Utility of a bacterial infection model to study epithelial–mesenchymal transition, mesenchymal–epithelial transition or tumorigenesis

Abstract

DCLK1 and Lgr5 have recently been identified as markers of quiescent and cycling stem cells in the small intestinal crypts, respectively. Epithelial–mesenchymal transition (EMT) is a key development program that is often activated during cancer invasion and metastasis, and also imparts a self-renewal capability to disseminating cancer cells. Utilizing the Citrobacter rodentium (CR)-induced transmissible murine colonic hyperplasia (TMCH) model, we observed a relative decrease in DCLK1 expression in the colonic crypts, with significant shift towards stromal staining at peak (12 days post infection) hyperplasia, whereas staining for Lgr5 and Msi-1 increased several fold. When hyperplasia was regressing (days 20–34), an expansion of DCLK1 +ve cells in the CR-infected crypts compared with that seen in uninfected control was recorded. Purified colonic crypt cells exhibiting epigenetic modulation of the transforming growth factor-β (TGFβ), Wnt and Notch pathways on 12 or 34 days post infection formed monolayers in vitro, and underwent trans-differentiation into fibroblast-like cells that stained positive for vimentin, fibronectin and DCLK1. These cells when trypsinized and regrown in soft agar, formed colonospheres/organoids that developed into crypt-like structures (colonoids) in Matrigel and stained positive for DCLK1. Mice exhibiting 12 or 34 days of TMCH were given azoxymethane once for 8 h (Gp1) or weekly for 3 weeks (Gp2), and subjected to crypt isolation. Crypt cells from Gp1 animals formed monolayers as well as colonospheres in soft agar and nodules/tumors in nude mice. Crypt cells isolated from Gp2 animals failed to form the monolayers, but developed into colonospheres in soft agar and nodules/tumors in nude mice. Thus, both hyperplasia and increased presence of DCLK1 +ve cells promote cellular transformation in response to a second hit. The TMCH model, therefore, provides an excellent template to study how alterations in intestinal stem cells promote trans-differentiation, crypt regeneration or colon carcinogenesis following bacterial infection.

INTRODUCTION

The epithelial cells of the mammalian colon are arranged in millions of crypts, whereas the stem cells, located at the base of the crypt, are responsible for epithelial tissue renewal.^1,2 We have recently identified DCLK1 and Lgr5 as markers of quiescent and cycling stem cells, respectively.³ Investigations of neoplastic tissues have provided evidence of self-renewing, stem-like cells within the tumors called cancer stem cells (CSCs) that facilitate de novo tumor growth.⁴ Cancer progression is manifested by the presence or absence of tumor metastasis and dissemination of cancer cells with self-renewal capability.⁵ As epithelial cells in general are nonmotile, cancers of the epithelial cells called carcinomas must undergo a process of epithelial–mesenchymal transition (EMT) at the invasive front to produce single migratory cells that lose E-cadherin expression. This is concomitant with deregulation of the Wnt pathway and a selective loss of the basement membrane.⁶ More recently, normal mammary epithelial cells are shown to adopt the CD44^high/CD24^lowexpression profile when exposed to transforming growth factor-β (TGF-β1) or when EMT-inducing transcription factors Snail or Twist are conditionally overexpressed.⁷ Together, these studies suggest that it is not unprecedented for EMT to generate mesenchymal cells with characteristics of stem-like cells.

During EMT, the epithelial cells lose their differentiated characteristics, including apical–basal polarity, and acquire mesenchymal features instead, including motility and invasiveness.⁸ The mesenchymal state is associated with the capacity of cells to migrate to distant organs and maintain stemness, allowing their subsequent differentiation into multiple cell types either during development or during initiation of metastasis.⁹ Similarly, the reverse process, called mesenchymal–epithelial transition (MET), is not only essential for normal tissue and organ development but also involved in colorectal carcinogenesis, and it seems to have an important role in colonic crypt regeneration.^10,11

Multiple signaling pathways including TGFβ, Wnt, Notch and NF-κB (nuclear factor kappa-light-chain-enhancer of activated B cells) in tandem with transcriptional factors such as Slug, Snail, Twist, Zeb1 and 2 suppress the expression of E-cadherin and induce EMT, resulting in loss of cell–cell adhesion and increased tumor progression and migration.^12–16 Recently, histone modification and chromatin remodeling are shown as the key regulators of eukaryotic transcription, and therefore are excellent targets for pathogenic infection.¹⁷ Similarly, hypoxia, which is a frequent feature of the microenvironment of infected tissues,^18,19 induces hypoxia-inducible factor-1α-mediated activation of histone deacetylase (HDAC)3, and is recently shown to be essential for EMT and metastasis.²⁰ These alterations in signaling pathways during an enteric infection can cause tissue/organ damage and can also promote the acquisition of malignant phenotype.^21–23 In the intestine, for example, a recent study demonstrated expression of EMT markers during the pathogenesis of fistulae in Crohn’s disease.²⁴ Similarly, lipopolysaccharide, the main component of the Gram-negative bacterial cell wall, induces EMT of intrahepatic biliary epithelial cells.²⁵ However, a possible link between enteric infection-induced changes in intracellular signaling and trans-differentiation of hyperproliferating colonic crypt cells to generate stem-like cells has not been established.

The enteric pathogen Citrobacter rodentium (CR) naturally infects mice using a mechanism similar to those employed by attaching and effacing (A/E) bacterial pathogens Entero-pathogenic E. coli (EPEC) and Entero-hemorrhagic E.coli (EHEC).^26,27 The genes regulating the A/E lesion formation in CR are carried on a pathogenicity island termed the locus of enterocyte effacement that encodes transcriptional regulators and structural components of a type III secretion system.²⁸ CR is an A/E pathogen that causes increased proliferation in the distal colon of adult outbred mice, without associated injury or significant histological inflammation.²⁹ In genetically susceptible strains, clinical signs such as retarded growth, diarrhea, dehydration, coat ruffling, hunched pasture and high mortality have been reported.²⁹ Utilizing the CR-induced transmissible murine colonic hyperplasia (TMCH) model, we have shown previously that hyperplasia of the colonic crypts in the outbred mice is associated with activation of the Wnt/β-catenin,^30–34 Notch^35,36 and NF-κB^37,38 pathways, which are integral to EMT as well. In the inbred mice, a dual phase following CR infection, wherein crypt hyperplasia due to activation of the MEK/ERK/NF-κB pathways, precedes acute/chronic inflammation characterized by expression of distinct cytokines/chemokines.³⁹ As pathways such as those activated during TMCH, besides TGFβ superfamily, may trigger the process of EMT, we hypothesized that CR infection-induced changes in intracellular signalingin vivo,^30–39 as well as chromatin remodeling to impose a specific transcriptional profile, may be sufficient for isolated colonic crypt cells to survive in culture and undergo EMT or MET-like changes in vitro and tumorigenesis in vivo. This hypothesis was tested in the current study.

