RNA binding protein DDX5 directs tuft cell specification and function to regulate microbial repertoire and disease susceptibility in the intestine

Abstract

Objective

Tuft cells residing in the intestinal epithelium have diverse functions. In the small intestine, they provide protection against inflammation, combat against helminth and protist infections, and serve as entry portals for enteroviruses. In the colon, they had been implicated in tumorigenesis. Commitment of intestinal progenitor cells to the tuft cell lineage requires Rho GTPase Cell Division Cycle 42 (CDC42), a Rho GTPase that acts downstream of the Epidermal Growth Factor Receptor (EGFR) and Wnt signaling cascades, and the master transcription factor POU class 2 homeobox 3 (POU2F3). This study investigates how this pathway is regulated by the DEAD box containing RNA binding protein DDX5 in vivo.

Design

We assessed the role of DDX5 in tuft cell specification and function in control and epithelial cell-specific Ddx5 knockout mice (DDX5^ΔIEC) using transcriptomic approaches.

Results

DDX5^ΔIEC mice harbored a loss of intestinal tuft cell populations, modified microbial repertoire, and altered susceptibilities to ileal inflammation and colonic tumorigenesis. Mechanistically, DDX5 promotes CDC42 protein synthesis through a post-transcriptional mechanism to license tuft cell specification. Importantly, the DDX5-CDC42 axis is parallel but distinct from the known IL-13 circuit implicated in tuft cell hyperplasia, and both pathways augment Pou2f3 expression in secretory lineage progenitors. In mature tuft cells, DDX5 not only promotes integrin signaling and microbial responses, it also represses gene programs involved in membrane transport and lipid metabolism.

Conclusion

RNA binding protein DDX5 directs tuft cell specification and function to regulate microbial repertoire and disease susceptibility in the intestine.

Introduction

Intestinal epithelial cells (IEC) lining the gastrointestinal tract are essential for nutrient absorption and provide barrier protection for the host¹. IECs are functionally heterogeneous. One subset called tuft cells is a major sensor of microbial challenges in the intestine. They express various unique surface molecules, including receptors that bind directly to noroviruses and receptors sensing microbial metabolites, including succinate^2–4. During helminth and protist challenges, tuft cells secrete IL-25 to activate type 2 innate lymphoid cells (ILC2s). In a positive feedback circuit, activated ILC2s produces interleukin 13 (IL-13) to drive tuft cell hyperplasia and ensure robust microbial clearance^{2 5–9}. Recent studies also revealed important roles of tuft cells in regulating inflammation and tumorigenesis in different intestine sections. In the ileum, reduction in tuft cell numbers is associated with elevated inflammation in the Crohn’s disease (CD) patients¹⁰. Succinate activation of the tuft cell circuit is found be protective against T cell mediated ileitis in a mouse model¹⁰. In the colon, tuft cells and their unique expression of the pro-survival doublecortin-like kinase 1 (DCLK1) have been implicated in tumorigenesis^{11 12}.

IECs have diverse turnover rates ranging from 3–5 days for enterocytes to 3–6 weeks for Paneth cells^{13 14}. They are replenished by progenitors originating from the intestinal stem cells (ISCs) residing in the crypt^{13 15}. Progenitor differentiation to specific IEC lineages is influenced by local nutrient availability, microbial composition, and gradients of Epidermal Growth Factor (EGF), Wingless-related integration site (Wnt), Bone Morphogenetic Protein (BMP), as well as cytokines from the immune system^16–18. Tuft cells in the small intestine and colon arise from the SOX4⁺ and ATOH1⁺ progenitors respectively^19–21. Commitment to the tuft cell lineage requires the expression of the POU domain transcription factor, POU2F3²². A recent study has implicated Cell Division Cycle 42 (CDC42) in tuft cell biogenesis²³. CDC42 is a ubiquitously expressed member of the Rho GTPase family. Upon activation by EGF and Wnt signaling cascades^{24,23 25 26}, CDC42 regulates actin cytoskeleton organization, polarity, proliferation, and migration^{27 28}. Mutations in Cdc42 have been linked to pediatric immunodeficiency²⁹ and severe developmental delay in patients with Takenouchi-Kosaki syndrome (TKS)^{30 31}. Global knockout of Cdc42 in mice is embryonic lethal³². Studies using mice where Cdc42 is specifically knocked out in IECs showed that CDC42 is required for ISCs growth and survival. Biogenesis of tuft cell, but not other secretory IEC subsets, is CDC42 dependent²³. However, little is known about how CDC42 contributes to tuft cell differentiation and upstream mechanisms involved in maintaining proper CDC42 expression in IECs. Here, we report that CDC42 promotes POU2F3 expression and tuft cell specification in the intestine, and this circuit is post-transcriptionally regulated by the RNA binding protein called DEAD-box helicase 5 (DDX5).

DDXs can directly bind to RNA substrates and utilize ATP hydrolysis energy to unwind RNA duplexes, facilitate RNA annealing, organize RNA-protein complex assembly,³³ and promote post-transcriptional processing³⁴. They can also partner with transcription factors to regulate gene expression^35–38. Our previous study revealed that DDX5 is the highest expressed member of the DDX family in the intestine epithelium³⁹. Overexpression of DDX5 predicts worse relapse-free survival in CRC patients^40–42, and knockdown of DDX5 can significantly inhibited the growth of cancer cells in xenograft models^{43 44}. We have recently reported that knocking out DDX5 in IECs protected against dextran sulfate sodium induced colitis as well as tumor formation in mice on a susceptible background³⁹. Given the emerging role of intestinal tuft cells in regulating intestinal inflammation and tumorigenesis, we asked whether changes in tuft cell specification and/or function may underly the contribution of DDX5 to intestinal inflammation and tumorigenesis. Here, we report an essential role of DDX5 in maintaining CDC42-POU2F3 levels in secretory lineage progenitors to drive tuft cell specification in both small intestine and colon. In differentiated IECs, including tuft cells, DDX5 promotes anti-microbial gene programs to regulate microbial composition and disease susceptibilities in models of ileitis and tumorigenesis.

Methods

Mice

C57BL/6 wild-type (Stock No: 000664) and Villin1Cre (Stock No: 021504) mice were originally obtained from the Jackson Laboratory. Previously described Ddx5^flox mice were obtained from Dr. Frances Fuller-Pace’s laboratory⁴⁵. Heterozygous mice were bred to yield 6–8 week-old Ddx5^+/+ Villin1Cre⁺ (subsequently referred to as wild-type, WT^IEC) and Ddx5^fl/fl Villin1Cre⁺ (referred to as DDX5^ΔIEC) cohoused littermates for experiments. For the tumor studies, Apc^flox mice were obtained from Dr. Eric Fearon’s laboratory⁴⁶ to generate Apc^fl/+ Ddx5^+/+ Villin1Cre⁺ and Apc^fl/+ Ddx5^fl/+ Villin1Cre⁺ (referred as APC^ΔIEC DDX5^WT) and Apc^fl/+ Ddx5^fl/fl Villin1Cre⁺ (APC^ΔIEC DDX5^ΔIEC) mice. Three (3) month old wildtype and mutant mice were given either 150 mM succinate (Alfa Aesar 41983–30) or control water (containing NaCl at 300 mM to match sodium molarity with succinate treatment) for 30 days. Colons were harvested from 120–130-day old mice to assess tumor burden. Tumor measurements were determined by double-blinded analyses using ImageJ. All animal studies were approved and followed the Institutional Animal Care and Use Guidelines of the University of California San Diego. Our vivarium at UC San Diego is kept under specific pathogen free (SPF) conditions. Regular serology and PCR tests are used to monitor and ensure the absence of EDIM, MHV, MPV, MVM, TMEV, fur mites and pinworms. MNoV is normally present in our vivarium, litters born to MNoV⁺ parents naturally acquire the virus from their environment prior to weaning and were used for experiments between 8–12 weeks of age.

The spatial transcriptomic dataset described were performed on tissues obtained from female mice. DCLK1 immunohistochemistry experiments were performed on tissues from female mice. All other experiments reported in this study were obtained from both male and female mice in similar ratios.

