METHODS

Subsequent to the original publication of the TCO linkage on chromosome 19p13.2,¹⁰ we have identified eight additional families in which at least one tumour was noted to have cell oxyphilia, including a large fNMTC pedigree from Austria (Tyrol) with six patients with NMTC and 12 patients with goitre (fig 1). On the basis of the presence of the oxyphilia phenotype, these families were excluded from our NMTC1 mapping study.¹³ Additionally, we identified two families from our NMTC1 mapping study¹³ as being relevant here: one small family (201; three affected) identified as having “tall cells”, a histological phenotype very similar to oxyphilia,¹⁰ and an informative Italian family (102) in which all tumours have not yet been examined in order to include/exclude the presence of oxyphilia or fvPTC. Histopathological diagnosis was confirmed in all cases. The families are summarised in table 1. Blood samples and family history were obtained from family members after informed consent and DNA extracted (Puragene Kit, Gentra Systems) The protocol for this study was approved by the French committee for the Protection of Persons in Biomedical Research.

Genotyping

As nine of the 10 families were known to have at least one tumour with cell oxyphilia (or a similar feature), we genotyped these families for the TCO region on chromosome 19 as well as the NMTC1 region on chromosome 2q21. Microsatellite markers genotyped in these families included on 2q21: D2S2224, D2S2271, D2S2215, D2S2160, D2S112, AFMa272zg9, D2S2256, D2S114, D2S368, D2S2196; and on 19p13: D19S1034, D19S901, D19S884, D19S391, D19S916, D19S586, D19S394 using standard PCR techniques with fluorescent primers and genotyped using a 377 Genetic Analyzer (Applied Biosystems).

Statistical analysis

Single-locus linkage analysis

A multipoint LOD score was calculated using the Vitesse analysis program¹⁴ restricted to three markers on chromosome 19p13 (D19S884–2.5 cM–D19S391–4.3 cM–D19S586) and three on chromosome 2q21 (D2S2215–3.45 cM–D2S1260–3.69 cM–D2S112) due to the size of one of the families. Genetic distances were taken from Marshfield genetic maps¹⁵ or the genetic distance database¹⁶ and the marker order was confirmed using the human genomic sequence.¹⁷ Allele frequencies were those used in our previous linkage studies of these regions^9,13 calculated from the larger data sets using one individual chosen from each family at random.

The genetic model chosen for each locus assumed a frequency of 0.01 for a dominantly inherited susceptibility allele at each locus and a 15-fold increased risk of NMTC (including follicular and Hurthle cell adenoma) above the assumed background rate of 0.2%. In addition, we assumed that the susceptibility allele resulted in a three-fold increased risk of goitre independent of age or sex, with a prevalence in non-carriers of the allele of 5%. Although this model is no doubt an oversimplification of the true underlying genetic basis, it captures the essential features of the effects of these two loci and satisfies constraints imposed by observed familial risk and disease prevalence. Under this model, each locus would be expected to produce a familial relative risk of 2.16, which represents about 37% of the overall familial risk of 8.0 under a multiplicative model. Together these two loci are predicted to account for 58% of the total observed familial risk of NMTC under a multiplicative model, thus leaving room for one or more additional genes contributing to the familial aggregation.

Two-locus linkage analysis

Two-locus analysis was carried out by the TMLINK program.¹⁸ We assumed the joint presence of the TCO and NMTC1 susceptibility loci described above, and further assumed that the loci acted multiplicatively in terms of cancer risk (that is, an individual carrying both susceptibility alleles would have a 225-fold increased risk) and additively in terms of the risk of goitre. We also performed analyses in which only patients with cancer/adenoma were considered as affected and looked at both multiplicative and additive models. In addition to the two locus analyses, we also calculated the LOD scores at each locus, assuming the other locus was in the model. Due to the size of the Tyrolean (103) family, to allow the calculation of a two-locus LOD score that used all available genetic data, the haplotypes of the family were recoded as a single informative locus with each haplotype assumed to be rare (fig 1A,B).

RESULTS

Evidence for linkage at the TCO and NMTC loci

For the 19p31.2 region, thought to contain the TCO locus, the overall maximum multipoint heterogeneity LOD score for the three markers in this region for the 10 families was 1.56 at marker D19S884, with an estimated 70% of the families linked to this region, providing further evidence of the presence of a fNMTC susceptibility gene at this locus. This was not surprising, as most of the families were selected for the presence of oxyphilia, which is a known phenotypic marker of the TCO locus. Somewhat more surprising, however, was our finding of linkage to the NMTC1 locus in these pedigrees. The total LOD score using the three markers tested in this region was 2.85 at marker D2S2215, with no evidence of locus heterogeneity. This provides further strong evidence for the role of a locus in this region of 2q21 for fNMTC susceptibility. A major portion of the linkage signal at both 19p12 and 2q21 was due to a large Tyrolean pedigree (family 103). On 2q21, in this family (103) a patient who was affected with goitre (patient 103.27) is a recombinant. This patient was assumed to have received the “unaffected” haplotype because if they had received the “affected” haplotype this would be in contradiction with the critical 2q21 region outlined in the NMTC1 study.¹³ Figure 1 shows the haplotypes for both regions in this family.

