Diverse BRAF Gene Fusions Confer Resistance to EGFR-Targeted Therapy via Differential Modulation of BRAF Activity

This article is featured in Highlights of This Issue, p. 515

Cancer genomes are often subject to genomic instability, which can result in various genomic rearrangements, including translocations (1). Some genomic rearrangements can lead to oncogenic transformation, in particular when tumor suppressor genes are being disrupted or oncogenic fusion genes are created (2). Fusion genes are chimeric genes resulting in proteins with altered or novel functions. Alternatively, fusions may facilitate the connection of a strong promoter to the coding sequence of a second gene, causing high expression of a protein such as kinases (1, 2).

Recent advances in sequencing technologies and bioinformatic solutions have enabled the straightforward identification of novel fusion genes (2, 3). This has revealed that their frequency of occurrence varies between many cancer types, ranging from high frequencies in breast cancer (14.7%) to low frequencies in uveal melanoma (0.16%; ref. 2), but also highlighted the vast diversity of fusion partners (>10,800 unique fusion configurations, Quiver Fusion Database).

Oncogenic fusion proteins frequently interfere with signaling pathways that regulate cellular differentiation and proliferation and therefore are excellent targets for personalized cancer therapy. Fusions that involve the BRAF oncogene are mutually exclusive with other oncogenic mutations in the MAPK pathway (e.g., BRAF^V600E), suggesting that BRAF fusion genes promote constitutive MAPK signaling (4). Indeed, previous studies have shown that loss of the auto-inhibitory N-terminal domain of BRAF promotes BRAF kinase activity independent of upstream RAS signaling activity resulting in enhanced downstream MAPK signaling (5, 6). Furthermore, specific BRAF fusion genes have been implicated in acquired resistance to targeted therapies (7). For example, the AGAP3-BRAF fusion has been reported to induce resistance to the BRAF inhibitor vemurafenib in a patient with BRAF^V600E-mutated melanoma that was previously responsive to treatment (7). Hence, BRAF fusion genes may represent an important and understudied mode to activate oncogenic MAPK signaling.

BRAF fusion genes have been identified in multiple cancers with a wide variety of 5′ fusion partners (>60 different partners published, some of which have only been described once; refs. 4, 8). The large variety of BRAF fusion partners complicates straightforward discrimination between oncogenic effects that are shared versus actions that are influenced or even determined by the unique fusion partner. Therefore, in contrast to common BRAF hotspot mutations, this high diversity of fusion gene configurations impedes the clinical interpretation of BRAF fusion genes in relation to their oncogenic potential and treatment responses.

To explore cellular and molecular effects of different BRAF fusion configurations that have recently been identified in colorectal cancers (e.g., AGAP3-BRAF, DLG1-BRAF, and TRIM24-BRAF; ref. 4), we employed patient-derived colorectal cancer organoids (CRC PDO) that are representative patient models (9). We here show that expression of different BRAF fusions in CRC PDOs renders overall resistance to targeted inhibition of the MAPK pathway, similar as the common BRAF^V600E mutation. However, notable differences in signaling activity and sensitivity toward MAPK-targeting drugs were observed between the different BRAF oncogenes, for example, BRAF fusions and BRAF^V600E. On the basis of quantitative phosphoproteomics and RNA sequencing, we show that 5′ fusion partners mainly promote canonical oncogenic BRAF activity by replacing the auto-inhibitory N-terminal region, but also impose unique features that can influence, among others, strength of downstream pathway activation.

BRAF (fusion) genes were cloned from patient RNA (4) into the pInducer20 vector (10). TRIM24-BRAF was ordered from Twist Bioscience. Gateway cloning (Invitrogen) was performed with BRAF (fusion)-specific primers (Supplementary Table S1) according to the manufacturer's protocol and correct insertion of fusion gene into the pInducer20 was verified by Sanger sequencing.

