Microbiota-derived IPA mitigates post-stroke neuroinflammation by inhibiting TREM2-dependent pyroptosis

Abstract

Ischemic stroke remains the leading cause of long-term disability globally, underscoring the urgent need for novel therapeutic strategies. Here, we explore a microbiota-gut-brain axis that provides valuable insights for achieving this objective. Utilizing a distal middle cerebral artery occlusion (dMCAO) mouse model, we observed a marked reduction in Duncaniella muris (D. muris) post-stroke, alongside dysregulated tryptophan metabolism, characterized by elevated levels of indole-3-lactic acid (ILA) and decreased indole-3-propionic acid (IPA). D. muris supplementation restored metabolic balance by converting ILA to IPA, leading to significant improvements in neurological recovery. Mechanistically, IPA exerted neuroprotective effects by attenuating neuroinflammation through TREM2-dependent modulation of microglial activation, promoting an anti-inflammatory phenotype and inhibiting NLRP3 inflammasome-mediated pyroptosis. These findings highlight the therapeutic potential of the D. muris-IPA-TREM2-pyroptosis axis as a novel target for ischemic stroke treatment, providing a basis for future microbiome-based interventions aimed at improving stroke outcomes.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-025-03660-8.

Introduction

Ischemic stroke represents a leading cause of global mortality and long-term disability, imposing substantial socioeconomic burdens. While intravenous thrombolysis with tissue plasminogen activator (tPA) and endovascular therapy remain the cornerstone treatments, their clinical utility is limited by narrow therapeutic windows and stringent eligibility criteria, limiting their benefit to only a small subset of stroke patients [1–7]. This underscores the critical need to elucidate novel pathophysiological mechanisms to expand therapeutic options.

Recent studies have highlighted the emerging role of the gut microbiota in the pathogenesis of ischemic stroke through the microbiota-gut-brain axis [8, 9]. Stroke-associated dysbiosis is marked by reduced microbial diversity and altered composition in both experimental models and patients [9–14]. Notably, stroke severity correlates with specific gut microbiome alterations [10, 15]. Although antibiotics exhibit neuroprotective effects in animal models, their clinical translation faces ethical and practical challenges. Instead, microbiota-derived metabolites, particularly those involved in tryptophan metabolism, have emerged as key mediators of microbiota-gut-brain communication [8]. Metabolites such as serotonin, kynurenine, and indole derivatives, can cross the blood-brain barrier and modulate neuroinflammation by acting as endogenous ligands for the aryl hydrocarbon receptor (AhR), thereby regulating immune responses and endothelial function [16–20]. We previously identified a significant downregulation of the tryptophan metabolite 3-hydroxy-kynurenamine (3-HKA) in stroke patients and animal models, with demonstrated neuroprotective effects in vivo [3]. Nevertheless, the precise gut microbiota alterations following ischemic stroke and the functional contribution of microbiota-derived metabolites to stroke pathogenesis remain poorly defined.

Neuroinflammation plays a pivotal role in the brain injury following ischemic stroke, with microglia being the primary mediators [21, 22]. Activated microglia can adopt either a pro-inflammatory or anti-inflammatory phenotype, profoundly affecting the neuronal damage and repair [23–25]. We previously demonstrated that promoting anti-inflammatory microglial polarization mitigates post-stroke brain injury [25]. However, how the gut microbiota and its metabolites influence microglia-mediated neuroinflammation after ischemic stroke remains unclear.

In this study, we investigated the interplay between gut microbiota dysbiosis and ischemic stroke progression, focusing on the role of the microbiota-gut-brain axis. Using a distal middle cerebral artery occlusion (dMCAO) model, we identified significant alterations in gut microbial composition, particularly a reduction in Duncaniella muris (D. muris), accompanied by disrupted tryptophan metabolism. This dysbiosis led to the accumulation of neurotoxic indole-3-lactic acid (ILA) and a decrease in neuroprotective indole-3-propionic acid (IPA). Remarkably, D. muris supplementation restored metabolic balance by converting ILA into IPA, improving neurological recovery. Mechanistically, IPA promoted an anti-inflammatory microglial phenotype by regulating TREM2-mediated pyroptosis, thereby attenuating neuroinflammation. Together, these findings highlight the therapeutic potential of targeting the gut microbiota and its metabolites, particularly through D. muris and IPA supplementation, as a novel strategy for ischemic stroke treatment.

Results

Gut microbiota dysbiosis occurred in a mouse model of ischemic stroke

Microbiome-related abnormalities, including alterations in community composition, have been observed in stroke patients; however, stroke-associated microbial signatures remain poorly defined [14, 15, 26]. To identify microbiome signatures influencing stroke pathogenesis, we performed 16S rRNA gene sequencing and nanopore metagenomic sequencing on fecal samples collected from mice at days 1 and 7 after dMCAO. The taxonomic composition and diversity of the gut microbiome were analyzed. At day 1 after stroke, α-diversity, as assessed by the Chao index, Sobs index, and Shannon index, was significantly reduced in the dMCAO group compared to the shams (Fig. 1a). Simultaneously, a significant difference in the β-diversity of the gut microbiota was observed between the two groups of mice (Fig. 1b). We next quantified dysbiosis using the microbial dysbiosis index (MDI), where higher values reflect greater microbiome disruption. The MDI was significantly elevated in the dMCAO group at day 1 compared to the shams (Fig. 1c).

Fig. 1.

Dysbiosis of the gut microbiota in mice post-stroke. a α-diversity analysis (Chao, sobs, and Shannon) at ASV level in mice at day 1 post-stroke. *p < 0.05, **p < 0.01, ***p < 0.001, dMCAO group vs. sham group, n = 8 mice/group. b Principal co-ordinate analysis showing the β-diversity of the gut microbiota at ASV level in mice at day 1 post-stroke. dMCAO group vs. sham group, n = 8 mice/group. c Microbial dysbiosis index (MDI) of the gut microbiota in mice at day 1 post-stroke. dMCAO group vs. sham group, n = 8 mice/group. d Stacked bar plots showing normalized abundance of microbiota taxa at phylum level in mice at day 1 post-stroke. e Linear discriminant analysis effect size (LEfSe) identifying the top 20 differentially abundant genera (LDA > 2) 1 day post-stroke. n = 8 mice/group. f Top 10 differentially abundant bacteria species day 1 post-stroke (P < 0.05 in false discovery rate (FDR), |fold change| (FC) > 2). *P < 0.05, **P < 0.01, ***P < 0.001, dMCAO group vs. sham group, n = 8 mice/group. g Stacked bar plots showing normalized abundance of microbiota taxa at phylum level in mice at days 1 and 7 post-stroke. h Total 10 differentially abundance bacteria species at day 7 post-stroke (FDR < 0.05, |FC| > 2). *P < 0.05, dMCAO group (n = 9) vs. sham group (n = 5). i Experimental design framework for oral administration of bacteria. j Rotarod test. *P < 0.05, dMCAO+L. johnsonii group vs. dMCAO+Vehicle group; ^P < 0.05, dMCAO+L. murinus group vs. dMCAO+Vehicle group; #P < 0.05, dMCAO+D. muris group vs. dMCAO+Vehicle group; analyzed by one-way ANOVA followed by LSD multiple comparison tests at individual time points; analyzed by repeated measures ANOVA followed by LSD multiple comparison tests (bracket). n = 11 mice/group. k-l Representative TTC stained sections and quantification of infarct volume at day 7 post-stroke. *P < 0.05, **P < 0.01, ***P < 0.001, vs. dMCAO+Vehicle group by one-way ANOVA followed by LSD multiple comparison tests. n = 6 mice/group

Furthermore, the composition of the gut microbiota was analyzed at the phylum level. At day 1 post- stroke, Bacillota and Deferribacterota abundances were significantly reduced, while Pseudomonadota was significantly elevated in dMCAO mice compared to shams (Figs. 1d and S1). Linear discriminant analysis effect size (LEfSe) analysis at the genus level identified 20 differentially abundant taxa with LDA > 2 between dMCAO and sham groups (Figs. 1e and S1). Lactobacillus, Duncaniella and Muribaculum were significantly enriched in shams, whereas Bacteroides, Escherichia, and Enterobacter exhibited significant enrichment in dMCAO mice (Figs. 1e and S1). At the species level, D. muris and probiotic bacteria such as Lactobacillus johnsonii (L. johnsonii) and Ligilactobacillus murinus (L. murinus) were markedly reduced in dMCAO mice, whereas the opportunistic pathogen Escherichia coli (E. coli) was elevated (Figs. 1f and S1).

Moreover, gut microbiota dysbiosis persisted in dMCAO mice up to day 7 post-stroke. Both α- and β-diversity remained significant altered in dMCAO mice compared to shams (Figure S2a-b). At the phylum level, Bacillota and Deferribacterota abundances were still reduced, whereas Pseudomonadota remained elevated, consistent with changes observed on day 1 post-stroke (Figs. 1g and S2c). Notably, by day 7, Pseudomonadota abundance in dMCAO mice showed partial recovery relative to day 1 levels (Fig. 1g). Importantly, the top 10 differentially abundant bacterial species in the dMCAO group at day 7 mirrored those at day 1, including persistent reductions in D. muris, L. johnsonii, and L. murinus, as well as an increase in E. coli compared to shams (Fig. 1h). These findings suggested that D. muris, L. johnsonii, L. murinus, and E. coli may have play distinct biological roles in stroke pathophysiology.

