BRAF status modulates Interelukin-8 expression through a CHOP-dependent mechanism in colorectal cancer

Inflammation might substantially contribute to the limited therapeutic success of current systemic therapies in colorectal cancer (CRC). Amongst cytokines involved in CRC biology, the proinflammatory chemokine IL-8 has recently emerged as a potential prognostic/predictive biomarker. Here, we show that BRAF mutations and PTEN-loss are associated with high IL-8 levels in CRC models in vitro and that BRAF/MEK/ERK, but not PI3K/mTOR, targeting controls its production in different genetic contexts. In particular, we identified a BRAF/ERK2/CHOP axis affecting IL-8 transcription, through regulation of CHOP subcellular localization, and response to targeted inhibitors. Moreover, RNA Pol II and an open chromatin status in the CHOP-binding region of the IL-8 gene promoter cooperate towards increased IL-8 expression, after a selective BRAF inhibition. Overall, our data show that IL-8 production is finely and differentially regulated depending on the tumor genetic context and might be targeted for therapeutic purposes in molecularly defined subgroups of CRC patients. Conciatori et al find that BRAF mutations and PTEN-loss promote IL-8 production in colorectal cancer cell (CRC) lines and identify a genetic-context-dependent BRAF/ERK2/CHOP molecular axis that controls IL-8 transcription. These data may assist in the identification of drugs to target CRC.

C ytokine networks contribute to the development and progression of cancer, particularly in Colorectal Cancer (CRC), in which inflammation represents a critical aspect of disease progression 1 . Within the Tumor MicroEnvironment (TME), both stromal and cancer cells release cytokines/chemokines and growth factors, thereby contributing to the cytokine networks, which modulate the inflammatory/immunologic milieu of cancer tissues 2 .
InterLeukin (IL)-8, also referred to as CXCL8, is a proinflammatory CXC ELR + chemokine; through the binding to its cell-surface G protein-coupled receptors, CXCR-1 and CXCR-2, IL-8 plays multiple roles in cancer, driving the activation of key signaling pathways in both stromal and intestinal epithelial cells, to promote or increase proliferation, angiogenesis and metastasis 3 . Recent evidence has shown a correlation between IL-8 overexpression and both Vascular Endothelial Growth Factor (VEGF)-independent tumor angiogenesis and chemoresistance in preclinical models, thus highlighting a role for IL-8 expression in CRC 4 . High levels of IL-8 are observed in the serum and cancer tissue of CRC patients and these levels significantly increase according to a worsening clinical stage and tumor grade [5][6][7] . A recent meta-analysis has suggested that high levels of IL-8 expression are significantly associated with poor prognosis in CRC patients (HR = 1.54, 95% CI 1.03-2.32) and correlate with advanced stage, lymphatic and liver metastasis, and resistance to antiangiogenic agents 8,9 .
IL-8 expression can be regulated at the transcriptional and post-transcriptional levels by multiple intracellular signaling pathways, including the Mitogen-Activated Protein Kinase (MAPK) and PhosphoInositide 3-Kinase (PI3K) 10,11 . However, the molecular mechanisms and transcription factor networks through which signaling pathways regulate IL-8 expression in specific, genetically defined, cancer contexts remain to be defined.
In this study, we have investigated the relationships between BRAF mutations/loss of Phosphatase and TENsin homologue deleted on chromosome 10 (PTEN) and IL-8 production and found that both alterations contribute to high levels of IL-8 production in a panel of genetically characterized CRC cell lines. IL-8 production was profoundly influenced by modulation of the MAPK pathway: in particular, BRAF inhibition by dabrafenib abrogated IL-8 production in BRAF-mut cell lines, but paradoxically increased it in BRAF-wt contexts, while MEK inhibition and Extracellular signal-Regulated Kinase (ERK)2 silencing selectively abrogated IL-8 production, regardless of the genetic background of the CRC cell lines examined. Double PI3K/ mammalian Target Of Rapamycin (mTOR) inhibition, on the other hand, did not substantially affect IL-8 production. Dabrafenib-induced IL-8 modulation was found to be mechanistically related to nuclear export of the C/EBP HOmologous Protein (CHOP) transcription factor in BRAF-mut cells and to CHOP nuclear retention and promoter binding in BRAF-wt contexts. Overall, our data highlight a BRAF/ERK2/CHOP regulatory axis which regulates both basal and drug-induced IL-8 expression in CRC models.

