Objective Cancer-associated fibroblasts (CAFs) influence the tumour microenvironment and tumour growth. However, the role of CAFs in colorectal cancer (CRC) development is incompletely understood.
Design We quantified phosphorylation of STAT3 (pSTAT3) expression in CAFs of human colon cancer tissue using a tissue microarray (TMA) of 375 patients, immunofluorescence staining and digital pathology. To investigate the functional role of CAFs in CRC, we took advantage of two murine models of colorectal neoplasia and advanced imaging technologies. In loss-of-function and gain-of-function experiments, using genetically modified mice with collagen type VI (COLVI)-specific signal transducer and activator of transcription 3 (STAT3) targeting, we evaluated STAT3 signalling in fibroblasts during colorectal tumour development. We performed a comparative gene expression profiling by whole genome RNA-sequencing of fibroblast subpopulations (COLVI+ vs COLVI–) on STAT3 activation (IL-6 vs IL-11).
Results The analysis of pSTAT3 expression in CAFs of human TMAs revealed a negative correlation of increased stromal pSTAT3 expression with the survival of colon cancer patients. In the loss-of-function and gain-of-function approach, we found a critical role of STAT3 activation in fibroblasts in driving colorectal tumourigenesis in vivo. With different imaging technologies, we detected an expansion of activated fibroblasts in colorectal neoplasias. Comparative gene expression profiling of fibroblast subpopulations on STAT3 activation revealed the regulation of transcriptional patterns associated with angiogenesis. Finally, the blockade of proangiogenic signalling significantly reduced colorectal tumour growth in mice with constitutive STAT3 activation in COLVI+ fibroblasts.
Conclusion Altogether our work demonstrates a critical role of STAT3 activation in CAFs in CRC development.
- colorectal cancer
Statistics from Altmetric.com
Significance of this study
What is already known about this subject?
IL-6 and IL-11 are signal transducer and activator of transcription 3 (STAT3) activators and frequently found upregulated in colorectal cancer (CRC).
STAT3 activation in tumour epithelial cells is critically involved in the pathogenesis of CRC.
Cancer-associated fibroblasts (CAFs) can modulate the tumour microenvironment in CRC.
What are the new findings?
Phosphorylation of STAT3 (pSTAT3) expression in CAFs is inversely correlated with the survival in CRC patients.
IL-6 and IL-11 induce partly overlapping gene expression patterns in fibroblast subsets in CRC.
STAT3 activation in CAFs plays a key role in the cross-talk to the tumour epithelium.
STAT3 activation in the COLVI+ fibroblast subset controls intestinal tumour growth in vivo.
How might it impact on clinical practice in the foreseeable future?
pSTAT3 in CAFs could serve as a prognostic tool in CRC patients and CAF activation might evolve as therapeutic target in CRC.
Colorectal cancer (CRC) belongs to the most frequent malignant tumours worldwide and its therapeutic options are still limited.1 Although multiple efforts have contributed to a better characterisation and stratification of the disease,2 the molecular mechanisms guiding CRC growth are incompletely understood.
Tumour stromal cells have gained increasing attention during the last years, as they can potently influence the tumour microenvironment, thereby modulating growth and progression of tumours.3 4 Of note, special characteristics of the tumour stroma such as T cell infiltration have been correlated with prognosis in CRC.5
Fibroblasts of the tumour stroma, so-called cancer-associated fibroblasts (CAFs), are a heterogeneous cell population with distinct cellular origin and expression of different marker like vimentin (VIM), fibroblast-specific protein-1 (FSP-1) or platelet-derived growth factor receptor-α (PDGFR-α). In addition, these cells have similarities with activated fibroblasts like the expression of α-smooth muscle actin (α-SMA).6 On the one hand, CAFs can promote tumour progression by release of growth factors such as epiregulin5 and hepatocyte growth factor.7 On the other hand, some CAFs have also the capacity to repress tumour development.8 Moreover, CAF-derived molecules can exert opposing effects on tumour growth and progression depending on the tumour stage and particular molecular context, for example, transforming growth factor β (TGF-β_ was reported to suppress tumour initiation and early tumour growth, whereas it promotes tumour progression and metastasis in later stages of the disease.4 9
The signal transducer and activator of transcription 3 (STAT3) is a pleiotropic transcription factor which plays an important role in the regulation of genes involved in survival, proliferation, migration and apoptosis.10 11 STAT3 can be activated by a variety of cytokines, including IL-6 and IL-11, which are frequently present in CRC.10 12 Activation of epithelial STAT3 has been demonstrated to regulate intestinal homeostasis and to influence CRC development.11–14 Inactivation of STAT3 in haematopoietic cells has been shown to suppress CRC.15
However, the role of STAT3 signalling in fibroblasts in CRC has not been clarified yet. Here, we provide strong evidence that STAT3 activation in tumour fibroblasts is critical for the development of CRC.
