Abstract
During the evolution of sex chromosomes, the Y degenerates and its expression gets reduced relative to the X and autosomes. Various dosage compensation mechanisms that recover ancestral expression levels in males have been described in animals. However, the early steps in the evolution of dosage compensation remain unknown, and dosage compensation outside of the animal kingdom is poorly understood. Here, we studied the evolutionarily young XY system of the plant Silene latifolia. We show that dosage compensation is achieved in this plant by a genomic imprinting mechanism where the maternal X chromosome is upregulated in both males and females. This is the first time such a situation is observed in any organism. It could be nonoptimal for females and may reflect an early stage of dosage compensation evolution, which strikingly resemble the first stage of the path proposed by Ohno for the evolution of X inactivation in mammals.
One Sentence Summary In the plant Silene latifolia, reduced expression from the Y chromosome is compensated by upregulation of the maternal X chromosome in both sexes.
Sex chromosomes have repeatedly evolved from a pair of autosomes in various taxa (1). Y chromosomes initially have similar gene content compared to the X, but ongoing Y degeneration leads to reduced expression of Y-linked genes and to eventual gene loss. This is thought to negatively impact male fitness due to imbalanced doses in protein networks involving both autosomes and sex chromosomes. Dosage compensation allows to counterbalance the negative effects of Y degeneration and takes diverse shapes in various animal species (2–4). Depending on the study, dosage compensation sometimes refers to ancestral expression recovery in XY males, sometimes to expression equality between sexes (hereafter sex equality) and sometimes to both phenomenons. Although, sensu stricto, dosage compensation should refer to the recovery of balanced expression levels between autosomes and sex chromosomes in both sexes. For clarity, we will use the terms ancestral expression recovery and sex equality whenever appropriate. In Drosophila, the X chromosome is upregulated specifically in males, resulting in complete ancestral expression recovery along with sex equality (5). In placentals, on the other hand, one X is inactivated in females, resulting in sex equality without ancestral expression recovery, except for a few dosage-sensitive genes whose expression was doubled (6–10). The situation is similar in Caenorhabditis elegans, where both X chromosomes are downregulated in hermaphrodites and only a few genes have their expression doubled for complete dosage compensation (11). In some mammals, the differential expression of alleles depending on the parent of origin, hereafter genomic imprinting (12), serves as a dosage compensation mechanism (6). Indeed, while in most placentals X inactivation is random, in marsupials it is consistently the paternal X chromosome that is inactivated (13). This is also the case in mice placenta (14). However, it is unknown whether genomic imprinting has been involved in the early steps of dosage compensation evolution in mammals. In a seminal work, Ohno hypothesized that X chromosome expression should be doubled first (15), reestablishing ancestral expression levels in XY males. Then, the resulting upregulation in XX females would select for X inactivation or downregulation in a second step. However, in order to understand these early steps of dosage compensation evolution and the order in which they happened, species with young sex chromosomes must be studied.
The plant Silene latifolia is an ideal model to study early steps of sex chromosome evolution thanks to its X/Y pair that evolved ~4 Mya (16). However, it appears that dosage compensation is poorly known in plants (17). In S. latifolia, only sex equality has been studied so far. Equal expression levels were observed for males and females for some genes in spite of Y expression degeneration (18–23). However, the mechanisms through which sex equality is achieved and whether ancestral expression was recovered in S. latifolia males are unknown. In order to address these questions, we have developed a methodology relying on (i) the use of an outgroup without sex chromosomes as an ancestral autosomal reference (17) in order to determine whether X chromosome expression increased or decreased in S. latifolia, (ii) the use of specific methods to study allele-specific expression while correcting for reference mapping bias (discussed in 17), and (iii) a statistical framework to quantify dosage compensation (17).
Because only 25% of the large and highly repetitive S. latifolia genome has been sequenced so far (23), we used an RNA-seq approach based on the sequencing of a cross (parents and a few offspring of each sex), to infer sex-linked contigs (i.e. contigs located on the non-recombining region of the sex chromosome pair) (24). X/Y contigs and X-hemizygous contigs (X-linked contigs without Y allele expression) were inferred separately for three tissues: flower buds, seedlings and leaves (Supplementary Table S2). About half of the inferred sex-linked contigs could be validated by independent data (Supplementary Table S2 and Materials and Methods). Contigs with significant expression difference between males and females were removed for further analysis (Supplementary Table S2 and Materials and Methods) as they are likely to be involved in sex-specific functions and therefore are not expected to be dosage compensated (25). X-hemizygous contigs have previously been characterized as having partial sex equality (23) and are analyzed separately due to limitations of the RNA-seq approach. Our results suggest that X-hemizygous genes are less dosage sensitive than X/Y genes, which could explain their poor level of dosage compensation (Supplementary Text S1).
