General features of cancer genes

癌症基因的一般特征

Mutations in at least 291 human genes are causally implicated in oncogenesis (see supplementary information S1 (table)). All encode proteins. No mutated RNA genes have yet been shown to be involved in oncogenesis. Assuming that there are approximately 25,000 coding genes in the human genome - as indicated on the Ensembl Human Genome Browser web page (see online links box) - it seems that mutations in more than 1% genes are so far known to be involved in cancer pathogenesis.

Most were identified by positional cloning without any previous hypothesis of biological function. By this strategy, the cancer gene is first localized to a small part of the genome and, subsequently, genes within the delimited interval are screened for mutations. The primary positional clues have been diverse and include chromosomal rearrangements that are visible in metaphase spreads of cancer cells, DNA copy-number changes in cancer cells, (which are detected by various molecular approaches) and, for cancer susceptibility genes, genetic-linkage analyses of families with several cases of cancer.

A small number of cancer genes have been detected through biological assays. Most notable among these has been the NIH-3T3 transformation assay, in which total human DNA is introduced into a line of mouse fibroblasts, and cells that incorporate certain classes of mutated human cancer genes acquire the transformed phenotype^{19, 20}. Mutations in the remainder of human cancer genes have been identified through analysis of plausible candidates based on known biological features of cancer cells. However, these constitute a small minority of known cancer genes.

Approximately 90% of cancer genes show somatic mutations and 20% show germline mutations. Both somatic and germline mutations have been reported in 10% of all cancer genes (Fig. 1). In general, the spectrum of neoplasms that are associated with germline mutations in a particular gene is similar to that reported with somatic mutations. There are, however, several notable exceptions to this rule. For example, somatic mutations in Tp53 are found in more than half of colorectal cancers, yet germline mutations do not apparently cause a predisposition to colorectal cancer^{21, 22}. Similarly, germline mutations in STK11 are associated with a predisposition to hamartomas of the gastrointestinal tract and to colorectal, pancreatic and ovarian neoplasms (the Peutz-Jegher syndrome)²³. However, somatic mutations of STK11 have only been found in lung adenocarcinomas, which are not usually considered to be components of the Peutz-Jegher syndrome²⁴. Several genes with germline mutations that cause cancer predisposition show very few, if any, somatic mutations in sporadic cancers of the same type, such as BRCA1 and BRCA2 in breast cancer^{25, 26}. The reasons for differences between the tumour spectrum that is associated with somatic mutations and the spectrum that is associated with germline mutations are generally unknown.

The most common class of somatic mutation that is registered in the cancer-gene census involves chromosomal translocations that result in a chimeric transcript or apposition of one gene to the regulatory regions of another gene - usually immunoglobulin or T-cell-receptor genes. This mutation type is common in leukaemias, lymphomas and mesenchymal tumours. However, several examples have now been reported among epithelial neoplasms, including thyroid papillary carcinoma (RET and NTRK1, both with several partners), thyroid follicular carcinoma (PAX8 and PPAR ), renal papillary carcinoma (PRCC and TFE3) and breast secretory carcinomas (ETV6 and NTRK3)^27-30. Because two genes are structurally rearranged in each chromosomal translocation, the number of translocated cancer genes, compared with other types of mutated cancer gene, is exaggerated in the census. Moreover, certain genes, such as MLL (mixed-lineage leukaemia), are highly promiscuous and form chimeric transcripts with a large number of partners³¹. As a consequence of this ability to form chimeric genes with more than one partner, 'networks' of translocation partners can be discerned. For example, MLL can form a chimeric cancer gene by chromosomal translocation with CREB-binding protein (CREBBP; also known as CBP). In addition to MLL, CBP can also form a chimeric cancer gene with RUNXBP2 (also known as ZNF220). RUNXBP2, in turn, can form a chimeric cancer gene with EP300, and this gene can from a chimeric cancer gene with MLL.

For some genes, several different types of mutations have been associated with cancer. Notably, many of the different classes of mutation are found in most of the recessive cancer genes, because the result is usually inactivation of the encoded protein, and there are many mutation types that can achieve this end. However, for dominantly acting cancer genes, in which the encoded protein is usually activated, the patterns of mutation are much more restricted, so only one class of mutation is usually found.

Indeed, if more than one class of mutation occurs in a dominantly acting cancer gene, each type of mutation might be associated with a particular type of cancer. For example, two classes of mutation result in constitutive activation of the RET kinase - chromosomal translocations and base substitutions that lead to missense amino-acid changes. Rearrangements of RET are found in papillary thyroid carcinoma, whereas base sub-stitutions are found in medullary thyroid cancer³². Occasionally, a single allele of a cancer gene might require more than one mutation for full biological effect. For example, in-frame deletions within the extracellular domain of EGFR are common in malignant gliomas³³. Usually, the rearranged EGFR allele is also amplified. Similar increases in copy number occur on alleles mutated by base substitutions, for example MET in renal papillary carcinoma³⁴.

More than 70% of cancer genes with somatic mutations in the census are associated with leukaemias, lymphomas and mesenchymal tumours, even though these account for less than 10% of human cancer incidence. Why have so many more genes been associated with these relatively rare tumour types than with the more common epithelial cancers? Some of this imbalance is attributable to the number of genes that are subject to chromosomal translocations in leukaemias, lymphomas and mesenchymal tumours, with the attendant double counting of genes (see above). Another explanation is that it might have been easier, in the past, to identify cancer genes by studying leukaemias, lymphomas and mesenchymal neoplasms than by studying epithelial cancers. If this interpretation is correct, it indicates that many more cancer genes remain to be identified in association with common epithelial cancers. Alternatively, there might be fundamental biological differences between the group of leukaemias, lymphomas and mesenchymal tumours and the group of common epithelial neoplasms, such that the repertoire of mutated genes that is necessary to generate common epithelial neoplasms is more restricted.

90% of somatic mutations in cancer genes are dominant at the cellular level. Again, this is predominantly determined by the frequency of chromosomal translocations in leukaemias, lymphomas and mesenchymal tumours, almost all of which act in this way. If cancer genes that contribute to oncogenesis by chromosomal translocation are excluded, equal numbers of somatically mutated cancer genes are dominant and recessive at the cellular level. By contrast, 90% of germline mutations that result in cancer predisposition are recessive at the cellular level, presumably because many cancer-causing mutations that might act in a dominant fashion would cause embryonic lethality.

There is at least one example of a cancer gene that is able to act through both dominant and recessive mechanisms. PRKAR1A - a regulatory subunit of protein kinase A - is a translocation partner with RET. This translocation activates the RET kinase in papillary carcinoma of the thyroid³⁵. By contrast, germline mutations that inactivate PRKAR1A are associated with predisposition to Carney complex, a rare syndrome that is characterized by endocrine tumours and MYXOMA of the heart³⁶.