Metronidazole treatment

For the antibiotic treatment studies, WT^IEC and DDX5^△IEC adult mice were treated with 1% sucrose alone or 1% sucrose containing 2.5 g/L of metronidazole in the drinking water for two weeks. New beddings were provided on day 7 to limit microbe reacquisition via the fecal-oral route. Mice were monitored and weighed 2–3 times per week and fecal/cecal material were collected to assess the level of protist and bacteria clearance. Pestle-homogenized fecal or cecal materials were digested at 75°C for 10min using the KAPA Express Extract kit (R&D). 1μL of the soluble fraction was used as template for qPCR quantitation.

Anti-CD3ɛ mediated model of ileitis

WT^IEC and DDX5^ΔIEC mice were given 300mM succinate-NaCl or equal molar of NaCl in drinking water for 2 weeks and subsequently i.p. injected 3 times with 15ug of anti-CD3ɛ per mouse every other day for 5 days⁴⁷.

Histology and immunohistochemistry

Tissues harvested from the ileum and colon were fixed overnight in 10% formalin (Research Products International) at room temperature. Paraffin-embedded tissues were sectioned into 5 μm slices for Periodic Acid-Schiff (PAS) or immunohistochemical (IHC) staining (see Supplementary Table 1 for antibody information). Briefly, paraffin sections were de-paraffinized and rehydrated with Tris-buffered saline (TBST, pH 7.8 with 0.1% Tween-20) washes between each step. Sections were blocked first against endogenous peroxidases (immersed for 30 minutes in 0.3% H₂O₂), and then blocked against endogenous biotin using unlabeled streptavidin (Jackson ImmunoResearch, 016-000-114) and excess free biotin. Antigen retrieval was induced by heating the slide twice for 5 minutes in 10 mM sodium citrate buffer pH 6.0 (Sigma-Aldrich), followed by 20 minutes of cooling. Finally, sections were blocked against non-specific hydrophobic interactions with 1% bovine serum albumin (BSA; Biotium) in TBST. Staining was then performed with either the negative control IgG antibody or anti-DCLK1 (1:1000) antibodies (Abcam) overnight in a humid chamber at 4°C. The next day, sections were washed with TBST and then sequentially overlaid with biotinylated goat anti-rabbit (Jackson ImmunoResearch, 111-065-045) at 1:500, followed by HRP-labeled Streptavidin (Jackson ImmunoResearch, 16-030-084) at 1:500. Substrate was then overlaid with 3-amino-9-ethylcarbazole from Vector labs following manufacturer directions for 30 minutes followed by nuclear counterstain with Mayer’s hematoxylin. Images were acquired using the AT2 Aperio Scan Scope (UCSD Moores Cancer Center Histology Core). Three intestinal regions per tissue image were randomly selected for QuPath analysis. DCLK⁺ tuft cells were determined by QuPath Positive Cell Detection⁴⁸ (minimum area = 10μm², maximum area = 400μm², intensity threshold = 0.4). Mucin⁺ goblet cells were assessed similarly (minimum area = 40μm², maximum area = 800μm², intensity threshold = 0.6). The average score from three regions examined in each tissue were included in the final graph.

Electron microscopy

Cardiac perfusion was performed using 5mL Ringer’s solution (Fisher Scientific, Cat. #50-980-246), followed by 5ml of fixation buffer containing 2% Paraformaldehyde/2.5% Glutaraldehyde in 0.1M Sodium Cacodylate buffer solution (BIOTREND, Cat.# 15960–01). Intestinal tissues were harvested and kept in the fixation buffer in room temperature for 3 additional hours. Fixed tissues were processed and imaged on JEOL 1400 plus at the UC San Diego Electron Microscopy Core Facility.

Intestinal epithelial and lamina propria lymphocyte harvest

Steady-state intestinal epithelial and lamina propria lymphocytes were harvested as previously described⁴⁹. To isolate IECs, intestinal tissues were first incubated in 5 mM EDTA (Invitrogen) in HBSS (Gibco) containing 1 mM DTT (Invitrogen) for 20 minutes at 37°C with 200rpm agitation, and then incubated in a second wash of 5 mM EDTA in HBSS without DTT for another 20 minutes at 37°C with agitation. Single-cell suspensions released into the EDTA solutions were pooled and confirmed to contain over 85% EpCAM⁺ IECs as reported previously³⁹. For the lamina propria, IEC-depleted tissues from post-EDTA washes were digested in a HBSS solution containing 10% Fetal Bovine Serum (Peak Serum), 1.0 mg/ml Collagenase D (Roche), 100 μg/ml DNase I (Sigma-Aldrich), and 50 U/ml dispase (Worthington Biochemical Corporation) at 37°C for 30–45 minutes. Lamina propria mononuclear lymphocytes were purified from the interphase of a 40:80% Percoll^™ (Cytiva; formerly GE Healthcare Life Sciences) gradient.

Intestinal crypt isolation and organoid culture

Small intestinal and colonic crypts were isolated according to the manufacturer’s recommendation (STEMCELL, technical bulletin #28223). Briefly, 20cm of small intestine proximal to the stomach were harvested from 6–8 week-old WT^IEC and DDX5^ΔIEC littermates and cut into 2mm pieces. After 20 washes in cold PBS, tissues were resuspended in 25mL room temperature Gentle Cell Dissociation Reagent (STEMCELL, #07174) and incubated at room temperature for 15 minutes on a rocking platform at 20 rpm. The pellets enriched with intestinal crypts were resuspended in cold PBS containing 0.1% BSA. Isolated colonic crypts were embedded in Corning® Matrigel® Matrix (Corning^™ 356231) and seeded onto pre-warmed, non-treated 24-well plates (CytoOne® by StarLab) and overlaid with conditioned media (STEMCELL, #6005) as described previously⁵⁰. Organoids were treated with recombinant mouse IL-13 (BioLegend, 20ng/ml), NaCl (Sigma, 15μM), DMSO (Sigma), or ML141 (Cayman Chemical, 10μM). Organoid pictures were imaged using a Keyence bz-x800 microscope at 20X magnification with image stacks capturing the entire organoid volume.

Retroviral transduction

Authenticated Platinum-E (Plat-E) cells purchased from ATCC were grown in DMEM medium (Gibco) containing 1% L-glutamine (Gibco) and 10% fetal bovine serum (FBS; Thermo Fisher Scientific). Plat-E cells were transfected with MSCV constructs using TransIT®-LT1 Transfection Reagent (Mirus) following the manufacturer’s instructions. Virus-containing supernatant collected three days after the infection was concentrated using Retro-X concentrator (Takarabio). Virus was resuspended in organoid medium containing 8μg/ml polybrene. Dissociated organoids were resuspended in virus-containing medium for spin inoculation (600g at 32°C for 60 min) followed by a six-hour incubation at 37°C. Transduced cells were collected and seeded in Matrigel. Media was changed every 2–3 days. Organoids were analyzed by flow cytometry and/or RT-qPCR 7 days post-transduction.

Flow cytometry analysis

For analysis of tuft cells in intestinal epithelium and organoids, cells were surface stained with LIVE/DEAD Fixable Cell stain (ThermoFisher, L34957), and fluorescent conjugated antibodies against EpCAM, CD24, CD45, Siglec-F, and CD90.1 (see Supplementary Table 1 for detailed information, 1:400 in PBS) for 30 minutes. For intracellular staining of CDC42, cells were fixed/permeabilized (ThermoFisher Cat: 00-5521-00), incubated with the anti-CDC42 antibody (ThermoFisher, 1:50) for 1h in room temperature, followed by the anti-rabbit IgG (H+L) Cross-Adsorbed Secondary Antibody (Invitrogen, 1:400) for 1h in 4°C. IECs were defined as live CD4⁻CD8⁻Epcam⁺CD45^low. Successfully transduced organoid cells were defined as CD90.1⁺. Proportion (%) of IECs in the tuft cell lineage were defined as Siglec-F⁺CD24⁺ as previously described⁵¹. Flow cytometry data was analyzed with FlowJo (version 10.8.1).

