A novel cancer risk prediction score for the natural course of FA patients with biallelic BRCA2/FANCD1 mutations

Abstract

Biallelic germline mutations in BRCA2 occur in the Fanconi anemia (FA)-D1 subtype of the rare pediatric disorder, FA, characterized clinically by severe congenital abnormalities and a very high propensity to develop malignancies early in life. Clinical and genetic data from 96 FA-D1 patients with biallelic BRCA2 mutations were collected and used to develop a new cancer risk prediction score system based on the specific mutations in BRCA2. This score takes into account the location of frameshift/stop and missense mutations relative to exon 11 of BRCA2, which encodes the major sites for interaction with the RAD51 recombinase, and uses the MaxEnt and HBond splicing scores to analyze potential splice site perturbations. Among 75 FA-D1 patients with ascertained BRCA2 mutations, 66 patients developed 102 malignancies, ranging from one to three independent tumors per individual. The median age at the clinical presentation of peripheral embryonal tumors was 1.0, at the onset of hematologic malignancies 1.8 and at the manifestation of CNS tumors 2.7 years, respectively. Patients who received treatment lived longer than those without. Using our novel scoring system, we could distinguish three distinct cancer risk groups among FA-D1 patients: in the first, patients developed their initial malignancy at a median age of 1.3 years (n = 36, 95% CI = 0.9–1.8), in the second group at 2.3 years (n = 17, 95% CI = 1.4–4.4) and in the third group at 23.0 years (n = 22, 95% CI = 4.3—n/a). Therefore, this scoring system allows, for the first time, to predict the cancer manifestation of FA-D1 patients simply based on the type and position of the mutations in BRCA2.

Introduction

Fanconi anemia (FA) is a rare inherited childhood disorder clinically characterized by congenital abnormalities, bone marrow dysfunction and a high propensity for a variety of cancers (1–3). Genetically, FA is caused by homozygous or compound heterozygous biallelic mutations in 22 DNA repair genes, 20 of which are autosomal recessive genes, with one X-chromosomal recessive and one autosomal dominant gene. These correspond to 22 FA complementation groups (FA-A to -C, -D1, -D2, -E to -G, -I to -J and -L to -W) (4,5). The associated gene products function in a large network of DNA damage signaling and repair, termed the FA-BRCA pathway, that is predominantly responsible for the efficient repair of DNA interstrand crosslink (ICL) lesions (4). Consistent with this, defective DNA repair and genome instability are the hallmark of FA, yielding a characteristic cellular hypersensitivity and chromosomal aberrations/breaks in the presence of the DNA ICL agents mitomycin C or diepoxybutane (6). This hypersensitivity is the basis for the so-called ‘chromosome breakage test’ that is still the standard test for diagnosing FA (6).

BRCA2 (OMIM ^*600185), a major breast and ovarian cancer predisposing gene (7,8), was first recognized as the FANCD1 FA gene (OMIM #605724) in 2002 based on the presence of biallelic mutations in four children from different cancer kindreds (9). Compared with the vast majority of FA patients belonging to the more common classical complementation groups FA-A/-C/-G (1–3), biallelic mutations in BRCA2 occur in < 5% of FA patients worldwide and are often associated with very severe clinical phenotypes, including extensive congenital malformations and very early onset of one or more aggressive cancers (10–13). In a first review of 27 FANCD1 patients in 2007, Alter et al. estimated a 97% probability for these patients to develop a cancer of any kind by the age of 5.2 years (14). In addition to myelodysplastic syndrome (MDS)/acute myeloid leukemia, which is the most frequent malignancy in FA (1,12,13), FANCD1 patients specifically experience many embryonal cancers (15,16), including Wilms tumors (12,17), neuroblastoma (17,18), medulloblastoma (10,11), hepatoblastoma (19,20), glioblastoma (17,20,21) and others (14,22).

BRCA2 is a large protein of 3418 amino acids that is encoded by 27 exons (7) and interacts with many essential DNA repair proteins, including those involved in mediating homologous DNA recombination (HDR) repair, such as PALB2, BRCA1 and RAD51 (23–25). Notably, BRCA2 directly binds PALB2 through amino acids 10–40 (24), which are encoded by sequences in exons 2 and 3 of BRCA2 (26). The BRCA1 and BRCA2 proteins do not directly interact but instead are linked together by PALB2 (27). While BRCA1 is involved in early stages of the DNA damage response, an important function of PALB2 is to recruit the BRCA2 protein to partially processed sites of DNA damage within nuclei (24,26,27). BRCA2 has a central role in regulating the assembly of the RAD51 recombinase into a nucleoprotein filament with single-strand DNA, which initiates HDR repair via strand invasion of duplex DNA (28). BRCA2 directly interacts with RAD51 through two domains (26): exon 11 encodes the main RAD51-binding domain with eight highly conserved BRC motifs (23) and the C-terminus of BRCA2 encoded by exon 27 contains an accessory RAD51-binding region (23,29,30). Another key functional domain in BRCA2 is a DNA-binding domain (DBD), which is encoded by exons 15–26 and spans from amino acids 2482 to 3184 (26). Interestingly, all of the consensus pathogenic/likely pathogenic (P/LP) missense BRCA2 mutants in ClinVar that do not affect mRNA splicing or the translation start, except for c.67G > T (p.Asp23Tyr) and c.91T > C (p.Trp31Arg) in the PALB2-binding domain, reside in the DBD.

