A total of 292 somatic base substitutions were in protein-coding sequences ( and Supplementary Table 4). Of these, 187 caused amino acid changes (non-synonymous), including 172 that were missense and 15 nonsense, and 7 affected highly conserved bases at splice sites. There were 105 silent (synonymous) substitutions. One somatic substitution was found in the microRNA hsa-mir-518d in the central region of the stem structure. None of the 66 confirmed small insertions and deletions was in coding sequences.
On the assumption that most silent mutations are biologically inert, the ratio of non-synonymous to synonymous substitutions in protein-coding sequences can be used to estimate the extent of selection overall on non-synonymous changes6,18
. The ratio in COLO-829 was 1.78, not significantly different from that expected by chance (P
=0.5). Thus, the substantial majority of mutations do not seem to be subject to positive or negative selection. However, this test will be insensitive to small numbers of selected mutations.
Several individual substitutions highlighted candidate novel cancer genes. For example, two heterozygous missense mutations (het, p.S229L; het, p.D283H) were in SPDEF, which encodes a member of the ETS transcription factor family19
expression is associated with disease progression in prostate, breast and ovarian cancer20,21
. Sequencing SPDEF through 48 additional, untreated metastatic melanomas revealed a further somatic mutation (p.W158*). A missense mutation (het, p.G297E) was identified in the matrix metalloproteinase gene MMP28
. Missense mutations in the family of matrix metalloproteinases were recently reported in melanoma, including several in MMP28
). A missense mutation (het, p.N561K) was also identified in UVRAG
, which was originally identified in a complementation assay for ultraviolet light sensitivity in xeroderma pigmentosum group C cells, has recently been shown to have an important role in autophagy and has been proposed as a tumour suppressor gene23,24
. To determine the significance of these and other mutations, analysis of additional melanomas will be required.
Of the 37 somatic rearrangements mapped to the base pair, 19 interrupted protein-coding genes. No in-frame fusion genes were predicted, but one gene (MAGI2) had an exonic deletion that creates a predicted in-frame rearranged transcript. PCR with reverse transcription (RT–PCR) across the exon–exon rearrangement boundary showed that the chimaeric transcript is expressed, but the biological significance of this rearrangement remains to be clarified.
Rearrangements occur at a high frequency in cancer over fragile sites. In COLO-829, rearrangements were found in FHIT and WWOX, which overlie the fragile sites FRA3B and FRA16D, respectively. Both rearrangements are contained within individual introns, however, and are not predicted to alter the sequences of the encoded proteins.
There are several modest copy-number reductions and increases that affect large genomic regions and which alter the copy number of many protein-coding genes (). Of particular interest, however, is the relatively restricted region of high (8- to 12-fold) copy number increase on chromosome 3p which contains four complete genes: RARB, TOP2B, NGLY1
). A ~0.5-megabase (Mb) region on chromosome 15 is also amplified to 4–6 copies and contains MKRN3
. None of these genes has previously been implicated in cancer development through amplification, although another member of the retinoic receptor family, RARA
, is consistently rearranged in acute promyelocytic leukaemia25
. A 12-kilobase (kb) internal homozygous deletion of the recessive cancer gene PTEN
is predicted to cause premature termination and is presumably implicated in the development of COLO-829. However, the significance of a 40-kb homozygous deletion that removes the start ATG codon of CNOT1
, a repressor of transcriptional activation by oestrogen receptor26
, is unknown.
Of the three previously identified driver mutations in COLO-829, genome-wide sequencing revealed BRAF V600E in addition to the PTEN deletion described above. A 2-bp deletion in CDKN2A was not detected, but was found in sequence reads when a targeted search was made.