STAT3 is activated in CAFs of human CRC and can serve as prognostic marker
Initially, we studied the stimulation of primary human colon fibroblasts with IL-6 or IL-11 which resulted in a strong phosphorylation of STAT3 (pSTAT3) as evaluated by immunofluorescence (IF) (figure 1A) indicating that this cell type can be directly activated by these two cytokines. Next, we analysed STAT3 activation of CAFs in human CRC by IF for pSTAT3 and VIM on cross-sections demonstrating a significantly higher number of spindle-shaped pSTAT3+VIM+ cells in human CRC tissue as compared with control colon tissue (figure 1B, online supplementary figure S1A). To assess STAT3 activation in CAFs in CRCs at a larger scale, we conducted IF and immunohistochemistry (IHC) staining for pSTAT3, VIM and CD45 on tissue microarrays (TMAs) carrying human colon cancer tissue of a 375 patient cohort which predominantly included specimens of stage I-III CRC. To distinguish CAFs from immune cells and tumour epithelial cells, semi-automated digital analysis was used, as shown in online supplementary figure S1B. Interestingly, we observed that whereas some of the tumours stained strongly positive for pSTAT3 in VIM+CD45– tumour stromal cells, other CRCs showed little or no staining, suggesting that there is substantial heterogeneity among CRCs with respect to STAT3 activation in CAFs (figure 1C). As our work included a cohort of patients with long-term follow-up, we could perform Kaplan-Meier analysis comparing the prognosis of patients with pSTAT3high versus pSTAT3low in CAFs. Strikingly, we detected a significant difference with respect to the overall survival (OS) of patients with pSTAT3low versus pSTAT3high in CAFs in stage I–III CRC patients, highlighting an inverse correlation of pSTAT3 expression in CAFs with OS (figure 1D–H). Thus, pSTAT3 activation of CAFs is found in human CRC and our data suggest that it could serve as useful prognostic tool in stage I–III CRC patients.
STAT3 is activated in murine CAFs of experimental CRC
Corresponding to the analysis of human specimens, we detected significant enrichment of pSTAT3+VIM+ cells in the stroma of tumours from the azoxymethane/dextran sodium sulfate (AOM/DSS) mouse model (figure 2A–B). As different cell types may contribute to pSTAT3 levels in total tumour tissue, we next analysed purified CAFs from the AOM/DSS model, thereby confirming that CAFs display elevated pSTAT3 levels (figure 2C). Of note, STAT3 activation could be further promoted by additional stimulation with IL-6. This is in agreement with a strong activation of STAT3 in colon fibroblasts and CAFs on stimulation with IL-6 or IL-11 (figure 2D). Next, we analysed the supernatants on cell stimulation with hyper-IL-6 (a designer fusion protein of the soluble IL-6 receptor (sIL-6R) and IL-6) which does not cross-react in quantitative IL-6 analysis by ELISA. Here, we detected a potent upregulation of IL-6 levels on stimulation with hyper-IL-6 or IL-11 (figure 2E). This finding was in agreement with a strong activation of STAT3 in CAFs on stimulation with IL-6 or IL-11 (figure 2F). Thus, observations in mice correspond with the human data and indicate that STAT3 is frequently activated in colon fibroblasts in the presence of IL-6 and IL-11.
Colon fibroblasts and CAFs targeted by ColVI-Cre mice
To validate efficient targeting of colon fibroblasts and CAFs, we generated reporter mice expressing the red fluorescent protein tdTOMATO under control of the ColVI promoter (tdTomatoColVI ) (figure 3A). Of note, tdTOMATO was highly expressed in the small and large intestine of unchallenged mice, but not in control mice as demonstrated by fluorescence microscopy of cross-sections (online supplementary figure S2A) and by multispectral fluorescence imaging (online supplementary figure S2B).
Fibroblasts purified from the colon and CAFs isolated from AOM/DSS tumours of reporter mice showed expression of tdTOMATO in about 20%–25% of cells (figure 3B), reflecting a VIM+ and PDGFR-α+ subpopulation of intestinal fibroblasts. Whereas CAFs isolated from colon tumours of AOM/DSS-treated reporter mice showed a strong expression of tdTOMATO with a similar relative frequency among fibroblasts (20%–25%) (figure 3C), the tdTOMATO protein was highly enriched in tumours as compared with tumour-free colon tissue as visualised by IF (figure 3D). Interestingly, CAFs expressing tdTOMATO formed dense net-like structures in colon tumours as demonstrated by IF. Additional imaging technologies including label-free multiphoton microscopy (online supplementary figure S3A) and raster-scanning optoacoustic mesoscopy (RSOM) (online supplementary figure S3B) confirmed the strong expression and spatial distribution of tdTOMATO expressing fibroblasts in colon tumours. IF analysis revealed coexpression of tdTOMATO with different fibroblast markers such as collagen type VI (ColVI), VIM, α-SMA and FSP-1, but not with epithelial marker E-cadherin (figure 3E). Thus, tdTomatoColVI reporter mice demonstrated efficient targeting of CAFs in colon tumours enabling to characterise the localisation and spatial distribution of CAFs within the tumour stroma.
