Loss of Sox4 results in increased proliferation and decreased ISC function

Complete deletion of Sox4 results in embryonic lethality ²⁷. To enable examination of Sox4 function in adult intestinal epithelium, we crossed previously generated Sox4^fl/fl mice to mice expressing Cre recombinase driven by the villin promoter (vilCre), which is constitutively expressed in the intestinal epithelium beginning at E13.5 ^{28, 29}. To examine Sox4 expression, we isolated CD44− and CD44+ populations from CD326 (EPCAM)+ control and Sox4cKO adult intestinal epithelium by FACS and confirmed CD44 enrichment by qPCR (Supplementary Figure 1A–C). CD44 marks crypt cells as well as EECs in the villi (Supplementary Figure 1A). As expected, control animals exhibited significant enrichment of Sox4 in the CD44+ population, consistent with restricted expression of Sox4 in the intestinal crypts (Supplementary Figure 2A). Sox4cKO animals did not express detectable levels of Sox4 mRNA or protein (Supplementary Figure 2A, B).

To assay the role of Sox4 in ISC/TA proliferation, we examined KI67 and EdU incorporation, which was administered as a single, 2hr pulse. Sox4cKO crypts exhibited significantly higher percentages of total proliferating cells as well as cells in S-phase (Figure 2A, B). This increase in proliferation was reflected by increased crypt length in Sox4cKO intestines (Supplementary Figure 2C). In organoid cultures, Sox4cKO crypts grew at the same rate as control crypts, suggesting that hyperproliferation in the absence of Sox4 may be masked by exogenous growth factors required for organoid culture (Figure 2C).

Open in a separate window

Figure 2.

Loss of Sox4 results in increased proliferation and decrease in functional stemness, with no change in ISC numbers.

(A) Sox4cKO intestines exhibit increased total proliferation by KI67 staining, as well as (B) increased numbers of cells in S-phase, as indicated by EdU (scale bar represents 100um). (C) Whole crypts isolated from Sox4cKO intestines form organoids that undergo similar growth rates to control crypts (scale bar represents 100um). (D) Numbers of OLFM4+ ISCs remain unchanged in Sox4cKO crypts, while (E) Lgr5 is downregulated in CD44+ populations from Sox4cKO intestines. Functional stemness, as assayed by organoid forming ability, is significantly decreased in CD44+ populations isolated from Sox4cKO intestines (scale bar represents 25um; asterisks indicate significance; p < 0.05).

Due to its known role in maintenance of stem/progenitor pools in other tissues, we sought to determine if Sox4 regulates the size or function of the ISC pool. OLFM4, which is an established marker of ISCs, was detected by immunofluorescence to assess ISC numbers, which were unchanged between control and Sox4cKO intestines (Figure 2D) ³⁰. Next, we FACS-isolated CD44− and CD44+ cells from control and Sox4cKO mice and conducted single cell organoid forming assays to test ISC function ³¹. As expected, CD44− cells failed to form organoids regardless of Sox4 status (Figure 2E). CD44+ cells from Sox4cKO mice produced approximately 50% fewer organoids than controls, demonstrating a significant loss of stemness (Figure 2E).

To determine if this functional deficit was coincident with an increase in apoptosis in Sox4cKO samples, we examined cleaved-caspase-3 expression in the crypts of control and Sox4cKO intestines, as well as the number of apoptotic cells in CD44− and CD44+ populations of cells prepared for FACS. In both cases, we found no significant difference in apoptosis between control and Sox4cKO samples (Supplementary Figure 3A, B). However, CD44+ cells from Sox4cKO mice expressed significantly lower levels of Lgr5, consistent with decreased stem cell function (Figure 2F). Consistent with immunofluorescence results, Olfm4 mRNA levels were unchanged between control and Sox4cKO samples (Figure 2F). Together, these data indicate that Sox4 does not regulate ISC numbers or apoptosis, but is required for normal ISC function and Lgr5 expression in the intestinal crypt.

Sox4 regulates secretory lineage allocation

We next sought to determine if Sox4 regulates differentiation and intestinal lineage allocation. Goblet cells, detected by expression of MUC2, were found at similar numbers between Sox4cKO and control intestines (Figure 3A). CHGA⁺ EECs were found in significantly reduced numbers in the villi of Sox4cKO mice (Figure 3B). Numbers of DCLK1⁺ tuft cells were significantly reduced in crypt and villus compartments of Sox4cKO intestines (Figure 3C). Paneth cells were increased in number in Sox4cKO intestines (Figure 3D), and located at higher positions in Sox4cKO crypts, but determined to be non-proliferative by co-staining with KI67 (Supplementary Figure 4A and 4B). Staining patterns for sucrose isomaltase (SIS), a brush-border enzyme associated with absorptive enterocytes, were unchanged between control and Sox4cKO samples (Supplementary Figure 5A). Additionally, villus height was unchanged, and no change was observed for Hes1, which specifies absorptive enterocyte fate (Supplementary Figure 2C and Supplementary Figure 5B).

