Introduction
Colorectal cancer (CRC) remains a leading cause of cancer-related mortality worldwide. Existing treatment relies on surgery with the use of perioperative chemotherapy/chemoradiotherapy as adjuncts. However, given the highly heterogenous nature of CRC, recent research has focused on employing a precision medicine approach to identify targeted therapies against molecularly segregated CRC. This has been successful in a small subgroup of patients with advanced stage mismatch repair (MMR) deficient disease through the adoption of immune checkpoint inhibitors [
1].
The development of the consensus molecular subtypes (CMS) and cancer cell intrinsic subtypes (CRIS) for CRC in the last decade have enabled subtyping of patients with distinct differences in the tumor biology and prognosis [
1‐
3]. The identification of more clinically translatable histology-based subtyping methods including Tumor Stroma Percentage (TSP) and Glasgow Microenvironment Score (GMS) have been useful tools in identifying independently prognostic phenotypes [
2,
4‐
6]. The establishment of novel and repurposed therapeutics best applicable to each subtype remains a key area of research. There is a profound association between high TSP and poor outcome in CRC, which has been extensively validated. However, the mechanisms underlying this are poorly understood [
7].
In malignancy overexpression of the transcription factor signal-transducer and activator of transcription 3 (STAT3) has been identified in ovarian, pancreatic, gastric, and lung cancer and associated with poor patient outcomes [
8‐
11]. Hyperactivation of the pathway is linked to many hallmarks of cancer including epithelial to mesenchymal transition (EMT), proliferation, angiogenesis, and metastases [
12,
13]. These data suggest that STAT3 could be a promising therapeutic target for CRC patients, and JAKi could be repurposed for use in combination with existing therapies in CRC. However, it is not yet clear if there is a specific subset of CRC patients likely to respond to pathway inhibition.
We hypothesise that histological changes observed in high TSP cases could be underpinned by dysregulation of targetable cellular signaling pathways such as JAK/STAT3. Signal transduction is canonically initiated by IL6 binding membrane bound or soluble IL6R, which causes activation of one of four JAK proteins (JAK1, JAK2, JAK3, TYK2) within the cell [
14,
15]. This ultimately leads to phosphorylation of STAT3 at tyrosine 705, whereby STAT3 homodimerizes and translocates to the nucleus to act as a master regulator of cancer-promoting genes [
14]. STAT3 can become maximally activated through phosphorylation at serine 727 [
16]. Janus kinase inhibitors (JAKi), which prevent activation of STAT3 are already used in the clinic for patients with myeloproliferative disorders [
17].
In this study we aim to elucidate the therapeutic potential of inhibiting signal transduction using 3 repurposed JAKi in recapitulative disease models. We subsequently aim to establish a prognostic relationship of JAK/STAT3 expression relative to TSP in 3 retrospective CRC cohorts complemented by multi-omic interrogation of the impact of JAK/STAT3 overexpression.
Material and methods
Cell lines
Cell lines utilized included DLD-1, HT29, SW620, HCT116, SW837 and CCD18Co which were obtained through American Type Tissue Culture (ATCC) and cultured according to guidelines. For detection of pSTAT3tyr705, a solid phase sandwich ELISA kit was utilized (DYC4607B-2, R&D Systems, Minneapolis, MN, USA). Samples were prepared in 96 well plates and conditioned medium harvested 48 h post-treatment. For WST-1 cell viability assays, cells were plated, left to adhere overnight, and treated using a range of concentrations of Ruxolitinib (#S1378), Tofacitinib (#S5001) or AZD1480 (#S2162) for 72 h (Selleckchem, Houston, TX, USA). WST-1 reagent (#11,644,807,001, Roche, Basel, Switzerland) was added for 1 h and absorbance read at 450 nm using a TECAN Infinite PRO (TECAN Group Ltd, Zurich, Switzerland). Treated sample data were normalized to vehicle control samples. All experiments were performed to n = 3. Cells were tested for mycoplasma before commencing experiments.
Silencing of STAT3
To knockdown STAT3 in HCT116 cells, scramble sequence or si-RNA for STAT3 (4,390,844, Invitrogen, Waltham, MA, USA) was transfected into the cells with lipofectamine RNAiMAX (13,778,100, Invitrogen, Waltham, MA, USA) for 24 h before the cells were used for cell viability assay.