Constitutive activation of STAT3 in colon fibroblasts promotes tumourigenesis in mice
In a gain-of-function approach, ColVI-Cre mice were crossbred with conditional transgenic mice allowing for constitutively active STAT3 protein (STAT3C) expression (R26Stat3CSTOPfl/fl ) resulting in mice expressing STAT3C in ColVI+ fibroblasts (R26Stat3CColVI ) (figure 4A). R26Stat3CColVI mice were evaluated in the AOM/DSS model. Interestingly, R26Stat3CColVI mice displayed increased tumour formation as compared with littermate controls by mini-colonoscopy and light sheet fluorescence microscopy (LSFM) (figure 4B, online supplementary figure S4A and online supplementary videos S1, S2). Macroscopic inspection (figure 4C) and quantification of tumours revealed that both the number of tumours and the tumour load of R26Stat3CColVI mice in the AOM/DSS model were increased (figure 4D). Microscopic analysis confirmed the presence of large tumours in R26Stat3CColVI mice (figure 4E).
Interestingly, tumour tissue from R26Stat3CColVI mice showed a significantly increased number of Ki67+EpCAM+ cells indicating enhanced proliferation as compared with tumour tissue from controls (figure 4F). Additionally, coculture experiments with intestinal epithelial organoids (IEOs) from wildtype mice and fibroblasts carrying constitutive active STAT3 resulted in a significantly increased proliferation of the IEOs as compared with cocultures using wildtype fibroblasts (figure 4G).
Thus, constitutive activation of STAT3 in fibroblasts was associated with increased proliferation of tumour epithelia and resulted in higher tumour loads in an experimental CRC model.
STAT3 inactivation in colon fibroblasts reduces the development of colon tumours in mice
To address the effects of STAT3 inactivation in fibroblasts during colorectal tumourigenesis, ColVI-Cre mice were crossed with Stat3fl/fl mice (Stat3ΔColVI ). Stimulation of colon fibroblasts with IL-6 resulted in potent STAT3 phosphorylation in cells from control Stat3fl/fl mice, but not from Stat3ΔColVI mice confirming efficient inactivation of STAT3 in target cells in Stat3ΔColVI mice (figure 5A).
Next, we studied the development of colorectal tumours in Stat3ΔColVI mice in the AOM/DSS model. Strikingly, we observed that Stat3ΔColVI mice were highly protected from colorectal tumour growth as compared with littermate controls by mini-colonoscopy (figure 5B). This difference could be also detected by ex vivo analysis of colon tissue using blood vessel staining with anti-CD31 and LSFM (online supplementary figure S4B). Stat3ΔColVI mice exhibited both reduced tumour numbers and average tumour sizes, as compared with littermate controls suggesting a critical role of STAT3 activation in tumour fibroblasts during both tumour initiation and tumour growth promotion (figure 5C–E).
Corresponding to our previous findings, significantly decreased proliferation of tumour cells was detected in tumour tissue from Stat3ΔColVI mice as compared with controls and demonstrated by IF of the proliferation marker Ki67 (figure 5F). Next, we performed coculture experiments of IEOs with STAT3-deficient fibroblasts and controls, respectively. Whereas the presence of fibroblasts promoted the proliferation of IEOs, this increase was highly diminished when cocultured fibroblasts were STAT3 deficient (figure 5G).
In orthotopic models for CRC in which either Stat3fl/fl or Stat3ΔColVI fibroblasts were injected together with MC-38 tumour cells or APTAK organoids into the submucosa of Rag1 –/– mice, tumour load was significantly reduced when Stat3ΔColVI fibroblasts were used for the injection (online supplementary figures S5 A-B).
In summary, our loss-of-function approaches suggested a critical role of STAT3 activation in CAFs for the proliferation of IECs in vitro and the growth of colon tumours in vivo.