Western blot

Cells were lysed in a 25 mM Tris pH 8.0 (G-Bioscience, 100 mM NaCl (G-Bioscience), 0.5% NP40 (G-Bioscience) solution with protease inhibitor (Life technologies) for 30 min on ice. Samples were spun down at 14,000 × g for 15 min, and soluble protein lysates were harvested. 30–50μg protein were loaded in each lane of a SDS-PAGE gel. Blots were blocked in Odyssey Blocking buffer (Li-CoR Biosciences) and probed for CDC42. Following incubation with IRDye secondary antibody (Li-CoR^® Biosciences), infrared signals on each blot were collected on the Li-CoR Odyssey CLX. The primary antibodies used in this study are listed in Supplementary Table 1.

cDNA synthesis and qPCR

Total RNA was extracted with the RNeasy Plus kit (QIAGEN) and reverse transcribed using iScript^™ Select cDNA Synthesis Kit (Bio-Rad). Real time RT-PCR was performed using iTaq^™ Universal SYBR^® Green Supermix (Bio-Rad). For primary IECs, results were normalized to mouse Gapdh. For organoid studies, results were normalized to β-actin. Primers were designed using Primer-BLAST to span across splice junctions, resulting in PCR amplicons that span at least one intron. Primer sequences are listed in Supplementary Table 2.

RNA-seq analysis

Intestinal epithelial cell transcriptomes from WT^IEC and DDX5^△IEC littermates were previously reported³⁹. Gene set enrichment analysis was carried out using the pre-ranked mode of the Gene Set Enrichment Analysis (GSEA) software with default settings⁵². The gene list from DEseq2 was ranked by calculating a rank score of each gene as −Log₁₀(p-value) × sign (Log₂ Fold Change), in which Fold Change is the fold change of gene expression in DDX5 ^{△ IEC} over that found in WT^IEC. De novo motif enrichment in the promoter regions of differentially expressed genes (log₂ fold change cutoffs of ≥ 0.5 or ≤−0.5 and p-values <0.05) was performed using the HOMER (v4.10) “findMotifs.pl” command line with the following parameters: “-nogo -start −1000 -end 500”⁵³. All other promoters of genes expressed in those particular cells were used as background. The top 5 motifs ranked by the lowest p-value that were found in at least 3% of the differentially expressed “Target” genes were illustrated. The tumor transcriptomic sequencing data reported in this paper is available on GEO (GSE146014, reviewer access token: gbahcsuozruptyf).

For metatranscriptomic analysis of ileal associated microbial populations, reads from the DDX5^ΔIEC and WT^IEC IEC RNAseq dataset that were not mapped to the mouse genome were assigned with taxonomic labels using Kraken v1. The standard Kraken database encompassing annotated bacterial, archaeal, and viral genomes was used for classification of sequences with the command: “kraken --classified-out /path/to/classified_fq --unclassified-out /path/to/unclassified.fq --db $DBNAME --paired --fastq-input pair1.fa pair2.fa > /path/to/results”. A Kraken report was generated with the the command: “kraken-report --$DBNAME kraken.output”⁵⁴. Differential microbial counts were assessed by DEseq2 cut off of p<0.05 with the Wald test and Log2 fold change (DDX5^ΔIEC/WT^IEC)>1.5 or <−1.5.

Spatial transcriptomics

Female mouse ileum samples were coiled into “Swiss rolls” and frozen immediately with chilled isopentane. The frozen tissues were embedded in Optimal Cutting Temperature (OCT) compound and cut in a pre-cooled cryostat at 20mM thickness onto two 6.5mm × 6.5mm capture areas with 5000 oligo-barcoded spots. Frozen tissue was tested for RNA quality with RIN > 7.0 (Tapestation). Slides were fixed, H&E stained, and imaged on Keyence (bz-x800) at 20x magnification. Tissues were permeabilized for 24 minutes. Reverse transcription and second strand synthesis were then performed on RNAs extracted from each tissue. cDNAs were quantified by qRT-PCR using Powerup SyBr Green master mix and the Bio-Rad system. The library integrity was confirmed by the Agilent TapeStation. Pooled libraries were sequenced on NovaSeq S4 (Illumina) using 150 base-pair paired-end dual-indexed configuration. Each sample was sequenced to a depth of 100 million. Spaceranger software (version 3.1.0) from 10X Genomics was used to process, align and summarize unique molecular identifier (UMI) counts against the mm10 human reference genome. Loupe Browser (version 5.0) was used for differential gene expression analysis and generation of visuals. The spatial transcriptomic sequencing data reported in this paper is available on GEO (GSE184564, reviewer access token: obaruukivrmpxad).

Ribosome pull-down assay

Small intestine IECs were lysed in polysome extraction buffer (10 mM HEPES, pH 7.4, 150 mM KCl, 5 mM MgCl2, 1% NP40, 2 mM dithiothreitol, 80 U/ml RNaseOUT, 100 μg/ml cycloheximide, and protease inhibitors). Cell extracts were subject to anti ribosome IP_overnight with 2 μg anti-RPL10A antibodies (Abcam) as described previously³⁹. Level of RPL10A associated transcripts in pull-down was calculated as fraction of input for each sample.

Statistical analysis

All values are presented as mean ± standard deviation (SD). Significant differences were evaluated using GraphPad Prism 8 software (GraphPad). The Student’s t-test was used to determine significant differences between two groups with normal distribution. A two-tailed p-value of <0.05 was considered statistically significant in all experiments.

Results

DDX5 regulates tuft cell populations in the murine small intestine and colon

In our previous report, we showed that DDX5 is highly expressed in the intestinal epithelium and regulates intestinal inflammation. However, the exact mechanism remained to be elucidated. In a recent studies activation of the tuft cell circuit is also linked to intestinal inflammation¹⁰. Therefore, we asked whether DDX5 may regulate intestinal inflammation at least in part by contributing to the differentiation and function of tuft cells and/or other IEC subsets. To address this question, we performed lineage-specific Gene Set Enrichment Analysis (GSEA) on our previously reported transcriptomes of steady-state IECs obtained from two matched pairs of WT^IEC and DDX5^△IEC male littermates³⁹. A loss of the tuft cell signature was found in both the small intestinal and colonic epithelia of DDX5^△IEC mice (Figure 1A–B and Supplementary Figure 1A). IECs from independent pairs of male and female WT^IEC and DDX5^△IEC littermates were used to confirm significant reductions in the expressions of two tuft cell-specific transcripts, Pou2f3 and Dclk1. In contrast, comparable expressions of select transcripts previously shown to be enriched in other IEC subsets, including Maf (enteroendocrine cells), Muc2 (goblet cells), Lyz1 (Paneth cells), Lgr5 and Ki67 (ISCs), were found in control and DDX5^ΔIEC mice (Supplementary Figure 1B). These results suggest that DDX5 may have a unique role in regulating tuft cell differentiation and/or function in the intestine.

Figure 1. Epithelial DDX5 promotes the tuft cell gene program in the intestine.

A. Heatmap summarizing the GSEA results on the indicated IEC gene subsets in the RNAseq dataset obtained from the ileal (SI) and colonic IECs of two pairs of male WT^IEC and DDX5^ΔIEC mice (GSE123881). NES, normalized enrichment score; FDR, false discovery rate.

B. GSEA enrichment plots of the tuft cell gene set in WT^IEC and DDX5^ΔIEC ileal (SI) and colonic IECs from A. NES, normalized enrichment score; NOM P, normalized nominal p-value.

C. Representative images from immunohistochemistry analysis of DCLK1⁺ cells in intestinal sections from female WT^IEC and DDX5^ΔIEC mice. Scale bar represents 300μm.

D. DCLK1⁺ tuft cell counts on ileal (SI) and colonic sections from 3 pairs of female WT^IEC (black, n=3) and DDX5^ΔIEC (red, n=3) mice analyzed according to the automated workflow described in Supplementary Figure 2A. Mice were fasted overnight to minimize the impact of varying food intake prior to tissue harvest. Each dot represents the result from one mouse. No primary antibody controls are shown in grey. * p<0.05 (unpaired t-test).