Mouse organoids
Organoid lines derived from
VillinCreER; KrasG12D/+;
Trp53fl/fl;
Notch1TG/+ (KPN) and
VillinCreER; Apcfl/+; KrasG12D/+; Trp53fl/fl;TgfbrIfl/fl (AKPT) mouse models (1 line each) were obtained from the Sansom Laboratory [
18]. The experiment was performed according to UK Home Office regulations (Project Licenses 70/8646, PP3908577, 60/4183) and was reviewed by local animal welfare and an ethical review committee at the University of Glasgow. Mice were housed in conventional cages at constant temperature (19–23 °C) and humidity (55% ± 10%) under a 12-h light–dark cycle and were allowed access to standard diet and water ad libitum. A KPN male mouse on a C57BL/6J (generation N8) background of 8 weeks of age were induced with a single intraperitoneal injection of 2mg tamoxifen and aged until clinical endpoint as evidenced by anemia, hunching and/or weight loss to generate small intestinal tumor organoid lines as previously described [
18]. Intracolonic induction in a 12-week-old female mouse
villinCre
ER Apcfl/+ KrasG12D/+ Trp53fl/fl Tgfbr1fl/fl on a C57BL/6 background (generation N6/N7) was performed under general anesthesia. Three 70µl 100uM dose of 4-hydroxy tamoxifen (H7904-5MG from Sigma) were injected from the mid-colon to distal colon into the sub-mucosa via a colonoscope (Karl Storz TELE PACK VET X LED endoscopic video unit). At clinical endpoint (weight loss with the presentation of hunching) colonic tumor tissue was collected from a single tumor that had grown and organoid cell lines were generated as previously described [
18]. KPN and AKPT tumors present as CMS4 disease. Organoids were cultured as described in supplementary data 2 [
16].Organoids were grown in advanced DMEM was supplemented with B27 (12,587,001, Thermo Fisher Scientific, Waltham, MA, USA), N2 (17,502,001 Thermo Fisher Scientific, Waltham, MA, USA), murine Noggin (250–38, Peprotech, London, UK) and EGF (AF-100–15, Peprotech, London, UK). Organoids were grown in 20µL domes of Cultrex® type R1 basement membrane extract (BME) (R&D Systems, Minneapolis, MN, USA) in 6 well plates. For WST-1 assays organoids were grown in 5µL domes in 96 well plates and assays were completed as above. IF staining post-drug treatment for Caspase 8 (NB100-56116, Novus Biologicals, Abingdon, UK), and Ki67 (14–5698-80, Invitrogen, Waltham, MA, USA) was performed as previously described (20).
Patient-derived organoids (PDOs) and explants
Establishment and culture of PDOs was performed as previously described [
17]. For drug screening experiments tumor PDOs were seeded as fragments in 5μL Matrigel (Corning Inc, Corning, NY, UCA) in 96 well plates and grown for 48 h using IntestiCult™ human organoid medium (#06010, Stemcell Technologies, Vancouver, Canada) supplemented with RHO/ROCK pathway inhibitor (72,308. Stemcell Technologies, Vancouver, Canada). Medium was removed and replaced with either fresh medium, 0.1% DMSO, 1 µM-100 µM Tofacitinib, 1 µM-100 µM Ruxolitinib or 1 µM-100 µM 5-fluouracil (5FU) in triplicate. Plates were incubated for 72 h at 37 °C 5% CO
2. WST-1 reagent (#11,644,807,001, Roche, Basel, Switzerland) was added, plates were incubated at 37 °C 5% CO
2 for 2 h and absorbance read at 450 nm using a TECAN Infinite PRO (TECAN Group Ltd, Zurich, Switzerland). After drug treatments, 3 patient-derived organoid Sanger lines (25, 31, 37) were expanded sufficiently to perform IF staining for markers of mid-phase apoptosis (Caspase 8, NB100-56116, Novus Biologicals, Abingdon, UK) and proliferation (Ki67, #M7240, Agilent Technologies Santa Clara, CA, USA). IF staining was performed as previously outlined (20).
