The canonical pathway represents the most thoroughly characterized mechanism driving pyroptosis in the septic lung. It is initiated when intracellular pattern recognition receptors, such as absent in melanoma 2, pyrin, or various nucleotide-binding oligomerization domain-like receptors (NLRP1, NLRP3 and NLRC4), detect conserved pathogen-associated molecular patterns or endogenous DAMPs released during cellular stress (5,45,46) (Fig. 4A). Unlike stochastic inflammatory triggers, the assembly of the NLRP3-apoptosis-associated speck-like protein containing a caspase recruitment domain (ASC)-caspase-1 inflammasome functions as a high-fidelity molecular amplifier that dictates the intensity of the early pulmonary cytokine storm (47,48). Upon successful assembly of this multiprotein complex, mature caspase-1 processes pro-IL-1β and pro-IL-18 into their bioactive forms, while concurrently cleaving GSDMD to liberate the N-terminal pore-forming domain (GSDMD-NT) (49). In the complex landscape of S-ALI, this pathway serves as the primary engine for early hyper-inflammation and the metabolic priming of the alveolar vascular endothelium (50,51). The significant individual variability in the activation thresholds of these cytosolic sensors explains the inconsistent clinical efficacy of broad-spectrum anti-inflammatory agents, suggesting that patients with a hyper-primed canonical axis require interventions specifically targeting inflammasome assembly rather than downstream effector cytokines (52).

For patients suffering from Gram-negative bacterial sepsis, the noncanonical pathway provides a direct, autonomous sensing mechanism that bypasses the classical receptor-mediated signaling hierarchy (53,54). In this axis, human caspase-4/5 (or murine caspase-11) functions as an intracellular receptor that binds directly to cytosolic LPS, a process significantly facilitated by interferon-inducible guanylate-binding proteins that destabilize pathogen-containing vacuoles (55,56) (Fig. 4B). This direct protein-LPS interaction represents a rapid-response executioner of endothelial lysis, occurring independently of the kinetic delays typically associated with the assembly of large inflammasome platforms (57). Furthermore, caspase-11-mediated potassium efflux can secondarily trigger the NLRP3 inflammasome, creating a potent, self-reinforcing feed-forward loop that accelerates the collapse of the alveolar-capillary barrier (58). From a precision medicine perspective, recognizing the dominance of this LPS-caspase axis is essential, as it identifies a subset of patients who remain refractory to NLRP3-specific inhibition, but may demonstrate clinical responsiveness to strategies targeting noncanonical executioners.

The molecular plasticity between silent apoptosis and explosive pyroptosis is governed by the strategic recruitment of apoptotic caspases to the gasdermin execution machinery. As illustrated in Fig. 4C, various stimuli including influenza A virus, tumor necrosis factor-α (TNF-α), death receptors, reactive oxygen species and chemotherapeutic drugs can trigger this pathway (59-62). These stimuli activate the extrinsic apoptotic pathway through death receptors, leading to the recruitment of adaptor proteins, such as FADD and RIPK1, which activate pro-caspase-8 into active caspase-8. Simultaneously, the intrinsic (mitochondrial) pathway is engaged via cytochrome c release from the mitochondria, resulting in the activation of caspase-9. Both initiator caspases (caspase-8 and caspase-9) then converge to activate the executioner caspases, primarily caspase-3, as well as caspase-6 and caspase-7. These executioner caspases cleave gasdermin family proteins, including gasdermin B (GSDMB), gasdermin C, GSDMD and gasdermin E (GSDME), to generate their pore-forming N-terminal fragments (23,60).

Caspase-3, traditionally viewed as a terminal executioner of non-inflammatory apoptosis, can cleave GSDME to generate GSDME-N fragments, which effectively switches the mode of cell death to pro-inflammatory pyroptosis. The clinical trajectory of S-ALI is profoundly influenced by the tissue-specific expression profiles of GSDME, as high baseline expression in pulmonary epithelial cells converts otherwise resolving apoptotic signals into localized tissue lysis and inflammatory amplification (63-65). Similarly, caspase-8 functions as a molecular rheostat under conditions of TNF-α stimulation or pathogenic stress, bridging these death pathways via the direct cleavage of GSDMD in specific immune cell subsets (66). This molecular crosstalk provides a profound stratification marker for clinicians in the ICU; evaluating the GSDME signature in pulmonary cells predicts whether a patient will exhibit a hyper-inflammatory phenotype or a more favorable resolving death phenotype (66).

