Over the past two decades, the important role of gut microbiota has been recognized in the establishment and maintenance of health, as well as in the occurrence and progression of various diseases. A well-maintained gut microbiota diversity is essential for a normal life, whereas alterations to it (dysbiosis) have impacts on the gut-brain axis, leading to a variety of neurological diseases, such as neuropsychiatric disorders (Ochoa-Repáraz et al., 2020; Ni et al., 2021). Gut microbiota alterations are also noticed in patients suffering from different types and areas of chronic pain, including visceral pain, inflammatory pain, headache, neuropathic pain, and chronic widespread pain (Newlove-Delgado et al., 2019; Yang et al., 2019; Chen et al., 2020; Guida et al., 2020; Freidin et al., 2021). Based on these studies, we believe that two possibilities should not be neglected: one is that chronic pain may lead to dysbiosis, and the other is that gut microbiota may be a possible way to regulate chronic pain conditions.

Opioids are a major component in the treatment of chronic pain. While they do relieve pain, the long-term use of opioids is often accompanied by multiple systems disorders in the whole body, including the nervous, cardiovascular, digestive, immune, and endocrine systems (Farmer et al., 2018; de Vries et al., 2020; Hadland et al., 2021; Rosoff et al., 2021; Cai et al., 2022). Due to the side effects mentioned above, it has long been a public concern to reduce opioid use (Sandhu et al., 2018). Recent studies also demonstrated that the long-term use of opioid use may result in increasing comorbidity and behavioral changes due to dysbiosis (Meng et al., 2013, 2015; Wang et al., 2018). Furthermore, opioid receptors are found in both the digestive tract and the central nervous system, which suggests their important role in the modulation of the gut-brain axis (Zhang and Roy, 2021). Notably, observational animal studies also indicate the importance of the gut microbiota in opioid tolerance (Kang et al., 2017; Lee et al., 2018; Zhang et al., 2019).

Looking at the importance of gut microbiota in the gut-brain axis and the critical role of opioids in pain management, it is, therefore, crucial to reveal the causal direction between gut microbiota alteration and prescription opioid use (POU) as well as multisite chronic pain (MCP). A clear causal direction not only guides us on how to maintain a healthy gut microbiota composition but also provides us with a possible strategy for reducing opioid use. Conventionally, well-designed randomized controlled trials are the golden standard for inferring a causal relationship between gut microbiota and POU/MCP; however, they are difficult to implement due to ethical and legal restrictions. Mendelian randomization (MR) is an alternative approach to assess the causal relationship between exposures and outcomes, using genetic variants as unconfounded proxies for exposures (Emdin et al., 2017). Since genetic variants are randomly distributed during meiosis yielding, the MR approach is conceptually similar to a randomized controlled study and can minimize confounding such as social and economic factors (Emdin et al., 2017). Based on the advantages of MR mentioned above, we applied a two-sample bi-directional MR approach to explore the causal relationships between gut microbiota and POU as well as MCP. Metabolites are one of the important bridges between gut microbiota and the central nervous system (Eicher and Mohajeri, 2022); thus, MR analysis was also conducted to find the potential associations between metabolites and POU, as well as MCP.

After clumping and discarding SNPs at linkage disequilibrium (LD) r² > 0.001 within a 10,000 kilobase pairs window, we identified a total of 2,368 and 2,601 SNPs for gut microbiota traits with POU and MCP at the suggestive significance level p < 1 × 10⁻⁵, respectively. Steiger filtering and harmonization were performed after removing SNPs with F-statistics smaller than 10. SNPs that exhibit either pleiotropy or heterogeneity were removed according to the results of MRPRESSO and leave-one-out analysis. At last, 1,999 and 2,188 SNPs were selected for gut microbiota traits with POU and MCP respectively, as shown in Data Sheet 2 in the Supplementary material. After MR analyses, a total of 36 causal associations were identified, including seven gut microbiota traits with POU, five gut microbiota traits with MCP, 12 metabolites traits with POU, and 12 metabolites traits with MCP (Tables 1, 2; Supplementary Tables S4, S5).

In this study, potential bi-directional associations between host genetic-driven gut microbiota and prescription opioid use (POU), as well as multisite chronic pain (MCP) were evaluated. A total of seven and five gut microbiota traits were identified to be causally associated with POU and MCP, respectively. In the other direction, no clear evidence for any significant causal association was found between POU and gut microbiota as well as MCP and gut microbiota. Besides, metabolites including lipid metabolites and neurotransmitters were found to exhibit causal relationships with POU and MCP.

