Multiple Single Nucleotide Polymorphisms on Human Chromosome 19q13.2–3 Associate with Risk of Basal Cell Carcinoma¹

SNPs³ are the plankton of genetics. Individually they are tiny, 1-base differences in a 3-gigabase human genome, but by their multitude they constitute a rich source for genetic research. In the wake of the Human Genome Project, myriads of SNPs have surfaced, and it is now estimated that a SNP occurs approximately every 1000 bases. So no matter which part of the genome you are interested in, there are likely to be multiple SNPs located in the vicinity.

SNPs often effectively dichotomize human populations, making them useful for tracking genes of importance to disease in outbred populations by linkage disequilibrium. Specifically, typing of SNPs may facilitate the locating of regions of previously unknown importance for cancer risk. Thus, we have previously reported that polymorphisms in the gene XPD seemed associated with the occurrence of BCC (1). This finding has later been corroborated by the association of polymorphisms in XPD with BCC in a different population (2) and by the association of markers in the same gene with three other cancers, malignant melanoma (3), glioma (4), and lung cancer (5).

In this paper, we present evidence that alleles of several other polymorphisms in the chromosomal region 19q13.2–3, encompassing the genes RAI and XPD, are associated with occurrence of BCC. We are therefore convinced that a chromosomal variation influencing the risk of getting BCC and possibly other cancers must be located in this region, and we see the present paper as a first step toward identifying this variation.

Of the genes that we have investigated, XPD, ERCC1, and LIG1 relate to DNA repair and are probably directly involved in preventing cancer. FOSB, which is a homologue of an oncogene, and GLTSCR1, which may be a tumor suppressor gene, might well also be involved in carcinogenesis or its prevention. For RAI, which seems to be involved in control of transcription, one might construct a relation to cancer prevention, but no experimental data to that effect are available. The remaining genes are unlikely to play a direct role in cancer. They were chosen simply because they were located in the chromosomal region of interest.

The groups of Caucasian Americans with and without BCC have been described previously (2, 6). DNAs for the analyses were purified from frozen lymphocytes obtained from blood.

Genotypes of polymorphisms in CKM exon 8 (position 20076, rs#4884) and RAI intron 1 (position 875, rs#1970764), ERCC1 exon 4 (position 19007; Ref. 7), FOSB exon 4 (position 34621, rs#1049698), SLC1A5 exon 8 (position 60620, rs#1060043), LIG1 exon 6 (position 111, rs#20580), and GLTSCR1 exon 1 (position 20775, rs#1035938) were detected using LightCycler (Roche Molecular Biochemicals, Mannheim, Germany; Ref. 8). The positions refer to the following accession numbers in GenBank: RAI, L47234; CKM, AC005781; ERCC1, M63796; FOSB, M89651; SLC1A5, AC008622; LIG1, L27710; and GLTSCR1; AC010519. The rs numbers refer to National Center for Biotechnology Information’s database over SNPs, dbSNP. PCR was performed by rapid cycling in a reaction volume of 20 μl with 0.5 μm each primer, 0.045 μm anchor and sensor probe, 3.5 mm MgCl₂, 7 ng of genomic DNA, and 2 μl of LightCycler DNA Master Hybridization probe buffer (Roche Molecular Biochemicals, catalogue number 2158 825). This buffer contains Taq DNA polymerase, dNTP mix, and 10 mm MgCl_2. Table 1 shows the design of primers and fluorogenic probes for LightCycler. Some of the primers were copied from Ref. 8. Hobolth DNA (Hillerød, Denmark) and TIB-Molbiol (Berlin, Germany) synthesized the primers and probes, respectively. In some cases, the reaction mixture also contained 5% DMSO. The temperature cycling consisted of denaturation at 95°C for 2 s, followed by 46 cycles consisting of 2 s at 95°C, 10 s at 57°C, and 30 s at 72°C. The last annealing period at 72°C was extended to 120 s. The melting profile was determined by a temperature ramp from 50°C to 95°C with a rate of 0.1 degree/s. For RAI intron 1, we ran the melting profile three times and used the last curve.

