the breadth of expression (τ), which measures how often the gene is expressed through the life cycle and/or in how many different tissues it is transcribed (Lipinska et al. 2015).

We calculated the breadth of expression of Ectocarpus genes using gene expression data collected for two types of tissues and at different stages of the life cycle. Globally, PAR genes exhibited greater expression specificity than either LG04 genes or autosomal genes in general (Mann–Whitney U test, P < 0.003) (supplementary fig. S5, Supplementary Material online). Sporophyte-biased PAR genes had τ values that were significantly higher than those of unbiased PAR or autosomal genes (Mann–Whitney U test, P < 6.5 × 10−5) (supplementary fig. S5, Supplementary Material online).

Fifty-one sporophyte-biased and 18 gametophyte-biased genes were identified on the PARs (supplementary table S3, Supplementary Material online). A significant proportion (∼50%) of the PAR sporophyte-biased genes were located in the two life cycle gene clusters mentioned above. In these clusters, 9 (sctg_266) and 13 out of 19 (sctg_96) contiguous genes exhibited sporophyte-specific expression (fig. 2; supplementary fig. S4, Supplementary Material online). Clustering analysis confirmed that the distribution of sporophyte-genes on the PAR was not random (Runs test, P < 2.2 × 10−16). The sporophyte-biased genes in the two clusters included a duplicated pair of adjacent genes for which there was one copy in each cluster (supplementary table S3, Supplementary Material online). The regions corresponding to the clusters, which are not closely linked to the SDR (fig. 2), exhibit lower recombination rates (on average 9 cM/Mb) than the average PAR rate. However, genes located on the clusters did not exhibit different characteristics from other sporophyte-biased genes located outside the clusters and did not differ from unbiased PAR genes (supplementary fig. S4, Supplementary Material online). The remaining sporophyte-biased genes were distributed along the PAR in triplets (1), pairs (5), or individually (16) (supplementary fig. S4, Supplementary Material online). Neither functional domains nor orthologues in public databases were detected for most of these genes and it was therefore not possible to identify any enrichment with respect to function. However, possible roles in protein–protein interactions (leucine rich repeats, tetratricopeptide repeats, or ankyrin repeats motifs) were predicted for 7 of the 51 sporophyte-biased PAR genes.

Fewer than 12% of the genes in the Ectocarpus genome (i.e., 1,947 genes) exhibits sex-biased gene expression (Ahmed et al. 2014), including 31 that are located in the PAR (supplementary fig. S4 and table S3, Supplementary Material online). This latter set of genes did not display any unusual structural characteristics compared with unbiased PAR genes (supplementary fig. S4, Supplementary Material online). There was also no significant tendency for generation-biased genes on the PAR to be also sex-biased (χ2 test, P = 0.25). Nonetheless, 12 of the 69 generation-biased on the PAR exhibited both generation- and sex-bias and there was a marked correlation between the precise type of life cycle generation-bias and the type of sex-bias: all seven of the genes that were both gametophyte-biased and sex-biased were male-biased, whereas four out of five of the genes that were both sporophyte-biased genes and sex-biased were female-biased (supplementary table S3, Supplementary Material online).

The Ectocarpus PAR Is Enriched in Young Genes

Recently evolved genes (referred to as “orphan” genes) tend to exhibit similar features to those that we observed for the PAR genes, including shorter coding regions, fewer exons, lower expression, and weaker codon bias compared with older genes (Arendsee et al. 2014; Palmieri et al. 2014). We therefore investigated whether gene age might be one of the factors underlying the unusual features of the PAR. Complete genome resources are currently insufficient to identify orphan genes, which are defined as having evolved within a species or group of species, in Ectocarpus but we were able to distinguish “young genes” from “old genes” by carrying out BLASTp comparisons with other complete stramenopile genomes, including the recently published Saccharina japonica genome (Ye et al. 2015), and sequences in the public databases. Young genes were defined as having no BLASTp match (10−4 E value cutoff) with any of these other genomes (indicating that they are likely to have evolved since the split from the most recent common ancestor, about 100 Ma; Silberfeld et al. 2010). The PAR was significantly enriched in young genes compared with all the autosomal linkage groups (34% vs. 10%, χ2 test with Bonferroni correction, P = 1.5 × 10−14). On average, young genes tended to be smaller and to have higher tissue specificity than old genes and their coding regions were smaller with lower codon adaptation index (CAI) and GC3 (Mann–Whitney U test, supplementary table S4, Supplementary Material online). When gene age was factored out by comparing only old genes or only young genes between the PAR and the autosomes, the PAR genes still exhibited higher percentage TE, lower GC content, longer gene size, shorter coding regions (significant for old genes only), shorter exons (significant for old genes only), and longer introns (Mann–Whitney U test, supplementary table S4, Supplementary Material online). Taken together, these analyzes indicated that the unusual features of the PAR genes could be explained in part by enrichment in young genes. However, when age is corrected for, PAR genes still exhibit markedly different features to autosomal genes. Interestingly, the proportion of young genes that showed generation-bias expression patterns was higher on the PAR than on the autosomes (52% vs. 28%, χ2 test, P = 4.18 × 10−7).

