Introduction

Colorectal cancer (CRC) is a heterogeneous disease arising through several discrete evolutionary pathways. The best-known and most-studied pathway to CRC is the canonical pathway, in which cancer originates from conventional adenomatous polyps bearing APC (adenomatous polyposis coli) mutation 1,2. Recently a new “alternative” pathway through serrated adenoma—the serrated pathway—has been uncovered. Mice studies have established that the BRAFV600E mutation is a driver mutation in the serrated pathway 3-5. In patients, BRAFV600E mutation is found in 50-67% of serrated CRC 6 and 10-15% of all CRCs 7.

The “Goldilocks principle” applies to mutant APC-driven and mutant BRAF-driven intestinal tumorigenesis: a threshold of oncogenic signaling needs to be reached for dysplastic lesions to form, but optimum tumor development requires “just-right” levels of oncogenic signaling, with too much being as detrimental as too little. In the canonical pathway to CRC, the primary driving force is mutant APC-mediated activation of Wnt/β-catenin signaling 8, and the “just-right” level of Wnt/β-catenin signaling optimal for tumor formation is achieved largely by the selection for specific APC mutant proteins based on their residual β-catenin-downregulating activity 9-12. The selection for APC mutations in the intestine is influenced by the underlying basal/physiological level of Wnt activity and stem-cell number, and APC mutation spectra vary throughout the intestinal tract resulting in different APC mutation spectra in the proximal and distal CRCs 10,11. In addition to the different mutation spectra, the ‘optimal’ thresholds for proximal and distal cancers are also variable 11.

BRAFV600E drives tumorigenesis through constitutive downstream ERK1/2 activation 13, but hyperactivation of ERK induced by oncogenic BRAFV600E is not tolerated in the intestine: high ERK activation, induced by transgenic expression of oncogenic BRAF (BRAFV600K) or by activation of two BRAF alleles in BRAFV600E/V600E mutant mice, engages tumor suppressive mechanisms, causing loss of stem cells and induction of differentiation and senescence 14,15. Lowering ERK activation by treatment with ERK or MEK (mitogen-activated protein kinase kinase) inhibitor counteracted BRAFV600E-induced organoid disintegration 14,16. It is therefore presumed that maintaining ERK activation within a narrow threshold range to avoid engaging tumor suppression is pivotal for mutant BRAF to exhibit the strongest transforming activity. However, despite being highly anticipated 16, the existence of in vivo intrinsic fine-tuning of mutant BRAF-induced ERK activation has never been experimentally examined. Given that over 60 mutations have now been identified in BRAF 13,17, theoretically, mutation-selection could be a way to achieve optimal ERK activation. However, because the V600E mutation accounts for about 90% of BRAF mutation seen in human cancer 18, mutation-selection is not the primary means to achieve the “just-right” levels of oncogenic ERK signaling. Normally, ERK activation is self-limiting by the rapid inactivation of upstream kinases and delayed induction of dual-specific MAKP phosphatases (MKPs/DUSPs) 19. Although feedback inhibitors of ERK signaling including DUSPs are overexpressed in BRAFV600E-expressing cells, the ERK signaling pathway is refractory to upstream feedback inhibition 20. EGFR is a core receptor upstream of the MAPK kinase axis. In vitro cell culture studies show that all activating BRAF mutants are RAS-independent 21: neither RAS inhibition 21 nor EGFR inhibition 22,23 was able to inhibit mutant-BRAF-induced ERK phosphorylation in BRAF-mutant human CRC cell lines.

In this study, we addressed the question that whether BRAFV600E-induced ERK activation is still tuneable during tumorigenesis in vivo. If yes, what are the factors involved in the regulation? Can BRAFV600E-induced ERK activation be fine-tuned to a “just-right” level optimal for tumor initiation? Our study identified FAK as a key regulator of BRAFV600E-induced ERK activation in mutant BRAF-induced serrated tumor formation/initiation and revealed that FAK loss allows BRAFV600E-induced ERK signaling to reach the permissive threshold “just-right” for cecal tumors to form.

Results

FAK expression is reduced in BRAFV600E-mutant serrated lesions in humans and mice

FAK is a cytoplasmic non-receptor tyrosine kinase involved in many aspects and types of cancer 24. To determine the role of FAK in mutant BRAF-induced serrated CRC, we first evaluated FAK protein expressions in human BRAFV600E-mutated serrated tumors (11 cases). We examined tissue sections containing BRAFV600E-mutant CRCs, sessile serrated adenoma/polyps (SSA/P)s, and adjacent histologically normal colon from the same tissue block. Results of immunohistochemistry (IHC) staining showed that FAK protein levels were lower in SSA/Ps (5/5) than in normal intestines and CRCs (5/5) (Fig. 1a). FAK expression was more complex in CRCs. FAK levels in CRCs were either similar to (6/11) or lower (4/11) or higher (1/11) than that of the normal intestines (Fig. 1a, b). FAK was mainly localized in the cytoplasm (Fig.1b). In mice, compared to the neighboring normal mucosa or stroma in the tumor, Fak protein levels were substantially decreased in carcinomas in the colon (Fig. 1c) and adenomas/polyps in the small intestine (SI) (Fig. 1d) in Vill-Cre;BRAFV600E/+ (BC) mice. The downregulation of FAK in human and mouse polyps suggests that FAK loss may play a role in BRAFV600E-induced tumor formation/initiation.

FAK downregulation in serrated tumors.

a Summary of FAK IHC staining in 11 human BRAFV600E-mutant CRC samples. N represents normal colon; P represents polyp; C represents carcinoma; NA, not applicable; ↔ represents no change; ↑ represents an increase. ↓ represents a decrease. b Representative IHC staining of BRAFV600E-mutant patient SSA/P, serrated colorectal adenoma, and adjacent normal tissues. c IHC staining of Fak in small intestine tumors in a 12-month-old BC mouse. d Representative IHC staining of Fak in colon tumor in 12-month-old BC mice. Scale bars in b: 1 mm (upper panel) and 100 μm (lower panel). Scale bars in c and d: 100 μm.