Explants were derived from CRC and adjacent normal tissue from surplus resection tissue from patients undergoing surgery with curative intent within Greater Glasgow and Clyde NHS hospitals through the Glasgow Biorepository (ETHICS). Tissue was chopped into small pieces (3 mm) and place in DMEM overnight at 5%CO2 37 °C. Medium was replaced after 24 h with vehicle control, Ruxolitinib or Tofacitinib containing medium and left for 48 h. Explants were washed in PBS and fixed ion 4% PFA for 1 h before being transferring to ethanol and embedded into paraffin blocks. Sections were cut at 4 μM thickness and put onto glass slides. Sections were stained via IHC for Ki67 (1:150) (Agilent Technologies Dako), MHCI (1:400) (BNB120-6405 Novus Biologicals, Littleton, Ontario, CA) using a Leica Bond Rx (Leica Biosystems, Wetzlar, Germany).
Colorectal cancer clinical cohorts
We investigated 3 independent CRC resected patient cohorts, including the TransSCOT clinical trial cohort, with patient characteristics outlined below and conformed to REMARK guidelines.
Cohort 1: Glasgow combined array (ethical approval WS/16/0207)
Cohort 1 consisted of 1030 CRC patients who underwent surgery with curative intent in Glasgow Royal Infirmary, Western Infirmary or Stobhill hospitals, Glasgow between 1997 and 2007. Patients were graded using the 5th edition of TNM staging. The cohort comprised patients with stage I-IV disease. In cohort 1 32% (n = 230) patients were < 65 years of age and 68% (n = 494) were > 65 years of age with a median age of 69. This cohort consisted of 49% female and 52% male patients. Clinical follow up data were last updated in 2017 from NHS Greater Glasgow and Clyde Safe Haven data. At this time, 324 patients (32%) had died of primary colorectal cancer, 332 patients (33%) had died of other causes and 355 patients (35%) were still alive. Cancer-specific survival (CSS), (date of surgery until last follow up) was used as a clinical endpoint throughout this study. Mean follow up time was 139 months.
Cohort 2: The TransSCOT clinical trial cohort [19] (ethical approval 07/S0703/136 until 2019, 16/WS/0207 2019 onwards)
In addition to retrospective cohorts, we analysed the prospective TransSCOT clinical trial cohort (cohort 2) [
19]. The TransSCOT clinical trial cohort consisted of 2912 patients from the SCOT clinical trial. Patients were graded using the 7th edition of TNM staging. The cohort consisted of stage III and high-risk stage II patients (≥ 1 of T4 disease, tumor obstruction with or without preoperative perforation, < 10 lymph nodes harvested, poor tumor differentiation perineural invasion or extramural venous invasion or lymphatic invasion). In the TransSCOT cohort 54% of patients were < 65 years of age. Patients were randomly assigned to a treatment arm, with both CAPOX (capecitabine and oxaliplatin) and FOLFOX (bolus and infused fluorouracil with oxaliplatin) regimens utilized over 3- or 6-month duration. The clinical outcome measure used for cohort 3 was disease-free survival (DFS) (date of surgery/randomisation until date of recurrence or all-cause mortality). Patients were followed up for at least 3 years and at this time there were 2221 (76%) patients alive and 691 (24%) patients who had died of cancer. The mean follow-up time was 35 months.
Cohort 3: Glasgow royal infirmary array (ethical approval MREC/01/0/36)
Cohort 3 consisted of 784 CRC patients who underwent surgery with curative intent within Greater Glasgow and Clyde health board between 1997–2013. This cohort comprised patients diagnosed with stage II-IV disease and clinical outcome was measured via CSS. In this cohort 33% (n = 233) patients were < 65 and 67% (n = 484) patients were > 65 years of age at the time of surgery. Of the patients included 45% were female and 55% were male. Follow up was updated in 2020 from NHS Greater Glasgow and Clyde Safe Haven. At this time 275 (35%) were alive, 231 (30%) had died of primary cancer and 277 (35%) had died of other causes. Mean follow up time was 89 months.
