In early sepsis, M1 macrophages primarily rely on glycolysis, exhibiting enhanced glycolytic flux and diminished oxygen consumption (31,35), which disrupts the TCA cycle. This reprogramming enables rapid generation of energy and metabolic intermediates necessary for macrophage activity and function during infection (36). Upregulation of glycolytic enzymes, including pyruvate kinase M2 (PKM2), glucose transporter 1 (GLUT1), hexokinase (HK), phosphofructokinase-1 and 6-phosphofructose-2-kinase/fructose-2,6-bisphosphatase 3 (PFKFB3), along with increased lactate production, further amplifies glycolytic activity and promotes M1 polarization (37,38).

Mechanistically, the metabolic-transcriptional interface in septic macrophages is orchestrated by the functional transformation of PKM2 and its synergy with HIF-1a (39,40). Under inflammatory cues, PKM2 undergoes post-translational modifications, specifically phosphorylation and acetylation, triggering its transition to a dimeric form that translocates to the nucleus (39,41). Elevated serum PKM2 levels have been shown to be strongly associated with disease severity and organ damage (42). Once in the nucleus, PKM2 acts as a co-activator for HIF-1a to directly promote the transcription of pro-inflammatory genes and essential glycolytic enzymes (16,43). Notably, this axis is further stabilized by the accumulation of succinate resulting from TCA cycle disruption at the succinate dehydrogenase (SDH) site (15,44). Intracellular succinate prevents HIF-1α degradation by inhibiting prolyl hydroxylases, thereby driving IL-1β production (45,46). Furthermore, extracellular succinate functions as a signaling molecule via G protein-coupled receptor 91 (GPR91) to sustain the pro-inflammatory M1 phenotype (47). Together, the PKM2-HIF-1α-succinate axis represents a metabolic checkpoint that bridges mitochondrial dysfunction with persistent glycolytic reprogramming.

HIF-1α upregulates glycolytic enzyme genes [such as HK2, PFKFB3 and lactate dehydrogenase (LDH)] and promotes GLUT1 synthesis. Moreover, it inhibits pyruvate entry into mitochondria and enhances its conversion to lactate by increasing LDH expression, thereby fueling glycolysis (48-50). Glycolysis, in turn, enhances HIF-1α translation and stability while promoting GLUT1 expression, creating a self-reinforcing loop that sustains glycolytic reprogramming (16,43). In the cytoplasm, dimeric PKM2 interacts with molecules such as high mobility group box-1 (HMGB1), enhancing transcription of glycolytic enzymes (GLUT1, LDH and HK) and skewing immune cells toward glycolysis, thereby amplifying the inflammatory response of M1 macrophages (40,41,51). These changes promote the release of late-phase pro-inflammatory mediators such as HMGB1 from macrophages (52). Extracellular lactate is taken up primarily via monocarboxylate transporters, facilitating HMGB1 lactylation through a p300 and CREB-binding protein-dependent mechanism.

Lactate can also be recruited to the nucleus via GPR81 to stimulate HMGB1 acetylation. Lactylated/acetylated HMGB1 is released via exosomes, increasing endothelial permeability and accelerating the progression of polymicrobial sepsis (53). HMGB1, a nuclear protein released extracellularly, exacerbates immune responses through TLR stimulation, direct cytotoxicity and platelet activation, contributing to disseminated intravascular coagulation and functioning as a damage-associated molecular pattern (54). Early clinical studies demonstrated elevated HMGB1 levels in sepsis, positively associating with disease severity and mortality (55,56). A previous study of 218 critically ill patients (145 with sepsis, 73 without) also reported a positive correlation between blood HMGB1 and lactate levels (r=0.144; P=0.035), supporting the interplay between HMGB1 and lactate during sepsis (57). The clinical correlation observed between PKM2, lactate and HMGB1 levels forms a metabolic inflammatory feedback loop in the pathogenesis of sepsis. The elevation of these markers is not only a result of tissue damage, but also an indicator of the enhanced glycolytic state before organ dysfunction. This indicates that the therapeutic window for glycolytic inhibitors or PKM2 modulators must be strictly aligned with the initial surge of these circulating biomarkers, as their peak likely represents a point of no return for HMGB1-driven vascular failure.

