In Vitro and Ex Vivo Analysis of CHRNA3 and CHRNA5 Haplotype Expression

Genome-wide association studies implicate variations in CHRNA5 and CHRNA3 as being associated with nicotine addiction (NA). Multiple common haplotypes (“risk”, “mixed” and “protective”) exist in Europeans; however, high linkage disequilibrium between variations in CHRNA5 and CHRNA3 makes assigning causative allele(s) for NA difficult through genotyping experiments alone. We investigated whether CHRNA5 or CHRNA3 promoter haplotypes, associated previously with NA, might influence allelic expression levels. For in vitro analyses, promoter haplotypes were sub-cloned into a luciferase reporter vector. When assessed in BE(2)-C cells, luciferase expression was equivalent among CHRNA3 haplotypes, but the combination of deletion at rs3841324 and variation at rs503464 decreased CHRNA5 promoter-derived luciferase activity, possibly due to loss of an SP-1 and other site(s). Variation within the CHRNA5 5’UTR at rs55853698 and rs55781567 also altered luciferase expression in BE(2)-C cells. Allelic expression imbalance (AEI) from the “risk” or “protective” haplotypes was assessed in post-mortem brain tissue from individuals heterozygous at coding polymorphisms in CHRNA3 (rs1051730) or CHRNA5 (rs16969968). In most cases, equivalent allelic expression was observed; however, one individual showed CHRNA5 AEI that favored the “protective” allele and that was concordant with heterozygosity at polymorphisms ∼13.5 kb upstream of the CHRNA5 transcription start site. Putative enhancer activity from these distal promoter elements was assessed using heterologous promoter constructs. We observed no differences in promoter activity from the two distal promoter haplotypes examined, but found that the distal promoter region strongly repressed transcription. We conclude that CHRNA5 promoter variants may affect relative risk for NA in some heterozygous individuals.


Introduction
Nicotine is the main addictive agent in cigarette smoke [1]. Nicotine addiction (NA) can be thought of in terms of an ''imbalance" between the rewarding and aversive effects of nicotine self-administration (through cigarette smoking). Nicotinic acetylcholine receptors (nAChR) are a family of hetero-and homo-pentameric protein complexes that form ligand-gated ion channels found at the neuromuscular junctions and in the central nervous system. When activated by acetylcholine, or other agonists such as nicotine, the nAChRs allow influx of Ca 2+ and other ions into neurons. A variety of nAChRs are expressed in brain, including (a4b2) 2 * (where * indicates differential incorporation of subunits such as b2, a4, or a5) subtypes thought to be the major high affinity brain receptor for nicotine; however, other subtypes including those containing the a5, a3 and b4 subunits are also expressed in brain [2]. Importantly, those that contain the a3 and a5 subunits are expressed in brain regions involved in reward, emotion, learning and memory (nucleus accumbens, amygdala and prefrontal cortex) as well as in regions involved in aversion (medial habenula). Studies in mice implicate (a4b2) 2 * nAChRs in behaviors related to nicotine addiction [3,4]. Moreover, a recent study of a5 subunit knockout mice revealed the importance of a5 containing nAChRs in nicotine self-administration through rescue of a5 expression in the medial habenula [5]. Further, the utility of varenicline (a partial agonist at (a4b2) 2 * nAChRs) in smoking cessation in humans [6] also demonstrates the importance of (a4b2) 2 * nAChRs in human NA.
Numerous genetic studies in human populations implicate a role for the CHRNA5 and CHRNA3 genes in NA [7][8][9][10][11][12][13][14][15][16][17][18]. All studies identify the same alleles as increasing risk for NA, but the effect size is small (odds ratio ,1.3, or ,1-2 cigarettes per day in quantitative analyses). CHRNA5 and CHRNA3 are located in tandem on human chromosome 15 and encode the a5 and a3 nAChR subunits, respectively. These genes are in high linkage disequilibrium (LD) with one another, forming a large haplotype block [14], which complicates identification of the causative polymorphism(s) for NA.
One coding single nucleotide polymorphism (cSNP) studied extensively is a mis-sense variant in CHRNA5 (rs16969968; G.A) that creates a D398N amino acid change, with the minor ''A'' allele being the NA risk allele. Comparison of the mouse (a4b2) 2 # (where # indicates human a5-D398 or a5-N398) nAChRs expressed in HEK293T cells revealed similar EC 50 values for both nAChR isoforms, but diminished maximal response for the (a4b2) 2 a5-N398 isoform when epibatidine was used as an agonist [15]. Human (a4b2) 2 a5-N398 nAChR isoforms have diminished calcium permeability and more extensive desensitization at 3 seconds after acetylcholine application compared to (a4b2) 2 a5-D398 isoforms [19]. Importantly, whereas the rs16969968 variant has a high minor allele frequency (MAF) in the European population (MAF = 0.42), it is not as prevalent in the African American (MAF = 0.04), Asian (MAFs = 0.01-0.03) and sub-Saharan African (MAF = 0.00) populations [13,15,20,21]. The fact that each of these populations can exhibit NA suggests that other variants in LD with rs16969968, or loci outside the CHRNA5-A3-B4 cluster, influence the NA phenotype in individuals of European and non-European ancestries.
