Ancient transposable elements transformed the uterine regulatory landscape and transcriptome during the evolution of mammalian pregnancy

Abstract

A major challenge in biology is determining how evolutionarily novel characters originate; however, mechanistic explanations for the origin of new characters are almost completely unknown. The evolution of pregnancy is an excellent system in which to study the origin of novelties because mammals preserve stages in the transition from egg laying to live birth. To determine the molecular bases of this transition, we characterized the pregnant/gravid uterine transcriptome from tetrapods to trace the evolutionary history of uterine gene expression. We show that thousands of genes evolved endometrial expression during the origins of mammalian pregnancy, including genes that mediate maternal-fetal communication and immunotolerance. Furthermore, thousands of cis-regulatory elements that mediate decidualization and cell-type identity in decidualized stromal cells are derived from ancient mammalian transposable elements (TEs). Our results indicate that one of the defining mammalian novelties evolved from DNA sequences derived from ancient mammalian TEs co-opted into hormone-responsive regulatory elements distributed throughout the genome.

Figure 1. Evolution of the Endometrial Transcriptome in Tetrapods

(A) Parsimony reconstruction of gene expression gain and loss events in the tetrapod endometrium. The numbers above branches indicate the number of genes recruited into (+) or lost (−) from endometrial expression in the stem lineages of Mammalia (light blue), Theria (blue), and Eutheria (red). Branch lengths are proportional to gene expression recruitment and loss events inferred by Wagner parsimony (see inset scale bar). (B) Expression of recruited genes in the major endometrial cell types. Recruited, total number of genes recruited into endometrial expression in the stem lineages of Mammalia, Theria, and Eutheria (n = 2,502).

Figure 2. Recruited Genes Are Enriched in Immune, Signaling, and Reproductive Functions

(A) Top 10 anatomical systems in which the expression of recruited genes is enriched. Anatomical system, center; −log₁₀ p value of enrichment (hypergeometric), left. Stacked bar chart shows the number of genes recruited into endometrial expression in the Mammalian (light blue), Therian (blue), and Eutherian (red) stem lineage that are expressed in each anatomical system. (B) Manhattan plot of −log₁₀ p values (hypergeometric test) for mouse knockout phenotypes enriched among recruited genes. Phenotypes are grouped by anatomical system. Horizontal red line indicates the dataset-wide false discovery rate (FDR q value = 0.1). (C) Word cloud of mouse knockout phenotypes enriched among genes recruited into endometrial expression in the Mammalian (light blue), Therian (blue), and Eutherian (red) stem-lineage. Abnormal phenotypes are scaled to the log₂ enrichment of that term (see inset scale). (D) Manhattan plot of −log₁₀ p values (hypergeometric test) for GO terms enriched among recruited genes. Go terms are grouped into biological process, cellular component, and molecular function. Horizontal red line indicates the dataset-wide false discovery rate (FDR q value = 0.1). (E) Word cloud of GO terms enriched among genes recruited into endometrial expression in the Mammalian (light blue), Therian (blue), and Eutherian (red) stem lineage. GO terms are scaled to the log₂ enrichment of that term (see inset scale).

Figure 3. Expression Dynamics of Recruited Genes

(A) Tissue specificity of recruited and ancestrally expressed genes across 27 tissues (all) and how specific the expression those genes are in the uterus. 0, not tissue specific; 1, tissue specific; p values shown for paired t tests, lines connecting comparisons are colored by the t value of the comparison (see scale at right). (B) Expression level (log₂ transcripts per million [TPM]) of recruited and ancestrally expressed genes; p values shown for paired t tests. (C) Relative expression (log₂ fold change [FC] of secretory samples relative to proliferative samples) of recruited and ancestrally expressed genes throughout the human menstrual cycle. Mean ± variance. (D) Response of recruited and ancestrally expressed genes expressed in human endometrial stromal cells to trophoblast conditioned media (TCM); log₂ fold change (FC) in gene expression 3 and 12 hr after treatment relative to control conditioned media (0 hr). Mean ± variance. (E) Relative expression (log₂ fold change [FC]) of recruited and ancestrally expressed genes in the endometria of women with unexplained infertility compared to fertile controls; p values shown for F test for variance.