Fak deletion promotes BRAFV600E-induced cecal tumor formation

Previous mice studies show that Fak deletion suppresses mammary tumorigenesis 25,26, mutant Apc-induced intestinal tumorigenesis 27, skin tumor formation 28, and hepatocarcinogenesis 29. To address the functional significance of FAK downregulation in BRAFV600E-induced serrated tumor formation/initiation, we generated the Vill-Cre;BRAFV600E/+;Fakfl/fl (FBC) mice. The Cre-mediated recombination efficiency was confirmed by tdTomato-reporter expression in intestinal crypts in Vill-Cre;Rosa26LSL-tdTomato/+ mice (Supplementary Fig. 1a). Deletion of Fak in the intestinal epithelium was further confirmed by IHC staining of the intestine in FBC mice (Supplementary Fig. 1b).

Similar to that seen in BC mice, compared to the BRAFV600E/+ (B) mice, the FBC mice exhibited hyperplasia throughout the intestine (Fig. 2a) and thickened small and large intestines (Supplementary Fig. 1c). In BC mice, intestinal tumors were primarily developed in the small intestine at 9 months or older (Fig. 2b). Fak loss had minimal impact on tumor incidence in the small intestine and the colon, however, it greatly enhanced BRAFV600E-induced cecal tumor formation: cecal tumor incidence increased from 0% (0/15) in 9-month or older BC mice to 100% (16/16) in FBC mice (Fig. 2c). Cecal adenoma/polyp started to develop in 3-month FBC mice and after 6 months, all mice (4/4) developed cecal tumors and 25% of the tumors (1/4) were carcinomas (Fig. 2c, d). At 9 months or older, 100% of the mice developed cecal tumors with a high incidence (13/16) of carcinoma (Fig. 2c, d, and Supplementary Fig. 1d). IHC staining confirmed that while the stroma showed strong Fak staining, tumor cells were Fak negative (Fig. 2e), hence validating that tumors were originated from Fak-deleted epithelial cells. Of note, no tumor metastasis was found in FBC mice. FBC mice were aged up to 434 days, and the life span of FBC mice was similar to that of BC mice.

Fak loss enhances BRAFV600E-driven cecal tumorigenesis in mice.

a Representative hematoxylin and eosin (H&E) staining of the small intestine, cecum, and colon from indicated 6-week-old mice. b Summary of tumor incidence at small intestine and colon in indicted mice at the indicated age. c Summary of tumor incidence and tumor stage at cecum in indicated mice at the indicated age. d H&E staining of the cecum in BC mice and cecal serrated adenoma/polyp and carcinoma in FBC mice at the indicated age. e Representative IHC staining of Fak in cecal tumors in FBC mice. Scale bars: 100 μm.

Together, these results revealed that Fak deletion promotes, rather than inhibits, BRAFV600E-induced cecal tumor formation. BRAF-mutant CRCs are primarily located in the right colon including the cecum 30. The same primary tumor location suggests that the FBC model truthfully recapitulates human BRAF-mutant serrated CRCs at least by location.

The molecular feature of the cecal tumors in FBC mice closely resembles human SSA/Ps

To characterize the molecular signatures of the cecal tumor in FBC mice, we performed whole-exome sequencing on paired tumors (n=2) and neighboring mucosa. No additional driver mutations were detected in the cecal tumors (Supplementary Table 1), implying that cecal tumor formation in FBC mice does not require additional driver mutations. To evaluate the relevance of FBC cecal tumors to humans, we performed RNA-sequencing (RNA-seq) and Gene Set Enrichment Analysis (GSEA) to determine whether FBC cecal tumors exhibited similar gene expression signatures as human SSA/Ps 31. The results showed that upregulated genes in human SSA/Ps were significantly enriched in cecal tumors in FBC mice (Fig. 3a). Downregulated genes in human SSA/P were also reduced in FBC tumors (Fig. 3b). Together, these results suggest that the FBC cecal tumors greatly resemble human serrated lesions at the molecular level.

Molecular characterization of cecal tumors in FBC mice.

a GSEA plot showing enrichment of human SSA/Ps signature genes (upregulated genes in SSA/Ps) in FBC cecal tumors vs normal cecal mucosa of B mice. b GSEA plot showing that downregulated genes in human SSA/Ps were also reduced in FBC cecal tumors. c Microsatellite instability status of FBC mice cecal mucosa and cecal carcinomas. Each column represents one sample.

About 50% of BRAF-mutated CRCs exhibit defective DNA mismatch repair 18. The results of microsatellite instability (MSI) analysis indicated that most FBC cecal tumors were microsatellite stable (MSS) (Fig. 3c). It has been shown that mismatch repair deficiency accelerates BRAF-driven serrated tumorigenesis 32. Maximizing the oncogenic activity of BRAFV600E without mismatch repair gene mutation and additional driver mutations suggests that in FBC mice, Fak loss created a “just-right” environment optimal for MSS serrated cecal tumor to form.

Fak loss increases intestinal stemness by upregulating Lgr4 levels in FBC mice

We explored the molecular mechanism underlying Fak loss-enhanced cecal tumor formation. Consistent with a prior report 27, we did not detect any abnormalities in the intestine in Vill-Cre; Fakfl/fl mice, implying that FAK loss by itself is not a driving force for intestinal tumorigenesis. A prior study showed that upon TGFβ (transforming growth factor β) receptor inactivation, BRAFV600E-induced right-sided tumorigenesis is supported by microbial-driven inflammation 33. To test the role of inflammation in FBC tumor formation, we compared sub-cryptal proprial neutrophil infiltration using myeloperoxidase (MPO) as a neutrophil marker for IHC staining. The results showed that consistent with prior findings 33, the number of MPO-positive cells was significantly higher in BC mice than in B mice, however, Fak loss did not further increase neutrophil infiltration in FBC mice (Supplementary Fig. 2a). Consistent with this, GSEA results showed that there was no difference in the expression of inflammatory response genes 34 in FBC mice and BC mice (Supplementary Fig. 2b). Together, these findings imply that Fak loss promotes tumor formation not by enhancing intestinal inflammation.

