A Novel Germline Mutation in BAP1 Predisposes to Familial Clear-Cell Renal Cell Carcinoma

Abstract

Renal cell carcinoma (RCC) clusters in some families. Familial RCC arises from mutations in several genes, including VHL, which is also mutated in sporadic RCC. However, a significant percentage of familial RCC remains unexplained. Recently, we discovered that the BAP1 gene is mutated in sporadic RCC. BAP1, which encodes a nuclear deubiquitinase, is a two-hit tumor suppressor gene. Somatic BAP1 mutations are associated with high-grade ccRCC and poor patient outcomes. To determine whether BAP1 predisposes to familial RCC, we sequenced the BAP1 gene in 83 unrelated probands with unexplained familial RCC. We identified a novel variant (c.41T>A; p.L14H), which cosegregated with the RCC phenotype. The p.L14H variant disrupts a highly conserved residue in the catalytic domain, a domain frequently targeted by missense mutations. The family with the BAP1 variant was characterized by early-onset clear cell RCC, occasionally of high Fuhrman grade, and lacked other features that characterize von Hippel-Lindau syndrome. These findings suggest that BAP1 is a familial RCC predisposing gene.

Introduction

Approximately 5% of all renal cell carcinoma (RCC) is familial (1). Several genes, including VHL, MET, FLCN, FH and genes encoding the SDH subunits B/C/D have been identified as causative (2-4). However, the genetic basis of a significant percentage of familial RCC remains unknown. There is precedent for genes mutated in the germline (i.e. VHL) that are also mutated in the sporadic setting, and thus somatically mutated genes may explain familial RCC if mutated in the germline.

BAP1 (BRCA1 associated protein-1) is a tumor suppressor gene that encodes a nuclear deubiquitinase (5-7). BAP1 functions as a classic two-hit tumor suppressor gene and is somatically mutated in uveal melanoma and mesothelioma (8, 9). Somatic mutations in BAP1 have also been recently identified in renal cell carcinoma (RCC) of clear cell type (ccRCC) (10, 11). We found that BAP1 is inactivated in approximately 15% of sporadic ccRCCs, and that BAP1 mutations are associated with high Fuhrman grade and poor patient survival (11, 12).

Materials and Methods

Patient Samples

Eighty-three unrelated individuals with a predisposition to RCC defined as early onset RCC, multifocal or bilateral tumors, and/or a family history of RCC, were analyzed. Peripheral blood samples were obtained from UT Southwestern Medical Center (UTSW) (n=6), UT Health Science Center at San Antonio (UTHSCSA) (n=4), Cleveland Clinic (n=26), and the National Cancer Institute (NCI) (n=47) (Table 1). De-identified samples of germline DNA were provided by the different institutions for BAP1 sequencing under a protocol approved by the UTSW Institutional Review Board (IRB). Patients from other institutions were recruited under their own IRB protocols.

Table 1.

