Colitis-induced IL11 promotes colon carcinogenesis

Abstract

Colitis increases the risk of colorectal cancer; however, the mechanism of the association between colitis and cancer remains largely unknown. To identify colitis-associated cancer promoting factors, we investigated gene expression changes caused by dextran sulfate sodium (DSS)-induced colitis in mice. By analyzing gene expression profiles, we found that IL11 was upregulated in DSS-induced colitis tissue and 2-amino-1-methyl-6-phenylimidazo[4,5-b]-pyridine (PhIP)/DSS-induced colon tumours in mice as well as in human colorectal cancer. By characterizing the activation/phosphorylation of STAT3 (pSTAT3), we found that pSTAT3 was induced transiently in colitis, but maintained at higher levels from hyper-proliferative dysplastic lesions to tumours. Using the IL11 receptor (IL11Rα1) knockout mice, we found that pSTAT3 in the newly regenerated crypt epithelial cells in colitis is abolished in IL11Rα1+/− and −/− mice, suggesting that colitis-induced IL11 activates STAT3 in colon crypt epithelial cells. Moreover, colitis-promoted colon carcinogenesis was significantly reduced in IL11Rα1+/− and −/− mice. To determine the roles of the IL11 in colitis, we found that the inhibition of IL11 signalling by recombinant IL11 antagonist mutein during colitis was sufficient to attenuate colitis-promoted carcinogenesis. Together, our results demonstrated that colitis-induced IL11 plays critical roles in creating cancer promoting microenvironment to facilitate the development of colon cancer from dormant premalignant cells.

Introduction

Colorectal cancer (CRC) is the third most commonly diagnosed cancer and the second leading cause of cancer death and accounts for about 1 in 10 cancer cases and deaths worldwide (1). Elucidating the mechanism of colon carcinogenesis is of profound importance in developing effective prevention and therapy strategies to reduce CRC and related death. CRC risk factors include modifiable factors, including body weight, lifestyle and die, and non-modifiable factors, including age, sex, family history, inherited genetic risk, prior adenomatous or sessile serrated polyps, inflammatory bowel disease and diabetes mellitus (2). Across several factors, inflammation is a common key element that increases the risk of CRC (3).

The most commonly used colon inflammation model is the colitis induced in rodents by supplementing drinking water with low molecular weight dextran sodium sulfate (DSS) (4–6). Administration of DSS leads to massive epithelial damage and robust inflammatory response in the colon, recapitulating several key features of human colitis, such as bloody diarrhoea, mucosal ulcerations, infiltrations with neutrophils and granulocytes and body weight loss. DSS-induced colitis is also commonly used to promote carcinogen- and genetic-induced colon tumorigenesis in rodent models (7–10). It involves multiple factors including not only colon epithelial damage, repair and regeneration but also infiltration of inflammatory cells, the elevation of inflammatory cytokines and chemokines, invasion of gut microorganisms and oxidative stress, all of which are demonstrated to play critical roles in promoting tumorigenesis. Consequently, colitis produces a pro-tumorigenic microenvironment that triggers the activation of premalignant cells (10–13). However, the detailed molecular mechanisms of the generation of cancer-promoting microenvironment in colitis tissues remain largely unknown.

Previously, we established a colon carcinogenesis model induced by a dietary carcinogen, 2-amino-1-methyl-6-phenylimidazo[4,5-b]-pyridine (PhIP; the most commonly occurring and abundant heterocyclic amine produced in high-temperature cooked meat and fish), in CYP1A-humanized (hCYP1A) mice, which resemble humans in activating PhIP to an active carcinogen (14–17). In this model, DSS-induced colitis is required for colon tumour development after PhIP treatment. In fact, the requirement of DSS-induced colitis is also found in the widely-used azoxymethane (AOM)/DSS-induced colon carcinogenesis model (6,9,18,19). In the genetic model such as the mice carrying a loss-of-function mutation in one Apc allele (i.e. Apc+/min), large numbers of tumours are developed in the small intestine, while colon tumour incidence is very low but is significantly increased by DSS-induced colitis (20). These experimental data demonstrate that colitis plays important roles in promoting colon tumorigenesis. We uncovered that PhIP/DSS-induced tumours carry dominant active mutations of β-catenin, which lead to the activation of Wnt signalling, a key cellular signalling pathway in human and rodent colon tumorigenesis (14). Mutation of β-catenin resulted from a single nucleotide replacement, such as G to A/C/T at codon 32 or 34 in exon 3. Since the mutagenic activity of PhIP is caused by the DNA-adduct formed by PhIP at the 8-position carbon of deoxyguanine base (21), these β-catenin mutations are highly likely caused by PhIP. Therefore, in PhIP/DSS-induced carcinogenesis, PhIP induces oncogenic mutations that create premalignant cells, but these cells remain quiescent until being activated by DSS-induced colitis. However, the mechanisms of cancer promotion by colitis are not clear. Therefore, identifying cancer promoting factor(s) produced by colitis could provide insights on the molecular mechanism behind CRC promotion by inflammation.

In this study, we investigated gene expression profiles in the DSS or PhIP/DSS-induced colitis tissues and PhIP/DSS-induced colon tumours of hCYP1A mice. By correlating the up- and down-regulated genes with human CRC genomic data in The Cancer Genome Atlas (TCGA) studies, we identified that IL11 was activated by DSS-induced colitis and maintained a high level of expression in PhIP/DSS-induced tumours. We further demonstrated that loss of IL11 signalling in IL11Rα1 knockout mice attenuated PhIP/DSS or AOM/DSS-induced colon carcinogenesis, and the inhibition of IL11 signal during colitis by the administration of recombinant IL11 antagonist mutein was also sufficient for reducing tumorigenesis. Collectively, our data demonstrate that elevated IL11 creates cancer promoting condition in colitis tissues.

