The EMT factor ZEB1 paradoxically inhibits EMT in BRAF-mutant carcinomas

Despite being in the same pathway, mutations of KRAS and BRAF in colorectal carcinomas (CRCs) determine distinct progression courses. ZEB1 induces an epithelial-to-mesenchymal transition (EMT) and is associated with worse progression in most carcinomas. Using samples from patients with CRC, mouse models of KrasG12D and BrafV600E CRC, and a Zeb1-deficient mouse, we show that ZEB1 had opposite functions in KRAS- and BRAF-mutant CRCs. In KrasG12D CRCs, ZEB1 was correlated with a worse prognosis and a higher number of larger and undifferentiated (mesenchymal or EMT-like) tumors. Surprisingly, in BrafV600E CRC, ZEB1 was associated with better prognosis; fewer, smaller, and more differentiated (reduced EMT) primary tumors; and fewer metastases. ZEB1 was positively correlated in KRAS-mutant CRC cells and negatively in BRAF-mutant CRC cells with gene signatures for EMT, cell proliferation and survival, and ERK signaling. On a mechanistic level, ZEB1 knockdown in KRAS-mutant CRC cells increased apoptosis and reduced clonogenicity and anchorage-independent growth; the reverse occurred in BRAFV600E CRC cells. ZEB1 is associated with better prognosis and reduced EMT signature in patients harboring BRAF CRCs. These data suggest that ZEB1 can function as a tumor suppressor in BRAF-mutant CRCs, highlighting the importance of considering the KRAS/BRAF mutational background of CRCs in therapeutic strategies targeting ZEB1/EMT.

In addition, more than 90% of all sporadic CRCs exhibit aberrant activation of the Wnt pathway, chiefly through gain-of-function mutations of the APC gene, nuclear translocation of β-catenin and Despite being in the same pathway, mutations of KRAS and BRAF in colorectal carcinomas (CRCs) determine distinct progression courses.ZEB1 induces an epithelial-to-mesenchymal transition (EMT) and is associated with worse progression in most carcinomas.Using samples from patients with CRC, mouse models of Kras G12D and Braf V600E CRC, and a Zeb1-deficient mouse, we show that ZEB1 had opposite functions in KRAS-and BRAF-mutant CRCs.In Kras G12D CRCs, ZEB1 was correlated with a worse prognosis and a higher number of larger and undifferentiated (mesenchymal or EMT-like) tumors.Surprisingly, in Braf V600E CRC, ZEB1 was associated with better prognosis; fewer, smaller, and more differentiated (reduced EMT) primary tumors; and fewer metastases.ZEB1 was positively correlated in KRAS-mutant CRC cells and negatively in BRAFmutant CRC cells with gene signatures for EMT, cell proliferation and survival, and ERK signaling.On a mechanistic level, ZEB1 knockdown in KRAS-mutant CRC cells increased apoptosis and reduced clonogenicity and anchorage-independent growth; the reverse occurred in BRAF V600E CRC cells.ZEB1 is associated with better prognosis and reduced EMT signature in patients harboring BRAF CRCs.These data suggest that ZEB1 can function as a tumor suppressor in BRAF-mutant CRCs, highlighting the importance of considering the KRAS/BRAF mutational background of CRCs in therapeutic strategies targeting ZEB1/EMT.
RAS synergizes with Wnt signaling to promote progression in CRCs (25,26).ZEB1 is induced by and synergizes with the Wnt pathway in CRC activating or repressing target genes depending on cell status and/or promoter (27)(28)(29).ZEB1 is also downstream of RAS and BRAF in lung carcinomas and melanomas and mediates some of the signaling of oncogenic RAS in lung carcinomas (15,(30)(31)(32)(33). ZEB1 associates with poorer survival in most carcinomas, including CRCs (13,15,16,29,34,35); however, the role of ZEB1 in BRAF-mutant CRCs or a potential differential role of ZEB1 in CRCs based on the mutational status have not been explored.
