Integration of immunogenic cell death in the treatment landscape of non-small cell lung cancer: harnessing the power of the immune system.

Facts

ICD effectively activates antitumour immune responses by releasing DAMPs such as CALR exposure, ATP secretion, and HMGB1 release, in a sequential cascade.

ICD is triggered by therapies via distinct mechanisms: chemotherapy primarily through endoplasmic reticulum stress; radiotherapy via DNA breakage and activation of the cGAS-STING pathway, and targeted therapy through precise interference with oncogenic signal.

Combining ICD-inducing treatments with immunotherapies, such as immune checkpoint inhibitors, represents a promising strategy to enhance immune activation and optimise the overall efficacy of ICD-based therapies.

To improve the clinical translation of ICD, future research should focus on identifying reliable predictive biomarkers and addressing immune suppression in advanced cancers to maximise therapeutic outcomes.

Open questions

How do specific factors within the tumour microenvironment determine the immunomodulatory function (immune activation vs. suppression of the immune system) of particular DAMPs, such as annexin A1?

How can parameters such as the radiotherapy dose and fractionation regimen, and the timing of its combination with immunotherapy, be precisely optimised to maximise immunostimulatory effects and minimise inhibitory ones?

Do antigen-specific T cell immune responses elicited by ICD induced via different mechanisms differ qualitatively in terms of their breadth, potency, and memory formation?

Which clinically detectable biomarkers (or combinations thereof) can stably and specifically predict successful induction of ICD and its ultimate clinical efficacy?

Introduction
Non-small cell lung cancer (NSCLC) is responsible for the majority of cancer-related deaths worldwide, comprising over 85% of lung cancer cases [1–3]. In 2022, there were nearly 2.5 million new diagnoses and 1.8 million deaths, representing 18.7% of the global cancer burden [4, 5]. Despite recent advances in conventional therapies such as surgery, chemotherapy and radiotherapy, the five-year survival rate for NSCLC remains low, ranging from 19 to 24% overall and less than 7% for advanced cases. Furthermore, more than 60% of patients experience recurrence or metastatic disease [6]. This challenging scenario highlights the limitations of traditional approaches in addressing tumour heterogeneity, systemic toxicity, drug resistance and the immunosuppressive tumour microenvironment [7]. To overcome these poor outcomes, immunogenic cell death (ICD) is being actively explored as a novel immunological mechanism [8].
ICD is a form of regulated cell death that is triggered by stress. It involves the release of damage-associated molecular patterns (DAMPs), such as the exposure of calreticulin (CALR), the release of adenosine triphosphate (ATP) and high mobility group box-1 protein (HMGB1). This activates dendritic cells (DCs), inducing tumour antigen-targeted T-cell responses [9–13]. ICD offers dual therapeutic action: direct tumour cytotoxicity and the induction of long-term immune memory, which impedes metastasis. Certain therapeutic approaches in clinical practice can trigger ICD, for example, specific chemotherapeutics (e.g. doxorubicin) induce the release of DAMPs via endoplasmic reticulum (ER) stress [9, 14], converting ‘cold’ tumours to ‘hot’ tumours. Radiotherapy (RT), molecular targeted therapy and photodynamic therapy (PDT) can also directly induce ICD [15–17]. Although immune checkpoint inhibitors (ICIs) cannot directly induce ICD, they can enhance ICD-triggered immune responses by relieving T-cell inhibition. This creates a synergistic effect that amplifies immune activation and therapeutic efficacy [18–20].
In light of the therapeutic potential of ICD in NSCLC, this review systematically elucidates the molecular mechanisms of ICD and critically evaluates preclinical and clinical progress in ICD-inducing therapies. By emphasising the pivotal role of ICD in remodelling the tumour microenvironment and overcoming immunosuppression, the review advances a new perspective on overcoming the limitations of conventional NSCLC treatments.