Beyond signaling pathways, macrophages in sepsis-induced acute lung injury (ALI) often overexpress the Sprouty RTK signaling antagonist 4 (Spry4) gene. Notably, Spry4 deficiency has been shown to alleviate lung injury by activating the calcium/calmodulin dependent protein kinase kinase 2 pathway, contrasting with reports that Spry4 can aggravate sepsis progression (58). Another key metabolic regulator in sepsis is the PI3K/Akt/mTOR pathway, which also contributes to ALI pathogenesis (59). PI3K activation leads to downstream Akt activation, enhancing GLUT1 expression and glucose uptake. Activated Akt further increases HIF-1α expression via mTOR, explaining the preference for glycolysis even under aerobic conditions (60). In addition, the PI3K/Akt/mTOR axis overactivation not only prioritizes glycolytic flux, but also governs the intracellular availability of essential amino acids such as glutamine (Gln) and arginine, effectively linking glucose consumption to the protein synthesis machinery required for cytokine production. Taken together, these findings underscore the glycolysis-driven metabolic reprogramming in macrophages mediated by AMPK inhibition, PI3K/Akt/mTOR activation and PKM2 functional transformation, which promotes M1 polarization and inflammatory mediator release, ultimately driving immunometabolic imbalance and multi-organ damage in sepsis.

Under physiological conditions, the TCA cycle maintains a balance between isocitrate dehydrogenase (IDH)-catalyzed conversion of isocitrate to α-ketoglutarate (α-KG) and SDH-catalyzed conversion of succinate to fumarate (14,31). In sepsis, the TCA cycle undergoes notable remodeling, characterized by two metabolic changes at the IDH and SDH sites (15,61).

In M1 macrophages, the activity of key TCA enzymes such as IDH and SDH is reduced, disrupting the mitochondrial oxidative respiration chain. This leads to partial TCA cycle blockade, enhanced glycolysis and accumulation of intermediates such as succinate and citrate (61). Cytosolic ATP-citrate lyase cleaves accumulated citrate to provide acetyl-CoA for fatty acid synthesis, supporting production of inflammatory mediators such as prostaglandins and nitric oxide (NO) (62,63). Furthermore, excess citrate-derived acetyl-CoA affects epigenetic regulation by serving as a core donor for histone acetylation, which upregulates genes involved in inflammatory responses, such as IL-6 and IL-1β (64,65). Epigenetically, mitochondrial sirtuin 3 deficiency leads to hyperacetylation of TCA enzymes, increasing lactate and nicotinamide adenine dinucleotide production and contributing to sepsis-induced myocardial dysfunction (66).

Enhanced glycolysis and TCA disruption jointly increase lactate production. Lactate can serve as a precursor for histone lactylation [such as histone H3 lysine 18 lactylation (H3K18la], upregulating M2-related genes including arginase 1 (Arg1). This mechanism may serve a role in late-stage sepsis immunosuppression (67). M2 macrophages maintain an intact TCA cycle. Clinical data show elevated H3K18la levels in peripheral blood mononuclear cells from critically ill patients with sepsis, positively associating with Arg1 mRNA expression and disease severity (68). This suggests that early inflammatory responses driven by TCA interruption and its metabolites may lay the groundwork for later epigenetic reprogramming and phenotypic shifts. Other studies demonstrated that alveolar macrophages (AMs) in early ALI exhibit weak glycolytic capacity, predominantly displaying an M2 phenotype that relies on OXPHOS for cytokine production during lipopolysaccharide (LPS) activation (69). However, under extreme hypoxia, such as when ALI progresses to acute respiratory distress syndrome, HIF-1α activation promotes a shift in AMs to an M1 phenotype, transitioning from OXPHOS to glycolysis (70,71).

The PPP branches from glycolysis (at glucose-6-phosphate and fructose-6-phosphate) and serves two key roles in septic macrophages (13,72): Synthesis of ribose-5-phosphate and generation of nicotinamide adenine dinucleotide phosphate hydrogen (NADPH).