Using human samples of primary CRC with BRAF mutations, CRC-established cell lines, transgenic mouse models for Kras G12D and Braf V600E intestinal tumors, and a Zeb1-deficient mouse, we found that ZEB1 is a tumor-promoting factor and induces an EMT phenotype in mouse Kras-mutant CRCs but, surprisingly, ZEB1 inhibits the EMT reprogramming of cancer cells and functions as a tumor suppressor in Braf-mutant CRCs.ZEB1 determines a better prognosis in patients harboring oncogenic BRAF metastatic CRC (mCRC).Our results show that ZEB1 functions as a tumor suppressor in BRAF-mutant CRCs, highlighting the need to assess the mutational background of CRC before using therapies that inhibit the expression and/or function of ZEB1.
ZEB1 paradoxically inhibits EMT and promotes histologically differentiated tumors in Braf V600E mice.We then conducted the pathological analyses of the colon and small intestine tumors generated in the different mouse models.The analysis revealed that whereas in the KVZ +/+ mice the tumors were mainly grade I-III adenocarcinomas, most tumors formed in the KVZ +/-mice corresponded to benign hyperplasia and tubular adenomas (Figure 2A).The lesions or tumors found in the small intestine of KVZ +/+ mice displayed higher malignancy grades than those in KVZ +/-mice.In contrast, the tumors in the small intestine of BVZ +/-mice corresponded to serrated adenomas and carcinomas with a greater loss of epithelial cell polarity than the lesions found in BVZ +/+ counterparts.Collectively, the downregulation of Zeb1 in Kras G12D mice resulted in more differentiated lesions and tumors, while the downregulation of Zeb1 in Braf V600E mice yielded less differentiated lesions and tumors.
High-grade tumor budding is a morphologic proxy of EMT and an independent prognostic factor associated with higher CRC recurrence, metastasis, and cancer-related death (44,45).We found that the tumors formed in the colon of KVZ +/+ mice had moderate-to high-grade tumor budding (Bd2 and Bd3), whereas those in KVZ +/-mice had lower or no tumor budding (Bd1) (Figure 2A).Conversely, lesions in BVZ +/+ mice had low tumor budding (Bd1) compared with the intermediate tumor budding (Bd2) found in BVZ +/-.
Intestinal tumor initiation and progression involve the deregulation of the homeostatic mechanisms controlling, inter alia, cell proliferation and/or apoptosis (46).In line with the reduced tumorigenesis in KVZ +/-mice, the hyperplastic mucosa in these mice expressed lower levels of the proliferation marker KI67 than in KVZ +/+ mice (Figure 2B).Conversely, compared with lesions in BVZ +/+ mice, those in BVZ +/-mice expressed higher levels of KI67.Relative to tumors in KVZ +/+ mice, tumors in KVZ +/- mice exhibited lower levels of nuclear β-catenin (a marker of aberrant Wnt signaling), Alcian blue (an acidic mucin marker of goblet cells), and lysozyme (a Paneth cell marker, which is upregulated in adenomas and carcinomas; ref. 47) (Figure 2B).Conversely, BVZ +/-tumors displayed higher expression of β-catenin and lysozyme than BVZ +/+ tumors.As in KVZ +/-mice, lesions in BVZ +/-had lower expression of Alcian blue.Altogether, these results suggest that ZEB1 promotes cell viability in Kras G12D tumors but has the opposite effect in Braf V600E ones.
To further investigate a possible differential regulation of signal transduction by ZEB1 in KRAS-and BRAF-mutant human CRC cells, we used the LS174T (KRAS G12D , WT BRAF) and RKO (WT KRAS, BRAF V600E ) CRC cell lines.RKO cells expressed higher levels of ZEB1 mRNA than did LS174T cells (Figure 4A) and the knockdown of KRAS in LS174T cells or of BRAF in RKO cells downregulated ZEB1 expression (Figure 4B).Conversely, the overexpression of BRAF V600E in LS174T cells or of KRAS G12D in RKO cells upregulated ZEB1 (Figure 4C).Inhibition of either MEK signaling with the inhibitor PD98059 or of PI3K signaling with LY294002 downregulated ZEB1 protein in both cell lines (Figure 4D and Supplemental Figure 1B), suggesting that KRAS and BRAF induce ZEB1 through the same upstream MAPK and PI3K signaling pathways.Although ZEB1 knockdown inhibited pAKT in both LS174T and RKO cells, it reduced pERK1/2 in the former but upregulated it in the latter (Figure 4E and Supplemental Figure 1C).Taken together, these data suggest that ZEB1 activates ERK signaling in mutant KRAS CRC cells but inhibits it when BRAF is mutated.
