Obesity exacerbates colitis-associated cancer via IL-6-regulated macrophage polarisation and CCL-20/CCR-6-mediated lymphocyte recruitment

Colorectal cancer (CRC) is one of the most lethal cancers worldwide in which the vast majority of cases exhibit little genetic risk but are associated with a sedentary lifestyle and obesity. Although the mechanisms underlying CRC and colitis-associated colorectal cancer (CAC) remain unclear, we hypothesised that obesity-induced inflammation predisposes to CAC development. Here, we show that diet-induced obesity accelerates chemically-induced CAC in mice via increased inflammation and immune cell recruitment. Obesity-induced interleukin-6 (IL-6) shifts macrophage polarisation towards tumour-promoting macrophages that produce the chemokine CC-chemokine-ligand-20 (CCL-20) in the CAC microenvironment. CCL-20 promotes CAC progression by recruiting CC-chemokine-receptor-6 (CCR-6)-expressing B cells and γδ T cells via chemotaxis. Compromised cell recruitment as well as inhibition of B and γδ T cells protects against CAC progression. Collectively, our data reveal a function for IL-6 in the CAC microenvironment via lymphocyte recruitment through the CCL-20/CCR-6 axis, thereby implicating a potential therapeutic intervention for human patients.

T he current obesity epidemic not only accounts for the increased incidence of classical comorbidities such as type 2 diabetes mellitus, but also predisposes to the development of certain cancers-primarily those that require an inflammatory tumour microenvironment (TME) 1 . One cancer type that is strongly associated with obesity is colorectal cancer (CRC) [2][3][4] . Globally, CRC is the second most diagnosed cancer in females and the third in males with 14.1 million new cancer cases and 8.2 million deaths in 2012 5 .
Obesity-induced alterations in microbiota composition and stem cell modulation have been demonstrated to promote CRC development 6,7 , but therapeutic strategies targeting these putative drivers of CRC might have unpredictable side effects. It is wellestablished that obesity is associated with a chronic, low-grade inflammatory state 8 that could also contribute to CRC development. However, the role of obesity-induced inflammation in CRC development is unknown. Importantly, obesity therapeutic strategies that reduce inflammation can be easily conducted in patients via dietary and lifestyle intervention 9 . Thus, reducing obesity-associated inflammation might represent a convenient strategy to prevent obesity-induced CRC.
In obesity, immune cells such as macrophages, T cells and B cells infiltrate the white adipose tissue. Activation of these cells causes local and systemic increases of inflammatory cytokines, such as tumour necrosis factor (TNF) and interleukin (IL)-6. Elevated cytokine levels are typically associated with obesity and propagate the obesity-associated inflammatory state [10][11][12][13] .
IL-6 acts via its membrane-bound IL-6 receptor (IL-6R) composed of IL-6Rα that mediates specificity and the common signalling chain of IL-6-type cytokines glycoprotein 130 (GP130) 14 . Though previously excluded, also ciliary neurotrophic factor (CNTF), another IL-6-type cytokine, can act as an alternative ligand for the IL-6R under certain circumstances, which might explain different outcomes when investigating IL-6 and IL-6R knockout mice 15 . Moreover, cell types that are not expressing IL-6Rα can be rendered IL-6-sensitive via IL-6 transsignalling mechanisms where a soluble IL-6Rα (sIL-6Rα) is shedded from the cell surface and acts with IL-6 on GP130-expressing cells 16 . Interestingly, such IL-6 transsignalling prevents obesity-induced recruitment of macrophages into adipose tissue that paradoxically failed to improve systemic insulin sensitivity 17 . On the other hand, enhanced central sIL-6Rα signalling improved energy and glucose homoeostasis in obesity 18 . Thus, different modes of signalling can affect various cell types that even do not express the necessary receptors. Moreover, we have demonstrated previously that IL-6 exerts beneficial effects in lean mice by limiting hepatic inflammation, whereas the chronic low-grade elevation of IL-6 in obesity abrogates these functions, presumably via the development of IL-6 resistance [19][20][21][22] . Moreover, IL-6 signalling can polarise macrophages towards an anti-inflammatory M2 phenotype, whereas IL-6Rα deficiency leads to largely arrested macrophages in the proinflammatory M1 state 19 . Notably, M2 macrophages functionally overlap with tumour-associated macrophages, indicating that IL-6 might have a detrimental role in carcinogenesis 23,24 .
Indeed, IL-6 promotes CAC development via its action in intestinal epithelial cells (IEC) [25][26][27][28] . Furthermore, in the classical aetiology of CAC, the initial development of inflammatory bowel diseases (IBD) such as colitis ulcerosa and Crohn's disease are also associated with increased IL-6 level in circulation 29 . This suggests that induction of IL-6 could be a common mechanism shared between obesity-induced and IBD-induced disease progression. However, how the low-grade nature of IL-6 in obesity impacts on CRC development and progression has not been investigated yet.
