Synonymous but Not Silent: A Synonymous VHL Variant in Exon 2 Confers Susceptibility to Familial Pheochromocytoma and von Hippel-Lindau Disease

Abstract

Context

von Hippel-Lindau (VHL) disease, comprising renal cancer, hemangioblastoma, and/or pheochromocytoma (PHEO), is caused by missense or truncating variants of the VHL tumor-suppressor gene, which is involved in degradation of hypoxia-inducible factors (HIFs). However, the role of synonymous VHL variants in the disease is unclear.

Objective

We evaluated a synonymous VHL variant in patients with familial PHEO or VHL disease without a detectable pathogenic VHL mutation.

Design

We performed genetic and transcriptional analyses of leukocytes and/or tumors from affected and unaffected individuals and evaluated VHL splicing in existing cancer databases.

Results

We identified a synonymous VHL variant (c.414A>G, p.Pro138Pro) as the driver event in five independent individuals/families with PHEOs or VHL syndrome. This variant promotes exon 2 skipping and hence, abolishes expression of the full-length VHL transcript. Exon 2 spans the HIF-binding domain required for HIF degradation by VHL. Accordingly, PHEOs carrying this variant display HIF hyperactivation typical of VHL loss. Moreover, other exon 2 VHL variants from the The Cancer Genome Atlas pan-cancer datasets are biased toward expression of a VHL transcript that excludes this exon, supporting a broader impact of this spliced variant.

Conclusion

A recurrent synonymous VHL variant (c.414A>G, p.Pro138Pro) confers susceptibility to PHEO and VHL disease through splice disruption, leading to VHL dysfunction. This finding indicates that certain synonymous VHL variants may be clinically relevant and should be considered in genetic testing and surveillance settings. The observation that other coding VHL variants can exclude exon 2 suggests that dysregulated splicing may be an underappreciated mechanism in VHL-mediated tumorigenesis.

A synonymous von Hippel-Lindau (VHL) variant causes pheochromocytoma or VHL disease as a result of exon 2 skipping.

von Hippel-Lindau (VHL) disease is an autosomal, dominantly inherited syndrome caused by a pathogenic mutation of the VHL gene and is characterized by variable clinical manifestations that can include renal cell carcinomas (RCCs), hemangioblastomas of the central nervous system, and pheochromocytomas (PHEOs) (1). VHL disease type 1 confers higher risk for RCC and lower risk for PHEO, whereas in VHL disease type 2 (A, B, or C), there is higher risk for PHEO, in association with hemangioblastoma (2A), RCC (2B), or PHEO only (2C) (1).

VHL is the substrate recognition component of a ubiquitin ligase that leads to degradation of hypoxia inducible factors (HIFs), a function that has been largely regarded as a key contributor to the tumor-suppressor properties of VHL (2). HIFs are induced in response to low oxygen levels, and their abnormal activation in sporadic and hereditary cancers lacking VHL function is required for tumor growth (2, 3). The VHL gene is comprised of three exons that encode two naturally occurring transcription forms. The full-length transcript includes exons 1, 2, and 3 (NM_000551.3) and is capable of producing two protein isoforms: p213 and through an alternative start site, p160, both considered to be tumor suppressors (2). The second, shorter transcript (NM_198156.2) includes exons 1 and 3 but lacks the in-frame exon 2 and is predicted to produce a protein with 172 amino acids (p172). As exon 2 codes for a substantial portion of the HIF-binding domain, the shorter transcript is not expected to function as a tumor suppressor (4–6). The regulatory mechanisms underlying VHL splicing and transcript generation have not been fully established. However, others have previously reported intronic, missense, and more recently, synonymous VHL variants that change splicing and lead to exon 2 skipping, highlighting that this may be a mechanism of oncogenicity in VHL disease (7).

Synonymous variants, also referred to as silent mutations because of their presumed neutral effect, are characterized by a modification of the DNA sequence, which does not directly alter the resulting amino acid (8). Nevertheless, synonymous variants are not always silent and can be pathogenic by modifying splicing (8–11). Typically, this occurs when the variant is located near an exon-intron border. However, synonymous variants within exons can also act as driver mutations by creating and/or abolishing splicing motifs associated with exon skipping and/or exon retention (11, 12).

