Microbiome-Targeted Reduction of Circulating Trimethylamine N-Oxide Mitigates Ischemic Stroke Risk

1 Elevated plasma trimethylamine N-oxide (TMAO) is an independent predictor of major 2 adverse cardiovascular events and ischemic stroke. While inhibition of microbial TMA 3 production has been explored, concerns regarding off-target effects and limited efficacy in 4 complex microbial ecosystems have hindered clinical translation. Here, we report a 5 microbiome-based therapeutic strategy based on the direct enzymatic degradation of 6 intestinal TMA by Paracoccus aminovorans BM109. Through targeted screening, we 7 identified BM109 as a commensal strain harboring a comprehensive set of enzymes capable 8 of metabolizing TMA and TMAO into non-toxic end products under both aerobic and 9 anaerobic conditions. In a chronic high-choline diet murine model, oral administration of 10 BM109 resulted in a 38% reduction in systemic TMAO levels. In a rat model of transient 11 middle cerebral artery occlusion (tMCAO), short-term pre-treatment reduced cerebral infarct 12 size by 58% and significantly improved neurological outcomes. These effects were 13 accompanied by favorable safety observations, including the absence of hemolytic activity 14 and intestinal tissue damage. Collectively, our findings establish BM109 as a promising live 15 biotherapeutic product that targets the gut microbiome-host metabolic axis. By reducing the 16 systemic TMAO burden, BM109 represents a potential strategy for modulating 17 cardiometabolic and cerebrovascular risk. 18 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 3

1 The human body maintains its function and health through intricate metabolic processes 2 and reciprocal interactions with commensal microbes, collectively termed the microbiome (1, 3 2). Studies estimate that approximately 40 trillion microbial cells inhabit the human body (3), 4 outnumbering human cells in a ratio of approximately 1.3:1 (4). This vast microbial 5 community profoundly influences human health, functioning as an integral metabolic organ. 6 Advances in microbiome research have highlighted the role of the gut microbiome in a wide 7 range of systemic diseases beyond the gastrointestinal tract, including metabolic disorders 8 (5), neurodegenerative conditions (6), kidney dysfunction (7) and cardiovascular diseases 9 (CVDs) (8). 10 Among the diverse functions of the gut microbiome, its role in cardiovascular diseases 11 has gained increasing attention, particularly in conditions such as stroke (9), atherosclerosis 12 (10) and thrombosis (11). Beneficial gut microbes contribute to cardiovascular health by 13 producing short-chain fatty acids (SCFAs) (12, 13). SCFAs, particularly butyrate, produced 14 by species such as Roseburia intestinalis, exhibit atheroprotective effects, mitigating 15 inflammation and improving vascular function (14). In addition, certain gut bacteria exert 16 antioxidant and anti-inflammatory properties, preventing oxidative stress-related endothelial 17 damage (15, 16). 18 Recent studies indicate that gut microbial modification and cardiovascular disease can 19 mutually affect each other (17). Metabolomics studies have illustrated that patients with 20 cardiovascular disease display altered composition and diversity of gut microbiome, different 21 from the eubiotic state (18-20). In addition, several studies postulate that an increase or 22 decrease in specific bacteria can lead to changes in gut metabolites, such as bile acid or 23 trimethylamine (TMA), which can potentially threaten cardiovascular health (21-24). Studies 24 have demonstrated that trimethylamine (TMA), a gut microbiome–derived metabolite, 25 contributes to cardiovascular disease through its hepatic conversion to trimethylamine N-26 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 4 oxide (TMAO) by flavin-containing monooxygenase 3 (FMO3) (25). TMAO promotes 1 atherosclerosis and thrombosis by enhancing vascular inflammation, lesion formation, 2 immune cell activation, and platelet reactivity (23, 26). Mechanistically, TMAO exacerbates 3 endothelial dysfunction by triggering mitochondrial ROS production and inhibiting the Nrf2-4 mediated antioxidant defense system, leading to vascular inflammation and pyroptosis (27). 5 Clinically, elevated circulating TMAO levels are consistently associated with increased 6 cardiovascular risk and adverse outcomes, and have been identified as an independent 7 predictor of major adverse cardiovascular events (MACE) across multiple cohort studies (28-8 34). Elevated TMAO levels pose additional health risks beyond cardiovascular pathology. In 9 patients with chronic kidney disease (CKD), impaired renal clearance results in persistently 10 high plasma TMAO levels, exacerbating cardiovascular complications, one of the leading 11 causes of mortality in this population (7, 35, 36). Despite the strong and consistent 12 associations observed in epidemiological studies, interventional approaches aimed at 13 lowering circulating TMAO levels remain limited. 14 A major challenge in targeting gut-derived TMA and TMAO lies in their continuous 15 production within the gut environment, largely independent of microbiome composition. A 16 wide range of bacterial taxa, including members of the genera Clostridium, Shigella, 17 Klebsiella, and Citrobacter, are capable of metabolizing dietary substrates such as choline, 18 betaine, carnitine and related compounds to generate TMA (37, 38). Given that TMA-19 producing microbes are taxonomically diverse and widely distributed within the gut, selective 20 eradication is unfeasible. Additionally, dietary restriction of TMA precursors such as choline, 21 betaine, lecithin, and carnitine is impractical, as these nutrients play important roles in 22 human physiology. 23 Since direct suppression of TMA biosynthesis remains challenging, an alternative 24 strategy is to degrade TMA post-production to prevent its systemic absorption. In this study, 25 we sought to identify commensal gut bacteria with intrinsic TMA-degrading capabilities. 26 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 5 Through targeted screening of human fecal microbiota, we isolated a commensal species 1 exhibiting robust enzymatic activity against TMA and TMAO. We characterized its metabolic 2 pathways, assessed its efficacy in reducing TMAO levels under aerobic and anaerobic 3 conditions, and validated its in vivo therapeutic potential using both a diet-induced and a 4 disease-relevant animal model. This study provides a proof-of-concept for a microbiome-5 based TMAO-lowering therapy through direct degradation of microbial TMA. 6 7 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 6

1 Isolation and Identification of a Robust TMA-Degrading Bacterial Species from Human 2 Feces. Methylotrophic bacteria are known to utilize C1 compounds such as methane and 3 methanol, and can also metabolize trimethylamine (TMA) and trimethylamine N-oxide 4 (TMAO), which contain three methyl groups (39). Based on this metabolic feature, we sought 5 to isolate fecal bacteria with TMA-degrading capacity. 6 As outlined in Fig. 1A, pooled human fecal samples were screened under conditions 7 favoring methylotrophic growth, yielding multiple candidate colonies. These isolates were 8 subsequently evaluated for their ability to degrade TMA. Among the screened candidates, 9 several strains exhibited measurable TMA consumption, with a subset demonstrating robust 10 degradation activity after 24 h incubation (Fig. 1B). These strains were selected for further 11 characterization, and TMA levels were quantified using an optimized picric acid-based assay 12 (40), as described in Methods. 13 Nine strains were selected as primary candidates based on their consistent growth on 14 MeOH-containing plates. Their TMA-degrading capacities were evaluated by quantifying 15 residual TMA levels relative to the initial concentration. Among these, seven strains 16 degraded more than 50% of TMA, although their activities varied (Fig. 1B). Notably, strain 9 17 exhibited near-complete TMA degradation within 24 h under aerobic conditions. In contrast, 18 strains 3 and 5 showed minimal TMA degradation despite robust colony formation, indicating 19 that growth under methylotrophic conditions does not necessarily correlate with TMA-20 degrading activity in liquid culture. 21 Species identification based on 16S rRNA gene sequencing revealed that strains 1, 2, 22 and 7 were Klebsiella aerogenes, strain 8 was Escherichia coli, and strain 4 was 23 Pseudomonas aeruginosa (Fig. 1C), all of which are known opportunistic pathogens (41, 24 42). In contrast, strain 9 (hereafter referred to as BM109) was identified as Paracoccus 25 aminovorans (99.91% sequence identity to JCM7865), a member of the alpha-26 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 7 Proteobacteria distinct from the other isolates (43). Notably, Paracoccus species have been 1 reported to be present in the human gastrointestinal tract (44, 45). 