The P18T patient-derived organoids were previously established and characterized (11, 12). P18T colorectal cancer organoids and HEK293 cells (ATCC, ordered April 4, 2018 and thereafter not authenticated) were cultured as described previously (11, 13). For selection of cells stably expressing BRAF (fusion) genes, cells were grown in culture medium containing 400 μg/mL G418 (Santa Cruz Biotechnology). P18T organoid line was validated by SNP array and confirmed Mycoplasma negative with the Mycoplasma PCR ELISA Kit (Roche; last test was performed September 12, 2019) and organoids were kept in culture for 10 passages until final experiments were performed. HEK293 were kept in culture for 16 passages until final experiments were performed.

Each BRAF (fusion) gene construct was stably integrated into the genome of patient-derived P18T organoids or HEK293 cells utilizing lentiviral transduction resulting in polyclonal BRAF (fusion) gene-expressing lines. Virus was produced with HEK293T cells and after 3 days virus was sterile filtered (45 μm) and concentrated, and target cells were infected.

Western blotting was performed as described before (14). Membranes were probed with antibodies directed against HA (RRID:AB_631618), Vinculin (RRID:AB_477629), GAPDH (RRID:AB_2107445), beta-catenin (RRID:AB_397555), pERK (RRID:AB_331646), ERK (RRID:AB_390779), pMEK (RRID:AB_331648), MEK (RRID:AB_823567), pRAF1 (AB_2492224), RAF1 (RRID:AB_390808), pAKT (RRID: AB_2315049), and AKT (RRID: AB_1147620).

Phenotypic drug screen was performed as described previously (14). In brief, 5-day-old organoids were cultured with medium containing either 1 μg/mL doxycycline or ddH₂O, together with 1 μmol/L of afatinib or DMSO. After 7 days, cell viability was visualized by microscopy with Hoechst 33342 (Life Technologies) and 1.5 mmol/L DRAQ7 (Cell Signaling Technology catalog no. 7406) staining. For calculating organoid viability and size, organoids were scored by morphology and analyzed by automated brightfield morphometry using OrganoSeg (15).

For immunofluorescence, organoids and HEK293 cells were probed with an HA-antibody (RRID:AB_390929). Hoechst 33342 was added together with secondary antibodies to stain for DNA. Images were captured with a Leica SP8X microscope using a 40 × objective. Postacquisition analyses of phenotypes were performed manually using ImageJ.

The 4-day drug screen was performed as described previously (14). Organoids were treated with afatinib, selumetinib, encorafenib, navitoclax (Selleck Chemicals), SCH772984 (MedChem Express), and dabrafenib (Bio-Connect). Organoid size was measured by integrating Hoechst signal and contrast using Columbus Cellular Imaging and Analyses (Perkin Elmer). Multiple identical drug combinations were averaged. Dose–response curves were generated using GraphPad software by performing nonlinear regression (curve fit), assuming a standard Hill equation [chosen method: log(inhibitor) vs. Response, constrain top = 100].

Paired-end RNA-sequencing was performed by Macrogen (Korea) on the NovaSeq platform (2 × 100 bp, 60M reads per sample). Data were processed with our in-house RNA analysis pipeline (https://github.com/UMCUGenetics/RNASeq, v.2.4.0, default settings). Principal component analysis, Euclidean distance-based clustering, and differential expression calculations were performed with the DESeq2 package (16). Geneset overrepresentation analysis (ORA) was performed on WebGestalt (17). RNA sequencing data were deposited to EGA with the dataset identifier EGAS00001003558.

For SILAC labeling, HEK293 cells were cultured in high-glucose DMEM (Thermo Fisher Scientific) lacking lysine and arginine supplemented with Lys-0/Arg-0 or Lys-8/Arg-10 (Silantes). Mass spectrometry was performed by the Proteomics Facility (UMC Utrecht). Raw files were analyzed with the MaxQuant software version 1.6.1.0 (18). Phosphorylation sites were analyzed in Perseus software (Version 1.5) using MaxQuant normalized H/L ratios. Data were deposited in the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD013461.

A detailed description of Materials and Methods is described in the Supplementary Materials and Methods.