Overview of immunogenic cell death (ICD)
ICD is characterised by the ER stress–dependent and tightly coordinated release of DAMPs. Hallmark features of this process include the early surface exposure of ‘eat-me’ signals, such as CALR and heat shock proteins (HSPs) [9, 21, 22], the regulated secretion of the chemotactic ‘find-me’ signal ATP [23], and the late passive release of the alarmin HMGB1 [24] (Table 1 provides a comparative overview of these key molecules and their roles in ICD). Together, these signals drive DCs recruitment, maturation and antigen uptake, thereby facilitating effective antigen processing and presentation. This occurs predominantly through MHC class I–restricted cross-presentation to CD8⁺ cytotoxic T lymphocytes, while MHC class II–dependent activation of CD4⁺ T cells provides supportive immune functions. Ultimately, these coordinated events give rise to a robust, tumour-specific adaptive immune response [25, 26].
The core feature of ICD is the sequential release and precise regulation of DAMPs. This process is initiated by ER stress, which drives the translocation of ER-resident chaperones, including CALR and the HSPs HSP70 and HSP90, to the cell surface, thereby generating an ‘eat-me’ signal. Surface-exposed ER chaperones subsequently promote phagocytosis by antigen-presenting cells (APCs) through engagement of low-density lipoprotein receptor-related protein 1 (LRP1/CD91) [27, 28]. Notably, although therapy-induced CALR exposure shares certain conserved features, the signalling pathways that culminate in ICD are stimulus-specific. For instance, both anthracycline-based chemotherapy and hypericin-based PDT depend on a phosphoinositide 3-kinase (PI3K)-regulated secretory pathway to facilitate CALR externalisation [29]. However, chemotherapy-induced ICD uniquely involves PERK-dependent phosphorylation of eukaryotic initiation factor 2α (eIF2α) and activation of a caspase-8–dependent pathway, whereas hypericin-based PDT triggers BAX/BAK activation through B cell receptor-associated protein 31 (BAP31), highlighting mechanistic divergence in ICD initiation across therapeutic modalities [14].
During the early stages of ICD, in addition to CALR exposure, ATP is actively secreted through PERK/PI3K-dependent secretory pathways in dying cells where plasma membrane permeabilisation has not yet occurred [29–31]. As the cell death process progresses, the integrity of the plasma membrane gradually deteriorates. Activated Caspase 3 cleaves pannexin 1 (PANX1), forming channels in the plasma membrane that facilitate ATP release partially dependent on these PANX1 channels. Following membrane permeabilisation, the ATP release mechanism shifts to rely primarily on extensive passive diffusion through these channels [23, 32]. Extracellular ATP acts as a key ‘find-me’ signal, eliciting immune responses through a concentration-dependent biphasic mechanism. For example, low concentrations (EC₅₀ < 1 μM) activate the purinergic P2Y2 receptor on APCs to recruit monocytes, while high concentrations (EC₅₀ > 100 μM) stimulate the purinergic P2X7 receptor, promoting NLRP3 inflammasome activation, interleukin-1β (IL-1β) secretion, and enhanced γδ T cell and CD8⁺ T cell immune effect [30, 33, 34].