Evolution of the PAR Genes

The rate and pattern of evolution of Ectocarpus genes was analyzed by comparing sequences from the reference strain (Ectocarpus sp. lineage 1c Peru) with orthologous sequences from another Ectocarpus species (E. siliculosus lineage 1a). Compared with a set of 88 genes from LG04, the 84 PAR genes that were analyzed displayed, on average, significantly elevated values for nonsynonymous to synonymous substitution ratios (dN/dS) (Mann–Whitney U test, P < 0.001). However, when the generation-biased genes (40 genes) were removed from the data set, no significant difference in mean dN/dS ratios was detected between the PAR and autosomal gene sets (fig. 4A). Moreover, the sporophyte-biased PAR genes showed dN/dS ratios that were significantly higher than sporophyte-biased genes on LG04 (Mann–Whitney U test, P = 2.268 × 10−5), indicating that the increased evolutionary rates were related to the fact that these generation-biased genes were located on the PAR. The faster rate of evolution of the sporophyte-biased PAR genes was due to an increase in the rate of nonsynonymous substitutions (dN) and not to a decrease in the rate of synonymous substitution (dS) (fig. 4B and C) (Mann–Whitney U test, P < 0.01). Finally, note that although the average dN/dS ratio for unbiased PAR genes was similar to that of the autosomal gene set, the average values for both dN and dS were significantly greater than for the autosomal genes (Mann–Whitney U test, P < 0.01).

F

Rates of evolution of PAR (generation-biased and unbiased) genes compared with autosomal genes (LG04). Pairwise dN, dS, and dN/dS ratios were calculated by comparing orthologous gene sequences from Ectocarpus sp. (lineage 1c Peru) and Ectocarpus siliculosus (lineage 1a). (A) Ratio of nonsynonymous to synonymous substitutions (dN/dS). (B) Nonsynonymous substitutions (dN). (C) synonymous substitutions (dS). Letters shared between groups indicate no significant difference (Mann–Whitney U test, P < 0.01).

Interestingly, there was a weak, negative correlation between expression breadth and dN/dS for the PAR genes (Spearman’s ρ = 0.206, P = 0.0526). In other words, PAR genes with higher dN/dS tended to exhibit a lower breadth of expression.

Of the 40 sporophyte-biased PAR genes analyzed, 24 had dN/dS ratios that were greater than 0.5, which could be an indication of adaptive evolution (Swanson et al. 2004). To perform a maximum likelihood analysis of positive selection (PAML), we searched for orthologues of the sporophyte-biased genes using transcriptome data for two additional Ectocarpus species (Ectocarpus fasciculatus lineage 5b and Ectocarpus sp. lineage 1c Greenland; supplementary table S5, Supplementary Material online). Complete sets of four orthologues from the four species were obtained for only seven of the sporophyte-biased PAR genes and the PAML analysis was therefore carried out using these sets. For one of these comparisons both pairs of models (M1a–M2a, M7–M8) suggested positive selection (Esi0096_0082, ω = 0.86, P < 0.05).

Codon-usage bias has been observed in almost all genomes and is thought to result from selection for efficient and accurate translation of highly expressed genes (Kanaya et al. 2001). Optimal codons have been described for Ectocarpus (Ahmed et al. 2014) and a weak but significant correlation was noted between codon usage bias and gene expression levels (Wu et al. 2013). In accordance with these findings, the genes on the PARs, which were expressed at a lower level, on average, than autosomal genes (fig. 3A; Mann–Whitney U test, P = 6.06 × 10−5), showed significantly lower frequency of optimal codons (CAI) compared with autosomal genes (Mann–Whitney U test, P = 2.0 × 10−5). Interestingly, we found that genes in the regions close to the SDR tended to have higher CAI than more distal genes, although the significance is borderline (Spearman’s ρ = −0.15, P = 0.028).

