Cross-Sectional Gene-Smoking Interaction Analysis in Relation to Subclinical Atherosclerosis-Results From the IMPROVE Study

Background:

Smoking is associated with carotid intima-media thickness (C-IMT). However, knowledge about how genetics may influence this association is limited. We aimed to perform nonhypothesis driven gene-smoking interaction analyses to identify potential genetic variants, among those included in immune and metabolic platforms, that may modify the effect of smoking on carotid intima-media thickness.

Methods:

We used baseline data from 1551 men and 1700 women, aged 55 to 79, included in a European multi-center study. Carotid intima-media thickness maximum, the maximum of values measured at different locations of the carotid tree, was dichotomized with cut point values ≥75, respectively. Genetic data were retrieved through use of the Illumina Cardio-Metabo- and Immuno- Chips. Gene-smoking interactions were evaluated through calculations of Synergy index (S). After adjustments for multiple testing, P values of <2.4×10⁻⁷ for S were considered significant. The models were adjusted for age, sex, education, physical activity, type of diet, and population stratification.

Results:

Our screening of 207 586 SNPs available for analysis, resulted in the identification of 47 significant gene-smoking synergistic interactions in relation to carotid intima-media thickness maximum. Among the significant SNPs, 28 were in protein coding genes, 2 in noncoding RNA and the remaining 17 in intergenic regions.

Conclusions:

Through nonhypothesis-driven analyses of gene-smoking interactions, several significant results were observed. These may stimulate further research on the role of specific genes in the process that determines the effect of smoking habits on the development of carotid atherosclerosis.

Subclinical atherosclerosis is an asymptomatic, chronic condition that is easily undiagnosed until a clinical event occurs, such as myocardial infarction or stroke.¹ Carotid intima-media thickness (C-IMT), assessed with B-mode ultrasound, a noninvasive method, has been shown to be a valid surrogate marker for subclinical atherosclerosis,² and a predictor for future cardiovascular disease (CVD).^3,4

Previous studies indicate that genetic susceptibility plays an important role in the pathogenesis of atherosclerosis.^5–8 The reported proportions of heritability of carotid atherosclerosis vary between 2% and 78%.⁸ Part of this heritability is likely to be explained by gene-environment interactions.⁸ There are hopes from the scientific community and healthcare that personalized medicine, such as knowledge of how genetic background can interact with modifiable factors and thereby influence cardiovascular risk, will be able to contribute to improved prevention of CVD.

Among the risk factors for premature atherosclerosis, smoking has been identified as a major determinant of atherosclerotic development.^9–11 Studies have shown that smoking exposure and duration of smoking cessation can affect carotid artery structure in all phases of atherosclerosis.^12,13 In an earlier investigation based on data from a European multi-center study IMPROVE (Carotid Intima Media Thickness [IMT] and IMT-Progression as Predictors of Vascular Events in a High Risk European Population), smoking was found to be a major determinant of C-IMT.¹⁴

Previous studies that have investigated gene-smoking interactions behind carotid atherosclerosis were generally performed with a candidate gene approach, and the results are inconclusive.^15–33 Only 2 studies evaluated gene-smoking interactions with an explorative approach, using the whole genome, one based on 669 Hispanics, mainly women, residing in New York,³⁴ and the other based on 1776 men from West Africa.³⁵ These studies are insufficient to detect all important gene-smoking interactions due to their limited sample size. In addition, it is doubtful whether the results can be generalized to populations of other ancestries.

Hence, we aimed to explore gene-smoking interactions behind carotid subclinical atherosclerosis in a multi-center study including men and women of European ancestry. We limited the search for interactions to include genetic variants available via platforms for genetic studies of cardiovascular, metabolic, and immune traits.

Methods

The data that support the findings of this study are available from the corresponding author upon reasonable request.

The Institutional review board at each recruitment center (Karolinska Institutet, Stockholm, Sweden; University of Milan, Milan, Italy; University of Kuopio and Kuopio Research Institute of Exercise Medicine, Kuopio, Finland; University Hospital Groningen, Groningen, The Netherlands; University of Perugia, Perugia, Italy; Groupe Hôpital Pitie-Salpetriere, Paris, France) approved the study. Written informed consents for general participation and for the genotyping were provided by all participants. The study was performed in accordance with the Helsinki Declaration.

