Genotype-Cancer Association in Patients with Fanconi Anemia due to Pathogenic Variants in FANCD1 (BRCA2) or FANCN (PALB2)

Abstract

Fanconi anemia (FA) is the most common inherited bone marrow failure syndrome and a cancer predisposition disorder. Cancers in FA include acute leukemia and solid tumors; the most frequent solid tumor is head and neck squamous cell carcinoma. FA is a primarily autosomal recessive disorder. Several of the genes in which biallelic pathogenic variants cause FA are also autosomal monoallelic cancer predisposition genes e.g. FANCD1 (BRCA2) and FANCN (PALB2). We observed that patients with FA due to biallelic or homozygous pathogenic variants in FANCD1 and FANCN have a unique cancer association. We curated published cases plus our NCI cohort cases, including 71 patients in the FANCD1 group (94 cancers and 69 variants) and 16 patients in the FANCN group (23 cancers and 20 variants). Only patients in FANCD1 and FANCN groups had one or more of these tumors: brain tumors (primarily medulloblastoma), Wilms tumor and neuroblastoma; this is a genotype-specific cancer combination of tumors of embryonal origin. Acute leukemias, seen in all FA genotypes, also occurred in FANCD1 and FANCN group patients at young ages. In silico predictions of pathogenicity for FANCD1 variants were compared with results from a mouse embryonic stem cell-based functional assay. Patients with two null FANCD1 variants did not have an increased frequency of cancer nor earlier onset of cancer compared with those with hypomorphic variants. Patients with FA and these specific cancers should consider genetic testing focused on FANCD1 and FANCN, and patients with these genotypes may consider ongoing surveillance for the specific cancers.

Introduction

Fanconi anemia (FA) is a rare inherited bone marrow failure and cancer predisposition syndrome associated with defective DNA repair. Many patients with FA have characteristic birth defects, bone marrow failure and a dramatically high risk of cancer. Most patients have biallelic (homozygous or compound heterozygous) inheritance of pathogenic variants in one of at least 22 FA/BRCA DNA repair pathway genes; FANCB is X-linked recessive and FANCR is autosomal dominant [1]. FANCA is the most frequently mutated gene, followed by FANCC and FANCG [1]. Patients with FA are at a significantly increased risk of cancers such as acute myeloid leukemia (AML), head and neck squamous cell carcinoma and others [2]. FANCD1 (BRCA2) and FANCN (PALB2) are also cancer predisposition genes when inherited in heterozygous autosomal dominant form, as are FANCS (BRCA1), FANCO (RAD51C) and FANCJ (BRIP1), all of which are part of the downstream FA/BRCA-related DNA repair pathway [1, 3, 4].

More than 2600 patients with FA have been reported in the literature, with approximately 600 cancers (Alter, unpublished). We focused on FA cases with inheritance of variants in FANCD1 or FANCN. We have included all published or electronically pre-published cases before December 31, 2019 and our unpublished cases from the NCI Cohort (www.marrowfailure.cancer.gov) [2, 5]. We report in vitro analyses of FANCD1 variants seen in a subset of cases [6–8], and in silico analyses for all variants in FANCD1 and FANCN. We compared the FANCD1 variant classifications in silico and in vitro and confirmed the importance of functional testing for variant curation. We determined for the first time that brain tumors, Wilms tumor, and neuroblastoma are almost exclusively seen in patients with FA and FANCD1 and FANCN genotypes, and that these tumors have not been reported in those with other FA genotypes.

Methods

We searched PubMed (https://pubmed.ncbi.nlm.nih.gov) for cases due to homozygous or biallelic variants in FANCD1 or FANCN published through December 31, 2019 using the terms: “Fanconi anemia”, AND “BRCA2”, “PALB2”, “FANCN”, “FANCD1”, “cancer”, “Wilms tumor”, “brain tumor”, “neuroblastoma”. Cases without gene information were not included. FANCD1 variant information was available for all but four cases with phenotype information reported to be due to FANCD1. We added unpublished cases from the National Cancer Institute Inherited Bone Marrow Failure Syndromes cohort (NCT00027274) (See Supplemental Tables 1 and 2).