Open in a separate window

Figure 3.

Sox4 is required for proper secretory lineage allocation.

(A) MUC2+ goblet cells remain unchanged in Sox4cKO intestines, while (B) CHGA+ EECs are found at decreased numbers in Sox4cKO villi, and (C) DCLK1+ tuft cells are decreased in both crypt and villus compartments of Sox4cKO intestines (scale bar represents 100um). (D) Paneth cell numbers are increased in Sox4cKO crypts (scale bar represents 50um). (E) Assessment of EEC differentiation by immunofluorescence against markers associated with specific EEC subtypes reveals that Sox4 is required for proper specification of a subset of EEC lineages. Conversely, numbers of cells positive for serotonin (5-HT) and Substance P (SUB-P) are unaffected in Sox4cKO intestines. (asterisks indicate significance; p < 0.05). (F) The top 20 most significantly downregulated, GO-identified terms for biological processes in RNA-seq of Sox4cKO vs. control CD44+ populations are consistent with EEC function.

In order to examine gene regulatory changes in Sox4cKO samples in an unbiased manner, we subjected FACS isolated CD44^- and CD44⁺ populations to RNA-seq. Consistent with histological observations, decreased expression of genes associated with EEC subpopulations, including pro-EEC transcription factors Isl1, Nkx2.2, and Pax6, was the predominant change observed in Sox4cKO CD44⁺ populations (Supplementary Figure 1D and Supplementary Table 1) ^{32, 33}. There were no significant gene expression changes in the CD44^- populations consistent with the restricted expression pattern of SOX4 in CD44⁺ intestinal crypts and CD44+ status of EECs (Supplementary Figure 1D). Gene set enrichment analysis (GSEA) revealed a decreased gene expression signature consistent with pancreatic beta cells, another well-characterized endocrine population (Supplementary Figure 1E and Supplementary Table 2). MAPK signaling, which was recently shown to inhibit EEC and tuft lineages, was enriched, as was IL-6 JAK/STAT signaling, which is associated with increased epithelial proliferation and may reflect increased Paneth cell numbers through Paneth cell-specific expression of IL-6R (Supplementary Figure 1E) ^34–36. Collectively, transcriptomic data support the phenotypic changes in lineage allocation observed in Sox4cKO mice, and indicate that these changes occur in CD44+ cells.

Since the EEC lineage is comprised of a diverse complement of subpopulations, we re-examined Sox4cKO intestines for the presence of hormones produced by specific EEC subtypes. Compared to the modest reduction in CHGA⁺ cell numbers, we found substantial and significant decreases in EEC subpopulations expressing GIP, CCK, SST, GCG, PYY, and GLP-2 (Figure 3E). Interestingly, numbers of 5-HT and SUBP-expressing populations remained unchanged (Figure 3E). Gene ontology (GO) analysis of significantly downregulated genes in Sox4cKO CD44+ cells was consistent with a loss of EEC populations, and yielded terms for biological processes involving hormone regulation, signaling, and nutrient homeostasis (Figure 3F).

Since Notch repression is required for EEC specification, we asked if Sox4 is activated following downregulation of Notch. To do so, we treated wild type organoids with the gamma secretase inhibitor DAPT. Since Sox4 has been reported to be regulated by Wnt in colon cancer cell lines and inhibited by Bmp in hair follicle progenitors, we also treated organoids with WNT3A and Noggin, a Bmp antagonist ^{37, 38}. While Sox4 levels and cell numbers did not respond to Wnt activation or Bmp inhibition, we found that Notch inhibition drove a 2-fold increase in Sox4 expression that coincided with decreased expression of Ascl2, and ISC marker, and increased expression of Ngn3, an early EEC transcription factor (Supplementary Figure 6). Together, these data demonstrate that Sox4 is required for proper secretory lineage allocation. Furthermore, Sox4 is upregulated following Notch inhibition, consistent with early secretory differentiation responses, and required for proper specification of EEC subpopulations.

