An acetylation-dependent switch underlies host disease tolerance during streptococcal infection

Abstract

Building on our finding that Streptococcus pyogenes pyruvate dehydrogenase (PDH) suppresses host disease tolerance (DT) via short-chain fatty acid (SCFA)-mediated modulation of host acetyl-CoA and IL-10 levels, we characterize the global transcriptomic and epigenetic mechanisms underlying this immunometabolic manipulation. Combining new histological and ultrastructural analyses with an in-depth re-analysis of single-cell and bulk RNA-seq datasets to more comprehensively characterize the DT response, we show that PDH deficiency is associated with broad immunologic rewiring, characterized by intracellular bacterial containment within phagocytes, expansion of pro-resolving myeloid cells, and altered cell–cell communication. Metabolic analysis of ΔPdh-infected tissues revealed a shift away from acetyl-CoA metabolism towards glycolysis and the coordinated activation of a multi-faceted DT program, encompassing hypoxia signaling, iron handling, and the NRF2-mediated antioxidant response. Crucially, a focused re-analysis of existing transcriptome datasets from Histone Deacetylase (HDAC)-inhibited macrophages suggested that Trichostatin A (TSA) abrogates the protective transcriptome in ΔPdh infection, indicating that acetylation-dependent repression functions as a key regulator of the host DT response. By integrating new experimental data with advanced computational analyses, our work reveals a bacterial strategy of metabolic-epigenetic crosstalk, suggesting acetylation as a critical control point for mitigating infection-associated tissue damage.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-42565-1.

Introduction

The severity of infectious diseases is governed not only by pathogen burden but also by the host’s ability to limit collateral tissue damage, a process known as disease tolerance (DT)^1,2. While immune-mediated resistance mechanisms aim to reduce pathogen load, DT pathways maintain tissue homeostasis without directly impacting microbial fitness, thereby preserving organ function and promoting host survival during infection^3,4. These protective mechanisms encompass cytoprotective programs such as enhanced stress resistance, regulation of iron homeostasis, modulation of cellular metabolism, and control of inflammatory resolution^5–9. Despite their importance, the molecular mechanisms that underpin DT, and particularly how successful pathogens may actively manipulate them, remain incompletely understood.

Streptococcus pyogenes, also known as Group A Streptococcus (GAS), is a Gram-positive human pathogen responsible for an estimated 600 million cases of pharyngitis and 700,000 invasive infections worldwide each year^10–13. GAS causes a spectrum of diseases ranging from self-limiting pharyngitis and impetigo to life-threatening necrotizing soft tissue infections and streptococcal toxic shock syndrome¹⁴. Its virulence is closely tied to remarkable metabolic plasticity^15–17, which allows GAS to thrive in diverse host niches by dynamically regulating carbon flux through its central fermentation pathways^17–19. This flexibility enables the bacterium to adapt to fluctuations in nutrient availability and host-imposed stress, and we hypothesized that it also directly influences the host’s immune response, making metabolic regulation a key determinant of disease severity²⁰.

In our recent study, we demonstrated that GAS utilizes the pyruvate dehydrogenase (PDH) complex for aerobic mixed-acid fermentation (ARMAF) to actively suppress the host’s DT response and exacerbate tissue damage^21,22. We established a direct mechanism: PDH activity drives the production of immunomodulatory short-chain fatty acids (SCFAs), notably acetate and formate, which we have previously shown to be produced in an active PDH-dependent manner during infection. The functional significance of this metabolic activity is underscored by the significantly reduced levels of acetate and formate in ΔPdh infection compared to WT and by the fact that supplementation of these metabolites restores WT-like phenotypes, including host acetyl-CoA levels and repression of interleukin-10 (IL-10), a key immunosuppressive and tissue-repair-promoting cytokine^21,22. This work suggested bacterial central carbon metabolism as a critical regulator of host–pathogen dialogue.

However, this discovery raised several fundamental questions: While we identified IL-10 as a critical target, the global transcriptional reprogramming and systems-level immunologic changes underlying the PDH-dependent alteration in DT were undefined. Specifically, it was unclear: (1) whether the effect on IL-10 was part of a broader immunoregulatory shift involving canonical DT pathways, such as those regulated by NRF2^23,24 or HIF-1α²⁵; (2) how the altered bacterial metabolome reshaped the immune landscape at a cellular level, including cell–cell communication networks and myeloid cell functional states; (3) how bacterial SCFAs influenced host cell metabolic pathways to attenuate inflammation; and (4) the precise epigenetic mechanism by which bacterial metabolites and host acetyl-CoA converge to regulate transcription. Although we showed that the histone deacetylase (HDAC) inhibitor Trichostatin A (TSA) suppressed IL-10²¹, it remained unclear whether this was mediated through histone acetylation or via acetylation of non-histone regulatory proteins^26–29, and what the full transcriptomic impact of this modulation would be.

To address these gaps, we conducted an in-depth, systems-level analysis that integrated new experimental data with advanced computational re-analysis. We performed histological and transmission electron microscopy (TEM) analyses to characterize the cellular and ultrastructural changes in infected tissue. We also conducted an advanced re-analysis of our existing single-cell RNA sequencing (scRNA-seq) dataset from infected tissue²¹, applying sophisticated computational tools to dissect immune cell communication, metabolic states, and transcriptional trajectories. Furthermore, to demonstrate the role of acetylation, we conducted a re-analysis of bulk RNA-seq data from macrophages treated with TSA²¹. This combined approach allowed us to move beyond correlation to mechanism.

Our results show that PDH-dependent manipulation of host DT extends beyond IL-10 suppression, encompassing a coordinated cytoprotective transcriptome, a reprogrammed myeloid compartment, and a distinct host metabolic state. Crucially, we demonstrate that these effects appear to be integrated through acetylation of non-histone regulatory proteins, suggesting bacterial metabolism as a key regulator of the host’s epigenetic response to infection. This work provides a framework for targeting bacterial carbon flux to enhance DT.

Results

In-depth analysis of the host immune landscape reveals the transcriptional basis of PDH-mediated DT manipulation

Our prior work established that aerobic mixed-acid fermentation (ARMAF) via PDH enables GAS to suppress IL-10 and exacerbate tissue damage²¹. While we previously reported that ΔPdh infection results in smaller, contained lesions with a dense inflammatory infiltrate²¹, the global transcriptional reprogramming underlying this altered disease tolerance phenotype remained undefined.

To gain a system-level understanding of the host pathways manipulated by bacterial carbon flux, we conducted an advanced re-analysis of our existing single-cell RNA sequencing (scRNA-seq) dataset from WT- and ΔPdh-infected skin tissue²¹, moving beyond initial clustering to demonstrate system-level changes in immune coordination and cellular states (Fig. 1a, schematic of experimental procedure aligned with the timeline of infection progression). This was contextualized with new histological examination, which confirmed the stark pathological divergence: WT infection generated large, expanding ulcers with extensive necrotic cores, whereas ΔPdh infection produced smaller, contained lesions with more preserved tissue architecture and a dense inflammatory infiltrate at the margin (Fig. 1b). Quantification of lesion area²¹ and inflammatory infiltrate density confirmed these observations (Fig. 1c). This early containment phenotype is critical, as longitudinal analysis has shown that ΔPdh lesions stall after day 1 and do not expand further, facilitating faster bacterial clearance and complete tissue healing by 12 days post-infection (dpi), in contrast to the progressive tissue damage caused by WT infection²¹.

