Attempts to understand the genetics of cancer have revealed, in considerable depth, mechanisms of tumorigenesis and drug resistance. Rapid technological advances will further improve our ability to study genes and their impact on cancer and its treatment. However, the interpretation of this story requires ever-increasing care, with important implications for clinical trial design.

The genetics underlying cancers are relevant for both inherited and non-inherited (sporadic) cancer. Inherited predisposition can be proven (as in the case of a person proven to have inherited a pathological mutation in a cancer predisposition gene), inferred (from a pedigree consistent with a dominantly inherited mutation in a high-penetrance predisposition gene, even if such a gene has not yet been documented in that family) or surmised, in a group of families in which the incidence of two related cancer types, for example, seems to be higher than predicted for that general population.

On the other hand, a ‘sporadic’ case of cancer is diagnosed in a family with no other relevant family history. However, we know that in some families one case may be the harbinger of a subsequent family constellation of related cancers, for which an inherited family-specific mutation in a relevant cancer predisposition gene may eventually be identified. For example:

1. a young case of osteosarcoma, suggestive of a ‘Li Fraumeni’ family in which a mutation in the guardian of the genome, p53 may be present, if subsequent cases of breast cancer are diagnosed^1,2;
2. a woman developing triple-negative breast cancer, aged less than 35 years at diagnosis³, followed by other breast and/or ovarian cancer cases; or
3. a woman with high-grade serous ovarian cancer, aged less than 50 years at diagnosis⁴, followed by other breast and/or ovarian cancer cases.

High-penetrance founder mutations, such as in BRCA1/2, occurring in either the Ashkenazi Jewish⁵ or Icelandic⁶ populations, involve a higher than expected rate of carriage of the relevant mutations compared with the general population. More confusingly, even with mathematical modelling, are population effects of what are presumed to be moderate-strength or low-risk cancer predisposition alleles, the latter having been identified in genome-wide association studies.⁷ At present, the impact of mutations in moderate-risk predisposition genes (CHEK2, ATM, BRIP1, PALB2 and NBS1)⁸ or low-risk alleles are difficult to interpret. It is even more difficult to accurately estimate, at a population level, the precise impact of such genes on the rate of cancer in a given population. It is thought these alleles may act in concert, for example, with three or more genes potentially predisposing to a particular type/subtype/s of cancer (polygenic risk).⁸ However, to accurately determine the effects of potentially inheriting a set of such alleles, for one woman, or one family, remains beyond our reach. It is worth noting, therefore, that proving that a particular case of cancer is indeed ‘sporadic’ could be quite difficult.

Much has been learned from studying inherited cancer predisposition genes. The way in which our knowledge of cancer genetics has developed over the last three decades has illustrated many salient points important for our general understanding of disease today. We have stumbled across important genes and later learned their function, by following the flags of rare, obviously inherited cancers. In doing so, we have learned much that is now of critical relevance for larger populations of patients who have cancer that is deemed to be sporadic in origin. This is exemplified by the discovery of genes underlying some breast and ovarian cancers, as described below.

Of great importance are the functions associated with the list of inherited, in many cases rare, cancer predisposition genes.⁹ Together these functions highlight most of the important ‘machinery-gone-wrong’ malfunctions involved in the genesis, propagation and, often, drug resistance underlying the lethality of sporadic cancer. Historically, the function of many of these genes was elucidated following a lengthy period of basic research, made possible by the identification of the gene underlying a striking inherited familial phenotype, for example that of the Retinoblastoma gene (Rb)¹⁰, or of familial breast and ovarian cancer (BRCA1)¹¹ or familial female and male breast cancer (BRCA2).¹²

Many genes important enough to cause a high-penetrant cancer syndrome are tumour suppressor genes with crucial roles in regulation of the cell cycle, DNA repair and programmed cell death or apoptosis.9 Much rarer are examples of inherited oncogenes, such as the RET oncogene that causes the MEN2 syndrome, including tumours such as medullary thyroid cancer.¹³ What we do not understand well, is why some genes, involved in such critical processes, affect only certain tissue types. Most of the known familial cancers have ‘diagnostic’ specificity, for example: BRCA1 is associated with breast, ovarian and prostate cancers¹¹ (perhaps all hormonally linked); BRCA2 is associated with breast, ovarian (to a lesser extent than for BRCA1), prostate, melanoma and pancreatic cancer¹⁴; indeed, the guardian of the genome itself, p53, is associated predominantly only with soft-tissue sarcoma, breast cancer and haematologic malignancies¹⁵; PTEN is involved in Cowden Syndrome, involving thyroid cancer and breast cancer¹⁶; and the Lynch Syndrome genes (MLH1, MSH2, MSH6) are involved in bowel, uterine and ovarian cancer and cancers of the uroepithelial tract.¹⁷ Such tissue specificity is relatively strict and yet the molecular basis for it is largely unknown.