Clinical and Pathological Characteristics for Probands by Institution

		UTSW	UTHSCSA	Cleveland	NCI	Total
		(n = 6)	(n = 4)	(n = 26)	(n = 47)	(n = 83)
Gender
	Male	4/6 (67%)	2/4 (50%)	7/26 (27%)	29/47 (62%)	42/83 (51%)
	Female	2/6 (33%)	2/4 (50%)	19/26 (73%)	18/47 (38%)	41/83 (49%)
Race
	Caucasian	6/6 (100%)	4/4 (100%)	20/23 (87%)	40/47 (85%)	70/80 (88%)
	Asian	0/6 (0%)	0/4 (0%)	1/23 (4%)	3/47 (6%)	4/80 (5%)
	African American	0/6 (0%)	0/4 (0%)	0/23 (0%)	2/47 (4%)	2/80 (3%)
	American Indian	0/6 (0%)	0/4 (0%)	1/23 (4%)	0/47 (0%)	1/80 (1%)
	Other	0/6 (0%)	0/4 (0%)	1/23 (4%)	2/47 (4%)	3/80 (4%)
Ethnicity
	Hispanic	1/6 (17%)	2/4 (50%)	2/21 (10%)	1/46 (2%)	6/77 (8%)
	Non-Hispanic	5/6 (83%)	2/4 (50%)	19/21 (90%)	45/46 (98%)	71/77 (92%)
Mean Age at Dx (range)		49.7 (24- 70)	48.3 (26- 59)	54.5 (30-79)	52.2 (33- 75)	52.5 (24- 79)
Laterality
	Right	5/6 (83%)	1/2 (50%)	9/18 (50%)	18/47 (38%)	33/73 (45%)
	Left	1/6 (17%)	0/2 (0%)	8/18 (44%)	15/47 (32%)	24/73 (33%)
	Bilateral	0/6 (0%)	1/2 (50%)	1/18 (6%)	14/47 (30%)	16/73 (22%)
Focality
	Unifocal	5/5 (100%)	2/2 (100%)	14/17 (82%)	28/47 (60%)	49/71 (69%)
	Multifocal	0/5 (0%)	0/2 (0%)	3/17 (18%)	19/47 (40%)	22/71 (31%)
Histology
	Clear Cell	6/6 (100%)	3/3 (100%)	12/17 (71%)	47/47 (100%)	68/73 (93%)
	Papillary	0/6 (0%)	0/3 (0%)	1/17 (6%)	0/47 (0%)	1/73 (1%)
	Chromophobe	0/6 (0%)	0/3 (0%)	1/17 (6%)	0/47 (0%)	1/73 (1%)
	Oncocytic	0/6 (0%)	0/3 (0%)	1/17 (6%)	0/47 (0%)	1/73 (1%)
	Transitional Cell	0/6 (0%)	0/3 (0%)	1/17 (6%)	0/47 (0%)	1/73 (1%)
	Tubulopapillary	0/6 (0%)	0/3 (0%)	1/17 (6%)	0/47 (0%)	1/73 (1%)
Fuhrman Grade
	I	0/3 (0%)	0/1 (0%)	3/15 (20%)	2/45 (4%)	5/64 (8%)
	II	1/3 (33%)	0/1 (0%)	8/15 (53%)	35/45 (78%)	44/64 (69%)
	III	2/3 (67%)	0/1 (0%)	2/15 (13%)	5/45 (11%)	9/64 (14%)
	IV	0/3 (0%)	1/1 (100%)	2/15 (13%)	3/45 (7%)	6/64 (9%)
Mean Tumor Size (range)		5.9 (3.3- 8.0)		4.4 (1.5- 10.0)	5.1 (1.2- 14.5)	5.0 (1.2- 14.5)
pT
	1	3/5 (60%)	0/1 (0%)	12/17 (71%)	31/47 (66%)	46/70 (66%)
	2	0/5 (0%)	0/1 (0%)	2/17 (12%)	9/47 (19%)	11/70 (16%)
	3	2/5 (40%)	0/1 (0%)	3/17 (18%)	6/47 (13%)	11/70 (16%)
	4	0/5 (0%)	1/1 (100%)	0/17 (0%)	1/47 (2%)	2/70 (3%)
pN