2 To evaluate TMA-degrading efficacy under gut-relevant conditions, BM109 was cultured 3 in LB medium and assessed under both aerobic and anaerobic conditions. BM109 degraded 4 ~95% of TMA within 24 h under aerobic conditions (Fig. 1D). Under anaerobic conditions 5 with nitrate as an alternative electron acceptor, BM109 degraded ~97.9% of TMA within 48 h 6 (Fig. 1E), despite reduced growth. These results demonstrate robust TMA-degrading activity 7 across oxygen conditions, supporting its potential functionality in the gut environment. As a 8 control, Paracoccus denitrificans, a related species within the Paracoccus genus, exhibited 9 markedly lower TMA-degrading activity, highlighting the distinct metabolic capability of 10 BM109. 11 Genomic Structure and Comparative Analysis of BM109. The entire genome of BM109 12 was sequenced and assembled into four contigs, two of which contained both tRNA and 13 rRNA gene sequences. The sequencing generated a total of 14,415,666 reads with an 14 average depth of 473.9, and the total genome size was determined to be 4,297,217 bp. 15 Genome annotation revealed 4,095 coding sequences (CDS) across all four contigs, with 16 detailed features summarized in Table 1. The genome sequence data have been deposited 17 in the NCBI BioProject database under accession number PRJNA891425. 18 Comparative genomic analysis revealed high 16S rRNA sequence similarity (99.91%) 19 between BM109 and Paracoccus aminovorans JCM7685 (Fig. 1C), but substantial 20 divergence at the whole-genome level. Only 7.47% of contig 1 and limited regions of contig 2 21 aligned with the reference JCM7685 genome (see Supplementary Information), indicating 22 significant genomic variation. These findings suggest that BM109 represents a distinct strain 23 within the P. aminovorans species. The overall genome structure of BM109 was similar to 24 that of JCM7685, consisting of four contigs with contig 1 as the largest. However, notable 25 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 8 differences in GC content were observed. In particular, plasmid III of BM109 exhibited a 1 markedly lower GC content (56.9%) compared to the other contigs, whereas JCM7685 2 showed relatively uniform GC content across its genome. This discrepancy suggests that 3 BM109 may have acquired plasmid III through horizontal gene transfer from an external 4 source. 5 Functional Genomics of TMA/TMAO Metabolism. Our genome annotation analysis 6 revealed multiple enzymes encoded in BM109's genome that are associated with the 7 metabolism of TMA and TMAO, as well as their downstream products. These findings 8 enabled us to construct a functional metabolic pathway for the sequential degradation of 9 TMA and TMAO into non-toxic compounds, including CO2 (Fig. 2A). 10 Two genes located on contig 2 were annotated as encoding TMA dehydrogenase 11 (TMADH, 2_00604 and 2_00975), an enzyme responsible for converting TMA into 12 dimethylamine (DMA) and formaldehyde (Fig. 2B, highlighted in blue). The presence of two 13 genes encoding the same enzymatic function indicates the importance of genetic 14 redundancy in maintaining critical metabolic functions. The enzymes labeled as "1" and "2" 15 in Fig. 2A correspond to TMA monooxygenase (TMM) and TMAO reductase (TMAORD), 16 respectively (Fig. 2B, highlighted in green). BM109 encodes both enzymes, enabling it to 17 oxidize TMA into TMAO and reduce TMAO back into TMA. This dual functionality allows 18 BM109 to maintain a dynamic TMA/TMAO balance, ensuring metabolic flexibility. 19 The hydrolysis of TMA produces formaldehyde, a toxic byproduct that must be further 20 metabolized to avoid harm. BM109 contains a gene encoding formaldehyde dehydrogenase 21 (FADH, 1_00675), which catalyzes the conversion of formaldehyde to formate, a less toxic 22 intermediate (Fig. 2B, highlighted in orange). Furthermore, two gene clusters encoding 23 formate dehydrogenase catalyze the oxidation of formate into CO2 (Fig. 2B, highlighted in 24 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 9 gray). This sequential pathway ensures complete detoxification of formaldehyde, 1 underscoring BM109’s ability to mitigate TMA-derived toxicity. 2 A gene encoding methanol dehydrogenase (1_02795) was also identified in the BM109 3 genome, explaining its ability to metabolize methanol to CO2 (Fig. 2B, highlighted in yellow). 4 This metabolic feature likely contributed to BM109’s growth under methanol-rich conditions 5 during the initial screening process. Methanol metabolism also produces NADH (46), a 6 molecule essential for energy production via respiratory pathways. This capacity likely 7 contributed to BM109’s growth under methanol-rich conditions during the initial screening 8 process. 9 To evaluate the uniqueness of BM109’s genetic repertoire, we conducted a comparative 10 genomic analysis using BLAST searches. Based on the reconstructed TMAO-to-CO2 11 metabolic pathway (Fig. 2A), we defined a set of 22 proteins encoded by BM109 that are 12 involved in TMA and TMAO metabolism. Notably, this set includes enzymes encoded by 13 multi-gene operons, in which individual gene products were considered separately, resulting 14 in a total of 22 proteins for analysis. The analysis identified 15 Proteobacteria species 15 harboring homologous or similar sequences to these proteins, including nine species within 16 the Alpha-proteobacteria (Fig. 3). However, none of the examined species possessed the 17 complete set of genes required for the full TMAO-to-CO2 metabolic pathway. Notably, even 18 P. aminovorans JCM7685, the closest relative of BM109, lacked the gene encoding TMAO 19 reductase, which was further confirmed by PCR analysis (Fig. S1). 20 Among the Alpha-proteobacteria, P. limosus exhibited partial coverage of the pathway, 21 encoding more components than most related species, including P . denitrificans, but still 22 lacking key enzymes required for complete TMA/TMAO metabolism. Similarly, Beta- and 23 Gamma-proteobacteria species possessed only incomplete subsets of the pathway, missing 24 essential enzymes for full metabolic conversion (Fig. 3). 25 26 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 10 Collectively, these results demonstrate that BM109 uniquely harbors a complete and 1 functionally integrated genetic repertoire for TMA and TMAO metabolism, distinguishing it 2 from other Proteobacteria species. 3 Oral delivery of BM109 reduces TMA and TMAO levels in vivo. To evaluate the in vivo 4 efficacy of BM109, mice were orally administered BM109 while being fed a normal chow diet 5 (Normal), high-choline diet (HCD), or high-fat diet (HFD) (Fig. 4A). Mice were maintained on 6 these diets for 45 weeks to model long-term dietary exposure. 7 As shown in Fig. 4B–E, HFD had no significant effect on TMA or TMAO levels. In 8 contrast, HCD feeding markedly increased both metabolites in feces and serum, establishing 9 a suitable model to assess the effects of BM109. In HCD-fed mice, fecal TMA levels were 10 significantly elevated and were reduced by 83% following five days of BM109 administration 11 (Fig. 4B). Similarly, fecal TMAO levels were substantially decreased after BM109 treatment 12 (Fig. 4C), indicating efficient degradation of TMA within the intestinal environment. 13 We next examined TMA and TMAO levels in mouse serum to assess systemic effects of 14 BM109. In HCD-fed mice, the TMAO concentration reached ~3,666 ng/mL (Fig. 4E), while 15 the TMA level was ~1,473 ng/mL (Fig. 4D). Interestingly, the ratio of TMAO to TMA in serum 16 was opposite to that observed in feces (~56,931 ng/mL vs. ~121,724 ng/mL). This result 17 further confirms that a significant amount of blood TMAO is produced from the hepatic 18 enzyme flavin-containing monooxygenase 3 (FMO3) using intestinal TMA absorbed through 19 the portal vein and delivered to the liver (25). Upon daily treatment with BM109 cells, blood 20 levels of TMA and TMAO decreased by ~30% and ~38%, respectively, in the HCD-fed group 21 (Fig. 4D and E). Furthermore, TMAO levels in the Normal diet group also decreased by 22 ~40% after BM109 treatment (Fig. 4E). 23 Notably, BM109 retained its TMA-degrading activity in vivo without supplementation of 24 exogenous electron acceptors, suggesting that the gut environment provides sufficient 25 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 11 conditions, potentially including endogenous or diet-derived nitrate to support its metabolic 1 function. Collectively, these findings demonstrate that BM109 effectively degrades intestinal 2 TMA and reduces circulating TMAO levels, supporting its therapeutic potential for TMAO-3 associated metabolic disorders. 