Here we investigate the role of obesity-induced IL-6 during development and progression of CAC in mice. We demonstrate that macrophage-specific IL-6Rα inactivation strongly ameliorates CAC in obesity. This is owing to a reduction of the chemoattractant CC-chemokine-ligand-20 (CCL-20) derived from M2 macrophages, which in turn facilitates recruitment of B cells and γδ T cells into the TME in a CC-chemokine-receptor-6 (CCR-6) dependent manner. Thus, we identify IL-6R signalling in macrophages as an important mediator of colon carcinogenesis during obesity.

Results
Diet-induced obesity increases CAC development. In a first experiment, we aimed at elucidating whether diet-induced obesity affects colon inflammation and CAC. To model obesity-induced CAC in mice, we exposed cohorts of C57BL/6 mice to either normal chow (NCD) or high-fat diet (HFD) feeding from weaning on. As expected, 8-week-old HFD-fed animals exhibited increased body weight and body fat content with increased serum insulin and leptin levels as well as impaired glucose homoeostasis (Supplementary Fig. 1a-e).
Next, we investigated expression of inflammatory cytokines as well as markers for immune cells in the colon tissue of these mice. This analysis revealed the elevated expression of inflammatory cytokines Il6, Tnf, Il1β and Il10 in the colon of obese mice compared with lean controls (Fig. 1a). Examination of immune cell markers in the obese colon revealed increases in T cells (Cd3δ), regulatory T cells (Tregs) (forkhead-box-protein P3 (Foxp3)), B cells (Cd19) but similar expression of the common macrophage marker epidermal growth factor-like modulecontaining mucin-like hormone receptor-like 1 (Emr1) (Fig. 1b). Thus, HFD-feeding accelerates inflammation and lymphocytes in the colon of mice.
Next, we subjected cohorts of NCD-and HFD-fed mice to the established colitis-associated adenoma protocol induced by azoxymethane (AOM) injection directly followed by three repetitive cycles of 2.5% dextrane sodium sulphate (DSS) in the drinking water over 7 days ( Supplementary Fig. 1f) 30 . Strikingly, obese mice either died spontaneously or had to be killed during the first DSS cycle, as these animals lost more than 20% of their initial body weight owing to symptoms associated with exaggerated colitis (Supplementary Fig. 1g, h). As the standard CAC protocol turned out to be highly lethal for obese mice, we adjusted the conditions accordingly to ensure survival of all cohorts throughout the treatment period. Therefore, we injected 8-week-old mice with AOM (day 1) and supplemented the drinking water with 1.5% DSS from day 5 for 5 days. At day 10 of the 1.5% AOM/DSS protocol the animals received normal water until the end of the experiment at day 62 (non-tumour, tumour) (Fig. 1c). To examine the colitis phase animals were killed with 10 weeks of age at day 13 (colitic) and as control with 8 weeks of age at day 0 (non-colitic) of the 1.5% AOM/DSS protocol.
Notably, lean mice that developed colitis exhibited an adapted colonic microenvironment at day 13 that resembled the inflammatory state of non-colitic HFD-fed mice (Fig. 1a, b, d, e). Strikingly, HFD in colitic mice increased Il6 expression and to a lower extent Tnf, Il1β and Il10 (Fig. 1d). Although colitic mice, independent of diet, had similar presence of T cells and B cells in colons compared with non-colitic HFD-fed colons, expression of the Treg transcription factor FoxP3 was increased suggesting a shift in T-cell composition (Fig. 1b, e). Macrophages increased only in tendency in obese colitic colons (Fig. 1e). Thus, HFD feeding mimics the inflammatory phenotype found in colitis.
