Bmi1-Cre^ER;Rosa26^eYFP (Bmi1^Ctrl) mice and Bmi1-Cre^ER; Rosa26^eYFP;Klf4^fl/fl (Bmi1^ΔKlf4) mice were used as described previously (10). Three to four months old mice were used in this study. Tamoxifen was dissolved in a corn oil (37.5 mg/ml) and administrated by single intraperitoneal injection (9 mg per 40 g of body weight). All studies and procedures involving animal subjects were approved by the Stony Brook University Institutional Animal Care and Use Committee and conducted strictly in accordance with the approved animal handling protocol.

Proximal parts of small intestines were swiss-rolled as described previously (10), and paraffin embedded blocks were cut into 5 μm thick sections. Tissues were de-paraffinized in xylen, rehydrated in ethanol gradient and incubated in a 10 mM Na-citrate buffer (pH 6.0) at 120℃ for 10 min in a pressure cooker in order to retrieve antigen. Sections were then washed with water, incubated for 1h at 37℃ in a blocking solution (5% Bovine serum albumin and 0.01% Tween 20 in 1x Tris-based phosphate-buffered saline [PBS]), and incubated with primary antibodies for GFP (AvesLabs), KLF4 (R&D Systems), MUC2 (GeneTex), chromogranin A (Abcam), Lysozyme (Leica biosystems). EdU was administered 3 h prior to euthanasia. EdU labelled cells were stained using Click-IT EdU imaging kit (Thermo Fisher Scientific) according to manufacturer’s instruction. Tissues were also counterstained with Hoechst 33258 to visualize nuclei.

Proximal small intestine was harvested from mice at 48 h after tamoxifen injection. Intestinal epithelial cells were dissociated as previously described (14). EYFP⁺ cells were sorted by flow cytometry (BD FACSARIA III; BD Biosciences, San Jose, CA) from Bmi1^Ctrl and Bmi1^ΔKlf4 mice. Sorted eYFP⁺ cells were embedded in Matrigel (Corning) and dispensed into 24-well plates as 20∼30 μl droplets. Organoid culture medium was prepared using L-WRN cells as previously described (15) and supplemented with 1x N2 suplement (Theromo Fisher Scientific), 1x B27 supplement (Thermo Fisher Scientific), gastrin I (10 nM) (Sigma-Aldrich), recombinant human epidermal growth factor (50 ng/ml) (Thermo Fisher Scientific), trans-forming growth factor β inhibitor A83-01 (500 nM) (Tocris Bioscience, Bristol, United Kingdom), N-acetylcysteine (1 mM) (Sigma-Aldrich) and antibiotic cocktail Primocin (100 μg/ml) (Thermo Fisher Scientific). GSK3β inhibitor CHIR99021 (10 μM) (Tocris) and ROCK inhibitor Y-27632 (10 μM) (Sigma-Aldrich) were also added during the first 2 days of the culture. The media were changed every 2 days. Live organoids were imaged using inverted microscope (Eclipse Ti2, Nikon).

Cultured organoids were washed with PBS. Matrigel was dissolved with Cell Recovery Solution (Corning) and collected organoids were resuspended in iPGell (GenoStaff, Tokyo, Japan). The iPGell-embedded organoids were fixed in 4% paraformaldehyde and embedded in paraffin.

Statistical analysis was carried out using GraphPad Prism version 8.3.0 for Windows (GraphPad). Student t test, Dunn’s multiple comparisons test with Kruskal-Wallis test, Spearman correlation were used. Values of p<0.05 were considered significant. All data are shown as Mean± SEM.

As previously reported, KLF4 is expressed mainly in the terminally differentiated epithelial cells lining the intestinal villi with the highest levels of expression within the proximal intestinal epithelium and also in a small population of the cells within the intestinal crypts of Bmi1^Ctrl mice (Fig. 1A) (12, 16). BMI1⁺ ISCs exist predominantly in the crypts of the proximal intestinal epithe-lium. Therefore, in this study, we focused on the KLF4 function in BMI1⁺ ISC lineage within this region. Immunostaining revealed that the goblet cell marker, MUC2 is expressed in 74.1±4.0% of KLF4⁺ cells in crypts (Fig. 1B and 1C). We also performed immunofluo-rescent staining (IF) of KLF4 and chromogranin A (an enteroendocrine cell marker) and lysozyme (a Paneth cell marker) (Fig. 1D and 1E). Co-expression rates of KLF4 in MUC2⁺, chromogranin A⁺ and lysozyme⁺ cells are 79.0±2.0%, 3.3±1.7% and 0.7±0.7%, respectively (Fig. 1F). Thus, KLF4 is mainly expressed in goblet cells in the crypt of small intestine.