Results
BRAF mutations and PTEN-loss correlate with IL-8 production. We investigated the role of the MAPK and PI3K pathways in the regulation of IL-8 expression in a panel of 28 CRC cell lines, characterized for BRAF, Kirsten RAt Sarcoma (KRAS), PTEN and PI3K gene status (Table 1 and Supplementary Table 1). PTEN expression was also investigated at the mRNA and protein level (Supplementary Fig. 1 and Table 1): PTEN protein expression was completely absent in 12 CRC cell lines, moderate in 7, and strong in 9 of the tested cell lines. Cell lines carrying PTEN deletions or inactivating mutations or completely lacking PTEN protein expression are referred to as PTEN-loss 12 .
Cell culture media of the 28 CRC cell lines (BRAF-mut = 12; KRAS-mut = 9; PI3K-mut = 13; PTEN-loss = 12, Table 1 and  Supplementary Table 1) were analyzed by ELISA assay under standardized culture conditions (see Methods) (Fig. 1a). Statistical analysis showed a statistically significant correlation between IL-8 expression and BRAF status: indeed, the presence of a BRAF V600E mutation predicted IL-8 levels higher than 257.5 pg/mL with 58.33% sensitivity and 93.75% specificity (Area Under the Curve (AUC) = 0.76) and the Receiver Operating Characteristic (ROC) curve-based prediction algorithm based on BRAF mutation had 52% accuracy in predicting IL-8 production (p = 0.004) (Fig. 1b). Statistical analysis also showed a trend towards a statistically significant association with PTEN-loss: indeed, PTEN-loss predicted an IL-8 value higher than 42.5 pg/mL with a 64.29% sensitivity and 78.57% specificity (AUC = 0.71) and the ROC curve-based prediction algorithm based on PTEN-loss had 43% accuracy in predicting IL-8 production (p = 0.05) (Fig. 1b). Moreover, the highest levels of IL-8 were observed in CRC cell lines carrying both BRAF V600E and PTEN-loss (n = 5): indeed, combined BRAF/PTEN analysis predicted an IL-8 value higher than 46 pg/mL with 87.50% sensitivity and 80.00% specificity (AUC = 0.88) and the ROC curve-based prediction algorithm based on these two alterations had 68% accuracy in predicting IL-8 production (p = 0.002) (Fig. 1b). Conversely, KRAS and PI3K mutation status were not correlated with IL-8 expression ( Supplementary Fig. 2).
We also confirmed the specific correlation between IL-8 expression and PTEN status by using X-MAN™ isogenic HCT116 cell lines (HCT116 and HCT116 PTEN −/− ). To this purpose, cell culture media of X-MAN™ isogenic HCT116 cell lines were analyzed after 24 h of culture in serum-free medium by Human Angiogenesis Antibody Array, which revealed the selective expression of IL-8 only by HCT116 PTEN −/− (Fig. 1c). This result was confirmed by IL-8 specific ELISA assay (Fig. 1d).
In order to verify the specific correlation between BRAF/PTEN status and IL-8 production, we also evaluated the levels of VEGF and IL-6, two other pro-angiogenic soluble factors involved in CRC progression. Differently from IL-8 production, no significant correlation was observed between BRAF status and VEGF production ( Supplementary Fig. 3). However, in normoxic conditions of growth, PTEN-loss predicted VEGF levels higher than 621 pg/mL with a 92.86% sensitivity and 57.14% specificity (AUC = 0.77) and the ROC curve-based prediction algorithm based on PTEN-loss had 50% accuracy in predicting VEGF production (p = 0.01). KRAS-wt status also predicted VEGF levels higher than 627 pg/mL with a 80% sensitivity and 75% specificity (AUC = 0.75) and the ROC curve-based prediction algorithm based on KRAS-wt had 55% accuracy in predicting VEGF production (p = 0.01). IL-6 was not detectable in the culture media of the CRC cell lines examined ( Supplementary Fig. 4).
Taken together, these results demonstrate that both BRAF and PTEN status specifically determine IL-8 expression.