Colitis compromises the gut barrier function, reflected by rectal bleeding, by mounting immune responses against commensal    Fig. 1 Obesity enhances colonic inflammation and CAC tumour burden in C57BL/6 mice. a, b qPCR analysis of indicated gene expression in distal colons derived from 8-week-old non-colitic NCD (n = 5) and HFD-fed (n = 5) C57BL/6 mice; following results are presented relative to non-colitic NCD. c Scheme of the 1.5% AOM/DSS protocol of mice fed a NCD or a HFD from weaning (3 weeks of age) until the end of the experiment with 17 weeks of age. 8-week-old mice were injected with 10 mg/kg AOM i.p. at day 1 of the 1.5% AOM/DSS protocol. From day 5 on, mice received 1.5% DSS in the drinking water for 5 days. Mice were either killed with 8 weeks of age at day 0 (non-colitic), 10 weeks of age at day 13 for analysis of colitic conditions (colitic) or with 17 weeks of age at day 62 (non-tumour and tumour) for analysis of tumourigenesis. d, e qPCR analysis of indicated gene expression in distal colons of 10-week-old NCD (n = 5) and HFD-fed (n = 5) colitic C57BL/6 mice; results are presented relative to non-colitic NCD. f Normalised albumin concentration in faeces of NCD (n = 10) and HFD-fed (n = 10) C57BL/6 mice under non-colitic (day 0) and colitic (day 13) conditions. g Representative endoscopic pictures of 13-week-old NCD and HFD-fed C57BL/6 mice of the 1.5% AOM/DSS protocol. h Tumour number of 17-week-old NCD (n = 18) and HFD-fed (n = 13) C57BL/6 mice counted at day 62 of the 1.5% AOM/DSS protocol. AOM, azoxymethane; DSS, dextran sodium sulphate, NCD, normal chow diet; HFD, high-fat diet; CAC, colitis-associated colorectal cancer. centre line: median; box limits: 1st and 3rd quartiles; whisker: maximum to minimum, *p ≤ 0.05, **p ≤ 0.01 and ***p ≤ 0.001 two-tailed unpaired Student's t-test a, b, h or two-way ANOVA followed by Fisher LSD d-f factors such as bacterial lipopolysaccharides (LPS) 30 . To examine gut barrier function and rectal bleeding, we measured the presence of blood-abundant albumin in faeces of mice 31 . This analysis showed elevated immunoreactive albumin in the faeces of HFD-fed mice both under non-colitic and colitic conditions, indicating impaired gut barrier function (Fig. 1f). Further colonic investigation using endoscopy 32 revealed elevated presence of dysplastic neoplasia in colitic HFD mice compared with NCD-fed mice (Fig. 1g). Importantly, macroscopic investigation of colons demonstrated an increased tumour burden in obese mice when compared with their lean littermates at day 62 ( Fig. 1h and Supplementary Fig. 1i). Furthermore, colons of diet-induced obese mice were shortened, which is indicative of elevated inflammation and colitis ( Supplementary Fig. 1j) 30 . Moreover, we observed an exaggerated immune cell infiltration into tumours of obese mice as determined by histochemistry using F4/80 and CD3 antibodies to identify macrophages and T cells ( Supplementary  Fig. 1k, l). Taken together, diet-induced obesity exaggerated colonic inflammation, impaired gut barrier function and enhanced CAC development. Furthermore, HFD-feeding per se induced colonic inflammation, and colitis in lean mice increased inflammation to levels of non-colitic HFD-fed mice. Importantly, of the investigated cytokines, only IL-6 was upregulated in colitis of obese mice implicating that IL-6 might have a role in obesityinduced CAC development.
IL-6Rα deficiency in myeloid cells reduces CAC development. Next, we asked whether the obesity-induced elevation in IL-6 is driving CAC development. First, to identify the cell type that   week-old c NCD (n = 6-7) or d HFD (n = 6-7)-fed C57BL/6 mice injected with either PBS, 500 µg anti-IL-6 antibody or 150 μg sGP130Fc at day 3 of the 1.5% AOM/DSS protocol. e Representative immunofluorescent staining of macrophages using F4/80 antibody (red) counterstained by Hoechst 3342 (blue) in colons of 17-week-old HFD-fed Il6rα Fl and Il6rα KO mice at day 62 of the 1.5% AOM/DSS protocol. f, g qPCR analysis of indicated gene expression in tumours (n = 6-7) of 17-week-old NCD or HFD-fed Il6rα Fl , Il6rα KO and Il6rα myl-KO mice at day 62 of the 1.5% AOM/DSS protocol, results are presented relative to non-colitic distal colon of Il6rα Fl NCD at day 0. h Relative cell number (%) of CD11c + / CD11b − dendritic cells, CD11c + /CD11b + and CD11c − /CD11b + macrophages and CD11c − /CD11b − lymphocytes from lymphocyte gate SSC/FSC in colitic colons of NCD-fed Il6rα Fl and Il6rα KO mice at day 13 of the 1.5% AOM/DSS protocol. i Representative FACS plots of CD11c and CD11b expression in immune cells from h. AOM, azoxymethane; DSS, dextran sodium sulphate; NCD, normal chow diet; HFD, high-fat diet; CAC, colitis-associated colorectal cancer. Centre line: median; box limits: 1st and 3rd quartiles; whisker: maximum to minimum, *p ≤ 0.05, **p ≤ 0.01 and ***p ≤ 0.001 two-tailed unpaired Student's t-test h, one-way a-d or two-way ANOVA followed by Fisher LSD f, g. Scale bar, 100 μm responds to IL-6 we induced CAC in lean and obese mice with conditional ablation of the IL-6Rα in all cells 33 (Il6rα KO ), in IECs (Il6rα IEC-KO ) 27 , in mature T cells 34 (Il6rα T-KO ) or in the myeloid lineage 19 (Il6rα myl-KO ) using the 1.5% AOM/DSS protocol (Fig. 1c). As a consequence of Cre recombinase expression, exons 2 and 3 of the Il6rα FL allele are excised to conditionally inactivate IL-6Rα in the respective cell type ( Supplementary Fig. 2a-d). In line with a previous report employing IL-6 knockout mice, IL-6Rα deficiency protected against CAC development under lean and under obese conditions (Fig. 2a, b, Supplementary  Fig. 2e-j) 25 . The cell type-specific inactivation of IL-6 signalling revealed that IEC-specific and T cell-specific inactivation of   IL-6Rα did not decrease CAC development. However, myeloid lineage-specific IL-6Rα deficiency protected against CAC (Fig. 2a, b, Supplementary Fig. 2e-j). Macrophages can provide IL-6 and sIL-6Rα in CAC, but IL-6 signalling intrinsic in macrophages might also have a role in CAC 25 . To investigate whether neutralising IL-6 or sIL-6Rα affects CAC, we injected anti-IL-6 antibody or the designer cytokine soluble GP130 (sGP130Fc) before the colitis phase at day 3 of the 1.5% AOM/DSS protocol in lean and obese C57BL/6 mice. C57BL/6 mice had similar body weight loss in colitis independent of anti-IL-6 and sGP130Fc treatment under NCD and HFD conditions ( Supplementary  Fig. 2k, l). Of note, tumour numbers were only reduced in tendency upon neutralisation with anti-IL-6 in lean and obese mice ( Fig. 2c, d, Supplementary Fig. 2m-r). On the other hand, lean and obese mice with blockade of sIL-6Rα signalling had similar tumour numbers independent of neutralisation ( Fig. 2c, d).