Data from genetic engineering of murine Brca2 demonstrated the importance of BRCA2–RAD51 interactions by showing that deletions/truncations that remove the BRC repeats or the C-terminal accessory RAD51-binding domain are embryonic lethal (31–33). Cells derived from early gestation stages of these mice show decreased cellular proliferation and hypersensitivity to ionizing radiation (31–33). Still, some engineered mice, e.g. with deletion of exon 27 (34), which encodes the accessory RAD51-binding site and putative nuclear localization signals, are viable but display an increased incidence of malignancies and cellular genomic instability. Thus, DNA repair is compromised but not completely lost in these cells or mice with at least one hypomorphic BRCA2 allele that includes exon 11 (34–37).

Therefore, the simple existence of human FANCD1 patients with biallelic mutations in BRCA2 strongly suggests that at least one of the alleles in these individuals provides some minimal residual protein function. It has been demonstrated in FA that the affected FANC gene, the type of alteration and residual functions of the genes/alleles can influence the clinical manifestations (38–42). Specifically in FA-D1, some BRCA2 mutations are reported to be associated with specific tumor types: the IVS7 splice site mutations (c.631+1G > A, c.631+2T > G) are associated with an early development of acute myeloid leukemia (AML) (13,43), while the c.658_659delGT and c.5946delT (old BIC nomenclature c.6174delT) (44) mutations in BRCA2 (NM_000059.4) are typically associated with the development of solid cancers, especially brain tumors (15–17). Nevertheless, it has not been shown at this point whether the presence of residual activity in functionally impaired/mutated alleles of BRCA2 can be correlated with the severity of the cancer-prone phenotype and whether this residual activity may be a reliable predictor for the age of onset of malignancies.

Here, we summarized all known cases of patients with biallelic alterations in BRCA2. By using only patients with biallelic BRCA2 mutations that were classified as P/LP, we were able to generate an easy-to-use algorithm, based on the location and types of the BRCA2 mutations, which reliably predicts the severity of clinical cancer phenotypes in FA-D1 patients. We believe that this risk assessment tool will be helpful for the clinical care of individual BRCA2/FANCD1 patients and their families.

Results

The genetic and clinical data of 96 patients with two alterations/variants in the BRCA2 gene are shown in Supplementary Material, Table S1. Thirty-one patients were reported as male (32%) and 49 as female (51%). For 16 patients, gender information was not available (17%). Seventy-nine of these 96 patients were reported to have a positive chromosomal breakage test, indicative of FA, while the information was not provided for the others. Eight patients were siblings of confirmed FA-D1 patients and were assigned to the FA-D1 complementation group based on the presence of the same two BRCA2 germline alterations from their affected siblings. Nine patients were classified as FA-D1 based on the combination of germline alterations/mutations in BRCA2 and the clinical manifestations. The genetic spectrum of these 96 patients encompassed 56 different deletion/nonsense/frameshift, 16 missense and 15 splice site alterations, for which the locations are depicted in Figure 1. For one patient, the specific BRCA2 mutations were not reported, but the clinical data was still added to Supplementary Material, Table S1 (45).

Malformations, malignancies and specific alterations/mutations

Information about the presence of congenital malformations was not available for 18 patients. Seventy-three of the remaining 78 (94%) patients, as shown in Supplementary Material, Table S1, were born with at least one FA-associated congenital malformation, excluding cafe-au-lait spots and skin abnormalities, as shown in (46). Fifty-nine patients had at least one malformation of the head (including microcephaly); 36 patients showed limb malformations (including thumbs and radii) and 13 patients had malformations of the cardiovascular system, 26 of the urogenital and 21 of the gastrointestinal tract. A polymalformative syndrome [vertebral defects, anal atresia, cardiac defects, tracheo-esophageal fistula, renal anomalies and limb abnormalities (VACTERL)-associated] was described in 16 patients (Supplementary Material, Table S1).

Strikingly, these 96 patients developed 124 malignancies (Table 1), of which 35% were hematological, 24% CNS and 28% embryonal tumors. The high lifetime malignancy risk of 83% resulted in a medium life expectancy of 4.6 years for the entire cohort (95% CI = 3.2–7.0 years). In addition, patients with at least one malignancy lived on average only 3.8 years (95% CI = 2.8–5.0 years).