Sox4 contributes to tuft cell differentiation and hyperplasia

Since a subset of SOX4^high cells were found to co-localize with DCLK1 and DCLK1⁺ tuft cells were decreased in Sox4cKO intestines, we sought to examine the regulatory role of Sox4 in tuft cell differentiation more closely. IL-13 is known to induce tuft cell differentiation, and is produced by Group 2 innate lymphoid cells (ILC2) that are recruited to the intestinal epithelium in response to tuft cell secreted IL-25. This mechanism produces an extrinsic feed-forward loop that signals through unknown intrinsic response elements in ISCs and progenitors to drive tuft cell hyperplasia following parasitic helminth infection ^{2, 5}. We treated wild type intestinal organoids with IL-13 to determine its effect on Sox4. As previously reported, IL-13 induced expression of tuft cell-specific Dclk1 (Supplementary Figure 7A). Compellingly, IL-13 also induced a dose-dependent upregulation of Sox4, driving an approximately 3-fold increase at the optimal dose (Supplementary Figure 7A). Levels of Sox4 upregulation correlated closely with Dclk1 induction. As expected, goblet cell Muc2 was the only non-tuft lineage specific gene also upregulated in response to IL-13 (Supplementary Figure 7B) ². To determine if Sox4 is required for proper Dclk1 upregulation in response to IL-13, we next treated Sox4cKO and control organoids with IL-13. While control organoids exhibited robust induction of Dclk1 and the tuft cell transcription factor Pou2f3, Sox4cKO organoids failed to significantly upregulate these genes (Figure 4A).

Open in a separate window

Figure 4.

Sox4 regulates tuft cell differentiation and is required for tuft cell hyperplasia following helminth infection.

(A) Tuft cell markers Dclk1 and Pou2f3 are upregulated in control, but not Sox4cKO organoids, in response to treatment with recombinant IL-13 (different letters indicate statistically significant differences, p < 0.05). (B) 7 days post-infection with N. brasiliensis, control intestinal crypts exhibit a significant expansion of SOX4^high cells (scale bar represents 20um). (C) Control mice demonstrate expected tuft cell hyperplasia at 7dpi, where Sox4cKO intestines have significantly fewer tuft cells as quantified by DCLK1, COX1, and COX2 (ND = no data; COX1+ cells are not found in intestinal crypts. Scale bar represents 100um). (D) Sox4cKO mice also fail to fully clear adult worms by 7dpi (asterisks indicate significance, p < 0.05, n=3 mice per group).

Next, we asked if Sox4-driven tuft cell responses were physiologically necessary for clearance of helminth infection. Sox4cKO and control animals were infected with N. brasiliensis, a helminth known to induce tuft cell-mediated type II immune reactions ^{2, 4, 5}. Intestines were examined 7 days post-infection (d.p.i.), at the peak of tuft cell hyperplasia, when most worms are typically cleared from wild type mice ⁵. We first examined SOX4⁺ cells in control mice. Helminth infection drove a 4-fold increase in the number of SOX4⁺ cells per crypt (Figure 4B). As expected, DCLK1⁺ cell numbers were largely upregulated in control animals (Figure 4C). In contrast, Sox4cKO intestines exhibited an attenuated increase in tuft cell numbers (Figure 4C). To determine if impaired tuft cell hyperplasia in Sox4cKO animals affected worm burden at 7d.p.i., we quantified the number of adult worms present in the intestines of infected animals. In order to match worm clearance data with tuft cell hyperplasia in the same mouse, duodenal and ileal tissues were harvested to establish worm counts and jejunal tissues were used for histological studies, discussed above. Control animals exhibited a near-total clearance of adult worms, while Sox4cKO animals bore a significantly higher worm burden at 7d.p.i. (Figure 4D). These data implicate Sox4 as a regulator of tuft cell differentiation at physiological baseline, with an important role in tuft cell hyperplasia during helminth infection.

Sox4− and Atoh1-expressing cells form overlapping and distinct populations

Early tuft cell specification remains a paradox in intestinal epithelial differentiation. Atoh1 is accepted as the master regulator of early secretory fate in the intestine, and is de-repressed following loss of Notch signaling ³⁹. While tuft cells fail to differentiate properly in the presence of constitutively active Notch, multiple lines of evidence demonstrate that a majority of tuft cells are not derived from Atoh1-expressing progenitors ^{3, 6}. These data suggest the existence of Notch-repressed, Atoh1-independent tuft cell regulatory pathways that remain uncharacterized. Since our data demonstrate that Sox4 is upregulated following Notch inhibition and is involved in tuft cell differentiation, we hypothesized that it might be expressed in an Atoh1-independent manner. We examined SOX4 expression in Atoh1^GFP intestines, and found that 55.3 ± 6.1% of SOX4^high cells were Atoh1-negative, with the remainder co-localizing with Atoh1-GFP (Figure 5A).