Gene expression profiling reveals different functional changes on STAT3 activation in colon fibroblast subpopulations
To address the potential mechanisms by which STAT3 activation in ColVI+ and ColVI– fibroblasts might guide tumour growth, we performed whole genome expression analysis by RNA-seq. Here, we purified colon fibroblasts from tdTomatoColVI reporter mice and sorted them in two cell populations (tdTOMATO+ and tdTOMATO–) by fluorescence-activated cell sorting (FACS) sorting (online supplementary figure S6B). Then, both tdTOMATO+ and tdTOMATO– fibroblast populations were left unstimulated or separately stimulated with the STAT3 activators IL-6 or IL-11, and RNA-seq analysis of 12 samples was performed (online supplementary figure S6A). Expression profiles were compared in groups, and advanced significance analysis revealed 99 transcripts on IL-6 stimulation and 524 transcripts on IL-11 stimulation with significant regulation (p≤0.05 after Benjamini-Hochberg correction). Hierarchical clustering as visualised by heatmap analysis revealed groups of transcripts with similar expression patterns (figure 6A). Interestingly, the stimulation by IL-6 or IL-11 resulted predominantly in an upregulation of genes and demonstrated similar patterns, as indicated in the summary of genes with highest upregulation or downregulation (for both conditions) (figure 6B–C). Of note, immune response-related genes (eg, Il4ra), proliferation-associated genes (eg, Bcl3) and the protease inhibitor Serpina3i belonged to the genes with highest upregulation in both conditions. In addition, the STAT inhibitor Socs3 was found among the top induced genes in both conditions, indicating the induction of counter-regulatory mechanisms on STAT3 activation. In contrast, diverse ribosomal proteins (eg, Rps19-ps7) were downregulated on stimulation by IL-6 or IL-11. Whereas the transcriptional regulation displayed substantial overlap between IL-6 and IL-11 stimulation, we detected various genes that were differentially influenced by either IL-6 or IL-11, for example, Osmr and Il1r1 by IL-6; and IL20ra and various additional serpin family members by IL-11. Interestingly, IL-6 did not only activate STAT3 but it also promoted the transcriptional induction of STAT3 mRNA expression (figure 6B). In addition, IL-6 was one of the 20 most upregulated transcripts on stimulation with IL-11 (figure 6C). Taken together, these findings suggest several mechanisms amplifying STAT3 signalling induced by IL-6 and IL-11. Correspondingly, gene ontology (GO) enrichment analysis revealed that several GO terms such as STAT3 pathway, molecular mechanisms of cancer or CRC metastasis were significantly upregulated in both of the stimulation groups (online supplementary figure S6C-D).
RNA-seq experiments were designed to compare functionality between different fibroblast subpopulations (COLVI+ vs COLVI–) on sorting. In fact, 78 and 229 transcripts were significantly upregulated on stimulation with IL-6 and IL-11, respectively, in COLVI+ (tdTOMATO+), but not in COLVI– (tdTOMATO–) fibroblasts (figure 6D). Interestingly, several transcription factors (eg, Myc, Fos) and kinases (eg, PI3KR5, AKT2) were found in these groups of transcripts and a GO-based analysis revealed a significant enrichment of signalling pathways including platelet-derived growth factor (PDGF) and vascularendothelial growth factor (VEGF) (angiogenesis), STAT3 and IL-6 signalling (figure 6D). By contrast, the list of transcripts with selective upregulation in COLVI– (tdTOMATO–) fibroblasts included genes coding for ribosomal proteins (eg, RPS12, RPS15A, RPS24) and were associated with GO terms including EIF2 signalling, mechanistic target of rapamycin (mTOR) signalling and interferon signalling (figure 6D).
In summary, RNA-seq analysis demonstrated various functional changes on STAT3 activation by IL-6 or IL-11 in colon fibroblast subpopulations.
In vivo inhibition of angiogenesis reduces tumour load in R26Stat3CColVI mice and in vivo activation of STAT3 promotes tumour development in Stat3ΔColVI mice
Based on the transcriptional patterns associated with PDGF and VEGF signalling regulated by STAT3 activation, we hypothesised that STAT3 activation in COLVI+ fibroblasts influences angiogenesis. Therefore, we evaluated tumours from Stat3ΔColVI and R26Stat3CColVI mice with respect to vascularisation. Interestingly, we observed that tumours from Stat3ΔColVI mice showed a diminished vessel density, as shown by anti-CD31 blood vessel staining and LSFM (figure 7A–B). Correspondingly, tumours from R26Stat3CColVI mice revealed a higher vessel density compared with tumours from littermate controls (figure 7C–D).
Next, we wondered whether angiogenesis could also represent a druggable downstream target during colon tumour growth promotion mediated by STAT3 activation in COLVI+ fibroblasts in vivo. Hence, we studied R26Stat3CColVI mice in the AOM/DSS model using the receptor tyrosine kinase (RTK) inhibitor dovitinib for the blockade of PDGFR and VEGR signalling (figure 7E). Strikingly, interference with proangiogenic signalling by dovitinib significantly reduced tumour growth suggesting that the promotion of neovascularisation/angiogenesis contributes to the fibroblast-mediated promotion of tumour development in R26Stat3CColVI mice (figure 7F–G, online supplementary figure S7A). Thus, in vivo inhibition of angiogenesis counter-regulated enhanced tumour development in R26Stat3CColVI mice.