Here, we report five independent cases/families of PHEOs or VHL disease associated with an identical synonymous VHL variant located within exon 2 and present data supporting a pathogenic role of this variant through increased exon 2 skipping. Our analysis of existing cancer datasets carrying VHL mutations suggests that exclusion of exon 2 may occur more broadly in VHL exon 2 variants.

Materials and Methods

Patients

Patients with the diagnosis of familial or sporadic PHEO and their relatives were enrolled on the University of Texas Health San Antonio (San Antonio, TX) Institutional Review Board-approved study, Genetic Analysis of Pheochromocytoma, Paraganglioma and Associated Conditions (NCT03160274), and provided their signed consent. Written consent was also obtained from the additional patients and their families, as approved by their respective Institutional Review Board Committees at Oita University (Oita, Japan), Kindai University (Osaka, Japan), and Tsukuba University (Kyoto, Japan). Control samples included in Pheo-Type profiling had been previously institutionally approved, as reported (13). Diagnoses of PHEO and VHL disease were performed according to current clinical guidelines (14, 15). Follow-up was defined from the date of initial PHEO diagnosis to the date of last follow-up in patients with PHEO. Clinical, biochemical, and histological information related to PHEO, RCC, and hemangioblastoma was obtained from medical records and is summarized in Table 1.

Table 1.

Clinical Features of Five Families Carrying a Germline c.414A>G, p.Pro138Pro VHL Variant

Family ID	1	2	3	4	5
Number of affected individuals in family	7	5	3	3	1
Proband age at onset (y)	32	53	31	20	47
Age at diagnosis of affected individuals (range in y)	12-73	29-64	31 to ∼50	20-27	47
Gender of affected individuals (F:M)	2:5	1:4	0:3	0:3	1:0
VHL-related phenotype (number of affected individuals)	PHEO (7), HB (1/7), renal cyst (1/7)a	PHEO (5), PGL(1/5), HB(1/5)	PHEO (3)	PHEO (3), HB (1/3)	PHEO+PGL (1)
Initial disease manifestation	PHEO	PHEO	PHEO	PHEO	PHEO
Number of individuals with multiple PHEOS (% affected individuals)	0 (0%)	3 (60%)b	0%	2 (67%)	1 (100%)
Secreted catecholamine	NE	NE	Non-secretingc	NE	NE
Malignancy	No	Yesd(lung metastases)	No	No	No
Other VHL manifestation	Spinal HB	Spinal HB	No	Spinal HB	No
VHL disease subtype	2A	2A	2C	2A	2C
Other manifestations (number of individuals)	Thyroid nodule (1), bladder carcinoma (1)	No	No	Parathyroid nodule (1), esophageal carcinoma (1)	No
Follow-up time	30 y	18 y	10 mo	25 y	20 y

Abbreviations: F, female; HB, hemangioblastoma; ID, identification; M, male; NE, norepinephrine; PGL, paraganglioma.

^aRenal cyst and HB were detected in two different patients.

^bOne patient had bilateral PHEO and PGL.

^cPatient with bilateral PHEO and PGL developed lung metastasis and was lost to follow-up.

^dOnly proband data available; lost to follow-up.

Methods

Genetic screening

Genomic DNA was extracted from peripheral blood, saliva, and fresh frozen tumor samples using conventional methods. Whole exome sequencing of the proband’s germline DNA was prepared using NimbleGen 44Mb SeqCap EZ Exome Kit v2, sequenced on a HiSeq 2000, and analyzed using filtering parameters previously reported (16), with the exception that synonymous variants of PHEO susceptibility genes were not excluded. Sanger sequencing of VHL from germline and tumor genomic DNA of affected and unaffected relatives was conducted as published (16). RNA was isolated from peripheral leukocytes and fresh frozen PHEO using TRIzol, according to the manufacturer’s guidelines. cDNA was generated as we previously reported (17). Sanger sequencing of leukocyte and tumor VHL cDNA was performed by amplification of VHL transcripts using primers spanning exon 1 and exon 2 (downstream of the c.414 variant) or exon 1 and exon 3 (primer sequences available upon request).