Immunohistochemical assessment of JAK1, JAK2, STAT3, pSTAT3tyr705 and pSTAT3ser727
Immunohistochemical staining was performed to detect IL6R (#ab128008, Abcam, Cambridge, UK), JAK1 (#3344, Cell signaling, Danvers, MA, USA), JAK2 (#3773, Cell signaling, Danvers, MA, USA), STAT3 (#9132, Cell signaling, Danvers, MA, USA), pSTAT3tyr705 (#9131, Cell signaling, Danvers, MA, USA) and pSTAT3ser727 (#9134, Cell signaling, Danvers, MA, USA) in cohort 1. Staining for pSTAT3tyr705 was subsequently performed in cohort 2 and cohort 3 as validation. Briefly, sections were dewaxed and rehydrated through a series of graded alcohols. Antigen retrieval was performed using TRIS–EDTA pH9 (pSTAT3tyr705, pSTAT3ser727) or citrate buffer pH6 (JAK1, JAK2, STAT3) by heating for 5 min under pressure. Endogenous peroxidases were blocked in 3% H2O2 for 10 min. Sections were rinsed in water and then blocked using 5% casein (JAK1), 10% casein (STAT3) or 5% horse serum (pSTAT3tyr705) at room temperature. Primary antibodies were added (JAK1 (1:100), JAK2 (1:100), STAT3 (1:300), pSTAT3tyr705 (1:50), pSTAT3ser727 (1:400) and incubated overnight at 4 °C. Sections were washed in tris-buffered saline (TBS) and incubated for 30 min in ImmPRESS (Vector Laboratories Inc, Burlingame, CA, USA) (JAK1, JAK2, STAT3, pSTAT3ser727) or Envision (Agilent Technologies, Santa Clara, CA, USA) secondary (pSTAT3tyr705). Sections were washed again in TBS, and DAB substrate was added for 5 min. After rinsing in water, sections were counterstained in Harris haematoxylin, dipped in acid alcohol, blued in Scots tap water, dehydrated through a series of alcohols, and placed in Histoclear. Mounting was performed using Omnimount (HS-110, SLS, Nottingham, UK). Scanning was performed using a Hamamatsu Nanozoomer (Hamamatsu, Hertfordshire, UK) at X20 and images were visualized on the NDP Platform (Hamamatsu, Hertfordshire, UK).
The weighted histoscore method was utilised to semi-quantitatively measure the intensity of staining detected in the tumor cell nuclei (pSTAT3
tyr705, pSTAT3
ser727 and JAK1, JAK2) and expression in any part of stromal cells (pSTAT3
tyr705). The proportion of nuclei were assessed for negative, weak, moderate, and strong staining for each marker by a single observer (KP) blinded to clinical outcomes. The following calculation allowed for a score ranging from 0–300 to be determined for every core; (0* % negative) + (1* %weak) + (2*%moderate) + (3*%strong). The TMA included 3 cores per patient to account for tumor heterogeneity. For validation of manual scores 10% of cores were scored digitally using QuPath [
19] by a second observer (SAB) blinded to clinical outcomes. Scatter plots were constructed to visualize correlation between manual and digital scores and for each marker intra-class correlation coefficients of > 0.7 were achieved (Additional file
1, Figure S
1).
Immunohistochemical assessment of, pSTAT3tyr705 in AKPT and KPN murine tumors
Immunohistochemical staining to detect pSTAT3tyr705 was performed on a KPN and an AKPT tumor to determine constitutive activation of the pathway in vivo. A KPN male mouse on a C57BL/6J (generation N5-6) background of 10 weeks of age were induced with a single intraperitoneal injection of 2mg tamoxifen and aged until clinical endpoint as evidenced by anemia, hunching and/or weight loss to generate small intestinal tumor for IHC staining. Intracolonic induction in a 16-week-old female mouse villinCreER Apcfl/+ KrasG12D/+ Trp53fl/fl Tgfbr1fl/fl on a C57BL/6 background (generation N6/N7) was performed under general anesthesia. Three 70µl 100uM dose of 4-hydroxy tamoxifen (H7904-5MG from Sigma) were injected from the mid-colon to distal colon into the sub-mucosa via a colonoscope (Karl Storz TELE PACK VET X LED endoscopic video unit). At clinical endpoint (weight loss with the presentation of hunching) colonic tumor tissue (spontaneous SI tumour) was collected from a single tumour for IHC staining.