However, anti-IL-6 and sGP130Fc treatment in lean mice reduced afflicted colonic area ( Supplementary Fig. 2m), which is in line with a previous report, demonstrating IL-6 deficiency reducing tumour area in CAC 25 . Of note, other IL-6-type cytokines such as CNTF that have not been blocked by antibody-mediated IL-6 depletion and blockade of sIL-6Rα via sGP130Fc might compensate in CAC development. Nevertheless, neutralising IL-6 and sIL-6Rα agents are less effective than genetic IL-6Rα ablation and in line with our previous experiments might even suggest macrophage-intrinsic IL-6 signalling functions impacting on CAC.
In this context, we have previously demonstrated that IL-6Rα deficiency in the myeloid compartment impedes M2 macrophage polarisation 19 . Furthermore, reports allocate M2-like macrophages a tumour-promoting function via angiogenic and immunomodulatory factors 35 . On the other hand, M1-like macrophages combat tumours via their proinflammatory and cytotoxic potential 36 . As we observed similar total macrophage numbers in tumour tissues independent of genotype ( Fig. 2e, f), we next asked whether IL-6Rα deficiency might alter specific macrophage subsets in CAC. Thus, we examined gene expression to identify M1/inducible nitric oxide synthase (iNos)-and M2/arginase 1 (Arg1)-expressing macrophages on isolated tumour tissue derived from control, Il6rα KO and Il6rα myl-KO mice. This analysis revealed that IL-6Rα deficiency in macrophages largely prevented M2-like polarisation as evidenced by reduced Arg1 expression, whereas iNos expression and thus M1 polarisation remained unaltered (Fig. 2g). Thus, a compromised M2 polarisation might hamper CAC in IL-6Rαdeficient mice. In line with this finding, fluorescence-activated cell sorting (FACS) analysis of immune cells revealed increases in CD11c + M1 macrophages in colitis colons of Il6rα KO mice, whereas the lymphocytes were reduced in Il-6rα-deficient colons (Fig. 2h, i). These findings suggest that a shifted macrophage polarisation capacity towards the inflammatory M1 lineage is contributing to the reduced CAC development found in Il6rα KO and Il6rα myl -KO mice.
Altered macrophage composition might affect DSS-induced colitis, impairments of gut barrier function and proliferation of IECs by providing growth factors and cytokines. In line with this evidence, increased inflammation and IEC death in a three times DSS-mediated colitis of IL-6-deficient mice is a result of impaired IL-6-mediated signal transducer and activator of transcription 3 (Stat3) action providing survival capacities in normal and premalignant IECs 25 . However, colitis symptoms such as weight loss and faecal albumin were comparable between control and Il6rα KO mice under NCD and HFD conditions, which might be a consequence of the mild HFD-adapted CAC protocol applied here ( Supplementary Fig. 3a-d). Moreover, even colitis-associated proliferation as examined via western blot analyses against the proliferation marker proliferating cell nuclear antigen (PCNA) was similarly detectable in control and IL-6Rα-deficient samples during colitis ( Supplementary Fig. 3e, f). In contrast, IL-6Rαdeficient tumours had less PCNA immunoreactivity compared with control tumours, whereas tumours from HFD-fed control mice had increased PCNA levels indicative of increased proliferation compared with NCD-fed controls ( Supplementary  Fig. 3g, h). Furthermore, decreased proliferation of IL-6Rαdeficient tumours was confirmed via Ki67 staining (Supplementary Fig. 3i, j). Thus, whereas colitis and colitis-associated proliferation is not affected by IL-6Rα deficiency, tumours of Il6rα KO mice exhibited less PCNA and Ki67 reactivity. Hence, the altered macrophage composition in IL-6Rα-deficient mice is not affecting the colitis phase of CAC but impacts on later stages of tumourigenesis.