Materials and methods

Animal experimental procedures

The hCYP1A mice, carrying human CYP1A gene in place of the mouse Cyp1A gene, were generated as described previously (15). Heterozygous IL11Rα1 knockout mice (IL11Rα1+/−), originally developed by Nandurkar et al. (22), were obtained from Jackson Lab (Bar Harbor, ME). Since IL11Rα1−/− mice are infertile, IL11Rα1+/− mice were used for breeding. The IL11Rα1+/− mice were also bred with hCYP1A mice to generate hCYP1A:IL11Rα1+/− mice, used to produce IL11Rα1+/+, +/− and −/− in hCYP1A mice. All animal procedures were in accordance with the animal study protocol approved by the Rutgers University Institutional Animal Care and Use Committee. Throughout the study, mice at the age of 5 weeks were switched from a normal lab chow diet to the semi-purified AIN93M diet (Research Diets, New Brunswick, NJ). At the age of 6 weeks, the mice were treated with two doses of PhIP (FUJIFILM Wako Chemicals USA, Co., Richmond, VA; 100 mg/kg body weight; a week apart) by oral gavage or two doses of AOM (Midwestern Research Institute, Kansas City, MO; 5 mg/kg body weight; a week apart) by intraperitoneal injection. After 1 week, the mice were administered 1.5% (w/v) DSS (MP Biomedicals, Solon, OH) in drinking water for 5 days. Then, mice were returned to regular drinking water and continuously fed AIN93M diet. At the end point, the mice were sacrificed for collecting the colon for further analyses. To collect cells or RNA, only the descending colon epithelia were used.

To study the effect of IL11 antagonist mutein IL11(W147A), mice were administrated with recombinant IL11(W147A) daily at the dose of 300 µg/kg BW for 2 weeks by intravenous (IV) injection through the mouse tail vein. PBS was used as a vehicle control.

Human colon cancer samples

A total of 9 frozen de-identified human colon cancer samples with their adjacent non-cancerous tissues as the controls were obtained from Rutgers Cancer Institute of New Jersey for RNA extraction. The expression levels of IL11 were determined using RNA-sequencing conducted by BGI US (Cambridge, MA). This study is in accordance with the study protocol approved by the Institutional Review Board of Rutgers, the State University of New Jersey.

Isolation of EpCAM⁺ and EpCAM⁻ cells

To separate colon crypt epithelial cells from stromal cells, we used EpCAM as the positive selection marker to isolate epithelial cells from colon tissues and tumours. The isolation was carried out by positively selecting magnetic MicroBeads-labelled cells using a MACS column (Miltenyi Biotec, Auburn, CA). In brief, tissues were minced and treated with collagenase and hyaluronidase (StemCell Technologies Inc., Vancouver, Canada) to prepare the single-cell suspension. Then, the dead cells were removed first using Dead Cell Removal Kit (Miltenyi Biotec). After leukocyte subtypes expressing EpCAM was depleted of leukocytes using mouse CD45 MicroBeads (Miltenyi Biotec), EpCAM⁺ epithelial cells were selected using mouse EpCAM MicroBeads (Miltenyi Biotec). EpCAM⁻ cells were combined with the cells isolated using CD45 MicroBeads to comprise total stromal cells (referred to as EpCAM⁻ cells in this study).

RNA extraction, RT-qPCR, RNA microarray and RNA-sequencing

Total RNA was extracted using RNAeasy Mini Kit (Qiagen). The fresh or frozen tumours/tissues were homogenized in lysis buffer using Omni Bead Raptor 24 (Omni International Inc., Kennesaw, GA). The isolated EpCAM⁺ and EpCAM⁻ cells were directly lysed by adding lysis buffer. RNA was then purified according to the manufacturer’s protocol.

The expression levels of IL11 in tissue samples were determined by RT-qPCR. In brief, the cDNA of the total RNA was first synthesized by reverse transcription using the SuperScript III First-Strand Synthesis SuperMix (Thermo Fisher Scientific; Waltham, MA), and IL11 and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) were then amplified by real-time PCR using mouse IL11 and GAPDH primer sets, respectively. The sequences of IL11 qPCR primers are 5′-TGTTCTCCTAACCCGATCCCT-3′ and 5′-CAGGAAGCTGCAAAGATCCCA-3′. The sequences of GAPDH qPCR primers are 5′-AGGTCGGTGTGAACGGATTTG-3′ and 5′-TGTAGACCATGTAGTTGAGGTCA-3′. Real time PCR was done using the Power SYBR Green PCR Master Mix (Thermo Fisher Scientific), and each reaction was performed in triplicates on the ViiA 7 Real-Time PCR System (Thermo Fisher Scientific).

The expression profiles of the PhIP/DSS-induced colon tumours were investigated by RNA microarray using RNA samples extracted from 7 tumours from 3 mice. Matching normal controls were the rest colon epithelia after removing all tumours. The microarray, including data collection and analysis, was conducted by Ocean Ridge Biosciences (Deerfield Beach, FL). The expression profiles of colitis tissues were investigated by RNA-sequencing using the RNA samples extracted from descending colon epithelia collected from mice on Day 1, 3, and 7 after DSS or PhIP/DSS treatment. The same aged mice without any treatment were used as controls. Three mice were used for each time point and treatment. Library preparation, RNA sequencing and data analysis were carried out by BGI US.

Tissue preparation and histopathological staining

To prepare tissues for histopathological characterization, colons were rolled and embedded in paraffin for the preparation of 4 µm-thick serial sections after being fixed in 10% formalin overnight, as described previously (15). Histopathological analyses of intestinal tumours were characterized according to well-recognized standards using hematoxylin and eosin (HE) stained slides (23). Immunohistochemical (IHC) staining was performed as described previously (24). The sources and dilution information of primary antibodies used for IHC staining are listed in Supplementary Table 1, available at Carcinogenesis Online).

