Expression reduction in mammalian X chromosome evolution refutes Ohno's hypothesis of dosage compensation

Abstract

Susumu Ohno proposed in 1967 that, during the origin of mammalian sex chromosomes from a pair of autosomes, per-allele expression levels of X-linked genes were doubled to compensate for the degeneration of their Y homologs. This conjecture forms the foundation of the current evolutionary model of sex chromosome dosage compensation, but has been tested in mammals only indirectly via a comparison of expression levels between X-linked and autosomal genes in the same genome. The test results have been controversial, because examinations of different gene sets led to different conclusions that either support or refute Ohno's hypothesis. Here we resolve this uncertainty by directly comparing mammalian X-linked genes with their one-to-one orthologs in species that diverged before the origin of the mammalian sex chromosomes. Analyses of RNA sequencing data and proteomic data provide unambiguous evidence for expression halving (i.e., no change in per-allele expression level) of X-linked genes during evolution, with the exception of only ∼5% of genes that encode members of large protein complexes. We conclude that Ohno's hypothesis is rejected for the vast majority of genes, reopening the search for the evolutionary force driving the origin of chromosome-wide X inactivation in female mammals.

Fig. 1.

Comparison of expression levels of orthologous genes from human and chicken. (A) Human X:chicken XX expression ratios for one-to-one orthologs in six male (M) or four female (F) tissues, when the medians of human AA:chicken AA expression ratios are normalized to 1. The null hypothesis of X:XX = AA:AA is rejected in each tissue (P < 10⁻⁹, Mann–Whitney U test). There are 325 gene pairs for calculating X:XX and 10,171 gene pairs for calculating AA:AA. The central bold line shows the median, the box encompasses 50% of genes, and the error bars include 90% of genes. (B) Ratios of chicken XX:AA median expressions and human X:AA median expressions in 10 tissues. Error bars show 95% confidence intervals estimated by bootstrapping genes 1000 times. The null hypothesis of X:AA = XX:AA is rejected in each tissue (P ≤ 0.001, bootstrap test). (C) Same as A, except that only genes actively expressed (RPKM >1) in the indicated chicken tissue are considered. The null hypothesis of X:XX = AA:AA is rejected in each tissue (P < 10⁻⁹, Mann–Whitney U test). (D) Same as B, except that only genes actively expressed in the indicated chicken tissue are considered. The null hypothesis of X:AA = XX:AA is rejected in each tissue (P < 0.001, bootstrap test).

Fig. 2.

Evolution of across-tissue expression profiles for human X-linked genes that have platypus and chicken one-to-one orthologs. (A) Expression levels, measured by log₂(RPKM), of 241 human X-linked genes and their one-to-one orthologous genes in platypus and chicken across 10 tissues. Each row is a gene. (B) Comparable expression profile similarity between chicken and platypus and that between chicken and human for human X-linked genes that have platypus and chicken one-to-one orthologs. Expression profile similarity between species is measured by Pearson’s correlation in expression level across 10 tissues. Each circle is a gene, and the numbers of circles above and below the diagonal are not significantly different (P > 0.09, binomial test).

Fig. 3.

Testis-biased expressions of evolutionarily young X-linked genes in humans. (A) Expression levels of old and young human X-linked genes across 10 tissues. (B) Proportions of human old and young X-linked genes that are actively expressed (RPKM >1) in each tissue. Error bars show one SE of the observed proportion. (C) Expression evolution of 12 young human X-linked genes generated by autosome (A)–derived gene duplication since therians diverged from the platypus. RPKM values relative to the median RPKM of all genes in a transcriptome are shown. In A and C, each row is a gene, and log₂(TBI) is presented in the last column. TBI, testis bias index.

Fig. 4.

Comparison between X-linked genes and autosomal genes for protein abundances in a human cell line using (A) human genes that have one-to-one chicken orthologs and (B) human genes whose chicken one-to-one orthologs are likely to be house-keeping. The median protein abundance for each gene bin is shown. The X-linked and autosomal genes are separately divided into 100 equal-size bins, based on the ranks of protein expression levels, and the top 25 bins are compared.

Fig. 5.

Comparison of expression levels of orthologous genes from opossum and chicken. (A) Opossum X:chicken XX expression ratios for one-to-one orthologs in 11 tissues, when the medians of opossum AA:chicken AA expression ratios are normalized to 1. The null hypothesis of X:XX = AA:AA is rejected in each tissue (P < 0.0003, Mann–Whitney U test). There are 197 gene pairs for calculating X:XX and 9,612 gene pairs for calculating AA:AA. The central bold line shows the median, the box encompasses 50% of genes, and the error bars include 90% of genes. (B) Ratios of chicken XX:AA median expressions and the corresponding opossum X:AA ratios. Error bars show 95% confidence intervals estimated by bootstrapping genes 1,000 times. The null hypothesis of X:AA = XX:AA is rejected at P = 0.001, 0.001, <0.001, <0.001, 0.039, 0.001, 0.005, <0.001, 0.003, <0.001, and 0.005, respectively, for the 11 tissues shown from left to right (bootstrap test). (C) Same as A, except that only those genes that are actively expressed in chicken tissues (RPKM >1) are considered. The null hypothesis of X:XX = AA:AA is rejected in each tissue (P < 0.0003, Mann–Whitney U test). (D) Same as B, except that only actively expressed genes in chicken tissues are considered. The null hypothesis of X:AA = XX:AA is rejected at P = 0.001, <0.001, <0.001, <0.001, 0.008, <0.001, 0.016, 0.005, 0.002, <0.001, and <0.001, respectively, for the 11 tissues from left to right (bootstrap test).

Fig. 6.

Human X-linked genes encoding members of large protein complexes are subject to up-regulation, compared with those encoding members of small complexes. Each open symbol is a protein complex, whereas the solid symbols show median values. Within each complex, the median expression ratio of human X-linked genes to their chicken orthologs (X:XX) is divided by the median expression ratio of autosomal genes (AA:AA). The null hypothesis of equal log₂((X:XX):(AA:AA)) values between large and small complexes is true with probabilities of 0.03, 0.17, 0.07, 0.01, 0.05, 0.05, 0.01, 0.03, 0.26, and 0.002, respectively, for the 10 tissues shown from left to right (Mann–Whitney U test).