In a recent RNA-sequencing screen of primary colon tumors, we identified BRAF fusion genes containing different 5′ partner genes (4). To characterize and cross-compare the influence of the 5′ partner between different BRAF fusions, we used CRC PDOs (P18T line) with a mutational background characteristic for BRAF fusion–positive colorectal cancer tumors, that is, non-functional APC and TP53, and KRAS wild-type (4, 11). We employed this platform (Fig. 1A) to stably integrate a panel of BRAF fusion genes (AGAP3-BRAF, DLG1-BRAF, and TRIM24-BRAF) as well as a wild-type BRAF (BRAF^WT), a truncated BRAF gene containing only the kinase domain (BRAF^Kinase), and a BRAF gene with the canonical V600E mutation (BRAF^V600E; Fig. 1B). BRAF (fusion) gene expression was under the control of doxycycline-inducible Tet-On activity to ensure controlled and selective construct expression (Fig. 1C).

We confirmed selective mRNA expression of all BRAF variants upon doxycycline administration (Fig. 2A). Furthermore, the presence of the various BRAF proteins was visualized by immunoblot staining against the C-terminal HA-tag (Fig. 2B). In the absence of doxycycline, no BRAF (fusion) mRNA or protein was detected, demonstrating tight control of gene expression by tetracycline-responsive promoters.

Whereas various studies have shown that BRAF fusions promote MAPK pathway activation by effectuating loss of the auto-inhibitory N-terminal domain of BRAF (4, 19), specific effects imposed by the 5′ partner on the fusion protein have only been studied to a limited extent (20). Indeed, the 5′ fusion partners of BRAF, for example, AGAP3 (21), DLG1 (22), or TRIM24 (23) manifest distinct biological functions in their original conformation and a potential carry-over toward the fusion is likely.

First we characterized the effects of fusion gene expression during normal culture conditions on MAPK pathway activation, which is implicated in cellular growth and proliferation (24). Normal culture conditions of CRC PDOs include EGF that induces baseline levels of MAPK pathway activity. Only the expression of DLG1-BRAF resulted in enhanced pERK levels, while the other BRAF fusions, as well as the well-known oncogenic mutant BRAF^V600E did not deviate from baseline pERK levels under normal culture conditions (Fig. 2C). In agreement with previous literature (25), none of the BRAF (fusion) proteins had an effect on AKT activity (Supplementary Fig. S1). Importantly, BRAF (fusion) protein levels vary between organoid lines (Fig. 2B), potentially influencing observed phenotypes. The expression levels of the BRAF fusions do not correlate with the levels of ERK phosphorylation, which indicates that there is a specific influence of the 5′ fusion partner on the downstream MAPK pathway activation (Supplementary Fig. S2).

We hypothesized that the 5′ fusion partners can redirect the subcellular localization of the BRAF fusion proteins, thereby affecting the efficiency by which the BRAF kinase domain can activate downstream MEK. Therefore, we visualized the intracellular localization of the BRAF variants by using immunofluorescent staining against the C-terminal HA-tag. Indeed, we observed differences in the cellular localization of the different BRAF fusions (Fig. 2D). The BRAF^WT, BRAF^Kinase, and BRAF^V600E proteins exhibited, as expected, a diffuse localization pattern throughout the cytoplasm (26). Whereas a similar localization pattern was observed for the AGAP3-BRAF and TRIM24-BRAF fusion proteins, DLG1-BRAF fusions were primarily localized to the plasma membrane. This is most likely due to the retained L27 domain of DLG1 in the DLG1-BRAF fusion, which triggers native DLG1 localization likewise to the apical plasma membrane (27).

On the basis of the observed differences between the BRAF fusions with respect to ERK activation, we aimed to characterize the type and extent of unique intracellular signaling pathways that are specifically activated by a BRAF fusion, in comparison with BRAF^V600E and BRAF^WT.