Figure 4. Ancient Mammalian TEs Are Enriched in Regulatory Elements Active in Decidualized Human Endometrial Stromal Cells

(A) Proportion of DNaseI-seq, FAIRE-seq, H3K27ac-seq, and H3K4me3-seq peaks that do not overlap annotated TEs (gray), overlap “ancient mammalian” TEs, i.e., those that integrated into the genome in the stem lineage of Mammalia, Thera, or Eutheria (green), and other TEs (light green). (B) Proportion of TEs in DNaseI-, FAIRE-, H3K27ac-, and H3K4me3-seq peaks and the human genome (hg18) by age class. (C) Word cloud of the 150 most enriched TEs among ancient mammalian elements, in DNaseI-, FAIRE-, H3K27ac-, and H3K4me3-seq peaks. The size of the transposon names corresponds to its enrichment (see inset 3-fold scale). Colors indicate lineage in which the TE originated: Mammalian-specific (light blue), Therian or Mammalian (green), Therian specific (blue), Eutherian or Therian (purple), Eutherian specific (red). (D) Number of enriched TEs among ancient mammalian elements in DNaseI-, FAIRE-, H3K27ac-, and H3K4me3-seq peaks by age class.

Figure 5. Genes Associated with Ancient Mammalian TE-Derived Regulatory Elements Are More Strongly Differentially Regulated during Decidualization than Genes without TE-Derived Regulatory Elements

(A) MA-plot showing changes in gene expression 48hrs after treatment of human endometrial stromal cells with cAMP/Progesterone. Red dots indicate significantly differentially expressed genes (n = 7,337, N = 16,551). (B) Cartoon of the GREAT association rule used to associate regulatory elements with nearby genes: regulatory element peaks located +5,000 to −1,000 bp of transcription start sites (TSSs) were defined as the basal regulatory domain (filled boxes represented below each TSS) for each gene (regardless of other nearby genes). Basal regulatory domains were extended in both directions to the nearest gene’s basal domain (thin lines) but no more than 100 kb. Under this association rule, for example, the regulatory element peak located between Gene B and Gene C would be associated with these genes but not Gene A. (C) Recruited genes associated with ancient mammalian TE-derived regulatory elements (+) are more strongly differentially regulated upon cAMP/progesterone treatment (decidualization) than genes without TE-derived regulatory elements (−). Genes are grouped into all genes expressed in DSCs (gray), Eutherian recruited genes (red), Therian recruited genes (blue), Mammalian recruited genes (light blue), and ancestrally expressed genes (black). F is the ratio of variances from a two-sample F test, and p is the significance of the two-sample F test.

Figure 6. Ancient Mammalian TEs Are Enriched in Binding Sites for TFs that Mediate Hormone Responsiveness and Endometrial Cell Identity

(A) Distribution of enriched (red) and depleted (blue) TFBSs in ancient TE-derived segments of FAIRE-, DNaseI-, H3K27ac-, and H3K4me3-seq peaks compared with TF ChIP-seq peaks with no TE overlap. Fifty and 75 TFBSs were significantly enriched and depleted, respectively, at FDR = 5%. (B) Word cloud of TFBSs enriched in ancient mammalian TE segments of ChIP-seq peaks compared to peaks with no TE overlap. Colors indicate TFs that mediate hormone responses (purple), remodel chromatin (light purple), have known functions in endometrial cells or that mediate immune responses (green), or with general regulatory functions (light green). Data shown for ≥ 1.2-fold enriched TFs at FDR = 5%. (C) Word cloud of TF motifs enriched in ancient mammalian TE segments of ChIP-seq peaks compared with peaks with no TE overlap. Colors indicate motifs enriched in FAIRE- (red), DNaseI- (tan), H3K27ac- (dark blue), and H3K4me3-seq peaks (light blue). Data shown for ≥ 3-fold enriched motifs at FDR = 1 × 10⁻⁴ and E1 × 10⁻⁴.

Figure 7. Ancient Mammalian TEs Globally Remodeled PGR Binding Site Architecture across the Genome

(A) The number of PGR ChIP-seq peaks that contain Mammalian-specific (light blue), Therian-specific (blue), and Eutherian-specific (red) TEs in primary cultures of hESCs treated with 10 nM 17β-estradiol, 100 nM medroxyprogesterone acetate, and 1 mM 8-bromo-cAMP. (B) Word cloud of ancient mammalian TEs enriched within PGR ChIP-seq peaks. Colors follow (A). Inset scale (10×) shows 10-fold enrichment. Only elements with R2-fold enrichment are shown. (C) Genes associated with ancient mammalian TE-derived PGR binding sites (+ PGR) are more strongly differentially regulated in hESCs upon decidualization than genes associated with ancient mammalian TE as a whole (+) and genes without TE-derived regulatory elements (−). Eu Recruit, Eutherian recruited gene; Th Recruit, Therian recruited gene; Mam Recruit, Mammalian recruited gene; Anc, ancestrally expressed gene. F is the ratio of variances from a two-sample F test. (D) Genes associated with ancient mammalian TE-derived PGR biding sites (+ PGR) are more strongly dysregulated by PGR knockdown in DSCs than either genes associated with ancient mammalian TE as a whole (+) and genes without TE-derived regulatory elements (−). F is the ratio of variances from a two-sample F test.