Impact of YAQ001 on the Gut Microbiome, Permeability and Markers of Systemic Inflammation in Patients With Cirrhosis

ABSTRACT

Background & Aims

YAQ001 is a novel, non‐absorbable, gut‐restricted, multiporous, carbon bead adsorbent. Pre‐clinical studies and a clinical trial suggested its therapeutic potential and safety in patients with cirrhosis. This substudy aimed to explore the effects of YAQ001 on the gut microbiome and their exploratory association with biomarkers of gut permeability and systemic inflammation.

Methods

Twenty‐eight‐patients with cirrhosis were randomized (double‐blind) to receive YAQ001 (4 g once daily) or placebo. Gut microbiome and biomarkers of inflammation and gut permeability were assessed at baseline, after 4 and 12 weeks. Changes in biomarkers and relationship with microbiome composition and gene profiling were assessed. The effect of YAQ001 on Klebsiella pneumoniae biofilms was assessed in vitro.

Results

Overall alpha and beta diversity remained unaltered. YAQ001 increased the abundance of bacteria associated with improved gut health such as Adlercreutzia equolifaciens , a bacterium commonly depleted in liver disease, and decreased the abundance of bacteria associated with infections and poor outcomes such as Klebsiella pneumonia, and Streptococcus mutans . YAQ001 impacted positively on virulence factors such as siderophores, fimbriae structures, and lipopolysaccharides that are associated with inflammation and invasion. Antibiotic resistance genes decreased significantly in the YAQ001 group. These changes were associated with a reduction in markers of gut permeability and systemic inflammation. YAQ001 prevented K. pneumoniae biofilms in vitro.

Conclusion

YAQ001 was associated with favourable compositional shifts and trends towards reduced virulence‐related gene signatures in the gut microbiome, alongside trends towards improved inflammatory biomarkers. The in vitro anti‐biofilm effect of YAQ001 warrants further mechanistic investigation. The findings position YAQ001 as a potential microbiome therapeutic.

Key Points

• In a randomized, double‐blind, placebo‐controlled trial, oral YAQ001 (4 g once daily) was safe and well tolerated in patients with cirrhosis and did not alter overall gut microbial diversity, distinguishing it mechanistically from antibiotics.

• YAQ001 was associated with a nominal increase in beneficial commensals such as Adlercreutzia equolifaciens and Faecalibacterium prausnitzii, and a nominal decrease in pathogens linked to poor outcomes in cirrhosis, including Klebsiella pneumoniae and Streptococcus mutans.

• Metagenomic analysis showed exploratory reductions in virulence factor genes (adhesion, siderophores, fimbriae, lipopolysaccharide) and antimicrobial resistance genes, particularly those linked to Streptococcaceae and Enterobacteriaceae.

• These microbiome changes were accompanied by trends towards improved gut permeability (lactulose/L‐Rhamnose ratio), systemic inflammation (CRP, CXCL10, ACE) and endothelial function (ADMA).

• In vitro, YAQ001 particles prevented K. pneumoniae biofilm formation and bound bacteria in a polarized intestinal epithelial model, providing mechanistic plausibility and supporting its development as a non‐antibiotic prophylactic strategy against biofilm‐associated infections.

• Taken together, these findings position YAQ001 as a candidate non‐antibiotic microbiome therapeutic in cirrhosis, one that may selectively reduce pathogenic species, virulence and resistance genes without the selective pressure that drives antimicrobial resistance, pending confirmation in larger trials.

Lay Summary

YAQ001 is an experimental bead taken by mouth that binds to harmful substances and bacteria in the gut without killing them. In this study, we analysed blood, stool, and urine samples from patients with liver disease to explore its effects. The results showed trends towards beneficial changes in gut bacteria composition, virulence, and antimicrobial resistance genes, which were associated with trends in reduced inflammation markers, and laboratory tests indicated that YAQ001 could prevent harmful bacteria from forming protective biofilms. However, because this was a small study, these findings are preliminary and not yet conclusive. Larger studies are needed to confirm whether these effects of YAQ001 translate into improved patient outcomes.

Abbreviations

ACE

angiotensin‐converting enzyme

ACLF

acute‐on‐chronic liver failure

ADMA

asymmetric dimethyl arginine

AMR

antimicrobial resistance

ARGs

antibiotic resistance genes

BAs

bile acids

BSA

bovine serum albumin

CA

cholic acid

CDCA

chenodeoxycholic acid

CRP

C‐reactive protein

DCA

deoxycholic acid

FBP

fibronectin binding protein

FPKM

fragments per kilobase of transcript per million fragments mapped

FreqRPKM

frequency reads per kilobase of exon model per million mapped reads

GCA

glycocholic acid

GCDCA

glycochenodeoxycholic acid

GC–MS

gas chromatography–mass spectrometry

GLCA

glycohyocholic acid

IGC2

integrated gut catalogue 2

IHMS

International Human Microbiome Standards

KEGG

Kyoto Encyclopedia of Genes and Genomes

KO

KEGG orthologue

LCA

lithocholic acid

LPS

lipopolysaccharides

MDR

multidrug resistance

MELD

Model for End‐Stage Liver Disease

METEOR

Metric for Evaluation of Translation with Explicit ORdering

MOMR

MetaOMineR

MSPs

metagenomic species pan‐genomes

NMR

nuclear magnetic resonance

NOS

nitric oxide synthase

PCA

principal component analyses

PCR

polymerase chain reaction

PE

phycoerythrin

PFBBR

pentafluorobenzyl bromide

PROJECT

Periodic Refocusing of J Evolution by Coherence Transfer

PsaA

pneumococcal surface adhesin A

PURGE

Presaturation Utilizing Relaxation Gradients and Echoes

rRNA

ribosomal RNA

SCFAs

short‐chain fatty acids

STORMS

strengthening the organization and reporting of microbiome studies

TBA

tetrabutylammonium

TCDCA

taurochenodeoxycholic acid

TLCA

taurolithocholic acid

UDCA

ursodeoxycholic acid

UPLC/Q‐TOF‐MS

ultra‐performance liquid chromatography–quadrupole time‐of‐flight mass spectrometry