We next evaluated the roles of cellular senescence, apoptosis, cell proliferation, and Lgr5 expression in cecal tumorigenesis in FBC mice. The results indicated that BRAFV600E was insufficient to trigger senescence evaluated by SA-β-galactosidase staining or apoptosis evaluated by the TUNEL staining in BC mice (Supplementary Fig. 2c, d). Bromodeoxyuridine (BrdU) incorporation assays confirmed mutant BRAF-induced hyperproliferation. However, Fak loss did not further enhance the BrdU incorporation rate (Supplementary Fig. 2e). These results indicated that Fak deletion promotes tumor formation not through modulating cellular senescence, apoptosis, and cell proliferation.

Given that BRAFV600E drives tumorigenesis through constitutive downstream ERK1/2 activation 13, we examined the impact of Fak loss on ERK pathway transcriptional output. GSEA analysis showed that ERK pathway output was significantly increased in BC mice (Fig. 4a), which was consistent with the earlier report 20, but Fak loss did not further enhance it (Fig. 4f). Wnt pathway activation 32 and activation of transcription co-factor YAP have been implied in BRAFV600E-induced serrated tumorigenesis 33. In this study, our GSEA results also showed that the expression of intestinal Wnt signature genes 35 and YAP target genes 36 were significantly higher in BC mice than in B mice (Fig. 2b, c), again, Fak loss did not further enhance the activations (Fig. 2g, h). Together, these findings excluded the possibility that Fak loss promotes cecal tumor formation through enhancing ERK pathway output and activation of the Wnt pathway and the YAP pathway.

BRAFV600E mutation and Fak loss-mediated changes in signaling pathways.

GSEA analysis showing upregulation of MAPK signature (a), intestinal WNT signaling (b), YAP/TAZ target gene signature (c) and intestinal differentiation signature (d), and downregulation of intestinal stem cell signature (e) in the cecum of BC mice vs B mice (n=4 per group). GSEA plots revealed no significant change in MAPK signature (f), intestinal WNT signaling (g), YAP/TAZ target gene signature (h), and intestinal differentiation signature (i) in the cecum of FBC mice vs BC mice, but enrichment of stem cell signature in FBC mice (j) (n=4 per group). k GSEA analysis showing upregulation of upregulated fetal spheroid markers in the cecum of BC mice vs B mice, and further enrichment in the cecum of FBC mice vs BC mice (n=4 per group). l Immunoblotting analysis of LGR4 in the cecum from indicated 6-week-old mice. m Representative in situ hybridization (ISH) staining of tumor sections from ApcMin/+, BC, and FBC mice using Lgr4 and Lgr5 probes. Scale bars: 100 μm.

In mice, BRAFV600E poorly initiates colon cancer due to oncogenic BRAF-induced tissue differentiation and loss of intestinal stem cells 15. Consistent with this, GSEA results showed increased expressions of intestinal differentiation signature genes 37 (Fig. 4d) and decreased expressions of intestinal stem cell signature genes 38 (Fig. 4e) in BC mice. Fak deletion did not reverse BRAFV600E-induced tissue differentiation (Fig. 4i) but significantly enhanced intestinal stemness (Fig. 4j). These results revealed that Fak deletion promotes BRAFV600E-induced cecal tumor formation through increasing intestinal stemness.

The adult stem cell marker Lgr5 and its relative Lgr4 are R-spondin receptors mediating R-spondin signaling and are critical for intestinal stemness 39,40. Mutant BRAF reduces Lgr5 expression in the intestinal crypt 15,33. Our results confirmed the downregulation of Lgr5 in the cecum crypt in BC mice, and we found that Fak loss did not restore Lgr5 expression in FBC mice (Supplementary Fig. 2f). These results thus excluded the possibility that Lgr5 mediates Fak loss-induced intestinal stemness.

Prior studies show that the fetal type of intestinal stem cells has a strikingly different transcriptome than that of adult intestinal stem cells and the receptor LGR4, but not LGR5, is essential for the cells 41. In VillinCreER;BrafLSL-V600E/+;Alk5fl/fl mice, the proximal colonic tumors exhibit fetal intestinal signature 33. Consistent with the notion that mutant BRAF-driven right-sided colonic tumors are fetal progenitor phenotypes, GSEA results confirmed enrichment of the fetal-type transcriptomic signatures 41 in cecal mucosa in BC mice, and the fetal signature was further enriched in FBC mice (Fig. 4k). Accordingly, the immunoblotting analysis showed that the protein level of Lgr4 was increased in the intestine epithelium in FBC mice (Fig. 4i). Consistent with the fact that intestinal Lgr5 expression was low in FBC mice (Supplementary Fig.2f), FBC tumors mainly expressed Lgr4 but not Lgr5, whereas BC tumors and Apcmin/+ tumors expressed both Lgr5 and Lgr4 (Fig. 4m). These results suggest that in FBC mice upregulated Lgr4 mediated the intestinal stemness increase.

FAK loss downregulates EGFR-dependent ERK phosphorylation to increase Lgr4 mRNA expression and stability

We addressed how Fak loss mediates Lgr4 increase. A prior study suggested that Wnt signaling maintains quiescent intestinal stem cell pools through suppression of the MAPK pathway in the intestine 42. Given the fact that Fak loss did not jeopardize ERK pathway transcriptional output (Fig. 4f), Fak loss may increase intestinal stemness by inhibiting ERK phosphorylation. To test, we first compared the levels of phosphorylated ERK across the intestines in B mice, BC mice, and FBC mice. As anticipated, BRAFV600E increased p-ERK levels throughout the intestine (Fig. 5a). FAK is known to be positively involved in ERK1/2 activation 24. Consistent with this, in FBC mice, FAK deletion suppressed mutant BRAF-induced elevation of p-ERK (Fig. 5a). The decoupling of ERK pathway output (no change) and the level of p-ERK (reduced) upon Fak loss is in line with a prior report suggesting that the level of ERK phosphorylation does not truthfully reflect ERK pathway activation 20.