MAPK-dependent regulation of IL-8 expression. We next investigated whether specific inhibitors targeting BRAF, MEK, and PI3K/mTOR (dabrafenib, trametinib, and gedatolisib, respectively) could modulate IL-8 expression. To this purpose, four CRC cell lines (SNU1235, SNU1047, HT29, and LS180), differing for BRAF and PTEN status (BRAF V600E /PTEN-loss, BRAF-wt/PTEN-loss, BRAF V600E /PTEN-competent, BRAF-wt/ PTEN-competent) were exposed to increasing concentration of drugs for 24 h. As shown in Fig. 2a, b, double inhibition of PI3K/ mTOR by gedatolisib minimally affected IL-8 release, regardless of BRAF and PTEN status (see also Supplementary Fig. 5); similar to gedatolisib, selective PI3K (using alpelisib), AKT (using MK226), and mTOR (using everolimus) inhibition decreased IL-8 production by less than 50%, independent of the genetic background of the cell lines tested ( Supplementary Fig. 6). Effects of selective BRAF inhibition on IL-8 production were profoundly influenced by BRAF-mutational status: dabrafenib strongly inhibited IL-8 production in BRAF V600E cell lines (SNU1235 and HT29; Fig. 2a), while it increased IL-8 levels in BRAF-wt cell lines (SNU1047 and LS180; Fig. 2b); dabrafenib effects on IL-8 production did not differ qualitatively according to PTEN status (Fig. 2a, b and Supplementary Fig. 5). Conversely, MEK inhibition by trametinib (Fig. 2a, b) and ERK1/2 inhibition by SCH772984 ( Supplementary Fig. 7) profoundly suppressed IL-8 expression in all the tested cell lines, regardless of their genetic background. As shown in Fig. 2c the combination of dabrafenib and trametinib prevented dabrafenib-induced IL-8 upregulation in BRAF-wt contexts (SNU1047 and LS180), but did not further increase IL-8 inhibition as compared to trametinib alone in either genetic context or dabrafenib alone in BRAF V600E cell lines. Specific MAPK pathway-dependent IL-8 regulation was further confirmed by VEGF expression analysis: indeed, with the notable exception of BRAF-wt/PTEN-competent LS180 cell line, VEGF levels under normoxic conditions were much less affected by pathway inhibitors, regardless of the genetic background of the tested cell lines ( Supplementary Fig. 8a).
The role of individual MAPK pathway elements was further examined by BRAF, MEK, ERK1, and ERK2 silencing, using short interfering (si) RNA and short hairpin RNA (sh) RNA. As shown in Fig. 3a, IL-8 production was downregulated after BRAF, MEK, or ERK2 silencing, regardless of BRAF-mutational status in three of the four analyzed cell lines. More variable effects, namely IL-8 level increase after BRAF silencing and lack of IL-8 decrease after ERK2 silencing, were observed in the LS180 cell line (BRAF-wt/ PTEN-competent), which also harbors a KRAS G12D mutation (Table 1 and Supplementary Table 1). ERK1 silencing did not affect IL-8 expression in any of the tested cell lines. VEGF production was consistently less affected by the silencing of MAPK elements, regardless of BRAF and PTEN status, thus further suggesting a specific role for the MAPK pathway in the regulation of IL-8 expression in CRC ( Supplementary Fig. 8b).
To further analyze the mechanisms of dabrafenib-induced differential IL-8 regulation, we evaluated the effects of dabrafenib in the presence of specific silencing of MAPK elements (Fig. 3b).
CHOP-dependent transcriptional regulation of IL-8. We next analyzed IL-8 mRNA levels after dabrafenib and trametinib treatment by Real Time quantitative Polymerase Chain Reaction (RT-qPCR). As shown in Fig. 4a, mRNA modulation closely paralleled protein expression data: dabrafenib differentially affected IL-8 mRNA expression depending on the genetic context  Fig. 9a), suggesting a potential involvement of both Activator Protein (AP)-1 and CHOP in IL-8 regulation.