VF

virulence factors

1. Introduction

Cirrhosis pathogenesis has been closely linked to the microbiome, particularly in advanced disease stages [1, 2]. Bacterial products such as endotoxins translocate from the gut to the liver and systemic circulation, driving dysregulated inflammatory responses and subsequent organ injury [3]. Current interventional strategies to modulate this process are limited to the use of oral antibiotics, which aim to reduce bacterial overgrowth and endotoxemia in cirrhosis [4, 5]. Sometimes those treatments necessitate their chronic administration for secondary prophylaxis of hepatic decompensation [6, 7]. Although antibiotics may provide short‐term benefits, they do not address the underlying gut microbiome dysbiosis and can exacerbate long‐term complications, such as the development and accumulation of antimicrobial resistance (AMR) of virulent strains [8]. AMR occurs when bacteria evolve to survive exposure to drugs that would normally kill them or stop their growth. In patients with cirrhosis, the risk of AMR is particularly high due to recurrent infections, frequent hospitalizations, and repeated antibiotic courses, which together create strong selective pressure for resistant organisms. AMR has been recognized as one of the greatest challenges facing modern medicine [9]. Antibiotic administration is particularly common among liver disease patients due to their compromised immune systems. Compounding this issue, many pathogenic bacteria in the gut form biofilms—structured communities of bacteria encased in a self‐produced protective matrix that adheres to mucosal surfaces. Biofilms act as physical barriers, reducing antibiotic penetration and promoting persistent colonization, which complicates infection management in cirrhosis. They also serve as reservoirs for endotoxins and pro‐inflammatory factors, thereby perpetuating systemic inflammation and organ injury. Therefore, there is a critical need for safer, alternative therapeutic strategies that could be used for both primary and secondary prophylaxis, offering a more effective and sustainable approach to managing cirrhosis.

YAQ001 is a novel, non‐absorbable, gut‐restricted, multiporous, carbon bead adsorbent with a defined pore size range (250–500 μm) designed to target gut‐related pathophysiological processes. Preclinical studies in cirrhosis models demonstrated its ability to reduce fibrosis, severity of portal hypertension, and organ dysfunction, likely through the restoration of a dysbiotic gut microbiome and reduction in endotoxemia [10]. By adsorbing bacterial toxins and physically sequestering bacteria, YAQ001 may disrupt biofilm formation and reduce the bacterial burden without exerting direct antimicrobial activity, thereby avoiding the selective pressure that drives AMR. This mechanism offers a potential advantage over traditional antibiotics, as it may reduce virulence and inflammation while preserving the ecological balance of the gut microbiome. A Phase 2, randomized, placebo‐controlled trial (CARBALIVE‐SAFETY STUDY) assessed the safety and efficacy of oral YAQ001 in patients with decompensated cirrhosis [11]. In this trial, 28 patients with cirrhosis were randomly assigned (double‐blind) to receive either YAQ001 (4 g once daily) or placebo. Samples, including blood, urine, and stool, were collected at baseline, 4 and 12 weeks. The primary endpoints focused on safety and tolerability, whereas secondary endpoints included organ function and nutritional status. No serious adverse events or significant differences in adverse events between groups were observed. Gastrointestinal‐related adverse events were noted but were comparable between groups [10]. Exploratory endpoints investigated microbiome composition and functionality, as well as local and systemic inflammatory indices and markers of intestinal barrier permeability.

This sub study, conducted as part of the CARBALIVE‐SAFETY trial, aims to comprehensively explore the effects of YAQ001 on the gut microbiome, including its composition and functional aspects such as metabolism, virulence factors (VFs), antibiotic resistance genes, and its effect on bacterial biofilms. We also investigated whether these microbiome changes are associated with changes in markers of gut inflammation, metabolism, and gut permeability, as well as their potential impact on systemic inflammation and bile acid metabolism to better understand the systemic implications of these changes in the context of cirrhosis.

2. Materials and Methods

2.1. Study Design and Patients

A group of 28‐patients with liver cirrhosis was randomly assigned into Active and Placebo groups as described (NCT03202498) (Figure S1) [10]. The study protocol and the informed consent form was reviewed by the members of the European Liver Patients Association, who provided critical inputs. The protocol and the informed consent were modified accordingly.

Serum and stool samples were collected at three different timepoints: baseline, Week 4, Week 12. Human stool samples were collected strictly following Strengthening The Organization and Reporting of Microbiome Studies (STORMS) guidelines [12]. In serum samples, we measured biomarkers associated with gut permeability, systemic inflammation, nutrition, intestine integrity, kidney function, and endothelial function. We also performed targeted metabolomics in serum samples. For faecal samples we measured biomarkers associated with gut inflammation, gut integrity, and metabolism (Figure 1A). All the subjects who took at least 70% of the treatment doses (2 sachets/day for Cohort 1) and did not present major protocol non‐compliances were included in this analysis. All measures of efficacy, including laboratory tests, were also performed on this population.

FIGURE 1.

Experimental design and changes in biomarkers that changed significantly with YAQ001 treatment. (A) The scheme shows experimental design. (B) Serum biomarkers show improvement with YAQ001 treatment; the boxplots show the Log₂‐fold‐change in Week 4 and Week 12 in relation to baseline in Active and Placebo groups. Statistical significance was assessed using a two‐sided Mann–Whitney U‐test.

2.2. Core Laboratory Analyses

Details of the analysis are available in Supporting Information section.

Luminex assay: The Luminex assays for the measurement of circulating cytokines were performed in accordance with the manufacturer's instructions. Faecal cytokine analysis and D‐lactate assays were performed as previously described [13].

Bile acids (BAs) analysis: Cholic acid (CA), chenodeoxycholic acid (CDCA), deoxycholic acid (DCA), lithocholic acid (LCA), ursodeoxycholic acid (UDCA), and their glycine and taurine conjugated species were identified and quantified by ultra‐performance liquid chromatography—quadrupole time‐of‐flight mass spectrometry (UPLC/Q‐TOF‐MS).

Analysis of short‐chain fatty acids (SCFAs) in stool: These were measured using Gas Chromatography–Mass Spectrometry (GC–MS).

2.3. Nuclear Magnetic Resonance (NMR) Analysis

Aliquots of plasma (from blood collected into lithium heparin tubes) and urine underwent NMR analyses using standardized protocols.

2.4. Analysis of Alterations on the Microbiome

2.4.1. DNA Extraction

For faecal samples, 200 mg of pre‐homogenized material was processed using a modified International Human Microbiome Standards (IHMS) protocol [14]. Concentration was measured by Qubit, and quality was verified by 16S ribosomal RNA (rRNA) polymerase chain reaction (PCR) and electrophoresis. Extraction blanks and positive controls were included for quality assurance.

2.4.2. Metagenomic Sequencing

Stool samples were subjected to metagenomic sequencing using the Illumina HiSeq 2000 platform (Illumina, San Diego, CA), generating paired end reads (2 × 100 bp). After quality filtering, each sample yielded a minimum of 10 million reads, with an average of 12.5 ± 2.1 million reads (mean ± SD). Only high‐quality reads fulfilling these standards were retained for downstream bioinformatic analyses.

