Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Apr 1.
Published in final edited form as: Obesity (Silver Spring). 2022 Mar 8;30(4):943–952. doi: 10.1002/oby.23386

Examination of serum metabolome altered by cigarette smoking identified novel metabolites mediating smoking-BMI association

Ruiyuan Zhang 1, Xiao Sun 1, Zhijie Huang 1, Yang Pan 1, Adrianna Westbrook 2, Shengxu Li 3, Lydia Bazzano 1, Wei Chen 1, Jiang He 1, Tanika Kelly 1, Changwei Li 1
PMCID: PMC8957487  NIHMSID: NIHMS1772266  PMID: 35258150

Abstract

Introduction:

We hypothesize that an untargeted metabolomics study will identify novel mechanisms underlying smoking associated weight loss.

Methods:

We performed cross-sectional analyses among 1,252 participants in the Bogalusa Heart Study and assessed 1,202 plasma metabolites for mediation effects on smoking-BMI associations. Significant metabolites were tested for associations with smoking genetic risk scores (GRSs) among a subset of participants (n=654) with available genomic data, followed by direction dependence analyses (DDA) to investigate causal relationships between the metabolites and smoking and BMI. All analyses controlled for age, sex, race, education, alcohol drinking and physical activity.

Results:

Compared to never smokers, current and former smokers had 3.31 kg/m2 and 1.77 kg/m2 lower BMI after adjusting for all covariables. Twenty-two xenobiotics and 94 endogenous metabolites were significantly associated with current smoking. Eight xenobiotics were also associated with former smoking. Forty metabolites mediated the smoking-BMI associations, and five showed causal relationships with both smoking and BMI. These metabolites, including 1-oleoyl-GPE (18:1), 1-linoleoyl-GPE (18:2), 1-stearoyl-2-arachidonoyl-GPE (18:0/20:4), alpha-ketobutyrate, and 1-palmitoyl-GPE (16:0) mediated 26.0% of the association between current smoking and BMI.

Conclusions:

We catalogued plasma metabolites altered by cigarette smoking and identified 5 metabolites that partially mediated the association between current smoking and BMI.

Keywords: Smoking, Obesity, Body-Mass Index, BMI, Mediation Analyses, Metabolomics

Introduction

Obesity is a major public health challenge worldwide as it increases the risk of mortality and a series of morbidities including type 2 diabetes, dyslipidemia, cardiovascular disease, and stroke. 1, 2 In the United States, the prevalence of obesity increased from 30.5% in 1999–2000 to 42.4% in 2017–18. 3 Effective intervention strategies and/or therapies are urgently needed to curb the trend of obesity.

Many studies showed that cigarette smoking is related to weight loss. Compared to non-smokers, smokers usually have a lower weight, and quit smoking leads to weight gain. 4, 5 Given the increased risk of cardiovascular disease and cancer, 6 cigarette smoking should not be promoted to reduce body weight. However, understanding the mechanisms underlying smoking associated weight loss will help discover novel treatment targets for obesity. As the main component of cigarette, nicotine plays an important role in weight loss.5 Nicotine could increase leptin sensitivity and its downstream transduction cascades, leading to higher resting metabolic rate and energy expenditure. 7 Nicotine also regulates weight loss by triggering the release of catecholamines. 8 A meta-analysis of 31 randomized controlled trials among 5650 smokers found that nicotine patches or spray supplemented therapies were associated with less weight gain compared to other non-nicotine supplemented smoking cessation treatments. 5 Besides nicotine, cigarettes contain many other gradients, such as formaldehyde, hydrogen cyanide, and ammonia. 9 Multiple biological pathways may be involved in smoking related weight loss. 10

Metabolomics can detect subtle changes in molecular profiles related to endogenous pathological processes, as well as dietary, environmental, and gut microbial influences. As the endpoint of internal and external molecular processes, examination of the human metabolome provides an opportunity to discover biological pathways linking cigarette smoking to weight loss. Previous studies have identified many circulating metabolites that were associated with cigarette smoking and/or body mass index (BMI), 10–12 however, very few studies investigated the mediating roles of metabolites in the effect of cigarette smoking on body weight. Therefore, we conducted an untargeted metabolomics study to identify serum metabolites that mediate the association between smoking and BMI among participants of The Bogalusa Heart Study.

Method

Study Population

The Bogalusa Heart Study (BHS) was designed to investigate the early natural history of cardiovascular disease among a biracial sample (35% African American and 65% white) of residents from Bogalusa, Louisiana. 13 Since 1973, BHS participants were surveyed every 3–4 years from childhood through adulthood. The current study included 1,261 participants who were screened at least twice during childhood and twice during adulthood. Blood samples, smoking status, and BMI were collected during the 2013–2016 visit cycle.

