The MicroRNA miR-223 constrains colitis-associated tumorigenesis by limiting myeloid cell infiltration and chemokine expression

Abstract

Aberrant intestinal inflammation plays a critical role in the development of colitis-associated colorectal cancer (CAC), yet the mechanisms controlling tumor development by the myeloid immune compartment are not fully understood. While altered microRNA (miRNA) expression is observed in CAC, it is also unclear how myeloid-specific microRNA’s impact on the inflammatory process that underpins the continuum from ulcerative colitis (UC) to tumorigenesis. Here, we report that miR-223 acts to limit myeloid-driven inflammation in the Azoxymethane-Dextran Sodium Sulfate (AOM-DSS) model of CAC in mice. In this model, miR-223^−/y mice present with significantly larger tumors with an enhanced proliferative signature. Immunoprofiling showed that miR-223^−/y mice have significantly increased colonic myeloid immune infiltrate (neutrophils, monocytes and macrophage) following AOM-DSS. This was accompanied by an increased inflammatory chemokine and cytokine signature for monocytes and neutrophils. Bone marrow chimera studies demonstrate that myeloid-expressed miR-223 is responsible for the enhanced tumor proliferation and inflammatory response. RNA sequencing identified several pathways that could be contributing to the development of CAC in miR-223^−/y mice, including the IL-6-IL-17a cytokine family and STAT3 signalling. Lastly, neutrophil depletion with an anti-GR1 antibody (Ly6G/Ly6C) during the initial phase of the AOM-DSS model reduced the tumor burden in miR-223^−/y mice. Collectively, our data indicates that miR-223 is an important regulator of mucosal inflammation and acts to constrain the progression from UC to CAC by limiting myeloid-associated inflammation.

Introduction

Colorectal cancer (CRC) is the third most frequently diagnosed cancer worldwide, with approximately 1.9 million new cases diagnosed annually and incidence rates are continuing to increase. Globally, CRC is also the second leading cause of cancer related deaths(1). A major risk factor for the development of CRC is a diagnosis of an inflammatory bowel disease (IBD), in particular ulcerative colitis (UC). This disease is characterized by aberrant intestinal inflammation, which results in colonic epithelial cells becoming damaged overtime, and can eventually lead to colitis associated colorectal cancer (CAC). Additionally, UC patients with increased duration and extent of disease have a heightened risk for developing CAC(2, 3). This further highlights the critical role of inflammation in both the initiation and the progression of CAC.

Myeloid associated inflammation is emerging as a critical driver of tumorgienesis(4). Under normal physiological conditions the crosstalk between myeloid and intestinal epithelial cells acts to maintain the integrity of the epithelial barrier(5). However, during chronic inflammation this process is disrupted, and the epithelial barrier is often irreparably damaged. Consequently, this further promotes a pro-tumorigenic enviroment(4, 6). In CAC, chronic inflammation is also a major contributor to the tumor microenvironment (TME) once a tumor has been established. The TME consists of a host of different cell types, which together with their associated cytokines, metabolites (i.e. lactic acid) and growth factors, interact with the surrounding tissue to enhance tumor growth(7). Among these cells are tumor associated macrophage (TAMs) and neutrophils. TAM’s are the largest myeloid cell type in the TME(8), where they can be polarised by environmental cues to differentiate into mature macrophages that can be broadly characterized as either pro- or anti-inflammatory. Pro-inflammatory macrophages have traditionally been associated with anti-tumor activity, while anti-inflammatory macrophages have been shown to promote tumor growth(9). Recent evidence has suggested that while macrophage can alter their phenotype in response to signals from their local environment, the majority of TAMs within the TME act to promote tumorigenesis(9, 10). Thus, identifying mechanisms that both control excessive myeloid inflammation and the regulation of TAM function are of therapeutic interest.

MicroRNAs (miRNAs) are small non-coding RNA molecules, approximately 18–22 nucleotides in length, that are essential for the post-transcriptional regulation of gene expression. This regulation is primarily executed via complementary base pairing to the 3’ untranslated region (UTRs) of their target mRNA(11). As such, miRNAs can be critical for the regulation of several biological processes, including inflammation. Additionally, the aberrant expression of miRNAs has previously been reported separately in UC, CAC, and TAMs(12–14)

MiR-223 has emerged as key regulator of the innate immune system during intestinal inflammation. This miRNA is specifically expressed in the myeloid compartment, is induced during myeloid differentiation and is essential for regulating myeloid cell activity(15). Additionally, miR-223–3p expression is increased in patients with active UC and miR-223^−/y mice have exacerbated colitis that is driven by inflammatory monocytes and neutrophils(16). In CRC, miR-223 in increased in patient biopsies and correlates with more advanced disease(17). Yet despite this correlation, a direct role for myeloid-expressed miR-223 in the progression from UC to CAC remains to be characterized.

In this study, we investigated the role of miR-223 in the development of CAC, using the chronic azoxymethane (AOM)-dextran Sodium Sulfate (DSS) mouse model of CAC. We show that miR-223^−/y mice have increased susceptibility to CAC, with an increased proliferative signature, larger tumor burden and a significant infiltration of monocytes and mixed cytokine profiles. RNA sequencing on miR-223^−/y colonic tissue also showed that there is enhanced expression of several pathways previously identified to be central to the progression from UC to CAC, particularly expression of IL-6 and IL-17a cytokine family members and STAT3 signaling. Collectively, these data implicate that enhanced early myeloid inflammation plays a key role in the onset of experimental CAC and that miR-223 constrains the progression to CAC.

Materials & Methods

Mouse strains

miR-223^−/y mice on a C57BL/6J background were previously describes(16) and bred in house at the University of Colorado Anschutz Medical Campus. C57BL/6J mice were purchased from The Jackson Laboratory. After weaning (21 days) mice were co-housed and experimental colitis induced at 8 weeks-of-age. All mice were bred under specific-pathogen free conditions and faecal samples tested negative for Helminth, Protozoa and Heliobacter species. Animal procedures were approved by the Institutional Animal Care and Use Committees at the University of Colorado Denver.

Bone marrow chimera

Bone marrow chimeric mice were generated as previously describes(16, 18). Briefly, femurs and tibias from C57BL/6J (WT) and miR-223^−/y donor mice were removed and flushed through a 70-μm cell strainer with cold RPMI 1640 to harvest BM cells. Recipient mice (8–12 wks of age) were irradiated with a total dose of 1,000 rads (500 rads twice, 4h apart). Immediately after irradiation, 1 × 10⁷ BM cells/mouse were intravenously injected via the retro-orbital route in 0.1 ml 0.9% sodium chloride. Mice were housed in microisolator cages for 8 wks to recover before induction of the AOM-DSS colitis-associate tumorigenesis. To assess BM reconstitution, spleens were excised at necropsy, and red blood cells were lysed, followed by flow cytometry for CD45⁺ leukocytes expressing either CD45.1 (clone A20) or CD45.2 (clone 104; eBioscience).