To gain further insights into the molecular mechanisms that promote CAC via obesity-induced IL-6, we compared global gene expression in obese control vs IL-6Rα-deficient tumour samples by microarray. Although examination of upregulated genes failed to clearly identify pathways, a detailed analysis of downregulated genes revealed that the TME has been altered in the absence of IL-6 signalling (Fig. 3a, Supplementary Table 1). Gene Ontology analyses revealed that cell recruitment was impaired in knockout tumours compared with controls. Here, important chemoattractants such as Ccl20 and Ccl5 were downregulated in the microarray of IL-6Rα-deficient tumours (Fig. 3b). Ccl20 and its unique receptor Ccr6 were among the 30 genes that were downregulated the highest in IL-6Rα-deficient tumours (Fig. 3a,  Supplementary Table 1). However, although expression of Ccl5 and its receptors Ccr1 and Ccr3 were not affected by IL-6Rα deficiency in CAC, expression levels of Ccl20 and of its unique receptor Ccr6 were reduced in knockout tumours (Fig. 3c) (Fig. 3d). Moreover, Ccl20 and Ccr6 increased in obese control tumours and were downregulated in non-tumour tissue and tumours of obese IL-6Rα-deficient mice (Fig. 3e-g). Thus, we assumed that chemoattraction of cells via the CCL-20/CCR-6 axis might affect CAC, whereas the source for CCL-20 is unclear.
CCL-20/macrophage inflammatory protein-3α derives from macrophages in numerous types of tumours, which causes the recruitment of CCR-6-expressing lymphocytes [37][38][39] . To this end, we Magnetic-activated cell sorting (MACS)-separated macrophages and IECs from colitic colons derived from control, Il6rα KO and Il6rα myl-KO mice and subjected them to qPCR to examine Ccl20 expression. This analysis revealed increased Ccl20 expression in control macrophages isolated from colitic colons when compared with IL-6Rα-deficient macrophages (Fig. 3h), whereas IL-6Rα inactivation via LysM-Cre had a tendency towards reduced Ccl20 expression (Fig. 3h). However, also control IEC samples exhibited increased Ccl20 expression in colitis either via IEC-derived Ccl20 expression or potentially via contaminating macrophages in the samples ( Supplementary  Fig. 4a). From our previous experiments that demonstrate IL-6 signalling in macrophages as crucial mediator in CAC, we speculated that intestinal macrophages are a main source of CCL-20 in CAC. Therefore, IL-6 signalling in macrophages might either directly control CCL-20 expression on the transcriptional level or indirectly via its ability to polarise towards M2 macrophages. To test these hypotheses, we generated bone marrow-derived macrophages (BMDM) from control C57BL/6 mice and polarised them towards M1 via LPS/interferon-γ (IFNγ) treatment or to M2 via IL-4/IL-6 treatment. Examination of CCL-20 in the supernatant revealed that IL-6 stimulation did not directly induce CCL-20 expression in both M1 and M2 (Fig. 3i). However, LPS stimulation mimicking invading commensals increased CCL-20 in M2 compared to LPS-stimulated M1 macrophages (Fig. 3i). This experiment revealed that IL-6 indirectly controls CCL-20 expression in macrophages via its ability to polarise towards M2-type macrophages.
To confirm that IL-6Rα-deficient macrophages fail to produce CCL-20, we subjected control and Il6rα KO BMDM to M1 and M2 protocols and stimulated them with IL-6 and LPS, respectively. Consistent with our previous finding, IL-6Rα-deficient M2 had a blunted ability to differentiate into M2 macrophages substantiated by a compromised upregulation of M2 markers Arg1 and Il4rα upon IL-6 stimulation ( Supplementary Fig. 4b, c). Of note, M1 differentiation was similar between control and IL-6Rαdeficient macrophages as revealed by iNos expression (Supplementary Fig. 4d). Whereas IL-6 stimulation of M1 and M2 showed undeterminable Ccl20 expression, LPS-stimulated control M2 increased Ccl20 expression compared with the blunted response in IL-6Rα deficiency (Fig. 3j, k). Furthermore, control M2 secreted more CCL-20 compared with IL-6Rαdeficient BMDM in the M2 protocol under LPS-stimulated conditions (Fig. 3l). Thus, CCL-20 expression in macrophages is not directly controlled via IL-6-induced transcriptional regulation but instead, IL-6 polarises macrophages in the TME towards M2 capable to produce CCL-20 upon exposure to commensal antigens such as LPS.
Collectively, these experiments suggest that CCL-20 is expressed by M2-type macrophages in colitis and that the inability of IL-6Rα-deficient macrophages to polarise towards M2-type impedes on CCL-20 expression and presumably on recruitment of CCR-6-expressing cells. Thus, the CCL-20/CCR-6 axis might recruit cell types that promote CAC development.
