Genetic Associations of Inflammatory Cytokines and Gut Microbiota-derived Metabolites with Sepsis and Acute Kidney Injury: A Bidirectional and Two-step Mendelian Randomization Study
Saurie Hernandez
Department of Internal Medicine, Omoa Private Clinic, Omoa, Honduras.
Indri Yani Septiana
Department of Physiology, CMHC Research Center, Palembang, Indonesia.
Ayesh Mahmood
Department of Palliative Therapy, Nukus Private Hospital, Nukus, Uzbekistan.
Jason Willmare
Department of Neurology, Diffa State Hospital, Diffa, Niger.
Akmal Hasan *
Department of Neurology, Phlox Institute, Indonesian Medical Research Organization, Palembang, Indonesia.
Despian Januandri
Department of Neurology, Phlox Institute, Indonesian Medical Research Organization, Palembang, Indonesia.
Novalika Kurnia
Department of Pediatrics, Phlox Institute, Indonesian Medical Research Organization, Palembang, Indonesia.
Leonardo Simanjuntak
Department of Obstetrics and Gynecology, CMHC Research Center, Palembang, Indonesia.
Maximillian Wilson
Department of Internal Medicine, Mananjary State Hospital, Mananjary, Madagascar.
Rifki Sakinah Nompo
Department of Nursing, Sekolah Tinggi Ilmu Kesehatan Papua, Sorong, Indonesia.
Vidhya Sathyakirti
Department of Radiology, Queensland State Hospital, Queensland, Australia.
*Author to whom correspondence should be addressed.
Abstract
Background: Whether elevated cytokines and gut metabolites in sepsis and acute kidney injury (AKI) represent upstream contributors or downstream correlates of impaired renal clearance remains unresolved.
Aims: To evaluate bidirectional and two-step mediated genetic associations of 21 cytokines and 21 gut metabolites with sepsis and AKI.
Study Design: Two-sample bidirectional and exploratory two-step mediation Mendelian randomisation (MR) study (STROBE-MR).
Place and Duration of Study: Summary-level analysis of SCALLOP (2023), YFS/FINRISK (2017), CLSA (2023), UK Biobank–FinnGen (2021), and FinnGen R12 (2024; analysed through October 2026).
Methodology: GRCh38 instruments (per 1-SD biomarker or log-odds liability; F=20.4–5,240.6) from SCALLOP (N=14,824), YFS/FINRISK (N≤8,293), and CLSA (N=8,299) were tested against overlapping FinnGen R12 AKI (8,383/480,448) and sepsis (17,133/439,048; polygenic P<5×10-6, mean F=23.5) and UKB–FinnGen AKI (7,695/474,571), with MRlap, UKB sepsis, and eGFR-adjusted multivariable MR.
Results: Only sepsis-to-AKI survived FDR correction (FinnGen polygenic OR=1.281, 95% CI 1.188–1.380, FDR=2.02×10-10; strict k=3 OR=1.298, 95% CI 0.963–1.751; UKB–FinnGen OR=1.188, 95% CI 1.044–1.352; non-overlapping UKB sepsis OR=1.214, 95% CI 1.068–1.380), exceeding reverse AKI-to-sepsis (Weighted Median OR=1.027). No biomarker survived FDR (BH-FDR=0.162–0.538) or replicated across both AKI datasets. Nominal AKI signals included IL-12B (OR=1.065, 95% CI 1.006–1.127), taurocholenate sulfate (OR=1.089, 95% CI 1.013–1.172), glycochenodeoxycholate 3-sulfate (OR=1.124, 95% CI 1.021–1.238), N-acetylkynurenine (OR=1.062), and butyrylcarnitine (OR=0.958, 95% CI 0.923–0.995); nominal sepsis signals included TRAIL (OR=1.094, 95% CI 1.024–1.168), 6-bromotryptophan (OR=1.091), CCL19 (OR=0.917, 95% CI 0.861–0.977), and cis-IL6R (OR=0.858; uncorrected pind=0.0135–0.0461, BH-FDRind≥0.0828). Uremic toxins, effector cytokines, and 14 reverse models were null.
Conclusion: Genetic liability to sepsis is associated with AKI risk, whereas cytokine and gut metabolite signals represent exploratory, nominal leads that do not survive FDR correction.
Keywords: Sepsis-associated acute kidney injury, mendelian randomisation, two-step mediation, FinnGen, pQTL, cytokines, gut metabolites, bile acids