We report functional analysis of variants within haplotypes of CHRNA5 and CHRNA3 in people of European ancestry. CHRNA5 and CHRNA3 promoter-59UTR haplotypes were examined in vitro, in the human neuroblastoma cell line (BE(2)-C) which expresses both CHRNA5 and CHRNA3, to study the influence of promoter and 59UTR polymorphisms on transcription and translation, respectively. The reward pathway is likely involved in whether an individual who initiates smoking is compelled to continue smoking. Thus, the potential for CHRNA5 or CHRNA3 allelic expression imbalance (AEI) in brain regions important in the reward pathway (prefrontal cortex (PFC); amygdala (Amyg) or nucleus accumbens (Nacc)) were evaluated in post-mortem brain tissue from individuals with one ''risk'' and one ''protective'' haplotype ( Figure 1A) [14]. We report AEI of CHRNA5 observed in Amyg and Nacc of one individual who was also heterozygous at distal (,13.5 kb upstream of the transcription start site (t.s.s.)) promoter polymorphisms (rs880395 and rs7164030) as recently reported [22]. Taken together, our results serve as an independent  [23] was used to download the version 2, release 21 CEU dataset between Chr15:76638000-76705000 from the HapMap project. The method of confidence intervals was used to generate haplotype blocks [24]. (A) The relative position of HapMap SNPs, numbered 1-58, along each gene is shown schematically at the top. Five major (.5%) and two rare (,5%) haplotypes of the gene cluster are shown with risk alleles in each haplotype associated with NA boxed in black. Those SNPs that were genotyped for haplotype analysis shown in Figure 1B are also boxed in black. The first or ''risk'' haplotype contains multiple SNPs associated with NA including a mis-sense coding SNP (cSNP) (rs16969968-[A]) in CHRNA5 and a synonymous cSNP (rs1051730-[T]) in CHRNA3 (Note: The alleles for CHRNA3 are those of the opposite strand such that, for example, rs1051730 is an ''A'' rather than a ''T''). The ''mixed'' haplotypes (second and fifth) contain many SNPs associated with NA, but not rs16969968-[A] or rs1051730-[T] (Note: The only difference between the second and fifth ''mixed'' haplotypes shown is at rs1948 in CHRNB4. Thus, the ''mixed'' haplotypes have a combined frequency of 24.2%+5.8% = 30% as defined in [14]. The third and fourth haplotypes are considered ''protective,'' containing no SNPs associated with NA. The sixth and seventh ''rare'' haplotypes also have a mixed allele structure with respect to SNPs associated with NA. Note that because it is not in the HapMap dataset, rs3841324 is not shown in the haplotype block structure shown in Figure 1A. (B) Haplotype analysis of CHRNA5 and CHRNA3 including rs3841324. Eighty European-Americans were genotyped at 6 polymorphisms (numbered 1-6) across the CHRNA5 and CHRNA3 (''A3'') genes. Haplotype and linkage disequilibrium (LD) analysis was done using Haploview 4.2 [23]. A solid spine of LD was used to define haplotype blocks. The three common and one rare haplotypes containing the normal ([N]) or deletion ([D]) allele at rs3841324 are shown with the r 2 values of the LD plot indicated in grayscale squares. Darker squares represent greater LD between SNPs. As in Figure 1A, risk alleles are outlined with black boxes. Note the ''mixed'' haplotype containing rs3841324-[D] (39.3%) in Figure 1B is equivalent to the second or fifth ''mixed'' haplotype (summation of 30%, see above) shown in Figure 1A [14]. Also note that the cSNP rs8040868 in CHRNA3 is not included with the haplotype blocks because it falls outside of the defined block 1 structure. doi:10.1371/journal.pone.0023373.g001 confirmation of those of Smith et al. [22] and support the idea that relative risk for NA must be assessed not only based on the presence of the mis-sense cSNP rs16969968-[A] in CHRNA5, but also on the regulation of relative allelic expression levels.

Results
The CHRNA5 deletion (rs3841324-[D]) associates with the ''mixed'' European-American haplotype Haplotype analysis indicated that the rs3841324-[D] allele is in strong LD (D' = 0.965, r 2 = 0.884) with the rs588765-[T] allele, but weaker LD (D' = 0.961, r 2 = 0.343) with the rs16969968-[G] allele ( Figure 1B). Our analysis places rs3841324 mainly on the ''second'' or ''mixed'' haplotype (frequency of 30% in European-Americans) defined in [14] (''mixed 1'' or ''mixed 2'' in Figure 1A). Notably, when rs3841324 is included in the haplotype analysis, we observe 3 common haplotypes ( Figure 1B) similar to the observation made by Saccone et al. [10]; wherein joint analysis of rs588765 and rs16969968 defined 3 common haplotypes in European-Americans with different effects on risk. This implies that the cSNP at rs16969968 is not the only factor conferring risk for NA in this gene cluster.

The SP-1 transcription factor binds within the rs3841324 deletion
The rs3841324 deletion removes a few transcription factor binding sites, including SP-1 [2,25]. Thus, we used the deleted region as probe in an SP-1-specific EMSA analysis. EMSA analysis using the sequence encompassed by the CHRNA5 promoter deletion at rs3841324 as probe revealed specific binding of purified recombinant human SP-1 (rhSP-1) transcription factor ( Figure S1). A doublet of rhSP-1 binding is present without competitor or when an AP-2 consensus sequence is used as competitor, but is absent when either cold probe or an SP-1 consensus sequence is used as competitor. This indicates that highly pure rhSP-1 binds the SP-1 site within the 22-base pair deleted region of the CHRNA5 promoter. HeLa nuclear extract, used as a positive control for the EMSA, caused a larger mobility shift of the probe suggesting that other factors in addition to SP-1 (perhaps Erg-2 or AP-2 which also have predicted consensus binding sites within the deleted region) might interact with the CHRNA5 promoter ( Figure S1).