To identify all downstream phosphosites that are regulated by a BRAF fusion, we performed an unbiased SILAC-based phosphoproteomics screen (Supplementary Fig. S3A). Because organoids are not a suitable platform to scale to sufficient protein quantities required for phosphoproteomic screens, and organoid growth kinetics may potentially be affected in SILAC medium, we opted to use HEK293 cells instead. HEK293 cells are a well-characterized cell system with an unperturbed MAPK pathway (28), a major prerequisite to not mask potential BRAF (fusion)-induced effects on MAPK signaling, and are compatible with growth in both types of SILAC media, which is not trivial for most well-known cancer cell lines. Like our organoid lines, we confirmed inducible BRAF fusion gene expression by mRNA RT-PCR and Western blot analysis (Supplementary Fig. S4A and S4B). In contrast to P18T organoids, weak BRAF fusion gene expression was detected in the uninduced state (−doxycycline) of HEK293 cells (Supplementary Figs. S4A and S4B and S5), presumably because of supplemented FBS in HEK293 culture medium (29). Importantly, however, except for GFP- and BRAF^WT-expressing cells, a strong increase in pERK levels was detected in HEK293 cells upon expression of all oncogenic BRAF variants (Supplementary Fig. S4C). Whereas a significant increase in ERK phosphorylation during unperturbed culture conditions was only observed upon DLG1-BRAF expression in our CRC PDOs (Fig. 2C), this was not the case in HEK293 cells. Presumably, this discrepancy may be the result of technical differences between the 2D and 3D models, such as different integration and expression efficiencies of the fusion proteins, or may be attributed to the inherent biological differences between the two model systems. In support of the latter, the distinctive localization patterns of BRAF fusion proteins observed in organoids (Fig. 2D) was absent in unpolarized HEK293 cells (Supplementary Fig. S5; ref. 24). Moreover, the MAPK signaling pathway is already active in organoids prior to induction of the fusion variants, while largely inactive in HEK293 cells, potentially masking their effects during normal growth conditions. As a result, we now reveal phosphorylation targets of the different oncogenic BRAF variants as a result of intrinsic capacity, rather than induced by differences in subcellular localization.

To identify targets directly downstream of BRAF (fusion) protein signaling by phosphoproteomic analysis, we measured the earliest timepoint of BRAF (fusion) protein expression upon doxycycline-mediated induction. Time-course measurements of HA-tagged protein expression and ERK phosphorylation in the GFP- and DLG1-BRAF–expressing cell lines identified 4 hours as the minimal duration of doxycycline exposure for protein expression and robust downstream MAPK pathway activation (Supplementary Fig. S6A and S6B). Same kinetics was confirmed in the remaining BRAF fusion cell lines (Supplementary Fig. S6C).

In general, we observed that expression of BRAF fusions, BRAF^V600E, and BRAF^Kinase mainly induced an increase in protein phosphorylation in 4 hours (Fig. 3A, Table 1, Supplementary Figs. S3B and S6D). BRAF^WT expression only resulted in one significantly downregulated phosphosite belonging to the BRAF protein itself, indicative of a negative feedback response induced by BRAF overexpression (Fig. 3A). In contrast to BRAF^V600E and BRAF fusions, the presence of BRAF^Kinase resulted in a rather low number of upregulated phosphosites (Table 1, Supplementary Fig. S3B).

To explore kinase activities responsible for the deregulated targets within the phosphoproteomics data, we used kinase-substrate enrichment analysis (KSEA; ref. 30). With the exception of GFP, we noticed that all BRAF variants showed a significant activation of kinases involved in the MAPK signaling pathway, including MEK1/2, ERK1/2, and Raf (Fig. 3B, Supplementary Fig. S3C, Supplementary Table S2). In addition, BRAF fusions and BRAF^V600E also showed significant activation of kinases involved in cell-cycle progression and DNA damage response, such as CDK1, CDK2, Aurora kinase B, and CHEK1 (Fig. 3B, Supplementary Table S2).

