Single Cell Analysis of Human Colonoids Exposed to Uranium-Bearing Dust

Tissue Collection and Colonoid Generation

Human colonoid studies were reviewed and approved by the Johns Hopkins University School of Medicine institutional review board (IRB number NA_00038329) and University of New Mexico institutional review board (IRB number 18-171). Colonic biopsies were obtained from healthy individuals undergoing screening colonoscopies who had given informed written consent. All colonoids were generated from proximal colonic biopsies from patients undergoing routine, screening colonoscopy and with no pathophysiology. Colonic crypt isolation and colonoid generation were prepared as previously reported.^41,42 Isolated crypts were resuspended in Matrigel (Corning) and $30\text{-}\mathrm{\mu }\mathrm{L}$ droplets were plated in a 24-well plate (Corning). After polymerization at 37°C, $500\mathrm{\mu }\mathrm{L}$ of organoid expansion media was added for 2 d [Advanced Dulbecco’s modified Eagle medium (DMEM)/Ham’s F-12 (ThermoFisher), $100\mathrm{U}/\mathrm{mL}$ penicillin/streptomycin (Quality Biological), $10\text{\hspace{0.17em}mM}$ HEPES (ThermoFisher), and 1× GlutaMAX (ThermoFisher), with 50% vol/vol WNT3A conditioned medium (ATCC CRL-2647), 15% vol/vol R-spondin1 conditioned medium (cell line kindly provided by Calvin Kuo, Stanford University, currently available through Sigma-Aldrich), 10% vol/vol Noggin conditioned medium (cell line kindly provided by Gijs van den Brink, Tytgat Institute for Liver and Intestinal Research, and described in Heijmans et al.⁴³), 1× B27 supplement (ThermoFisher), $1\text{\hspace{0.17em}mM}$ N-acetylcysteine (MilliporeSigma), $50\text{\hspace{0.17em}ng}/\mathrm{mL}$ human epidermal growth factor (ThermoFisher), $10\text{\hspace{0.17em}nM}$ (Leu-15) gastrin (AnaSpec), $500\text{\hspace{0.17em}nM}$ A83-01 (Tocris), $10\mathrm{\mu }\mathrm{M}$ SB202190 (MilliporeSigma), $100\mathrm{mg}/\mathrm{mL}$ primocin (InvivoGen), $10\mathrm{\mu }\mathrm{M}$ CHIR99021 (Tocris), and $10\mathrm{\mu }\mathrm{M}$ Y-27632 (Tocris)]. All conditioned media were generated following, in precise detail, the protocol from the vendor or initial publication. Briefly, cells for conditioned media were grown to 80% confluency, upon which the media was replaced with Advanced DMEM with fetal bovine serum (FBS) (WNT3A) or without FBS (R-spondin1 and Noggin). Media were not changed until collection. Conditioned media from WNT3A cells were collected on days 4 and 7, combined, filtered through a $0.2\mathrm{\mu }\mathrm{M}$ filter, aliquoted, and stored at $-80°\mathrm{C}$ for up to 6 months. Conditioned media from R-spondin1 and Noggin cells were collected on day 7, filtered through a $0.2\mathrm{\mu }\mathrm{M}$ filter, aliquoted, and stored at $-80°\mathrm{C}$ for up to 6 months.

After 2 d, the organoid expansion media (without CHIR99021 and Y-27632) was replaced every other day. Colonoids were passaged every 10–14 d by harvesting in Cultrex Organoid Harvesting Solution (Bio-Techne, Minneapolis, MN) at 4°C with shaking for 45 min, as previously described.⁴⁴ All colonoid cultures were maintained at 37°C and 5% ${\mathrm{CO}}_2$. Unless noted, colonoid lines have been passaged $>20$ times.

Brightfield Imaging

Colonoids plated in Matrigel in 24-well plates were imaged during experiments on a Zeiss Axio Observer A1 inverted microscope (Zeiss) with images captured on CellSense imaging software (Olympus). As shown in previous intestinal organoid studies, propidium iodide (PI) was used to quantify changes in cell death.^44–47 PI and DAPI (ThermoFisher) were added at a final concentration of $2\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ and $1\mathrm{\mu }\mathrm{g}/\mathrm{mL}$, respectively. Colonoids were placed at 37°C for 1 h before imaging. Images were viewed and processed using OlyVia (Olympus). Total cells and PI+ cells were counted in ImageJ. $n=3$ independent experiments were performed, with at least 1,000 cells counted in each condition.