The rs55853698 and rs55781567 CHRNA5 59UTR polymorphisms affect luciferase reporter activity The CHRNA5 59UTR SNPs rs55853698 (T.G) and rs55781567 (C.G) do not directly impact transcription factor binding sites [7] (data not shown), although rs55853698 flanks an SP-1 binding site and rs55781567 flanks a C/EBP-b and ikaros-2 binding site. Because translation of transcribed CHRNA5 mRNA from different promoter haplotypes (i.e. ''risk'' versus ''protective''; see below) might be influenced by the two 59UTR SNPs, we tested all combinations of these SNPs for effects on luciferase activity in BE(2)-C cells while keeping the promoter sequence constant. We found no statistically significant difference in luciferase activity derived from the ''GG'' and ''TC'' 59UTR haplotypes (p = 0.46) in BE(2)C cells (Figure 3). In contrast, the ''GG'' 59UTR haplotype showed statistically significant differences in luciferase activity versus the ''GC'' (p, 0.0025) and ''TG'' (p,0.0001) 59UTR haplotype constructs. Likewise, luciferase activity from the ''TC'' 59UTR haplotype was significantly different from the ''TG'' haplotype (p,0.0001). Taken together, these data suggest influences of both the rs55853698 and rs55781567 SNPs in the CHRNA5 59UTR on luciferase expression in BE(2)-C cells ( Figure 3).

CHRNA3 promoter haplotypes do not affect relative transcription in vitro
No statistically significant differences in relative firefly luciferase activity were observed among the analogous (i.e. pGL4(NNA3) or pGL4(XNNA3)) constructs of the CHRNA3 promoter haplotypes tested ( Figure 4).

CHRNA5
, but not CHRNA3, shows AEI ex vivo The SNaPshot system was used to assess AEI in post-mortem brain tissue from individuals heterozygous at the CHRNA3 and CHRNA5 SNPs rs1051730 and rs16969968, respectively. This method is based on an expected value of 50% for gDNA amplified from heterozygotes. AEI in the mRNA-derived (cDNA) samples is indicated by deviation of either SNP allele from the expected 50% value. Because we wanted to assess AEI derived specifically from the ''risk'' and ''protective'' haplotypes, and the CHRNA5 rs3841324-[D] allele associated with the ''mixed'' haplotype ( Figure 1B) [14,27], all individuals tested for AEI of CHRNA5 were rs3841324-[N] homozygotes.
Analysis of individual sample means (cDNA versus gDNA) indicated statistically significant (p,0.05; ANOVA with Tukey-Kramer HSD post-hoc) AEI for both CHRNA3 and CHRNA5 ( Figure 5); however, when all determinations for a particular brain region were grouped, there was no statistical evidence of AEI for CHRNA3 (PFC: Levene's test for equal variance, p = 0.0382; Welch's test for equal means, p = 0.2189; Amyg: Levene's, p = 0.0012; Welch's, p = 0.8486; Nacc: Levene's, p,0.0001; Welch's, p = 0.1022). Likewise, there was no statistical evidence for AEI in grouped determinations of CHRNA5 in PFC (Levene's, p = 0.0839; Welch's, p = 0.0606). In contrast, for CHRNA5, grouped determinations for Amyg and Nacc revealed statistically significance differences in allelic expression (Amyg: Levene's, p = 0.0232, Welch's, p = 0.0025; Nacc: Levene's, p = 0.0006; Welch's, p = 0.050). In both cases, individual ''7'' (,80%G expressed) appears to be driving the statistical result ( Figure 5B) because statistical significance is no longer observed (i.e. p.0.05) when this individual is removed from the analysis. Whereas the smoking history of individual ''7'' could not be determined, we note that neither diagnosis (normal, schizophrenic or Alzheimer's) nor smoking history of an individual was concordant with whether an individual showed AEI. Whereas the 80:20 AEI between CHRNA5 alleles in both the Nacc and Amyg of individual ''7'' is probably biologically significant ( Figure 5B), the biological relevance of the ''suggestive'' AEIs for CHRNA3 and CHRNA5 in individuals other than ''7'' remains unclear as they were observed in only one brain region, were in opposite directions in different brain regions from the same individual or were small enough to be within the precision limits of the assays ( Figure 5). Indeed, CHRNA3 promoter haplotypes for those individuals studied were not in concordance with any suggestive AEI of CHRNA3 ( Figure 5A and Table S2). Importantly, both methods used for determining AEI in each sample (standard curve versus formula of [28]) gave similar results, with individual ''7'' showing CHRNA5 AEI ( Figures S2A and S2B).
CHRNA5 AEI is concordant with heterozygosity at 59 distal promoter SNPs, but not with an intronic SNP  Figure 1B). However, because individual ''7'' was homozygous for the ''C'' allele at rs588765 (data not shown), such recombination could not explain this observation. Further, sequencing of amplified gDNA from this individual's CHRNA5 core promoter (which was heterozygous for the ''W'' SNP (rs503464) at 2166, the ''K'' SNP (rs55853698) at 2123 and the ''S'' SNP (rs55781567) at 276; data not shown) showed no evidence of a rare variant that might alter a binding site for transcription factors. A recent report found a positive correlation between CHRNA5 AEI in PFC and the minor alleles of three SNPs (rs880395, rs905740 and rs7164030) in the extremely 59 distal CHRNA5 promoter region (,13.5 kb 59 of the t.s.s.) [22]. We amplified this region of CHRNA5 from gDNA of all individuals tested for AEI and genotyped two of the three SNPs (rs880395 and rs7164030) by RFLP analysis. Whereas individual ''7'', who showed convincing AEI of CHRNA5 ( Figure 5B), was heterozygous at both rs880395 and rs7164030 ( Figure S3), all other individuals in our study, who showed no statistically significant CHRNA5 AEI ( Figure 5B and Figure S2B), were homozygous for the major alleles at these SNPs ( Figure S3). These results serve as an independent confirmation of those reported previously [22].
