Acceso abierto

A Mendelian Randomization Study: Roles of Gut Microbiota in Sepsis – Who is the Angle?


Cite

Introduction

Sepsis refers to a lethal organ dysfunction that is attributed to the dysregulated responses of the host to infections, serving as a leading reason for morbidity and mortality worldwide, so the World Health Organization has recognized it as a global health emergency. The pathogenesis of sepsis remains unclear to date (Anggraini et al. 2022), including imbalance of inflammatory response, immune dysfunction, coagulation dysfunction, and other aspects. The process of its onset is a gradual sequential reaction, and further development may lead to the development of multiple organ dysfunction syndrome. The gastrointestinal tract is the most unique among them. On the one hand, the gastrointestinal tract is the organ first affected by sepsis, and it is affected by both endogenous and exogenous factors, leading to functional disorders. On the other hand, the gastrointestinal tract can act as an intermediate hub, causing damage to other distant organs and accelerating the progression of sepsis (Longhitano et al. 2020).

The microbiota distributed in the gastrointestinal tract is commonly referred to as the gut microbiota (GM), which exerts an essential effect on the maintenance of physiological functions such as metabolism, immunity, and endocrine in the human body (Klingensmith and Coopersmith 2023). There is increasing evidence that the imbalance of GM leads to a higher risk of sepsis and worse outcomes (Adelman et al. 2020; Kang and Thomas 2021; Kullberg et al. 2021; Miller et al. 2021; Niu and Chen 2021). Recent evidence has demonstrated that gut microbiota regulates host physiological homeostasis mediators. Therefore, maintaining or restoring microbiota and metabolite composition might be a therapeutic or prophylactic target against critical illness (Niu and Chen 2021). Fecal microbiota transplantation and supplementation of probiotics are microbiota-based treatment methods that are limited in evidence-based efficacy. The detailed roles of these microorganisms in the sepsis pathophysiology need further investigation.

The causality involving outcomes together with exposure was probed into through a Mendelian randomization (MR) study with instrumental variables (IVs) selected from single-nucleotide polymorphisms (SNPs). Compared to randomized controlled trials, MR analysis can more quickly and more economically determine the cause and effect involving exposure factors plus outcomes (Skrivankova et al. 2021). The correlation of GM with sepsis has not been investigated using MR analysis so far. Therefore, the causal relationship between GM and sepsis, as well as the concrete functions of diverse GM taxa in sepsis, deserves in-depth investigation. Therefore, this MR study was performed by virtue of massive summary statistics from genome-wide association studies (GWASs) concerning GMs and sepsis to distinguish GM taxa with possible influences on sepsis to give support to current evidence while providing novel opinions for the treatment plus prevention of sepsis.

Experimental
Materials and Methods
Study design

A two-sample MR study was carried out to exploit the association of GM with sepsis. Fig. 1 illustrates the whole design of our research.

Fig. 1.

The whole workflow of Mendelian randomization (MR) analysis.

Data sources

The MiBioGen study provided GM-related GWAS data (Kurilshikov et al. 2021), a multiethnic genome-wide meta-analysis on GM with most enormous scale. In this study, the sequencing profiles for 16S rRNA gene targeting the variable regions such as V4, V3–V4, and V1–V2, together with the data of whole-genome genotyping collected from 18,473 individuals (with a predominant proportion of European ancestry) in 25 cohorts, were utilized. Finally, 122,110 host genetic variants were obtained from GWASs, which were mapped to the abundance-associated genetic loci in 211 taxa (16 classes, 131 genera, 20 orders, 9 phyla, and 35 families) through the analysis of variations in GM taxa among diverse populations and GWAS analysis by taxa step by step.

