SUBJECTS AND METHODS

Patients and clinical evaluation

Seventeen affected probands with a family history of VHL (15 from Brazil, one from Portugal, and one from Ecuador) and three Brazilian VHL patients without a family history (defined by clinical criteria) were included in this study. Fifteen of these were clinically evaluated at the Department of Oncogenetics of the Hospital do Câncer - Sao Paulo, Brazil, and the others were referred by other centres that were following the families. All families received appropriated pre- and post-test genetic counselling.

Blood samples from the affected probands were collected following receipt of signed informed consent. Clinical status was determined by physical examinations (including ophthalmoscopy), radiological evaluations, and laboratory testing according to methods described elsewhere.^3,5 Previous medical records of affected members and follow up were reviewed and registered in a database.

Amplification and sequencing of the VHL gene

Genomic DNA was extracted from 10 ml of peripheral blood using phenol-chloroform by standard procedures after separation of mononuclear cells and stored at 4°C. The three exons of the VHL gene were amplified by PCR in a total volume of 25 μl containing 50 ng of template DNA, 1 × PCR buffer (20 mmol/l Tris-HCl, 50 mmol/l KCl), 200 μmol/l each dNTP (Promega, USA), 1.5 mmol/l MgCl₂, 0.25 units of Taq DNA polymerase (Gibco-BRL Life Technologies, USA), and 0.5 μmol/l each primer set. Primers used for the amplification of the coding region of the VHL gene were: 1F 5‘ CCATCCTCTACCGAGCGCGCG 3‘ and 1R 5‘ GGGCTTCAGACCGTGCTATCG 3‘ for exon 1, yielding a fragment of 522 bp; 2F 5‘ TGCCCAGCCACCGGTGTG 3‘ and 2R 5‘ GTCTATCCTGTACTTACCACAACA 3‘ for exon 2, 232 bp; and 3F 5‘ CACACTGCCACATACATGCACTC 3‘ and 3R 5‘ ACTCATCAGTACCATCAAAAGCTG 3‘ for exon 3, 383 bp. Reactions were incubated at 95°C for five minutes and amplified for 35 cycles of 95°C for one minute, 58°C for one minute (exons 2 and 3), or 68°C (exon 1) for one minute, and 72°C for one minute, with a final extension cycle of 72°C for five minutes. PCR products were sequenced in both strands with Big Dye terminators (Applied Biosystems) and sequences were collected and analysed on an ABI Prism 377 (Applied Biosystems). The same PCR primers were used in the sequencing reactions, with the exception of exon 3, which was sequenced in the sense direction with an internal primer 3I 5‘ CGTTCCTTGTACTGAGACCCT 3‘.

Southern blot analysis

Germline rearrangements involving part or the entire VHL gene were detected by quantitative Southern blotting using g7 cDNA¹¹ and a β globin probe, as control, as previously described.³³ In brief, 10 μg of genomic DNA were digested with the endonucleases EcoRI and AseI and DNA fragments were separated by electrophoresis. The DNA was transferred to a nylon membrane and hybridised to random primer labelled probe specific for the VHL gene (g7) and for the β globin gene. Filters were exposed to x ray film (Kodak X-AR, USA) in a film cassette with two intensifying screens at −70°C for one to three days. Quantification was possible by comparing the intensity between the two bands. DNA from normal subjects and from patients with known rearrangements in the VHL gene (kindly provided by Dr Catherine Stolle, University of Pennsylvania, USA) was used as negative and positive controls, respectively. In normal subjects, a single 9.7 kb EcoRI and AseI fragment is detected by hybridisation with the g7 probe, whereas in subjects with partial deletions of the VHL gene, a slower or faster migrating fragment was detected in addition to the normal sized fragment. Complete deletion of the VHL gene produced a single 9.7 kb band with 50% relative intensity after normalisation, which was confirmed by densitometry of the bands.

RESULTS

Twenty VHL probands and their families were evaluated in this study, of whom 17 presented with a family history (15 from Brazil, one from Portugal, and one from Ecuador). Fifteen familial VHL and two VHL patients were classified as having type 1 VHL (without PH) and two familial VHL and one VHL patient were classified as having type 2 VHL (with PH). Ten families were large enough (four or more affected subjects from at least two generations) for full phenotype analysis (table 1).

We were able to define germline VHL gene mutations in all probands who fulfilled the clinical criteria for VHL. In 16 cases (80%), these were point mutations involving either the coding region or exon boundaries of the VHL gene (nine missense, three frameshift, one in frame deletion, two splice defect, one nonsense), while four (20%) were large deletions of which three were partial deletions and one a complete deletion of the VHL gene (table 1). Eight of the 16 point mutations (50%) have not been previously described and thus represent novel mutations. The point mutations were not distributed uniformly along the VHL gene and most were concentrated within exons 1 and 2 (87.5%). The 13 point mutations associated with type 1 VHL occurred in this region in contrast to the findings of others.^29,31,42 We were able to analyse the germline mutation linkage to disease in nine families (families 2, 8, 9, 12, 14, 15, 16, 18, and 20). Four of these families carried known point mutations described before by others, two presented large deletions, and three had novel point mutations. The novel mutations 238A>C, 344delA, and IVS1+7G>A were detected in five, three, and two affected family members, respectively. The pathogenic nature of the IVS1+7G>A remains to be confirmed at the mRNA level; however, the fact that this alteration was not detected in one unaffected first degree relative (after clinical and radiological screening) suggests that it may be associated with the disease. Additionally, none of the novel mutations were represented in 120 alleles from normal subjects in the Brazilian population suggesting that these base changes are not common polymorphisms.