Metabolomics profiling and quality control

Among the 1,261 BHS participants, a random sample of 64 participants were selected to have their blood samples collected twice to serve as blind duplicates. Thus, a total of 1,325 fasting serum samples were processed using an untargeted, ultrahigh performance liquid chromatography-tandem mass spectroscopy-based metabolomic quantification protocol. Detailed metabolomics profiling and quality control measures were described previously. 14 In brief, a technical replicate generated by taking a small fraction of every sample and a process blank were used to remove system artifacts, mis-assignments, and background noises. Missing values for each metabolite were imputed by the minimum detection limit of the platform. 15 Variations due to instrument inter-day tuning differences were corrected by a normalization step. A total of 956 known and 510 unknown metabolites were identified, unknown compounds were tagged beginning with X and followed by numbers.

We calculated Spearman correlation coefficient (reliability coefficient) for each metabolite among the 64 blind duplicate samples. Metabolites with reliability coefficient ≤ 0.3 or missing rate / below-detection-rate ≥ 80% were removed (n=264). A total of 1,202 metabolites were included in the analyses. There are 167 metabolites with a missing rate / below-detection-rate between 50% and 80%, and these metabolites were categorized as 1=missing or below detection limit, 2=greater than detection limit but below median, and 3=equal or greater than median.14 The rest of metabolites (n=1,035) were normalized.

Smoking status and covariables

Demographic information (age, sex, race, and education) and lifestyle behaviors (smoking, drinking, and physical activity) were collected by survey questionnaires. Education level was categorized into high school or less and more than high school. Smoking and drinking status were measured as never, former and current smokers/drinkers. Physical activity was estimated using the International Physical Activity Questionnaire. 16 Height and weight were measured twice (to ±0.1 cm and to ±0.1 kg, respectively) during a physical examination, and the mean height and weight were used to calculate BMI in kg/m2. Total calory intake was estimated based on the dietary data collected by the Youth/Adolescent Questionnaire. 17

Genotype imputation and smoking genetic risk score (GRS)

We also performed Mendelian randomization analyses to infer causal relationships between smoking and circulating metabolites among 654 BHS participants with both genome-wide genotype data and metabolomics data. Genome-wide data was genotyped by the Illumina BeadChip Human670K platform and imputed to the 1000 genome reference panel using standard method.18, 19 After imputation, SNPs with an imputation quality r2 < 0.3 or minor allele frequency < 0.01 were removed. GRS for smoking was calculated based on findings from a large-scale genome-wide association study of smoking among 1.2 million participants of European ancestry 20. This study identified 378 variants independently associated with ever regularly smoking. A weighted GRS for smoking was constructed by combining risk alleles for smoking weighted by their corresponding effect sizes reported in the original study. An unweighted GRS was also constructed by counting the number of risk alleles of the 378 variants.

Statistical Analysis

Characteristics of the study participants were displayed for the overall sample and by smoking status. Categorical variables were presented as numbers and percentages, and continuous variables were shown as mean and standard deviation. Continuous variables were checked for normality and log-transformed if necessary.

Mediation analyses:

We evaluated the mediation effects of circulating metabolites on smoking-BMI associations by three linear regression models: One modeled the effect of smoking on an individual metabolite, and the other two modeled the effects of smoking on BMI with and without controlling for the metabolite. Mediation effect of a metabolite was estimated by multiplying the effect of smoking on the metabolite (noted as a) by the effect of the metabolite on BMI after controlling for smoking (noted as b). The 95% confidence interval (CI) and p value for a mediation effect were estimated using a Monte-Carlo method with 110,000 times simulation based on the coefficients and standard errors of a and b. 21 All models controlled for age, sex, race, education, drinking, and physical activity. Significant metabolites were determined by Bonferroni correction for 1,202 metabolites (p<0.05/1202/2=2.08E-5). Since smoking could inhibit calory intake and lead to decrease of body weight, for significant metabolites, we conducted sensitivity analyses by further controlling for total calory intake. Because smoking patterns are different among race and sex subgroups in our study population, we further conducted post-hoc subgroup analyses by race and sex, applying the same mediation model to investigate the robustness of our main findings and to identify race-/sex-specific metabolites. The overall mediation effect of all significant metabolites was assessed by adding them all in a mediation model. To avoid collinearity among significant metabolites, we also performed principal component analyses for all significant metabolites and evaluated mediation effects of the first 6 principal components of the significant metabolites.

Mendelian randomization analyses:

Due to the cross-sectional design of the current study, mediation analysis could not determine the causal directions among the exposure, mediators, and outcome. To identify the causal relationship between smoking and significant metabolites identified in the mediation analyses, we conducted Mendelian randomization analyses among 654 BHS participants with both genome-wide genotype and metabolomics data. We tested the association between smoking GRS and a metabolite using multivariable linear regression controlling for the same covariates as in the mediation analyses. Metabolites significantly associated with smoking GRS were further evaluated for causal relationships with BMI.