In order to illustrate many of the principles outlined above, the story begins with the analysis of the very well-documented large kindreds involving breast and ovarian cancer that resulted, in 1994, in the identification of BRCA1 (so-named because it was the first gene to be associated with inherited breast cancer).¹¹ It was thought the identification of BRCA2, ^{3, 4} and so on would rapidly follow and yet even BRCA2 remained elusive until pedigrees involving another discrete phenotypic feature were studied, that of the inclusion of male as well as female breast cancer.¹² In fact, BRCA3 and 4 remain as yet undiscovered, in the sense that, although we have identified additional breast cancer susceptibility genes associated with breast and/or ovarian cancer, such as ATM, PALB2 and RAD51C and D, each of these accounts for only a very small number of families and is thought to confer mostly moderate risk (higher risk in some families), leaving the bulk of families, especially, with ‘breast cancer only’ as yet unexplained in a genetic sense.

Following the discovery of these genes, much basic research ensued that rapidly defined essential roles for BRCA1¹⁸ and BRCA2¹⁹ in DNA repair, particularly in the high-fidelity pathway of homologous recombination (HR). The defect in DNA repair was present only in the cancer cell (which had obligate loss of the wildtype allele), but not in the rest of the patient’s body (in which the cells were heterozygote for the gene concerned and therefore retained normal protein and DNA function). It was logical to hypothesise that a defect in one important DNA repair pathway might be exploited by creating a ‘synthetic lethal’ setting, with the addition of a pharmacologically driven defect in a complementary DNA repair pathway.^20,21 This approach led to the development of potent inhibitors of PARP, important for Base Excision Repair (reviewed in²²). Indeed, results to date of PARP inhibitors in high-grade serous ovarian cancer are impressive, both in patients documented to carry a mutation in BRCA1 or BRCA2²³ and in patients with ‘sporadic’ high-grade serous ovarian cancer (reviewed in²²). It is likely that up to approximately 50 per cent of high-grade serous ovarian cancers have a ‘BRCAness-like’ defect, due to a variety of molecular mistakes in wiring of the DNA repair machinery.²⁴ A functional test or easy molecular test to identify this phenotype is not yet available, but would be of great utility.

The likelihood that we could improve the targeting of PARP inhibitors in ovarian cancer is suggested by two important recent observations:

1. 1. In ovarian cancers, established mutations in either BRCA1 or BRCA2 can, under pressure to evade treatment, with cisplatin or PARP inhibitor, undergo ‘mutation reversion’.^25-27 This means any cancer cell in which by chance the DNA is subsequently altered to result in the BRCA1 or BRCA2 mutation going back ‘in frame’ at the DNA level, will be able to survive treatment and clonally expand, causing drug resistance and tumour recurrence.
   2. Another source for resistance to PARP inhibitor treatment lies in the recent demonstration that a related DNA repair pathway, called non-homologous end joining (NHEJ), known as the ‘poor sister’ of DNA repair because it is not of high fidelity, is unexpectedly essential for the killing by PARP inhibitors in cells lacking DNA repair by HR.²⁸ Initially counterintuitive, it appears in the absence of HR, PARP normally prevents this poor-sister error-prone pathway from getting out of control. But, in the presence of a PARP inhibitor and the absence of functional HR, this NHEJ pathway does indeed spiral out of control, causing marked genomic instability that results in death of that cell. As such, any HR-deficient cell lacking intact machinery for NHEJ would be predicted to be resistant to PARP inhibitor therapy. It would seem logical that in this setting, PARP inhibitor therapy would be best avoided for that patient.

Understanding the molecular pathways that result in successful killing of cancer cells, by targeting a specific weakness in that cancer cell, is likely to drive the great therapeutic success stories of the next few decades. However, in order to take advantage of the current ‘genomics explosion’ we must ensure that clinical trial design evolves to enable tissue analysis both pre- and post-therapeutic exposure to putative targeted drugs. Only in this way will we collect tissue, both clinically and molecularly annotated, and – together with molecular pathologists and genomicists – be in a position to fast-track novel therapies with a sufficient understanding to ensure a high response rate and a low drug-resistance or drug-failure rate. In this way, the genetics of cancer has the potential to address the great unknowns that currently preclude effective cancer treatment, ultimately, resulting in improvement in treatment outcomes for our patients.