During the terminal phase of ICD, progressive plasma membrane permeabilisation permits the passive release of HMGB1 [24]. In monocytes and macrophages, extracellular HMGB1 exerts cytokine-like activities, the magnitude and context of which are regulated by inflammatory cues such as tumour necrosis factor (TNF), lipopolysaccharide and IL-1β [32]. The immunostimulatory function of HMGB1 is mediated through its interaction with pattern recognition receptors (PRRs), most notably Toll-like receptor 2 (TLR2) and Toll-like receptor 4 (TLR4) [35]. Among these pathways, activation of the HMGB1–TLR4–MYD88 signalling axis is particularly critical for the initiation of adaptive immune responses. Engagement of this pathway enhances antigen processing and presentation efficiency in DCs, thereby reinforcing innate immune co-stimulatory signals and ultimately promoting robust antitumour adaptive immunity [36, 37].
In addition to the aforementioned classic DAMPs, other signalling molecules such as type I interferons (IFN-I) and annexin A1 (ANXA1) also play an important and complex role in ICD. IFN-I are essential innate immune mediators that are triggered by cytosolic DNA during infection. DNA binds to cyclic GMP-AMP synthase (cGAS), catalysing the formation of the second messenger cyclic GMP-AMP (cGAMP). This activates the ER protein stimulator of interferon genes (STING). STING then recruits and phosphorylates TANK-binding kinase 1, which leads to the phosphorylation of nuclear factor kappa-B (NF-κB). This ultimately drives the expression of IFN-I and cytokine [38]. Secreted IFN-I then signals through the IFN-α/β receptor, inducing an antiviral/antibacterial state in neighbouring cells and promoting the activation of macrophages, DCs, and natural killer (NK) cells, thereby enhancing overall immune defence [39]. ANXA1 can limit inflammatory spread or enhance chemotherapy-induced ICD by binding to the formyl peptide receptors FPR2 or FPR1, respectively. The latter process potentially involves the activation of PRRs, such as TLR3/TLR4 [40]. Notably, ANXA1 exhibits context-dependent immunomodulation. In NSCLC, it binds to poly ADP-ribose polymerase 1 and upregulates Stat3-mediated PD-L1, thereby promoting immune evasion, proliferation, and metastasis [41]. Therefore, in NSCLC, ANXA1 may not act as an ICD-associated danger signal, but it could potentially be targeted to modulate tumour progression.
In summary, ICD links tumour cell death to antitumour immunity by orchestrating the release of DAMPs. Its core mechanism involves the exposure of CALR and the release of ATP and HMGB1, which collectively activate innate and adaptive immune responses, remodel the TME, reverse immune suppression and amplify antitumour effects. Future efforts should clarify how tumour heterogeneity affects ICD induction, identify biomarkers to predict ICD efficacy, and develop combination strategies based on the underlying mechanisms to enable precise, clinically translatable ICD induction.