Direction dependence analysis (DDA):

Given very few GWAS on metabolites, GRS for metabolites were not available for Mendelian randomization analyses. We applied DDA to investigate the causal directions between metabolites and BMI. The DDA was developed to determine causal relationships between non-normally distributed variables in cross-sectional studies.22 If an explanatory variable (X) is the cause of an outcome variable (Y), DDA tests the following three assumptions: 1) the outcome variable (Y) should be more normally distributed than the explanatory variable (X); 23 2) the true model (X→Y) should have a more normally distributed residual compared to the reverse model (Y→X); 24 and 3) the true model (X→Y) should hold the homoscedasticity while the reverse model (Y→X) does not. 22 We examined the distributions of BMI and significant metabolites identified in the Mendelian randomization analyses and performed DDA to test causal directions of the metabolites-BMI associations. A significant causal relationship in DDA is a model that satisfies all the three assumptions.

Result

Characteristics of the study participants

A total of 1,252 BHS participants were included in the mediation analyses. As shown in Table 1, participants were middle-aged (mean age=48.2 years) and more likely to be females (58.7%) and whites (65.7%). About half of the participants (49.1%) had more than high school education, the majority were ever drinkers (87.7%), and 55.8% of the participants were current drinkers. About half of the participants (49.1%) ever smoked, and 19.8% were current smokers. Participants had a mean BMI of 31.4 kg/m2. Compare to never smokers, former and current smokers had lower BMI and education levels but a higher physical activity level (Table 1).

Table 1.

Characteristics of Bogalusa Heart Study Participants

Overall (N=1,252) Never Smoker (N=637) Former Smoker (N=367) Current Smoker (N=248)
Body mass index, kg/m2, mean (SD) 31.4 (7.8) 32.5 (8.0) 30.7 (7.0) 29.7 (7.9)
Age, years, mean (SD) 48.2 (5.3) 48.0 (5.2) 49.0 (5.2) 47.4 (5.5)
Female, % 735 (58.7) 406 (63.7) 207 (56.4) 122 (49.2)
Black, % 429 (34.3) 211 (33.2) 117 (31.9) 101 (40.7)
Education, >High school, % 613 (49.1) 372 (58.4) 165 (45.3) 76 (30.6)
Drinking status, %
 Current drinker 699 (55.8) 340 (53.4) 214 (58.3) 145 (58.5)
 Former drinker 400 (31.9) 182 (28.6) 134 (36.5) 84 (33.9)
 Never drinker 153 (12.2) 115 (18.1) 19 (5.2) 19 (7.7)
Physical activity, ET, mean (SD) 7,145.0 (7187.3) 6,781.6 (6918.8) 7,146.8 (7048.6) 8,044.5 (7952.4)
Smoking GRS, weighted −0.09 (0.24) −0.11 (0.24) −0.05 (0.24) −0.05 (0.23)
Smoking GRS, unweighted 399.3 (12.3) 397.9 (12.4) 401.1 (12.0) 400.9 (11.7)
Open in a new tab

Note: MET=Metabolic Equivalents; SD=standard deviation.

Serum metabolome altered by cigarette smoking

A total of 116 metabolites of known identifies and 52 unknown metabolites showed significant associations with current cigarette smoking after adjusting for age, sex, race, education, drinking, and physical activity (Figure 1 and supplementary Table S1). The 116 known metabolites included 56 lipids, 22 xenobiotics, 14 amino acids, 10 cofactors and vitamins, 6 carbohydrates, 6 nucleotides, and 2 peptides. The top 10 metabolites with the lowest p values were 3 tobacco metabolites, 3 benzoates, food component 4-vinylguaiacol sulfate, and chemicals 3-acetylphenol sulfate, tryptophan 7-hydroxyindole sulfate, and sterol 3beta-hydroxy-5-cholestenoate. Although the levels were much lower in former smokers than current smokers, 8 of the xenobiotics, including 3 tobacco metabolites, 3 benzoates, chemical 3-acetylphenol sulfate, and food component 4-vinylguaiacol sulfate, were also associated with former smoking.

Figure 1.

Figure 1.

Open in a new tab

Volcano Plots for the Association between Metabolites and Current Smoking (a) or Former Smoking (b).

Mediation analysis results

Compared to never smokers, current and former smokers had 3.31 kg/m2 (95% CI: 2.15–4.47) and 1.77 kg/m2 (95% CI: 0.75–2.79) lower BMI after adjusting for age, sex, race, education, drinking, and physical activity (Figure 2). A total of 31 metabolites of known identities and 9 unknown metabolites significantly mediated the BMI differences between current and never smokers (Figure 3 and supplementary Table S2). Among the 40 metabolites, only 7-hydroxyindole sulfate and X-24811 significantly mediated the BMI differences between former and never smokers (Table S2), and 22 metabolites showed nominal (P<0.05) mediation effects for former smoking. Furthermore, most of the (36 out of 40) metabolites mediated smaller percentage of BMI differences for former smoking than for current smoking.

Figure 2. Effect of Smoking on BMI before and after Adjusted for Mediators.

Figure 2.

Open in a new tab

Note: PC=principal component. First 6 PCs were used because they could explain 50% of variation for all mediators.

Black lines are 95% CIs of the effect of smoking; blue circle is the effect of current smoking; red circle is the effect of former smoking.