EMT factors cross-regulate each other and ZEB1 is downstream of other EMT factors (50,51).We examined the expression of other EMT factors and whether they were differentially modulated by ZEB1 in LS174T and RKO cells.The EMT factors SNAI1 and TWIST, but not ZEB2, were expressed in LS174T cells, but all of them were barely detectable (particularly ZEB2 and TWIST) in RKO cells (Figure 4F).The downregulation of ZEB1 did not alter SNAI1 and TWIST mRNA levels in LS174T and RKO cells (Figure 4F).
ZEB1 inhibits cell death and promotes clonogenicity and migration in KRAS G12D CRC cells but not in BRAF V600E CRC cells.ZEB1 knockdown reduced cell viability and cell cycle progression in LS174T cells but not in RKO cells (Figure 5, A and B, and Supplemental Figure 1D).Further analyses showed that the knockdown of ZEB1 increased apoptosis in LS174T cells but not in RKO cells (Figure 5C).ZEB1 mediates RAS/AKT-induced resistance to anoikis (anchorage-independent survival) that allows migratory cancer stem cells to shed from the primary tumor, invade the surrounding stroma, and eventually metastasize (52,53).We found that the downregulation of ZEB1: (a) reduced the anchorage-dependent 2D clonogenicity of LS174T cells, whereas it slightly increased it in RKO cells (Figure 5D); and (b) inhibited the 3D cell growth of LS174T cells but promoted it in RKO cells (Figure 5E).In sum, these results suggest that ZEB1 has opposing functions on anchorage-independent cancer cell growth, promoting it in KRAS G12D cells but inhibiting it in BRAF V600E ones.
ZEB1 triggers a more motile phenotype in cancer cells, thus increasing their migratory capacity (27)(28)(29)40).Accordingly, transient and stable knockdown of ZEB1 inhibited the migration of LS174T cells in both wound healing and Transwell assays; however, ZEB1 knockdown had no significant effect in RKO cells (Figure 5, F and G).
We also tested the role of ZEB1 in the in vivo tumorigenic capacity of KRAS-and BRAF-mutant CRC cells using a xenograft model.LS174T and RKO cells with basal and downregulated levels of ZEB1 were xenotransplanted in immunodeficient nude mice and tumor formation was evaluated over time.In line with other experiments in this study, the downregulation of ZEB1 in LS174T cells inhibited their tumorigenic capacity (tumor volume) (Figure 5H), whereas ZEB1 downregulation in RKO cells promoted it (Figure 5I).
ZEB1 determines different gene signatures in KRAS G12D and BRAF V600E CRC cells.The gene signature associated with ZEB1 in KRAS-and BRAF-mutant CRC cells was explored by RNA-Seq.LS174T and RKO CRC cells were transiently transfected with a control siRNA or a specific siRNA against ZEB1 (28) to generate LS174T CTL , RKO CTL , LS174T ZEB1KD (where KD refers to knockdown), and RKO ZEB1KD transgenic cell lines.RNA-Seq bioinformatics analysis revealed 304 differentially expressed genes (DEGs) between LS174T CTL and LS174T ZEB1KD , and 205 DEGs between RKO CTL and RKO ZEB1KD cells (Figure 6A and Supplemental Table 1).There were 44 DEGs between RKO ZEB1KD and RKO CTL cells relative to the DEGs in LS174T ZEB1KD versus LS174T CTL cells.These DEGs are involved in the transcriptional regulation of pluripotent stem cells, RTK signaling, cell-to-cell junction organization, and cell metabolism (Supplemental Table 1).Importantly, of these din (n = 9 and 8 in Kras G12D ; 10 and 9 in Braf V600E ), vimentin (n = 9 in Kras G12D ; 10 in Braf V600E ), and fibronectin (n = 9) (in red) counterstained with DAPI (blue) in the colon of KVZ +/-(blue) and BVZ +/-(green) in comparison with their WT ZEB1 counterparts (in black).Individual stainings are shown in Supplemental Figure 1A.Scale bar: 20 μm.An Unpaired t test was used to determine statistical significance.P values are reported in Supplemental  6B), suggesting that most of the genes regulated by ZEB1 in CRC cells are specific to either KRAS G12D or BRAF V600E oncogenes.As with the stable downregulation of ZEB1 (Figure 4F), the transient downregulation of ZEB1 in both cell lines did not alter the expression of other EMT factors (e.g., SNAI1, SNAI3) (Supplemental Table 2).
