The Fanconi anaemia pathway orchestrates incisions at sites of crosslinked DNA


  • No conflicts of interest were declared.


Fanconi anaemia (FA) is a rare, autosomal recessive, genetically complex, DNA repair deficiency syndrome in man. Patients with FA exhibit a heterogeneous spectrum of clinical features. The most significant and consistent phenotypic characteristics are stem cell loss, causing progressive bone marrow failure and sterility, diverse developmental abnormalities and a profound predisposition to neoplasia. To date, 15 genes have been indentified, biallelic disruption of any one of which results in this clinically defined syndrome. It is now apparent that all 15 gene products act in a common process to maintain genome stability. At the molecular level, a fundamental defect in DNA repair underlies this complex phenotype. Cells derived from FA patients spontaneously accumulate broken chromosomes and exhibit a marked sensitivity to DNA-damaging chemotherapeutic agents. Despite complementation analysis defining many components of the FA DNA repair pathway, no direct link to DNA metabolism was established until recently. First, it is now evident that the FA pathway is required to make incisions at the site of damaged DNA. Second, a specific component of the FA pathway has been identified that regulates nucleases previously implicated in DNA interstrand crosslink repair. Taken together, these data provide genetic and biochemical evidence that the FA pathway is a bona fide DNA repair pathway that directly mediates DNA repair transactions, thereby elucidating the specific molecular defect in human Fanconi anaemia. Copyright © 2011 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Pathobiology of human Fanconi anaemia

The clinician Guido Fanconi first described Fanconi anaemia (FA) in 1927, soon after completion of his medical training. Three children succumbed to anaemia associated with microcephaly, skin hyperpigmentation and hypoplasia of the testis. He soon realized that this disease affected all blood lineages and was associated with developmental defects in many organs and a predisposition to cancer development 1, 2. This devastating illness affects approximately 1 in 360 000 live births, with patients having median survival of 20 years 3. They exhibit a wide range of clinical features, each of incomplete penetrance. The key clinical features observed in FA patients can be broadly divided into three categories, developmental abnormalities, stem cell defects and neoplasia.

Developmental abnormalities

FA patients range from being developmentally normal to having such severe abnormalities that the pregnancy results in spontaneous abortion or perinatal lethality 4–6. The majority of FA patients are compromised, with 60–70% having at least one congenital malformation 7. The archetypal abnormalities observed in FA are radial-ray defects, ranging from hypoplasia to complete absence of the radius 2, 3. However, in the most severe cases developmental abnormalities simultaneously affect many organ systems 4, 8, 9. The congenital malformations observed in FA probably represent the final outcome of inappropriate cell death during embryogenesis. A strong clue that this is indeed likely to be the case is the observation that inactivation of p53 in FA pathway-deficient zebrafish embryos greatly reduces the extent of developmental defects. It is likely that this increase in p53-dependent apoptosis is due to the inability of Fanconi cells to repair DNA damage that occurs during development 10.

Stem cell defects

Deficiencies in two stem cell pools contribute to the phenotype observed in FA. First, there is a defect in the haematopoietic stem cell population, resulting in aplastic anaemia, a major cause of morbidity and mortality 7, 11. Peripheral blood counts are often normal at birth before patients develop progressive pancytopenia. The first indication of bone marrow dysfunction is often macrocytosis, which is followed by thrombocytopenia, then leukopenia, and finally progresses to full-blown pancytopenia 11. The extent and severity of developmental malformations is associated with the occurrence of pancytopenia 12. This known association is consistent with the notion that embryonic haematopoiesis can be compromised in FA 13. These important clinical observations suggest that the FA fetus may be exposed to a stochastic source of DNA damage in utero, driving both developmental abnormalities and the depletion of embryonic haematopoietic reserves. The depletion of haematopoietic reserves during development may subsequently promote a more rapid progression towards aplastic anaemia within the first decade of life.

Second, patients have defects in the primordial germ cell pool, the unique population of cells that eventually differentiates into sperm and ova. Male Fanconi patients have greatly reduced fertility and sperm count and an increased frequency of sperm with abnormal morphology. Female patients are also infertile, with an increased prevalence of primary ovarian failure and premature menopause 3, 7, 14–16. Despite this, in a very few instances both FA males and females can give rise to progeny.