Paternal and maternal allele expression levels in males and females were estimated for sex-linked and autosomal contigs in S. latifolia after correcting for reference mapping bias (Materials and Methods). Then, the corresponding allelic expression levels in one or two closely related outgroups without sex chromosomes were used as a reference to polarise expression changes in S. latifolia. For autosomal contigs, expression levels did not differ between S. latifolia and the outgroups (Figure 1). This is due to the close relatedness of S. latifolia and the outgroups (~5My, Supplementary Figure S1), and validates the use of the outgroups as a reference for ancestral expression levels. We used the ratio of Y over X expression levels in males as a proxy for Y degeneration in order to categorize X/Y contigs. As the Y allele degenerates (paternal allele in blue in Figure 1), the expression of the X allele in males increases (maternal allele in red in Figure 1). This is the first direct evidence for ancestral expression recovery in S. latifolia, i.e. ancestral expression levels are reestablished in males in spite of Y expression degeneration. In females, the maternal X allele expression increases with Y degeneration (gray bars in Figure 1), in a way similar to the maternal X allele in males. The paternal X allele in females, however, maintained its ancestral expression level regardless of Y degeneration (black bars in Figure 1). Overall, sex equality is not achieved in S. latifolia due to upregulation of sex-linked genes in females compared to ancestral expression levels. All results were confirmed on two other tissues, as well as when analysing independently validated contigs only (although statistical power is sometimes lacking due to the low number of validated contigs, Supplementary Figures S2-S7).
The maternal X allele is upregulated both in males and females (Figure 1 and Supplementary Figures S2-S7), suggesting genomic imprinting is playing a role in dosage compensation in S. latifolia. In order to statistically test this observation, we used a linear regression model with mixed effects (Materials and Methods). The outgroups were used as a reference and expression levels in S. latifolia were then analyzed while accounting for the variability due to contigs and individuals. The joint effect of the parental origin and the degeneration level was estimated, which allowed to compute the difference in expression between the maternal and paternal alleles in females for different degeneration categories (Figure 2). In autosomal contigs, the maternal and paternal alleles are expressed in a similar way in females, showing a global absence of genomic imprinting for these contigs. However, for X/Y contigs, the difference between the maternal and paternal X in females increases with Y degeneration. Results were confirmed on two other tissues as well as when analysing independently validated contigs only (Supplementary Figures S8-S13).
Previous studies that showed sex equality between sexes in S. latifolia could have been explained by simple buffering mechanisms, where one copy of a gene is expressed at a higher level when haploid than when diploid, due to higher availability of the cell machinery or adjustments in gene expression networks (23, 26, 27). However, the upregulation of the X chromosome we observe here in S. latifolia males cannot be explained by buffering mechanisms alone, otherwise the maternal X in females would not be upregulated. Rather, our findings suggest a specific dosage compensation mechanism relying on genomic imprinting has evolved in S. latifolia. This mechanism looks like an evolutionary convergence with marsupials, although the mechanistic details are very different between the two species (in marsupials the paternal X is inactivated, while in S. latifolia the maternal X is upregulated).
An exciting challenge ahead will be to understand how upregulation of the maternal X is achieved in S. latifolia males and females at the molecular level. Chromosome staining suggests that DNA methylation could be involved. Indeed, one arm of one of the two X chromosomes in females was shown to be hypomethylated, as well as the same arm of the single X in males (28) (Supplementary Figure S14). Based on our results, we hypothesize that the hypomethylated X chromosome corresponds to the maternal, upregulated X. Unfortunately, parental origin of the X chromosomes was unknown in this study (28). It would be insightful in the future to study DNA methylation patterns in S. latifolia paternal and maternal X chromosomes, along with a homologous autosomal pair in a closely related species without sex chromosomes. The methylation pattern observed by chromosome staining suggests that dosage compensation in S. latifolia could be a chromosome arm-wide phenomenon. However, the position of genes along the X chromosome is required to fully confirm this hypothesis with expression data.
Female upregulation of the X chromosome compared to autosomes was previously reported in Tribolium castaneum, but later shown to be due to biases from inclusion of gonads in whole body extracts (4). Our findings thus provide the first characterization of such a situation, which could be deleterious for females. This suggests that reduced expression of sex-linked genes in males is more deleterious than upregulation in females. The young age of S. latifolia sex chromosomes may explain this potentially suboptimal pattern, and sex equality might evolve only at a later stage, following the evolutionary path proposed by Ohno for placentals (15).
Author contributions
Aline Muyle, Niklaus Zemp, Alex Widmer and Gabriel Marais conceived the study and experimental design. Niklaus Zemp and Alex Widmer prepared and sequenced the plant material. Aline Muyle ran SEX-DETector on the RNA-seq datasets for the three tissues, analysed the data, prepared Tables and Figures and wrote the Supplementary Material with inputs from other authors. Niklaus Zemp generated the X chromosome genetic map (with help from Aline Muyle for the mapping and genotyping part). Radim Cegan, Jan Vrana and Roman Hobza did the Y chromosome flow cytometry sorting and sequencing. Clothilde Deschamps did the first assembly of the sorted Y chromosome and improved it with RNA-seq data with the help of Cecile Fruchard. Aline Muyle did the blasts to validate the inferences of SEX-DETector. Raquel Tavares did the GO term analysis. Aline Muyle and Frank Picard did the statistical analyses of the data. Gabriel Marais and Aline Muyle wrote the main text of the manuscript with inputs from other authors.
Acknowledgments
This project was supported through an ANR grant ANR-14-CE19-0021-01 to G.A.B.M and SNSF grants 141260 and 160123 to A.W.