Research progress of ICD in NSCLC
It has been demonstrated that chemotherapy, RT, molecular targeted therapy and PDT can induce ICD in NSCLC, thereby activating systemic antitumour immunity. This section provides a comprehensive review of the mechanisms, clinical efficacy, and future directions of these methods in inducing ICD in the context of NSCLC.

Chemotherapy-induced ICD
Conventional chemotherapeutic agents such as anthracyclines, taxanes, and platinum-based drugs can induce ICD and activate antitumour immunity in NSCLC. However, their non-selective cytotoxicity not only kills tumour cells but also damages immune cells and induces systemic toxicity. By contrast, targeted chemotherapy, as exemplified by antibody-drug conjugates (ADCs), enables the precise delivery of cytotoxic payloads and the induction of ICD, offering a key strategy to mitigate the systemic toxicity of conventional chemotherapy and achieve precise immune modulation. This section systematically elucidates the similarities and differences in ICD induction by these agents, evaluates the efficacy and safety of their clinical translation, and considers future research directions.
In the treatment of NSCLC, certain anthracyclines, taxanes and platinum-based agents can induce ICD via various mechanisms and with varying potency (Fig. 1). For example, anthracyclines such as doxorubicin (DOX) trigger the release of ATP and HMGB1 and the exposure of CALR, thereby activating T cells and M1 macrophage polarisation. This results in a time-lagged ICD effect [42–44]. Among taxanes, paclitaxel (PTX) promotes the release of DAMPs via mitochondrial oxidative stress, enhancing DCs maturation and T/NK cell proliferation [45, 46]. Docetaxel (DOC) increases HMGB1 release in a reactive oxygen species (ROS) -dependent manner, activating the NF-κB pathway and synergising with platinum drugs to enhance ICD [47, 48]. Among platinum-based agents, cisplatin (CDDP) alone induces the highest levels of DAMPs, while carboplatin (CARBO) has weaker effects, but enhances ATP release and CALR exposure when combined with taxanes. This results in stronger immune activation and antitumour responses in both vitro and in vivo settings [45, 48]. Future investigations should optimise combination strategies according to the heterogeneity of the ICD and clarify the underlying synergistic mechanisms.
In clinical practice, chemotherapeutic agents that induce ICD exhibit distinct efficacy and toxicity profiles (see Table 2 for more clinical data). Li et al. reported that, compared to inhalable CDDP, inhalable DOX demonstrated superior reduction in cancer nodule size (5.88% ± 3.98% vs. 4.15% ± 2.92%) and lower haematological and alopecia toxicity (e.g. anaemia 25% vs. 45%, alopecia 1% vs. 10%) [49]. Schiller et al. found that combinations such as PTX plus CDDP/CARBO and DOC plus CDDP had a comparable objective response rate (ORR) of 17–21% and median survival of 7.4–8.1 months in advanced NSCLC. However, they differed significantly in terms of toxicity: PTX + CDDP was associated with more pronounced haematological toxicity (e.g. a 75% risk of grade 3–4 neutropenia) and gastrointestinal reactions. In contrast, PTX + CARBO presented a lower risk of gastrointestinal toxicity and febrile neutropenia. For DOC + CDDP, hypersensitivity reactions require attention, with a 7% risk of grade 3–4 events [50]. Furthermore, Belani et al. observed an ORR of 43% (95% CI: 24–63%) and a median overall survival (OS) of 13.9 months (range 1–35+ months) with DOC + CARBO, albeit with notable haematological toxicity (a 79% risk of grade 4 neutropenia) [51]. Therefore, treatment selection should be individualised based on these distinct toxicity profiles.
Single chemotherapeutic agents often cause severe systemic toxicity due to their limited selectivity. In contrast, ADCs deliver cytotoxic payloads directly to tumour cells via antibody-mediated targeting. This allows for the potent localised killing of cells and the induction of ICD, while minimising toxicity to healthy tissues. Preclinical studies have shown that ADCs such as luveltamab tazevibulin (Fig. 1), 1959-sss/DM4 and datopotamab deruxtecan trigger the release of DAMPs (e.g. HMGB1, ATP and CALR), thereby activating antitumour immunity (see Table 3 for more details and examples) [52–54]. In clinical trials (see Table 2 for more clinical data), tusamitamab ravtansine (DM4 payload) achieved an ORR of 20.3% with CEACAM5-high non-squamous NSCLC [55]. Datopotamab deruxtecan has demonstrated significant efficacy in the treatment of advanced/metastatic NSCLC, achieving an ORR of 35.8% (95% CI: 27.8–44.4) in the TROPION-Lung05 trial. Furthermore, it has shown superior median progression-free survival (PFS) compared to DOC (4.4 vs. 3.7 months; HR: 0.75, p = 0.004) in TROPION-Lung01 trial. Additionally, it has been associated with a lower incidence of grade ≥3 treatment-related adverse events (25.6% vs. 42.1%). However, manageable risks such as adjudicated drug-related interstitial lung disease/pneumonitis (3.6–8.8%) require monitoring [56, 57]. Thus, ADCs represent a promising strategy for balancing efficacy and safety in NSCLC by inducing targeted ICD.
In summary, chemotherapeutic agents, including DOX, PTX, DOC, CDDP, and CARBO, can induce ICD in NSCLC through heterogeneous mechanisms. While combination regimens of these agents show comparable efficacy but distinct toxicity profiles, informing personalised selection, ADCs enable precise chemotherapy by targeting ICD induction and demonstrate promising clinical potential. Further studies are needed to elucidate the mechanistic differences in chemotherapy-induced ICD, explore timing-based combinations, and optimise ADCs application, with the aim of advancing more effective and safer individualised therapies.