*Significant mediators include 1-oleoyl-GPE (18:1), 1-linoleoyl-GPE (18:2), 1-stearoyl-2-arachidonoyl-GPE (18:0/20:4), alpha-ketobutyrate, and 1-palmitoyl-GPE (16:0).

** Alpha-ketobutyrate was removed

Figure 3.

Figure 3.

Open in a new tab

Volcano Plots for the Mediation Effects of Metabolites on the Associations of Current Smoking (a) and Former Smoking (b) with Body Mass Index.

The 40 metabolites together explained 90.7% of the BMI differences between current and never smokers and 76.9% of the BMI differences between former and never smokers (Figure 2). The first 6 principal components of the 40 metabolites captured 50% variations of the metabolites and explained 87.3% of the BMI differences between current and never smokers and 67.5% of the BMI differences between former and current smokers (Figure 2). After further controlling for calories intake in sensitivity analyses, these metabolites still mediated smoking-BMI associations (p value < 0.0001) (supplementary Table S3). Metabolite 3beta-hydroxy-5-cholestenoate had the largest mediation effect (−2.52 kg/m2, 95% CI: [−3.10, −1.97]) on the association between current smoking and BMI.

In subgroup analyses by race and sex groups, most of the significant metabolites identified in the overall participants remained nominally significant (p<0.05) and had the same effect directions, except for 1 metabolite for White participants, 12 metabolites for Black participants, 3 metabolites for men, and 7 metabolites for women (Table S4). We also identified 11 race-/sex-specific metabolites that significantly mediated the BMI difference between current smokers and never smokers, including 5 metabolites in White population, 1 metabolite in men, and 5 metabolites in women (Table S5).

Mendelian randomization and DDA results

Smoking GRS was associated with 9 metabolites (P<0.05), of which, 5 metabolites showed causal associations with BMI in DDA (Table 2 and supplementary Table S6). These metabolites together mediated 26.0% of the association between current smoking and BMI, and 13.6% of the association of former smoking with BMI (Figure 2). The association directions with smoking GRS were consistent with smoking for all metabolites except for the alpha-ketobutyrate (Table 2 and supplementary Table S1). The rest 4 metabolites together mediated, respectively, 14.4% and 4.0% of the associations of current and former smoking with BMI (Figure 2). Four metabolites, alpha-ketobutyrate, 1-palmitoyl-GPE (16:0) [LPE (16:0)], 1-oleoyl-GPE (18:1) [LPE (18:1)], and 1-linoleoyl-GPE (18:2) [LPE (18:2)] mediated the BMI decreasing effect of smoking with effect sizes ranging from −0.63 to −0.44 kg/m2. Interestingly, 1-stearoyl-2-arachidonoyl-GPE (18:0/20:4) [PE (18:0/20:4(5Z,8Z,11Z,14Z)] mediated the BMI increasing effect of smoking (mediation beta=0.39 kg/m2, 95% CI: 0.18 to 0.65). Of these metabolites, LPE (16:0) showed the strongest mediation effect, explaining 19.0% of the association between current smoking and BMI. None of the race-/sex-specific metabolites were significant in the Mendelian randomization analyses or DDA.

Table 2.

Findings of Causal Inference using Mendelian Randomization and Direction Dependence Analyses for Significant Metabolites Identified in the Mediation Analyses.

 Proportion mediated Smoking GRS, weighted
 Smoking GRS, unweighted
 Suggested causal directions by direction dependence analyses
Biochemical Beta (s.e.) P Beta (s.e.) 1 P Observed variables Residuals Independence properties of errors
Amino Acid
alpha-ketobutyrate2 19.6% 0.27 (0.14) 0.05* 0.05 (0.03) 0.05* Metabolite→BMI Metabolite→BMI Metabolite→BMI
Homoarginine 14.8% −0.16 (0.07) 0.02* −0.03 (0.01) 0.01* Cannot determine Cannot determine BMI→Metabolite
Cofactors and Vitamins
beta-cryptoxanthin −16.0% −0.39 (0.26) 0.14 −0.10 (0.05) 0.05* Metabolite→BMI Metabolite→BMI Cannot determine
Lipid
1-linoleoyl-GPE (18:2) 17.7% 0.18 (0.08) 0.03* 0.04 (0.02) 0.02* Metabolite→BMI Metabolite→BMI Metabolite→BMI
1-oleoyl-GPE (18:1) 13.4% 0.19 (0.10) 0.06 0.04 (0.02) 0.03* Metabolite→BMI Metabolite→BMI Metabolite→BMI
1-palmitoyl-GPE (16:0) 19.0% 0.16 (0.08) 0.05* 0.04 (0.02) 0.02* Metabolite→BMI Metabolite→BMI Metabolite→BMI
1-stearoyl-2-arachidonoyl-GPE (18:0/20:4) −11.8% 0.22 (0.09) 0.01* 0.05 (0.02) 0.01* Metabolite→BMI Metabolite→BMI Metabolite→BMI
sphingomyelin (d17:2/16:0, d18:2/15:0) 27.9% −0.16 (0.08) 0.03* −0.03 (0.02) 0.04* Cannot determine Cannot determine Metabolite→BMI
sphingomyelin (d18:2/21:0, d16:2/23:0) 16.2% −0.13 (0.07) 0.04* −0.03 (0.01) 0.03* Cannot determine Cannot determine Cannot determine
Open in a new tab

