Characteristics of Colon Crypt Stem Cells in Preserved Epithelial Cell Clumps in Dextran Sulfate Sodium-induced Mouse Model of Ulcerative Colitis

Crypt stem cells rescue the colorectum from refractory ulcers in ulcerative colitis (UC). We previously reported that clumps of a few epithelial cells were scattered in ulcer regions in a dextran sulfate sodium (DSS)-induced mouse model of UC; however, the origin of the clumps is unknown. We determined the immunohistochemical expression of stem-cell markers in the epithelial clumps in female Balb/c mice administered DSS in drinking water for 6 days, followed by withdrawal of DSS for 6 days. Similar to the characteristics of UC, the ulcers were more severe in the distal region close to the anus than in the proximal region of the colorectum. Quantitative analyses revealed that the epithelial clumps appeared in relation to the severity of ulcer, and they expressed the cell adhesion molecules E-cadherin and β-catenin. Among stem cell markers, the epithelial clumps primarily expressed + 5 cell marker Dll1, followed by + 4 cell marker Bmi1 and crypt stem cell marker Lgr5 in that order. Nuclear expression of Sox9, but not Ki-67 and β-catenin was identied in the clumps. The present results suggest that the epithelial clumps comprised crypt-derived stem cells with the potential to regenerate crypts, which could serve as a potential treatment strategy for UC.


Introduction
Epidemiological analyses raise concerns regarding the increased incidence of patients with in ammatory bowel disease (IBD) worldwide; the incidence is estimated to have increased more than 2-fold in selected industrialized countries of the Western world between 2010 and 2030, thus representing global IBD burden 1 . Ulcerative colitis (UC) is an IBD, characterized by in ammation and ulceration of the mucosa limited to the colorectum in patients with repeated relapse and remission; it leads to adverse effects on the quality of life of the patients and the risk of colon cancer 2,3 . The disease activity of UC is assessed by clinical, biochemical, endoscopic, or histologic analyses, and remission at each level has been discussed to achieve therapeutic goals, that is, mucosal healing 4 . Histological remission has recently been recognized as an ultimate indicator of mucosal healing in UC [5][6][7] . To accomplish this, intestinal crypt stem cells play a central role in crypt regeneration, which is anticipated to be a potential therapeutic strategy for patients with UC 8 .
An animal model using dextran sulfate sodium (DSS)-treated mice have been widely accepted for analyzing ulcers and subsequent regeneration in a time-dependent manner and for determining histological healing 9 . Initial mucosal damage may be immediately rescued by restitution, which is the migration of monolayered epithelial cells from the surrounding mucosa reserved at the ulcer edge to cover the denuded area 10 . We and other researchers reported a monolayer of restituted epithelium covering the surface of ulcer areas during the recovery period after DSS administration in mice [11][12][13][14] . Together with the observation of restitution, we reported that small cell clumps remained under the lamina propria after induction of ulcers in DSS-treated mice 11,12 . The epithelial clumps appeared alongside or before the period of restitution. These clumps consisted of a few epithelial cells; immunohistochemical studies demonstrated that they expressed cytokeratin AE1/AE3 and lacked the mesenchymal marker vimentin; cell proliferation markers proliferating cell nuclear antigen (PCNA) and 5bromo-2'-deoxyuridine (BrdU); and mucin, as indicated by PAS reaction 11 . However, the origin of the epithelial clumps is unknown.
Mucosal healing is regulated by stem cell lineages in the crypt of the small and large intestines: activecycling/proliferating intestinal stem cells (aISC) or crypt base columnar (CBC) cells and quiescent/slowcycling reserve ISC (rISC) or label-retaining cells (LRC) 15 . aISC/CBC cells expressing leucine-rich repeatcontaining G-protein coupled receptor 5 (Lgr5) are essential for the maintenance of mucosal homeostasis; they reside in the crypt bottom as self-renewing stem cells generating daughter progenitors, which differentiate into absorptive and secretory lineages [16][17][18] . The cells that reside in the fourth position and the fth to fteenth positions from the crypt bottom are known as + 4 and + 5 cells, respectively, and are characterized by the expression of the progenitor cell markers B lymphoma Mo-MLV insertion region 1 homolog (Bmi1), and Notch ligands jagged1, delta-like1 (Dll1), respectively. These cells are categorized as rISC; they can differentiate into functional absorptive and secretory epithelial cells and revert to stem cell lineages when aISC is lost or injured 19,20 . Sex-determining region Y (SRY)-box transcription factor 9 (SOX9) is a negative regulator of the Wnt-beta-catenin pathway, which regulates stemness in aISC and rISC 21,22 . In this study, we hypothesized that the strayed epithelial clumps might be derived from in ammation-resistant stem cells or their daughter progenitor cells, which have the potential to regenerate crypts and mucosal healing in refractory ulcers of the colorectum in a DSS-induced UC model. To test this hypothesis, we conducted morphometric and immunohistochemical studies of the epithelial clumps in the proximal regions with slight-to-moderate ulcers and in the distal regions close to the anus with more severe ulcers in the colorectum and elucidated whether the epithelial clumps maintain the characteristics of Lgr5-expressing crypt stem cells or that of Bmi1-, Dlli1-, or Sox9-expressing daughter progenitor cells.