	0	1/2 (50%)	1/1 (100%)	2/2 (100%)	9/10 (90%)	13/15 (87%)
	1	1/2 (50%)	0/1 (0%)	0/2 (0%)	1/10 (10%)	2/15 (13%)
M
	0	3/3 (100%)		2/2 (100%)	37/47 (79%)	42/52 (81%)
	1	0/3 (0%)		0/2 (0%)	10/47 (21%)	10/52 (19%)
Other Tumors (Proband)
	Breast	0/6 (0%)	0/4 (0%)	11/26 (42%)	0/47 (0%)	11/83 (13%)
	Thyroid	0/6 (0%)	0/4 (0%)	6/26 (23%)	4/47 (9%)	10/83 (12%)
	Prostate	0/6 (0%)	0/4 (0%)	0/26 (0%)	3/47 (6%)	3/83 (4%)
	Uterus	0/6 (0%)	0/4 (0%)	2/26 (8%)	0/47 (0%)	2/83 (2%)
	Thymoma	0/6 (0%)	0/4 (0%)	1/26 (4%)	1/47 (2%)	2/83 (2%)
	Pancreas	1/6 (17%)	0/4 (0%)	0/26 (0%)	0/47 (0%)	1/83 (1%)
	Bladder	1/6 (17%)	0/4 (0%)	0/26 (0%)	0/47 (0%)	1/83 (1%)
	Esophagus	1/6 (17%)	0/4 (0%)	0/26 (0%)	0/47 (0%)	1/83 (1%)
	Ovary	0/6 (0%)	0/4 (0%)	1/26 (4%)	0/47 (0%)	1/83 (1%)
	Cervix	0/6 (0%)	0/4 (0%)	1/26 (4%)	0/47 (0%)	1/83 (1%)
	Lung	0/6 (0%)	0/4 (0%)	0/26 (0%)	1/47 (2%)	1/83 (1%)
	Pheochromocytoma	0/6 (0%)	2/4 (50%)	0/26 (0%)	0/47 (0%)	2/83 (2%)
	Melanoma	0/6 (0%)	0/4 (0%)	1/26 (4%)	1/47 (2%)	2/83 (2%)
	Paraganglioma	0/6 (0%)	1/4 (25%)	0/26 (0%)	0/47 (0%)	1/83 (1%)
	Carcinoid	0/6 (0%)	0/4 (0%)	0/26 (0%)	1/47 (2%)	1/83 (1%)
Familial RCC
	1st Degree Relatives
	0	4/6 (67%)	2/4 (50%)	13/26 (50%)	10/47 (21%)	29/83 (35%)
	1	0/6 (0%)	2/4 (50%)	13/26 (50%)	28/47 (60%)	43/83 (52%)
	2	2/6 (33%)	0/4 (0%)	0/26 (0%)	6/47 (13%)	8/83 (10%)
	3	0/6 (0%)	0/4 (0%)	0/26 (0%)	3/47 (6%)	3/83 (4%)
	2nd Degree Relatives
	0	5/6 (83%)	4/4 (100%)	14/26 (54%)	29/47 (62%)	52/83 (63%)
	1	1/6 (17%)	0/4 (0%)	12/26 (46%)	12/47 (26%)	25/83 (30%)
	2	0/6 (0%)	0/4 (0%)	0/26 (0%)	6/47 (13%)	6/83 (7%)
Germline Mutation Testing *
	VHL	6/6 (100%)	4/4 (100%)	2/26 (8%)	22/47 (47%)	34/83 (41%)
	SDHB	0/6 (0%)	4/4 (100%)	12/26 (46%)	19/47 (40%)	35/83 (42%)
	SDHC	0/6 (0%)	0/4 (0%)	11/26 (42%)	19/47 (40%)	30/83 (36%)
	SDHD	0/6 (0%)	0/4 (0%)	11/26 (42%)	19/47 (40%)	30/83 (36%)
	PTEN	0/6 (0%)	0/4 (0%)	25/26 (96%)	1/47 (2%)	26/83 (31%)
	BRCA1/2	0/6 (0%)	0/4 (0%)	5/26 (19%)	0/47 (0%)	5/83 (6%)
	MET	0/6 (0%)	0/4 (0%)	1/26 (4%)	3/47 (6%)	4/83 (5%)
	FLCN	0/6 (0%)	0/4 (0%)	0/26 (0%)	7/47 (15%)	7/83 (8%)
	FH	0/6 (0%)	0/4 (0%)	0/26 (0%)	4/47 (9%)	4/83 (5%)
	TSC1/2	0/6 (0%)	0/4 (0%)	0/26 (0%)	3/47 (6%)	3/83 (4%)