MR-Base database (https://www.mrbase.org) (Hemani et al. 2018), which has incorporated the results of GWASs involving 605 traits in the UK Biobank so far, was utilized to extract the GWAS summary statistics for sepsis. It comprises 11 billion pieces of SNP-trait associations uncovered by 1673 GWASs, with regular updates. The cohorts in UK Biobank (https://www.ukbiobank.ac.uk) enrolled nearly 500,000 adults from the UK in 2006–2010. The participants were permitted to obtain extensive data on phenotypes and genotypes. According to different quality control standards, a total of three GWAS summary statistics of sepsis from MR-Base were obtained, including 2 GWAS that Hospital Episode Statistic coded only, performed using regenie v2.2.4 (Mbatchou et al. 2021) and 1 GWAS performed using MRC IEU UK Biobank pipeline. The ids are ieu-b-4980 with 11643 cases and 474841controls, ieu-b-5088 with 1896 cases and 484588 controls, ieu-b-69 with 10154 cases and 454764 controls, respectively. The hospital admission data were coded by ICD and linked to the UK Biobank to identify the sepsis cases, including ICD-10 codes A02, A40, A39, and A41. The whole participants are of European descent.

The other two from the FinnGen database (https://finngen.gitbook.io/documentation) (Kurki et al. 2023) were employed to offer the remaining summary statistics of GWAS. FinnGen is a large-scale genomic study in Finland aiming to create a comprehensive genetic and health data database from up to 500,000 participants. The project aims to identify new genetic targets for preventing and treating common diseases involving cancer, diabetes, and cardiovascular disease. This database is open to researchers worldwide and has already yielded important discoveries. The identifiers for their entries in MR-Base are finn-b-O15_PUERP_SEPSIS and finn-b-ASTHMA_PNEUMONIA_AND_SEPSIS, respectively (Table I).

Details of data sources.

Trait Population Sample_size Case Control
ieu-b-5088 Sepsis (under 75) European 462,869 11,568 451,301
ieu-b-4980 Sepsis European 486,484 11,643 474,841
ieu-b-69 Sepsis European 462,918 10,154 454,764
finn-b-O15_PUERP_SEPSIS Puerperal sepsis European 121,441 2,286 119,155
finn-b-ASTHMA_PNEUMONIA_AND_SEPSIS Asthma-related pneumonia or sepsis European 140,994 5,545 135,449
Data processing

A procedure integrating distance < 500 kb, European ancestry, and LD r2 < 0.3 was applied to determine the GM-associated lead SNPs at the genome-wide significance level (p < 5.0 × 10−8). Then, the same SNPs were picked up from GWAS summary statistics of sepsis to conduct subsequent MR analysis.

Two-sample MR

R-based “TwoSampleMR” package (version v0.5.6) (Hemani et al. 2018) was employed to accomplish the two-sample MR analysis, with five models in total, namely, MR-Egger, Wald ratio, weighted median, inverse-variance-weighted (IVW), and weighted mode established. The lead SNPs, after clumping, were employed as IVs. For the purpose of guaranteeing the robustness of data and the precision of results, the IVs for the models in this study were selected based on three major principles: 1) IVs had relationships to GM taxa (p < 5.0 × 10−8), 2) IVs were only associated with sepsis through their relation to risk factors, 3) there was no correlation between IVs and confounders. At last, SNPs were excluded in the case of linkage disequilibrium (clump = 500 kb, r2 = 0.3).

Our study used the IVW approach as a primary analytical method when there were two or more instrumental variables, and the Wald ratio model was utilized for data analysis with merely one IV existing. If the IVW model produced a significant result, the GM taxa would be assumed to have a causal link with sepsis. The other three models were also utilized as references, and if the favorable outcomes were reproduced, we considered the results more reliable. Cochran Q test plus MR-Egger regression were employed to evaluate the heterogeneity and pleiotropic effect, respectively, to avoid violating MR assumptions. Causal effects were considered credible in the case of p > 0.05 for MR-Egger intercept, together with p > 0.05 for Cochran’s Q test. A randomeffect IVW model was employed when significant heterogeneity (p < 0.05) emerged. Leave-one-out analysis was utilized to amine the promotion of individual SNP to the causal estimate, which was repeated by sequentially discarding the SNPs related to each exposure. In our study, the causality involving GM and sepsis was appraised by the combination of standard errors (SE), odds ratio (OR), β, and confidence interval (CI).