Full materials and methods are available in Supplemental Materials.

Results

Baseline characteristics of all study participants and by their smoking status are presented in Table 1. The current smokers were younger, less physically active, and educated than nonsmokers. Smokers had also higher levels of total cholesterol, triglycerides, LDL-C (low-density lipoprotein cholesterol), blood glucose, and High-sensitivity C-reactive protein. However, their level of uric acid and creatinine were lower than in nonsmokers.

Table 1.

Characteristics of the IMPROVE Study Participants by Smoking Status

In total, 207 586 genetic variants were available for analyses. Results from the main analysis investigating gene-smoking interaction in relation to C-IMT_max cut off at the 75th percentile are shown in Table 2. We found 47 single nucleotide polymorphisms (SNPs) significant (P for Synergy index <2.4×10⁻⁷) after Bonferroni correction. All the aforementioned interaction results were synergistic, with Synergy index point estimates in the range between 3.3 and 5.8 (Table S1). Compared with the reference group of nonsmokers without the risk variant, the odds for having C-IMT >75th percentile associated with smoking and having the risk variant were ≈3 to 4-fold higher (Table 2). Of the 47 significant SNPs, 28 were in protein coding genes, 2 in noncoding RNA and the remaining 17 in intergenic regions (Table 3). None of the 47 SNPs involved in the interactions identified in our study were among the published quantitative trait locus data included in the Genotype-Tissue Expression (accessed March 25, 2022).

Table 2.

Significant Gene-Smoking Interaction Results* After Bonferroni Adjustment for Multiple Testing in Relation to C-IMT_max With Cutoff at the 75th Percentile

Table 3.

Genes in Proximity to the Genetic Variants Included in the Significant Gene-Smoking Interaction Results Observed for C-IMT_max With Cutoff at the 75th Percentile

Additional analysis that used C-IMT_max cutoff at the 50th percentile resulted in the identification of 146 SNPs for which a significant synergistic interaction with smoking was observed (Table S2). Among those SNPs, 75 were in protein coding genes, 21 in noncoding RNA, and the remaining 50 in intergenic regions (Table S3). Two of these significant SNPs (rs6032180 in LOC105372631 and rs3744761 in PLCD3) were found both when using the 75th and the 50th percentile C-IMT_max cutoff values.

Analyses of gene-smoking interactions that also considered data where the number of observations for each of the possible combinations of the exposures considered are <10 resulted in the identification of additional significant results for the C-IMT_max, cutoff 75th percentile (Table S4), and for C-IMT_max, cutoff 50th percentile (Table S5). All the observed interactions were synergistic. Of the SNPs that appeared in these results, 130 are located in protein coding genes, 43 in long noncoding RNA, and 84 in intergenic regions (Table S6).

We observed no significant results of interaction on the multiplicative scale.

Discussion

In this population of European descent at high risk of CVD but free of clinical manifestations of CVD, our nonhypothesis-based analyses of gene-smoking interactions resulted in the identification of several genetic variants that may have a role in the process behind the effects of smoking on the development of carotid atherosclerosis. Among the 47 SNPs identified in the main analyses, 8 SNPs (Figure 1) are located in any of 7 coding genes that in previous research have been linked to atherosclerosis development: rs72676073 in the interleukin 23 receptor (IL23R), rs9877192 in the LIM (Lin-11, Islet-1, and Mec-3) domain containing preferred translocation partner in lipoma (LPP), rs2278392 in the 5-hydroxytryptamine receptor 4 (HTR4), rs10810371 in the tetratricopeptide repeat domain 39B (TTC39B), rs7068194 and rs12251673 in the 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase (PFKFB3), rs2511241 in the purinergic receptor P2Y2 (P2RY2), and rs915064 in the potassium voltage-gated channel subfamily H member 5 (KCNH5).^36–42 None of these coding genes were identified in 2 previous studies that evaluated gene-smoking interactions with an explorative approach in relation to carotid atherosclerosis.^34,35 These 2 studies were based on the whole genome and assessed interaction on the multiplicative scale only; significant findings of interaction with smoking were observed for a few genetic variants (rs112017404; rs144170770; rs4941649; rs1192824; rs77461169; rs3751383)^34,35 that were not available in the Cardio-Metabo- and Immuno-Chips.

