CHRNA7 Gene and Response to Cholinesterase Inhibitors in an Italian Cohort of Alzheimer’s Disease Patients

Variants in CHRNA7 gene (cholinergic receptor nicotinic alpha 7) have been found to influence susceptibility to Alzheimer’s disease (AD) [1], the risk of conversion from mild cognitive impairment to AD [2] and the clinical response to cholinesterase inhibitors (ChEI) [3 –5].

This gene, located on chromosome 15q14, encodes for α7-nAChR, the major subunit of the acetylcholine receptor in the central nervous system, and plays an important role in cholinergic neurotransmission. This receptor could be involved in the pathogenesis of AD, as it binds with high affinity to toxic amyloid-β₄₂ [6], finally resulting in its internalization, andaccumulation; furthermore results of a recent study allow to hypothesize that amyloid-β₄₂ deposition could in turn increase the expression of α7-nAChR thus inducing a positive feedback loop [7]. This gene is thus a plausible candidate gene for exploring genetic determinants of differential response to ChEI treatments designed to improve cholinergic neurotransmission [8].

A recent study reported an association of APOE4 allele and rs6494223, an intronic variant of CHRNA7 gene, with response to ChEI in a Brazilian cohort of 177 AD patients [4]. A role for CHRNA7 in the modulation of the outcome of ChEI therapy had also been highlighted in a previous work [3] carried out in a Chinese cohort of 204 AD patients, where association was found for variant rs8024987, in particular for galantamine-treated AD-affected women.

The current study sought to replicate findings for rs6494223 and rs8024987. We next tested the influence of CHRNA7 on response to ChEI at gene-level, by assessing the single and overall contribution of 84 common variants; finally, since the expression level of CHRNA7 is critical in the pathogenesis of AD and the protective effect of ChEI was found to be linked to the upregulation of α7-nAChR in cultured rat cortical neurons [9], we investigated whether there was an association to response to ChEI of eQTL variants (eSNPs), found to modulate expression of CHRNA7 in publicly available datasets.

Our population consisted of an Italian cohort of mild and severe AD patients treated with ChEI; the study was approved by the local ethics committee, and all patients signed the informed consent. This cohort had been genotyped on Illumina Human 660W-Quad BeadChips (Illumina, San Diego, CA, USA) and investigated in previous pharmacogenomic studies [10, 11]. The final quality-controlled dataset (see for details [11]) consisted of 169 patients (complete cohort): we next selected from the complete cohort a subset (mild cohort) retaining only mild-AD patients (Mini-Mental State Examination (MMSE) at baseline ≥20), to be consistent with the selection criterion adopted in the study performed by Braga et al. [4].

Response to treatment was defined as reported by Braga et al. [4] (response_1): patients were deemed as responders (R) if ΔMMSE ≥0 (difference in MMSE between six-months follow-up and baseline), and non-responders (NR) if ΔMMSE <0. According to this criterion, the analyzed dataset consisted of 77 R and 92 NR, whereas in the mild cohort (n = 100) we obtained 42 R and 58 NR. We next used a more stringent outcome definition, similarly to Weng et al. [3] (response_2), who defined as R those patients having ΔMMSE ≥2 at 6 months follow-up (n = 33) and NR the others (n = 136). We used a longer (1-year) follow-up than Weng’s study [3], since a 6-month follow-up would have yielded a more unbalanced dataset (12 R, 157 NR).

Single-SNP analysis was carried out in PLINK 1.7 [12], fitting logistic regression models under additive allele coding. For gene-level analysis, we used the set-based test as implemented in PLINK. The gene region was defined as encompassing the gene ± 20 kb (hg18 assembly) to include possible regulatory variants. Genotyped samples were imputed on HapMap II rel. 22 CEU reference panel, yielding a final set of 84 SNPs. Details about imputation and gene-based analysis, as well as statistical power analysis, can be found in the Supplementary Material.

eSNPs were defined as those SNPs significantly modulating CHRNA7 expression level in brain tissue from SNPExpress database (http://igm.cumc.columbia.edu/SNPExpress) [13], at a significance level of p = 1*10^–4; 72 of these variants were available on our array, and hence tested for association.

Clinical and demographic variables were screened for association with response to ChEI, namely age and MMSE score at treatment start, sex, and APOE status (Table 1). Male sex turned out to be associated with response_1 to treatment in the mild cohort (p = 0.03, OR = 2.62, 95% CI 1.15–6.12), and a lower baseline MMSE was observed in R both in mild (p = 0.018) and complete cohort (p = 0.03). As regarding response_2 we found that age at treatment start was higher in R than in NR (p = 0.003). We consequently adjusted all the mentioned analyses for relevant baseline predictors (age, sex, and baseline MMSE).