The pattern of sterility among FA patients is also shared with FA gene knockouts in mice 17. In fact, all FA mouse models to date exhibit the same fertility defect 18–27. FA mutant mice are born with extremely few germ cells in either the testis or the ovary. The reduced number of germ cells is due to failure in the development of primordial germ cells during embryogenesis 20, 28. This is likely to result in premature exhaustion of the germ cell reserve, culminating in primary ovarian failure and loss of spermatogenesis in the testis. This process is to some degree stochastic and therefore, on rare occasions, both male and female FA mice can parent litters; thus, the FA DNA repair pathway is not essential for germ cell development 18, 21.

It is unknown why only these two stem cell pools are affected in FA. However, this may reflect a source of endogenous DNA damage which is generated within, or in close proximity to, these stem cell pools.


FA patients also bear a significant predisposition to neoplasia—both to leukaemia and also to solid tumour formation. Interestingly up to 34% of patients with FA have myelodysplastic syndrome (MDS) 11, 29, 30. A proportion of these patients go on to develop bone marrow failure but a significant proportion develop leukaemia. In total 9% of FA patients develop leukaemia, with a median age of 14 years at diagnosis 7, 31–33. All leukaemias observed are acute, with the vast majority (>90%) being acute myeloid leukaemia (AML) 33. The prevalence of AML in the FA population is in stark contrast to non-FA patients, in which the vast majority of childhood leukaemias are acute lymphoblastic leukaemia (ALL) 12. In the non-FA population, MDS is typically a disease of the elderly and frequently progresses to AML 34. A common feature of myelodysplasia is the accumulation of cytogenetic alterations, which include chromosomal translocations, inversions and deletions 35. The same pattern of chromosomal alterations, but with increased frequency and complexity, are observed among myelodysplastic FA patients 29. The genetic alterations in MDS and AML in FA affect many oncogenes, including RUNX1, that have previously been implicated in non-FA MDS/AML 36. It is therefore likely that the genomic instability in FA drives the cytogenetic alterations leading to both myelodysplasia and AML.

Additionally, FA patients also bear a significant predisposition to solid tumour formation, with a cumulative incidence of 76% by the fifth decade 33. This is a significant tumour burden and several patients present with multiple independent primary tumours. The majority of solid tumours are squamous cell carcinomas of head and neck, oesophagus and the ano-genital region 32, 37, 38. Patients diagnosed in adulthood often have no features of FA except genomic instability and predisposition to neoplasia.

There is a mounting body of evidence that the leukaemic clone frequently originates during in utero development 39–42. The most compelling evidence to support this comes from the high rate of leukaemia concordance among monozygotic twins who shared a monochorionic placenta 43–46. Frequently both twins develop leukaemia and share a unique non-constitutive DNA sequence at the site of a chromosomal translocation—a marker of the monoclonality of the leukaemia 47, 48. Leukaemic cells can readily pass between twins who share a monochorionic placenta via the circulation 49. This shows that the leukaemic clone frequently arises during embryonic life and that chromosomal translocations are early events in the progression of leukaemia.

As already discussed, the severity of developmental defects correlate with bone marrow failure. We speculate that the embryo is exposed to a source of DNA damage that is not repaired in FA and results in cell death. The ultimate outcome of this DNA damage may be developmental abnormalities and the depletion of embryonic haematopoietic reserves. Furthermore, these genetic alterations may be the initiating event in the genesis of a leukaemic clone that will present as AML in early childhood. However, if the fetus is exposed to less DNA damage, then development proceeds normally and the FA phenotype is not revealed until solid tumour formation occurs in adulthood.

Fanconi anaemia is due to a defect in DNA repair

Cells derived from FA patients undergo premature replicative senescence, spontaneously accrue chromosomal aberrations (Figure 1a) and accumulate in G2 phase of the cell cycle 2, 50. Additionally, FA cell lines are exquisitely sensitive to a class of DNA damaging agents—interstrand crosslinkers. These compounds, such as cisplatin, are often used clinically as chemotherapeutic agents and act by covalently bonding opposite strands of DNA together 2, 51. The chemistry of guanine disposes it to react with crosslinking agents and the majority of DNA interstrand crosslinks (ICL) occur between guanines 52, 53. These lesions are extremely toxic to cells, as they pose an absolute block to both DNA transcription and DNA replication as depicted in Figure 1b. The failure of a cell to repair such lesions ultimately results in chromosomal breaks and complex radial structure formation 54.