Statistical analysis

One-way Analysis of Variance (ANOVA) followed by Dunnett's multiple comparison test was used to compare differences among treatment groups and control groups on the GraphPad Prism 5 software. Student's t-test was used to determine the difference between the two groups. The expression levels of IL11 in human colon cancer samples and their matching controls were analyzed using the paired analysis with student t-test. Statistical significance was indicated by P-value <0.05 in the two-tailed comparison.

Results

IL11 is upregulated in DSS-induced colitis tissues and PhIP/DSS-induced tumours in mice and CRC in humans

To characterize colitis-associated gene expression, we extracted RNA from colon epithelia from mice 1, 3 and 7 days after DSS or PhIP/DSS treatment as well as untreated mice as controls for RNA-sequencing (GEO accession no. GSE137387). These time points were selected based on the histological features of this model as characterized previously (16): DSS-induced colon epithelium injury (i.e. ulceration) reached a maximum in 1–3 days after DSS treatment and the regenerated crypts can be clearly identified as early as on Day 7. The results showed that there were ~1500 genes up- and down-regulated by at least 1.0 of log2 fold from the untreated controls to the treated samples. Principal component analysis of the expression profiles distinguished the DSS and PhIP/DSS treatments from the control, but not from each other, in the first two component analyses (Supplementary Figure 1, available at Carcinogenesis Online), indicating that DSS was the major factor causing these changes.

Because the activation of dormant premalignant cells is likely a critical driving force supporting the continuous growth of tumours in such a rapid tumorigenesis model, we expected that the cancer-promoting factors should remain active during tumour development. Therefore, we compared the gene expression changes of colitis tissues with those of tumours from a microarray study consisting of 7 PhIP/DSS-induced colon tumours with their adjacent colon epithelial tissues as the controls (GEO accession no. GSE137292). We found that there were 511 genes up-regulated and 264 genes down-regulated consistently between colitis tissues and tumours by at least 0.5 of log2 fold. Next, to identify the potential drivers from these 775 genes, we compared them with the gene expression data of human CRC in TCGA studies available in the Oncomine (www.oncomine.org) (25) since relevant key cancer promoting factors are expected to be critical in both mice and humans. Using the Oncomine-identified top 1% up-regulated and down-regulated genes in TCGA colon adenocarcinoma study, we narrowed down the colitis-induced changes to a list of 18 genes relevant to mouse tumours and human CRC (Table 1). Among these genes, IL11 draws our attention because IL11 was reported to be the predominant STAT3-activating cytokine in gastrointestinal tumours (26). Therefore, it is highly likely that IL11, a critical player in tumours, is also the factor produced by colitis and contributing to the pro-tumorigenesis microenvironment.

Table 1.

The fold changes of the genes up- and down-regulated in DSS-induced colitis tissues and PhIP/DSS-induced colon tumors in hCYP1A mice and also ranked in the top 1% up-and down-regulated genes of CRC in TCGA study

	Day 1		Day 3		Day 7		Week 8
Gene Symbol (mouse)	DSS vs Controla	PhIP/DSS vs Controla	DSS vs Control1	PhIP/DSS vs Controla	DSS vs Control1	PhIP/DSS vs Controla	Tumours vs Controlsb
Upregulated genes
Tgfbi	3.97	21.18	13.99	2.25	14.47	7.37	4.40
Psat1	3.98	5.99	3.24	2.32	1.99	1.88	2.46
Slc4a11	10.16	18.15	5.98	4.94	12.20	4.24	3.38
Mmp7	19.31	29.63	176.73	90.98	122.07	108.56	7.18
Il11	3.37	9.13	6.98	2.43	5.26	2.43	12.19
Srpx2	3.22	14.37	5.53	4.55	8.04	7.60	3.27
Igf1	2.38	5.39	5.11	2.34	5.30	3.91	2.00
Pi16	2.19	15.71	12.22	1.82	6.03	1.82	2.24
Jam2	1.63	3.98	2.62	1.82	6.23	4.56	3.91
Trib3	3.95	8.98	6.55	2.73	4.18	3.65	2.22
Downregulated genes
Sepp1	0.35	0.29	0.43	0.49	0.60	0.69	0.47
B3gnt7	0.34	0.13	0.30	0.55	0.23	0.43	0.14
Sult1a1	0.36	0.24	0.21	0.29	0.12	0.19	0.03
Pdcd4	0.63	0.36	0.55	0.59	0.50	0.69	0.39
Synpo	0.54	0.38	0.48	0.54	0.57	0.60	0.40
Trpm6	0.35	0.14	0.12	0.27	0.19	0.28	0.14
Bmp3	0.28	0.21	0.44	0.22	0.44	0.40	0.29
Usp2	0.52	0.27	0.29	0.49	0.34	0.41	0.40

^aBased on the RNA-sequencing results of the pooled samples collected from 3 mice.

^bThe average of 7 tumours to their matching controls.

To explore the importance of IL11 in CRC, we examined IL11 expression in the available genomic data. According to Human Protein Atlas (27) (HPA; www.proteinatlas.org), the expression level IL11 is very low in normal human colon and rectum (Figure 1A). Similar results are also found in the Genotype-Tissue Expression (GTEx) project (https://gtexportal.org; Supplementary Figure 2, available at Carcinogenesis Online). In contrast, IL11 is significantly upregulated in all digestive tract cancers, including oesophagal, stomach, colon and rectum, in the available 36 TCGA (Figure 1B). According to TCGA CRC data in the Oncomine, IL11 is upregulated significantly in all subtypes, namely the adenocarcinoma and mucinous adenocarcinoma in cecum, colon, rectum and rectosigmoid junction (Figure 1C). In addition, upregulation of IL11 is found in 8 out a total of 9 other published transcriptome datasets of the colon and/or rectum cancer deposited in the Oncomine (Supplementary Table 2, available at Carcinogenesis Online). But the expression levels of IL11 were not found to be associated with the overall survival of CRC patients according to the data available in PROGgeneV2 (genomics.jefferson.edu/proggene (28)). Furthermore, the HPA data showed that all the examined colon cancer tissues were positive for IL11 receptor (IL11RA) (Figure 1D), although its expression levels were found to be lower in cancer than the normal tissues (Supplementary Figure 3, available at Carcinogenesis Online). Together, these data suggest that IL11 signalling plays critical roles in human CRC.