Next, we explored phosphotargets that are common to all BRAF fusions, versus unique targets per BRAF variant. We identified 289 phosphosites and 298 proteins that are shared as substrate (direct or indirect) by all BRAF fusions, BRAF^V600E, and BRAF^Kinase (Supplementary Fig. S3D). As indicated already, most of the overlapping targets include members of the MAPK signaling pathway, such as ERK1/2 (MAPK3/1), BRAF, and MEK2 (MAP2K2). Moreover, these phosphotargets turned out to be most strongly upregulated (Fig. 3A, Supplementary Fig. S3B). Besides shared downstream targets, we found that each BRAF fusion also induced the phosphorylation of a unique set of substrates (Fig. 3A, Supplementary Fig. S3B). Intriguingly, however, the degree by which these unique phosphosites are phosphorylated was generally lower when compared with the fold change (FC) of shared phosphosites (Supplementary Fig. S3B).

Next, when analyzing which pathways are represented by the set of downstream phosphotargets [ORA (WebGestalt); ref. 17], we noticed a high degree of overlap between the different BRAF fusion and BRAF^V600E-expressing cells (Fig. 3C). In particular, the pathways mainly converged on MAPK signaling pathway regulation and cell-cycle progression (Fig. 3C and D).

Together, from a protein intrinsic point of view, these data indicate that the different BRAF fusions have largely similar substrates as BRAF^V600E and activate similar signaling pathways, mainly involving MAPK pathway activation and cell-cycle progression. To complement the phosphoproteomic analysis and to improve our understanding of the global impact of BRAF (fusion) genes, we performed Western blot analysis and RNA-sequencing (Materials and Methods) on CRC PDOs. Western blot analysis showed that doxycycline-induced expression of BRAF^V600E, AGAP3-BRAF, and DLG1-BRAF in CRC PDOs indeed resulted in increased MEK-ERK phosphorylation and confirmed varying degrees of MAPK pathway activation between oncogenic BRAF variants (Supplementary Fig. S7A). On the basis of gene expression analysis (Materials and Methods) of organoids expressing fusion genes, we indirectly confirmed the activation of MAPK signaling by BRAF fusions, specifically for the DLG1-BRAF fusion. The DLG1-BRAF fusion induced a large amount of differentially expressed genes in CRC PDOs that were mainly involved in the cell cycle and signaling by Rho GTPases (Supplementary Fig. S7B–S7D; Supplementary Tables S3 and S4; refs. 31, 32). Indeed, MAPK pathway activation promotes cell proliferation (33) and migration (33, 34), which corresponds to the strong effects of DLG1-BRAF on ERK activation. Furthermore, we confirmed that the expression of genes implicated in cell-cycle regulation, such as CHEK1, CHEK2, and CCNB2 was similarly regulated in organoids expressing DLG1-BRAF and in a tumor sample positive for the DLG1-BRAF fusion, corroborating the validity of these findings (Materials and Methods, Supplementary Fig. S8A and S8B).

Besides KRAS, oncogenic mutations in BRAF (e.g., V600E) have been associated with resistance to anti-EGFR–targeted therapy in metastatic colorectal cancer (mCRC; refs. 35, 36). BRAF fusions may also influence the sensitivity of tumor cells to anti-EGFR–targeted therapy due to constitutive activation of downstream MAPK signaling. Therefore, we challenged the CRC PDOs with a pan-HER inhibitor (afatinib) for 7 days and scored viability by microscopy (Fig. 4A). Consistent with previous patient-derived colorectal cancer studies (13), most of the GFP- and BRAF^WT-expressing organoids died upon EGFR inhibition, while the majority of BRAF^V600E-mutant organoids showed resistance to similar treatment (Fig. 4B; Supplementary Fig. S9A). In addition, we observed that expression of all BRAF fusion variants provided resistance to afatinib (Fig. 4C and D; Supplementary Fig. S9A). Interestingly, BRAF^Kinase organoids exhibited an intermediate phenotype to anti-EGFR–targeted treatment compared with the BRAF fusion organoids, showing few large surviving organoids together with significant cell death (Fig. 4B and D; Supplementary Fig. S9A and S9B). As BRAF^Kinase expression is comparable with expression of BRAF fusions (Fig. 2B), these data suggest that a protein domain at the N-terminal side of the BRAF kinase domain is important to maximize BRAF activity.