However, when codon usage analysis was carried out specifically for the groups of sporophyte-biased and unbiased genes, the CAIs were significantly lower only for the sporophyte-biased genes on the PAR, compared with all other genes (Mann–Whitney U test , P < 0.004) (supplementary fig S6, Supplementary Material online). Analysis of the Drosophila genome identified a positive correlation between codon bias and recombination rate (Haddrill et al. 2007; Campos et al. 2012). Ectocarpus PAR genes located in regions with low recombination rates had significantly lower CAI (Mann–Whitney U test, P = 0.01879), but we found no significant difference in CAI for sporophyte-biased genes located in PAR with low versus average-to-high recombination rates. Therefore, the local recombination rate does not explain the low codon usage bias of the sporophyte-biased PAR genes (supplementary fig. S7, Supplementary Material online).

A Model for the Spread of Generation-Biased Alleles Located in the PAR

In XY or ZW systems, it has been argued that the excess of sex-biased genes often observed on X (or Z) chromosomes may result from SA selection (e.g., Vicoso and Charlesworth 2006). For example in XY systems, alleles with recessive or partially recessive effects that increase male fitness at a cost to female fitness are expected to spread more easily on the X than on autosomes; modifiers that decrease the expression of these genes in females may then spread, leading to an excess of male-biased genes on the X. We developed a theoretical model to explore whether a similar scenario (involving generation-antagonistic rather than SA selection) could explain the excess of sporophyte-biased genes observed on the PARs. This would imply that alleles increasing the fitness of sporophytes but with a fitness cost to gametophytes would spread more easily in the PAR than on autosomes.

The model (detailed in the supplementary material S1, Supplementary Material online) considers a selected locus located at a recombination distance r from the SDR of a UV sex-determination system, at which two alleles (denoted A and a) have different effects on the fitness of sporophytes, female gametophytes and male gametophytes. The different events of the life cycle are diploid selection, meiosis (recombination), haploid selection (within each sex), and fertilization (random union of gametes); the fitnesses of the different genotypes are given in table 1 (note that the results only depend on relative fitnesses within each ploidy phase and sex, as we assume that selection takes place independently among females, males, and sporophytes). The model is similar to that recently proposed by Immler and Otto (2015) but, while these authors explored conditions under which selection favors decreased recombination between a PAR locus and the SDR, we focus on the conditions for the spread of a rare allele (say allele a) at the selected locus, as a function of r, and the fitness effect of the allele on sporophytes (s) and on female (s) and male (s) gametophytes. We focus on generation-antagonistic alleles (s and s = (s + s)/2 have opposite signs), because the spread of such alleles may result in an increase in the frequency of genes that are differentially expressed in sporophytes and gametophytes (generation-biased genes) (table 1).

Table 1.

Fitnesses of the Different Genotypes at the Selected Locus.

	AA	Aa	Aa	A	a
Sporophyte	1	1+h s_d	1+s_d
Female gametophyte				1+s_f	1
Male gametophyte				1+s_m	1

Overall, our analysis (fig. 5) shows that genomic localization has little effect on the spread of alleles when selection is similar in both sexes (s ≈ s). However, when selection differs between the sexes (and in particular when the gametophyte-deleterious allele is neutral or slightly beneficial in one of the sexes), the model indicates that a sporophyte-beneficial allele benefits from sex-linkage, as this allows the allele to avoid being inherited by the sex where it is disfavored. Linkage to the SDR is also predicted to benefit the gametophyte-beneficial allele, but to a lesser extent since this allele still suffers from a fitness cost in the sporophytic generation. This can be seen on figures 5B and C: reducing the recombination rate between the selected locus and the sex-determining locus (from to solid curves for r = 0.5 to the dotted curves for r = 0.01) increases the parameter regions where alleles increase in frequency when rare, this effect being more marked for the sporophyte-beneficial allele (blue curves) than for the gametophyte-beneficial allele (red curves; note that the scale of the x axis is logarithmic). Therefore, taking into account the possibility of sex-differences in selection, being in the PAR expands the range of parameters allowing generation-antagonistic mutations to spread, but more so for sporophyte-beneficial, gametophyte-deleterious alleles than for gametophyte-beneficial, sporophyte-deleterious alleles. Again, this effect is generated by the fraction of generation-antagonistic mutations that is differentially selected in males and females. This model could thus explain the observed excess of sporophyte-biased gene expression in the PAR, assuming that reduced expression in gametophytes would have evolved secondarily to prevent the expression of alleles that are deleterious in at least one sex (note that complete linkage to the SDR would be another means to resolve this conflict).