Beyond endogenous cell-autonomous signaling, the granzyme-mediated pathway highlights the critical role of external immune executioners in driving irreversible pulmonary tissue lysis. Perforin-delivered granzymes (particularly GZMA and GZMB) from over-activated cytotoxic T-lymphocytes or natural killer cells can directly cleave gasdermin family members, such as GSDMB or GSDME, in an action that is entirely independent of conventional caspase activity (67-72) (Fig. 4D). In the hyper-active immune phase often observed in early sepsis, this pathway represents a significant mechanism of collateral damage where the adaptive immune system of the exacerbates alveolar wall destruction and tissue dysfunction. Identifying this lymphocyte-driven lysis shifts the focus of personalized therapy toward immunomodulatory stratification, particularly for those critically ill patients who exhibit abnormally high cytotoxic activity in bronchoalveolar lavage fluid, effectively moving the field beyond the traditional constraints of innate immune-centric treatment models (73).

Alveolar macrophages function as the frontline sensors of the pulmonary microenvironment. In the initial phase of S-ALI, macrophage pyroptosis functions as a high-potency 'molecular detonator' that amplifies pathogen-driven signals into an expansive cytokine storm (76). The activation of the NLRP3 inflammasome facilitates the rapid, unconventional secretion of IL-1β and high mobility group box 1, which orchestrate neutrophil recruitment and prime the lung parenchyma for secondary damage (77). This macrophage-centric execution accounts for the systemic hyper-inflammation observed in 'hyper-inflammatory' subphenotypes (78). The intensity of this detonator phase suggests that early-window targeting of macrophage-specific gasdermin activation could preemptively halt the inflammatory cascade before irreversible structural destruction occurs.

The structural integrity of the alveolar-capillary barrier is fundamentally dependent on the semi-permeable microvascular endothelium. In the evolution of S-ALI, endothelial pyroptosis serves as the primary driver for the catastrophic loss of vascular barrier function, leading to the hallmark accumulation of protein-rich edema (26). The intracellular sensing of LPS triggers non-canonical signaling through caspase-11, which executes endothelial lysis via GSDMD-mediated pore formation (79). This cell-specific destruction explains the clinical transition to profound hypoxemia, as endothelial pore formation represents a decisive biophysical transition point (80). Furthermore, the release of pro-thrombotic factors following membrane rupture exacerbates microcirculatory dysfunction, linking pyroptosis to the coagulation abnormalities frequently observed in patients with sepsis (81).

While endothelial damage drives leakage, the pyroptosis of alveolar epithelial cells (AECs), particularly type II cells, dictates the failure of respiratory mechanics. The depletion of AECII through pyroptotic pathways directly impairs pulmonary surfactant production, contributing to alveolar collapse and refractory atelectasis (82). Unlike apoptosis, this regulated lysis prevents effective lung repair and potentially initiates pathways driving late-stage fibroproliferation (83). Recognizing the epithelial compartment as a terminal executioner provides a translational rationale for strategies that specifically preserve AEC integrity. Such protection is essential for maintaining mechanical stability in patients undergoing invasive mechanical ventilation, where epithelial loss increases the risk of ventilator-induced lung injury (84).

Critical to the exacerbation of S-ALI is the molecular crosstalk between these cellular compartments, often converging on the assembly of the PANoptosome (85,86). This integrated death mode involves the coordinated activation of pyroptosis, apoptosis and necroptosis, allowing the cell to bypass individual signaling bottlenecks (87). The discovery of PANoptosis explains the 'therapeutic escape' observed in clinical trials, where inhibiting a single protease fails due to the compensatory activation of parallel death modes (83). For instance, selective caspase-1 inhibitors (such as VX-765) in inflammatory lung injury models and related acute respiratory distress syndrome (ARDS) trials have exhibited limited sustained efficacy, as the blockade of pyroptosis leads to the compensatory activation of apoptosis and necroptosis pathways via caspase-8 and RIPK3/MLKL, resulting in persistent tissue damage and cytokine release (88). Similarly, attempts at single caspase-8 inhibition in sepsis-associated contexts have demonstrated incomplete protection due to upregulation of alternative PANoptosome components (89). Integrating PANoptosome dynamics into the S-ALI framework underscores the necessity for 'combination precision medicine' that addresses these convergent molecular hubs to effectively mitigate multifaceted tissue destruction (90).