E. Summary of flow cytometry results of WT^IEC and DDX5^ΔIEC ileal (SI) and colonic epithelium. IECs were defined as live CD4⁻CD8⁻Epcam⁺CD45^low. Proportion (%) of IECs in the tuft cell lineage defined as Siglec-F⁺CD24⁺ are graphed. See Supplementary Figure 3A for gating strategy. Each dot represents the result from one mouse. * p<0.05, ** p<0.01 (paired t-test).

F. Human DCLK1 expression in colonic biopsies with DDX5 expression below (low) or above (high) the mean of each respective dataset (GSE83687⁵⁵ and GSE75214⁵⁶).

Next, we investigated whether the loss of tuft cell gene signature in DDX5^ΔIEC IECs may be due to a loss of the intestinal tuft cell population. Indeed, immunohistochemistry analysis revealed a significant reduction of DCLK1⁺ tuft cells in the DDX5^△IEC small intestine and colon (Figure 1C–D, workflow described in Supplementary Figure 2A). In contrast, Periodic acid-Schiff staining (PAS) of the small intestine and colon sections from control and DDX5^△IEC mice showed comparable crypt density and goblet cell numbers (Supplementary Figure 2B–D). Consistent with the histology results, flow cytometry analysis of the small intestine and colonic IECs (defined as CD45⁻EpCAM⁺) also confirmed a significant reduction of tuft cells (defined as Siglec-F⁺CD24⁺, similar to a previous report⁵¹) in the DDX5^△IEC IECs from male and female mice (Figure 1E, gating strategy detailed in Supplementary Figure 3A). These results demonstrate that DDX5 promotes tuft cell biogenesis in the murine small intestine and colon. In two human colonic IEC transcriptomic datasets previously reported^{55 56}, elevated DCLK1 expression was also associated with higher DDX5 expression (Figure 1F), suggesting regulation of the tuft cell gene program by DDX5 is likely evolutionarily conserved.

Impaired expression of the tuft cell commitment factor, Pou2f3, in DDX5^△IEC IEC secretory lineage progenitors

In the small intestine, ATOH1 and SOX4 expressing IECs are progenitors of the secretory lineages, including tuft cells, goblet cells, Paneth cells, and enteroendocrine cells^{20 57 58}. Next, we asked whether DDX5 contribute to the generation of these progenitor populations and/or their ability to express factors required for the generation of each secretory lineage, including Pou2f3-Gfi1b for tuft cells, Spdef for Goblet cells, Sox9 for Paneth cells, and Neurod1 for enteroendocrine cells. Spatial transcriptomics (10X Visium) analysis of ileal sections revealed similar number of Atoh1^high and Sox4^high spots were present in the WT^IEC and DDX5^△IEC tissues (Figure 2A). As expected, Pou2f3 transcripts were more abundantly found in Sox4^hi spots than Atoh1^hi spots in the wildtype ileum. However, its expression was greatly reduced in both Sox4^high and Atoh1^hi spots from the DDX5^△IEC ileum (Figure 2B). In contrast, Atoh1^high spots from the DDX5^△IEC ileum showed higher levels of Neurod1, but similar levels of Gfl1, Spdef, and Sox9 compared to Atoh1^high spots from WT^IEC ileum (Supplementary Figure 4A). Similar differential gene expression analysis on Lgr5^hi spots, but revealed only a relatively minor footprint of DDX5 in the ISCs (Supplementary Figure 4B and Supplementary Table 3), consistent with our earlier observation from our bulk IEC analysis (Supplementary Figure 1B). These results suggest that DDX5 may regulate tuft cell biogenesis by promoting Pou2f3 expression in the secretory lineage progenitors.

Figure 2. Spatial transcriptomic analysis revealed altered secretory lineage progenitor signature in the DDX5^ΔIEC mice.

A. Left: H&E staining of the small intestinal swiss rolls from one pair of female WT^IEC and DDX5^ΔIEC littermates used for the spatial transcriptomic analysis (Visium, 10X Genomic). Right: Loupe browser displaying Atoh1^high or Sox4^high spots on each tissue section. Scale bar: 2mm.

B. Heatmap indicating the average expression of IEC lineage-specifying genes in Atoh1^high or Sox4^high spots indicated in A.

C. Schematic of establishing intestinal crypt organoid cultures.

D. Left: Representative bright field images of organoids established from small intestinal crypts of WT^IEC and DDX5^ΔIEC mice following the protocol outlined in⁵⁰. Scale bar, 100μm. Right: Normalized mRNA expressions of Pou2f3 and Vil1 from WT^IEC (n=2) and DDX5^ΔIEC (n=2) small intestinal organoids cultured over 4–7 passages. Each dot represents the result from an independent organoid experiment. Results were averaged from two independent biological experiments. * p<0.05 (unpaired t-test).

E. Left: Representative bright field images of organoids established from colonic crypts of WT^IEC and DDX5^ΔIEC mice following the protocol outlined in⁵⁰. Scale bar, 100μm. Right: Normalized mRNA expressions of Pou2f3 and Vil1 from WT^IEC and DDX5^ΔIEC colonic organoids cultured over 4–7 passages. Each dot represents the result from an independent organoid experiment. Results were averaged from two independent biological experiments. * p<0.05 (unpaired t-test).

F. Left: HOMER de novo analysis identified the top 5 motifs present at the promoters (defined as −1 kb to +500 bp from the transcription start site) of all ileal IEC DDX5-dependent genes identified in Figure 1A³⁹. Promoters from all other IEC expressed genes were used as background. Right: The presence (red) or absence (black) of the motifs indicated on the left at the tuft cell DDX5-dependent gene promoters.

G. Average log₂ fold change of the transcript abundance of tuft cell-expressed genes harboring SIX, HMX, POU2F, NFY motifs, or None of the above motifs at their promoters in ileal IECs from two independent pairs of WT^IEC and DDX5^ΔIEC littermates. Each dot represents one gene.

Consistent with findings from the spatial transcriptomic studies, small intestine crypts enriched with ISC and progenitors isolated from DDX5^ΔIEC mice had lower Pou2f3 mRNA abundance compared to those found in crypts obtained from their WT^IEC littermates (Supplementary Figure 4C). In addition, DDX5^ΔIEC small intestinal crypts cultured ex vivo over 4–7 passages in the presence of the Wnt pathway agonist R-spondin (Rspo), epidermal growth factor (EGF), the BMP inhibitor Noggin (as previously described¹⁶), penicillin, streptomycin, and gentamicin developed into organoids and maintained lower Pou2f3 expression compared to those derived from WT^IEC mice (Figure 2C–D). Similar observations were found in DDX5^ΔIEC colonic organoids (Figure 2E). No significant change in organoid viability and size were found between cultures obtained from control or DDX5^ΔIEC crypts (Supplementary Figure 5A). Furthermore, tuft cell-specific genes harboring a promoter POU2F motif had the greatest loss in the expression in the DDX5^ΔIEC tissue (Figure 2F–G). Together, these results demonstrate an epithelial cell-intrinsic role of DDX5 in maintaining Pou2f3 expression in vivo and ex vivo.

DDX5 is dispensable for IL-13 driven tuft cell hyperplasia

Differentiation of IEC subsets, including tuft cells, is heavily influenced by the local microbial composition and immune cytokine profiles. In the presence of protozoan Trichomonas muris, for example, the small intestinal tuft cell population expands rapidly in an IL-13-dependent manner^{3 4}. Therefore, we asked whether the loss of tuft cell numbers in the DDX5^ΔIEC intestine may be explained by a reduction in local IL-13 level, difference in Trichomonas abundance, and/or expressions of IL-13 receptors in the ISC and progenitors. At the RNA level, we found a similar abundance of transcripts encoding Il13 expressed by immune cells in the lamina propria of cohoused WT^IEC and DDX5^ΔIEC littermates (Supplementary Figure 6A). Furthermore, cohoused WT^IEC and DDX5^ΔIEC littermates acquired similar levels of Trichomonas species from their parents prior to weaning (Supplementary Figure 6B).