Induction of experimental colitis-associated colorectal cancer (AOM-DSS) and antibody depletion studies

8 week old mice from indicated genotypes were injected intraperitoneally with 10mg/kg of azoxymethane (AOM) (Sigma Aldrich). 3 days later and to assess survival in one study (supplemental Figure 1A), mice were administered dextran sodium sulphate (DSS) (3% w/v, 36–50,000 kDa; MP Biomedicals) in drinking water ad libitum for 6 days. All mice were then returned to drinking water ad libitum and monitored for 50 days for survival. For all remaining in vivo studies, mice received dextran sodium sulphate (DSS) (2% w/v, 36–50,000 kDa; MP Biomedicals), administered in drinking water ad libitum for 6 days. This was followed by a two week recovery period where DSS was removed (drinking water only). This cycle of DSS was performed three times (chronic AOM-DSS). Upon necropsy, colons were removed and flushed with cold PBS. Their length and weight was recorded. Colon tissue was either snap-frozen or fixed in 10% buffered formalin.

To assess the effects of neutrophils and myeloid derived suppressor cells on the development of CAC, an acute phase model of AOM-DSS was performed where both cell types were depleted. Here, one round of AOM (i.p.; 10mg/kg) was administered and followed 3 days later by DSS (2% w/v) for 6 days, as above. This was then followed by a two week recovery period where DSS was removed (drinking water only). Next, miR-223^−/y mice were randomly assigned to 2 groups of either IgG control or a treatment group with 20μg (i.p) anti-GR1 (Lyg6/Ly6C: RB6–8C5 clone, BioXcell, USA), every 3 days for 3 weeks (8 total injections). All mice were euthanized 24h following the last injection.

Colon pathology and Immunohistochemistry

Formalin-fixed and paraffin embedded (FFPE) tissue was cut into 7μm sections and baked at 60°C for 24hrs. Slides were deparaffinized by xylene and rehydrated using an ethanol gradient (100 – 70%). Samples were placed in hematoxylin for 4 minutes, followed by washing with water for 1 minute. Slides were immersed in an acid wash for 1 minute and then rinsed with water for 1 minute. A bluing reagent was added to slides for 1 minute before rinsing for 30 seconds with water. Slides where placed in eosin for 15 seconds and then washed in water until clear. Slides were dehydrated via sequential immersion in ethanol (70 – 100%) and xylene, then a coverslip was added. Samples were imaged using a Motic Panthera L Series microscope.

Immunohistochemistry

FFPE tissues were paraffin embedded cut into 7μm sections and deparaffinized as above. Antigen retrieval was performed using a citrate-based solution in a pressure cooker for 10 minutes at 6 psi, slides were cooled with water and then washed with PBS for 5 minutes. 0.2% Trition X was added to slides for 10 minutes, followed by 5% BSA/PBS for 20 minutes. Samples were then incubated overnight at 4°C with primary antibodies; Rabbit anti-mouse b-catenin (610154) from BD Bioscience; Rabbit anti-mouse STAT3 (ab68153, Abcam), Rabbit anti-mouse CD68 (97778, Cell Signaling), Rabbit anti-mouse Ly6G (87048, Cell Signaling), Rabbit anti-mouse PCNA (13110, Cell Signalling), Rabbit anti-mouse CCR2 (ab27305, Abcam). Slides were washed in PBS and secondary antibody (Vector ABC kit) was added as per manufactures instructions (Vector Labs). Staining was detected using ImmPACT NovaRed as per manufactures instructions (Vector Labs). Slides were washed in PBS and then counterstained with hematoxylin for 1 minute. Slides were dehydrated via sequential immersion in ethanol (70 – 100%) and xylene. Samples were imaged using either a Motic Panthera L Series or Invitrogen Evos M5000 microscope.

Histology Scoring and quantification

H&E images were used to determine the number of tumors present in each sample. Motic Images Plus Software was also used to determine the area of each tumor, this was used to classify tumors into either: dysplastic, <100,000μm², small 100,000 – 300,000μm², medium 300,000 – 500,000μm², large 500,000–1,000,000μm² or very large tumor >1,000,000μm². Immunostaining was quantified using QuPath software (version 0.3.1). Images were imported, image type was set to brightfield (H-DAB) and stain vectors were set using the automatic detection function. Staining was detected based on optical density sum. The number of positive cells was recorded for each image. Statistical analysis was performed using GraphPad Prism.

Intestinal lamina propria cell isolation and Flow Cytometry

Resected colon tissues were washed with cold phosphate-buffered saline, mesenteric fat excised and the tissue opened longitudinally. Intestines were then cut into 2 cm pieces, washed with cold PBS and vortexed in IEL solution (PBS/15 mmol/L HEPES/1 mmol/L EDTA) to remove epithelial cells. Following a PBS wash, tissues were enzymatically digested (RPMI, 5% FBS, 10 mmol/L HEPES, 0.07 mg/ml liberase (Roche), 10 μg/ml DNAse I (Roche)] at 37°C with gentle agitation for ~45min. Debris was extracted using a 100μm filter and the cell suspension was washed with cold RPMI/10%FBS twice. Single-cell suspensions of lamina propria mononuclear cells were were blocked with1μg/μl of Fc block (CD16/32; eBiosciences, CA) and then stained for 2 hours at 4 degrees with anti-mouse antibodies including; Ly6G (1A8), Ly6C (AL-21), SiglecF (E50–2440) (BD-Biosciences, CA); MHC-II (M5/114.15.2), CD19 (6D5), CD103 (3E7), FoxP3 (FJK-16s), CX3CR1 (SA011F11) (Biolegend, CA) and CD4 (RM4–5), CD8 (53–6.7), CD11b (M1/70), CD11c (N418), CD45 (30-f11), CD64 (X54–5/7.1) (eBiosciences, CA). Viable cells were gated up using AquaVi Live/Dead staining (Invitrogen, NY). Flow cytometric analysis was performed using a BD FACSCanto^™ II (BD Biosciences, CA). Data files were further analyzed using FLOWJo software (Tree Star Inc, Ashland, OR, USA).

microRNA and mRNA transcriptional analysis by RT-PCR

Total RNA or separated fractions of miRNA and mRNA were isolated from snap-frozen colonic tissue using QIAzol Reagent and RNeasy kit following manufactures instructions (Qiagen). Concentration of extracted RNA was determined by using a NanoPhotometer N60 and then samples were equalized. RNA was then transcribed into cDNA using the High-Capacity cDNA Reverse Transcription Kit (Thermofisher) or miRCURY LNA RT Kit (Qiagen) as per manufactures instructions. Transcript levels were determined by real-time RT-PCR using specific primer sets (Table 1. miRNA: SyberGreen, Qiagen; mRNA: TaqMan, Applied Biosystems). Expression levels of mRNA were normalized to endogenous 18S, while RNU-6 was used to normalize miRNA expression profiles.

Table 1.