Figure 1.

Molecular basis of Fanconi anaemia. (a) Deficiency in Fanconi anaemia pathway results in increased frequency of gross chromosomal aberrations (arrows) following treatment with DNA crosslinking agents. Murine Slx4 deficient metaphase spread following treatment with MMC. (b) DNA interested crosslinks covalently bond opposite strands of DNA together forming extremely toxic lesions to cells as they block both DNA transcription and replication. (c) Following DNA damage FANCI is phosphorylated. The FA core complex then monoubiquitantes both FANCD2 and FANCI through the action of FANCL. Upon monoubiquitination FANCD2 and FANCI are recruited to chromatin where they promote DNA repair

FA is genetically complex

To date 15 genes have been identified, the biallelic disruption of which results in human disease as outlined in Table 155–58. The majority of human FA genes have been identified by complementation analysis and cell fusion experiments. In this process, cell lines derived from FA patients are fused; those fusions that do not complement the sensitivity to crosslinking agents both belong to the same complementation group 2. Each distinct complementation group is given a letter prefixed with FANC. Complementation analysis was performed by transfection of a cell line from each complementation group with a cDNA library and selected with media containing a crosslinking agent. The plasmid containing the cDNA was recovered from clones and sequenced to reveal the identity of the putative Fanconi gene 59.

Table 1. Fanconi anaemia complementation groups
Complementation groupGene aliasChromosome locationMolecular functionReferences
FANCA 16q24.3Core complex76, 142
FANCB Xp22.2Core complex143
FANCC 9q22.32Core complex76, 144
FANCD1BRCA213q13.1Homologous recombination145, 146
FANCD2 3p25.3Core complex substrate, exonuclease activity69, 147
FANCE 6p21.31Interacts with FANCD280
FANCF 11p14.3Core complex148
FANCG 9p13.3Core complex149
FANCI 15q26.1Core complex substrate72
FANCJBRIP117q23.2G-quadreplex resolution86
FANCL 2p16.1E3 ubiquitin ligase75
FANCNPALB216p12.2Homologous recombination150, 151
FANCM 14q21.2Helicase/translocase78
FANCORAD51C17q22Homologous recombination, RAD51 paralogue58
FANCPSLX416p13.3Holliday junction resolvase18, 56, 57

Many attempts have been made to assess whether there is a correlation between complementation group and the severity of the human phenotype 60. To date, no clear evidence has been presented that the complementation group affects disease severity. This analysis is greatly complicated by the fact that some complementation groups contain very few patients, while others represent a spectrum of alleles with varying degrees of mutation 61, 62.

Studying the concordance of different aspects of the FA phenotype in several monozygotic twins and consanguineous families with FA has proved to be a fruitful avenue of research. These children, despite having identical mutations, display a variable penetrance of the phenotype, eg one twin with multiple congenital abnormalities whilst the other is comparatively intact 63–65. In some cases the difference in phenotype, despite having an identical genotype, can be explained by somatic mosacism 64–66. This phenomenon occurs when each allele contains a different mutation. It is therefore possible for one allele to act as a template for homologous recombination, allowing the other allele to be repaired. If this process occurs during embryonic development, it results in a mosaic individual in which some cells are Fanconi-deficient while others will be Fanconi-competent 66, 67.

Although a compelling argument somatic mosacism does not explain the variability of the FA phenotype. In other monozygotic twins with different phenotypes somatic mosacism was not detected. Additionally, in mouse models of Fanconi anaemia, in which somatic mosacism is not possible by virtue of both alleles carrying identical deletions, there is also a variable penetrance of the phenotype 18–27. Hence, there is a stochastic element that contributes to the severity of the phenotype that is likely to be due to an endogenous or exogenous source of DNA damage.

The Fanconi core complex—an E3 ubiquitin ligase

The Fanconi DNA repair pathway is evolutionarily conserved among Metazoa 68. Bioinformatic analysis of the first identified Fanconi gene products revealed them to be orphans, resembling only one another, and yielding little insight into the biological role of these gene products. A major advance was the discovery that two Fanconi proteins, FANCD2 and FANCI, are mono-ubiquitinated following DNA damage 69. Mono-ubiquitinated FANCD2 and FANCI localize to chromatin and form foci at the presumed sites of DNA damage 70–73.

