Corona Virus Disease 2019 (COVID-19) as a global pandemic continues to spread rapidly across the world causing serious concerns. The individuals with COVID-19 infection could develop fevers, coughing, difficulty in breathing, pneumonia and even death. The prognosis of COVID-19 could be greatly improved with the prevention from the development of severe symptoms. The severity of symptoms varied among the patients which may be attributed to immunity, a combination of basic diseases, The angiotensin-converting enzyme 2 expression and genetic factors (1–5). Nevertheless, the mechanism needs to be further studied.

The gastrointestinal tract is the largest immunological organ in human body and plays an essential role in immunity regulation (6, 7). The microbiome as an important immune regulator in the gastrointestinal system controls host immunity by preserving intestinal mucosa and producing immune regulatory metabolites (such as short chain fatty acids) (8). The intestinal bacteria have been suggested to be closely associated with COVID-19 infection status and severity (9). In the COVID-19 patients, the abundance of Faecalibacterium prausnitzii, Clostridium butyricum is decreased, while the abundance of Enterobacter and Enterococcus is increased (10). In addition, the abundance of Coprobacillus, Clostridium ramosum, and Clostridium hathewai in the feces is positively associated with the severity in hospitalized COVID-19 patients, while the abundance of Faecalibacterium prausnitzii is negatively correlated (11). Moreover, the severity of COVID-19 has also been reported to be associated with disturbance of various metabolic pathways which are directly or indirectly associated with the systemic inflammatory response observed in patients with severe COVID-19 (12–15). And the metabolites could accurately predict the course of COVID-19, such as tryptophan, kynurenine and 3-hydroxykynurenine, the metabolites of kynurenine pathway (16). In a recent study on immune metabolism, proinflammatory cytokines and chemokines were found to be closely related to metabolites originating in tricarboxylic acid cycle, amino acid metabolism, purine and pyrimidine metabolism and primary bile acid metabolism in severe COVID-19 patients (17). However, most of previous studies are correlational studies, and they are unable to determine whether the selected intestinal bacteria or metabolites has a potential causal relationship with COVID-19 infection and severity. Searching for the causal intestinal bacteria and metabolites of severe COVID-19 could benefit the mechanistic investigation of severe COVID-19 and help the clinical decision before the onset of serious illness.

Causal analysis acts a statistical method allowing the analysis of causal relationship between exposure factors and outcomes. Mendelian randomization (MR) analysis is currently the most commonly used causal inference method in the clinical research frequently using genome-wide association studies (GWAS) summary statistics data. It incorporates germline variants and infers causality based on the principle of random allocation of alleles transferred from parent to offspring at the time of gamete formation, which is less likely to be affected by confounding factors (18). In short, two types of defined phenotypes are studied in MR analysis, one is the exposure factors (candidate causal factors), such as the abundance of intestinal bacteria and another is outcome (affected factors), such as severe COVID-19. Briefly, MR analysis identifies SNPs that are significantly associated with exposure factors at a given p-value threshold (p<1×10^-5, etc.). Linkage disequilibrium analysis is then applied to select the representative SNPs which can present other SNPs in a high linkage disequilibrium region. These selected SNPs are used as the final instrumental variables (IVs). Based on the IVs, the outcome population is stratified into two groups: the exposure group consisting of individuals who carry the risk alleles of exposures, and the control group consisting of individuals who do not carry the risk alleles. Finally, the differences of outcomes between the two groups are evaluated to infer the causal effects of exposures on outcomes. This causal inference process achieves an effect like that of Randomized Controlled Trials (RCT) and provides a higher level of clinical evidence than correlation analysis (19).

Here, we applied MR analysis to examine the relationship between the gut microbiome, plasma metabolome and severe COVID-19. We found 13 taxa and 53 metabolites act as the causing factors associating with the severity of COVID-19. Moreover, 11 mediated relations among them had been identified. This is the first study to systematically assess the causal relationship between intestinal microbiome, plasma metabolome and the development of severe COVID-19.