The polymorphisms in XRCC1 exon 7 (position 26651; Ref. 7), XRCC1 exon 17 (position 36189; Ref. 7), RAI exon 6 (position 8786, rs#6966), and XPD intron 4 (position 19244, rs#1618536) were typed simultaneously on a ABI Prism 310 sequenator (Applied Biosystems, Foster City, CA) using the SNaPshot technique (9). The position numbers refer to the following accession numbers in GenBank: XRCC1, L34079; RAI, L47234; and XPD, L47234. The relevant primers are listed in Table 2. The PCR reaction consisted of 1 μl of purified genomic DNA, 1 pmol of each primer (DNA Technology, Aarhus, Denmark), 12.5 nmol of each dNTP (Bioline, London, United Kingdom), 100 nmol of MgCl₂ (Bioline), and 0.15 μl of BIOTAQ DNA polymerase (Bioline) in a total volume of 20 μl of water. The program consisted of 4 min at 96°C, followed by 25 cycles of 96°C for 30 s, 60°C for 30 s, and 72°C for 60 s. The last cycle was followed by 72°C for 6 min. The primers and dNTPs were removed in reactions containing 2 units of SAP (Roche Molecular Biochemicals), 2 units of exonuclease I (New England Biolabs, Beverly, MA), and 9 μl of PCR reaction in a total volume of 14 μl of water. The reactions were incubated at 37°C for 60 min and 72°C for 15 min. The SNaPshot reactions contained 1 μl of SNaPshot Ready Reaction Mix (Applied Biosystems), 0.5 μl of each of the SNaPshot primers (XRCC-e7-ss1, 4 pmol/μl; XPD-i5-cp1, 0.5 pmol/μl; RAI-e7-cp1, 1 pmol/μl; XRCC-e17-ss1, 2 pmol/μl), 2 μl of the purified PCR product, and 1.5 μl of buffer [200 mm Tris-HCl and 5 mm MgCl₂ (pH 9.0)]. The reactions were cycled 25 times: 96°C for 10 s, 50°C for 5 s, 60°C for 30 s. The primers and dNTPs were removed in a reaction containing 1 unit of SAP, 0.8 μl of 10× SAP buffer, and 5 μl of SNaPshot reaction in a total volume of 8 μl of water. Two μl of purified product were added to 10 μl of concentrated deionized formamide (Amresco), incubated for 5 min at 95°C, and analyzed on the sequenator. The two markers in XRCC1, in exon 7 and exon 17, could not be reliably scored, and thus we excluded these.

The polymorphism in XRCC1 exon 10 (position 28152 - A399G, Ref. 7) was analyzed using the ABI Prism 7700 sequence detection system (Applied Biosystems). The position refers to GenBank accession number L34079. PCR primers and Taqman probes (Table 3) were designed using Primer Express v 1.0 (Applied Biosystems). TAG-Copenhagen Aps (Tagc.com, Copenhagen, Denmark) synthesized the primers, and Applied Biosystems synthesized the fluorescent Taqman probes. The reactions were performed in MicroAmp optical tubes sealed with MicroAmp optical caps (Applied Biosystems) containing a 10-μl reaction volume (1× Taqman buffer A; 2.5 mm MgCl₂; 200 μm each of dATP, dCTP, and dGTP; 400 μm dUTP; 800 nm each primer; 200 nm each probe; 0.01 unit/μl AmpErase UNG; and 0.025 unit/μl AmpliTaq Gold Polymerase). Thermal cycler conditions were as follows. Tubes were incubated at 50°C for 2 min followed by 10 min at 95°C. The incubation was succeeded by 45 cycles of 95°C for 15 s and 64°C for 1 min. One sample of each genotype from a different cohort (1) previously analyzed for XRCC1 exon 10 (results not published) using restriction enzyme analysis (10) was used as reference sample.