Fig. 1.

Histological and single-cell transcriptomic re-analysis reveals the global transcriptional basis of enhanced disease tolerance in ΔPdh-infected tissue. (a) Schematic of the murine ulcer infection model and experimental timeline. (b) Hematoxylin and eosin (H&E) staining of infected skin sections (representative images). Infection zones are outlined by dashed red lines. Labels: E – epidermis; D – dermis; PC – panniculus carnosus; FA – fascia; M – muscle; ES – eschar; C – core; MA – margin. (c) Quantification of histological features in WT- and ΔPdh-infected skin lesions. Density of the inflammatory infiltrate (cells/mm²) at the lesion margin. Each data point represents one animal. *p < 0.05, two-tailed Student’s t-test. (d) UMAP projection of integrated scRNA-seq data (dataset from Xu et al., Nat Commun, 2025) from WT- and ΔPdh-infected tissues, showing transcriptional clustering across major immune cell populations. Abbreviations: MAC, macrophages; Neu, neutrophils; M-MdM, monocyte-derived macrophages; NK, natural killer cells; DCs, dendritic cells. (e) Volcano plot of differentially expressed genes (DEGs) between WT- and ΔPdh-infected tissue (across the integrated dataset), identified in this study using Seurat’s FindMarkers() with a Wilcoxon rank-sum test and Benjamini–Hochberg correction. Genes with adjusted p < 0.05 and |log₂(fold change)|> 1 were considered significant. Red and blue points represent up- and downregulated genes in ΔPdh relative to WT; gray points indicate non-significant genes. (f) Gene Ontology (GO) enrichment analysis of DEGs (criteria as in e) performed in this study using the clusterProfiler R package. Dot size indicates the number of DEGs per term; color scale reflects adjusted p-value; gene ratio = number of DEGs in term/total genes annotated to that term. (g) Reactome pathway enrichment analysis of upregulated DEGs (criteria as in (e)) performed in this study using the Enrichr platform (Reactome 2024 database). Pathways were ranked by adjusted p-value, and bar length represents –log₁₀ of the adjusted p-value.

We next applied advanced computational approaches to the scRNA-seq data to dissect the molecular basis of these differences. Unsupervised clustering of the integrated dataset recapitulated the major immune populations previously annotated in this model²¹, including macrophages (MAC), neutrophils (Neu), monocyte-derived macrophages (M-MdM), dendritic cells (DCs), natural killer (NK) cells, T cells, and B cells (Fig. 1d). A comprehensive differential gene expression analysis across the integrated dataset revealed broad transcriptional differences in ΔPdh-infected tissue compared to WT (Fig. 1e). This analysis identified 428 upregulated and 265 downregulated genes (adjusted p < 0.05, |log₂FC|> 1), including key mediators such as Ftl1 and Csf1 (upregulated) and Ly6a and Igkc (downregulated) (Fig. 1e). Subsequent functional enrichment analysis was performed using two complementary approaches. Gene Ontology (GO) analysis revealed broad processes related to immune regulation and cellular stress responses (Fig. 1f). In parallel, Reactome pathway analysis highlighted curated immune signaling pathways, most notably revealing an enrichment for 'IL-10 signaling’ (Fig. 1g). This suggests that the influence of altered bacterial metabolism induces a broad, functionally coherent shift in the host immune transcriptome.

ΔPdh infection remodels cellular architecture and immune communication networks

To define how these transcriptomic shifts alter cellular pathophysiology at a systems level, we combined new ultrastructural analysis with advanced computational modeling of intercellular communication. We performed new transmission electron microscopy (TEM) to visualize host–pathogen interactions at high resolution. We focused on 3 dpi, a time point where ∆Pdh lesions are contained but bacterial burdens are equivalent to WT, a hallmark of DT²¹. While our previous work indicated intracellular localization of ΔPdh bacteria within macrophages²¹, the TEM analyses reveal a broader containment phenotype. WT bacteria were frequently associated with lytic damage and disrupted host cells, consistent with a cytotoxic pathogenesis strategy (Supplementary Fig. 1). In striking contrast, ΔPdh bacteria were contained within intact phagocytes (Fig. 2a), with individual phagocytes harboring numerous bacteria while maintaining cellular integrity. Quantification of bacterial localization revealed a significantly higher proportion of intracellular bacteria within phagocytes in ΔPdh-infected tissue compared to WT (Fig. 2b). Critically, this containment phenotype is not due to enhanced bacterial clearance, as we have previously established that WT and ΔPdh infections yield similar tissue burdens at 3 dpi²¹. These findings provide ultrastructural evidence that phagocytes can persistently contain ΔPdh without undergoing rapid lysis, whereas WT bacteria overwhelm host defenses.

Fig. 2.

ΔPdh infection alters cellular ultrastructure and immune communication networks. (a) Transmission electron micrographs of formalin-fixed tissue sections from C57BL/6 J mice infected with WT or ΔPdh S. pyogenes (n = 5 mice per group, this study). Representative images show cellular ultrastructure and bacterial localization (arrow). (b) Quantitative analysis of bacterial localization by TEM. Percentage of bacteria observed in an intracellular location in WT- and ΔPdh-infected tissue sections. n = 3 fields of view per mouse section. Statistical significance was assessed using a two-tailed Student’s t-test. ***, p < 0.001. (c) Cell–cell communication networks inferred by CellChat from integrated scRNA-seq data (dataset from Xu et al., Nat Commun, 2025). Ribbons represent predicted ligand-receptor interactions between immune cell types; red edges denote enhanced signaling in ΔPdh infection, blue in WT. Node size corresponds to the number of incoming and outgoing interactions. (d) UMAP showing log-normalized transcription of Il10 across the integrated scRNA-seq data. Each point represents an individual cell, colored by expression level. (e, f), Principal component analysis (PCA) of chemokine-related (e) and cytokine-related (f) gene expression across WT and ΔPdh conditions. (g) Partial Least Squares Discriminant Analysis (PLS-DA) was performed on the expression matrices of cytokine and chemokine-related genes across WT and ΔPdh-infected tissues. The top genes contributing to this separation, ranked by VIP scores, are listed in Supplementary Table 3.

We next applied CellChat, a computational approach not previously used on this dataset, to model ligand-receptor signaling networks and assess global immune coordination. This analysis revealed a striking difference in cell–cell communication that was not apparent from the UMAP of scRNAseq (Fig. 2c). WT infection was characterized by prominent autocrine signaling loops among neutrophils and macrophages, reflecting a disordered, self-amplifying inflammatory response (Supplementary Table 1). Conversely, ΔPdh infection featured enhanced paracrine crosstalk, with strengthened interactions between macrophages and neutrophils and between dendritic cells and T cells, pointing to a more orchestrated and regulated immune program (Supplementary Table 2). In this ΔPdh-specific network, T cells, NK cells, M-MdM, and monocytes emerged as potential key signal coordinators, exhibiting strong outgoing signaling edges that may drive the altered cytokine landscape and myeloid cell reprogramming observed in the tolerant state.