2.4.3. Gene and Species Abundance Profiling

The raw reads for all samples were trimmed using Alien Trimmer 0.4.0 with parameters ‐k 10 ‐l 45 ‐m 5 ‐p 40 ‐q 20 and Illumina contaminant oligonucleotides [15]. Trimmed samples were mapped against the Integrated Gut Catalogue 2 (IGC2) [16] using Bowtie integrated into the Metric for Evaluation of Translation with Explicit ORdering (METEOR) pipeline (https://forgemia.inra.fr/metagenopolis/meteor). This generated gene count matrices that were then subjected to downsizing (10 million reads) and normalization (reads per kilo base per million mapped reads (RPKM method)) to generate the gene frequency matrix for downstream analysis. The resulting gene matrices were then projected on previously reconstructed metagenomic species (MGSs) [17, 18] using the top 50 marker genes per MGS to calculate MGS abundances for each sample. Genes included in the IGC2 catalogue are associated with MGSs; we used the taxonomic annotations of the MGSs to assign taxonomic annotations to the genes.

2.4.4. Functional Annotation of Genes in IGC2 Catalogue

Protein sequences of the IGC2 catalogue were annotated for Antibiotic Resistance Genes using the DeepARG software [19], using a minimum probability cutoff of 0.9. Proteins of IGC2 catalogue were annotated to Kyoto Encyclopedia of Genes and Genomes (KEGG) orthologs using Diamond against KEGG database [20, 21] Best‐hit alignments with e‐value < 10⁻⁵. Annotation of VFs was performed using blastp Best‐hit alignments with an e‐value cutoff of 1e10⁻¹⁰ and a minimum percentage identity of 65% against the set of sequences provided in the virulence factor database (VFDB) [22]. Annotation of Secondary BA genes was performed using best‐hit alignments of blastp running against sequences of genes and suggested cutoffs as previously reported (BaiA, HDHA, BaiL, BaiB, BaiCD, BaiH, BaiF, BaiK, BaiG, e‐value 1e‐10; BaiJ, e‐value 1e‐5) [23]. To construct the KO, VF, and ARG abundance tables, we started from the gene count table, grouped genes sharing the same annotation (KO, VFID, or ARG), and summed their abundances.

2.5. Analysis of Anti‐Biofilm and Epithelial Anti‐Adhesion Activity of YAQ001 Against Klebsiella pneumoniae Biofilm Formation and Treatment Assays

Klebsiella pneumoniae was selected for the in vitro experiments because it is: (i) one of the most frequently isolated Gram‐negative pathogens in cirrhotic patients with spontaneous bacterial peritonitis and bacteremia, (ii) a strong biofilm former, and (iii) a representative member of the Enterobacteriaceae family, which is often enriched in the dysbiotic gut microbiome of patients with cirrhosis. Although other organisms such as E. coli and Enterococcus spp. are also relevant, K. pneumonia serves as a robust model organism for assessing anti‐biofilm activity due to its well‐characterized biofilm formation pathways. Details of the analysis are available in Supporting Information section.

2.5.1. Biofilm Formation and Treatment Assays

Klebsiella pneumonia (TOP52) was pre‐cultured statically in TSB at 37°C for 48 h and seeded into 24‐well polystyrene plates (1 mL/well) at an OD₆₀₀ = 0.1. Plates were incubated statically at 37°C. Total biofilm biomass was quantified by crystal violet (CV) staining. For the biofilm prevention assay, YAQ001 particles (0.625–10 mg/mL in PBS) were added at the time of inoculation, and biofilms were incubated for 72 h. The antibiofilm effect of YAQ001 was investigated against K. pneumonia mature biofilms (pre‐formed for 48 h) followed by the incubation of YAQ001 particles (0.625–10 mg/mL in PBS) for 24 h. Antibiotic controls were included using both meropenem and ciprofloxacin at 1× to 10 000× the minimum inhibitory concentration (MIC). Antibiotics used in synergy with 10 mM EDTA were also tested. MICs were determined by the broth microdilution method (Figure S6).

2.5.2. Polarized Intestinal Epithelial Model and Infection

A polarized human intestinal epithelial layer was established by co‐culturing Caco2 colorectal epithelial cells and HT29‐MTX‐E12 mucous‐producing, goblet‐like colorectal cells at a 3:1 ratio on Snapwell inserts for 14 days. Medium was refreshed every 2–3 days until maturation. For infection, K. pneumonia TOP52 (GFP‐tagged) [24] was grown statically in TSB at 37°C for 48 h and adjusted to an optical density of OD₆₀₀ of 0.001 or 0.005 corresponding to approximately 1 × 10⁶ CFU/mL and 2 × 10⁶, respectively. The concentration of K. pneumonia (1 × 10⁶ CFU/mL) used in the biofilm formation assay was chosen based on preliminary optimization experiments, which demonstrated that this inoculum density yields consistent and reproducible biofilm formation within 24 h without reaching stationary‐phase saturation.

The apical chamber was infected with K. pneumonia (TOP52) in antibiotic‐free medium for 14 h at 37°C. For treatment, YAQ001 particles (10 mg/mL in PBS) were applied simultaneously with bacteria to a total volume of 500 μL. A non‐YAQ001 treated control was included for comparison. CFU/ml of planktonic and mucosal associated bacteria were quantified the following day. Confocal microscopy was performed with a Leica SP8 confocal microscope and images were analysed using LeicaX Software.

2.6. Statistics

Differential abundance analysis of MSP abundance and genes abundance between Placebo and Active arms was performed using Mann–Whitney U‐tests, two tailed, with a cutoff p‐value of 0.05; no features were identified as significant after applying False Discovery Rate correction; therefore, only p values before correction were employed. No prevalence or abundance filters were applied in abundance tables. For comparing the effect on biomarkers between Active arm and Placebo, first for each subject we measured the Log‐2 fold change after 4 and 12 weeks in relation to baseline. Then we compared the change between Active and Placebo samples at the same week using a Mann–Whitney U‐test. Correlations between MSP abundance and measured biomarkers were performed using Spearman correlation test.

3. Results

No significant differences in clinical and biochemical variables, prognostic scores such as the Child‐Pugh score [median (IQR) 7 (7–8) vs. 7 (7–8)] or the Model for End‐Stage Liver Disease (MELD) scores [median (IQR) 13.2 (10.2–16.1) vs. 12.6 (9.7–13.5)] were observed between Placebo and Active groups. The rate of adverse events between the two groups was similar. As shown in the previous publication, the nutritional state and associated markers were similar between groups [16]. Patient details are described in Table S1.