A total of five positive causal associations were identified, including Adlercreutzia, Allisonella, and Dialister at the genus level with POU, Family Clostridiaceae1, and order Gastranaerophilales with MCP. Findings from our study were consistent with a cross-sectional study showing that genera Adlercreutzia and Dialister are significantly related to chronic back pain (Dekker Nitert et al., 2020). The underlying mechanism has not been fully explored; however, it may include the role of gut microbiota in regulating inflammation-driven hyperalgesia, impaired reward behavior, and other neuropsychiatric disorders (Lee et al., 2018; Valles-Colomer et al., 2019; Zhuang et al., 2020). A recent animal study found that gut microbiota and one of its products, valeric acid, may play a crucial role in neuroinflammation(Lai et al., 2021). An increasing number of studies has shown that altered gut microbiota is associated with inflammatory cytokines such as IL-6, IL-8, IFN-γ, and TNF-α (Lin et al., 2019). As is known, inflammatory cytokines play a crucial role in the process of hyperalgesia and thus strengthen the sense of pain, which may also increase the demand for analgesics. Another underlying mechanism is the neuropsychiatric disorders pathway. Many studies found evidence of relationships between gut microbiota and neuropsychiatric diseases and neurodegenerative diseases, for example, major depressive disorder, schizophrenia, and Parkinson’s disease (Zhuang et al., 2020; Rosoff et al., 2021; Ning et al., 2022). One of the crucial symptoms of these diseases is the abnormality in reward processing, which often accompanies with chronic pain and also means increased demand for opioids (Pellissier et al., 2018).

Interestingly, protective causal effects of gut microbiota on POU, as well as gut microbiota on MCP were also found in our study. A total of seven negative relationships were identified, including Anaerofilum, Anaerostipes, ChristensenellaceaeR, and LachnospiraceaeNC2004group at the genus level with POU, genus ErysipelotrichaceaeUCG003, family Actinomycetaceae, and order Actinomycetales with MCP. These associations are broadly supported by previous studies. As is known, Anaerofilum, Anaerostipes, ChristensenellaceaeR.7group, and LachnospiraceaeNC2004group are all Gram-positive and anaerobic bacterial genera from the class Clostridia and the phylum Firmicutes. Class Clostridia is a producer of short-chain fatty acids (SCFAs), mainly including butyrate. It is widely recognized that SCFA butyrate may reduce oxidative damages and suppress inflammation, thus alleviating pain perception and further reducing the demand for pain medications (Russo et al., 2018).

In the other direction, no clear evidence was found to reveal any significant causal relationship between POU and gut microbiota nor MCP and gut microbiota. This was contradicted with traditional studies that generally suggest gut microbiota alteration as a result of chronic pain and long-term use of pain medications such as opioids (Banerjee et al., 2016; Acharya et al., 2017), thus supporting the fact that these previous observational associations may be due to confounding or reverse causality. Santanu et al. showed that long-term use of morphine was significantly associated with gut microbiota alterations and resulted in an increase in the abundance of Gram-positive pathogenic bacterial strains (Banerjee et al., 2016). In another cross-sectional study of patients with diabetes, a similar conclusion was yielded that the use of opioids was accompanied by changes in the gut microbiota, including levels of Bifidobacterium (Barengolts et al., 2018). In addition, Gicquelais et al. performed a 16S rRNA gene sequencing, using fecal samples from 46 patients in an outpatient addiction clinic. Results from this study showed a decrease in Roseburia (butyrate-producing bacteria) and Bilophila in the prescription opioid use group but not in the nonopioid control group (Gicquelais et al., 2020). The reason that our results were different from the above-mentioned observational studies may be due to the small sample sizes (ranging from 46 to 99), which were susceptible to unmeasured confounders such as diet and health status, leading to biased conclusions.