The canonical NLRP3-caspase-1 axis serves as a primary molecular detonator in S-ALI, with the intensity of its activation being closely associated with alveolar-capillary barrier disruption (93). Distinct regulatory clusters are governed by endogenous rheostats, such as heat shock factor 1 and heat shock protein 8, which modulate NLRP3 ubiquitination to prevent the irreversible formation of the ASC speck complex (94,95).

These regulatory clusters differ significantly in molecular signaling pathways and differentiation states, indicating a dramatic increase in transcriptional heterogeneity during sepsis progression. The intermediate and transitional state of NLRP3 priming is evident, as multiple genes and pathways show consistent shifts from early priming to terminal execution. A priming cluster shared across models demonstrates potent differentiation potential according to the transcriptional differentiation trajectory. Pathway enrichment points to the stem-like characteristic of this subpopulation, suggesting it as the potentially pioneering force of lineage plasticity for canonical pyroptosis. A vast repertoire of pharmacological agents has demonstrated robust efficacy in preclinical models (96-122); however, the discordance between murine success and clinical reality underscores the challenge of target engagement in the human septic niche. For example, colchicine suppresses NLRP3 inflammasome assembly and oxidative stress in LPS-induced ALI (123), while metformin inhibits GSDMD activation and downstream inflammatory amplification in CLP models (96), both demonstrating consistent attenuation of pyroptosis-associated lung injury in vivo. A critical barrier in translating pyroptosis-targeted strategies from murine models to human sepsis lies in species-specific divergence in inflammasome signaling, particularly the absence of direct orthology between murine caspase-11 and human caspase-4/5, which exhibit distinct activation thresholds and ligand sensitivities to cytosolic LPS (53). This molecular non-equivalence complicates the extrapolation of caspase-11-centered findings from LPS- or CLP-induced murine models of SALI to human patients in the ICU, where caspase-4/5-driven noncanonical inflammasome activation may follow different kinetics and cellular distribution (55).

Several therapies with potential applicability in the ICU setting have been described. These include anakinra (IL-1 receptor antagonist), which is already used in critically ill patients and has favorable pharmacologic feasibility due to its subcutaneous or intravenous administration, relatively predictable pharmacokinetics even in organ dysfunction, and established safety profile in sepsis trials (3). Mesenchymal stromal cells (MSCs) have also been administered intravenously in multiple ARDS ICU trials, although their feasibility is limited by the need for cell preparation and variable engraftment (124). By contrast, direct NLRP3 inhibitors (e.g., DFV890) and GSDMD inhibitors remain largely investigational, with limited data on ICU pharmacokinetics, potential drug interactions and tissue penetration in critically ill patients with altered hemodynamics and pulmonary vascular injury (125).

In addition, pharmacokinetic and pharmacodynamic constraints in critically ill patients, including altered drug absorption, organ dysfunction and unpredictable tissue distribution within injured pulmonary microvasculature, further limit reliable target engagement despite robust preclinical efficacy signals (126).

The expression of NLRP3-associated gene modules gradually increases along the histological transition in experimental models, while clinical translation remains limited by the lack of patient stratification (127,128). Clinical applicability of this axis depends on the early identification of patients with high inflammasome activity through plasma GSDMD cleavage products and IL-18 levels, preferably within 24-48 h of admission to the ICU. Emerging clinical signals from NLRP3 inhibitors, such as DFV890 and the soluble urokinase plasminogen activator receptor (suPAR)-guided administration of anakinra suggest that identifying hyper-pyroptotic endotypes is the only viable path to translating these findings into survival benefits (125,129). Therapeutic efficacy is likely to be maximized during the early hyperinflammatory phase, whereas late-stage administration risks exacerbating immunosuppression. The full pharmacological landscape of inhibitors targeting the canonical NLRP3-caspase-1-GSDMD axis is synthesized in Table II. Table II provides a comprehensive summary of various pharmacological agents targeting different nodes (priming, assembly and execution) of the canonical inflammasome pathway. It lists specific agents (e.g., MCC950, metformin, colchicine and natural compounds), their experimental inducers (mainly LPS or CLP), molecular mechanisms, and key outcomes in murine models, such as reduced pulmonary damage, reduced pyroptosis, reduced NLRP3 activation, reduced GSDMD cleavage, and improved survival, along with associated risks and limitations.