IL-6Rα-deficient tumours have reduced lymphocyte quantity. CCR-6, the unique receptor for CCL-20, is mainly expressed on T and B lymphocytes and CCR-6 binding to CCL-20 causes chemoattraction of these cells towards the CCL-20 source [40][41][42][43] . In line with this fact, investigation of gene expression data from the microarray demonstrated that lymphocyte-specific genes, such as Cd19, Cd22, Cd79A, B cell activating factor receptor (Baff-R), Burton's tyrosine kinase (BTK) and paired box protein 5 (Pax5) for B cells and Cd3ε, Cd3δ, Cd4 and Il2rα for T cells, were downregulated in IL-6Rα-deficient tumours (Figs. 3a and 4a, Supplementary Table 1). Furthermore, Gene Ontology analysis revealed a reduced quantity of B cells and T cells in CAC in the absence of IL-6Rα signalling (Fig. 4a). To further validate these findings, we performed qPCR and immunohistochemistry of tumours to detect Treg, αβ T cells, γδ T cells and B cells (Fig. 4b-e). Indeed, the qPCR experiments confirmed the reduction of these lymphocyte subsets in IL-6Rα-deficient tumours when compared with controls (Fig. 4b). Notably, these data are in line with our previous observation of reduced lymphocyte counts in IL-6Rα-deficient mice in colitis (Fig. 2h, i). To quantify CCR-6 + lymphocyte recruitment in 1.5% AOM/DSSinduced colitis, we examined total cell numbers from non-colitic and colitic control as well as IL-6Rα-deficient colons by FACS (Fig. 4f-h). Although consistent to our previous data, similar macrophage numbers were observed in control and IL-6Rαdeficient colons, IL-6Rα deficiency largely attenuated lymphocyte recruitment in colitis (Fig. 4f). Using specific antibodies in FACS to detect CD19-expressing B cells and γδ as well as αβ T-cell receptor-expressing T cells revealed that these lymphocyte subsets were reduced in the colitis colons of knockout animals (Fig. 4g). Furthermore, co-staining with CCR-6 antibody revealed that the majority of B cells and γδ T cells in colitis express CCR-6, whereas only a minority of αβ T cells express the CCR-6 receptor (Fig. 4h) when compared with their total numbers (Fig. 4g). Taken together, these findings clearly demonstrate the reduced presence of CCR-6-expressing lymphocytes in IL-6Rα-deficient CAC and imply an impaired chemoattractance of these cells into the CAC TME of IL-6Rα-deficient animals.
CCR-6 + lymphocytes promote CAC development. To directly examine whether the CCL-20/CCR-6 axis is necessary for lymphocyte recruitment during colitis, we investigated CCR-6deficient mice (Ccr6 KO ) in our CAC model 44 . During the colitis phase, CCR-6-deficient mice exhibited reduced weight loss (Fig. 5a). Importantly, colons from colitic control Ccr6 WT mice had increased Ccr6 expression at day 13 compared with non-colitic controls, whereas consistently Ccr6 expression was absent in Ccr6 KO mice (Fig. 5b). Colitic colons derived from Ccr6 KO animals had reduced expression of markers for Tregs, αβ T cells, and B cells, but not of γδ T cells and macrophages (Fig. 5c). Concomitantly, expression of Il6, Tnf and Il1β were reduced in colitis colons of CCR-6-deficient mice (Fig. 5d). Thus, these analyses demonstrate that CCR-6-expressing cells are recruited during colitis and that CCR-6 deficiency largely protects against colitis. Consistently, CCR-6-deficient animals failed to develop CAC, both under lean and obese conditions (Fig. 5e, f, Supplementary Fig. 5a-c). Therefore, we conclude that CCL-20 released in the colitis phase recruits CCR-6-expressing lymphocytes that promote CAC tumourigenesis. However, which CCR-6expressing lymphocyte population, that is recruited to the colon in colitis promotes CAC, still has to be investigated. B cells, γδ T cells and Tregs are known to express CCR-6 [40][41][42][43] .