Data recording and tabulations as well as ANOVA analysis were performed with the program SPSS (SPSS, Chicago, IL). Fisher’s exact test was performed online.⁴ Bonferroni adjustment was performed online.⁵ A global P was calculated from the individual levels of significance (p_i) using the formula:

where ln is the natural logarithm, n is the number of tests, and 2n is the degrees of freedom for χ^2. The χ^2-test was performed using Excel (Microsoft, Redmond, WA). The same program was used to calculate the odds ratios and related parameters. We used the program Arlequin,⁶ which simulates the genotype data using Markov chains, to calculate the haplotype frequencies and the Ps for linkage disequilibrium. The haplotype frequencies were then converted to Lewontin’s normalized linkage disequilibrium D’ using Excel and the formulas given online.⁷

We have investigated the occurrence of the alleles of 10 new SNPs on chromosome 19q13.2–3 among Caucasian Americans with or without BCC. There is no scientific reason why we have used different techniques for typing different polymorphisms. The choices simply reflect which equipment was available to us at the times. Table 4 summarizes the results. Most SNPs showed no association with disease, but two markers (namely, RAI intron 1 and RAI exon 6) around the gene RAI, which is immediately 5′ to XPD, showed associations with BCC. In particular, the association of the RAI intron 1^A allele with cancer was fairly strong (P = 0.004; two-sided).

When we focused on those persons with early BCC (onset before 50 years of age), the difference in the distribution of the marker RAI intron 1 became even more pronounced (P = 0.0002; two-sided). As this would indicate, the average onset of BCC was later among cases carrying at least one RAI intron 1^G allele than among those homozygous for RAI intron 1^A [51.9 years (n = 10) versus 45.1 years (n = 43); P = 0.022; two-sided]. The heterozygotes also seemed to have fewer and later cases than expected. A similar analysis indicated that RAI exon 6^T also was associated with reduced risk of BCC. Not surprisingly, the two markers were in linkage disequilibrium, and the protective alleles were correlated.

Table 5 lists the frequencies of the rare alleles in controls and in all cases as well as in young cases (<50 years at onset). It is quite clear that the RAI intron 1^G allele is underrepresented in the cases, and there are suggestions that several of the neighboring SNP variants also differ in frequency between cases and controls.

Calculations of odds ratios with confidence intervals and Ps for comparisons of BCC among those with two RAI intron 1^A versus those with at least one RAI intron 1^G are shown in Table 6 for both all BCC cases and early-onset BCC. Clearly, the odds ratios were less than 1, in keeping with the idea that the RAI intron 1^G allele was protective. Similar calculations, where the cases and controls were subdivided according to their report of BCC occurring in their family, are shown in Table 7. The association of the SNP with the risk of BCC seemed to be present irrespective of family history.

We have made similar calculations for SNP RAI exon 6. Here the T allele was marginally associated with reduced risk of BCC (results not shown).

The previously reported values for DNA repair capacity (6) appeared linearly related to the marker RAI exon 6 in those persons without a family history for BCC [7.5 ± 2.2 (n = 110), 8.4 ± 2.8 (n = 37), and 9.4 ± 1.8 (n = 2) for AA, AT, and TT, respectively; P = 0.035, ANOVA]. Among those persons with a family history, there also was a relation to RAI exon 6, but it was nonlinear [8.1 ± 1.8 (n = 24), 6.4 ± 2.0 (n = 7), and 8.8 ± 0.9 (n = 2) for AA, AT, and TT, respectively; P = 0.043, ANOVA].

Table 8 lists the normalized linkage disequilibria between the markers in cases and controls together (this investigation and Ref. 2) and also lists the Ps for the linkage disequilibria. Many of the markers were in strong linkage disequilibrium. Presumably, in reality, they are located close together on chromosome 19. We have organized the markers so that strong disequilibria correspond to short distances. The map corresponds to the current National Center for Biotechnology Information map of chromosome 19.

In the present paper, we have searched for an association of BCC with 10 different SNPs on chromosome 19q13.2–3. The alleles of neighboring SNPs correlated moderately, and one or two of the SNPs were individually associated with disease.