Although a clear heterogeneity existed in the clinical manifestations of the disease, presumably due to the residual activity of at least one or both BRCA2 alleles, it should be noted here that the data sets reported were not always complete: out of 96 patients, only 81 had biallelic loss-of-function mutations [class 4 or 5 as defined by the International Agency for Research on Cancer (IARC)], 27 had extensive malformations according to Guardiola et al. (47), 63 had early onset of a malignancy (under 5 years of age) and 68 had a positive FA breakage test.

The pregnancies of four patients with biallelic pathogenic variants were terminated prematurely due to the presence of loss-of-function BRCA2 mutations and congenital anomalies or malignancies (48–50). In addition, nine female patients with primary gonadal dysfunction were reported as carriers of biallelic germline alterations in BRCA2 (51–56). However, of these, only two sisters (47) and two individuals (54,56) carried pathogenic mutations, had a positive FA breakage test and were phenotypically well described. Although all nine patients are included in Supplementary Material, Table S1, we consider only these four women to be reliably classified as FA-D1 patients.

Six patients with clinical signs of FA had at least one benign BRCA2 alteration (IARC class 1: 4x c.1114A > C/p.Asp372His and 1x c.7469T > C/p.Ile2490Thr; class 2: 1x c.6842-12A > G). Remarkably, three of them were reported to have a positive FA breakage test, thus raising the possibility that the patients might have belonged to another FA complementation group. Seven patients had a positive chromosomal breakage test and were harboring a variant of uncertain significance (VUS) on one allele (3x c.7802A > G/p.Tyr2601Cys; c.7847C > T/p.Ser2616Phe; 2x c.8471G > C/p.Arg2824Thr; c.316G > A/p.Gly106Arg). However, a majority of them developed a first malignancy only as adults, thus raising concerns whether they had been properly allocated to the FA-D1 group. For one patient with AML at the age of 2 years, neither the exact BRCA2 mutations nor a chromosomal breakage test were mentioned (Supplementary Material, Table S1) (45).

Finally, subtracting the patients with incomplete or inconclusive data, a total of 75 patients with biallelic deleterious BRCA2 mutations could definitely be assigned to the FA complementation group FA-D1, based additionally on chromosomal breakage testing and/or a characteristic clinical phenotype (early cancer diagnosis and/or congenital malformations). We considered this strict definition of an FANCD1 patient to be a prerequisite to being able to perform reliable genotype–phenotype analyses, as we care for a small boy who carries two established germ-line mutations on the same BRCA2 allele and has a normal clinical phenotype and a normal FA test (Hanenberg et al., unpublished data).

Cumulative incidence of any (first) malignancy and overall survival

In an effort to identify valid genotype–phenotype correlations, we decided to focus our analysis on the 75 patients reliably assigned to the FA-D1 complementation group. Only 9 of these 75 patients (12%) did not develop any malignancy at the time of reporting. The other 66 patients developed 102 malignancies, ranging from one to three independent tumors (considering MDS and AML as closely related entities). As shown in Figure 2A, during the lifetime of these patients, one neoplasia occurred in 39 patients (52%), two malignancies in 18 (24%) and three tumors in 9 patients (12%). For these 75 patients, the median age at the first malignancy was 1.9 years (95% CI = 1.3–3.0 years) (Fig. 2B) and the calculated median overall survival (OS) 3.8 years (95% CI = 2.8–5.0 years) (Fig. 2C). When excluding patients with simultaneous manifestation of two malignancies, the most common first malignancies in FA-D1 patients were peripheral embryonal tumors (n = 21) followed by hematologic malignancies (n = 18) and CNS malignancies (n = 17) (Fig. 2D). The Kaplan–Meier analysis in Figure 2D demonstrated that the median age at the manifestation of peripheral embryonal tumors was 1.0 years (95% CI = 0.6–1.9), for hematologic malignancies 1.8 years (95% CI = 1.1–3.0) and for CNS tumors 2.7 years (95% CI = 1.8–3.9). A statistical analysis revealed that the differences in the age at diagnosis of the first malignancy between embryonal tumors versus CNS were significant (p < 0.05 using log-rank testing). Strikingly, all cases of first malignancies, diagnosed as MDS/AML, CNS neoplasia and peripheral embryonal tumors occurred within 5.5, 4.8 and 3.8 years of age, respectively (Fig. 2D).

Treatment appears to prolong the life of FANCD1 patients

According to the available information for 56 patients, 40 patients received treatment for their tumor(s) including chemotherapy, surgery, irradiation and/or hematopoietic stem cell transplantation (HSCT). The median OS after the malignant diagnosis was 1.5 years (95% CI = 0.9–2.7) for the patients who received therapy and 0.3 years (95% CI = 0.1–1.2) for those without any treatment (Fig. 2E). This accounted for a hazard ratio (HR) of 0.29 (95% CI = 0.15–0.56; p < 0.001) when receiving treatment.