Primer probe list

TaqMan Primers (Applied Biosystems by Thermo Fisher Scientific)
Target	Cat no.
Acida (AID)	Mm01184115_m1
Arg1	Mm00475988_m1
CD274 (pdl1)	Mm03048248_m1
CD273 (pdl2)	Mm00451734_m1
Ccl2	Mm00441242_m1
Ccl7	Mm00443113_m1
Chi3l3	Mm00657889_m1
Csf1	Mm00432686_m1
Csf2	Mm01290062_m1
Cxcl1	Mm04207460_m1
Cxcl2	Mm00436450_m1
Ifny	Mm01168134_m1
Il-11	Mm00434162_m1
Il-17a	Mm00439618_m1
Il-17f	Mm00521423_m1
Il-1b	Mm00434228_m1
Il-1rn	Mm00446186_m1
IL-21	Mm00517640_m1
IL-22	Mm01226722_m1
Il-4	Mm00445259_m1
Il-6	Mm00446190_m1
Irf4	Mm00516431_m1
Irf5	Mm00496477_m1
Lif	Mm00434762_m1
Nos2	Mm00440502_m1
Relna	Mm00445109_m1
Tgf-b1	Mm00441727_m1
Tnf	Mm00443258_m1
18S	Hs99999901_a1
Syber Green Primers (miRCURY miRNA Assay by Qiagen)
Target	Cat no.
Has-miR-223–3p	YP00205986
U6 snRNA	YP00203907

Western immunoblot analysis

To measure protein content, colonic biopsies were freeze-thawed in T-Per Tissue Protein Extraction Reagent with Phosphatase and Protease Inhibitors (Abcam). Samples were then processed using TissueLyser LT as per manufactures instructions (Qiagen). Supernatant protein concentrations were quantified using Pierce BCA Protein Assay Kit (Thermo Scientific), resuspended in reducing 4x Laemmli sample buffer (Bio-Rad) and heated to 98°C for 10 min. Samples were resolved on a 8%, 10% or 12% polyacrylamide gel and semi-dry transferred to nitrocellulose membranes. The membranes were blocked for 30 mins at room temperature in TBS-T supplemented with 5% non-fat milk or 5% BSA. Membranes were subsequently incubated with, Rabbit anti-mouse NF-kappaBp65 (8242), Rabbit anti-mouse Cyclin D1 (55506), Rabbit anti-phospho-STAT3 (Tyr705) (9145), Rabbit anti-c-Myc (18583), Rabbit anti-PCNA (13110), Rabbit anti-Phospho-S6 Ribosomal Protein (Ser235/236) (4858), Rabbit anti-Phospho-AKT (Ser473) (4060), Rabbit anti-B-catenin Ser33/37/Thr41 (9561) and Rabbit anti-B-catenin Ser552 (5651) from Cell Signalling Technology; Rabbit anti-STAT3 (AB68153) from Abcam or anti-TGFa (nbp2–34296) from Novus Biologicals. All primary antibodies were incubated overnight at 4°C. Membranes were then washed three times in TBS-T for 15 minutes. Secondary antibodies including, goat anti-rabbit IgG-HRP (AP307), goat anti-mouse IgG-HRP (AP308) or donkey anti-goat IgG-HRP (AP180) from Sigma Aldrich were incubated for 1hr at room temperature. Membranes were washed three times in TBS-T for 15 minutes and proteins were then detected by enhanced chemiluminescence (Thermo Fisher Scientific). B-actin was used as a loading control with anti-mouse b-actin (sc-1616) from Santa Cruz Biotech and donkey anti-goat IgG-HRP (AP180) from Sigma Aldrich.

Bulk mRNA sequencing and bioinformatic analysis

Colon samples were stored in RNAlater (Ambion) at −80 °C and total RNA was isolated using the RNeasy kit (Qiagen). Poly-A mRNA sequencing was performed by Novogene. Messenger RNA was purified from total RNA using poly-T oligo-attached magnetic beads. First strand cDNA was synthesized using random hexamer primers, second strand cDNA was synthesized using either dUTP for directional library or dTTP for non-directional library. The library was checked with Qubit and real-time PCR for quantification and bioanalyzer for size distribution detection. Quantified libraries were be pooled and sequenced on illumina platform (polyA mRNA sequencing, with 40M PE150 reads). Quality control: Raw data in fastq format was processed by removing reads containing adapter, ploy-N and low quality reads from raw data. Q20, Q30 and GC content of the clean data was determined. Clean data with high quality was used for all downstream analyses. FeatureCounts v1.5.0-p3 was used to calculate the reads numbers mapped to each gene. Reads mapping to the reference genome: Reference genome and gene model annotation files were downloaded from genome website directly. Index of the reference genome was built using Hisat2 v2.0.5 and paired-end clean reads were aligned to the reference genome using Hisat2 v2.0.5 mapping tool. Quantification of gene expression level: FeatureCounts v1.5.0-p3 was used to count the reads numbers mapped to each gene. And then FPKM (fragments Per Kilobase of transcript sequence per Millions base pairs sequenced) of each gene was calculated based on the length of the gene and reads count mapped to this gene. Differential expression and enrichment analysis: DESeq2 R package (1.20.0) was used to determine the differential expression of two conditions/groups using a model based on the negative binomial distribution. The p-values that were generated were adjusted using the Benjamini and Hochberg’s approach for controlling the false discovery rate. Genes with adjusted P-value <=0.05 were classified as being differentially expressed. To further analyse differently expressed genes the clusterProfiler R package was used to perform Gene Ontology (GO) enrichment, with correction for gene length bias. The KEGG database (http://www.genome.jp/kegg/) was used to identify biological system involved and clusterProfiler R package tested the statistical enrichment of differential expression genes in KEGG pathways. The DO (Disease Ontology) database was used to identify disease association and The DisGeNET database integrated human disease-related genes. ClusterProfiler software was used to test the statistical enrichment of differentially expressed genes in the Reactome pathway, the DO pathway, and the DisGeNET pathway. GO, KEGG, Reactome, DO and DisGeNET data sets were used for independent Gene Set Enrichment Analysis (GSEA) (http://www.broadinstitute.org/gsea/index.jsp).

Statistical Analysis

Outside of RNASeq analysis all other statistical analyses were performed using ANOVA or a two-tailed Student’s t-test using GraphPad Prism 10.1 software. Data is expressed as mean ± standard error of the mean (SEM). Statistical significance was set at p<0.05 or padj<0.05.

Results

Increased tumor size in miR-223^−/y mice following experimental CAC.

The miRNA miR-223 is increased in biopsies from patients with UC during active inflammation and correlates with more severe disease in CAC patients(16), but its role in the progression from UC to CAC is yet to be elucidated. To determine the role of miR-223 in the development of intestinal tumorigenesis associated with UC, an AOM-DSS model of CAC was utilized. Notably, our initial studies using a higher percentage of DSS (3%w/v) highlighted a central role for miR-223 in constraining excessive inflammation and tissue destruction during the treatment course, with 66% of mice dying over the length of the protocol in miR-223^−/y mice (Supplemental Fig 1A). Tapering this dosing regimen to a lower DSS-insult (2%w/v) for further chronic studies, we performed a series of experiments with 3-repeated rounds of DSS and recovery following an initial AOM-insult. This model mimics the recurring periods of active colitic flare and then recovery and remodeling in patients with Ulcerative Colitis. We assessed spleen weight, colon weight and colon length (Supplemental Fig 1 B–D). These serve as markers of systemic inflammation, gross tumor burden and lingering physiologic implications of colitis induced by DSS, respectively. Differences between treatment groups with this initial analysis was negligible. Notably, when the number of macroscopically visible tumors was also assessed (Fig 1A) and dysplasia and tumor size was determined by microscopy, our data indicates that miR-223^−/y mice developed more frequent and larger tumors when compared to WT counterparts (Fig 1 C–F). This suggests that miR-223 acts to constrain inflammatory driven progression from UC to CAC in the AOM-DSS model.