The 59 distal CHRNA5 promoter region acts as a strong transcriptional repressor To investigate the function (enhancer or repressor) of the distal promoter region containing the three SNPs (rs880395, rs905740 and rs7164030) associated with AEI [22], we amplified the ''risk'' or ''protective'' (i.e. GCA) and the ''mixed'' (i.e. ATG) haplotypes of the region ( Figure 1A) and created heterologous promoter constructs. We found that this 59 distal promoter region acts as a strong repressor of transcription when joined upstream of the HSV-TK promoter ( Figure 6). Importantly, and somewhat unexpectedly, we observed no statistically significant difference in relative promoter activity from the two haplotype constructs (p = 0.183) in BE(2)C cells ( Figure 6).
We observed no statistically significant differences in relative CHRNA3 promoter activity among haplotypes studied in human neuroblastoma BE(2)-C cells ( Figure 4) and no statistically significant AEI of CHRNA3 mRNA in the brain regions examined ( Figure 5A and Figure S2A). Taken together, these data suggest that overall relative risk for the development of NA among heterozygotes for the ''risk'' and ''protective'' haplotypes does not depend on CHRNA3 AEI due to the promoter and/or cSNPs (within the mRNA) that comprise each examined haplotype. Further, no concordance between CHRNA3 promoter haplotypes and any ''suggestive'' AEI was observed (Table S2). One caveat to these studies is the observation that CHRNA3 expression is regulated by promoter methylation [30]. Thus, we cannot exclude differential promoter methylation within different brain regions for those samples that showed ''suggestive'' AEI of CHRNA3.
In contrast to CHRNA3, CHRNA5 promoter haplotypes showed statistically significant differences in relative promoter activity when compared in BE(2)-C cells; however, these statistical differences were not between the ''risk'' and ''protective'' promoter haplotypes, but rather between the ''risk'' or ''protective'' haplotype and the ''mixed'' promoter haplotype containing the deletion at rs3841324 (Figure 2). The rs3841324-[D] allele has no effect on transcription in TE671 cells [25], but decreases transcription by ,40% in HEK293T cells [25]. Similarly, in three of four lung cancer cell lines examined, rs3841324- [D] caused an ,80% decrease in CHRNA5 promoter-derived luciferase activity [26]. In contrast to these in vitro data, ex vivo data from post-mortem brain (PFC) and resected lung tissue indicated that CHRNA5 rs3841324-[D] homozygotes generate ,2.5-3-fold higher CHRNA5 steady-state mRNA levels than CHRNA5 rs3841324-[N] homozygotes [26,27]. Thus, whereas our data in BE(2)-C cells agree with most of these in vitro studies, indicating decreased relative CHRNA5 promoter activity from the promoter with a deletion at rs3841324, the ex vivo studies suggest the opposite effect [26,27].
The rs3841324 deletion encompasses a few transcription factor binding sites, including SP-1 [2,25], a transcription factor which activates transcription. The large shift of the probe with HeLa nuclear extract might indicate binding of factors other than SP-1 because the mobility shift is greater than with rhSP-1 alone ( Figure  S1). Nonetheless, although we cannot exclude the impact of other factors, such as Erg-2 and AP-2, which might regulate brain expression of CHRNA5 by binding within the promoter deletion at rs3841324, our EMSA data indicate that pure rhSP-1 binds within this deleted region ( Figure S1). Given that SP-1 stimulates transcription, the loss of a binding site for SP-1 caused by the rs3841324 deletion could explain, in part, the decreased in vitro transcriptional activity from the CHRNA5 rs3841324-[D] construct in most cell types, but not the enhanced activity observed from this allele ex vivo [26,27].
In the study of post-mortem brain [27], the higher CHRNA5 steady-state mRNA levels from the rs3841324-[D] allele were associated with the rs588765-[T] allele in intron 1 of CHRNA5. Furthermore, CHRNA5 distal promoter (,13.5 kb upstream of the t.s.s.) polymorphisms correlated with the levels of CHRNA5 mRNA alleles expressed ex vivo [22], a finding that we confirmed in our limited CHRNA5 AEI experiments ( Figure 5B and Figure S2B). Thus, an alternative and simple explanation for the contradictory (in vitro versus ex vivo) findings may be the exclusion of regions of the CHRNA5 gene (distal promoter or intron 1 where rs588765 is located) from the constructs used in the in vitro transfection studies that are necessary for determining the relative allelic transcriptional activity in vivo.
Because the rs3841324-[D] allele is in ''substantial LD'' with the minor alleles (''ATG'') of the distal promoter SNPs [22] and is also in strong LD with the ''mixed'' European-American haplotype ( Figure 1A), it was hypothesized that transcription factors binding putative ''enhancer'' elements within the distal portion of the CHRNA5 promoter might increase expression of the rs16969968-[G] allele from the ''mixed'' haplotype in vivo [22]. We restricted our analysis of CHRNA5 AEI to individuals with two rs3841324-[N] alleles at the CHRNA5 promoter. As such, we assessed the effect of all SNPs within the ''risk'' or ''protective'' haplotypes on the relative allelic mRNA levels of CHRNA5 by interrogating heterozygotes for these two haplotypes. As with CHRNA3, most individuals showed no AEI in steady-state CHRNA5 mRNA in the post-mortem brain regions examined ( Figure 5B). However, one individual (''7'') showed an 80:20 %G:%A ratio of CHRNA5 steady-state mRNA in favor of the ''protective'' rs16969968-[G] allele ( Figure 5B). Neither recombination between the rs3841324-[N] allele and the intronic rs588765-[T] allele, which was found to be associated with higher steady-state mRNA levels in PFC [27], nor a rare variant in this individual's CHRNA5 core promoter explained this finding. However, we found heterozygosity at rs880395 and rs7164030 ( Figure S3) in this individual's distal CHRNA5 promoter that is concordant with the observed AEI ( Figure 5B) and with correlative findings in other studies [22]. Notably, in contrast to the ''mixed'' haplotype where the minor alleles (''ATG'') are associated, haplotype analysis indicates that the major alleles (''GCA'') at these distal promoter SNPs most often associate with the ''risk'' and ''protective'' haplotypes ( Figure 1A).
