Fig 1.
Schematics of the two scenarios of transcriptional divergence between species.
In each panel we show the expression level over time of human and mouse orthologs, representing transcriptionally divergent genes we study in this work. In each of these panels, the human orthologous genes is induced following immune stimulation (such as following dsRNA or IFN treatment), while the mouse orthologous gene remains transcriptionally unchanged. The difference between the two scenarios is in the expression level of the mouse gene: In the left panel, the mouse gene level remains low or the gene is not expressed before and after stimulation. In the right panel, the mouse ortholog is already expressed in high levels before stimulation, and remains in this level following stimulation. The two panels represent functionally different scenarios: In the left panel, the mouse ortholog is not expressed and is not part of the immune defense program, unlike its human ortholog. In the right panel, the mouse ortholog is constitutively highly expressed, such that it can immediately function against invading viruses without the need to be induced, unlike the human ortholog that its levels are increased only after infection. The silhouettes were created in BioRender.(2025) https://BioRender.com/i59u303.
Fig 2.
A cross-species in vitro immune stimulation analysis shows an overall similarity of transcriptional response across primate and rodent cells with accumaltive diveregnce.
(A) Phylogenies of the two cross-species fibroblast dsRNA-stimulation systems used in this analysis. Top: 4-species stimulation system: Dermal fibroblasts from two primates and two rodents were profiled with and without dsRNA stimulation using RNA-seq, along with ChIP-seq of histone marks and IFNB stimulation profiling [11]. Bottom: 10-species stimulation system: Dermal fibroblasts from nine primates and mouse were profiled with and without dsRNA stimulation using RNA-seq [19]. (B) Scatter plot of principal component analysis (PCA) of expression levels (count values from Salmon mapping [55]). The proportion of variance explained by the principal components is indicated in parentheses. (C) Hierarchical clustering of pairwise Spearman’s rank correlations between fold change in response to dsRNA-stimulation values in 1-to-1 orthologous genes that are DE in at least one of the species (2,771 and 7,711 genes, top and bottom, respectively). The silhouettes were created in BioRender.(2025) https://BioRender.com/i59u303.
Fig 3.
A method to identify divergent genes induced in one species and constitutive in the other.
(A) Two separate DE analyses between human and mouse samples were performed in stage 1: One in the control-unstimulated conditions (top) and another in the dsRNA-stimulated conditions (bottom). (B) A representative volcano plot of DE analysis between ortholog gene expression of human and mouse in samples taken in the same condition, as described in stage 1. Each dot represents DE values (Fold Change and Q-value) of a one-to-one orthologous gene between human and mouse. Colored dots represent genes that differ in expression between human and mouse (FDR-corrected P-value<0.001 and with |Fold Change| > 1), either in the control-unstimulated condition or in the dsRNA-stimulated condition. Orange dots are genes more highly expressed in mouse than in human, and blue are genes expressed more highly in human than in mouse in this specific condition. (C) We employed a filtering procedure to identify genes uniquely upregulated only in one species from the blue and orange sets: First, we filtered genes that differ between human and mouse in both dsRNA-stimulation and control-unstimulated conditions (these genes are genes whose divergence is unrelated to the antiviral response since they are higher in one species regardless of immune stimulation). Next, we excluded genes that are not significantly upregulated in response to dsRNA stimulation in either human or mouse (this ensures removal of genes that are insignificantly upregulated in response to dsRNA treatment in at least one species). The silhouettes were created in BioRender.(2025) https://BioRender.com/i59u303.
Fig 4.
Four groups of human-mouse transcriptionally divergent genes and their patterns of expression.
(A) A volcano plot of DE analysis between human and mouse in dsRNA-stimulated conditions. Colored dots show gene sets after the filtering procedure described in Fig 3. The right group (in purple) includes 224 genes induced in stimulation in mouse, but remain constitutively low in human (‘human constitutive-low’). The left group (light blue) includes 166 genes induced in stimulation in human, but remain constitutively low in mouse (‘mouse constitutive-low’). Bottom: Gray bars illustrate each group’s pattern of expression in human and mouse orthologs in both stimulated and unstimulated conditions. (B) The same as in A but comparing human and mouse in the control-unstimulated condition. The right group (turquoise) includes 104 genes induced in stimulation in human, but remain constitutively high in mouse (‘mouse constitutive-high’). The left group (light turquoise) includes 84 genes induced in stimulation in mouse, but remain constitutively high in human (human constitutive-high). Bottom: Gray bar plots illustrate each group’s pattern of expression in human and mouse orthologs in both conditions. (C) An upset plot showing intersection between various gene groups used in this work. The silhouettes were created in BioRender.(2025) https://BioRender.com/i59u303.
Fig 5.
Hierarchically clustered heatmaps based on gene expression levels.