Figure 1. Increased tumor size and proliferative capacity in miR-223^−/y mice following AOM-DSS.

Schematic overview and timeline of the AOM-DSS study. A. number of macroscopic polyps per colon. Following H&E staining B. the number of tumors per section, C. quantification of dysplasia (μm²) per section, D. quantification of tumor size (μm²) per section was determined at the time of necropsy following treatment with AOM-DSS. E. Tumor grading according to size from H&E micrographs in WT or miR-223^−/y mice treated with chronic AOM/DSS F. Representative 20x H&E micrograph of a tumor from WT or miR-223^−/Y mice treated with chronic AOM/DSS. All data is expressed as mean ± SEM; *, P<0.05; **, P<0.01; ***, P<0.001 versus the indicated counterpart (Unpaired, one-tailed students t-test). C + D: WT AOM-DSS: 16 tumors and 2 dysplastic regions counted from 17 mice; miR-223^−/Y AOM-DSS: 27 tumors and 9 dysplastic regions counted across 7 mice. Data is presented from 2 independent experiments. G. Western immunoblot assessment of PCNA, cyclin-D1, c-Myc, p-AKT^(Ser473), p-S6^(Ser235/236) and β-actin in colon biopsies (n=3 mice/group). H. Quantification of western immunoblots for PCNA, Cyclin D1 and c-Myc and p-AKT. All data is expressed as mean ± SEM; *, P<0.05; **, P < 0.01; ***, P<0.001 versus the indicated counterpart (One-way ANOVA). n=3 mice/group.

Enhanced proliferation in colon tissues of miR-223^−/y mice following AOM-DSS

As miR-223^−/y mice had significantly larger tumors compared to WT when treated with chronic AOM/DSS, we hypothesized that miR-223 deficiency results in enhanced proliferation. To test this, colonic tissue from AOM-DSS treated mice was analyzed for the expression of proliferation markers via western immunoblotting. Firstly, WT and miR-223^−/y mice did not express a marked increase in the proliferation markers assessed at baseline. In contrast, miR-223^−/y mice presented with significantly higher expression of proliferating cell nuclear antigen (PCNA) and cyclin-D1 and c-Myc compared to WT controls (Fig 1 G&H). An increase in other proliferative markers such as phospho-AKT (Ser473) was replicated in the tumors of both genotypes compared to non-treated controls (Fig 1G&H). Of note, this increased proliferative signature appears to be a selective as other markers associated with tumor growth, nutrient sensing and proliferation were unchanged across AOM-DSS genotypes, namely the mTOR signaling phospho-ribosomal S6 (Ser235/236) (Fig 1G) or NF-kB (Supplemental Fig 3B). This was also evident for β-catenin isoforms which failed to show differences in expression across genotypes (as measured by western immunoblot and immunohistochemistry) (Sup Fig. 3A&B). Collectively, these data indicates that miR-223 plays a role in controlling the proliferation capacity of the AOM-DSS-derived tumors.

miR-223^−/y regulates the infiltration of CCR2⁺ monocytes and associated inflammatory cytokine profile following AOM-DSS

To identify what cells are potentially implicated in the increased tumorigenic and proliferative signature observed in the miR-223^−/y mice, we performed immunohistochemical and RT-PCR analysis on colon samples from AOM-DSS treated mice. Immunostaining of CCR2⁺ monocytes, CD68⁺ macrophage, CD206+ “M2-like” macrophage and Ly6G⁺ neutrophils were equivalent in non-inflamed WT and miR-223^−/y colons at baseline (Supplemental Fig 2). Following chronic AOM-DSS protocol, quantification of immunostaining on unaffected tissue showed that significantly more CCR2⁺ monocytes, CD68⁺ macrophage and Ly6G⁺ neutrophils infiltrate into the colons of miR-223^−/y mice compared to WT (Fig 2 A–H). Similar numbers of CD68⁺ macrophage and Ly6G⁺ neutrophils where observed in the tumors across groups (Fig 2. A–D). However, there were significantly more CCR2⁺ monocytes detected within the tumors of the miR-223^−/y mice compared to the WT when treated with AOM-DSS (Fig 2. E–F). Of note, we observed an increase in miR-223 expression in the colons of WT mice following AOM-DSS (Fig 3A). While not conclusive, given that miR-223 was described by our work and others as predominantly expressed from myeloid progenitors in the bone marrow(15, 16, 19, 20), we hypothesised that this expression represents the influx of new myeloid cells (eg. neutrophil and monocytes) into the colon.

Figure 2. Macrophage, neutrophils, and monocytes infiltrate into the colon of miR-223^−/y mice treated with chronic AOM-DSS.

A. Representative 20x micrographs of unaffected tissue and tumor of WT and miR-223^−/y mice treated with chronic AOM-DSS after immunohistochemical staining for Ly6G B. Quantification of Ly6G immunostaining. C. Representative 20x micrographs of unaffected tissue and tumor of WT and miR-223^−/y mice treated with chronic AOM/DSS after immunohistochemical staining for CD68 D. Quantification of CD68 immunostaining. E. Representative 20x micrographs of unaffected tissue and tumor of WT and miR-223^−/y mice treated with chronic AOM-DSS after immunohistochemical staining for CD206 F. Quantification of CD206 immunostaining. E. Representative 20x micrographs of healthy tissue and tumors WT and miR-223^−/y mice treated with chronic AOM-DSS after immunohistochemical staining for CCR2 F. Quantification of CCR2 immunostaining. All data is expressed as mean ± SEM; *, P<0.05; **, P<0.01; ***, P<0.001 versus the indicated counterpart (One-way ANOVA). A+B. WT AOM-DSS n=16; miR-223^−/Y AOM-DSS n=4; C+D. WT AOM-DSS n=14; miR-223^−/Y AOM-DSS n=4; E+F. WT AOM-DSS n=11; miR-223^−/Y AOM-DSS n=4; G+H. WT AOM-DSS n=12; miR-223^−/Y AOM-DSS n=4; with 3 fields of view quantified per animal.

Figure 3. miR-223^−/y mice have increased chemokine and cytokine signature for monocytes and neutrophils following AOM-DSS.