Mean size (cm), grade and stage were determined on the largest tumor per patient. Dx, diagnosis; pT and pN, pathologic tumor and node stage according to American Joint Committee on Cancer, 2010 edition; M, clinical metastases.

^*Germline mutation testing refers to the number of individuals tested for mutation in each indicated gene. All genetic tests were negative.

DNA extraction and BAP1 Sequencing

Germline DNA extraction (UTSW samples) and sequencing were performed as previously described (11). For BAP1 loss of heterozygosity studies, DNA from fresh frozen (or microdissected formalin fixed paraffin embedded) tumor tissue was extracted using Maxwell 16 Tissue DNA Purification Kit or Maxwell 16 FFPE Tissue LEV DNA Purification Kit (Promega, Madison, WI). Redesigned BAP1 primers were used for PCR amplification and sequencing of microdissected tumor DNA: forward primer 5′-GCCTGCCTGACCATCACC and reverse primer 5′-AAGGAAAGCAGTAGGGAAGGA.

BAP1 Immunohistochemistry

Five micron formalin-fixed paraffin-embedded sections were deparaffinized and blocked with methanol-30% H₂O₂. After antigen retrieval by boiling in citrate buffer, slides were incubated with monoclonal anti-BAP1 antibody (C-4; Santa Cruz Biotechnology, TX) diluted 1/150. Then, slides were immunostained with avidin-biotin-peroxidase complex and developed with diaminobenzidine. Harris’ hematoxylin was used to counterstain the slides. Non-immune mouse immunoglobulin was used as a negative control. Expression was evaluated as positive or negative. Staining was considered positive when more than 10% of the nuclei showed immunoreaction.

Results and Discussion

During studies that led to the discovery of somatic BAP1 mutations in ccRCC, a germline variant (c.121G>A; p.G41S) was identified in one individual (II:1) that had two first-degree and a second-degree relative with RCC and who had previously tested negative for a VHL mutation (Supplementary Figure S1) (11). The variant (c.121G>A; p.G41S) was not found in the 1,000 Genomes Project (13). However, cosegregation studies suggested that it was not responsible for the cancer predisposition in this family (Supplementary Figure S1). Nevertheless, BAP1 mutations have been previously observed in the germline, where they predispose to uveal and cutaneous melanoma as well as mesothelioma (14-18). Some of these families also exhibited other tumor types, although at much lower frequencies, including RCC (16, 17). On the basis of the somatic and germline associations of BAP1 mutations and RCC, we sought to determine whether germline BAP1 mutations were associated with familial RCC.

We examined 83 unrelated probands with familial RCC. Phenotypic information is summarized in Table 1. The samples were predominantly from Caucasian individuals of non-Hispanic origin. The mean age at diagnosis was 52.5 years. 22% of the individuals had bilateral tumors and RCC was multifocal in 31% of individuals. 93% of the individuals examined had RCC of clear cell type. The mean size of the largest tumor in each individual was 5 cm. Most of these tumors were pT1 (66%) and over 75% were of low Fuhrman nuclear grade (Grade I-II). Lymph node and distant metastases were rare (13% and 19% respectively). Other primary tumors reported in probands included those of the breast, thyroid, and prostate. 66% of individuals had at least 1 first degree relative with RCC, and 37% had at least 1 second degree relative with the disease. Individuals in this cohort had been previously evaluated for mutations in other RCC predisposition genes. 41% had been evaluated for germline VHL mutations and had tested negative. A similar percentage had tested negative for germline mutation of the SDH complex genes. 31% of the individuals had tested negative for germline PTEN mutations. Fewer individuals had been screened for germline mutations in a variety of other genes including BRCA1, BRCA2, MET, FLCN, FH, TSC1 and TSC2. In all patients, the cause of the familial RCC predisposition remained unknown.

We sequenced the BAP1 gene in peripheral blood DNA from the 83 probands. The coding sequence and intron/exon junctions were analyzed by Sanger sequencing as previously described (11). One out of the 83 DNA samples failed to sequence. Among the rest, a high quality sequence tracing in at least one direction was obtained for 100% of amplicons. Only two samples were detected with germline variants. A germline missense variant (c.869A>G; p.N290S) was found in a UTSW female with unilateral, unifocal, ccRCC diagnosed at age 24 who had previously tested negative for germline mutations in VHL and PTEN. The variant was predicted to be benign by PolyPhen-2 (19) and conservation studies across species showed that the corresponding asparagine was replaced by a serine in some species. Thus, the variant was not analyzed further.

A second germline missense variant (c.41T>A; p.L14H) was detected in an NCI proband. This individual belonged to Family NCI-1326, which included 5 individuals diagnosed with RCC. This kindred is characterized by early onset and aggressive clear cell RCC. Of the 5 affected members, three died at 36, 48 and 58 years of age (Figure 1). The proband (IV:1) underwent a right radical nephrectomy for a multifocal RCC at age 44 at an outside hospital. Pathologic evaluation revealed an 8.0 cm solid, high grade (Fuhrman nuclear grade III) ccRCC as well as three additional smaller cystic lesions with clear cell kidney cancer. In addition, an angiomyolipoma and a renal medullary fibroma were found. Because of the family history and multifocality, the patient was referred to NCI.

Figure 1. Family NCI-1326: an early onset, aggressive form of bilateral, multifocal solid and cystic clear cell kidney cancer.