Animals
Four-week-old female BALB/cAnNCrlCrlj mice (n = 34, Charles River Laboratories Japan, Inc., Kanagawa, Japan) were acclimated for 9 days to the testing environment under controlled temperature (23 ± 3 °C), humidity (50 ± 20%), and light (12:12-h light-dark cycle) conditions. Three to four mice were maintained in an individual plastic cage with enrichment and safe harbor and provided free access to a basal diet (Oriental MF, Oriental Yeast, Tokyo, Japan) and tap water. Food consumption and water intake were measured twice a week per cage, and body weight was measured daily during the study. All procedures in this study were conducted in accordance with the Guidelines for Proper Conduct of Animal Experiments (Science Council of Japan, June 1, 2006), and the protocols were approved by the Animal Care and Use Committee of the Tokyo University of Agriculture and Technology (Fuchu, Tokyo, Japan). The study was carried out in compliance with the ARRIVE guidelines.

Experiment design of DSS-induced colitis
Animals were divided into two groups according to body weight at the initiation of the study: the negative control group (n = 4) and DSS group (n = 30). Negative controls (untreated mice) received sterile distilled water. In the DSS-treated group, the mice received 5% (w/v) DSS dissolved in sterile distilled water as drinking water for 6 days, from Day 1 to Day 6, followed by withdrawal of DSS for 6 days for induction of colitis. The DSS solution was prepared twice during the administration. The DSS dose was selected on the basis of a preliminary study. The mice were anesthetized on Day 6 (control, n = 2; DSS-treated, n = 11) and Day 12 (control, n = 2; DSS-treated, n = 10) using iso urane inhalation, euthanized via exsanguination, and subsequently necropsied. The other mice showed moribundity and lethality during the study, following which they were necropsied.

Assessment of Disease Activity Index of Colitis
Each mouse was evaluated daily for diarrhea, fecal blood (Uropaper® III Eiken, Eiken Chemical Co., Inc., Tokyo, Japan), and body weight loss for the sake of being used as the disease activity index (DAI), as previously described 11,23 (Supplemental Table 1). Fecal blood was scored by cutting the collected feces in half, putting them in a PCR tube containing physiological saline for 30 seconds, and subsequently dipping a Uropaper ® III Eiken in the tube for another 1 minute. DAI was calculated by summing the scores for diarrhea, fecal blood, and body weight loss.