To investigate the role of gut microbiota dysbiosis in stroke, we administered D. muris, L. johnsonii, L. murinus, and E. coli individually via gavage to mice after dMCAO, and subsequently evaluating the effects of this treatment (Fig. 1i). Neurological recovery was assessed using the rotarod test, and infarct volume was measured via TTC staining. Interestingly, compared to vehicle treatment, the rotarod test revealed that D. muris treatment increased the duration of time spent on the rotarod at days 7, 14, and 21 post-stroke, indicating that D. muris improved post-stroke motor coordination and learning ability (Fig. 1j). Consistently, the D. muris-treated mice exhibited a lower infarct volume than the vehicle-treated mice at day 7 post-stroke (Fig. 1k-l). Similarly, L. johnsonii and L. murinus treatment improved neurological recovery and reduced infarct volume compared to vehicle, supporting the beneficial effects of these species in other diseases contexts [27, 28] (Fig. 1j-l). In contrast, E. coli treatment impaired neurological recovery and increased infarct volume, corroborating previous findings [15] (Figs. 1k-l and S2d). Together, these findings support the link between stroke-induced gut dysbiosis, characterized by probiotic reduction and enrichment of opportunistic pathogens. Notably, we proposed D. muris as a previously unrecognized probiotic candidate with therapeutic potential in stroke recovery.

To investigate whether the transplanted D. muris was able to colonize the gut and remain the dominant bacterial population in the recipient mice, we quantified D. muris levels in the fecal samples collected on days 7 (Figure S2f) and 14 (Figure S2g) post-stoke. Our findings show that D. muris abundance significantly increased by day 7 post-stroke and remained elevated at day 14, indicating successful colonization and persistence of the transplanted bacterium in the gut.

Disorders of tryptophan metabolism occurred in a mouse model of ischemic stroke

To investigate whether gut microbiota dysbiosis affects metabolic processes following stroke, we performed untargeted metabolomic profiling of fecal samples from the dMCAO (day 1) and sham groups. Partial least squares discriminant analysis (PLS-DA) indicated that the metabolomic composition was striking differences between the two groups (Figure S3a). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis revealed that ischemic stroke significantly altered tryptophan metabolism (Fig. 2a).

Fig. 2.

Disorders of tryptophan metabolism were observed in mice after ischemic stroke. a KEGG pathway enrichment diagram of metabolites from the feces of the sham group and dMCAO groups at day 1 post-stroke. The size of the bubbles in the diagram indicates the number of metabolites enriched in each pathway, and the color of the bubbles reflects the significance level of the enrichment, represented by the p-value. dMCAO group (n = 5) vs. sham group (n = 6). b Heatmap showing the relative abundance of significantly differentiated tryptophan metabolites in the serum of mice in the sham and dMCAO groups at day 1 post-stroke. The left panel shows a heatmap (blue to red colour: low to high abundance), while the right panel presents a bubble plot illustrating the differential enrichment of metabolites between groups. The color of the bubbles indicates the significance p-value, and the size of the bubbles reflects the fold change. P < 0.05, dMCAO group vs. sham group, n = 8. c ILA, IPA, and IAA concentrations were measured by targeted mass spectrometry in the serum of mice in the sham and dMCAO groups at day 1 post-stroke. ***P < 0.001, dMCAO group vs. sham group; Student’s t test. n = 8. d ILA, IPA, and IAA concentrations were measured in the peri-infarct cortex of mice in the sham and dMCAO groups at day 1 post-stroke. ***P < 0.001, dMCAO group vs. sham group; Student’s t test. n = 8. e The heatmap displayed the correlation strength (negative to positive: the color from blue to red) between the significantly differentiated tryptophan metabolites in the serum of mice, as assessed using Spearman correlation. *P < 0.05, **P < 0.01, ***P < 0.001, n = 8 mice/group. f Experimental design framework for the administration of tryptophan metabolites. g Rotarod test. *P < 0.05, dMCAO+IPA group vs. dMCAO+Vehicle group; #P < 0.05, dMCAO+IAA group vs. dMCAO+Vehicle group; analyzed by one-way ANOVA followed by LSD multiple comparison tests at individual time points; analyzed by repeated measures ANOVA followed by LSD multiple comparison tests (bracket). n = 11 mice/group. h-i Representative TTC stained sections at day 7 post-stroke, and quantified cerebral infarction size. *P < 0.05, **P < 0.01, ***P < 0.001, vs. dMCAO+Vehicle group by one-way ANOVA followed by LSD multiple comparison tests. n = 6 mice/group

We next conducted targeted metabolomic analysis on serum samples from the dMCAO (day 1) and sham groups to further characterize tryptophan metabolism following stroke. The analysis revealed that 10 tryptophan metabolites, including indole-3-lactic acid (ILA), kynurenine, and indole-3-carboxaldehyde, were upregulated, whereas 4 metabolites, including indole-3-propionic acid (IPA), indole-3-acetic acid (IAA), and 3-hydroxy-kynurenamine, were downregulated in the mice post-stroke (Figs. 2b-c and S3b). Notably, compared to the sham group, IPA and IAA levels were reduced, and ILA levels increased in the peri-infarct cortex—a pattern that was consistent with the results observed in serum samples (Fig. 2d). Among these, IPA exhibited the most pronounced reduction following stroke. IPA, an indole analog specifically produced by the gut microbiota, has been shown to promote axonal regeneration and provide protection against radiation toxicity [29], with ILA serving as its metabolic intermediate. Additionally, IAA has been reported to reduce the inflammatory response, thereby enhancing intestinal barrier function and alleviating colitis [30]. Moreover, ILA levels were significant negatively correlated with both IPA and IAA (Fig. 2e). Based on these insights, we selected three representative tryptophan metabolites—IPA, ILA, and IAA—for further investigation of their biological roles in stroke pathology.

To monitor the role of tryptophan metabolism in stroke, we administered IPA, ILA, and IAA individually to mice after dMCAO and subsequently evaluated the effects of these treatment (Fig. 2f). Notably, the rotarod test demonstrated that IPA and IAA treatments increased the time spent on the rotarod at days 7, 14, and 21 post-stroke compared to the vehicle-treated mice, suggesting that IPA and IAA facilitated neurological recovery following stroke (Fig. 2g). Consistently, both IPA and IAA treatments reduced infarct volume compared to vehicle treatment at day 7 post-stroke (Fig. 2h-i). However, no significant improvement in neurological recovery was observed in the ILA-treated mice compared to the vehicle control group (Fig. 2g). Interestingly, ILA treatment further increased infarct volume compared to vehicle treatment at day 7 post-stroke (Fig. 2h-i), suggesting that ILA may exacerbate stroke severity. In the transient middle cerebral artery occlusion (tMCAO) model, we also observed that IPA treatment reduced infarct volume at day 7 post-stroke (Figure S3c). These results confirm the dysregulation of tryptophan metabolism post-stroke and highlight IPA as a potential metabolic marker and therapeutic candidate for improving stroke outcomes.

D. muris improved post-stroke neurological recovery by catabolizing ILA to IPA

Gut microbiota metabolized nutrients ingested by the host, producing a variety of microbiota-derived metabolites that signal beyond the gastrointestinal tract and modulate brain function via the gut-brain axis [31]. Among these metabolites, tryptophan (Trp)—an aromatic essential amino acid containing an indole moiety—was catabolized through three primary pathways: serotonin, kynurenine (Kyn), and indole pathway. Notably, IPA, IAA and ILA were generated from tryptophan via the indole pathway, which was almost entirely mediated by microbial enzymes [32]. To validate the correlation between gut microbiota and tryptophan metabolites, we conducted Spearman correlation analysis between the differentially abundant gut microbial species identified post-stroke and 14 altered tryptophan metabolites. Among them, D. muris exhibited significant correlations with most of these metabolites reduced abundance negatively correlating with elevated ILA levels and positively correlating with decreased IPA levels in dMCAO mice (Figs. 3a and S4a-b).

Fig. 3.

D. muris improved post-stroke neurological recovery by metabolizing ILA to IPA. a The heatmap displayed the correlation strength (negative to positive: the color from blue to red) between the changes in the abundance of different gut microbiota and the alterations in the expression of various tryptophan metabolites, as assessed using Spearman correlation. *P < 0.05, **P < 0.01, ***P < 0.001, n = 12. b Schematic pathway of microbial Trp metabolism to IPA. Enzymes encoded by D. muris are labeled in green; non-encoded enzymes in gray. c The gene map displayed the Aldh and acdA genes in the D. muris genome, with the GC content shown in the lower track. d Experimental design framework for oral administration of D. muris. e ILA and IPA concentrations were measured by targeted mass spectrometry in the serum of mice from the sham, dMCAO+Vehicle and dMCAO+D. muris groups at day 7 post-stroke. ***P < 0.001, vs. dMCAO+Vehicle group by one-way ANOVA followed by LSD multiple comparison tests. n = 6 mice/group. f An experimental design framework was established for the administration of D. muris or IPA under tryptophan-deficient diet conditions. g Rotarod test. *P < 0.05, dMCAO+D. muris+Trp group vs. dMCAO+Vehicle group; #P < 0.05, dMCAO+IPA group vs. dMCAO+Vehicle group; analyzed by one-way ANOVA followed by LSD multiple comparison tests (7 d) or Dunnett’s multiple comparison tests (14 d); analyzed by Kruskal-Wallis test followed by Dunn’s post hoc analysis (21 d); analyzed by repeated measures ANOVA followed by Dunnett’s multiple comparison tests (bracket). n = 11 mice/group. h-i Representative TTC stained sections at day 7 post-stroke, and quantified cerebral infarction size. ***P < 0.001, vs. dMCAO+Vehicle group by one-way ANOVA followed by LSD multiple comparison tests. n = 6 mice/group. Trp, tryptophan; IAM, indole-3-acetamide; IAA, indole-3-acetic acid; IAld, indole-3-aldehyde; IPYA, indole-3-pyruvate; ILA, indole-3-lactic acid; IA, indole acrylic acid; IPA, indole-3-propionic acid; TA, Tryptamine; IAAld, Indole-3-Acetaldehyde; Aldh, Aromatic lactate dehydrogenase; Arat, Aromatic amino acid aminotransferase; ldhA, D-2-hydroxyacid dehydrogenase; Fldh, Indolelactate dehydrogenase; fldH, phenylactate dehydrogenase; Fldbc, Indolelactoyl-CoA dehydratase alpha/beta; Flda, Cinnamoyl-CoA:indolelactate CoA-transferase; FldI, R-indolelactate dehydratase activator; acdA, acyl-CoA dehydrogenase; Acda, Indoleacrylate reductase; Ipd, Indolepyruvate decarboxylase; Tmo, Tryptophan 2-monooxygenase; Tdc, L-tryptophan decarboxylase; Dao, Diamine oxidase; Amie, Amidase; Ald, Aldehyde dehydrogenase; IaaH, indoleacetamide hydrolase; IaaDH, Indoleacetaldehyde dehydrogenase