Figure 1.

IL11 is upregulated in human digestive tract cancer.. (A) IL11 expression in normal human tissues is analyzed using the data from HPA, which included RNA-sequencing results from multiple individuals (https://www.proteinatlas.org). The data is presented as mean TPM (transcripts per million), corresponding to the mean value of different individual samples in each tissue. Color-coding is based on tissue groups, each consisting of tissues with common or relevant functional features according to HPA. (B) The expression levels of IL11 in human normal and cancer tissues are plotted according to the result of the analysis of the available 36 TCGA datasets generated by Oncomine (Oncomine.org). TCGA-colon adenocarcinoma (COAD), -colorectal adenocarcinoma (COADREAD) and -rectum adenocarcinoma (READ) studies are labelled as red. Abbreviations of TCGA studies are listed in Supplementary Table 4, available at Carcinogenesis Online. (C) The expression levels of IL11 in human normal tissues and CRC subtypes in TCGA study are plotted by Oncomine. The sample numbers are labelled above the data. (D) The percentages of patients with IL11RA+ lesions is graphed according to the result of IL11RA IHC staining in human cancers from HPA. (E–F) The IL11 expression levels in 9 pairs of human colon cancer samples and their adjacent non-cancerous tissues were determined by RNA-sequencing. The fold changes of each pair are summarized in (E). The expression levels as reported by RNA-sequencing result (reads per kilobase per million mapped reads; RPKM) are summarized in (F) for each pair. The numbers inside the drawing boxes are patient ID numbers with the information listed in Supplementary Table 3, available at Carcinogenesis Online. * indicated the increasing trend from non-cancerous tissue to cancer in the paired comparison (P < 0.01).

To obtain direct evidence of the involvement of IL11 in CRC, we determined the IL11 levels in 9 human colon cancer samples and their adjacent non-cancerous tissues by RNA-sequencing (GEO accession No. GSE137327) (the gender and age of patients and the pathological characteristics of cancers are listed in Supplementary Table 3, available at Carcinogenesis Online). The result showed that the IL11 level in colon cancer was higher than its matching non-cancerous control by >0.5 Log2 fold in 6, reduced by >0.5 Log2 fold in 1 and unchanged in 1 out 9 patients (Figure 1E). Using the pairing comparison, these data display an increasing trend in the IL11 expression in colon cancer (P < 0.01; Figure 1F).

IL11 activates STAT3 in the newly regenerated crypts in colitis

Since colitis tissues are known to consist of a variety of stromal cells and few epithelial cells (16), we investigated whether colon crypt stromal and epithelial cells express IL11 and respond to IL11. We performed qPCR to analyze the cells separated by epithelial cell surface marker EpCAM. The EpCAM⁺ cells represent epithelial cells, whereas the EpCAM⁻ cells represent stromal cells, including inflammatory cells. We found that IL11 was upregulated significantly in EpCAM⁻ cells on Days 1, 3, 7, 10 and 14 after PhIP/DSS treatment and was decreased to normal after 21 days (Figure 2A and B). In the EpCAM⁺ cells, IL11 was upregulated significantly on Days 7 and 10 and then was reduced to normal after 14 days (due to the loss of crypts during colitis, there were not sufficient EpCAM⁺ cells in tissues collected on Days 1 and 3 for extracting RNA; Figure 2A). To determine the possible involvement of PhIP, we measured the levels of IL11 by ELISA in the sample treated with DSS, PhIP, or PhIP/DSS and found that PhIP neither induced IL11 nor influenced its induction by DSS (Supplementary Figure 4, available at Carcinogenesis Online).

Figure 2.

IL11 is upregulated in PhIP/DSS-induced tumours and DSS-induced colitis tissues. (A) Schematic illustration of the experimental design of PhIP/DSS-induced carcinogenesis in hCYP1A mice. (B) Expression levels of IL11 in colitis tissues and tumours were determined by qPCR using RNA extracted from EpCAM⁻ and EpCAM⁺ cells of colon epithelia collected from mice (n = 3) on 1, 3, 7, 10, 14 and 21 days after the termination of PhIP/DSS treatment. The controls (n = 3) were the same aged mice without any treatment. Tumours (n = 5) were collected from the mice on 10 weeks after PhIP/DSS treatment. Expression levels of IL11 in all samples were normalized by the corresponding GAPDH levels. (C) Colitis colons were collected on 1, 3, 10 and 14 days after DSS treatment for H&E staining and IHC staining of EpCAM, Ki67 and pSTAT3. Colons collected from the mice without treatment were used as the controls. Black bars represent 50 μm. This experiment was repeated twice with 3–4 mice on each time point and the similar results were obtained.

We further characterized the cells responding to IL11 using IHC staining of phosphorylation of STAT3 (pSTAT3), since IL11 induces pSTAT3 as its key downstream signalling event. In the H&E stained slides, we found that colon epithelium was damaged severely with an almost complete loss of crypts on Days 1 and 3 after a 5-day DSS treatment (Figure 2C). After the loss of crypts, the space was filled with stroma cells. Normal crypt structure was regenerated with an intact layer of epithelial cells on Day 10. The loss of crypt epithelial cells was also characterized and confirmed by IHC staining for epithelial cell specific markers, EpCAM (Figure 2C) and pan-cytokeratin (Supplementary Figure 5, available at Carcinogenesis Online). The IHC staining of pSTAT3 was positive in colitis tissues, mainly in stromal cells on Days 1 and 3. In the newly regenerated colon epithelium, pSTAT3 staining was positive in crypt epithelial cells featured with a stronger signal in cells close to the lumen and significantly reduced in cells at the crypt bottom. Since the crypt bottom half consists of stem cells and amplification zone cells (as characterized by the IHC staining of Ki67 in Figure 2C), the stronger pSTAT3 positive cells on Day 10 were most likely to be the differentiated epithelial cells. In contrast, pSTAT3 staining was negative in the control colon epithelial crypts and the regenerated crypts on Day 14. These data suggest that the activation of STAT3 is a transient event associated with the upregulated expression of IL11 in colitis.