To assess whether the Trichomonas species present in our mouse colony were implicated in the tuft cell specification, we treated cohoused WT^IEC and DDX5^ΔIEC littermates with metronidazole in drinking water for 14 days to deplete majority of the Trichomonas species together with other metronidazole sensitive bacterial species (Supplementary Figure 6B). Interestingly, WT^IEC mice treated with metronidazole had no significant change in the proportion of tuft cells in the small intestine and colon as measured by flow cytometry (defined as Siglec-F⁺CD24⁺). Likewise, the proportion of tuft cells and epithelial Pou2f3 transcript levels in the metronidazole-treated DDX5^△IEC mice remained lower than levels found in WT^IEC mice under similar treatment (Supplementary Figure 6C). These results suggest that the metronidazole-sensitive Trichomonas and bacteria species present in our mouse colony were not involved in the establishment of tuft cell in the WT^IEC and DDX5^ΔIEC mice. It remained to be investigated whether specific metronidazole-insensitive microbe(s) and/or metabolites present in our colony contribute to the DDX5-dependent generation of tuft cell we observed.

Furthermore, we found no defect in the ability of DDX5^ΔIEC ISC and secretory lineage progenitors to express genes encoding receptors for IL-13 (Supplementary Figure 7A). Therefore, we hypothesized that DDX5 is likely dispensable for tuft cell hyperplasia in response to IL-13. To test this possibility, WT^IEC and DDX5^ΔIEC small intestinal and colonic organoids were treated with recombinant IL-13 for three days. At the RNA level, both WT^IEC and DDX5^ΔIEC organoids upregulated Pou2f3 to a similar extent, resulting in a ≥2 fold increase in tuft cell numbers compared to vehicle-treated organoids (Supplementary Figure 7B–D). Together, these results suggest that DDX5 is dispensable for type 2 immune cytokine-driven tuft cell hyperplasia.

DDX5 promotes Cdc42 translation to induce tuft cell biogenesis

Next, we asked whether DDX5 promotes Pou2f3 expression and drives tuft cell biogenesis by directly binding to Pou2f3 mRNA and/or transcripts encoding its upstream regulators. To this end, we examined the DDX5-associated RNA interactome identified by the enhanced cross-linked immunoprecipitation (eCLIPseq) experiment on IECs derived from the steady-state WT small intestine³⁹. To our surprise, we did not identify any DDX5 footprint on the Pou2f3 transcript, suggesting that DDX5 likely regulate Pou2f3 expression indirectly. Of the eleven known molecules previously implicated in Pou2f3 expression and tuft cell biogenesis (Gfi1b, Mtor, Rptor, Stat6, Ogt, Tas1r3, Sox4, Hes1, Dll1, Atoh1, and Cdc42), only the Cdc42 transcripts were bound by DDX5 (Figure 3A). CDC42 is a GTP-binding protein previously implicated in ISC survival, growth, as well as tuft cell biogenesis²³. In the absence of DDX5, IECs had reduced CDC42 protein abundance (Figure 3B). Spatial transcriptomics analysis revealed that Cdc42 is ubiquitously expressed across all IEC subsets, and most abundantly found in Lgr5^hi ISCs and Sox4^hi progenitors (Figure 3C). Murine retrovirus carrying a CDC42 (isoform 1, V1) expression construct partially rescued CDC42 protein levels in DDX5^ΔIEC small intestinal organoids (Supplementary figure 8A–B) and restored the Siglec-F⁺CD24⁺ tuft cell population to levels found in WT^IEC cultures (Figure 3D–E). These results demonstrate that DDX5 promotes tuft cell biogenesis by regulating CDC42.

Figure 3. DDX5 regulates Pou2f3 expression and tuft cell specification by promoting Cdc42 translation.

A. Left: Schematic representation of the enhanced CLIP (eCLIPseq) experiment. Right: Integrative Genomics Viewer (IGV) browser view of DDX5 eCLIPseq signals at the Cdc42 loci from two independent experiments.

B. Representative western blot for CDC42 in whole cell lysates of small intestinal IECs from WT^IEC and DDX5^ΔIEC mice. Experiments were repeated three times using independent biological samples with similar results.

C. Violin plots displaying expression of Dclk1, Pou2f3, Ddx5, and Cdc42 in small intestinal stem cells (ISC, Lgr5^hi), Atoh1^hi and Sox4^hi secretory lineage progenitors, or tuft cells from the spatial transcriptomics dataset described in Figure 2A.

D. Summary of the % of SiglecF⁺CD24⁺ tuft cells in WT^IEC and DDX5^ΔIEC small intestinal organoids 7 days post-transduction with retrovirus (marked by CD90.1⁺) carrying the indicated expression vectors. Each dot represents the result from an independent organoid experiment. n.s. not significant, * p<0.05, ** p<0.01 (unpaired t-test).

E. Representative flow cytometry plots from D.

F. Heatmap indicating the average expression of known signaling molecules upstream of CDC42 activation in ISCs, progenitors, and tuft cells from Figure 2A.

G. Fold change of Pou2f3 mRNA in WT^IEC (n=6) or DDX5^ΔIEC (n=5) small intestine organoids cultured in the presence of CDC42 inhibitor ML141 (10μM) for 24 hours relative to WT^IEC cells treated with vehicle control (DMSO). Each dot represents the result from one mouse. n.s. not significant, ** p<0.01, **** p<0.001 (unpaired t-test).

H. IGV browser view of WT^IEC and DDX5^ΔIEC ileal IEC RNAseq-derived sample reads mapped to the Ddx5 and Cdc42 loci (GSE123881).

I. Fold change of ileal Cdc42 mRNA in DDX5^ΔIEC mice (n=3) relative to values found in WT^IEC mice (n=3). Each dot represents the result from one mouse. n.s. not significant (unpaired t-test).

J. Ribosome RPL10A enrichment from Cdc42 mRNA in small intestinal IECs. Results shown represent the average of two independent experiments. Each dot represents the result from one mouse. * p<0.05 (unpaired t-test).

EGF and R-spondin in the organoid culture media engage EGFR and Wnt signaling pathways known to drive CDC42 activity. At the RNA level, small intestinal Sox4^hi progenitors abundantly expressed Egfr, Fzd2, and Lrp5 (Figure 3F). Therefore, we asked whether CDC42 GTPase activity may be implicated in the generation of tuft cells in the organoid culture system. Consistent with this possibility, DDX5^ΔIEC small intestinal organoids transduced with the dominant negative CDC42^T17N or CDC42^G12V expression constructs failed to restore the Siglec-F⁺CD24⁺ tuft cell population (Figure 3D), suggesting that the GTPase activity of CDC42 is required for DDX5-mediated tuft cell biogenesis. To bypass the essential function of CDC42 in ISC reported previously^{23 59}, established WT organoids were passaged, seeded on fresh Matrigel, and allowed to differentiate normally for 48hrs and then subjected to CDC42 inhibition by ML141. In this assay, acute ML141 treatment did not result in measurable difference in organoid survival or growth (Supplementary Figure 9A). We also did not detect significant change in the abundance of Klf4, Lyz1, and Lgr5 transcripts in ML141 treated organoids (Supplementary Figure 9B). However, Pou2f3 expression was significantly diminished in WT organoids transiently treated with ML141 for 24hrs (Figure 3G). Intriguingly, ML141 treated wildtype small intestinal and colonic organoids remained responsive to IL-13 by expanding their tuft cell population to a similar extent as control treated cultures (Supplementary Figure 9C–D). Together, these results confirm that the DDX5-CDC42 axis is dispensable for IL-13-induced tuft cell expansion ex vivo.