In this study, the causal relationships of gut microbiome and plasma metabolome for the severity of COVID-19 were investigated using MR analysis. We found that 13 taxa (Butyrivibrio, Howardella, Oxalobacter, Ruminiclostridium 6, Ruminococcus torques, etc.) and 53 metabolites (2-stearoylglycerophosphocholine, alpha-glutamyltyrosine, carnitine, myo-inositol, etc.) had a potential causal relationship with severe COVID-19. Pathway analysis of the 53 identified metabolites suggested that they were significantly enriched in pathways of ascorbate and aldarate metabolism, beta oxidation of very long chain fatty acids and oxidation of branched chain fatty acids. Mediation analysis among the identified exposures found that the associations of Howardella, Ruminiclostridium 6, myo-inositol and N-acetylalanine with severe COVID-19 were likely to be mediated by one or more exposures. After multiple testing correction of cis-MR results, only Ruminococcus torques had a trend of positive causal relationship with the severe COVID-19. The increased abundance of Ruminococcus torques could be a contributor to the incidence of severe respiratory symptoms in COVID-19 patients. The results of the colocalization analysis revealed that the abundance of Ruminococcus torques and the expression of RASIP1 in colon tissue shared a causal factor and had a high colocalization probability with the occurrence of severe respiratory symptoms, implying that they both played important roles in the development of severe COVID-19.

Myo-inositol has been reported to downregulate the expression of IL-6 levels inhibiting the downstream inflammatory response (29). Furthermore, myo-inositol, as precursor of inositol-phosphate, stimulates surfactant production in lung tissue, and thus could represent a potential preventive strategy for COVID-19 (29, 30). Consistent with this, we provided causal evidence for directionally consistent effects of myo-inositol on severe COVID-19. Bi-directional Mediation analysis results indicated that myo-inositol mediated 13.7% effect of Howardella on severe COVID-19, while the mediation effect of myo-inositol via Butyrivibrio was 12% for severe COVID-19.

A recent study indicates that gut microbiome of patients with post-acute COVID-19 syndrome are characterized by higher levels of Ruminococcus gnavus ( 31), which has been shown to promote inflammatory responses and impair barrier functions by producing inflammatory polysaccharides (32). We also showed that Ruminococcus gnavus causally increases the risk of severe COVID-19 using univariate MR analysis. Furthermore, Mediation analysis results revealed that Ruminococcus gnavus mediates more than one third effect (36.8%) of N-acetylalanine on severe COVID-19.

Notedly, Ruminiclostridium 6 was previously found to have a strong positive correlation with the levels of ghrelin (33), which exerts immunomodulatory functions in COVID-19 infection, such as the suppressive effects on pro-inflammatory cytokine production including IL-1 β, IL-6 and TNF- α (34). Therefore, it is conceivable that the causal effect of Ruminiclostridium 6 on severe COVID-19 may result from ghrelin.

Ruminococcus torques, also known as Mediterraneibacter torques, is an anaerobic and gram-positive intestinal bacteria which belongs to the genus Mediterraneibacter in the family Lachnospiraceae. According to earlier research, Ruminococcus torques was positively associated with intestinal paracellular permeability and gastrointestinal disorders (35, 36). Increased intestinal permeability could cause endotoxemia and activate the inflammatory response, which ultimately raises the risk of various diseases including severe illness in COVID-19 patients (37, 38). Additionally, an increase in the abundance of Ruminococcus torques is associated with constipation and diarrhoea in children with autism, and the presence of gastrointestinal symptoms has been demonstrated to be an independent risk factor for severe COVID-19 (39, 40). Therefore, Ruminococcus torques could have a potential role in the development of severe respiratory symptoms in COVID-19 patients.

Ras interacting protein 1 (RASIP1) is a vascular-specific GTPase signaling regulator involved in a variety of functions, including the Rho signal transmission pathway. RASIP1 regulates the stability of vascular endothelial connections, which is relevant to vascular barrier function, and mediates the regulation of Rho in intrinsic barrier function through Rap1 (41, 42). RASIP1 depletion reduces the barrier function of vascular endothelial cells induced by Rap1 (43). The disruption of endothelium barrier can result in chronic inflammation, atherosclerosis and vascular leakage, as well as the development and progress of COVID-19 (44). Interestingly, Ruminococcus torques and RASIP1 were both associated with cell permeability, and in our investigation, they seemed to have a very strong colocalization. Previous database analysis results indicated that rs35866622 decreased the abundance of Ruminococcus torques while increased the expression of RASIP1indicating a negative association relationship. As a result, increased abundance of Ruminococcus torques coupled with the decreased RASIP1 expression were associated with the disruption of cell barrier and increased permeability, thereby ultimately increase the risk of COVID-19 worsening which was consistent with our MR results.