This rewired communication network was coupled to a shift in cytokine expression (Supplementary Fig. 2, 3). Re-analysis of the scRNA-seq data confirmed that, similar to our previous discovery²¹, elevated Il10 transcription was a defining feature of ΔPdh infection, originating primarily from macrophages, neutrophils, and dendritic cells (Fig. 2d, Supplementary Fig. 2). This increase was consistent with our prior quantification of IL-10 levels in infected tissue and macrophages^21,22. This finding is integrated within a systems-level analysis, suggesting IL-10 induction as a central component of the broader, coordinated DT program. WT lesions, in contrast, exhibited a more diffuse and significantly lower IL-10 transcription. To further characterize these immune landscapes, we performed principal component analysis (PCA) on chemokine and cytokine gene expression. The analysis revealed clear separation between WT and ΔPdh infections for both chemokine (Fig. 2e, Supplementary Fig. 4) and cytokine (Fig. 2f) expression profiles. Notably, the 95% confidence ellipses indicated that WT-infected samples cluster more tightly, reflecting a consistent, stereotyped cytokine response, whereas ΔPdh-infected samples exhibit greater heterogeneity, suggesting a broader spectrum of activation states (Fig. 2f). To complement this analysis and better capture condition-specific patterns, we performed supervised partial least squares discriminant analysis (PLS-DA), which showed enhanced separation of chemokines and cytokines (Fig. 2g, Supplementary Table 3). Consistent with this separation, we observed a selective elevation of specific pro-inflammatory mediators during ΔPdh-infection (Supplementary Fig. 5). Together, these findings suggest that ΔPdh infection does not simply suppress inflammation but instead reprograms immune signaling toward a more coordinated and focused response.

Advanced transcriptional profiling reveals a coordinated pro-resolving reprogramming of myeloid cells in ΔPdh infection

Building on our initial observation of myeloid expansion in ΔPdh-infected tissue²¹, we performed a focused re-analysis of the scRNA-seq data to characterize the functional state of these cells. This deeper examination revealed a comprehensive reprogramming of the host immune landscape. The most prominent feature was a dramatic amplification of the innate myeloid compartment, with a two-fold increase in both pro-inflammatory M1 and anti-inflammatory M2 macrophages and all neutrophil subsets (Table 1, Fig. 3a). This was accompanied by a concomitant reduction in B cell populations, including memory B cells, plasmablasts, and transitional B cells (Table 1). CD4⁺ T cells showed a modest increase, whereas CD8⁺ T cells and regulatory T cells remained largely unchanged (Table 1).

Table 1.

Immune cell % from WT‑ and ΔPdh‑infected tissues.

Immune cell subset*	WT infection	ΔPdh infection	Fold change (ΔPdh/WT)
Dendritic cells
mDC	0.80	0.3	0.38
pDC	0.10	0.00	0.00
B cells
Memory B cells	16.64	4.16	0.25
Plasmablasts	3.10	0.96	0.31
Transitional B cells	10.30	2.96	0.29
T cells
CD4 +	0.72	1.96	2.27
CD8 +	3.40	2.7	0.79
MAIT	0.00	0.04	NA
Tfh	0.04	0.04	1.00
Th1	0.06	0.04	0.67
Th17	0.00	0.02	NA
Th2	0.00	0.02	NA
Treg	0.38	0.32	0.84
γδTCR + Vδ2 +	0.24	0.22	0.92
Macrophages
M1	3.60	12.42	3.45
M2	4.52	8.80	1.95
Neutrophils
Neu1	4.08	8.34	2.04
Neu2	3.86	8.54	2.21
Neu3	3.44	4.74	1.38
Monocytes
Classical	1.22	0.72	0.59
Intermediate	1.22	1.18	0.97
Non-classical	0.02	0.02	1.00
NK cells
Immature	0.16	0.2	1.25
Mature	0.06	0.1	1.67
Effector	0.02	0.00	0.00

* Values indicate the percentage of each annotated immune cell type out of the total cells sequenced by scRNA‑seq. Cell identity was assigned via marker‑based clustering and canonical gene expression profiles. Rare populations (e.g. MAIT, Th17) are reported at low percentages (< 0.05%).

Fig. 3.

ΔPdh infection drives a coordinated pro-resolving reprogramming of myeloid cells. (a) Quantification of major immune cell populations from scRNA-seq data, showing the relative abundance in WT vs. ΔPdh-infected tissue. Bars represent the percentage of total cells for each population. (b, d) Heatmaps of the top 15 differentially expressed genes (DEGs) in neutrophils (b) and macrophages (d), comparing ΔPdh vs WT infection. DEGs were identified using the Wilcoxon rank-sum test with Bonferroni correction (adjusted p < 0.05). Expression values are z-scored across cells (blue to red). (c, e) Pathway enrichment analysis of DEGs upregulated in ΔPdh-infected neutrophils (c) and macrophages (e). DEGs were defined as in b/d. Enrichment was performed using the Enrichr platform with Reactome 2024 and WikiPathways databases, and pathways were ranked by adjusted p-value (Fisher’s exact test). Bar length represents –log10 of the adjusted p-value; only pathways with adjusted p < 0.05 are shown.

Given the predominance of neutrophils and macrophages in the inflammatory infiltrate, we next focused on characterizing their transcriptional programs in response to ΔPdh infection. UMAP visualization revealed that neutrophils and macrophages from ΔPdh-infected lesions formed distinct clusters separate from their WT counterparts (Supplementary Fig. 6). ΔPdh-infected neutrophils exhibited upregulated Srgn (serglycin), Fau (ribosomal protein S30 fusion), Ftl1 (ferritin light chain 1), Fth1 (ferritin heavy chain 1), and Gadd45b (growth arrest and DNA-damage-inducible beta), consistent with oxidative stress management and cell survival (Fig. 3b)^30–33. Conversely, they downregulated a set of pro-inflammatory and tissue-disruptive effectors, including Retnlg (resistin-like gamma), Lcn2 (lipocalin-2), Plac8 (placenta-specific 8), Mmp8 (matrix metalloproteinase-8), Ifitm6 (interferon-induced transmembrane protein 6), S100a6 (S100 calcium-binding protein A6), Mt-Col1 (mitochondrial cytochrome c oxidase subunit 1), and Wfdc21 (WAP four-disulfide core domain protein 21)^34–36. Macrophages displayed a partially overlapping transcriptional program with neutrophils (Fig. 3d). Genes upregulated in ΔPdh-infected macrophages included Ftl1, Fau, Fth1, Tyrobp (DAP12), and Gadd45b, suggesting coordinated stress resilience and metabolic reprogramming. In contrast, ΔPdh-infected macrophages downregulated Plac8, Actb (beta-actin), S100a6, Saa3 (serum amyloid A3), Retnlg, S100a8, and Lcn2, pointing to suppression of inflammatory and matrix-degrading activity.

Pathway analysis of upregulated DEGs, using Reactome and WikiPathways, aligned with these transcriptional shifts. ΔPdh-infected neutrophils were enriched for processes linked to iron handling, hepcidin regulation, and apoptosis (Fig. 3c), while ΔPdh-infected macrophages showed enrichment for pathways governing neutrophil degranulation, extracellular matrix remodeling, and myeloid differentiation (Fig. 3e). Additionally, a heatmap of genes from these top enriched pathways further illustrates the coordinated transcriptional changes (Supplementary Fig. 7). Together, these results suggest how PDH-mediated bacterial metabolism reprograms myeloid cells toward stress-adaptive, tissue-reparative states.