The malformation status can predict the patients’ mortality risk

Information on the malformation status was available for 65 of the 75 FANCD1 patients. Based on the published data, only three of 65 patients did not have any congenital malformations (excluding skin anomalies as malformations). As the degree and severity of malformations has been linked to clinical parameters, at least for some of the FA genes (38,40–42), we separated the 65 patients into two groups: 1) those with limited and 2) those with extensive malformations based on the criteria used by Guardiola et al. (47). As shown in Figure 3A, both patient groups showed a similar incidence of malignancies (limited n = 40, median age 1.9, 95% CI = 1.3–3.0; vs. extensive n = 25, 1.3 years, 95% CI = 1.2–4.3, respectively). We did not notice statistically significant differences in the OS (Fig. 3B), with a median of 3.1 years for the group with limited malformations (95% CI = 2.7–5.3) and a median of 2.6 years for the extensive malformation group (95% CI = 2.0–8.0).

Next, we defined the extensive malformation group in a different way, as patients needing neonatal surgery, having gastrointestinal, skeletal or CNS malformations and/or having been born with polymalformative syndrome; the remaining patients were then assigned to the limited malformation group. Our Kaplan–Meier analysis revealed that this approach also did not have any predictive value for the incidence of malignancy in both groups of patients (‘limited’ n = 35, median age of onset 1.8, 95% CI = 1.3–3.0 vs. ‘extensive’ n = 30, median of 1.8 years at onset, 95% CI = 1.2–3.1) (Fig. 3C). However, the group with less severe malformations and no requirement for surgery showed a significantly decreased mortality risk (OS: ‘limited’ median age 4.5, 95% CI = 2.5–12.0; vs. ‘extensive’ 2.7 years, 95% CI = 2.0–4.6, respectively; HR 0.5, 95% CI = 0.3–0.9, p < 0.05) (Fig. 3D).

A simple genetic scoring algorithm distinguishes three clinically distinct cancer risk subgroups

The group of 75 FA patients had 45 BRCA2 alleles with nonsense/frameshift/stop mutations, 11 alleles with splice site mutations and 10 with amino acid exchanges/missense mutations. In total, 58 of 75 patients were compound heterozygous mutation carriers (Supplementary Material, Table S1). In order to understand whether the type(s) of the two mutations influences the clinical onset of malignancies, we first analyzed the age at onset of the first malignancy and OS for all 75 patients. As shown in Supplementary Material, Figure S1, the curves for patients with biallelic frameshift/stop mutations (n = 37, median age of cancer incidence 1.8 years, 95% CI = 1.3–3.0) were comparable to those of patients with biallelic splice site (n = 6, median age 2.0, 95% CI = 0.4—n/a) and also to those of patients with combinations of pathological frameshift/stop, splice and missense mutations (n = 31, median age 1.9, 95% CI = 1.3–4.9). The OS in all three groups was also comparable (Supplementary Material, Figure S1). Next, we analyzed in the largest subgroup of patients with biallelic fs/stop mutations, whether the position of the mutations relative to exon 11, containing the BRC repeats for RAD51 binding, influences the incidence of malignancies. Interestingly, patients with both mutations located before or in exon 11 developed the first malignancy significantly earlier (median age at malignancy onset 1.3 years of age, 95% CI = 0.7–2) than those with one fs/stop mutation located after exon 11 (median age at malignancy onset 2.65 years of age, 95% CI = 1.4—n/a, Fig. 4A) and also displayed shorter survival (Fig. 4B). Patients with both mutations located after exon 11, although fewer in number, had a significantly later onset of the first malignancy (median age 5.3 years of age, 95% CI = 3.0—n/a, Fig. 4A), suggesting an additive benefit of each mutated BRCA2 allele/haplotype occurring downstream of exon 11 in delaying the onset of cancer.

For genotype–phenotype predictions, we then created a mutation scoring system (Table 2), where each allele was assigned between 0 and 2 points depending on the following two criteria: 1) Frameshift/truncating and stop codon mutations (nonsense variants) were assigned 0 and missense mutations +1 point, as we expect the latter ones to be functionally less deleterious. Importantly, stop/frameshift and missense mutations located downstream of exon 11 obtained an extra point (+1), as any hypomorphic protein potentially expressed from this allele would contain the essential RAD51-binding domains located in exon 11 of BRCA2 (23,57) (Fig. 1). 2) The splice site alterations were analyzed using MaxEnt score based on the maximum entropy principle to characterize the intrinsic splice site strength (58). Here, we assumed that changes in the MaxEnt score (∆MaxEnt) > 4 will affect the functionality of the splice site. Splice donor site alterations were additionally analyzed by the HBond score (HBS) (https://www2.hhu.de/rna/html/hbond_score.php), which calculates hydrogen bonding between a splice donor sequence and all 11 nucleotides of the 5′ end of an U1 snRNA (59,60). Splice donor variants ∆HBS > 2 have a high likelihood to change the splicing pattern at the site. Using the changes in the scores from both algorithms, splice site mutations were characterized as severe (predicted to have strongly reduced splicing strength = > 0 point added) or mild (+1 point).