From the 53 metabolites found to increase the risk of severe COVID-19, we pinpointed the key pathways including ascorbate and aldarate metabolism, beta oxidation of very long chain fatty acids and oxidation of branched chain fatty acids. In these signals, vitamins (ascorbate and aldarate metabolism) have been reported responding to the risk of COVID-19 and its severity. Vitamin C is a potential antiviral agent and may improve immunity. Supplementation with high-dose vitamin C could increase the survival rates of patients with severe COVID-19 by decreasing inflammation and pathogen infectivity and viral yield, improving immune response, alleviating tissue and organ damage (45). Numerous evidences confirm vitamin D insufficiency is associated with greater severity of COVID-19 infection, even the more recent Omicron subvariant of COVID-19 (46–48). Vitamin D administration has been found to be associated with less severe COVID-19 and resulted in a decreased risk of death and admission to intensive care units in patients with COVID-19 (49). In addition, fatty acid metabolism is a crucial event for many viruses to complete their life cycle, and a common consequence of infection by many viruses is to change the nature of lipid metabolism usually from fatty acid oxidation to fatty acid synthesis (50). Fatty acid oxidation is the most powerful pathway to generate energy, and significant impairment in fatty acid oxidation has been reported in patients with post-acute COVID-19 syndrome (51, 52). Our results indicated that the severe COVID-19 causal associated metabolites were significantly enriched in pathways of beta oxidation of very long chain fatty acids and oxidation of branched chain fatty acids. Therefore, fatty acid metabolism offers another promising target to control the COVID-19 infection extent.

It is also worth noting that 5 metabolites (2-tetradecenoyl carnitine, carnitine, cis-4-decenoyl carnitine, decanoylcarnitine and octanoylcarnitine) in the carnitine metabolism pathway were identified to be causal associated with severe COVID-19. Consistent with our findings, a UPLC-MS/MS-based widely targeted metabolomics study also revealed several carnitine family members are significantly reduced in severe COVID-19 patients versus healthy controls subjects and mild COVID-19 patients (53). Carnitine metabolism balance plays an important role in maintaining normal physiological functions through its anti-inflammatory, antioxidative, anti-apoptotic, anti-fibrosis and biomembrane-stabilizing properties (54). Carnitine deficiency occurs in multiple diseases such as sepsis, advanced liver cirrhosis and endocrine disorders (54). Severe COVID-19 patients usually exhibit metabolic disorders and multiple organ dysfunctions, the downregulated carnitine in the severe patients may contribute to impaired organ function. Additionally, alanine, as another important metabolites for COVID-19 severity, was revealed by the 4 identified metabolites (aspartylphenylalanine, leucylalanine (X-14189), leucylalanine (X-14304) and N-acetylalanine). A key physiological function of alanine is to transport pyruvate and glutamate from the muscles to the liver, a process known as the glucose–alanine cycle. Data from patients with different severity grades of COVID-19 show that circulating pyruvate level was the strongest determinants of severe COVID-19 (55), and a meta-analysis indicated that elevated glutamate was associated with an increased risk of COVID-19 severity (56).

In this study, we set 131 taxa and 452 metabolites as exposure factors, severe COVID-19 as the outcome variable, and then grouped the COVID-19 cohort according to the IVs selected for each exposure factor. Then, by analyzing the differences in the risk of severe COVID-19 between the exposed and control groups, we could infer the causal relationship between exposure factors and severe COVID-19. Our study showed that MR analysis could sort out some intestinal bacteria and metabolites that have a potential relationship with COVID-19 severity. In addition, by combining two-stage MR analysis and mediation analysis, we successfully linked gut microbiome and plasma metabolome, and constructed some pathways from intestinal bacteria to severe COVID-19 through plasma metabolites or from plasma metabolites to severe COVID-19 through intestinal bacteria. Our study provides potential biomarkers associated with severe COVID-19 and can benefit the mechanistic investigation of severe COVID-19.

This study has some limitations. Firstly, our casual inference results of intestinal bacteria might be affected by confounding factors such as race, diet, and disease status. Additionally, no experiments have been conducted to validate the causal associations of this study. Therefore, mechanistic studies will be necessary in the future to verify these relationships.

In conclusion, our comprehensive MR analyses identified 13 human intestinal taxa and 53 human serum metabolites potential causal associated with COVID-19 severity. We also found 11 mediated relations among the identified intestinal taxa and serum metabolites. These causal taxa and metabolites potentially served as clinical biomarkers for risk stratification and prognostication of severe COVID-19 and would benefit the mechanism mechanistic investigation of severe COVID-19.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The study was approved by the ethics committee of Institute of Clinical Pharmacology, Central South University.

XL, LZ and J-QL participated in research design. HY, LZ, J-QL, SZ and WZ participated in the writing of the paper. H-XH, PX, X-HC and Y-DQ participated in the performance of the research. XL and HY prepared the figures. All authors contributed to the article and approved the submitted version.