Targeting the canonical NLRP3 inflammasome pathway has shown promising results in preclinical studies (Table II).

Lineage plasticity is particularly evident in the non-canonical caspase-4/5/11 pathway (86). Unlike the canonical axis, this pathway directly senses cytosolic LPS within pulmonary endothelium, triggering rapid GSDMD-mediated membrane permeabilization (130). Distinct clusters of endothelial cells co-exist in both acute and subacute phases of lung injury compared to myeloid-dominant canonical models. These clusters differ significantly in marker genes, molecular signaling pathways and differentiation states, indicating a marked increase in transcriptional heterogeneity during endothelial transformation. The intermediate and transitional state of non-canonical activation is evident, as multiple pathways show consistent shifts along the histological transition from localized inflammation to systemic alveolar flooding (4). An upstream humoral-cellular crosstalk cluster shared among different sepsis models demonstrates potent differentiation potential according to the transcriptional differentiation trajectory. Pathway enrichment points to the stem-like characteristic of this subpopulation, suggesting it as the potentially pioneering force of lineage plasticity for non-canonical pyroptosis. The complement component 3a/C5a-C5aR axis serves as a key modulator, while vilobelimab has shown clinical success in severe respiratory failure (4). Notably, vilobelimab-mediated C5a blockade represents a clinically translatable example of complement-targeted intervention that has progressed beyond preclinical validation into randomized human ARDS trials, highlighting a partial bridge between murine non-canonical inflammasome models and ICU feasibility (131). The significant activation of non-canonical clusters occurs across disease phases. Current therapeutic strategies for this axis remain limited and require careful temporal control to avoid compromising host defense mechanisms in the early phase of sepsis. Targeting intratumoral heterogeneity or stem-like endothelial nodes has been proposed as a novel strategy to overcome treatment resistance. The full list of inhibitors targeting the non-canonical inflammasome pathway is synthesized in Table III. Table III summarizes therapeutic strategies targeting the non-canonical (caspase-4/5/11) pathway in S-ALI, including upstream complement modulation, caspase-11 inhibition and cellular therapies. Table III lists representative agents, molecular mechanisms, key outcomes (such as reduced endothelial pyroptosis, reduced vascular leakage, reduced lung injury and improved survival), as well as risks and limitations for each approach.

The concept of lineage plasticity in the septic lung is further reflected in the integrated PANoptosome complex, which enables a distinct phenotype of cell death with potential for state-shifting (87). Distinct clusters of dying cells where pyroptosis co-exists with apoptosis and necroptosis are more evident compared to traditional single-mode death models (90). These clusters differ significantly in marker genes, molecular signaling pathways and differentiation states, indicating a dramatic increase in transcriptional heterogeneity during death-mode switching. The intermediate and transitional state of GSDME-mediated injury is evident, as multiple pathways show consistent shifts from early apoptosis to terminal lysis (83). An epithelial cluster shared across injury groups demonstrates potent differentiation potential according to the transcriptional differentiation trajectory. Gene expression and pathway enrichment point to the stem-like characteristic of this subpopulation, suggesting it as the potentially pioneering force of lineage plasticity for PANoptosis in S-ALI. Caspase-3 and caspase-8 function as critical hubs for death-mode switching (132). For instance, ligustrazine suppresses both caspase-8 and NLRP3-caspase-1 signaling simultaneously, illustrating a multi-node regulatory strategy capable of modulating pyroptosis-apoptosis crosstalk within PANoptotic networks in septic lung injury (133). The significant activation of compensatory pathways occurs when one executioner is inhibited (134). This coordinated activation highlights the limitation of single-pathway inhibition and supports the need for multi-target or adaptive therapeutic strategies. Current therapies focusing on a single pathway underestimate the high plasticity of septic cells. Targeting intratumoral heterogeneity or stem-like protective cells, such as MSCs acting as biological rheostats, has been proposed as a novel strategy to overcome treatment resistance (135).