Fak loss inhibits ERK phosphorylation and upregulates Lgr4.

a and b Immunoblotting analysis of intestinal mucosa lysates from indicated bowel subsites in indicated 6-week-old mice. c Immunoblotting analysis of cecum lysates from 6-week-old BC mice treated with vehicle or EGFR inhibitor erlotinib for 4 hours. Each lane represented a single mouse. d Immunoblotting analysis of cecum lysates from 6-week-old BC mice treated with vehicle or FAK inhibitor PF-562271 for 4 hours. Each lane represented a single mouse. e Immunoblotting analysis of lysates from freshly isolated cecal crypts and cecal organoids treated with DMSO, MEK inhibitor PD0325901, or erlotinib, respectively as described in Method. f qRT-PCR of Lgr4 using lysates from HT-29 cells treated with the vehicle and MEKi for 4 hours. Data presented as mean ± SD (***P<0.001; Student’s t-test, two-tailed). g qRT-PCR of Lgr4 using cecum lysates from BC mice treated with vehicle or MEKi for 6 hours. Data presented as mean ± SD (**P<0.01; Student’s t-test, two-tailed). Abrogation of ERK phosphorylation at T202/Y204 in the cecum was confirmed by western blot. h qRT-PCR of Lgr4 in cecum from BC and FBC mice (n=3 per group). Data presented as mean ± SD (P value calculated using two-tailed Student’s t-test). i Immunoblotting analysis of the lysates from HT-29 cells treated with cycloheximide (100 μg/ml) and/or MEK inhibitor PD0325901 (10 μM) as indicated.

We next examined how Fak loss altered BRAFV600E-induced phosphorylation of ERK. A prior study found that FAK promotes EGFR signaling 43, raising the possibility that FAK regulates ERK phosphorylation through EGFR. We then evaluated Egfr activation (represented by phosphorylated EGFR at tyrosine 1068) in the mice. The results showed that the level of phosphorylated EgfrY1068 was increased in BC mice throughout the intestine (Fig. 5b). In FBC mice, Fak deletion moderately reduced BRAFV600E-induced Egfr activation (Fig. 5b) and suppressed Egfr downstream signal transduction as evidenced by the decreased levels of phosphorylated c-RafS338 and MEK1/2S217/221 in FBC mice (Supplementary Fig. 3a). To validate that EGFR indeed regulates BRAFV600E-induced ERK phosphorylation, we treated BC mice with the EGFR inhibitor erlotinib. Erlotinib treatment, without significantly reducing ERK pathway output (Supplementary Fig. 3b), indeed suppressed phosphorylation of C-RAF, MEK, and ERK (Fig. 5c). Of note, Fak deletion had no impact on the level of p-EGFR and p-ERK in control mice (Supplementary Fig. 3c). Inhibition of Fak kinase activity by FAK inhibitor PF-562271 did not affect the phosphorylation of Egfr and ERK (Fig. 5d), implying that the kinase activity of Fak is not involved in the FAK/EGFR/ERK regulation in BRAFV600E-induced serrated tumorigenesis.

FAK complexes with activated EGFR to promote EGFR signaling 43. We assessed whether FAK interacts with EGFR in BRAFV600E-mutant cells. The results of co-immunoprecipitation using lysates from cecal mucosa confirmed the Fak-Egfr interaction and revealed that the Fak-Egfr interaction was increased in BC mice and inhibition of Egfr appeared not to affect the Fak-Egfr binding (Supplementary Fig. 3d). ERK phosphorylation is refractory to EGFR inhibition in human BRAFV600E-mutant CRC cell lines 22,23, however, the FAK-EGFR interaction was still detected in HT29 CRC cells and the interaction was not affected by either EGFR inhibition or FAK inhibition (Supplementary Fig. 3e). These results indicated that FAK/EGFR interaction alone is not sufficient for FAK getting involved in the regulation of MAPK signaling.

The contradictory results seen in BC mice and human BRAFV600E-mutant CRC cell lines could result from the differences between in vitro culture systems and in vivo. To test, we examined whether inhibition of Egfr leads to ERK inhibition in freshly isolated cecal crypts from BC mice and BC cecal organoids. The results showed that inhibition of Egfr did not reduce ERK phosphorylation, confirming that the contradictory findings resulted from in vitro and in vivo. We speculate that the lack of certain stromal factors in vitro is responsible for the EGFR’s inability to transmit its signal to activate ERK.

Finally, we examined whether and how a reduction in ERK phosphorylation increases Lgr4 expression/stemness. Our results showed that treatment with MEK inhibitor increased the mRNA expression of LGR4 in human BRAFV600E-mutant CRC HT29 cells (Fig 5f) and BC mice (Fig. 5g), uncovering a negative association between the level of ERK phosphorylation and mRNA expression of Lgr4. Of note, inhibition of ERK activation in BC mice was confirmed by the abrogation of ERK phosphorylation (Fig 5g) and suppression of ERK pathway transcriptional output (Supplementary Fig. 4). This negative association was further supported by our observation that the mRNA levels of Lgr4 were higher, albeit not statistically significant, in FBC mice than in BC mice (Fig 5h). Regulation of Lgr4 protein stability represents an important mechanism of modulating Lgr4 function 44. Our cycloheximide chase analysis results showed that inhibition of ERK phosphorylation by MEK inhibitor treatment dramatically enhanced Lgr4 protein stability in BRAFV600E-mutant CRC cell line HT29 cells (Fig 5i). This finding revealed the inverse correlation between the level of ERK phosphorylation and the protein stability of Lgr4. Together, these results suggest that Fak loss lowers BRAFV600E-induced ERK phosphorylation to increase Lgr4 mRNA expression and protein stability, thereby enhancing intestinal stemness and cecal tumor formation.