Previous studies have shown that constitutive MAPK pathway activation plays an essential role in anti-EGFR therapy resistance in mCRC (13, 37). To validate that BRAF fusion genes can promote MAPK pathway signaling independent of external EGF stimulation, we investigated ERK and AKT activity upon afatinib treatment. In accordance to the observed phenotypes, the afatinib-sensitive lines (GFP and BRAF^WT) failed to phosphorylate ERK upon EGFR inhibition while afatinib-resistant lines (BRAF fusions, BRAF^V600E, and BRAF^Kinase) were able to sustain ERK phosphorylation (Fig. 4E). Along similar lines, we confirmed sustained phosphorylation by BRAF fusions and BRAF^V600E of additional targets identified with the phosphoproteomics screen, like phosphorylated MEK (pMEK) and pCDK2 (Fig. 4F). Intriguingly, this was especially apparent during EGFR inhibition as common targets of the MAPK pathway seem already activated to significant levels during normal growth conditions (Fig. 2C). In addition, no influence of BRAF fusions on pAKT levels was observed (Supplementary Fig. S10). Intriguingly, although BRAF^V600E- and BRAF^Kinase-expressing organoids were able to sustain ERK phosphorylation, the degree of ERK phosphorylation was affected by EGFR inhibition. This is concordant with two recent studies, which show that KRAS-mutant lung cancer cells still require upstream receptor signaling for tumor growth and survival (35, 38). Dependency on continuous upstream signaling input at the receptor level, however, was not observed in organoids expressing the BRAF fusion genes. Furthermore, the DLG1-BRAF fusion consistently showed stronger ERK phosphorylation, both at unperturbed (Fig. 2C) and afatinib-treated conditions (Fig. 4E), pointing toward an enhanced capacity to activate the downstream MAPK pathway, possibly due to its localization at the plasma membrane.

To identify drugs that could be used for the treatment of patients with mCRC with BRAF fusion genes, we tested a panel of MAPK pathway–targeting agents in a drug screen (Fig. 5A).

Previously, a synergistic effect was observed with dual inhibition of the MAPK pathway on BRAF- and RAS-mutant colorectal cancer cells (13, 39). Therefore, we investigated whether a similar effect could be achieved with combinatorial targeting of the MAPK pathway. Organoids expressing BRAF^WT or GFP were highly sensitive to afatinib (pan-HERi) treatment in this drug screen assay, while BRAF^V600E, BRAF^Kinase, and BRAF fusion–expressing organoids showed resistance, with BRAF^V600E-, DLG1-BRAF-, and TRIM-BRAF–mutant organoids showing resistance at even high concentrations (Fig. 5B, Supplementary Fig. S11A). MEK inhibition (selumetinib) had a similar inhibitory effect on BRAF^WT, BRAF^V600E, BRAF^Kinase, and BRAF fusion–expressing organoids, which was slightly lower compared with its inhibitory effect on the control GFP-expressing line (Fig. 5C, Supplementary Fig. S11A). Combining selumetinib with afatinib resulted in an additive effect on BRAF^Kinase- and AGAP3-BRAF–expressing organoids that already exhibited higher sensitivity to afatinib compared with organoids with the other BRAF fusions (Fig. 5D; Supplementary Fig. S11A). We observed only a minor additive effect of this combinatorial treatment on highly afatinib-resistant DLG1-BRAF and TRIM-BRAF fusions, as well as BRAF^V600E mutants, compared with afatinib or selumetinib alone. We conclude that, unlike mutant KRAS, the combination of pan-HER and MEK inhibition only has an additive effect in some, that is, AGAP3-BRAF and BRAF^Kinase, but not all oncogenic BRAF-expressing colorectal cancer organoids.

Next, we tested whether combining a MEK inhibitor with an ERK inhibitor has additive or synergistic effect. All oncogenic BRAF variants showed approximately similar sensitivity to the specific ERK1/2 inhibitor SCH772984 (40), which was comparable with the control GFP-expressing organoids (Fig. 5E, Supplementary Fig. S11A). Combinatorial targeting of MEK (selumetinib) and ERK (SCH772984) improved overall sensitivity, without major differences between the lines (Fig. 5F, Supplementary Fig. S11A).