Gene set enrichment analysis (GSEA) of gene ontology annotations indicated that, compared with RKO ZEB1KD cells, LS174T ZEB1KD cells expressed lower levels of cell cycle checkpoints and higher levels of genes associated with apoptosis and activation of pERK and RAF-independent MAPK1/3 signaling (Figure 6C and Supplemental Table 3).ZEB1 knockdown also has opposing effects on ROBO signaling and  G12D and BRAF V600E or an empty control vector (Vect).Α-Tubulin (α-tub) was included as a loading control.(D) Lysates from KRAS-and BRAF-mutant CRC cells incubated with PD98059 (PD), LY294002 (LY), or with their solvent (Untr) were immunoblotted for ZEB1 along with GAPDH as control of equal loading.(E) Left: Expression of pAKT and total AKT in KRAS-mutant and BRAF-mutant CRC cells interfered with either siCtl (CLT) or with a siRNA against ZEB1 (ZEB1 KD).Right: As in the left panel but for pERK and total ERK.(F) As in A, but for relative expression levels of ZEB1, ZEB2, SNAI1, and TWIST mRNA expression in KRAS-mutant and BRAF-mutant CRC cells stably interfered with lentivirus against ZEB1 KD (red bar) and compared with cells interfered with by a control vector (CTL) (black bar).Expression of EMT factors was expressed relative to ZEB1, which was set at 100.At least 3 independent experiments were done, or ≥ 3 values were used for an unpaired t test of statistical significance.P values are reported in Supplemental Table 16.***P ≤ 0.001, **P ≤ 0.01.
ZEB1 inhibits the EMT signature in BRAF CRC cells and patients with CRC.Notably, the downregulation of ZEB1 increased the GSEA EMT signature in BRAF-mutant CRC cells (Figure 6F and Supplemental Table 3) where leading-edge genes of the EMT signature were upregulated (Supplemental Figure 1E).The opposing role of ZEB1 over EMT in KRAS-and BRAF-mutant CRC was validated through the mRNA assessment of epithelial-E-cadherin (CDH1), tight junction protein ZO-3 (TJP3), occludin (OCLN), claudin-1 (CLDN1)-and mesenchymal-vimentin (VIM) and fibronectin III domain-containing protein 4 (FNDC4)-genes in LS174T and RKO cells stably interfered with an shRNA against ZEB1 or an shRNA control (Figure 6G; ref. 54).The downregulation of ZEB1 in RKO upregulated mesenchymal markers like VIM and downregulated epithelial genes like CDH1 and CLDN1, but it also downregulated the mesenchymal marker FNDC4.LS174T cells where ZEB1 has been downregulated displayed a more epithelial phenotype, with increased expression of OCLN and CLDN1 and downregulation of VIM.Again, these data support that ZEB1 exerts opposing effects on the regulation of the EMT program depending on the KRAS or BRAF mutational background.
To determine whether the tumor suppressor signature and the inhibition of EMT associated with ZEB1 in Braf-mutant mouse models also occurred in patients with CRC, we retrospectively analyzed 41 BRAF-mutant mCRC according to their ZEB1 expression (Supplemental Table 4).Analyses of the gene signatures associated with human BRAF and KRAS-mutant CRC revealed that relative to KRAS-mutant CRC, ZEB1 expression in BRAF-mutant CRC was associated with increased expression of genes related to angiogenesis, and immune and decreased apoptosis signature (Figure 7A and Supplemental Tables 5 and 6).In line with our discussed results in mouse CRC models and human CRC cell lines, we found that compared with patients with KRAS-mutant CRC, patients with BRAF-mutant CRC had a reduced EMT signature (Figure 7A).
