Fanconi anaemia genes and susceptibility to cancer

Abstract

Fanconi anaemia (FA) is a rare recessive disorder associated with chromosomal fragility, aplastic anaemia, congenital abnormalities and a high risk of cancer, including acute myeloid leukaemia and squamous cell carcinomas. The identification of 11 different FA genes has revealed a complex web of interacting proteins that are involved in the recognition or repair of DNA interstrand crosslinks and perhaps other forms of DNA damage. Bi-allelic mutations in BRCA2 are associated with a rare and highly cancer-prone form of FA, and the DNA helicase BRIP1 (formerly BACH1) is mutated in FA group J. There is little convincing evidence that FA heterozygotes are at increased risk of cancer, but larger studies are needed to address the possibility of modest risk effects. Somatic inactivation of the FA pathway by mutation or epigenetic silencing has been observed in several different types of sporadic cancer, and this may have important implications for targeted chemotherapy. Inhibition of this pathway represents a possible route to sensitization of tumours to DNA crosslinking drugs such as cisplatin.

The Fanconi anaemia syndrome

Fanconi anaemia (FA) is an inherited disorder associated with progressive aplastic anaemia, multiple congenital abnormalities and predisposition to malignancies including leukaemia and solid tumours (Fanconi, 1967). The developmental abnormalities include radial aplasia, hyperpigmentation of the skin, growth retardation, microphthalmia and malformation of the kidneys. The disorder generally presents as aplastic anaemia between the ages of 5–10 years, but the diagnosis may be made much earlier if characteristic developmental abnormalities are present or if there is a family history, or much later if the haematological symptoms are very mild. FA is inherited mainly as an autosomal recessive trait, but is genetically heterogeneous, with multiple complementation groups that include an X-linked form (Meetei et al., 2004). It is a rare disease with an incidence of 1 in 200 000–400 000 live births (Joenje and Patel, 2001), and a heterozygote frequency of around 1 in 250. However, it is more common in some populations, with carrier frequencies of about 1 in 90 reported in Ashkenazi Jews and the Afrikaans-speaking population of South Africa (Rosendorff et al., 1987; Verlander et al., 1995). At the cellular level, FA is considered to be a chromosomal fragility syndrome. Cells from FA patients are hypersensitive to DNA interstrand crosslinking agents such as mitomycin C and diepoxybutane (Schroeder et al., 1964), and the increased chromosomal breakage elicited by these agents forms the basis for a laboratory diagnostic test. Clinical, genetic and functional aspects of FA have been the subject of detailed reviews (Joenje and Patel, 2001; Tischkowitz and Dokal, 2004; Thompson et al., 2005; Taniguchi and D’Andrea, 2006). The main focus of this review is the association of mutations in FA genes with susceptibility to cancer.

Genetics of FA

Complementation analysis of cell lines from different FA patients has led to the description of at least 12 groups, named FA-A, B, C, D1, D2, E, F, G, I, J, L and M, with the corresponding genes named as FANCA-FANCM (Table 1). The FA genes were cloned using a variety of strategies including functional complementation of FA cell lines, positional cloning, candidate gene screening and the identification of novel proteins from immunoprecipitates of other FA proteins. The FA proteins form nuclear multi-protein complexes referred to collectively as the FA pathway (Taniguchi and D’Andrea, 2006). Interest in this pathway was stimulated by the discovery that the gene for FANCD1 is BRCA2 (Howlett et al., 2002); mono-allelic mutations in BRCA2 cause susceptibility to breast and other cancers (this issue, 2006), whereas bi-allelic mutations cause Fanconi anaemia. Most FA genes have a wide spectrum of mutations in FA patients that include deletions, frameshifts, stop codons, splice-site mutations and missense mutations (Joenje and Patel, 2001). FA-A is the most common group world-wide, but there are population-specific differences as a result of founder mutations. Thus, for example, most Ashkenazi Jewish patients are from group C and are homozygous for a splice-site mutation (IVS4+4A>T) in FANCC (Whitney et al., 1993), and other examples have been identified in Afrikaners, Spanish gypsies and sub-Saharan Africans (Tipping et al., 2001; Callen et al., 2005; Morgan et al., 2006).