Gene-expression profiling

Tumor RNA from archival formalin-fixed paraffin-embedded PHEO and additional control PHEOs with known genetic background was prepared using Qiagen and analyzed using a custom NanoString panel (Pheo-Type) on the nCounter GX gene-expression system (NanoString Technologies), in accordance with the manufacturer’s protocol. The Pheo-Type NanoString panel was comprised 46 probe sets targeting 40 genes (13). NanoString gene-expression data were z score normalized using a larger dataset of 90 other PHEO tumors, including 39 tumors of known genotype, previously analyzed on the same platform. A support vector machine, trained on a compendium of z score-normalized Affymetrix data, were used to test the independent NanoString-profiled samples, classifying each tumor into one of four PHEOs and paragangliomas classes [RTK, MAX-like, VHL, and succinate dehydrogenase (SDH)], as previously described (13). The z score-normalized NanoString data of samples with known genotype were used for Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP) clustering of samples using the umap-learn python package [v0.3.2; https://github.com/lmcinnes/umap (arXiv:1802.03426)] with parameters minimum distance = 0.25, n_neighbors = 5, and correlation distance metric.

Real-time quantitative RT-PCR of SLC2A1, HK2, VEGFA, RET, and PNMT transcripts was performed using PHEO cDNA from an affected individual (from Family 1, IV:3) and control tumors carrying known pathogenic mutations of EPAS1 (n = 2), NF1 (n = 1), RET (n = 1), SDHB (n = 1), TMEM127 (n = 1), and VHL (n = 4). Expression values were calculated using the ΔΔCt method, normalized to TATA-binding protein (Tbp) and β-actin housekeeping controls.

For VHL transcript-specific quantification, primers flanking VHL exons 1 and 3 that yielded two transcripts were used to estimate relative abundance of two VHL transcript isoforms using regular PCR and visualization by gel electrophoresis. Real-time quantitative RT-PCR, using SYBR Green Master Mix on a StepOne Real-Time PCR System (Applied Biosystems), was used for absolute VHL transcript quantification using isoform-specific primers (sequences available upon request), calculated as above (17).

TCGA pan-cancer analysis

To evaluate the percentage of transcripts of VHL lacking exon 2 across cancer samples, we used the percentage spliced in (PSI) matrix of alternative splicing across 32 The Cancer Genome Atlas (TCGA) cancer types of 8705 patients (8). PSI provides a quantitative, event-base assessment of exon use representing the percentage of a gene’s mRNA transcripts that include a specific exon junction, thus identifying differential isoform processing without prior knowledge of splicing patterns in each gene (18). PSI scores were calculated using SpIAdder (19), applying default criteria (minimal of three reads per junction). For the VHL gene, low PSI represents exon 2 skipping. Other reads that may have involved other sets of junction reads were not included in this calculation. We downloaded the mutation annotation file from https://gdc.cancer.gov/node/905/ in https://gdc.cancer.gov/about-data/publications/PanCanAtlas-Splicing-2018. Exon 2 PSI values were plotted against mutation genomic location.

Results

A synonymous VHL variant (c.414A>G, Pro138Pro) is associated with PHEOs and VHL disease

The proband of family 1 (III:3; Fig. 1a), a 32-year-old woman, was diagnosed with a normetanephrine-secreting, 5-cm right PHEO with no other clinical features of VHL disease. Her paternal grandfather (I:1) had been diagnosed with an adrenal tumor and hypertension in his 40s. In addition, four other relatives developed unilateral PHEOs (Fig. 1a): her father (II:1) at age 73, her 34-year-old brother (III:2), her paternal aunt in her 40s (II:4), and the aunt’s 12-year-old son (III:7). Subsequently, the son of III.7 (IV:3) was also diagnosed with PHEOs at 23 years of age. All affected individuals underwent adrenalectomy and remain asymptomatic, 1 to 30 years after surgery (Table 1). In addition, a renal cyst was detected in individual III.2 at the time of the PHEO diagnosis, whereas II.1 developed multiple spinal cord hemangioblastomas after the adrenal tumor resection. No RCCs were diagnosed.

Figure 1.