Altered bacterial carbon flux reprograms host cell metabolism through acetyl-CoA availability

Our previous work established that PDH-mediated ARMAF in GAS is a major source of the immunomodulatory SCFAs acetate and formate, and that loss of PDH (ΔPdh) markedly reduces their production²¹. Supplementation with acetate and formate during ΔPdh infection of macrophages restores IL-10 levels²¹. We therefore hypothesized that the reduction in bacterial-derived acetate and formate creates a metabolic deficit that reprograms host cell acetyl-CoA metabolism and energy utilization. To test this, we computed module scores for major metabolic pathways from our scRNA-seq dataset. ΔPdh infection shifts in glycolytic gene expression, with significantly elevated glycolysis scores compared to WT across multiple cell types (Fig. 4a; Supplementary Fig. 8; see Supplementary Table 4 for per-cell-type statistics). This suggests a potential metabolic reprogramming that warrants future investigation. Consistent with our previous measurements of reduced acetate, formate, and acetyl-CoA in ΔPdh-infected macrophages^21,22, acetyl-CoA metabolism was transcriptionally suppressed, as indicated by reduced acetyl-CoA module scores (Fig. 4b; Supplementary Fig. 8; Supplementary Table 4). No major changes were observed in downstream mitochondrial pathways, including oxidative phosphorylation and fatty acid β-oxidation (Supplementary Fig. 9).

Fig. 4.

Altered carbon metabolism in WT- and ΔPdh-infected tissue. (a, b) Single-cell module scores for glycolysis (a) acetyl-CoA metabolism (b). Module scores were calculated using Seurat’s AddModuleScore() function based on curated gene sets, reflect the relative activity of the pathway in each cell. The glycolysis module included Hk1, Pfkfb3, Ldha, Eno1, Hk2, Aldoa, Gapdh, Pgk1, Pkm, and Gpi; the acetyl-CoA metabolism module included Acly, Acss2, Acaca, Slc25a1, Acss1, Fat, Fabp4, Sirt3 and Pdha1. Each point represents a single cell, colored by its log-normalized module score, reflecting relative pathway activity. c, Pseudotime analysis of Acly and Acss2 expression along neutrophil (Neu1 → Neu3) and macrophage (MAC1 → MAC2) differentiation trajectories. Line plots show log-normalized expression levels across pseudotime. Trajectories were inferred using Monocle 3 from scRNA-seq data.

To understand the dynamics of this metabolic shift, we performed pseudotime analysis on neutrophil and macrophage differentiation trajectories. This revealed that ΔPdh-infected neutrophils and macrophages upregulated acetyl-CoA synthetase 2 (Acss2) more strongly than WT infection, whereas WT-infected neutrophils preferentially induced ATP-citrate lyase (Acly) (Fig. 4c). A UMAP visualization of these trajectories is provided in Supplementary Fig. 10. These data indicate that loss of bacterial PDH activity shifts host cells toward alternative acetyl-CoA synthesis routes, likely reflecting a compensatory response to reduced bacterial SCFAs. In addition, ΔPdh-infected lesions exhibited increased expression of Histone Deacetylase 1 (Hdac1) and reduced levels of the histone acetylation mark H3K9ac1 (Supplementary Fig. 11; Supplementary Table 4), consistent with impaired acetyl-CoA availability for chromatin modification. Together, these findings demonstrate that bacterial carbon flux directly regulates host immunometabolism, diverting immune cells toward glycolysis and away from mitochondrial acetyl-CoA-dependent programs that may contribute to the immunoregulatory state observed in ΔPdh infection.

ΔPdh infection activates a coordinated DT program

Having established the metabolic and cellular reprogramming in ΔPdh infection, we next sought to define the comprehensive DT transcriptome that underlies the attenuated tissue damage. To systematically identify DT-associated genes, we curated a gene set encompassing established cytoprotective pathways^37–39 and performed module scoring analysis on our scRNA-seq data.

This analysis revealed broad induction of a coordinated DT program in ΔPdh-infected lesions (Fig. 5a), reflected in significantly higher DT module scores across macrophages, neutrophils, and dendritic cells compared to WT infections (quantified by cell type in Fig. 5b; Supplementary Table 4, Supplementary Fig. 12). This protective program encompassed multiple cytoprotective pathways, as visualized by the coordinated upregulation of key genes in the ΔPdh condition (Fig. 5a, Supplementary Fig. 13), including those involved in hypoxia signaling (Hif1a, Vegfa), iron homeostasis (Hmox1, Fth1), the NRF2-mediated oxidative stress response (Nfe2l2), and AMPK–mTOR signaling (Prkaa1). The increased activity of these complete pathways was confirmed by elevated module scores across immune cells (Supplementary Fig. 14). Together, this indicates that ΔPdh infection orchestrates a coordinated transcriptional program to limit tissue damage and maintain cellular homeostasis. Functionally, this transcriptional reprogramming was linked to pro-resolution outcomes. ΔPdh-infected lesions showed enhanced activation of wound healing pathways, most prominently VEGF-driven angiogenesis (Vegfa, Kdr) and ALOX-mediated lipid mediator synthesis (Alox5, Alox15), whereas other canonical repair pathways, such as PDGF or TGFβ signaling, were not significantly induced (Fig. 5c, Supplementary Fig. 15). Macrophage polarization analysis of the macrophage subsets (clusters MAC1 and MAC2) revealed that WT-infected macrophages were skewed toward inflammatory M1-like states (IL-1β, TNF-α, Nos2), whereas ΔPdh-infected macrophages favored regulatory, pro-repair M2-like states (IL-10, Arg1, Mrc1) (Fig. 5d–f). Finally, we assessed regulated cell death, a major driver of immunopathology. ΔPdh infection selectively upregulated genes associated with apoptosis (Bax, Fas, Tnfrsf1a) and autophagy (Sqstm1, Map1lc3b), processes that facilitate controlled clearance of damaged cells. In contrast, genes associated with lytic and highly inflammatory pathways, pyroptosis (e.g., Casp1, Nlrp3, Gsdmd) and ferroptosis (e.g., Acsl4, Gpx4), did not show a consistent pattern of marked upregulation (Supplementary Fig. 16). Furthermore, while autophagy-related genes such as Map1lc3b were elevated, Becn1 expression did not differ substantially between conditions. This precise modulation of cell death likely limits collateral tissue damage and may represent a hallmark mechanism of DT.

Fig. 5.