FAK’s influence on oncogenic MAPK-driven intestinal tumorigenesis depends on FAK’s impact on ERK phosphorylation

Fak loss reduced ERK phosphorylation in FBC mice (Fig. 5a) but not in control mice with wild-type BRAF (Supplementary Fig. 3c). To determine whether FAK is involved in other oncogenic MAPK-driven tumors, we generated Vill-Cre;KrasLSL-G12D/+ (KC) mice and Vill-Cre;KrasLSL-G12D/+;Fakfl/fl (FKC) mice. In KC mice, the endogenous expression of oncogenic Kras induces serrated hyperplasia, however, high ERK activation-induced senescence prevents hyperplasia progression into dysplasia 45. As shown in Fig. 6a, no tumor was found in KC mice (n=6, 9-months-old) and FKC mice (3-month-old, n=3; 6-month-old, n=3; 9-month-old, n=4). Immunoblotting results confirmed that Fak loss failed to influence the phosphorylation of Egfr or ERK (Fig. 6b). The co-immunoprecipitation results showed that Fak complexed with Egfr in KC mice similarly as in BC mice (Fig.6c), implying that the noninvolvement of Fak was not due to the lack of Fak/Egfr interaction. A recent preprint (https://doi.org/10.1101/2020.07.02.185173) suggests that “EGFR network oncogenesis cooperates with weak oncogenes in the MAPK pathway”, which inspired us to propose the notion that EGFR participates in the regulation of ERK phosphorylation only when the p-ERK level is relatively low. In KC mice, KRASG12D induces extremely high levels of ERK phosphorylation, high enough to cause intestinal senescence 45. Given the level of increased p-ERK in KC mice, one would expect that ERK phosphorylation is EGFR-independent. The EGFR independence was confirmed by our results showing that pharmacologic abrogation of EGFR activation had no impact on KRASG12D-induced ERK phosphorylation in KC mice (Fig 6d). Our notion was further supported by clinical findings. Anti-EGFR therapy is excluded for patients with KRAS-mutant CRC, supporting that EGFR has minimum impact on downstream MAPK signaling upon KRAS mutation. However, when ERK activation is inhibited by KRASG12C inhibitors, EGFR signaling acts as the dominant mechanism of colorectal cancer resistance to KRASG12C inhibitors 46.

ERK activation is FAK/EGFR-independent in KC mice.

Representative hematoxylin and eosin (H&E) staining of the small intestine, cecum, and colon from indicated 9-month-old mice. b Immunoblotting analysis of intestinal mucosa lysates from indicated bowel subsites in indicated 6-week-old mice. c The cecal mucosa lysates from 6-week-old KC and BC mice were used for immunoprecipitation and immunoblotting with the indicated antibodies. d Immunoblotting analysis of cecum lysates from 6-week-old KC mice treated with vehicle or EGFR inhibitor erlotinib for 4 hours. e and f Comparison of FAK expression levels between CRCs with indicated mutations by analysis of TCGA RNA-sequencing dataset. Data were analyzed for statistical significance using a Student t-test. g Diagram of the “just-right” MAPK signaling model in the serrated pathway.

To address whether FAK downregulation is specific to human BRAF-mutant CRCs, we compared FAK expression levels in CRCs with different driver mutations using the TCGA database. TCGA analysis revealed that FAK mRNA levels were significantly lower in BRAF-mutated CRCs than in APC-mutated CRCs or KRAS-mutant CRCs (Fig. 6e). This result is consistent with the result seen in mice, again, it suggests that FAK is not involved in the regulation of KRAS-mutant CRCs.

In mice, mutant BRAF-induced ERK activation is cancer stage-dependent with significantly higher levels of phosphorylated ERK in high-grade dysplasia and carcinoma 3, suggesting that different tumor stages may require different levels of p-ERK. If FAK is a key regulator of ERK phosphorylation in mutant BRAF-induced serrated tumorigenesis in patients, one would expect the level of FAK may increase as the tumors progress. Consistent with this notion, we observed that FAK levels were higher in BRAF-mutant CRCs than in BRAF-mutant polyps (Fig 1a), TCGA analysis (Fig. 6e) further confirmed that FAK expression was restored to a level similar to normal intestines, albeit still significantly lower than in APC mutant or KRAS mutant CRCs (Fig. 6e).

In patients, BRAF mutations are divided into two groups: Activator and amplifier mutation 47. In CRC, the majority (80%-90%) of activating mutations in BRAF are V600E 18. Among these mutants, based on their kinase activities, BRAFV600E belongs to the high-activity mutants, and the rest of the mutants except G595R (with impaired BRAF kinase activity in vitro but still induce constitutive ERK activation in vivo,) are intermediate activity mutants 48. If mutant BRAF-induced ERK phosphorylation needs to reach a “just-right” level via FAK downregulation in patients, one would expect that the degree of FAK downregulation is BRAF mutant activity-dependent, and there could be a correlation between the activity of BRAF mutants and the degree of FAK reduction. Consistent with this speculation, TCGA data analysis confirmed that CRCs with BRAFV600E mutation had lower FAK expression than CRCs with non-V600E mutations and BRAF wild-type CRCs (Fig. 6f). Although the differences between V600E and non-V600E groups were not statistically significant due to limited sample numbers, they might be biologically relevant.

Discussion

The current study finds that in BRAFV600E-mutant intestinal epithelium, elevating the p-ERK level to a minimum threshold is sufficient to maximize the pathway transcriptional output, i.e., only lowering the p-ERK level below the threshold will significantly abrogate the ERK pathway transcriptional output. Due to the negative association between ERK phosphorylation and intestinal stemness, any increase in ERK phosphorylation will decrease intestinal stemness (Fig 6g). In BRAFV600E-mutant intestinal epithelium, ERK phosphorylation is EGFR/RAS/c-RAF-dependent. The involvement of EGFR provides an opportunity for non-MAPK pathway factors such as FAK to participate in the regulation of ERK phosphorylation to influence the biological outcomes of BRAF mutation. This study has established the first “just-right” MAPK signaling model of BRAFV600E-induced tumor formation (Fig. 6g). Our results show that by lowering BRAFV600E-induced ERK phosphorylation, Fak loss, without jeopardizing the ERK pathway transcriptional output, enhances mRNA expression and protein stability of Lgr4, thereby increasing intestinal stemness and promoting cecal tumor formation in mice.