Table 1 Fanconi anaemia genes and proteins

There is some evidence for a correlation between either complementation group or mutation type and the severity of the clinical phenotype. The IVS4+4A>T mutation is associated with a severe FA phenotype in Ashkenazi Jews (Yamashita et al., 1996; Gillio et al., 1997), but the same mutation produces a much milder phenotype in Japanese patients, suggesting that modifier genes or environmental factors contribute to disease severity (Futaki et al., 2000). A comparison of the more common FA groups A, C and G found a higher incidence of acute myeloid leukaemia (AML) or myelodysplastic syndrome (MDS) in FA-G (Faivre et al., 2000), and FANCD1/BRCA2 mutations produce a particularly cancer-prone phenotype (see below). Variations in clinical severity are consistent with the finding that different patient-derived mutations in the FANCA gene have variable effects on the function of the FA pathway (Adachi et al., 2002). The clinical phenotype may also be modified by somatic mosaicism in which a functional FANC allele is generated by intragenic recombination or gene conversion in compound heterozygotes or, remarkably, in mutation homozygotes by a compensatory secondary mutation in cis which restored the function of the mutant allele (Waisfisz et al., 1999).

Function of the FA pathway

The association of the cellular phenotype of FA with increased chromosome breakage and hypersensitivity to DNA interstrand crosslinkers and of the clinical phenotype with a strong predisposition to cancer led to the assumption that the FA proteins had some role in the recognition or repair of DNA damage. However, the identification of the first FA genes provided few functional insights since their encoded proteins had no obvious connections with known protein families. Several did, however, contain nuclear localization signals, consistent with a role in DNA repair (Table 1). The identification of the FANCD1 gene as BRCA2 provided the first direct link between FA proteins and DNA repair, given this protein's involvement in homologous recombination and the regulation of the RAD51 recombinase. Some of the newer FA genes have been more informative, with FANCL having a PHD finger motif that is associated with ubiquitin ligase activity (Meetei et al., 2003), FANCJ (BRIP1) being a DNA-dependent ATPase and 5′–3′ helicase (Levitus et al., 2005; Levran et al., 2005), and FANCM being related to the archeal protein Hef, which has functional endonuclease and helicase domains (Meetei et al., 2005).

A group of at least eight FA proteins (A, B, C, E, F, G, L and M) form a nuclear complex (the FA core complex) that is required for the monoubiquitination of the FANCD2 protein. The ubiquitinated form of FANCD2 is translocated to sites of DNA damage in chromatin, where it is found in a complex with BRCA2/FANCD1 and FANCE (Wang et al., 2004), but its function is currently unknown. The FANCG protein also appears to function both inside and outside of the FA core complex, since it is found in a complex with BRCA2 and the homologous recombination repair protein XRCC3 that is independent of core complex formation or the ubiquitination of FANCD2 (Hussain et al., 2006). There appears to be extensive cross-talk between the FA pathway and other DNA repair pathways, since FA proteins have been found in complexes with BLM, RPA, TopoIIIα, NBS1, BRCA1 and ATR (reviewed by Surralles et al. (2004) and Taniguchi and D’Andrea (2006)). A broader function has been proposed for the FA proteins in monitoring oxidative stress and in the stabilization of stalled replication forks that arise during recombination and repair (Thompson et al., 2005). A current model of the FA pathway is shown in Figure 1. Although we now have more information about the nature and interactions of these proteins, it is clear that their precise roles in DNA repair have not yet been defined.

Figure 1

A model of the Fanconi Anaemia pathway. A large nuclear complex of at least eight FA proteins are required for the monoubiquitination of FANCD2. The FANCI protein, which has not yet been identified, does not appear to be needed for complex formation but is required for the modification of FANCD2. The Fanconi anaemia-associated protein FAAP100 has also not been identified, and may be encoded by a novel FA gene. The active ubiquitinated form of FANCD2 associates with DNA repair proteins such as BRCA2 in chromatin at sites of DNA damage, but its precise role is not known.

Other functions of individual FA proteins that are not directly related to DNA repair have been proposed, such as prevention of oxidative damage by reactive oxygen species (Cumming et al., 2001), and in the regulation of apoptosis and signalling in response to cytokines and growth factors (Fagerlie et al., 2001). A large number of interactions between FA and other proteins have been described (Reuter et al., 2003), but the functional significance of many of these remains to be established.