Statistical analysis

R 4.0.2 software (R Core Team 2020) was used for statistical analysis. F-statistic was evaluated to determine the presence or absence of weak IV bias, where the IVs based on genetic variations displayed vague relationships to exposure. The following formula presented below was used to calculate the F-statistic: F=R2(N2)1R2 $$F = {{{R^2}(N - 2)} \over {1 - {R^2}}}$$ where R2 stands for the phenotype variance interpreted using inverse variances and obtained from PRSice calculation (Choi et al. 2019). N and K indicate the sample size and number of IVs, respectively.

Given the number of IVs remaining below 10, R2 equals to: R2=2×EAF×(1EAF)×β2 $${R^2} = 2 \times EAF \times (1 - EAF) \times {\beta ^2}$$

In other cases, the formula can be changed to: R2=2×EAF×(1EAF)×β22×EAF×(1EAF)×N×SE2+2×EAF×(1EAF)×β2 $${R^2} = {{2 \times EAF \times (1 - EAF) \times {\beta ^2}} \over {2 \times EAF \times (1 - EAF) \times N \times S{E^2} + 2 \times EAF \times (1 - EAF) \times {\beta ^2}}}$$ where EAF, β, and SE refer to the effect allele frequency of IV and the impact degree and standard error of IV on exposure, respectively (Papadimitriou et al. 2020).

Results
Selection of IVs

In subsequent investigations, the IVs for exposure were determined as the SNPs at genomewide statistical significance threshold (p < 5.0 × 10−8) after clumping. In the present study, F-statistics for each IV far exceeding 10 indicated an extremely low possibility of weak bias on IVs. Detailed information of instrument variables was listed in Table SI.

Two-Sample MR Analysis

Table II exhibited the MR results of various causality evaluation methods for GM and sepsis. The results of IVW analysis demonstrated that Actinobacteria and Bifidobacteriaceae influence sepsis at different taxa levels. When using outcome coded ieu-b-69 (N = 10,154 cases and 454,764 controls), Actinobacteria at the phylum (β = –0.34, SE = 0.10, p = 0.0008) and class (β = –0.23, SE = 0.07, p = 0.0011) levels displayed negative relationships to the sepsis risk. When using outcome coded ieu-b-4980 (N = 11,643 cases and 474,841 controls), Actinobacteria at the phylum level (β = –0.22, SE = 0.10, p = 0.027) was also negatively associated with sepsis. In Bifidobacteriaceae, comparable outcomes can be observed. When using outcome coded ieu-b-69 (N = 10,154 cases and 454,764 controls), Sepsis risk and Bifidobacteriaceae at the order (β = –0.20, SE = 0.06, p = 0.0021), family (β = –0.20, SE = 0.06, p = 0.0021), and genus (β = –0.20, SE = 0.06, p = 0.0007) levels were all inversely affiliated. The results of the Wald ratio model showed that Tyzzerella genu [OR (95%CI) = 0.6902[0.4907,0.9708], p = 0.0331] and Gastranaerophilales order [OR (95%CI) = 0.5907[0.3516,0.9926], p = 0.0468] were negatively connected with sepsis. Fig. 2 and Fig. S1 illustrated each exposure’s predicted degree of impact on outcomes. There was no horizontal pleiotropy heterogeneity in addition to horizontal pleiotropy observed from IVs and outcomes. Furthermore, all F-statistics above 10 verified a low probability of a weak IV bias. The detailed results are presented in Table SII.

Fig. 2.

Forest plot of estimated effect sizes at genome-wide significance level (p < 5.0 × 10−8) each exposure on outcomes.

CI – Confidence interval, F – F statistics, MR – Mendelian randomization, MR method – the type of model applied in MR analysis, Nsnp – number of SNPs, OR – odds ratio, P – p-value of causal estimation in different MR methods, Ple_pval – pleiotropy, Q_pval _ heterogeneity, R2 _ phenotype variance explained by IVs

Mendelian randomization (MR) analysis of gut microbiota on sepsis at genome-wide significance level (p < 5.0 × 10−8).

