Recent evolution of the developing human intestine impacts metabolic and barrier functions

Abstract

Diet, microbiota, and other exposures place the intestinal epithelium as a nexus for evolutionary change; however, little is known about genomic changes associated with adaptation to a uniquely human environment. Here, we interrogate the evolution of cell types in the developing human intestine by comparing tissue and organoids from humans, chimpanzees, and mice. We find that recent changes in primates are associated with immune barrier function and lipid/xenobiotic metabolism, and that human-specific genetic features impact these functions. Enhancer assays, genetic deletion, and in silico mutagenesis resolve evolutionarily significant enhancers of Lactase (LCT) and Insulin-like Growth Factor Binding Protein 2 (IGFBP2). Altogether, we identify the developing human intestinal epithelium as a rapidly evolving system, and show that great ape organoids provide insight into human biology.

At birth, the intestinal tract of humans and other animals becomes exposed to the outside world. Epithelial cells lining the intestinal tract directly interface with a dynamic and unpredictable lumenal environment, facilitating nutrient absorption, coordinating systemic physiology, and forming a barrier to viruses, bacteria, and other microorganisms. Prior to birth, the gene regulatory networks (GRNs) that underlie these epithelial functionalities must orchestrate the development, differentiation, and maturation of diverse intestinal epithelial cell types that constitute this environmental interface. Tissue-resident intestinal epithelial stem cells (ISCs) emerge during development and maintain the capacity to differentiate into each one of the functional epithelial cell types (1–3), which have absorptive or secretory activity, into adulthood. The absorptive lineage includes diverse enterocytes and microfold cells, which absorb luminal nutrients and antigens, respectively (4–6). More recently characterized BEST4+ enterocytes have diverse roles in both the small intestine and colon, including facilitating mucus hydration, balancing pH and electrolyte levels, and small molecule metabolism (7, 8). The secretory lineage comprises mucus-producing goblet cells, antimicrobial peptide-producing Paneth cells, hormone-secreting enteroendocrine cells (EECs), and cytokine-secreting Tuft cells (4–6, 9–11). Each of these cell types may have different proportions, states, and functions over an individual’s lifespan (5, 6, 9, 12).

GRNs that orchestrate intestinal epithelium development and function have ancient origins and persist under continuous evolutionary pressure. While all vertebrates share a core set of stem, absorptive, and secretory cell type functionality, along with conserved anterior-posterior functional domains (13–16), the dynamic pressures related to diet composition, frequency of meals, and exposure to microorganisms have shaped the intestinal tract of vertebrate lineages throughout hundreds of millions of years of evolution (17, 18). In the past 15 million years, the human intestine has adapted to diverse environmental conditions since the divergence from chimpanzees and the other great apes. Recently, cooking and the transition from a hunter-gatherer culture to agricultural societies with extensive consumption of dairy, meat, and carbohydrate-rich crops has had a substantial impact on all cell types within the intestine. Deep and recent changes impact the modern human intestine morphologically, metabolically, and immunologically (19–24). Morphologically, the small intestine to colon volume ratio has expanded in humans compared to other great apes, whereas the overall relative size of the human gut has reduced (18, 25, 26). Some of these recent changes are known to impact intestinal epithelium function, including variants involved in Crohn’s disease (CD), inflammatory bowel disease (IBD), and hypolactasia (27–29). An evolutionary perspective on the developing human intestine therefore has human health relevance and is crucial for shedding light on modern human gastrointestinal disorders (30).

Human gut cell atlases at the single-cell transcriptome and chromatin accessibility level have expanded our understanding of the diversity of human intestinal cell types and their role in development, homeostasis, and disease (12, 31, 32). However, interpretation of human intestinal gene regulatory mechanisms in the context of evolution and how these mechanisms may differ between species is unclear. We lack a fundamental view of how intestinal epithelial cell type functions emerged in evolution and how they changed in the human lineage. Here, we set out to explore the evolutionary history of the developing human intestinal epithelium. We integrate a reference atlas of the developing human intestine covering transcriptome and accessible chromatin modalities and use comparative genomics to catalog ancient and recent features. Human and chimpanzee organoids and comparisons to human and mouse developing tissues elucidate human-specific features of intestinal epithelial development and we use functional assays to study several of these regions in developing human enteroids. Altogether, each intestinal epithelial cell type has distinct evolutionary pressures, and human-specific genetic changes impact a wide range of human intestinal physiology including metabolic and barrier function. Many of these changes likely prepared the human intestine for their particular external environment.

Evolutionary dynamics of the developing human intestinal epithelium

To provide a reference atlas of the developing human small intestine, we integrated single-cell multiome (transcriptome and accessible chromatin) data from the developing human proximal small intestine (11 and 18 post-conception weeks, PCW) with scRNA-seq data (8 to 19 PCW) (9, 32) using canonical correlation analysis (CCA) (33, 34) and projected the cells to a two dimension Uniform Manifold Approximation and Projection (UMAP) embedding (Fig. 1, A and B; table S1 and S2). The integrated atlas includes 52,928 high-quality single-cell or nucleus transcriptomes, including matched chromatin accessibility for 14,558 nuclei. We annotated 21 molecularly distinct cell clusters grouped into five major cell classes (epithelial, mesenchymal, immune, neural, and endothelial) (Fig. 1B and fig. S1, A to C, and Data S1(35)).

Fig. 1. Enterocytes of the developing human small intestinal epithelium are rapidly evolving.

A, Schematic describing human proximal small intestine (duodenum) specimens profiled using scRNA-seq (light gray) or scMultiome (dark gray). The data was used to assess nonsynonymous to synonymous substitution (dN/dS) ratios of expressed genes and deepest ancestor of accessible chromatin. B-C, UMAP of developing human intestinal single-cell transcriptome data with each cell colored by annotated cell class and numbered by the cell type (B, See also Fig. S1), and average normalized nonsynonymous to synonymous substitution ratio (dN/dS) in protein coding regions of expressed genes across primates (C). D, Violin plots show the distribution of average normalized dN/dS ratio per cell type. M1-M4, mesenchymal subtype 1–4. EN, enteric neuron. E, Gene ontology enrichment analysis for enterocyte enriched genes with high dN/dS scores. F, Ancestry enrichment plot with x-axis depicting nodes grouped into vertebrate (nodes 12–14), tetrapod (8–11), mammal (5–7), primate (1–4), and y-axis depicting the enrichment of intestine open chromatin regions mapped to their deepest ancestry. Colors depict cell class (top) or intestinal epithelial cell type (bottom). G, Heatmap shows the scaled enrichment of specific gene ontologies which is defined as the difference between the -log10-transformed binomial test Benjamini-Hochberg(BH)-corrected P values of the primate evolutionary group of regulatory regions and the mean of others.