RESULTS

Effect of CR-induced crypt hyperplasia on colonic stem cells

We have shown previously that a biphasic response of progression (days 6–12) and regression (days 20–34) of hyperplasia correlates with changes in crypt lengths in NIH:Swiss outbred mice.^30–38 Most crypt cells in the gut mucosa have a residence time of only 3–5 days in the crypt. We therefore hypothesized that sustained proliferation of the colonic crypts on day 12 and beyond ^30–38 may either be due to overproduction of stem and/or progenitor cells. We therefore set out to determine the changes in expression of stem and/or progenitor cell markers during the progression and regression phases of colonic crypt hyperplasia. When total RNA isolated from the crypts on days 6, 12 and 34 was subjected to real-time RT–PCR, DCLK1 mRNA expression on day 6, and particularly, on day 12 decreased significantly, compared with uninfected control [Figure 1a(i)]. On day 34, however, DCLK1 expression was higher in the crypts than that either on day 6 or 12 [Figure 1a(i)]. The changes in DCLK1 mRNA correlated with protein levels and paralleled Klf4 expression, which is predominantly expressed by differentiated epithelial cells (Supplementary Figure 1A). During immunostaining for DCLK1, both uninfected normal and day-6 crypts exhibited staining that were similar, whereas on day 12, relative staining for DCLK1 decreased in the crypt [Figure 1a(ii)]. Intriguingly, there was a significant shift in DCLK1 staining towards the subepithelial and submucosal regions at this time point [Figure 1a(ii)]. While the hyperplasia was regressing on days 20–34, DCLK1 +ve cells expanded in the crypt, with a gradual increase on days 20,27 and 34, compared with that seen in the uninfected control [Figure 1a(ii)]. Even though these findings are consistent with DCLK1 being a marker of quiescent stem cells,³ they also suggest that DCLK1 may not be the reservoir for rapidly proliferating cells at peak hyperplasia. We therefore investigated whether expression of Lgr5, a marker of rapidly cycling cells, and Msi-1, a marker of progenitor cells, change, following CR infection. During real-time PCR analysis, Lgr5 expression increased significantly on days 6, 12 and 20 compared with the uninfected control, followed by a decline on days 27–34 [Figure 1b(i)], and correlated with protein levels at these time points (Supplementary Figure 1A). Immunostaining for Lgr5 [Figure 1b(ii)] along with Msi-1 (Figure 1c) increased significantly between days 6 and 20 compared to uninfected control, with a gradual decline on days 27–34. These changes were specific to Citrobacter, as a CR mutant (ΔescV) lacking functional type-three secretion system that prevents CR from injecting its effector proteins into host cells failed to exhibit β-catenin, NF-κB-p65 or Notch-dependent increases in crypt hyperplasia (Supplementary Figures 1B and C). Thus, at least two distinct populations of stem cells following CR infection may differentially regulate crypt hyperplasia.

Figure 1.

Effect of CR infection on stem cell markers expression. (ai) Total RNA isolated from the distal colonic crypts of uninfected normal (N) or days 6, 12 and 34 post-CR-infected mice (D6-D34) was subjected to real-time PCR. Bar graph showing fold change in DCLK1 expression (*P<0.03 versus day 12; n = 3 independent experiments). (aii) Paraffin-embedded sections prepared from uninfected normal (N) and days 6–34 post-infected mouse distal colons were stained with antibody specific for DCLK1 and were analyzed via light microscopy. (Bar = 100 μm; n = 3 independent experiments). (bi) Real-time PCR analysis of total RNA isolated from colonic crypts of group of mice described in (ai). Bar graph showing fold change in Lgr5 expression (*P<0.05 versus uninfected control; n = 3 independent experiments). (bii and c) Paraffin-embedded sections prepared from uninfected normal (N) and days 6–34 post-infected mouse distal colons were stained with antibodies specific for Lgr5 and Msi-1, respectively (Bar = 100 μm; n = 3 independent experiments).