Classification Bacterial taxa Outcome No. SNP MR method Beta SE OR 95% CI P F Horizontal pleiotropy Heterogeneity
Actinobacteria phylum ieu-b-4980 4 IVW −0.222 0.1 0.801 0.658–0.975 0.027 190.05 0.47 0.76
4 Weighted median −0.245 0.116 0.783 0.623–0.983 0.035
ieu-b-5088 4 IVW −0.23 0.099 0.795 0.655–0.964 0.02 190.04 0.6 0.94
4 Weighted median −0.25 0.109 0.779 0.629–0.965 0.022
ieu-b-69 4 IVW −0.338 0.101 0.713 0.585–0.869 0.001 190.04 0.55 0.74
4 Weighted median −0.367 0.118 0.693 0.550–0.873 0.002
ieu-b-69 6 IVW −0.231 0.071 0.794 0.691–0.912 0.001 352.59 0.44 0.02
6 Weighted median −0.241 0.089 0.786 0.660–0.936 0.007
Bifidobacteriaceae order ieu-b-4981 7 IVW 1.051 0.368 2.861 1.391–5.885 0.004 390.71 0.18 0.3
7 Weighted median 1.024 0.49 2.783 1.065–7.276 0.037
ieu-b-69 7 IVW −0.199 0.065 0.819 0.722–0.930 0.002 388.52 0.55 0.74
7 Weighted median −0.208 0.081 0.812 0.692–0.952 0.01
ieu-b-69 7 IVW −0.231 0.071 0.819 0.722–0.930 0.001 388.52 0.55 0.74
7 Weighted median −0.208 0.082 0.812 0.691–0.954 0.012
ieu-b-69 9 IVW −0.204 0.065 0.815 0.724–0.912 0.002 441.31 0.83 0.9
9 Weighted median −0.226 0.079 0.798 0.683–0.931 0.004
Tyzzerella genus finn-b-ASTHMA_PNEUMONIA_AND_SEPSIS 1 Wald ratio −0.371 0.174 0.69 0.490–0.970 0.033
Gastranaerophilales order finn-b-015_PUERP_SEPSIS 1 Wald ratio −0.526 0.265 0.59 0.352–0.992 0.047

Beta – estimated causal effect coefficient in different MR methods, CI – confidence interval, F-statistics, heterogeneity – p-value of heterogeneity analysis, horizontal pleiotropy – p-value of horizontal pleiotropy analysis, MR method – the type of model applied in MR analysis, No. SNP – number of SNPs, OR – odds ratio, p – p-value of causal estimation in different MR methods, SE – standard error of coefficient estimate

Discussions

The causal correlation of sepsis with GM was evaluated through two-sample MR analyses in this study, the first research using MR analyses to assess this relationship. We found that at different taxa levels, Actinobacteria and Bifidobacteriaceae are causally associated with low risk of sepsis, indicating their protective effect on sepsis. Other species of GM indicated potential protective effects against sepsis, including Tyzzerella genus and Gastranaerophilales order.

Sepsis substantially threatens the public health due to its high morbidity and mortality. Unfortunately, over the past decades, antibiotics and supportive care represent the only established targeted therapy of sepsis (Evans et al. 2021). As shotgun metagenomic sequencing, 16S rRNA, and other culture-independent methods are emerging and developing (Durazzi et al. 2021), increasing evidence now suggests that the GM exerts crucial effects on sepsis from the aspect of the pathophysiological process (Chancharoenthana et al. 2023). Sepsis can significantly influence the composition, diversity, and balance of normal GM due to the pathophysiological changes such as gut hypoperfusion and impaired mucosal integrity and the clinical interventions, including antibiotics for treating sepsis. At the same time, its development may be facilitated by these factors in turn, forming a vicious cycle (Miller et al. 2021). GM alteration increases sepsis susceptibility before sepsis onset via multiple mechanisms, such as stimulating the pathogenic intestinal bacteria to expand, initiating strong proinflammatory responses of the immune system, and inhibiting beneficial microbial products (short-chain fatty acids and so on) from generation. Once sepsis is on set, GM disruption is exacerbated while the susceptibility to end-organ dysfunction is enhanced. According to finite evidence, the reduction of sepsis risk plus the improvement of sepsis outcomes may be attributed to selective decontamination of the digestive tract, prebiotics, fecal microbial transplantation, probiotics, and other microbiome-based therapies, while the safety of such therapies has been rarely considered (Adelman et al. 2020). Research on specific microbiome changes during sepsis and the mechanism of action on sepsis have therefore been identified as a critical problem to explore.