3.1. Effect of YAQ001 in Biomarkers of Systemic Inflammation, Gut Permeability and Endothelial Function

3.1.1. Systemic Inflammation and Endothelial Function

Of the large number of markers of systemic inflammation measured, we observed nominal improvements in the following serum biomarkers in the Active arm in comparison to the Placebo group based on uncorrected p‐values; these findings are exploratory and did not survive FDR correction. Given the clinically stable nature of the studied patients, many of the measurements fell within the normal range and the associated changes in the biomarkers were small. Angiotensin‐converting enzyme (ACE) and CXCL10 (Figure 1B and Figure S2A, Table S2) showed a nominal decrease after 12 weeks in the YAQ001 and no improvement in the Placebo group. C‐reactive protein (CRP) showed a nominal at Week 4. ACE is involved in the conversion of angiotensin I into the vasoconstrictor angiotensin II, promoting fibrosis and increased portal pressure within the liver [25]. CRP functions as part of the innate immune system by binding to damaged cells and pathogens, rapidly increasing in concentration in the bloodstream when tissue injury or infection occurs. Persistently high CRP levels in cirrhotic patients have been associated with complications such as decompensation, bacterial infections, and mortality. CXCL10 acts as a pro‐inflammatory chemokine, attracting immune cells to the site of injury contributing to hepatic inflammation and fibrosis. Asymmetric dimethyl arginine (ADMA) showed a nominal improvement at Week 12 in the active arm. Increase in ADMA has been associated with impaired liver function. ADMA inhibits endothelial nitric oxide synthase (NOS) activity resulting in endothelial dysfunction and abnormalities of the hepatic microcirculation and portal hypertension [26, 27]. Additionally, we observed a reduction in cytokeratin‐18, M30 in the Placebo group, but the numbers were within the normal range (Figure S2F, Table S2).

3.1.2. Gut Permeability

We found a nominal improvement in the Lactulose/L‐Rhamnose ratio, which is a marker of gut permeability (Figure 1B) based on uncorrected p‐values, and this finding requires validation in future studies.

3.1.3. Gut Metabolism

We observed a nominal reduction in the acetic acid, propionic acid, and butyric acid concentrations in the Active arm group (Figure S2D) and no differences in the nutrition biomarkers (Figure S2E). However, the mean value of the concentration at the end of the treatment in the Active arm is 63.5, 7.3, and 14.5 μmol/g wet weight for acetic acid, propionic acid, and butyric acid, respectively, which fall within the normal healthy ranges.

3.1.4. Markers of Gut Inflammation

Although there were trends towards lower pro‐inflammatory cytokines and endotoxin and increased anti‐inflammatory cytokines in the faeces of the patients in the Active arm, the differences were not statistically different in any of the biomarkers measured in faecal samples (Figures S3 and S4). These null findings further support the exploratory nature of the positive signals observed elsewhere.

3.2. Effect of YAQ001 on the Composition of the Microbiome

To evaluate the changes in the composition of the gut microbiome to the treatment and placebo, we performed shotgun metagenomics on 70 samples, with an average of 15 million reads. The raw sequences were filtered and mapped against the gut non‐redundant microbial Integrated Gene Catalogue 2 (Methods, Table S3) [16]. After the normalization based on the determined gene counts for each sample, the metagenome species abundances were calculated using the gene counts. After YAQ001 treatment, the species richness between the Placebo and Active groups was similar across all three visits (Figure 2A). Additionally, when comparing the distribution of the Bray Curtis distance within individuals at different timepoints between Placebo and Active groups, no significant differences were observed (Figure 2B). This could indicate that changes in the overall microbial community composition through time are similar in both Active and Placebo groups.

FIGURE 2.

Effect of YAQ001 treatment on microbiome species and functions. (A) Boxplot showing species richness in Active and Placebo arms at baseline, Week 4 and Week 12. No statistically significant difference was observed between Active and Placebo groups. (B) Boxplots show beta diversity between time points within the same individual comparing Active vs. Placebo groups. No statistically significant difference was observed between Active and Placebo groups. (C) Differentially abundant species between Active and Placebo arms, the heatmap shows the log₂‐fold change in mean values of species abundance between Active and Placebo arms. (D) The cladogram shows the KEGG functional categories with at least three genes increasing or decreasing after YAQ001 treatment in both Week 4 and Week 12, external ring shows the number of genes increasing (red) or decreasing (blue). Statistical significance was assessed using two‐sided Mann–Whitney U‐test and was reported as (*) p < 0.05.

Comparing the effect of YAQ001 with Placebo at the level of individual microbes, we observed nominal increases (uncorrected p < 0.05) in most bacteria that increased in abundance in the YAQ001 group and have been associated with promoting gut health, such as Adlercreutzia equolifaciens , a commensal bacterium with anti‐inflammatory properties commonly depleted in liver disease [28]. Also, we observed a nominal increase in Faecalibacterium prausnitzii , a commensal bacterium and significant butyrate producer with anti‐inflammatory properties. In contrast, most of the species showing nominal decreased abundance with YAQ001 treatment have been associated with infections and poor outcomes in liver disease patients, such as K. pneumoniae [29, 30] and Streptococcus mutans (Figure 2C, Table S4) [31]. Given that no microbiome changes remained significant after FDR correction, these species‐level associations should be interpreted as exploratory and hypothesis‐generating. For additional visualization of species that were differentially abundant between Placebo and Treatment arms specifically at Week 12 (with the coefficient representing the effect of Placebo on abundance), see Figure S7.

To reveal the functional landscape of the changes in the gut microbiome, we performed differential abundance analysis of genes grouped by KEGG orthologue (KO) annotations between Active and Placebo arms (Methods, Table S5). Among the nominally significant changes (uncorrected p < 0.05), the functional categories with most genes changing in abundance were the ABC transporters, two‐component systems, and biofilm formation genes, all related to external bacterial components. We also found the metabolic pathways of arginine biosynthesis, cysteine and methionine, galactose, methane, oxidative phosphorylation, and biosynthesis of non‐ribosomal peptide siderophores had at least three nominally depleted enzymes in the Active arm group (Figure 2D). Notably, within the biofilm formation category, only genes decreasing in abundance were observed in the Active arm. Biofilm formation plays a crucial role in pathogenesis, allowing bacteria to adhere to surfaces, providing protection against the immune system and antibiotics, leading to difficult‐to‐treat infections. These observations preliminarily suggest that the effect of YAQ001 treatment might impact the membrane or extracellular components of the affected species, though these functional predictions require confirmation in independent cohorts.

3.3. Effect of YAQ001 on Microbial Genes Associated Virulence and Antimicrobial Resistance

As observed above, bacterial species showing nominal decreases in abundance in the Active arm have been described as pathogenic. Consequently, we performed differential abundance analysis of the microbiome genes coding for virulence factors (VFs) (Tables S6 and S7) comparing Active vs. Placebo groups (Methods). VFs are bacterial components providing pathogenic bacteria an increased capacity to infect the organism [32]. Among the differentially abundant VF genes identified using uncorrected p‐values, all showed a decrease in abundance in the Active arm compared with the Placebo group, with the greatest decrease at Week 12 (Figure 3A,B); however, these results did not survive FDR correction and are therefore exploratory. For a complementary view of VFs that were differentially abundant between Placebo and Treatment arms at Week 12 (with the coefficient representing the effect of Placebo on abundance), see Figure S8.

FIGURE 3.