Ribose-5-phosphate provides nucleotide precursors for macrophage proliferation (72). NADPH produced by glucose-6-phosphate dehydrogenase (G6PD; the rate-limiting enzyme of PPP) supports macrophage survival during infection. NADPH also maintains redox balance and fuels reactive oxygen species (ROS) production via NADPH oxidase to eliminate pathogens (13). Notably, M1 macrophages exploit both glycolysis and the PPP to meet their ATP demands (14,44). During sepsis, downregulation of carbohydrate kinase-like protein (CARKL/SHK), a factor co-localized with G6PD, increases PPP activity. CARKL/SHK is typically highly expressed in M2 macrophages (15,73). Inhibiting CARKL drives macrophages toward M1 polarization, enhancing ROS-mediated bacterial killing but also increasing the risk of tissue damage (13).

During sepsis, macrophage lipid uptake capacity is enhanced, increasing fatty acid metabolic activity (76). This uptake depends largely on the expression of fatty acid transporters such as cluster of differentiation 36 and carnitine palmitoyltransferase 1a (CPT1a). Upregulation of these transporters facilitates efficient fatty acid acquisition (63,77). Additionally, mitochondrial STAT3 further promotes FAO by stabilizing CPT1a via ubiquitin specific peptidase 50-mediated deubiquitination (78), enhancing fatty acid entry into mitochondria. Lipid metabolites can also act as immune signaling molecules; for example, saturated fatty acids such as palmitic acid promote pro-inflammatory responses in macrophages via TLR signaling, increasing tumor necrosis factor-α (TNF-α) and IL-6 production (79,80). These pro-inflammatory cytokines, in turn, further enhance fatty acid uptake and activate metabolic regulators such as AMPK and mTOR (81,82). Although AMPK activation can promote autophagy and FAO to counteract inflammation (77,83), its activity is typically suppressed in sepsis.

In late-stage sepsis, macrophages rely mainly on mitochondrial OXPHOS and FAO for tissue repair (15,84). However, reliance on FAO often leads to imbalances. First, anti-inflammatory cytokines such as IL-4 and IL-10 may contribute to immune suppression (10). Second, rapid glycogen depletion causes transient hyperglycemia and increased triglyceride lipolysis, elevating free fatty acid (FFA) and glycerol levels and manifesting as impaired FAO in sepsis. Dysfunction of PPARα and the glucocorticoid receptor exacerbates metabolic imbalance, leading to ketone body and glucose deficiency and promoting inflammation (85-87).

Excessive lipid accumulation disrupts intracellular homeostasis and aggravates inflammation. However, not all lipid metabolites are detrimental. For instance, ketone bodies directly protect cells and negatively regulate inflammation by activating the β-hydroxybutyrate receptor GPR109A and inhibiting the NOD-like receptor domain-containing protein 3 inflammasome (88,89). Prostaglandin E2 (PGE2) inhibits TNF-α and IL-6 production, while lipoxin A4, derived from arachidonic acid, inhibits PGE2 signaling and promotes M2 macrophage polarization (90). Moreover, other fatty acid metabolites also suppress inflammatory responses. Unsaturated fatty acids promote tissue repair by activating anti-inflammatory pathways such as PPARγ (91). Omega-3 fatty acids attenuate sepsis-induced inflammation and oxidative stress by increasing notch receptor 3 expression via downregulation of micro-RNA (miR)-1-3p and blocking the Smad pathway, thereby mitigating intestinal epithelial injury (92). Nevertheless, persistent lipid metabolic dysregulation can sustain M1 polarization, perpetuating chronic inflammation and tissue damage (3,63,93,94).

Clinically, lipid metabolism is linked to sepsis prognosis. Elevated serum FFA levels are positively associated with disease severity and can exacerbate sepsis by activating specific inflammatory pathways, further contributing to MODS (95). Alterations in essential fatty acid metabolism may disrupt the balance between pro- and anti-inflammatory mediators (such as eicosanoids and cytokines), leading to immune dysregulation in sepsis (96). Additionally, elevated heart-type fatty acid-binding protein serves as a biomarker for early diagnosis and prognosis of sepsis-induced cardiomyopathy (97). These findings suggest that targeting lipid metabolic pathways may improve macrophage function and attenuate sepsis-induced inflammation. Therapeutically, mTOR inhibitors such as rapamycin promote FAO and autophagy to limit tissue damage and prevent excessive immune responses (98). Preclinical studies show that PPARα agonists improve survival in septic mice by restoring FAO and reducing lipotoxicity (87,99). Although this strategy awaits validation in human trials, it highlights the potential of PPAR activation in alleviating sepsis-related metabolic disorders (100).