High-level activation of oncogenes (e.g. KRAS, BRAF, and c-MYC) triggers intrinsic tumor suppression 45,49-52. Genetic abrogation of tumor suppressors such as p53 or p16 revokes the tumor-suppressive barrier thereby facilitating oncogene-induced tumorigenesis 4,45,50,51. Cooperation with other oncogenic stimulation, such as co-expression of c-MYC and KRAS, ultraviolet radiation on melanocytes expressing BRAFV600E, can also break the suppressive barrier 53,54. In cellular models 55,56, overexpression of MKP/DUSPs evades high ERK activation-induced tumor suppression. Whether and how the suppressive barrier can be avoided or reduced in vivo has never been experimentally tested. The current study is the first demonstration that mutant BRAF-induced activation of ERK signaling is tuneable in vivo and by tuning ERK activation to alter the suppressive barrier, FAK regulates BRAF transforming activity.

In BRAF-mutated melanoma, a complete shutdown of the MAPK pathway is necessary for significant tumor response 57. In patients with BRAFV600E-mutated CRCs, a combination of encorafenib, cetuximab, and binimetinib (MEK inhibitor) treatment increased the response rate to 26% 58, highlighting the importance of complete ERK pathway inhibition. However, the inverse correlation between the level of phosphorylated ERK and the level of stemness/Lgr4 expression seen in mutant BRAF expressing intestinal epithelial cells let us speculate that inhibition of ERK phosphorylation may cause stemness increases in BRAF-mutated CRC cells. The molecular mechanisms underlying ERK phosphorylation inhibition-mediated stemness increase remain to be determined. Given the importance of cancer cell stemness in treatment resistance 59, we propose that the optimal treatment outcome can only be achieved when the inhibition of ERK phosphorylation-mediated stemness increase is simultaneously suppressed.

In sum, the current study reveals the existence of a balance—between the level of phosphorylated ERK, the level of ERK pathway output, and the level of intestinal stemness. Our results show that the “just-right” balance optimal for BRAFV600E-induced cecal tumor formation can be achieved through FAK alteration. Achieving optimal treatment response in BRAF-mutated CRC patients, though, may require abrogation of the p-ERK-stemness regulatory link. That said, the current study could have profound implications for the development of new anticancer agents and new treatment approaches for patients with BRAF-mutated CRC.

Methods

Mice and Treatment

All animal procedures were performed according to protocols approved by the Institutional Animal Care and Use Committee at the University of Pittsburgh. Mice were fed a standard diet (diet ID 5P75; Purina LabDiet, St. Louis, MO). Fakfl/fl mice were received from the Mutant Mouse Resource & Research Centers (MMRRC, cat. no. 009967-UCD). Villin-Cre (cat. no. 021504), BrafLSL-V600E/+ (cat. no. 017837), KrasLSL-G12D/+ (cat. no. 008179) and Rosa26-tdTomato (cat. no. 007914) mice were obtained from the Jackson Laboratory. Genotyping was performed according to the protocols provided by MMRRC and the Jackson Laboratory. Villin-Cre and BrafLSL-V600E/+ mice were crossed to get the BC mice. The littermates harboring BrafLSL-V600E allele were used as controls whenever available. To get the FBC mice, Fakfl/fl mice were first crossed with Villin-Cre mice and BrafLSL-V600E/+ mice, respectively. The offspring Villin-Cre;Fakfl/+ and BrafLSL-V600E/+;Fakfl/+ mice were further crossed with Fakfl/fl mice to get the Villin-Cre;Fakfl/fl (FC) and BrafLSL-V600E/+;Fakfl/fl (FB) mice. The FBC mice were finally obtained by crossing FC and FB mice. Same strategy was used to generate the FKC mice. BC, FBC, KC and FKC mice were euthanized at the indicated age to evaluate the tumor formation. Villin-Cre mice and Rosa26LSL-tdTomato/LSL-tdTomato mice were crossed to get the Villin-Cre; Rosa26LSL-tdTomato/+ mice.

For Bromodeoxyuridine (BrdU) labeling, six-week-old mice were given BrdU (MilliporeSigma) at a dose of 100 mg/kg by intraperitoneal injection two hours before harvesting. For inhibitor treatment, six-week-old mice were given vehicle (a mixture of 50% DMSO and 50% PEG 400), PF-562271 (60 mg/kg in the vehicle), or Erlotinib (100 mg/kg in vehicle) by a single oral gavage four hours (for immunoblotting) or six hours (for qRT-PCR analysis of ERK output genes) before harvesting. MEK inhibitor PD0325901 was given to mice by a single oral gavage at a dose of 25 mg/kg in the vehicle six hours before harvesting. All experiments were performed in both male and female mice.

Cell culture and treatment

HT-29 cells were obtained from the American Type Culture Collection (ATCC) and cultured in DMEM supplemented with 5% fetal bovine serum, 100 units/ml penicillin, and 100 μg/ml streptomycin, in a 37°C humidified incubator containing 5% CO2. The cells were treated with DMSO, PF-562271 (5 μM), or erlotinib (10 μM) for one hour before being harvested for immunoprecipitation.

Protein stability assay

HT-29 cells were seeded twenty-four hours before the experiments. The cells were treated with 100 μg/ml cycloheximide (Selleck Chemicals), 10 μM MEK inhibitor PD0325901, or their combination as indicated. Then the cells were harvested, and the whole cell lysates were used for immunoblotting.