Cancer in FA homozygotes

Cancer incidence

FA patients have a high risk of leukaemia and solid tumours. A literature review of over 1300 reported cases from 1927 to 2001 found that 9% had leukaemia, mainly AML, 7% had MDS, 5% had solid tumours and 3% had liver tumours (Alter, 2003). The distribution of the solid tumours was very unusual compared to the general population, since more than half were of the head and neck, oesophagus or vulva. The liver tumours were generally not malignant, and were thought likely to be related to prolonged androgen therapy. A retrospective cohort study of 145 FA patients showed that nine had developed AML and 14 had solid tumours, most of which were of the head and neck, oesophagus, vulva or cervix (Rosenberg et al., 2003). The ratio of observed to expected cancers was 50 for all cancers, 48 for solid tumours and 785 for leukaemia. The cumulative incidence to age 48 was 10% for leukaemia and 29% for a solid tumour. A larger study of 754 patients in the International Fanconi Anemia Registry (IFAR) found neoplasms in 23%, with haematological neoplasms (mainly AML or MDS) in 15.9% and non-haematological tumours in 10.5%, with about 3% of patients having more than one neoplasm during their lifetime (Kutler et al., 2003b). The most common non-haematological tumours were squamous cell carcinomas (SCC) of the head and neck, vulva and cervix. Importantly, only three of the 19 patients with a head and neck SCC had a history of alcohol or tobacco abuse, which are the major risk factors for these tumours. The cumulative incidence of haematological and non-haematological neoplasms was 33 and 28%, respectively by age 40. The standardized incidence ratio for head and neck SCC (HNSCC) in the IFAR registry was 500, with a median patient age of 31 years (Kutler et al., 2003a).These patients had poor tolerance for radiotherapy and chemotherapy, and the tumours were aggressive. Examples have been reported of young cases of apparently sporadic HNSCC with no known risk factors, who had a severe reaction to radiotherapy and who were later diagnosed as FA (Alter et al., 2005). There is also evidence that conditioning for stem cell transplants in FA patients is associated with an increased risk of SCC of the head, neck and oesophagus (Rosenberg et al., 2005).

Cancers in FA-D1

It is clear that FA is generally associated with strong predisposition to leukaemia and solid tumours, especially of the head and neck. However, clinical and molecular characterization of FA-D1 families with bi-allelic mutations in BRCA2 has revealed an even stronger cancer-prone phenotype in the 16 confirmed D1 kindreds reported to date (Howlett et al., 2002; Offit et al., 2003; Hirsch et al., 2004; Wagner et al., 2004; Reid et al., 2005). Early onset leukaemia was common, with a median onset at 2.2 years of age compared to 13.4 years for all other FA patients in the IFAR registry (Wagner et al., 2004); this was mainly AML, but there were also cases of T- and B-cell acute lymphoblastic leukaemia. There were nine cases with brain tumours (mainly medulloblastomas), and five with Wilms tumours (summarized in Reid et al. (2005)). It was proposed initially that FA-D1 would occur in individuals who had at least one partially functional BRCA2 allele (Howlett et al., 2002). This would be consistent with the fact that complete knockout of brca2 is an early embryonic lethal in the mouse (Ludwig et al., 1997; Sharan et al., 1997), whereas some mice homozygous for a C-terminal truncating brca2 mutation are viable but succumb to lymphomas (Connor et al., 1997; Friedman et al., 1998). The frequency of BRCA2 mutation carriers in the general population is estimated to be around 1 in 1000 (Thompson and Easton, 2004), which corresponds to a mutant allele frequency of 0.0005 and a predicted homozygote frequency in the absence of selection of 1 in 4 million. As the incidence of FA in most populations is around 1 in 250,000 and FA-D1 accounts for about 3.3% of all FA cases (Levitus et al., 2004), this corresponds to an incidence of FA-D1 of 0.53 in 4 million, which is about half of the predicted incidence. Thus, it is possible that only a subset of mutant BRCA2 alleles is compatible with survival to term in humans. Perinatal lethality has been observed in fancd2 and fancc knockout mice (Houghtaling et al., 2003; Carreau, 2004), so this may be a general feature of the FA pathway. However, an assessment of the location of BRCA2 mutations in FA-D1 patients in comparison with BRCA2 carriers does not support a bias towards carboxy-terminal mutations in FA-D1 (Reid et al., 2005). Also, the nature of the mutations in many of the reported FA-D1 cases would predict early truncation of the BRCA2 protein expressed from both alleles, and its absence has been confirmed in some cases by Western blot analysis (Hirsch et al., 2004).