As shown in Table 3, we analyzed a total of 11 splice site mutations, seven splice donor and four splice acceptor site alterations. Based on the changes in the MaxEnt and HBond scores, two splice site alterations were classified as mild (+1) and nine as severe (+0). We also indicated in Table 3, in the last column on the right, the clinical course and the second BRCA2 mutation for each patient.

Using this scoring system, we assigned between 0 and 4 points (0–2 points for each BRCA2 allele) to each of the 75 patients (Supplementary Material, Table S1). We were thereby able to distinguish three patient groups having scores of 0, 1 and  > 1, respectively. Every patient in the group with a score of 0 developed tumors and 50% experienced multiple malignancies (Table 4). Remarkably, although all patients with a score of 1 experienced malignancies, only 29% developed more than one tumor. In contrast, only 59% of patients with scores > 1 developed malignancies at all; four of these patients, however, probably owing to their extended life expectancy, also developed more than one malignancy (Table 4). The median age at the diagnosis of the first malignancy for a score of 0 was 1.3 years (n = 36, 95% CI = 0.9–1.8), while patients with a score of 1 and with higher scores (>1) showed median incidences of their first malignancies at 2.3 years (n = 17, 95% CI = 1.4–4.4) and 23.0 years (n = 22, 95% CI = 4.3—n/a), respectively. The overall log-rank testing analysis revealed that the differences between the three groups were highly significant at a level of p < 0.001 (scores 0 vs. 1 p = 0.002; scores 0 vs. 2 p < 0.001; scores 1 vs. 2 p < 0.001) (Fig. 4C). The Kaplan–Meier survival analysis showed a median OS of 2.3 years for group 1 with score 0 (n = 36, 95% CI = 1.8–4.0), 3.8 years for group 2 with a score of 1 (n = 17, 95% CI = 2.8—n/a; HR 0.44, 95% CI = 0.2–0.9, p < 0.05) and 35.0 years for group 3 with a score > 1 (n = 22, 95% CI = 8.0—n/a; HR 0.15, 95% CI = 0.1–0.3, p < 0.001). These differences were significant by log-rank (Mantel-Cox) testing: scores 0 versus 1 p < 0.05; scores 0 versus 2 p < 0.001; scores 1 versus 2 p < 0.05 (Fig. 4D).

Missense mutations in the BRCA2 DBD, located between amino acids 2482 and 3184, can be severely damaging for BRCA2 protein function (26,61–63). Therefore, we analyzed the effect of assigning only one point to missense variants in the DBD and the reanalyzed the data for the three groups. As shown in Figure 4E for the incidence of the first malignancy (cumulative scores of 0, 1 and  > 1 with median ages of 1.3, 3.5, and 5.0 years, respectively), and Figure 4F for the survival of patients, this scoring facilitated a better separation of the curves in the first five years, but did not discriminate well later in life. Importantly, the differences between the three curves of patients with scores 0, 1 and  > 1 were also statistically significant (p < 0.05). However, it is critical to point out here that all missense mutations in our patient group are located in the DBD. As we believe that biological differences exist between having an allele that encodes a truncated protein, often at low levels due to nonsense mediated decay or aberrant splicing, or having a full-length protein with a substitution of one residue, we favor our initial approach for the scoring algorithm shown in Table 2 and Figure 4C and D.

Discussion

Our work summarizes the largest cohort of patients with two alterations in the BRCA2 gene so far. These 96 patients experienced 124 malignancies. Based on the tumor spectrum and a medium life expectancy of only 4.6 years, the analysis confirms that the FA-D1 complementation group is phenotypically different from most FA complementation groups except for the FA-N patients (16). However, the phenotypic similarities between FA-D1 and FA-N patients make sense given that the BRCA2 and PALB2 proteins encoded by FANCD1 and FANCN, respectively, physically interact and are functionally related (24,26,27). Twenty-one patients were excluded from the genotype–phenotype analysis, either due to prenatal death or abortion, insufficient data sets or the presence of at least one BRCA2 alteration that was classified as either VUS or even benign/likely benign. Although some of these 21 FA patients were phenotypically similar to the genetically ascertained FANCD1 patients, others showed surprisingly mild clinical phenotypes. This raises the possibility that either these patients were not true FA-D1 patients or that the underlying mutation in the second BRCA2 allele had not been found, possibly as a result of large genomic deletions (64).