Comparative metatranscriptomic analysis of the gut microbiota between the sham and dMCAO groups revealed significant alterations in the abundance of bacterial genes involved in tryptophan catabolism. Specifically, we observed marked upregulation of microbial genes encoding enzymes responsible for the conversion of Trp to ILA, including amino acid aminotransferase (Arat), D-2-hydroxyacid dehydrogenase (ldhA), and indolelactate dehydrogenase (Fldh) (Figs. 3b and S4c). Conversely, bacterial genes critical for the subsequent metabolism of ILA to IPA, notably acyl-CoA dehydrogenase (acdA) and indoleacrylate reductase (Acda), were significantly downregulated following cerebral ischemia (Figs. 3b and S4c). This dysregulation of the tryptophan metabolic pathway resulted in the accumulation of ILA along with reduced IPA levels, likely due to the depletion of bacterial taxa capable of metabolizing ILA to IPA.

Next, we examined whether the D. muris genome encodes genes previously reported in the microbial conversion of Trp to IPA (Fig. 3b). Basic Local Alignment Search Tool (BLAST) analysis of the D. muris genome identified coding sequences for Arat, ldhA and aromatic lactate dehydrogenase (Aldh), which are known to catalyze the conversion of Trp to ILA (Figs. 3b-c and S4d). Notably, genomic analysis confirmed the presence of a coding sequence for acdA, the key enzyme responsible for the conversion of ILA to IPA, within the D. muris genome (Figs. 3b-c and S4d). However, coding sequences of the key enzymes involved in the metabolism of tryptophan to IAA, including amidase (Amie), indoleacetamide hydrolase (IaaH), aldehyde dehydrogenase (Ald), and indoleacetaldehyde dehydrogenase (IaaDH), were not identified in the genome of D. muris (Fig. 3b).

We subsequently assessed the ability of D. muris to produce IPA in culture medium containing various concentrations of tryptophan. IPA production increased in a dose-dependent manner with increasing tryptophan concentration, confirming that D. muris promoted tryptophan metabolism to IPA (Figure S4e). Supplementation with ILA resulted in a dose-dependent increase in IPA production by D. muris (Figure S4f). In contrast, supplementation with tryptophan resulted in minimal to no production of IAA by D. muris (Figure S4g). Notably, dMCAO model mice presented significantly lower IPA levels and higher ILA levels than mice in the sham control group did. Treatment with D. muris effectively restored the levels of both metabolites to near-normal levels by day 7 after stroke (Fig. 3d-e). On the basis of these findings, our study demonstrates that D. muris has a remarkable ability to convert the neurotoxic metabolite ILA into the neuroprotective compound IPA following ischemic stroke.

Furthermore, to investigate whether the neuroprotective effects of D. muris after stroke were mediated through IPA production, we performed a tryptophan deprivation experiment on dMCAO model mice (Fig. 3f). Our findings revealed that under tryptophan-deficient conditions, treatment with D. muris + Trp or IPA significantly improved behavioral recovery at 7, 14, and 21 days after stroke, whereas treatment with D. muris alone had only modest effects (Fig. 3g). Similarly, D. muris + Trp or IPA treatment significantly reduced the infarct volume in dMCAO model mice under tryptophan-deficient conditions on day 7 after stroke (Fig. 3h-i). These findings demonstrate that D. muris, which harbors the acdA gene, enhances post-stroke neurological recovery through the enzymatic conversion of ILA to IPA.

IPA ameliorated microglia-mediated neuroinflammation post-stroke

Microglia, the resident immune cells of the central nervous system, play a pivotal role in initiating post-stroke inflammatory responses. After the onset of stroke, activated microglia rapidly migrate to the ischemic lesion and exhibit dynamic polarization between pro-inflammatory and anti-inflammatory phenotypes [33, 34]. As demonstrated in previous studies, shifting microglial polarization toward the anti-inflammatory phenotype significantly enhances neurological recovery following stroke [25]. To investigate the impact of IPA on microglial polarization, we assessed the expression of pro-inflammatory markers (CD16, IL-1β, IL-18, IL-6, and TNF-α) and anti-inflammatory markers (CD206, TGF-β, IL-10, Arg-1, and YM-1) via qPCR at day 3 post-stroke. Compared with sham controls, vehicle-treated mice presented significantly upregulated expression of pro-inflammatory genes, whereas IPA treatment markedly reduced their expression (Fig. 4a). Moreover, compared with vehicle treatment, IPA treatment significantly elevated the mRNA expression of anti-inflammatory genes on day 3 following stroke (Fig. 4b). Similarly, in vitro, IPA reversed the oxygen‒glucose deprivation (OGD)-induced upregulation of CD16 in microglia (Figure S5b). In addition, compared with the OGD group, the IPA-treated group exhibited increased mRNA expression of CD206 in microglia (Figure S5c).

Fig. 4.

IPA ameliorated microglia-mediated neuroinflammation post-stroke. a mRNA expression of pro-inflammatory markers (CD16, IL-1β, IL-18, IL-6, and TNF-α) of microglia at day 3 post-stroke. *P < 0.05, **P < 0.01, ***P < 0.001, analyzed by one-way ANOVA followed by LSD multiple comparison tests (CD16 and IL-18) or Dunnett’s multiple comparison tests (IL-1β, IL-6, and TNF-α). n = 6. b The mRNA expression of anti-inflammatory markers (CD206, TGF-β, IL-10, Arg-1 and YM-1) of microglia at day 3 post-stroke. *P < 0.05, **P < 0.01, ***P < 0.001, analyzed by one-way ANOVA followed by LSD multiple comparison tests (TGF-β, Arg-1 and YM-1) or Dunnett’s multiple comparison tests (IL-10); analyzed by Kruskal-Wallis test followed by Dunn’s post hoc analysis (CD206). n = 6. c-d The percentage of CD86-positive cells were detected by flow cytometry in the peri-infarct region at days 3 post-stroke. **P < 0.01, ***P < 0.001, analyzed by one-way ANOVA followed by LSD multiple comparison tests. n = 12. e-f The percentage of CD206-positive cells were detected by flow cytometry. **P < 0.01, by one-way ANOVA followed by Dunnett’s multiple comparison tests. n = 12. g Iba1 (green) and CD16 (red) double-immunostaining was used to determine the microglia of pro-inflammatory phenotype in the peri-infarct region at day 3 post-stroke. Bar = 50 μm. h Quantification of CD16+Iba1+/Iba1+ cells. ***P < 0.001, by one-way ANOVA followed by LSD multiple comparison tests. n = 6. i-j The concentrations of IL-1β and IL-6 in the plasma samples from mice at day 3 post-stroke. ***P < 0.001, analyzed by one-way ANOVA followed by LSD multiple comparison tests (IL-1β) or Dunnett’s multiple comparison tests (IL-6). n = 6

Flow cytometry and immunofluorescence staining further confirmed the ability of IPA to modulate microglial polarization. At day 3 post-stroke, the ratio of cells positive for CD86 (a pro-inflammatory phenotypic marker) in the peri-infarct region was greater in vehicle-treated mice than in sham control mice (Figs. 4c-d and S5a). Compared with vehicle treatment, IPA treatment drastically decreased the ratio of CD86-positive cells and increased the ratio of CD206-positive cells (Figs. 4c-f and S5a). Then, we measured the proportion of microglia with a pro-inflammatory phenotype via double staining for CD16 and Iba1. At day 3 after stroke, the ratio of CD16⁺Iba1⁺ cells to Iba1⁺ cells in the peri-infarct area was significantly greater in vehicle-treated mice than in sham-operated control mice, whereas IPA treatment reversed this effect (Fig. 4g-h). Ultimately, our findings demonstrate that IPA facilitates the transition of microglia to an anti-inflammatory phenotype following stroke.

Microglia contribute to modulating the inflammatory microenvironment through cytokine secretion. The concentrations of IL-1β, IL-18, IL-6, and TNF-α were elevated in plasma samples following cerebral ischemia, while the ischemia-induced upregulation of these cytokines was reversed in the IPA-treated group compared to the vehicle-treated group (Figs. 4i-j and S5d-e). This anti-inflammatory effect was recapitulated in vitro, where OGD-stimulated BV2 microglia presented marked increases in secreted IL-1β, IL-18, IL-6, and TNF-α levels. Compared with the OGD group, the IPA-treated group exhibited similar suppression of cytokine production in the BV2 cell supernatant of (Figure S5f). Collectively, our data demonstrate that IPA promotes microglial polarization toward an anti-inflammatory phenotype after stroke, thereby alleviating the inflammatory microenvironment.