To quantify the evolutionary forces on each cell type, we calculated the normalized expression weighted non-synonymous to synonymous substitution ratios (dN/dS) on the primate tree similar to previously described (36). We observed that epithelial and immune cells had higher dN/dS ratios than other cell types (Fig. 1, C and D and fig. S1D), indicating that expressed protein-coding genes of epithelial and immune cells likely experience less constraint than those of other cell types. Among the different epithelial cell types (in total 12,269 cells), including stem cells, enterocytes, tuft cells, BEST4⁺ cells, enteroendocrine cells (EECs), and goblet cells, enterocytes exhibited top dN/dS scores (Fig. 1D and fig. S1E). Genes with enriched expression in the epithelium but with broad expression in most epithelial cell types had lower dN/dS scores compared to the genes enriched in specific cell types (fig. S1F). This suggests constraints on core epithelial processes and different modes of selection, potentially additive across functions in different cell types, acting on cell type-specific functions. Enterocyte-enriched genes with high dN/dS scores are involved in lipid and xenobiotic metabolism (Fig. 1E), including Cytochrome P450 3A4 (CYP3A4), which has been extensively studied in the intestinal epithelium due to its ability to metabolize a wide variety of chemical compounds (37). Antibacterial humoral response genes also demonstrated high substitution rates, including Interleukin 32 (IL32), which functions as a proinflammatory cytokine (38). In contrast, ubiquitously expressed epithelial genes with low dN/dS scores are involved in general processes such as cell-cell adhesion, miRNA processing and regulation of macroautophagy (fig. S1H, and table S3). These analyses indicate that cells within the developing human intestinal epithelium have experienced distinct evolutionary pressures and that developing enterocytes rapidly evolved in primates.

To assess the evolutionary history of cell type regulatory repertoires, we inferred the evolutionary age of each accessible chromatin region from genome-wide alignments between 34 species, spanning more than 500 million years and 14 phylogenetic branches from the human-lamprey ancestor to extant humans (39) (Fig. 1A and fig. S2A, table S4 and Data S1) (see Methods). We highlight two intronic regions within the Basonuclin Zinc Finger Protein 2 (BNC2) locus with enterocyte and stem cell enriched chromatin accessibility and differential deepest ancestry (vertebrates and mammals) (fig. S2B). Single nucleotide polymorphisms (SNPs) within these regions are linked to Neanderthal introgression (40). We highlight additional ancient regulatory regions with ubiquitous epithelial accessibility (HES1 promoter (13)) or with cell type enrichment in enterocytes (TSHZ1 intron) or enteroendocrine cells (NPAS3 intron), and show that these regions maintain accessible chromatin in the zebrafish intestinal epithelium (13), supporting their ancient conservation (fig. S2C). We find that regions with epithelium-enriched accessibility have a deepest alignment to more recent evolutionary ancestors compared to regions enriched in other cell classes (endothelial, immune, mesenchymal and neural) (Fig. 1F and fig. S2D). Within the epithelial cell types, enterocyte-enriched regions are over-represented for originating in recent ancestors (Fig. 1F and fig. S2D). Consistent with these analyses, we observe that epithelial- and enterocyte-enriched regions have lower phastCons scores (fig. S2E). Taken together, epithelial- and enterocyte-enriched regulatory regions might experience less evolutionary constraints, while neuron- and EEC-enriched regulatory regions show higher genomic sequence stability.

Genomic Regions Enrichment of Annotations Tool (GREAT) (41, 42) analysis on open chromatin regions that originated in vertebrates (nodes 12–14), tetrapods (8–11), mammals (5–7), and primates (1–4) (table S3) revealed notable enrichments for each group. For example, vertebrate conserved terms were related to cell fate commitment, anterior-posterior patterning, and chromatin organization (fig. S2F), and included ancient regulatory regions near master regulators of intestinal cell type differentiation, such as MAF (43) and SPDEF (44, 45). Conversely, regions dated to more recent nodes within mammals and primates were enriched in terms related to barrier defense, nutrient transport, and metabolic activities, suggesting turnover or birth of regulatory regions controlling genes involved in response to a dynamic luminal environment (Fig. 1G). For example, SULT2A1 (Bile salt sulfotransferase) has enterocyte-enriched gene expression, and an enterocyte-enriched accessible chromatin region with ancestry only shared within primates (fig. S2G). The enzyme catalyzes the sulfate conjugation of many hormones, neurotransmitters, drugs, and xenobiotic compounds (46–48). Another example was SLC19A3 (also known as Thiamine-transporter-2; THTR2) which is responsible for the absorption of thiamine (Vitamin B1), a micronutrient involved in diverse metabolic functions that humans must acquire from external sources (fig. S2G) (49, 50). We observed that genes nearby regions with deepest ancestry only within primates had higher dN/dS scores than genes nearby regions with ancestry to more ancient nodes (fig. S2H). Altogether, these data suggest that the regulatory mechanisms of each intestinal epithelial cell type experienced different degrees of evolutionary change on various phylogenetic branches, and that enterocytes of the developing intestine have many recent changes that occurred within primates.

Human intestinal organoids recapitulate gene regulatory networks underlying in vivo intestinal development

We assessed if human pluripotent stem cell (PSC)-derived intestinal organoids (HIOs) (51) and developing intestine epithelial stem cell-derived enteroids (collectively termed organoids) recapitulate the gene regulatory architecture of the primary human developing intestinal epithelium. We generated single-cell transcriptome and accessible chromatin data from in vitro and transplanted HIOs (tHIOs) (51, 52) (5 cell lines; paired data from the same cell suspensions: scRNA-seq, 37,404 cells; scATAC-seq, 14,120 nuclei) and developing enteroids (scMultiome, 7,379 cells) (Fig. 2, A and B, fig. S3, A to C, and fig. S4A). While HIOs comprise epithelial and mesenchymal cells, with a small proportion of immune and neural cells (fig. S3, D and E, and fig. S4, B to D; Data S1), developing enteroids are purely epithelium. We identified stem cells, enterocytes, and BEST4+ cells, as well as secretory lineage cells consisting of goblet cells, enteroendocrine cells, and Paneth cells in tHIOs and developing enteroids (fig. S3, A, D and F). We integrated tHIO (7,481) and enteroid (6,612) small intestine epithelial cells with primary developing tissue epithelium using canonical correlation analysis (CCA) (33, 34) and visualized the integration in a UMAP embedding (Fig. 2C). We observed correspondence between cells from primary developing tissue and organoids and identified expressed genes and accessible chromatin regions with robust cell type enrichment independent of sample source (Fig. 2, D and E, fig. S3, G and H, fig. S5A; table S5).