Note: s.e.=standard error; bold=significant p value

1

The effect size per 10 unit increase

2

The direction of association is different between smoking and smoking GRS for this metabolite

Discussion

In the biracial, middle-aged, and cohort with overweight/obesity of BHS, current and former smokers had lower BMI compared to never smokers. We provided a comprehensive catalog of external metabolites associated with and endogenous metabolites in response to cigarette smoking. Among many xenobiotics associated with current smoking, eight continuously existed among former smokers. We identified 40 metabolites that mediated associations between current/former smoking and BMI. Most of them were robust in race-/sex-specific subgroups. We also identified 11 race-/sex-specific metabolites. Causal inferences using Mendelian randomization analyses and DDA provided evidence for five metabolites that mediated the effect of current smoking on BMI. These findings provided mechanistic insight of the smoking-BMI association.

Three lipids, LPE(16:0), LPE(18:1), and LPE (18:2) mediated the BMI decreasing effect of smoking. LPEs are metabolic products of PE by phospholipase A2. 25 In our study, significant LPE metabolites was positively associated with smoking and negatively associated with BMI. The findings are consistent with those of previous studies. 26–30 Lietz et al. showed that LPEs had the largest increment among all lipid species in aortas of mice exposed to cigarette smoking. 26 Ren et al. exposed rat to different concentration of tobacco smoking and found that LPE (16:0), LPE (18:1), LPE (18:0), and LPE (20:4) were up-regulated. 30 Meanwhile, an untargeted metabolomic study among more than 1,000 Mexican Americans discovered that LPE was negatively associated with BMI. 27 Another human study also found that LPE metabolism is impaired among individuals with obesity. 29 More importantly, in a randomized controlled trial, LPE (18:1) and LPE (18:2) were increased while BMI was controlled by dietary intervention. 28 Our study provided further evidence that the three chemicals mediated the BMI decreasing effect of current smoking. Two mechanisms potentially explain the impact of LPEs on BMI. First, LPEs could stimulate Ca2+ signaling and elevate intracellular Ca2+ level, 31, 32 which can inhibit the differentiation of pre-adipocytes into mature adipocyte in both cell lines33, 34 and animal models. 35 Second, obesity is characterized by chronic inflammation, 36 and LPEs could regulate body weight through inhibiting the production of pro-inflammatory cytokines and promoting the secretion of anti-inflammation cytokine IL-10.37, 38 Future studies tracing changes of the three metabolites after quit smoking and their subsequent impact on BMI will provide better evidence on their roles in body weight regulation.

Alpha-ketobutyrate was negatively associated with cigarette smoking and positively associated with BMI, and therefore, mediated the BMI decreasing effect of smoking. However, its association with smoking GRS was in an opposite direction as with cigarette smoking. It is possible some other ingredients of cigarette smoke suppressed the release of this metabolite. Alpha-ketobutyrate is a by-product of the reaction to form cysteine from cystathionine by cystathionase, and can be oxidized to propionyl-CoA by branched chain α-ketoacid dehydrogenase (BCKD). 39 Oral recombinant methioninase that catabolize methionine to alpha-ketobutyrate prevented weight gain for mice under high fat diet. 40 However, blood alpha-ketobutyrate or its derivative alpha-hydroxybutyrate was elevated in insulin-resistant or diabetic individuals with obesity. 39, 41 Larger scale of Mendelian randomization studies are needed to confirm the association between smoking GRS and alph-ketobutyrate. Meanwhile, longitudinal studies evaluating the association of this metabolite with body weight change are needed.

Interestingly, PE(18:0/20:4(5Z,8Z,11Z,14Z)) was positively associated with cigarette smoking and mediated its BMI increasing effect. PE is the second most abundant glycerophospholipid in eukaryotic cells and is essential for various cellular processes. 42 Padmavathi et al. found that smoking could up-regulate the levels of PE and down-regulate phosphatidylcholine (PC) in males. 43 Although no association between PE and body weight has been reported in previous studies, several investigations found that the deficiency of phosphatidylethanolamine N-methyltransferase (PEMT), a small integral membrane protein that convert PE to PC, decreased the risk of high-fat diet induced obesity and insulin resistance by inhibiting the differentiation of preadipocytes into mature adipose cells. 44, 45 The role of PE(18:0/20:4(5Z,8Z,11Z,14Z)) in body weight regulation warrants further investigation.

