Reply to the Letter to the Editor from Wang

In Response: To date, all studies of the 15q25 locus have reported statistical associations of several single nucleotide polymorphisms (SNP) mapping in the CHRNA5 locus region with either lung cancer risk, smoking habit, or CHRNA5 mRNA levels, although the functional SNP(s) and the pathologic mechanism(s) have not been clarified.

In our work, we reported the novel observation of an association between CHRNA5 mRNA levels in normal lung and the D398N variant (rs16969968), a coding change in the nicotinic receptor that we confirmed to be associated with lung cancer risk (1) and that was linked to ∼2.5-fold lower CHRNA5 mRNA levels in individuals homozygous for the risk allele (N398; n = 69; P = 8.04e-06; Fig. 4 in ref. 1). Because the 15q25 region contains several polymorphisms showing high linkage disequilibrium with each other, we did not make conclusions about the existence of any biological association between D398N and CHRNA5 mRNA levels, as Wang et al. (2) have interpreted.

Wang et al. (2, 3) recently showed that several variants either 5′ upstream of or within the CHRNA5 gene are more strongly associated with the variability in CHRNA5 mRNA expression than is the rs16969968 SNP (D398N) in the human frontal cortex, with the insertion/deletion variant rs38411324, located in the 5′ region near the gene, showing the strongest association with CHRNA5 mRNA expression. The authors selected the tagged SNP rs588765, mapping in the intron 1 of the gene, for genotyping in all of the case-control series because of a higher genotyping success rate.

Using the same 69 samples, we previously analyzed and the PCR primers described by Bierut et al. (4), we carried out an association analysis of rs3841324 with CHRNA5 mRNA levels in normal lung; indeed, we found a stronger association of the rs3841324 than the rs16969968 (D398N; P = 1.25e-08; Fig. 1), consistent with the findings of Wang et al. (3) in brain tissue. However, given the small (only 25 kb) distance between rs3841324 and rs16969968, as well as the expected highly significant linkage disequilibrium of these polymorphisms (D′ = 1.0; r^2 = 0.45; n = 69; analysis by JLIN; ref. 5), the higher association of the rs3841324 than of the D398N with CHRNA5 mRNA levels would be better interpreted as an indication of a closer distance of the first polymorphism with the functional polymorphism affecting CHRNA5 mRNA levels, and not as a proof of biological association. Moreover, the diplotype analysis proposed by Wang et al. (2) cannot be considered sufficient to provide biological mechanism, because it also consists in an association analysis and its interpretation is complicated by the small number of samples and the variability of the mRNA expression.

In conclusion, we think that further studies are needed to identify the functional variant(s) modulating CHRNA5 mRNA levels.