Recent single-cell landscape studies have identified a specific subpopulation called LAMs, which differs from the classical M1/M2 classification system (22-24). The characteristic of LAM is the TREM2 dependent transcriptional program, which serves as a metabolic sensor to coordinate lipid uptake, lysosomal function and energy homeostasis. In sepsis, TREM2 serves a double-edged sword role in this subgroup. On the one hand, TREM2 signaling in LAMs helps prevent systemic lipotoxicity, promote the clearance of apoptotic cells and improve sepsis outcomes in liver and heart injury models (23,24). On the other hand, overactivation of TREM2⁺ LAMs can lead to systemic hypercholesterolemia and increase susceptibility to sepsis by over-activating the SHP1/BTK axis, which in turn impairs FAO (22). Studies have found that knocking out TREM2 in macrophages reduces inflammation, organ damage, triglyceride accumulation and enhances FAO, improving survival in septic mouse models (25). This indicates that LAMs represent the metabolic adaptation of macrophages to the lipid-rich sepsis microenvironment, and its functional direction depends on the severity and stage of infection.

Cholesterol, a sterol lipid, is a precursor for steroid hormones, bile acids and oxysterols (101,102), and also regulates various cellular functions while forming structural components of cell membranes along with cholesteryl esters (103). Both cholesterol and its lipoprotein carriers possess immunomodulatory properties, binding and neutralizing endotoxins to prevent PAMPs activation of TLRs (104).

During sepsis, serum levels of total cholesterol, high-density lipoprotein cholesterol (HDL-C) and low-density lipoprotein cholesterol (LDL-C) are markedly reduced (105,106), associating with increased mortality risk. Early clinical data indicate that LDL-C levels are lowest at diagnosis, while HDL-C levels typically reach their nadir around the third day in hospital (107). However, previous studies suggest that while low levels of HDL-C may be a key contributor to mortality risk, reduced LDL-C concentrations may not be causative (106,108). A cohort study investigating the non-HDL-C/HDL-C ratio in patients with sepsis revealed a U-shaped relationship with 28-day mortality, both excessively high and low ratios were associated with increased mortality (109), highlighting the importance of monitoring this ratio. The U-shaped relationship between the comprehensive cholesterol ratio and mortality rate indicates that lipid homeostasis can serve as a buffering system. In addition to serving as energy precursors, lipoproteins also act as innate scavengers of pathogenic endotoxins. The precipitous drop in HDL-C observed in non-survivors reflects a critical failure in this protective capacity, positioning cholesterol profiles as functional indicators of the host's residual innate defense reserves rather than a simple metabolite. Animal studies demonstrate hypocholesterolemia during sepsis, with serum cholesterol levels inversely associated with inflammatory markers (110-112). Although well-documented, the mechanisms underlying hypocholesterolemia in sepsis remain unclear.

Cholesterol levels intricately influence macrophage signaling, particularly through lipid raft domains that modulate TLR4 and TLR9 signaling (113). Impaired cholesterol transport (such as due to ATP-binding cassette protein A1/ATP-binding cassette protein G1 defects) hinders efflux, leading to cellular cholesterol accumulation and lipid raft enlargement. This heightens macrophage sensitivity to TLR4 signaling and LPS, exacerbating inflammatory responses (113-117). Cholesterol metabolites also activate nuclear receptors such as hepatocyte nuclear factor 4a and liver X receptors. In infected macrophages, altered lipid metabolism can activate these receptors, modulating inflammatory mediator expression (118).

Aromatic amino acids tyrosine, tryptophan and phenylalanine, are among the most markedly altered metabolites in sepsis. A previous study noted marked increases in intermediates such as phenylpyruvate and L-phenylalanine, highlighting prominent dysregulation of aromatic amino acid metabolism in sepsis (120).