Variants were converted to current HGVS nomenclature (http://www.hgvs.org) on a consistent transcript (FANCD1, NM_000059.3; FANCN, NM_024675.3). Genomic positions were determined using BRCAExchange (https://brcaexchange.org) and the ClinGen allele registry (http://reg.clinicalgenome.org/redmine/projects/registry/genboree_registry/landing). rsIDs were identified from the dbSNP database (https://www.ncbi.nlm.nih.gov/snp/). ANNOVAR (https://annovar.openbioinformatics.org) and Human Splice Finder (HSF) (http://www.umd.be/HSF3/index.html) were used for in silico predictions [9]. The ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/) and BRCAExchange databases were queried for the presence and pathogenicity of the variants (Supplemental Tables 3 and 4). Analyses and graphs were created using STATA16 (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC.) and Microsoft Excel for Office 365. A subset of FANCD1 variants were evaluated by in vitro methods as described by Biswas et al. [6–8].

Variants were classified as null, deleterious hypomorph, benign or variant of unknown significance (VUS). For FANCD1, variants were assigned as null if one of the following criteria were met: if they failed to rescue the lethality of Brca2^ko/ko mouse embryonic stem cells (mESC) in the functional studies [6], OR the variant was a nonsense stop gain or frameshift OR splice acceptor/donor site variant which was predicted deleterious by HSF. Missense variants were categorized as deleterious hypomorph if one of the following criteria was met: partially rescued the lethality of Brca2^ko/ko mESC and/or viable cells were sensitive to DNA damaging agents [6] OR a missense variant with a ClinVar pathogenic or likely pathogenic designation OR a missense variant with a BRCAExchange pathogenic designation OR a missense variant with a positive in silico score (CADD>25 and REVEL>0.5 and MetaSVM=D) [10–12]. Variants were categorized as benign (neutral) if they completely rescued the lethality of Brca2^ko/ko mESC and cells expressing these variants did not exhibit sensitivity to DNA damaging agents [6]. All of these variants were also benign or likely benign by ClinVar, BRCAExchange and in silico tools. Variants were assigned to be VUS if designated VUS by ClinVar and in silico tools predicted VUS. Results from the mESC based assay took priority over in silico predictions in variant classification. Two variants had an in silico score of null, but functional assay data indicated deleterious hypomorph, and these variants were categorized as hypomorph (highlighted yellow in Supplemental Table 3). Three variants were predicted to be deleterious hypomorph by in silico tools, but by functional assay data were classified as null and thus were scored as null (highlighted green in Supplemental Table 3) [8]. Variants analyzed by Guidugli et al. by homology-directed DNA repair (HDR) assay are noted in Supplemental Table 3, and analyses agreed with the functional assay and in silico scoring [13, 14].

Cases of Fanconi anemia due to FANCD1 variants were classified into subsets based on the combination of variant alleles in each patient: no protein production, i.e. null-null; some protein production, i.e. null-hypomorph or hypomorph-hypomorph; possibly heterozygous i.e. null-benign, benign-hypomorph, VUS-hypomorph or VUS-null; other i.e. benign-benign or unknown. For analysis we compared null-null with all other variant combinations as a group.

For FANCN: 15 of the 20 variants were present in ClinVar and designated as pathogenic or likely pathogenic; they were stop gain, frameshift or splice variants and thus assigned a score of null. Two of the splice variants were predicted deleterious by HSF, one was not, but no protein production was documented by Reid et al. through western blotting [15]. Three variants were not reported in ClinVar and were frameshift or large deletion and thus assigned a null score. One variant was reported by Byrd et al. and these authors demonstrated it to be a partially functional hypomorph (Supplemental Table 4) [16].

Results

Cases

We identified 71 patients with FANCD1 variants and 16 patients with FANCN variants. Sixty-seven of the 71 patients with FANCD1 variants had detailed variant information provided; there were 69 distinct variants (Supplemental Tables 1 and 3). Sixteen patients were identified with FANCN variants and all had variant information documented in the literature, with a total of 20 variants (Supplemental Tables 2 and 4). Ninety-four cancers were documented amongst the patients with FANCD1 variants and 23 cancers in the patients with FANCN variants (Figure 1 and Supplemental Tables 1 and 2). Several reports discussed patients with FANCD1 and FANCN variants but phenotype descriptions were not included. In 2008, Ameziane et al. described a comprehensive genetic subtyping approach combining genetic testing, protein expression and functional testing to improve upon complementation grouping [17]. These authors identified variants in two FANCD1 cases and one FANCN case. Chandrasekharappa et al. described one patient with damaging FANCD1 variants and oropharynx squamous cell carcinoma at 45 years old, but they did not report whether this patient had positive chromosome breakage or other FA phenotypic features [18]. In 2017, Johnson-Tesch et al. reported the central nervous malformations of seven FANCD1 group patients with FA, but no specific variants were included [19]. Cases in these three reports were not included in our analyses due to limited information on genotype and phenotype.