Mechanistically, DDX5 did not alter IEC Cdc42 transcript abundance (Figure 3HI). Ribosome capture assay revealed that optimal translation machinery engagement on Cdc42 mRNA transcripts in small intestinal IECs was DDX5-dependent (Figure 3J). Intriguingly, retrovirus carrying wildtype DDX5 or the helicase dead DDX5^DEAD, expression constructs were able to rescue the Siglec-F⁺CD24⁺ tuft cell population in DDX5^ΔIEC small intestinal organoids to levels found in WT^IEC cultures (Figure 3D). This suggests that DDX5 helicase activity is dispensable for promoting Cdc42 ribosomal engagement. Together, results from our genetic, genomic, and pharmacological experiments demonstrate that epithelial DDX5 binds Cdc42 transcripts and promotes CDC42 protein translation in a helicase activity-independent manner to control tuft cell biogenesis.

DDX5 negatively regulates tuft cell lipid and protein metabolic programs

Despite a significant loss of tuft cells in the DDX5^△IEC intestine, both immunohistochemistry and flow cytometry analyses demonstrated a small population of tuft cells can be generated in the absence of DDX5 (Figure 1C–E). Next, we tested whether DDX5^△IEC tuft cells may be morphological and/or functional distinct from those with intact DDX5 expression. Electron microscopy revealed similar brush-like microvilli structures characteristic of elongated tuft cells in both WT^IEC and DDX5^△IEC small intestine (Figure 4A), suggesting that DDX5 does not influence tuft cell morphology.

Figure 4. DDX5-dependent tuft cell gene programs in the small intestine.

A. Representative electron microscopy images of small intestinal tuft cells from WT^IEC and DDX5^ΔIEC mice. Scale bar: 5μm.

B. Loupe browser screenshots displaying quantified Dclk1^hi spots on the small intestinal sections from WT^IEC and DDX5^ΔIEC mice described in Figure 2A. Scale bar: 2mm.

C. Gene ontology analysis revealed the top 7 pathways downregulated (black) or upregulated (red) in DDX5^ΔIEC Dclk1^hi cells from B.

D. Violin plots displaying expressions of DDX5-dependent genes—with or without DDX5-associated transcripts—involved in the antimicrobial response (Defa3 and Defa17), viral response (Duox2 and Hspb1), transport (Slc6a19), proteolysis (Anpep and Cnbp2), or lipid metabolism (Apoa1) in WT^IEC and DDX5^ΔIEC Dclk1^hi cells from C.

E. Log fold changes (DDX5^ΔIEC/ WT^IEC) in the expression of tuft cell DDX5-dependent genes participating in pathways identified in C. Small intestinal IEC DDX5-associated transcripts identified in eCLIP analysis are indicated with a blue star (GSE124023).

Previous reports clustered the small intestinal tuft cells into two main subsets: one having higher expression of genes related to neuronal development (type 1), and the other (type 2) enriched with immune-related genes, such as the Protein Tyrosine Phosphatase Receptor Type C gene encoding the pan-immune cell marker CD45^{1 60}. Of the few DDX5^△IEC tuft cells present in the small intestine, their proportion of CD45 low- and mid-subpopulations were similar to those found among WT^IEC tuft cells (Supplementary Figure 10A). GSEA of the small intestinal IEC transcriptomes³⁹ also confirmed that both type 1 and 2 programs were significantly impaired in the DDX5^△IEC epithelium (Supplementary Figure 10B).

To fully elucidate IEC subset-specific DDX5-dependent gene programs, we performed differential gene analysis on the spatial transcriptomic datasets obtained from the WT^IEC and DDX5^△IEC small intestinal sections discussed in Figure 2A. We uncovered 17.5% of the DDX5-induced transcripts and 3.8% of the DDX5-repressed genes were shared among tuft, enteroendocrine, goblet, and Paneth cells (Supplementary Figure 11A–B and Supplementary Table 4). The numbers of DDX5-regulated genes in goblet and enteroendocrine cells is lower than those found in tuft and Paneth cells. Notably, DDX5^△IEC small intestinal tuft cells had reduced expression of genes involved in microbial responses (Figure 4B–D), including a loss of Cd300lf transcripts encoding a surface molecule used by the Murine norovirus (MNoV) as a key docking site for host entry² (Supplementary Figure 12A). As a result, DDX5^ΔIEC mice raised by MNoV⁺ parents had significantly fewer MNoV transcripts in their small intestinal epithelium compared to their cohoused wildtype littermates (Supplementary Figure 12A). In addition, meta-transcriptomic analyses of the WT^IEC and DDX5^△IEC intestinal epithelium transcriptomes revealed significantly lower mucosal-associated reads mapping to viral and bacterial genomes from the DDX5^ΔIEC small intestine and colon, respectively (Supplementary Figure 12B–C). In particular, species of Helicobacters and Prevotella were significantly reduced in the DDX5^ΔIEC colon (Supplementary Figure 12D and Supplementary Table 5).

While none of the DDX5-dependent genes involved in microbial responses harbor a DDX5 eCLIPseq footprint, ten of the genes involved in transmembrane transport and protein and lipid metabolism that were upregulated in DDX5^△IEC tuft cells were direct targets of DDX5 (indicated with a blue star on Figure 4E). These results reveal a novel role of DDX5 as a repressor of transmembrane transport and protein and lipid metabolic programs in tuft cells of the small intestine. Together, these results extend the role of DDX5 in shaping IEC subset gene programs and modulating the microbial community in the intestine.

Succinate-induced tuft cell hyperplasia protects against ileitis and restores colon tumorigenic potential in DDX5^△IEC mice

Small intestinal tuft cells have been recently reported to provide protection against T-lymphocyte mediated ileitis upon anti-CD3ɛ challenge¹⁰. Therefore, we hypothesized that the reduced tuft cell numbers in the small intestine of DDX5^ΔIEC mice may result in enhanced susceptibility to ileitis. Consistent with our hypothesis, 50% of the DDX5^ΔIEC mice challenged in this model succumbed to the disease by day 18 (Figure 5A). Of the ones that survived, mononuclear immune cells from their ileal lamina propria had elevated transcripts encoding the inflammatory cytokine TNF (Figure 5B). Pre-administration of succinate, a microbial-derived metabolite known to promote tuft cell hyperplasia in an IL-13 dependent manner^{4 6}, in drinking water was able to partially restore Dclk1 expression in the small intestine epithelium of DDX5^ΔIEC mice (Figure 5C), similar to previous reports^{3 10}, dampened Tnf expression, and significantly protected them against mortality from the anti-CD3ɛ challenge (Figure 5A–B).

Figure 5. Succinate-induced tuft cell hyperplasia protects DDX5^ΔIEC mice against exacerbated ileitis and restores tumorigenic potential in the DDX5^ΔIEC colon.

A. Survival plot of mice treated with succinate-NaCl (Succ., WT^IEC, n=6; DDX5^ΔIEC, n=7) or equal molar of NaCl (Veh., WT^IEC, n=4; DDX5^ΔIEC, n=6) in drinking water for 2 weeks and subsequently challenged by by 15μg of anti-CD3ɛ per mouse for 5 days. * p<0.05 (Mantel-Cox test).

B. Normalized mRNA expression of Tnf in small intestinal lamina propria (LP) mononuclear cells isolated from mice from A on day 19. Each dot represents the result of one mouse. n.s. not significant, * p<0.05 (paired t-test).

C. Normalized mRNA expression of Dclk1 in the small intestinal IECs isolated from mice from A on day 19. One dot represents the result of one mouse. * p<0.05 (unpaired t-test).

D. Representative images of DCLK1 staining in colon from mice treated with NaCl (Veh.) or succinate-NaCl in drinking water for 30 days. Scale bar represents 100μm.

E. Left: Representative images of colonic tumors from 120 days old APC^ΔIEC DDX5^WT and APC^ΔIEC DDX5^ΔIEC mice treated with succinate-NaCl or equal molar of NaCl in drinking water on day 90 for 30 days. Right: Summary of colonic tumor counts from treated APC^ΔIEC DDX5^WT and APC^ΔIEC DDX5^ΔIEC mice. Each dot represents the result from one mouse. * p<0.05 (unpaired t-test). Scale bar represents 1cm.