Virulence factors decreasing after YAQ001 treatment. (A) Heatmap showing the log₂‐fold change in mean values virulence factors between Active and Placebo. (B) The boxplots show the sum of the abundance values of all differentially abundant virulence factors in Active and Placebo groups. (C) Alluvial plots show the proportions of the mean abundance values of the virulence factors genes and their associated family of samples at Week 12 in the Active (left) and Placebo arm (right). (D) Correlation network between virulence factor genes and measured serum biomarkers: Oval nodes represent genes with virulence factor annotations within a Species. Rectangle nodes represent the measured serum biomarkers; Edges represent correlations between a gene and a biomarkers; positive correlations are shown in red, negative correlations in blue; edge width is proportional to absolute correlation value. Only correlations with p values < 0.002 are shown. Statistical significance in (A) was assessed using two‐sided Mann–Whitney U‐test and was reported as (*) p < 0.05. Correlations in (D) were calculated using Spearman correlation test.

The overall set of decreasing VF genes was linked to adhesion, nutritional factors, effector delivery systems, immune modulation, regulation, invasion, and antimicrobial resistance categories (Figure 3A, Table S8). The two most frequent categories were adhesion and nutritional factors. Adhesion VF enable bacteria attachment, invasion, and biofilm formation; for example, fibronectin binding protein (FBP) promotes fibronectin‐mediated collagen recruitment, which leads to matrix deposition on and between streptococcal cells to induce the formation of large bacterial aggregates [33]. In the nutritional VF category, we found mostly genes related to metal ion uptake, like pneumococcal surface adhesin A (PsaA), whose key function is the transport of Mn²⁺ and Zn²⁺ into the cytoplasm of the bacteria; PsaA mutants show a marked impact on the capacity to colonize and increased susceptibility to oxidative damage [34]. Also, within this category, we observed the Sal gene, a major trigger of inflammation and bacterial dissemination during K. pneumonia lung infection. Those genes linked to adhesion and nutritional factor categories showed the greatest reduction in the relative abundance at the end of the treatment (Figure 3C). A broad range of bacterial families was associated with the changing VFs where the Streptococcaceae family showed the greatest changes after treatment (Figure 3C).

Interestingly, we identified significant negative correlations based on nominal significance levels between the gene abundance of VFs, like FBP and PsaA, within the Streptocaccaceae family and the ACE biomarker, one of the nominally improving biomarkers after YAQ001 treatment (Figure 3D); these correlation findings are preliminary and require replication.

Identifying VFs as one of the key phenotypic changes and as the YAQ001 treatment nominally reduced some of the opportunistic pathogens, we hypothesized that the treatment could result in a reduction in the AMR genes in the gut microbiome. ARGs pose a significant concern when carried by pathogenic bacterial strains. Moreover, the ARG and VF genes are intertwined and can overlap. Therefore, next we annotated the ARGs using DeepARG software [19]. Then we investigated the impact of treatment by performing a differential abundance analysis of predicted ARGs (Tables S9 and S10) between the Active and Placebo groups. Among those with nominal significance (uncorrected p < 0.05), most of the differentially abundant ARGs showed a decrease in abundance in the Active group (Figure 4A,B). In the multidrug resistance (MDR) family, we observe RamA and OqxB showed the largest (mean log₂ fold change < −7); both have an efflux pump mechanism and are also linked to the Enterobacter and Streptococcus species (Figure 4A,C). Beta‐lactam resistance was another family of genes with nominally dramatic decrease after treatment with significant genes PBP1a and PBP1b linked to the Streptococcus pneumoniae . As observed with VFs, ARGs linked to the Streptococcaceae and Enterobacter family showed the greatest reduction in the Active group (Figure 4A,C). These ARG findings are exploratory and should be validated in future studies with more stringent statistical correction (see Figure S9 for additional visualization of ARGs differentially abundant between Placebo and Treatment arms at Week 12, with the coefficient representing the effect of Placebo on abundance).

FIGURE 4.

Profile of the antibiotic resistance genes after YAQ001 treatment. (A) Heatmap showing the log₂‐fold change in mean values predicted ARGs between Active and Placebo. Stars (*) p < 0.05 statistical significance was reported as (*) p < 0.05. (B) The boxplots show the sum of the abundance values (RPKM) of all differentially abundant ARGs in Active and Placebo groups. (C) Alluvial plots show the proportions of the mean abundance values of the ARGs and their associated family of samples at Week 12 in the active (left) and placebo arm (right).

3.4. Effect YAQ001 on Faecal and Serum BAs and Microbiome Genes Linked to Its Metabolism

The intestinal microbiota can influence the size and composition of the bile acid pool in serum and alterations in the secondary bile acid pool have been observed in liver cirrhosis patients [35]. Therefore, we proceeded to test for differential abundance of genes reported to participate in the secondary bile acid metabolism (Table S11) and compared these to changes in faecal and serum secondary bile acid concentration (Tables S12 and S13) in relation to baseline in Active and Placebo groups. After YAQ001 treatment, only the BaiCD gene changed, showing a nominal decrease in abundance in the Active group (Figure S4). In serum, taurolithocholic acid (TLCA) and glycolithocholic acid (GLCA) showed a trend towards increasing in the active arm but not in placebo at Week 12. In faecal samples, glycocholic acid (GCA), taurochenodeoxycholic acid (TCDCA), glycochenodeoxycholic acid (GCDCA), and lithocholic acid (LCA) showed trends to an increase in the Active arm compared to Placebo at Week 12 (Figure S4). These trends are descriptive only and do not constitute statistically significant findings.

3.5. Effect of YAQ001 on the Host Metabolome

The impact of YAQ001 on metabolic profiles was evaluated through targeted metabolomic analysis at Week 4 and 12 relative to baseline (Figure S5, Table S14). In Active group, nominally significant alterations in metabolite levels were observed compared to the Placebo group. At week plasma tyrosine and phenylalanine showed a marked reduction in the Active arm. These changes persisted through Week 12, demonstrating a sustained modulation of metabolic activity in response to YAQ001 treatment. In contrast, the Placebo group exhibited no significant shifts in metabolic profiles at either time point. These results preliminarily suggest that YAQ001 induces early and persistent modifications in the host metabolome, though these metabolomic findings also require confirmation with appropriate multiple‐testing correction in future studies.

3.6. YAQ001 Particles Prevent Biofilm Formation of Klebsiella pneumonia TOP52

In order to define the potential mechanisms associated with the selective depletion of pathogenic bacteria such as K. pneumonia with YAQ001, we explored its effect on biofilms generated by this organism in in vitro experiments. Crystal violet staining confirmed time‐dependent K. pneumonia TOP52 in vitro biofilm formation, with maximal biomass detected at 72 h (Figure 5A,B). YAQ001 particles did not significantly reduce biofilm biomass against mature biofilms (48 h) (Figure 5D). Antibiotics, either alone or combined with EDTA, failed to significantly disrupt established biofilms (Figure 5B). However, in the biofilm prevention assays, YAQ001 particles reduced biofilm formation in a concentration‐dependent manner, with a significant effect at 10 mg/mL (p < 0.05) (Figure 5F). Meropenem had no biofilm prevention effect at any tested concentration, whereas only ciprofloxacin significantly prevented biofilm formation only at the highest concentrations tested (100×, 1000×, and 10 000× MIC). Experimental schematics of antibiofilm and biofilm prevention assays are described accordingly (Figure 5C,E). In vitro experiments are not subject to FDR concerns and are interpreted as mechanistic support for the exploratory microbiome observations.