When we tested directly the two haplotypes of this 59 distal CHRNA5 promoter region, however, we found that it acts as a strong repressor, rather than an enhancer, of transcription from the HSV-TK promoter ( Figure 6). Quite unexpectedly, there were no statistically significant differences between the promoter activities of the two haplotype constructs ( Figure 6). We cannot exclude the possibilities that elements within this 59 distal region of CHRNA5 might act differently when placed at a greater distance from the transcription start site (normally ,13.5 kb upstream of the t.s.s.) or when linked to the native CHRNA5 promoter which is not a TATA-containing promoter [2] as is the HSV-TK promoter. Nonetheless, taken together with the finding that the ''risk'' and ''protective'' promoter haplotypes showed no statistically significant difference in promoter activity (Figure 2), the observation of 80:20 AEI of CHRNA5 in individual ''7'' ( Figure 5B) might also suggest an expression effect of SNPs within the coding, other than rs16969968, or untranslated regions of the CHRNA5 transcript.
A recent meta-analysis of NA genome-wide association studies revealed a strong association signal from rs55853698 in the CHRNA5 59UTR [7]. We analyzed combinations of the CHRNA5 59UTR SNPs rs55853698 and rs55781567, on the rs3841324-[N]:rs503464-[A] CHRNA5 promoter background, in BE(2)-C cells. Whereas neither rs55853698 nor rs55781567 is predicted to alter a transcription factor binding site [7], both flank predicted transcription factor binding sites (see Results above). There was no statistically significant difference in activity observed from the ''risk'' ''GG'' versus ''protective'' ''TC'' haplotype constructs ( Figure 3). Interestingly, the ''risk'' ''GG'' construct showed a statistically significant difference in activity compared to both the ''TG'' and ''GC'' constructs ( Figure 3) suggesting an influence of rs55853698 and rs55781567, respectively. Thus, these data suggest a possible role for 59UTR SNPs in translational regulation of firefly luciferase in BE(2)-C cells, and perhaps a translational regulatory component to CHRNA5 expression. Given the occurrence of the ''GG'' or ''TC'' diplotypes on the ''risk'' or ''protective'' and ''mixed'' European haplotypes, respectively, the biological significance of this effect remains unclear. We conclude that, in addition to rs16969968, overall risk for NA in populations of European ancestry must be assessed based on the contributions of other SNPs within the extended haplotype of CHRNA5 which may influence expression levels of CHRNA5 mRNA, and perhaps also a5 protein.

Subjects
Post-mortem brain samples were obtained from either the University of Pennsylvania Brain Bank (n = 6; PFC) or the Golden Brain Bank at the Department of Veterans Affairs (VA) Medical Center -Coatesville, PA (n = 6; Amyg and Nacc). For both brain banks, subjects were excluded for prolonged agonal state (unresponsive.24 hrs), ventilator support (.12 hrs), and history of major, known CNS disease apart from the disease recorded (normal, schizophrenia or Alzheimer's). Patients for whom the post-mortem interval was.24 hrs were excluded. Patients were excluded on the basis of i) alternate or ambiguous DSM-IV diagnosis, ii) substance abuse (of illicit substances, but not ethanol or nicotine), iii) neurological disorder predating the onset of psychiatric symptoms (e.g., epilepsy, traumatic brain injury, etc.), iv) subsequent neurological disorders that would significantly compromise brain function (e.g., anoxia, stroke) and confound interpretation of postmortem findings. All subjects were of European ancestry as determined at time of autopsy. All samples were obtained with approval from the respective institutional review boards (by an IRB at the University of Pennsylvania or by the Veterans Integrated Service Network 4 Multi-Site IRB at Coatesville VAMC) and following all brain bank user agreements.

Haplotype analysis of CHRNA5 and CHRNA3 SNPs in the European-American population
Genomic DNA (gDNA) from individuals of European ancestry from the University of Pennsylvania Brain Bank (n = 80) were genotyped for six polymorphisms across the CHRNA5 and CHRNA3 genes. For CHRNA5, the normal ([N]) versus 22 base pair deletion ([D]) allele at rs3841324 in the CHRNA5 core promoter was genotyped by PCR using primers (see Table S1 for all PCR primers and parameters used) that flanked the deletion site. Alleles were separated on either a 5% non-denaturing polyacrylamide or a 2.5% agarose gel system. For SNPs rs588765, rs637137, and rs16969968 in CHRNA5 and SNPs rs1051730 and rs8040868 in CHRNA3, genotypes were determined using TaqManH SNP genotyping assays C_18826_10, C_5866_10, C_26000428_20, C_9510307_20 and C_261698_10; respectively (Life Technologies, Carlsbad, CA) in a standard 5 ml PCR using an Applied Biosystems, Inc. (ABI) 384-well block 9700 under standard cycling parameters. Plates were post-read on an ABI 7900HT using the SDS 2.2.2 software. All calls were entered into an Excel spreadsheet prior to analysis of LD and haplotype structures using Haploview 4.2 software [23].