To determine whether the activation of STAT3 is IL11-dependent in colitis, we crossed IL11Rα1+/− mice with hCYP1A mice and generated hCYP1A:IL11Rα1+/− mice, which produced IL11Rα1+/+, +/− and −/− offspring for our study. We collected and characterized colons from these mice on Day 1, 3, 7, 10, 14 and 21 after PhIP/DSS treatment. All mice display similar colitis-induced injury: crypts were lost on Days 1 and 3, but regenerated thereafter (Figure 3). By pSTAT3 IHC staining, we found that the colitis-induced pSTAT3 in the newly regenerated cryptic epithelial cells on Day 10 was significantly reduced in IL11Rα1+/− and −/− mice, while that in stromal cells was only reduced slightly (Figure 3). This finding was validated by the results obtained by Western-blot analysis using colitis tissues (Supplementary Figure 6, available at Carcinogenesis Online). Our results demonstrate that the colitis-induced activation of STAT3 in regenerated crypts is primarily resulted from the upregulated IL11, whereas the activation of STAT3 in stromal cells is only partially induced by IL11.

Figure 3.

The activation of STAT3 in crypt epithelial cells in colitis tissues is significantly reduced in IL11Rα1 +/− and −/− mice. Colitis tissues from the PhIP/DSS-treated hCYP1A:IL11Rα1 +/+, +/− and −/− mice were characterized by H&E staining and pSTAT3 IHC staining. Colons collected from the mice without treatment were used as the controls. Black bars represent 50 μm. This experiment was repeated twice with 3 mice on each time point and the similar results were obtained.

STAT3 is activated in the PhIP/DSS-induced earlier dysplastic lesions and tumours

Next, we performed qPCR to examine the expression of IL11 in tumours and found that IL11 was found significantly upregulated in both EpCAM⁻ and EpCAM⁺ cells (Figure 2A), demonstrating that IL11 is expressed in both tumour and stromal cells. By IHC staining of pSTAT3, we found that STAT3 was activated in tumours collected on 6, 7 and 8 weeks after DSS treatment: all tumour cells were positive for pSTAT3, while only a portion of stromal cells were positive; in contrast, the adjacent normal colon crypts were negative for pSTAT3 (Figure 4A). Similar results were found by Western blot (Supplementary Figure 7, available at Carcinogenesis Online). In a recently published AOM/DSS-induced colon carcinogenesis study, pSTAT3 was found increased in tumors and to a less extent in adjacent tissues (29). While the increased pSTAT3 in tumors is consistent with our result, the increased pSTAT3 in adjacent non-tumor tissues could be due to multiple cycles of DSS treatment which is different from our study using a single cycle of DSS treatment. Multiple cycles could increase fibrosis (30,31), and IL6 or IL11 produced by the activated myofibroblasts (32,33) may cause the activation of STAT3 in the epithelium.

Figure 4.

STAT3 is activated in the PhIP/DSS-induced tumours as well as earlier hyper-proliferative lesions. (A) The activation of STAT3 was characterized by pSTAT3 IHC staining in tumours (>2.0 mm in diameters) collected 6, 7 and 8 weeks after PhIP/DSS treatment. These are the representatives of 16 tumours randomly selected from in multiple experiments consisting of a total of 45 mice. (B) Colon collected on 7 days after PhIP/DSS treatment was characterized for newly regenerated crypt by IHC staining of Ki67, β-catenin and pSTAT3 for identifying proliferative areas, active Wnt signalling and STAT3, respectively. (C–E) Serial sections of colons, collected on 14, 17 and 21 days after the PhIP/DSS treatment, were characterized by IHC staining of Ki67, β-catenin and pSTAT3 for identifying proliferative areas, active Wnt signalling and active STAT3, respectively. Black and red bars represent 50 and 200 μm, respectively. These are the representatives of over 30 lesions in two experiments consisting of 4 male and 4 female mice on each time point.

To determine the activation of STAT3, we characterized the pSTAT3 in colons collected Days 7–21 after the PhIP/DSS treatment, along with the Ki67 and β-catenin staining for identifying the lesions with hyper-proliferation and active Wnt signalling. On Day 7, crypts were regenerated and the newly regenerated crypts displayed well-organized hierarchical structure, featured with the proliferating cells located in crypt bottom (Figure 4B). Positive nuclear staining of pSTAT3 was found in a few crypt epithelial and stromal cells, whereas most normal crypt cells were weakly positive for β-catenin staining on the peripheral membrane. On Days 14, 17 and 21, normal crypts showed the similar Ki67 and β-catenin staining patterns as that on Day 7, but the signalling intensity and numbers of positive pSTAT3-stained cells were significantly lower (Figure 4C-E). On Day 21, positive pSTAT3 staining was only found in a few stromal cells. Furthermore, through examining these samples thoroughly, we located earlier lesions such as dysplastic crypts. On Day 14, we found a few hyper-proliferative atypical crypts characterized as low grade dysplastic crypts, featured with branched crypts and irregular epithelial cells as reported previously (16). Ki67 staining was positive in most lesioned cells, and the positively stained cells were no longer restricted to the crypt bottom (Figure 4C), suggesting that the lesioned cells no longer exit from cell cycle for terminal differentiation. The staining of β-catenin was found intensified in lesioned cells, especially in the nuclei (Figure 4C). As characterized by DNA sequencing using the samples collected by lase captured microdissection, the accumulated β-catenin has resulted from dominant active mutations (16). In this lesion, most cells including stromal cells displayed stronger nuclear pSTAT3 than those in the adjacent normal crypts (Figure 4C). On Days 17 and 21, we also identified hyper-proliferative lesions using Ki67 staining (Figure 4D–E). These lesions were characterized as high grade dysplastic crypts, featured with cribriform structures and back-to-back glands, epithelial cells with an enlarged nucleus and marked reduction of inter-glandular stroma (16). Ki67 was positive in almost all lesioned cells and Ki67-positive cells were no longer restricted to crypt bottom, which is similar to the low grade dysplastic lesions on Day 14. Intensified β-catenin staining was found in all lesioned cells and nuclear localization became more prominent. Positive nuclear staining of pSTAT3 remained strong in lesioned cells, while the pSTAT3 staining in adjacent normal crypts became negative (Figure 4D–E). These results suggest that, along with the progression of the earlier low grade dysplastic crypts to tumours, the lesioned cells are not only featured with the accumulation of nuclear β-catenin and hyper-proliferation, but also the activation of STAT3.