To formally prove their involvement, luciferase (luc) genes containing different IL-8 promoter constructs were co-transfected with pRL-TK into the four CRC model cell lines. Consistent with the known role of AP-1 and Nuclear Factor kappa-light-chainenhancer of activated B cells (NF-κB)/Nuclear Factor for IL-6 expression (NF-IL-6), lack or mutation of NF-κB and NF-IL-6 binding sites reduced luc activity, as compared to a full-length IL-8 promoter (546-luc) ( Supplementary Fig. 9b,c) 13 . However, a minimal NF-κB/ NF-IL-6 promoter (98-luc) was not able to sustain basal transcription and the lack of a portion or the complete CHOP-binding sites partially or completely abrogated luc activity, respectively (Fig. 4b, c). Overall, these data confirm the known involvement of AP-1 and strongly indicate CHOP as a relevant player in IL-8 transcription in CRC cell lines, both in basal conditions and in response to MAPK inhibition.

CHOP localization according to BRAF-selective inhibition.
According to our hypothesis, in cellular fractionation experiments dabrafenib down-or upregulated CHOP in the nuclear compartment in BRAF V600E (HT29) or BRAF-wt contexts (LS180), respectively ( Fig. 5a; see also Supplementary Fig. 10a for additional data in the SNU1235 and SNU1047 cell lines); conversely, trametinib treatment downregulated CHOP in the nucleus regardless of BRAF status ( Fig. 5a and Supplementary  Fig. 10a). Immunofluorescence experiments confirmed that dabrafenib caused CHOP redistribution to the perinuclear region in the BRAF V600E context, whereas it upregulated nuclear CHOP in BRAF-wt cell lines ( Fig. 5b and Supplementary Fig. 10b). MEK inhibition, on the other hand, caused CHOP nuclear exclusion regardless of BRAF status ( Fig. 5b and Supplementary Fig. 10b).
CHOP and RNA Pol II mediate IL-8 transcriptional activation. We next analyzed the potential role of CHOP-mediated IL-8 transcription in response to pharmacological MAPK inhibition. A constitutive physical interaction between CHOP and RNA Polymerase II (RNA Pol II) was observed regardless of the pharmacological treatment, as assessed by co-immunoprecipitation experiments conducted in the LS180 BRAF-wt cell line (Fig. 6a). CHOP and RNA Pol II recruitment to the IL-8 promoter in response to dabrafenib and trametinib treatment was further studied using Chromatin ImmunoPrecipitation (ChIP) assays. As shown in Fig. 6b, recruitment of CHOP to the CHOP-binding site of the IL-8 promoter did not change significantly upon drug treatment; however, selective BRAF inhibition by dabrafenib specifically increased the recruitment of RNA Pol II to the CHOP-binding site of the IL-8 promoter, without increasing its CHOP-independent binding to the IL-8 promoter TATA box (Fig. 6c, d). Moreover, a marked increase in the CHOP-binding site activity of the IL-8 promoter in response to dabrafenib, but not in response to trametinib, was observed after ChIP with an anti-acetylated histone H4 (Fig. 6e), suggesting increased IL-8 promoter accessibility to the CHOP-RNA Pol II complex in response to selective BRAF inhibition in a BRAF-wt context.

Discussion
Our data show that BRAF mutations and PTEN-loss promote (in a non-mutually exclusive fashion) high levels of constitutive IL-8  production in a panel of human CRC cell lines; however, BRAF mutations, but not PTEN status, specifically dictated the response of CRC cells to selective pathway inhibitors, particularly MAPK pathway inhibitors, in terms of IL-8 production. Indeed, to the best of our knowledge we describe here for the first time a drugsensitive, genetic-context-dependent BRAF/ERK2/CHOP molecular axis, tightly controlling IL-8 transcription in CRC models. IL-8 has recently emerged as a putative prognostic/predictive biomarker in CRC. In preclinical models, IL-8 promotes both tumor and endothelial cell proliferation, migration, angiogenesis and decreases sensitivity to the cytotoxic effects of oxaliplatin 4 . Clinically, IL-8 expression correlates with CRC progression and development of liver metastases and is associated with resistance to antiangiogenic therapy 9,14,15 . Regulation of IL-8 expression occurs at three different levels: repression and activation of the gene promoter, mRNA stabilization, and post-translational cleavage of its precursor 16 . It is now well established that the MAPK pathway is the main regulator of IL-8 expression: ERK and c-Jun N-terminal Kinase (JNK) promote AP-1-and NF-κB-mediated IL-8 transcription, whereas p38 stabilizes IL-8 mRNA 17,18 . Consistent with a dominant role of the MAPK pathway in IL-8 regulation, genetic or pharmacologic MEK inhibition invariably abrogated IL-8 production, regardless of the genetic background of the examined CRC cells 19 . Similarly, ERK2 silencing almost invariably downregulated IL-8 expression. As extensively revised by Buscà, ERK1 and ERK2 are generally described as homologous molecules and seem to be functionally redundant, though differential roles for ERK1 and ERK2 in terms of cell proliferation, colony formation, epithelial-mesenchymal transition and cell invasion have been described 20 . Here we show that only ERK2, but not ERK1, gene silencing results in IL-8 downregulation; similar results have been reported with the gp130 subunit of the promiscuous IL-6 receptor, whose expression is selectively controlled by ERK2, but not by ERK1 21 .