(a) Four-generation pedigree of Family 1: the synonymous VHL variant, c.414A>G, p.Pro138Pro, segregates with affected family members. (b) Distance of nucleotide 414 (A, *) relative to the canonical donor and acceptor splice sites of VHL exon 2 [(top) exonic sequences are shown in uppercase; intronic sequences in lowercase]; (middle) results from whole exome sequencing and (bottom) Sanger sequencing of the proband’s III:3 germline DNA reveal a heterozygous, synonymous VHL variant, c.414A>G, p.Pro138Pro, in exon 2. (c) NanoString-based gene-expression data of archival PHEO from III:3 (*) classifies this tumor within the VHL-type group. UMAP was used to cluster the z score-normalized NanoString data of 43 PHEOs of known genotype, color coded as indicated, showing tight clustering of the PHEO proband among other VHL-mutant tumors. (d) Real-time RT-PCR quantification of HIF2α transcription targets HK2, SLC2A1 (Glut1), and VEGFA, induced in pseudohypoxic PHEOs, and of PNMT and RET transcripts, induced in cluster 2 (kinase signaling) PHEOs using cDNA from another individual from Family 1 (IV:3). Control PHEOs with classic VHL mutations or those with mutations in RET, TMEM127, or NF1 were included (n = 9 or n = 6, respectively, for the HIF targets or cluster 2 targets). Values shown represent the ΔΔCt values normalized to housekeeping genes (TBP or β-actin, respectively) from two independent experiments performed in duplicates. (e) Sanger sequencing of DNA from the PHEO from individual IV:3 indicates loss of the wild-type (WT) allele.

The proband had been previously reported to carry “no pathogenic variants” of PHEO susceptibility genes after whole exome sequencing of her germline DNA revealed no remarkable missense or truncating mutations in PHEO susceptibility genes (16). However, a heterozygous, synonymous variant was detected in exon 2 of the VHL gene, c.414A>G, p.Pro138Pro (NM_000551.3), located 49 nucleotides upstream of the nearest exon-intron border (Fig. 1b). Sanger sequencing of germline DNA from 16 relatives showed that the c.414A>G variant segregated with the disease in this family (Fig. 1a). This variant is absent in the Genome Aggregation Database (or gnomAD) or other normal reference databases.

The NanoString expression profiling classified the proband PHEO as “VHL-mutant” (Fig. 1c) using a support vector machine classifier trained on a large compendium of independent microarray data representing known PHEOs and paragangliomas subtypes and genotypes. Furthermore, UMAP clustering, with an additional 43 PHEO tumors of known genotype, showed tight clustering of the PHEO proband among other VHL tumors (Fig. 1c). To validate these findings, we quantified genes known to be differentially expressed in these tumors in a second PHEO from this family (individual IV:3), along with nine controls carrying distinct driver mutations. In agreement with the Pheo-Type profiling results, the second VHL c.414A>G-carrier PHEO showed induction of HIF targets VEGF, HK2, and SLC21A(Glut1) but low expression of RET and PNMT, similar to the pseudohypoxic profile observed in other VHL-mutant PHEOs (Fig. 1d), as previously established (20, 21). Moreover, we detected loss of the wild-type (WT) VHL allele in DNA from this tumor (Fig. 1e), consistent with the expected pattern seen in other PHEOs with VHL loss of function. Taken together, these findings suggested that the VHL c.414C>T, p.Pro138Pro, variant is pathogenic.

Genotype-phenotype correlations of the VHL c.414A>G variant

Next, to examine the impact of this synonymous variant, we expanded our analysis to additional patients and families with PHEO or VHL disease without a detectable pathogenic VHL variant and identified four additional, unrelated cases; a single individual; and three families with an identical germline VHL c.414A>G, p.Pro138Pro, which segregated with the disease phenotype. A summary of the clinical information from these cases is displayed in Table 1.

In all, 19 affected individuals were identified in five families: three kindred (including the original family described above) had VHL type 2A disease (PHEO and hemangioblastoma), whereas in the fourth family and the single individual, PHEO was the only manifestation (VHL type 2C). Notably, all clinically affected carriers had PHEO. No renal carcinomas were detected in any of these patients or relatives, despite follow-up periods of up to 30 years. In these patients, the clinical presentation of PHEOs was similar to that related to classic VHL mutations (15, 22): median age at disease onset was 36.6 years (±13.2), the tumors secreted norepinephrine, and they were often multiple (six individuals had bilateral PHEOs or PHEO and paraganglioma) and rarely malignant (5%; one of 19 developed metastatic disease; Table 1). Unique to this cohort, there was an almost fourfold predominance of male (M) over female (F) patients in these families (15 M:4 F). A sex imbalance is not known to occur in typical VHL disease (1, 15, 23). In addition, we identified three asymptomatic carriers, all women older than 45 (43 to 84 years) in family 1, which might suggest lower penetrance of this variant than that reported for classic VHL mutant alleles (15). Both of these results could be biased by the small number of cases. We next evaluated the frequency of the c.414A>G variant in disease settings by scanning the literature and publicly available databases, including Catalogue Of Somatic Mutations In Cancer (or COSMIC), ClinVar, and the VHL Universal Mutation Database (in which synonymous variants are recorded) and found nine entries for the c.414A>G variant. Although the clinical information is limited, the data suggest that this variant is causal of the PHEO/VHL phenotype in patients (7, 24).