Evidence of CR—induced EMT

Bacterial infections, such as those associated with Helicobacter pylori, increase EMT in the gastric cancer setting and in a series of human gastric cell lines.²³ Whether a non-neoplastic, hyperproliferating colonic epithelium is also predisposed to similar fate following infection with an enteric pathogen is not known, and was investigated next. Colonic crypt cells isolated from uninfected normal mice did not form monolayers in vitro and remained as single cells in the absence of intestinotrophic factors such as R-spondin or Noggin, following a 30-day period (Figure 2a). Day-12 and -34 cells, on the other hand, not only formed monolayers but also exhibited significant trans-differentiation into fibroblast-like mesenchymal cells within 6–12 days post plating (Figure 2b). These changes were not driven by contamination from the mesenchymal cells during the isolation process, as the purified crypts (Supplementary Figures 1D and E) were negative for α-smooth muscle actin expression either in uninfected or in 6–34-day post-infected crypt cellular extracts (Supplementary Figure 1F), whereas cells isolated from uninfected mouse distal colon not only failed to grow beyond 1–2 weeks in culture conditions that included D-valine⁴⁰ or citrulline,⁴¹ but never changed their phenotype (data not shown). When the cells in monolayers were stained for markers of EMT, they stained positive for vimentin but negative for E-cadherin, suggesting an EMT-like process in vitro (Figure 2d). These changes were also confirmed via western blotting (Supplementary Figure 1G). Interestingly, as depicted in Figure 2d, the cells also stained positive for both DCLK1 and fibronectin, further suggesting a stem cell-driven EMT-like phenomenon. The phenotypic changes in day-12 crypts were apparently driven by CR-induced increases in: (i) β-catenin and its downstream targets Jagged-1 and cyclinD1, with a concomitant decrease in E-cadherin [Figure 2e(i), e(ii)]; (ii) Notch intracellular domain and Hes-1, with concomitant decrease in Atoh-1 [Figure 2f(i), f(ii)] and (iii) TGFβ along with increases in Slug and Snail, the repressors of E-cadherin [Figure 2g(i), g(ii)]. To validate CR’s role in promoting EMT-like processes in vitro, we utilized Young Adult Mouse Colon (YAMC) cells that are immortalized, but remain non-clonogenic in soft agar and nontumorigenic in nude mice.⁴² YAMC cells in control medium displayed cuboidal morphology and had robust E-cadherin expression by both immunofluorescent labeling and western blot [Figure 3a(i), a(ii)]. In contrast, the same cells became spindle shaped, lost E-cadherin expression and exhibited increased vimentin expression following CR infection [Figure 3a(i), a(ii)]. Interestingly, CR-infected cells also stained positive for DCLK1 that colocalized with vimentin, whereas uninfected cells were negative for either of these proteins (Figure 3b). As vimentin is a target of β-catenin/Tcf4 signaling,⁴³ cells treated with β-catenin small interfering RNA failed to exhibit vimentin staining in response to CR infection, which paralleled the lack of vimentin staining in cells treated with Notch blocker dibenzazepine (DBZ),³⁵ whereas E-cadherin staining was not detected under either condition (Figure 3c). During would-healing assay, CR infection compared with uninfected control increased the cellular migration, whereas β-catenin small interfering RNA almost completely, and ΔescV and DBZ partially, inhibited cell migration [Figure 3d(i), d(ii)]. Consistent with hypoxia being a frequent feature of the microenvironment of infected tissues due to activation of the hypoxia-inducible factor and NF-κB pathways,^18,19 CR-infection-induced Hif-1α expression that coincided with relative increases in Hif-1α levels during hypoxia, in addition to induction of HDAC3, β-catenin, Slug and Snail (Supplementary Figure 2A). During wound-healing assay, N-acetylcysteine, an inhibitor of reactive oxygen species activity that is elevated during hypoxia,⁴⁴ inhibited CR-induced cell migration under normoxic conditions [Figure 3e(i), e(ii)]. During hypoxia, although CR-induced cell migration was recorded within 6 h, CR only partially rescued N-acetylcysteine-induced inhibition of cell migration [Figure 3f(i), f(ii)] due in part to decreases in β-catenin, Slug and Snail (Supplementary Figure 2A). The spindle-shaped morphology recorded during normoxia and, more so, during hypoxia (Supplementary Figure 2B) was specific to CR infection, as noninfected cells remained mostly cuboidal even under hypoxic conditions despite loss of E-cadherin, whereas vimentin staining in CR-infected cells was similar under both conditions (Supplementary Figure 2C). Thus, CR infection seems to facilitate phenotypic change that promotes cellular migration.

Figure 2.

Culture of colonic crypt cells in vitro. (a–c) Colonic crypt cells from uninfected normal (N) or 12 (CR₁₂) and 34 (CR₃₄) days post-infected mice were isolated and cultured for 0–30 days (N, CR₁₂) or 0–24 days (CR₃₄) in insulin/transferrin/selenium-supplemented DMEM. Evidence of EMT in cultured cells. Colonic crypt cells in culture were stained for vimentin and E-cadherin and fibronectin and DCLK1. (ei, fi, gi) Total and/or nuclear colonic crypt extracts from uninfected normal (N) or CR-infected mice (CR) were subjected to western blots with antibodies for indicated proteins. (eii, fii, gii) Representative bar graphs showing relative levels of indicated proteins following normalization with either actin or lamin B (*P<0.05 versus N; n = 3 independent experiments).

Figure 3.

Evidence of CR infection-induced EMT in primary cells. (ai) Uninfected normal (N) YAMC cells were treated with control media or with CR (at 90:1 multiplicity of infection) for 3 h followed by washing to remove bacteria. At 24 h post infection, cells were stained with antibodies for E-cadherin and vimentin, whereas nuclei were stained with DAPI. (aii) Western blot showing relative levels of E-cadherin, vimentin, Slug and Snail in N and CR-infected cells. (b) Colocalization of vimentin with DCLK1. Uninfected (N) or CR-infected YAMC cells were stained with antibodies for DCLK1 and vimentin. Please note the significant colocalization of vimentin with DCLK1 in CR-infected cells. (c) Signaling via the Notch and Wnt/β-catenin pathways regulates EMT. CR-infected cells were treated with either Notch blocker DBZ (CR + DBZ) for 24 h or transfected with small interfering RNA (siRNA) to β-catenin (CR + siβ-cat) for 48 h followed by staining for E-cadherin and vimentin while nuclei were stained with DAPI. (di) Wound-healing assay. Uninfected (N) YAMC cells were infected with either wild-type CR or ΔescV mutant, whereas wild-type CR-infected cells were also transfected with siRNA to β-catenin or treated with DBZ followed by wound assay for 12 h. Please note that almost-complete inhibition of cell migration occured with β-catenin siRNA, whereas ΔescV and DBZ partially inhibited cell migration. (dii) Representative bar graph showing relative migration at 12 h compared with 0 h in indicated samples (*P<0.05 versus control; **P<0.05 versus CR; n = 3 3 independent experiments). (ei, fi) Effect of hypoxia on wound healing. Uninfected (N) or CR-infected YAMC cells under normoxic (5% CO₂, 95% O₂) or hypoxic (5% CO₂, 1% O₂, 94% N₂) conditions were treated as indicated, followed by wound-healing assay (n = 3 independent experiments). (eii, fii) Representative bar graphs showing relative migration at 12 h (eii; *P<0.05 versus control; **P<0.05 versus CR) and 6 h (fii, *P<0.05 versus control; **P and ***P<0.05 versus CR), respectively, compared with 0 h in indicated samples (n = 3 independent experiments).