At a difference with MEK/ERK inhibition, pharmacologic BRAF inhibition affected IL-8 differentially, according to BRAFmutational status. Paradoxical downstream MAPK activation has been extensively described in BRAF-wt genetic contexts, thus MAPK-dependent downregulation and upregulation of IL-8 production in response to dabrafenib in BRAF-mut and BRAFwt contexts is theoretically expected [22][23][24][25] . Consistent with the recently reported ability of combined BRAF/MEK inhibition to offset paradoxical MAPK activation in BRAF-wt models, the combination of dabrafenib and trametinib effectively prevented dabrafenib-induced IL-8 upregulation in both BRAF-wt cell lines SNU1047 and LS180 22 . However, BRAF silencing, alone or in combination with dabrafenib treatment, had strikingly different effects in these two models. In the SNU1047 cell line, BRAF silencing inhibited both constitutive and dabrafenib-induced IL-8 production, consistent with a model in which a kinase-inhibited, but not an absent, BRAF protein can heterodimerize with CRAF and paradoxically activate downstream elements of the MAPK pathway 25 . Conversely, in the LS180 cell line, which harbors a KRAS G12D mutation, BRAF silencing paradoxically upregulated IL-8 production and combined BRAF silencing and dabrafenib treatment synergistically increased IL-8 levels. In CRC an intricate relationship exists between Epidermal Growth Factor Receptor (EGFR) and RAS family signaling 26 . Although we have not addressed this specific issue experimentally, it is possible that KRAS-wt cells (whether BRAF-mut or -wt) mostly rely on EGFR signaling to feed both basal and stimulated MAPK activation; in this context, BRAF protein expression and kinase activity may have a more prominent role in the activation of MAPK signaling, as opposed to KRAS-mut contexts, where CRAF, but not BRAF, is essential to allow the signaling flow to downstream elements of the cascade 22,27 . It is interesting to note that ERK2 silencing is not effective at inhibiting IL-8 production in the KRAS-mut LS180 cells and that dabrafenib-stimulated IL-8 production is further increased by ERK1 or ERK2 silencing in BRAF-wt contexts. Overall, our data suggest that individual MAPK elements, namely BRAF and ERK2, may play different roles in regulating IL-8 production in BRAF-wt CRC cells, depending on KRAS and, possibly, EGFR family activation status.