Figure 1.

Visualization of 8 selected significant results from interaction analyzes. The bars show odds ratio point estimates for the risk of having carotid intima-media thickness (C-IMT) above the 75th percentile associated with (a) the genetic risk variant without the presence of smoking, (b) smoking without the presence of the genetic risk variant, and (c) the genetic risk variant in combination with smoking. Reference category is nonsmoking without the presence of the genetic risk variant. These 8 SNPs are located in coding genes previously linked to the development of atherosclerosis.

Scientific support for relevance of the IL23R gene seems to be emerging; it encodes for a protein, interleukin 23 receptor, involved in the cascade of proinflammatory mediators which may in turn play a role in the development of atherosclerosis.³⁶ Further, the IL23R gene has been previously related to autoimmune disease^43,44 and smoking behavior.⁴⁵ It has been found to synergically interact with smoking in relation to sarcoidosis, an autoimmune disease, in a Swedish population-based case-control study.⁴³ The HTR4 and P2RY2 genes may also possibly be of particular interest. These proteins belong to the family of serotonin and purinergic receptors, respectively. The activation of extracellular nucleotide purinergic receptors, such as ATP, has been suggested to stimulate inflammatory mediators⁴⁶ and regulate the expression of vascular cell adhesion molecule, which is thought to be important for the pathogenesis of atherosclerosis.³⁹ The HTR4 gene has been noted to associate to C-IMT in a previous study based on the IMPROVE study material using a candidate gene approach.⁴¹

Among the 47 SNPs identified in our main analysis of interaction as well as in our additional analyses that used the 50th percentile cutoff, there is a SNP (rs3744761), located in a protein coding gene, the phospholipase C delta 3 (PLCD3) gene, which may be of particular interest due to its link to hypertension. This gene has been identified in the Global Blood Pressure Genetics Consortium genome-wide association study (GWAS) including >34 000 study participants, as one of 8 genes linked to hypertension.⁴⁷ Hypertension, in turn, has been consistently associated with increased C-IMT in several studies including the IMPROVE.^14,48 The identification of the PLCD3 gene in the Global Blood Pressure Genetics was not confirmed in a later larger GWAS: the International Consortium for Blood Pressure (≈200 000 study participants including also Global Blood Pressure Genetics participants).⁴⁹ A possible explanation for this lack of replication may relate to underlying gene-smoking interaction.

Among the 146 significant interaction results generated from analyses that used the 50th percentile C-IMT_max cutoff, 75 are in protein coding genes. Among those, perhaps the most interesting finding involves the APOB (apolipoprotein B) gene (rs550619 and rs570877). The APOB gene encodes for the well-known APOB protein involved in the transportation and metabolism of lipids such as LDL-C, which in turn seems to play a fundamental role in CVD pathophysiology.⁵⁰ Findings from recent Mendelian randomization studies suggest APOB as the predominant lipoprotein trait that accounts for a causal mechanism that links LDL-C to CVD.^51,52 Also, levels of APOB have been noted to increase in relation to smoking tobacco,⁵³ however, not consistently.⁵⁴

The remaining significant results (not discussed above) from analyses based on the C-IMT_max 75th or 50th percentile cutoffs, involve SNPs located in genes previously discussed in relation to: (1) regulation of cardiometabolic factors and related diseases such as obesity, hypertension and diabetes (eg, COBLL1; HFM1, CXCR1; COL21A1, DOCK3; DGKB, BMP1; IDE; KCNQ1; and KCNQ1-AS1, ZC3H10),^55–64 (2) endothelial inflammation and dysfunction (eg, TNFAIP8L1, CCNY, GSE1),^65–67 (3) vascular smooth muscle cell proliferation (eg, VEGFA),⁶⁸ (4) inflammatory diseases (eg, PSORS1C1),⁶⁹ (5) risk of CVD hard end point such as atrial fibrillation and venous thromboembolism (eg, ZFPM2; LMO7),^70,71 and (6) addiction behavior including nicotine dependence (eg, SP140L, THSD7B).^72,73

