Tracing the cis-regulatory changes underlying the endometrial control of placental invasion

Abstract

Among eutherian (placental) mammals, placental embedding into the maternal endometrium exhibits great differences, from being deeply invasive (e.g., humans) to noninvasive (e.g., cattle). The degree of invasion of placental trophoblasts is positively correlated with the rate of cancer malignancy. Previously, we have shown that fibroblasts from different species offer different levels of resistance to the invading trophoblasts as well as to cancer cell invasion. Here we present a comparative genomic investigation revealing cis-regulatory elements underlying these interspecies differences in invasibility. We identify transcription factors that regulate proinvasibility and antiinvasibility genes in stromal cells. Using an in vitro invasibility assay combined with CRISPR-Cas9 gene knockout, we found that the transcription factors GATA2 and TFDP1 strongly influence the invasibility of endometrial and skin fibroblasts. This work identifies genomic mechanisms explaining species differences in stromal invasibility, paving the way to therapies targeting stromal characteristics to regulate placental invasion, wound healing, and cancer dissemination.

Keywords: cancer malignancy; endometrium; evolution of cancer; placenta; stomal invasibility.

Fig. 1.

Modeling the genomic basis of interspecies expression differences using cis-regulatory elements. (A) Illustration showing the observation that placental invasibility and cancer malignancy are inversely related, along the epitheliochorial to hemochorial axis. (B) Schematic showing the gain and loss of binding sites in the promoter regions of a given gene in different species and its effect on gene expression. β is a coefficient estimating the extent of effect gain of a binding site has on gene expression. (C) The distribution of P values for individual TFBS motifs, when fit using the model described in B where occurrence is tested for an effect on interspecies gene expression differences. The overrepresentation of low P values beyond the distribution expected by chance (red line) is evidence of a real effect. (D) Histogram for the effect size (beta) found for different TFBS motifs that were found to be statistically significant (FDR < 0.05). Beta represents the amount the scaled (square root of TPM) expression changes with one binding site, on average. Positive beta implies an enhancer effect, while negative beta implies a repressor effect on gene expression of downstream genes. (E) TFs whose putative binding sites affect the variation in expression across selected eutherian species. Each point represents a single TF binding motif. The statistical significance is plotted in terms of the FDR on the y axis, while the effect size beta (increase or decrease in expression) for additional binding sites is plotted on the x axis. Statistically significant motifs are labeled in green. (F) Illustration of the effect of downstream effect of MGA binding sites on the expression of individual genes. Here each point represents a gene whose expression was independently tested for a linear relationship to the number of binding MGA binding sites. The gene-specific effect size (beta) and significance (FDR) are plotted on the x and y axes, respectively. (G) Distribution of the number of binding site motifs matched in the promoter regions of genes in different species, shown here for the MGA binding site motif. (H) Illustration of the effect of MGA binding sites on the expression of LRCH3, across different species listed in the order of placental invasiveness. (I) The R-squared (fraction of variance explained) for the most statistically significant TFs affecting global ESF gene expression across species. The colors correspond to the effect size, with blue for an overall increase of gene expression with the particular TF binding in its promoter region and orange for an overall decreasing effect for additional binding.

Fig. 2.

Finding genes correlated with the ELI of the endometrial stroma across eutherian mammals. (A) Phylogenetic hierarchical representation of selected eutherian species whose ESFs were analyzed in the manuscript. The tree is balanced for invasive and less invasive placental species. Color legend denotes the extent of placental invasion ranked from 0 to 3. (B) Schematic explaining the ELI principle, showing that among the epitheliochorial placental mammals, the stroma has evolved to resist trophoblast invasion with secondary manifestation in other tissues resulting in decreased cancer malignancy. (C) Statistical significance of the expression of individual genes for correlation with invasibility among different species, evaluated using a simple linear model (x axis) and a phylogenetic linear model (y axis). P values are larger under the phylogenetic linear model, but similar ordering of the genes under the two methods is visible. (D) Illustration of the linear fit test used to probe whether the gene expression of a particular gene (in this case, PKD1) in ESFs across species is correlated with the invasibility. (E) Illustration of a gene whose expression is anticorrelated with invasibility. (F and G) Heat maps of expression across species for the genes most correlated with the invasibility phenotype across species, either positively (ELI^up genes) or negatively (ELI^dn genes).

Fig. 3.