Epigenetics and EMT

Epigenetic mechanisms have recently been implicated in driving the process of EMT.²⁰ We investigated whether epigenetic modulation of β-catenin, E-cadherin or its repressor Snail may be contributing towards trans-differentiation of isolated crypt cells in culture. Analysis of HDAC profile in the crypts revealed significant increases in both HDAC3 on days 6 and 12, and HDAC4 on days 12 and 34 compared with the uninfected control [Figure 4a(i)]. In the crypt-denuded lamina propria representing the mesenchyme,³⁹ HDAC3 levels remained elevated on days 6–34, whereas HDAC4 levels were higher on days 6–12 before declining on day 34 [Figure 4a(ii)]. Increases in HDAC3 on days 6 and 12 were associated with robust expression of H3K4me3, and its methylase SMYD3 in the crypt. In the crypt-denuded lamina propria, although changes in H3K4me3 were subtle, SMYD3 increased significantly on day 12, followed by a decline on day 34 [Figure 4a(i),a(ii)]. Polycomb repressive complex protein Enhancer of Zeste Homolog 2 (EZH2) increased significantly in the crypt on day 6 post infection, followed by a declining trend on days 12 and 34, and coincided with expression of EZH2 target protein H3K27me3 (Figure 4b) that marks the repressive chromatin signature. Indeed, expression of Wnt antagonist Dickkopf (Dkk)-2 decreased significantly on days 6 and 12 (Figure 4c), whereas downregulation of H3K27me3 on day 12 in the crypt coincided with upregulation of β-catenin, cyclinD1, Slug and Snail (see Figures 2e–g). Interestingly, an almost-complete loss of H3K27me3 expression was observed in the crypt-denuded lamina propria on days 6 and 12 compared with the uninfected control (Figure 4d), which coincided with trans-differentiation of cultured crypt epithelial cells into mesenchymal cells (see Figure 2). As proof-of-principle, knockdown of EZH2 significantly inhibited Topflash reporter activity but not NF-κB signaling due to downregulation of β-catenin [Figure 4e(i), e(ii)], suggesting EZH2-induced derepression of Wnt antagonists (for example, Dkk-2, Wif-1 and so on, not shown) in the process. During chromatin immunoprecipitation assay in the crypt on day 6, EZH2’s accumulation around β-catenin, Jagged-1, H3K4m3-methyltransferase SMYD3 and Snail promoters resulted in decreased levels, whereas E-cadherin levels slightly increased (Figure 4f). Similar chromatin immunoprecipitation assay in the crypt on day 12 revealed H3K4me3’s accumulation around β-catenin, Jagged-1, SMYD3 and Snail promoters, resulting in upregulation, whereas E-cadherin levels slightly decreased (Figure 4g), further supporting EMT-like phenomenon.

Figure 4.

Effect of CR infection on epigenetic parameters. (a) Nuclear colonic crypt or crypt-denuded lamina propria extracts from uninfected normal (N) or days 6, 12 and 34 post-CR-infected mice were subjected to western blots with antibodies for indicated proteins (n = 3 independent experiments). (b) Western blots showing relative abundance of H3K27m3, EZH2 and H3K36m3 in the crypt nuclear extracts. H3 was used as loading control (n = 3 independent experiments). (c) Real-time PCR showing fold change in Dkk-2 expression in the crypt isolated from uninfected normal (N) or days 6, 12 and 34 post-CR-infected mice. *P<0.05 versus N (n = 3 independent experiments). (d) Western blots showing relative abundance of H3K27m3, EZH2 and H3K36m3 in the crypt-denuded lamina propria. H3 was used as loading control (n = 3 independent experiments). (ei) Effect of knocking down EZH2 on Wnt and NF-κB pathways in vitro. 293HEK cells were transfected with either TOPflash or FOPflash reporters as well as NF-κB reporter, respectively. Cells were treated with control siRNA or EZH2-siRNA for 48 h followed by measurement of reporter activity using Renilla luciferase as internal control (*P<0.05 versus control; n = 3 independent experiments). (eii) Western blots showing relative levels of EZH2, H3K27m3 and β-catenin following knockdown of EZH2. Actin was used as loading control (n = 3 independent experiments). (f, g) Chromatin immunoprecipitation (ChIP) assay. Crypt genomic DNA was extracted from uninfected normal or days 6 (f) or 12 (g) post-infected distal colons. A ChIP assay was performed with antibodies specific for EZH2 (f) and H3K4me3 (g), respectively. The DNA purified after ChIP was evaluated by semiquantitative PCR using specific primers that recognize the genomic DNA sequences of indicated markers. The amount of DNA after the ChIP assay was normalized to the input DNA level. The bar graphs represent the fold enrichment relative to input (n = 3 independent experiments).

Isolated crypt cells with the ability to form spheroids and/or organoids in vitro

As the ability to self-renew is a hallmark of stem cells and the formation of epithelial-like colonies (spheroids) is an index of self-renewal, we next explored the possibility that CR-infected colonic crypt cells will be better suited to form spheroids (called ‘colonospheres’ from now onwards) in vitro compared with their uninfected counterparts. As depicted in Figure 5a(i), freshly isolated crypt cells from uninfected normal distal colon failed to form colonospheres, which is consistent with the inability to form monolayers in the absence of intestinotrophic factors. Crypt cells from days 12 to 34, on the other hand, formed well-developed colonospheres both at 5 and 8 weeks post plating, whereas day-6 crypt cells only developed colonospheres at 8 weeks [Figure 5a(i)]. Figure 5a(ii) is a representative bar graph showing average number of colonospheres formed/well at indicated times compared with the uninfected controls. Next, we delineated the possibility of cells growing in monolayers to form organoids in vitro. As depicted in Figure 5b(i), cells from days 12 to 34, when trypsinized and regrown in soft agar, transitioned into well-developed organoids at 5 weeks and more so at 8 weeks, whereas day-6 cells developed smaller organoids. Figure 5b(ii) is a representative bar graph showing average number of organoids formed/well at indicated times compared with uninfected controls. Interestingly, within 3 weeks, the same cells from days 12 or 34 when trypsinized and embedded in Matrigel developed into colonoids[Figure 5c(i)]. Figure 5Cii is a representative immunohistochemistry of day-12 colonoid showing H&E, PAS, Ki-67 and DCLK1 staining, whereas 5D is a representative western blot, showing relative levels of DCLK1 and Lgr5 in the colonoids. Thus, CR-infected colonic crypt cells not only undergo trans-differentiation into EMT-like cells but also retain the ability to form organoids/colonoids in vitro in the absence of known epithelial mitogens.

