Breast cancer patients suggestive of Li-Fraumeni syndrome: mutational spectrum, candidate genes, and unexplained heredity

Eighty-three unrelated study subjects were selected from among a pool of female patients with BC who conformed to the German consortium criteria for HBOC [18] and whose pedigree had been established during genetic counseling in the tumor genetics outpatient clinic at Hannover Medical School, Germany, between 2002 and 2015. Prior to study inclusion, all patients had a negative test result for pathogenic single-nucleotide variants or small indels (Sanger sequencing) as well as gross genomic rearrangements (multiplex ligation-dependent probe amplification [MLPA], SALSA MLPA kits [MRC Holland, Amsterdam, The Netherlands], P002B/P087 for BRCA1, P045-B3/P077 for BRCA2, and P056-C1 for TP53) within the genes BRCA1, BRCA2, and TP53. Final inclusion of study participants was based on personal medical or family history characteristics suggestive of Li-Fraumeni-like traits and conformance to at least one of the hitherto existing Li-Fraumeni-related criteria (LFS/LFL/Chompret criteria). Classification into LFS/LFL and Chompret criteria was performed either on the particular proband undergoing mutation analysis or on another index case within the same family. The majority of patients were assumed to be of European-Caucasian ancestry. All patients signed informed consent forms, and the project was approved by the research ethics committee of Hannover Medical School (approval number 3528).

Among our cohort of 83 patients with BC, 50 unrelated women had early-onset BC (age at diagnosis ranging from 19 to 34 years, median age 29 years), 9 index cases were diagnosed with bilateral BC or two primary independent breast carcinomas, 4 individuals harbored at least one independent neoplasm besides BC, in 10 families either the index patient or one of her relatives had a sarcoma, and in 16 families either the index patient or a relative was diagnosed with a brain tumor (multiple entries being possible).

One woman (patient 58), who is additionally affected by the class 4 splice donor site variant RECQL4:c.1390+1G>C, was identified to carry a CDKN2A/p14^ARF nonsense variant in exon 2 of 3 total exons [CDKN2A:p.(Arg98*)]. The CDKN2A locus encodes two distinct proteins, p14^ARF and p16^INK4, that are defined by translating the common second exon in alternate reading frames. The variant we detected is not predicted to affect expression or function of p16^INK4, a tumor suppressor implicated in the CDK4/CDK6-RB1 pathway; instead, p14^ARF expression is expected to be abolished via nonsense-mediated RNA decay (NMD). The protein p14^ARF acts upstream of TP53 by binding directly to MDM2, an E3 ubiquitin ligase controlling the activity and stability of TP53. Because p14^ARF promotes MDM2’s degradation, TP53 is stabilized and accumulates, and a TP53 response manifests in elevated levels of p21^CIP1, inducing cell cycle arrest [46]. PVs in human p14^ARF exon 2 have been reported to disrupt its nucleolar localization and impair its ability to block nuclear export of MDM2 and TP53 [47]. The exon 2 variant in our cohort was identified in an index patient whose grandfather was diagnosed with hypopharyngeal carcinoma, which is in line with CDKN2A PVs predisposing to tobacco-related cancers such as orolaryngeal cancer, next to its predominant role in hereditary melanoma, pancreatic cancer, and further tumor entities [MIM:600160]. Of note and with particular relevance to LFS and LFL, germline CDKN2A PVs have recently been demonstrated to account for a subset (8 of 190) of hereditary sarcoma cases negative for germline TP53 PVs [17]. Conversely, however, sarcomas are rare in CDKN2A mutation carriers, and the affected family of our cohort also does not present with sarcoma cases. In a somatic context, CDKN2A PVs have previously been associated with the mutational landscape of BC [48].