The colitis-promoted carcinogenesis is attenuated in IL11Rα1+/− and −/− mice

To determine whether the IL11 signalling is required for DSS-induced colitis promoted carcinogenesis, we treated hCYP1A:IL11Rα1 +/+, +/− and −/− mice with PhIP and DSS (illustrated in Figure 5A) and found that colon carcinogenesis was significantly reduced in both female and male hCYP1A:IL11Rα1 +/− and −/− mice (Figure 5B). In female mice, tumour incidence was reduced from 93.3% of IL11Rα1+/+ to 86.7% of IL11Rα1+/− and 46.7% of IL11Rα1−/−, and tumour multiplicity was reduced from an average of 5.27±3.61 of IL11Rα1+/+ to 2.60±2.47 of IL11Rα1+/− and 1.27±1.94 of IL11Rα1−/− (mean±standard derivation). In male mice, tumour incidences were 100% of IL11Rα1+/+, 75% of IL11Rα1+/− and 46.7% of IL11Rα1−/−, and tumour multiplicities were 7.29±3.67 of IL11Rα1+/+, 3.50±3.41 of IL11Rα1+/− and 1.20±1.52 of IL11Rα1−/−. Similar results were also obtained in non-hCYP1A mice treated with AOM and DSS (Supplementary Figure 8, available at Carcinogenesis Online). These data suggest that the reduction or loss of IL11 signalling attenuates colitis-promoted colon carcinogenesis in mice.

Figure 5.

Loss of IL11 signalling during attenuates the PhIP/DSS-induced carcinogenesis. (A) The timeline to treat hCYP1A:IL11Rα1 +/+, +/− and −/− mice with two doses of PhIP followed by a 5 day DSS treatment. (B) Tumour numbers in the colon of female and male hCYP1A:IL11Rα1 +/+, +/− and −/− mice 8 weeks after the PhIP/DSS treatment. * and ** indicate the difference between two groups with statistical significance (ANOVA; *, P < 0.001; ** P < 0.01). This experiment was repeated once and a similar result was obtained. (C) Recombinant mouse IL11 and IL11(W147A) were used to treat human colon cancer DLD1 cells. The response was analyzed by Western blot for pSTAT3 and STAT3. The sample loading was monitored by the levels of GAPDH. (D) Inhibitory activity of IL11(W147A) was evaluated by treating DLD1 cells together with IL11. The effect of IL11(W147A) was determined by the reduction of IL111-induced pSTAT3. (E) The schedule of the PhIP/DSS treatment and the daily administration of IL11(W147A) during colitis. (F) Colons were collected and characterized for pSTAT3 using IHC staining from PhIP/DSS treated mice receiving either PBS or IL11(W147A) on 1, 5 and 10 days after DSS treatment. This experiment was repeated once with the samples collected for Western blot analysis (Supplementary Figure 12, available at Carcinogenesis Online). (G–H) Colons were collected and examined for tumorigenesis from PhIP/DSS treated mice receiving either PBS or IL11(W147A) in 8 weeks after PhIP/DSS treatment. Tumour numbers and sizes were summarized in E and F, respectively (M-male mice were used in this experiment). * indicates the difference with statistical significance between two groups (student t-test; P-value = 0.0016). This experiment was repeated once and a similar result was obtained. (I) The schedule of the PhIP/DSS treatment and the daily administration of IL11(W147A) in the last 2 weeks before collecting tumours. (J) Colon tumours were collected IL11(W147A) treated mice and characterized by the IHC staining of pSTAT3, Ki67 and cleaved-Caspase 3 (C-Casp3). Black bars represent 50 μm. This experiment was repeated once and a similar result was obtained.

In addition, we performed pSTAT3 IHC staining for tumours developed in the IL11Rα1−/− mice and found that these tumours display strong positive nuclear pSTAT3 staining (Supplementary Figure 9, available at Carcinogenesis Online), suggesting that alternative pathways can activate STAT3 and promote the development of these fewer tumours in IL11Rα1−/− mice.

Inhibition of IL11 signalling by IL11(W147A) mutein during colitis is sufficient to inhibit the PhIP/DSS-induced carcinogenesis

Although above study using the IL11Rα1 knockout mice demonstrated that the IL11 signalling is critical for the PhIP/DSS or AOM/DSS-induced colon carcinogenesis, the result from the knockout mice cannot distinguish the roles of IL11 signalling in non-cancerous tissues such as colitis from that in the advanced tumours. The critical roles of IL11 in the latter conditions were demonstrated in several published studies (26,29,32,34). Herein, we were specifically interested in whether the inhibition of IL11 signalling during colitis, a non-cancerous but cancer promoting condition, is sufficient to inhibit colitis-associated carcinogenesis.