Bonferroni adjustment is a formula that specifies how low the lowest of a series of Ps must be to make the tests as a whole significant, i.e., it is a way of correcting for mass significance when multiple tests are made. Bonferroni adjustment allows the different tests to be partly related and includes a way to correct for this by using the mean absolute correlation coefficient of the variables as a parameter. When we performed Bonferroni adjustment for the multiplicity of tests (in total, 10 tests) of variables with a mean absolute correlation coefficient of 0.14, the results indicated that any P < 0.007 would make the whole set significant at the level of 0.05. The P for an association of the marker RAI intron 1 with cancer was lower (P = 0.004), i.e., this association was still significant after correction for the multiplicity. Leaving out the correction for the correlation among the variables did not change this conclusion. We have also calculated the global P from the 10 Ps for associations between all markers and BCC using a formula (Ref. 1; see “Materials and Methods”) that assumes no correlation between variables. The result was a global P of 0.045. Thus, either way, the result is globally significant. Moreover, using the formula (1) on early skin cancers further reduced this global P considerably (P = 0.004). Consequently, it is unlikely that our results were caused by mass significance; rather they indicate the presence of a genetic variation influencing the risk of BCC.

In a previous paper we suggested that such a causative variation influencing the risk of skin cancer should be found in the 5′ end of XPD or 5′ to the gene (2). In the present paper we show that two markers next to or inside the gene RAI, located immediately 5′ to XPD, are associated with risk of BCC. Thus, a total of three closely linked markers seem to confer resistance: RAI intron 1^G; RAI exon 6^T; and XPD exon 6^C. As one would expect the three alleles associated with protection tended to occur together. Presumably, the causative gene variation is located nearby. The data suggest fairly strongly that the etiological mutation resides in a block of linkage disequilibrium that encompasses at least XPD and RAI. However, deviations from monotonicity are known to occur in maps of linkage disequilibrium, so the boundaries must be viewed with caution (11). Neither of the polymorphisms in RAI changes protein coding. Also, we would like to point out that a slight ambiguity exists with respect to the positions of the markers RAI intron 1 and RAI exon 6. Previous maps of the region put the markers at different locations in the gene.

Our results are strong evidence that occurrence of BCC, the most common cancer among Caucasians, is genetically nonrandom. This is not at issue with the fundamental concept that carcinogenesis is a stochastic process, but it does indicate that the parameters governing this process differ between persons. Ultimately, the treatment of the majority of the population as homogeneous vis-à-vis cancer may prove fallible.

The present data, which locate a gene influencing cancer occurrence to a certain region of chromosome 19q13.2–3, only relate to BCC. However, it is likely that whatever gene variation causes the differences also influences other forms of cancer. First, many protective mechanisms against cancer are rather general in nature, e.g., DNA repair. Secondly, a number of studies of other cancer forms have also found associations of markers in this region with disease (3, 4, 5). In this context, it should be noted that others have described evidence of a glioma tumor suppressor gene in the same general region (12). We think this is a different gene because the present map of chromosome 19 puts it at a fair distance (13). Neither XRCC1 nor LIG1 seems to play a role in our study.

These investigations were started with the notion that variations in nucleotide excision repair genes could influence the risk of getting cancer, and it is still quite possible that a difference in the 5′ end of XPD (for instance, in a regulatory region) modulates repair. Alternatively, RAI, which is known to play a role in transcription (14), may be involved in regulation of genes for DNA repair.

We have reported previously that the mRNA levels of XPD and ERCC1 in lymphocytes are closely correlated and also correlate with the DNA repair capacity in such cells (15). It is intriguing that the region described here as related to occurrence of skin cancer may span regulatory regions for both genes and that a genetic variation in this region is correlated to DNA repair capacity.

Definitive identification of the causative gene will likely require sequencing of the genomic region on disease-associated versus non-disease-associated chromosomes. We have embarked on this project.

We thank Thrine Schneidermann for expert technical assistance, Helle Binderup for help with the SNaPshot typing, and Dr. Olfert Landt (TIB-Molbiol) for advice on the design of probes for the LightCycler.