A. Relative mRNA expression of miR-223–3p. B. Relative mRNA expression of the pro-inflammatory macrophage markers nos2, tnf, il-1β and S100a8. C. Relative mRNA expression of the anti-inflammatory macrophage markers arg1, chi3l3, retnla, il-4, and irf4. D. Relative mRNA expression of the monocyte chemokines ccl2 and ccl7; neutrophil chemokines cxcl1 and cxcl2. E. Relative mRNA expression of the myeloid growth factors m-csf1, il-34 and gm-csf. All data is expressed as mean ± SEM; *, P<0.05; **, P<0.01; ***, P<0.001 versus the indicated counterpart (One-way anova). A. n=5–6 mice/group, B-E. n=3–6 mice/group.

Next, to better understand the inflammatory microenvironment that miR-223 might regulate in the context of CAC progression we performed a series of RT-PCR experiments on colon tissue from WT and miR-223^−/y mice following an AOM-DSS course. Untreated uninflamed WT and miR-223^−/y served as controls. We assessed gene expression for markers associated with both pro- and anti-inflammatory macrophage, neutrophil and monocyte functional responses, in addition to cytokines and chemokines for these respective cell types, given the differential tissue numbers. For all markers assessed, there was no differential expression between untreated WT and miR-223^−/y at baseline, while non-significant trends of increased expression were observed for some pro-inflammatory mediators and cytokines such as NOS2 and GM-CSF compared to untreated WT controls (Fig 3B+E). Consistent with the increase in tumor burden, miR-223^−/y mice present with a significantly increased cytokine and chemokine expression profile, following AOM-DSS challenge compared to all other groups assessed. Interestingly, there is a mixed profile of both classical pro-inflammatory and colitis-associated cytokines (nos2, tnf, il-1b and s100a8; Fig 3B) and TAM-associated markers (arg1, chi3l3 and irf4; Fig 3C). miR-223^−/y mice further demonstrated a marked increase in chemokine expression for both monocytes (ccl2 and ccl7; Fig 3D) and neutrophils (cxcl1 and cxcl2; Fig 3D) and interestingly, myeloid growth factors (m-csf1, IL-34 and gm-CSF) (Fig 3E). These later data indicates that in a miR-223-deficient intestine under AOM-DSS conditions, both the recruitment, maturation and survival factors for monocytes and neutrophils are significantly enhanced. This phenocopies the previously described cytokine profile of miR-223^−/y during the initial phase of DSS-colitis in mice and collectively with the data presented here indicates that miR-223 maintains a consistent or enhanced myeloid cell recruitment to the colon following the AOM-DSS challenge(16). Of note, not all tumor associated cytokines or growth factors assessed were altered by miR-223-deficiency, for example, vascular growth factors that are implicated in tumor survival, namely vegfa and vegfc are unchanged in whole tissue analysis (Supplemental Fig 1E+F). As are genes for macrophage activation (irf5) and tumor associated immunoregulation (tgfb1 and il-1rn) (Supplemental Fig G-I). Collectively, these data suggest that miR-223 acts as an important break to limit monocyte and neutrophil recruitment and their associated cytokine responses in the colon, with implications for the progression from UC to CAC.

miR-223 deficiency results in the upregulation of pathways associated with poor prognosis in AOM-DSS.

To further elucidate why induction of CAC with chronic AOM-DSS results in miR-223^−/y mice exhibiting more severe disease and increased proliferation, we performed RNA sequencing on whole colon samples. Pearson’s correlation matrix analysis showed that all samples within each group were similar (Fig. 5A). Co-expression analysis showed that there was overlap between the three treatment groups, but also identified unique mRNA’s that were only expressed by individual groups (Fig. 5B). Analysis of the differentially expressed genes between miR-223^−/y AOM-DSS and WT AOM-DSS showed that mRNA expression varied between the two groups (Fig. 5C–D). Additionally, application of unbiased gene set enrichment (GSEA) KEGG pathway analysis showed that both the Inflammatory bowel disease pathways (mmu05321) (Fig 5D–F) and colorectal cancer pathways (mmu05210) (Fig. 5G–I) were significantly altered between genotypes with positive enrichment in the miR-223^−/y AOM-DSS mice compared to WT AOM-DSS controls. Further GSEA analysis revealed that selected pathways associated with poor clinical prognosis in CRC and CAC, are significantly positively enriched in the miR-223^−/y AOM-DSS treated group, including ‘p53 signalling’, ‘apoptosis’, ‘DNA damage’, ‘regulation of PTEN stability’ (Fig. 5J). It is noteworthy that the phenotype of increased proliferation and tumor burden described earlier in miR-223^−/y was not representative of all CRC signalling pathways assessed. While the reactome analysis identifies signalling pathways that were upregulated in the miR-223^−/y group with clinical relevance to CAC and CRC progression (Supplemental Fig 4A+B), those of the TGFa–EGF cytokine family were also markedly increased in miR-223^−/y mice compared to WT counterparts at the protein level across groups (Supplemental Fig 3 B+C).

Figure 5. Bone marrow chimera identifies hematopoietic deficiency of miR-223 enhances susceptibility and inflammatory signature to AOM-DSS.

Schematic overview and timeline of the bone marrow chimera AOM-DSS study. A. number of macroscopic polyps per colon. Following H&E staining B. representative image of colon polyps. C. number of tumors per section, D. quantification of dysplasia (μm²) per section, E. quantification of tumor size (μm²) per section was determined at the time of necropsy following treatment with AOM-DSS. F. Tumor grading according to size from H&E micrographs in WT or miR-223^−/y mice treated with chronic AOM/DSS. RT-PCR was used to measure relative mRNA expression of the following markers from colonic tissues of indicated genotypes; G. pro-inflammatory markers, nos2, tnf, il-1b, il-6. H. anti-inflammatory markers arg1, chi3l3, retnla, il-4, irf4. I. monocyte chemokines ccl2 and ccl7 and neutrophil chemokines cxcl1 and cxcl2. All data is expressed as mean ± SEM; *, P<0.05; **, P<0.01; ***, P<0.001 versus the indicated counterpart (one-way ANOVA). C-F; WT→WT: 4 tumors and 3 dysplastic regions from 5 mice; WT→miR-223^−/y: 1 tumor and 3 dysplastic regions from 7 mice; miR-223^−/y→WT: 19 tumors and 7 dysplastic regions from 5 mice; miR-223^−/y→miR-223^−/y: 10 tumors and 7 dysplastic regions from 4 mice. G-I. n=5–6 mice/group.