VHL c.414A>G promotes exon 2 skipping

To identify potential mechanisms through which the c.414A>G variant might affect VHL function, we PCR amplified a PHEO for which cDNA was available (IV.3). The predominant product detected was a transcript that lacked the entire in-frame exon 2 (e1 + e3; Fig. 2a and 2b), identical to a transcript previously recognized in VHL (NM_198156.2; Fig. 2b) (4). This pattern contrasted with other PHEOs, including those with known pathogenic VHL mutations, in which the full-length (e1 + e2 + e3) transcript was the most abundant isoform (Fig. 2a). No other products were obtained from these reactions. Transcript-specific quantification revealed an almost 20-fold higher ratio of the exon 2-lacking isoform relative to the full-length transcript in this tumor compared with control PHEOs carrying other mutations (Fig. 2c). We next used semiquantitative (Fig. 2d) and isoform-specific real-time quantitative RT-PCR (Fig. 2e) of leukocyte cDNA obtained from both c.414A>G carriers and noncarriers to measure the two VHL transcripts and found a threefold-higher ratio of the shorter (e1/e3) transcript relative to full-length VHL (P = 0.001). This ratio reflected the simultaneous increase in the expression of the short isoform and decreased expression of the full-length VHL transcript in carriers compared with leukocyte cDNA from WT relatives (Fig. 2d and 2e). Remarkably, this aberrant ratio was detected both in affected and in clinically unaffected c.414A>G carriers (Fig. 2d). The less marked increase ratio in leukocyte, compared with tumor, cDNAs is likely a result of the presence of the WT VHL allele in peripheral leukocytes of the variant carriers.

Figure 2.

(a) Gel electrophoresis of VHL cDNA PCR products generated using primers spanning exons 1 and 3 showing the abundance of the VHL full-length (top band) and/or a transcript lacking exon 2 (bottom band) isoforms in tumor cDNA from four non-VHL-mutated tumors (EPAS1, TMEM127, SDHB); VHL c.414A>G PHEO from patient IV:3 (arrow); and four classic VHL-mutated tumors. β-Actin PCR serves as a loading control. (b) Sanger sequencing of PCR products from PHEO cDNA from patient IV:3 using primers spanning exon 1 and exon 3 shows expression of a VHL transcript that lacks exon 2 (NM_198156.2). Dashed, vertical line indicates border of exon 1 and exon 3. (c) Isoform-specific transcription of VHL in PHEO cDNA from non-VHL-mutated PHEOs (“Non-VHL,” including SDHB, EPAS1, and TMEM127 mutants); PHEO from individual IV:3 carrying the VHL c.414A>G variant; and classic VHL-mutated tumors (VHL-mutant) shown as fold change of VHL (left) short e1 + e3, (middle) full-length e1 + e2 + e3, and (right) ratio of short/full-length transcripts normalized to housekeeping gene (β-actin). Values are the result of repeat PCRs performed in duplicate. (d) Gel electrophoresis of PCR products generated as in a, using peripheral leukocyte cDNA as templates from the indicated c.414A>G carrier and WT individuals from Family 1. TBP PCR is used as a loading control. (e) Fold change of VHL (left) short e1 + e3, (middle) full-length e1 + e2 + e3, and (right) ratio of short to full-length transcripts from real-time PCR of leukocyte cDNA of the individuals shown in d. Individual III:1 WT sample was chosen as reference. Values are the result of repeat PCRs performed in duplicate. P Values were calculated by t test: *P < 0.05, **P < 0.005. (f) Sanger sequencing of PCR products from (top) leukocyte genomic DNA or (bottom) cDNA of a c.414A>G carrier (IV:3) using primers in exon 1 and exon 2 downstream of the variant.