FIGURE 5.

YAQ001 prevents biofilm formation. (A) Biofilm mass quantified at 24‐, 48‐ and 72‐h time points to determine biofilm maturation. Klebsiella pneumonia was cultured in 24‐well plates, and YAQ001 or antibiotics were added either at (E) inoculation (prevention) or (C) after 48 h of biofilm formation (antibiofilm assay). (B) Biofilm biomass was quantified by crystal violet staining at indicated time points. (D) Antibiofilm effect of YAQ001 was tested against mature biofilms (48 h). (F) Biofilm prevention assay was used to determine the effect of YAQ001 particles to inhibit K. pneumonia. Data represents mean ± SD of three independent experiments. An ordinary one‐way ANOVA with Tukey multiple comparisons was used, and statistical significance was reported as (*) p < 0.05, (**) p < 0.01, (****) p < 0.0001.

3.7. K. pneumonia TOP52 Binds to the Surface of YAQ001 Particles and May Prevent Mucosal/Epithelial Association in a Polarized Human Intestinal Model

To investigate bacterial‐epithelial interactions, a polarized intestinal epithelial model was generated using a Caco‐2/HT29‐MTX‐E2 co‐culture and infected with K. pneumonia TOP52 with and without YAQ001 particles (10 mg/mL) simultaneously (Figure 6A). After 14 h treatment, colony‐forming units (CFU) of the planktonic fraction did not differ between YAQ001‐treated and untreated conditions (Figure 6B). In contrast, we observed potentially lower CFUs/ml in the mucosal fraction when treated with YAQ001 compared to the untreated control (data not shown). Via confocal imaging we observed bacterial attachment to both YAQ001 and non‐YAQ001 treated cells. In YAQ001‐treated cells we observed strong bacterial attachment and aggregation of K. pneumonia TOP52 to YAQ001 particles even after multiple PBS washes (Figure 6D). The data suggest that the lower mucosal association may be attributed to this mechanism. No bacteria were observed in the uninfected controls (Figure 6C). These in vitro mechanistic findings provide supportive evidence for the exploratory in vivo observations.

FIGURE 6.

Klebsiella pneumonia (TOP52) binds to YAQ001 particles in a human polarized intestinal epithelial infection model. (A) A schematic of polarized intestinal epithelial infection with K. pneumonia TOP52‐GFP with YAQ001. Caco‐2/HT29‐MTX‐E2 co‐cultures were maintained for 14 days prior to infection with K. pneumonia TOP52‐GFP. YAQ001 (10 mg/mL) was applied simultaneously with infection, and outcomes were assessed after 14 h. (B) YAQ001 particles and bacteria were applied to the apical chamber, and planktonic vs. mucosal fractions were collected for CFU analysis. (C) Representative confocal images showing uninfected and infected co‐cultures with and without YAQ001 particles. YAQ001 particles are denoted with a white dashed circle. (D) Side view and 3D images indicating bacterial localization with YAQ001. DNA shown in cyan (DAPI), Actin shown in magenta (phalloidin‐647) and K. pneumonia TOP52‐GFP shown in green.

4. Discussion

YAQ001 is a novel, orally administered, gut‐restricted therapeutic approach to sequester substances and metabolites of 1–500 nm in the gut lumen without exerting any direct antimicrobial activity [10]. In animal models of cirrhosis, it has been shown to prevent progression of fibrosis, reduction in bacterial translocation, improvement of organ function, occurrence of acute on chronic liver failure with widespread effects on severity of endotoxemia, gene expression in multiple organs and positive effects on the microbiome [10]. The CARBALIVE‐SAFETY clinical trial was a randomized double‐blind trial of YAQ001 in patients with cirrhosis, in which its safety and tolerability were confirmed [10]. In this study, we used the blood, stool and urine samples obtained in the above clinical trial to explore the biological effect of YAQ001 in these patients. The most notable biological observation associated with YAQ001 was noted in the abundance and functionality of the gut microbiome and its impact on gut permeability and some markers of systemic inflammation. We observed a nominal decrease of pathogenic bacterial species and a nominal increase of autochthonous bacteria without major changes in the global microbiome composition. This was associated with exploratory functional changes in the microbial genes involved in membrane‐associated proteins linked to the virulence capabilities, antibiotic resistance and biofilm formation. In in vitro studies, YAQ001 prevented the formation of biofilm. These preliminary changes in the microbiome were associated with trends towards favourable effects on biomarkers representing the intestinal barrier permeability, inflammation and endothelial function. Importantly, these findings should be interpreted as exploratory and hypothesis‐generating, requiring validation in larger independent cohorts.

The metagenomic composition studies showed no significant differences in bacterial diversity, with either YAQ001 treatment or Placebo. This is an important observation as it provides further evidence that YAQ001 does not exert an antibiotic effect and distinguishes it mechanistically from other microbiome targeting treatments such as antibiotics. In a previous in vitro study, no effect of YAQ001 on growth characteristics of the bacteria was observed in contrast to antibiotics, which prevented bacterial proliferation [10]. Shifts in Klebsiella and Streptococcus in humans compared favourably with shifts in Veillonella and Roseburia species in the mouse gut, indicative of a less hostile environment. These data are of particular importance as cirrhosis patients are prone to developing complications such as variceal bleeding, hepatic encephalopathy, ascites, bacterial infections and mortality, which have been associated with gut microbial changes. In cirrhosis patients with progressive disease, pathogenic bacteria such as Klebsiella and Streptococcus species become more abundant and can cause infections and progression of liver disease [36, 37].

As has been shown previously, we observed the presence of multiple microbial VFs, which are known to impact negatively on the outcome of cirrhosis patients [38]. Although these factors are more pronounced in patients with more severely decompensated disease, it was interesting to note their presence in this population with well compensated cirrhosis. A key observation of this study was the exploratory association of YAQ001 with virulence genes, which encode for multiple virulence factors that extend to mechanisms which allow invasion of the host through secretion of fimbriae, generation of biofilm, secretion of siderophores, attachment processes, and secreted lysins. Although the exact mechanism through which YAQ001 exerts this effect is not clear, it may do so by providing a physical barrier selectively decreasing the abundance of virulent species without exerting a selective pressure as is the case with antibiotic administration, which in the long run increases the risk of infection of resistant strains. However, this mechanistic interpretation remains hypothesis generating and requires independent confirmation.