To determine whether the inhibition of IL11 signalling during colitis is sufficient to inhibit cancer promotion by colitis, we used an IL11 mutein, IL11(W147A), which was reported to function as an antagonist (35). This mutant, carrying a single amino acid replacement W147A in the GP130-binding motif, retains the activity to bind the receptor but fails to assembly the active IL11:IL11 receptor:GP130 complex. We produced recombinant IL11(W147A) and test it in three IL11-responsive cell lines, DLD-1, HT-29 and RKO (identified from commonly used human colon cancer cell lines; Supplementary Figure 10, available at Carcinogenesis Online). While IL11 effectively activated STAT3, IL11(W147A) failed to activate STAT3 in DLD-1 (Figure 5C), HT-29 and RKO cells (Supplementary Figure 11, available at Carcinogenesis Online). When used together with IL11 to treat DLD-1 cells, IL11(W147A) inhibited the IL11-induced activation of STAT3 (Figure 5D). Therefore, consistent with previous data (35), IL11(W147A) acts as a competitive inhibitor.

We determined whether IL11(W147A) inhibits carcinogenesis by injecting the PhIP/DSS-treated mice with IL11(W147A) at a daily dose of 300 μg/kg body weight through the tail vein for two weeks, starting on the day when DSS was used (Figure 5E). PBS was used as the control. We first collected colons from mice on Day 1, 5 and 10 after DSS treatment, and found that IL11(W147A) effectively reduced pSTAT3 levels in the newly regenerated crypt epithelial cells, while pSTAT3 in the stroma also showed reduction but to a lesser extent (Figure 5F). The reduced pSTAT3 by IL11(W147) on Day 10, the earliest time point when crypts were regenerated completely, was validated by Western blot analysis (Supplementary Figure 12, available at Carcinogenesis Online). In 7 weeks, we found that the treatment of IL11(W147A) during colitis significantly reduced the tumour incidence from 100% to 67% and tumour multiplicity from 4.29±1.80 to 0.67±5.20 (Figure 5G). However, the average tumour sizes were not changed (Figure 5H), suggesting that tumours escaped from IL11 inhibition presumably through an alternative pathway. This result demonstrates that the inhibition of IL11 signalling during colitis significantly inhibits colitis-promoted colon carcinogenesis.

We also determined whether targeting IL11 signalling in tumours is an effective therapeutic strategy by injecting the PhIP/DSS-treated mice with IL11(W147A) in the last 2 weeks before collecting tumours (Figure 5I). We found that IL11(W147A) reduced the pSTAT3 to different levels among the tumours examined (Figure 5J). Using the samples with reduced pSTAT3, we further characterized cell proliferation and apoptosis by IHC staining of Ki67 and cleaved-Caspase 3, respectively. While we did not see any effect on cell proliferation (data not shown), we found that apoptosis, as indicated by positive cleaved-Caspase 3 staining, was increased in some area in a tumour in mouse No. 3, in which pSTAT3 was reduced the most (Figure 5J). However, increased apoptosis was not found in the other tumours (a tumour in mouse No. 1 was shown as an example). This data suggested extensive inhibition of IL11 triggered apoptosis in tumours.

Discussion

Here we demonstrated that IL11 is a colitis-induced cancer promoting factor. In colitis, IL11 is first upregulated in stromal cells and then in the regenerated crypt epithelial cells, and the activation of STAT3 in regenerated epithelial cells is primarily activated by IL11. After the regeneration, IL11 is reduced to basal levels. Indeed, IL11 has also been reported to be expressed in an inflammation-associated fibroblast cells in ulcerative colitis (UC) tissues (36). Since the administration of recombinant IL11 improves intestinal tissue repair in a variety of murine injury models (37–45), colitis-induced IL11 could promote the proliferation of intestinal cells for the regeneration. However, such action could also activate dormant premalignant cells, which creates a cancer-promoting condition through activating STAT3. In fact, we found that inhibition of IL11 signal during colitis is sufficient to attenuate colitis-promoted colon carcinogenesis.

Colitis constitutes of the crosstalk among a variety of cells in the mucosa by means of direct contacts and autocrine/paracrine regulations through growth factors, cytokines and chemokines as well as the interactions with gut microorganisms (11,12). In addition to producing reactive oxygen and nitrogen species and causing DNA damage, inflammation produces cellular regulatory signals activating quiescent premalignant cells and promoting tumour development (11,12). Cytokines activating NFkB, STAT3 and AP-1 in premalignant cells have been suggested to promote cancer through regulating genes that increase cell proliferation and inhibit apoptosis (11,12,46). Excessive STAT3 activation is associated with the poor survival of the patients (29,47). The importance of STAT3 in mediating cancer promotion is demonstrated by the reduced AOM/DSS-induced tumour incidence and multiplicity in the intestinal-specific STAT3 knockout mice and the enhanced carcinogenesis in the mice expressing dominant active STAT3 (29,48). While IL6 had been proposed to be the cytokine activating STAT3 in CRC (11,12,49,50), significantly positive association was found with the expression levels of IL11 but not IL6 (26). Consistently, we did not find a significant increase of IL6 in TCGA CRC database. Although the increased IL6 and activation of STAT3 are well-recognized in colitis (50,51), we found that IL11 is the only IL6 family member upregulated in DSS-induced colitis. In spite of this discrepancy, our finding is in line with the recent finding of an UC-specific subset of fibroblasts expressing IL11 in UC patients (36). Together, these data support that IL11/STAT3 is an important signal in colon pathogenesis including colitis and tumorigenesis.