F. Working model. Left: DDX5-CDC42-POU2f3 axis regulating progenitor commitment toward the tuft cell lineage. Right: DDX5-dependent gene programs in differentiated tuft cells.

In the colon, multiple IEC subsets, including DCLK1⁺ tuft cells, have been reported to harbor tumorigenic potential^{11 12}. We recently reported that DDX5^ΔIEC mice on the intestinal tumor-susceptible Apc^ΔIEC background⁴⁶ harbored fewer intestinal tumors³⁹. Transcriptomic analysis of control and DDX5 deficient tumors revealed pathways as well as a subset of DDX5-dependent genes that were similarly regulated in tuft cells (Supplementary Figure 13A–C and Supplementary Table 6). Therefore, we tested whether succinate supplementation can also promote colonic tuft cell programs, including Dclk1 expression, and restore the tumorigenic potential of the APC^ΔIECDDX5^ΔIEC mice. Immunohistochemistry analysis confirmed partial restoration of the DCLK1⁺ population in succinate treated DDX5^ΔIEC mice (Figure 5D). In the tumor susceptible background APC^ΔIEC, succinate treated DDX5^ΔIEC mice had higher colonic tumor counts than those treated with vehicle control alone (Figure 5E). Altogether, these results highlight the critical roles of epithelial DDX5 in protecting against ileal inflammation yet contributing to colonic tumorigenesis (modeled in Figure 5F).

Discussion

Tuft cells in the small intestine and colon arise from the SOX4⁺ and ATOH1⁺ progenitors respectively^19–21. Commitment to the tuft cell lineage requires the expression of the POU domain transcription factor, POU2F3²². Induction of Pou2f3 is best characterized in response to IL-13 and IL-4 stimulation during helminth and protist challenges^{2 5–9}. However, the molecular mechanisms that control Pou2f3 expression in the absence of IL-13 and IL-4 stimulation remain elusive. Here, we report that CDC42 and its GTPase activity is essential for Pou2f3 expression. Consistent with our findings, epithelial specific knockout of CDC42 resulted in a loss of intestinal tuft cells, but no change to other secretory IEC subsets²³. In the small intestine, we show that Cdc42 is the abundantly expressed in ISCs and secretory lineage progenitor cells. Optimal translation of the Cdc42 transcript rely on the RNA binding protein DDX5 in a helicase-activity independent manner. Knocking out DDX5 specifically in IECs results in reduced CDC42 protein abundance and loss of tuft cell populations in both the small intestine and colon. Ex vivo, retroviral expression of wildtype CDC42 rescued the tuft cell biogenesis defect in small intestinal DDX5^ΔIEC organoids.

Surprisingly, the DDX5-CDC42 axis is dispensable for tuft cell hyperplasia in response to IL-13. WT organoids cells treated with the CDC42 inhibitor and DDX5^ΔIEC organoids remain capable of responding to IL-13, upregulating Pou2f3 expression, and inducing tuft cell expansion. Future studies will be needed to investigate whether these parallel DDX5-CDC42 and IL-13-induced pathways recruit shared or distinct transcription factors and/or machineries to drive Pou2f3 expression in the secretory lineage progenitors. Compared to the secretory lineage progenitors, DDX5 has a limited transcription footprint on ISCs. We also did not find abnormalities in growth and survival in the DDX5^ΔIEC crypts containing ISCs and progenitors, unlike those observed in cells from the CDC42^ΔIEC mice²³. We speculate that this likely suggest that the remaining CDC42 levels in the DDX5^ΔIEC epithelium is sufficient to maintain ISC growth and survival and/or CDC42 expression in ISC may be DDX5-independent. Future studies that genetically ablate Ddx5 in specific IEC progenitor and/or IEC subsets will be needed to more definitively address whether the contribution of DDX5 to tuft cell differentiation and function may be cell-autonomous and/or indirect.

In the small intestine, tuft cells protect against ileal inflammation in a mouse model and a loss of tuft cells is identified in Crohn’s disease patients¹⁰. As expected, DDX5^ΔIEC mice with reduced intestine tuft cell biogenesis were more susceptible to disease in a model of ileitis. Administration of succinate partially restored tuft cell numbers in DDX5^ΔIEC mice and allowed for greater resilience during ileitis. Compared to the small intestine, the colon epithelium harbored fewer tuft cells and their physiologic and pathologic functions are not well understood. Results from this study suggest that the DDX5-CDC42 axis is also implicated in colonic tuft cell biogenesis. In contrast to the anti-inflammatory role of DDX5 in the small intestine IECs, we previously reported that DDX5 in the colonic IECs promotes inflammation in a DSS-induced model of colitis³⁹. Regional specific roles of DDX5 in intestine inflammation may be related to distinct microbe-host interactions and/or distinct gene programs regulated by DDX5 in colonic IECs. Future single-cell transcriptomic studies will be needed to fully elucidate the mechanisms underlying these specificities. Development of additional IEC-subset specific knockout models will be needed for definitively attributing specific DDX5 regulated programs to the various phenotypes observed here.

In summary, this study uncovered an epithelial cell intrinsic pathway involving the RNA binding protein DDX5 and its ability to regulate the CDC42-POU2F3 axis critical for tuft cell biogenesis that is parallel but distinct from the IL-13 induced hyperplasia responses and can serve as new molecular targets for the restoration of mucosal homeostasis in disease settings.

Supplementary Material

Supp1

Supplementary Figure 1: IEC-subset specific gene expression in the ileal and colonic epithelium.

A. Expression of genes determined by RNAseq (DESeq2) analysis of the ileal and colonic IECs from two pairs of WT^IEC and DDX5^ΔIEC mice as described in Figure 1A. Proximal enterocytes (Prox Ent), distal enterocytes (Dist Ent); Paneth cells (P), enteroendocytes (E.E). Black: log₂ fold changes (DDX5^ΔIEC/WT^IEC) > 0.5; blue: log fold changes (DDX5^ΔIEC/WT^IEC) < −0.5.

B. Subset-specific gene expression analysis by RT-qPCR of ileal (left) and colonic (right) IECs from WT^IEC (n=5) and DDX5^ΔIEC mice (n=5). Each dot represents the result from one mouse pair.

Supplementary Figure 2: Intestinal crypt density and goblet cell numbers are DDX5-independent.

A. DCLK1⁺ cell quantitation workflow in QuPath⁴⁸.

B. Representative Periodic acid-Schiff (PAS) staining of WT^IEC and DDX5^ΔIEC small intestines. Scale bar represents 100μm.

C. WT^IEC (n=3) and DDX5^ΔIEC (n=3) mice Ileal and colonic crypt densities. Each dot represents the result from one mouse. n.s. not significant (unpaired t-test).

D. WT^IEC (n=3) and DDX5^ΔIEC (n=3) mice ileal and colonic goblet cell counts. Each dot represents the result from one mouse. n.s. not significant (unpaired t-test).

Supplementary Figure 3: Tuft cell gating strategy.

A. Small intestines were incubated in 5 mM EDTA containing HBSS for 20 minutes at 37°C for the isolation of IECs. SSC: side scatter (an index for cell granularity). Tuft cells were defined as live EpCAM⁺CD4⁻CD8β⁻CD45^low-midSiglec-F⁺CD24^{+ 4}.

FSC-A: area under forward scatter (an index for cell size).

Supplementary Figure 4: DDX5 dependency of IEC subset defining genes in the intestinal ISC and secretory lineage progenitors.

A. Violin plots of expression from the indicated genes in Atoh1^hi or Sox4^hi cells from the spatial transcriptomic analysis from figure 2A.

B. Venn diagrams of distinct and overlapping DDX5-dependent genes (p<0.05) among ISC and the secretory lineage progenitors.

C. Normalized RNA expressions of Pou2f3 and Vil1 in small intestinal crypts isolated from two pairs of WT^IEC and DDX5^ΔIEC mice. Each dot represents the result from one mouse. ** p<0.01 (unpaired t-test).