Fig. 1.

Cases of FA due to FANCD1 and FANCN variants were identified from the literature and combined with four unpublished cases. There were 71 patients with FANCD1 variants and 16 with FANCN variants. The number of cancers in each genotype exceed the number of patients because some patients had multiple cancers. The types of cancer were the same in both groups.

Variant Pathogenicity and Type

The majority of the 69 FANCD1 variants in the 71 patients were null (51) or deleterious hypomorph (13). Three variants were predicted benign and two were VUSs (Table 1A). Twenty-one of the 71 variants were evaluated by mESC-based functional assay, and five of the 21 were re-scored based on these results (Supplemental Table 3, highlighted, #6, 10, 51, 52, 61) [6–8]. The most common variant in the cohort (occurred 14 times), a small frameshift deletion, c.658_659delGT (886delGT) had an in silico score of null, but functional assay data showed it to be a deleterious hypomorph, as it generates alternatively spliced transcripts that encode a partially functional protein [8]. A second variant (c.582G>A), a stop gain, had an in silico score of null, but was also shown to be a deleterious hypomorph by functional testing. These two variants are highlighted yellow in Supplemental Table 3. Three variants (c.8009C>T, c.8057T>C, c.9004G>A) were predicted to be deleterious missense by in silico tools, but functional assay data showed they were null (Supplemental Table 3, highlighted green) [7, 8]. Nineteen of 20 FANCN variants were predicted null variants (Supplemental Table 4). The one non-null variant was a deleterious hypomorph described by Byrd et al. [16]. These authors describe that this variant (c.2586+1G>A, p.Thr839_Lys862del) lacked six amino acids and was still able to interact with BRCA2 and retain partial activity.

Table 1.

Classification of FANCD1 and FANCN Variants and Variant Allele Combinations in Patients

	Variants
A. Genotype	Total	Null	Hypomorph	Benign	VUS*
FANCD1 (BRCA2) (71 patients)	69	51	13	3	2
FANCN (PALB2) (16 patients)	20	19	1	0	0
B. Variant Allele Combinations	Null-Null	Null-Hypomorph	Hypomorph-Hypomorph	Null-Benign	Hypomorph-Benign	Hypomorph-VUS	Benign-Benign	Unknown
FANCD1 (BRCA2)	29	29	3	3	0	2	1	4
FANCN (PALB2)	14	2	0	0	0	0	0	0

^*VUS= variant of unknown significance.

For analyses Null-Null group was compared with all other variant combinations as a single group.

Patients with FANCD1 variants were categorized by the combination of alleles that were present (null-null, null-hypomorph etc., see Methods) (Table 1B). These subsets were based on the predicted amount/type of protein produced. Twenty-nine patients had no predicted protein production and were classified as null-null (Table 1B, dark grey). Thirty-five patients had one or both alleles predicted to produce some protein and were hypomorphic based on in silico tools and/or in vitro testing (Table 1B, medium grey). Five patients had only one deleterious allele with the other allele benign or a VUS (Table 1B, light grey). All these patients were confirmed to have FA by positive chromosome breakage. Four patients had unknown alleles and one patient was predicted to be homozygous benign; none of these reports indicated whether the patient was confirmed to have FA based on chromosome breakage (Supplemental Table 1). Fourteen of the 16 patients with FANCN variants were null-null; two patients (siblings) were null-hypomorph (Supplemental Table 2) [16].

A congenital clinical phenotype combination often noted in FA is the VACTERL-H (Vertebral, Anal, Cardiac, Tracheo-esophageal fistula, Esophageal atresia, Renal, upper Limb and Hydrocephalus) association [20]. Six patients with FANCD1 variants had at least three of the eight criteria for VACTERL-H, and five of those six patients had two null variants (Supplemental Table 1). For FANCN, two patients had VACTERL-H criteria and were null-null (Supplemental Table 2). Not all the components of VACTERL-H were described in all the reported cases as either present or absent so it is not possible to determine whether VACTERL-H association is more common amongst patients with two null alleles, but there is a tendency toward that in the FANCD1 group.