Pedigree from familial renal cancer kindred NCI-1326. The proband IV:1 was initially diagnosed with kidney cancer at the age of 44. Individual IV:4 was diagnosed with RCC at 40 and individual IV:6 died of metastatic RCC at age 36. Individuals II:1 and III:4 died of metastatic RCC at 48 and 58 years of age, respectively. Two individuals with a history of RCC for whom samples were available (IV:1 and IV:4) had a germline BAP1 variant. Two other individuals (IV:3 and IV:5) who were screened with abdominal imaging and were found to have no evidence of RCC, were negative for the germline BAP1 gene variant. Individual IV:2 (omitted from pedigree) is the spouse of IV:1.

To determine the genetic basis of the RCC predisposition in this family, the proband was evaluated at the NCI for germline mutations of the following genes: VHL, MET, FLCN, TSC1, TSC2, SDHB, SDHC, SDHD, and FH. No germline mutations were detected in any of these genes. In addition, germline chromosome 3 translocation familial renal carcinoma (20) was excluded by karyotype analysis.

At the NCI, the proband underwent the first of three left partial nephrectomies at the age of 46. At that time he had a resection of a 3 cm ccRCC that was found to be Fuhrman grade III, a 1.5 cm ccRCC (also Fuhrman grade III), and a 2 centimeter atypical cyst with clear cell lining (Figure 2). Subsequently, the patient underwent active surveillance for 8 years at which point surgical intervention was recommended due to tumor growth (Figure 3A). At the age of 54, when his largest renal tumor had reached 3.46 cm, he underwent a second left partial nephrectomy for removal of a ccRCC (Fuhrman grade II) and three renal cysts, two of which were lined by “atypical clear cells” (Figure 4 A-C). The patient was then managed by active surveillance for almost 4 years. During this time a tumor grew to 3.7 cm (Figure 3B) necessitating a third left partial nephrectomy at age 57 with removal of 4 separate ccRCC (three Fuhrman grade II and one Fuhrman grade III) as well as a renal cyst. At his most current evaluation in May of 2012, he had excellent renal function and was without evidence of metastatic disease.

Figure 2. Bilateral, multifocal cysts and solid kidney cancer in proband IV:1 from Family NCI-1326.

Axial abdominal computed tomography (CT) scans from the proband (IV:1, NCI-1326 kindred) following right radical nephrectomy showing multifocal left renal lesions, indicated by arrows in A-D. Subsequently, at the age of 46 the individual underwent the first of three left partial nephrectomies, for the surgical removal of a 3 cm ccRCC (Fuhrman grade III), a 1.5 cm ccRCC (Fuhrman grade III) and a 2 cm atypical cyst with clear cell lining.

Figure 3. Rapid growth of renal tumors in proband IV:1 from Family NCI-1326.

Rapid growth rate of proband IV:1 renal tumors prior to second (A), and third left partial nephrectomies (B).

Figure 4. Recurrent multifocal cysts and solid kidney cancer in proband IV:1 from Family NCI-1326.

(A-C) Second left partial nephrectomy from proband (IV:1, NCI-1326 kindred). Axial abdominal computed tomography (CT) scan (A), and histology slides from removed lesions, a ccRCC (B) and an atypical renal cyst with clear cell lining (arrows) (C).

Because of the family history, individual IV:4 underwent screening and a central mass was detected in her right kidney at age 40 for which she underwent a right radical nephrectomy. This tumor was a 3.8 cm ccRCC, Fuhrman grade II. Twelve years after the operation she remains without evidence of disease. Individual IV:6 presented at the age of 36 with metastatic RCC and died a short time later.

Given the finding of a BAP1 missense germline variant (c.41T>A, p.L14H) in individual IV:1, we performed cosegregation studies. This variant was not found in dbSNP137, Sanger Institute Catalogue of Somatic Mutations in Cancer (COSMIC) (21) or 1000 Genomes Project databases (13). The same variant was found in germline DNA from affected individual IV:4 (Figure 5A). In contrast, no germline BAP1 variant was detected in unaffected individuals IV:3 and IV:5 (Figure 5A). No DNA could be obtained for individual IV:6, who had died. Thus, the BAP1 c.41T>A (p.L14H) variant cosegregated with the RCC predisposition in this family. The odds against random segregation are 5; under a dominant mode of inheritance and assuming full penetrance, the backward odds (22, 23) are 16 and the LOD score is 1.2.