TREM2 activation drove IPA-induced microglial reprogramming toward an anti-inflammatory phenotype post-stroke

Triggering receptor expressed on myeloid cells 2 (TREM2), an immunoglobulin superfamily transmembrane receptor predominantly localized to microglia, serves as a critical regulator of microglial activation, proliferation, and phagocytic function [35]. Recent studies have demonstrated that TREM2 upregulation in activated microglia after ischemic stroke attenuates neuroinflammation, indicating its neuroprotective potential [36, 37]. We observed increased expression of the TREM2 protein following cerebral ischemia, indicating the activation of TREM2, which is consistent with the findings of previous studies (Fig. 5a). Interestingly, compared with vehicle treatment, IPA treatment further increased TREM2 activation in the ischemic brain (Fig. 5b). Colocalization analysis of TREM2 and Iba1 in the penumbra at day 3 after stroke revealed a greater number of TREM2⁺ microglia, which was further amplified by IPA treatment (Fig. 5c-d). qPCR analysis of TREM2 mRNA expression corroborated these findings (Fig. 5e). These results suggest that microglial TREM2 is activated after cerebral ischemia and that IPA exerts its effects by promoting TREM2 activation.

Fig. 5.

IPA modulated microglia towards an anti-inflammatory phenotype by activating TREM2 following stroke. a Protein levels of TREM2 in each group at day 3 post-stroke. ***P < 0.001, analyzed by Student’s t-test. n = 5. b Protein levels of TREM2 in each group at day 3 post-stroke. *P < 0.05, **P < 0.01, analyzed by Dunnett’s multiple comparison tests. n = 5. c Iba1 (green) and TREM2 (red) double-immunostaining in the peri-infarct region at day 3 post-stroke. Bar = 50 μm. d Quantification of TREM2+Iba1+/Iba1+ cells. ***P < 0.001, by one-way ANOVA followed by LSD multiple comparison tests. n = 6. e The mRNA expression of TREM2 at day 3 post-stroke. **P < 0.01, ***P < 0.001, analyzed by one-way ANOVA followed by LSD multiple comparison tests. n = 6. f mRNA expression of pro-inflammatory markers (CD16, IL-1β, IL-18, IL-6, and TNF-α) in BV2 microglial cells in vitro. *P < 0.05, **P < 0.01, ***P < 0.001, analyzed by one-way ANOVA followed by Dunnett’s multiple comparison tests (CD16, IL-1β, IL-18, and TNF-α) or LSD multiple comparison tests (IL-6). n = 5. g The mRNA expression of anti-inflammatory markers (CD206, TGF-β, IL-10, Arg-1 and YM-1) in BV2 microglial cells in vitro. *P < 0.05, **P < 0.01, ***P < 0.001, analyzed by one-way ANOVA followed by Dunnett’s multiple comparison tests (CD206) or LSD multiple comparison tests (TGF-β, IL-10, Arg-1 and YM-1). n = 5

To investigate IPA’s regulatory role in microglial polarization via TREM2 activation, we performed lentivirus-mediated TREM2 knockdown in BV2 microglial cells, with successful knockdown confirmed at mRNA levels (Figures S6a-b). As expected, OGD significantly elevated the expression of pro-inflammatory genes (CD16, IL-1β, IL-18, IL-6, TNF-α) in microglia, an effect that was effectively attenuated by IPA treatment (Fig. 5f). However, TREM2 deficiency abolished IPA’s suppressive effects on pro-inflammatory phenotype microglia. Furthermore, IPA promoted the expression of anti-inflammatory microglial markers (CD206, TGF-β, IL-10, Arg-1, YM-1) following OGD, an effect that was similarly negated by TREM2 knockdown (Fig. 5g). Collectively, these findings demonstrate that IPA drives microglial polarization toward an anti-inflammatory phenotype through TREM2-dependent mechanisms.

IPA suppressed TREM2-dependent pyroptosis following stroke

To investigate the downstream mechanisms underlying the therapeutic effects of IPA on stroke, we performed transcriptomic analysis of brain tissues from sham-operated, vehicle-treated, and IPA-treated mice following dMCAO. Comparative analysis revealed 598 differentially expressed genes (DEGs) common to both the dMCAO versus sham groups and the dMCAO + IPA versus dMCAO groups (Figure S7a-b). KEGG pathway analysis of these 598 DEGs revealed significant enrichment of several inflammatory pathways, suggesting that IPA treatment modulates the inflammatory response following ischemic stroke (Fig. 6a). Additionally, a detailed examination of the pyroptosis signaling pathway revealed marked activation of the NLRP3 inflammasome following ischemia, with notable changes observed in the expression of the Caspase-1, IL-1β, and NLRP3 genes, which was significantly inhibited by IPA treatment (Fig. 6b). The mRNA levels of NLRP3 and Caspase-1 were increased after stroke compared with after sham surgery but were significantly reduced by IPA treatment (Fig. 6c-d).

Fig. 6.

IPA suppressed TREM2-dependent pyroptosis following stroke. a KEGG pathway analysis of the 598 common genes identified from the Venn diagram (dMCAO vs. sham and dMCAO+IPA vs. dMCAO) revealed significant enrichment in inflammatory pathways, indicating that IPA treatment modulates the inflammatory response following ischemic stroke. n = 4. b Heatmap depicting the expression of representative genes involved in the pyroptosis pathway, derived from transcriptomic analysis. n = 4. c mRNA expression of NLRP3 at day 3 post-stroke. ***P < 0.001, analyzed by one-way ANOVA followed by LSD multiple comparison tests. n = 6. d mRNA expression of Caspase-1 at day 3 post-stroke. ***P < 0.001, analyzed by one-way ANOVA followed by LSD multiple comparison tests. n = 6. e-j Protein levels of NLRP3, pro-caspase-1, p20 caspase-1, GSDMD and GSDMD-N in each group at day 3 post-stroke. *P < 0.05, **P < 0.01, ***P < 0.001, analyzed by one-way ANOVA followed by LSD multiple comparison tests (NLRP3, pro-caspase-1, GSDMD and GSDMD-N) or Dunnett’s multiple comparison tests (p20 caspase-1). n = 5. k Protein levels of NLRP3, pro-caspase-1, p20 caspase-1, GSDMD and GSDMD-N in each group. l-q Protein levels of NLRP3, pro-caspase-1, p20 caspase-1, GSDMD and GSDMD-N in each group. *P < 0.05, **P < 0.01, ***P < 0.001, analyzed by one-way ANOVA followed by LSD multiple comparison tests (NLRP3, p20 caspase-1, GSDMD and GSDMD-N) or Dunnett’s multiple comparison tests (pro-caspase-1). n = 5

We further assessed the protein expression of key components of the NLRP3 inflammasome and pyroptosis machinery, including NLRP3, pro-caspase-1, p20 caspase-1, gasdermin D (GSDMD), and N-terminal GSDMD (GSDMD-N). Our results indicated that the expression of NLRP3, pro-caspase-1, and p20 caspase-1 increased following cerebral ischemia, suggesting activation of the NLRP3 inflammasome, and this effect was reversed by IPA treatment (Fig. 6e-h). Furthermore, the level of GSDMD-N, a pore-forming fragment of GSDMD, was found to be increased after ischemia, and IPA treatment effectively reversed this increase (Fig. 6e and i-j). To further validate the effects of IPA on microglial pyroptosis, BV2 microglia were cultured in vitro under OGD conditions. Consistent with the in vivo findings, we observed that OGD led to an increase in the expression of NLRP3, pro-caspase-1, p20, caspase-1, GSDMD, and GSDMD-N in microglia, and this increase was abolished by IPA treatment (Figs. 6k and S6c-g).

Given that IPA promotes anti-inflammatory microglial polarization via a TREM2-dependent mechanism, we next tested whether TREM2 is required for the IPA-mediated inhibition of pyroptosis. Lentivirus-mediated knockdown of TREM2 in BV2 microglia effectively abolished the suppressive effects of IPA on the expression of NLRP3 inflammasome components, including NLRP3, pro-caspase-1, and p20 caspase-1 (Fig. 6l-o). Furthermore, TREM2 knockdown abolished the inhibitory effect of IPA treatment on pyroptotic pore-forming proteins, including GSDMD and GSDMD-N (Fig. 6l and p-q). These results demonstrate that IPA suppresses pyroptosis after stroke through a TREM2-dependent mechanism.

IPA enhanced post-stroke neurological recovery through TREM2 activation

To validate these findings in vivo, we performed intracerebroventricular injection of a microglia-specific adeno-associated virus to achieve global TREM2 knockdown (Fig. 7a). Comprehensive sensorimotor assessments (rotarod, modified neurological severity score (mNSS), corner test) were conducted pre- and post-stroke (up to 21 days). As previously reported, IPA treatment significantly enhanced rotarod performance at days 7, 14, and 21 post-stroke. However, microglial TREM2 ablation negated IPA-mediated improvements in motor coordination and learning (Fig. 7b). Consistently, IPA-treated mice demonstrated significantly reduced neurological deficit scores (Fig. 7c) and fewer right turns (Fig. 7d) compared to vehicle-treated mice, while microglial TREM2 ablation reversed these benefits (Fig. 7b-d). Quantitative gait assessment using the Catwalk XT system further demonstrated that IPA treatment significantly improved sensorimotor function, evidenced by enhanced average running speed (Fig. 7e), increased stride length in the left front paw (LF), and expanded pawprint area in the left front paw (Fig. 7f-g). However, genetic ablation of TREM2 abolished these IPA-mediated improvements in sensorimotor recovery.