As the above findings suggested that STAT3-induced and fibroblast-induced IL-6 might contribute to the proliferation of epithelial cells and the growth of tumours, we evaluated the effect of IL-6 and IL-11 injection on tumour growth in the AOM/DSS model in Stat3ΔColVI mice (figure 7H). The tumour load was significantly increased in mice treated with IL-6 and to a lesser extent in mice given IL-11, suggesting a critical role of both fibroblast-derived cytokines for tumour development (figure 7I–J). Furthermore, the proliferation marker Ki67 was significantly increased in epithelial cells of AOM/DSS tumours from mice treated with IL-6 or IL-11 (figure 7K).
Thus, in vivo inhibition of angiogenesis reduces tumour load in R26Stat3CColVI mice, while IL-6 and IL-11 promote tumour development in Stat3ΔColVI mice.
Although previous studies identified CAF activation signatures in CRC patients with poor prognosis,16 the molecular signalling pathways driving CAF activation in CRC have been incompletely understood. Here, we uncovered a key role of the transcription factor STAT3 in inducing CAF activation. By analysing 375 human CRC tissue samples, we detected augmented STAT3 activation in CAFs of subgroups of CRC patients. Subsequent quantification of pSTAT3 in CAFs using TMAs and digital pathology revealed that increased pSTAT3 in CAFs is associated with reduced OS in CRC patients indicating that STAT3 activation in these cells is inversely correlated to clinical prognosis. Studies in experimental CRC models using gain-of-function and loss-of-function approaches identified a crucial regulatory role of STAT3 activation in fibroblasts for tumour growth. Finally, studies in organoid systems and cell culture identified IL-6 and IL-11 as central regulators of STAT3 activation in these cells and demonstrated that STAT3 proficient fibroblasts drive proliferation and expansion of intestinal epithelial cells. Collectively, these findings uncover a crucial regulatory role for STAT3 in controlling activation of tumour-associated fibroblasts and identify STAT3 in CAFs as a potential target for therapy in CRC.
Cellular heterogeneity within individual tumours is a common finding in tumour tissue and local differences of epithelial STAT3 activation between the centre and the invasion front can be found in CRC. In fact, higher epithelial pSTAT3 was observed at the invasive front of tumours as compared with the tumour core.12 Correspondingly, the inverse correlation of elevated pSTAT3 expression in CAFs with clinical prognosis for stage I–III CRC in our dataset was observed with TMA samples obtained from the tumour centre, but STAT3 activation in CAFs might be even more relevant at the invasion front. In line with this concept, a critical role of CAF-derived signals originating from the tumour edge during tumour expansion and therapy has been demonstrated.17 Whereas previous work by others demonstrated the negative prognostic value of high STAT3 activation in tumour epithelial cells in CRC,18 our work highlights the contribution of STAT3 activation in CAFs in guiding tumour progression. Moreover, we identified IL-6 and IL-11 as key activators of STAT3 that are known to be induced in CRC tissue.9 12 14 In agreement with a cytokine-driven activation loop of STAT3 controlling CRC progression, previous data demonstrated a negative outcome of CRC in patients with elevated serums levels of IL-6.19 CAFs have been suggested to control tumour initiation, expansion and stemness in CRC.7 16 20 To explore the functional role of STAT3 in CAFs, we analysed the role of STAT3 activation in tumour fibroblasts during colorectal carcinogenesis focusing on the chemical-based AOM/DSS model using genetically modified mice.21 ColVI-Cre mice targeting collagen VI expressing cells were chosen as an outstanding functional relevance with a critical role in intestinal tumourigenesis in vivo was previously demonstrated for this subpopulation of mesenchymal cells.22 However, targeting different or partly overlapping fibroblast subsets by other Cre-strains could be also interesting as earlier studies reported divergent outcomes when inhibitor of nuclear factor kappa-B kinase subunit beta (IKKβ) was conditionally targeted by different Cre-lines (ColVI-Cre vs Col1a2-CreER).22 23
Interestingly, we could demonstrate a critical role of pSTAT3 activation in ColVI+ fibroblasts. Specifically, loss-of-function and gain-of-function experiments using genetically modified mice with COLVI-specific STAT3 targeting demonstrated that STAT3 activation in fibroblasts plays a critical role during colorectal tumourigenesis in vivo. While constitutive activation of STAT3 in ColVI+ fibroblasts augmented tumour growth in vivo, deletion of STAT3 in these cells markedly impaired growth of colorectal tumours. Correspondingly, a significantly decreased tumour growth was observed in orthotopic tumour models after cotransfer of MC-38 tumour cells or APTAK organoids along with Stat3ΔColVI fibroblasts, as compared with cotransfer with wildtype fibroblasts. Although MC-38 tumour cells and APTAK organoids were established on a B6 background, our experiments were performed in Rag1 –/– recipient mice as transplantation studies into immunocompetent syngeneic hosts are typically challenged by a lower engraftment rate and reduced tumour sizes when compared with transplantations into immunodeficient hosts.24 25 Altogether, our findings provide further evidence for a critical role of STAT3 activation in CAFs in CRC.