The majority of the other Fanconi gene products (FANCA, FANCB, FANCC, FANCE, FANCF, FANCG, FANCL and FANCM) form a large nuclear core complex, as shown in Figure 1c 74–77. The core complex, through FANCL, acts as an E3 ubiquitin ligase mono-ubiquitinating FANCD2 and FANCI following DNA damage 75. Mutations in components of the core complex, often found in human FA patients, result in destabilization of this multi-protein assembly and loss of the E3 ligase activity 78–80. FANCD2 and FANCI are paralogues that form a heterodimer distinct from the core complex 81. FANCI becomes phosphorylated and mono-ubiquitinated in response to DNA damage; however, only the phosphorylation modification is necessary to promote the mono-ubiquitination of FANCD2 and maintain cellular resistance to crosslinking agents (Figure 1c) 72, 82.

The majority of Fanconi gene products act together, in response to DNA damage, to mono-ubiquitinate the key downstream effecter molecules FANCD2 and FANCI. Following mono-ubiquitination these effectors are recruited to chromatin, promoting the DNA repair process.

Defining the role of the FA pathway in DNA repair

Significant insight into the function and molecular basis of FA has been gained by using reverse genetics in the tractable chicken lymphoblastoid cell system, DT40. Many components of the FA pathway have been disrupted in this system and all show an elevated level of spontaneous sister chromatid exchange (SCE) in addition to crosslinker hypersensitivity, chromosomal instability and cell cycle perturbation 78, 83–86. Both Fancd2- and Fancc-deficient cells show reduced levels of gene conversion and somatic hypermutation in the chicken immunoglobulin locus 83, 84. These data strongly suggested a role for the FA pathway in both homologous recombination (HR), which is required for gene conversion, and also in translesion synthesis (TLS), the process required for somatic hypermutation.

Using the power of this genetically tractable system, it was possible to generate cells that lack both a FA gene and also different components of the HR machinery. It was found that the FA pathway is epistatic to both Xrcc2 and Xrcc3 with respect to crosslinker sensitivity 83, 87, 88. This shows that the FA pathway acts together with Xrcc2 and Xrcc3 to promote HR and maintain genomic instability.

The role of the FA pathway in TLS was also defined using this system. It is not unexpected that there is a requirement for TLS in ICL repair, as the lesion involves both strands, therefore preventing either strand being used as the template. This prohibits the DNA from being copied by the normal replication polymerase. There are two specialized translesion polymerases, Rev1 and Rev3, required to maintain resistance to DNA crosslinking agents 83, 89. These specialized polymerases are recruited to the site of a stalled replication fork through their interaction with mono-ubiquitinated PCNA. A ubiquitin dead point-mutant of PCNA (K164R) is unable to efficiently repair DNA-crosslinked substrates, suggesting that the translesion polymerase is recruited through PCNA 90. Both Rev1 and Rev3 act in a Fanconi-dependent pathway of ICL repair 83. Rev1 is in fact a cytidine-transferase, ie it functions to incorporate a single cytosine and is not processive 91. As DNA interstrand crosslinks occur mainly between guanines, insertion of a cytidine opposite a crosslinked guanine will result in error-free repair, making it an attractive candidate to repair crosslinked DNA 53, 91. Despite this, the catalytic activity of Rev1 is not required to maintain tolerance to interstrand crosslinks 92. It is likely that Rev1 is required to recruit Rev3 or another translesion polymerase to the site of damage-facilitating lesion bypass. It is possible that Fancd2 can directly recruit a TLS enzyme via its mono-ubiquitination modification. Despite the power of somatic genetics in defining a role for the FA pathway in HR and TLS, it was not possible to determine the temporality of events or the specific molecular defect in FA.