Genotype Cancer Association

There was a clear trend in the types of cancers reported in patients with FANCD1 and FANCN variants. Acute leukemia (primarily acute myeloid leukemia) is common among all FA genotypes. Brain tumor, neuroblastoma and Wilms tumor were reported almost exclusively in patients with FANCD1 or FANCN variants (Table 2A). This specific type of cancer association is seen only in patients with FANCD1 and FANCN genotypes, and we termed these the “D1-N associated cancers”. The brain tumors were primarily medulloblastomas (n=21), although one astrocytoma and three glioblastoma were also reported. One patient with a single FANCA variant was reported to have a brain tumor, but the second allele was not documented [21]. This patient may have unidentified biallelic FANCD1 or FANCN pathogenic variants, in addition to the reported FANCA variant, as single pathogenic FA variants are present in the general population [22]. Two cases of neuroblastoma were documented in other FA genotypes, in one patient with biallelic FANCS (BRCA1) and another with biallelic FANCA variants. All three of these patients were confirmed to have FA with positive tests for chromosome breakage [23–25]. Wilms tumor has been reported only in patients with FANCD1 or FANCN genotype. There are six additional reports of patients with FA and Wilms tumor, but the genotypes were not reported in those cases. Five patients with FANCD1 did not have any cancers when reported (last known age 2, 6, 11, 17 and 30 years old) (Supplemental Table 1, Column: No Cancers). Ten other types of cancer were reported in 16 patients with FANCD1, three types of cancer in four patients with FANCN variants, and three types of cancer in six cases with FANCS variants (Table 2B). Colorectal adenocarcinoma was the most frequent other cancer (n=4) (Table 2B). The ages at colorectal cancer ranged from 24–33 years old at diagnosis. After the end of data collection we were made aware of two additional unpublished colorectal adenocarcinoma cases in patients in group FANCD1 (one NCI IBMFS cohort case, one case presented in abstract form).

Table 2.

Cancer Types Among Patients with Fanconi Anemia by Genotype

A. D1-N Associated Cancers	FANCD1 (BRCA2)	Median Age (range) of Cancer*	FANCN (PALB2)	Median Age (range) of Cancer*	FANCS (BRCA1)	FANCA	Other Genes	Gene Unknown**
Total Reported Cases of FA with Cancer(s)	66	2.3 (0–35)	16	1.4 (0.3–12)	7	~300	~400	~1900
Total D1-N Associated Cancers	78	2 (0–16)	19	1 (0.7–4)	2	23		13
Brain Tumor	25	3 (0–13)	7	1.9 (1–4)	0	1***		3
Wilms tumor	16	1.2 (0.4–6.6)	6	1 (0.9–1.4)	0	0		6
Acute Leukemia	31	2 (0.5–16)	4	1.2 (0.9–2)	1	21		~200
Neuroblastoma	6	1 (0.2–1.4)	2	0.8 (0.7–0.9)	1	1		4
B. Other Cancers	FANCD1 (BRCA2)	FANCN (PALB2)	FANCS (BRCA1)	References
Total Reported Cases of FA with Cancer(s)	66	16	7
Total Other Cancers	16	4	6
Colorectal Adenocarcinoma	4			[49]
Rhabdomyosarcoma	3			[21], [50]
Breast	1		4	D1 [49] S [51, 52, 53, 54]
Clear Cell Kidney	1			[55]
Renal Tubular Papillary Adenoma	1			[55]
Hepatoblastoma	1			[56]
Lung	1			[57]
Teratoma	1			[57]
Intra-abdominal Tumor	1			[49]
Lymphoma	2	2		[16, 49, 57]
Kaposiform Hemangioendothelioma		1		[28]
Mosaic Clone		1		[50]
Chronic Lymphocytic Leukemia			1	[53]
Ovarian Cancer			1	[59]
C. FANCD1 (BRCA2)
Amongst Patients with Multiple Cancers (n=23)
	First Cancer	Second Cancer	Third Cancer
Brain Tumor	1	6	1
Wilms Tumor	10	1	1
Acute Leukemia	5	10	1
Neuroblastoma	2	2	0
Other	5	4	2
Amongst Patients with Only One Cancer (n=43)		Amongst All Patients with Cancer	First or Only Cancer
Brain Tumor	17	Brain Tumor	18
Wilms Tumor	4	Wilms Tumor	14
Acute Leukemia	15	Acute Leukemia	20
Neuroblastoma	2	Neuroblastoma	4
Other	5	Other	10
Patients with No Cancers (n=5)
D. FANCN (PALB2)
Amongst Patients with Multiple Cancers (n=6)
	First Cancer	Second Cancer	Third Cancer
Brain Tumor	1	2	1
Wilms Tumor	4	0	0
Acute Leukemia	0	3	0
Neuroblastoma	1	1	0
Other	0	0	0
Amongst Patients with Only One Cancer (n=10)		Amongst All Patients with Cancer	First or Only Cancer
Brain Tumor	3	Brain Tumor	4
Wilms Tumor	2	Wilms Tumor	6
Acute Leukemia	1	Acute Leukemia	1
Neuroblastoma	0	Neuroblastoma	1
Other	4	Other	4
Patients with No Cancers (n=0)

^*In years

^**Some cases had descriptions of phenotype consistent with FANCD1 or FANCN, but no genotype data.