Tryptophan metabolism is particularly relevant to immune escape mechanisms in sepsis (121). Inflammatory cytokines such as interferon-γ induce indoleamine 2,3-dioxygenase gene transcription in macrophages, enhancing tryptophan degradation via the kynurenine pathway. Kynurenine and its derivatives suppress T-cell proliferation and modulate macrophage activity, promoting immunosuppression (121-123). Notably, kynurenic acid, a kynurenine pathway metabolite, facilitates the transition from M1 to M2 macrophages by inhibiting NF-κB signaling and alleviating septic colon injury (124).

The phenylalanine/tyrosine ratio also reflects immune activation status, and studies have identified phenylalanine and histidine metabolism as among the most markedly altered in sepsis (120,125). Excessive phenylalanine can inhibit protein synthesis and exert toxic effects on antibodies (120). A study of 63 patients with sepsis also found strong associations between phenylalanine metabolism and sepsis-associated acute kidney injury.

In macrophages, Gln is converted to glutamate by glutaminase and further metabolized to a-KG, which enters the TCA cycle for energy production (75). Gln metabolism-driven glutathione (GSH) synthesis provides energy and intermediates while helping maintain redox balance (13,14). Studies show a close association between Gln and M2 polarization; α-KG restores pyruvate dehydrogenase activity and supports M2 differentiation (13,14). Additionally, inflammatory mediator-stimulated metabolic reprogramming in macrophages depends on Gln. Macrophages can activate mTOR and other pathways to enhance Gln transporter expression, forming a positive feedback loop that amplifies immune responses (126,127). In sepsis-induced muscle atrophy, Gln therapy activates the mTOR pathway to alleviate muscle degradation (128). These cellular mechanisms indicate the potential of Gln in restoring metabolic balance and reducing organ damage in sepsis.

Substantial evidence suggests that Gln regulates cellular metabolism and function through post-translational modifications such as acetylation and succinylation, particularly in burn-induced sepsis (129,130). Gln also reduces oxidative stress by rescuing dysfunctional mitochondrial ETC complexes, protecting hepatocytes from inflammation-induced injury, a protective mechanism in burn sepsis (131). Moreover, Gln supplementation reduces sepsis-induced cardiomyocyte apoptosis in rat models (132,133). However, translating these beneficial mechanisms to clinical practice has yielded complex results. Some previous studies question the benefits of Gln in critically ill intensive care unit (ICU) patients and even associate its supplementation with development of chronic critical illness (134,135).

In the ICU, the benefit of Gln supplementation is highly dependent on timing, dose and host baseline status. Regarding timing, a study of 1,223 critically ill patients found that patients receiving Gln had a trend toward higher 28-day mortality compared with non-recipients (32.4 vs. 27.2%; adjusted odds ratio 1.28; 95% CI 1.00-1.64; P=0.05). In-hospital and 6-month mortality were also significantly higher in the Gln group (both P=0.02) (136). Dose-dependence was also significant, as evidenced by the significantly higher frequency of high urea levels in the glutamine group (13.4 vs. 4.0%; P<0.001). A large multicenter randomized trial indicated that high-dose parenteral Gln (>0.5 g/kg/day) should be avoided early in critical illness. This caution is mainly due to the metabolic substrate overload that occurs during the hyperacute phase of sepsis. In patients with multiple organ dysfunction, especially kidney and liver damage, the body's ability to handle nitrogen load is markedly reduced. High doses of exogenous Gln can lead to excessive accumulation of metabolic byproducts such as ammonia and urea, which may exacerbate uremic toxicity or hepatic encephalopathy.

In addition, in the early stages of critical illness, the host may trigger large-scale catabolic reactions to mobilize endogenous amino acids. At this point, adding high doses of exogenous Gln may interfere with beneficial autophagy processes or inhibit the host's adaptive metabolic stress response, leading to increased mortality. The different results of large-scale Gln trials combine a fundamental clinical paradox, namely that the therapeutic benefits of amino acid supplementation depend on the metabolic stage of the patient rather than absolute dosage. The increase in mortality rate during the hyperacute phase indicates that exogenous nitrogen load may cause severe liver damage, while the same intervention during the recovery phase helps with redox balance and tissue repair. Therefore, a more cautious dosage of 0.3-0.5 g/kg/day should only be considered after the acute phase has subsided and organ function has stabilized (137). Regarding host status, patients with high metabolic demands (such as burns or trauma) may benefit from supplementation due to increased Gln consumption (elevating GSH and reducing oxidative stress). It may also benefit experimental neonatal endotoxemia and very low birth weight preterm infants. However, intracellular heat shock protein 70 deficiency due to Gln deprivation may increase post-sepsis mortality (138).