Consistently with findings reported in Braga’s study [4], we found that APOE ɛ 4 carriers had a worse clinical response than those without the ɛ 4 allele (OR = 0.47, 95% CI 0.19–1.07, p = 0.1), although results were not significant; as mentioned, we confirmed in the mild cohort that males were more likely to benefit from treatment (OR = 2.62, 95% CI 1.15–6.12, p = 0.03), as revealed by previous studies [14].

Rs8024987 was an imputed variant, whereas rs6494223 was genotyped on the array, with 100% call rate; both polymorphisms were in Hardy-Weinberg equilibrium (p = 0.29 and p = 0.35 for rs6494223 and rs8024987, respectively), as computed from exact test [15] implemented in PLINK. Genotypes counts and frequencies of the two SNPs in the two examined cohorts are reported in Table 2A.

We did not find association to response to ChEI (Table 2B) for rs6494223 (OR = 1.08, 95% CI 0.52–2.2, p = 0.83), and combination of our results with Braga’s [4] findings using a fixed-effect meta-analysis was not significant too (OR = 1.4, p = 0.1, I2 = 24%). In the larger cohort of 169 patients, we did not find a significant effect for rs6494223, although we observed a stronger effect size and consistency of direction (OR = 1.4, 95% CI 0.85–2.29, p = 0.17); for this cohort, the meta-analysis across the two studies was significant (OR = 1.57, p = 0.02, I2 = 0%).

We then investigated the impact of variant rs8024987 on response to ChEI as defined in response_2. Again, no significant effect was found, both on additive (OR = 0.79, 95% CI 0.38–1.65, p = 0.53) and dominant model (OR = 0.76, 95% CI 0.33–1.77, p = 0.53) (Table 2B). Upon stratification of the sample by sex, we did not find any association (female: OR = 1.13, 95% CI 0.52–2.47, p = 0.74; male: OR = 0.15, 95% CI 0.01–1.34, p = 0.08), but an opposite direction of effect was observed. We then explicitly test for different effect direction by examining an interaction model between sex and genotype, obtaining a similar trend as found in Weng’s study [3] (OR_INT = 0.14, 95% CI 0.01–1.41, p = 0.09), suggesting an effect modification induced by sex on the link between the variant and drug response.

To further examine the impact of rs8024987 we compared, analogously to Weng et al. [3] women using galantamine with GG or GC genotypes to women using non-galantamine medications (i.e., patients treated with rivastigmine or donepezil) and carrying wild type CC genotype. We found no association, although a tendency to better respond to drug in accordance with Weng et al. [3] was found in galantamine-treated GG or GC women as compared to non-carriers/non-galantamine-treated (OR = 1.5, p = 0.57, Fisher exact test).

Gene-level analysis did not support a global implication of CHRNA7 with response outcome, after summarization of 84 SNP test statistics and 10,000 permutations (p = 1, see Supplementary Material).

Single-variant analysis of eSNPs did not reveal evidence of association beyond Bonferroni significance level (α= 7*10^- 4): the top associated SNPs were rs13409654_A (OR = 4.07, 95% CI 1.29–12.8, p = 0.016) and rs10898159 _G (OR = 0.5, 95% CI = 0.35–0.92, p = 0.023) for response_2 and response_1 respectively. In particular, for rs13409654 each copy of the A allele, which we found increasing the probability of response to ChEI according to an additive model, is reported in SNPExpress database to be associated to an up-regulation of CHRNA7 (beta = 10.81, p = 8.2 *10^- 5).

In conclusion, we do not support the involvement of the two polymorphisms rs8024987 and rs6494223 in the response to ChEI in a study with similar sample size than previously published [3, 4]. However, for rs6494223 we observed a consistency in the effect direction, with individuals carrying T allele having higher odds of being responder if evaluated in the larger cohort of 169 patients comprising mild, moderate, and severe cases. Moreover, for this SNP the combined effect across the two studies yielded a significant association, and we confirmed a trend for the role of APOE4 in predicting a less effective response to treatment.

As regarding the other variant, rs8024987, we confirmed a trend for an effect modification of sex on response to treatment at this locus, with opposite genetic effect for women and men. The role of sex in response to ChEI has already been investigated in literature, and dissimilar response to ChEI between males and females has been reported [14, 16]. The underlying biological mechanism of the different response is still unclear, however sex hormones have been hypothesized to play a role in this difference: in particular Mitsushima [17] found a sex-specific acetylcholine release in the hippocampus of rats, that could be explained by sex-specific circulating sex steroids.

A confounding factor in our analysis could be the different ancestry between study cohorts (Italian, Brazilian, Chinese), or the occurrence of the winner’s curse phenomenon, for which the effect size of the variant is larger in the discovery than in the replication study.

We were finally unable to confirm an overall involvement of the gene, and impact of variants having an eQTL effect on its expression in brain tissue should be confirmed. We believe that it could be worth to further investigate this gene, especially in the broader context of interactions with other genes in acetylcholine pathway, to better characterize its role in personalized medicine of ChEI medications, and to establish consortium of clinical groups active in the field to increase the sample size and statistical power.