Molecular basis and tissue specificity

The association of haematopoietic stem cell failure and a high risk of cancer with a chromosomal instability syndrome is not surprising since excessive chromosome breakage might be expected to lead to unrepaired DNA damage and apoptosis, or to mutations which could confer a selective growth advantage to a progenitor cell. Why cells of myeloid origin are particularly affected, leading to MDS or AML, is not known; perhaps myeloid precursors are particularly susceptible to forms of DNA damage that are recognized or repaired by the FA pathway. The unusual spectrum of solid tumours, with a preponderance of squamous cell carcinomas of the oral cavity and vulva or cervix is also unexplained. Evidence both for and against an involvement of human papilloma virus (HPV) in SCCs from FA patients has been presented (Kutler et al., 2003c; van Zeeburg et al., 2004). If even a significant minority of these tumours are HPV positive, there will be a strong case for prophylactic HPV vaccination in FA patients. Another possible route to carcinogenesis is an interaction of the environmental toxins to which the mucous membranes of the oral cavity and genital areas are exposed with a constitutional deficiency of the FA patient to detoxify such agents or to repair DNA damage consequent to the exposure. In a clinical context, it will be important to establish what proportion of young patients who have HNSCC in the absence of major risk factors are in fact undiagnosed FA, since this will have a major impact on their clinical management. The highly cancer-prone phenotype of FA-D1 patients is not unexpected given the central role occupied by BRCA2 in homologous recombination repair. However, the high incidence of brain tumours in these patients is not characteristic of FA patients in general, or of BRCA2 mutation carriers, and remains to be explained, as does the association with familial Wilms tumour (Reid et al., 2005).

Cancer in FA heterozygotes

Epidemiology

The possibility that FA heterozygotes might be at increased risk of cancer was first suggested by Swift (1971), and it is a reasonable hypothesis that a modest reduction in DNA repair efficiency or loss of a second FA allele in some tissues might lead to the development of a tumour. Support for this hypothesis comes from the now well established increase in risk of breast cancer in heterozygotes for mutations in the ATM gene (reviewed in this issue, 2006). There is also some experimental support for differences in FA heterozygotes. The mean index of DEB-induced chromosome breakage was higher in obligate FA heterozygotes relative to controls (0.13±0.06 versus 0.07±0.09), although there was overlap between these groups (Pearson et al., 2001). Also, comet assays of FA heterozygotes showed an increased release of fragmented DNA after exposure to X-irradiation compared to controls (Djuzenova et al., 2001). However, Swift's original findings of an increased cancer risk in FA relatives were not confirmed in a later and larger study of relatives in 25 FA families (Swift et al., 1980). There were 48 deaths from cancer among blood relatives of the index cases, with 56.2 expected, and no excess of cancers were observed in obligate heterozygotes. A smaller study of the relatives of nine FA patients did not detect an increase in cancer risk (Potter et al., 1983). A recent study of 36 British families (Hodgson et al., 2004) found no increase in cancer in relatives of FA patients, with 55 cancers observed and 56.95 expected (O/E 0.97, 95% CI 0.71–1.23). Thus there is no current evidence from epidemiological studies that FA heterozygotes are at increased risk of cancer. However, the existing studies are small, and have insufficient power to detect a modest increase in cancer risk. The increase in the risk of breast and other cancers as a result of heterozygosity for BRCA2 mutations in the FA-D1 group would be difficult to detect in such studies, as this represents only about 3% of all FA families. A multi-centre, international collaboration is needed, and this could be strengthened by a study design that involved FA families with known mutations that would allow molecular testing for carrier status in relatives with a diagnosis of cancer. This would have the added advantage of assessing whether cancer risk was related to a subset of complementation groups and specific mutations.

