Background
Cigarette smoking, a major public health threat across the world, causes more than 8 million deaths globally each year. Despite higher awareness of cigarettes’ adverse effects and ongoing efforts on tobacco control, there still exist 22.3% of the global population being regular smokers (made up of 36.7% of the world’s men and 7.8% of all women) [
1]. Smoking is highly inherited with an estimated heritability of 44% (66% for males and 21% for females, respectively) [
2,
3] and meanwhile is influenced by postnatal environmental conditions (e.g., socioeconomic position, culture). Understanding the modifiable risk factors of smoking as well as its full spectrum of consequence is always an essential and challenging question, especially at the microscale and molecular levels. The term “gut microbiota” refers to all microorganisms that inhabit the human gastrointestinal tract, whose volume reaches trillion level [
4]. Due to its intricate and reciprocal symbiotic relationship with the host, the gut microbiota is closely related to human health, not just intestinal diseases [
5‐
7]. The diversity and quantity of intestinal microbiome are in a dynamic balance, which might be disturbed by various factors, such as genetics, aging, living habits, as well as environmental factors (e.g., cigarette smoke exposure) [
8,
9]. Indeed, mounting observational evidence has reported that tobacco use was associated with alterations in the gut microbiota composition [
10‐
12]. Taking
Bifidobacterium (the representative bacteria of probiotics) for example, existing population studies unanimously found that the abundance of
Bifidobacterium was significantly decreased in current smokers compared with non-smokers, regardless of the ethnicities [
13‐
17]. Instead, smoking cessation, even for short periods, could somewhat restore
Bifidobacterium abundance [
15]. In vivo and in vitro studies [
18,
19] also supported the inhibitory effect of cigarette smoke on the growth of
Bifidobacterium. However, population-level studies with higher evidence levels for causality are lacking.
Given the essential role of the gut microbiota in the regulation of the central nervous system (CNS) [
20], another interesting question is whether smoking behaviors are affected by the gut microbiota. Currently, the microbiota-gut–brain axis, i.e., gut microbiota changes may alter brain function, piqued significant research interest [
21]. A recent review summarized the evidence for the presence of bidirectional communications of the axis, and such crosstalk has been linked to major depressive disorder and other psychiatric disorders [
22]. Another recent study also identified that abundant genetic signals associated with the gut microbiome were enriched in the genes of neurological functions [
23]. In the meantime, the neurological function of the brain was further closely related to tobacco use. Two brain areas, the orbitofrontal cortex and the prefrontal cortex, could interact to turn nicotine cravings on or off [
24,
25]. The dopamine reward circuit in the limbic system of the brain was a widely accepted mechanism of tobacco use that withdrawal from nicotine, the main component of cigarettes, will cause a drastic decrease in dopamine secretion. Moreover, animal studies have directly shown that altering the gut microbiome could affect the reward- and stress-related behavior associated with substance abuse, including tobacco [
26‐
28]. Therefore, the gut microbiota has the possibility to affect smoking though the pathways of microbiota-gut-brain-smoking and on the contrary gut microbial homeostasis could be a potential target for addressing tobacco use via improving brain functions [
29]. However, the direct links from the microbiome to smoking behaviors which concordance with the gut-brain axis were largely unexplored.
Meanwhile, prior evidence has been found that the manner of communication between the microbiota and the brain involves autonomic nervous system with corresponding neurotransmitters (e.g., γ-aminobutyric acid (GABA), endorphins), bacterial metabolites (typically, short-chain fatty acids) [
21], etc. Amino acid metabolites and amino acid-related derivatives are essential sources of most important neurotransmitters. Therefore, associations between the microbiome and smoking could be bridged by relevant metabolites, such as tryptophan (the raw material for serotonin, as known as 5-hydroxytryptamine, 5HT).
Mendelian randomization (MR) is an increasingly used approach to integrate summary data of genome-wide association study (GWAS) to identify causal links between exposures and outcomes. The main reason for its advantage in inferring causality is that MR employs the genetic variants as instrumental variables. MR uses the facts that (1) genetic variants are randomly inherit one allele from each of the father and mother (namely the law of segregation assortment) and (2) alleles will be passed to offspring independently of each other (namely the law of independent assortment). Therefore, MR results are unlikely to be influenced by the environment that might confound the estimated relationship. Recently, the MiBioGen consortium [
30] released numerous microbiome abundance-associated loci, offering an unprecedented chance to explore the causality between the gut microbiota and tobacco use. Based on the knowledge above, we hypothesized that the gut microbiome links smoking behaviors and conducted a two-sample bi-directional MR analysis to elucidate the causal association between the gut microbiota and smoking phenotypes and further explore the potential role of several metabolites on these associations.