Ultimately, the clinical benefit of Gln is not absolute. Future research should move beyond the simplistic question of whether Gln is beneficial or harmful. Efforts should clarify these regulatory factors to identify sepsis subgroups most likely to benefit from Gln supplementation and develop individualized treatment strategies targeting related metabolic pathways, thereby improving sepsis outcomes.

In macrophages, arginine is metabolized via two primary pathways: The nitric oxide synthase (NOS) pathway producing NO, and the arginase pathway producing ornithine. Regarding phenotype, NO can prevent M1 macrophages from repolarizing to M2, while inducible NOS (iNOS) inhibition facilitates M1-to-M2 transition (139). By contrast, M2 macrophages metabolize arginine via Arg1, producing ornithine and urea.

In sepsis, macrophage iNOS expression increases in response to PAMPs (140). iNOS converts arginine to citrulline and NO. NO can react spontaneously with ROS to form reactive nitrogen and oxygen intermediates, serving as a key effector molecule in macrophage-mediated antibacterial and anti-inflammatory responses (31,140,141). However, excessive NO may also induce tissue damage (142). Additionally, arginine availability is often compromised in sepsis due to impaired de novo synthesis and accelerated catabolism, affecting arginine-dependent pathways such as mTOR signaling and thereby impacting macrophage function and survival (143,144). Arginine deficiency and reduced NO production are common in endotoxemia models and have been observed in ICU patients (145). Nonetheless, no consistent causal relationship has been established between low arginine levels and sepsis etiology (143), suggesting that arginine deficiency is not universal in all sepsis cases. Supplementing with arginine has been shown to reduce systemic inflammation and help maintain vascular homeostasis in sepsis (146,147). However, there is a complex paradox regarding the availability of arginine. Some researchers hypothesize that the reduction of arginine in host plasma during the acute phase of sepsis may be an adaptive defense mechanism to limit excessive NO production and subsequent oxidative stress substrates. On the contrary, long-term arginine deficiency can lead to immune dysfunction, such as impaired T cell proliferation and microcirculation (143,145,147). Therefore, the treatment goal should be to restore arginine to physiological levels, rather than reaching the level of an 'inflammatory cytokine storm' to support immune homeostasis (143,146,148).

Arginase activity is also altered during sepsis. Arg exists in two isoforms: Arg1 (cytosolic, liver-specific) and Arg2 (mitochondrial). Arg1 participates in the urea cycle, typically utilizing extracellular arginine to regulate its availability to neighboring cells (149). In M2 macrophages, Arg1 expression is promoted via the STAT6 pathway (150) and can be synergistically induced by IL-10 and LPS (151). In M2 macrophages, Arg1 expression is promoted via the STAT6 pathway and can be synergistically induced by IL-10 and LPS (152). Arg1 hydrolyzes arginine to ornithine and urea (153,154). Ornithine serves as a precursor for polyamine and proline synthesis; polyamines (such as putrescine and spermidine) participate in cell proliferation, while proline is crucial for collagen synthesis (155). Similar to Arg1, Arg2 expression is upregulated in human and mouse macrophages after LPS treatment (151). Arg2 downregulates succinate levels by promoting SDH (complex II) activity, further downregulating HIF-1α and IL-1 expression, promoting the M2 anti-inflammatory phenotype and enhancing mitochondrial respiration (156,157). However, in chronic inflammation models (such as atherosclerosis), Arg2 can also produce proinflammatory effects by increasing mitochondrial ROS. This suggests that it may have a similar dual role in sepsis, but further confirmation is needed (158). In summary, the balance between the NO and arginase pathways is disrupted; overactivation of iNOS consumes large amounts of arginine for NO production, affecting the arginase pathway and impairing normal macrophage functions such as phagocytosis and antigen presentation during the sepsis (159).