Leukaemia

Another approach to this problem has been to look for germline FA gene mutations in cancer patients who do not have FA in the expectation that a subset of cancers in the general population may be associated with mutations in the FA pathway. A screen for FANCC mutations in remission blood samples from cases of sporadic childhood leukaemias did not detect any pathogenic mutations in 97 cases of AML or 91 of ALL, but there was a marginally significant increase in the frequency of a single nucleotide polymorphism (S26F) in the AML group (Barber et al., 2003). This group also analysed the FANCG gene in 107 children with sporadic AML and found a pathogenic splice site mutation in one case together with an unclassified variant (S588F), although whether the second mutation was in cis or in trans could not be determined (Meyer et al., 2006). Mutation screens of FANCA in adult sporadic cases of AML detected a low frequency of both missense mutations (Condie et al., 2002) and heterozygous exonic deletions (Tischkowitz et al., 2004), but remission samples were not available to determine whether these were germline or somatic mutations. Taken together, these studies provide little support for a contribution of inherited FA gene mutations to susceptibility to AML, but a comprehensive screen for mutations in all 11 of the known FA genes would be required to resolve this question.

Breast cancer

The major role of BRCA2/FANCD1 in susceptibility to breast and other cancers such as cancer of the prostate and pancreas is well established (this issue, 2006). There have been several studies of the other known FA genes in relation to breast cancer susceptibility. Two different approaches have been used; one has been to look for highly penetrant FA gene mutations in families with multiple cases of breast cancer, and the other to analyse single nucleotide polymorphisms (SNPs) in FA genes for association with sporadic breast cancer. A comprehensive mutation screen of the FANCA, FANCC, FANCD2, FANCE, FANCF and FANCG genes was carried out in 88 familial breast cancer families who were negative for BRCA1/2 mutations (Seal et al., 2003). No clear pathogenic mutations were detected, but two missense variants were found (one in FANCA and one in FANCE) that were absent in controls and segregated with breast cancer. The authors concluded that FA gene mutations, other than in BRCA2, are unlikely to make a significant contribution to the risk of familial breast cancer. The FANCD2 gene was screened for mutations in an Australian cohort of 30 non-BRCA1/2 familial breast cancer families, but no pathogenic mutations were detected (Lewis et al., 2005). An association study of seven SNPs in the FANCA, FANCL and FANCD2 genes was carried out to determine whether any of these SNPs conferred susceptibility to breast cancer in the Spanish population (Barroso et al., 2006). A marginally significant association of a synonymous SNP in FANCD2 with breast cancer was observed, but independent replication of such findings is essential to validate their significance.

A gene that is variously known as BACH1/BRIP1/FANCJ and is located on chromosome 17q22-24 has been screened extensively for involvement in breast cancer susceptibility. This was originally identified as a protein that binds to the BRCT repeats of BRCA1, and was called BACH1 (Cantor et al., 2001). Its name was later changed to BRIP1 (BRCA1 interacting protein 1) since, unfortunately, another gene that encodes a leucine zipper transcription factor and is located on chromosome 21q22.11 was also called BACH1. BRIP1 was later identified as the gene that is mutated in FA-J patients, so is also called FANCJ. The original report on BRIP1/BACH1 found two missense mutations (P47A and M299I) in early onset familial breast cancer cases (Cantor et al., 2001), and the P47A mutation was later shown to abrogate the ATPase and helicase activity of the protein (Cantor et al., 2004).

Several subsequent studies screened the BRIP1 coding sequence in large numbers of BRCA1/2 negative familial breast cancer families from populations in Northern Europe (Luo et al., 2002; Karppinen et al., 2003; Vahteristo et al., 2006), North America (Rutter et al., 2003) and Australia (Lewis et al., 2005). This resulted in the detection of one frameshift mutation and several missense mutations, none of which segregated with breast cancer in the families. The conclusion from these studies was that germline pathogenic mutations in BRIP1 are extremely rare or absent in familial breast cancer.

SNPs in BRIP1 were also tested for association with sporadic breast cancer. A study of 19 SNPs in eight DNA repair genes found a marginally significant association of homozygosity for a non-synonymous SNP, S919P, in BRIP1 in breast cancer up to age 50 in a North American population, but the association was not significant when all cases up to age 70 were included (Sigurdson et al., 2004). The S919P SNP was not associated with breast cancer risk in two subsequent studies (Garcia-Closas et al., 2006; Vahteristo et al., 2006).