Discussion
To our best knowledge, this work is among the first to systematically evaluate the causal relationships between the gut microbiota and tobacco use from a genetic perspective. This two-sample MR study gave reasonably strong evidence that genetically predicted abundance of specific gut microbes play non-negligible roles in the occurrence and progression of cigarette smoking, in which, metabolites may be participating. As for the other direction, the MR confirmed and strengthened the role of smoking on the gut microbiota. Leveraging the power of molecular genetic markers as instrumental variables, the MR approaches largely avoided the interference of confounders (e.g., socioeconomic position, culture) and reversed causality which make regular observational study vulnerable [
45].
As mentioned in the introduction, the theory of the microbiota-gut-brain communication hints at a possible influence of the gut microbiota on smoking. Nevertheless, few studies had directly explored this theme. Leveraging the large-scale GWAS data sources, our MR study filled this knowledge gap from a novel angle. (1) Previous studies indicated that smoking would decrease the abundance of
Actinobacteria while our results found that a lower abundance of
Actinobacteria may cause an increase in the number of cigarettes smoked per day, i.e., worse smoking status. More severe smoking conditions may in turn cause a further diminishment in
Actinobacteria abundance, implying a potential positive feedback effect [
10,
55]. This might partially explain why smokers tend to increase the tobacco use. (2) In addition, observational research showed that, compared to infants from non-smoking families, those from smoking households had lower intestinal flora diversity and abundance, with
Bifidobacterium in particular [
56]. Interestingly, we found a lower abundance of
Bifidobacterium may induce an earlier age of smoking initiation. While conventional wisdom has expounded that early smoking in children may result from early exposure to third-hand smoke or imitation of father’s smoking behavior [
57,
58], our study provides new insights that the early smoking initiation may be proportionally explained by gut flora. Regulating the gut microbiota, such as probiotic intervention, might be an option to redeem bad effects caused by premature smoke exposure.
Rather than just concerning the causality between the gut microbiota and smoking, we also considered the possible involvement of metabolites in this process. (1) Our results suggested an attenuated significance of association between
Peptococcus and smoking after adjusting tryptophan and/or tyrosine, implying a potential metabolite-dependent mechanism of the microbiota on smoking that these two amino acids drove. Wen’s study, from another angle that using metabolomics and 16S rRNA gene sequencing analyses in the rat model, proved the correlation between
Peptococcus and key metabolic pathways, also including tryptophan metabolism and tyrosine metabolism [
59]. (2) MetOrigin is a bioinformatics tool, aiming to identify which bacteria and how they participate in certain metabolic reactions [
60]. A similar implication in our multivariable MR analysis that tryptophan may modify the effects of
Actinobacteria on smoking was also somewhat corroborated in this platform by a simple quick search, which supported the relationship between
Actinobacteria and tryptophan synthase.
There is growing evidence, albeit some indirect, providing possible biological explanations for the mechanisms of commensal gut microbiota on smoking, particularly probiotics such as
Bifidobacterium. (1) The vagus nerve is thought to be a major modulatory constitutive communication pathway between the intestinal bacteria and the brain.
Bifidobacterium longum have been found, via the vagus nerve, to send signals to the brain, leading to the secretion of a higher level of dopamine [
61]. Since dopamine is related to the brain's reward function, higher levels of dopamine will offset the euphoria of smoking or the pain of quitting, thereby reducing smoking addiction [
62]. (2) Neurotransmitters probably mediate the influences of the intestinal microbiome on smoking. For instance,
Bifidobacterium was reported to promote serotonin (5-HT) biosynthesis in colonic enterochromaffin cells by activating the CGA/ADRα2A cascade signal and regulating the TRP/TPH-OR pathways [
63,
64]. 5-HT has been the therapeutic target for addiction to alcohol, cocaine, or drug, so it may also be for smoking [
65]. Other neurotransmitters with similar functions and previously shown to be influenced by
Bifidobacterium also include GABA [
66] and noradrenaline [
67]. (3) The close link of metabolites (e.g., short-chain fatty acids [
68], metabolite acetate [
69]) or components (e.g., peptidoglycan [PGN]) of
Bifidobacterium with the CNS may explain its effect on smoking. Short-chain fatty acids are relevant to the morphology and function of microglia [
70], and metabolite acetate has therapeutic potential to prevent cognitive impairment [
69]. PNG can penetrate the blood-brain barrier, entering the brain, and communicating with the PGN-sensing molecules (Pglyrp2) in the amygdala [
71,
72]. Accordingly, changes in metabolite levels resulting from gut flora dysbiosis make an unavoidable effect on CNS, releasing fear- or anxiety-like emotions or triggering depression, subsequently elevating the risk of smoking initiation or failure to smoking cessation [
73‐
75]. The above biological evidence also explains, to some extent, why the relationship between the gut microbiota and smoking may be modified after adjusting for specific amino acids or short-chain fatty acids. A point worth noting is that these potential mechanisms are not fully evidenced. In addition, there is a non-negligible gap between nicotine cravings and the complex smoking behaviors/pattern observed at the population level. Future studies on the gut microbiome and smoking behaviors are anticipated. Certainly, before moving forward, specialized mechanistic investigations are needed to understand the distinct roles of individual taxa, as most of the currently available mechanistic explanations remain at the generalized whole-gut microbial level.