Uranium-Bearing Dust Analysis

UBD was obtained from a church attic on indigenous lands that lie adjacent to the 3,000-acre Jackpile mine in New Mexico. Permission was granted by the Environmental and Natural Resources Department of the Laguna Pueblo with assistance in obtaining the sample from the Southwest Information and Research Center. Samples of the collected dust were sieved using a $20\mathrm{\mu }\mathrm{M}$ filter. Transmission electron microscopy (TEM) analyses of the dust were carried out on all particulates $<20\mathrm{\mu }\mathrm{M}$ size fraction for an unbiased analysis. Particulates from this fraction, suspended in acetone, were deposited onto holey carbon film supported on Cu TEM grids using pipettes. For the TEM analysis, five $50\times 50\mathrm{\mu }\mathrm{M}$ TEM grid squares were studied in detail. The samples were air dried and then analyzed using a JEOL 2010F FEG TEM/STEM (JEOL USA) operating at $200\text{\hspace{0.17em}kV}$. The samples were imaged using a variety of different TEM and scanning transmission electron microscopy (STEM) techniques, including bright-field TEM and high-angle annular dark-field STEM imaging. Uranium-bearing particulates were identified on the JEOL 2010F by energy dispersive X-ray spectroscopy (EDS) using an Oxford Instruments AZtecEnergy EDS system (Oxford Instruments) coupled to an Oxford X-Max $80{\mathrm{mm}}^2$ silicon drift detector (SDD). The detailed distribution of U-bearing phases and their association with other minerals was obtained using STEM EDS spectral X-ray mapping. TEM and STEM images of two examples of these U-bearing grains are shown in Figure S1.

Treatment and Preparation of Single Cells from 3D Colonoids for Single-Cell RNA Sequencing

Colonoids used for sequencing are described in Table S1. UBD was diluted in organoid expansion media, then colonoids in Matrigel were treated overnight (18 h) with $50\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ UBD, with control colonoids receiving the same volume/volume of organoid expansion medium as the vehicle. Dosage was chosen based on prior studies^26,48–51 that showed molecular changes to cells or animals exposed to uranium, but with minimal cytotoxic damage, since acute uranium exposure is not linked to cytotoxicity. Next day, colonoids were harvested from Matrigel using Cultrex Organoid Harvesting Solution, then resuspended in $500\mathrm{\mu }\mathrm{L}$ TrypLE enzyme solution (ThermoFisher), transferred to a 24-well plate and spun at $600\text{\hspace{0.17em}rpm}$ at 37°C for 45 min in a spinoculator (Mixer HC; USA Scientific) to digest into single cells. Single cells were washed in $10\mathrm{mL}$ of Advanced DMEM/F12, pelleted again, resuspended in $1\mathrm{mL}$ of organoid expansion media and filtered with $40\text{-}\mathrm{\mu }\mathrm{m}$ Flowmi cell strainers (SP Bel-Art). Cells were transported to the UNM Analytical and Translational Genomics Core in the UNM Comprehensive Cancer Center for counting and processing. Average processing time from colonoids to library preparation was $<2.5$ hours, and the cells were highly viable ($>90\%$).

Single Cell and Library Preparation

Using the Cell Countess II FL (ThermoFisher), single cells and viability were counted by Trypan Blue exclusion. All reagents for single cell and library preparation were purchased from 10X Genomics and provided by the UNM Comprehensive Cancer Center. The Next GEM Chip G was loaded and put into the Chromium Controller for processing, creating the GEMs (gel beads-in-emulsion). cDNA synthesis of the GEMs was completed per the 10X Chromium Next GEM 3’ protocol. Synthesized cDNA was evaluated for quality and quantity on a DNA High-Sensitivity Chip on the Agilent BioAnalyzer machine. Next, cDNA was fragmented, adaptor ligated, and sample indexed (barcoded) with PCR. Once the libraries were completed, they were processed on a DNA High-Sensitivity Chip with the Agilent BioAnalyzer to ensure that the final libraries were $\sim 500$ bp in size. They were also processed with the KAPA Library quantification kit to verify the concentration of final libraries.