4 Neuroprotective Effects of BM109 in the tMCAO Rat Model. Building upon the findings in 5 the HCD diet mouse model, we next evaluated the therapeutic efficacy of BM109 in a 6 disease-specific context using a transient middle cerebral artery occlusion (tMCAO) rat 7 model. This experiment aimed to demonstrate BM109’s potential in mitigating ischemic injury 8 associated with elevated TMAO levels. As illustrated in Fig. 5A, rats were divided into groups 9 based on diet and treatment conditions. BM109 was administered orally for seven days prior 10 to the tMCAO surgery. 11 To assess the impact of diet on blood TMAO levels, we measured TMAO 12 concentrations after two weeks of feeding either a normal diet or HCD. HCD-fed animals 13 exhibited a marked increase in blood TMAO levels compared to their baseline levels before 14 HCD feeding (Fig. 5B). This trend was consistent with observations in mice (Fig. 4E). 15 Notably, among the 58 animals fed an HCD, 33 exhibited a substantial increase in blood 16 TMAO levels, with at least a three-fold elevation, as indicated by the double asterisks in Fig. 17 5B. To further evaluate whether BM109 treatment could lower systemic TMAO levels, we 18 analyzed this subset of animals with markedly elevated TMAO. As depicted in Fig. 5C, 19 BM109 administration resulted in a statistically significant reduction in blood TMAO levels (p 20 = 0.0057), whereas the control group showed no significant changes (p = 0.1686). These 21 findings further support the in vivo efficacy of intestinally delivered BM109 in reducing diet-22 induced TMAO accumulation. 23 Given this effect, we next examined how this reduction influenced the outcomes of 24 tMCAO. As shown in Fig. 5D, BM109 treatment led to a significant reduction in infarct size 25 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 12 across both diet groups. In animals fed a normal diet, BM109 administration reduced infarct 1 size by approximately 22% compared to controls (p = 0.049). This effect was even more 2 pronounced in HCD-fed rats, where BM109 treatment decreased infarct size by ~58% (30.01 3 ± 4.67 vs. 12.59 ± 3.40, p < 0.0001) compared to PBS-treated controls. These results 4 highlight the potential of BM109 as a therapeutic strategy for mitigating TMAO-induced 5 ischemic injury. Representative TTC-stained images from each group further illustrate these 6 differences, visually highlighting the neuroprotective effects of BM109 (Fig. S2). 7 Behavioral assessments further validated BM109’s efficacy. Garcia’s scores revealed 8 significant motor and sensory improvements in BM109-treated rats, particularly in the HCD 9 group (p = 0.0003, Fig. 6A). While Longa’s scores did not show significant differences (Fig. 10 6B), the modified neurological severity score (mNSS) confirmed BM109’s role in reducing 11 neurological deficits. Specifically, mNSS scores were significantly lower in BM109-treated 12 HCD-fed rats compared to PBS-treated controls (p = 0.0025, Fig. 6C). Together, these 13 findings demonstrate that BM109 effectively reduced HCD-induced TMAO elevation, leading 14 to substantial neuroprotection in the tMCAO model. By lowering systemic TMAO and 15 mitigating ischemic injury, BM109 shows promise as a therapeutic agent for TMAO-16 associated ischemic stroke. 17 Biosafety assessment of BM109. To evaluate BM109 as a potential live biotherapeutic 18 product (LBP), comprehensive biosafety assessments were conducted. First, hemolytic 19 activity was examined, and BM109 showed no hemolysis on blood agar (Fig. S3A). Second, 20 gastrointestinal safety was assessed in mice administered either PBS or BM109 (n=5). No 21 signs of inflammation, fluid accumulation, or tissue damage were observed, and colonic 22 length remained unchanged (Fig. S3B). Histological analysis further confirmed BM109’s 23 safety, as H&E staining revealed no significant differences in epithelial integrity, crypt 24 architecture, or inflammatory cell infiltration between PBS- and BM109-treated groups (Fig. 25 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 13 S3C). Finally, antibiotic resistance was evaluated against eight clinically and agriculturally 1 relevant antibiotics, including ampicillin, streptomycin, and vancomycin (Fig. S4). BM109 2 was sensitive to all tested antibiotics, although slight resistance to vancomycin was 3 observed. However, considering that BM109 is a Gram-negative bacterium, this is not a 4 significant concern, as vancomycin primarily targets Gram-positive organisms (47). 5 Collectively, these results confirm BM109’s safety as an oral LBP, demonstrating no 6 concerning antibiotic resistance and strong potential for mitigating TMA- and TMAO-7 associated disorders. 8 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 14

1 TMAO has been suggested as an independent risk factor and effective prognostic 2 marker for various cardiovascular and cerebrovascular diseases. In prospective 3 observational studies conducted in many countries, patients with higher blood TMAO levels 4 invariably exhibited a higher tendency to develop adverse long-term cardiovascular risk in 5 the USA (29, 48), Korea (30), Japan (34), China (49), Norway (50) and Spain (51). Notably, 6 TMAO showed a stronger correlation power than other blood test items, such as LDL-7 cholesterol and triglycerides, predicting future CVD events (29, 30, 48, 49). While the 8 predictive power of TMAO differs based on race of the target patients (48, 52), these results 9 demonstrate that TMAO is a metabolite whose amounts need to be decreased for the 10 prevention and treatment of CVD. 11 TMA, a precursor of TMAO, is enzymatically produced by commensal bacteria that 12 reside in the intestine. Choline-TMA lyase and carnitine oxygenase encoded by cutC and 13 cntA genes are mainly involved in this process (53, 54). Based on comprehensive 14 bioinformatics analysis, 1,107 and 6,783 out of 67,134 bacterial genomes contain cutC and 15 cntA genes, respectively (37). This analysis indicated that the cutC gene, in particular, has a 16 very low distribution across the entire bacterial kingdom. However, quantitative gene 17 amplification assays using 50 different human stool samples revealed that the cutC gene 18 was detected in all samples, whereas the cntA gene was detected in 26% of the samples 19 (37). This result suggests that the human intestine requires the presence of cutC and cntA 20 genes of microbial origin, with cutC at a higher frequency, to facilitate enhanced digestion 21 during the process of establishing symbiosis. From an ecological and evolutionary 22 perspective, this observation suggests that the human microbiome has co-evolved toward a 23 symbiotic relationship that favors TMA production, reflecting its adaptive significance in 24 optimizing digestive processes and nutrient metabolism. 25 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 15 Efforts to inhibit microbial TMA production have garnered significant attention due to its 1 contribution to systemic TMAO formation. Among these, a structural analog of choline, 3,3-2 dimethyl-1-butanol (DMB), has been extensively studied for its suppressive activity against 3 choline-TMA lyase (55). In animal studies, DMB treatment led to decreased TMAO levels 4 and subsequent amelioration of TMAO-induced adverse phenotypes (56, 57). However, 5 repeated treatment with DMB induces neurotoxic effects in adult mice (58). Furthermore, 6 DMB treatment failed to suppress TMA production in an in vitro model of human colon 7 fermentation (59). More recently, iodomethylcholine (IMC), a mechanism-based inhibitor of 8 choline-TMA lyase, has been shown to effectively reduce TMA and TMAO levels in 9 preclinical models by irreversibly inhibiting microbial TMA formation (60-62). However, such 10 approaches rely on broad enzymatic inhibition and may be limited by off-target effects and 11 challenges in selectively modulating complex gut microbial communities. Collectively, these 12 findings suggest that inhibition-based strategies may not fully address TMA production in the 13 complex gut environment. In contrast, our strategy focuses on the direct degradation of TMA 14 by a commensal bacterium, offering a distinct and potentially more targeted approach to 15 reducing systemic TMAO levels. 16 Lawrence et al. reported an association between circulating bacterial DNA profiles and 17 cardiovascular mortality in a cohort of 405 individuals (45). Blood samples obtained from 18 patients who died from CVD contained significantly lower levels of bacterial DNA originating 19 from the genus Paracoccus (p < 0.001). Since circulating bacterial DNA may originate, at 20 least in part, from the translocation of bacteria from the gut, this finding suggests that these 21 patients may have had lower intestinal abundance of Paracoccus. These observations are 22 consistent with our finding that a Paracoccus strain (BM109) exerts beneficial effects on 23 TMA metabolism and cardiovascular-related outcomes. While causality cannot be 24 established, this association raises the possibility that Paracoccus abundance may be linked 25 to cardiovascular health. 