Staining was performed using pSTAT3tyr705 antibody (#9131, Cell signaling, Danvers, MA, USA) at 1:100 on a Leica BOND Rx autostainer (Leica Biosystems, Wetzlar, Germany).
Assessment of pSTAT3tyr705 and pSTAT3ser727 colocalization in each compartment of the tumor microenvironment
Multiplex immunofluorescence was performed to determine the importance of colocalization of both phosphorylation sites of STAT3 in patients from cohort 1. TMAs were baked for 1 h at 60 °C. Dewaxing was performed using a PT module (Epredia, Runcorn, UK) by heating slides to 95 °C for 35 min in pH6 buffer (TA-999-DHBL, Dewax and HIER Buffer L, Epredia, Runcorn, UK). Staining was performed using a Lab Vision 480S Autostainer (Thermo Fisher Scientific, Waltham, MA, USA). Staining was performed using an UltraVision Quanto Detection System HRP kit (TL-060-QHL, Thermo Fisher Scientific, Waltham, MA, USA). Slides were washed in dH2O, quenched in 3% H2O2, washed in TBS and blocked in UVQ protein block (Thermo Fisher Scientific, Waltham, MA, USA). Sections were incubated with pSTAT3tyr705 (Cell signaling #9131) for 60 min and washed in TBS. UVQ amplifier (Thermo Fisher Scientific, Waltham, MA, USA) was applied for 10 min, washed off with TBS and then UVQ HRP (Thermo Fisher Scientific, Waltham, MA, USA) was added for 10 min. Sections were washed in TBS and incubated in Opal 480 at 1:300 (#SKU FP1500001KT, Akoya Biosciences, Marlborough, MA, USA) for 10 min before heating in the PT module. Slides were washed in dH2O, blocked in H2O2, washed in TBS, and blocked in UVQ protein block. Sections were incubated in pSTAT3ser727 antibody (Thermo Fisher Scientific, Waltham, MA, USA) at 1:400 for 30 min and then washed in TBST. UVQ amplifier was applied followed by washing in TBS, incubation with UVQ HRP, washing and addition of Opal 650 (SKU FP1496001KT, Akoya Biosciences, Marlborough, MA, USA) at 1:300 for 10 min. Sections were heated in the PT module for a final time, washed, incubated in DAPI for 5 min and then mounted using. Slides were scanned onto NDP using the Hamamatsu Nanozoomer.
Assessment of the mutational profile of TSPhigh patients with high and low pSTAT3 expression
Mutational profiling was performed on a subset of cohort 1 patients (
n = 252). DNA was extracted from formalin fixed paraffin embedded sections by NHS Tayside Centre for Genomic Analysis (NHS Tayside, Dundee, UK). DNA quality and concentration were determined using the Qubit assay (ThermoFisher, Massachusetts, USA). Sequencing was outsourced and performed by Glasgow Precision Oncology Laboratory (GPOL) using a custom panel of 196 cancer-associated genes (Additional file
6, Table S
2).
Assessment of the bulk transcriptional profile of TSPhigh patients with high and low pSTAT3 expression
Full CRC tissue resections from 49 patients from cohort 1 were profiled using TempO-Seq for detection of expression of the full transcriptome (~ 22,000 genes) (Biospyder Technologies, Carlsbad, CA, USA). This was performed as previously described [
20]. Normalisation of raw gene counts was performed in R Studio (RStudio, Boston, MA, USA) using DESeq2. The expression profile of high TSP/pSTAT3
tyr705 groups was analysed using publicly available software accessed at (
https://www.gsea-msigdb.org/gsea/msigdb/index.jsp). Accession numbers for the dataset are provided within the data availability statement of the manuscript.
NanoString digital spatial profiling (GeoMx™) of TSPhigh patients with high pSTAT3 expression
GeoMx™ profiling was performed on a subset of CRC patients from cohort 2 (n = 11) as previously described. A tissue microarray from cohort 3 was cut at 5 μm and baked for 30 min at 60 °C. Antigen retrieval was performed using a Leica BOND Rx autostainer (Leica Biosystems, Wetzlar, Germany) using ER2 buffer at pH9 at 100 °C for 10 min. Protein digestion was performed using proteinase K (0.1 μg/ml) for 15 min. In-situ hybridisation of RNA-directed DNA oligo probes (Nanostring Human Whole Transcriptome Atlas, 18,677 genes) was performed according to the manufacturer’s protocol.