Mechanics of replication coupled DNA-interstrand crosslink repair

There is substantial evidence that the repair of ICLs is coupled to DNA replication. First, regardless of which stage of the cell cycle that cells are treated with a crosslinking agent they always accumulate with a 4N complement of DNA 93. Additionally, chromosomal breakage, double-strand break (DSB) formation and FANCD2 mono-ubiquitination only occur following transit through S phase of the cell cycle 93, 94. An elegant cell-free system has been developed to allow the replication-coupled repair of a site-specific DNA interstrand crosslink to be followed in vitro95–97. In this system, a replicating plasmid containing a site-specific ICL is introduced into Xenopus egg extract and the repair monitored. This assay has huge advantages over traditional biochemical approaches to ICL repair, as for the first time it is possible to follow the replication-coupled repair of a single ICL 95, 98.

Repair is initiated when two replication forks converge upon the site of the crosslink. The replication forks halt 20–70 nucleotides (nt) from the site of the crosslink (Figure 2b). This pause may represent the time required to alter the composition of the replication machinery. Only one of the replication forks then advances, pausing 1 nt before the crosslink (Figure 2c). Dual incisions occur at both 5′ and 3′ sides of the DNA interstrand crosslink, releasing a sister chromatid (Figure 1d, e) 95. The exact sites of the incisions and the identity of the nucleases responsible remain unknown. The intact sister chromatid is repaired by replication beyond the crosslink. As the crosslinked base is adducted, Watson–Crick base pairing cannot be used to ensure the fidelity of the base inserted in the nascent strand. It is necessary to use a specialized process, translesion synthesis, to replicate over the adducted base (Figure 2f) 95, 96.

Figure 2.

Mechanism of replication coupled crosslink repair. (a) DNA interstrand crosslink distorts the topology of DNA helix (b) Replication forks converge upon interstrand crosslinking pausing ∼40 nt from the lesion. (c) One replication fork advances to the base adjacent to the crosslink. (d) Endonucleases are recruited to the site of the interstrand crosslink and make incisions on the opposite strand. These incisions are dependent upon monoubiquitinated FANCD2. (e) The incision results in double strand breaks and a broken sister chromatid that are repaired by homologous recombination in a RAD51 dependent manner. (f) Specialized translesion polymerases sythesise DNA over the site of the crosslink. (g) REV7, a specialised polymerasse, extends beyond the site of the crosslink. (h) The remaining adducted guanine base is repaired by homologous recombination or by nucleotide excision repair

The broken sister chromatid, which is generated by the incision events (Figure 2e) must now be repaired. This chromatid is repaired through homologous recombination, using the sister that has been repaired by TLS (Figure 2i) as the template. The double-strand breaks of the broken sister chromatid are resected, exposing ssDNA, onto which RAD51 is loaded, facilitating strand invasion and repair of the broken chromatid 97.

Following this step, one strand has been repaired; however, the opposite strand still contains an adducted guanine (Figure 2g). It remains unclear how this lesion is repaired, but this may be achieved by a number of mechanisms, including nucleotide excision repair and base excision repair.

A specific molecular defect in Fanconi anaemia

The Xenopus system can be used to elucidate the role of specific factors in ICL repair by using antibodies to deplete a factor and then monitor the progression of the repair process. This powerful biochemical system has revealed the nature of the specific ICL repair defect in FA 96. Through the depletion of FANCD2 (and the re-introduction of mutant FANCD2 protein, which cannot by mono-ubiquitinated) it was revealed that mono-ubiquitinated FANCD2 is required for incisions at the site of the crosslink (Figure 2e). In addition, the TLS step (Figure 2f) is also dependent on mono-ubiquitinated FANCD2. Finally, depletion of FANCD2 revealed a block in the HR event, indicating that HR acts downstream of FANCD2 97. The results of FANCD2 immunodepletion complement previous data generated in DT40 cells. The combination of these biochemical and genetic data place the FA, HR and TLS machinery in a common pathway to repair ICLs. The FA core complex-mediated mono-ubiquitination of FACND2 acts upstream of both TLS and HR 83, 87, 88, 95–97. This is a major advance, defining the nature of the specific molecular block in Fanconi anaemia.

Fanconi pathway and endonuclease-mediated incision

The work carried out in Xenopus clearly identifies endonuclease-mediated incisions as a key event controlled by the FA pathway. Despite this, the identities of the endonucleases responsible for incisions at the site of the crosslink remain unknown.