Pooling and Sequencing Quality Control

Sample concentration was determined using Agilent Tape Station 4200 and Invitrogen Qubit 4.0. Samples were diluted with molecular grade water, normalized, and pooled at $4\text{\hspace{0.17em}nM}$, according to Illumina protocol, prior to quantitative real-time polymerase chain reaction (qPCR) to determine cluster efficiency. Samples were loaded onto an Illumina NovaSEQ 6000 instrument and sequenced on S4 Flow cells at specified read length. Data were delivered in FASTQ file format.

Processing FASTQ Reads into Gene Expression Matrices

Raw FASTQ reads from sequencing were processed using Cell Ranger (version 2.0.0) software from 10X Genomics. In brief, sequencing reads were demultiplexed, unique molecular identifiers (UMI) were collapsed, and reads aligned to the human transcriptome GRCh38-3.0.0 as the reference.⁵²

scRNA-Seq Quality Control and Processing of 10X Sequencing Data

The raw data created by Cell Ranger was read into the Seurat R package for quality control and downstream processing. Low-quality cells expressing $<200$ genes and genes expressed in fewer than three cells were filtered out. After looking at the violin plots of the number of counts, features (genes), and percent mitochondria per sample, we further filtered cells expressing $>\mathrm{2,000}$ genes in control samples and cells expressing $>\mathrm{7,500}$ genes in UBD-treated samples to avoid probable doublets. Cells expressing $>20\%$ of mitochondrial genes were filtered out in control and UBD samples. To account for differences in sequencing depth across samples, the data was log-normalized and scaled.

Clustering, Visualization, and Cell Annotation

For cell clustering, we sorted the top 2,000 highly variable genes that differed from cell to cell for downstream analysis. The dimensionality of the data was reduced by principal component analysis (PCA) on the highly variable genes. Since Seurat clusters cells based on their PCA scores, the elbow plot was used to determine the top principal components (PCs) for use in clustering the data. From the elbow plot, the top 15 PCs were used in clustering. Cells were divided into unsupervised clusters using a shared-nearest-neighbor modularity optimization-based clustering algorithm. Cells were clustered at a resolution of 0.4 and visualized using Uniform Manifold Approximation and Projection (UMAP) plots.⁵³ Differentially expressed genes per cluster were calculated using the Seurat Wilcoxon signed-rank test. Cluster annotation was based off the top genes (log2 fold change $>0.25$ and $p>0.05$) and marker genes described in literature. Control and UBD-treated samples were annotated together to draw direct comparisons. Marker gene expression was visualized using violin plots and dot-plots (size of the dot reflects the percentage of cells expressing the gene, and color indicates relative expression). R version 4.0.4 was used for all data plotting and analysis.

Differences between Treatment

For comparisons between control and UBD-treated samples, the samples from the same subject were log normalized and scaled separately and were merged into one dataset for downstream analysis. Where necessary, all control samples were merged into one dataset, and all UBD-treated samples were merged into another dataset.

Percentage of Cells and Statistical Analysis

The relative abundance of each cell type was calculated as percentage of cells per condition (control and UBD). We tested for statistical significance using two-tailed $t$-test (GraphPad Prism 9.0).

Pathway Analysis

Differentially expressed (DE) genes among clusters or samples were ranked according to their expression with the upregulated ones at the top of the list and the downregulated ones at the bottom. The ranked gene list was imported onto gProfiler (Ensembl 109). Pathway analysis was done using gene ontology (GO) biological process and Reactome (version 84). After running the analysis, the gProfiler data was downloaded as a GMT file, while the results file was downloaded in the Generic Enrichment Map format. Both types of files were then imported into Cytoscape (version 3.9.1) where the EnrichmentMap app was used to generate the enrichment map. We used a $p$-value cutoff and q-value cutoff of 1 and an Overlap Coefficient cutoff of 0.5. Using DE genes from Seurat as input, protein–protein interactions of upregulated genes in the UBD-treated enteroendocrine cluster were determined using the STRING app on Cytoscape.

Immunofluorescence, RNAscope, and Confocal Imaging

Colonoids were harvested from Matrigel using Cultrex Organoid Harvesting Solution, then fixed for 40 min in 4% paraformaldehyde (Electron Microscopy Sciences). Immunostaining was performed as previously described.⁴⁴ In brief, fixed colonoids were permeabilized and blocked and incubated with primary antibodies at 4 degrees overnight. After washing, organoids were incubated with fluorescent secondary antibodies and DAPI for 1 h at room temperature, washed, and mounted. Primary and secondary antibodies used are listed in Table S2.