Hematopoietic deficiency of miR-223 enhances susceptibility to AOM-DSS

As we performed AOM-DSS studies in a germline hemizygous knockout strain of miR-223^−/y mice, we generated bone marrow chimeric animals to ascertain if the stromal/radio-resistant or hematopoietic compartment was responsible for the effects on inflammation-driven tumorigenesis. Our data confirm hematopoietic-derived miR-223 (miR-223^−/y → WT recipients) as driving exacerbated AOM-DSS, as indicated by increased macroscopic polyps, dysplasia and tumor numbers compared with controls (WT → miR-223^−/y recipients; Fig 5, A–F). However, a significant difference for colon length and spleen weight was not observed (Supplemental Fig 1J–L). To further confirm the effect of miR-223 derived from the hematopoietic compartment versus the stromal compartment we analysed the relative mRNA expression of markers associated with pro- and anti-inflammatory macrophage, monocytes and neutrophils. Expression of il-1β, nos2 and chi3l3 were increased in miR-223^−/y → WT compared to the stromal chimera (WT → miR-223^−/y) (Fig 5G+H). This was accompanied by increases in the neutrophil chemokines cxcl1 and cxcl2, with non-significant trends towards increased expression of monocyte chemokines, ccl2 and ccl7 (Fig 5I). Similarly to gene expression analysis of AOM-DSS alone, vascular growth factors (vegfa and vegfc), macrophage activation (irf5) and tumor associated immunoregulation (tgfb1 and il-1rn) genes are unchanged in the BM-chimera study across all groups (Supplemental Fig 1M–Q). While the bone marrow chimera study highlighted some genes that appeared to be regulated by stromal miR-223 interactions (e.g. nos2 and arg1), overall, our data suggest that miR-223 constrains the inflammatory tone of infiltrating myeloid cells to limit the progression towards tumorigenesis during experimental AOM-DSS.

Increased IL-6 cytokine family and STAT3 signalling in miR-223^-/y mice following AOM-DSS challenge.

RNA-sequencing analysis highlighted a consistent pattern of increased cytokine signalling and particularly an enriched signature for IL-6 family members and STAT3 signalling in miR-223^−/y following AOM-DSS when compared to WT AOM-DSS (Fig. 6A–C). Given the clinical importance to STAT3 signalling to CRC progression(21), we wanted to confirm these findings further. RT-PCR analysis showed that expression of IL-6 family members and STAT3-signalling cytokines are significantly increased in miR-223^−/y mice compared to WT controls, including il-6, il-11 and lif (Fig. 6D). We extended this analysis to include cytokines that signal via STAT3 and play a role in CRC and mucosal inflammation, namely il-22 and il-17a, with both also being significantly increased in miR-223^−/y mice (Fig. 6D). Lastly, the expression and localisation of active phospho-STAT3 and total STAT3 proteins were assessed by immunoblot and immunohistochemistry. A marked increase in p-STAT3 was observed in the uninvolved colon tissues of miR-223^−/y mice compared to WT controls following AOM-DSS, consistent with increased tissue expression of IL-6 family cytokines and myeloid inflammatory infiltrates (Fig 6D–I). Together, these findings suggests that IL-6 and p-STAT3 signalling are contributing to the development of CAC in miR-223^−/y mice.

Figure 6. IL-6 cytokine family and p-STAT3 signalling are increased in miR-223^−/y mice in the chronic AOM-DSS model of CAC.

A. Dot plot for Reactome pathways upregulated in miR-223^−/y AOM/DSS. B. Heat map of the differentially expressed mRNA in the cytokine signalling in the immune system pathway for miR-223^−/y AOM-DSS vs WT AOM-DSS (n=3/group). C. Heat map of the differentially expressed mRNA in the IL-6 signalling family for miR-223^−/y AOM-DSS vs WT AOM-DSS. D. Gene expression was analysed in colon tissue by RT-PCR for expression of il-6, il-11, lif, il-22 and il-17a. E. Western immunoblot assessment of p-stat3, stat3 and β-actin in colon biopsies. F. Representative 20x micrographs of unaffected and tumor tissue from WT and miR-223^−/y mice treated with chronic AOM-DSS after immunohistochemical staining for p-stat3. G. Quantification of p-stat3 staining. H. Representative 20x micrographs of unaffected and tumor tissue from WT and miR-223^−/y mice treated with chronic AOM-DSS after immunohistochemical staining for stat3. I. Quantification of stat3 staining. All data is expressed as mean ± SEM; *,P<0.05; **,P<0.01; ***,P<0.001 versus the indicated counterpart (One-way ANOVA). A, B, C, E, F: n=3 mice/group; D. n=3–6 mice/group; F&H: WT AOM-DSS n=5; miR-223^−/Y AOM-DSS n=4; G&I: WT AOM-DSS n=5; miR-223^−/Y AOM-DSS n=4 with 3 fields of view presented per animal.

Antibody-mediated depletion of infiltrating Ly6G⁺ myeloid cells with RB6–8C5 reduced early tumorigenesis in miR-223^−/y mice using an acute model of AOM-DSS.

Our previous work identified a direct role for miR-223 in controlling the early inflammatory response to colitis and in the context of bacterial infection of the lung(16, 22). While our bone marrow chimeric studies indicate a role for haematopoietic-derived miR-223 in controlling AOM-DSS pathology, we do not know which immune subset are responsible. Given the importance of the IL-6/IL-17a/STAT3 cytokine axis in recruiting neutrophils to mucosal tissues and the role of CXCR2⁺ neutrophils in both AOM-DSS and CAC(23), we wanted to test if miR-223 controlled neutrophil recruitment was responsible for the increased tumorigenesis in miR-223^−/y mice.

To directly test this we utilised a revised protocol for AOM-DSS with a shortened treatment plan (AOM + 1 round of DSS and 5 weeks recovery with water). This allowed for modest development of inflammation-driven neoplasia and a window to target immune cell populations important during the development phase of CAC. We treated miR-223^−/y mice with an IgG control or anti-GR1 (clone RB6–8C5) antibody every 3 days for 3 weeks. This regime selectively depleted neutrophils and myeloid-derived suppressor cells, without impacting on the frequencies of other colonic myeloid cells (Fig 7C–F). Importantly, anti-GR1 depletion studies largely blocked the development of colonic dysplasia and tumor development compared to the IgG-treated control group (Fig 7A–B). Of note, neither CD4⁺, CD8⁺ or CD19⁺ lymphocyte populations were significantly different in the WT vs miR-223^−/y groups nor were they altered in the anti-GR1 treatment groups (Supplemental Fig 4A–C). In addition, we assessed the expression of tumor promoting regulatory T cells (CD4⁺ FoxP3⁺) and their colonic frequency was not changes at this time point (Supplemental Fig 4, F+G). We also show a significant increase in IFNy signalling in the colons of miR-223^−/y mice compared to WT counterparts (bulk RNA-seq) (Supplemental Fig 4H+I), while RT-PCR showed enhanced expression of IFNy and PD-L1 in miR-223^−/y colons at baseline or following AOM-DSS (compared to WT counterparts) (Supplemental Fig 4, J–L). Of note, PD-L2 (thought to be expressed largely by TAM’s) was not significantly altered. Collectively, this data suggests that while the frequencies of major T cells are unchanged across groups, there is marked increase in effector T cell cytokines in the colons along with pro-tumor check-point ligands. Additionally, from our original bulk RNA-sequencing data (Fig 4), there is a series of pro-tumor pathways activated that may collectively contribute to enhances tumor growth observed, which the myeloid compartment of miR-223^−/y mice appears to control.

Figure 7. Antibody-mediated depletion of infiltrating Ly6G⁺ myeloid cells with RB6–8C5 reduced early tumorigenesis in miR-223^−/y mice in an acute model of AOM-DSS.