A high expression of ZEB1 determines better survival in patients with metastatic BRAF V600E CRCs.We then correlated the clinical characteristics and genotype distribution of 115 patients with BRAF-mutant mCRC with their expression of ZEB1.Patients whose tumors have lower ZEB1 expression had more liver metastases, high lactate dehydrogenase levels, and a poorer overall status as determined by the Eastern Cooperative Oncology Group (ECOG) performance status scale (55).Patients with BRAF mutation had higher ZEB1 expression compared with patients with RAS-mutant and double WT genotypes (Figure 7B and Supplemental Table 7).In line with our results in mice and human CRC cell lines, the analysis of ZEB2, SNAI1, SNAI2, and TWIST1 expression in these patients did not reveal any correlation with ZEB1 expression (Supplemental Figure 2A).Next, to assess whether, as found in mice (Figure 1), ZEB1 expression affects the survival of patients with BRAF-mutant CRC, 38 patients with metastatic CRC harboring BRAF V600E were segregated into 2 cohorts based on ZEB1 expression above or below the upper tertile (n = 11 with high ZEB1 expression [ZEB1-high] and 27 with low ZEB1 expression [ZEB1-low]).ZEB1-high in patients with BRAF V600E CRC associated with metastatic resection (47% vs. 5%) and a better ECOG performance status (ECOG PS >2) (0% vs. 19%) relative to patients with BRAF V600E CRC in the ZEB1-low cohort (Figure 7, C and D).In fact, patients in the ZEB1-low cohort had a more aggressive debut of the illness (P = 0.004) than those in the ZEB1-high cohort precluding any oncologic therapy in the ZEB1-high group (Figure 7C and Supplemental Tables 7 and 8).The response rate was 61% in the ZEB1-low cohort and 12% in the ZEB1-high group (P = 0.083) (Supplemental Table 8).
In patients with metastatic BRAF V600E CRC, ZEB1-high was associated with better overall survival on univariate analysis and multivariate analysis (Figure 7, D and E, Supplemental Table 9, and Table 1).Interestingly, median postprogression survival in ZEB1-low patients was only 1 month, compared with 13 months in ZEB1-high patients (P = 0.047).

Discussion
ZEB1 promotes tumor initiation and progression in both carcinomas and certain nonepithelial tumors (reviewed in refs.13, 16, and 18).Accordingly, ZEB1-high associates with poorer survival in patients with CRC (22,29) although the effect of ZEB1 based on the mutational status of CRCs had not been previously considered.Here, we found that both in human samples and mouse models of CRC, ZEB1 has a tumor-promoting role and determines poorer prognosis in mutant KRAS CRC but, surprisingly, it functions as a tumor suppressor and determines better prognosis in BRAF CRC (see schematic summary in Figure 8).In the Kras G12D CRC mouse model, ZEB1 induced more and larger intestinal lesions and tumors with a more dedifferentiated histological pattern.Conversely, in the Braf V600E CRC mouse model, ZEB1 determined not only fewer, smaller, and more differentiated primary CRC lesions and tumors but also fewer liver and lung metastases.
ZEB1 expression is upregulated by most developmental and oncogenic signaling pathways (e.g., Wnt, TGF-β, KRAS, Hippo, Notch); in turn, ZEB1 mediates some of the downstream protumoral functions of these pathways (27,28,32,54,(56)(57)(58)(59) (reviewed in refs.14, 15, and 18).Our results indicate that ZEB1 is downstream of RAS/BRAF signaling and regulates ERK and AKT phosphorylation.It is worth noting that regulation of ERK is cell type specific (60); for instance, KRAS G12D can induce the phosphorylation of ERK in Paneth cells but not in enterocytes; in turn, BRAF V600E , but not KRAS G12D , induces ERK phosphorylation in intestinal organoids.Conversely, ERK signaling also regulates ZEB1 expression (61,62).ZEB1 mediates some of the downstream effects of RAS in cancer cells like the maintenance of a stem-like phenotype, cell proliferation, and anoikis resistance (33,52,63).Our results here show that ZEB1 promotes cell viability, colony formation, anchorage-independent growth, and migration in KRAS G12D CRC cells but not in BRAF V600E CRC cells.
Although ZEB1 is best known for triggering an EMT, we found that ZEB1-high in BRAF/Braf-mutant CRCs paradoxically correlated with low tumor budding and a reduced EMT signature, the opposite than in KRAS-mutant CRCs.Our results also indicate that the reverse effects of ZEB1 in KRAS G12D CRC cells but not in BRAF V600E CRC cells are not related to a differential regulation of other transcription factors known to induce an EMT (e.g., ZEB2, SNAI1, TWIST1).