ΔPdh infection promotes DT programs in skin tissue. (a) Heatmap showing relative expression of DT–associated genes in ΔPdh-infection compared to WT. Gene expression was log-normalized and scaled across cells. (b) DT module scores across major immune cell types in WT vs. ΔPdh-infected tissues. A DT gene module score was calculated per cell as the mean log-normalized expression of a curated DT gene set (examples shown on plot: Hif1a, Epas1, Vegfa, Hmox1, Slc40a1, Fth1, Ftl1, Ahr, Cyp1a1, Cyp1b1, Il10, Tgfb1). Per-cell DT scores were z-scored across all cells to enable comparison across cell types, then summarized by major immune cell type and condition. Bars show mean ± SE for WT and ΔPdh infection. (c) Wound healing pathway module scores across major immune cell types in WT vs. ΔPdh-infected tissues. Module scores for six wound healing-associated pathways (TGF-β, VEGF, PDGF, ALOX, MMPs, and ECM remodeling) were calculated per cell as the mean log-normalized expression of curated gene sets for each pathway. Per-cell scores were summarized by major immune cell type and condition. Bars show mean ± SE. (d–f), Violin plots comparing macrophage polarization module scores between WT (turquoise) and ΔPdh-infected (red) cells. The M1 module included Il1β, Ccl2, Cxcl2, Nos2, Il6, Tnfα, and the M2 module included Il10, Arg1, Mrc1. Module scores were computed using Seurat’s AddModuleScore(). Each dot represents an individual cell; violin width reflects score distribution. Elevated scores indicate increased transcriptional activity of polarization-associated genes.

HDAC inhibitor transcriptome establishes acetylation as a master regulator of the immunoregulatory program

Our previous work used the HDAC inhibitor Trichostatin A (TSA) as a tool to demonstrate that acetylation plays a key role in the host response to bacterial metabolites, as TSA suppressed IL-10 expression in infected macrophages²¹. However, the global transcriptional impact of this epigenetic modulation and its influence on the broader DT program remained undefined. To define the acetylation-dependent transcriptome, we re-analyzed bulk RNA-seq data for WT- and ΔPdh-infected RAW 264.7 macrophages with and without TSA treatment (Fig. 6a). As anticipated, ΔPdh infection alone induced a starkly different transcriptional profile compared to WT, characterized by significant upregulation of IL-10, and downregulation of Csf2 (GM-CSF) and Irf4 (Interferon regulatory factor 4) (Fig. 6b). Gene Ontology enrichment analysis of the upregulated DEGs in ΔPdh-infected macrophages (adjusted p < 0.05, log₂FC > 1) revealed pathways linked to "regulation of response to biotic stimulus," "actin filament organization," and “lymphocyte differentiation” (Fig. 6c). The downregulated DEGs were less numerous and did not show significant enrichment for specific pathways^21,22. TSA treatment profoundly altered the macrophage response in both infections. While TSA had a lower effect on the already-acetylated WT transcriptome, it triggered extensive gene repression in ΔPdh-infected cells, markedly dampening the expression of key immunoregulatory genes, including Il10, Il1α, Pdk1, Cxcl2, and Cxcl3 (Fig. 6b, Supplementary Fig. 17). ΔPdh infection promoted a M1-like program, indicated by elevated Il6 and Nos2 expression compared with WT infection, but induced a stronger M2-like phenotype characterized by high Il10 levels (Fig. 6d). TSA treatment suppressed this polarization, reducing the expression of both M1 and M2 signature genes and effectively nullifying the distinct polarization states induced by the two bacterial strains. Furthermore, ΔPdh infection in vitro upregulated genes associated with both apoptosis and, in contrast to our in vivo data, pyroptosis (Fig. 6e). This discrepancy in cell death pathways highlights a limitation of the in vitro system, suggesting that the full DT program, which in vivo suppresses lytic pyroptosis, requires a more complex tissue microenvironment. Nonetheless, the central finding holds: TSA abrogates the distinct transcriptional programs driven by bacterial carbon flux, demonstrating that acetylation is a master regulator of the macrophage response to infection.

Fig. 6.

Re-analysis of the transcriptomic impact of HDAC inhibition on infected macrophages. (a) Experimental design: RAW 264.7 macrophages were pre-treated with TSA or DMSO, then infected with WT or ΔPdh for RNA-seq analysis (data re-analyzed). (b) Volcano plot of differentially expressed genes (DEGs) between ΔPdh‑ and WT‑infected macrophages. DEGs were identified using DESeq2 with Wald tests, with p‑values adjusted by the Benjamini–Hochberg method. Genes with adjusted p < 0.05 and |log₂FC|> 1 were considered significant. (c) GO enrichment analysis of upregulated DEGs (adjusted p < 0.05, log₂FC > 1), performed using clusterProfiler. Over-representation in GO Biological Process terms was tested via a hypergeometric test with Benjamini–Hochberg correction. Dot size reflects the number of DEGs associated with each GO term; color indicates adjusted p-value. Gene ratio = number of DEGs in the term/total genes annotated to the term. (d) Violin plots showing sample-level macrophage polarization module scores derived from the bulk RNA-seq dataset using Seurat’s AddModuleScore() function. The M1 module included Il1β, Ccl2, Cxcl2, Nos2, Il6, and Tnfα; the M2 module included Il10, Arg1, and Mrc1. Each violin represents the distribution of module scores across biological replicates within each condition. (e) Violin plots showing relative expression of gene sets associated with regulated cell death pathways in WT- and ΔPdh-infected macrophages. Module scores were computed using AddModuleScore() from curated gene sets: Apoptosis: Apaf1, Bax, Bcl2, Casp3, Casp8, Casp9, Fadd, Fas, Tnfrsf1a; Autophagy: Atg5, Atg7, Becn1, Dapk1, Map1lc3b, Sqstm1, Ulk1; Pyroptosis: Aim2, Casp1, Casp11, Casp4, Gsdmd, Il1β, Nlrp3, Pycard. Higher module scores reflect increased transcriptional activation of the corresponding cell death program.

Discussion

Our previous work established that PDH-dependent metabolism in GAS manipulates the host’s DT response by repressing IL-10 through a process involving bacterial-derived short-chain fatty acids (SCFAs) and host acetyl-CoA^21,22. In this study, we extend these findings to define the full immunometabolic reprogramming elicited by bacterial carbon flux. By combining new histological and ultrastructural analyses with advanced re-analysis of single-cell and bulk transcriptomic data, our analysis suggests new mechanistic insight. We show that loss of PDH activity reshapes the entire immune landscape through expansion of pro-resolving myeloid cells, altered cell–cell communication, and a metabolic shift away from acetyl-CoA metabolism. Crucially, we describe a coordinated cytoprotective transcriptional program encompassing hypoxia signaling, iron homeostasis, and NRF2-mediated oxidative stress responses, which we propose is regulated by an acetylation-dependent switch. Collectively, these findings suggest bacterial central carbon metabolism as a key orchestrator of host immune composition, metabolic rewiring, and epigenetic regulation, linking microbial metabolism to a tissue-protective transcriptome.