^***This patient only had a single deleterious FANCA variant identified.

Some cases had multiple cancers.

The sequence of multiple cancers is available in Supplemental Tables 1 and 2.

Patients with FANCD1 variants presented with diverse cancer types as the first or only cancer: 27% (18/66) brain tumor, 30% (20/66) leukemia, 21% (14/66) Wilms tumor, 6% (4/66) neuroblastoma, 15% (10/66) other cancer as well as 8% (5/66) none (Table 2C and Supplemental Table 1). However, patients who had multiple cancers primarily presented with Wilms tumor (10/23) first, with the second cancer most frequently acute leukemia (10/23) while the 43 patients with only one cancer tended to have brain tumor or leukemia as a the only cancer (17 and 15 respectively).

Patients with FANCN variants presented with Wilms tumor first and less frequently with acute leukemia (first cancer: 6/16 Wilms tumor, 4/16 brain tumor, 1/16 leukemia, 1/16 neuroblastoma, 4/16 other) (Table 2D and Supplemental Table 2). Overall, combining both genotypes, brain tumor, acute leukemia and Wilms tumor all presented approximately equally as the first cancer: 22, 21 and 20 respectively out of 82 patients with cancer compared with neuroblastoma which was the first cancer in only 5 of 82 across both genotypes.

No difference was seen in the number of D1-N associated cancers in patients with FANCD1 null-null (no protein production) (31 D1-N associated cancers in 27 patients with cancer) and the other variant combinations (some protein production) (47 D1-N associated cancers in 39 patients with cancer) (Table 3A). Twenty-three patients had multiple cancers and 7/23 (30%) had null-null variants compared with 16/23 (70%) who had other variant combinations (Table 3B, Supplemental Table 1). Forty-three patients had only one cancer and 19 of those patients had two null alleles(44%).

Table 3.

Cancer in Fanconi Anemia Patients according to FANCD1 and FANCN Variant Allele Combinations

D1-N Associated Cancers
A. Cancers by Variant Group		Number of Cases with Cancer	Brain Tumor	Wilms Tumor	Acute Leukemia	Neuroblastoma	Total D1-N Associated Cancers	Other Cancers*	None
FANCD1 (BRCA2)	Null-Null	27	11	5	11	4	31	3	3
	Other Variant Combinations**	39	14	11	20	2	47	13	2
FANCN (PALB2)	Null-Null	14	7	6	4	2	19	2	0
	Null-Hypomorph	2	0	0	0	0	0	2	0
B. Cancer Numbers by Variant Group	Number of Patients with:	Multiple Cancers	Single Cancer	No Cancers	Total
FANCD1 (BRCA2)	Null-Null	7	19	3	29
	Other Variant Combinations**	16	24	2	42
FANCN (PALB2)	Null-Null	6	8	0	14
	Null-Hypomorph	0	2	0	2

^*See Table 2 for non- D1-N associated cancers

^**Other Variant Combinations=Hypomorph-Hypomorph, Null-Hypomorph, Null-Benign, Hypomorph-Benign, Hypomorph-Variant of Unknown Significance, Benign-Benign, or Unknown

Most FANCN variants reported were null (19/20) (Table 1A and Supplemental Table 4), and thus most patients were genotypically null-null (14/16) (Table 1B and Supplemental Table 2). All reported patients with FANCN variants had cancer. Ten of these patients had single cancers and six patients had multiple cancers (Table 3B). Two patients with null-hypomorph variants had only one cancer (lymphoma) each (Table 2B). Wilms tumor was the most common first cancer amongst those patients with multiple cancers (Table 2D).

Age and Incidence

The age at onset of solid tumors in FANCD1 or FANCN genotype patients was significantly younger than in patients with other FA genotypes (Figure 2 and [2]). The cases of neuroblastoma and Wilms tumor all occurred in patients younger than five years old in both genotypes. The brain tumors all occurred in patients younger than five years of age with FANCN genotype. Most of the brain tumors in patients with FANCD1 genotype occurred before age six, but two were reported at older ages, 9 (medulloblastoma) and 13 (glioblastoma) years old. Leukemia occurred primarily below age five, particularly with FANCN genotype. Overall, there was not a significant difference in age at the development of D1-N associated cancers between FANCD1 and FANCN genotypes.