26 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 16 Our results (Fig. 4) demonstrate that both TMA and TMAO levels were markedly 1 increased in fecal samples from HCD-fed mice. This observation raises the possibility that 2 TMAO may be present within the intestinal environment, although its precise origin remains 3 unclear. While TMAO is primarily generated in the liver via flavin-containing monooxygenase 4 3 (FMO3), previous studies have suggested that certain gut bacteria possess TMA 5 monooxygenase activity, which could contribute to local TMAO formation (63). Notably, 6 BM109 uniquely encodes TMAO reductase, which catalyzes the reduction of TMAO to TMA, 7 followed by further metabolism via TMA dehydrogenase (Fig. 3). This metabolic capability 8 suggests that BM109 may reduce intestinal TMAO levels, thereby limiting its absorption into 9 the bloodstream. 10 Paracoccus species, including the BM109 strain explored in the current study, are 11 methylotrophic (64, 65). Consistent with this metabolic capability, BM109 degraded 10 mM 12 TMA within 24 h in vitro, a concentration exceeding typical physiological levels. This high 13 degradation capacity may be supported by downstream metabolic pathways that convert 14 toxic intermediates such as formaldehyde into formate and CO2. BM109 exhibited optimal 15 growth at 37 °C and retained activity under anaerobic conditions, supporting its potential 16 functionality within the intestinal environment. Genomic analysis further indicated the 17 presence of genes associated with nitrate-dependent anaerobic respiration, suggesting that 18 BM109 may utilize alternative electron acceptors available in the gut. Together with its 19 complete set of TMA and TMAO metabolic genes, these features support the ability of 20 BM109 to efficiently degrade TMA and TMAO under physiologically relevant conditions. 21 To evaluate the in vivo efficacy of BM109, we employed a high-choline diet (HCD) 22 model, which is known to elevate TMA and TMAO levels. A high-fat diet (HFD) model was 23 also included to assess the impact of dietary fat alone; however, HFD did not significantly 24 alter TMA or TMAO levels. BM109 administration effectively reduced systemic TMAO levels 25 by directly degrading intestinal TMA, thereby intercepting the microbial-host metabolic axis 26 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 17 linking gut microbiota to cardiovascular risk. Notably, plasma TMAO levels were consistently 1 decreased under both HCD and normal diet conditions following oral administration of 2 BM109. Short-term treatment (5~7 days) was sufficient to achieve significant reductions in 3 TMAO levels and was associated with improved outcomes in the tMCAO model. These 4 findings support the potential of BM109 as a microbiome-based therapeutic strategy for 5 reducing systemic TMAO levels. Future studies are warranted to evaluate the long-term 6 efficacy and durability of BM109 treatment in vivo. 7 Previous studies have reported a positive correlation between circulating TMAO levels 8 and infarct severity in experimental stroke models (66). Consistent with these findings, our 9 rat experiments demonstrated a relationship between elevated TMAO levels and increased 10 infarct size, which was most pronounced under high-choline diet (HCD) conditions. In these 11 models, dietary choline intake increased plasma TMAO levels, recapitulating TMAO-12 associated ischemic risk observed in humans. Prior work has also shown that enhancing 13 microbial TMA production, for example through microbiota transplantation enriched in 14 choline-utilizing pathways, exacerbates ischemic injury, whereas disruption of TMA-15 producing enzymes reduces TMAO levels and improves outcomes. While these approaches 16 primarily aim to reduce TMAO levels by limiting microbial TMA production, our strategy is 17 fundamentally distinct. BM109 directly degrades TMA after its formation, thereby reducing 18 the substrate available for hepatic conversion to TMAO. This mechanism targets the 19 upstream microbial metabolite pool, offering a complementary and potentially more efficient 20 approach to lowering systemic TMAO levels. By enzymatically eliminating TMA, BM109 21 represents a microbiome-based strategy to mitigate TMAO-associated ischemic injury. 22 Integration of these findings with prior studies supports the potential of BM109 as a 23 microbiome-based therapeutic strategy. In addition to lowering circulating TMAO levels, 24 BM109 conferred significant neuroprotective effects in ischemic models, linking metabolic 25 modulation to improved disease outcomes. Given the high risk of stroke recurrence (67, 68), 26 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 18 targeting TMAO may represent a viable approach for reducing recurrent ischemic events. 1 Accordingly, BM109 may have therapeutic potential not only in acute settings but also in 2 long-term risk management. Further studies are warranted to evaluate its durability and 3 clinical applicability in both primary and secondary prevention of stroke. 4 Although BM109 was isolated from human feces, members of the genus Paracoccus 5 are commonly associated with environmental niches such as soil (69). The use of non-6 traditional or environmental bacteria for therapeutic purposes has been increasingly 7 explored, particularly when a defined mechanism of action can be demonstrated (70). In 8 such contexts, functional activity, together with safety considerations, is a key determinant of 9 therapeutic potential. Our findings demonstrate that BM109 effectively degrades TMA and 10 reduces circulating TMAO levels in vivo, supporting its potential as a microbiome-based 11 strategy for modulating TMAO-associated cardiovascular and cerebrovascular risk. Further 12 studies are warranted to evaluate its long-term efficacy and clinical applicability. 13 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 19

1 Ethical Statements. All animal experiments in this study were performed in compliance with 2 the guidelines established by the Department of Animal Resources at the Yonsei Biomedical 3 Research Institute. This study was approved by the Institutional Animal Care and Use 4 Committee (IACUC) of Yonsei University College of Medicine under permit numbers 2024-5 0143 and 2021-0139 and 2023-0063. 6 Screening human fecal microbiome for bacteria with TMA-degrading capabilities. A 7 collection of human feces (Severance Hospital, IRB approval number, 4-2016-0850) was 8 used to yield pools of microbiome suspensions. Aliquots of fecal bacterial suspensions were 9 inoculated onto agar plates of DLB (Diluted Luria Bertani) plus 7% MeOH. DLB medium was 10 prepared by 20% dilution of LB medium [1% (w/v) tryptone, 0.5% (w/v) yeast extract, and 1% 11 (w/v) sodium chloride] in water. At 48 h post-inoculation, the colonies were recovered as 12 potential strains with methylotrophic features. Each colony was then inoculated in a broth 13 medium containing DLB+5 mM TMA to examine its growth and TMA-degrading capability. 14 After cultivation, the spent medium was collected for TMA quantification. Potential TMA 15 degraders were identified by sequencing the entire region of the 16S rRNA gene. 16 TMA degradation challenge with additional nutrition. To determine whether the BM109 17 strain could still degrade TMA in the presence of additional carbon and nitrogen sources, 18 BM109 was cultivated in LB broth supplemented with 10 mM TMA. NaNO3 (60 mM) was 19 added as an alternative electron acceptor to stimulate anaerobic growth. For aerobic 20 incubation, BM109 cells were grown by shaking (180 rpm) at 37 °C. Anaerobic cultivation 21 was achieved in an anaerobic chamber (Coy Lab Products, Grass Lake, MI), where a gas 22 mix (5% H2 and 95% N2) was used to fill the atmosphere. BM109 was grown statically inside 23 an anaerobic chamber. We harvested the aerobic culture after 24 h and the anaerobic 24 culture after 48 h to quantify the residual TMA. We also cultured Paracoccus denitrificans 25 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 20 ATCC17741 (PD) using the same protocol as a negative control. OD600nm values of each 1 bacterial culture were measured after incubation. 2 Colorimetric TMA quantification assay. Colorimetric TMA quantification was performed 3 using a modified picric acid-toluene assay (71, 72). Serially diluted TMA standards (10~1.25 4 mM) were prepared in LB for the standard curve. BM109 and PD cultures were centrifuged, 5 and supernatants were processed with formaldehyde, toluene, and potassium carbonate. 6 After vortexing and shaking, the toluene phase was mixed with picric acid-toluene solution, 7 and absorbance at 410 nm was measured using a microplate reader. 