Finally, organoids were exposed to the selective, ATP-competitive BRAF inhibitor encorafenib or dabrafenib (41). Similar to what has been reported previously (39), organoid growth was barely affected by the abrogation of BRAF kinase activity by encorafenib or dabrafenib, independent of whether oncogenic BRAF variants were present or not (Fig. 5G and I, Supplementary Fig. S11A). Surprisingly, combinatorial targeting of the MAPK pathway that includes BRAF inhibition, either combined with pan-HER or with MEK inhibitors, did not reveal significant improvements over pan-HER or MEK inhibition alone (Fig. 5H and J, Supplementary Fig. S11A).

Together, screening drug sensitivities across multiple oncogenic BRAF lines revealed intertumoral differences against the pan-HER inhibitor afatinib. AGAP3-BRAF–expressing organoids behaved overall similarly as drug sensitive BRAF^WT and the BRAF^Kinase mutants to combined EGFR and MEK inhibition, while DLG1- and TRIM24-BRAF fusions approximated resistance phenotypes of the BRAF^V600E mutant. In contrast, ERK or BRAF inhibition, alone or in combination, gave similar responses between all the lines, irrespective of oncogenic BRAF variants, BRAF^WT, or normal control (GFP).

BRAF fusions are recurrent events throughout various cancer types and form oncogenic drivers (8). Thus far, few studies have addressed the effect of BRAF fusion expression on intracellular signaling and cellular processes (19, 42). Moreover, while some results have been obtained for ALK and BRAF fusions (20, 43), the possible effects induced by specific 5′ fusion partners of BRAF have not been thoroughly investigated in a clinically relevant model system. It is generally assumed that the loss of the N-terminal domain is responsible for enhanced oncogenic BRAF activity (19). Whereas this is consistent with our findings showing enhanced MAPK pathway activation upon BRAF fusion gene expression, we systematically investigated the influence of 5′ partner genes on BRAF activity.

As opposed to previous studies describing fusion gene characterization in cell lines, we here used a patient-derived organoid model that is the closest representative of human colorectal tumors that is compatible with biochemical analysis of signaling pathway alterations by BRAF fusions. Using the organoid system, we observed that subcellular localization of the BRAF fusion protein, as well as the level of MAPK pathway activation were affected by the 5′partner. A distinct localization of DLG1-BRAF proteins was detected at the plasma membrane, possibly responsible for the enhanced activation of the MAPK pathway in unperturbed as well as in drug-treated conditions as compared with AGAP3-BRAF, TRIM24-BRAF, and the canonical BRAF^V600E mutation. Wild-type BRAF is usually expressed throughout the cytoplasm and gets recruited to the plasma membrane through activated Ras (44). Previous studies showed that localization of the BRAF protein to the plasma membrane potentiates BRAF signaling through proximity to downstream effectors (45). Mediated by its L27 domain, DLG1 is known to localize to junctions at the plasma membrane (27). Exactly this domain is retained in the DLG1-BRAF fusion and is likely to promote its plasma membrane localization independent of active RAS. Redirected subcellular localization showcases how the 5′ fusion partners can influence functionality of oncogenic BRAF. In concordance with previous studies, the unique localization of DLG1-BRAF at the plasma membrane was lost in unpolarized 2D HEK293 cells and underscores the advantage of using 3D tumor organoid models for assessing functional effects of oncogenes (46).