Pancreatic cancer

The involvement of BRCA2 in the FA pathway and its known association with an increased risk of pancreatic cancer prompted mutation screening of other FA genes in sporadic and familial pancreatic cancer. Mutation analysis of the FANCC and FANCG genes in 44 pancreatic tumours from young onset cases selected for loss of heterozygosity (LOH) at these genes detected the missense mutation D195V in FANCC (van der Heijden et al., 2003); the mutation was also present in the patient's germline DNA, but its effect on FANCC protein function is not known (Verlander et al., 1994). Some of the subsequent screens of germline DNA for mutations FANCC, FANCG and FANCA in familial pancreatic cancer did not detect any pathogenic mutations (Rogers et al., 2004a, 2004b). However, mutation screening of FANCC and FANCG in germline DNA from 421 unselected pancreatic cases identified heterozygous truncating mutations in FANCC in two young onset cases, and none in FANCG (Couch et al., 2005). The tumours from these two cases had LOH of the wild-type allele, and no truncating FANCC mutations were detected in 658 controls. As FANCC, FANCG and BRCA2/FANCD1 mutations together account for only about 21% of all cases of FA (Levitus et al., 2004), it is possible that constitutional heterozygosity for FA gene mutations may contribute to susceptibility to pancreatic cancer in a small but significant minority of young onset cases.

Somatic inactivation of the FA pathway

Acute myeloid leukaemia

Several studies have examined the question of whether somatic mutations in the FA pathway might contribute to the pathogenesis of sporadic cancers (Lyakhovich and Surralles, 2005). Reduced levels of FA proteins or absence of one or more FA proteins and their complexes was observed in five of 10 AML cell lines and 11 of 15 primary AML samples (Xie et al., 2000). One of the cell lines (CHRF-288) was hypersensitive to MMC and lacked the FANCF protein. A follow-up study of this cell line showed that the MMC sensitivity could be corrected by transfection with the FANCF gene, and that FANCF expression in these cells was silenced not by mutation, but by extensive methylation of its promoter region (Tischkowitz et al., 2003). Hypersensitivity to MMC and inactivation of the FA pathway with loss of FANCD2 ubiquitination and reduced amounts of nuclear FA proteins was reported in another AML cell line (Lensch et al., 2003). In this case, MMC sensitivity was partially corrected by transduction with the FANCA gene, but no mutation or methylation of FANCA was detected, so the basis of this acquired FANCA dysfunction was unresolved. The deletion of all or part of the FANCA gene in four of 101 adult sporadic AMLs was associated with a reduction in FANCA expression, but no pathogenic mutations were observed in the intact FANCA allele (Tischkowitz et al., 2004).

Other cancers

Interest in the possible involvement of the FA pathway in cancers other than AML was stimulated by the finding of cisplatin sensitivity and a lack of ubiquitinated FANCD2 in two of 25 ovarian cancer cell lines, which was corrected by transfection with FANCF (Taniguchi et al., 2003). The lack of FANCF expression was correlated with hypermethylation of the FANCF CpG island, and FANCF methylation was also observed in four of 19 primary ovarian tumours. A screen for FANCF hypermethylation using a methylation-specific PCR of bisulphite-modified DNA from primary tumours detected methylation in 13 of 89 head and neck SCCs and 22 of 158 non-small-cell lung carcinomas (Marsit et al., 2004), but the effect of this on the level of FANCF expression and the function of the FA pathway was not examined. FANCF hypermethylation was also observed in 27 of 91 (30%) primary cervical cancers, in three of nine cervical cancer cell lines and in 0 of 20 samples of normal cervical epithelia (Narayan et al., 2004); six of the nine cell lines had reduced expression of FANCF mRNA, and this correlated with hypersensitivity to MMC. Two pancreatic cancer cell lines have been described that are MMC sensitive and defective in FANCD2 monoubiquitination (van der Heijden et al., 2004). One line had acquired a homozygous deletion of FANCC and the other a truncating mutation in FANCG, and in both cases the cellular FA phenotype could be corrected by transduction with the relevant wild-type FA gene.

The association of pathogenic mutations or epigenetic silencing of FA genes with MMC sensitivity and lack of monoubiquitinated FANCD2 in cell lines from some AMLs and ovarian and pancreatic cancers, coupled with restoration of a functional FA pathway upon transduction with the relevant FA gene, is quite compelling evidence that inactivation of this pathway has a role in the development of a subset of these types of tumour. However, since these well-characterized examples of loss of the FA pathway are in transformed cell lines, they could be late events in tumour progression. Detailed examination of the function of this pathway in primary tumours will be required to resolve the timing of these events.