Belonging to the Actinobacteria phylum, Bifidobacteriales order, and Bifidobacteriaceae family (Mancino et al. 2019), the Bifidobacterium genus was first discovered by Tissier in 1899 and then separated from feces of a breastfeeding infant (Ventura et al. 2014). As an essential bacterial genus composing the intestinal flora in human beings, Bifidobacterium has such physiological functions as anti-tumor activity, promotion of nutrient absorption, immune regulation, and maintenance of intestinal micro-ecological balance. Bifidobacterium is generally reduced in critical ill patients and related to prognosis. Wei et al. (2021) found that the reduced abundance of Bifidobacterium had preferable effects on predicting in-hospital mortality of critically ill patients. Consistent with previous findings, recently a study to assess the association between GM composition and Day-28 mortality in critically ill patients indicated that survivors had a higher abundance of Bifidobacterium (Prevel et al. 2022). There are similar changes in sepsis. Peng et al. (2023) found Bifidobacterium reduced in different intestinal segments of mice with sepsis by cecum ligation and puncture modeling. Sun et al. (2023) found that patients with sepsis showed a decreased abundance of Bifidobacterium, too. Previous randomized controlled trials and metaanalyses examined the impact of probiotics on sepsis, most containing Lactobacillus and Bifidobacterium species, which yielded complex results. As corroborated by a meta-analysis on critically ill patients, probiotics were associated with alleviated ventilator-associated pneumonia plus infections. There was heterogeneity of inclusion criteria and probiotic species selected for the included trials, in which merely one subset was established explicitly for sepsis patients (Manzanares et al. 2016). The clinical conditions of preterm neonates with low birth weight and the roles of oral supplementation of probiotics in influencing Lactobacillus and Bifidobacterium populations in the gut were examined through a randomized, multicenter, double-blind, placebo-controlled trial. They found that the probiotic bacteria selected and appropriately characterized are safe for preterm neonates to normalize the disturbed GM and conducive to lowering the incidence rate of staphylococcal sepsis (Robertson et al. 2020). Whether GM disruption before the onset of sepsis aggravates the prognosis of humans with sepsis has not been clarified. Our study suggests the protective effect of Bifidobacterium on sepsis through MR analysis, indirectly confirming the results of other studies and providing reliable evidence for future clinical randomized controlled studies.

We found that the Gastranaerophilales order was negatively connected with sepsis, but the mechanism is unclear. The Gastranaerophilales order lacks the genes essential for aerobic respiration so that it can grow in human and animal guts with low or nearly no oxygen. It can be classified as non-photosynthetic Cyano-bacteria (Melainabacteria class) (Hu and Rzymski 2022). The proportion of Cyanobacteria phylum in the GM is less than 1%. Di Rienzi et al. (2013) utilized the human gut and groundwater to assemble the whole genomes of non-photosynthetic cyanobacterium-like organisms for the first time in 2013. Even though the correlations of diverse diseases with the composition of human GM have been explored in numerous studies, the association between health and intestinal Cyanobacteria and Gastranaerophilales in humans should be researched in detail. In the past years, a couple of researches indicated that gut Cyanobacterial abundance increased in many diseases, such as amyotrophic lateral sclerosis (Di Gioia et al. 2020), neurodevelopment disorder (Zhang et al. 2021), acute gastroenteritis (Xiong et al. 2021), Wilson’s Disease (Cai et al. 2020) and allergy rhinitis (Zhu et al. 2020). However, there is no observational study on the correlation between sepsis and Gastranaerophilales. We infer that the mechanism by which Gastranaerophilales participates in sepsis may be related to the metabolism of tryptophan, which belongs to the crucial amino acids supplied to human bodies by diets, containing the structure of an indole and metabolized by GM to produce indole and indole derivatives. Rosario et al. (2021) found Gastranaerophilales can contribute to increased concentrations of indole, which could be converted into indole-3-propionic acid (IPA). It is argued that IPA conjugates with the aromatic hydrocarbon receptor of intestinal epithelium cells or pregnane X receptor of lymphocytes to activate the immune system, thereby stimulating the intestinal hormone secretion along with gastrointestinal motility, enhancing the intestinal epithelial barrier, putatively modulating GM composition, and exerting anti-inflammatory, anti-oxidative or toxic effects on systemic circulation (Zhang et al. 2022). As Fang et al. (2022) uncovered, IPA ameliorates sepsis-induced mortality and decreases the serum levels of proinflammatory cytokines by modulating intestinal microbiota, which could provide a new potential therapeutic approach for sepsis. Consistent with the abovementioned research, Huang et al. (2022) also found that the IPA level in the feces of septic patients is significantly lower than in controls, and a lower IPA level was about poorer clinical outcomes. It was also discovered from animal models that there was a relationship between the survival of mice with sepsis and GM-derived IPA, and the treatment with IPA was able to prevent sepsis-related mortality while mitigating the organ injury plus bacterial burden resulting from sepsis.