Schematic overview and timeline of the acute AOM-DSS study. miR-223^−/y mice were treated with either 20μg (i.p) anti-GR1 (Lyg6/Ly6C: RB6–8C5 clone) or an isotype control, every 3 days for 3 weeks (8 total injections). Following H&E staining A. number of tumors per section, quantification of dysplasia (μm²) per section and quantification of tumor size (μm²) per section was determined at the time of necropsy following treatment with AOM-DSS. B. Tumor grading according to size (μm²) was determined from H&E micrographs in miR-223^−/Y mice treated either IgG or anti-GR1 antibody with representative H&E micrographs. Flow cytometry was performed on enzyme digested colons from WT and miR-223^−/Y mice post treatment to assess myeloid immune subsets. From live cells (AquaVi negative), CD45⁺ singlets, C. Neutrophils were identified Ly6G⁺ CD11b⁺. D. monocytes as Ly6G⁻ CD11b⁺ Ly6C⁺ MHCII⁻ (inflammatory/infiltrating) or Ly6C⁺ MHCII⁺ (intermediate/differentiating). E. CD64⁺ MHCII⁺ tissue resident macrophage were further subdivided into CX3CR1⁺ CD11b⁻ or CX3CR1⁺ CD11b⁺. F. Dendritic cell subsets were identified as CD64⁻ MHCII⁺ CD11c⁺ and further subdivided into CX3CR1⁺ CD103⁻ or CX3CR1⁻ CD103⁺ (migratory). All data is expressed as mean ± SEM; *, P<0.05 versus the indicated counterpart (one-way ANOVA). A+B. Effects of anti-GR1 on the volumetric size of tumors in miR-223^−/y mice following an acute regime of AOM-DSS; IgG: 15 tumors and 7 dysplastic regions from 6 mice; anti-GR1: 6 tumors 1 dysplastic region from 5 mice. C-H. n=3–5 mice/group.

Figure 4. miR-223^−/y mice present with an upregulation of pathways associated with poor prognosis in CAC.

A. Heat map of the Pearson correlation between all samples analysed via RNA sequencing. B. Venn diagram illustrating the overlap of mRNA shared between treatment groups. C. Volcano plot of differentially expressed mRNA between miR-223^−/y AOM-DSS vs WT AOM-DSS. D. Heat map of the top up-regulated differentially expressed mRNA associated with Inflammatory Bowel Disease between miR-223^−/y AOM-DSS vs WT AOM-DSS. E. Heat map of the most down regulated differentially expressed mRNA associated with Inflammatory Bowel Disease between miR-223^−/y AOM-DSS vs WT AOM-DSS. F. Enrichment plot for Inflammatory Bowel Disease (MMU05321) from GSEA analysis (NES: 1.209; FDR q-val: 0.473). G. Heat map of CRC associated mRNA that are up-regulated in miR-223^−/y AOM-DSS vs WT AOM-DSS. H. Heat map of CRC associated mRNA that are down-regulated in miR-223^−/y AOM-DSS vs WT AOM-DSS. I. Enrichment plot for Colorectal Cancer (MMU05210) from GSEA analysis (NES: 1.343; FDR q-val: 0.281) J. Enrichment plots for p53 signalling (MMU04115) (NES: 1.404; FDR q-val: 0.239), Apoptosis (MMU04210) (NES: 1.404; FDR q-val: 0.212), DNA damage bypass (R_MMU_73893) (NES: 1.368; FDR q-val: 0.308) and Regulation of PTEN stability and activity (R_MMU_8948751) (NES: 1.391; FDR q-val: 0.402) from GSEA analysis. (NES, normalized enrichment score; FDR, false discovery rate). n=3 mice/group.

Discussion

Global incidence of colorectal cancer, including CAC, continue to rise(24). However, our understanding of the underlying mechanisms behind CAC development remains incomplete. The myeloid associated miRNA, miR-223 is a critical regulator of innate immunity and its expression is elevated in patients with active UC,(16, 17) but little research has focused on its role in CAC. Collectively, our data suggests that this miRNA is important for controlling proliferation and limiting tumor development in the AOM-DSS model of CAC. Our data has demonstrated that deficiency in this miRNA results in increased tumor size and with increased monocytes, macrophage and neutrophils infiltrating into the colon, in conjunction with increased monocytes migrating into tumors. This suggests that in this model of CAC, excessive myeloid inflammation is contributing to a pro-tumorigenic microenvironment and that miR-223 acts to limit the progression of CAC.

Chronic inflammation is a well-established contributor to both the initiation and progression of cancer(25) including gastric(26, 27), esophageal(28, 29), and cervical cancers(30, 31). In CAC, the inflammatory microenvironment promotes the release of cytokines, such as IL-6, IL-1β, TNF-α and TGF-β, to promote tumor growth(32). Consequently, excessive NOS is produced, along with the presence of reactive oxidative species (ROS) which further boosts the production of pro-inflammatory cytokines(33). This creates a cycle that sustains inflammation, leading to DNA damage and ineffective tissue repair. In our model, miR-223^−/y mice had significantly higher expression of pro-inflammatory cytokines, iNOS, and markers of proliferation (c-Myc, PCNA and cyclin-D1) suggesting that miR-223 is critical for controlling the inflammatory response in CAC.

Persistent inflammation in CAC also increases the number of immune cells infiltrating into the colon, with neutrophils and monocytes among the most abundant(34). The ability for monocytes to subsequently differentiate into TAMs further ensures the inflammatory cycle continues via the production of inflammatory cytokines and growth factors. Specifically, CCL2, CCL7, TNF, TGF-β and M-CSF1 are critical for this process and are increased with miR-223-deficiency following AOM-DSS(9). This incites a pro-tumorigenic, immunosuppressive environment that limits the detection and elimination of cancerous cells. Accordingly, TAMs are associated with poor prognosis in many cancers(35). Due to the immunomodulatory ability of TAMs, they are currently being investigated as a potential therapeutic target for the treatment of CAC, along with many other cancers. Studies to date have focused on either reducing the presence of TAMs or altering their functionality, with research in CRC showing promising results(36, 37).