Open in a separate window

Figure 5.

Atoh1+ and Sox4+ exhibit overlapping and exclusive expression in secretory cells and ISCs.

(A) Co-detection of SOX4 and Atoh1, using a transgenic Atoh1^GFP allele, reveals that 55.3 ± 6.1% of SOX4+ cells co-express Atoh1 (yellow arrowhead indicates SOX4+/Atoh1+ cell; white arrowhead indicates SOX4+/Atoh1− cell. Scale bar represents 25um). (B) scRNA-seq ⁴⁰ identifies unique cell populations in the intestinal epithelium (C) Classifying cells based on Sox4 and Atoh1 expression status (Sox4+:Atoh1-, Sox4-:Atoh1+, and Sox4+:Atoh1+) reveals that Sox4+ populations contain the highest proportion of tuft cells, EECs, and ISCs, consistent with observed phenotypes in vivo and in vitro.

To examine Sox4 and Atoh1 populations more thoroughly, we reanalyzed previously published single cell RNA-seq (scRNA-seq) data from primary mouse intestinal epithelial cells ⁴⁰. Cells were visualized by tSNE dimensionality reduction and classified into previously characterized phenotypic populations (Figure 5B) ⁴⁰. Next, we queried the distribution of cell types within each of three populations of interest: (1) Sox4+:Atoh1− (n = 617 cells), (2) Sox4+:Atoh1+ (n = 88 cells), and (3) Sox4-:Atoh1+ (n = 162 cells). Within the Sox4+ population, 12.5% of cells were also Atoh1+. Next, we determined the cellular composition of each of our populations of interest. The Sox4+:Atoh1− population consisted primarily of ISCs and progenitor cells (88%), but also contained the highest proportion of tuft cells relative to other populations (Figure 5C). The Sox4+:Atoh1+ population was a mix of ISCs and secretory cell types, and contained the highest proportion of EECs relative to other populations (Figure 5C). As expected, the Sox4-:Atoh1+ population was comprised mostly of secretory cells, with the lowest proportion of ISCs and progenitor cells and the highest proportion of goblet and Paneth cells relative to other populations (Figure 5C). To ask the inverse question, we next binned scRNA-seq data by cell type and queried the percent distribution of Sox4 and Atoh1 within defined intestinal epithelial cell subpopulations. This analysis indicates that 46.1% of tuft cells are Sox4+:Atoh1− and 48.0% are negative for both Sox4 and Atoh1, while Sox4-:Atoh1+ and Sox4+:Atoh1+ populations represent 2.9% of total tuft cell numbers, each (Supplementary Figure 8). These data demonstrate that Sox4 and Atoh1 form distinct and overlapping populations, and that Sox4 is most strongly correlated with ISCs, progenitors, and tuft cells when expressed exclusively of Atoh1 and more strongly correlated with EECs when co-expressed with Atoh1.

Sox4 drives tuft cell differentiation independently of Atoh1

We next sought to determine if Sox4 is sufficient to drive secretory EEC and tuft cell differentiation. To this end, we generated stable transgenic organoids carrying a cumate-inducible Sox4 overexpression vector (Sox4OE). Gain of function was validated by immunofluorescent detection of SOX4 in empty vector control and induced Sox4OE organoids (Supplementary Figure 9). To validate organoid models for assessment of Sox4-dependent differentiation phenotypes, we included organoids isolated from Sox4cKO mice in our analyses. As expected, Sox4cKO organoids exhibited very low numbers of DCLK1+ tuft cells and CHGA+ EECs (Figure 6A, B). In contrast, overexpression of Sox4 resulted in a significant increase in numbers of both tuft cells and EECs, relative to empty vector controls (Fig 6A, B).

Open in a separate window

Figure 6.

Sox4 drives tuft cell and EEC differentiation independently of Atoh1.