A provocative candidate is the recently identified Fanconi-associated Nuclease 1 (FAN1) 99–102. FAN1 has been shown to protect Caenorhabditis elegans, chicken cells and human cancer cell lines from DNA crosslinking toxicity 99–102. It has also been shown that FAN1 interacts specifically with the mono-ubiquitinated form of FANCD2, and through this interaction is recruited to chromatin 99–102. Biochemically, FAN1 exhibits a 5′ flap endonuclease activity with an associated 5′ → 3′ exonuclease activity (Figure 3a) 99–101. This biochemical activity may be relevant to crosslink repair, as the replication fork extended to the–1 position results in a structure reminiscent of a 5′ flap (Figure 2c) 95. Therefore, FANCD2-dependent recruitment of FAN1 may represent the mechanism by which one nuclease is recruited to the site of DNA crosslink.

Figure 3.

Specificity of structure specific endonucleases linked to ICL repair. (a) FAN1 has robust endonuclease activity upon a 5′ flap structure cutting within the dsDNA. (b) XPF-ERCC1 and MUS81-EME1 both have activity toward splayed arms however, XPF is more potent on this structure and cleaves at the dsDNA:ssDNA junction. (c) MUS81-EME1 has greater activity on a 3′ flap structure. (d) The specificity of XPF-ERCC1 is altered by adding a psoralen interstrand crosslink to splayed arms. The nuclease no longer cuts at the junction between double and single stranded DNA but nicks within the duplex DNA. XPF-ERCC1 can subsequently cleave at the dsDNA:ssDNA junction. (e) The specificity of each endonucleases is not known on the substrate predicted from the model of replication coupled repair of crosslinked DNA based on the Xenopus system

However, depletion of FAN1 results in normal kinetics of γ-H2AX foci formation, a marker of DNA double-strand breaks. This is inconsistent, as depletion of the nuclease responsible for the incision event in Figure 2e should fail to generate double-strand breaks and therefore one would expect decreased γ-H2AX signal 99–102. Additionally, RPA foci and RAD51 filament formation is unimpaired following depletion of FAN1 and damage with crosslinking agents. This suggests that homologous recombination and the prerequisite double-strand break formation occurs in the absence of FAN1. Finally, the disruption of F1n1 in the DT40 system revealed a phenotype distinct from other chicken Fanconi knockout cell lines 103. Fan1-deficient cells do not show elevated levels of spontaneous SCE, a feature of all other FA mutant DT40 cells 78, 83–86. Moreover, cells deficient in both Fancc and FAN1 show additional sensitivity to crosslinking agents compared to either single mutant 103. Taken together, these data show that, whilst FAN1 is required to maintain genome stability from crosslinking agents, it is likely to act downstream of the incision step or in a distinct pathway from classical FA repair.

MUS81–EME1 is a heterodimeric endonuclease that has been implicated in the repair of ICLs 104–107. MUS81-deficient murine embryonic fibroblasts and human lymphoblasts are hypersensitive to crosslinking agents 104, 106, 107. However, there is no genetic evidence linking MUS81–EME1 to the Fanconi-dependent mechanism of crosslink repair. Mus81 knockout mice are fertile and do not exhibit primordial germ cell failure, which is a universal feature of all other Fanconi knockout mice 106, 107. This suggests that, although MUS81 is required to maintain resistance to DNA interstrand crosslinking agents, it does not function in a Fanconi-dependent mechanism.

The heterodimeric endonuclease XPF–ERCC1 has been linked to ICL repair in numerous studies. This endonuclease is involved in multiple DNA repair transactions, including nucleotide excision repair, which is the pathway required to repair DNA damage caused by UV irradiation 108. XPF–ERCC1 has a role in a subset of homologous recombination transactions, specifically single-strand annealing, a function shared with its ancient yeast orthologue 109–112. Consistent with the multiple roles of XPF–ERCC1, the knockout mice of both of these genes have severe and complex phenotypes not observed in other knockout mice of either components of the NER or the FA pathway 113, 114. Whilst XPF-deficient patient-derived fibroblasts exhibit only a mild sensitivity to UV, they are exquisitely sensitive to DNA crosslinking agents 115. In addition, cell lines from Xpf- and Ercc1-deficient mice, Chinese hamster ovary cell mutants and Dictyostelium mutants are all extremely sensitive to crosslinking agents 68, 113, 116.