Fig. 7.

IPA enhanced post-stroke neurological recovery through TREM2 activation. a An experimental design framework was established for the administration of IPA under TREM2 gene knockdown. b Rotarod test. *P < 0.05, dMCAO+IPA (TREM2-NC) group vs. dMCAO+Vehicle (TREM2-NC) group; analyzed by one-way ANOVA followed by LSD multiple comparison tests (7 d and 21 d) or Dunnett’s multiple comparison tests (14 d); analyzed by repeated measures ANOVA followed by Dunnett’s multiple comparison tests (bracket). n = 11 mice/group. c mNSS. *P < 0.05, dMCAO+IPA (TREM2-NC) group vs. dMCAO+Vehicle (TREM2-NC) group; analyzed by Kruskal‒Wallis test followed by Dunn’s post hoc analysis (7 d, 14 d and 21 d); analyzed by repeated measures ANOVA followed by Dunnett’s multiple comparison tests (bracket). n = 11 mice/group. d Corner test. *P < 0.05, dMCAO+IPA (TREM2-NC) group vs. dMCAO+Vehicle (TREM2-NC) group; analyzed by one-way ANOVA followed by LSD multiple comparison tests (7 d, 14 d and 21 d); analyzed by repeated measures ANOVA followed by Dunnett’s multiple comparison tests (bracket). n = 11 mice/group. e Average run speed measured in the gait test. *P < 0.05, dMCAO+IPA (TREM2-NC) group vs. dMCAO+Vehicle (TREM2-NC) group; analyzed by Kruskal-Wallis test followed by Dunn’s post hoc analysis (7 d and 14 d); analyzed by repeated measures ANOVA followed by Dunnett’s multiple comparison tests (bracket). n = 11 mice/group. f Stride length for the left front paw. *P < 0.05, dMCAO+IPA (TREM2-NC) group vs. dMCAO+Vehicle (TREM2-NC) group; analyzed by one-way ANOVA followed by Dunnett’s multiple comparison tests (7 d) or LSD multiple comparison tests (14 d); analyzed by repeated measures ANOVA followed by Dunnett’s multiple comparison tests (bracket). n = 11 mice/group. g Pawprint area for the left front paw. *P < 0.05, dMCAO+IPA (TREM2-NC) group vs. dMCAO+Vehicle (TREM2-NC) group; analyzed by one-way ANOVA followed by LSD multiple comparison tests (7 d and 14 d); analyzed by repeated measures ANOVA followed by Dunnett’s multiple comparison tests (bracket). n = 11 mice/group. h-i Representative TTC stained sections at day 7 post-stroke, and quantified cerebral infarction size. ***P < 0.001, dMCAO+IPA (TREM2-NC) group vs. dMCAO+Vehicle (TREM2-NC) group by one-way ANOVA followed by LSD multiple comparison tests. n = 6 mice/group

Quantitative analysis of infarct volumes at day 7 post-stroke demonstrated that IPA treatment significantly attenuated cortical lesion formation compared to vehicle controls, correlating with improved neurobehavioral outcomes. Notably, microglial-specific TREM2 ablation reversed this neuroprotective effect (Fig. 7h-i). Taken together, these results suggested that IPA promotes post-stroke functional recovery through TREM2-dependent mechanisms (Fig. 8).

Fig. 8.

Microbiota-Derived IPA Mitigates Post-Stroke Neuroinflammation by Inhibiting TREM2-Dependent Pyroptosis. (Created in https://BioRender.com)

Discussion

Our findings, in conjunction with emerging evidence, highlight the critical role of gut microbiota dysbiosis in modulating neurological outcomes following ischemic stroke. Recent studies have demonstrated that the gut microbial composition influences multiple neurological disorders, including stroke, Alzheimer’s disease (AD), Parkinson’s disease (PD), and amyotrophic lateral sclerosis (ALS) [8, 31, 38–42]. Notably, antibiotic-induced alterations in the gut microbiota protect against cerebral ischemia in mouse models [43], whereas age-related microbial shifts have been shown to adversely impact stroke prognosis [12]. Both preclinical and clinical investigations have revealed significant changes in post-stroke microbial diversity, with stroke severity correlating with the gut microbiota composition [10, 15], suggesting a potential role for the gut microbiota in determining neurological recovery. Our differential microbial abundance and LEfSe analyses revealed a marked reduction in the abundance of Lactobacillus genera. In addition, the abundances of the genera Escherichia and Enterobacter increased post-stroke. At the species level, D. muris, L. johnsonii, and L. murinus were significantly depleted as early as day 1, whereas E. coli—an opportunistic pathogen linked to worsened infarction outcomes [15]—was enriched, a pattern that persisted for at least 7 days. Importantly, L. johnsonii and L. murinus, previously implicated in host metabolic and immune regulation [28, 29, 44–48], were found to improve post-stroke neurological recovery upon exogenous supplementation, expanding their known therapeutic applications. Interestingly, D. muris supplementation improved post-stroke neurological recovery, suggesting that D. muris is a promising probiotic candidate warranting further investigation in stroke-related dysbiosis. These observations collectively reinforce the microbiota‒gut‒brain axis as a critical mediator of post-stroke neurological recovery, with microbial imbalance potentially exacerbating brain injury through the depletion of beneficial taxa (D. muris, L. johnsonii, and L. murinus) and the enrichment of pathogenic E. coli. Future studies should explore targeted probiotic interventions aimed at restoring microbial homeostasis and mitigating stroke-related deficits, for which D. muris has emerged as a novel therapeutic target.

Building upon the established role of gut microbiota dysbiosis in stroke recovery, our findings elucidate the critical function of microbial tryptophan metabolism in post-stroke pathophysiology. Tryptophan, an essential amino acid acquired through the diet, is metabolized by both the host and microbiota, serving as a key metabolic checkpoint for the gut and systemic homeostasis during health and disease [18, 27, 31, 49–53]. Tryptophan is metabolized primarily through three major pathways: the serotonin pathway, the kynurenine pathway, and the indole pathway [54, 55]. The indole pathway is almost exclusively mediated by microbial enzymes, which produce metabolites capable of crossing the blood‒brain barrier and mediating microbiota‒gut‒brain communication [32]. Metabolic profiling of serum from dMCAO model mice revealed significant dysregulation of tryptophan metabolism, characterized by pathological accumulation of ILA and depletion of the protective metabolites IAA and IPA. While prior studies reported beneficial effects of microbiota-derived ILA in depression and cancer models [32, 56], its detrimental impact involving expansion of the cerebral infarction volume was observed in our stroke model; however, the performance of ILA-treated mice in the rotarod test was not significantly different from that of vehicle-treated mice, highlighting the disease context-dependency of the actions of microbial metabolites. Conversely, IPA supplementation consistently improved neurological recovery, aligning with the established protective roles of IPA in stroke [57], atherosclerosis [58], radiation toxicity [59], trauma recovery [60], and cancer immunotherapy [29]. Our findings demonstrated significant post-stroke upregulation of microbial genes encoding enzymes involved in tryptophan catabolism to ILA, whereas genes responsible for the subsequent conversion of ILA to IPA were markedly downregulated. This dysregulation suggests that ischemic stroke may impair the capacity of the gut microbiota to metabolize ILA into IPA, leading to an imbalance in tryptophan-derived metabolite levels. D. muris demonstrated therapeutic potential through its ability to catalyze the conversion of the neurotoxic metabolite ILA to the neuroprotective metabolite IPA in a manner mediated by the acdA gene, thereby rectifying the tryptophan metabolite imbalance after stroke. Critically, this neuroprotective effect was abolished following tryptophan deprivation, confirming that IPA production is the primary mechanism underlying this effect. While exogenous IAA administration also enhanced neurological recovery, D. muris lacks the enzymatic machinery for IAA biosynthesis, further underscoring the central role of IPA in its therapeutic effect.