The immune landscape in ΔPdh-infected tissue revealed a dramatic expansion and reprogramming of the myeloid cells, consistent with patterns observed in other tolerance models where innate immune mechanisms take precedence over adaptive responses that can drive immunopathology¹. This shift, marked by a decrease in B cell populations alongside the expansion of macrophages and neutrophils, provides a cellular basis for skewing toward DT, potentially driven by altered chemokine signaling or resource allocation that favors innate effector recruitment. The concomitant increase in transcripts associated with both M1 and M2 macrophages (Table 1) further suggests a state of recalibrated inflammation, enabling concurrent bactericidal activity and tissue repair functions⁴⁰. Furthermore, the expanded neutrophil population, which exhibits enhanced bacterial uptake, appears to be functionally redirected. Their transcriptional profile, indicating stress management (e.g., Fth1, Gadd45b upregulation) and downregulation of tissue-disruptive effectors (e.g., Mmp8), suggests a fate geared toward containment rather than lytic, tissue-damaging clearance. The observed intracellular accumulation of ΔPdh bacteria within neutrophils likely stems from this host reprogramming, rather than a loss of bacterial anti-phagocytic defenses, as we did not observe significant changes in key virulence factors (e.g., SLO, SpeB, SLS) in the ΔPdh mutant. Together, this reprogramming of innate effectors represents a coordinated activation of classic DT modules, characterized by upregulation of Hmox1 (which generates the potent antioxidant biliverdin⁴¹), Fth1/Ftl1 (which limit iron-catalyzed ROS production via ferritin nanocage formation⁴²), and Nfe2l2 (which coordinates the expression of over 200 cytoprotective genes⁴³). This recalibration again represents balanced inflammation, evidenced by the concurrent, controlled expression of specific inflammatory mediators (Cxcl2, Cxcl3, Il1β)^44–46. This state permits essential leukocyte recruitment for bacterial clearance⁴⁷, while limiting excessive pro-inflammatory signaling that can drive tissue destruction. This controlled inflammatory profile parallels the “containment” programs observed in chronic infections such as tuberculosis, where host immunity restricts pathogen expansion without triggering self-damaging hyperinflammation^48–50.

While supplementation with acetate and formate restores a WT-like phenotype²¹, confirming a direct role for these metabolites, we acknowledge that deletion of the central metabolic node Pdh may have pleiotropic effects. The observed DT phenotype likely results from a combination of depleted immunomodulatory SCFAs and broader bacterial adaptations to altered carbon flux. Notably, acetate and formate have roles as both metabolic precursors for host acetyl-CoA and as signaling molecules that engage specific receptors and epigenetic regulators²⁶, a functional versatility that distinguishes them from other potential carbon substrates.

The metabolic crosstalk identified here reveals a sophisticated interkingdom signaling mechanism. PDH-derived acetate functions not merely as a carbon source but as a paracrine signaling molecule that integrates with host metabolism through multiple pathways: (1) direct assimilation into acetyl-CoA via acetyl-CoA synthetase (Acss2)^51,52, (2) modulation of G protein-coupled receptor signaling (e.g., Gpr43)^53–55, and (3) potential regulation of AMPK activity affecting mitochondrial function⁵⁶. Our results show that bacterial-derived SCFAs sustain host acetyl-CoA pools, suggesting bacterial central carbon metabolism upstream of a critical metabolic hub that coordinates cellular energy balance, biosynthetic activity, and epigenetic control. However, transcriptomic changes do not always reflect metabolite levels, and future comprehensive metabolomic profiling of glycolytic, TCA‑cycle, and fatty‑acid‑oxidation intermediates will be important to fully characterize the immunometabolic rewiring induced by altered bacterial carbon flux.

An important finding of this study points to an acetylation-dependent mechanism. Our re-analysis of the HDAC inhibitor transcriptome provides a crucial insight: the regulation of the DT program is mediated by an acetylation-dependent switch that likely targets non-histone proteins. While HDAC inhibitors typically promote gene activation by increasing histone acetylation, our observed repression suggests predominant non-histone targeting⁵⁷. Critically, we found that infection did not alter global histone H3 acetylation, ruling out a primary role for histone acetylation in this regulation²¹. This aligns with emerging recognition that HDACs regulate critical signaling proteins through deacetylation, including: (1) NF-κB p65, where acetylation enhances DNA binding and transcriptional activity^58,59; (2) STAT3, whose acetylation regulates dimerization and nuclear translocation⁶⁰; and (3) FoxO1, where acetylation modulates apoptotic and metabolic programs⁶¹. The requirement for combined HDAC class inhibition²¹ suggests GAS manipulates a redundant network of acetylation-dependent regulators, ensuring robust control over host DT programs. This may represent an elegant evolutionary strategy where pathogens target conserved regulatory nodes rather than individual pathways.

Our work supports a model where that bacterial metabolic flux, as governed by PDH, serves as an environmental sensor that directly shapes the host’s DT response. The enzymatic activity of PDH is known to be post-translationally regulated by local oxygen tension^62–66. This regulatory link leads us to propose that the ΔPdh mutant may phenocopy the metabolic state of WT bacteria in specific host niches. For instance, the attenuated, pro-tolerant host response we observe in ΔPdh infection might mirror the host–pathogen dynamic in an anaerobic abscess, where PDH activity is naturally low. Conversely, the tissue-destructive pathology of WT infection likely reflects its high PDH activity in more oxygenated environments, such as the bloodstream or tissue interfaces during necrotizing fasciitis. This model provides a metabolic-epigenetic framework to explain why the severity of streptococcal diseases varies dramatically between different host compartments. The bacterium’s central carbon metabolism acts as a translator of local environmental cues, which in turn rewire host immunometabolism and epigenetic programs to either exacerbate or mitigate tissue damage.

The discrepancy between in vivo and in vitro pyroptosis responses⁶⁷ highlights the critical importance of tissue microenvironmental context. The full DT program likely requires integrated signals from stromal cells (e.g., fibroblast IL-33 production^68,69), tissue-specific metabolites (e.g., lactate from Warburg metabolism⁷⁰), and physical constraints absent in monolayer cultures. This emphasizes the value of our combined approach, using in vivo single-cell data to define the physiological program and in vitro models to dissect specific mechanisms.

The therapeutic implications of targeting bacterial PDH extend beyond GAS. Many pathogens (e.g., Staphylococcus aureus^71,72, Pseudomonas aeruginosa^73,74) utilize similar fermentative pathways, suggesting potential broad applicability of PDH inhibition as an anti-virulence strategy. This approach could be promising given: (1) reduced selective pressure for resistance compared to bactericidal agents⁷⁵, (2) preservation of beneficial microbiota⁷⁶, and (3) potential synergy with conventional antibiotics by improving drug penetration through preserved tissue architecture. As we demonstrated with a bacterial-specific PDH inhibitor²¹, this strategy has the potential to mitigate tissue damage without affecting bacterial burden, suggesting its potential therapeutic value.

In conclusion, our findings demonstrate that PDH-dependent bacterial metabolism coordiates a multi-layered reprogramming of the host response. As summarized in Fig. 7, ΔPdh infection leads to: (1) contained lesion morphology; (2) expansion of pro-resolving myeloid cells; (3) a transcriptional shift in these cells toward stress-adaptive states; (4) a host metabolic shift away from acetyl-CoA metabolism; (5) a rewired cytokine/chemokine landscape favoring IL-10 and controlled inflammation; and (6) the coordinated activation of a multi-faceted disease tolerance program. This model proposes how bacterial carbon flux acts as a central regulator of host immunometabolism and epigenetic regulation, suggesting PDH as a key target for anti-virulence strategies.

Fig. 7.