In index patient 30, we detected a splice donor site variant in RUNX1 (RUNX1:c.97+1G>A) predicted deleterious by 4 of 5 splice prediction programs owing to complete loss of the native splice site. As of yet, this variant has not been described in the literature and was classified by us as a class 4 variant on the basis of in silico prediction. RUNX1 is a gene well known for its crucial role in the hematopoietic system and for its association with sporadic and familial leukemia [49]. The above-mentioned variant was identified in a family exclusively struck by BC. Of note, no hematologic abnormalities were reported in this family at the time of consultation. Even though the variant was detected in 43% of the sequencing reads (VAF 43%, depth 400 ×), we cannot exclude the possibility that this variant might be due to clonal hematopoiesis (i.e., undiagnosed hematologic disorder). In the case of a confirmed germline event, functional splicing assays would be needed to confirm the disruptive nature of the variant. In regard to BC pathology, RUNX1 has been suggested to be largely understudied [50], and genome-wide sequencing of cohorts of patients with BC have subsequently exposed RUNX1 as one of the most frequently mutated and/or deleted genes in BC in a somatic setting [51, 52]. Its role in BC progression has been related to estrogen signaling but remains elusive, likely including oncogenic rather than tumor-suppressive functions in mammary epithelial cells [53‐55]. A conclusive role for RUNX1 germline variations in BC pathogenesis cannot be assessed currently and needs further clarification.

The nonsense FANCI variant FANCI:p.(Arg1285*) in exon 37 of 38 total exons, located within the last 72 bp of the second to last exon and therefore imprecise in regard to NMD, was detected in patient 65, who was additionally found to carry a deletion of exons 3–8 of the PMS2 gene, which was verified by MLPA analysis. Because the maternal lineage includes two colorectal carcinomas in two of the index person’s second-degree relatives, diagnosed at 37 and 70 years of age, the PMS2 variant is primarily suggestive of being responsible for the familial cancer phenotypes. Nonetheless, the FANCI variant concerns a highly conserved arginine, has previously been described in patients with Fanconi anemia (FA), and has been found to impair DNA binding and ubiquitination of the ID2 complex, a dimeric complex formed by FANCI and FANCD2, which is relevant for DNA crosslink repair [56, 57]. Because the paternal lineage also shows incidence of malignant disease, it is conceivable that the FANCI and PMS2 variants may jointly combine in the index person’s genome, leading to a more severe phenotype, such as earlier age of onset of disease.

Furthermore, 6 patients of our cohort were identified to carry 5 very rare FANCA VUS: FANCA:p.(Trp22Gly) twice, FANCA:p.(Pro497Ala), FANCA:p.(Pro667Arg), FANCA:p.(Arg1144Trp), and FANCA:p.(Leu1230Val). All but one of these rare FANCA VUS were absent in a large cohort of 11,000 exome-sequenced individuals of HMGU (Helmholtz Zentrum München – Deutsches Forschungszentrum für Gesundheit und Umwelt) and are either not or very rarely found in datasets from the Genome Aggregation Database [58] (see Table 4 for details). Interestingly, clinical presentation of these heterozygous FANCA missense VUS carriers comprises some of the most striking LFL features, involving either very early-onset BC or exceptionally Li-Fraumeni-suggestive personal and family traits, such as a woman with triple primaries (malignant hemangiopericytoma at age 12 years, BC at age 28 years, and contralateral BC at age 47 years) or a woman whose sister and half-brother were diagnosed with two liposarcomas and a melanoma, respectively. The median age of onset of BC disease for these 6 FANCA VUS carriers was 29.5 years (27, 27, 28, 31, 35, and 44 years), whereas the median age of onset for the entire cohort was 33 years. In a biallelic context, the Fanconi family genes are associated with FA, a condition characterized by congenital abnormalities, bone marrow failure, and cancer predisposition already during childhood. FANCA gene PVs are by far the most common in FA, responsible for at least 60% of all cases of FA [59]. Heterozygous parents and siblings of patients with FA have not been found to exhibit an elevated incidence of malignant disease [60]; however, monoallelic FANCA variants have been investigated only sparsely in a disease context, and with inconsistent results [61‐63]. Considering the rarity of each FANCA VUS detected in the present study, and given the possibility of differing FANCA missense variants conferring vastly diverse effects on protein function, the disproportionate occurrence of these rare variants appears to be worthy of further investigation.