The BRAF V600E mutation confers poor prognosis in mCRC (64); consequently, tumors aligning with the consensus molecular subtype type 1 (CMS1), which is mainly enriched with both immune cells and BRAF signaling pathways enriched (yellow) or downregulated (blue) in RKO CTLvsZEB1KD versus LS174T CTLvsZEB1KD or in individual RKO CTLvsZEB1KD (BRAF-mut CRC) and LS174T CTLvsZEB1KD (KRAS-mut CRC).(D) Cnet plot of RKO CTLvsZEB1KD showing DEG associations.Genes are colored on the basis of the fold change associated.FDR < 0.05.(E) Fold-change expression of the indicated genes in KRAS-mutant (LS174T) and BRAF-mutant (RKO) CRC cells interfered with siZEB1 (ZEB1 KD) (red), siKRAS (KRAS KD) (blue), or siBRAF (BRAF KD) (green) in comparison with siCtl (0 baseline).(F) GSEA and box plot of an EMT signature in BRAF-mutant CRC cells interfered with siZEB1 (ZEB1 KD) in comparison with siCtl (CTL).(G) E-cadherin (CDH1), tight junction protein ZO-3 (TJP3), occludin (OCLN), claudin 1 (CLDN1), vimentin (VIM), and fibronectin III domain-containing protein 4 (FNDC4) relative mRNA expression in KRAS-mutant (LS174T) (blue) and BRAF-mutant (RKO) (green) CRC cells lentivirally interfered with siZEB1 (ZEB1 KD) in comparison with CTL quantified by qRT-PCR using GAPDH as the reference gene.The 0 value line represents the value of each gene in CTL.(E-G) Bars represent the mean of ≥3 independent experiments performed in triplicate with the SD.An unpaired t test was used to determine statistical significance.P values are included in Supplemental Table 16.***P ≤ 0.001, **P ≤ 0.01, or *P ≤ 0.05.(65,66).In a set of patients with metastatic BRAF-mutant CRC treated with targeted therapy (i.e., dabrafenib, trametinib, and panitumumab), those with the BRAFV600E-mutant (BM) 2 subtype signature (BM2) (characterized by a low EMT and high oxidative phosphorylation [OXPHOS] and G2M cell cycle signatures) had the poorest prognosis (67,68).Altogether, these clinical data suggest that, in patients with BRAF-mutant mCRC, the low EMT signature in the BM2 subtype could account for the poorer survival of the CMS1 subtype.A summary of published articles about BM subtype, best-observed response and survival of treated patients with BRAF-mutant mCRC is included in Supplemental Table 15.
The CMS4 subtype is characterized by higher ZEB1 expression in tumors with high levels of TGF-β and with a high stromal component (69).Alternatively, in non-TGF-β-driven tumors, EMT associates with WNT signaling (70).In fact, chemotherapy resistance in CRC preclinical models relied on EMT-WNT/MYC (71,72) and OXPHOS (73).Therefore, it can be hypothesized that in patients with BRAF mutation with ZEB1-low and whose cancer cells are exposed to a microenvironment with high competition for nutrients (e.g., CSM1; ref. 74), the increase in non-TGF-β/EMT and glycolysis/OXPHOS leads to a metabolic rewiring and poorer survival.
The expression and function of EMT factors are being targeted in several cancer therapy clinical trials, including in CRC (24).The present study highlights the need to assess the BRAF or KRAS mutational background of patients with CRC, and tentatively of other tumors, before attempting therapies targeting ZEB1.These results also stress the need to develop not only inhibitors of ZEB1 but potentially also activators of its expression and/or function.

Methods
Human samples.This study includes a retrospective cohort of 115 patients with mCRC enriched with BRAF mutations who were diagnosed at the Hospital Clínic of Barcelona (Barcelona, Spain).Eligibility criteria and basic clinical data of patients with mCRC are detailed in the Supplemental Methods.