In addition, we identified Tyzzerella on the genus level was negatively connected to sepsis. Tyzzerella genus is a member of Lachnospiraceae family, which is a potential research highlight. Yutin and Galperin (2013) first named Tyzzerella genus by 16S rRNA and ribosomal protein sequences in 2013. Nonetheless, the detailed functions of Tyzzerella have not been elaborated yet. Despite the minimal number of research describing Tyzzerella genus, patients at a high risk of cardiovascular diseases manifest an increased level of this genus, which has correlations with a high risk of lifetime cardiovascular diseases as well as a raised cardiovascular disease risk (Murphy et al. 2021). In addition, Tyzzerella genus is related to inflammatory bowel disease. Tyzzerella genus has been reported to be increased in ulcerative colitis patients from China (Zhu et al. 2022).

Compared with that in patients without Crohn’s disease, the abundance of Tyzzerella was remarkably increased in patients who have Crohn’s disease (Olaisen et al. 2021). Liu et al. (2021) observed that Tyzzerella served as a critical species with a notably elevated abundance in SARS-CoV-2 infection, and its AUCs for distinguishing patients with asymptomatic or symptomatic SARS-CoV-2 infection from healthy controls were 70%, and 75%, suggesting that as a bacterial taxonomic biomarker, Tyzzerella has potential diagnostic value for SARS-CoV-2 infection.

Moreover, one study showed that a higher intake of saturated fatty acids and trans-fatty acids leads to a prominently raised relative abundance of Tyzzerella, which was associated with fatty acid intake (Xu et al. 2022). Pitifully, the exact potential mechanism of Tyzzerella in mediating the adverse effects of these diseases is unclear. So far, no research has explored the relationship and mechanism of Tyzzerella genus and sepsis, but some studies have elucidated the relationship between Lachnospiraceae family and sepsis. Sun et al. (2023) examined the possible function of gut micro-biome and metabolites in affecting sepsis severity. They indicated that Lachnospiraceae were found to have promising benefits for sepsis progression. There was an intimate correlation between Lachnospiraceae species and other sepsis-related metabolites. Overall, further research on the role and mechanism of Tyzzerella genus in sepsis is needed.

In general, our study had several limitations. First, while most participants in the GWASs providing summary data for this study had European ancestry, a minority of GM data came from collections acquired through other ethnic groups, possibly leading to racial bias. Secondly, merely the genus level, instead of more specific levels like strain and species, was selected to explore bacterial taxa. More detailed and accurate results may be obtained by analyzing microbiome GWAS via more advanced shotgun metagenomic sequencing. Thirdly, genus was set as the lowest taxonomic level for the data about exposure, and suc a limitation prohibited further investigation into the causality between GM and sepsis from the aspect of species.

Conclusions

To sum up, the results of this study provide evidence for underlying causal protective effects of different taxa levels Actinobacteria and Bifidobacteriaceae, Tyzzerella genus, and Gastranaerophilales order on sepsis, which provide novel insights into GM functions on sepsis in terms of pathogenesis and treatment.

eISSN:
2544-4646
Idioma:
Inglés
Calendario de la edición:
4 veces al año
Temas de la revista:
Life Sciences, Microbiology and Virology