There is an emerging body of evidence linking XPF–ERCC1 with the FA pathway in crosslink repair. First, Xpf and Fancd2 are epistatic with respect to ICL repair in Dictyostelium, indicating that they are likely to act in a common pathway 68. Furthermore, in the absence of XPF–ERCC1 there is a defect in recruitment of FANCD2 to chromatin 116. Most compellingly, the phenotype of Ercc1-deficient mice has significant overlap with FA exhibiting growth retardation, blood cytopenias, developmental defects and sterility 113, 117, 118. It is important to note that the sterility observed in Ercc1-deficient mice is a pattern of primordial germ cell failure the unifying feature of all FA mouse models 118.

Biochemically, XPF–ERCC1 is an endonuclease making an incision in one strand of a splayed arm structure at the junction between dsDNA and ssDNA (Figure 3c) 108, 119, 120. This biochemical activity is compatible with cleavage at the side of ICL at which the replication fork has not extended (Figure 2d, right) 95. However, the biochemical activity of XPF–ERCC1 has been assayed on a splayed arm substrate containing a psoralen ICL (Figure 3d) 121. This modified substrate altered the biochemical activity of XPF–ERCC1, which now nicks the dsDNA four nucleotides before the psoralen ICL. Surprisingly, XPF–ERCC1 will subsequently make an incision at the junction between double- and single-stranded DNA 121. This altered the specificity and biochemical activity of XPF–ERCC1 on a crosslinked substrate suggests that XPF–ERCC1 alone may be sufficient to perform the dual incisions at the site of an ICL. This combined genetic and biochemical evidence makes XPF–ERCC1 the prime candidate to make the initial incision at the site of an ICL.

The recent discovery of the vertebrate orthologue of the yeast gene Slx4 elucidates an important new component to DNA crosslink repair machinery in vertebrates 122–125. Yeast Slx4 was identified in a synthetic lethal screen with Sgs1, the yeast orthologue of the Bloom's syndrome helicase 126. Human SLX4 interacts with the structure-specific endonuclease SLX1. The SLX4–SLX1 complex can resolve Holliday junctions in vitro; however, the biological significance of this biochemical activity remains uncertain.

SLX4 interacts with both the nucleases XPF–ERCC1 and MUS81–EME1, implicated in crosslink repair, and potentiates their biochemical activity in vitro122–125. Loss of Slx4 in yeast as well as in Drosophila, C. elegans, mouse and human cells leads to a hypersensitivity to DNA crosslinking agents 18, 56, 57, 125, 127. Remarkably, four kinships have recently been described that carry mutations in SLX4, and present clinically as Fanconi anaemia 56, 57. In addition, disruption of murine Slx4 recapitulates the key cellular and organismal features of Fanconi anaemia 18. These data suggest that Slx4 acts in a FA pathway-dependent DNA interstrand crosslink repair process.

Disruption of Slx4 in chicken cells has revealed it to be cellularly essential—at odds with yeast, mouse and human systems 18, 56, 57, 128, 129. This may be a peculiarity of chicken, due to its abnormal complement of nucleases—as it lacks both Mus81 and Ercc1 but has orthologues of Xpf and Eme1. In this system Xpf is cellularly essential, which is also in contrast to human, mouse and yeast 114, 128, 130. This makes the analysis of nucleases involved in crosslinking very difficult in chicken cells.

As already discussed, it is likely that XPF–ERCC1 acts in FA-dependent crosslink repair. Furthermore, Slx4-deficient cells have a defect in the recruitment of XPF–ERCC1 onto chromatin 18, 56. We therefore propose a model in which SLX4 has an analogous role in recruiting XPF–ERCC1 to the site of an interstrand crosslink to that of XPA in nucleotide excision repair (NER). In NER, XPA interacts with ERCC1, recruiting it to the site of UV-damaged DNA 131, 132. The loss of XPA sensitizes cells to UV irradiation but not to DNA crosslinking agents 115. It is possible to further delineate the role of ERCC1 in ICL repair from its role in NER, with a point mutant in ERCC1 which prevents its interaction with XPA. This mutant form of ERCC1 complements the sensitivity of ERCC1-deficient cells to DNA interstrand crosslinking agents, but not to UV irradiation 133. Conversely, loss of SLX4 does not sensitize cells to UV irradiation but leads to increased sensitivity to DNA interstrand crosslinking agents. It is therefore likely that SLX4 segregates the function of XPF–ERCC1 in ICL repair from its role in the repair of UV-induced damage.