Recruitment of B cells via CCL-20/CCR-6 promotes CAC. Next, we aimed at investigating which CCR-6 + cell type contributes to CAC development. A prominent CCR-6expressing lymphocyte population comprises mature B cells 41,45 and B cells exert macrophage-polarising functions in colitis. In particular, regulatory B1b-like B cells in the colon (CD5 − , IgM + , CD19high, B220low) provide IL-10 to polarise macrophages towards M2-like phenotype 46,47 and such B cells are induced by gut microbiota-driven IL-1β and IL-6 production 48 . Presumably, these B cells sense and control commensal bacteria during  51 . We subjected B cell-deficient JHT mice to our CAC protocol and examined colitis parameters and CAC tumour development. Body weight loss in colitis and colitis-induced expression of Il6, Tnf, Il1β and Il10 were reduced in JHT mice (Fig. 6a, b). Furthermore, besides the lack of B cells, less immune cells such as T cells, Tregs and macrophages were present in colitis colons of JHT mice (Fig. 6c). This is in line with the finding that control mice developed dysplastic lesions, whereas the absence of B cells ameliorated colon pathology as revealed by endoscopy (Fig. 6d). Ultimately, control mice developed CAC, whereas B cell deficiency largely protected against CAC (Fig. 6e,  f). Investigation of gene expression revealed that JHT mice exhibited decreased colonic expression of Il6, Il1β and Il10, whereas Tnf was unaltered (Fig. 6g). Furthermore, not only B cells were absent in JHT colons but also gene expression markers for T cells and Tregs were reduced (Fig. 6h, i). Consistent with this fact was a reduced Ccr6 expression accompanied with a decreased Ccl20 expression in JHT colons, suggesting that B cells ultimately impact on the CAC TME (Fig. 6j). Apparently, the decreased Ccl20 expression is in line with an impaired macrophage polarisation to Arg1-expressing M2 as well as iNos-expressing M1-like in JHT colons (Fig. 6k). Collectively, these analyses reveal that during colitis CCL-20-recruited CCR-6 + B cells promote CAC. Contrariwise, we have also investigated JHT mice in the CAC protocol under obese conditions where B cell-deficient mice died either spontaneously or had to be killed owing to excessive colitis symptoms ( Supplementary Fig. 5d, e). This discrepancy between diets in B cell-deficient mice hints to a more complex and dual role for B cells in colonic inflammation under different dietary exposures than hitherto assumed and even more strengthens our finding that CCR-6-expressing B cells affect CAC. Anti-γδTCR injection led to downregulation of γδTCR as obvious from the appearance of CD3 + αβ T cells and γδ T cells in the blood of these mice (Fig. 7a). Although most γδ T cells express CCR-6 during colitis in untreated mice, a proportion of γδ T cells in antibody-treated mice were negative for CCR-6 expression (Fig. 7b). Ultimately, lean control mice developed CAC, whereas γδTCR antibody injection drastically reduced tumour burden in control mice similar to IL-6Rα-deficient mice (Fig. 7c, Supplementary Fig. 6a-c). Notably, the reduced CAC burden in anti-γδTCR-treated control mice was largely recapitulated under obese conditions (Fig. 7d, Supplementary Fig. 6d-f). Analysis of tumour-derived gene expression revealed slightly reduced γδTCR upon anti-γδTCR antibody injection in lean mice (Fig. 7e). Our experiments are in line with previous reports assigning γδ T cells as crucial mediators in colonic inflammation by providing pathogenic IL-17 53 . Thus, recruitment of CCR-6 + γδ T cells promotes CAC and inhibition of γδ T cells might be a reasonable therapeutically strategy against CAC.
Another CCR-6-expressing T lymphocyte population that has been demonstrated to contribute to CAC are Treg cells, which control gut inflammation via their immunosuppressive functions 37 . We have shown that IL-6 signalling in effector T cells is required to release them from Treg-mediated suppression using the well-established Ovalbumin (OVA) model 34 . Suppression of effector function in IL-6Rα-deficient cells included a blunted Th1/ IFNγ and total absence of Th17 IL-17a response. However, the Th1 response could be restored in IL-6Rα-deficient T cells when Tregs were depleted by anti-CD25 antibody treatment. In order to examine whether Tregs also suppress effector functions in lean and obese IL-6Rα-deficient mice in our CAC model, anti-CD25 antibody was i.p. injected at day 3 of the 1.5% AOM/DSS protocol just prior to the colitis phase of CAC. This experimental procedure should provide only a transient Treg depletion during the colitis phase and should affect T effector cells to a lesser extent that upon activation upregulate CD25. Most likely, the single anti-CD25 antibody injection depletes Tregs only until anti-CD25 antibody is consumed/degraded where in the context of proceeding T-cell development these cells are able to repopulate the colon. Nevertheless, although in lean mice, anti-CD25 antibodymediated transient depletion of Treg cells exhibited only minor effects, it restored CAC development in obese IL-6Rα-deficient animals, whereas obese control mice had unaltered CAC (Fig. 7f, Supplementary Fig. 6g-m). Detection of markers for αβ T cells and Treg cells in tumours of the obese cohorts revealed a reduction of these cells in IL-6Rα-deficient mice that was largely unaltered when mice received anti-CD25 treatment (Fig. 7g). However, despite the fact that also other cells express Ifnγ and Il17a, the examination of T-cell effector molecules revealed decreased Ifnγ and Il17a expression in obese IL-6Rαdeficient tumours that have been restored upon Treg depletion (Fig. 7h, i). Thus, Treg depletion in obese IL-6Rα-deficient mice restores CAC development via restoration of T-cell effector functions.
Collectively, these experiments reveal a crucial role for the recruitment of CCR-6-expressing lymphocytes into the colon during the colitis phase to promote CAC. Knockout and antibody-inhibition/depletion models demonstrate that recruitment of CCR-6-expressing B cells and γδ T cells promote CAC development whereas CCR-6 + Tregs exert an immunosuppressive role in CAC under obese conditions by interfering with T-cell effector functions. B cells and IL-6 synergise in the polarisation towards CCL-20-expressing macrophages causing a vicious circle of CCR-6 + lymphocyte recruitment in CAC (Fig. 8). Overall, our experiments assign obesity-induced IL-6 an unappreciated role in the CAC TME by regulating macrophage polarisation and lymphocyte recruitment.