As the role of a tumourigenic microbiota has been well established in CRC,26 we performed our experiments with cohoused littermate control mice to demonstrate that our observations are stably based on genotypes and not driven by a transferable gut microbiota. Thus, STAT3 is a key driver of CAF activation and this finding may at least partially explain recent observations recognising a critical contribution of CAFs in shaping the tumour stroma and the growth conditions in colorectal tumours.5 27 As previous studies demonstrated a tumour-promoting role for STAT3 activation in IECs11 13 14 28 and myeloid immune cell populations (macrophages/dendritic cells (DCs)),15 29 our findings additionally highlight a crucial role of STAT3 activation in CAFs in CRC. In fact, our data provide further evidence that STAT3 activation in tumour stromal cells supports tumour growth in CRC, suggesting that blockade of STAT3 activation or efficient inhibition of cytokines mediating STAT3 activation in CRC might offer potential therapeutic opportunities.
The cytokines IL-6 and IL-11 which are both known to be abundantly expressed in CRC9 12 14 were found to activate STAT3 in tumour-associated COLVI+ fibroblasts and these cytokines were previously shown to activate several pathways critically involved in tumour development including STAT3, mTOR and extracellular signal-regulated kinase (ERK).5 14 30 Our comparative RNA-seq studies of fibroblast subpopulations suggested a preferential activation of the IL-6/STAT3 pathway in COLVI+ fibroblasts, while a significant upregulation of mTOR was noted in COLVI– fibroblasts. However, gene expression analysis of IL-6ra and IL6st by qPCR did not reveal any significant difference between COLVI– (tdTOMATO–) cells and COLVI+ (tdTOMATO+) cells (data not shown). In addition, RNA-seq studies clearly indicated that also COLVI– (tdTomato–) fibroblasts are responsive to IL-6 stimulation and display gene expression patterns that partly overlap with the COLVI+ (tdTOMATO+) subset. However, these findings cannot rule out other alterations in signal transduction potentially influencing the susceptibility of fibroblast subpopulations with respect to IL-6 stimulation and STAT3 activation. Thus, our findings highlight the existence of different fibroblast subtypes in colorectal tumours that exhibit characteristic gene profiles and may contribute to the different molecular phenotypes and clinical courses observed in individual CRC patients.2 Comparative gene expression profiling by whole genome RNA-sequencing of FACS-sorted fibroblast subpopulations (COLVI+ vs COLVI–) on STAT3 activation under different stimulation conditions (IL-6 vs IL-11) revealed the regulation of transcriptional patterns associated with fibroblast activation, cytokine signalling and angiogenesis (eg, IL-6, Serpina3, Wnt4, VEGF) in COLVI+ fibroblasts suggesting that these pathways may play a role in exerting STAT3 functions in colorectal neoplasias. Consistently, blockade of PDGFR and VEGFR signalling using the RTK inhibitor dovitinib led to suppression of tumour growth in vivo suggesting that STAT3 effects in colorectal neoplasia may be at least partially due to modulation of proangiogenic signalling. Thus, STAT3-induced activation of VEGF may contribute to the well-known effects of VEGF in controlling tumour growth and extracellular matrix remodelling in CRC.1
Our findings underline the potential relevance of STAT3 as therapeutic target in CRC. Indeed, strategies using STAT3 inhibitors have shown promising results in several malignancies.31 32 Recently, it was reported that an unselected patient population did not globally benefit from a blockade of STAT3 in therapy-refractory stage IV metastatic CRC.33 However, recent data also suggested that STAT3 could be an important target for the treatment of subgroups of CRC, for example, in tumours with elevated pSTAT3 expression in IECs.33 Notably, we observed a marked overlap between patients with high pSTAT3 levels in CAFs and activated epithelial STAT3 in our patient cohort (~80%) highlighting the potential clinical relevance of pSTAT3 quantification in CAFs. Nevertheless, further studies have to clarify whether STAT3 activation in CAFs or other cells populating the tumour stroma will be suited to predict medical responses of STAT3-targeted therapies in CRC. However, our findings in stage I–III CRC further point towards investigation of strategies targeting STAT3 in adjuvant settings. In addition, strategies directly interfering with surface receptors or signalling molecules in activated fibroblasts might be promising for CRC therapy and suggest new avenues for targeted interventions.