Statistical analyses
Statistical analyses of IHC data were performed in SPSS version 28 (IBM, NY, USA). Kaplan Meier survival curves were plotted to determine association between groups and cancer-specific survival. To determine associations with clinicopathological features chi-squared tests were performed. Significance was set to p < 0.05. Cut points for high and low expression were determined in R Studio v1.4 (RStudio, MA, USA) R build version 4.2.1 and packages Survminer (version 0.4.9), Maxstat (version 0.7–25), Tidyverse (version 1.3.2) and Survival (version 3.5–0). These analyses determined the optimal cut off point using log rank statistics based on cancer-specific survival for cohort 1 and cohort 3, and disease-free survival for the TransSCOT clinical trial cohort (cohort 3).
Mutational profiling data were analysed in R Studio v1.4 (RStudio, MA, USA) using the Maftools (version 2.18.0) package. Fishers’ exact tests were utilized to determine any differential patterns of mutation between pSTAT3tyr705 and GMS patient groupings.
Normalized counts from the GeoMx experiment were downloaded and analysed in RStudio (v2022.07.2) using R build version 4.2.1. Differential Gene Expression (DGE) was performed using the Exact Test as part of edgeR package. Volcano plots were generated using EnhancedVolcano. Heat maps were generated using ComplexHeatMap. Gene set enrichment analysis (GSEA) was performed using the fgsea package. Single Sample Gene Set Enrichment Analysis (ssGSEA) was performed using the GSVA package. ClusterProfiler was used to interrogate the Reactome curated database. Immune spatial deconvolution of GeoMx derived WTA data was performed using SpatialDecon package. Receptor Ligand gene pairs were obtained from the CellPhoneDB repository.
Data from cell viability assays were analysed using paired T tests in GraphPad Prism version 9 (GraphPad Software, San Diego, CA, USA). Raw optical density read outs were averaged and normalized to vehicle controls. Paired t tests were used to determine any significant differences between treatment groups. To establish differences in immune profiles of high and low pSTAT3 groups within TSPhigh cases Mann–Whitney unpaired non-parametric T tests were performed in GraphPad Prism version 9 (GraphPad Software, San Diego, CA, USA). Significance was set to p < 0.05.
All figures were constructed in Adobe Acrobat version 20 (Adobe Inc, San Jose, CA, USA).
Discussion
In this study a novel association between JAK/STAT3 signaling pathway activation and poor prognosis has been identified in a specific subset of patients with stromal-rich (TSP
high) CRC tumors. Previous research has implicated STAT3 in CMS1 and CMS3 tumors [
22]. However, here we have shown that high expression of the main marker of pathway activation pSTAT3
tyr705 was associated with poor outcome in TSP
high cases, which more closely resemble CMS4 (
p = 0.006).
This interaction between STAT3 signaling and tumor-stroma has been noted in previous literature. In mouse models of pancreatic ductal adenocarcinoma (PDAC), inhibiting STAT3 in the tumor resulted in stromal remodelling [
23]. In vitro cell line studies of PDAC have shown that conditioned medium from Cancer Associated Fibroblasts (CAFs) induced STAT3 expression in PDAC-3 tumor cells at the transcriptomic and protein level [
23‐
25]. Similarly in models of non-small cell lung cancer (NSCLC) culturing tumor cells with primary CAFs induces STAT3 in the tumor cells and induces EMT [
25].
In the present study, JAK inhibition reduced cell viability expression pf pSTAT3tyr705 in HCT116 and HT29 CRC cell lines. Similarly, mouse derived KPN organoids and a subset of PDOs showed a significant decrease in the percentage of viable cells, apoptosis and proliferation following JAK/STAT3 inhibition using Ruxolitinib and AZD1480. In this study we showed JAK inhibitors were significantly less effective in cells silenced for STAT3, however off target effects should be investigated in future studies. Further work is needed to understand if this difference in responses and unravel the mechanisms underlying why specific PDOs showed better responses to JAKi.