Mouse models.The following mouse models were used in the study: C57BL/6J (denoted as Zeb1 +/+ ), Zeb1 +/-, Kras LSL-G12D , Braf LSL-V600E , and Vil1 Cre .The last 3 models were purchased from The Jackson Laboratory.See the Supplemental Methods for additional details on these mice and their crossing.Xenograft studies were performed using athymic nude mice purchased from Charles River Laboratories.The list of primers used for mouse genotyping is detailed in Supplemental Table 10.
Cell lines and cell culture.LS174T and RKO CRC cells were cultured as described in Supplemental Methods.Where indicated, cell lines were stably or transiently interfered for ZEB1, KRAS, or BRAF using shRNA harboring lentiviral vectors or siRNA oligonucleotides, respectively, as described in the Supplemental Methods.The sequences of shRNA lentiviral constructs and siRNA oligonucleotides are included in Supplemental Tables 11 and 12, respectively.LS174T and RKO cells stably interfered with either a noncoding shRNA control or an shRNA specific against ZEB1 are referred to here as LS174T CTL , RKO CTL , LS174T ZEB1KD , and RKO ZEB1KD , respectively.
Cell viability and clonogenic assay.Cell viability, proliferation, apoptosis, clonogenic, and migration assays were assessed as detailed in Supplemental Methods.
Determination of protein and RNA levels.Determination of protein expression by Western blot and/ or immunostaining is described in Supplemental Methods.The identities and sources of primary and  13.Relative mRNA levels were determined by qRT-PCR.The DNA primers used in the qRT-PCR are included in Supplemental Table 14.
Bulk RNA-Seq and NanoString gene expression profiling.Gene expression profiles were assessed by RNA-Seq.A NanoString panel was used to interrogate gene expression on FFPE tissue.All procedures are detailed in Supplemental Methods.
Data availability.The RNA-Seq data have been deposited in the NCBI Gene Expression Omnibus under reference GSE123416.
Statistics.Statistical analysis was performed using SPSS 18.0 (IBM), SPSS 17.0 (IBM), or GraphPad Prism 8.0.1 (GraphPad Software).The type of statistical test used and the corresponding P value is indicated in Supplemental Table 16.Unless specified otherwise, the means and SD of data and the statistical significance of their differences were assessed with a nonparametric, unpaired Mann-Whitney test and 2-tailed Student's t test.Statistical analyses involving multiple comparisons relative to a shared control were carried out with a 95% CI using Dunnett's test or Tukey's test (when also including comparison between noncontrol conditions).Bonferroni's test with a 95% CI was used for time-specific comparisons.Xenograft volume analysis was analyzed with a 2-way ANOVA test.Qualitative variables such as demographic and clinical variables were analyzed with a χ 2 test to compare the groups of patients with high and low expression of ZEB1.In Kaplan-Meier survival analyses, differences in mouse (Figure 1, A  and B) and patient (Figure 7, D and E) survival probabilities were determined by log-rank test and Mantel-Cox methods using SPSS (IBM) and SAS software.Logistic regression analysis was used to identify possible explanatory variables involved in survival.In the analysis of progression-free survival, data from patients who were alive without disease progression were censored as of the time of the last imaging assessment.Radiological progression and death that occurred without disease progression were included as events.Postprogression survival was calculated from the time of progressive disease to the date of death or last follow-up.For the analysis of overall survival, data for patients without documented death at the date of cutoff were censored.In turn, censored mice refers to those euthanized at the indicated periods to harvest and analyze their tissues.Where appropriate, relevant comparisons were labeled in figures as significant at the following values: ***P ≤ 0.001, **P ≤ 0.01**, or *P ≤ 0.05.P values were nonsignificant when P > 0.05.The P values reported in all figures also are given in Supplemental Table 16.
Study approval.The use of human samples in the study was approved by the Clinical Ethics Research Committee at the Hospital Clinic of Barcelona (references HCB-2013/8674, HCB-2018/0633, and HCB-2019/0255).All patients and donors gave their informed consent for the use of samples in accordance with the principles of the Helsinki Declaration.The use of mice in the study followed the guidelines of the Animal Experimental Committee at the University of Barcelona School of Medicine and was approved under references CEEA 347/14 and 193/16.

Figure 4 .