Figure 5. Cosegregation and LOH studies of BAP1 variant in Family NCI-1326.

(A) Sequence chromatogram for proband IV:1 and affected individual IV:4 with the c.41T>A (p.L14H) BAP1 variant. Unaffected individuals IV:3 and IV:5 were negative for the BAP1 variant. (B) Sequence chromatograms of renal tumors from proband IV:1 displaying loss of heterozygosity (LOH) at the BAP1 locus for 2 different regions from one tumor (Tumors 1a and 1b) and mutant allele enrichment in another tumor (Tumor 3). Sequence chromatogram of individual IV:4 tumor showing LOH. Nx, nephrectomy.

The BAP1 variant (p.L14H) maps to the catalytic domain, a domain that is a frequent site of pathogenic missense mutations. Somatic mutations in the ubiquitin C-terminal hydrolase (UCH) domain have been reported in several RCC studies as well as in COSMIC and the Kidney Renal Cell Carcinoma (KIRC) dataset produced by The Cancer Genome Atlas (TCGA) (Figure 6A) (11, 21, 24). Leucine 14 is highly conserved (Figure 6B) and along with previously reported RCC mutations in neighboring residues, L14H is predicted to be deleterious by PROVEAN, SIFT, and PolyPhen-2 prediction tools (Figure 6C) (19, 25, 26).

Figure 6. Analysis of the novel p.L14H BAP1 variant.

(A) Schematic of BAP1 protein showing the position of the novel p.L14H missense variant (gray triangle) in comparison to known BAP1 missense mutations associated with sporadic RCC (black triangles). Data compiled from Peña-Llopis et al., Hakimi et al., COSMIC and KIRC (TCGA) (11, 21, 24). UCH, ubiquitin C-terminal hydrolase domain (blue); HBM, HCF-1-binding motif (yellow); ULD, Uch37-like domain (black); BRCA1, putative BRCA1-interacting domain (red); NLS, nuclear localization signal (green). (B) BAP1 amino acid conservation across species assessed with BioEdit’s ClustalW multiple alignment function. Protein sequences are from UniProt (Q92560, H9G0D9, F6TYN2, E2R9Z2, D3ZH56, Q99PU7, G1PS27, F6SMM8, F6RI15, Q5F3N6, Q52L14, H2UEV1, A1L2G3, Q7K5N4, Q17N72) and Ensembl (ENSPTRP00000025898) (33). (C) In silico predicted effects of the novel p.L14H missense variant and the surrounding sporadic RCC-associated missense mutations assessed with PROVEAN, SIFT and PolyPhen-2 prediction tools (19, 25, 26). (D) BAP1 structure model. Left Panel: Cartoon depiction of the BAP1 UCH domain (purple) noting p.Leu14 (red sphere) involved in organizing a flexible crossover loop and other flexible portions of the domain (salmon) that order upon ubiquitin substrate (cyan) binding. Uch37-like domain (ULD) is shown in green. Right Upper Panel: Zoom-in of wild type BAP1 leucine 14 residue with atom radii depicted in dots interacting with surrounding residues (side chains within 4 Å displayed in stick). Right Lower Panel: Zoom-in of BAP1 histidine 14 mutant with atom radii (dots) revealing clashes with surrounding residues.

We previously constructed a BAP1 structural model based on the related family members Uch-L3 and Uch37 (11). Leucine 14 maps to the first helix of the UCH domain and is physically adjacent to two previously identified pathogenic RCC mutations, p.G13V and p.H144N (Figure 6D). The leucine 14 side chain in the paralogue UCH-L3 helps organize a crossover loop and other flexible portions of the UCH domain that order upon ubiquitin binding, and forms a portion of the interaction surface for the ULD tail (27). Mutation of this residue to histidine is predicted to increase the effective volume of the side chain, possibly causing steric clashes with surrounding residues, and may prevent productive ubiquitin binding (Figure 6D).