Functional Analysis of CHRNA5 and CHRNA3 promoter-59UTR haplotypes in BE(2)-C cells Human BE(2)-C (CRL-2268) neuroblastoma cells (a sub line of human SK-N-BE(2) cells which express both CHRNA5 and CHRNA3 (GAD, unpublished data)) were obtained from the American Type Culture Collection (ATCC; Manassas, VA) and propagated in a 37uC humidified incubation chamber with 5% CO 2 in a 1:1 mixture of Eagle's Minimal Essential Medium (ATCC):Ham's F12 (Life Technologies) containing 10% fetal bovine serum (Thermo Scientific [Hyclone]; Rockford, IL). The CHRNA5 core promoter and 59UTR (-1440 to +1 at the AUG start codon) was amplified from gDNA of an individual heterozygous at rs3841324, but homozygous G/G at rs16969968, in an initial PCR using forward and reverse primers that generated 1798 bp amplicons, and that yielded the ''mixed'' and ''protective'' CHRNA5 promoter haplotypes (see Figure 1A). For CHRNA3, the core promoter and 59UTR (-1442 to +1 at the AUG start codon) was amplified from gDNA of different individuals in an initial PCR using forward and reverse primers that generated 1535 bp amplicons. Resulting CHRNA5 and CHRNA3 promoter amplicons were subcloned by T/A-cloning into pCRII-TOPO using the Dual Promoter T/A-TOPO-cloning kit (Invitrogen, Carlsbad, CA). Plasmid DNA from various clones was prepared by mini-preparation using the GeneElute Miniprep system (Sigma, Saint Louis, MO). Clones without mutations (determined by Sanger sequencing of each clone) were diluted to 1 ng/ml before use in a second ''half-nested'' PCR employing the same forward primers (which both had Xho I sites incorporated at their 59 ends) as in the initial PCR, but a new reverse primer with an incorporated Nco I site to facilitate subsequent subcloning (see below). We note that the Nco I site changes a ''G'' nucleotide to a ''C'' nucleotide at the -1 position in the CHRNA5 promoter. Resulting amplicons (from the second PCR) were cloned into pCRII-TOPO and plasmid DNA from various clones was prepared by mini-preparation and checked for unwanted mutations as above. pCRII clones without mutations were then digested with Xho I and Nco I prior to gel purification and subsequent sub-cloning of the promoters into Xho I/Nco I-cut pGL4.10[luc2] (Promega, Madison, WI, USA) using T4 DNA ligase (New England Biolabs, Beverly, MA). We performed three QuikChangeH (Agilent) site-directed mutagenesis reactions (see Table S1 for all QuikChangeH primers and PCR parameters) to generate the ''risk'' CHRNA5 promoter from the ''protective'' CHRNA5 promoter construct (i.e. pGL4(XXNA5iATC); see above). Prior to use in transfections, each construct was prepared by maxi-preparation using Purelink columns (Life Technologies) and the promoter region fully sequenced to check for unwanted mutations. Experimental firefly luciferase promoter constructs (1.96 mg) and pGL4.74[hRLuc/TK] (Promega) vector (0.04 mg) were transiently co-transfected into human BE(2)-C neuroblastoma cells using Trans-it LT1 reagent (Mirus, Madison, WI). After 48 hrs, firefly and Renilla luciferase expression levels were assessed using a ''Dual-Ready'' TD-20/20 luminometer (Turner Biosystems, Sunnyvale, CA). Firefly luciferase activity (in light units [LU]) was normalized by dividing by Renilla luciferase activity (in LUs). For each experimental series, duplicate or triplicate transfections were done and the normalized (firefly/Renilla) values (in relative LUs) were averaged and counted as one determination. At least five (CHRNA5) or three (CHRNA3) separate experimental series were conducted and determinations were averaged, graphed and used in statistical analysis.
To test if the CHRNA5 59UTR SNPs rs55853698 and rs55781567 might influence translation (of firefly luciferase), the original ''protective'' promoter construct (i.e. pGL4(XNNA5iA-TC); see above) was mutagenized to all four possible 59UTR 2-SNP combinations using the QuikChangeH site-directed mutagenesis kit (Agilent). Mutagenized plasmids were prepared by maxi-preparation and promoter regions sequenced in entirety to check for undesired (PCR-based) mutations before being transfected into human BE(2)-C neuroblastoma cells for analysis as described above. For each experimental series, triplicate transfections were done and the normalized (firefly/Renilla) values (in relative LUs) were averaged and counted as one determination. Four separate experimental series were done and the 4 determinations were then averaged, graphed and used for statistical analysis.
To test if the distal promoter region of CHRNA5, containing SNPs rs880395, rs905740 and rs7164030 ( Figure 1A), could function as an enhancer, we amplified an 852 bp region of the distal CHRNA5 promoter (,13.5 kb 59 of the t.s.s.) from gDNA of individual #7 ( Figure 5) who was heterozygous at rs880395 and rs7164030 ( Figure S3) by PCR with forward and reverse primers (Table S1) using the SequalPrep Long Distance PCR Kit (Life Technologies). PCR amplicons were then sub-cloned into the pCRII-TOPO vector using the Dual Promoter T/A-TOPO cloning kit (Life Technologies). Plasmid DNA from positive clones was prepared by mini-preparation using the GeneElute system (Sigma) and was fully sequenced to determine the orientation and identity of the putative enhancer haplotype sequence. Two haplotypes of rs880395, rs905740 and rs7164030 were identified: A-T-G and G-C-A, respectively. pCRII plasmid DNA of clones with the distal promoter haplotypes in the ''negative'' orientation were digested with Kpn I and Hind III restriction enzymes and the Herpes simplex virus thymidine kinase promoter (HSV-TK) fragment (removed from pGL4.74[hRLuc/TK] (Promega) by digestion with Kpn I and Hind III) was sub-cloned into the above pCRII vector using T4 DNA ligase (New England Biolabs). The resulting constructs were then digested with Eco RV and Hind III, the CHRNA5 ''A5 enhancer''-HSV-TK fragment gel purified and then ligated into pGL4.74[hRLuc/TK] (Promega) that had been digested with Acc 65I, blunted by a fill-in reaction with Klenow fragment (NEB) and then digested with Hind III. This resulted in the creation of plasmids pGL4(A5enhATG/TK) and pGL4(A5enhGCA/TK) which were prepared by maxi-preparation, fully sequenced and then used in transfection experiments. Because Renilla luciferase was the reporter for these assays, the pGL4.13[luc2/SV40] (Promega) vector (which expresses firefly luciferase from the SV40 promoter) was co-transfected and used to normalize the Renilla luciferase expression from the experimental plasmids. Briefly, experimental Renilla luciferase promoter constructs (1.96 mg) and pGL4.13[luc2/SV40] (Promega) vector (0.04 mg) were transiently co-transfected into human BE(2)-C neuroblastoma cells using Trans-it LT1 reagent (Mirus). After 48 hrs, firefly and Renilla luciferase expression levels were assessed using a ''Dual-Ready'' TD-20/20 luminometer (Turner Biosystems). Renilla luciferase activity (in light units [LU]) was normalized by dividing by firefly luciferase activity (in LUs). For this experimental series, duplicate or triplicate transfections were done (total of 5 replicate determinations for each experimental construct) and the normalized (Renilla/firefly) values (in relative LUs) were averaged and counted as one determination. Five separate experimental series were done and the 10 determinations were then averaged, graphed and used for statistical analysis.