As IPA is an effector molecule derived from D. muris, we further delineated the mechanism by which it is involved in attenuating neuroinflammation through targeted modulation of microglial activation. Previous studies established the capacity of IPA to target diverse immune cells (T cells, macrophages, and microglia) for immune response regulation [42, 43]. Microglia, as resident innate immune cells, undergo rapid activation after ischemia, aggregating in peri-infarct zones to release pro-inflammatory cytokines (TNF-α, IL-1β, and IL-6) and chemokines that exacerbate brain injury and impede neurovascular repair [24, 25]. Microglial heterogeneity exhibits substantial variation across neurodegenerative disease types, pathological stages, and neuroanatomical regions, rendering the simplistic M1/M2 classification system inadequate for comprehensive phenotypic characterization [61–63]. While the disease-associated microglia (DAM) classification system embodies the microglial heterogeneity observed in Alzheimer’s disease (AD) and amyotrophic lateral sclerosis (ALS) [61–63], this classification system is primarily applicable to neurodegenerative conditions, with distinct transcriptional signatures across pathologies. Recent identification of stroke-associated microglia (SAM) via single-cell RNA sequencing [64] represents an advance, but methodological limitations—including model variability (transient vs. permanent ischemia) and the use of single-omics—restrict its generalizability. Consequently, functional assessments of pro-inflammatory (CD16, IL-1β, IL-18, IL-6, and TNF-α) and anti-inflammatory (CD206, TGF-β, IL-10, Arg-1, and YM-1) markers remain instrumental in evaluating microglial phenotypes in stroke. Crucially, we demonstrated that IPA promoted anti-inflammatory microglial polarization while suppressing pro-inflammatory polarization by upregulating TREM2. The expression of TREM2, a key regulator of microglial lipid metabolism, phagocytosis, and inflammatory responses [35, 37, 65–67], was significantly elevated in the ischemic brain after IPA treatment, which is consistent with the neuroprotective effect of IPA in mitigating postischemic inflammation [37]. TREM2⁺ microglia in ischemic brain regions were shown to enhance glucose metabolic reprogramming, including by increasing oxidative phosphorylation, further reinforcing their neuroprotective phenotype [36]. Mechanistically, IPA-induced TREM2 signaling inhibited NLRP3 inflammasome-mediated pyroptosis—an inflammatory programmed cell death pathway distinct from necrosis and apoptosis. Pyroptosis is characterized by cellular swelling, the formation of pyroptotic pores in the plasma membrane, membrane rupture, and the subsequent release of inflammatory cytokines that amplify inflammatory cascades [33, 68–74]. NLRP3 inflammasome activation plays a pivotal role in initiating pyroptosis in microglia after stroke [75–78]. The NLRP3 inflammasome (comprising NLRP3, ASC, and pro-caspase-1) activates caspase-1, which cleaves GSDMD, generating NGSDMD-N, which oligomerizes and forms pyroptosis pores in the cell membrane, facilitating the release of IL-1β and IL-18 [71–73, 79, 80, 81]. These findings are consistent with our previous findings that suppressing NLRP3 inflammasome-driven pyroptosis promotes neurological recovery following ischemic stroke and traumatic brain injury [33, 82]. Notably, this study provides the first evidence that IPA suppresses microglial pyroptosis via TREM2-dependent NLRP3 inflammasome inhibition, extending its established neuroprotective mechanisms. Previous studies from our group showing that promoting microglial polarization towards an anti-inflammatory phenotype fosters angiogenesis and neuroplasticity during the post-stroke recovery phase—processes that are fundamental for long-term functional recovery [25]. Therefore, IPA’s improvement of stroke neurological function recovery may involve modulating microglia polarization to promote long-term neuroprotective effects. These findings establish IPA as a therapeutic agent that modulates the TREM2–NLRP3–pyroptosis axis to mitigate post-stroke neuroinflammation and promote recovery, offering novel mechanistic insights into gut microbiota-derived neuroprotection.

Although we identified microbial species and metabolites with neuromodulatory activity and demonstrated the existence of the microbiome‒gut‒brain axis in dMCAO model mice, several limitations warrant consideration. First, while we clearly demonstrated the ability of D. muris and IPA to improve stroke outcomes, other microbial species and metabolites may exert similar effects. Therefore, future research is needed to elucidate their mechanisms of action and therapeutic potential. Second, whether IPA promotes post-stroke neurological recovery through additional pathways remains to be investigated. Finally, the exclusive reliance on rodent models underscores the need for preclinical validation in nonhuman primates to assess the translational potential of IPA.

In summary, this work elucidates the D. muris–IPA–TREM2–pyroptosis axis as a promising therapeutic target for ischemic stroke. Through comprehensive investigation of the microbiota‒gut‒brain axis in an dMCAO model, we observed stroke-induced dysbiosis characterized by depletion of D. muris and disruption of tryptophan metabolism, manifesting as pathological accumulation of neurotoxic ILA and deficiency of neuroprotective IPA. Critically, D. muris supplementation rectified this metabolic imbalance through the enzymatic conversion of ILA to IPA, enhancing neurological recovery. Mechanistically, IPA attenuated neuroinflammation by reprogramming microglial activation through TREM2-dependent signaling, promoting anti-inflammatory phenotypes and suppressing NLRP3 inflammasome-mediated pyroptosis. This research opens new avenues for the development of microbiota-targeted therapies to mitigate stroke-related deficits and promote functional recovery.

Materials and methods

Animals and stroke model

Male C57BL/6 mice (8–10 weeks old, 20–24 g) were housed under specific pathogen-free conditions with controlled environmental parameters (22 ± 2 °C, 55 ± 5% humidity, 12 h light/dark cycle). Following a 3-day acclimation period with ad libitum access to food and water, all procedures were conducted in accordance with NIH guidelines and the ARRIVE reporting standards [83, 84] (protocol No. BNI202405004, approved by the Beijing Institute of Neurosurgery Animal Ethics Committee).

Distal middle cerebral artery occlusion (dMCAO): Cerebral ischemia was induced via permanent occlusion of the right distal middle cerebral artery (MCA) combined with ipsilateral common carotid artery (CCA) ligation, as previously reported [3]. Briefly, anesthesia was administered via intraperitoneal injection of Avertin (400 mg/kg, Sigma-Aldrich, Cat# T48402-25G). A neck incision was made to expose the right CCA, which was ligated using an 8 − 0 silk suture. A small craniotomy was performed to access the distal MCA, which was then coagulated with a cauterizer under a microscope. Throughout the procedure, the animals’ body temperature was maintained at 37.5 ± 0.5 °C. Sham-operated mice underwent similar procedures but without occluding the CCA or coagulating the distal MCA.

Transient Middle Cerebral Artery Occlusion (tMCAO): Transient cerebral ischemia was induced in mice via right MCA occlusion using an intraluminal filament technique. Animals were anesthetized, and a midline cervical incision was made to expose the carotid arteries. The external carotid artery (ECA) was ligated and transected, and a silicone-coated monofilament (A5-122250, CINONTECH, China) was advanced from the ECA stump into the internal carotid artery until mild resistance indicated MCA occlusion. Cerebral blood flow was monitored by laser Doppler flowmetry (Moor VMS-LDF), with occlusion defined as a reduction to < 25% of baseline. After 60 min of ischemia, the filament was withdrawn to permit reperfusion. Sham-operated mice underwent identical procedures excluding filament insertion. Body temperature was maintained at 37.0 ± 0.5 °C using a heating pad throughout the surgery.

16S rRNA gene sequencing and analysis

DNA extraction and PCR amplification

Total microbial genomic DNA was isolated from murine fecal samples using the FastPure Stool DNA Isolation Kit (MJYH, Shanghai, China). DNA integrity was verified by 1% agarose gel electrophoresis, and the DNA concentration and purity were assessed using a NanoDrop2000 spectrophotometer (Thermo Scientific, USA). For bacterial community profiling, the V1–V9 hypervariable regions of the 16S rRNA gene were amplified using the universal primers 27 F (5′-AGRGTTYGATYMTGGCTCAG-3′) and 1492R (5′-RGYTACCTTGTTACGACTT-3′). Fungal ITS regions were amplified with the primers ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA-3′) and ITS4R (5′-TCCTCCGCTTATTGATATGC-3′). Unique PacBio barcode sequences were appended to each primer pair for sample multiplexing. For PCR, each reaction (20 µL) consisted of 4 µL of 5× FastPfu buffer, 2 µL of 2.5 mM dNTPs, 0.8 µL each of forward and reverse primers (5 µM), 0.4 µL of FastPfu DNA Polymerase, 10 ng of template DNA, and nuclease-free water. The thermal cycling conditions were as follows: initial denaturation at 95 °C for 3 min; 27 cycles of 95 °C for 30 s, 60 °C for 30 s, and 72 °C for 45 s; and a final extension at 72 °C for 10 min (T100 Thermal Cycler, Bio-Rad, USA). Amplicons were purified using AMPure® PB beads (Pacific Biosciences, USA) and quantified via a Qubit 4.0 instrument (Thermo Fisher Scientific, USA).

Library preparation and sequencing

Equimolar concentrations of purified amplicons were pooled, and SMRTbell libraries were constructed using the SMRTbell Prep Kit 3.0 (Pacific Biosciences, USA) following the manufacturer’s protocol. High-fidelity (HiFi) sequencing was performed on the PacBio Sequel IIe platform (Pacific Biosciences, USA) by Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China). Circular consensus sequencing (CCS) was conducted using SMRT Link v11.0 to generate highly accurate reads.

Bioinformatic processing

HiFi reads were demultiplexed and filtered by length (16S: 1,000–1,800 bp; ITS: 300–900 bp). Denoising and amplicon sequence variant (ASV) calling were performed using DADA2 within the QIIME2 pipeline (v2020.2), achieving single-nucleotide resolution. To standardize the diversity metrics, the samples were rarefied to 6,000 sequences per sample, maintaining an average Good’s coverage of 97.90%. Taxonomic classification was conducted using a Naive Bayes classifier trained on the SILVA 16S rRNA database (v138). Functional metagenomic predictions were obtained via PICRUSt2, employing HMMER for sequence alignment, EPA-NG/Gappa for phylogenetic placement, castor for copy number normalization, and MinPath for pathway inference.

Statistical analysis

Microbial community analyses were conducted on the Majorbio Cloud Platform (https://cloud.majorbio.com). Alpha diversity (ASVs, the Chao1 index, and the Shannon index) and rarefaction curves were computed using Mothur v1.30.1. Beta diversity was assessed via Bray‒Curtis dissimilarity-based principal coordinate analysis (PCoA) (Vegan v2.5-3). PERMANOVA was used to test group differences (9,999 permutations). LEfSe analysis was used to identify differentially abundant taxa (LDA score > 2, P < 0.05). Distance-based redundancy analysis (db-RDA) was used to evaluate the contributions of clinical parameters to community structure, with forward selection via Monte Carlo permutation (P < 0.05). Cooccurrence networks were constructed using robust correlations (Spearman’s |R| > 0.6, P value < 0.01)

Metagenomic analysis

DNA extraction and quality control

Total genomic DNA was isolated from 0.2 g stool samples using the FastPure Stool DNA Isolation Kit (Magnetic Bead; MJYH, Shanghai, China) following the manufacturer’s protocol. The DNA concentration and purity were assessed using SynergyHTX (BioTek, USA) and a NanoDrop2000 spectrophotometer (Thermo Fisher Scientific, USA), respectively. Integrity was verified by 1% agarose gel electrophoresis.