Figure 5.

Crypt cells in culture exhibit colonosphere and organoid/colonoid formation. (ai) Colonic crypt cells were isolated from uninfected normal or 6, 12 and 34 days post-CR-infected (CR-D6, CR-D12, CR-D34) NIH:Swiss mice and directly embedded in 0.3% soft agar. A set of two independent experiments are shown. Please note the significant growth of day-12 and -34 colonospheres at 5 and 8 weeks post plating. (aii) is a representative bar graph showing average number of colonospheres formed/well at indicated times compared with uninfected controls. *P<0.05 versus control (n = 5 independent experiments). (bi) Cells from 6, 12 and 34 days post-CR-infected mouse distal colonic crypts growing as monolayers and showing evidence of EMT were trypsinized, and regrown in 0.3% soft agar for 5 and 8 weeks, respectively. (bii) is a representative bar graph showing average number of organoids formed/well at indicated time points compared with uninfected controls. *P<0.05 versus day 6 (n = 5 independent experiments). (ci) Trypsinized cells from 12 and 34 days as described in (bi) were regrown in Matrigel for 21 days, and were followed for colonoid formation in vitro. Please note the gradual appearance of crypt-like structures on day 21, suggesting a process of MET transition. (cii) A single colonoid from day 12 group was sectioned and stained for: H&E, PAS to label goblet cells, Ki-67 to label proliferating cells and for DCLK1 as a stem cell marker. (d) Western blots showing DCLK1 and Lgr5 levels in the colonoids at indicated time points (n = 3 independent experiments).

Signaling via the Wnt and Notch pathways regulates EMT

We have shown previously that CR-induced colonic crypt hyperplasia is associated with robust activation of the Wnt/β-catenin^30–35 and Notch^35,36 pathways, and interference with the Wnt/Notch crosstalk abrogates hyperplasia.³⁵ We therefore hypothesized that blocking these pathways in vivo will significantly reduce the ability of the isolated crypt cells to either undergo EMT or form colonospheres and/or organoids in vitro. We utilized nanoparticle-encapsulated small interfering RNA approach to target β-catenin (si-β-Cat-NP) while Notch signaling was targeted in vivo by γ-secretase inhibitor, DBZ as described.³⁵ As depicted in Figure 6a, uninfected colonic crypt cells failed to form the monolayers, whereas day-12 crypt cells exhibited monolayer formation as described elsewhere. In response to either with si-β-Cat-NP or DBZ for 10 days, we did not observe monolayer formation in either case [Figure 6a(i)], whereas a combination approach to include both inhibitors did not necessarily yield a synergistic response [Figure 6a(i)]. Western blotting revealed significant downregulation of both active and total β-catenin along with decreases in downstream targets Jagged-1, Slug and Snail with si-β-Cat-NP, whereas only Snail was downregulated with DBZ compared with CR + vehicle-treated samples [Figure 6a(ii), a(iii)]. During the colonosphere formation assay, day-12 cells formed well-developed colonospheres compared with the uninfected control [Figure 6b(i)]. si-β-Cat-NP completely blocked colonosphere formation, whereas DBZ treatment led to significant reduction in both number and size of colonospheres formed in vitro [Figure 6b(i)]. Interestingly, the combination approach was much more effective in blocking colonospheres formation [Figure 6b(i)]. Figure 6b(ii) is a representative bar graph showing average number of colonospheres formed/well in various treatment groups. Thus, CR-induced activation of both the Wnt/β-catenin and Notch pathways in vivo is apparently sufficient to provide the necessary ingredients needed for the process of trans-differentiation to begin in vitro, and that the interplay between the two pathways provides an interesting target to block EMT in vivo.

Figure 6.

Signaling via the Wnt/β-catenin and Notch pathways are critical for EMT and for colonosphere formation. (ai) CR-infected mice received intraperitoneal injections of either nanoparticle-encapsulated β-catenin siRNA (si-β-Cat) or γ-secretase inhibitor DBZ for 10 days, followed by crypt isolation and plating for monolayer formation. Please note the almost-complete lack of monolayer formation when cells were isolated from animals treated with either si-β-Cat or DBZ separately or in combination. (aii,aiii) Western blots for indicated proteins in the colonic crypt cellular extracts prepared from uninfected normal (N), CR-infected (CR), CR-infected + vehicle-treated (CR + V), CR-infected and either si-β-Cat (CR + si) or DBZ-treated (CR + DBZ) mice (n = 3 independent experiments). (bi) Colonic crypt cells isolated from above group of animals were tested for their ability to form colonospheres in vitro. Although day 12 cells formed colonospheres as expected, si-β-Cat-treated cells failed to form any colonosphere. DBZ alone also inhibited colonosphere formation, but the inhibition was more efficient when the two inhibitors were given in combination. (bii) A representative bar graph showing average number of colonosphere formed/well at indicated time points compared with uninfected controls. *P<0.05 versus uninfected control; **P<0.05 versus CR-D12 (ND, not detected; n = 5 independent experiments).

