The clinical and genomic distinctions of Class1/2/3 BRAF-mutant colorectal cancer and differential prognoses

BRAF mutations are the oncogenic drivers in colorectal cancer and V600 mutations (Class1), which lead to RAS-independent active monomers, are the most common mutation types. BRAF non-V600 mutants can be further classified as RAS-independent active dimers (Class2) and RAS-dependent impaired kinase (Class3). We retrospectively reviewed the mutational profiles of 328 treatment-naïve colorectal tumors with BRAF mutations detected using capture-based hybrid next-generation sequencing targeting 400 + cancer-related genes. The clinical and genetic distinctions of patients harboring Class1/2/3 BRAF mutations were investigated, which revealed that tumors with Class1 BRAF mutations showed more unique genomic profiles than those with Class2/3 mutations. Also, by using an external dataset from cBioPortal, we demonstrated that patients with Class3 BRAF mutations had the best survival outcomes compared to the other two subgroups. These findings promoted the development of anti-BRAF strategies by distinguishing BRAF mutant subgroups. Supplementary Information The online version contains supplementary material available at 10.1186/s40364-022-00443-8.


To the editor,
BRAF mutation is considered to be an oncogenic driver in colorectal cancer (CRC) and V600E is the most dominant BRAF mutation [1]. Mutations occurring at V600 of BRAF such as V600E/K/D/R/M cause the expression of RAS-independent active monomeric proteins, which are grouped as Class1 BRAF mutations. Other non-V600 BRAF mutations can be classified based on signaling mechanism and kinase activity, including RAS-independent active dimers (Class2) and RAS-dependent impaired kinase (Class3) [2,3]. The molecular features of the three classes of BRAF mutations varied and could lead to different treatment strategies [4]. For instance, canonical BRAF inhibitors, such as encorafenib and vemurafenib, directly target the monomeric BRAF protein, to which only the patients carrying Class1 BRAF mutations were sensitive [5]. Thus, comprehensively studying the distinctions in signaling transduction and other genetic features among patients carrying different classes of BRAF mutations may inspire the combination strategies of existing BRAF inhibitors and other targeted therapies. Also, the development of next-generation BRAF inhibitors targeting the dimeric BRAF protein is urgently needed.
Somatic mutation profiling revealed that the most frequently mutated gene in Class1 and Class3 was TP53 (66.9% and 84.9%) while that in Class2 was APC (75.9%, Fig.S1B). KRAS, NRAS, and APC mutations were significantly enriched in the Class2/3 subgroups, while the mutational frequency of RNF43 was significantly higher in the Class1 subgroup. The genes involved in PI3K signaling pathway, including PTEN and AKT3, were significantly more frequently mutated in Class2 (Fig. 1B). In addition, Wnt or RTK/RAS signaling pathway alterations were significantly more common in Class2/3 subgroups compared to Class1, but pathogenic cell cycle pathway alterations were only detected in Class1/2 (Fig. 1C). Then, we compared the allele frequencies (AFs) of BRAF mutations and concurrent KRAS/NRAS/ HRAS mutations that were reported to cause RAS signaling activation. As shown in Fig. 1D, the proportions of patients harboring RAS activating mutations in Class2 (31%) and Class3 (43%) were significantly higher than that in Class1 (7%, p < 0.001). In Class2/3, the AFs of BRAF mutations were commonly lower than the concurrent RAS activating mutations, indicating that these BRAF mutations might be subclonal. Furthermore, the mutational signature associated with nucleotide excision repair (NER) was significantly enriched in tumors with Class1 BRAF mutations, but the mutational signature related to the apolipoprotein B mRNA editing catalytic polypeptide-like (APOBEC) enzyme was more commonly identified in Class3 compared to Class1 (Fig. 1E). The tumor mutational burden (TMB) of patients harboring BRAF mutations was significantly higher than that of patients with wildtype BRAF (median: 5.8), while Class2 BRAF-mutant tumors demonstrated non-significantly higher TMB than the other two subgroups (median: 10.4 vs. 7.6 and 8.4, Fig. 1F). However, microsatellite instability-high (MSI-H) tumors were slightly more common in Class1 compared to Class2/3 (Class1: 11% vs. Class2: 7% vs. Class3: 4%, Fig. 1G), but significantly when compared with wildtype subgroup (5%, p < 0.001).