F

Conditions for the spread of generation-antagonistic alleles. The sporophyte-beneficial allele (a) increases in frequency when rare above the blue curves, while the gametophyte-beneficial allele (A) increases in frequency when rare above the red curves. Solid curves: r = 0.5; dashed curves: r = 0.1; dotted curves: r = 0.01. The strength of selection in males is fixed to 0.1, while the different panels correspond to different values of : 0.05 (A), 0 (B), and − 0.05 (C). Note that swapping and would yield exactly the same results, as the model assumes that both sexes are equivalent.

Discussion

The Ectocarpus PAR Does Not Exhibit an Increased Recombination Rate Compared with Autosomes but Does Exhibit Local Peaks of Recombination

PARs play a critical role in successful progression through meiosis in the heterogametic sex of most plant and animal species because at least one crossover is required for correct segregation of the sex chromosomes (e.g., Burgoyne et al. 1992; Wai et al. 2012), generating a strong selective force to maintain recombination in the PAR. Accordingly, in human males, PAR1 has a crossover rate that is 17-fold greater than the genome-wide average. In contrast, the recombination rate in females, where recombination is between homologous X chromosomes, is comparable to the genome-wide average (Page et al. 1987; Flaquer et al. 2008). In UV systems, meiosis occurs in the sporophyte and, consequently, there is no male or female meiosis and all meiotic events involve pairs of U and V chromosomes in which recombination can only occur in the PARs. This feature of UV systems might be expected to further increase overall recombination rates in the PAR, but measurement of the recombination rate along the Ectocarpus PAR indicated a mean rate that was not significantly different from that of the rest of the genome. The absence of a detectable increase in recombination rate is probably explained by the large relative size of the PAR in Ectocarpus, which occupies approximately 80% of the sex chromosome (Ahmed et al. 2014). Similarly, the PAR of the blood fluke, and the PAR of the emu, which represent a high proportion (57% and 75%, respectively) of their sex chromosomes, both exhibit average recombination rates that are similar to those of autosomes (Criscione et al. 2009; Janes et al. 2009). Therefore, there appears to be a general tendency for PARs that constitute a large proportion of the physical size of the sex chromosome not to exhibit increased recombination rates compared with autosomes.

Although the mean recombination rate along the Ectocarpus PAR was comparable to that measured for autosomes, recombination mapping identified two peaks of elevated recombination rates flanking the SDR. Fine scale mapping of recombination rates along all the Ectocarpus linkage groups will be required to determine whether this type of recombination peak is a specific feature of the sex chromosome or if such peaks occur in autosomes (e.g., surrounding regions of reduced recombination such as the centromeres). Recombination hotspots at borders of SDRs have been described for species with XY or ZW sexual systems, including humans (Flaquer et al. 2008), mice (Soriano et al. 1987), blood flukes (Criscione et al. 2009), medaka fish (Kondo et al. 2001), emu (Janes et al. 2009), flycatcher birds (Smeds et al. 2014), Populus (Yin et al. 2008), and papaya (Wai et al. 2012). A similar phenomenon has also been observed in fungal mating type chromosomes (Hsueh et al. 2006).

The PARs Recombine at Similar Levels to the Rest of the Genome but Exhibit Structural Characteristics Typical of Nonrecombining Regions

The Ectocarpus PARs exhibit a number of features that are typical of genomic regions with reduced levels of genetic recombination (Charlesworth D and Charlesworth B 2005), including increased TE content, decreased gene density, smaller average coding sequence size, larger average intron size, higher gene GC content, higher rates of both synonymous and nonsynonymous substitution (and higher dN/dS ratios), and lower average gene expression levels compared with autosomes. Paradoxically, despite these features, the mean recombination rate measured for the PAR was not significantly different from that of the autosomal part of the genome. Moreover, we found no evidence that the majority of the PAR genes (excluding sporophyte-biased genes) contained higher levels of suboptimal codons than autosomal genes. However, note that PAR gene coding regions are significantly shorter than those of autosomal genes and this might counteract any tendency for suboptimal codons to accumulate, because selective pressures on codon usage are typically stronger for genes that encode short proteins (Duret and Mouchiroud 1999). The PAR was found to be enriched in young genes compared with autosomes but while the presence of these genes contributes to some extent to the unusual features of this region, this enrichment alone does not explain all the unusual structural features of the PAR.