Through comparing the serum metabolome between current smokers and never smokers, we identify both external and endogenous metabolites associated with current smoking. The external metabolites included not only tobacco metabolites, but also benzoates, food components, food component/plant, chemicals, and xanthine. Some of the external metabolites are products of nicotine or additives of tobacco, while others may be a result of other co-existing behaviors of smoking, such as coffee drinking. Furthermore, although in a lower level than those among current smokers, eight xenobiotics including 3 tobacco metabolites, 3 benzoates, chemical 3-acetylphenol sulfate, and food component 4-vinylguaiacol sulfate were also present in former smokers. Origins of those metabolites in former smokers warrant further investigation. It is possible that some of the cigarette smoking related external metabolites can exist in the body for a long time. It is also possible that the metabolites in former smokers were due to exposure to second hand smoking or even information bias that some former smokers may still smoke. Finally, some of the metabolites may originate from lifestyle behaviors associated with smoking.

Our study has several strengths. First, to our knowledge, this is the first study that applied mediation analysis to investigate the three-way relationships between smoking, serum metabolome, and BMI. Furthermore, the identified relationships remained significant even after we further adjusted for total calorie intake in our sensitivity analysis. Third, we successfully demonstrated the causal directions for 5 metabolites, with supporting evidences from previous studies. However, there are also some limitations in our study. First, although we used various method to demonstrate the causal relationship, the cross-sectional design of our study prohibits temporal inferences. Longitudinal studies to investigate associations of the identified metabolites with long-term change in body weight are needed to delineate the temporal relationships. Second, the sample size, particularly for the Mendelian randomization analyses and subgroup analyses, was relatively small, which limited the ability of our analyses to identify metabolites that have small mediation effects.

To conclude, our study provided a comprehensive list of serum metabolites altered by cigarette smoking. We identified 40 metabolites mediated the cross-sectional associations between current smoking and BMI. Through causal inference using Mendelian randomization and DDA, we discovered five metabolites mediated the effects of cigarette smoking on BMI. The five metabolites together mediated 26.0% of the association between current smoking and BMI, and 13.6% of the association between former smoking and BMI. More studies are needed to identify additional metabolites mediating the effect of smoking on BMI. Furthermore, prospective cohort studies, metabolomics studies among quit smoking individuals, and animal studies will further delineate the effects of metabolites reported in our study on body weight regulation.

Supplementary Material

tS1-6

What is already known about this subject?

  • Obesity is a major public health challenge worldwide;

  • Cigarette smoking is related to weight loss, partly because of nicotine.

What are the new findings in your manuscript?

  • A total of 168 metabolites showed significant associations with current cigarette smoking.

  • Forty metabolites significantly mediated the BMI differences between current and never smokers, explained 90.7% of the BMI differences.

  • Five metabolites showed causal associations with BMI and smoking, mediating 26.0% of the association between current smoking and BMI, and 13.6% of the association of former smoking and BMI

How might your results change the direction of research or the focus of clinical practice?

  • More studies are needed to identify additional metabolites mediating the effect of smoking on BMI.

  • Furthermore, prospective cohort studies, metabolomics studies among quit smoking individuals, and animal studies will further delineate the effects of metabolites reported in our study on body weight regulation.

Funding information:

The work was supported by multiple grants from the National Institutes of Health, including awards R21AG051914 and R01AG041200 from the National Institute on Aging and award 1P20GM109036-01A1 from the National Institute of General Medical Sciences.

Footnotes

Conflict of Interest Statement: The authors have declared that no conflict of interest exists.