(See figure on next page.) Fig. 1 The genomic features of tumors with Class1 BRAF mutations are more unique than those with Class2/3 mutations. A The positions of detected mutations in each class (Class1: green; Class2: purple; Class3: orange) along the tyrosine kinase domain of BRAF gene are illustrated by the lollipop chart, whose frequencies are shown by the pie charts. B The proportion of patients with genomic alterations across three BRAF classes. All genes whose mutational frequencies were over 10% in either Class1 or 2 or 3 subgroup were included for analysis and significant genes (p < 0.05 based on Fisher's exact test) are grouped based on enrichment patterns indicated below the gene names (e.g., Class1 vs. Class2/3 subgroup includes genes whose mutational frequencies were significantly different in Class1 compared to both Class2 and Class3). Genes involved in PI3K, RTK/RAS, and Wnt signaling pathway are colored in orange, blue, and red, respectively, and the rest significant genes are black. C Due to the lack of survival data in our cohort, we utilized an external dataset from cBioPortal which contained 455 unresectable metastatic CRC patients to investigate the association between prognosis and BRAF mutation subtypes (wildtype: N = 396, Class1: N = 38, Class2: N = 8, Class3: N = 8, Others: N = 5) [6]. First, similar analyses on genetic characteristics including concurrent RAS-activating mutations, MSI-H, and TMB levels were performed to validate the findings observed in our cohort. As shown in Fig. 2A, Class1 BRAF V600E mutation was not concurrent with any RAS activating mutations, which were detected in 62% of patients with wildtype BRAF and 50% of those in Class3 (p < 0.001). Similar to our cohort, the proportion of MSI-H patients was the highest in Class1 (24%, Fig. 2B). Even though no significant difference in TMB was observed among those subgroups in this external cohort, a trend of higher TMB in BRAF-mutated tumors than BRAF-wildtype tumors was observed (Fig. 2C), which also supported the genetic similarity between the two cohorts. Then, we investigated the survival outcomes of patients with Class1/2/3 BRAF mutations. As shown in Fig. 2D-E, patients with Class1 BRAF mutants had the worst progression-free survival (PFS; median: 5.0 months) and overall survival (OS; median: 12.7 months) whereas patients in Class3 demonstrated the best prognosis (median PFS: 7.3 months; median OS: 31.9 months). Considering the restricted cohort size for patients with Class2/3 BRAF mutations, we exploited a second independent dataset for survival analysis [7]. A similar trend of better OS was observed in patients with Class 3 BRAF mutations versus the other two subtypes (Fig. S2), although the sample size of this cohort was still limited (Class1: N = 27, Class2: N = 5, Class3: N = 8) due to the low frequency of Class2/3 BRAF mutations.
Matsumoto et al. [8] reported a strong association between RNF43 mutations and BRAF V600E, which was consistent with our observation (Fig. 1B). However, other genes in the Wnt signaling pathways, e.g., APC, AMER1, and CTNNB1 showed the opposite trend and the percentage of patients with alterations in Wnt pathway were significantly higher in Class 2/3 than Class1 (Fig. 1C). Notably, the TMB levels of each BRAF subgroup in our cohort were not in accordance with the public cBioPortal cohort [6], which might be due to the restricted cohort size and the different targeted NGS panels used for TMB estimation. In addition, the large difference in cohort size between Class1 and Class2/3 both in our cohort and the external dataset should be noted, implying the warrant of further validation with a larger sample size of Class2/3 BRAF mutations. Also, the zygosity status (homozygous vs. heterozygous) of the BRAF mutations was not evaluated in the current NGS pipeline and the external dataset, which was reported to be associated with the response to BRAF inhibitor treatment [9]. Thus, to comprehensively interpret treatment sensitivity and prognosis between different zygosity status, further studies are needed. The prognosis of three BRAF subgroups analyzed here was supported by a previous study that reported the longest PFS and OS in patients with Class3 BRAF mutations [2]. As the available BRAF inhibitors only target BRAF V600 mutations, patients with Class2/3 BRAF mutations are not sensitive to them [10]. Encouragingly, a secondgeneration BRAF inhibitor, BGB-3245, targeting both monomer and dimer forms of active BRAF proteins is in the early phase of clinical trial (NCT04249843), which might expand the benefits to patients with non-V600 BRAF mutations. In addition, due to the high dependency on EGFR signaling, Class3 BRAF mutations were proven to respond to anti-EGFR therapy, indicating a bright future of combination therapies in patients with non-V600 BRAF mutations [4].
In conclusion, the CRC tumors harboring Class1 BRAF mutations demonstrated more unique genetic profiles than those with Class2/3 mutations and patients with Class3 BRAF mutations had the best survival outcomes among all BRAF subgroups. Our findings suggested the potential differential treatment strategies for patients with different BRAF mutation subtypes and emphasized the urgency of the development of anti-BRAF drugs, especially targeting the active dimeric BRAF proteins.