Fixed colonoids were prepared for whole-mount staining with RNAscope Multiplex Fluorescent version 2 using probes Hs-CHGA and Hs-PROX1 (Table S3). Sample preparation and labeling were performed according to the manufacturer’s protocol (Advanced Cell Diagnostics) with slight modifications as previously described.⁵⁴ Opal fluorophores (Perkin Elmer) and DAPI were used for visualization. Confocal imaging was carried out in the UNM Comprehensive Cancer Center Fluorescence Microscopy Shared Resource using an LSM 800 AiryScan confocal microscope (Zeiss) and UNM AIM Center using an LSM 900 AiryScan confocal microscope (Zeiss). Total cells and antibody- or oligo-positive cells were counted in ImageJ. Immunofluorescence and RNAscope were $n\ge 3$ independent experiments, with at least 1,000 cells counted in each condition.

Statistics

Data were represented as $\text{mean}\pm \text{standard\hspace{0.17em}error\hspace{0.17em}of\hspace{0.17em}the\hspace{0.17em}mean\hspace{0.17em}}\left(\text{SEM}\right)$. Statistical significances were calculated using Student’s $t$-test. Significance was represented as at least $p<0.05$. All experiments were performed on a minimum of three different colonoid lines derived from separate normal human subjects, with a total of seven different colonoid lines used throughout these studies for validation. $n$ refers to numbers of independent replicates performed. All analyses were performed on GraphPad Prism 9.0.

Additional Software

All graphical schematics within this manuscript were generated in BioRender.com under an academic plan and includes a publication license.

Mineralogy Analysis of the UBD

Mineralogy of the $<20\mathrm{\mu }\mathrm{M}$ size fraction in the UBD was based on an analysis of $\sim \mathrm{1,000}$ individual mineral grains or grain clusters and is shown in Figure S1. This fraction consisted of a variety of minerals that are commonly present in aeolian dust in arid environments including quartz, feldspar, micas/clay minerals, calcite, and gypsum.⁵⁵ Rare minerals included baryte, rutile, and iron oxides such as hematite and magnetite. In addition to these minerals, extensive searching of the TEM grids revealed the presence of submicron grains of U-bearing minerals. Mineralogy of two of the five $50\times 50\mathrm{\mu }\mathrm{M}$ TEM grid square areas showed no U-bearing minerals, but in the three other grid squares, several grains were found, often in close spatial association. Of relevance, a) no U-bearing grains were found with grain sizes $>0.5\mathrm{\mu }\mathrm{M}$, b) the grains always occurred in clusters intimately mixed with other minerals—no individual isolated grains were found, and c) the individual U-bearing minerals were as small as $50\text{\hspace{0.17em}nM}$ in size or less. Estimating the abundance of these grains was very challenging from TEM data, but our estimate was $\sim 0.01\%$. The mass fraction was significantly lower, because the grains were all submicron, a characteristic that is unique to the U-bearing minerals.^56,57

Cytotoxicity Assay of Colonoids Post-Acute UBD Exposure

To determine if UBD induced cell death or morphological changes in human colonoids, three biologically distinct human colonoid cultures were treated with vehicle (control, organoid expansion media) or $50\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ UBD (resuspended in organoid expansion media) overnight ($\sim 18$ h). Brightfield imaging of both conditions showed no differences in gross morphology after acute UBD exposure (Figure 1A). Propidium iodide (PI) staining was used to visualize dead cells with DAPI as the nuclear counterstain. No significant difference in PI-positive cells was observed between control (Figure 1B) and UBD-treated colonoids (Figure 1C) or found when PI-positive cells were quantified (Figure 1D).

scRNA-Seq Analysis of Colonoids Post-Acute UBD Exposure

To understand if UBD was changing specific cell lineages, we performed droplet-based single-cell RNA sequencing (scRNA-seq) on three biologically unique colonoid lines (Table S1) under control or UBD treated conditions. Despite the lack of changes in cell death and overall morphology, computationally there was a stark difference in cellular gene expression. Cells in control colonoids expressed $<\mathrm{3,000}$ genes while some UBD-treated cells expressed $>\mathrm{5,000}$ genes (Figure 1E).