Importantly, sequencing of leukocyte cDNA from variant carriers (heterozygotes for the variant) revealed that only the WT allele produced a full-length VHL transcript (Fig. 2f). This result indicates that the VHL c.414A>G variant promotes exon 2 skipping, resulting in the exclusive generation of the shorter VHL transcript (VHL e1 + e3) from the mutated allele. As the WT allele is deleted in tumor DNA (Fig. 1e), only the short VHL transcript is produced by the tumor, in favor of loss of VHL function, rather than a competing or dominant-negative effect of the short isoform.

Other VHL exon 2 variants may also promote exon 2 skipping

Next, we considered that other variants affecting VHL, beyond those that specifically target canonical splice sites and are thus expected to alter splicing, might also impact exon 2 expression. To that end, we performed an evaluation of VHL exon 2 skipping in TCGA pan-cancer samples containing both VHL sequence and transcript data (n = 10,019 tumors from 32 cancer types) for which PSI scores had been calculated (25). In this analysis, PSI scores reflect the percentage of junctions encoding exon 2 (Fig. 3a); low scores represent samples in which exon 2 was excluded. The cancers were divided into five groups (Fig. 3b): (i) all tumors, (ii) tumors with any VHL mutation (n = 213), (iii) tumors with mutations in exon 1 (n = 46), (iv) tumors with mutations in exon 2 (n = 74), and (v) tumors with mutations in exon 3 (n = 77). As expected, tumors with VHL mutations had lower PSI scores than VHL-WT tumors (P = 0.01). Specifically, those with mutations in exon 2 had the lowest PSI scores (P = 0.005 vs group 1; Fig. 3c), suggesting that these variants were associated with a transcript lacking exon 2. In contrast and in support of a specific effect of exon 2 mutations on the transcript preference, tumors with mutations in exon 3 had the highest exon 2 PSI scores (P = 0.0008 vs group 4). Of note, as tumors with no expression of exon 2 junctions were excluded from the analysis, these findings may underestimate the frequency of exon 2 skipping. In fact, two of the three VHL variants from the PHEO-paraganglioma dataset were from exon 2 and did not produce measurable scores. Interestingly, the sample with the lowest PSI score carried a synonymous exon 2 variant (c.354C>T,p.Leu118Leu; Fig. 3), suggesting that other synonymous variants targeting this exon may be prone to exon 2 exclusion.

Figure 3.

(a) Diagram illustrating the strategy used for calculation of the PSI scores of VHL transcript isoforms. Events (top) containing junctions flanking or spanning exon 2, covered by at least three reads, were included, whereas those (bottom) skipping exon 2 were excluded for PSI calculation of TCGA pan-cancer VHL transcripts. (b) PSI scores of TCGA pan-cancer tumor samples, distributed in the following groups: all tumors (All); only VHL-mutant tumors (VHL_mt); and tumors with mutations in exon 1 only (VHL_mt-E1), exon 2 only (VHL_mt_E2), or exon 3 only (VHL_mt_E3). Each point represents a tumor sample. Box plot represents median, second, and third quartiles, and whiskers represent the top and bottom quartiles. (c) Pairwise statistical comparisons of PSI scores using nonpaired t test: *PSI values were significantly lower in the VHL_mt and VHL_mt_E2 groups compared with the All group. **PSI values were significantly higher in the VHL_mt_E3 group compared with the VHL_mt_E2 group. N.A., not applicable; N.S., not significant.

Discussion

We identified a synonymous VHL mutation (c.414A>G, p.P138P, NM_000551.3) that confers susceptibility to PHEO and/or VHL disease. These results have implications for clinical genetic testing and surveillance and also offer insights that may be relevant for understanding mechanisms of tumorigenesis mediated by VHL disruption.

The most prominent phenotype of the five families carrying the c.414A>G variant was the high-risk of PHEOs, present in all clinically affected individuals and resembling classic VHL-deficient tumors. In three families with long-term follow-up (18 to 30 years), hemangioblastomas were later diagnosed in one of the PHEO carriers (14% to 33% of affected individuals in each family), but no RCC was detected in these five families. We noticed a predominance of men among the affected individuals in this series. In one family, we documented three asymptomatic c.414A>G carriers older than age 45, including an 84-year-old woman, suggesting that penetrance of this allele was not complete. As sex preference and low penetrance are not features of typical VHL disease (15); these findings have to be validated by observation of additional families.