A model for PDH-dependent coordination of host immunometabolism and DT. Schematic summary of ΔPdh versus WT S. pyogenes infection outcomes in murine skin tissue: (1) Lesion morphology and tissue damage: WT infection produces large, necrotic ulcers, whereas ΔPdh infection generates smaller, contained lesions with preserved tissue architecture. (2) Immune composition: ΔPdh infection expand a macrophage- and neutrophil-dominated innate program. (3) Neutrophil and macrophage polarization: ΔPdh infection reprograms myeloid cells toward stress-adaptive, pro-repair transcriptional states, with upregulation of ferritin genes (Ftl1, Fth1), Srgn, Fau, and Gadd45b, and downregulation of inflammatory and matrix-degrading effectors (Retnlg, Lcn2, Plac8, Mmp8, S100a6). (4) Host metabolism: Loss of bacterial PDH reduces acetate/formate production, suppressing acetyl-CoA synthesis and mitochondrial metabolism in immune cells, and limiting histone acetylation (H3K9ac1). (5) Cytokine and chemokine regulation: ΔPdh infection induces IL-10 while maintaining controlled neutrophil and monocyte recruitment, promoting coordinated, non-cytotoxic immune responses. (6) DT pathways: ΔPdh-infected tissue exhibits upregulation of hypoxia, iron homeostasis, NRF2 oxidative stress, AMPK–mTOR signaling, and regulated cell death via apoptosis/autophagy, while lytic pathways appear to be suppressed. Figure created with Biorender.com.

Methods

This study involves both new experiments and re-analysis of data previously generated in our laboratory²¹. New data presented here include the histological examination and TEM. Transcriptomic data are re-analyses of datasets originally published²¹, as indicated in the respective figure legends. Below, we describe the methodological details for all experiments. For methods that were previously published, we reference the original publication, with a focus on the new analytical approaches applied in this study.

Ethics

All animal experiments were approved by the Institutional Animal Care and Use Committee (protocol #16–1119 and #22–0307). Procedures were performed according to all institutional policies, Animal Welfare Act, NIH guidelines and American Veterinary Medical Association guidelines on euthanasia. Adult (6–8-week-old) female C57BL/6 J and derivative mice, were maintained at Washington University School of Medicine animal facilities. This study is reported in accordance with ARRIVE guidelines (https://arriveguidelines.org).

Bacterial strains

Where indicated, S. pyogenes HSC5⁷⁷ and isogenic mutant ΔPdh²² were cultured in Todd Hewitt + 1% Yeast Extract (THY) broth⁷⁸ or C medium supplemented with 0.2% glucose⁷⁸. Quantitation of S. pyogenes CFUs utilized THY medium solidified by the addition of 1.4% agar⁷⁸, which was incubated at 37 °C anaerobically using a commercial atmospheric container (GasPak™ EZ, catalog #BD 260,001). All cultures were seeded from overnight cultures in C medium to an initial OD₆₀₀ = 0.05, which were then incubated for the various times indicated in the text.

Mouse subcutaneous ulcer model of infection

Infection of 6–8-week-old C57BL/6 mice followed a well-established protocol²². As sex does not influence infection outcomes⁷⁹, only female mice were utilized in order to minimize the number of mice used⁷⁹. Briefly, mice received a subcutaneous injection of ~ 10⁷ CFU of the indicated bacterial strains into the thigh⁸⁰. Mice were euthanized by CO₂ with a flow rate of 30–70% of the chamber’s volume per minute, followed by cervical dislocation.

Macrophage infection

RAW 264.7 macrophages were grown in Dulbecco’s Modified Eagles Medium (DMEM, Sigma) with 10% fetal bovine serum supplemented with antibiotics (1% Penicillin–Streptomycin) at 37 °C in an atmosphere of 5% CO₂. For infection, the bacterial inoculum was prepared as described above for murine infection, and after replacing media with antibiotic free media, bacteria were added to macrophages cultured in 12-well dishes at an MOI of 10:1 which were then immediately subjected to centrifugation at 500 × g for 1 min. and placed in an incubator for 1 h. Wells were then washed twice with Dulbecco’s PBS, then DMEM containing antibiotics (100 μg/ml Gentamicin) was added for 5 min and washed 3 × with PBS before the cells were incubated for an additional 2–4 h in antibiotic-free DMEM. For supplementation, acetylation inhibitors (TSA: Sigma-Aldrich, T1952) were added at the beginning of infection. Cells were harvested using a cell scraper and collected by centrifugation at 1,000 × g for 5 min, were resuspended in 1 ml dH₂O at pH 11 and lysed using a cup sonicator for 20 s. Bacteria in the resulting suspension were collected by centrifugation (6,000 × g, 5 min) and resuspended in 100 μl of C-media for determination of CFUs. Net fold-change in CFU was determined by comparison to cells harvested after the addition of antibiotic-containing medium.

Transmission electron microscopy

For ultrastructural analyses, infected tissue samples were fixed in 2% paraformaldehyde/2.5% glutaraldehyde (Ted Pella Inc, Redding, CA) in 100 mM sodium cacodylate buffer for 2 h at room temperature and then overnight at 4 °C. Samples were washed in sodium cacodylate buffer and postfixed in 1% osmium tetroxide (Ted Pella Inc.) for 1 h at room temperature. After three washes in dH₂O, samples were stained en bloc in 1% aqueous uranyl acetate (Electron Microscopy Sciences, Hatfield, PA) for 1 h. Samples were then rinsed in dH₂0, dehydrated in a graded series of ethanol, and embedded in Eponate 12 resin (Ted Pella Inc). Ultrathin sections were cut to a thickness of 95 nm with a Leica Ultracut UCT ultramicrotome (Leica Microsystems Inc., Bannockburn, IL), stained with uranyl acetate and lead citrate, and viewed on a JEOL 1200 EX transmission electron microscope (JEOL USA Inc., Peabody, MA) equipped with an AMT 8-megapixel digital camera and AMT Image Capture Engine V602 software (Advanced Microscopy Techniques, Woburn, MA). Cell types were identified based on established ultrastructural morphology: neutrophils by multi-lobed nuclei and small, numerous granules; macrophages by their large, irregular shape, abundant cytoplasm, and phagocytic vacuoles. For quantification, bacteria in acquired images were scored as intracellular if surrounded by a host cell membrane, or extracellular if not.

Microscopy

Infected skin ulcers were fixed in 10% buffered formalin (Thermo Fisher Scientific) for 24 h at room temperature, followed by overnight storage in 70% ethanol at 4 °C. Samples were paraffin-embedded and sectioned (5 µm) by the Anatomic and Molecular Pathology Core Labs at Washington University. Hematoxylin and eosin (H&E) staining was performed by the core to evaluate tissue architecture and inflammatory infiltrates. Stained sections were imaged using a a Zeiss Axio Imager 2 microscope (Jena, Germany) equipped with a digital camera.

Single cell RNA-seq

This section describes the re-analysis of existing scRNA-seq data generated as previously described²¹. In this study, we re-analyzed the preprocessed data (available as Seurat-compatible.rds files) using advanced computational methods to gain deeper insights into the immune response. The data were reanalyzed in R (v4.3.1) using Seurat (v4.3.0.1). Low-quality cells and potential doublets were filtered out by excluding cells with < 200 or > 4,000 detected genes or > 10% mitochondrial reads. Genes expressed in fewer than 3 cells were also removed. The data were then log-normalized using a scale factor of 10,000, scaled, and dimensionality was reduced via PCA (20 principal components). The WT and ΔPdh-infected conditions were integrated using Seurat’s anchor-based workflow to mitigate batch effects and enable direct comparative analysis. Cells were clustered using the FindNeighbors() function (based on the first 20 PCs) and the FindClusters() algorithm at a resolution of 0.5.