We therefore concentrated our genotype–phenotype analysis on the development of a cancer-predictive score for the 75 patients that were reliably assigned to the FA-D1 complementation group. While 75 patients developed 102 malignancies, ranging from one to three independent tumors, nine patients did not develop any tumors during the observation time period. Importantly, approximately 36% of patients even experienced two or more malignancies during their life. For the group of 75 patients, the median age at the first malignancy was 1.9 years, and the calculated median OS was 3.8 years. The most common first malignancies in FANCD1 patients were peripheral embryonal tumors followed by hematologic malignancies and CNS malignancies. Importantly, it needs to be pointed out here that a clear bias exists between the type of the first malignancy and the OS/time to develop a second cancer. For example, localized Wilms tumors often manifested during the first year of life; however, the early stages of this embryonal malignancy can be treated well by nephrectomy and limited chemotherapy for complete local disease control (65), thereby prolonging the survival of these patients. In contrast, while CNS malignancies often remained untreated, a few centers decided to perform HSCT for FANCD1 patients with AML/MDS, resulting in a prolonged leukemia-free, but not necessarily cancer-free survival in < 30% of transplanted patients (66). Our entire cohort included 12 patients who had received HSCT and three survived longer after transplantation (Supplementary Material, Table S1). Therefore, although clearly influenced by the specific tumor characteristics, any sort of treatment of the patients in our cohort was associated with a significantly longer OS. However, successful treatment for the non-hematological malignancies usually implied surgical excision of the tumors, sometimes supported by adapted irradiation and/or chemotherapy with non-crosslinking agents (17,66–68). Of course, another compelling possibility to explain the increased OS of patients treated is that, owing to the overall dismal prognosis, only FA-D1 patients with less advanced malignancies would be considered for treatment at all.

In general, we favor the idea that at least one hypomorphic allele in BRCA2 has to be present to be compatible with sufficient fetal development and birth (16,33,69,70). The existence of several patients (n = 12) with two IARC class 4/5 mutations, both located before the essential exon 11, which contains the BRC repeats necessary for binding the RAD51 protein, would appear to contradict our hypothesis (26). However, the seminal publication by Biswas et al. from 2011 provided a clear experimental rationale for these patients (69): low levels of a naturally occurring alternatively spliced isoform of BRCA2 with normal function, exon ∆4–7, appear to be the main reason why patients with two null alleles (in/del, nonsense) before exon 11 developed to term and beyond. The observation that the most frequent mutation (n = 15) in our cohort, the c.658_659delGT deletion in exon 8, which leads to a frameshift and premature stop, only occurred in heterozygous constellations, among them eight patients who developed Wilms tumor and five patients with CNS neoplasia, is also supportive of our hypothesis. The second most common mutation in our patient cohort (n = 7) was the Ashkenazi Jewish founder deletion c.5946delT in exon 11, a loss-of-function frameshift deletion, which surprisingly was present in one patient in homozygous form (71).

Other mutations were more obviously associated with a milder clinical phenotype. For example, the missense mutation c.8524C > T; p.Arg2842Cys (class 4) in the highly conserved DBD was reported in homozygous constellation in a patient (score 4) from a consanguineous Turkish family who only suffered from primary ovarian insufficiency and showed slightly increased chromosomal breakage in her cells (54). The combination of c.8524C > T with a frameshift mutation in exon 11 (c.2330dupA) in two sisters (score 2) was associated with mildly elevated levels of chromosomal breakage and severe malformations at birth, but no malignancy or bone marrow failure occurring until the ages of 20 and 23, respectively (72). For this apparent haploinsufficiency of the c.8524C > T allele, it would be highly interesting to analyze how much of the missense BRCA2 protein is required for normal function(s) in cells. Another surprisingly mild genetic alteration in BRCA2 is the splice acceptor mutation c.8488-1G > A (old nomenclature IVS19-1G > A), predicted to result in partial or complete loss of exon 20, which encodes important amino acids in the DBD. However, a male patient with homozygous/biallelic c.8488-1G > A only had thumb abnormalities and no malignancies until 30 years of age (14). Mechanistically, the mutation in the original splice acceptor site leads to the usage of an alternative splice acceptor in exon 20 with an increased Max-Ent-score (from 3.3. to 4.3), leading to the loss of four amino acids, p.Trp2830-Lys2833del, in the DBD at the base of the tower domain of BRCA2 (72). However, the mild chromosomal breakage of the patient’s cells and the mild clinical presentations suggested that the mutant protein produced from two alleles was at least partially functional (14).

A great example of the huge difficulties that can arise, if one wants to classify a change in the BRCA2 coding sequence, is the c.7802A > G alteration. It is located outside of the splice donor site of exon 16 and creates a missense protein, p.Tyr2601Cys. As elegantly shown by Degrolard-Courcet et al. (73), the base pair substitution also creates an alternative splice donor site that competes with the natural donor site located four base pairs downstream. Although the HBond and the MaxEnt scores calculated that both splice sites are very comparable in their activities and therefore usage of the first splice donor site should prevail, the functional analysis revealed, surprisingly, that the downstream natural splice donor site contributed to more than 50% of the BRCA2 message from this allele (74). As it is known for BRCA1, that 20–30% of the normal message is able to maintain normal BRCA1 activity in cells (75), the authors of the functional studies labeled this splice donor site/missense alteration as a VUS (74). We therefore excluded patients with this mutation from our risk score analysis.