Linkage analyses were not possible in eight of the VHL families owing to the absence of live family members (families 1, 3, 4, 5, and 7) or relatives alive but not assessable (families 11, 17, and 19). Only in family 1 was genetic testing performed in a non-symptomatic boy, since no relatives were alive for testing. Although many members of this family had already presented with HB, the diagnosis of VHL was confirmed after necropsy of the proband’s father, who died at 37 years from complications following emergency surgery for cerebellar HB. The necropsy showed many occult lesions related to VHL, including RC, multiple pancreatic and renal cysts, and cystadenoma of the epididymis.

From the eight families in which linkage to disease could not be established, four carried mutations previously described as having a deleterious effect (table 1), two had partial deletions of the VHL gene, and two had novel missense mutations. Although these two missense mutations have not been previously described, they affect amino acid residues for which other different mutations have already been described.^28,29,31,43 Additionally, these mutations (Trp88Arg and Val130Phe) affect the HIF-1α binding site within the β domain of pVHL and the hydrophobic core of pVHL responsible for the structural integrity of the β sheets, respectively.³⁷ Thus, in both cases these novel mutations are predicted to result in significant alteration of pVHL function.

Three novel germline mutations were detected in all VHL patients with no family history: frameshift mutation 305delC, frameshift mutation 436delC, and missense mutation Thr124Ile. Both novel frameshift mutations introduce a premature stop codon leading to a truncated protein and the missense mutation was not detected in normal subjects from the Brazilian population. Following parental testing, two VHL patients were confirmed as having de novo VHL (families 6 and 10), since their mutation was not found in their parents (although no test to confirm the paternity was performed). However, in family 13, analysis of the proband’s parents showed that the mutation 436delC was present in two independent samples from his asymptomatic mother and thus it cannot be considered a de novo mutation. On complete clinical investigation, however, the proband’s mother was shown to have two occult HBs (one cerebellar and one spinal) as well as multiple pancreatic cysts.

In addition to the mutations described above, two variants that apparently do not interfere with the protein structure were found in three families: the synonymous mutation Pro61Pro (in families 9 and 17), previously reported as a known polymorphism, and the variant IVS2-94T>A (in family 16) that has not been previously reported. In all the families, except family 15, the variants did not segregate with the related VHL gene mutation. The allele frequency of each of these variants was shown to be at least 1% in normal DNA samples from the Brazilian population (a minimum of 120 alleles evaluated) (table 1). In contrast, none of the novel point mutations described in the present study were detected in normal subjects.

Genotype-phenotype correlations were consistent with those reported in other studies. Most of the mutations found in the 17 families classified as type 1 VHL resulted in a truncated protein (66%), whereas all three mutations found in families whose phenotype included PH (type 2 VHL) predicted a full length protein. There was no correlation between the type of germline mutation and the occurrence of HB or RA, since these manifestations were present in almost all families analysed. Two families harboured distinct mutations in codon 167, a VHL hotspot associated with high risk of RC, HB, and PH (type 2B). However, RC was not observed in these families, and we considered that a larger number of affected members and a longer period of follow up will be necessary for full phenotypic characterisation in these families. Eleven out of 17 type 1VHL families had a RC phenotype (table 1). Almost all mutations detected in these families affect residues in the HIF-1 binding site domain (residues 67-117), residues in the hydrophobic core important for VCB complex assembly, or are associated with the presence of a truncated form of pVHL. Interestingly, we have observed in large VHL families with a RC phenotype that although the principle cause of mortality in previous generations was HB, in younger generations this is now mostly the result of RC (data not shown).

DISCUSSION

We have studied von Hippel-Lindau disease in Brazil since 1998, and have identified 20 families with VHL. Many studies in Japanese and western populations have highlighted differences in the spectrum of VHL germline mutations around the world.^28,31,42 The heterogeneity of the Brazilian population, consisting mostly of descendants of Portuguese immigrants, African, and Brazilian Indians raises the possibility of a different spectrum of VHL germline mutations and risks of VHL associated lesions. The VHL families included in this study are representative of our population, since they came from many different regions of the country, including the south east, north east, and the central region. Also, one family from Ecuador and another from Portugal were included in our study. To our knowledge this is the first study in Portuguese speaking families and the first in Latin America.