Organoid culture and treatment

Mouse organoids were isolated according to the published protocol with some modifications 60. Briefly, the cecum of the BC mouse was rinsed with cold PBS, cut into small pieces, and washed eight times in cold PBS by gently pipetting. The fragments were incubated in 10 mM EDTA diluted in PBS for 8 minutes in a 37 °C tube rocker. Then the EDTA solution was removed and the tissue was pipetted 10 times in cold PBS. The supernatant was collected and centrifuged at 300 × g for 3 minutes at 4 °C. The cell pellet was washed with DMEM/F-12 medium and centrifuged at 400 × g for 3 minutes at 4 °C. The pellet was resuspended in Cultrex Reduced Growth Factor Basement Membrane Extract, Type R1 (R&D Systems), and seeded into a 24-well plate. Organoids were cultured using Mouse IntestiCult™ Organoid Growth Medium (STEMCELL Technologies) in a 37°C humidified incubator containing 5% CO2. The medium was changed every other day. For inhibitor experiments, the freshly isolated crypts (one hour after seeding) and organoids (five days after seeding) were treated with 10 μM EGFR inhibitor erlotinib and 10 μM MEK inhibitor PD0325901, respectively, for two hours. To isolate protein for immunoblotting after treatment, the crypt cultures were scraped and suspended in 500 μl of TrypLE Express containing 10 μM EGFR inhibitor or 10 μM MEK inhibitor and incubated at a 37°C water bath for 5 minutes with occasional agitation. After the addition of 500 μl of DMEM/F-12 medium, the crypt cultures were centrifuged at 400 × g for 3 minutes at 4 °C. The cell pellets were resuspended in cold PBS and centrifuged again. The final pellets were lysed in RIPA buffer (Alfa Aesar) supplemented with protease inhibitor and phosphatase inhibitor (Thermo Fisher Scientific). Crypt cultures treated with DMSO were used as controls. The lysates were quantified and resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and blotted with the indicated antibodies.

Immunoblotting and immunoprecipitation

After the mice were euthanized, the entire intestines were immediately removed and rinsed twice with ice-cold PBS. The mucosal layers of the small intestine (about 1 cm length), colon (about 1 cm length), and cecum (entire cecum, without appendix) were harvested by scraping with a blade and all procedures were performed on ice. The freshly collected tissue was lysed in RIPA buffer supplemented with protease inhibitor and phosphatase inhibitor. The lysates were quantified and resolved by SDS-PAGE and blotted with the indicated antibodies. SuperSignal Western Blot Enhancer (Thermo Fisher Scientific) was used to enhance the blotting signal when needed. To detect the interaction between FAK and EGFR, the tissue lysates were pre-cleared with Protein G-sepharose beads at 4°C for 30 min. The cleared lysates were incubated with anti-EGFR antibody conjugated to agarose (Santa Cruz Biotechnology) or EZview Red anti-HA affinity gel (MilloporeSigma) at 4°C for 4 hr. The immunoprecipitates were washed three times with lysis buffer containing 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% NP40, and 10% Glycerol, and subjected to SDS-PAGE followed by immunoblotting. The antibodies used for immunoblotting are shown in Supplementary Table 2. All experiments were independently repeated at least three times.

Immunohistochemistry, in situ hybridization, BrdU staining, TUNEL staining, and histopathology

The de-identified human colon tissue samples from BRAFV600E-mutated CRC patients were provided by the University of Pittsburgh School of Medicine, Department of Pathology tissue core. For mouse tissue sections, the mouse intestine was dissected out, rinsed twice with ice-cold PBS, fixed overnight in 10% neutral buffered formalin at 4°C, embedded in paraffin, and finally cut into 5-μm sections. The sections were deparaffinized in xylenes and rehydrated in graded alcohol solutions followed by washes in distilled water. Antigen retrieval was performed for 15 minutes in boiling pH 8 EDTA buffer (Abcam). The sections were allowed to cool to room temperature and then washed with PBS. The endogenous peroxidase was blocked with 3% hydrogen peroxide for 10 minutes. After washing with PBS, the sections were blocked with 20% goat serum diluted in PBS for 45 minutes. Sections then were incubated overnight at 4°C in a humidified chamber with primary antibodies diluted in 3% BSA. Primary antibodies used in this study are listed in Supplementary Table 2. The sections were washed with PBS and incubated with secondary antibodies for 1 hour at room temperature. Color visualization was performed with 3.3’-diaminobenzidine until the brown color fully developed. The sections were counterstained with hematoxylin, dehydrated, and coverslippped with permanent mounting media. The slides were scanned using the Aperio digital pathology slide scanner (Leica Biosystems). The images were analyzed using Aperio ImageScope software.

In situ hybridization (ISH) was performed using the RNAscope 2.5 HD Reagent Kit-BROWN (Advanced Cell Diagnostics) according to the manufacturer’s instructions. The following probes from Advanced Cell Diagnostics were used: Lgr5 (cat. no. 312171) and Lgr4 (cat. no. 318321).

BrdU staining was performed on formalin fixed paraffin embedded (FFPE) tissue sections using monoclonal anti-BrdU antibody (MilloporeSigma) as described by the manufacturer. For Terminal deoxynucleotidyl transferase dUTP nick-end labelling (TUNEL) staining, the FFPE tissue sections was deparaffinized, treated with proteinase K and labelled using the In Site Cell Death Death Detection Kit POD (MilloporeSigma) according to the manufacturer’s instructions. To quantify the results of BrdU, TUNEL and RFP staining, thirty crypts/villi per mouse were scored for three mice in each group.

Myeloperoxidase (MPO) was used as the marker for neutrophils. Ten random-chosen 500 μm-length cecum sections were evaluated for each mouse. MPO+ cells within the band of lamina propria, immediately beneath and surrounding the crypts were counted. Three mice in each group were analyzed. H&E-stained intestinal sections were evaluated for tumor stage by a board-certified GI pathologist (Dr. SF Kuan).

Quantitative Reverse-transcription PCR analysis

Total RNA was extracted from the mucosal layer of the mouse intestine or HT-29 cells using the RNeasy Mini Kit (Qiagen). The DNase-treated RNA was reverse-transcribed using SuperScript III reverse transcriptase (Invitrogen). The PCR reactions were performed on the CFX Connect Real-Time PCR Detection System (Bio-Rad Laboratories) using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad Laboratories). The PCR thermal cycle conditions were as follows: denature at 95 °C for 30 s and 40 cycles for 95 °C, 10 s; 60 °C, 30 s. The specificity of the PCR products was determined by the melting curve analysis. β-actin was selected as an internal reference gene. The sequences of PCR primers are shown in Supplementary Table 3.