At a difference with previous findings, we observed that NF-κB is important but not sufficient to promote IL-8 transcription in CRC models; however, it has been demonstrated that other transcription factors (such as AP-1) physically interact with NF-κB and functionally cooperate to promote IL-8 gene expression and might be targeted by MAPK regulation 13,28 . Indeed, our data show an expected modulation of c-Jun Ser73 phosphorylation in response to dabrafenib and trametinib 29 . To the best of our knowledge, here we report for the first time a more prominent role for the CHOP, also known GADD153, transcription factor in MAPK-dependent regulation of IL-8 transcription in CRC models, depending on their BRAF-mutational status. In the last decade, several groups demonstrated that CHOP promotes IL-8 gene transcription independently of NF-κB in several cellular contexts, such as T lymphocytes and cystic fibrosis bronchial epithelial cells 30,31 . Here, we report that CHOP resides in the nuclear compartment of untreated CRC cell lines, while treatment with dabrafenib or trametinib modulates CHOP subcellular localization and consequently IL-8 production. In the new model ( Fig. 7) we propose that CHOP and AP-1 are the key regulators of MAPK-dependent IL-8 gene transcription in CRC: in BRAF-wt contexts dabrafenib causes paradoxical ERK activation, nuclear CHOP accumulation and binding to the IL-8 promoter, which, together with increased AP-1 activation, results in increased IL-8 transcription; trametinib, on the other hand, shuts down ERK, AP-1, and CHOP activation, thereby downregulating IL-8 transcription; in BRAF-mut cell lines CHOP is retained in the cytoplasm, with a perinuclear distribution, after either BRAF or MEK inhibition, thereby explaining the downregulation of IL-8 expression. Interestingly, Oh and coll. have described a similar situation in which CHOP enhances DR5 transcription after paradoxical MEK/ERK activation induced by BRAF inhibition in RAS-mut cell lines 32 . Furthermore, we demonstrated that dabrafenib affects not only CHOP compartmentalization, but also IL-8 gene chromatin accessibility. Indeed, in a BRAF-wt context, specific BRAF inhibition increases the binding of the RNA Pol II to the CHOP-binding region in IL-8 promoter, hence resulting in IL-8 transcription. Consistently, the CHOP-binding region displays an increased histone H4 acetylation, which results in an open-chromatin status. As consequence, dabrafenib induces IL-8 promoter activation in a CHOP-dependent manner. Compelling evidence shows that oncogenic signaling controls gene transcription, not only by affecting transcription factors' recruitment, but also by affecting chromatin structure through posttranslational histone modifications 33 . Among them, signaling through the classical MAPK module controls histone acetylation and phosphorylation in a kinase-dependent and -independent manner 34 . Interestingly, it has been demonstrated that ERK2 acts as a DNA-binding protein, hence interfering with the binding of transcriptional factors 35 .
It is nowadays accepted that soluble factor networks are involved in tumor-stroma interactions; cytokines and chemokines production (including IL-8) may be sustained not only by cancer cells but also by stromal elements (namely fibroblasts, endothelial, and immune cells), in a bidirectional crosstalk 2 . In a complex TME, stromal and/or infiltrating immune cells may substantially contribute to IL-8 expression; IL-8 regulation, in turn, can be modulated by targeted agents, such as selective BRAF inhibitors, differentially in a genetically normal stromal compartment, as compared to the tumor cell compartment, in which the net effect of pathway inhibition appears to be dictated by the specific genetic landscape of the tumor. This should be taken into account in interpreting CRC biology, especially in response to molecularly targeted drugs.
Overall our data depict a complex regulation of IL-8 production in CRC and identify a BRAF/ERK2/CHOP axis which dictates the overall effect of pharmacological MAPK manipulation in different genetic contexts. In addition to response to pharmacological modulation, BRAF status deserves further investigation as a potentially predictive biomarkers of IL-8 production in vivo, in both tumor and infiltrating mononuclear cells. This, in turn, may help identify the clinical settings in which IL-8 targeting (with IL-8 or CXCR1/2 antagonists currently in clinical development) might be most promising as a therapeutic strategy.

Methods
Cell lines. CRC cell lines were kindly provided from Federica Di Nicolantonio (University of Turin, Turin, Italy) 36 . X-MAN™ HCT116 Parental and HCT116 PTEN −/− were generated by Horizon from homozygous knock-out of PTEN by deleting exon 5 which encodes the active site of the protein in the CRC cell line HCT116 (Horizon Discover, www.horizondiscovery.com) 12 . Isogenic cell lines Fig. 7 Working model of CHOP-dependent IL-8 regulation, after dabrafenib and trametinib treatment, according to BRAF status. a Upon addition of the BRAF inhibitor dabrafenib, signaling output from BRAF V600E is blocked and there is a transient suppression of ERK activation and MAPK signaling. In this genetic context, CHOP is exported to the cytoplasmic compartment, thus resulting in the downregulation of IL-8. b In contrast, in BRAF-wt contexts, even in the presence of dabrafenib, BRAF forms a complex with CRAF and hyperactivates CRAF itself, thereby driving paradoxical hyper-activation of both MEK and ERK. Due to CHOP nuclear import and histone acetylation, chromatin is accessible to CHOP, thereby increasing its binding to the IL-8 promoter, resulting in IL-8 transcription and protein production. c Trametinib inhibits MEK activity, thereby downregulating p-ERK levels: CHOP translocates from the nucleus to the cytosol and IL-8 gene is not transcribed, regardless of BRAF genetic status.