Library preparation and sequencing

DNA was fragmented to an average size of 350 bp using a Covaris M220 ultrasonicator (Gene Company Limited, China). Paired-end libraries were prepared with the NEXTFLEX Rapid DNA-Seq Kit (Bioo Scientific, USA) and sequenced on the Illumina NovaSeq™ X Plus platform (Illumina Inc., USA) using NovaSeq X Series 25B Reagent Kits by Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China).

Bioinformatic processing

Raw reads were processed on the Majorbio Cloud Platform (http://www.majorbio.com). Adapter sequences and low-quality reads (length < 50 bp or average Phred score < 20) were removed using fastp (v0.20.0). Human host contamination was filtered by aligning reads to the human reference genome (GRCh38) with BWA (v0.7.17). De novo assembly was performed using MEGAHIT (v1.1.2), retaining contigs ≥ 300 bp. Open reading frames (ORFs) were predicted with Prodigal (v2.6.3), and nonredundant gene catalogs were generated via CD-HIT (v4.7; 90% identity, 90% coverage). Gene abundance was quantified using SOAPaligner (v2.21) with 95% sequence identity.

Taxonomic and functional annotation

Taxonomic classification was performed by aligning nonredundant genes against the NCBI NR database using DIAMOND (v2.0.13; e-value cutoff: 1 × 10⁻⁵). Functional annotation was conducted using the Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), eggNOG, Carbohydrate-Active Enzymes (CAZy), and Pathogen-Host Interactions (PHI) databases and the Comprehensive Antibiotic Resistance Database (CARD). Differential abundance analysis across taxonomic and functional categories was performed using the Kruskal‒Wallis test.

Metabolomics analysis

For untargeted metabolomics analysis and targeted metabolomic analysis of the tryptophan pathway, previously validated methodologies were employed [3].

Antibiotic cocktail therapy

The mice were administered an antibiotic cocktail for 10 days, followed by polyethylene glycol administration to deplete the native intestinal microbiota. The antibiotic mixture included neomycin sulfate (100 mg/kg), ampicillin sodium (100 mg/kg), metronidazole (100 mg/kg), and vancomycin hydrochloride (50 mg/kg) dissolved in a 4 mM acetic acid solution. All reagents were purchased from Sigma‒Aldrich, and the prepared solution was administered via gavage. Fecal DNA analysis of both WT and ABX-treated mice revealed a substantial depletion of gut microbiota following ABX treatment, as shown in Figure S2e.

Bacterial strain colonization

Adult male mice were subjected to dMCAO and subsequently administered D. muris (HZB677234, HZBio Microbial Conservation), L. johnsonii (HZB196986, HZBio Microbial Conservation), L. murinus (HZB487586, HZBio Microbial Conservation), or E. coli (HZB196114, HZBio Microbial Conservation) via oral gavage. The bacteria were delivered at a dose of 10^8 colony-forming units per day for a duration of 7 consecutive days. Mice that received the same volume of physiological saline served as controls.

Quantitative PCR for microbiota analysis

Following microbiota colonization, fecal samples were collected from the mice, and genomic DNA (gDNA) was extracted. The gDNA concentration was then quantified and adjusted to 1 ng/mL for further analysis. The abundance of specific intestinal bacterial groups was measured using PowerUp SYBR Green Master Mix (Thermo Fisher) with group-specific 16 S rRNA gene primers. These primers were designed using NCBI Primer-BLAST to ensure specificity, with amplification efficiencies maintained between 90% and 105%. Total Eubacteria 16 S rDNA was used as an internal control to normalize the data.

Administration of ILA, IPA, and IAA

The compounds ILA (purity 99.99%, #HY-113099), IPA (purity 99.82%, #HY-W015229), and IAA (purity 99.94%, #HY-18569) were obtained from MedChemExpress (China). The mice were randomly divided into experimental groups using a computer-generated randomization sequence (SPSS v27.0). Following dMCAO, the animals were administered ILA, IPA, or IAA (40 mg/kg/d, dissolved in 0.9% saline) daily by gavage on the basis of previously established protocols [56, 85–87]. Treatment was initiated immediately after surgery and continued once daily for 7 consecutive days.

Neurobehavioral tests

Neurobehavioral assessments were performed by an investigator who was blinded to the treatment groups. Evaluations were conducted prior to dMCAO and continued up to 21 days after dMCAO. Sensorimotor deficits were evaluated using the rotarod test, the mNSS, the corner test, and gait analysis, as previously validated in the literature [3, 33].

Brain infarct volume

The infarct volume was quantified at 7 days after dMCAO using standardized TTC staining protocols. Coronal Sect. (8 slices/brain, 1 mm thickness) were incubated in 2% TTC (37 °C, 20 min) followed by fixation with 4% paraformaldehyde (PFA). Viable tissue was stained red, whereas infarcted regions remained unstained. Digital image analysis was conducted using ImageJ, with correction for edema-induced hemispheric swelling.

Immunofluorescence staining

The mice were anesthetized and transcardially perfused with saline, followed by cold 4% PFA in 0.1 M phosphate-buffered saline (PBS). After perfusion, the brains were cryoprotected in 30% sucrose in PBS for 48 h. Serial coronal sections were then cut at 15 μm thickness using a cryotome (Thermo Scientific, USA). The sections were permeabilized with 0.3% Triton X-100 for 15 min, followed by blocking with 10% normal donkey serum at 37 °C for 1 h. Primary antibodies were then applied, and the samples were incubated overnight at 4 °C. The primary antibodies used for immunofluorescence included goat anti-Iba1 (1:200, Abcam #ab289874), rat anti-CD16 (1:100, BD Biosciences # 553141), and rabbit anti-TREM2 (1:200, CST #55739) antibodies. On the following day, the sections were washed with PBS and incubated with the corresponding secondary antibodies (1:500 or 1:800, conjugated to Alexa Fluor 488 or 594, Jackson ImmunoResearch) at 37 °C for 1 h. Negative control sections were processed identically, excluding incubation with primary antibodies. Images of the stained sections were acquired by laser scanning confocal microscopy with a 40×/1.0 NA objective (Zeiss, Germany). For each region of interest, 3–4 sections per mouse were analyzed. The CD16⁺Iba1⁺/Iba1⁺ cell ratio and TREM2⁺Iba1⁺/Iba1⁺ cell ratio were analyzed using ImageJ software.

Quantitative real-time PCR (qRT‒PCR)

Total RNA was isolated from brain tissues and astrocytes using TRIzol reagent (Invitrogen), after which cDNA was synthesized with the PrimeScript™ First Strand cDNA Synthesis Kit (Takara). qRT‒PCR was performed using PowerUp SYBR Green Master Mix (Thermo Fisher) on a LightCycler 480 system (Roche). Primer pairs were designed via NCBI Primer-BLAST to ensure specificity, with amplification efficiencies maintained at 90–105%. Gene expression was normalized to the expression of GAPDH and analyzed using the 2^−ΔΔCt method, as previously described [3, 33]. The primer sequences for the target genes are detailed in Table 1.

Table 1.

Mouse primer sequences

Gene	Forward primer (5’- 3’)	Reverse primer (5’- 3’)
NLRP3	CCAGCCAGAGTGGAATGACA	AGCGGGAGACAAATGGAGAT
Caspase1	GGCACATTTCCAGGACTGACTG	GCAAGACGTGTACGAGTGGTTG
IL-1β	AACTCAACTGTGAAATGCCACC	CATCAGGACAGCCCAGGTC
IL-18	GACAGCCTGTGTTCGAGGATATG	TGTTCTTACAGGAGAGGGTAGAC
IL-6	CACTTCACAAGTCGGAGGCT	CTGCAAGTGCATCATCGTTGT
TNF-a	GGTGCCTATGTCTCAGCCTCTT	GCCATAGAACTGATGAGAGGGAG
CD16	TTTGGACACCCAGATGTTTCAG	GTCTTCCTTGAGCACCTGGATC
CD206	TCTTTGCTTTCCAGTCTCC	TGACACCCAGCGGAATTTC
TGF-α	TTCTCATTCCTGCTTGTGG	ACTTGGTGGTTTGCTACG
IL-10	TGCCTTCAGTCAAGTGAAGAC	AAACTCATTCATGGCCTTGTA
Arg-1	TCCTTAGAGATTATCGGAGCG	GTCTTTGGCAGATATGCAGG
YM-1	CAGGGTAATGAGTGGGTTGG	CACGGCACCTCCTAAATTGT
GAPDH	TTGATGGCAACAATCTCCAC	CGTCCCGTAGACAAAATGGT

Western blotting

Proteins were extracted from brain tissue and cultured cells using radioimmunoprecipitation assay (RIPA) lysis buffer (Solarbio #R0020) supplemented with 1% protease inhibitor cocktail (Sigma #P8340) and 1% phosphatase inhibitor (Applygen #P1260) at 4 °C. The protein concentrations were determined using the BCA protein assay kit (Thermo Scientific, USA). Equal amounts of protein (30 µg per sample) were separated by 10% SDS‒PAGE and transferred onto PVDF membranes (Roche, USA). The membranes were blocked for 1 h at room temperature with 5% nonfat milk in TBST, followed by overnight incubation at 4 °C with primary antibodies. The primary antibodies used were against NLRP3 (1:1000, Proteintech #30109-1-AP), pro-caspase-1 (1:1000, AdipoGen #AG-20B-0042), p20 caspase-1 (1:1000, AdipoGen #AG-20B-0042), GSDMD (1:1000, Abcam #ab219800) and TREM2 (1:1000, Proteintech #68723-1-Ig). GAPDH (1:10000, Bioworld Technology #AP0063) was used as a loading control. After primary antibody incubation, the membranes were washed three times with TBST and incubated for 1 h at room temperature with a DyLight 800-conjugated goat anti-rabbit IgG (H&L) secondary antibody (1:10000, Rockland #611-145-122). The protein bands were detected using an Odyssey infrared imaging system (LICOR Biosciences, USA), and the band intensities were quantified using ImageJ software. The expression levels of NLRP3, pro-caspase-1, p20 caspase-1, GSDMD, and TREM2 were calculated relative to those of GAPDH in both in vivo and in vitro samples.