Effect of cellular transformation on EMT and/or colonospheres/organoid formation in vitro

In the TMCH model, CR-induced hyperplastic state increases the susceptibility to the mutagenic effect of 1,2-dimethylhydrazine.⁴⁵ We therefore hypothesized that both the hyperplasia on days 6 and 12 as well as increased presence of DCLK1 + ve cells on day 34 will facilitate cellular transformation following a second hit. NIH:Swiss mice either uninfected or exhibiting 6, 12 or 34 days of TMCH were given azoxymethane (AOM) once for 8 h (Gp1) or weekly for 3 weeks (Gp2), followed by crypt isolation a week later. As depicted in Figure 7ai, crypt cells from uninfected mice receiving AOM acquired the ability to form monolayers in vitro compared with untreated controls. However, they failed to exhibit a continuous growth pattern that stalled completely within few days after plating. In contrast, both day-12 and -34 cells from Gp1 animals exhibited significant growth in culture compared with cells from the same time point without AOM [Figure 7a(i)], and retained the ability to form colonospheres/organoids [Figure 7a(ii)]. Figure 7a(iii) is a representative bar graph showing average number of colonospheres formed/well in various treatment groups. In Gp2 animals, we recorded some fascinating observations: neither uninfected normal nor day-12 or -34 cells exhibited any monolayer formation compared with cells without AOM [Figure 7b(i)], but exhibited significant growth as colonospheres when directly embedded in soft agar [Figure 7b(ii)]. Figure 7b(iii) is a representative bar graph showing average number of colonospheres formed/well in various treatment groups. During immunostaining of cells in monolayers from Gp1 animals, day-12 and -34 cells compared with either day-6 or uninfected cells exhibited significant co-staining for both DCLK1 and fibronectin [Figure 8a, b(ii)–b(iii)], whereas CD133, a CSC marker, also exhibited increased staining in response to AOM but failed to colocalize with an epithelial marker laminin (Supplementary Figure 3). Thus, AOM-induced complete cellular transformation may be inhibitory for monolayer formation in vitro, and that the EMT–MET, which we recorded on day-12 or 34 cells without AOM (see Figures 5b–d), may be more relevant in the regulation of hyperplasia than tumorigenesis.

Figure 7.

Partial or complete transformation of colonic crypt cells following mutagenic insult. (ai) Uninfected or days 12 (CR-D12) and 34 (CR-D34) CR-infected NIH:Swiss mice were given AOM (D12 + AOM, D34 + AOM) for 8 h. and their colonic crypts were isolated and tested for their ability to form monolayers. (aii) Cells growing as monolayers and showing evidence of EMT were trypsinized and regrown in 0.3% soft agar for 5–8 weeks. Please note the significant colonospheres/organoid formation, particularly from days 12 and 34 cells. (aiii) A representative bar graph showing average number of colonospheres/organoids formed/well at indicated time points. *P<0.05 versus day 6 (n = 5 independent experiments). (bi) Uninfected or days 12 (CR-D12) and 34 (CR-D34) CR-infected NIH:Swiss mice were given AOM (N + AOM, D12 + AOM, D34 + AOM) weekly for 3 weeks, and their colonic crypts were isolated and tested for their ability to form monolayers. Please note the lack of monolayer formation in N + AOM, D12 + AOM and D34 + AOM groups. (bii) Colonic crypt cells isolated from animal groups described in (bi) were directly embedded in 0.3% soft agar and followed for colonosphere formation for 21 days. (biii) A representative bar graph showing average number of colonospheres formed/well in each group. *P<0.05 versus control (n = 5 independent experiments).

Figure 8.

Monolayers in culture stain positive for markers of mesenchymal and cancer stem cells. (A) Uninfected normal (N) or days 6, 12 and 34 post-CR-infected NIH:Swiss mice were given AOM (N + AOM; D6 + AOM; D12 + AOM, D34 + AOM) for 8 h followed by crypt isolation, and cultured as monolayers. Cells growing as monolayers were stained for markers of mesenchymal and stem cells. Please note the significant co-staining of fibronectin with DCLK1 particularly on days 12 and 34 (arrowheads; n = 3 independent experiments). (bi–biii) Representative bar graphs showing percent cells positive for DCLK1 (bi; *P<0.05 versus N + AOM), fibronectin (Bii; *P<0.05 versus N + AOM) and DCLK1, and fibronectin colocalization (biii; *P<0.05 versus N + AOM), respectively (n = 3 independent experiments).

CR/AOM treatment promotes nodular and/or tumorigenic growth in vivo

Next in a xenograft study, we examined the nodular or tumorigenic growth potential of colonospheres generated from the two groups of animals. When injected subcutaneously, colonospheres from both Gp1 and Gp2 mice developed nodular/tumorigenic growth in athymic/nude mice. Specifically, the colonospheres collected from day-12 or -34 cells without AOM developed nodular growth, but the growth tapered off within 6 weeks (Figure 9a). Both colonospheres from Gp1 and particularly from Gp2 mice exhibited significant nodular and/or tumorigenic growth, which was sustained for at least 12 weeks (Figure 9a). Interestingly, colonospheres from Gp2 mice developed nodules/tumors quickly within 45 days post injection compared with 75 days for Gp1 mice, and the sizes of these nodules/tumors were also significantly different (Figure 9a). Figure 9b is a tumor-growth curve showing percent growth of the xenograft in the two groups versus control. When these nodules/tumors were fixed, and sectioned and stained for markers of epithelial/mesenchymal lineages along with a stem cell marker, we observed significant heterogeneity in cell types, as nodules/tumors were positive for both vimentin and cytokeratin along with DCLK1 (Figure 9c). Thus, both hyperplasia (on day 12 of TMCH) and increased expression of DCLK1 (on day 34 of TMCH) may be associated with cellular transformation and tumorigenic growth.

Figure 9.

CR/AOM treatment promotes nodular and/or tumorigenic growth in athymic nude mice. (a) Day-12 colonospheres from both Gp1 and Gp2 mice were dissociated and injected subcutaneously into athymic nude mice, and followed for tumor formation in vivo. Upper panel: kinetics of nodule/tumor formation from day-12 or -34 colonosphere in the absence of AOM; middle panel: nodular/tumor growth of day-12 or -34 colonospheres from group 1 (8-h AOM) mice; lower panel: complete kinetics of nodular/tumor growth from uninfected, days 6, 12 or 34 animals in group 2 (3 weeks of AOM injection). (b) Tumor growth curve. Line graph showing percent growth of the xenograft in the two groups versus control (^◆P and ^◆◆P<0.05 versus control; n = 3 independent experiments). (c) Paraffin-embedded sections prepared from a day-12 tumor were stained with antibodies specific for: vimentin, cytokeratin and DCLK1 (n = 3 independent experiments).