Reference

  • 1.Kopelman PG. Obesity as a medical problem. Nature. Apr 2000;404(6778):635–43. doi: 10.1038/35007508 [DOI] [PubMed] [Google Scholar]
  • 2.Blüher M Obesity: global epidemiology and pathogenesis. Nat Rev Endocrinol. 05 2019;15(5):288–298. doi: 10.1038/s41574-019-0176-8 [DOI] [PubMed] [Google Scholar]
  • 3.Hales CM CM, Fryar CD, Ogden CL. Prevalence of obesity and severe obesity among adults: United States, 2017–2018. NCHS Data Brief,. 2020;no 360 [PubMed] [Google Scholar]
  • 4.Albanes D, Jones DY, Micozzi MS, Mattson MEJAjoph. Associations between smoking and body weight in the US population: analysis of NHANES II. 1987;77(4):439–444. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Hsieh MT, Tseng PT, Wu YC, et al. Effects of different pharmacologic smoking cessation treatments on body weight changes and success rates in patients with nicotine dependence: A network meta-analysis. 2019;20(6):895–905. [DOI] [PubMed] [Google Scholar]
  • 6.Messner B, Bernhard DJA, thrombosis,, biology v. Smoking and cardiovascular disease: mechanisms of endothelial dysfunction and early atherogenesis. 2014;34(3):509–515. [DOI] [PubMed] [Google Scholar]
  • 7.Audrain-McGovern J, Benowitz NJCP, Therapeutics. Cigarette smoking, nicotine, and body weight. 2011;90(1):164–168. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Benowitz NLJPicd. Cigarette smoking and cardiovascular disease: pathophysiology and implications for treatment. 2003;46(1):91–111. [DOI] [PubMed] [Google Scholar]
  • 9.Avezov K, Reznick AZ, Aizenbud DJAoob. LDH enzyme activity in human saliva: the effect of exposure to cigarette smoke and its different components. 2014;59(2):142–148. [DOI] [PubMed] [Google Scholar]
  • 10.Gu F, Derkach A, Freedman ND, et al. Cigarette smoking behaviour and blood metabolomics. 2016;45(5):1421–1432. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Xie B, Waters MJ, Schirra HJJJoB, Biotechnology. Investigating potential mechanisms of obesity by metabolomics. 2012;2012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Qi Y, Jiang C, Cheng J, et al. Bile acid signaling in lipid metabolism: metabolomic and lipidomic analysis of lipid and bile acid markers linked to anti-obesity and anti-diabetes in mice. 2015;1851(1):19–29. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Li S, Chen W, Srinivasan SR, et al. Childhood cardiovascular risk factors and carotid vascular changes in adulthood: the Bogalusa Heart Study. 2003;290(17):2271–2276. [DOI] [PubMed] [Google Scholar]
  • 14.Li C, He J, Li S, et al. Novel Metabolites Are Associated With Augmentation Index and Pulse Wave Velocity: Findings From the Bogalusa Heart Study. American Journal of Hypertension. 2019;32(6):547–556. doi: 10.1093/ajh/hpz046 %J American Journal of Hypertension [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Zheng Y, Yu B, Alexander D, et al. Metabolomics and incident hypertension among blacks: the atherosclerosis risk in communities study. Hypertension. Aug 2013;62(2):398–403. doi: 10.1161/HYPERTENSIONAHA.113.01166 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Craig CL, Marshall Al Fau - Sjöström M, Sjöström M Fau - Bauman AE, et al. International physical activity questionnaire: 12-country reliability and validity. (0195–9131 (Print)) [DOI] [PubMed]
  • 17.Rockett HR, Breitenbach M, Frazier AL, et al. Validation of a youth/adolescent food frequency questionnaire. 1997;26(6):808–816. [DOI] [PubMed] [Google Scholar]
  • 18.Delaneau O, Marchini J, McVean GA, et al. Integrating sequence and array data to create an improved 1000 Genomes Project haplotype reference panel. Nature Communications. 2014/06/13 2014;5(1):3934. doi: 10.1038/ncomms4934 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Das S, Forer L, Schönherr S, et al. Next-generation genotype imputation service and methods. Nature Genetics. 2016/10/01 2016;48(10):1284–1287. doi: 10.1038/ng.3656 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Liu M, Jiang Y, Wedow RA-O, et al. Association studies of up to 1.2 million individuals yield new insights into the genetic etiology of tobacco and alcohol use. (1546–1718 (Electronic)) [DOI] [PMC free article] [PubMed]
  • 21.Monte Carlo method for assessing mediation: An interactive tool for creating confidence intervals for indirect effects. 2008.
  • 22.Wiedermann W, Li X. Direction dependence analysis: A framework to test the direction of effects in linear models with an implementation in SPSS. Behavior Research Methods. 2018/08/01 2018;50(4):1581–1601. doi: 10.3758/s13428-018-1031-x [DOI] [PubMed] [Google Scholar]
  • 23.Dodge Y, Yadegari I. On direction of dependence. Metrika. 2010/07/01 2010;72(1):139–150. doi: 10.1007/s00184-009-0273-0 [DOI] [Google Scholar]
  • 24.Wiedermann W, von Eye A. Log-linear models to evaluate direction of effect in binary variables. Statistical Papers. 2020/02/01 2020;61(1):317–346. doi: 10.1007/s00362-017-0936-2 [DOI] [Google Scholar]
  • 25.Makide K, Kitamura H Fau - Sato Y, Sato Y Fau - Okutani M, Okutani M Fau - Aoki J, Aoki J. Emerging lysophospholipid mediators, lysophosphatidylserine, lysophosphatidylthreonine, lysophosphatidylethanolamine and lysophosphatidylglycerol. (1098–8823 (Print)) [DOI] [PubMed]