We considered possible evolutionary mechanisms that might explain these unusual structural and functional features of the PAR and its constituent genes. Genetic linkage to the SDR is expected to influence the evolution of the PAR, but the effect should be limited to regions of the PAR that are very close to the SDR, unless selection is very strong (Charlesworth et al. 2014). This was not the case for the Ectocarpus PAR, as the unusual structural features were characteristic of the entire PAR and were not limited to regions adjacent to the SDR. To date, no mechanisms have been proposed which would allow the SDR to influence the evolution of linked, recombining regions over the distances observed here. It is not clear at present, therefore, whether the unusual structural features of the Ectocarpus PAR are related in some way to the presence of the SDR on the same chromosome or if they indicate that the evolutionary history of the PAR has been different from that of the other autosomes.

Preferential Accumulation of Sporophyte-Biased Genes on the PAR

The Ectocarpus PAR is enriched in sporophyte-biased genes compared with the autosomes and these sporophyte-biased genes appear to be evolving in a different manner to the other genes on the PAR. PAR genes in general showed elevated levels of both synonymous and nonsynonymous mutations compared with autosomal (LG04) genes whereas the sporophyte-biased PAR genes showed highly elevated rates of nonsynonymous mutations but a similar synonymous mutation rate to unbiased autosomal (LG04) genes. The elevated rate of nonsynonymous substitutions could be indicative of adaptive evolution, and indeed a signature of positive selection was detected for one out of the seven sporophyte-biased PAR genes that could be analyzed for this feature. However, while positive selection may be driving the evolution of some of the sporophyte-biased genes, this is unlikely to be the case for all of them. The set of sporophyte-biased PAR genes had a reduced content of optimal codons compared with an autosomal gene set, suggesting that the majority of these genes are under relaxed purifying selection. One possible explanation for the accumulation of nonoptimal codons in these genes is that they may escape haploid purifying selection (Lewis and Benson-Evans 1960; Lewis 1961; Lewis and John 1968), as they are completely silent during the gametophyte generation. Consequently, alleles with suboptimal codons will be masked in diploid heterozygous individuals and will not be selected against during the haploid phase.

Another possibility is that the lack of expression of the sporophyte-biased PAR genes during the gametophyte generation leads to relaxed selection due to the reduced breadth of expression of these genes. Breadth of expression, that is, the degree of tissue or developmental stage specificity, is known to effect nonsynonymous substitution rates (Duret and Mouchiroud 2000). However, this hypothesis alone is not sufficient to explain the higher evolutionary rates of sporophyte-biased genes, because gametophyte-biased PAR genes, which also have a reduced breath of expression, had similar nonsynonymous mutation rates to an average PAR gene.

Mathematical modeling was used to identify evolutionary mechanisms that might explain the preferential accumulation of sporophyte-biased genes in the PAR. Consistent with a recent model proposed by Immler and Otto (2015), we show that generation-antagonistic alleles spread more easily on the PAR than on autosomes if selection differs between males and females. The model presented here may explain our empirical observations that generation-biased genes accumulate preferentially on the PAR, provided that differences in expression between generations result from generation-antagonistic selection. However, note that there is evidence that the relationship between sex-biased gene expression and SA selection is complex (Innocenti and Morrow 2010; Parsch and Ellegren 2013) and this is likely also to be the case for the relationship between generation-biased gene expression and generation antagonistic selection. Generation-biased gene expression may therefore only be an approximate proxy for generation-antagonism.

Our model also predicts that sporophyte-beneficial, gametophyte-detrimental alleles tend to benefit more from linkage to the SDR than gametophyte-beneficial, sporophyte-detrimental alleles, in situations where selection is much weaker in one sex than in the other. Such a process might explain the prevalence of sporophyte-specific genes in the Ectocarpus PAR. Although this phenomenon should occur predominantly in regions that are tightly linked to the SDR (as the influence of the SDR is predicted to decrease rapidly with genetic distance), it may extend over a larger proportion of the PAR if the reproductive system involves partial clonality or inbreeding (thereby reducing effective recombination rates). Note that a recent field study identified both sexual populations and populations that were reproducing asexually (Couceiro et al. 2015), consistent with significant levels of asexual reproduction occurring under some conditions.