Fig. 2.

The age at diagnosis of the D1-N associated cancers in FA due to FANCD1 and FANCN variants. No significant difference is seen (all p > 0.05). A) Brain Tumor. B) Wilms Tumor. C) Acute Leukemia, D) Neuroblastoma. Dark grey, FANCD1. Light grey, FANCN. X is the median.

We hypothesized that the patients with FANCD1 with two null variants would have more cancers or earlier onset of cancers compared with other variant combinations (Figure 3). However, our data did not support this hypothesis. There were no significant (p>0.05) differences among the variant combinations for brain tumor, Wilms tumor, acute leukemia and neuroblastoma. Acute leukemia appeared to have slightly younger onset in patients with two null FANCD1 alleles, compared with those with other alleles, but this was not significant (p=0.92).

Fig. 3.

The age at onset according to the genotypes of the D1-N associated cancers in FA due to FANCD1. A) Brain Tumor B) Wilms Tumor. C) Acute Leukemia. D) Neuroblastoma. N-N= Null-Null; N-H or H-H= Null-Hypomorph or Hypomorph-Hypomorph; N-B, H-B or H-VUS= Null-Benign, Hypomorph-Benign, Hypomorph-VUS; B-B or unknown= Benign-Benign or unknown alleles. No significant difference is seen comparing the variant allele groups. There is a non-significant trend to earlier onset of acute leukemia in the null-null group compared with all other variant groups (p = 0.92).

Discussion

The types and ages at onset of cancer in patients with FA have not previously been strongly correlated with the FA complementation groups. Fiesco-Roa et al. reviewed physical findings and genotypes in literature cases of FA, and found that having two null alleles significantly increased the likelihood of having renal malformations, microcephaly, short stature or VACTERL-H [26]. These authors also determined that patients with FANCB and FANCD2 genotypes had the largest number of physical abnormalities. Jung et al. described a large cohort of FANCB patients and found that those with missense variants had longer overall survival and later onset of hematological symptoms compared with those with whole gene deletions or truncating variants [27].

Here we describe a specific genotype-cancer association in two other FA genotypes. Among all known FA genes FANCD1 and FANCN are the only genotypes with a specific cancer association. Patients with FANCD1 and FANCN genotypes are uniquely predisposed to embryonic tumors: brain tumor (most often medulloblastomas), Wilms tumor and neuroblastoma within the first two decades of life. These types of tumors are observed only among patients with FA who have mutations in FANCD1 and FANCN. This unique combination of tumors may in part be due the interactions of these proteins with cells and their role in both genome stability and the cell cycle. In the cell, FANCN (PALB2) protein interacts with FANCD1 (BRCA2) protein. PALB2 binds BRCA2 at its N-terminus and is necessary for the nuclear sub-localization of BRCA2. PALB2 is also essential for the function of BRCA2 for repair of double strand breaks by homologous recombination [28]. PALB2 deficient cells have decreased levels of BRCA2 due to its role in stabilizing BRCA2 [29]. PALB2 is also required for the maintenance of the G2/M checkpoint during the cell cycle [30]. It is possible that this dual function of BRCA2 and PALB2 in both DNA repair and the cell cycle predisposes patients with FANCD1 and FANCN to a more severe course and a different spectrum of tumors than patients with other FA genotypes. While the somatic alternations in the tumors of FANCD1 or FANCN patients have not been studied, Webster et al. recently presented work on somatic mutation in head and neck squamous cell carcinomas (HNSCC) from other patients with FA [31]. Interestingly, they found that FA HNSCC FA tumors are driven by structural variant gain and have relatively few single nucleotide variants. They noted amplifications of the TAZ and YAP1 in these tumors. YAP1 and TAZ are part of the Hippo signaling pathway which is important in several embryonic tumors including those seen in FANCD1 and FANCN patients [32]. While work remains to fully understand the connections between the FA/BRCA and Hippo pathway, it is an intriguing hypothesis linking genomic instability, structural variant acquisition and embryonic development. It is possible that this link is why these types of tumors arise early FANCD1 and FANCN patients.