8 Whole-genome sequencing of BM109 and genome analysis. Genomic DNA was purified 9 using G-spin Total DNA Extraction kit (iNtRON). Genome sequencing was performed using 10 PacBio Sequel and Illumina NovaSeq6000 platform. DNA is required to prepare size-11 selected approximately 15 kb SMRTbell templates with barcoded overhang adapters. For 12 PacBio Sequel sequencing, library was prepared using PacBio SMRTbell Express Template 13 Prep kit 2.0. The Sequel Sequencing Kit 3.0 and SMRT cells 1M v3 Tray was used for 14 sequencing. SMRT cells (Pacific Biosciences) using 600 min movies were captured for each 15 SMRT cell using the PacBio Sequel (Pacific Biosciences) sequencing platform by Macrogen 16 (Seoul, Korea). Reads from PacBio Sequel system were assembled using SMRTlink 17 10.1.0.119588 (73). Illumina raw reads were used for error correction. The assembly was 18 corrected using high quality Illumina reads by Pilon v1.21 (74). The whole-genome 19 sequence data for BM109 have been deposited in GenBank under accession number 20 PRJNA891425. 21 To investigate the TMAO metabolic pathway in the BM109 genome and its closely 22 related species, we employed a comprehensive analytical approach. Gene prediction and 23 annotation were performed using Prokka (version 1.14), InterProScan (version 5.30-69), and 24 PSI-BLAST (version 2.4.0), with the EggNog database (version 4.5) serving as a reference 25 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 21 for functional classification. To examine the BM109 genome in comparison with closely 1 related species, we retrieved genome sequences of related species from the NCBI 2 database. These genomes were subsequently analyzed using BlastKOALA, which enabled 3 the identification of genes relevant to the TMAO pathway. 4 The identified genes from these genomes were then aligned with the gene set of 5 BM109, allowing for the identification of conserved elements across species. This 6 comparative analysis provided key insights into the genetic basis of the TMAO pathway in 7 BM109 and its evolutionary relatives, highlighting conserved genetic features that may 8 underlie the functionality and adaptability of this pathway. 9 Chronic Dietary Model in Mice. For the chronic dietary model, C57BL/6J female mice (5 10 weeks old) were divided into three groups of ten mice each and fed one of the following diets 11 for 45 weeks: a chemically defined normal-chow diet (NC, 0.08% wt/wt total choline), the 12 same chemically defined diet supplemented with 1% wt/wt choline (HCD, Envigo 13 TD.130328), or a chemically defined 60% wt/wt fat diet (HFD, Envigo TD.06414). Female 14 mice were used in this study because hepatic expression of flavin-containing 15 monooxygenase 3 (FMO3), the key enzyme responsible for TMAO production, is markedly 16 reduced in male mice, resulting in significantly lower TMAO levels (75). After 45 weeks, each 17 group was split into subgroups of 5 mice. Control groups were gavaged with 100 μL of tap 18 water daily for five consecutive days. Experimental groups received 100 μL of BM109 19 suspension, which was prepared by harvesting activated BM109 cells after 24 h of 20 incubation, concentrating them to 5×10⁹ CFU/mL, washing the pellet, and resuspending it in 21 PBS before administration. 22 Transient Middle Cerebral Artery Occlusion (tMCAO) Model in Rats. For the tMCAO 23 model, male Wistar rats (8 weeks old) were randomly assigned to eight groups based on diet 24 (chow or choline-supplemented) and treatment (BM109 or vehicle) with or without tMCAO 25 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 22 surgery. The study followed ARRIVE guidelines, and sample size was determined using 1 MedCalc software. Focal ischemia was induced via tMCAO under isoflurane anesthesia 2 following procedures described previously (76). A monofilament was introduced through the 3 external carotid artery to occlude the middle cerebral artery for 90 minutes, after which 4 reperfusion was allowed. Sham-operated controls underwent the same procedure without 5 occlusion. Neurological function was assessed 24 h post-surgery using the Garcia test (77), 6 Longa score (78), and modified Neurological Severity Score (mNSS) (79). Brains were 7 collected 24 h post-surgery, sectioned, and stained with TTC. Infarcted areas were 8 quantified using image analysis software, calculating infarct volume as a percentage of total 9 brain area (80). Samples were collected at four time points: before diet administration, 10 before BM109 supplementation, before surgery, and before euthanasia. Plasma was 11 isolated for TMAO quantification. 12 Quantification of TMA and TMAO. 13 Blood and fecal samples in Figure 4 were collected 24 h after the final gavage. Blood 14 plasma was obtained by centrifugation, while fecal samples were homogenized in PBS, 15 centrifuged, and the supernatant was collected. Both plasma and fecal supernatants were 16 diluted in methanol (1:4 vol/vol), centrifuged, and the supernatants were analyzed for TMA 17 and TMAO using liquid chromatography-mass spectrometry (LC-MS) at the National 18 Instrumentation Center for Environmental Management (NICEM), Seoul National University. 19 The results were compared across six groups (NC, HC, HF), each with control and 20 experimental subgroups. Additionally, TMAO quantification in rat blood samples was 21 conducted at the Department of Laboratory Medicine, Severance Hospital following 22 established procedures (30). 23 Assessment of BM109 safety. 24 The hemolytic activity of BM109 was assessed by streaking the strain on a blood agar 25 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 23 plate, as described in a previous study (81). Intestinal and colon tissues were collected from 1 C57BL/6J mice following oral administration of BM109 (5×10⁸ CFU/day for five days) or 2 PBS. The tissues were fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned 3 for histological analysis. Hematoxylin and eosin (H&E) staining was performed to evaluate 4 potential tissue damage, inflammation, or histological abnormalities, following the methods 5 outlined elsewhere (82). The antibiotic susceptibility of BM109 was evaluated using the E-6 test method, following the Clinical and Laboratory Standards Institute (CLSI) guidelines (83). 7 Statistical analysis 8 Statistical analyses were performed using GraphPad Prism (version 10.4.1, GraphPad 9 Software, San Diego, CA, USA). Data are presented as mean ± standard deviation (SD), 10 and a p-value < 0.05 was considered statistically significant. Differences between paired 11 samples were analyzed using a paired t-test, while comparisons between multiple groups 12 were assessed using one-way and two-way ANOVA with Bonferroni post hoc tests. 13 Repeated measures mixed models were applied where appropriate. 14 15 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 24

1 This work was supported by grants from the National Research Foundation (NRF) of Korea, 2 funded by the Korean Government (2022M3A9F3017506 and 2019R1A6A1A03032869). 3 This research was supported by Korea Drug Development Fund funded by Ministry of 4 Science and ICT, Ministry of Trade, Industry, and Energy, and Ministry of Health and 5 Welfare (RS-2024-00443597). This research was also supported by the National Institute of 6 Health (NIH) research project (2024-ER2107-00). This work was supported by the National 7 Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. 8 2022R1A2C1007948). 9 Author contributions 10 JSY, CEY , and JBK contributed equally as co-first authors. All authors participated in 11 experimental design and data interpretation. JSY , CEY, JBK, MAA, and SGK performed the 12 experiments and analyses. CEY, YBK, and HSN conducted the in vivo stroke model 13 experiments and behavioral assessments. JSY , CEY, JBK, and SSY drafted the manuscript. 14 HSN and SSY supervised the study as co-corresponding authors. 15 Competing interests 16 BM109 is covered under a patent held by BioMe Inc. Sang Sun Yoon, a corresponding author, 17 is a professor at Yonsei University and the CEO of BioMe Inc., which is actively involved in the 18 development of BM109. The authors declare no other competing interests. 19 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 25