DISCUSSION

The EMT–MET processes have central roles in embryonic development, fibrogenesis and tumor progression.⁸ Many stimuli, signal transduction pathways, such as Wnt^46–49 and Notch,^50–53 and transcription factors, such as Slug and Snail, govern the acquisition of EMT. As many of these EMT-signaling pathways are also upregulated by microbial pathogens, these studies suggest that pathogens should be considered as potential EMT inducers. Indeed, modulation of TGFβ following microbial invasion of cultured cells is implicated in the EMT process.^52,54 However, little is known regarding our understanding of how these pathways coordinately suppress the epithelial phenotype and induce a mesenchymal program that generates cells with properties of stem cells, and whether the acquisition of EMT phenotype could be an early event triggered by enteric pathogens. In the current study, we demonstrate that CR infection-induced activation of the Wnt/β–catenin, Notch and TGFβ pathways in vivo promotes trans-differentiation of cultured crypt cells into fibroblast-like mesenchymal cells, and that these EMT-like events are likely sustained by transient epigenetic changes including histone modification and chromatin remodeling. The fact that blocking both of the Wnt/β-catenin and Notch pathways either separately or together blocked the monolayer-forming ability of these cultured cells rules out the possibility that the isolated cells may have been either fibroblasts or other mesenchymal cells that will simply outgrow the epithelial cells in culture. A corroborating study in vitro has shown that lipopolysaccharide induces EMT of intrahepatic biliary epithelial cells.²⁵ Moreover, pathogenic H. pylori strain 60190 is shown to induce EMT markers Slug and Snail,²³ whereas diffusely adherent E. coli also promotes an EMT-like behavior in vitro.^55–57 We now provide evidence of CR-induced EMT of immortalized, but primary, epithelial cells that promotes cell migration and intriguingly exhibits responses that mimic changes recorded during hypoxia. Given that hypoxia/Hif-1α regulates metastatic processes, especially EMT by directly interacting with β-catenin^44,58 or Notch intracellular domain,⁵⁹ it is not surprising that CR infection that mimics hypoxia and is associated with activation of the Wnt/β-catenin^30,34 and Notch^35,36 pathways, promotes neoplasia in response to a second hit.^36,45 Finally, several recent studies have shown that it is not unprecedented for EMT to generate cells with many of the properties of self-renewing stem cells (7–9). Most of these studies, however, rely on ectopic expression of either the Twist or Snail transcription factors, both of which are capable of inducing EMTs in epithelial cells.^60–62 We provide data that suggest that both HDAC3 and HDAC4 via differential expression of H3K4me3 and H3K27me3 may lie at the center of chromatin alterations in one case to repress Wnt antagonists, and in the other to activate genes that promote EMT (see Figure 4). This notion is supported by a recent finding by Wu et al.²⁰ who showed an essential role for HDAC3 in hypoxia-induced EMT. At the clinical level, our findings corroborate with studies both in patients with Crohn’s disease²⁴ and in 2,4,6-trinitrobenzene sulfonic acid (TNBS)-induced model of Crohn’s colitis⁶³ that exhibits EMT. As the process of EMT is usually associated with metastasis and frank malignancy, our findings suggest that this process may be an early event which, despite defying the conventional wisdom, could still be facilitated by a pathogenic insult. Thus, although TMCH is a self-limiting disease, the observed changes recapitulate series of events that precede cancer cell metastasis and tumor spread. We believe, however, that this epigenetically regulated EMT process may not reflect a classic EMT, but could be related to the stromal–epithelial crosstalk that is required to maintain crypt homeostasis in response to CR infection. Whether stromal presence of DCLK1 is the driving force behind this crosstalk is currently being investigated.

Stem-like cells in the primary epithelial tumor may undergo EMT to give rise to migrating CSCs, which may revert to stationary epithelial-like CSCs after a MET at the site of metastasis.⁶⁴ However, how MET is regulated in vivo is not known. On the basis of findings in colorectal cancer metastases, a transient EMT–MET process characterized by phenotypic plasticity could be triggered by environmental agents and may involve reversible epigenetic modifications in relevant genes.⁶⁴ In the current study, we have discovered significant colonosphere-forming ability of the freshly isolated, as well as trypsinized, cells in culture from infected mice, and blocking the Wnt/β-catenin and Notch pathways either separately or together completely abolishes colonosphere-forming ability. Intriguingly, the same cells when grown in Matrigel developed into colonoids, suggesting a role for microenvironment such as extracellular matrix, in the process. The fact that these colonoids exhibited positive staining for DCLK1 suggests an inherent ability of the DCLK1 + ve cells to promote the transient EMT–MET process in vitro. At the same time however, it is also conceivable that the non-stem cells in this population may spontaneously convert to a stem-like state, as is shown recently,⁶⁵ thereby promoting the EMT–MET process in vitro. Nevertheless, as no definitive proof exists for a transient EMT–MET, our studies, which despite not being a true reflection of macrometases, still represent the only evidence for such a process in a nonmalignant setting, and suggests that the stem cell plasticity observed during the EMT–MET process can be applied as a tool to study crypt regeneration following a pathogenic insult.

The CR-induced hyperplastic state increases the susceptibility to either mutagenic insult⁴⁵ or in mice heterozygous for Apc gene.^36,66 In the current study, we observed either monolayer and colonosphere formation during acute AOM injection or simply colonosphere formation in two-dimensional culture following chronic AOM treatment. Interestingly, colonospheres from day-12 and -34 cells when injected subcutaneously into athymic/nude mice developed nodules/tumors of comparable sizes, thereby confirming our earlier hypothesis. Whether the monolayer or nodule/tumor formation was a characteristic of stem cells was confirmed by staining for both DCLK1 and CD133. We have shown recently that DCLK1 regulates EMT of human pancreatic cells through a miR-200a-dependent mechanism, and that nanoparticle-based delivery of si-DCLK1 inhibits colorectal cancer tumor growth via a Notch-1-dependent mechanism.^67,68 Moreover, a link between EMT and stem cells is recently established.⁷ As stem cell state correlates with bivalent activating/repressing histone modification of H3K4me/H3K27me,⁶⁹ it is tempting to speculate that CR-induced epigenetic changes in either HDAC3/4 or EZH2 may be critical in regulating the epithelial–mesenchymal balance. Efforts are therefore underway to block HDAC3/4 or EZH2 or knockdown DCLK1, either in vivo or in vitro to see if either EMT or nodules/tumor formation could be interfered with.