Supplementary Figure 5: DDX5 is not involved in intestinal organoid survival and growth.

A. Viability and size of small intestinal organoids (46 WT^IEC from 5 random fields; 62 DDX5^ΔIEC organoids from 7 random fields) and colonic organoids (10 WT^IEC from 5 random fields; 11 DDX5^ΔIEC organoids from 5 random fields) two days after re-seeding of established cultures. Images were acquired under a 20X magnification.

Supplementary figure 6. Similar type 2 cytokines and trichomonas levels in WT^IEC and DDX5^ΔIEC mice.

A. RNA expressions of Il4, Il13, and Gata3 normalized to Gapdh in intestinal lamina propria lymphocytes from WT^IEC (SI n=4; colon n=3) and DDX5^ΔIEC (SI n=4; colon n=3) mice.

B. Levels of Trichomonas (left) and bacterial (right) genomes normalized to host genome detected in fecal or cecal contents. Each dot represents the result from one mouse. n.s. not significant, **** p<0.001 (t-test). Metro: metronidazole administered by drinking water.

C. Proportion of tuft cells and normalized RNA expression of Pou2f3 in the intestines of WT^IEC and DDX5^ΔIEC treated with or without metronidazole. Results shown are the average and standard deviation of two independent experiments. Each dot represents result from one mouse. * p<0.05 (t-test).

Supplementary Figure 7. IL-13 driven tuft cell hyperplasia is DDX5-independent.

A. Heatmap indicating the average expression of genes encoding receptors for IL-13 and IL-4 in ISCs, progenitors, and tuft cells from Figure 2A.

B. Normalized RNA expression of Pou2f3 in small intestine or colon organoids stimulated with IL-13 (20ng/ml) or vehicle (NaCl at 15μM) for 48hrs. Results shown are the average and standard deviation of two independent experiments. n.s. not significant, * p<0.05 (paired t-test).

C. Representative flow cytometry plots of SiglecF⁺CD24⁺ tuft cell populations from WT^IEC and DDX5^ΔIEC small intestinal and colonic organoids stimulated with IL-13 (20ng/ml) or vehicle (NaCl at 15μM) for 72hrs.

D. Summary of fold changes in the proportions of tuft cells in intestinal organoids stimulated with IL-13 compared to vehicle. Each dot represents the result from one independent experiment.

Supplementary Figure 8. Restoring CDC42 protein levels in DDX5^ΔIEC small intestinal organoids by retroviral transduction.

A. Representative retrovirus transduction efficiency in small intestinal organoids was assessed using flow cytometry to measure surface expression of CD90.1. Experiments were repeated three times using independent biological samples with similar results.

B. Left: Representative CDC42 protein abundance in transduced small intestinal organoids as assessed by flow cytometry. Right: Summary of CDC42^hi small intestine organoid IECs from three independent experiments. n.s. not significant, * p<0.05 (paired t-test).

Supplementary Figure 9. Transient inhibition of CDC42 activity did not alter wildtype intestinal organoid viability, growth, or response to IL-13 stimulation.

A. Viability and size of WT^IEC small intestinal organoids 24hrs post treatment (DMSO, n=553 from 4 random fields; ML141, n=551 from 4 random fields) and 72hrs post treatment (DMSO, n=259 from 2 random fields; ML141, n=267 from 2 random fields). **** p<0.001, n.s. not significant (unpaired, t-test).

B. Normalized RNA expressions of Lyz1, Klf4 and Lgr5 in WT^IEC small intestinal organoids cultured in the presence or absence of CDC42 inhibitor ML141 (10μM) for 24hrs. Each dot represents the result from one mouse. n.s. not significant (unpaired, t-test).

C. Representative flow cytometry plots of WT^IEC small intestinal or colonic organoids cultured for 72hrs in the presence or absence of recombinant IL-13 (20ng/ml) and/or CDC42 inhibitor ML141 (10μM).

D. Summary of fold changes in tuft cell frequencies among WT^IEC small intestinal and colonic organoids upon treatment with DMSO or ML141 inhibitor from C. Tuft cell lineage was defined as Siglec-F⁺CD24⁺ IECs⁴. Each dot represents the result from one independent experiment.

Supplementary Figure 10. DDX5 promotes both type 1 and type 2 tuft cell programs in the small intestine.

A. Left: Histogram of CD45 expression on gated ileal tuft cells (CD4⁻CD8⁻Epcam⁺Siglec-F⁺CD24⁺). Right: Proportion (%) of CD45^low or CD45^mid cells within the tuft cell lineage. Results from four pairs of WT^IEC and DDX5^ΔIEC cohoused littermates were graphed. n.s. not significant (unpaired t-test). B. GSEA of Type 1 or Type 2 tuft cell gene sets in ileal IEC transcriptomes from WT^IEC and DDX5^ΔIEC mice. NES, normalized enrichment score; NOM P, normalized p-value.

Supplementary Figure 11. DDX5-dependent and -independent genes among small intestinal IEC lineages.

A. Violin plots displaying expressions of IEC-lineage marker genes among IEC lineages assessed from the spatial transcriptomic dataset described in Figure 2A.

B. Venn diagrams of distinct and overlapping downregulated and upregulated DDX5-dependent genes (p<0.05) among small intestine IEC lineages.

Supplementary Figure 12. DDX5-dependent viral and bacterial populations in the murine intestine.

A. Left: Normalized RNAseq read counts of Cd300lf and Dclk1 in two pairs of ileal IECs from cohoused WT^IEC and DDX5^ΔIEC littermates. Right: Normalized MNoV RNA abundance from five pairs of ileal IECs from cohoused WT^IEC and DDX5^ΔIEC littermates. * p<0.05 (unpaired t-test).

B. Schematic of meta-transcriptomic analysis of ileal IEC RNAseq data from two pairs of WT^IEC and DDX5^ΔIEC mice.

C. Normalized transcript counts assigned to viral, bacterial and archaeal genomes by meta-transcriptomic analysis as illustrated in B. * p<0.05 (unpaired t-test).

D. Significantly altered microbial species in the DDX5^ΔIEC colons from C (DESeq2). p<0.05, Log2 fold change (DDX5^ΔIEC/WT^IEC)>1.5 or <−1.5.

Supplementary Figure 13. DDX5-dependent programs in colonic tumors.

A. Volcano plot depicting DDX5-dependent transcripts in murine colonic tumors. One dot represents one gene (black dot, pV<0.05).

B. Gene ontology analysis revealed the top 7 pathways downregulated (black) or upregulated (red) in DDX5^ΔIEC colonic tumor cells from A.

C. Heatmap of select DDX5-dependent transcripts and their expression in control or DDX5 deficient colonic tumors.

What is already known on this subject?

• A subset of intestinal epithelial cells known as tuft cells orchestrates mucosal defense against helminths and parasites, contributes to epithelial repair, and reduces ileal inflammation, but also serves as a viral reservoir and harbors tumor stem cell properties.

• Tuft cell specification requires Rho GTPase Cell Division Cycle 42 (CDC42) and the master transcription factor POU class 2 homeobox 3 (POU2F3). However, little is known about how this pathway is regulated in vivo.

• RNA helicase DDX5 promotes tumorigenesis in the small intestine and colon. However, little is known about its role in normal IEC differentiation and function.

What are the new findings?

• In the small intestine, knocking out DDX5 specifically in intestinal epithelial cells results in a dramatic reduction of intestinal tuft cell numbers and exacerbation of ileal inflammation,

• In the colon, knocking out DDX5 specifically in intestinal epithelial cells results in alteration of microbial composition and protection against colonic tumorigenesis.

• Mechanistically, DDX5 binds to and drives the translation of Cdc42 transcripts to maintain proper CDC42 level in the secretory progenitor cells required for Pou2f3 expression and tuft cell specification.

What is the foreseeable clinical impact?

• Our findings reveal an epithelial cell intrinsic (DDX5-CDC42-POU2F3) pathway in promoting intestinal tuft cell specification. Modulation of this pathway may have therapeutic potentials for combating ileal inflammation and colonic cancers.