Genomic basis for the regulation of genes correlated with stromal invasibility among eutherian species. (A) The distribution of P values for the effect of individual TFBS motifs, tested for the effect on expression of ELI-related genes, with the red line representing the distribution expected by chance. (B) The TFs whose binding sites affect the expression of genes purportedly involved in the invasibility of the endometrium (the ELI-related genes). Statistical significance (y axis, FDR) with the effect size (x axis, beta). (C) The fraction of variance (R-squared) of the expression across genes explained only by the occurrence of a particular binding motif aggregated over a balanced set of the most significant 100 ELI^up and 100 ELI^dn genes, shown here for the most significant TF binding motifs. The colors of the bars represent the effect size, green being higher. (D) Illustration of the linear fit used to calculate the statistical significance and effect size for an affected gene, UACA. Shown are the changes in gene expression across species with increasing binding sites for the TF GATA2. Species are ordered on the basis of their extent of placental invasion. (E) Specific genes likely to be regulated by GATA2. The fraction of variance explained (y axis) and the effect size (x axis, beta) are calculated for a linear model relating the (scaled) gene expression across species for a single gene to the number of GATA2 binding site motifs in the promoter regions among the different species. (F) ELI-correlated (ELI^up) and (G) ELI-anticorrelated (ELI^dn) genes whose variance is most explained by the number of specific binding site sequences. Beta is the common effect size found for the binding site sequence found for all ELI genes. For individual TF–gene pairs, a large extent of interspecies variance in the direction (or against the direction) of placental invasiveness could be explained by the cis-regulatory TFBS elements. (H) qPCR analysis showing reduction in expression of GATA2 72 h after Cas9/sgRNA transfection, (I) resulting in reduced expression of two genes with the highest R2 values, Sash1, Stdb1. n = 3 samples; error bars represent SEM. Statistical significance for H and I established at *P < 0.05, **P < 0.01, ***P < 0.001. (J) TFBSs whose occurrence is related to the expression of genes experimentally tested (2) for their role in modulating invasibility. Each point represents a binding site sequence motif, with the mean variance explained on the y axis and the number of genes with P < 0.05 for the correlation on the x axis.

Fig. 4.

Differences in the TFBS explained regulation of genes correlated (ELI^up) or anticorrelated (ELI^dn) with placental invasiveness in ESFs. (A) Histograms of the effect sizes (beta) of different binding sites, when tested on only ELI^dn and ELI^up genes. Only effect sizes for TFBSs that are statistically significant (P > 0) in the respective genes (ELI^dn or ELI^up). (B) Scatterplot shows the effect sizes (beta) of individual TFs for ELI^dn (y axis) and ELI^up (x axis) genes. Names of some selected TFBSs are labeled. Selected genes with a high fraction of variance explained by TFBS motif occurrences are shown as examples in the four tables next to each quadrant.

Fig. 5.

TFs predicted to enhance invasibility of stromal fibroblasts functionally control invasion in multiple cellular contexts. (A) Method to transfer nanopatterns mimicking collagen fibers using polyurethane (PUA) replica mold with capillary force lithography. (B) Nanofabricated scaffolds are used as substrates to pattern invading cells and fibroblasts using poly(dimethylsiloxane) stencils, creating an interface between two cell populations. (C) Schematic showing a directional movement of invading fronts, accompanied by formation of invasive forks, also shown in (D) phase contrast images for breast epithelial cells (MCF10A1) invading into BJ fibroblasts within 24 h. Scale bars in D and E correspond to 400 μm. (E) Live cell microscopy snapshots of invading MCF10A1 cells into the BJ monolayer in 24 h, with BJ cells subjected to CRISPR/Cas9-based gene KOs for GATA2, and TFDP1 TFs. Scr refers to the untargeted scrambled control. (F) Quantification for the extent of aerial invasion, normalized to the initial interface length. (G) Cumulative measurement of invading fronts calculated as the fraction of pixels occupied at every location in the direction of invasion. (Left) Schematic explains the calculation of fraction occupied at two different locations. (H) Quantification for the depth of invasion by each fork within 24 h invading into the stromal barrier, here BJ fibroblasts with the given gene KOs. Shown is automated identification of the tip of the invading forks (red circles); solid red bar refers to the mean length of invasion, and black dashed lines refer to the 95% confidence intervals. (I) Normalized extent of invasion of HTR8s into hESF monolayers with gene KOs. (J) Depth of invasion by each HTR8 fork within 24 h invading into hESF monolayers. (Right) Representative tips of invading forks (red circles). (K) Normalized extent of aerial invasion of A375 melanoma cells in BJ fibroblasts with gene KOs. In G–K, error bars represent SDs, statistical significance by t tests established at P < 0.05, and denoted by *P < 0.05, **P < 0.01; ***P < 0.001, ***P < 0.0001. Each dot represents an independent imaged interface.