Hierarchical clustering of gene expression data from all human and mouse individuals in the 4-species system in both stimulated and unstimulated conditions. In each panel, one the four previously defined divergent genes groups is shown, with expression levels in the human and mouse orthologs: (A) ‘human constitutive-low’, (B) ‘mouse constitutive-low’, (C) ‘human constitutive-high’ and (D) ‘mouse constitutive-high’. Columns are colored by species (blue and orange for human and mouse, respectively) and by conditions (dark and light for stimulated and unstimulated conditions, respectively) and represent one individual each. Gene levels are shown on a scale based on log(TPM). The clustering indicates that genes have similar levels in three ‘types’ of data, while the fourth is significantly different. For example, in A, both stimulated and unstimulated human samples cluster with the mouse unstimulated samples, while the stimulated mouse is the outgroup, with significantly higher levels of this gene set. This agrees with the patterns observed in Fig 4. The silhouettes were created in BioRender.(2025) https://BioRender.com/i59u303.
Fig 6.
Evolutionary characteristics of genes that transcriptionally diverge between human-mouse in dsRNA response.
Distributions of (A) percentage of sequence identity between human and mouse orthologs, (B) ratio of non-synonymous to synonymous substitutions (dN/dS), (C) gene evolutionary age (higher values denotes older age), and (D) rate of gene gain and loss across vertebrates (p-values for each gene’s rate are shown), for (from left to right): the set of DE genes (in grey) that includes all genes significantly upregulated in either human and/or mouse in response to dsRNA stimulation (FDR-corrected P-value<0.01 and FC > 0), human constitutive-low, mouse constitutive-low, mouse-constitutive-high, human-constitutive high. Group colors, sizes and genes are as in Fig 4. FDR-corrected P-values are shown, one-sided Mann–Whitney tests were performed for each of the 4 groups against all DE genes. For comparison, the values of the group of all 1-to-1 orthologs are also shown.
Fig 7.
Human-mouse divergence in dsRNA response is recapitulated at the chromatin level, in IFN response and across different cell types.
(A) Fraction of genes whose pattern of promoter activity agrees with gene expression, matching one of the 4 divergent groups (colored as in the legend) versus the fraction of genes among the rest of DE genes that display this pattern (in grey). For example, ~ 6% of human constitutive-low genes show the expected pattern at the chromatin level, and ~2% of all other DE genes show the same pattern (two left-most bars). FDR-corrected Fisher’s exact test P-values are shown. These tests show that the pattern of transcriptional response is recapitulated at the chromatin level in all four groups since the fraction of genes whose chromatin accessibility pattern agrees with the transcriptional pattern is higher than that observed in the set of all other DE genes (this is significant in three of four groups tested). (B) logFC distribution values based on DE analyses between human and mouse in dsRNA stimulation (before the arrow) and in IFN stimulation (after the arrow). In both cases the genes are partitioned into three boxplots based on three groups defined previously: “human constitutive-low”, “mouse constitutive-low” and “all other DE genes”. Thus, the same three groups of genes are shown before and after the arrow, but their FC values varies based on either dsRNA- or IFN-stimulation DE values. The analysis shows that the separation in logFC values in the dsRNA-stimulation between the three groups is also observed in the IFN-stimulation, suggesting similar transcriptional divergence between dsRNA response and IFN response. (C) The same as in B, but with basal conditions used in the DE analyses and showing: “mouse constitutive-high”, ‘human constitutive-high”, “all other DE genes”. In both B and C, FDR-corrected P-values are shown for one-sided Mann–Whitney tests, performed under the hypothesis that the FC distribution of the left group is higher than the right group.
Fig 8.
Divergence in dsRNA response across different primate and rodent species.
(A) Top: logFC distribution of differential expression analysis in dsRNA-stimulation condition between human and mouse (left) and rhesus macaque and rat (right) from the 4-species system. Separation in both boxplots is according to the groups originally defined from human-mouse divergence: “mouse constitutive-low”, “human constitutive-low”, “All other DE genes” (as defined in Figs 3 and 4). The right panel, which is based on rhesus versus rat data shows that the orthologs of two species diverge in a similar manner to the human and mouse orthologs. Thus, both left and right panels use the same three groups of orthologous genes, but the FC values are based on either human-mouse (left) or macaque-rat (right) DE analysis. Bottom: The same as in Top, but with DE preformed between human and mouse in basal conditions and showing “human constitutive-high”,”mouse constitutive-high” and “all other DE genes”. (B) The same as in A, but comparing the human-mouse divergence in response to dsRNA from the 4-species system to each one of the 9 primates-mouse divergence in response to dsRNA from the 10-species system (the primate species used is shown in the bottom, and its data is always compared to mouse). In both (A) and (B) we observed that the patterns of relative FC between the compared groups recapitulate the patterns from the human-mouse dsRNA comparison. In all analyses, FDR-corrected P-values are shown for one-sided Mann–Whitney test, performed under the hypothesis that the FC distribution of the left group is higher than the right group in each of the analyses.