In conclusion, deregulation of tumor-suppressor and/or stem cell-associated pathways (for example, Wnt, JAK-STAT, JNK, Notch and so on), as well as epigenetic reprogramming induced by bacteria, are possible causes of cancer development in epithelial niches.⁵⁴ We ^30–39 and others⁷⁰ have extensively shown that enteric pathogens such as CR can activate signaling pathways that are directly relevant in the early-onset EMT. In addition, the EMT–MET process can be used as a tool to study regenerative medicine for various debilitating diseases such as inflammatory bowel disease. Finally, our studies may also help us understand the molecular basis of microbial pathogenesis and in defining future therapeutic strategy in treating diseases with infectious etiology.

MATERIALS AND METHODS

Transmissible murine colonic hyperplasia

TMCH was induced in 5–6 week-old Helicobacter-free NIH:Swiss mice (Harlan, IN, USA) by oral inoculation with a 16-h culture of either wild-type CR or type-three secretion system- mutant ΔescV, as previously described.^30–39 Age- and sex-matched control mice received sterile culture medium only. Animals were euthanized on 6, 12, 20, 27 and 34 days post infection, and distal colons removed and portions of the colon were fixed for immunohistochemistry. Distal colonic crypts or crypt-denuded lamina propria were isolated as described.^{30–39,71,72}

AOM injection to induce cellular transformation

In the first strategy (group 1), uninfected normal (N) or CR-infected NIH:Swiss mice on 12 or 34 days post infection received either a single intraperitoneal injection of AOM in saline at 10 mg/kg body weight or saline alone as controls, and the animals were euthanized after 8 h. In the second strategy (group 2), N or CR-infected mice on 12 or 34 days post infection received three injections of AOM or saline (at 10 mg/kg body weight) intraperitoneally at weekly intervals, followed by euthanasia 1 week later. Crypts were isolated from these colons as described^{30–39,71,72} and used for either monolayer formation or for spheroid assay.

Monolayer formation of crypt cells, hypoxia and wound-healing assay

Isolated crypts were centrifuged at 1000 r.p.m. for 5 min. The pellets were washed with phosphate-buffered saline, resuspended in RPMI glutamax medium/0.5 U/ml dispase at 37 °C and shaken gently for 5 min. The cells were pelleted and resuspended in RPMI glutamax medium supplemented with 5% fetal calf serum plus penicillin and streptomycin, and incubated at 37 °C in 5% CO₂. Monolayer formation was followed for 0–30 days by replacing the medium at 48 h. For hypoxia (1.0% O₂) experiments, YAMC cells were seeded in normoxia and grown to subconfluence of ~60–70%, or as specifically needed. After 2 h of serum-free culture, cells were exposed to hypoxia in a hypoxic chamber (Thermal Tech, Orlando, FL, USA) for indicated duration. Cells were infected with CR (multiplicity of infection 1:30) for 3 h, washed and treated with reactive oxygen species inhibitor N-acetylcysteine (100 μmol/l). For wound-healing assay, YAMC cells (1 × 10⁶) seeded onto 12-well plates were scratched using a sterile pipette tip 3–6 h post seeding, and the same spots were photographed under phase contrast microscopy at 6–12 h post ‘wounding’.

Clonogenic assay

Freshly isolated crypt cells or cells growing as monolayers were trypsinized and pelleted at 1000 r.p.m. for 5 min at 4 °C before being suspended at an approximate density of 5000 cells/200 μl/well (48-well plate) in 2 × RPMI medium containing 0.3% soft agar with 5% fetal calf serum, without intestinotrophic mitogens such as R-spondin or Noggin. The cell suspensions were plated in 48-well plate above a layer of solidified 1% soft agar in plain 2 × RPMI medium. The plates were incubated at 37 °C under 5% CO₂ and the cells were followed for spheroid formation in RPMI glutamax medium plus 1% fetal calf serum at weekly intervals for 5–8 weeks. For colonosphere assay in Matrigel, same number of cells were resuspended in 100 μl of growth factor-reduced Matrigel in a 48-well plate, and colonosphere formation was followed as described.^73–75

Knockdown of Wnt/β-catenin and Notch signaling in vivo

For blocking β-catenin in vivo, we injected intraperitoneally β-catenin small interfering RNA incorporated into poly (lactide-co-glycolide) acid nanoparticles (PLGA-NPs) (at 4 μM/kg body weight) every alternate day for 10 days, followed by euthanasia at 12 days post infection. These PLGA-NPs were synthesized using a double emulsion solvent evaporation technique as described.⁷⁶ To block Notch signaling in vivo, we used a cell-permeable inhibitor of γ-secretase, DBZ (EMD Chemicals, Inc, Gibbstown, NJ, USA) as described.³⁵ Distal colonic crypts were isolated on 12 days post infection as described,^{30–39,71,72} and used for both monolayer formation and spheroid assays.

Xenograft study

For xenograft study, spheroids formed in soft agar were counted and resuspended in phosphate-buffered saline, and ~1000 spheroids were injected subcutaneously into the left flanks of athymic nude mice (BALB/cnu/nu; Harlan). Engrafted mice were inspected biweekly for nodule/tumor appearance by visual observation and palpation. Mice were euthanized when the tumor size reached a diameter of 0.5–1 cm or on 2 months post transplantation, and the tumors were analyzed as described.^{67,68,73,74,77}

Chromatin immunoprecipitation and western blotting

Chromatin immunoprecipitation assay was performed in the isolated crypts, utilizing the commercially available CHIP 1-day kit according to the manufacturer’s instructions (Qiagen, Valencia, CA, USA). Cellular or nuclear extracts prepared from the isolated crypts (30–50 μg protein/lane) were subjected to SDS–PAGE and western blotting with appropriate primary and secondary antibodies, respectively, as described.^{30–39,71,72}

Immunofluorescence/immunohistochemistry

Fluorescent microscopy on confluent cells or immunohistochemistry on 4-μm-thick paraffin-embedded sections prepared either from distal colon or organoids/colonoids were performed as described.^{30–39,71,72}