Tissue Preparation and Histological Evaluation of Colitis
The colorectum was removed from the large intestine and measured for length 12 . The colorectum was divided into two segments, distal and proximal regions, and xed in 4% paraformaldehyde in phosphate buffer solution (PFA). The xed colorectum was routinely embedded in para n, sectioned, and subjected to hematoxylin and eosin (H&E) staining. Histological evaluation was performed by scoring the extent of tissue injury based on in ammation, mucosal loss, and edema using H&E-stained specimens, as previously described 12 (Supplemental Table 2). Epithelial clumps at the lamina propria were observed as previously reported 11 and classi ed into seven types based on morphological characteristics. The clumps were subclassi ed into type 1 to 4, according to the number of nuclei: type 1 exhibited a single nucleus; type 2, two or three nuclei; type 3, four to nine nuclei; and type 4, 10 nuclei or more (Supplemental Fig. 1).
These clumps were additionally subclassi ed into types a and b, according to the presence of luminal structure: type a exhibited luminal structure, while type b lacked luminal structure. By combining types a and b, and types 1 to 4, there were 7 subtypes of epithelial clumps, except for type 4b. Furthermore, the number of epithelial clumps was evaluated by counting the number of clumps per millimeter of the total mucosal length on H&E sections using a vertical slide system (VS120, Olympus Corporation, Tokyo, Japan).

Double immunohistochemical staining
Double immunohistochemical staining was conducted as previously reported 24 using SOX9 and DLL1 as the primary antibodies. In brief, depara nization was performed as described above, and immunopositive reactions were visualized using the VECTASTAIN® Elite ABC kit and 3,3'diaminobenzidine. The sections were treated with 0.1 M glycine hydrochloride buffer twice for 25 minutes, followed by overnight incubation with the primary antibodies at 4 °C. An avidin-biotin-alkaline phosphatase complex method was followed using the VECTASTAIN® ABC-AP kit with an alkaline phosphatase substrate as a chromogen; Vector®Blue (Vector Laboratories Inc.), followed by using a water-soluble encapsulant.

Statistical analysis
Multiple comparison test, Tukey-Kramer test, or Steel-Dwass test was used to analyze the body weight, the length of the colorectum, and the number of epithelial cell clumps on the H&E-stained and immunostained sections. The length of the colorectum was statistically analyzed using Student's unpaired t-test or the Aspin-Welch test. Histological evaluation (in ammation, mucosal loss, and edema) was statistically analyzed using the Wilcoxon signed-rank test. All data are represented as the mean and standard deviation (SD), except for the DAI data.

DSS administration and subsequent withdrawal exacerbates disease activity index and shortens colon length
The body weight of the DSS-treated mice began to decrease on and after Day 6 and continued to decrease until the end of the study (Fig. 1A). The scores for diarrhea and fecal blood in the DSS-treated mice began to increase on and after Day 3 (Fig. 1B, C). The diarrhea score continued to increase until Day 8; the score for fecal blood peaked on Day 6 and decreased thereafter. DAI, as determined by summing the scores of these three parameters, peaked on Day 6 and decreased on and after Day 9 (Fig. 1D).
Shortening of the colon length is a reliable indicator of colitis 12 . Macroscopically, the mean colorectal lengths in the control mice were 9.3 and 11.3 cm (n = 2 each) on Day 6 and Day 12, respectively, whereas those in the DSS-treated mice were 6.4 and 7.2 cm on Day 6 (n = 11) and Day 12 (n = 10), respectively ( Fig. 2A). Clinical and macroscopic quanti cations suggested that DSS challenge induced colitis-related illness and macroscopic changes, as previously reported 12 . DSS administration and subsequent withdrawal induces more severe colitis in the distal colorectum Colitis was estimated by scoring in ammation, edema, and mucosal loss, and the scores were compared on Day 6 and Day 12 or the distal region close to the anus with the proximal region of the colorectum. In the distal region, the scores of in ammation and mucosal loss were almost the same between Day 6 and Day 12 (Fig. 2B-D), as the lesions were already severe even on Day 6. In the proximal region, the score of each parameter was signi cantly higher on Day 12 than on Day 6, as the lesions progressed in a timedependent manner. The score of each parameter was signi cantly higher in the distal region than in the proximal region on Day 6 and Day 12. In ammation and edema were sought to be better indicators than edema in the present study.