Young genes are 3-fold more abundant in the PARs than in autosomes. This enrichment is likely to be due to a combination of factors. As new genes are often derived from TEs (Tautz and Domazet-Loso 2011; Arendsee et al. 2014) the higher density of TEs in the PARs may play a role by permitting a higher rate of creation of new transcribed loci. This hypothesis is supported by the fact that, compared with the young autosomal genes, a greater proportion of the young genes in the PARs share homology with elements in the repeated fraction of the Ectocarpus genome (46.3% compared with 31.7%, Mann–Whitney U test, P = 0.038). Note, however, that additional factors are likely to be operating because this mechanism does not explain why the young PAR genes are enriched 2-fold in sporophyte-biased genes compared with young autosomal genes. Novel, transcribed loci are thought to arise at a high frequency in the genome but most of these loci are thought to be subsequently lost unless they are stabilized by selective forces (Tautz and Domazet-Loso 2011; Arendsee et al. 2014). It is possible that the mechanism considered in our model (where the excess of sporophyte-biased genes on the PAR results from the spread of sporophyte-beneficial, gametophyte-detrimental alleles) promotes the emergence of new genes with sporophyte-biased expression in the PAR. However, this mechanism alone does not seem sufficient to explain the high proportion of young PAR genes that are generation-biased (52%), as it seems unlikely that such a high proportion of the selectively advantageous new genes have generation-antagonistic effects.

Sporophyte-Biased Genes in the PAR Occur in Clusters

Almost half of the sporophyte-biased PAR genes are located in two gene clusters that are highly enriched in sporophyte-biased genes. Clustering of genes with related functions does occur in eukaryotic genomes, although to a lesser extent than in prokaryotes (Williams and Hurst 2002; Mugford et al. 2013), but the Ectocarpus genome as a whole does not exhibit unusually high levels of functional clustering (Cock et al. 2010). At present it is not clear what mechanisms led to the formation of these gene clusters on the PAR. Gene duplication has not played a major role in the evolution of these clusters although there are paralogous pairs of two genes across the two clusters. The model presented in this manuscript provides a possible mechanism for the accumulation of sporophyte-biased genes near the SDR and this could lead to clustering. However, neither cluster is adjacent to the SDR, although it is possible that the clusters have translocated to their current positions as a result of sex chromosome rearrangements.

Conclusion

We provide the first detailed analysis of the structural and evolutionary features of the PAR of a pair of UV sex chromosomes. We show that this PAR recombines at a rate that is not different from any other region of the genome, but remarkably, exhibits a number of structural and evolutionary features that are typically associated with regions of suppressed recombination. The PAR has significantly accumulated clusters of genes that are differentially expressed during the sporophyte versus gametophyte generation of the life cycle, and these generation-specific genes exhibit clear signs of accelerated evolution. We propose a mechanism that may explain some of the exceptional evolutionary features of these regions compared with autosomes.

Materials and Methods

Ectocarpus Strains and Culture Conditions

Ectocarpus strains were cultured as described (Coelho et al. 2012) and details are provided in supplementary material S1, Supplementary Material online. Supplementary table S5, Supplementary Material online, describes the Ectocarpus species used in this study. Note that, currently only three species are recognized within the genus Ectocarpus (E. siliculosus, E. fasciculatus, and E. crouanorium) (Peters et al. 2010) but there is increasing evidence that the taxa E. siliculosus represents a complex of several species. As the type specimen for E. siliculosus was isolated in England, we refer to the non-European strains related to E. siliculosus (such as the Peruvian and Greenland strains) as “Ectocarpus sp.”. The E. sp. lineage 1c Peru is the reference species of Ectocarpus used for the genome sequencing project and genetic map (Cock et al. 2010; Heesch et al. 2010). To study PAR recombination in an additional Ectocarpus species we used E. siliculosus lineage 1a. E. sp. lineage 1c Greenland and E. fasciculatus lineage 5b were used to evaluation of rates of gene evolution.

Generation of a Fine Recombination Map

A segregating population of 60 individuals that had been used for the genetic map (Heesch et al. 2010) and additional 220 individuals from a segregating population derived from a cross between strains Ec494 (male) and Ec568 (female) (Ahmed et al. 2014) were used to more precisely estimate recombination frequencies across the PAR. Simple sequence repeat markers for each of the 23 supercontigs of the sex chromosome (LG30) have been described previously (Heesch et al. 2010; Ahmed et al. 2014), and additional markers are described in supplementary table S2, Supplementary Material online, and in supplementary material S1, Supplementary Material online.