Fig. 2. Human intestinal organoids recapitulate gene regulatory features underlying intestinal epithelial development.

A, Schematic depicting the protocol to generate transplanted PSC-derived human intestinal organoids (tHIO) and developing small intestine epithelial stem cell derived enteroids, and multiomic comparison of mRNA expression and chromatin accessibility measurements of epithelial cell types of each specimen. WNRIFAG : Wnt3a, Noggin, R-spondin-3, IGF1, FGF2, A-83–01, Gastrin. B, tHIO stained with H&E to highlight tissue organization (top, scale bar, 100 μm) and bright field image of developing enteroids (bottom, scale bar, 200 μm). C, Integrated UMAP representation depicting contribution of each organoid/tissue source of developing human small intestine (left), tHIO (middle) and developing enteroid (right) based on scRNA-seq data with cells colored by cell types. D, Representative scATAC-seq peaks associated with the promoter or nearby regulatory element of the epithelial cell types of the developing enteroid (top), developing human small intestine tissue (middle) and tHIO (bottom). CHGA, SLC15A1 and MUC2 peaks are associated with enteroendocrine cells, enterocytes, and goblet cells, respectively. E, Heatmap showing per-source scaled accessibility of putative cell type-specific cis-regulatory elements (CREs) consistently detected in primary developing intestine tissue, developing enteroid, and tHIOs (ESC, iPSC). Cell type colors follow panel C. F, Schematic illustrating the single-cell STARR-seq approach used to test the enhancer activity of candidate CREs in human intestinal enteroids. Candidate CREs were pooled and co-transfected with a constitutively expressed GFP reporter to enable selection of transfected cells prior to scRNA-seq and normalization of CRE activity. mP, minimal promoter; ORF, open reading frame; pA, polyadenylation sequence; CMV, cytomegalovirus promoter; GFP, green fluorescent protein. G, Heatmaps showing the expression of putative target genes of selected CREs (left), chromatin accessibility of selected CREs (middle), and CRE activity measured with the scSTARR-seq reporter assay (right) in developing enteroids. Two distinct CREs were measured for CYP3A4 and are defined as (up, upstream CRE) and (dn, downstream CRE). BNC2ma denotes the BNC2 intronic CRE with mammalian ancestry (fig. S2B).

Next, we used the scATAC-seq data to identify regions demonstrating cell type-specific chromatin accessibility and implemented a multiplexed enhancer assay based on single-cell STARR-seq (53) in developing enteroids to assess the enhancer activity of these candidate cis-regulatory elements (CRE)(Fig. 2F). This approach utilizes a pool of vectors containing the self-transcribing CREs inserted downstream of a minimal promoter and open reading frame (ORF), along with a control vector that constitutively expresses GFP to enable selection of transfected cells prior to scRNA-seq and normalization of CRE activity. We transfected developing enteroids, sorted GFP⁺ cells, and used single-cell transcriptome sequencing to estimate CRE activity per cell type. In total we profiled over 14,000 single cells across two independent experiments and identified stem cells, enterocytes, goblet cells and EECs (fig. S4 E–G). We quantified the average CRE activity per cell type and identified 14 putative CREs that demonstrated cell type-enriched enhancer activity consistent with gene expression and chromatin accessibility measurements (Fig. 2G). These results showcase that cell type-enriched accessible chromatin regions captured in our data are predictive of cell type-specific cis-regulatory enhancer function.

We next combined single-cell transcriptome and accessible chromatin data from tHIOs to infer and visualize a GRN underlying human intestinal epithelial cell type diversification (54) (table S6). This analysis illuminated core transcription factors previously reported to guide intestinal epithelial cell type differentiation, including ASCL2 (55, 56) and SOX9 for stem and progenitor cells (57, 58), NEUROD1 (59) and NEUROG3 (60) for enteroendocrine cells, c-MAF (43) for enterocytes, ATOH1 and SPDEF for goblet cells (44, 45) and less explored BEST4⁺ cells are predicted to utilize MEIS1 and RBPJ transcription factors in their GRNs (fig. S5, B to F). Overall, we observed strong gene regulatory correspondence between organoid and primary developing intestinal tissue, confirming induced pluripotent stem cell (iPSC)-derived intestinal organoids and developing enteroids as robust models of intestinal development. Altogether, these data show that regulatory networks underlying primary developing intestinal epithelium can be explored using organoid models to understand the development, maturation, and evolution of intestinal epithelial cell types.

Chimpanzee intestinal organoids model great ape intestinal development

To understand human features of the gut compared to our closest living relatives, we generated chimpanzee intestinal organoids (CIOs) by directing the differentiation of 6 chimpanzee iPSC lines into intestinal tissue in vitro using the same temporal series of growth factors that reconstitute human embryonic intestinal development transitioning from pluripotency to definitive endoderm and midgut/hindgut progenitors (51) (Fig. 3A, and fig. S6A). We note that chimpanzee intestinal organoids initiate midgut-hindgut budding around one day earlier than human counterparts. Following spheroid collection and embedding in matrigel, we observe coordinated differentiation of pseudostratified epithelium, emergence of mesenchymal cells and early villus morphogenesis in vitro (Fig. 3, B and C, fig. S6C). The CIOs establish apical-basal polarity (Fig. 3C), with diverse mesenchyme formation surrounding the epithelial layer (Fig. 3, B and C, fig. S6C) and possess small intestine identity demonstrated by CDX2+SOX9+ intestinal epithelial cells (Fig. 3D, fig. S6B), and absence of foregut (SOX2) and colon (SATB2) marker expression (fig. S6B).

Fig. 3. Chimpanzee intestinal organoids model great ape intestinal development.