Importantly, both tissue resident macrophage in the colon and in particular CD68⁺ TAM’s in the tumor microenvironment, have been linked to the recruitment of neutrophils and myeloid-derived suppressor cells through the expression of chemokines CXCL1 and CXCL2(38, 39). While critical for anti-microbial tissue homeostasis, neutrophils have been increasingly linked with pro-tumor responses and tumor growth, using mechanisms such as metalloproteinases to degrade the tissue parenchyma and allow for tumor growth(40) (23, 41), and more recently being linked to tumor vascularization. Our data shows that while increased neutrophil chemokines are present along with increased neutrophils in tumor-adjacent tissues, neutrophil numbers are not altered within the tumor directly. It should be noted that this measurement was at end-stage of CAC development in this model (3 rounds of chronic AOM-DSS) and so may be indicative of the tumor stage compared to WT counterparts. The heightened cytokine and chemokine expression signature of miR-223^−/y mice might also underlie the increased proliferative state of the tumors in miR-223^−/y colons (possibly via pro-growth STAT3 from cytokines such as IL-6, IL-17a, IL-22 or via TGF-EGFR signaling). There appears to be clinical discrepancies as to the scoring parameters and hallmarks of tumor-associated neutrophils, in part due to shared cellular markers to identify neutrophils with other myeloid populations (unlike mouse Ly6G) (42). However, increased neutrophils are correlated with tumor score in CAC, with location associated with altered function (eg: angiogenesis)(43). In addition, altered neutrophil to lymphocyte ratios have been associated with heighted disease course and increased liver metastasis in CRC (44). While our bone-marrow chimera and antibody depletion studies go some way to demonstrate that heightened myeloid inflammation (and potentially tissue damage and altered repair mechanisms) underling the increased CAC phenotype of miR-223^−/y mice, some variations are present across experiments. For example, most of the inflammatory cytokines and chemokines that were measured show similar expression profiles in AOM-DSS treated colons, yet some important TAM-associated markers have differential expression between germline miR-223^−/y mice and the bone-marrow chimera studies. For example, while retlna, irf4 and chil3 all demonstrate similar expression patterns, arg1 shows discrepant expression profiles (increased in the chronic studies while decreased in hematopoietic groups of the bone-marrow chimera). This may be reflective of altered metabolism or mixed immune cell infiltration into the colon at the time of analysis. It is noteworthy that in the case of arg1 and arginine metabolism/Urea cycle, nos2 remains significantly increased across all studies and correlates with tissue inflammation and progression to CAC in this model.

In addition to a mixed myeloid inflammatory and TAM-associated gene expression profiles, miR-223^−/y mice present with significantly increased CCR2+ cells within the tumors at end-stage disease. With the majority of CCR2+ marking inflammatory monocytes in mice, this is an interesting finding that correlates with monocyte chemokine induction (ccl2, ccl7) and growth factors (csf-1 [m-csf] and il-34). Monocytes and monocyte-derived macrophage are found in multiple tumor types and are linked to tumor proliferations, metastasis, angiogenesis, and chemotherapy resistance (45, 46) (47) (48). Yet, the functional or regional significance of infiltrating monocyte subsets is not clear for solid tumors such as CAC or CRC. It is noteworthy that both monocyte and TAMs are mixed cell populations with differential expression of both inflammatory and pro-tumor markers, especially at an early stages of cancer (49). As such, early therapeutic intervention may have the capacity to reverse the inflammatory continuum of CAC and limit more chronic angiogenesis, invasion and epithelial-to-mesenchymal progression.

Understanding the collective role of different myeloid cell subsets to the development of CAC has been challenging, due to the heterogeneity of disease and varying onset of tumorigenesis. Yet, this avenue of research is critical to understanding CAC clinically and considering the difficulty in treating large colorectal adenocarcinomas with biologic therapeutics. For example, while there is increased expression of T cell-derived PD1 and PD1 ligands, PD-L1 and -L2 in AOM-DSS, PD1 antibody blockade had no therapeutic effect in this model (50). Along with clinical findings that gastrointestinal tumors are particularly refractory to check-point inhibitor blockade (51, 52), understanding how aberrant myeloid inflammation can control the development of solid tumors is of clinical importance going forward.

While many inflammatory pathways have been linked to tumor development in CAC, our identification of the STAT3-inducing IL-6-IL-17a cytokine family members in miR-223^−/y mice are interesting. The IL-6 family of cytokines are increasingly being recognized for their role in intestinal carcinogenesis. A variety of different cell types produce IL-6, IL-11 and LIF but during chronic inflammation the primary producers are monocytes, macrophage and fibroblasts following activation from NF-kB (53). Increased expression of IL-6 is correlated with metastasis, larger tumors, poor prognosis in CRC (54, 55) and with more severe inflammation in UC patients(56). Additionally, upregulation of IL-6 has also been shown in CAC (57) and associated with neutrophilia (58). This results in increased activation of STAT3 and the upregulation of its target gene cyclin-D1, promoting a proliferative microenvironment (59). Due to the evidence that aberrant IL-6 signaling is contributing to the progression of CAC, clinical investigations into its therapeutic viability are ongoing, with an IL-6 inhibitor Siltuximab in clinical trials and the STAT3 inhibitor Bruceantinol in development (60, 61). Thus, mining selective non-coding RNA networks to identify important therapeutic opportunities, such as miR-233 regulated mRNA circuits, may prove beneficial in elucidating druggable targets.

Aside from utility as predictive biomarkers (62), using microRNAs to identify critical signaling hubs would also provide a more selective therapeutic opportunity given their promiscuity and ability to bind hundreds of mRNA sequences (63). While our data suggests that miR-223 is acting as a tumor suppressor in CAC, its role in other cancers varies. In gastric cancer the overexpression of miR-223 is associated with increased proliferation, migration, and poor prognosis for patients (64, 65). In contrast, in lung cancer upregulation of miR-223 is linked to tumor suppression and can inhibit proliferation via a miR-223–3p-mutant p53 feedback loop (66) and by targeting IGF-1R pathway (67). In breast cancer some research suggests that the upregulation of miR-223 is associated with improved patient outcomes (68) as it acts to limit cancer cells by suppressing the EGF pathway (69). This is an intriguing finding as our data indicated that the EGFR ligand, TGFa, is markedly increased in miR-223-/y mice compared to WT counterparts. While we did not confirm if this is secreted and the processed form of TGFa, its downstream proliferative targets (c-MYC, clyclin-D1 and p-AKT) were increased in miR-223-/y mice and correlated with enhanced tumor growth. Others pro-tumorigenic pathways associated with miR-223 translational control includes Mef2c-β-catenin and Hippo/YAP signaling (70–73). miR-223 has been shown to contribute to tumor development in non-CAC colorectal cancers (CRC) (74–76), in addition, the expression of miR-223 in CRC is an indicator of metastasis and poor patient prognosis (77). Although CAC is also a cancer of the colon, the pathogenic sequence leading to disease differs significantly and our results show that miR-223 plays a protective role during inflammation-driven tumorigenesis. Due to miR-223 contribution to both the suppression and promotion of disease, its regulation for therapeutic purposes needs to be looked at in a disease specific manner. Our data suggests that restoring the expression of miR-223, or its regulated mRNA and thus protein circuits, could potentially lead to reduced tumor growth and prevent the progression of CAC, but this has yet to be examined in mouse models or in a clinical setting. Most recently, research using miR-223 as a therapeutic target has been explored in hepatocellular carcinoma, where treatment with miR-223 inhibited angiogenesis and prevented disease progression (78). In conclusion, we propose that miR-223 is essential for controlling myeloid driven inflammation in the progression from UC to CAC by suppressing the production of pro-inflammatory cytokines, limiting tumor growth, controlling proliferation and tumorigenic pathways.

Key Points.

• The microRNA, miR-223 regulates colitis-associated tumorigenesis in mice.

• miR-223^−/y mice have increased tumors and inflammatory cytokines and chemokines.

• Bone marrow chimera and antibody studies link miR-223^−/y phenotype to myeloid immune cells.

Data Availability

Bulk RNA sequencing data generated for this study is available through the Gene Expression Omnibus with accession number GSE245724.