Evidence for interaction between the TCO and NMTC1 loci

The finding of linkage to these two previously identified regions opens up the possibility that these two loci interact to produce a high risk of NMTC and other thyroid pathologies. In order to assess simultaneously the possibility of two loci interacting in these pedigrees, we performed two-locus linkage analysis in the 10 families using the genotype data from the TCO and NMTC1 loci. Table 2 summarises the results of the two-locus linkage analyses performed. Under the model with both cancer and other thyroid pathology considered as affected, the two-locus analysis resulted in an overall LOD score of 3.92, with a value of 4.51 achieved under heterogeneity at α = 0.70, indicating that there is strong evidence in favour of a two-locus inheritance model. The large Tyrolean family again contributed significantly to the LOD score, with an individual two-locus LOD of 3.21. Comparison of the individual LOD scores at each region showed a concordance of linkage results in nine out of the 10 families, with a Spearman rank correlation between the family LOD scores of 0.73. This is also reflected in the fact that when considering cancer only, the multiplicative model seemed to fit better than the additive model, at least for the particular models examined.

The notion of two susceptibility loci interacting in these pedigrees is further supported by an examination of the segregating haplotypes of the families in which the sharing of the NMTC cases is consistent for linkage with both loci. In the large Tyrolean family (fig 1), we see that all six cases of NMTC share both the chromosome 19 and chromosome 2 haplotypes. For the potentially related phenotype of goitre, in 5/12 cases they shared both loci, in four shared NMTC1 only but not TCO, and in the remaining three shared TCO but not NMTC1. Thus, all 12 cases of goitre carried at least one of the linked haplotypes. A similar haplotype segregation pattern is present among most of the other smaller linked families (fig 2 and data not shown).

DISCUSSION

The linkage analysis performed in the 10 families presented here provides further evidence that 19p13 and 2q21 harbour susceptibility genes for fNMTC. Among the 10 pedigrees presented here there is significant evidence for a two-locus inheritance pattern, indicating that among a subset of fNMTC pedigrees the TCO and NMTC1 loci may interact to result in increased risk of disease. However, the TCO locus appears not to be relevant to the majority of fNMTC^9,11 and neither the French family used to map the TCO locus nor the Tasmanian family used to map the NMTC1 locus presented with any evidence for linkage at the 2q21 or 19p13 loci, respectively.^9,13 Therefore, it is unlikely that the TCO and NMTC1 loci contribute to NMTC risk together among the majority of fNMTC pedigrees. Nevertheless, our results indicate that such an interaction may occur in a reasonably large subset of fNMTC and a two-locus mode of inheritance is consistent with the observations made of fNMTC^5,6,8,9 and segregation analysis in other complex genetic forms of hereditary cancer fit a two-locus model.¹⁹ 2q21 has been noted as an area of interest in a genome wide search on a Canadian MNG1 family,²⁰ and it would be interesting to review these data in the context of the two-locus model outlined here, especially with respect to the three NMTC cases of the Canadian MNG1 family.²⁰ Similarly 2q21 was implicated in a large family with Gardners syndrome²¹ in which NMTC has been found to be an associated tumour.²² Given the reported lack of biallelic inactivation of the APC gene in the thyroid carcinomas associated with familial adenomatous polyposis²³ and the reported cytogenetic alteration in the NMTC1 region,²¹ it is possible that the NMTC1 locus could be put forward as a modifier of NMTC susceptibility in this inherited syndrome. Examination of interaction models between the identified fNMTC loci and other genes that have been associated with NMTC risk, for example RET²⁴ and more recently thyroglobulin²⁵ in the wider familial NMTC without oxyphilia and sporadic NMTC may also be warranted.

Genetic heterogeneity has been reduced in fNMTC family sets by stratification on the presence of the histological variants of oxyphilia⁷ and fvPTC.¹³ Some families outlined here (for example families 118, 128, and 201; table 1) present with both the oxyphilia and fvPTC phenotypes. The evidence for linkage at both loci in this subset is therefore consistent with a stratification based on the presence of both the histological variants. Thorough histopathological data on all affected members of all pedigrees was not available for independent review to explore this observation further.

Loss of heterozygosity (LOH) on chromosome 2 has been noted in oxyphilic NMTC tumours²⁶ and at 2q22²⁷ suggesting further the NMTC1 locus may be implicated in the pathology associated with the TCO locus. In our own material from sporadic NMTC tumours, 35/71 (49%) oxyphilic tumours and adenomas and NMTC with oxyphilic features exhibited LOH at 2q21 while only 38% of these tumours showed LOH at 19p13. Similarly in 20 follicular adenomas/carcinomas, 30% and 15% showed LOH at TCO and NMTC1, respectively (Stankov, unpublished data). However, in neither of the two sets of tumours was the number of tumours showing concordant LOH at both loci different from that expected under independence. Unfortunately, material was unavailable to test for allelic imbalance in the familial cases of this study.

These results provide further evidence for fNMTC susceptibility genes at 2q21 and 19p13.2. Furthermore, there is evidence that TCO and NMTC1 may interact to increase risk in individuals that inherit both susceptibility genes. The identification of the NMTC susceptibility genes at both 2q21 and 19p13.2 will allow the contribution of these two genes to the risk of developing NMTC to be fully characterised.