A, Schematic showing the generation of chimpanzee intestinal organoids (CIOs) from pluripotent stem cells. Organoids mature after transplantation into the mouse kidney capsule. B, Brightfield image showing epithelial and mesenchymal CIO domains, scale bar, 100 μm. C, Immunofluorescence (IF) of CIOs showing epithelial (ECAD) part of the organoid with enterocytes (DPP4) polarized apically and smooth muscle cells (SM22) covering the outermost basal layer of the emerging mesenchyme, scale bar, 60 μm. D, Intestine epithelial cell (EPCAM) transcription factor expression (SOX9 and CDX2) in the in vitro zoomed-in images, scale bar, 20 μm, nuclei (DAPI). E, CIOs 12.5 weeks after transplantation (tCIOs) (top left), scale bar, 5 mm, H&E staining of a tCIO (top right), scale bar, 100 μm. IF staining of tCIO for epithelial cell type markers CDX2 for intestinal epithelium, MKI67 for proliferative cells, DPP4 for enterocytes, MUC2 for goblet cells, DEFA5 for Paneth cells and CHGA for enteroendocrine cells co-stained with the epithelial marker (ECAD, blue) and DAPI (gray), scale bars, 100 μm (bottom). F, UMAP embedding of tCIO epithelial cells based on scRNA-seq colored by cell types. scATAC-seq data was projected to the UMAP. G, Representative feature plots depicting gene expression (top) and chromatin accessibility (bottom) for enterocytes (APOA4), stem cells (LGR5), and enteroendocrine cells (CHGA). H, Top 20 cell type-enriched marker genes, and I, Top 50 cell type enriched accessible chromatin regions of tCIO epithelial cell types. J, Normalized intestinal epithelial cell type chromatin accessibility of the promoter or nearby regulatory element of tCIOs: BEST4+ cells (BEST4/MEIS1), enteroendocrine cells (CHGA), enterocytes (APOA4), goblet cells (MUC2), and stem cells (OLFM4) markers.

After observing limited maturation levels and cell type heterogeneity within in vitro CIOs, we transplanted CIOs derived from 2 iPSC lines into the kidney capsule of an immuno-compromised mouse host (tCIOs) and collected organoids after 12 weeks for further analysis (Fig. 3E). tCIOs exhibit extraordinary hallmarks of epithelial maturation similar to that in the developing human intestine, including the emergence of stereotypic crypt-villus architecture with CDX2+ intestinal epithelium, proliferative MKI67+ intestinal stem cells interspersed with DEFA5+ Paneth cells localized within crypt domains (Fig. 3E and fig. S6D). The differentiated DPP4+ absorptive enterocytes and secretory lineages consisting of CHGA+ enteroendocrine and MUC2+ goblet cells localized along the villus epithelium (Fig. 3E).

We generated paired scRNA-seq and scATAC-seq data from the same cell suspension to characterize in vitro and transplanted CIO cell type heterogeneity (fig. S7A). As with HIOs, the in vitro CIOs contain primarily epithelial and mesenchymal compartments, as well as rare neuronal cells (fig. S6, E to G and fig. S7, B and C). We uncovered enterocytes (ALDOB+), epithelial stem cells (LGR5+) and NKX2–3+ mesenchymal cells from the in vitro CIOs (fig. S6G and fig. S7C). In the single-cell sequencing data of tCIOs, we recovered diverse differentiated cell types observed in stainings as well as recently characterized BEST4+ cells (5) (Fig. 3F, fig. S6H and fig. S7B). We identified cell type marker genes and open chromatin regions and highlighted representative cell type-enriched features (Fig. 3, G to J, fig. S6, F to H, and fig. S7, C and D; Data S1). Collectively, these data show that chimpanzee intestinal organoids can provide insights into the major cell types and gene regulatory regions that underlie key aspects of intestinal cell type differentiation in great apes during development.

Organoids reveal human-specific gene expression in the developing intestinal epithelium

We integrated developing human, chimpanzee, and mouse intestinal scRNA-seq data to identify human features of intestinal epithelium development that are conserved or divergent across evolutionary time scales (Fig. 4A). We resolved enterocytes, enteroendocrine cells, goblet cells, and stem cells from all three species, with each cell type expressing known marker genes (Fig. 4B and fig. S8, A and B). Consistent with the phylogenetic distance, transcriptomes from human and chimpanzee intestinal organoid cell types were more similar to each other, and more divergent from their mouse counterparts (Fig. 4B and fig. S8A). We note that BEST4+ cells were absent from the developing mouse tissues (8), which is associated with depleted expression of the BEST4+ specific TFs MXD1 and MEIS1 and their target genes (fig. S8C and S8D). Gene ontology (GO) enrichment analysis on human and chimp conserved BEST4+ cell markers revealed peptide hormone secretion (Hypergeometric test nominal P=0.0086), regulation of growth hormone activity (P=0.036), cholesterol transportation (P=0.036) and storage (P=0.014) as well as a defense response to bacterium (P=0.01) as enriched terms (fig. S8E).

Fig. 4. Conservation and divergence between human and chimpanzee intestinal epithelial gene regulation.

A, Comparing single-cell sequencing data between human, chimpanzee, and mouse developing intestinal epithelium can illuminate similarities and differences between species. B, Developing human, chimpanzee and mouse epithelium integrated UMAP with cells colored according to cell type (left) or source (right). EC: Enterochromaffin cells. C, Barplots showing the number of differentially expressed genes (DEGs, left) and accessible chromatin regions (DARs, right) between human and chimpanzee for each intestinal epithelial cell type. D-E, Epithelial cell type average profiles of human-chimp differential and conserved gene expression (D) and chromatin accessibility (E) features. F, Heatmap indicates top 10 human-chimp differentially expressed enterocyte marker genes and nearby differentially accessible chromatin regions. G, Distinct human-specific pseudo temporal expression patterns along stem cell (SC) to enterocyte (Ent) trajectory (co-expression gene module index with number of genes in parenthesis) whereby each line depicting average expression patterns of each gene module in each species and shaded area depicting one standard deviation in expression patterns with corresponding KEGG pathway enrichment analysis (right). H, Expression profile of SLC5A12 along the stem cell to enterocyte differentiation trajectory (left) and the accessibility profile in enterocytes and stem cells of different sources of differentially accessible chromatin regions (right). Light grey color box highlights an SLC5A12 intronic region, chr11:26717719–26718218.

Comparing each epithelial cell type between human and chimp transplanted organoids, we identified 5,418 differentially expressed genes (DEGs) and 13,570 differentially accessible regions (DARs) (Fig. 4, C to F, and fig. S8, F and G; Data S1). We also compared gene expression based on read alignment to a human-chimp consensus reference, finding that 71% of the DEGs are identified in this alternative approach (fig. S8G). We further found that the direction of species differences in enterocytes and stem cells tend to be consistent between in vitro and transplanted organoids (Hypergeometric test nominal P value < 0.0001, fig. S8H). We highlight AIG1, RBP2 and SLC26A3 that are specifically high in human enterocytes in both in vitro and transplanted organoids (fig. S8I; Data S1).