Human colon cancer TMAs included a cohort of 375 patients.34 Clinical details can be found in online supplementary table 1. Briefly, tumour biopsies were taken from patients with primary colon cancer treated at the University Hospital Erlangen between 2002 and 2010. Patient information including age, gender and follow-up information like postoperative therapy and survival was recorded. IHC and IF staining, as described in the supplemental methods section of this paper (including online supplementary table 2 and 3 with primary and secondary antibodies), was performed on TMAs with around 100 patients per slide including three 0.6 mm core biopsies per tumour. TMA slides were scanned (Pannoramic MIDI II, Sysmex, Germany) and digital images were analysed. The open source software QuPath (V.0.1.2) for quantitative pathology was used for image analysis. Each slide was annotated using the automated TMA dearraying tool and the detection classifier to distinguish between tumour epithelium and stroma. pSTAT3 expression was measured by aid of a customised QuPath script, whereas CD45 expression was detected using the positive cell detection tool. Basically, both stainings were read as two categories (positive or negative). For the survival analysis, patients were stratified in two groups with low or high pSTAT3 expression in CAFs (cut-off: 2.5%). The detection was visually checked for all cores and then manually corrected in order to exclude staining artefacts. Further, CD45-expressing immune cells were manually excluded from the pSTAT3 quantification of tumour stromal cells by consideration of the two corresponding TMAs in parallel.
Mice with transgenic expression of the Cre recombinase under control of the ColVI promotor (ColVI-Cre)35 were crossed either with mice-bearing floxed Stat3 alleles (Stat3fl/fl )36 or with mice carrying a floxed stop cassette followed by a constitutive active Stat3 gene in the ROSA26 locus (R26Stat3CSTOPfl/fl )37 or with reporter mice expressing tdTOMATO. Rag1 –/– mice were described previously.38 In vivo experiments were performed with cohoused littermate controls and in accordance with protocols approved by the governments of Middle Franconia and Rhineland-Palatinate, Germany.
CRC in vivo models
Colitis-associated AOM/DSS model was performed as described previously.39 In orthotopic CRC models, MC-38 tumour cells40 or genetically engineered APTAK organoids (ApcΔ/Δ; Trp53Δ/Δ; Tgfbr2Δ/Δ; myristoylated human Akt; KrasG12D), provided by F Greten (Georg-Speyer-Haus Frankfurt, Germany), were endoscopically transferred along with colon fibroblasts into the submucosa of Rag1 –/– mice.
Briefly, APTAK organoids were generated from colon organoids of TgfβR2fl/fl x p53fl/fl mice with C57Bl6 background. The Apc deletion was introduced via CRISPR/Cas9 as previously described.41 Further, Wnt was removed from the culture media for selection of Apc–/– organoids. Tgfbr2 and Trp53 knockouts were generated by transient in vitro Cre expression (Addgene 11543) and organoids were selected with nutilin3 (an Mdm2 inhibitor). Finally, the organoids were stably infected with retrovirus expressing murine KrasG12D and myristoylated human Akt and selected with selection markers (puromycin and hygromycin).
The MC-38 tumour cells were cotransferred with fibroblasts at a ratio of 1:1 (105 cells/100 µL PBS). For cotransfer of APTAK organoids, we delivered~60–80 organoids (a half well from a 24-well plate) and 105 fibroblasts per injection in 100 µL PBS with 10% Matrigel.25 Tumour development was monitored and scored using the endoscopic ‘Coloview system’ (Karl Storz) as described before.39 42 Animals were sacrificed at day 12 (after MC-38 tumour cell cotransfer) and at day 17 (after APTAK organoid cotransfer) and the tumour size was scored (average diameter) on longitudinal opening of the bowel.
For in vivo cytokine treatments, Stat3ΔColVI mice and controls received recombinant IL-6 or IL-11 (2 µg per ip injection) (Immunotools) every 48 hours.
The tumour development of R26Stat3CColVI mice was monitored by mini-colonoscopy and in week 5 of the experiment animals were equally assigned to groups according to their tumour score as previously described.42 Then animals were treated with 30 mg/kg body weight/mouse of PDGFR/VEGFR inhibitor dovitinib (CHIR-258, Selleckchem) or vehicle po by a feeding needle every 48 hours.
Isolation of primary colon fibroblasts, CAF and intestinal organoids
Fibroblasts were purified from untreated mice and CAFs from AOM/DSS tumours as reported previously.5 43 Human fibroblasts were purified from biopsies obtained during colonoscopy using a similar protocol as for mice.44 IEOs were purified as described previously.43
Cell culture and coculture experiments
Fibroblasts were cultured in Dulbecco's Modified Eagle Medium (DMEM) F12 (Sigma-Aldrich) with 10% fetal calf serum (FCS) and 1% penicillin/streptomycin at 37°C with 5% CO2. Cells between the third and sixth passage were used for in vitro experiments. Stimulation of fibroblasts with IL-6, IL-11 or hyper-IL-6 was performed in DMEM F12 containing 1% penicillin/streptomycin. Hyper-IL-6 is a fusion protein of the sIL-6R and IL-645 provided by S Rose-John (University Kiel, Germany). IEOs were cultured as previously described.43 For coculture experiments, wildtype IEOs and colon fibroblasts were mixed resulting in ∼30 IEOs and 1.5×104 fibroblasts per well of an 8-well μ-chamber slides (ibidi) and cultured together for 4 days in Matrigel (Corning) using IEO medium.