Across 3 large CRC patient cohorts, high expression of pSTAT3tyr705 was associated with reduced CSS. This included data from the TransSCOT clinical trial cohort where we were able to show that patients with high pSTAT3tyr705 and TSPhigh had significantly better responses to FOLFOX over CAPOX regimes, and to treatment for 12 weeks over 24 weeks duration. This result needs to be validated in a subsequent cohort, however, highlights the potential for improving outcomes to existing therapeutics in the pSTAT3 high TSPhigh group.
To understand association with other phenotypes, mutational, bulk transcriptional, spatial, and immune profiling revealed differences in the tumors of patients with high pSTAT3 and TSPhigh. In the mutational analyses of patient cohort 1, there was an enrichment of AMER1 in the high pSTAT3tyr705 group, but significantly less SMAD4 mutations. These data suggested that aberrant JAK/STAT3 signaling in the CRC setting is unlikely to be mutationally driven. At the bulk transcriptomic level, differences were observed between patients with high/low pSTAT3tyr705 in patient cohort 1, however the biological relevance if this needs further investigation, due to overlap between groups despite statistical significance. Pro-tumorigenic Hallmark pathway TNFA signaling via NF-κB, and Apoptosis enriched in TSPhigh patients classified as high pSTAT3tyr705. Digital spatial profiling of a subset of patients from patient cohort 3 enabled interrogation of differential gene expression within the panCK positive, αSMA positive and TME areas of high pSTAT3 TSPhigh cases. These data have highlighted the milieu of dysregulated signaling in different spatial compartments of stromally dense STAT3 activated tumors. There was an upregulation of genes associated with immune exclusion, decreased anti-tumor immune response, altered neutrophil biology, non-classical fibroblast deposition and increased hypoxia in TSPhigh tumors with high pSTAT3. This was in concordance with data from the protein level from cohort 1 which showed high tumoral expression of PDL1 in TSPhigh tumors classified as high for pSTAT3tyr705. However, the biological relevance if this needs further investigation, due to overlap between groups despite statistical significance. Future research is required to validate these mechanisms in vitro/in vivo, and to determine if inhibition of STAT3 signaling can reverse the adverse phenotypes acquired in the high pSTAT3 TSPhigh tumors.
While the effect of STAT3 activation in tumor cells on the inflammatory infiltrate is well characterized, the effects on the tumor-stroma is less well-studied [
26]. The interplay between all 3 components of the TME is likely responsible for driving phenotypes which predict poor prognosis. We hypothesise that a feedback loop arises, whereby cytokines secreted by CAFs activate STAT3 in the tumor cells, which causes transcription of genes which promote stromal cell proliferation and recruitment of CD66b + cells to the TME. The increased stromal component is linked to tumor budding, hypoxia and EMT [
4,
27‐
29]. Activation of STAT3 in tumor cells causes increased expression of PDL1 resulting in immune evasion [
30]. Blocking STAT3 activation in patients with stromal-rich tumors could suppress these hallmarks of cancer and prevent tumor progression.
There is evidence that IL6 can not only activate STAT3 but also other STAT family members including STAT1. High expression of STAT1 generally confers good prognosis in solid tumors through promotion of anti-tumor immunity [
31]. Indeed, in previous tissue-based studies the ratio of STAT1:STAT3 in tumor cells was highly prognostic in CRC, and in prostate cancer loss of STAT1 associated with increased recurrence [
32,
33]. Further mechanistic work is needed to determine whether blocking STAT3 switches signal transduction to STAT1, in this study
STAT1 gene expression was upregulated in low STAT3 TSP
high tumors (
p = 0.0025). Future work should include coculture models to investigate JAKi effect on tumor cells in the presence of CAFs and various immune cell populations.
In conclusion, patients with TSPhigh tumors have the worst outcomes, regardless of segregation strategy including transcriptional level (CMS4/CRIS-B) or the histological level. This subgroup of patients warrants urgent identification of novel therapeutic options to improve survival. From this study, we have shown that JAK inhibitors should be investigated using a stratified medicine approach for patients with stromal-rich TSPhigh tumors.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.