Electrophoretic mobility shift assays (EMSA)
Because an SP-1 site is located within the DNA sequence encompassing the CHRNA5 promoter deletion at rs3841324, we used the deleted region of the CHRNA5 promoter as probe in an SP-1-specific EMSA analysis. Briefly, 100 pmol each of TOP (59-GATTGGGCGGGGCCAGAGGGAAATAGGGG-39) and BOTTOM (59-CCCCTATTTCCCTCTGGCCCCGCCCAAT-C-39) oligomers were made double-stranded by heating to 75uC for 10 min in 1X oligomer annealing buffer (10 mM Tris-HCl; pH 7.5, 1 mM EDTA, 100 mM NaCl), then slow-cooled in 10uC steps of 65uC for 5 min, 55uC, 45uC, and 35uC, each for 30 sec, then held at 25uC until further use. Double-stranded DNA oligomers (100 pmol) were radio-labeled in the presence of c-[ 32 P]-ATP using T4 polynucleotide kinase under standard conditions. The reaction was stopped by addition of 1 ml of 0.5 M EDTA and 89 ml TE (10 mM Tris-HCl; pH 7.6, 1 mM EDTA) prior to purification through a G-25 Sephadex spin column (Roche, Nutley, NJ). EMSA was performed essentially as outlined in the Gel Shift Assay System (Promega) instructions. Briefly, 300 ng of recombinant human (rh)SP-1 (Promega) were incubated for 10 min at 25uC in the absence or presence of 50 or 100 molar excess unlabeled double-stranded probe, specific (SP-1) or non-specific (AP-2) binding site oligomers prior to addition of 125 fmol radio-labeled double-stranded probe. As a positive control, 10 mg of HeLa cell nuclear extract was used in the gel shift assay. After addition of radio-labeled probe, proteins (HeLa or rhSP-1) were allowed to bind for 20 min at 25uC. Reaction products were separated on a 4% non-denaturing polyacrylamide gel, the gel dried and exposed to a Phosphor screen. Bands were visualized using a Typhoon Phosphor imager (GE Healthcare, Piscataway, NJ).

Assessment of CHRNA5 and CHRNA3 AEI in post-mortem brain
The SNaPshot (Life Technologies) SNP genotyping system was used to determine the specific levels of ''risk'' or ''protective'' alleles expressed from CHRNA5 (at 16969968) or from CHRNA3 (at rs1051730). SNaPshot has been adapted successfully by a number of investigators for the purpose of determining AEI [22,28,[31][32][33][34]. Because the SNaPshot system relies on post-hoc analysis of PCR amplicons of cDNA, care was taken to design primers for the amplification step that were within regions without known SNPs. Furthermore, we determined the cycle thresholds for CHRNA5 (Hs00181248_m1, ABI) and CHRNA3 (Hs00609519_ m1, ABI) using quantitative RT-PCR gene expression assays (which do not distinguish relative allelic contributions to steadystate mRNA levels) on cDNA libraries created from post-mortem brain mRNA. From these qRT-PCR values, care was then taken to ensure that we were in the linear range of PCR amplifications for the AEI assays using SNaPshot. Total RNA was isolated from post-mortem brain tissue regions of individuals (PFC, n = 6; Amyg or Nacc, n = 6) heterozygous (i.e. with one ''risk'' and one ''protective'' allele) at rs16969968 and rs1051730 using the TrizolH reagent (Life Technologies) method. Prior to reverse transcription, 2-5 mg of total RNA was treated with DNAse I (Sigma-Aldrich, Saint Louis, MO) for 15-30 mins at 25uC. DNAse I digestion was stopped by adding EDTA, heating reactions to 70uC for 10 mins and quenching on ice before reverse transcription with random hexamers and SuperScript IIH reverse transcriptase using the First-strand cDNA synthesis kit (Life Technologies). First strand cDNA was treated with RNase H at 37uC for 20 min before use in exon-specific PCR which amplified portions of exon 5 from CHRNA5 or CHRNA3 cDNAs containing rs16969968 and rs1051730, respectively. For CHRNA5 or CHRNA3, PCR primers (Table S1) amplified a 198 bp or 160 bp amplicon, respectively, that was then used in SNaPshot analysis.