There remain many unanswered questions relating to the mechanics of Fanconi-dependent incisions at the site of an ICL. Significant insights will be gained through immunodepletion of these nucleases complexes in the cell-free repair of a DNA ICL in Xenopus egg extract 96, 97. It is essential to biochemically assess the activity of both FAN1 and XPF–ERCC1 on the substrates predicated in the ICL repair model (Figure 2d). As illustrated by XPF–ERCC1, the biochemical activity of the endonuclease can be altered by the introduction of a crosslink into the DNA substrate 121. It is also necessary to test the biochemical activity of complexes implicated in crosslink repair. The FAN1–FANCD2–Ub and SLX4–XPF–ERCC1 complexes may have significantly altered specificities and biochemical activities compared to the catalytic subunits alone.

Fanconi anaemia and endogenous DNA damage

A striking feature of Fanconi anaemia is the stochastic nature of the developmental abnormalities, bone marrow failure and neoplasia. This is almost certainly due to the nature of the DNA damage. Cisplatin and Mitomycin C, the drugs used to induce ICLs, are not common environmental sources of DNA damage and it is unlikely that a DNA repair pathway has evolved to protect against these mutagens. The source of DNA damage may be endogenous products of cellular or physiological metabolism or an as-yet unidentified environmental mutagen.

A potential source of endogenous DNA damage are reactive aldehydes. It has been shown that specifically cells deficient in the Fanconi repair pathway are hypersensitive to both formaldehyde and acetaldehyde 134–136. Importantly, there are differences in the genetic requirements for the repair of aldehyde-mediated DNA damage and interstrand crosslinks. Whilst there is a requirement for homologous recombination (Xrcc2, Xrcc3) and translesion synthesis (Rev1 and Rev3) in crosslink repair, there is no such requirement in the repair of aldehyde-damaged DNA 135, 136. This suggests that the nature of the DNA lesion caused by reactive aldehydes may not be a traditional interstrand crosslink.

These small reactive molecules are not only common environmental mutagens but are also frequently produced by cellular and organismal metabolism 137. Moreover, mice deficient in both Aldh2 (the enzyme responsible for the catabolism of several aldehydes, but most importantly acetaldehyde) and Fancd2 potentiates the Fanconi phenotype 135. These double-deficient mice have reduced in utero survival, haematopoietic dysfunction and a strong predisposition to acute leukaemia. This shows that the levels of reactive aldehydes generated endogenously are sufficient to cause DNA damage and to precipitate the phenotype observed in Fanconi anaemia 135. Intriguingly, there appears to be a special requirement for aldehyde catabolism and Fanconi DNA repair during embryonic development. It is possible that some process during development increases the likelihood of exposing cells to a source of aldehyde-mediated DNA damage. Acetaldehyde can cause many forms of DNA damage, including base damage, inter- and intrastrand crosslinks and protein adducts 138–141. It will be critical to our understanding of FA to show which of these DNA lesions is repaired by the FA pathway.

Conclusions and future perspective

Although FA presents with a complex clinical picture, this phenotype is due to a simple and defined defect in DNA repair. Recent work has elucidated the probable enzymatic machinery necessary to make the incisions at the site of the DNA lesion. It remains to clearly define the biochemical steps in the repair transaction, and it is likely that depletion of the nucleases XPF–ERCC1, MUS81–EME1 and FAN1, in addition to the regulator of nucleases, SLX4, will shed light on the role of these proteins in repair. Reactive aldehydes are likely to be the source of endogenous DNA damage repaired by the FA pathway. The next key step in the field will be to define the nature of DNA modification caused by these reactive molecules that the FA pathway evolved to counteract.


We are very grateful to Juan Garaycoechea, Frederic Langevin and Michael Hodskinson for critically reading the manuscript. GPC also wishes to thank Ivan Rosado for so many (constructive) arguments. The MRC is thanked for financial support.

Author contributions

Both authors wrote the paper and had final approval of the submited manuscript.