AMR is a particular issue in patients with cirrhosis who have greater susceptibility to development of infections due to cirrhosis‐associated immune dysfunction [3, 39]. There are several mechanisms that contribute to this immune dysfunction but alterations in the microbiome composition, gut permeability, translocation and inflammation all seem to be important [3]. ARGs confer microbial communities with the ability to resist antibiotics and represent quorum‐sensing and secretion systems that allow microbiota to survive through adaptations in genotypic responses to antimicrobials. The gut microbiome acquires ARGs leading to colonization with pathogenic organisms [40, 41, 42, 43]. The data in our study showed a nominal reduction in the abundance of ARGs in the group treated with YAQ001 compared with placebo, with effects on ARGs associated with several antibiotics including beta‐lactams, aminoglycosides, MLS and multidrug resistance genes. The data preliminarily suggest that the reduction in the abundance of the Streptococcus and Klebsiella species most likely contributes to the reduced burden of ARGs. As with the other metagenomic findings, these ARG observations are preliminary and require further studies for confirmation.

The exploratory changes in the metagenomic composition associated with YAQ001 treatment led us to speculate that some of the differentially abundant extracellular bacterial components, which under ordinary conditions facilitate the binding and invasion, might be interacting with the surface of the non‐absorbable carbon beads. This property possibly led to a nominal selective decrease of the strains enriched in such elements, along with other genomic elements commonly observed in those strains, such as AMR genes. In turn, this would ameliorate the gut damage inflicted by such bacterial species, and in consequence, lead to an improvement in gut health. To explore the potential operative mechanism through which YAQ001 may alter the composition of the microbiome and its functions, we explored its effect on K. pneumoniae biofilms. The study demonstrated that YAQ001 particles effectively prevented K. pneumoniae biofilm formation in vitro at the highest concentration tested. Crystal violet assays indicated concentration‐dependent inhibition of biofilm formation by YAQ001, although its effect on pre‐formed biofilms was limited and variable—consistent with the known tolerance of mature biofilms to antimicrobials. These findings emphasize the constraints of antibiotic‐based treatments in managing biofilm‐associated infections and highlight the promise of non‐antibiotic alternatives like YAQ001 for biofilm prevention. These in vitro mechanistic data are not subject to the multiple‐testing concerns and provide supportive biological plausibility for the exploratory in vivo microbiome observations.

The biomarkers related to systemic inflammation showing nominal improvement included ACE, CRP, and CXCL10. Systemic inflammation is central to the pathogenesis of complications of cirrhosis, most markedly acute decompensation and acute on chronic liver failure [44]. ACE is involved in the conversion of angiotensin I into the vasoconstrictor angiotensin II, promoting fibrosis and increased portal pressure [45]. We observed nominal strong positive correlations between ACE and virulence factors associated with the species of the Streptococcaceae family, suggesting a preliminary link between virulence factors and increased ACE levels. CXCL10 is a pro‐inflammatory chemokine which has been shown to correlate with acute decompensation, ACLF severity and complications in subjects with severe portal hypertension [46]. C‐reactive protein has been identified as an independent predictor of mortality and therefore represents a significant target for intervention. Immune cell activation is driven by bacterial translocation in cirrhosis which is a potential mechanism for this observation. These data are broadly consistent with observations in a previous study in animal models [10].

Endothelial dysfunction is a feature of cirrhosis, which is associated with systemic inflammation [47]. It was therefore not surprising to note a nominal reduction in ADMA in the YAQ001 treated patients, which is commensurate with the reduction in the severity of portal hypertension and improvement of blood pressure and renal dysfunction observed in animal studies. No such correlate was observed in the present study as the patients had relatively compensated cirrhosis.

Existing data reveal that acute decompensation is associated with a highly inflamed and permeable gut barrier [48]. A nominal decrease in the Lactulose/L‐Rhamnose ratio after YAQ001 treatment suggests a potential trend towards a stronger gut barrier. Although the exact mechanism has not been elucidated, there were trends towards a positive modulation of faecal cytokines, which are likely contributors. The lack of significant changes in faecal cytokines most likely reflects the relative normality of the cytokine concentrations, reducing the ability to distinguish any impact of YAQ001.

The data described must be interpreted considering some limitations. The relatively small sample size reduces the power of statistical tests, allowing only the detection of trends towards improvement in several markers but in many cases not reaching statistical significance. Moreover, and most importantly, no microbiome alterations remained statistically significant after FDR correction for multiple comparisons. Consequently, all microbiome‐related findings—including species shifts, VF and ARG changes, and functional pathway analyses—are largely exploratory and hypothesis‐generating. In our previous study in animal models, the employed dose was higher, and the animals were more unwell, which may account for a more significant effect of YAQ001. Further studies are required to determine an optimal dose. There were trends towards reduction in short chain fatty acids, but the concentrations remained well within the normal range, an observation which should be further assessed. Finally, there was a trend towards higher values of the M30 component of cytokeratin 18, a marker of apoptosis, in the YAQ001 treated patients. As the values were within the normal range, the significance of this observation is difficult to gauge.

In conclusion, these exploratory findings support the potential of YAQ001 to address key pathophysiological mechanisms underlying complications of cirrhosis, particularly its effect on the microbiome composition, its virulence and AMR, which were associated with significant impact on markers of inflammation, endothelial function and gut permeability. These observations require confirmation in larger clinical trials with appropriate statistical correction for multiple comparisons. Nevertheless, the consistent in vitro mechanistic data provide biological plausibility, and the overall pattern of findings supports the development of YAQ001 as a novel prophylactic strategy against biofilm‐associated infections, with promising applications in intestinal colonization control and other clinical settings as a microbiome therapeutic.

Author Contributions

Conception and design: Jane Macnaughtan, Jude A. Oben, Nathan Davies, Rajeshwar P. Mookerjee, Susan Sandeman, Fausto Andreola, Rajiv Jalan; administrative support: Karen Church, Carrie Morgan, Thomas Avery, Michal Kowalski, Daniel Green; provision of study materials or patients: Javier Martinez, Cesar Jimenez, Giacomo Zaccherini, Katherine Husi, Miguel Rodriguez‐Gandia, Adria Juanola, Elsa Sola, Helena Cortez‐Pinto, Reiner Wiest, Francois Durand, Paolo Caraceni, Vicente Arroyo, Victor Vargas, Agustin Albillos, Gautam Mehta, Annarein J. C. Kerbert; collection and assembly of data: Jinxia Liu, Fernando Garcia‐Guevara, Jane Macnaughtan, Nicholas Yuen, Abeba Habtesion, Alexandra Phillips, Francesco De Chiara, Ganesh Ingavle, Tom Baldwin, Paul Cordero‐Sanchez, Junpei Soeda, Aarti Jalan, Amir Gander, Lindsey Ann Edwards, Nathan Davies, Fausto Andreola, Rajiv Jalan; data analysis and interpretation: Jinxia Liu, Fernando Garcia‐Guevara, Yi Jin, Frederick Clasen, Theo Portlock, Jia V. Li, Simon Eaton, I. Jane Cox, Roberto Elosua, Joan Vila, Marco Pavesi, Rajeshwar P. Mookerjee, Saeed Shoaie, Rajiv Jalan; manuscript writing: Jinxia Liu, Fernando Garcia‐Guevara, Jane Macnaughtan, Rajeshwar P. Mookerjee, Saeed Shoaie, Rajiv Jalan; final approval of manuscript: All authors.