A target for chemotherapy?

As DNA interstrand crosslinking (ICL) agents are an important class of chemotherapeutics, it is possible that selective inhibition of the FA pathway in tumours may sensitize them to drugs such as cisplatin and melphalan. Transfection of a dominant-negative form of FANCA into cancer cell lines disrupted FANCD2 monoubiquitination and produced a two- to threefold sensitization of the cells to cisplatin (Ferrer et al., 2004). Melphalan-resistant myeloma cells had reduced ICL formation and enhanced ICL repair, and small interfering RNA (siRNA) knock down of FANCF in melphalan-resistant cells partially reversed their drug resistance (Chen et al., 2005). A cell-based screening strategy has been developed which identified several small-molecule inhibitors of the FA pathway that may have utility in sensitizing cancer cells to ICL-based therapy (Chirnomas et al., 2006). Finally, since BRCA2-deficient cells and tumours are acutely sensitive to inhibitors of poly(ADP-ribose) polymerase (Bryant et al., 2005; Farmer et al., 2005), it would be interesting to see whether this property is shared by tumours that are deficient in other components of the FA pathway.

Conclusions

There has been major progress in unravelling the complexities of the FA pathway in the past few years, with the identification of new FA genes, the discovery that FANCD1 is BRCA2, and that the FANCJ and FANCM proteins have DNA helicase or DNA translocase activity. However, much remains to be done to define the functions of these proteins and to understand how they contribute to DNA recombination and repair. The use of proteomics to dissect the components of FA protein complexes (Meetei et al., 2005) and model organisms such as chicken DT40 cells to knockout combinations of FA genes (Yamamoto et al., 2003; Bridge et al., 2005; Mosedale et al., 2005) are beginning to yield valuable insights into how these proteins combine to recognize or repair DNA damage.

A scheme for the involvement of germline and somatic mutations of the FA genes in the development of cancer is shown in Figure 2. In homozygotes the lack of a functional FA pathway will lead to mutations or chromosomal breakage and rearrangements in some cells. The cellular response in most cases is likely to be apoptosis, but some cells will acquire a proliferative advantage and develop into a tumour. The unusual spectrum of solid tumours in these patients is unexplained. Heterozygotes for a germline FA mutation would require a somatic mutation or loss of heterozygosity involving the other allele of the same FA gene to disrupt the pathway and generate genomic instability. At present, it seems unlikely from the limited data available that heterozygosity for mutations in any of the FA genes except BRCA2 confers strong susceptibility to cancer, but a large collaborative epidemiological and molecular study is needed to examine the possibility that mutation carriers have a modest increase in cancer risk. Extensive mutation screening of the full complement of FA genes in germline DNA from young onset cases of sporadic cancers such as AML and squamous cell carcinomas of the head, neck and cervix, although laborious, will also help to address this question. The majority of the population will have two functional copies of their FA genes, and would therefore require somatic mutation of both alleles of the same FA gene to inactivate the pathway. An important exception to this is the FANCB gene which is located on the X chromosome (Meetei et al., 2004). All males are effectively FA carriers, and only one somatic mutation will therefore be required for loss of function of the FA pathway, and in females one copy would be subject to X-inactivation. The biological significance of somatic mutations of FA genes and the silencing of FANCF by hypermethylation in some sporadic cancers needs further investigation. If a broader role for inactivation of the FA pathway in cancer can be established, it may have important implications for future therapeutic strategies.

Figure 2

Fanconi anaemia genes in cancer. FA patients (FA^−/−) have bi-allelic FA gene mutations in their germline, which results in genomic instability and additional mutations in their somatic cells and the development of leukaemia and solid tumours. The BRCA2^−/− genotype strongly predisposes to cancer. In FA heterozygotes (FA^+/−) a somatic mutation in the other allele of the corresponding FA gene would inactivate the FA pathway and could lead to apoptosis or to the acquisition of further mutations and a proliferative advantage. Cells with no FA germline mutations (FA^+/+) would require multiple mutations or epigenetic events to inactivate the FA pathway.

The advances of the past few years have moved the Fanconi anaemia syndrome from the study of a relatively obscure chromosome breakage disorder into the mainstream of DNA repair and cancer research, and have the potential to yield novel insights into the cellular response to DNA damage and the pathogenesis of some forms of human cancer.