BAP1 is a two-hit tumor suppressor gene, and we performed studies for loss of heterozygosity (LOH). Three tumors (including two samples from different regions of one tumor) were examined from the proband. DNA was extracted from the different tumors and sequenced for the BAP1 variant (c.41T>A). Two samples from a surgery in 2008 showed clear LOH (Figure 5B, Tumors 1a and 1b). From a surgery in 2012, two tumors were examined. One did not show appreciable LOH, possibly due to contamination by non-malignant cells or the acquisition of a somatic mutation in the other BAP1 allele (Tumor 4, data not shown). Enrichment of the mutant allele relative to the wild-type allele was observed in the other tumor from the proband 2012 surgery (Figure 5B, Tumor 3). In addition, a tumor was evaluated from individual IV:4. This tumor also showed enrichment of the mutant allele consistent with LOH (Figure 5B, IV:4 Tumor 1). Additionally BAP1 protein expression in the tumors was investigated by immunohistochemistry. Immunohistochemical staining was negative for BAP1 protein in the tumor from individual IV:4 (Figure 7 and insert A) and Tumors 3 and 4 from the proband (data not shown). In contrast, nuclear BAP1 staining was observed in the normal adjacent epithelium (Figure 7 asterisk and insert B). All the tumors examined, including Tumor 4 that failed to show LOH at the DNA level, were negative for BAP1 by immunohistochemistry. Thus, the immunohistochemical data confirm loss of BAP1 protein and provide further evidence for “two-hit” inactivation of BAP1 in the tumors from this family. Given that BAP1 loss by IHC is observed in 15% of sporadic ccRCC (11), the probability that all 3 tumors would be BAP1 negative by chance alone is 0.0034. These data are most consistent with the notion that (1) p.L14H is a causative mutation, (2) p.L14H abrogates protein expression in tumors, and (3) mutations abolishing expression of the second allele are uniformly present.

Figure 7. Loss of BAP1 protein in renal tumor from individual IV:4.

Low power view of renal tumor from individual IV:4. The tumor involves the medullary area of the kidney and shows negative staining for BAP1. Note the pelvic transitional epithelium that stains positive for BAP1 (*) as positive internal control (150X). Insert A: High power image showing the renal tumor with negative BAP1 immunohistochemical staining (200X). Insert B: High power image of BAP1 immunohistochemical staining showing pelvic transitional epithelium with positive nuclear staining as a positive internal control (250X).

Previous VHL mutation testing of germline DNA from the proband of family NCI-1326 was negative. However, the VHL and BAP1 genes are both on chromosome 3p so we asked whether somatic mutations in VHL could be detected in the tumors. VHL mutation analysis (Supplementary Methods) showed VHL mutation in some tumors (IV:1, Tumor 1 and IV:4, Tumor 1), but not in others (IV:1, Tumors 3 and 4) (Supplementary Figure S2 and data not shown). Thus, mutations in VHL and BAP1 may cooperate in the development of at least some of the tumors.

Together, these data suggest that the novel BAP1 p.L14H missense variant is the cause of the underlying cancer phenotype in the family described. First, the variant cosegregated with the RCC phenotype. Second, the variant targets the catalytic domain, which is a common site of missense mutations including pathogenic somatically-acquired mutations in neighboring residues pG13V and p.H144N. Third, L14 is highly conserved across species and in silico analyses suggest that a histidine substitution at this position precludes productive ubiquitin binding. Fourth, consistent with BAP1 function as a two-hit tumor suppressor gene, the variant was associated with LOH in tumors from the proband and an affected sister by DNA sequence analysis. Finally, albeit indirectly, the absence of BAP1 protein, by immunohistochemistry, in all tumors examined suggests that the BAP1 p.L14H variant abrogates protein expression. Furthermore, these findings confirm that even in the samples in which LOH was not observed, BAP1 function was lost.

Although it is not possible to generalize from a single kindred, our data suggest that BAP1 mutations will be found, albeit infrequently, in familial RCC, and thus, it seems fitting to comment on the phenotypic aspects of the family we report. The family in which the novel p.L14H variant cosegregated with RCC demonstrated a cancer phenotype characterized by an early-onset, aggressive form of ccRCC. The proband is noted to have had bilateral, multifocal disease with multiple solid and cystic lesions with a rapid rate of growth. Individual IV:6 died of early onset metastatic disease at the age of 36. Two other individuals (II:1 and III:4) died of metastatic RCC at 48 and 58 years of age, respectively. Another family member, who inherited the p.L14H variant, developed early-onset ccRCC at age 40. Overall, this is consistent with our previous research findings, which have shown an association of BAP1 loss with high tumor grade and poor survival (11, 12). Additionally, other studies have correlated BAP1 loss with metastases in other BAP1-associated cancers, particularly uveal melanoma (9). Taken together, these observations support the idea that germline BAP1 mutations may be associated with aggressive familial ccRCC.