Senescence-associated (SA) β-Galactosidase Staining

After the mice were euthanized, the cecum was immediately removed and rinsed with ice-cold PBS. The tissues were frozen in dry ice after the excess liquid was carefully removed using filter paper. Then the tissues were embedded in OCT compound and cut into 10-μm sections. The assays were performed using the Senescence β-Galactosidase Staining Kit (Cell Signaling Technology) according to the manufacturer’s instructions. The sections were counterstained with hematoxylin before being dehydrated and coverslipped with mounting media.

MSI Analysis

The DNA was extracted from FFPE tissue sections using QIAamp DNA FFPE Tissue Kit (Qiagen). Cecal hyperplasia samples were from 6-week-old FBC mice. Cecal tumor samples were from 9-14.5-month-old FBC mice. Cecal tissue of 6-week-old B mice was used as control. According to a prior report 61, five microsatellite repeat markers, Bat24, Bat26, Bat30, Bat37 and Bat64, were used for MSI analysis. PCR amplification was carried out in a multiplex reaction using HSTaq polymerase (Takara Bio, Japan), with primer concentrations 0.5 μM. The thermal cycling conditions were as follows: initial denaturation at 95°C for 5 minutes; followed by 35 cycles of 95°C for 30 s, 60°C for 30 s, and 72°C for 30 s; then a final extension step at 68°C for 30 minutes. PCR fragments were analyzed by capillary electrophoresis, ABI3130XL (Life Technologies), and the GeneMapper ID3.2 program (Life Technologies). Tumor samples with greater or equal 40% MSI were classified as MSI-high (MSI-H), less than 40% as MSI-low (MSI-L), and samples without alterations were classified as MSS.

RNA-seq and data analysis

Total RNA was extracted from the cecal tissues of indicated mice using the RNeasy Mini Kit (Qiagen). After DNase I treatment and performing quality control (QC), 200 ng of high-quality total RNA was proceeded to library construction. Oligo(dT) magnetic beads were used to isolate mRNA. The mRNA was fragmented randomly by adding fragmentation buffer, then the cDNA was synthesized using mRNA template and random hexamers primer. Short fragments are purified and resolved with EB buffer for end repair and single nucleotide A (adenine) addition. After that, the short fragments were connected to sequencing adapters. The double-stranded cDNA library was completed through size selection and PCR enrichment. Agilent 2100 Bioanaylzer and ABI StepOnePlus Real-Time PCR System were used in the quantification and qualification of the sample library. Finally, the qualified RNA-seq libraries were sequenced using Illumina NovaSeq6000 in CD Genomics (Shirley, NY) after pooling according to its effective concentration and expected data volume. The FastQC tool was used to perform basic statistics on the quality of the raw reads. Sequencing adapters and low-quality data were removed by Cutadapt (version 1.17). The alignment tool Salmon (version 0.13.1) was employed to quantify transcript expression based on mm10 reference genome. Output files from Salmon were imported into R (V.4.2.0) and analyzed by DESeq2 package (V1.36.0) to identify differentially expressed genes. All genes were ranked by log2(fold change) and used to check the gene set enrichment by using clusterProfiler[4] (V.4.4.1) in R. The following gene sets were used: MAPK signature 20; intestinal Wnt signature 35; cancer YAP/TAZ target gene signature 36; intestinal differentiation signature 37; intestinal stem cell signature 38; the Hallmark Inflammatory Response gene set (Broad Institute) 34; upregulated fetal spheroid markers 41; upregulated and downregulated genes in human SSA/P 31 (only genes in human SSA/Ps with fold increase>2 or fold decrease<-2 with FDR<0.05 were used).

Whole exome sequencing

DNA was extracted from the cecal tumor of 12-month-old FBC mice using DNeasy Blood & Tissue Kits (Qiagen). Sequencing libraries were generated using Agilent SureSelect mouse All Exon Kit (Agilent Technologies) following the manufacturer’s instructions and index codes were added to attribute sequences to each sample. DNA samples were sonicated using a hydrodynamic shearing system (Covaris) to generate 180-280bp fragments. The remaining DNA overhangs were converted into blunt ends by exonuclease/polymerase. After the adenylation of 3’ ends, DNA fragments were ligated with adapter oligonucleotides. The fragments with adapters on both ends were selectively enriched using PCR. Then the library was hybridized in the liquid phase with biotin-labeled probes, followed by the capture of the exons using streptomycin-coated magnetic beads. Captured libraries were enriched by PCR to add index tags to prepare for hybridization. The resulting products were then purified using the AMPure XP System (Beckman Coulter) and quantified using the Agilent High Sensitivity DNA Assay on the Agilent Bioanalyzer 2100 System. The qualified libraries were sequenced using Illumina NovaSeq6000 in CD Genomics (Shirley, NY) after pooling according to its effective concentration and expected data volume. For the alignment step, BWA is utilized to perform reference genome alignment with the reads contained in paired FASTQ files. For the first post-alignment processing step, Picard tools are utilized to identify and mark duplicate reads from BAM file. The variant calling was performed by using GATK HaplotypeCaller.

Analysis of CRC patient data

TCGA RNA-seq data and mutation data of all cancer types were collected from Xena database (https://xenabrowser.net/datapages/), i.e., TCGA Pan-Cancer (PANCAN), which includes 376 CRC tumor samples and 51 matched normal samples. Expression data for FAK and mutation data for BRAF were extracted for analysis. The difference between the two groups was evaluated using the Student t-test (two-tailed, pairwise).

Author contributions

Conceptualization: CG, EC, JH

Investigation: CG, HG, SFK, CC, XL, JW, FE, JH

Visualization: CG, HG, CC, XL, JH

Funding acquisition: FE, JH

Supervision: EC, JH

Writing – original draft: CG, JH

Writing – review & editing: CG, RES, EC, JH

Data and materials availability

All data are available in the main text or supplementary materials.