The full list of agents targeting apoptotic caspase-mediated pyroptosis is synthesized in Table IV. Table IV summarizes inhibitors of the apoptotic caspase-mediated pathway in S-ALI, detailing targeting nodes (primarily caspase-3, caspase-8 and caspase-9), representative agents (e.g., irisin, resveratrol, ligustrazine, MSCs), molecular mechanisms and key outcomes in murine models, such as reduced caspase-3 activity, reduced lung injury, reduced microvascular permeability and reduced epithelial pyroptosis, along with associated risks and limitations.

Although human evidence for the modulation of pyroptosis in S-ALI remains limited, it is increasingly promising. Circulating GSDMD fragments, IL-1β, IL-18 and inflammasome-associated signatures have been reported to be elevated in patients with sepsis in the ICU and in COVID-19-associated ARDS (56,136). Notably, these signals appear to be more prominent in hyperinflammatory bacterial sepsis than in certain viral ARDS cohorts, suggesting substantial inflammatory heterogeneity across patient populations (137).

Current biomarker-guided studies and key translational and clinical investigations targeting complement-inflammasome-pyroptosis-associated pathways have begun to provide early, yet heterogeneous evidence for clinical efficacy in critically ill patients. In particular, vilobelimab (anti-C5a monoclonal antibody) has been evaluated in randomized clinical trials in severe COVID-19-associated ARDS, where treatment was shown to be associated with a reduction in 28-day mortality and trends toward improved ventilator-free days in selected hyperinflammatory subgroups, although overall effects across unselected populations remained variable (138,139).

In addition to vilobelimab, earlier clinical attempts to broadly suppress inflammatory injury in sepsis and ARDS, without targeting specific pyroptosis-related pathways or applying biomarker-based stratification, have largely failed to demonstrate consistent improvements in mortality or organ failure scores, highlighting the limitations of non-selective anti-inflammatory approaches in critical illness. These findings collectively indicate that the therapeutic modulation of pyroptosis-related signaling requires both pathway specificity and patient stratification to achieve clinical efficacy.

Among candidate therapeutic nodes, upstream NLRP3 inflammasome blockade may provide the broader suppression of caspase-1 activation, IL-1β/IL-18 maturation and downstream pyroptotic amplification during the early hyperinflammatory stage of S-ALI (140). By contrast, direct GSDMD inhibition may more specifically prevent membrane pore formation and terminal cell lysis, while preserving part of the upstream immune sensing cascade (9). The modulation of the IL-1 pathway represents a downstream anti-inflammatory strategy that may attenuate cytokine-driven tissue injury, but does not completely suppress inflammasome activation or pyroptotic execution (141). Compared with conventional broad-spectrum anti-inflammatory therapies, pyroptosis-targeted interventions provide the potential advantage of mechanism-guided precision modulation by directly interrupting regulated inflammatory cell death pathways closely linked to alveolar barrier disruption, endothelial leakage, and maladaptive innate immune amplification in S-ALI (140).

Among the currently available translational strategies, substantial heterogeneity exists regarding both mechanistic specificity and clinical applicability. Upstream inflammasome-directed approaches, such as NLRP3 inhibition, may provide the broader suppression of caspase-1 activation and downstream cytokine maturation during the early hyperinflammatory phase; however, they may also increase the risk of excessive immune suppression when administered during later septic stages (142). By contrast, downstream IL-1 pathway blockade appears to represent a comparatively safer and clinically feasible strategy, particularly in biomarker-enriched hyperinflammatory subgroups, as illustrated by anakinra-related research (129). Complement-targeted interventions, such as vilobelimab may be particularly relevant in endothelial-dominant inflammatory phenotypes characterized by excessive C5a activation and neutrophil-driven tissue injury (138). Notably, several translational studies summarized in Table V have demonstrated that non-stratified anti-inflammatory interventions have frequently failed to achieve consistent survival benefits, highlighting that therapeutic efficacy is highly dependent on appropriate endotype selection, the timing of intervention and the preservation of essential host defense mechanisms. Collectively, current evidence supports a transition from empiric broad immunosuppression toward biomarker-guided precision modulation of pyroptosis-associated inflammatory pathways in S-ALI.