RNA-Seq

RNA-Seq analysis was carried out to compare the relative abundances of PAR gene transcripts at several different developmental stages of the life cycle (immature and fertile male and female gametophytes and two tissues of the sporophyte generation, namely basal filaments and upright filaments). The RNA extractions and processing of sequenced reads were performed as previously described in Ahmed et al. (2014) and Lipinska et al. (2015) (see supplementary table S6, Supplementary Material online, for the sequencing and mapping statistics and supplementary material S1, Supplementary Material online, for details of the Materials and Methods).

Differential expression analysis between male and female gametophytes, as well as between gametophyte (male and female libraries as replicates) and sporophyte was performed with the DESeq package (Bioconductor) (Anders and Huber 2010) using an adjusted P value cutoff of 0.1 and a minimal fold-change of 2 (see supplementary material S1, Supplementary Material online, for more details). The PAR was also analyzed for the presence of duplicated genes. Duplicated gene pairs were detected as described in Cock et al. (2010).

Evaluation of Rates of Gene Evolution

To estimate evolutionary rates of PAR genes, we searched E. siliculosus lineage 1a genomic data for orthologues of E. sp. lineage 1c Peru genes by retaining best reciprocal BLASTn matches with a minimum e value of 10 × 10−10. Sequences that produced a gapless alignment that exceeded 100 bp were retained for pairwise dN/dS (ω) analysis using PAML (codeml, F3x4 model, runmode = −2). To detect PAR genes under positive selection, we used transcriptomic and genomic data from four different Ectocarpus species (detailed in supplementary material S1 and table S5, Supplementary Material online). Nucleotide alignments (with a minimum length of 100 bp) were constructed using ClustalW implemented in Mega6 (Larkin et al. 2007; Tamura et al. 2013), curated manually when necessary and transformed to the PAML4 required format using perl fasta manipulation scripts (provided by Naoki Takebayashi, University Alaska Fairbanks). Nonsynonymous (dN) and synonymous (dS) rates were estimated by the maximum likelihood method available in CODEML program (PAML4 package). Effective number of codons and CAI were calculated using CAIcal server (http://genomes.urv.es/CAIcal/) (Puigbo et al. 2008).

Classification of “Old” and “Young” Genes

To determine the effect of gene age on various structural parameters, Ectocarpus genes were classified as “old genes” or as “young genes” based on the presence or absence, respectively, of homologous sequences in seven complete stramenopile genomes or in the NCBI database (excluding Ectocarpus sequences; February 2015). For the stramenopiles, BLASTp searches were carried out against the following complete deduced proteomes: Thalassiosira pseudonana (diatom; Thaps3 assembled and unmapped scaffolds, http://genome.jgi-psf.org/Thaps3/Thaps3.download.ftp.html; Armbrust et al. 2004), Phaeodactylum tricornutum (diatom; Phatr2 assembled and unmapped scaffolds, http://genome.jgi-psf.org/Phatr2/Phatr2.download.ftp.html; Bowler et al. 2008), Aureococcus anophagefferens (Pelagophyceae; http://genome.jgi-psf.org/Auran1/Auran1.download.ftp.html; Gobler et al. 2011); Nannochloropsis oceanica (Eustigmatophyceae; https://bmb.natsci.msu.edu/BMB/assets/File/benning/genome_assembly.txt; Vieler et al. 2012), Nannochloropsis gaditana (Eustigmatophyceae; http://www.nature.com/ncomms/journal/v3/n2/full/ncomms1688.html; Radakovits et al. 2012), Phytophthora capsici (oomycete; http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3551261; Lamour et al. 2012), and S. japonica (http://124.16.129.28:8080/saccharina/; Ye et al. 2015). Recent estimates indicate that all these species diverged from the Ectocarpales lineage more than 100 Ma (Brown and Sorhannus 2010; Silberfeld et al. 2010). Genes were classified as old genes if their protein sequences detected a BLASTp match with an E value of less than 10−4 in any of the subject genomes.

Supplementary Material

Supplementary material S1, tables S1–S7, and figures S1–S7 are available at Molecular Biology and Evolution online (http://www.mbe.oxfordjournals.org/).