We replicated earlier findings that chromosome 20 harbors a genetic element influencing ND (Han et al., 2010
). In our previous genome-wide linkage scan (Loukola et al., 2008
), LOD score of 2.36 was obtained for DSM-IV
ND at 20p13. In the current study, linkage with a LOD score of 3.8 was found with a marker residing in 20p11.21 (Study 1). Sex-specific analyses revealed that the signal was predominantly driven by females (MLOD score of 3.4), with the linkage locus shifting to 20q11. When the sample size was increased (Study 2), evidence for linkage remained invariable (MLOD score of 3.3 in 20q11.23 for females).
Multiple genes reside under the linkage peaks. Among others, chromosome 20p11.21 harbors genes coding for Type 2 cystatins (CST
s), extracellular secreted polypeptides that are broadly distributed and found in most body fluids (Dickinson, Thiesse, & Hicks, 2002
). Interestingly, there is a resemblance between cystatins and the α subunits of nAChRs, as both contain four cysteine residues forming two disulfide bonds, which in nAChRs play a critical role in agonist binding (Steinlein, Weiland, Stoodt, & Propping, 1996
). Chromosome 20p11.21 also contains GGTLC1
(gamma-glutamyltransferase light chain), a transpeptidase crucial in the metabolism of glutathione.
The female-specific linkage locus on 20q11.23 harbors, among others, SRC
(proto-oncogene tyrosine protein kinase SRC), which plays a role in the regulation of embryonic development and growth, BLCAP
(bladder cancer–associated protein), which encodes an apoptosis stimulating tumor suppressor protein, and NNAT
(neuronatin isoform beta), suggested to regulate ion channels during brain development and thus being one of the many forming and maintaining factors of the nervous system (Dou and Joseph 1996
). There is no evidence for any of these genes mentioned to have a role in the development of ND. The genes located at the linkage peak area are presented in the Supplementary Figure 1
A gene earlier associated with ND measures, CHRNA4
, resides on 20q13.2–13.33, nearly 26-Mb downstream. However, as our linkage signal peak is rather wide, the multipoint 1-LOD drop region defining the 90% confidence region (Dupuis & Siegmund, 1999
) covers a region of ~15 cM, and the multipoint signal peaks at 20q13.12, the possibility that the signal is caused by a genetic element around CHRNA4
, cannot be ruled out.
The chromosome 20q signal is repetitively identified in linkage and candidate gene association studies of smoking behavior, but no single genome-wide association study so far has found an association at the region (The Tobacco and Genetics Consortium, 2010
). Even the three large meta-analyses of smoking quantity, smoking persistence, and smoking initiation did not observe any association to this region (Liu et al., 2010
, The Tobacco and Genetics Consortium, 2010
, Thorgeirsson et al., 2010
). Thus, the effect of the CHRNA4
locus on ND may be heterogeneous or sex specific. Allelic heterogeneity would suggest that a locus has multiple variants affecting a phenotype, and these alleles may be rare in the population at large. Linkage analysis examining the cosegregation of marker alleles may detect such effects, but association analysis can identify only common variants influencing the phenotype.
Sex differences are apparent in the association of this chromosomal area to ND. The fact that genome-wide association meta-analyses of smoking-related traits have so far not performed sex-specific analyses may be another explanation for the lack of significant association findings for this chromosome 20 region. In a study including families of European American and African American ancestry, Li et al. (2005)
found that the association between CHRNA4
and ND was strongest in African American females. However, the association has also been observed in a sample of Chinese male smokers (Feng et al., 2004
). Both studies used FTND (Heatherton et al., 1999
) as a continuous measure of ND.
The FTND affection status measure showed no linkage with chromosome 20 region markers in our study, and only a minuscule linkage signal with FTND as continuous variable was observed. It has been proposed that the two different measures of ND, DSM-IV
ND and FTND, measure the phenomenon from partly different points of view. This is supported by the fact that in clinical trials, DSM-IV
ND and FTND rarely yield consistent results (diFranza et al., 2010
; Moolchan et al., 2002
; Piper, McCarthy, and Baker, 2006
). It is likely that the FTND provides a stronger measure of physical and pharmacological dependence, whereas the DSM-IV
ND measures more thoroughly the behavioral and cognitive factors, for example loss of control in terms of smoking behavior, underlying ND. On the basis of these arguments, and considering that females are more prone to the pressure of social factors than males, our results are consistent with the assumption of the differences between the two ND measures.
We observed changes in LOD scores when we increased the sample size by combining two subsamples of the NAG study. However, this is not unusual. LOD scores are known to be sensitive to changes (Hodge and Greenberg, 1992
). Despite the changes in LOD scores, the signal exists.
The meta-analysis of 15 linkage scans of 10,253 family members identified multiple chromosomal regions, including chromosome 20, associated (at nominal significance levels) with smoking behavior based on FTND and MaxCig24 measures (Han et al., 2010
). As for genome-wide association studies, also linkage studies require large sample sizes to study complex traits. For the same reason, the information content for markers analyzed in Study 1 were slightly lower than in Study 2 ( and ). ND as a complex trait seems to require larger samples in order to increase the information contents of the markers.
A limitation of our study is the low number of participating parents leading to incomplete family structures and decreased power in the linkage analysis. This is due to the fact that the twins were relatively old (mean age of 57 years) at the time of the data collection and thus the family members included are mostly siblings. In addition, as the three-symptom diagnostic threshold of DSM-IV
does not provide a perfect accuracy in the diagnosis of ND, some subjects’ affection status may have been misclassified. No biochemical verification of the smoking status was performed as the analysis was based on affected only; it is unlikely that any nonsmokers would have claimed to be smokers in the extensive interview. Essentially, we did not replicate our MaxCigs24 finding (Saccone et al., 2007
), which suggests that the original finding was either a false positive or, among other possible explanations, the variance or patterns of transmission for MaxCigs24 was different in consequential ways once subjects were added to the linkage families. The sample may be underpowered to resolve the nonreplication, and more work is needed to be done to resolve the results.
In conclusion, our results provide further evidence that chromosome 20 harbors genetic elements influencing ND. The comparison of our results with the literature supports the hypothesis that the locus has multiple mutant alleles influencing smoking behavior.