Our data show that the c.414A>G variant promotes exon 2 skipping, despite its location away from the canonical splice-site regions (Fig. 1b). Exon 2 encodes for a substantial portion of the β domain necessary for HIF binding and its ubiquitin-mediated degradation under conditions of oxygen sufficiency (2). Thus, a VHL transcript lacking exon 2 would be expected to produce a protein that lacks the ability to recognize its substrate, leading to HIF accumulation and induction of its downstream targets, as seen in classic loss-of-function VHL mutations (4, 26). In favor of the key functional relevance of this domain, selective deletion of exon 2 is sufficient to cause VHL disease (27). Furthermore, expression of a recombinant VHL form lacking exon 2 in cells and in mouse xenografts supports an oncogenic role of this product (5). However, it remains to be established if this transcript is, in fact, translated and functional in vivo (28).

We found that the WT VHL allele is deleted in tumor DNA from patients carrying the c.414A>G variant, a pattern that is similar to other PHEOs with classic loss-of-function VHL mutations. In addition, our expression data from two variant-carrying tumors agree with results from a third PHEO from an independent family carrying the same variant (7), which shows a pseudohypoxic profile indistinguishable from classic VHL-mutant tumors. These results argue for loss-of-function VHL being the main mechanism of tumorigenesis as a result of the c.414A>G variant, as opposed to a dominant-negative effect of the shorter isoform. However, currently, we cannot completely exclude that; if translated, this shorter variant might possess functions that contribute to tumor development.

Our analysis of pan-cancer data suggests that variants targeting exon 2, regardless of being located near splice sites, appear more vulnerable to exon 2 skipping. This finding is in agreement with earlier observations in renal carcinoma (26) or VHL disease (7). Of note, the pan-cancer dataset comprises only somatic variants and is biased toward frameshift/truncating mutations. In contrast, type 2 VHL disease, the clinical subtype associated with PHEOs, is typically enriched for missense VHL mutations. Evaluation of exon 2 exclusion in datasets that involve PHEOs and/or tumors from patients with VHL type 2 disease should be conducted to re-examine whether the trend that we observed in sporadic tumors is validated in these clinical contexts.

Based on our findings and those of Lenglet et al. (7), we propose that the c.414A>G variant status be modified from “variant of unknown significance” to “pathogenic.” Currently, the clinical phenotype associated with this variant is not fully defined. Long-term follow-up and detailed clinical evaluation of these and additional families will be required to establish the penetrance, phenotypic range, and disease aggressiveness, resulting from the c.414A>G variant. Until then, we recommend that carriers undergo full-surveillance protocols for VHL disease. As the knowledge improves, more specific recommendations for screening may be developed. Importantly, these results suggest that synonymous VHL variants may explain the phenotype of other patients with PHEO and VHL disease for which no pathogenic VHL mutation has been identified. Of practical translational value, as the abnormal ratio of VHL isoform transcripts was promptly detected in peripheral leukocyte cDNA of c.414A>G variant carriers, regardless of their clinical status, we propose that this approach can be adopted for new germline variants in which there is a suspicion of abnormal splicing and when frozen tumor is unavailable for analysis.

More broadly, our analysis of somatic mutations from cancer datasets suggests that variants targeting VHL exon 2 may be prone to exon 2 skipping. Recently, a large screen of rare genetic variants revealed a high frequency of splice aberrations, resulting from noncanonical changes that were often not a priori predicted to alter splicing (27). In addition, synonymous variants from the pan-cancer datasets were found to contribute to cancer to a much higher degree than previously suspected, often through aberrant splicing (8). Our current findings support both observations and further suggest that exon 2 exclusion may be an underappreciated mechanism to inactivate VHL functionally.

Glossary

Abbreviations:

HIF

hypoxia-inducible factor

MAX-like

related to the Myc associated factor X gene

PHEO

pheochromocytoma

PSI

percentage spliced in

RCC

renal cell carcinoma

RTK

reeptor tyrosine kinase

SDH

succinate dehydrogenase

TBP

TATA-binding protein

TCGA

The Cancer Genome Atlas

UMAP

Uniform Manifold Approximation and Projection for Dimension Reduction

VHL

von Hippel-Lindau; WT-wild type