Main

To determine how alterations in BRAF and KRAS might affect one another, we systematically evaluated mutations in these genes in 330 colorectal tumours (Table 1). We identified 32 mutations in BRAF: 28 cases with thymine-to-adenine (T–A) transversions at nucleotide position 1,796 (corresponding to an amino-acid swap of glutamate for valine at residue 599; V599E), and one case each of a guanine-to-thymine (G–T) transversion at nucleotide 1,382 (R461I), a T–G transversion at nucleotide 1,385 (I462S), a G–A transition at nucleotide 1,388 (G463E), and an A–G transition at nucleotide 1,798 (K600E). All but two of these mutations seemed to be heterozygous, and in all 20 cases for which normal tissue was available, the mutations were shown to be somatic. In the same set of tumours, there were 169 mutations in KRAS, including alterations to codons 12, 13, 59 and 61. No tumour exhibited mutations in both BRAF and KRAS.

Table 1 BRAF and KRAS mutations in colorectal tumours
Full size table

Mutations in either BRAF or KRAS occurred in all Duke's stages of cancer (results not shown) and also in premalignant lesions. Mutations in both genes seemed to be more common in adenomas larger than 1 cm across than they were in smaller adenomas.

There was also a striking difference in the frequency of BRAF mutations between cancers with and without mismatch-repair (MMR) deficiency (P < 10−6, χ2 test; Table 1). All but one of the 15 BRAF mutations identified in MMR-deficient cases resulted in a V599E substitution.

These results provide strong support for the hypothesis that BRAF and KRAS mutations are equivalent in their tumorigenic effects2. Both genes seem to be mutated at a similar phase of tumorigenesis, after initiation but before malignant conversion. Moreover, we found no tumour that concurrently contained both BRAF and KRAS mutations. In view of the large number of mutations of both genes found in colorectal cancers, this observation is highly statistically significant (P < 10−6, χ2-test) and cannot be easily explained in other ways. This conclusion could not have been reached through the study of melanomas or of most other tumour types in which only one of the two genes is commonly mutated. It is consistent with biochemical observations3 and was suggested by Davies et al.1.

Our results also show that MMR-deficient tumours have a very high incidence of BRAF mutations and a lower incidence of KRAS mutations compared with MMR-proficient colorectal cancers. This is consistent with the idea that both tumour types progress through the same biochemical pathways, but that the mutation spectrum depends on the nature of the underlying genetic instability4. The V599E mutation is the most frequent nucleotide substitution ever identified in a repair-deficient tumour.

The only other tumour type with a BRAF-mutation frequency as high as that seen in MMR-deficient colorectal cancers is melanoma1. Melanomas and MMR-deficient colorectal cancers also share a high incidence of mutations in the oncogene that encodes β-catenin5,6. It will be interesting to see whether melanomas have a repair defect that makes them susceptible to the types of mutation found in MMR-deficient colorectal cancers, and to determine what structural or sequence elements surrounding BRAF codon 599 make it prone to mutagenesis in a repair-deficient background.