Given that metabolites may play a crucial role in the interaction between gut microbiota and POU/MCP, MR analysis was conducted to find their potential associations. It was surprising to find that a series of lipid metabolites may increase the risk of POU. Higher levels of triacylglycerol, diacylglycerol, and cholesterol ester potentially link to the risk of POU and/or MCP, implying their important roles in the microbiota-host crosstalk. These results are consistent with a previous study that showed higher levels of triacylglycerol in patients with chronic pain would increase pain severity (Sibille et al., 2016). Interestingly, prior studies also showed that prebiotics or probiotics could be adopted in the management of lipid metabolites regulation (Barengolts, 2016). Furthermore, a recent MR study also showed a causal relationship between fecal microbiota and triglyceride (TAG) concentration, which strongly indicates that the levels of lipid metabolites, especially TAGs, are associated with gut microbiota composition (Liu et al., 2022). From the MR analysis in this study, a group of neurotransmitters, such as 5-hydroxyindoleacetic acid (5-HIAA), inositol, sphingomyelin, and cyclic adenosine monophosphate, were observed to possibly act as protective factors in reducing the risk of POU or MCP. As the primary metabolite of serotonin, a higher level of 5-HIAA was observed to increase significantly after morphine treatment and had shown to be associated with gut microbiota alteration in animal studies (Guzman et al., 2014; Yue et al., 2021). This may indicate the possible role of 5-HIAA in mediating the interaction between gut microbiota and POU. In this study, an elevated level of inositol was noticed to be potentially associated with a lower risk of MCP. This result coincided with a recent study showing that spliced inositol-requiring enzyme 1[α] (IRE1[α])-X-box-binding protein (XBP1) is a mediator in the process of prostaglandin biosynthesis and plays a role in controlling pain (Chopra et al., 2019). In conclusion, from the perspective of the effects of metabolites on POU and MCP, we believe that lipid metabolites and neurotransmitters may be more worthy of our attention, which may also provide us with another possible way to reduce POU and MCP.

This study has several strengths. Firstly, this is a two-sample bi-directional designed MR analysis to assess the causality between gut microbiota and POU as well as gut microbiota and MCP. This design can minimize the potential bias due to confounding as well as reverse causality in traditional observational studies. Secondly, summary-level data on POU and gut microbiota are the largest GWASs to date and there were no overlapping samples in datasets between exposures and outcomes, which may avoid bias caused by the winner’s curse phenomenon (Bowden and Dudbridge, 2009). Thirdly, since the GWAS study of gut microbiota, POU, and MCP was the largest to date, this study thus has enough statistical power to detect causal effects. The statistical power to identify an odds ratio of 1.2 per 1 relative abundance increment in the gut microbiota on POU with p < 0.05 was between 80 and 100%. Fourthly, evidence for causal inference was confirmed by the consistent direction and magnitude of effect estimates across three MR methods, including IVW, weighted median, and MR-Egger. Moreover, the horizontal pleiotropic test including the MR-PRESSO outlier test and MR-Egger intercept did not present any pleiotropy. Cochrane’s Q-test and leave-one-out analysis showed no Heterogeneity. The results of sensitivity analyses provide us with strong evidence of the robustness of MR estimates.

Nevertheless, some limitations should be noted when interpreting findings from this study. To start with, while we leveraged summary statistics from the largest GWAS on POU, gut microbiota, and metabolites, limited SNPs meeting genome-wide significance were identified, which may result in weak genetic instruments. To solve this problem, the p value threshold for SNPs was relaxed to (p < 5 × 10⁻⁶) for POU, and (p < 1 × 10⁻⁵) for gut microbiota and metabolites, to include additional SNPs. The F-statistics were then calculated to quantify the strength of each SNP and only SNPs with F-statistics larger than 10 were enrolled for further analysis. Secondly, the biological actions of the selected SNPs were not fully studied, making it hard to identify all potential pleiotropy. However, we ruled out the possible affecting factor of chronic intestine disorders. Furthermore, sensitivity analyses including MRPRESSO and MR Egger intercept did not detect any horizontal pleiotropy in this study. Finally, although the gut microbiota GWAS adopted in this study is the largest at present, nevertheless, its sample size is still relatively small, and the loci studied are relatively limited. So, further studies should be conducted to confirm our results when larger GWAS statistics are available.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary material.

LL, JL, JZ, and DL were responsible for the concept and design. JQ, FW, XB, and WM assisted with carrying out the analyses. LL and JL drafted the early version of the manuscript. JZ and DL jointly supervised the study. All authors contributed to the article and approved the submitted version.

This work was supported by grants from the China Postdoctoral Science Foundation (2021M700176) and the Sun Yat-sen Memorial Hospital Launch Foundation (YXQH2022017).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.