1 1. J. S. You et al., Commensal-derived metabolites govern Vibrio cholerae pathogenesis 2 in host intestine. Microbiome 7, 132 (2019). 3 2. M. Y . Yoon et al., A single gene of a commensal microbe affects host susceptibility to 4 enteric infection. Nat Commun 7, 11606 (2016). 5 3. R. Sender, S. Fuchs, R. Milo, Revised Estimates for the Number of Human and 6 Bacteria Cells in the Body. PLoS Biol 14, e1002533 (2016). 7 4. E. Bianconi et al., An estimation of the number of cells in the human body. Ann Hum 8 Biol 40, 463-471 (2013). 9 5. Y. Ding, Z. Zhang, K. Wang, C. Jiang, The microbiome regulates host metabolic health 10 and diseases through microbial enzymes. Nat Rev Gastroenterol Hepatol, (2026). 11 6. Y. Y . Ma, X. Li, J. T. Yu, Y. J. Wang, Therapeutics for neurodegenerative diseases by 12 targeting the gut microbiome: from bench to bedside. Transl Neurodegener 13, 12 13 (2024). 14 7. M. Wang et al., The Gut Microbial Metabolite Trimethylamine N -oxide, Incident CKD, 15 and Kidney Function Decline. J Am Soc Nephrol 35, 749-760 (2024). 16 8. Y. Makkieh et al. , The gut -heart axis: Exploring the role of the gut microbiome in 17 cardiovascular health - A focused systematic review. Am Heart J Plus 61, 100687 18 (2026). 19 9. I. Szegedi, D. Bomberak, Z. Eles, L. Loczi, Z. Bagoly, Cardiovascular disease and 20 microbiome: focus on ischemic stroke. Pol Arch Intern Med 135, (2025). 21 10. S. Feng, Y. Jiang, J. Jiang, H. Bian, R. Zhu, Effects of diet -modulated gut microbiota 22 and microbial metabolites in atherosclerosis. Biomed Pharmacother 198, 119328 23 (2026). 24 11. M. P . Khuu et al., The gut microbiota in thrombosis. Nat Rev Cardiol 22, 121-137 (2025). 25 12. A. Piccioni et al., Gut Microbiota and Environment in Coronary Art ery Disease. Int J 26 Environ Res Public Health 18, (2021). 27 13. H. X. Yang et al., Gut microbiota-derived butyrate prevents aortic dissection via GPR41. 28 Acta Pharmacol Sin 46, 3230-3243 (2025). 29 14. K. Kasahara et al. , Interactions between Roseburia intestina lis and diet modulate 30 atherogenesis in a murine model. Nat Microbiol 3, 1461-1471 (2018). 31 15. K. T. Kim et al., Antioxidant and Anti-Inflammatory Effect and Probiotic Properties of 32 Lactic Acid Bacteria Isolated from Canine and Feline Feces. Microorganisms 9, 33 (2021). 34 16. J. Zhou et al. , Fecal Microbiota Transplantation in Mice Exerts a Protective Effect 35 Against Doxorubicin-Induced Cardiac Toxicity by Regulating Nrf2 -Mediated Cardiac 36 Mitochondrial Fission and Fusion. Antioxid Redox Signal 41, 1-23 (2024). 37 17. M. Troseid, G. O. Andersen, K. Broch, J. R. Hov, The gut microbiome in coronary artery 38 disease and heart failure: Current knowledge and future directions. EBioMedicine 52, 39 102649 (2020). 40 18. S. Fromentin et al. , Microbiome and metabolome features of the cardiometabolic 41 disease spectrum. Nat Med 28, 303-314 (2022). 42 19. W. H. W. Tang, F. Backhed, U. Landmesser, S. L. Hazen, Intestinal Microbiota in 43 Cardiovascular Health and Disease: JACC State-of-the-Art Review. J Am Coll Cardiol 44 73, 2089-2105 (2019). 45 20. Y. Talmor-Barkan et al., Metabolomic and microbiome profiling reveals personalized 46 risk factors for coronary artery disease. Nat Med 28, 295-302 (2022). 47 21. N. Kazemian, M. Mahmoudi, F. Halperin, J. C. Wu, S. Pakpour, Gut microbiota and 48 cardiovascular disease: opportunities and challenges. Microbiome 8, 36 (2020). 49 22. R. G. Pushpass, S. Alzoufairi, K. G. Jackson, J. A. Lovegrove, Circulating bile acids as 50 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 26 a li nk between the gut microbiota and cardiovascular health: impact of prebiotics, 1 probiotics and polyphenol-rich foods. Nutr Res Rev, 1-20 (2021). 2 23. Z. Wang et al., Gut flora metabolism of phosphatidylcholine promotes cardiovascular 3 disease. Nature 472, 57-63 (2011). 4 24. L. Cui, T. Zhao, H. Hu, W. Zhang, X. Hua, Association Study of Gut Flora in Coronary 5 Heart Disease through High-Throughput Sequencing. Biomed Res Int 2017, 3796359 6 (2017). 7 25. S. H. Kim, M. Y. Yoon, S. S. Yoon, TMAO and the gut microbiome: implications for the 8 CVD-CKD-IBD axis. Ann Med 57, 2522324 (2025). 9 26. A. B. Roberts et al., Development of a gut microbe -targeted nonlethal therapeutic to 10 inhibit thrombosis potential. Nat Med 24, 1407-1417 (2018). 11 27. L. Xia, Z. Wang, X. Chen, TMAO Induce s Vascular Endothelial Cells Pyroptosis 12 Through TET2-CYTB-ROS Pathway. J Inflamm Res 18, 8719-8733 (2025). 13 28. L. Xiu, P . Zhao, X. Gu, B. Yu, Role of Trimethylamine N -Oxide in Assessing Plaque 14 Instability of the Culprit Lesion in Chinese Patients With ST -Elevation Myocardial 15 Infarction: Insights From a 7 -Year Long-Term Follow-Up Study. Clin Transl Sci 18, 16 e70357 (2025). 17 29. W. H. Tang et al. , Intestinal microbial metabolism of phosphatidylcholine and 18 cardiovascular risk. N Engl J Med 368, 1575-1584 (2013). 19 30. H. S. Nam et al., Elevation of the Gut Microbiota Metabolite Trimethylamine N -Oxide 20 Predicts Stroke Outcome. J Stroke 21, 350-352 (2019). 21 31. X. Yu et al., Trimethylamine N-oxide predicts cardiovascular events in coronary artery 22 disease patients with d iabetes mellitus: a prospective cohort study. Front Endocrinol 23 (Lausanne) 15, 1360861 (2024). 24 32. T. Suzuki, L. M. Heaney, D. J. Jones, L. L. Ng, Trimethylamine N -oxide and Risk 25 Stratification after Acute Myocardial Infarction. Clin Chem 63, 420-428 (2017). 26 33. T. Suzuki et al. , Association with outcomes and response to treatment of 27 trimethylamine N-oxide in heart failure: results from BIOSTAT-CHF. Eur J Heart Fail 21, 28 877-886 (2019). 29 34. Y. Kinugasa et al., Trimethylamine N-oxide and outcomes in patients hospitalized with 30 acute heart failure and preserved ejection fraction. ESC Heart Fail 8, 2103-2110 (2021). 31 35. T. Shafi et al., Trimethylamine N -Oxide and Cardiovascular Events in Hemodialysis 32 Patients. J Am Soc Nephrol 28, 321-331 (2017). 33 36. P. Andrikopoulos et al. , Evidence of a causal and modifiable relationship between 34 kidney function and circulating trimethylamine N-oxide. Nat Commun 14, 5843 (2023). 35 37. S. Rath, B. Heidrich, D. H. Pieper, M. Vital, Uncovering the trimethylamine -producing 36 bacteria of the human gut microbiota. Microbiome 5, 54 (2017). 37 38. K. A. Romano, E. I. Vivas, D. Amador -Noguez, F. E. Rey, Intestinal microbiota 38 composition modulates choline bioavailability from diet and accumulation of the 39 proatherogenic metabolite trimethylamine-N-oxide. mBio 6, e02481 (2015). 40 39. J. Colby, L. J. Zatman, Trimethylamine metabolism in obligate and facultative 41 methylotrophs. Biochem J 132, 101-112 (1973). 42 40. C. Loechel, A. Basran, J. Basran, N. S. Scrutton, E. A. Hall, Using trimethylamine 43 dehydrogenase in an enzyme linked amperometric electrode. Part 1. Wild -type 44 enzyme redox mediation. Analyst 128, 166-172 (2003). 45 41. A. F. Hallett, R. Cooper, Respiratory infection in an intensive care unit. S Afr Med J 52, 46 1095-1098 (1977). 47 42. H. Gu et al. , A case report of Klebsiella aerogenes -caused lumbar spine infection 48 identified by metagenome next-generation sequencing. BMC Infect Dis 22, 616 (2022). 49 43. T. Urakami, H. Araki, H. Oyanagi, K. Suzuki, K. Komagata, Paracoccus aminophilus 50 sp. nov. and Paracoccus aminovorans sp. nov., which utilize N,N-dimethylformamide. 51 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 27 Int J Syst Bacteriol 40, 287-291 (1990). 1 44. H. S. Han et al., Correlations of the Gastric and Duodenal Microbiota with Histological, 2 Endoscopic, and Symptomatic Gastritis. J Clin Med 8, (2019). 3 45. G. Lawrence et al. , The blood microbiome and its association to cardiovascular 4 disease mortality: case-cohort study. BMC Cardiovasc Disord 22, 344 (2022). 5 46. T. K. Le, Y. J. Lee, G. H. Han, S. J. Yeom, Methanol Dehydrogenases as a Key 6 Biocatalysts for Synthetic Methylotrophy. Front Bioeng Biotechnol 9, 787791 (2021). 7 47. A. Zhou et al., Synergistic interactions of vancomycin with different antibiotics against 8 Escherichia coli: trimethoprim and nitrofurantoin display strong synergies with 9 vancomycin against wild -type E. coli. Antimicrob Agents Chemother 59, 276 -281 10 (2015). 11 48. Y. Lee et al., Longitudinal Plasma Measures of Trimethylamine N -Oxide and Risk of 12 Atherosclerotic Cardiovascular Disease Events in Community -Based Older Adults. J 13 Am Heart Assoc 10, e020646 (2021). 14 49. Z. Dong et al., The Association between Plasma Levels of Trimethyla mine N-Oxide 15 and the Risk of Coronary Heart Disease in Chinese Patients with or without Type 2 16 Diabetes Mellitus. Dis Markers 2018, 1578320 (2018). 17 50. M. Troseid et al. , Microbiota -dependent metabolite trimethylamine -N-oxide is 18 associated with disease sev erity and survival of patients with chronic heart failure. J 19 Intern Med 277, 717-726 (2015). 20 51. C. Roncal et al., Trimethylamine-N-Oxide (TMAO) Predicts Cardiovascular Mortality in 21 Peripheral Artery Disease. Sci Rep 9, 15580 (2019). 22 52. Y. Yazaki et al., Ethnic differences in association of outcomes with trimethylamine N -23 oxide in acute heart failure patients. ESC Heart Fail 7, 2373-2378 (2020). 24 53. Y. Q. Yang, W. H. Deng, R. Z. Liao, Mechanistic Insights into Choline Degradation 25 Catalyzed by the Choline Trimethylamine-Lyase CutC. J Phys Chem B 129, 5438-5448 26 (2025). 27 54. W. K. Wu et al., Gut microbes with the gbu genes determine TMAO production from L-28 carnitine intake and serve as a biomarker for precision nutrition. Gut Microbes 17, 29 2446374 (2025). 30 55. Z. Wang et al., Non-lethal Inhibition of Gut Microbial Trimethylamine Production for the 31 Treatment of Atherosclerosis. Cell 163, 1585-1595 (2015). 