Graph-based clustering using the Louvain algorithm and marker-gene analysis revealed few clusters in control colonoids that primarily represented cells founds in the deep crypt, as shown in the representative control colonoid UMAP (Figure 2A). Clusters were manually identified using marker genes and published datasets^58–60 as cycling transit amplifying (TA) cells, stem cells, and two clusters of secretory progenitor cells (Excel Tables S2–S4). The colonoids represented nondifferentiated epithelia since they were maintained in organoid expansion media that contained WNT3A/Rspo-1/Noggin.

In contrast, UBD-treated colonoids maintained in identical media, had a higher degree of cell type variability, as shown in the representative UBD-treated colonoid UMAP. They grouped into clusters (Excel Tables S5–S7) that included differentiated cell types and were identified as secretory progenitors, goblet cells, absorptive colonocytes, cycling TA cells, absorptive progenitors, stem cells, and enteroendocrine (EEC)–enterochromaffin cells and EEC–L cells (Figure 2B). Since the three colonoid lines were from biologically unique donors, differential gene and clustering analysis was performed on each control and UBD-treated colonoid set to first determine if there were major differences due to genetic variability. Similar clustering patterns and differentially expressed genes were seen in all three colonoid lines for control (mainly deep crypt cells) and UBD-treated (mainly differentiated cells) conditions. Thus, control and UBD-treated colonoid datasets were merged to directly compare cell proportions in each cluster (Figure 2C; Excel Table S8). There was an increase in all types of differentiated cells but a decline in proliferative (stem and cycling TA) cells in the UBD-treated compared to control colonoids (Figure 2D).

To visualize differences within clusters between control and UBD-treated colonoids, the datasets were merged, clustered, and split according to treatment. UBD-treated cells (blue) did not entirely overlap with control cells (red) (Figure S2). Volcano plots of control and UBD-treated colonoids for all three colonoid lines showed significantly upregulated (red) and downregulated (blue) genes by treatment (Figure S3). In UBD-treated colonoids, genes associated with secretory lineage (such as TFF3, CHGA, PCSK1N, MUC2, FCGBP) were upregulated, while cell cycle and mitochondrial genes were downregulated (in relation to control). In contrast, mitochondrial genes were downregulated in UBD-treated colonoids (in relation to control colonoids).

Measurement of EECs in UBD-Treated Human Colonoids

Disruption of EECs and hormones have been reported in various gut-related and systemic metabolic diseases.^61–63 Among the cell type differences observed in UBD-treated colonoids, there was an expansion in EECs (Figure 2C, cluster 6). Since EECs are rare and control colonoids did not show major biological variation when clustered independently, we merged the control and -treated colonoids for downstream analysis. We identified EECs by the expression of chromogranin A (CHGA) and chromogranin B (CHGB), both well characterized pan-EEC markers. Feature plots and dot plots were used to identify CHGA- and CHGB-expressing cells (Figure 3A) and to quantify CHGA and CHGB expression (Figure 3B), respectively. Total number of EECs were 3.6-fold higher, with higher transcript levels of CHGA (4.4 fold) and CHGB (6.3 fold) in UBD-treated colonoids compared to controls (Figure 3C). These differences were quantified via immunostaining of CHGA-expressing cells (Figure 3D,E). Additionally, detection of RNA oligos via RNAscope in situ hybridization showed a significant difference in CHGA+ oligos per cell in UBD-treated colonoids compared to controls (Figure 3F,G).

Evaluation of Serotonin Producing-Enterochromaffin Cells and Peptide YY-Producing L Cells upon UBD Exposure

Enterochromaffin (EC) and L cells are the two most abundant EEC subtypes in the colon; however, there are at least 12 subtypes described in the small bowel and colon.⁶⁴ We further analyzed these two major EEC subtypes in our scRNA-seq datasets. EC cells were marked by expression of tryptophan hydroxylase 1 (TPH1), the enzyme that catalyzes the rate-limiting step in synthesis of serotonin (5-HT) from tryptophan,⁶⁵ while L cells were marked by expression of peptide tyrosine tyrosine (PYY).⁶⁶ Feature plots of merged control and UBD-treated colonoids showed a 5.4-fold expansion of EC cells and a 3.4-fold expansion of L cells following UBD treatment (Figure 4A–C). A scatter plot of average gene expression in EECs by treatment showed that PYY was nearly absent in control colonoids but, like CHGA, was highly expressed in UBD-treated colonoids (Figure S4).