Cell clusters were annotated based on the expression of canonical immune cell marker genes. Major populations were identified as follows: macrophages (Adgre1, Cd68, Csf1r), neutrophils (S100a8, S100a9, Cxcr2), dendritic cells (Itgax, H2-Ab1, Cd74), natural killer cells (Nkg7, Klrb1c), T cells (Cd3d, Cd3e, Cd8a, Cd4), and B cells (Cd79a, Ms4a1). Subpopulations including M1 (Nos2, Il1b) and M2 (Arg1, Retnla) macrophages, neutrophil subsets (Neu1–3), dendritic cell lineages (cDC1, cDC2, pDC), and T cell subsets (CD4⁺, CD8⁺, Tregs) were defined using curated panels of established markers. All annotations were rigorously validated by inspecting differential expression statistics, visualizing gene expression patterns on dot plots and UMAP feature plots, and cross-referencing with cell-type signatures from the ImmGen and CellMarker v2.0 databases.

RNA sequencing

This section describes the re-analysis of existing bulk RNA-seq data generated as previously described²¹. In this study, we re-analyzed the gene count data to investigate the global transcriptional impact of HDAC inhibition. Differential expression analysis was performed using DESeq2 (v3.15) with Wald tests, identifying genes with adjusted p-values < 0.05 and absolute log2 fold-changes > 1. Pathway enrichment analysis was performed using Reactome, with P-values corrected for multiple testing via the Benjamini–Hochberg procedure²².

Differential gene expression

Differential gene expression analyses between WT and ΔPdh-infected groups within each cell type were performed on the RNA assay using the FindMarkers() function in Seurat with the Wilcoxon rank-sum test. Genes with an adjusted p-value < 0.05 (Benjamini–Hochberg correction) and an absolute log₂-fold change (|log₂FC|) ≥ 0.25 were considered significantly differentially expressed. Volcano plots were generated with ggplot2 (v3.5.1) to visualize the results. Key differentially expressed transcripts were projected onto UMAP embeddings using FeaturePlot() to validate their cell-type-specific expression patterns. To specifically assess transcriptional divergence in inflammatory programs, a principal component analysis (PCA) was performed using a curated set of cytokine and chemokine genes. Finally, heatmaps of the top 15 DEGs (by log₂FC) per cluster were generated for neutrophils and macrophages to illustrate their distinct activation states.

Immunometabolic module scoring

Pathway activity was quantified at the single-cell level using the AddModuleScore() function in Seurat. Curated gene sets for key immunometabolic pathways (e.g., glycolysis, oxidative phosphorylation, hypoxia, iron homeostasis) were assembled from established literature sources. The specific composition of each gene set is detailed in the corresponding figure legends. The resulting module scores were visualized using violin plots (for distribution across conditions), UMAP feature plots (for spatial embedding), and grouped bar charts (for comparative analysis of mean scores). Quantitative statistical analysis of module scores (glycolysis, acetyl-CoA metabolism, Hdac1, H3K9ac1, DT module) was performed for each major cell type. The mean module score was calculated separately for WT and ΔPdh conditions within each annotated cell cluster. Differences between conditions were assessed using the Wilcoxon rank-sum test. Effect sizes were calculated as Cohen’s d. These per-cell-type statistics are compiled in Supplementary Table 4.

Pathway and enrichment analysis

Significant DEGs were submitted for pathway enrichment analysis using clusterProfiler (v4.6.2), ReactomePA (v1.42.0), and Enrichr. Gene Ontology (GO) biological process terms were obtained using enrichGO() and Reactome pathways via enrichPathway(), using Mus musculus (mmusculus) as the background organism. KEGG and WikiPathways were queried via the Enrichr API. Over-representation was assessed using hypergeometric tests or Fisher’s exact tests, with Benjamini–Hochberg adjustment (adjusted p < 0.05). Enrichment results were visualized as dot plots or bar plots, with adjusted p-values and gene ratios indicated.

Cell–cell communication analysis

Intercellular signaling networks were inferred using CellChat (v1.6.1) and its curated mouse ligand–receptor database. Communication probabilities were computed with computeCommunProb(), aggregated into signaling pathways with aggregateNet(), and compared between conditions to identify differentially altered signaling. Results were visualized using circle plots and heatmaps, where edge width reflected interaction strength and node size represented the number of interactions per cell group. Condition-specific edges were colored to highlight signaling pathways enhanced in WT versus ΔPdh infections.

Pseudotime trajectory analysis

Cellular differentiation trajectories were reconstructed using Slingshot (v2.4.0) on the UMAP embeddings and pre-annotated cell clusters. Starting states (roots) were designated based on the expression of precursor-associated markers: for neutrophils, Neu1 was set as the root state and Neu3 as the terminal state; for macrophages, MAC1 was set as the starting state and MAC2 as the terminal state. Gene expression dynamics for key targets (Acly, Acss2) were modeled as smoothed functions along the inferred pseudotime. Differential expression patterns between WT and ΔPdh-infected conditions were visualized and compared using gene-specific line plots.

Supervised multivariate analysis (PLS-DA)

To identify transcriptional features that best discriminate WT and ΔPdh infection conditions, PLS-DA was performed on curated cytokine and chemokine gene expression matrices. Log-normalized expression values were used as input. PLS-DA was conducted using the mixOmics package in R, with model components selected based on variance explained and group separation. Variable loadings were extracted to identify genes contributing most strongly to condition separation, and Variable Importance in Projection (VIP) scores were calculated to rank genes by their contribution to the model. Model performance was assessed by cross-validation, and permutation testing indicated no evidence of overfitting.

Statistical analysis

All analyses were performed in R (v4.3.1). For two-group comparisons, the Wilcoxon rank-sum test was used for non-normally distributed data, and the Student’s t-test was used for normally distributed variables, as assessed by the Shapiro–Wilk test. For comparisons across more than two groups, the Kruskal–Wallis test was used, followed by Dunn’s post-hoc test for pairwise comparisons with p-value adjustment. P-values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate correction. An adjusted p-value < 0.05 was considered statistically significant. Data are presented in figures as the mean ± standard error of the mean (SEM) unless otherwise noted in the figure legend. All data visualization was performed using ggplot2 (v3.5.1), ComplexHeatmap (v2.14.0), and standard plotting functions in Seurat.

Author contributions

WX, SKP, MGC conceptualized the project. WX, SKP designed the research studies. WX, SKP conducted experiments and acquired data. SKP, SG analyzed data. SKP, WX prepared figures. WX, SKP wrote the original draft. MGC reviewed and edited the manuscript.

Funding

This work was supported by a startup fund from the Joan C. Edwards School of Medicine at Marshall University (to WX), National Institutes of Health Grants R56 AI070759, R21 AI163825 (to MGC).

Data availability

The single-cell RNA sequencing and bulk RNA sequencing data re-analyzed in this study were originally generated in our previous work²¹ and are available under NCBI BioProject ID PRJNA917702 (scRNA-seq) and PRJNA1145018 (bulk RNA-seq). All other data generated in this study (including histology and transmission electron microscopy images) are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.