Clumps of epithelial cells appear in ulcer region in the colorectum
We quantitatively examined the epithelial clumps in the lamina propria of the ulcer site in both the distal and proximal regions of the colorectum (Fig. 3A). In the DSS group, the number of clumps increased on Day 12 in both the distal and proximal regions when compared with that on Day 6; there was a signi cant difference between the values on Day 6 and Day 12 in the proximal region (Fig. 3B). The number of clumps was signi cantly higher in the distal region than in the proximal region on Day 6; this parameter was associated with the severity of colitis.

Clumps of epithelial cells express markers of crypt stem cells and daughter progenitor cells
We rst con rmed that the epithelial cells in the normal crypts, remaining crypts, and regeneration crypts expressed the epithelial cell adhesion markers E-cadherin and β-catenin and the stem cell/daughter progenitor cell markers LGR5, BMI1, DLL1, and SOX9 (Supplemental Fig. 2). The epithelial clumps expressed these cell markers; E-cadherin and β-catenin were restricted to the cell-cell border, and LGR5 and DLL1 were observed in the cytoplasm, while SOX9 and BMI1 were detected in the nucleus (Fig. 4A-C, 5A-C). We observed that the expression of each marker varied among the epithelial clumps; therefore, we counted each clump with negative staining or positive staining at a low or high level per unit of length. Ecadherin low or β-catenin low clumps were a major population, while E-cadherin high or β-catenin high clumps were a minor population (Fig. 4D, E, G, H). The number of E-cadherin low or β-catenin low clumps were signi cantly increased in the proximal region on Day 12 compared with the values on Day 6, as the disease progressed. SOX9 low clumps were more abundant in both the distal and proximal regions on Day 6 than on Day 12, while the number of SOX9 negative clumps were signi cantly increased on Day 12 compared with that on Day 6 (Fig. 4F, I). Lgr5 low clumps were identi ed as the most minor population in both the distal and proximal regions on Day 6 and Day 12 (Fig. 5D, G). BMI1 low clumps were identi ed in both the distal and proximal regions on Day 6 and Day 12, and BMI1 high clumps were detected in the distal region on Day 12 (Fig. 5E, H). DLL1 high and DLL1 low clumps appeared to be more in number than LGR5 low or BMI1 low clumps on Day 6 and Day 12 (Fig. 5F, I). The clumps did not express the myo broblast marker αSMA, macrophage marker Iba1, or cell proliferation marker Ki-67.
To determine the coexpression of SOX and DLL1 in the clumps, we con rmed the expression of these markers in normal crypts, residual crypts, and regenerative crypts (Supplemental Fig. 3). SOX9/DLL1coexpressing clumps were evident in the distal region on Day 6 ( Fig. 6A-C). DLL1-single expressing clumps were signi cantly increased in the distal region on Day 12, compared with the SOX9/DLL1coexpressing cells. The percentage of SOX9 low DLL1 low or SOX9 high DLL1 low clumps was higher than that of SOX9 high Dll1 high or SOX9 low Dll1 high clumps (Fig. 6D).