Previous reports (14–16) already tried to correlate the severity of BRCA2 mutations/alterations with the occurrence of malignancies, either by the numbers of tumors developed or by the onset/age at diagnosis. While two earlier reports had few patients, a recent report with 71 FANCD1/BRCA2 individuals used an elaborate classification scheme (16), combining multiple in silico tools and online databases (ClinVar, BRCAExchange) with functional testing of several variants to rescue the lethality of Brca2^ko/ko mouse embryonic stem cells in vitro (69). Using the assignments null, deleterious hypomorphic, benign or VUS for each allele, the authors then compared patients with two null alleles to the other FA-D1 patients in their cohort (16). Although a specific genotype-cancer association with predisposition to embryonic tumors (including medulloblastomas, Wilms tumors and neuroblastoma) clearly existed, this classification system still was not able to predict the onset of cancer in this large cohort (16).

In contrast, we established a straightforward scoring system in our ascertained BRCA2-mutated cohort of 75 patients, which is based solely on the genetic alteration in each unambiguously mutated BRCA2 allele. As shown in Figure 4C, patients assigned to these three groups significantly differed at the ages of onset of their malignancies, even between the true ‘null’ patients (score 0) and those with only one hypomorphic allele (score 1). As it is important to prospectively validate any scoring system in an independent cohort (76), we are aware that our approach currently is only an approximation and specific functional assessments for different mutant and normal proteins and splice patterns might be necessary for fine-tuning (26). Although the general risk for FA-D1 patients to develop any malignancy is very high, our algorithm allowed us to distinguish three different risk groups, with scores of 0, 1 and  > 1, that significantly differed in median ages of 1.3, 2.3 and 23.0 years, respectively, at onset of the first malignancy. In addition, while all patients in the risk groups with scores of 0 and 1 experienced their first malignancy until 4.5 and 5.5 years of age, respectively, only roughly 60% of patients (13/22) in the third group (with scores > 1) developed a tumor by 33 years of age. Here, the finding that as many as 31% of these patients (4/13) in the third group experienced a second malignancy is probably a direct consequence of their prolonged life expectancy. Despite an overlap in the tumor types, the tumor incidence for FANCD1 patients with a least one milder allele is still considerably higher compared with healthy heterozygous BRCA2 mutation carriers (77), but distinctive from other FA-D1 patients with little or no residual BRCA2 function (scores 0 and 1).

Conclusion

Using our algorithm, we were able to assign a cohort of 75 FA-D1 patients to three cancer risk groups which presumably have no (score 0), minimal (score 1) or at least moderate (score  > 1) BRCA2 protein function. These three groups clearly differ in the onset of their malignancies and in their OS. Ideally, this assignment should be confirmed experimentally by systematically testing each mutant and maybe also each variant allele in vitro in non-malignant human BRCA2 deficient cells, considering also the impact of the protein levels for each mutant or natural transcript isoform in the assays. It will also be interesting to apply our algorithm to patients from BRCA2 germline mutation pedigrees and to determine if the residual activity of the inherited mutated BRCA2 allele is informative for the onset of malignancies in heterozygous mutation carriers. Finally, reliably predicting the cancer risk for a specific BRCA2 mutation, either in heterozygous or in homozygous configuration, would be an important tool for planning the optimal care for our patients and their families.

Materials and Methods

Study subjects

Published cases with biallelic alterations in BRCA2 were identified using the terms ‘Fanconi anemia’ and ‘BRCA2’ on PubMed. In some cases, additional information was obtained by personal communication with the authors of the publications (see Acknowledgements). In addition, we included data on three unpublished FANCD1/BRCA2-mutated patients from France. From the data available until January 2022, we extracted the following variables: gender (sex), BRCA2 mutation status (HGVS and HGVSp, relative to the reference transcript NM_000059.4), FA chromosome breakage testing, malformations, type(s) of malignancies, age(s) at diagnosis of the malignancies, treatment information, survival status and/or age of death. The term alteration or variant was used for any change found in the genomic sequence of BRCA2, while alterations recognized as pathogenic/associated with impaired function in the ClinVar database were exclusively described as mutations. This retrospective study was approved by the local ethics committee.