Notably, a substantial proportion of currently available human evidence is derived from COVID-19-associated ARDS cohorts rather than classical bacterial sepsis-induced ALI (137). Although these studies provide valuable proof-of-concept support for inflammasome and pyroptosis modulation, significant differences exist between viral and bacterial inflammatory programs, including pathogen-recognition pathways, neutrophil predominance, complement activation patterns and immune exhaustion trajectories (137). Therefore, caution is required when extrapolating COVID-19-based findings directly to heterogeneous sepsis populations.

Nevertheless, a substantial discordance persists between robust preclinical efficacy and relatively modest clinical outcomes. This discrepancy may partly reflect the dual role of pyroptosis in host defense and tissue injury, the existence of narrow therapeutic timing windows and the marked immunological plasticity observed during different stages of sepsis (143). A number of experimental inhibitors fail to translate clinically as they insufficiently account for patient stratification, inflammatory endotype, and stage-specific immune status. Patients with excessive inflammasome activation or elevated circulating IL-1β/IL-18 signatures may represent the most plausible candidates for pyroptosis-targeted interventions. Based on these signals, a biomarker-guided stratification framework may be used to identify a 'pyroptosis-high' inflammatory subphenotype in patients with S-ALI/ARDS (144), whereas inappropriate late-stage suppression could potentially aggravate sepsis-associated immunosuppression. A biomarker-guided stratification framework for S-ALI is defined as follows: Candidate pyroptosis-related biomarkers include circulating GSDMD cleavage products (GSDMD p30), IL-1β and IL-18, primarily measured in plasma or serum samples obtained within 24-48 h of admission to the ICU (9). Bronchoalveolar lavage fluid may be used in research settings; however, is not routinely feasible in critically ill patients. Real-world application is constrained by delayed laboratory turnaround time, variability in sampling timing, and dynamic changes in inflammatory status over the disease course. Biomarker interpretation should also consider confounding factors, such as acute kidney injury, extracorporeal membrane oxygenation and multi-organ dysfunction, which may significantly alter circulating inflammatory profiles (145).

Based on this framework, patients can be stratified into hyperinflammatory pyroptosis-high and hypo-inflammatory endotypes to support mechanism-guided therapeutic selection (9). In the early hyperinflammatory phase, characterized by elevated levels of GSDMD p30, IL-1β, and IL-18, therapeutic strategies are primarily directed toward inhibition of the canonical inflammasome axis, including NLRP3 and caspase-1 blockade, as well as downstream IL-1 signaling modulation. By contrast, in later stages or in patients progressing toward immunosuppression, excessive inflammasome inhibition should be avoided, and supportive or immune-restorative approaches may be more appropriate (9). In endothelial-predominant injury phenotypes, targeting the C5a-complement axis may provide additional benefit, whereas in mixed or highly plastic PANoptotic states, combination or multi-target strategies are likely required to achieve effective disease control (134).

To provide a coherent overview of translational readiness, representative human studies and ongoing or terminated clinical trials relevant to pyroptosis-modulating strategies in sepsis, ARDS and S-ALI are summarized in Table V.

Table V summarizes human evidence and clinical trials for pyroptosis-related interventions in S-ALI. It details various interventions/targets (e.g., circulating GSDMD as biomarker, NLRP3 inhibitors, anakinra, MSCs and vilobelimab), trial phases, target populations, mechanistic pathways, specific biomarkers for target engagement (such as plasma GSDMD-NT, IL-1β and IL-18), patient endotype stratification frameworks, as well as risks and limitations of each approach.

Of note, these studies collectively demonstrate that successful clinical translation depends not only on pathway inhibition itself, but also on appropriate patient stratification, inflammatory endotype characterization, therapeutic timing and the preservation of antimicrobial host defense.