Discussion
The present study revealed four major ndings on preserved epithelial clumps in a mouse model of UC. First, the epithelial clumps were detected in the lamina propria in the ulcer region in H&E-stained sections with easy microscopic identi cation for pathologists; they can be classi ed into four subgroups according to various morphological characteristics in terms of the number of nuclei and intracellular space in the cytoplasm or intercellular crypt-like structure (Supplemental Fig. 1). Second, most epithelial clumps expressed markers for the epithelial cell adhesion molecules E-cadherin (approximately 80%) and β-catenin (approximately 60%) (Fig. 4G, H). Third, the epithelial clumps heterogeneously expressed the aISC marker Lgr5 and rISC markers Bmi1 and Dll1; a major population of the epithelial clumps expressed Dll1 (50-80%), a part of the population expressed Bmi1 (approximately 20% or less), and a minor population expressed Lgr5 (Fig. 5G-I). Fourth, a population of epithelial clumps expressed the stem cell niche marker Sox9; however, its expression was notably reduced with disease progression (approximately 50% on Day 6 and a low percentage on Day 12) (Fig. 4I). Furthermore, the coexpression pro les of Sox9 and DLLI indicated that a major population of the epithelial clumps was Dll1-single-positive clumps, and a minor population was SOX9 low DLL1 low or SOX9 high DLL1 low clumps (Fig. 6C, D). The results of the present study suggest that the epithelial clumps were derived from the crypt base in the villous-crypt axis; the heterogenous cellular properties of the stem cells or their daughter cells in these clumps might be dependent on the degree of resistance to in ammation-mediated epithelial loss.
The intestinal epithelial crypt has a capacity of constant renewal by proliferating stem cells owing to maintenance of homeostasis 25 . Lgr5 + aISC rapidly proliferates for self-renewal and the generation of all functional epithelial lineages under normal conditions 15,26 . We detected a small number of Lgr5 + epithelial clumps in the ulcer region; these results might be supported by the evidence that Lgr5 + aISC/CBC cells in the colon crypt are particularly susceptible to mucosal damage in mice with DSSinduced colitis 27 and radiation-exposed mice 28 . The Lgr5 high Ki-67 + ISC/CBC cells in the small intestine are active with rapid cell cycling for controlling self-renewal and producing progenitor cells; in contrast, the colon Lgr5 low Ki-67 − ISC/CBC cells are quiescent with the potential to retain epithelial self-renewal over long periods of time 29,30 . The negative nuclear expression of Ki-67 in the epithelial clumps might be consistent with the presence of quiescent Lgr5 + ISC/CBC cells in the colon crypt. The epithelial clumps were classi ed as Lgr5 high or Lgr5 low ; the percentage of Lgr5 low epithelial clumps was higher than that of the Lgr5 high epithelial clumps (Fig. 5D, G). Several researchers have reported that Lgr5 low Ki-67 − ISC/CBC cells generate enteroendocrine cells, goblet cells, tuft cells, and deep crypt secretory (DCS) cells and constitute ISC/CBC cells by self-renewal 20, 31-33 . Lgr5 is directly regulated by Wnt signaling: Lgr5 is targeted to the β-catenin complex, committing to nuclear-translocated β-catenin 34 . Lgr5 high βcatenin nuclear+ cells may lead to premalignant conditions caused by the activation of Wnt signaling 35 ; however, no nuclear translocation of β-catenin was identi ed in the epithelial clumps. Thus, Lgr5 low Ki-67 − β-catenin nuclear− epithelial clumps were assumed to be quiescent for the sake of potentially regenerating the crypt and healing the mucosa over long periods of time.
Dll1 high cells are called + 5 cells and are categorized as rISC or label-retaining cells (LRC) with quiescent or slow-cycling status; however, after aISC/CBC cell are damaged, the cells can quickly differentiate into secretory lineage cells, that is, tuft cells, enteroendocrine cells, goblet cells, and Paneth cells 20 . The DCS cells in the colon express the Notch ligand Dll1 and likely serve as a niche for stem cells to regulate their homeostasis 36 . In addition, transit amplifying (TA) cells are rapidly cycling Dll1 + Ki-67 ± daughter progenitor cells; they undergo differentiation into an absorptive lineage, that is, enterocytes 33,37 .