The NCI-1326 family kidney cancer phenotype is, in some ways, similar to the von Hippel-Lindau (VHL) phenotype, i.e., the presence of bilateral, multifocal clear cell tumors and cysts, and tumors within cysts. However, the kidney cancer phenotype in this family appears to differ from the typical VHL phenotype in several ways. The renal tumors in proband IV:1 grew faster than might have normally been expected in a VHL patient. In the experience at NCI, the average growth rate of VHL-associated RCC is 3 mm per year, which is consistent with the growth rate of sporadic small renal masses (28). In addition, there are a number of high Fuhrman grade ccRCCs in this family, which are uncommonly seen in VHL tumors, particularly those less than 3 cm in size.

There are several clinical implications for the physician when determining a differential diagnosis for germline genetic analysis. Our findings and the fact that somatic BAP1 mutations are often associated with ccRCC of high grade may encourage clinicians to look for these features when deciding whether to proceed with germline BAP1 mutation analysis. Identifying cues associated with germline BAP1 mutation will be important as these mutations are rare. Additionally, our family (although not all individuals) displayed a cancer phenotype that was noted to be “VHL-like” with bilateral, multifocal disease and multiple renal cysts. Thus, clinicians may consider BAP1 analysis for individuals who test negative for germline VHL mutations and display a more “aggressive” phenotype that lacks other VHL hallmark findings such as hemangioblastomas, pancreatic cysts, and pheochromocytomas.

Prior to this report, germline BAP1 mutations were reported to predispose to several additional cancers including uveal and cutaneous melanoma and mesothelioma (14-17). However, other tumor types have been found in these pedigrees such as lung carcinomas, meningiomas, and cholangiocarcinomas (14, 16-18, 29). We did not observe any of these cancers in our family. The reason for this is unclear. However, these data are in keeping with the previous literature in which, initially, BAP1 germline mutations were associated with two clinically distinct syndromes of familial melanoma and mesothelioma (17, 18). These separate cancer phenotypes were later demonstrated to overlap (29, 30). As germline BAP1 mutation families continue to be described, the phenotype may be clarified. No clear genotype/phenotype correlation has been observed, and familial mutations frequently result in early truncation of the BAP1 protein (14, 15, 17, 18, 29, 31, 32). It seems plausible that germline BAP1 mutations produce a cancer susceptibility syndrome in which the penetrance of each given feature (RCC, melanoma or mesothelioma) depends upon additional environmental or genetic factors. Nevertheless, the presence of other BAP1-associated malignancies in RCC families may increase suspicion for a germline BAP1 mutation.

Our overall germline BAP1 mutation frequency of 1.2% suggests that, although germline BAP1 mutations may predispose to RCC, the frequency of these mutations is low but in keeping with that of other studies. For comparison, several other studies describing germline BAP1 mutations in a variety of cohorts describe an overall germline frequency of approximately 3.8% (range of 1.9% in a group of individuals with uveal melanoma to 8.0% in a subset of apparent sporadic mesotheliomas) (14-17, 30). Overall, these data show that germline BAP1 mutations predispose to several tumor types and the degree of tumor susceptibility conferred by BAP1 mutation may vary across tissues.

There are several limitations to this study. First, among the 82 probands successfully evaluated, only one possibly pathogenic BAP1 variant was found. Second, the functional significance of p.L14H remains to be examined. Third, cosegregation studies could be performed in just four individuals and the LOD score is low. However, given the loss of BAP1 proein by IHC in all tumors examined (n=3) and a probability that this finding would be from chance alone of 0.0034, our results strongly suggest that the variant identified is responsible for the RCC predisposition observed. Despite these limitations, given the role of BAP1 in sporadic ccRCC, the established role of BAP1 as a familial tumor suppressor gene, and the evidence presented, we believe that BAP1 represents a novel familial RCC gene.

In conclusion, we report for the first time a novel BAP1 germline missense variant that appears to predispose to familial, early-onset, aggressive ccRCC. Our data suggest that BAP1 may be the causative gene for renal tumor development in a subset of inherited kidney cancer patients, although the frequency of BAP1 gene mutations in RCC families appears to be low.