  • 26.Lietz M, Berges A, Lebrun S, et al. Cigarette-smoke-induced atherogenic lipid profiles in plasma and vascular tissue of apolipoprotein E-deficient mice are attenuated by smoking cessation. Atherosclerosis. 2013/07/01/ 2013;229(1):86–93. doi: 10.1016/j.atherosclerosis.2013.03.036 [DOI] [PubMed] [Google Scholar]
  • 27.Weir JM, Wong G, Barlow CK, et al. Plasma lipid profiling in a large population-based cohort. J Lipid Res. 2013;54(10):2898–2908. doi: 10.1194/jlr.P035808 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Bondia-Pons I, Martinez Ja Fau - de la Iglesia R, de la Iglesia R Fau - Lopez-Legarrea P, et al. Effects of short- and long-term Mediterranean-based dietary treatment on plasma LC-QTOF/MS metabolic profiling of subjects with metabolic syndrome features: The Metabolic Syndrome Reduction in Navarra (RESMENA) randomized controlled trial. (1613–4133 (Electronic)) [DOI] [PubMed]
  • 29.del Bas JM, Caimari A, Rodriguez-Naranjo MI, et al. Impairment of lysophospholipid metabolism in obesity: altered plasma profile and desensitization to the modulatory properties of n–3 polyunsaturated fatty acids in a randomized controlled trial. The American Journal of Clinical Nutrition. 2016;104(2):266–279. doi: 10.3945/ajcn.116.130872 [DOI] [PubMed] [Google Scholar]
  • 30.Ren X, Zhang J, Fu X, et al. LC-MS based metabolomics identification of novel biomarkers of tobacco smoke-induced chronic bronchitis. 2016;30(1):68–74. doi: 10.1002/bmc.3620 [DOI] [PubMed] [Google Scholar]
  • 31.Lee J-M, Park S-J, Im D-SJB, therapeutics. Calcium signaling of lysophosphatidylethanolamine through LPA1 in human SH-SY5Y neuroblastoma cells. 2017;25(2):194. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Hisano K, Kawase S, Mimura T, et al. Structurally different lysophosphatidylethanolamine species stimulate neurite outgrowth in cultured cortical neurons via distinct G-protein-coupled receptors and signaling cascades. 2021;534:179–185. [DOI] [PubMed] [Google Scholar]
  • 33.Suh YH, Kim Wh Fau - Moon C, Moon C Fau - Hong YH, et al. Ectopic expression of Neuronatin potentiates adipogenesis through enhanced phosphorylation of cAMP-response element-binding protein in 3T3-L1 cells. (0006–291X (Print)) [DOI] [PubMed]
  • 34.Ntambi JM, Takova TJD. Role of Ca2+ in the early stages of murine adipocyte differentiation as evidenced by calcium mobilizing agents. 1996;60(3):151–158. [DOI] [PubMed] [Google Scholar]
  • 35.Wang C-H, Chen Y-F, Wu C-Y, et al. Cisd2 modulates the differentiation and functioning of adipocytes by regulating intracellular Ca2+ homeostasis. Human Molecular Genetics. 2014;23(18):4770–4785. doi: 10.1093/hmg/ddu193 [DOI] [PubMed] [Google Scholar]
  • 36.Monteiro R, Azevedo I. Chronic Inflammation in Obesity and the Metabolic Syndrome. Mediators of Inflammation. 2010/07/14 2010;2010:289645. doi: 10.1155/2010/289645 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Wang X, Li Y-F, Nanayakkara G, et al. Lysophospholipid Receptors, as Novel Conditional Danger Receptors and Homeostatic Receptors Modulate Inflammation-Novel Paradigm and Therapeutic Potential. J Cardiovasc Transl Res. 2016;9(4):343–359. doi: 10.1007/s12265-016-9700-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Hung ND, Kim Mr Fau - Sok D-E, Sok DE. 2-Polyunsaturated acyl lysophosphatidylethanolamine attenuates inflammatory response in zymosan A-induced peritonitis in mice. (1558–9307 (Electronic)) [DOI] [PubMed] [Google Scholar]
  • 39.Adams SH. Emerging Perspectives on Essential Amino Acid Metabolism in Obesity and the Insulin-Resistant State. Advances in Nutrition. 2011;2(6):445–456. doi: 10.3945/an.111.000737 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Tashiro Y, Han Q, Tan Y, et al. Oral Recombinant Methioninase Prevents Obesity in Mice on a High-fat Diet. (1791–7549 (Electronic)) [DOI] [PMC free article] [PubMed]
  • 41.Macotela Y, Emanuelli B Fau - Bång AM, Bång Am Fau - Espinoza DO, et al. Dietary leucine--an environmental modifier of insulin resistance acting on multiple levels of metabolism. (1932–6203 (Electronic)) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Calzada E, Onguka O, Claypool SM. Phosphatidylethanolamine Metabolism in Health and Disease. Int Rev Cell Mol Biol. 2016;321:29–88. doi: 10.1016/bs.ircmb.2015.10.001 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Padmavathi P, Reddy VD, Kavitha G, Paramahamsa M, Varadacharyulu N. Chronic cigarette smoking alters erythrocyte membrane lipid composition and properties in male human volunteers. Nitric Oxide. 2010/11/01/ 2010;23(3):181–186. doi: 10.1016/j.niox.2010.05.287 [DOI] [PubMed] [Google Scholar]
  • 44.Presa N, Dominguez-Herrera A, van der Veen JN, Vance DE, Gómez-Muñoz A. Implication of phosphatidylethanolamine N-methyltransferase in adipocyte differentiation. Biochimica et Biophysica Acta (BBA) - Molecular Basis of Disease. 2020/10/01/ 2020;1866(10):165853. doi: 10.1016/j.bbadis.2020.165853 [DOI] [PubMed] [Google Scholar]
  • 45.Vance DEJBeBA-M, Lipids CBo. Physiological roles of phosphatidylethanolamine N-methyltransferase. 2013;1831(3):626–632. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

tS1-6
NIHMS1772266-supplement-tS1-6.docx (78KB, docx)

RESOURCES