We validated EC and L cell expansion upon acute UBD exposure by immunostaining for 5-HT (Figure 4D) and PYY (Figure 4F), respectively. Quantification of 5-HT showed that there were higher numbers of EC cells in UBD-treated colonoids compared to controls (Figure 4E). Similarly, the proportion of PYY expressing L cells were also higher in UBD-treated colonoids when compared to controls (Figure 4G). Interestingly, we observed PYY+ basal extensions (Figure 4F, arrows) from UBD-treated L cells that were absent in control colonoid L cells.

Subclustering of EECs in Control and UBD-Treated Colonoids

To further determine the spectrum of EECs that results from acute UBD exposure, we reanalyzed the EEC clusters from the combined control and UBD-treated colonoids. The EEC clusters were subclustered at a higher resolution of 1 instead of 0.4, then identified by comparing the differentially expressed genes (Excel Tables S15 and S16) with published EEC datasets.^59,67 This resulted in the control colonoid EECs separating into three clusters identified as EEC progenitors, X cells, and EC cells (Figure 5A, left UMAP). The UBD-treated EECs subclustered into eight distinct clusters identified as EEC progenitors, X cells, EC cells, N cells, K cells, L cells, and enterochromaffin-late cells (Figure 5A, right UMAP). Subclustering also identified two subtypes of L cells in the UBD-treated colonoids. A small group of L cells expressed both GCG (proglucagon) and PYY (circled cluster, solid arrow) while a larger group of L cells expressed only PYY (circled cluster, dashed arrow) (Figure 5B).

Gene Expression Profile Analysis of EECs in Control and UBD-Treated Colonoids

We evaluated the transcriptome of the newly formed UBD-induced EECs compared to control colonoid EECs. We compared gene profiles of the three common EEC clusters between control and UBD-treated colonoids [progenitors (control, 0; UBD, 6), X cells (control, 1; UBD, 1), and EC cells (control, 2; UBD, 0)] (Figure 6A). The most notable differences were observed in the EEC progenitor clusters. The topmost expressed genes in the EEC progenitor cluster in control (Figure 6A, cluster 0) were associated with secretory cells. In contrast, genes in EEC progenitors in UBD-treated cells (Figure 6A, cluster 6) were associated with the cell cycle.

Comparing relative expression amounts of select EEC-associated genes, we found that CHGA, TPH1, REG4, AGR2, PYY, and FABP1 transcripts had higher expression in the UBD-treated colonoids compared to control colonoids (Figure 6B, arrows). We performed differential gene analysis on EECs from UBD-treated colonoids to further characterize the transcriptomic changes. Over 700 genes were differentially expressed in EECs upon UBD treatment, of which 67% were upregulated compared to control. Network analysis of top upregulated genes in UBD-treated EECs showed interactions between mitochondrial and ribosomal genes (Figure 6C). We performed pathway analysis in gProfiler using the genes that were upregulated in UBD-treated EECs. Upregulated pathways from gProfiler were imported into Cytoscape for enrichment analysis. The enrichment map of UBD treated EECs shows upregulation of multiple pathways related to WNT signaling (Figure 6D).

Analysis of EEC Differentiation Transcription Factors

Based on our result indicating that EEC hyperplasia in UBD-treated colonoids may be related to upregulated WNT signaling, we sought to understand the mechanism of EEC differentiation in the context of UBD treatment. NEUROG3 and NEUROD1 are the early and late transcription factors that drive EEC differentiation in mammals.⁶⁸ Both transcription factors are considered WNT-independent, which is contrary to our network analysis of EECS in UBD-treated colonoids. Dot plot analysis showed that transcript levels of NEUROG3 and NEUROD1 were not very different between UBD treatment and control (Figure 7A).

Thus, we searched for a possible WNT-dependent transcription factor that may be driving UBD-induced EEC differentiation. In Drosophila, the transcription factor Prospero drives differentiation of midgut stem cells into EECs.⁶⁹ The mammalian homolog of Prospero is PROX1 and is WNT-dependent.⁷⁰ Therefore, we hypothesized that UBD treatment leads to activation of WNT signaling and upregulates PROX1 expression, driving EEC differentiation from secretory progenitor cells (Figure 7B). From our scRNA-seq analysis, PROX1-expressing cells were higher in number in UBD-treated colonoids compared to control colonoids (Figure 7C,D). Additionally, in UBD-treated colonoids, PROX1 transcripts were higher in cycling TA and secretory progenitor cells (Figure 7A). This PROX1 difference was validated via immunostaining (Figure 7E,F) and RNAscope (Figure 7G,H).