Two class 4–5 aberrations (patients 59 and 79) and one class 3 copy number change (patient 63) were identified in WRN, RECQL4, and BLM — 3 of 5 genes belonging to the human family of RecQ helicases, comprising enzymes that drive the unwinding of DNA in an ATP- and Mg²⁺-dependent manner and which are essential for genome maintenance and stability. In an autosomal recessive setting, each of these genes confers very rare and complex syndromes, namely Werner syndrome (WRN), Bloom syndrome (BLM), and Rothmund-Thomson syndrome or RAPADILINO syndrome (RECQL4). In regard to LFS, in which sarcoma incidence is being considered somewhat of a hallmark within the tumor spectrum, a particularly striking fact about the above-mentioned syndromes is that the spectrum of malignancies conferred by them is dominated by sarcomas as well [64, 65]. Moreover, for all three proteins, a firm association with TP53 has been described [65, 66]. Whereas biallelic impairment is considered mandatory for the full spectrum of the syndromes to develop, WRN and BLM have additionally been suggested as BC susceptibility genes in a monoallelic setting [67‐69], and dominant-negative effects or gain-of-function processes have previously been proposed for specific missense mutant WRN or BLM proteins [70].

To our knowledge, this is the first study focusing on a German population-based cohort of familial BC patients suggestive of LFL but negative for causative TP53 as well as BRCA1/2 germline PVs to be systematically panel-tested for aberrations in further cancer susceptibility genes. One hypothesis for the emergence of an LFL phenotype without detectable TP53 PV is the simultaneous and additive occurrence of dual or multiple clearly pathogenic aberrations in more than one cancer susceptibility gene. Our study design was able to confirm double heterozygosity in 3 cases. In regard to patients with BC, double heterozygosity for germline PVs in BC predisposition genes has been detected and discussed before [67]. This type of oligogenic or polygenic model might explain the fact that often the index patients of our cohort happen to be burdened by cancer incidences from both family lineages. The model would also explain why the family mode may not follow a strict autosomal dominant inheritance pattern, which is usually considered a hallmark of classic LFS.

With the current state of knowledge and excluding all VUS, the presently described aberrations, occurring in 10 (12%) of our cohort’s individuals, may in part explain or contribute to some of the corresponding phenotypes. However, the gravity and the spectrum of malignant disease in many of our cohort’s families seem to suggest that additional PVs in modifying genes or as yet unknown cancer predisposition genes may play a role in some of them. Importantly, these results indicate that even after testing a fairly extensive number of cancer susceptibility genes with a panel design covering both sequence alterations and copy number changes, a large number of LFL BC cases remain unexplained. Few studies have approached the quest for further susceptibility genes in individual TP53-negative LFS/LFL cases, either by specifically testing single genes associated with the spectrum of tumors appearing in LFS or by performing whole-exome sequencing. Associations have been suggested for LFS-associated brain tumors with nonsense PVs in CASP9 (caspase-9) [71], for a POT1 (protection of telomeres 1) missense variant with cardiac and breast angiosarcomas [72], and, as mentioned above, for CDKN2A PVs with hereditary sarcoma cases [17].

Apart from the idea of aberrations in additional susceptibility genes being responsible for TP53-negative LFL cases, novel mechanisms for TP53 impairment are also emerging. These include structural variants that cannot easily be detected by conventional approaches such as intron 1 rearrangements [73], variants in the far-off 3′ UTR affecting microRNA binding [74] (a region not covered by the Illumina gene panel used in our study), and novel splicing PVs [75]. Whereas the severe phenotype known for LFS is expected to result predominantly from missense or isoform-specific variants conferring a dominant-negative effect of the mutated over the wild-type proteins, TP53 haploinsufficiency (e.g., loss of function or gene dosage effects due to splice variants or 3′ UTR variants and deletions, respectively) is emerging as a likely mechanism for a more subtle phenotype, one that may in fact be described as LFL, as has previously been suggested [75]. Testing patients without variants in coding TP53 regions, such as the ones in our cohort, for the above-mentioned variants would constitute a crucial future undertaking.

The strengths of this study include a well-characterized cohort of LFL BC patients and the dual-method study design of NGS and array-based CGH to cover a larger number of aberrations. Limitations include the following:

1. Segregation analyses would be mandatory in order to characterize the impact of the detected variants in affected and unaffected family members and to predict the penetrance associated with these variants; unfortunately, DNA was not available for any of the family members of affected germline mutation carriers.

3. Functional assays would be helpful in determining the consequences of the detected VUS and novel aberrations.