From the results of our analyses restricted to cell counts of 10 or below, the identification of a SNP located in the PIN2/TERF1 interacting, telomerase inhibitor 1 (PINX1) gene is potentially interesting, because this gene was previously identified in GWAS of subclinical atherosclerosis⁶ and carotid plaque.⁷ However, it was not found to interact with smoking in a previous study on C-IMT using a candidate gene approach.³³ The study addressed multiplicative interactions only.

An important advantage of our study is that we did not limit the gene-smoking interaction analyzes to involve SNPs identified in previous GWAS of C-IMT. It is possible that a gene itself is not associated with C-IMT but becomes important only when smoke exposure occurs. Interestingly, none of the SNPs we have identified as significantly involved in smoking interaction are among the significant findings reported in previous GWAS in relation to C-IMT or smoking behavior.^5,7

Limitations

Our study, just like other exploratory studies, cannot determine which findings are truly positive, and as to whether there are other true effects we did not detect. However, we used the most conservative approach available to adjust for multiple testing, which increases the likelihood that reported findings are true positive. Further, to our knowledge, our study is the largest to date investigating gene-smoking interaction in relation to subclinical atherosclerosis with an explorative approach. Interactions we may have failed to identify should be of a smaller magnitude than those we have identified. Concerning our positive findings, replication analyses using an external study material would have been a good complement. However, no suitable material for replication analyses was available. Another study limitation is that our genetic data were extracted from genetic chips which do not encompass the whole genome; our results are thus limited to genes related to cardiovascular and immunologic traits which means that some of the relevant SNPs related to smoking predisposition may not have been included. An additional limitation is that our results may not be generalized to populations other than those with European ancestry and at high risk of CVD. Finally, there is also a limitation linked to the fact that our chosen method for interaction analyses requires dichotomization of exposure variables; the results may have been diluted because we included former smokers in the same category as the current smokers. However, smoking cessation is considered a risk factor for CVD.⁷⁴ Further, studies on the relation between smoking cessation and C-IMT have not shown any clear decreased risk of C-IMT progression.⁷⁵

Conclusions

In this European population at high risk of CVD, we identified several significant gene-smoking interactions in relation to C-IMT. Further research in this field is urged to build strong scientific evidence that may open new possibilities for improving cardiovascular prevention through personalized recommendations or drug development.

Article Information

Acknowledgments

The authors thank all the study participants of the IMPROVE study (Carotid Intima Media Thickness [IMT] and IMT-Progression as Predictors of Vascular Events in a High Risk European Population). The authors also thank Max Vikström and Paolo Frumento for their statistical support.

Sources of Funding

The IMPROVE study (Carotid Intima Media Thickness [IMT] and IMT-Progression as Predictors of Vascular Events in a High Risk European Population) was supported by the European Commission (contract number: QLG1- CT- 2002- 00896; to Drs Tremoli, Baldassarre, Giral, Kurl, Pirro); Ministero della Salute Ricerca Corrente, Italy (2017; 2018; to Dr Baldassarre). The IMPROVE was also funded by the Swedish Research Council (8691 and 0593), European Commission (contract number: QLG1- CT- 2002- 00896), the Swedish Heart-Lung Foundation, the Swedish Foundation for Strategic Research, the Stockholm County Council (project 562183), and the British Heart Foundation (RG2008/008). The present study was supported by the Swedish Research Council (project 2016-02815 to Dr Leander). Dr Strawbridge is supported by United Kingdom Research and Innovation-Health Data Research United Kingdom Fellowship (MR/S003061/1).

Disclosures

None.

Supplemental Material

Supplemental Methods

Tables S1–S6

References 76–86