Funding

This study was performed with support from a grant from the EU H2020, Grant Agreement 21 number: 634579—CARBALIVE—H2020‐PHC‐2014‐2015/H2020‐PHC‐2014 programme; J.L. was supported by a grant from National Natural Science Foundation of China (82570690).

Ethics Statement

France: Agence Nationale de Sécurité du Médicament et des Produits de Sante (ANSM), EudaMed (CIV‐PT‐18‐01‐022784, 18 May 2018). Italy: Ministero Della Salute (AIFA), EudaMed (CIV‐PT‐18‐01‐022784, 22 May 2018). Portugal: National Authority of Medicines and Health Products, I.P. (INFARMED), EudaMed (CIV‐PT‐18‐01‐022784, 23 March 2018). Spain: Agencia Española de Medicamentos y Productos Sanitarios AEMPS, EudaMed (CIV‐PT‐18‐01‐022784, 08 August 2018). Switzerland: Swissmedic –CH, EudaMed (CIV‐PT‐18‐01‐022784, 25 May 2018). UK: Medicines & Healthcare products Regulatory Agency (MHRA) (CI/2018/001 (cirrhosis), 25 April 2018).

Conflicts of Interest

Rajiv Jalan is the inventor of OPA, which has been patented by UCL and licensed to Mallinckrodt Pharma. He is also the founder of Yaqrit Discovery, Hepyx Limited (spin‐out companies from University College London), and Cyberliver. He has research collaborations with Yaqrit Discovery. Yaq‐001 was licensed by Yaqrit Ltd. from UCL. JRM has received consultancy fees from EnteroBiotix and Cultech, and speaker fees from Falk Forum.

Supporting information

Figure S1: CONSORT diagram. Of 34 assessed individuals, 6 were excluded, and 28 were randomized (14 active, 14 placebo). By Week 4, 2 placebo participants had discontinued (1 AE, 1 noncompliance), leaving 26 with available data (14 active, 12 placebo). By Week 12, a further 3 participants had dropped out (1 active due to noncompliance; 2 placebo due to AE and noncompliance), giving 23 completers (13 active, 10 placebo). Week 12 primary efficacy analyses were performed on the per‑protocol population (n = 23), with intention‑to‑treat sensitivity analyses for the primary outcomes.

Figure S2: Serum biomarkers. Log₂ fold change after 4 and 12 weeks in relation to baseline of serum biomarkers in Active and Placebo groups. (a) Systemic inflammation. b, Gut permeability. (c) Intestine, Kidney and Endothelial function. (d) Metabolism. (e) Nutrition. (f) Cell death. Stars show p‐value < 0.5, Mann‐Whitney U‐test, alternative = less, comparing with the other arm at the same week.

Figure S3: Faecal biomarkers. Log₂ fold change after 4 and 12 weeks in relation to baseline of faecal biomarkers in Active and Placebo groups. Stars show p‐value < 0.5, Mann–Whitney U‐test, alternative = less, comparing with the other arm at the same week.

Figure S4: Secondary bile acids. Log₂ fold change after 4 and 12 weeks in relation to baseline of secondary bile acids, in Active and Placebo groups. (a) Secondary bile acids measured in serum samples, (b) secondary Bile acids measured in faecal samples. Stars show p‐value < 0.5, Mann–Whitney U‐test, alternative = less, comparing with the other arm at the same week.

Figure S5: Targeted metabolomics. Log₂ fold change after 4 and 12 weeks in relation to baseline of targeted metabolomic measured metabolites in Active and Placebo groups. Stars show p‐value < 0.5, Mann–Whitney U‐test, alternative = less, comparing with the other arm at the same week.

Figure S6: Minimum inhibitory concentrations (MICs) of meropenem and ciprofloxacin against planktonic Klebsiella pneumoniae .

Figure S7: Scatter plots showing abundance over time for species differentially abundant between Placebo and Treatment arms at Week 12 using Mann–Whitney U‐test and p < 0.05. Each panel reports the p‐value and coefficient using a mixed with time and treatment arm as fixed effect, and subject as random effect. The coefficient represents the effect of Placebo on abundance. Species with estimated p‐value < 0.05 are marked with an asterisk *, none remained significant after FDR correction.

Figure S8: Scatter plots showing abundance over time for virulence factors differentially abundant between Placebo and Treatment arms at Week 12 using Mann–Whitney U‐test and p < 0.05. Each panel reports the p‐value and coefficient using a mixed with time and treatment arm as fixed effect, and subject as random effect. The coefficient represents the effect of Placebo on abundance. Virulence factors with estimated p‐value < 0.05 are marked with an asterisk *, none remained significant after FDR correction.

Figure S9: Scatter plots showing abundance over time for antimicrobial resistance genes differentially abundant between Placebo and Treatment arms at Week 12 using Mann–Whitney U‐test and p < 0.05. Each panel reports the p‐value and coefficient using a mixed with time and treatment arm as fixed effect, and subject as random effect. The coefficient represents the effect of Placebo on abundance. Antimicrobial resistance genes with estimated p‐value < 0.05 are marked with an asterisk *, none remained significant after FDR correction.

Table S1: Patient characteristics.

Table S2: Measured biomarkers.

Table S3: Meteor‐counting report.

Table S4: MSP abundance table.

Table S5: Abundance table of genes grouped by KO annotation.

Table S6: Gene abundance table of genes with homology to virulence factors.

Table S7: Virulence factors differential abundance.

Table S8: Correlations of virulence factors vs. serum biomarkers.

Table S9: Gene abundance table of genes annotated as antimicrobial resistance genes.

Table S10: Antimicrobial resistance genes differential abundance.

Table S11: Gene abundance table annotated secondary bile acid genes.

Table S12: Measured faecal secondary bile acids.

Table S13: Measured serum secondary bile acids.

Table S14: Metabolomics.

Liu J., Garcia‐Guevara F., Macnaughtan J., et al., “Impact of YAQ001 on the Gut Microbiome, Permeability and Markers of Systemic Inflammation in Patients With Cirrhosis,” Liver International 46, no. 11 (2026): e70933, 10.1111/liv.70933.

Data Availability Statement

The data that support the findings of this study are openly available in European Nucleotide Archive (ENA) at https://www.ebi.ac.uk/ena/browser/view/PRJEB104381.