Flow cytometry analysis

The mice were deeply anesthetized and transcardially perfused with sterile PBS before brain tissue and blood collection. Blood was collected in 50 mM EDTA, while brain tissue was homogenized in ice-cold PBS containing DNase I (0.05 mg/mL, Sigma‒Aldrich #DN25) and collagenase (5 mg/mL, Sigma‒Aldrich) using a gentle MACS™ dissociator (Miltenyi Biotec). The myelin debris was removed via 30% Percoll gradient centrifugation. Viable cells were quantified via trypan blue (0.4%) exclusion on a hemocytometer. Single-cell suspensions were stained with Zombie NIR™ viability dye (1:100, BioLegend) to exclude dead cells, followed by Fc receptor blockade using anti-CD16/32 (1:50, BD Biosciences). Surface markers were labeled with fluorochrome-conjugated antibodies against CD45 (1:100), CD11b (1:100), P2RY12 (1:50), and CD86 (1:100) (all from BioLegend). After being washed, the cells were fixed and permeabilized using Cytofix/Cytoperm™ (BD Biosciences) for intracellular staining with an anti-CD206 antibody (1:50, BioLegend). Samples were analyzed on a BD FACSAria™ III flow cytometer, and the data were analyzed using FlowJo™ software (v10.5). The gating strategy for microglial identification is illustrated in Figure S5a.

Cytometric bead array (CBA)

The concentrations of IL-1β, IL-18, IL-6, and TNF-α in the astrocyte culture supernatants were determined using a BD CBA Mouse Soluble Protein Master Buffer Kit (BD Biosciences). Standard curves were prepared from samples that had been freshly serially diluted, and capture beads were mixed according to the manufacturer’s protocol. Briefly, 50 µL of bead suspension was combined with 50 µL of standards or samples in assay tubes and incubated for 1 h at room temperature in the dark. After adding 50 µL of PE detection reagent, the samples were incubated for an additional hour, washed with 1.0 mL of buffer, and centrifuged at 200 × g for 5 min. The pellet was resuspended in 300 µL of wash buffer and analyzed on a BD FACSAria II flow cytometer. The data were processed using FCAP Array software (v3.0, BD Biosciences).

Transcriptomic analysis

Total RNA was isolated from tissue samples using TRIzol^® reagent following the manufacturer’s protocol. The RNA integrity and concentration were assessed using an Agilent 5300 Bioanalyzer and NanoDrop 2000 spectrophotometer, respectively. Only high-quality RNA samples (OD260/280: 1.8–2.2, OD260/230 ≥ 2.0, RQN ≥ 6.5, 28 S:18 S ratio ≥ 1.0, total RNA > 1 µg) were selected for library preparation. mRNA sequencing libraries were constructed using the Illumina^® Stranded mRNA Prep Ligation Kit with poly(A) selection and fragmentation. Double-stranded cDNA was synthesized using random hexamer primers (SuperScript cDNA synthesis kit, Invitrogen), followed by end repair, adapter ligation, and size selection (300 bp fragments). PCR amplification was performed with Phusion DNA polymerase (15 cycles), and libraries were quantified via Qubit 4.0 before paired-end sequencing (PE150) on the NovaSeq X Plus platform (NovaSeq Reagent Kit). The raw reads were processed using fastp for quality trimming, and the clean reads were aligned to the reference genome (HISAT2) and assembled (StringTie). Differential gene expression analysis was conducted using DESeq2/DEGseq, with thresholds of |log2FC| ≥ 1 and FDR < 0.05 (DESeq2) or FDR < 0.001 (DEGseq). Functional enrichment analyses (GO and KEGG enrichment analyses) were performed using Goatools and Python scipy (Bonferroni-corrected P < 0.05). Alternative splicing events were identified via rMATS, with a focus on exon inclusion/exclusion, alternative splice sites, and intron retention.

Cell culture and experimental treatments

BV2 microglia (China Center for Type Culture Collection, RRID: CVCL_0182) were cultured in DMEM (Gibco) supplemented with 10% FBS (Gibco) and 1% penicillin‒streptomycin (Solarbio) at 37 °C in a 5% CO₂ humidified atmosphere. The cells were passaged every 48 h.

For experimental treatments, BV2 cells were subjected to OGD and/or treatment with isopropyl alcohol (IPA, 50 µM) for 6 h after 24 h of subculture. Three experimental groups were established: the control group (untreated cells), OGD group (cells exposed to OGD only) and OGD + IPA group (cells treated with OGD and 50 µM IPA).

OGD model

To mimic ischemic stroke in vitro, BV2 microglia were washed and incubated in glucose-free DMEM (Gibco) supplemented with 0.58 mg/mL L-glutamine and 1% penicillin‒streptomycin. The cells were then placed in a hypoxic chamber (5% CO₂, 95% N₂) at 37 °C for 6 h. Following OGD, the cells were returned to normal culture medium and maintained under normoxic conditions (5% CO₂/95% air) for 6 h.

In vitro lentiviral transduction

A lentiviral vector (hU6-MCS-CBh-gcGFP-IRES-puromycin, Genechem) encoding a short hairpin RNA targeting TREM2 (NM_031254.4) was used to establish TREM2-knockdown (TREM2-KD) BV2 microglia. Control cells were transduced with a GFP-expressing lentivirus (TREM2-NC). Both groups of cells were infected at a multiplicity of infection (MOI) of 5. After 48 h of incubation, the transduced cells were subjected to OGD. The transduction efficiency was verified by visualizing GFP expression via immunofluorescence, and the knockdown efficacy was quantitatively assessed via qRT‒PCR.

Intracerebroventricular virus delivery

To achieve microglia-specific TREM2 knockdown in vivo, an AAV vector (GV858 Iba1-promoter-EGFP-mir155(TREM2)-WPRE.miR-9.T.miR-129-2-3p.T.SV40p, Genechem) was stereotaxically injected into the right lateral ventricle of 6-week-old C57BL/6 mice. Control animals received an AAV carrying a scrambled sequence (GV858 Iba1-promoter-EGFP-mir155(MCS)-WPRE.miR-9.T.miR-129-2-3p.T.SV40p, TREM2-NC). The mice were anesthetized and secured to a stereotaxic frame, and 3 µL of viral suspension (1 × 10¹³ vg/µL, total dose 3 × 10¹³vg) was infused over 10 min at the following coordinates relative to bregma: anteroposterior (AP) = − 0.3 mm, mediolateral (ML) = + 1.0 mm, and dorsoventral (DV) = − 3.0 mm. The needle was kept in place for 10 min after injection to ensure diffusion of the virus. The animals were allowed to recover in their home cages and were monitored until subsequent procedures were performed. Three weeks after virus delivery, the mice underwent dMCAO. Successful microglial transduction was confirmed by EGFP immunofluorescence.

D. muris-containing medium

D. muris-containing medium was prepared by modifying standard chopped meat broth with carbohydrates (DSMZ medium 110) to exclude tryptophan components. In this customized medium, peptone, beef extract, yeast extract, and minced beef were replaced with yeast nitrogen base (YNB) and an amino acid supplement.

Statistical analyses

The data are presented as the means ± SD. The statistical analyses used for the metabolomics and proteomics data are provided in the respective sections. For normally distributed data, parametric tests were used; two-tailed Student’s t tests were used for comparisons between two groups, and one-way ANOVA followed by the least significant difference (LSD) test (for equal variances) or Dunnett’s test (for unequal variances) was used for comparisons among three or more groups. Nonnormally distributed data were analyzed using the Kruskal‒Wallis test with Dunn’s post hoc test for multiple group comparisons. Repeated-measures ANOVA followed by the LSD multiple comparison test was used to analyze data from experiments involving two categorical independent variables and one dependent variable. Sample sizes were calculated by the ‘resource equation’ method [88]. To control for Type I errors in multiple comparisons, adjustments were made using the LSD, Dunnett’s, Dunn’s, and Bonferroni’s methods. The sample size (n) for each analysis is specified in the corresponding figure legend, and statistical significance was set at P < 0.05. All statistical analyses were conducted and all graphical representations were generated using SPSS version 27.0 (SPSS, Chicago, IL), and curve fitting was performed using GraphPad Prism version 8.0.2 (GraphPad Software).

Authors’ contributions

Jun-Min Chen: Writing - original draft, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Cong Zhang, Lu-Lu Yu, Jian-Xu Sun, and Jiang-Hao Zhang: Writing - original draft, Methodology, Investigation, Formal analysis, Data curation. Lu Chen, Fei Zhu and Guang Shi: Methodology, Investigation, Formal analysis, Data curation. An-Chen Guo and Lan Yang: Writing - review and editing. Jian-Ping Wu, Tie-Shan Tang and Qun Wang: Supervision, Project administration, Funding acquisition, Writing - review and editing.

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information.

Declarations

Competing interests

The authors declare no competing interests.