We performed gene ontology analysis of DEGs and nearby genes of DARs and identified enrichment in metabolic and immune-related processes (fig. S8, J and K). For example, metabolism associated with glutathione and lactate tend to be enriched in human-high features (fig. S8K). We found that the amino acid substitution rates (dN/dS) of DEGs associated with these terms are higher than non-DEGs associated with the same term (fig. S8J). In addition, human-chimpanzee DARs close to genes associated with these terms are more enriched in mammal or primate nodes than non-DARs associated with the same term (fig. S8K). Collectively, these results reveal processes being modified through both protein-coding and gene regulatory changes during recent evolution. We also identified TFs that showed concordant species differential expression and TF binding site motif enrichment in DARs, which together might underlie observed differences in gene expression (fig. S8L). We note that chimpanzee organoids mature slightly faster than human counterparts, however enterocyte and stem cell DEGs and functional enrichments are consistent after differentiation pseudotime alignment (fig. S9). Overall, we identified biological processes and related molecular features that are similar and different between the human and chimpanzee organoid models of the intestinal epithelium.

To expand the evolutionary scale of intestinal epithelial gene expression comparisons, we analyzed differential gene expression between developing human and mouse tissues for each epithelial cell type and found that tHIOs reliably recapitulate the human differential features (fig. S10, A to J, and Data S1). Using mice as an outgroup, we identified co-expression gene modules with human-specific expression profiles along the stem cell to enterocyte differentiation trajectory, and in goblet cells and EECs (Fig. 4G and fig. S11, A and B; table S7; Data S1). KEGG pathway enrichment revealed human-specificity in xenobiotic metabolism, lipid processing and epithelial barrier (Fig. 4G; table S3). For example, human enterocytes showed elevated expression of SLC5A12, RRAS2 and CBLB and decreased expression of PLEC, along with concordant accessibility differences between human and chimpanzee in nearby regulatory regions (Fig. 4H and fig. S11, B and C). SLC5A12 (Solute carrier family 5 member 12) is a sodium-coupled monocarboxylate transporter involved in lactate transport and associated with lipid metabolism and inflammation (61, 62). RRAS2 (RAS Related 2) and CBLB (Cbl Proto-Oncogene B) are associated with C-type lectin receptor signaling (63, 64) and PLEC (Plectin) is linked to hemidesmosomes (65)). Together, these data illuminate human-specific gene expression and regulation underlying metabolic and barrier functions.

Cataloging gene regulatory region innovations in the developing human intestine

We established a composite functional and evolutionary annotation for each putative human intestinal epithelial regulatory region incorporating scores for cell type specificity, ontology enrichment, ancestry, differential accessibility, variation in modern and archaic humans, and other evolutionary signatures (Fig. 5A). We classified epithelial cell-enriched regions according to whether they overlap with 1) evolutionarily constrained regions, 2) GWAS variants, 3) human-chimp differentially accessible regions, 4) modern human-specific single nucleotide changes (SNCs) as compared to great apes and archaic humans, 5) human accelerated regions (HAR) and 6) loci under positive selection (PS) (Fig. 5B). We then performed GREAT analysis to identify biological processes overrepresented in developing intestinal epithelium putative regulatory regions harboring constrained elements, GWAS hits, and human selection signatures, and visualized ontology hierarchy of the enriched terms and their parent terms (Fig. 5, C and D; table S3).

Fig. 5. Small intestinal epithelial regulatory regions harbor human selection signatures that impact diverse functions.

A, Schematic to annotate each putative regulatory region with information on cell type specificity, functional ontology, evolutionary ancestry, evolutionary constraint and selection, functional species comparison, and hominid variation. B, Epithelial marker region classification based on overlap with features, 1: Constrained, 2: GWAS variant, 3: Differentially accessible (DA) between human and chimp, 4: Single nucleotide change specific to modern humans (SNC), 5: human accelerated region (HAR), 6: Positive selection signature (PS). SNC, HAR and PS are collectively termed as human selection signatures. C, Willow plot showing ontology hierarchy of biological processes (nodes) overrepresented in developing intestinal epithelium enriched regulatory regions harboring constrained and human selection signatures. Nodes are colored and labeled based on ontology hierarchy, edges represent ontology term connections in the ontology hierarchy. Ontology terms are labeled to explain the graph, and also highlight significantly enriched processes (BH-corrected binomial test P < 0.05). D, Willow plot colored based on term enrichment of human selection signatures. E, Scatter plot of developing intestine epithelium enriched accessible regions scored by the epithelial accessibility specificity (Tau, x-axis) and by the composite analysis score for genomic signatures (y-axis, See Methods). F, Accessibility profiles of loci with human selection signature in enterocytes. G, Schematic showing enhancer assay to evaluate the activity of LCT enhancer (chr2:135850827–135851326) in human intestinal enteroids (top). Boxplot of per-cell STARR-seq activity (log2 scale) in enterocytes of the putative LCT enhancer carrying either the C variant (−13910C, ancestral, lactose intolerant) or the T variant (−13910T, SNP, lactase persistence) in developing human enteroids (bottom). *, Mann-Whitney U Test P = 1.652e−06. H, Nucleotide importance scores of the LCT enhancer for enterocytes (ChromBPNet derived) (top), overlap with genomic signatures (middle) and in silico saturation mutagenesis showing predicted effect of SNPs on chromatin accessibility (bottom). Predicted TF-binding sites are shown by boxes of the plot on the top, for repressors (ZBTB7A) and activators (GATA4, HNF4, AP1). SNP under positive selection (PS) for lactase persistence (ancestral C, selected T) is highlighted and computational conversion from C to T is predicted to break the ZBTB7A motif and increase chromatin accessibility. I, Schematic of two guide CRISPR/Cas9 targeting of LCT enhancer (chr2:135850727–135851426) for deletion in developing enteroids (top left). Quantitative real time PCR results for LCT mRNA detection in wild type and enhancer knockout (n = 3 biological replicates) (top right). RNA levels were normalized to TBP and expressed as fold over wild type average. Error bars denote SEM. *, t-test P = 0.0037. Representative images of wild type and MCM6/LCT putative enhancer knockout enteroids after 7 days in differentiation media (bottom). J, Boxplot of per-cell STARR-seq activity (log2 scale) in stem cells and enterocytes of the putative PDX1 enhancer carrying either the human or chimp variants in developing human enteroids (top). *, Mann-Whitney U Test P = 1.5e−12. Quantitative real time PCR results for PDX1 mRNA detection in wild type and enhancer knockout (n = 4 biological replicates) (bottom). RNA levels were normalized to TBP and expressed as fold over wild type average. Error bars denote SEM. *, t-test P = 0.0068.