Next, we investigated the effect of Klf4 deletion in BMI1⁺ ISC population on goblet cells differentiation. Bmi1^Ctrl and Bmi1^ΔKlf4 mice were injected with tamoxifen and the proximal small intestinal tissues were collected on days 2, 3, and 7 post-treatment and stained for eYFP (BMI1⁺ cells and their lineage), KLF4 and MUC2 (Fig. 2A). The number of eYFP⁺ cells increased in both Bmi1^Ctrl and Bmi1^ΔKlf4 mice after tamoxifen injection (Bmi1^Ctrl vs. Bmi1^ΔKlf4, 2.0±0.1 vs. 2.5±0.2, 4.0±0.2 vs. 6.3±0.2, 9.0±0.6 vs. 9.4±0.6 at days 2, 3, and 7, respectively) (Fig. 2B). At 3 days after tamoxifen injection, the number of eYFP⁺ cells per eYFP crypt in Bmi1^ΔKlf4 mice was significantly higher than Bmi1^Ctrl mice (6.3±0.2 and 4.0±0.2, respectively; p<0.001) (Fig. 2A and 2B). Importantly, the number of KLF4⁺ cells in Bmi1-eYFP⁺ cells in Bmi1^ΔKlf4 mice was significantly lower than Bmi1^Ctrl mice on days 3 and 7 after tamoxifen injection (Fig. 2C). The number of goblet cells marked by MUC2 stain per eYFP⁺ crypt in Bmi1^Ctrl mice was significantly higher than in Bmi1^ΔKlf4 mice on day 7 (Fig. 2D). Thus, goblet cell differentiation was significantly inhibited in BMI1⁺ ISC derived lineage after Klf4 deletion in BMI1⁺ ISC during homeostasis. Deletion of Klf4 within BMI1⁺ cell population did not affect differentiation of other secretory lineages marked by chromogranin A and lysozyme (data not shown). Three days after tamoxifen administration, the percentage of EdU⁺ cells in eYFP⁺ cells in Bmi1^ΔKlf4 mice was also significantly higher than in Bmi1^Ctrl mice (46.7±2.1% and 20.7±0.9%, respectively; p<0.001) (Fig. 2E and 2F). Taken together these data suggest that Klf4 deletion in BMI1⁺ ISC leads to increased cell proliferation in BMI1⁺ ISC derived cell lineages and a reduction of goblet cell differentiation.

To confirm relationship between KLF4 expression and goblet cell differentiation, we counted BMI1-eYFP positive cells from proximal small intestinal crypts of Bmi1^Ctrl and Bmi1^ΔKlf4 mice 7 days after tamoxifen injection. In Bmi1^Ctrl mice 18.0±2.4% of eYFP⁺ cells co-expressed MUC2 while in Bmi1^ΔKlf4 mice only 9.2±1.9% (Fig. 3A). Additionally, in Bmi1^Ctrl mice, 60∼70% of MUC2 positive cells co-expressed KLF4. The Spearman correlations analysis showed that KLF4 and MUC2 are significantly positively correlated (Spearman correlation=0.9429, p<0.01) (Fig. 3B).

To determine the effect of Klf4 deletion from BMI1⁺ cells on tissue regenerative process, we used organoids derived from fluorescence-activated cell sorting-isolated eYFP⁺ cells. Cultured organoids obtained from Bmi1^Ctrl and Bmi1^ΔKlf4 mice showed budding formation around culture day 5 (Fig. 4A and 4B) with no significant difference between these two groups. However, on day 7 organoids obtained from Bmi1^ΔKlf4 mice showed increased in the proliferation in comparison to organoids from Bmi1^Ctrl mice. This is in accordance with our previous in vivo observation that demonstrated that Klf4 deletion in BMI1⁺ cells led to their increased proliferation (10). Immunostai-ning of cultured organoids demonstrated that MUC2⁺ cells are found in organoids derived from Bmi1^Ctrl mice, and 73.9% of MUC2⁺ cells co-expressed KLF4 (Fig. 4C). Organoids derived from Bmi1^ΔKlf4 showed decreased number of MUC2⁺ and KLF4⁺ cells (Bmi1^Ctrl vs. Bmi1^ΔKlf4, 11.2±1.1% vs. 1.4±0.5% and 13.4±1.6% vs 2.1±0.8% for KLF4⁺ and MUC2⁺ cells, respectively) (Fig. 4D and 4E). Thus, these data demonstrate that KLF4 regulates goblet cell differentiation from BMI1⁺ ISC.

In this study, we have shown that goblet cell differentiation is regulated by KLF4 in BMI1⁺ ISC lineage during homeostasis. Initially, BMI1⁺ ISC was identified as quiescent and radiation-resistant cell located around +4 position in the proximal small intestinal crypt (4, 6). Afterward, several studies observed controversial results that Bmi1 was expressed throughout the crypt using mRNA in situ hybridization and BMI1 expressing cells were located in crypt cells, including goblet, enteroendocrine, Paneth and Lgr5⁺ cells (8, 17, 18). These data and our results suggest that BMI1⁺ ISCs are partially responsible for the intestinal homeostasis not only in injury state but also under homeostasic condition.

Goblet cells are the most abundant intestinal secretory lineage, comprising ∼15% of the small intestinal epithelial cells (2, 19). Although turnover time of MUC2⁺ goblet cells is slower in the crypt compared to goblet cells along the villi, newly produced MUC2 expressing goblet cell will disappear within 24 h (2, 19). In this study, MUC2 positive cells were detected in BMI1⁺ cell lineage at 7 days after tamoxifen injection. Thus, BMI1⁺ cells act as not only rISC but also active intestinal stem cells (aISC) with self-renewal and multipotency during homeo-stasis. Previous studies indicated that KLF4 had a critical role in the development of goblet cells and thefunction was regulated by the Notch pathway (11, 12, 20, 21). Another study suggested that KLF4 cooperated with Hippo signaling effectors, YAP and TAZ in promoting differentiation into goblet cells (13). In these studies, the effect of Klf4 deletion was included in both differentiated cells and ISC. Here, we have investigated the direct effect of BMI1⁺ ISC specific Klf4 deletion using lineage tracing mouse model from Bmi1 locus and ex vivo organoid model.

In conclusion, we have shown that KLF4 regulates goblet cell differentiation in BMI1⁺ intestinal stem cell lineage during homeostasis.