Considering that Dll1 + epithelial clumps were the most common type of epithelial clumps according to the present study, we therefore propose that Dll1 high epithelial clumps might maintain a potential for reviving stemness, followed by mucosal healing in the ulcer region.
Sox9 belongs to the Sox family of transcription factors, and it is required in the small intestine for the differentiation of the secretory lineage and proliferation of rISC 25 . Additionally, Sox9 plays a role in negative feedback with respect to the activity of the Wnt-β-catenin pathway; the overexpression of Sox9 inhibits the Wnt-β-catenin pathway, which promotes tumorigenesis, while appropriate expression of Sox9 is required for Wnt-β-catenin pathway activity in rISC 21 . Roche et al 22 proposed that Sox9 expression levels were categorized as high or low according to intensity in ISC: Sox9 high rISC is quiescent and slowcycling under a homeostatic state; however, it can be converted to Sox9 low aISC, facilitating epithelial repair, when aISC is lost after damage. Although the role of Sox9 in the colon is not fully understood at present, it is proposed that Sox9 is committed to morphogenesis in the colon epithelium and that it plays a role in the differentiation of the secretory lineage, especially goblet cells 38,39 . Coexpression analyses of Sox9 and markers of aISC or rISC would con rm the potential of restoring the colon crypt. Although the coexpression analysis of Sox9 and Dlli has not been determined in normal and injured colon crypts to our knowledge, we speculated that SOX9 low DLL1 high clumps might have the potential to restore the crypts in ulcer sites. Bmi1 + quiescent ISC (qISC)/LRC or + 4 cells play a crucial role in facilitating regeneration in injured crypts, indicating that they compensate for the absence of Lgr5 + cells in the small intestine 28,40 ; however, they tend to be insensitive to the absence of Lgr5 + cells in the distal small intestine and colon 40 . Sox9 directly binds to the promoter of Bmi, thus regulating the gene expression of Bmi1 41 . Bmi1 is an epigenetic regulator, which restricts cell proliferation and shows a strong correlation with Sox9 expression 42 . As damage-sensitive SOX9 + clumps markedly reduced between Day 6 and Day 12, Bmi1 + clumps might exhibit a low potential for mucosal healing in the DSS-induced UC model.
Considering the above results, we propose a model for the formation of epithelial clumps in the colorectum in a DSS-induced mouse model. In ammatory stimuli impair Lgr5 + aISC/CBC cells and Bmi1 + qISC/LRC (or + 4 cells), as these cells are more sensitive than Dll1 + rISC (or TA/+ 5 cells), which are relatively resistant to in ammatory stimuli and which can survive as small clumps of epithelial cells in the ulcer region. Furthermore, the colon crypt base contains Sox9 + DCS cells, instead of Paneth cells; DCS cell-derived clumps can be preserved at an early phase of colitis and subsequently impaired by longlasting in ammation. The present study indicated that clumps of epithelial cells heterogeneously expressed markers of aISC/CBC cells and their daughter rISC in the colon crypt. The results suggested that the Dll1 + crypt-derived clumps were resistant to DSS-induced colitis; they might have retained the potential to regenerate the crypt and heal the mucosa. Transplantation of primary crypt stem cells and induced human pluripotent stem cells to the colorectum might enhance mucosal healing in patients with UC; however, this approach has some limitations, including interruption with gut micro ora and transplant rejection by the immune system 8 . Induction of regenerative crypts from epithelial clumps with stem cell potential would be an alternative experimental approach to stem cell therapy; therefore, further study is required to observe the morphological changes in epithelial clumps and the relationship between the clumps and regenerative crypts through the process of mucosal healing in the DSS-induced UC model. Figure 1 Disease activity index (DAI) in mice administered 5% DSS in drinking water for 6 days, followed by withdrawal of DSS for 6 days. (A) Body weight (g). (B) Diarrhea (score). (C) Fecal blood (score). (D) DAI (score) is calculated by summing the scores for diarrhea, fecal blood, and body weight loss. The data represent the mean values for body weight, DAI, diarrhea scores, and fecal blood scores. N=2, control group (CTL); N=30, DSS-treated group.       Coexpression distribution of SOX9 and DLL1 at high or low expression level.