Light sheet fluorescence microscopy
Blood vessel staining with AF647-conjugated rat anti-mouse CD31 antibody (clone MEC13.3, BioLegend), the perfusion of mice and preparation of tissue samples for LSFM imaging was performed as described previously.46 Z-stacks were acquired with the UltraMicroscope II Light Sheet Microscope (LaVisionBioTec) at the IMCES of the University Hospital Essen (Germany) and the OICE of the University of Erlangen-Nuremberg (Germany). 3D reconstruction of LSFM data and quantification of blood vessel density was performed with customised ImageJ macros kindly provided by B Schmid of the OICE Erlangen-Nuremberg (Germany). 3D animation of LSFM data was performed as described previously.47
Confocal laser scanning microscopy
To detected morphological changes of blood vessels after antiangiogenic therapy, blood vessels were measured using fluorescein isothiocyanate (FITC)-dextran and confocal laser scanning microscopy as previously described.48 Wildtype mice in an AOM/DSS experiment were treated with the PDGFR/VEGFR inhibitor dovitinib (CHIR-258, Selleckchem) or vehicle. Mice were anaesthetised through an ip injection of a solution with 15% ketamine and 10% xylazine in sterile isotonic saline. Each mouse was injected (intravenous) with 50 µL of 5% FITC-dextran 70 (Sigma-Aldrich) in PBS. The distal colon seized with tumours was carefully exposed by laparotomy. The confocal microscope probe of the CLE-device (Cellvizio LAB with Mini-Z probe, Mauna Kea Technologies) was oriented directly on the tumour. The blood vessel architecture in colon tissue of a healthy animal was used as control. The fractal dimension D, which represents the complexity of a structure and is higher for normal blood vessels compared with tumour blood vessels, was calculated with ImageJ.
Raster-scanning optoacoustic Mesoscopy
RSOM is a non-invasive, lable-free, high-resolution optoacoustic (OA) imaging technology, which allows imaging of tdTOMATO expressing cells at high frequencies (85–87 MHz). The RSOM system (RSOM Explorer P50, iThera Medical GmbH, München, Germany) was used, as previously described.49 The harvested colon of AOM/DSS treated tdTomatoColVI reporter mice and control mice was rinsed with PBS and opened longitudinally for RSOM measurement.
Quantitative data are displayed as mean values per group±SD and significant differences are indicated as * p≤0.05, ** p≤0.01. For statistical analysis, non-parametric tests were used including two-tailed Mann-Whitney test for comparisons of two groups and Kruskal-Wallis test in combination with Dunn’s multiple comparison post-test for comparisons of more than two groups. Statistical analyses of Kaplan-Meier survival curves were done with Gehan-Breslow-Wilcoxon test. All statistical analyses were performed with Prism V.5.
We wish to thank K Enderle, K Hofmann, T Kisseleva and I Zoeller-Utz for excellent technical assistance; P Tripal and T Fraaß (Optical Imaging Centre Erlangen, University of Erlangen-Nuremberg) for excellent assistance with imaging and the Imaging Center Essen (IMCES); A Ekici (Institute of Human Genetics, University of Erlangen-Nuremberg) for the RNA-sequencing data analysis; M Mroz and D Schönhöfer (Core Unit for Cell-Sorting of the University Hospital Erlangen) for cell sorting.
Correction notice This article has been corrected since it published Online First. The author names and affiliations have been updated.
Contributors CH, KS, AS, CIG, BS, SW, O-MT, VK, MJW, HF, SM, AG, MG, RG, SR-J, SK, GK, MV, AH, FG and CN provided reagents, protocols, samples or designed experiments. MP generated the genetically engineered APTAK organoids. CB analysed RNA-sequencing data. CH, KS, O-MT, VK and FK performed experiments. CH, MFN and CN analysed, discussed and interpreted data. CH, MFN and CN wrote the manuscript.
Funding This study was funded by the DFG (FOR2438 to CN and MFN, NE1927/2-2 to CN, SFB1181-C02 to CN and TRR241-A08 to CN), by the Interdisciplinary Centre for Clinical Research Erlangen (to CN), by the FAU Emerging Fields Initiative (to CN) and by the Rudolf Bartling Stiftung (to CN).
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.