Discussion
The steadily increasing obesity epidemic and the simultaneous incidence of obesity-associated comorbidities such as colorectal cancer requires a detailed understanding of molecular and cellular mechanisms that are affected to develop potential novel therapies.
Here, we show that obesity exacerbates AOM/DSS-induced CAC via increased expression of inflammatory cytokines. However, our experimental set up was unable to address whether in obesity preneoplastic lesions occur earlier than in lean mice. We demonstrate that obesity-induced impairment of gut barrier function recruits and activates immune cells that promote CAC. In the colitis phase of CAC, IL-6-polarised M2-like macrophages express the chemokine CCL-20 that recruits CCR-6-expressing lymphocytes, further promoting CAC progression. Our in vitro experiments suggest that intestinal M2 macrophages express CCL-20 in colitis via exposure to commensal antigens such as LPS and not as a consequence of direct IL-6-activated transcriptional traits. Though IL-6-type cytokine-activated STAT3 can directly activate CCL-20 expression 54 , CCL-20 expression is mainly regulated via inflammatory signalling and nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB)-controlled transcription [55][56][57][58] . Given that LPS-activated NF-κBcontrolled gene expression in M1 macrophages has only minor effects on CCL-20 expression suggests extra transcription factors active in M2 macrophages that act in concert with LPS-induced NF-κB to increase CCL-20 expression. Hence, the substantially reduced Ccl20 expression in colons of Il6rα KO mice is a consequence of collapsed M2-type polarisation of IL-6Rα-deficient macrophages in CAC, thereby compromising CCR-6-expressing lymphocyte recruitment in colitis to promote tumourigenesis. CCR-6-expressing B cells, γδ T cells and Treg cells 40,43 have either low or undetectable IL-6Rα expression underlining our observation that the protection against CAC in IL-6Rα-deficient mice is not a direct consequence of IL-6 action in these cells but is rather an indirect effect through compromised chemoattraction. Nevertheless, these cells might still be able to receive IL-6 signalling in control mice via IL-6-transsignalling mechanisms that we have avoided by using IL-6Rα-deficient mice devoid of both membrane and soluble signalling capacities. Despite the crucial role of direct IL-6 function on the approval of Treg-mediated suppression of αβ T cells, we estimate that Tregs also inhibit other immune cells by a similar mechanism as T-cell-specific inactivation of IL-6 signalling using CD4-Cre is unable to prevent CAC development. In line with this assumption, γδ T cells in the colon were known to be suppressed via Tregs and γδ T cells are not affected by CD4-Cre-mediated recombination 59 . Although γδ T cells can promote CRC via allocation of IL-17 to transformed tumour cells 60,61 , they are also able to restrain αβ T-cell responses in cancer by providing inhibitory programmed cell death 1 ligand to αβ T cells 62 . Collectively, our study assigns obesity-induced IL-6 as a modulator of the TME in CAC via macrophage polarisation and successive lymphocyte recruitment via the CCL-20/CCR-6 axis. In light of the steadily increasing obesity epidemic, novel treatment options to combat obesity-associated disorders are desperately required highlighting our study to interfere with inflammatory signalling and to inhibit corrupt cell types as important contribution to global health issues.
Induction of CAC in AOM/DSS model and cell depletion. 2.5% AOM/DSS protocol: 8-week-old mice fed a NCD or a HFD were i.p. injected with 10 mg/kg body weight AOM (A5486, Sigma-Aldrich) at day 1 of the 2.5% AOM/DSS protocol ( Supplementary Fig. 1f). The animals were exposed to three repetitive cycles of 2.5% DSS (MW = 36,000-50,000, 0216011080, MP Biomedicals) in the drinking water for 7 days from day 1-7, day 22-28 and day 43-49 (DSS water was refreshed every second day) to induce colitis. Animals were killed by CO 2 with 17 weeks of age at day 62 for tumour analysis. Mice that lost more than 20% of their body weight during the colitic phases were killed.
1.5% AOM/DSS protocol: 8-week-old NCD or HFD-fed mice were i.p. injected with 10 mg/kg body weight AOM (A5486, Sigma-Aldrich) at day 1 of the 1.5% AOM/DSS protocol (Fig. 1c). At day 5 mice were exposed to 1.5% DSS (MW = 36,000-50,000, 0216011080, MP Biomedicals) in the drinking water for 5 days (DSS water was refreshed at day 7) and provided with normal drinking water from day 10 on until the end of the experiment at day 62. Animals were killed with 8 weeks at day 0 for non-colitic control tissue or at day 13 for colitic analysis or at day 62 for tumour analysis. The colon length was measured from caecum till the anus. The distal part of the colon was used for non-colitic and colitic analysis. At day 62 tumours were counted in a blinded fashion by number, size (<2 mm, >2 mm) and percentage of the afflicted area of the distal colon. Tumour and non-tumour tissue from the distal colon was separated and used for tumour analysis.