Cell lines were routinely maintained in RPMI 1640 or DMEM medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine, and antibiotics (Pen/Strep) (all from Euroclone, Milan, Italy) in a humidified atmosphere with 5% CO 2 at 37°C.
All cell lines tested negative for mycoplasma contamination.
The final concentration of drugs was obtained by dilution with culture medium. Standardization and assessment of cell culture media. In order to overcome the possibility that a different number of cells alter the results of chemokines analysis, growth curves for each CRC cell line were assessed and cells were seeded at different cell concentrations to evaluate their rate growth. For cell counting, Thoma chamber was used. Standardization: all cell lines were plated into 60 × 15 dishes (BD Falcon, Oxford, UK) to have about 1×10 6 cells after 48 h of plating. After 24 h from the plating, the culture medium was replaced by serum-free medium, and after 24 h, media were collected and cells were counted.
Assessment: CRC cell culture media were analyzed in triplicate as per the manufacturer's instructions using human IL-8 (Enzo Life Sciences, Farmingdale, NY, USA) and VEGF (R&DSystems, MN, USA) specific ELISA. Absorbance was read at 450 nm. IL-8 and VEGF expression was represented as pg/mL and then related to the control. HCT116 Parental and HCT116 PTEN −/− culture media were analyzed by Human Angiogenesis Antibody Array (RayBiotech, Norcross, GA, USA), according to the manufacturer's protocol.
Full, uncropped blot/gel images are available in Supplementary Fig. 11.
The Immunofluorescence. SNU1235, SNU1047, HT29, and LS180 cells were seeded on 22x22 mm coverslips and medium was replaced by serum-free medium after 24 h from plating. After 24 h from the replaced medium, the grown cells were fixed in 4% formaldehyde in Phosphate-Buffered Saline (PBS) for 10 min, permeabilized in PBS containing 0.5% Triton X-100 for 5 min, and blocked with PBS containing 5% Bovine Serum Albumin (BSA) and 0.3% of Triton X-100 for 90 min. Coverslips were incubated overnight with primary antibodies (CHOP and phosphorylated Thr202/Tyr204 ERK1/2, Cell Signaling Technology Inc., Beverly, USA) in 1% BSA, 0.3% Triton X-100 in PBS. Coverslips were washed three times in PBS and then incubated with one of the following secondary antibodies for 1 h at a dilution of 1:400: goat anti-rabbit and goat anti-mouse (A11034 and A11032 respectively, Alexa Fluor ® , Invitrogen, Carlsbad, CA, USA). After washing, the nuclei were counterstained with DAPI. Cells were then washed twice with PBS and observed under Zeiss Axiovert 200 M fluorescence microscope at 60X magnification. Specific fields were photographed with a digital camera equipped with Zeiss Axiovision acquisition software.
IP. Chip-Grade Protein G Magnetic Beads (Thermo Fisher Scientific, Rochester, NY, USA) were incubate with 5 μg of CHOP (Cell Signaling Technology Inc., Beverly, USA) overnight at 4°C. Precleared beads were than incubated overnight at 4°C with 1 μg of protein. The immunoprecipitates were collected and after 2 wash in CHAPS buffer resuspended in 35 µl of the same buffer and Ladder buffer 1X. The immune complexes and 20 μg of protein total cell extract were analysed by WB.