The spectrum of known mutations in the VHL gene is enormous and includes missense and nonsense mutations, deletions or insertions of one to several nucleotides, splice site mutations, and rearrangements consisting of partial or complete deletion of the gene. Here, we found that by complete sequencing of the coding regions of the VHL gene, followed by quantitative Southern blotting when a sequence mutation could not be detected, we were able to reproduce the results obtained by Stolle et al,³³ in that we detected germline mutations in all 20 distinct VHL families analysed. Because of the relatively short length of the VHL gene, no screening method based on single stranded DNA conformation, such as SSCP or CDGE, was used before sequencing. Since not all investigators achieved the same rate of mutation detection, we consider that both methodology and patient selection are important for achieving a high efficiency of mutation detection in VHL. We believe that the high rate of mutation detection achieved in this work can also be explained by the stringent clinical criteria used for family selection. All patients selected for the study had the diagnosis of VHL confirmed by clinical, pathological, or radiological evidence.

We found that molecular screening was particularly invaluable in situations where the diagnosis of VHL could not be achieved by clinical criteria alone, such as where family data were not available, or where patients had a negative family history. In addition, the genetic test was important to determine the status of the parents in the apparently de novo VHL patients. In one case, it was possible to detect the same mutation in the mother of the patient, with implications for the risk assessment and screening of other family members. Indeed, we emphasise here that the study of the patient’s parents is essential for the characterisation of de novo mutations, and for risk assessment and counselling of other relatives. Recent studies have shown that about 23% of probands have parents and sibs with no history of the disease at the time of the diagnosis.⁹

We have contributed eight novel germline mutations, including four missense mutations, three frameshift mutations, and one splice site mutation (table 1). Five novel mutations occurred with a familial pattern. However, two of them could not be analysed further in other family members: the missense mutation Trp88Arg, because the mother of the proband had died and was adopted, and the missense mutation Val130Phe, because there were no available relatives at the time of the study. Although the mutation IVS1+7G>A does not affect the invariant 3‘ AG splicing donor site, segregation and population analysis presented in this work suggests it can be a pathogenic mutation. Splicing mutations outside the invariant receptor or donor site have been reported in many cancer susceptibility genes, including VHL, TP53, BRCA1, and hMLH1, and they represent as many as 27% of the splice sites mutations reported in these genes (HGMD online). None of the novel point mutations described in this study were represented in the normal population, excluding the possibility that they might be common polymorphisms.

Germline mutations in the VHL gene are distributed throughout the coding region of the gene. Although other groups have found that the germline mutations of the VHL gene tended to concentrate in exon 1 and exon 3, we observed in our series that most point mutations (87.5%) occurred in exon 1 (43.7%), exon 2 (31.3%), and the boundaries of these exons (12.5%). We found large deletions of the VHL gene in four of our families (20%), including three partial deletions and one complete deletion. Previous studies showed a frequency of partial VHL gene deletion of 25% and complete VHL gene deletion in 9% in the VHL families.³³

In general, genotype-phenotype correlations observed in this study were consistent with those reported by others.^{12,29,30,36,44} It is known that distinct phenotypes can result from almost identical mutations.⁴⁵ Here we found that the mutation Ser80Arg detected in a large family conferred a high risk for RC, HB, and RA, but no apparent risk for PH. A similar missense mutation in this same codon, Ser80Gln, has been reported to be associated with type 2C VHL (only PH). ⁴⁶ It should be emphasised that new genotype-phenotype correlations are emerging for VHL and further studies such as the one presented here are necessary. Recently functional studies have shown that phenotypes could be related to loss of specific pVHL functions.

Here, we observed an increased risk of RC in most of our families, based on the family phenotype or the presence of high risk mutations. Most of these mutations directly affected the HIF-1 binding site of the pVHL and/or the proper assembly of the VCB complex. Functional studies strengthen the notion that HIF-1 deregulation plays a causal role in HB and RC and contributes to RC pathogenesis in the setting of VHL disease.^38,47,48

Although the phenotype in VHL may vary among families, intrafamilial variability can also be observed,^34,35 suggesting that other factors could be involved, such as environmental exposure, modifier genes, or single nucleotide polymorphisms (SNPs).⁴⁹ All these factors could influence the clinical spectrum of VHL in different populations, and so far there is no satisfactory explanation for the clinical heterogeneity observed in VHL. Recently, the cyclin D1 genotype was investigated as a modifier gene in VHL, and it was found that the presence of the nucleotide G at position 870 was associated with an increased risk of RA and HB.^50,51 More molecular and epidemiological studies will be necessary for complete understanding of the phenotype associated with VHL.

The improvement of life expectancy of VHL patients in recent years is the result of the refinement of image based diagnostic methods and the development of screening programmes. We argue that this trend is now likely to be strengthened by the incorporation of accurate mutational analysis, risk assessment, and genetic counselling. Here, we have confirmed that it is now reasonable to expect to find VHL mutations in all affected subjects, but that the spectrum of mutations is highly population dependent. Furthermore, the phenotypic implications of specific mutations are also likely to be influenced by the genetic background of the population. Thus, we have contributed eight novel germline mutations in the VHL gene found in the Brazilian families whom we suspect may have unique features in respect to risk of malignancies. Continued, careful study of these and other VHL families in Latin America will be required to confirm this altered disease profile, which would have significant implications for genetic counselling and screening programmes of VHL disease on this continent.