32 56. G. Liu et al., Inhibition of Microbiota -dependent Trimethylamine N-Oxide Production 33 Ameliorates High Salt Diet-Induced Sympathetic Excitation and Hypertension in Rats 34 by Attenuating Central Neuroinflammation and Oxidative Stress. Front Pharmacol 13, 35 856914 (2022). 36 57. G. Wang et al. , 3,3 -Dimethyl-1-butanol attenuates cardiac remodeling in pressure -37 overload-induced heart failure mice. J Nutr Biochem 78, 108341 (2020). 38 58. J. Mao et al., Repeated 3,3 -Dimethyl-1-butanol exposure alters social dominance in 39 adult mice. Neurosci Lett 758, 136006 (2021). 40 59. P. Day-Walsh et al., The use of an in-vitro batch fermentation (human colon) model for 41 investigating mechanisms of TMA production from choline, L -carnitine and related 42 precursors by the human gut microbiota. Eur J Nutr 60, 3987-3999 (2021). 43 60. Y. Tang et al., Intestinal metabolite TMAO promotes CKD progression by stimulating 44 macrophage M2 polarization through histone H4 lysine 12 lactylation. Cell Death Differ 45 33, 314-326 (2026). 46 61. W. Zhang et al. , Inhibition of microbiot a-dependent TMAO production attenuates 47 chronic kidney disease in mice. Sci Rep 11, 518 (2021). 48 62. P. Pathak et al., Small molecule inhibition of gut microbial choline trimethylamine lyase 49 activity alters host cholesterol and bile acid metabolism. Am J Phy siol Heart Circ 50 Physiol 318, H1474-H1486 (2020). 51 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 28 63. D. Fennema, I. R. Phillips, E. A. Shephard, Trimethylamine and Trimethylamine N -1 Oxide, a Flavin -Containing Monooxygenase 3 (FMO3) -Mediated Host -Microbiome 2 Metabolic Axis Implicated in Health and Disease. Drug Metab Dispos 44, 1839-1850 3 (2016). 4 64. J. Czarnecki, D. Bartosik, Diversity of Methylotrophy Pathways in the Genus 5 Paracoccus (Alphaproteobacteria). Curr Issues Mol Biol 33, 117-132 (2019). 6 65. S. C. Baker et al., Molecular genetics of the genus Paracoccus: metabolically versatile 7 bacteria with bioenergetic flexibility. Microbiol Mol Biol Rev 62, 1046-1078 (1998). 8 66. W. Zhu et al. , Gut microbes impact stroke severity via the trimethylamine N -oxide 9 pathway. Cell Host Microbe 29, 1199-1208 e1195 (2021). 10 67. J. Xue et al. , Residual Risk of Trimethylamine -N-Oxide and Choline for Stroke 11 Recurrence in Patients With Intensive Secondary Therapy. J Am Heart Assoc 11, 12 e027265 (2022). 13 68. Y. Y . Chen, Z. S. Ye, N. G. Xia, Y. Xu, TMAO as a Novel Predictor of Major Adverse 14 Vascular Events and Recurrence in Patients with Large Artery Atherosclerotic Ischemic 15 Stroke. Clin Appl Thromb Hemost 28, 10760296221090503 (2022). 16 69. R. Kumar, B. Singh, V. K. Gupta, Biodegradation of fipronil by Paracoccus sp. in 17 different types of soil. Bull Environ Contam Toxicol 88, 781-787 (2012). 18 70. T. Brodmann et al. , Safety of Novel Microbes for Human Consumption: Practical 19 Examples of Assessment in the European Union. Front Microbiol 8, 1725 (2017). 20 71. S. Mohri, M. Kanauchi, Isolation of Lactic Acid Bacteria Eliminating Trimethylamine 21 (TMA) for Application to Fishery Processing. Methods Mol Biol 1887, 109-117 (2019). 22 72. X. Heng, W. Liu, W. Chu, Identification of choline -degrading bacteria from healthy 23 human feces and used for screening of trimethylamine (TMA)-lyase inhibitors. Microb 24 Pathog 152, 104658 (2021). 25 73. C. S. Chin et al., Nonhybrid, finished microbial genome assemblies from long -read 26 SMRT sequencing data. Nat Methods 10, 563-569 (2013). 27 74. B. J. Walker et al. , Pilon: an integrated tool for comprehensive microbial variant 28 detection and genome assembly improvement. PLoS One 9, e112963 (2014). 29 75. B. J. Bennett et al. , Trimethylamine -N-oxide, a metabolite associated with 30 atherosclerosis, exhibits complex genetic and dietary regulation. Cell Metab 17, 49-60 31 (2013). 32 76. J. W. Jung et al., Mild hypercapnia before reperfusion reduces ischemia -reperfusion 33 injury in hyperacute ischemic stroke rat model. J Cereb Blood Flow Metab , 34 271678X241296367 (2024). 35 77. J. H. Garcia, S. Wagner, K. F. Liu, X. J. Hu, Neurological deficit and extent of neuronal 36 necrosis attributable to middle cerebral artery occlusion in rats. Statistical validation. 37 Stroke 26, 627-634; discussion 635 (1995). 38 78. E. Z. Longa, P. R. Weinstein, S. Carlson, R. Cummins, Reversible middle cerebral 39 artery occlusion without craniectomy in rats. Stroke 20, 84-91 (1989). 40 79. J. Chen et al. , Intravenous administration of human umbilical cord blood reduces 41 behavioral deficits after stroke in rats. Stroke 32, 2682-2688 (2001). 42 80. D. W. McBride, D. Klebe, J. Tang, J. H. Zhang, Correcting for Brain Swelling's Effects 43 on Infarct Volume Calculation After Middle Cerebral Artery Occlusion in Rats. Transl 44 Stroke Res 6, 323-338 (2015). 45 81. K. Takada, A. Fukatsu, S. Otake, M. Hirasawa, Isolation and characterization of 46 hemolysin activated by reduc tant from Prevotella intermedia. FEMS Immunol Med 47 Microbiol 35, 43-47 (2003). 48 82. K. Lee et al., The ferrichrome receptor A as a new target for Pseudomonas aeruginosa 49 virulence attenuation. FEMS Microbiol Lett 363, (2016). 50 83. R. P . Rennie, L. Turnbull, C. Brosnikoff, J. Cloke, First comprehensive evaluation of 51 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 29 the M.I.C. evaluator device compared to Etest and CLSI reference dilution methods for 1 antimicrobial susceptibility testing of clinical strains of anaerobes and other fastidious 2 bacterial species. J Clin Microbiol 50, 1153-1157 (2012). 3 4 5 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 30 1 Table 1. Overall landscape of BM109 whole genome and comparison with P. 2 aminovorans JCM7685 genome. BM109 genome is constituted in 4 different contigs, with 3 the first contig being the largest chromosome 1. Overall genome architecture is similar to 4 that of JCM7685. GC content (%) represents the total guanine and cytosine ratio in each 5 DNA contig. CDS stands for coding sequence and indicates the number of genes coding for 6 polypeptide products. rRNA and tRNA indicate the number of genes encoding rRNA and 7 tRNA, respectively. 8 9 10 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 31 Figure legends 1 Figure 1. Schematic procedure of screening for TMA-degrading fecal microbes. (A) 2 Schematic representation of the screening procedure. Human fecal samples were pooled to 3 generate a microbiome suspension, which was plated onto DLB agar supplemented with 7% 4 MeOH to enrich for methylotrophic bacteria. After 48 h of incubation, isolated colonies were 5 inoculated into liquid DLB medium supplemented with 10 mM TMA to assess their growth 6 and TMA-degrading capacity. After 24 h of cultivation, culture supernatants were collected 7 for TMA quantification using a picric acid-based colorimetric assay. *DLB refers to 1/5 diluted 8 LB medium. (B) Quantification of residual TMA in culture supernatants after 24 h of aerobic 9 growth in DLB supplemented with 10 mM TMA. The amount of remaining TMA was 10 compared to the initial input and expressed as a ratio (mean ± SD, n = 3). Strain No. 9 11 exhibited the highest TMA degradation efficiency (red arrow). (C) Bacterial strains capable of 12 degrading >60% of the initial TMA were selected for species identification via 16S rRNA 13 gene sequencing. Strain No. 9 was identified as Paracoccus aminovorans, while the other 14 strains belonged to Klebsiella aerogenes, Pseudomonas aeruginosa, and Escherichia coli. 15 BM109 and a control strain (P. denitrificans) were grown in LB amended with 10 mM TMA 16 aerobically with shaking for 24 h (D) and anaerobically without shaking for 48 h (E). To 17 support anaerobic respiratory growth, 60 mM NO3- was added in the anaerobic culture 18 media. Concentrations of TMA detected in the supernatants were compared with the initially 19 added TMA and presented with ratio values (mean ± SD, n=3). *p < 0.001 versus values 20 obtained from control growth. Under both conditions, BM109 exhibited significantly greater 21 TMA degradation compared to the control strain, while bacterial growth was comparable 22 between the two strains. 