Formaldehyde cross-linking and ChIP. Formaldehyde was added directly to cell culture media of 6 × 10 6 LS180 cells, at a final concentration of 1% at room temperature for 8 min and the cross-linking was stopped by the addition of glycine to a final concentration of 0.125 M. Cells were rinsed twice with cold PBS, incubated with 5 ml of cold PBS containing protease and phosphatase inhibitors (Thermo Fisher Scientific, Rochester, NY, USA), and then scraped. Cells were collected by centrifugation at 1.000 rpm for 10 min at 4°C. Pellets were resuspended in 5 vol of lysis buffer (5 mM piperazine N, N bis zethone sulfonic acid pH 8, 10 mM KCl, 85 mM KCl, 0.5% NP-40, protease and phosphatase inhibitors) and incubated on ice for 20 min. Nuclei were collected by centrifugation at 3,000 rpm, resuspended in sonication buffer (1% SDS, 10 mM EDTA, 50 mM Tris HCl pH 8) and incubated on ice for 10 min.
Input of chromatin was collected before the first wash from the supernatant of each IgG samples and was processed with the eluted immunoprecipitates beginning at the crosslink reversal step.
Samples were centrifuged at 14,000 rpm for 5 min and supernatants were transferred in clean tube. Crosslinks were reversed by addition of NaCl to a final concentration of 200 mM by incubation at 65°C for 4 h with shaking. 5 μl of Proteinase K Solution (Qiagen, Hilden, Germany) were added to samples and incubated for 1 h at 43°C.
DNA extraction. Genomic DNA was extracted using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The extracted DNA was quantified and its quality assessed using NanoDrop ® (Thermo Fisher Scientific, Rochester, NY, USA) and Qubit ® (Thermo Fisher Scientific, Rochester, NY, USA) platforms according to manufacturer' instructions. Multiplex PCR libraries were generated from 10 ng of DNA per sample using the Ampliseq technology (Ion Ampliseq Library Kit v2.0, Thermo Fisher Scientific, Rochester, NY, USA). Each library was barcoded with the Ion Xpress Barcode Adapters 1-16 kit and 17-32 kit (Thermo Fisher Scientific, Rochester, NY, USA). Library concentration was evaluated with Qubit 2.0 fluorometer using high sensitivity Qubit Assay Kit (Thermo Fisher Scientific, Rochester, NY, USA). Each diluted library (100 pM) was clonally amplified on to Ion Sphere Particles (ISP) using emulsion PCR (emPCR) in an Ion Chef System (Thermo Fisher Scientific, Rochester, NY, USA) according to the manufacturer's instructions. Enriched ISPs were loaded on to 530 chips accommodating thirty-two tumor samples on a single chip per sequencing run.
Sequencing was performed on an Ion S5 Sequencer using an Ion 530 Chip and an Ion 530 kit-Chef (all from Thermo Fisher Scientific, Rochester, NY, USA).
Data analysis and reporting. The raw data were analyzed using the Torrent Suite software (version 5.10.1) (Thermo Fisher Scientific, Rochester, NY, USA) through default analysis parameters. Variant Caller version 5.10.1.20 and Coverage Analysis version 5.10.0.3 plug-ins (Thermo Fisher Scientific, Rochester, NY, USA) were used for variant calling and sequencing coverage analysis, respectively. A minimum sequencing depth of 250× was considered as adequate sequencing depth, and an allelic frequency of 5% was used as a cut-off for variants.
Ion Reporter™ Server hosting informatic tools (Ion Reporter™ Software version 5.4) was used for variant analysis, filtering, and annotations. The Integrative Genomics Viewer was used to visualize the read alignment and the presence of variants against the reference genome as well as to confirm variant calls by checking for strand biases and sequencing errors.
Only mutations reported in the Sanger Institute Catalogue of Somatic Mutations in Cancer (COSMIC) database (http://www.sanger.ac.uk/cosmic) were considered.
Statistics and reproducibility. Results are expressed as average of three independent experiments or as representative experiment out of three independent experiments performed with similar results. Optimal cut-off and performance characteristics (sensitivity, specificity, AUC) were evaluated by computing ROC curves. AUC under 0.70 were not considered as relevant. The associations between variables were tested by two-sided Pearson Chi Square test or Fisher exact test, when appropriate. Mean comparison of more than two groups was made by ANOVA, when appropriate. The SPSS ® (21.0), R ® (2.6.1), MedCalc ® (13.0) statistical programs were used for all analyses. p-value of < 0.05 was considered statistically significant.

Data availability
All data supporting the findings of this study are available in the article along with Supplementary Information