Figure 4. KRAS and BRAF induce ZEB1 through ERK-and AKT-dependent mechanisms.(A) ZEB1 and CDH1 mRNA in KRAS (LS174T) and BRAF (RKO)-mutant CRC cells were quantified by qRT-PCR using GAPDH as a reference gene.Bars represent the mean with the SD of at least 3 independent experiments.(B) ZEB1 mRNA expression in mutant (mut) KRAS and BRAF CRC cells interfered with a nontargeting siRNA (CTL) or with specific siRNAs against KRAS (KRAS KD), BRAF (BRAF KD), or ZEB1 (ZEB1 KD).Dunnett's comparison test was used.ZEB1, ERK, KRAS, and BRAF proteins were analyzed by Western blot.(C) ZEB1, KRAS, and BRAF proteins were analyzed by Western blot in cell lysates of KRAS-and BRAF-mutant CRC cells overexpressed with lentiviral vectors for KRAS G12D and BRAF V600E or an empty control vector (Vect).Α-Tubulin (α-tub) was included as a loading control.(D) Lysates from KRAS-and BRAF-mutant CRC cells incubated with PD98059 (PD), LY294002 (LY), or with their solvent (Untr) were immunoblotted for ZEB1 along with GAPDH as control of equal loading.(E) Left: Expression of pAKT and total AKT in KRAS-mutant and BRAF-mutant CRC cells interfered with either siCtl (CLT) or with a siRNA against ZEB1 (ZEB1 KD).Right: As in the left panel but for pERK and total ERK.(F) As in A, but for relative expression levels of ZEB1, ZEB2, SNAI1, and TWIST mRNA expression in KRAS-mutant and BRAF-mutant CRC cells stably interfered with lentivirus against ZEB1 KD (red bar) and compared with cells interfered with by a control vector (CTL) (black bar).Expression of EMT factors was expressed relative to ZEB1, which was set at 100.At least 3 independent experiments were done, or ≥ 3 values were used for an unpaired t test of statistical significance.P values are reported in Supplemental Table 16.***P ≤ 0.001, **P ≤ 0.01.

Figure 5 .
Figure 5. ZEB1 inhibits cell death and promotes clonogenicity, migration, and tumorigenesis in KRAS G12D but not in BRAF V600E CRC cells.(A) Cell viability of KRAS (LS174T) and mutant (mut) BRAF (RKO) CRC cells stably infected with lentivirus with an shRNA control (CTL) (black) or against ZEB1 (ZEB1 KD) (red).Cell viability by an MTT assay (n ≥ 5) is represented as the mean with the SD.(B) Opposite regulation of cell cycle progression by stable knockdown of ZEB1 in KRAS-and BRAF-mutant CRC cells.Share of cells in G2/M, S, GO/G1, or DNA fragmentation.(C) KRAS-and BRAF-mutant CRC cells transiently transfected with siCtl (CTL) or siZEB1 (ZEB1 KD) were assessed for apoptosis (n ≥ 4).(D) As in C, but for 2D clonogenicity.The plating efficiency of siCtl cells was set to 100 (n = 4).(E) 3D anchorage-independent growth of KRAS-and BRAF-mutant CRC cells as in C. Mean relative number of colonies with the SD (n ≥ 4).Right: Original magnification, ×100.(F) Cell migration in KRAS-and BRAF-mutant CRC cells, as in C, assessed by wound healing assays (n ≥ 3 in triplicate).(G) Cell migration in KRAS-and BRAF-mutant CRC cells, as in F, assessed by Transwell assays (left; n ≥ 3 in triplicate).(H) ZEB1 promotes tumorigenesis in KRAS-mutant CRC xenografts.Left: Tumor volume of 6 mice (3 males and 3 females) s.c.engrafted with cells stably infected with lentivirus encoding an shRNA control (CTL) (black) or against ZEB1 (ZEB1 KD) (red).Two-way ANOVA test was used for comparison.Right: Ex vivo tumor volume and images.(I) ZEB1 reduces tumorigenesis in BRAF mut CRC xenografts.As in H, but with BRAF-mutant CRC cells.Unless stated, an unpaired t test was used; Bonferroni's test was used in A and F. P values are reported in Supplemental Table16.***P ≤ 0.001, **P ≤ 0.01, or *P ≤ 0.05.

Table 1 . Multivariate analysis of overall survival in patients with BRAF V600E mCRC
secondary Abs are included in Supplemental Table