We identified many regions with high composite annotation scores representing cell type-enriched functionality with signatures of recent evolutionary change in humans (Fig. 5, E and F). For example, Lactase encoded by the LCT gene is an enzyme responsible for the digestion of the milk sugar lactose. Lactase persistence into adulthood is a well-known positively selected trait in specific modern human populations (19, 29), with single-nucleotide polymorphisms (SNPs) within MCM6 intronic regions reported to be associated with regulation of LCT (19, 29). One of these regions shows enterocyte-enriched accessibility in developing human and chimp organoids, and is one of the top hits of our composite annotation analysis (Fig. 5, E and F). Among the other top hits, an intronic region of VLDLR (Very low density lipoprotein receptor) overlaps a HAR, SNC, and human ancestor quickly evolved region (HAQER) (53) and has decreased accessibility in human stem cells and enterocytes compared with chimpanzee cells (Fig. 5E and fig. S12A). VLDLR coordinates systemic uptake of circulating VLDL cholesterol produced in the liver and secreted into the bloodstream (66), and has been identified as an entry receptor for alphaviruses (67). Another example, a region within the long non-coding RNA PLUT (PDX1 associated lncRNA, transcription upregulator) also overlaps a HAR (chr13:27851751–27851810), and PDX1 (Pancreatic duodenal homeobox-1) (68) is a key transcription factor controlling duodenal patterning and development(69) (Fig. 5, E and F). We also identify an intronic region within the lipid metabolism regulator AIG1 that harbors a SNC specific to modern humans. This gene is also specifically upregulated in human enterocytes compared with chimpanzee and mice (fig. S12B). Collectively, we provide a framework for comprehensive functional and evolutionary annotation of putative regulatory regions in the developing human intestinal epithelium.

Intestinal organoid validation of a cis-regulatory region linked to hypolactasia

We sought to establish the functional relevance of one of these putative human gene regulatory regions using an organoid model. The intronic region (chr2:135850727–135851426) within MCM6 (between exon 12 and 13) has been hypothesized to act as an enhancer for the LCT gene based on genetic linkage and enhancer assays in immortalized cell lines (19, 70, 71). A positively-selected SNP locus (chr2:135851076, rs4988235, C>T) is reported as causative in hypolactasia, a common enzyme deficiency leading to lactose malabsorption manifesting in adults (72), where the ancestral (C) and selected (T) alleles are associated with lactose-intolerant and lactase-persistent phenotypes, respectively. However, the experiments to establish enhancer function for this region have not been performed in a human multicellular model, which would allow assessment of enhancer cell type specificity. We used an enhancer assay to assess the activity of the LCT enhancer carrying either the ancestral (C) or selected (T) variant in developing intestinal enteroids (Fig. 5G). We found that both variants enhance transcription in enterocytes of developing human enteroids, with a slightly stronger activity for the (T) allele (Fig. 5G), consistent with observations from luciferase assays in adult immortalized cell lines (70).

We trained a convolutional neural network on chromatin accessibility in enterocytes, enabling interrogation of enhancers at base-pair resolution (73) (table S8). Implementing this model on the LCT enhancer predicts several activator binding sites, including known master regulators of small intestine epithelial gene expression (HNF4A, GATA4, AP1) (Fig. 5H). These binding sites are not affected by nucleotide positions that differ between human and chimpanzee, associated with disease, or under positive selection. The model also predicts a repressive motif matching with a member of the Zinc finger and BTB domain containing TF-family (e.g. ZBTB7A) (74). Using in-silico mutagenesis, the model predicts that the selected variant (T) disrupts the binding site of the ZBTB7A repressor, leading to an increase in accessibility (Fig. 5H). These data highlight how deep learning models can be used to interpret the potential effects of single-base changes with evolutionary implications.

We then used CRISPR/Cas9 editing to introduce a homozygous deletion of the LCT enhancer, generated developing human enteroids in differentiation conditions, and measured LCT expression using quantitative real time PCR (Fig. 5I and fig. S12C). The deletion led to a 3-fold reduction of LCT mRNA (Fig. 5I). The reduction of LCT expression by the enhancer KO is larger than that of MCM6, within which the enhancer is located, and we confirmed that the expression reduction is not due to a difference in enterocyte differentiation, as indicated by the enterocyte marker FABP1 (fig. S12D). These data show that this region is a bona fide enhancer of LCT in the developing human intestinal epithelium. Notably, this enhancer is accessible in both human and chimpanzee developing intestinal organoids, underlying its likely role regulating LCT expression during mammalian and primate suckling periods. Altogether, we provide a framework for comprehensive analysis and functional validation of regulatory regions and specific variants in the developing human intestinal epithelium.

Functional assessment of cis-regulatory regions with human-specific genetic changes

We next established the functional relevance of several cis-regulatory regions with human-specific changes. Intestinal organoid reporter assays using human or chimpanzee regions associated with PDX1, a transcription factor regulating intestinal cell identity, revealed significantly stronger enhancer activity of the human variant (Fig. 5J). Additional enhancer reporter testing for two intronic regions within AIG1 (Fig. 5F) and SLC5A12 (Fig. 4H) that displayed differential accessibility, did not show significantly higher enhancer activity of the human variants (fig. S12E). We next used genetic deletion experiments to assess the functional relevance of the PDX1-associated CRE. PDX1-CRE knock-out resulted in significantly reduced PDX1 gene expression in duodenal enteroids (Fig. 5J). We note that the chromatin-accessible HAR located within PDX1-CRE also overlaps the lincRNA PLUT1, which has been shown to regulate PDX1 expression (68).