23 Figure 2. Metabolic pathway of TMA and TMAO degradation and corresponding gene 24 annotations in BM109. (A) Schematic representation of TMA and TMAO metabolism. TMA, 25 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 32 generated from nutrient digestion, is either oxidized to TMAO by TMA monooxygenase 1 (enzyme 1) or converted into dimethylamine (DMA) and formaldehyde by TMA (DMA) 2 dehydrogenase (enzyme 3). DMA undergoes further metabolism into monomethylamine 3 (MMA) and formaldehyde. Formaldehyde, a toxic intermediate, is processed through multiple 4 detoxification steps. Formaldehyde dehydrogenase (enzyme 6) catalyzes the conversion of 5 formaldehyde into formate, which is subsequently oxidized to CO2 by formate 6 dehydrogenase (enzyme 7). Methanol, another potential substrate, can also be converted 7 into formaldehyde by methanol dehydrogenase (enzyme 8). Additionally, TMAO reductase 8 (enzyme 2) enables the reduction of TMAO back to TMA, and TMAO demethylase (enzyme 9 2’) facilitates alternative TMAO degradation. Detoxification of formaldehyde also involves 10 glutathione (GSH)-dependent pathways, with S-hydroxymethyl glutathione synthetase 11 (enzyme 9), dehydrogenase (enzyme 10), and hydrolase (enzyme 11) contributing to 12 formate production. (B) Corresponding genes identified in the BM109 genome encoding 13 enzymes involved in TMA and TMAO metabolism. Each enzyme in panel A is matched with 14 specific gene loci found in BM109, including multiple gene clusters encoding key metabolic 15 enzymes. Notably, two distinct genes (Contig 2_00604 and Contig 2_00975) encode TMA 16 dehydrogenase, and two loci encode formate dehydrogenase, with one organized as a four-17 gene operon. 18 Figure 3. Comparative genomic analysis of BM109 and closely related species based 19 on the presence of TMA and TMAO metabolism-related genes. A total of 15 strains, 20 including two Paracoccus aminovorans strains (BM109 and JCM7685), were identified as 21 the closest phylogenetic relatives of BM109, and their phylogenetic relationships were 22 constructed based on 16S rRNA gene sequences. These species were classified into three 23 major groups: nine species from alpha-Proteobacteria, two from beta-Proteobacteria, and 24 four from gamma-Proteobacteria. The heatmap on the right represents the presence and 25 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 33 sequence identity of genes associated with TMA and TMAO metabolism across these 1 species. Homologous genes identified via BLAST-based searches using BM109 protein 2 sequences are shown in dark blue (≥60% sequence identity), whereas genes with lower 3 sequence identity but annotated with a similar function are shown in light blue. The four 4 genes within the operon spanning 1_02817~20 in BM109 are labeled 5b-1, 5b-2, 5b-3, and 5 5b-4. Genes with conserved presence across multiple species are outlined in orange. This 6 analysis highlights the genomic distinction of BM109 and provides insights into the 7 evolutionary distribution of key metabolic genes. 8 Figure 4. Effects of oral treatment of BM109 cells on changes in TMA and TMAO levels 9 in mouse models. (A) Mice were assigned to three dietary groups: normal diet (ND), high-10 choline diet (HCD), and high-fat diet (HFD), which they were fed for 45 weeks. Each group 11 was subsequently divided into two subgroups: one receiving a vehicle control and the other 12 BM109 (5×108 CFU/day) for 5 consecutive days via oral gavage. (B, C) TMA and TMAO 13 concentrations in fecal samples. (D, E) TMA and TMAO concentrations in serum samples. 14 Data are expressed as mean ± SD (n=5 per group). In each bar, values of mean ± SD are 15 presented. *p < 0.001 vs. control treatment, **p < 0.05 vs. control treatment. 16 Figure 5. Neuroprotective effects of BM109 in a transient middle cerebral artery 17 occlusion (tMCAO) model. (A) Experimental timeline and group assignments. Male Wistar 18 rats (8 weeks old) were fed either a normal diet (ND) or high-choline diet (HCD) for 3 weeks, 19 with BM109 (5×10¹⁰ CFU/day) or vehicle control (15% glycerol) administered during the final 20 7 days. On the day of surgery, animals underwent a 90-minute tMCAO procedure and were 21 sacrificed 24 h post-reperfusion for infarct analysis and behavioral assessments. (B) Plasma 22 TMAO levels at baseline and after dietary intervention. TMAO levels were analyzed in the 23 normal chow (n=60) and HCD (n=58) groups. Statistical analysis was performed using one-24 way ANOVA. Among the HCD group, 33 rats exhibited more than a threefold increase in 25 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 34 TMAO levels compared to baseline (** in the figure). (C) Effect of BM109 on plasma TMAO 1 levels in high-TMAO rats (n=33) from panel B (** in the figure). These rats were divided into 2 a vehicle group (n=18) and a BM109-treated group (n=15). Pre- and post-gavage TMAO 3 levels are shown for each group. BM109 significantly reduced TMAO levels compared to the 4 vehicle group, with p-values indicated (paired t-test). (D) Infarct size (%) measured by 5 triphenyltetrazolium chloride (TTC) staining to evaluate ischemic brain injury. Control and 6 BM109-treated groups are shown in blue and red, respectively. Statistical significance was 7 determined using paired t-test, with p-values indicated above the comparisons. 8 Figure 6. BM109 improves neurological outcomes in a (tMCAO) model. (A) Garcia 9 score assessed sensorimotor function and neurological performance. (B) Longa score 10 evaluated stroke severity based on motor deficits. (C) Modified Neurological Severity Score 11 (mNSS) assessed overall neurological impairment, including motor, sensory, and reflexive 12 functions. In all panels, red circles represent the control group, and blue circles represent the 13 BM109-treated group. Black horizontal lines indicate the mean values for each group. p-14 values are directly indicated in the figure for statistical comparisons. 15 16 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 35 1 2 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 36 1 2 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 37 1 2 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 38 1 2 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 39 1 2 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 40 1 2 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 41 [Supplementary information] 1 2 3 4 5 6 Microbiome-Targeted Reduction of Circulating Trimethylamine N-7 Oxide Mitigates Ischemic Stroke Risk 8 9 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 42 1 2 3 4 Figure S1. PCR confirmation of the gene encoding TMAO reductase in BM109, but not 5 in JCM7685. 6 Genomic DNA extracted from BM109 (lanes 1 and 3) or JCM7685 (lanes 2 and 4) was PCR 7 amplified for the presence of the TMAO reductase-encoding gene. Primer sequences used for 8 PCR reactions are shown in the table. 9 10 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 43 1 2 3 Figure S2. BM109 reduces ischemic injury in a tMCAO model. 4 Representative coronal brain sections stained with 2,3,5 -triphenyltetrazolium chloride (TTC) 5 to visualize infarcted areas 24 h post-reperfusion. The schematic on the left illustrates the 6 sectioning method, where four coronal slices (1, 3, 5, and 7) were obtained at 2-mm intervals. 7 White areas indicate infarcted (ischemic) tissue, while red regions represent viable tissue. 8 Brain sections are shown for rats fed a normal chow diet (left) or a high -choline diet (HCD, 9 right), with control (vehicle-treated) and BM109-treated groups. BM109 administration visibly 10 reduced infarct size under both dietary conditions. Scale bar = 5 mm. The schematic 11 illustration on the left was created using BioRender.com. 12 13 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 44 1 2 3 Figure S3. Safety assessment of BM109 administration in vivo. 4 (A) Hemolysis assay on blood agar plate showing that BM109 is non-hemolytic, indicating no 5 hemolytic activity. (B) Gross examination of the intestinal tract from mice administered either 6 PBS (left) or BM109 (5×10⁸ CFU, right). No visible abnormalities, inflammation, or structural 7 damage were observed in the BM109 -treated group. (C) Histological analysis of the small 8 intestine and colon following BM109 administration. Representative hematoxylin and eosin 9 (H&E)-stained sections of t he small intestine (top) and colon (bottom) show no significant 10 histopathological differences between the PBS and BM109-treated groups. 11 12 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 45 1 2 3 Figure S4. Antibiotic susceptibility testing of BM109. 4 E-test strips were used to determine the minimum inhibitory concentrations (MICs) of BM109 5 against eight clinically and agriculturally relevant antibiotics. BM109 was streaked onto Müller-6 Hinton agar plates, and E -test strips impregnated with a gradient of each antibiotic were 7 applied. The inhibition zones indicat e BM109’s susceptibility to each antibiotic. The tested 8 antibiotics include ampicillin (AM), streptomycin (SM), chloramphenicol (CL), clindamycin 9 (CM), gentamicin (GM), kanamycin (KM), tetracycline (TC), and vancomycin (VA). BM109 10 exhibited susceptibility to all tested antibiotics, with slight resistance observed for vancomycin, 11 which is expected for Gram-negative bacteria. 12 13 14 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint 46 1 Supplementary information. Genome alignment summary of BM109 contigs with reference 2 Paracoccus aminovorans genomes and plasmid sequences 3 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprintthis version posted April 16, 2026. ; https://doi.org/10.64898/2026.04.15.718846doi: bioRxiv preprint