(A) Tuft cells, as identified by DCLK1, are rare in control organoids, absent in Sox4cKO organoids, and upregulated in Sox4OE organoids. (B) Similar results are observed for CHGA+ EECs, indicating that Sox4 is sufficient to increase tuft and EEC differentiation (scale bars represent 50um; different letters indicate statistically significant differences, p < 0.05, n=25 organoids per group). (C) 4-OHT treatment downregulates Atoh1 in control and Sox4OE organoids, resulting in (D) loss of Muc2 expression, which demonstrates loss of Atoh1-driven transcription. (E) Loss of Atoh1 in organoids also results in downregulation of Dclk1, which is rescued to normal levels in Sox4OE organoids, regardless of Atoh1 expression. (F) Chga is not affected by Atoh1 deletion in organoids, but is upregulated by Sox4OE in the absence of Atoh1 (different letters indicate statistically significant differences, p < 0.05, n=3 different wells of organoids per group). Cell counts of (G) DCLK1+ tuft cells and (H) CHGA+ EECs demonstrate correlation between changes in gene expression and lineage specification in Atoh1iKO and Sox4OE organoids. Notably, CHGA+ cells numbers are increased in Sox4OE organoids both with and without Atoh1 expression (scale bars represent 50um; different letters indicate statistically significant differences, p < 0.05, n=25 organoids per group).

Since scRNA-seq demonstrated differential distribution of Sox4 and Atoh1 across progenitors and secretory cell types, we hypothesized that Sox4 may function independently of Atoh1 in secretory differentiation. To test this, we generated Sox4OE organoids derived from Atoh1^fl/fl:vilCre^ER mice. Atoh1 recombination was induced by treating organoids with 1uM 4-OHT once a day for three days and collecting for analysis 24hr following final treatment with 4-OHT. Analysis of Atoh1 expression confirmed the effectiveness of this strategy in generating Atoh1iKO organoids (Figure 6C). To assess whether residual ATOH1 protein was exerting continued effects in Atoh1iKO organoids, we next analyzed Muc2 expression, which is associated with Atoh1-dependent goblet cell differentiation. We found that Muc2 levels were also significantly downregulated in Atoh1iKO organoids, suggesting no residual ATOH1 function at the time of organoid analysis (Figure 6D).

Next, we examined the effect of Atoh1 and Sox4 on tuft and EEC differentiation. Surprisingly, Dclk1 was significantly downregulated following loss of Atoh1, in direct conflict with reported in vivo data demonstrating an increase in tuft cells in Atoh1-null intestines (Figure 6E) ¹⁰. However, treatment of Atoh1iKO organoid cultures with IL13 restored Dclk1 expression to control levels, in contrast to the loss of IL13 responsiveness in Sox4cKO organoids (Supplementary Figure 10). Overexpression of Sox4 was also sufficient to restore Dclk1 expression levels in the absence of Atoh1 (Figure 6E). Unexpectedly, Chga expression levels remained unchanged following loss of Atoh1, and were upregulated by Sox4 in Atoh1iKO organoids (Figure 6F). Notably, Sox4OE failed to rescue or upregulate Muc2 expression levels, suggesting specificity to tuft and EEC lineages (Figure 6D). To refine these observations, we conducted whole mount staining on Atoh1iKO and Sox4OE organoids and directly quantified tuft and EEC numbers in whole organoids. Quantification by immunofluorescence recapitulated gene expression results, demonstrating upregulation of DCLK1+ tuft cells and CHGA+ EECs in Sox4OE organoids regardless of Atoh1 expression (Figure 6G and H). Together, these data demonstrate that Sox4 is sufficient to drive an increase in tuft and EEC differentiation and can do so independently of Atoh1.

The authors would like to thank Drs. Rashmi Chandra and Roger Liddle (Duke University, Durham, NC) for providing antibodies against CCK and PYY; Dr. Veronique Lefebvre for providing Sox4^fl mice; Brian Golitz and Noah Sciaky for assistance with Illumina sequencing and analysis; Dr. Pablo Ariel and the UNC Microscopy Services Laboratory for assistance with confocal microscopy; Alexei Kouminov for technical support. RNA-seq data associated with this manuscript have been deposited in NCBI’s Gene Expression Omnibus and are available under GEO accession number GSE90795. This work was funded by the National Institutes of Health, R01 DK091427 (Magness), the Center for Gastrointestinal Biology and Disease P30 DK034987 (Magness), National Institutes of Health, R01 AI119004 (Reinhardt), and National Institutes of Health, R01 CA1428260, R01 DK092306 (Shroyer). A.D.G. was partially supported by the UNC GI Division Basic Science Training Grant, T32 DK007737–20 (Sartor), National Institutes of Health, K01 DK111709 (Gracz), and a Research Scholar Award from the American Gastroenterological Association (Gracz). Y-H.L. was partially supported by National Institutes of Healthy, F99 CA212433.