We performed an unbiased phosphoproteomic screen to identify proteins that are differentially phosphorylated by each of the different BRAF (fusion) genes. We observed that BRAF fusion signaling mainly converges on the same signaling targets and pathways (e.g., MAPK) as the oncogenic BRAF^V600E mutant, with very few BRAF fusion–specific targets. On the other hand, whole-transcriptome expression analysis identified shared as well as fusion-specific effects. Intriguingly, most of the deregulated genes that all BRAF fusions have in common are presumably beyond the traditional effects of MAPK pathway activity. ERK activity levels (with the exception of enhanced levels in DLG1-BRAF) were similar between all lines, which may be attributed to the already default active MAPK signaling pathway due to EGF presence at normal culture conditions. The expression of the DLG1-BRAF fusion uniquely affected a subset of genes, which mainly converged on the cell cycle. Similarly, cell-cycle–related genes were deregulated in a patient with DLG1-BRAF fusion–positive colorectal cancer, substantiating this observation. Together, this data show that BRAF fusion genes commonly impact gene sets involved in cell proliferation and migration and that specific BRAF fusions can affect unique and distinct sets of genes.

In clinical practice, patients with mCRC are treated with anti-EGFR–targeting mAbs, given that the patient does not harbor oncogenic KRAS or NRAS mutations (47, 48). In addition, oncogenic mutations in BRAF (e.g., V600E) are also associated with anti-EGFR therapy resistance in colorectal cancer (49). Here, we observed that all BRAF fusion genes tested were able to confer resistance to targeted inhibition of EGFR with the small-molecule inhibitor afatinib. Our data highlight that BRAF fusions activate the MAPK pathway to equal or even more pronounced levels than oncogenic BRAF^V600E, both at unperturbed and afatinib-treated growth conditions. Furthermore, sustained activity of the downstream MAPK signaling pathway in the presence of afatinib is in concordance with resistance mechanisms observed in patients that are insensitive to EGFR-targeting agents (50). These findings emphasize the clinical relevance of this study that suggest to include BRAF fusions to genetic screening programs for patients with colorectal cancer to assist personalized therapy design.

In conclusion, we show in a patient relevant model system that 5′ fusion partners can impose a unique influence on the oncogenic effects of BRAF, among others by redirecting its subcellular localization. Nevertheless, all BRAF fusion genes showed insensitivity toward targeted inhibition of EGFR family members. Therefore, we provide a strong incentive to include BRAF fusion genes to genetic screening programs for patients with colorectal cancer amenable for anti-EGFR therapy.

No potential conflicts of interest were disclosed.

Conception and design: C. Stangl, J.B. Post, M.J. Koudijs, E.E. Voest, H.J.G. Snippert, W.P. Kloosterman

Development of methodology: C. Stangl, J.B. Post, E.E. Voest, W.P. Kloosterman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Stangl, H.R. Vos, R.M. van Es

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Stangl, J.B. Post, M.J. van Roosmalen, N. Hami, H.R. Vos, E.E. Voest, H.J.G. Snippert

Writing, review, and/or revision of the manuscript: C. Stangl, J.B. Post, E.E. Voest, H.J.G. Snippert, W.P. Kloosterman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Stangl

Study supervision: E.E. Voest, W.P. Kloosterman

Other (performed viral transductions): I. Verlaan-Klink

This work was funded by the Oncode Institute, which was partly financed by the Dutch Cancer Society, and was funded by the gravitation program CancerGenomiCs.nl from the Netherlands Organization for Scientific Research, by grants from the Dutch Cancer Society [Koningin Wilhelmina Fonds (KWF), UU 2013-6070 and UU 2012-5710], by a SU2C-DCS International Translational Cancer Research Dream Team Grant (SU2C-AACR-DT1415), and by a ERC starting grant (to H.J.G. Snippert), and a Dutch Cancer Society grant. Stand Up To Cancer is a division of the Entertainment Industry Foundation. Research grants were administered by the American Association for Cancer Research, the Scientific Partner of SU2C. The authors thank KWF for funding this project. Furthermore, we thank all the members of the Kloosterman, Voest, Snippert, and Bos laboratories for fruitful discussions and support. We thank the Proteomics facility (UMCU) and Macrogen (Korea) for their help with the Phosphoproteomics and RNA sequencing experiments. We thank Glen Monroe and Hans Bos for critical reading of the article and Francis Blokzijl for providing a script for the RNA sequencing analysis. We thank Robert Coebergh, Jan Ijzermans, Anieta Sieuwerts, and John Martens for collaborations that led up to this work.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.