In the other direction, our findings strengthened and extended existing observational evidence, suggesting that tobacco smoking could disrupt the homeostasis of the intestinal microbiota. (1) Our study supported that initiation of smoking could increase
Intestinimonas abundance, which showed consistency with the results obtained in a previous experimental study. Qu and colleagues observed an elevated level of
Intestinimonas after exposure to NNK plus BaP in mice [
52]. Notably, NNK and BaP, the products of smoking, are the major risk elements for inducing cellular carcinogenesis of lung cancer [
76]. (2) The MR results confirmed the roles of smoking for a higher abundance of
Catenibacterium and a lower abundance of
Ruminococcaceae which were observed from two cross-sectional studies. In a Bangladeshi population, a study exhibited that the relative abundance of
Catenibacterium was significantly higher in current smokers compared with never-smokers, showing a dose-response relationship with packs of cigarettes smoked per day [
53]. Enrolling 116 healthy male subjects from China, an observational study revealed that smoking could lower the abundance of
Ruminococcaceae, which was independent of BMI and age [
54]. Importantly, MR design allows for more reliable results with the highest evidence hierarchy other than randomized controlled trials (RCT) [
45]. (3) In addition, Wang et al. reported that cigarette smoking significantly reduced the level of
Lactococcus [
77]. The current MR study further pointed out that the younger the year of smoking initiation, the greater this reduction. (4) Apart from the above, there also appeared several significant evidence for the effect of smoking on
Eisenbergiella,
Pasteurellaceae,
Christensenellaceae,
Haemophilus,
Romboutsia, and
Coriobacteriaceae, which were rarely addressed or not clearly understood before. (5) Nevertheless, it was noteworthy that for some microbial taxa, such as
Bifidobacterium and
Actinobacteria, the existing literature reported the impact of smoking on these taxa, while our work did not provide corresponding strong causal evidence, although most of the effect estimates were consistent in the direction. The main mechanisms by which smoking affects the gut microbiota include the following: raising the pH of the intestinal environment [
18], inducing chronic low-grade inflammation or inflammation-related diseases [
78], as well as promoting oxidative stress [
79].
Several limitations of our study should be acknowledged. Firstly, to reduce the potential effect of weak IV bias, we applied a stricter
P-value cutoff (1e − 06), compared with 1e − 05 which was used in the original paper [
31] and another recent paper [
80]. Thus, it may result in insufficient statistical power, a critical reason for false negatives. Because of the large number of microbial taxa, as well as the hierarchical structure (the abundance could be highly correlated for a microbial strain), and correlations among smoking phenotypes, the multiple comparison adjustment, especially global multiple corrections, may be excessive, further affecting the false negative. Therefore, causality cannot be completely ruled out in negative results, which should be treated with caution. Secondly, since the majority of participants in the GWAS of tobacco use were ancestrally Europeans, extrapolation of the results in the present study to other ethnic groups might be limited. Thirdly, although most of the participants of the gut microbial GWAS were ancestrally Europeans, the ethnic proportion was not perfectly matched between the two samples (i.e., the exposure GWAS and the outcome GWAS dataset), which may result in some levels of inconsistency in LD correlations. Fourthly, smoking is predominantly prevalent among men, and the composition of the gut microbiota also somewhat varies by gender. However, our work cannot analyze the two genders separately. Likewise, the estimates of a lifetime effect of the gut microbiota on smoking provided by MR cannot deliver much clinical meaningful for age-specific interventions. The limited sample size may also prevent us from providing a sufficiently precise estimate as well as 95% confidence intervals for clinical practice. It would be helpful to perform a gender- or age-specific MR analysis especially with larger sample size in future endeavors. Finally, the metabolites analyzed in multivariable MR were detected in human serum. We think that more appropriate and direct information may be generated from fecal samples, but unfortunately, this kind of data is currently lacking. Additionally, direct analysis of all metabolites may leave the hypothesis without sufficient biological evidence, whereas a biologically informed selection may reduce significant findings. Of a certainty, there still exist other neurotransmitters such as norepinephrine as well as other more important short-chain fatty acids such as propionate and butyrate, but the summary data of itself or its related metabolites is also lacking.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.