The impact of malformations on the clinical phenotype

For classifying the malformations of each patient as limited or extensive, we first used the number of anatomic sites involved as previously established by (47), which was used for scoring malformations in a cohort of transplanted FA patients based on the number of anatomical sites involved, which was based on (46). Skin abnormalities (hypo- and hyperpigmentations, café au lait spots) were not considered to be contributing as they were present in more than 90% of patients. The following were defined as individual anatomical areas: 1) head, including eyes (small eyes, strabismus, epicanthal folds, hypertelorism), ears (deafness, abnormal shape, atresia, dysplasia, low set, canal stenosis, abnormal middle ear), face (microcephaly, micrognathia, triangular face) and neck (short, Sprengel); 2) limbs, including thumbs and radii abnormalities (absent, hypoplastic, short, supernumerary, bifid, or low set for thumbs and absent or hypoplastic for radii), hypoplastic thenar eminence, clinodactyly, polydactyly, absent first metacarpal, abnormal or short fingers and dysplastic ulnae; 3) kidneys; 4) gastrointestinal tract; 5) urogenital tract and 6) cardiovascular system. Even if more than one malformation was observed in one defined anatomic site (e.g. ears, eyes and neck), this site was considered only once when assessing the extent of malformations. Accordingly, a malformative syndrome was considered as extensive if at least three sites were involved, including at least one deep organ (kidneys, gastrointestinal or urogenital tract and cardiovascular system).

Alternatively, we defined congenital malformations as severe if 1) they needed to be operatively corrected immediately after birth, including malformations of the airways (bilateral choanal atresia), gastrointestinal malformations (atresia of esophagus with tracheo-esophageal fistula, duodenal atresia, anal atresia with or without (recto-vesicular/recto-vaginal) fistula, intestinal duplication cysts/mesenteric lymphangioma) and congenital CNS malignancies, or 2) intracranial congenital malformations (holoprosencephaly, cerebellar hypoplasia, tethered spinal cord, abnormal CNS gyration, (partial) corpus callosum agenesia, absence of septum pellucidum, hydrocephalus, cerebral aqueduct stenosis and cerebral ventriculomegaly) or malformations of the axial skeleton (right-convex scoliosis and ribs anomalies, vertebral anomalies, sacral hemivertebra or Sprengel deformity), or 3) polymalformative syndrome/VACTERL syndrome, were present. Skin abnormalities (hypo- and hyperpigmentations, café au lait spots), microcephaly or intrauterine growth restriction were not considered to be distinguishing as they were present in the majority of patients. Patients with malformations as described in 1), 2), or 3) were assigned to the group with extensive malformations, while the group of patients lacking the characteristics described in 1), 2), or 3) were considered to have limited malformations.

Influence of BRCA2 mutations on the clinical phenotype

We created a mutational scoring system (Table 2) with 0–2 points per allele to evaluate the association/influence of type and location of each BRCA2 alteration/mutation on the onset of malignancies and on the OS. Points from both alleles were summarized and correlated with previously mentioned parameters.

Statistical analysis

Statistical analysis and data visualization were performed using R (version 4.0.3; packages: cmprsk 2.2–10, ggfortify version 0.4.11, ggplot2 version 3.3.3, ggpubr 0.4.0, gtsummary version 1.3.5, knitr version 1.30, stats version 4.0.3, survival version 3.2–7, survminer version 0.4.8) (78). OS was defined as the time (in years) from birth until death of any cause or last follow-up of the patient. Date of birth was defined as the date of FA onset. The following endpoints were calculated for all patients: incidence of malignancy and OS time estimated according to the Kaplan–Meier method and compared using log-rank testing. The prognostic factors evaluated were mutation scores, age at the initiation of treatment and age at the onset and type of first malignancy. Treatment was defined as a binary variable (treated or not). P-values < 0.05 were considered significant.

Acknowledgements

We are in debt to Claire Pluchart, Pediatric Hematology Unit, American Mémorial Hospital, CHU Reims, France; Abdulrahman Alsultan, Department of Pediatrics, College of Medicine, King Saud University, Riyadh, Saudi Arabia; Karel Svojgr, Department of Pediatric Hematology and Oncology, Charles University in Prague, Second Faculty of Medicine and University Hospital Motol, Prague, Czech Republic; Jaroslav Sterba, University Hospital Brno Department of Pediatric Oncology, Brno, Czech Republic; Candice Feben, National Health Laboratory Service & The University of the Witwatersrand, Johannesburg, South Africa; Amar Gajjar, St. Jude Children’s Research Hospital, Memphis, USA; and Laurence Faivre, Centre de Génétique, Hôpital d’Enfants, Dijon University Hospital for valuable additional information about the published patients. The authors would like to apologize to all colleagues whose excellent work we could not cite owing to space limitations.

Conflict of Interest statement. The authors declare that they have no conflicts of interest.

Funding

Our research was supported, in part, by the Deutsche Krebshilfe e.V. and the Essener Elterninitiative zur Unterstützung krebskranker Kinder e. V.; by a Junior Clinical Scientist Grant, provided by the Medical Faculty of the University of Duisburg-Essen [I.R.]; by the Centre de Référence Maladie Rares “Aplasies Médullaires”, the Association Française de la Maladie de Fanconi (AFMF), the CONECT-AML (Collaborative Network for Children and Teenagers with Acute Myeloid Leukemia) program (INCa-ARC-LIGUE_11905); by the ANR as “THEMA, the French National Center for Precision Medicine in Leukemia” [L.L, J.S.] and by the Department of Defense grant W81XWH-18-1-0269 [P.R.A.].

References