A region proximal to the RBP2 promoter is differentially accessible between human and chimpanzee intestinal enterocytes (Fig. 5F), and we confirmed differential protein expression in enterocytes using IF in primary human duodenum, tHIOs, and tCIOs (fig. S13, A and B). RBP2 harbors SNCs associated by GWAS to blood pressure traits (75), has been shown to participate in fatty acid metabolism in the intestine and Rbp2-deficient mice are prone to develop obesity (76). Homozygous genetic deletion of the RBP2-CRE significantly reduced RBP2 gene expression in duodenal enteroids (fig. S13C). Enteroids lacking the CRE failed to upregulate RBP2 under lipid challenge (2% and 10% lipid concentrations), mirroring Rbp2-deficient mice (76). This disruption coincided with dysregulated lipid metabolism, including elevated GIP (intestinal incretin hormone) and suppressed HMGCS2 (ketogenesis enzyme)(77) (fig. S13D). These data provide functional support that selection for RBP2 expression in the intestinal epithelium might underlie susceptibility to metabolic disease and obesity, and organoid models could be used to further explore functional relevance of ancestral and modern human alleles.

The IGFBP2 locus, encoding a critical regulator of insulin-like growth factor signaling, harbors two human accelerated regions (HARs) with chromatin accessibility differences between human and chimpanzee intestinal stem cells (Fig. 6A). IGFBP2 was among the differentially expressed genes between human and chimpanzee intestinal stem cells (Fig. 6B) and IGFBP2 mRNAs were enriched in tHIO crypts compared with tCIOs (Fig. 6, C and D). Two HARs—one intronic (IGFBP2i) and one distal (IGFBP2d)—overlap enhancers harboring activator motifs (KLF5/6, HNF4, AP1) as predicted by deep learning (Fig. 6E and fig. S13F). To test the function of each of these putative IGFBP2 enhancers, we used CRISPR/Cas9 editing to introduce deletions of each region separately, generated developing human intestinal enteroids in differentiation conditions, and measured IGFBP2 expression using quantitative real-time PCR. The deletions led to 20-fold and 2-fold reductions of IGFBP2 mRNA, for intronic and distal regions, respectively (Fig. 6F). Deep learning analysis of the IGFBP2 distal locus predicts a strong cumulative decrease in chromatin accessibility with the introduction of chimpanzee variants (fig. S13G). Concordantly, comparative reporter assays in intestinal enteroids showed that the human distal IGFBP2 regulatory region has a 6-fold enhancement of activity compared to the chimpanzee counterpart (Fig. 6G). Deep learning prediction further suggests that two human-specific single nucleotide changes as compared to most other mammals within the HAR innovates a binding site of Caudal Type Homeobox 2 (CDX2) (Fig. 6E), a central regulator of small intestinal epithelial biology(78). These data suggest that signalling through the insulin-like growth factor pathway might underlie human-specific features of small intestinal development or physiology.

Fig. 6. Functional evaluation of two human accelerated regions at the IGFBP2 locus.

A, Genome tracks of chromatin accessibility in developing human (Dev. H.) and transplanted human (tHIO) or chimp (tCIO) intestinal organoids across the IGFBP2 locus for enterocytes and stem cells. The functionally tested intronic (IGFBP2i) and distal (IGFBP2d) putative enhancers are highlighted (gray bars (left) and zoom in panels (right)). B, Expression profile of IGFBP2 along the stem cell to enterocyte differentiation trajectory. C, Representative images of RNAscope in situ hybridization for IGFBP2 in transplanted human (tHIO) or chimp (tCIO) intestinal organoids. Boxes indicate crypt areas displayed in the zoom in panels on the right side of each sample. Epithelial marker (ECAD, blue) and DAPI (gray), scale bar, 100 μm. D, Quantification of IGFBP2 mRNA in tHIO (n=21) and tCIO (n=20) crypts. *, unpaired t-test, P = 0.0123. E, Nucleotide importance scores of the putative IGFBP2 distal enhancer (IGFBP2d) for enterocytes (ChromBPNet derived) (top), overlap with human -chimp single nucleotide changes (SNCs) (middle) and in silico saturation mutagenesis showing predicted effect of SNCs on chromatin accessibility (bottom). Predicted TF-binding sites are shown by boxes of the plot on the top for transcriptional activators (AP1, HNF4, CDX2). SNCs within a Human Accelerated Region (HAR) in the CDX2 motif are highlighted and computational conversion to the chimp variants is predicted to negatively impact the CDX2 motif and decrease chromatin accessibility. F, Quantitative real time PCR results for IGFBP2 mRNA detection in wild type and enteroids carrying CRISPR/Cas9 deletion of IGFBP2 putative intronic (IGFBP2i, left) and distal (IGFBP2d, right) enhancers. RNA levels were normalized to TBP and expressed as fold change over wild type average. Error bars denote SEM. IGFBP2i *, t-test P = 0.006. IGFBP2d *, t-test P =0.0063. G, Boxplot of per-cell STARR-seq activity (log2 scale) in stem cells and enterocytes of the putative IGFBP2 distal enhancer carrying either the human or chimp variants in developing human enteroids. *, Mann-Whitney U Test P < 2.2e−16.

Discussion

The intestinal epithelium interfaces with an extraordinarily complex and dynamic luminal environment. Diverse cell types evolved, and continue to change, in animals to mediate nutrient absorption, barrier function, and a myriad of other important physiological functions required to maintain this critical interface with the outside world. Gene regulatory networks underlying human intestinal development prepare the epithelium for imminent microbial colonization and nutrient exposures at birth. These networks undoubtedly have ancient origins; however, it has been unclear to what extent the gene regulation underlying development has changed during the course of human evolution.

We find that each intestinal epithelial cell type has gene expression and regulatory features possessing distinct evolutionary histories. Enterocytes, a diverse group of intestinal epithelial cells specializing in nutrient absorption and barrier function, have selection signatures in the range of immune cells, and immune cells are known to be among the most rapidly evolving cell types across animal tissues (23, 79). Enteroendocrine cells, which secrete hormones and coordinate inter-tissue functions, have comparatively slower evolutionary rates. The gene regulatory regions that underlie enterocyte gene expression also have signatures of dynamic change and turnover. Together, these findings indicate that developing enterocytes have undergone a rapid evolutionary change in humans. This has human health relevance, as regions of the genome that have rapid changes influence disease risk (30) and are associated with metabolic and gastrointestinal disorders (19, 80–83). Evolution involves trade-offs that confer benefits but also create vulnerabilities. Similarly, changes that were adaptive in particular environmental conditions may pose disease risks as cultures and environments change. Our catalog of evolutionary signatures overlapping gene regulatory regions of the developing intestine provides a basis for studying human disease susceptibilities.