Several published studies demonstrated the importance of IL11 in the development and progression of gastrointestinal cancer. Ernst's group first reported that IL-11 was increased in gastric tumours developed in GP130(Y757F/Y757F) mice, and IL11Rα1 knockout in this mice prevented STAT3 activation and inhibited the development of gastric tumours (34). Using IL6 and IL11 knockout mice, they further demonstrated that IL11 is predominant factor activating STAT3 in gastrointestinal tumours and the loss of IL11 reduced tumorigenesis (26). They also generated reciprocal bone marrow chimeric mice and demonstrated that colon tumorigenesis is independent of IL11-responsive hematopoietic cells. Calon et al. demonstrated that IL11, produced by TGFβ-stimulated fibroblasts, promotes colon cancer metastasis (32). Heichler et al revealed that IL6/IL11-activated STAT3 in cancer-associated fibroblasts (CAF) plays critical roles in the development of CRC and is inversely associated with the survival of CRC patients (29). The promotion of angiogenesis was found to be an important mechanism (29). Herein, we demonstrated that IL11 is an important colitis-induced factor producing cancer promoting microenvironment. Together, these data suggest that IL11 plays important roles in multiple steps of colon cancer development and progression: constituting cancer promoting microenvironment, maintaining tumour development, promoting angiogenesis and establishing metastasis niche.

Heichler's study on CAF (29) also underlined the importance of the IL6/IL11/STAT3 signal in stromal cells. In fact, IL-11 can be produced by not only cancer cells but also fibroblast, CAFs and myeloid cells (32,34,52,53). In DSS-induced colitis, an extensive loss of crypt epithelial cells produces stromal/inflammatory cell-enriched epithelium (16). On one hand, these stromal/inflammatory cells may promote tissue repair and regeneration (37–45); on the other hand, these cells create a cancer-promoting microenvironment. Our data indicate that IL11 produced by the non-epithelial source is an earlier event than that by crypt epithelial cells. Indeed, a subset of fibroblast cells that express IL11 is suggested to be a key inflammatory hub in colon epithelium of UC patients (36). Since IL11 is also expressed in tumour stromal cells, it is possible that IL11-expressing stromal cells extend their role from activating premalignant cells to sustaining tumour growth by becoming an important component of tumour stroma. Therefore, non-epithelial source of IL11 in colitis needs to be identified. To determine the expression levels in different cell populations separated using fluorescence-activated cell sorting with different cell surface markers would be informative.

Putative STAT3 binding sites are identified in the promoter of IL11 of human and mouse, and the promoter can be activated by IL6 through STAT3 binding site (34), suggesting that IL11/STAT3 signal can constitute a positive feedback regulation of IL11 expression. Using a reporter assay, we found that the STAT3-mediated activation of IL11 promoter can be enhanced by active β-catenin (Supplementary Figure 13, available at Carcinogenesis Online). So it is highly likely that a positive feedback mechanism of IL11 expression, which is initially produced in stromal cells and then also in newly regenerated epithelial cells during colitis, is strengthened in premalignant and tumour cells that carry an active Wnt signal to maintain higher levels of IL11 expression, resulting in excessive activation of STAT3.

In this study, we identified ~1,500 genes differentially expressed in colitis tissues (the data is available in the processed data file deposited in NCBI GEO with accession No. GSE137387). Among them, there are 18 genes consistently regulated in colitis tissues and tumours in mice and CRC in human. In addition to IL11, the other 17 genes could be involved in cancer promoting too. Among the upregulated genes, Tgfbi, Psat1, Sl4a11, Srpx2 and Mmp7 promote cancer cell proliferation and invasion (54–60). The expression levels of Tgfbi, Slc4a11 and Jam2 are associated with poor prognosis in CRC patients (56,61,62). Pi16 and Mmp7 are well-recognized tissue remodelling factors and involved in inflammation (59,60,63,64). Mmp7 expression can be induced by active WNT signal resulted from APC mutation in colon cancer cells (65). Trib3 interacts with β-catenin/TCF4 to increase expression of genes associated with cancer stem cells and tumorigenesis (66). Igf1 is one of the well-studied regulatory signal acting as a driving force in CRC (67). Among the downregulated genes, Sepp1, B3gnt7, Pdcd4 and Bmp3 are reported to carry the tumour suppressor activity (68–71) whereas Trpm6 is found downregulated in CRC but has no defined function (72). The roles of Sult1a1, Synpo and Usp2 in cancer are not defined yet. Overall, most of these genes are associated with CRC; however, their roles in colitis-associated cancer promotion need to be further studied.

In summary, we identify that IL11, a pleiotropic factor involved in hematopoiesis, adipogenesis and female fertility, is an important colitis-induced cancer promoting factor. Together with previous studies establishing the importance of IL11/STAT3 signal in gastrointestinal cancer (26,29,32,34), our results support that IL11 plays critical roles in multiple steps of the development and progression of CRC.

IL11 was found upregulated in DSS-induced colitis tissue and the inhibition of IL11 signaling during colitis was sufficient to attenuate colitis-promoted carcinogenesis, demonstrating that colitis-induced IL11 creates cancer-promoting microenvironment to activate dormant premalignant cells.

Glossary

Abbreviations

AOM

azoxymethane

CRC

colorectal cancer

DSS

dextran sulfate sodium

GAPDH

glyceraldehyde 3-phosphate dehydrogenase

hCYP1A

CYP1A-humanized

HPA

human protein atlas

IHC

immunohistochemical

PhIP

2-amino-1-methyl-6-phenylimidazo[4,5-b]-pyridine

pSTAT3

phosphorylation of STAT3

TCGA

the cancer genome atlas

Funding

National Institutes of Health Cancer Center Support Grant (P30CA072720to H.W.); John L. Colaizzi Chair Endowment Fund (to C.S.Y.).

Acknowledgements

This study was supported by a pilot project award from the Rutgers Cancer Institute of New Jersey. We thank the Biospecimen Repository and Histopathology Service of the Rutgers Cancer Institute of New Jersey for providing the de-identified frozen human colon cancer samples in accordance with the IRB approval and Diane Hanrahan for assistance and support in this study. We also thank Dr. Lanjing Zhang and William Yan for critically reading the manuscript and many other members of Dr. Chung S.Yang's group for helpful discussion.

Conflict of Interest Statement: None declared.