We identified 29 patients with biallelic null FANCD1 variants and had hypothesized a priori that these patients would have more numerous cancers and/or earlier onset of cancer, but this was not supported by the evidence as the cancer rate in these patients was not above those with other FANCD1 alleles. Five patients with no cancer had a similar distribution of variant allele combinations as those with cancer. A genotype-phenotype correlation based on null alleles has been described in several other (non-FA) autosomal recessive disorders. Hamo et al. studied autosomal recessive polycystic kidney disease and found that patients with two truncating variants had the most severe perinatal phenotype [33]. Similarly, patients with two nonsense alleles in alpha-L-iduronidase have the most severe phenotype in mucopolysaccharidosis type 1 [34]. Carcao et al. determined that patients with hemophilia A with null mutations had their first adverse event at a younger age, but concluded that variant type made only a small contribution to phenotype [35]. In our analysis, patients with two null FANCD1 alleles were not more likely to have a D1-N associated cancers, nor did the cancers occur at earlier ages.

Survival of patients who are at risk of developing multiple tumors is predicated on surviving the first cancer. Survival with the types of tumors seen in D1-N patients varies greatly in the general non-FA population, with 5-year survival rates between 40 and 90% depending on tumor type and a variety of other host and tumor characteristics. Additionally, the effect of cytotoxic cancer therapies for these cancers on patients with a known genome instability disorder, and subsequently the risk of a second cancer developing, is yet to be quantified. We did note that fourteen patients had Wilms tumor as their first cancer. Four of these had Wilms tumor as their only cancer, while ten subsequently developed a second cancer. Wilms tumor is curable in most affected children but survival is dependent on stage, histology, age at diagnosis and molecular features [36]. Perhaps Wilms tumor is a common first cancer because its overall excellent prognosis favors survival to development of a second cancer, compared with patients who develop a first cancer with overall worse prognosis, such as brain tumors or acute leukemias. We identified a surprising number of colorectal cancer cases (notably in young adults), a risk that was not previously recognized and warrants further investigation. No patients with FANCD1 or FANCN mutations have been reported to have had head and neck or gynecological squamous cell carcinoma, which are common among the other FA genotypes[2]. This may be due to the fact that few of those patients in the D1-N groups survive to the third decade of life when the squamous cell carcinomas are often seen in patients with FA with other genotypes, or because the tumor profile is unique in the FANCD1 and FANCN genotype patients.

Previous reports suggested that patients with FANCD1 variant had brain tumors prior to six years of age [5, 37, 38]. However, we identified two case reports of brain tumors occurring later, at nine and thirteen years of age [39, 40]. This highlights the importance of ongoing surveillance. Prior reports also suggested that patients with two c.6174delT FANCD1 variants were not viable [37]; Steinberg-Shemer and colleagues recently reported a FANCD1 patient who was homozygous for c.6174delT, and had AML at six months old and medulloblastoma at three years old [41]. This variant is also the second most common FANCD1 variant in the reported cases with eight occurrences out of 69 total variants.

Many of the variants reported here were classified based on in silico tools alone, and while the tools continue to improve they will never be as accurate as true functional assay or in vivo/in vitro testing. This was seen in several of the FANCD1 variants for which we have both in silico and functional assay data available (See Supplemental Table 3 highlighted variants). This is particularly important for nonsense variants for which in silico tools often predict a null effect, while in vitro testing may reveal a deleterious hypomorph, such as the 886delGT allele. This allele is predicted in silico to be null, but is a hypomorph as it generates alternatively spliced transcripts that encode a partially functional protein based on in vitro studies [8]. The variant predictions are also based on the assumption that there are no other variants within the gene that modify or regulate the given allele. Ikeda et al. first described a patient in the FANCD1 group with a c.8732C>A allele and a c.8415G>T allele who had AML [42]. Subsequently, Biswas et al. showed that the c.8415G>T was actually neutral in the mESC-based functional assay [6]. Bakker et al. returned to preserved cell lines from this patient and sequenced longer 5’ and 3’ UTR regions [43]. These authors identified a splicing mutation, c.−40+1G>A, which lead to downregulation of FANCD1, which was in fact the disease-causing allele in that patient (Supplemental Tables 1 and 3). This variant analysis highlights the importance of functional testing for variant classification. Five patients in the current report had only one deleterious allele while the other allele was benign or a VUS (Supplemental Tables 1 and 3). It is possible that these patients may be heterozygous carriers, or have a second unidentified deleterious variant, or the variant predictions may be inaccurate, or the patients were clinically misclassified as having FA.

Our study is limited by primarily literature-based case ascertainment. Patients with FA and FANCD1 and FANCN variants who do not have cancer may be less likely to be reported. In addition, reports of multiple cancers in patients with a cancer predisposition syndrome are predicated on the patient surviving the first cancer.