We show that both iPSC- and developing intestinal stem cell-derived organoids can be used to study intestinal epithelial gene regulation at prenatal stages of development. We propose regulomes that underlie human cell type identity, highlighting sets of transcription factors and gene regulatory regions enriched in each cell type, and identify several new cis-regulatory elements using a multiplexed single-cell enhancer activity assay. The data show that FSC- and iPSC-derived organoids are useful models of developmental gene regulation, providing well characterized models and high-information content GRNs that can be used to understand neonatal-related disorders such as necrotizing enterocolitis and rare developmental diseases associated with barrier function and nutrient metabolism.

Given the inaccessibility of developing tissue from chimpanzees and other great apes, we used chimpanzee iPSCs to recapitulate intestinal development, illuminating similarities and differences with our closest living relatives. We found significant differential expression of genes associated with lipid metabolism, cholesterol absorption, and immune response when comparing human and chimpanzee enterocytes. Combining functional genomic data with comparative genomic analysis of human populations, archaic hominids, great apes, primates, and mammals provided perspectives on intestinal epithelial features that are human-specific (84). This integrative analysis identified diverse regulatory regions nearby nutrient transporters (SLC5A12, SLC5A1, SLC19A3), enzymes involved in xenobiotic and nutrient metabolism (SULT2A1, AIG1, LCT), signaling and receptor molecules associated with nutrient uptake (RBP2, IGFBP2, VLDLR), and transcription factors regulating small intestine development and cell type differentiation (PDX1).

Some of these regions have ancestral alignments deep in the vertebrate phylogeny, whereas others are recently born with deepest ancestry in primates. We identified regions that have differential accessibility between cell types and between human and chimpanzee cell counterparts. We tested several of these regions using enhancer assays in developing human enteroids, and focused on a well-studied region within the MCM6 gene that has been linked to the emergence of lactase persistence. Previous studies using luciferase reporters have suggested that lactase persistence SNPs contribute to lactase gene expression (19, 70, 71) based on experiments performed in colorectal cancer cell lines. Here, we show that the genetic deletion of the MCM6 intronic region significantly decreases baseline LCT expression. Additionally, reporter assays with the −13910C and −13910T alleles show differential activity. These experiments were performed in developing cell states where we expect both alleles to be functional due to evolutionarily conserved exposure to milk upon birth. Our analyses predict small intestinal master regulators (HNF4A, GATA4, AP1) orchestrate LCT expression, and suggest a novel role for a ZBTB domain containing TF to function as a repressor of Lactase gene expression dependent on the genetic background of the individual and stage of development. It will be interesting to explore these predicted TF-enhancer interactions and understand how these LCT enhancer alleles function under different conditions and at different ontogenetic stages.

We functionally validated two HARs located at the IGFBP2 locus, each which harbors changes that introduce new transcription factor binding sites. While the functional effects of enhancing IGFBP2 expression in stem cells of the developing human intestine are unclear, the insulin-like growth factor pathway coordinates proliferation and growth (85, 86), and it is provocative to imagine that changes in this pathway underlie region proportion differences along the gastrointestinal tract between humans and other apes. Additionally, we functionally validate a differentially accessible HAR at the PDX1 locus, a regulator of insulin signaling and proximal small intestine identity(69, 87). These data together with reports of transient insulin-producing enteroendocrine cells in the fetal human intestine (88), Igf1 signaling altering villus height (89), and high concentrations of IGF in human amniotic fluid(90) and breast milk(91), suggest recent evolutionary modifications within the insulin and insulin-like signaling pathways impact human-specific features of the developing human small intestinal epithelium. We also identify changes within the RBP2 promoter that might enhance retinol transport, which cascades into transcriptional control of diverse nutrient transporter expression (92, 93). Together these findings suggest mechanisms underlying recent evolutionary changes in nutrient acquisition within the human small intestine.

In summary, stem cell-derived organoids from neonatal, juvenile and adult individuals provide an inroad into exploring unstudied human-specific alleles. These data bring a new focus on a fascinating interface between humans and our external environment at a stage of life where barriers are most susceptible. These results have a broader implication for the evolution of other anatomical changes, such as the expanded human brain. Modifications to the molecular components of our intestinal epithelium regulating nutrient uptake may have enabled other unique aspects of human biology by providing additional nutrients previously unavailable in sufficient quantities. The molecular and physiological basis for how nutrient absorption, processing, and distribution has evolved to enable brain expansion and other unique aspects of human biology are yet to be understood.

Materials and methods

All materials and methods information is available in the Supplementary Materials.

Data and materials availability:

The accession number for the STARR-seq transfection plasmid library is ArrayExpress:E-MTAB-14972. The accession number for the raw and processed single cell sequencing data of developing human intestine tissue and human and chimpanzee intestine organoids is ArrayExpress:E-MTAB-15112. The accession number for the raw and processed FLASH-seq data of human and chimpanzee single spheroid is ArrayExpress:E-MTAB-15107. Published developing mouse intestine scRNA-seq data is accessible via ArrayExpress:E-MTAB-13201.

Additional supporting materials (Data S1(35)) contains 1) comprehensive annotation of open chromatin regions detected in developing human or chimp intestine organoids or tissue; 2) cell type enriched features, including (A) developing human proximal small intestine tissue cell type marker genes and regions, (B) in vitro and transplanted HIO cell type enriched genes and regions, (C) in vitro and transplanted CIO cell type enriched genes and regions, (D) developing mouse epithelial cell type marker genes; 3) amino acid substitution rate (dN/dS) and epithelial cell type expression specificity (Tau) of epithelial cell class and specific cell type marker genes; 4) differentially expressed genes (DEG) between species, including (A) human-chimp DEGs of each intestinal epithelial cell type of in vitro or transplanted organoids, (B) human-mouse DEGs of epithelial cell type between developing human and mouse tissues, (C) developing human and mouse tissue DEG along stem-cell-to-enterocyte pseudotime trajectory, (D) human-chimp DEGs based on consensus genome; (E) human-chimp DEGs after stem cell to enterocyte differentiation pseudotime alignment; 5) collected evolutionary selection signatures for composite annotation of open chromatin regions; 6) Seurat objects of scSTARR-seq experiments and the table for the calculation of the normalized CRE activity; 7) Seurat objects and 8) stem cell to enterocyte pseudotime (Pt) aligned expression matrix of human, chimpanzee, mouse data.

The core scripts of single cell data analysis are available at https://github.com/devsystemslab/Dev_intestine_evo/ (94).