Standard curves for SNaPshot analysis were created by analyzing mixtures of exon 5 amplicons from gDNA from individuals homozygous at rs1051730 (CC or TT) or at rs16969968 (GG or AA) combined in multiple CC:TT or GG:AA ratios (100:0, 90:10, 80:20, 70:30, 60:40, 50:50, 40:60, 30:70, 20:80, 10:90 and 0:100). To determine AEI, the exon 5 region of CHRNA5 or CHRNA3 was amplified from cDNA and gDNA from the same (heterozygous) individual and then PCR products were analyzed using the SNaPshot Multiplex Kit (Life Technologies). Briefly, 0.01 to 0.04 pmol of diluted gDNA and cDNA PCR products were thermocycled for 25 cycles in duplicate singleplex reactions in the presence of the appropriate PAGE-purified primer: (59-CTAGAAACACATTGGAAGCTGCGCTC-39 for rs16969968; 59-ATCAAAGCCCCAGGCTA-39 for rs1051730) and fluorescently-labeled dideoxyNTPs. Reaction products were analyzed on an ABI 3730 bioanalyzer using a POP7 matrix and LIZ 120 standards. Peak areas were calculated with Peak Scanner 1.0 software (Life Technologies). The ratio of the peak areas of each SNP was taken from the mixed homozygous allele reactions and used to generate a standard curve in Excel. For experimental (cDNA) and control (gDNA) reactions from heterozygous individuals, the ratio of each peak area was calculated and then the percent ''G'' (rs16969968) or percent ''C'' (rs1051730) allele was determined using the standard curve for each respective SNP. To verify that our method of AEI determination was accurate, we also quantified each sample using the formula: where f (a) is the relative frequency of the 'a' allele expression, H a and H b are the peak heights of alleles 'a' and 'b' in cDNA and k corrects for the unequal amplification of SNP alleles in a heterozygous gDNA sample and is equal to 'a/b' where 'a' and 'b' are the peak heights of the two alleles in gDNA [28].
Restriction fragment length polymorphism (RFLP) analysis of the CHRNA5 upstream promoter region An 852 bp region of the distal CHRNA5 promoter (,13.5 kb 59 of the t.s.s.) was amplified by PCR. For RFLP analysis, each amplicon was digested with either Alu I (rs880395) or Dde I (rs7164030). Digestion products were resolved on a 5% nondenaturing polyacrylamide gel, post-stained with ethidium bromide and visualized and documented under UV trans-illumination.

Statistical Analysis
Relative promoter activities were compared by multi-way ANOVA followed by Tukey's ''Honestly Significant Difference'' (HSD) post-hoc analysis. AEI in specific brain regions was assessed statistically by Welch's ANOVA because of suspected heterogeneous variance, which was confirmed for most groups by Levene's test. JMP 8.0 software was used for all statistical analyses and a pvalue,0.05 was considered statistically significant in all tests. Figure S1 CHRNA5 promoter InDel (rs3841324) sequence binds SP1. Radio-labeled double-stranded oligomers encompassing the CHRNA5 promoter deletion (rs3841324, 2227 to 2248) were used as probe for binding to 300 ng of pure recombinant human SP-1 (rhSP-1) protein or 10 mg of HeLa nuclear extract. HeLa and rhSP-1 specific bands are indicated by double gray and double black arrows, respectively. The single black arrow indicates the radio-labeled free probe (''dsA5pIn-Del*''). Right triangles at the top of the schematic indicate lanes with increasing molar excess (50X or 100X) of competitors. Probe alone (lane 1); probe with HeLa nuclear extract only (lane 2); probe with rhSP-1 only (lane 3); probe with rhSP1 and 50 or 100 molar excess of cold probe as competitor (lanes 4 and 5, respectively); probe with rhSP-1 and 50 or 100 molar excess of cold SP-1 consensus oligomer as specific competitor (lanes 6 and 7, respectively); probe with rhSP1 and 50 or 100 molar excess of cold AP-2 consensus oligomer (which, like the SP-1 site, is GC-rich) as non-specific competitor (lanes 8 and 9, respectively). (TIF) Figure S2 Allelic expression imbalance (AEI) of CHRNA3 and CHRNA5 in post-mortem brain regions.

Supporting Information
The SNaPshot system was used to assess the relative allelic expression of either CHRNA3 (A) or CHRNA5 (B) in complementary DNA (cDNA) preparations derived from mRNA of postmortem brain regions, including the prefrontal cortex (PFC, black symbols), amygdala (Amyg, blue symbols) or nucleus accumbens (Nacc, red symbols). Allelic expression was quantified for each sample using the formula: f (a)~Ha=(HazkHb) where f (a) is the relative frequency of the 'a' allele expression, H a and H b are the peak heights of alleles 'a' and 'b' in cDNA and k corrects for the unequal amplification of SNP alleles in a heterozygous gDNA sample and is equal to 'a/b' where 'a' and 'b' are the peak heights of the two alleles in gDNA [28]. The averages 6 standard deviations of at least two determinations for each point are shown. The dashed lines at 40% and 60% represent the boundaries of what are considered AEI-positive determinations (i.e. differences of greater than 20%) using this method of calculation [28].  Figure S3 Concordance between CHRNA5 AEI (80:20 %G:%A) and heterozygosity at SNPs rs880395 and rs7164030. An 852 bp region of the CHRNA5 distal promoter region (,13.5 kb upstream of the t.s.s.) containing distal promoter SNPs rs880395and rs7164030 was amplified by PCR and amplicons digested with either Alu I (cuts the minor allele at rs880395) or Dde I (cuts the minor allele at rs7164030). The Alu I and Dde I digests are shown. Asterisks (*) indicate the presence of the minor allele product in both digests. Only individual ''7'' (with arrow) showed the presence of the minor alleles at both of these distal promoter SNPs and also showed CHRNA5 allelic expression imbalance (see Figure 5B). (TIF)