Patients with FA due to FANCD1 and FANCN pathogenic variants warrant regular tumor surveillance including bone marrow aspirate and biopsies, brain magnetic resonance imaging (MRI), abdominal ultrasounds, urine catecholamines, and chest radiograph (https://www.stjude.org/disease/hereditary-neuroblastoma.html). We have proposed a suggested surveillance regimen for FANCD1 and FANCN patients (Table 4) [44, 45]. However, it is important to note that these are based on expert review of the literature and at this time are not evidence based. These recommendations are built from review of median and range of the presentation of these tumors in published cohorts, guidelines for other inherited syndromes and guidelines for individuals with heterozygous pathogenic variants in FANCN and FANCD1. It is important to note that the multiple malignancies that patients with FANCD1 and FANCN FA are predisposed to may not be preventable or curable despite rigorous screening. The benefits of intensive surveillance remain to be proven in patients with FANCD1 and FANCN FA in whom the risk of multiple sequential malignancies is extremely high as are increased toxicities of chemotherapy. The risk benefit ratio of intensive screening as well as preemptive hematopoietic cell transplant should be discussed with the families along with the possibility of prenatal testing before further pregnancies. Prenatal or preimplantation testing with genetic counseling are major components of the care for these families. It is important for all screening to be discussed thoroughly with a patient’s health professional for an individualized assessment of the risks and benefits of tumor surveillance. This is of particular importance for neuroblastoma where it has been shown that in the general population there is no reduction in mortality with early detection [46, 47]. In contrast to prior literature, we have suggested that brain tumor surveillance might continue beyond six years of age, since the oldest case reported was age 13 years [39]. Patients with FA due to FANCD1 (BRCA2) variants should be followed according to the current National Comprehensive Cancer Network (NCCN) guidelines for BRCA2 heterozygous individuals as they enter adulthood including clinical breast exam, breast MRI, mammography, discussion of risk-reducing surgery and prostate cancer screening (https://www.nccn.org/), and patients with FA due to FANCN (PALB2) should follow similar guidelines as recently described by Tischkowitz et al. [48]. The uncertain risk of colorectal cancer in young adulthood might be discussed, and possible early surveillance, such as colonoscopy, could be considered, pending further data.

Table 4.

Suggested^* Surveillance for FANCD1 and FANCN Patients

Cancer	Testing/Imaging	Frequency	Age**
All	History and physical exam	Yearly	All
Acute Myeloid Leukemia	Complete Blood Count	Every 3–4 months or with symptoms	All
	Bone Marrow Aspirate and Biopsy	Yearly or with a change in CBC	1 yo. +
Wilms Tumor and Neuroblastoma	Abdominal Ultrasound	Every 6 months	6 mo. – 10 yo.
Neuroblastoma	Urine Catecholamines	Every 6 months	6 mo. – 5yo.
Brain Tumor	Brain MRI	Yearly	1–16 yo.
Colorectal Cancer	Fecal Occult Blood and Colonoscopy	Clinician discretion	18 y.o +
Adult Tumors***
Breast Cancer	Breast MRI	Yearly	25 yo. +
	Mammography	Yearly (6 months alternating with MRI)	30 yo. +
Ovarian Cancer	Transvaginal Ultrasound and serum CA-125	Clinician discretion	30 yo. +
Pancreatic Cancer	Magnetic resonance cholangiopancreatography	Clinician discretion	50 yo. +
Melanoma	Screening exam with dermatologist	Yearly	18 yo. +
Prostate Cancer	Digital Rectal Exam	Yearly	40 yo. +

^*These are only suggested guidelines based on expert opinion and are at this time are not evidence-based screening

^**yo. years old, mo. months old

^***Based on the National Comprehensive Cancer Network (NCCN) guidelines for those with monoallelic pathogenic variants

We have now identified a genotype-cancer association (brain tumor [particularly medulloblastoma], Wilms tumor, and neuroblastoma) seen only in patients with Fanconi anemia due to FANCD1 and FANCN pathogenic variants, which has implications for genetic testing, patient identification and cancer surveillance. Patients with FA and FANCD1 or FANCN variants might consider cancer surveillance at very young age, particularly focused on the unique D1-N associated cancers. Physicians treating patients with the D1-N associated cancers should be suspicious about the possibility of an FA diagnosis. Families with FANCD1 and FANCN variants need thorough genetic counseling and cancer screening for both the patient with FA and the heterozygous family members who are at risk of autosomal dominant BRCA pathway related cancers.

Highlights.

• There is a unique genotype-cancer association in FANCD1 and FANCN Fanconi anemia.

• FANCN and FANCD1 Fanconi anemia patients often develop embryonic tumors early in life.

• Mouse embryonic stem cell-based functional testing improves FANCD1 variant curation.