Radiotherapy (RT)-induced ICD
RT represents a classical modality capable of inducing features of ICD through the generation of DNA damage, ER stress and ROS, thereby promoting the exposure or release of DAMPs, including CALR, ATP and HMGB1. The capacity of radiotherapy to induce ICD is widely regarded as a key immunological basis underlying the so-called abscopal effect [58]. Clinically, this phenomenon refers to the regression of distant, non-irradiated tumour lesions following local irradiation of a primary site, reflecting a systemic antitumour immune response elicited by radiotherapy-induced immune activation [59–62]. Notably, this immunomodulatory effect is markedly dose-dependent.
Early studies demonstrated that increasing radiation doses were associated with enhanced ICD-related DAMP release. For instance, Golden et al. observed a dose-dependent increase in CALR exposure, ATP secretion and HMGB1 release when radiation doses were escalated from 2 to 20 Gy in vitro breast cancer models. However, accumulating evidence indicates that RT-induced ICD does not follow a linear dose–response pattern. Rather, RT alone generally induces limited ICD, with an apparent immunogenic dose window beyond which further dose escalation fails to enhance, or may even suppress, immunogenic signalling [63]. In NSCLC models, such as A549 cells, fractionated regimens (e.g. 8 Gy × 3 or 4 Gy × 2) have been shown to robustly induce ICD markers, an effect that can be further potentiated by radiosensitisers, including Bio-MnO₂ nanoparticles or ZnO-Au@mSiO₂ constructs [15, 64].
In clinical practice, patients with early-stage, inoperable NSCLC are frequently treated with high-dose stereotactic body radiotherapy (SBRT). A retrospective analysis by Koshy et al. demonstrated superior 3-year OS in patients receiving a biologically effective dose exceeding 150 Gy compared with lower-dose regimens (55% vs. 46%), although the maximum tolerated dose was not defined in that study [65]. Subsequently, Bezjak et al. established 12.0 Gy per fraction as the maximum tolerated SBRT dose, with a relatively low incidence of dose-limiting toxicity (7.2%; 95% CI: 2.8–14.5%). Comparable 2-year local control, OS, and PFS rates were observed with fraction doses of 11.5 or 12.0 Gy [66]. While SBRT achieves excellent tumour control primarily through direct cytotoxicity, its capacity to optimally activate systemic antitumour immunity may be limited, underscoring a potential divergence between regimens designed for maximal local control and those favouring ICD.
Notably, RT exerts dual regulatory effects on the TME. While inducing ICD, it can also upregulate immune checkpoint molecules such as PD-L1, thereby promoting immunosuppression. Preclinical studies confirm that specific doses of RT increase PD-L1 expression on tumour cells, DCs and macrophages, and impair CD8⁺ T-cell function within the TME [67]. There is also clinical support for this, as NSCLC patients with a history of RT show a significantly higher PD-L1 positivity rate (tumour cell cutoff ≥25%) than those without (43.7% vs. 37.1%) [68]. This provides a rationale for combining RT with ICIs, yet clinical outcomes vary (see Table 2 for more clinical data). For example, the phase II DOLPHIN trial reported favourable survival benefits (12-month PFS: 72.1%; median PFS: 25.6 months; ORR: 90.9%) with 60 Gy of RT plus durvalumab [69]. Conversely, a retrospective cohort study by Mander et al. found that palliative RT (e.g. 8 Gy × 1 or 20 Gy × 5) combined with pembrolizumab resulted in a significantly decreased PFS compared to no RT (HR = 6.254) [70]. This discrepancy likely stems from differences in RT dose and fractionation, highlighting the importance of optimising these parameters for successful combination therapy.
RT-induced immunosuppression extends beyond PD-L1 upregulation to include suppression of the cGAS-STING pathway. Vanpouille-Box et al. discovered that radiation doses exceeding 12–18 Gy per fraction stimulate the production of the DNA exonuclease Trex1, which breaks down cytosolic DNA. This inhibits cGAS-STING activity and IFN-β production [71], suggesting that high-dose RT may hinder immune activation driven by ICD. Notably, Xue et al. confirmed that cGAS-STING activation enhances radiosensitivity in NSCLC [72]. In this context, nanomaterial-based strategies designed to potentiate cGAS–STING signalling represent a promising approach to optimise RT-induced ICD. For instance, Bio-MnO₂ nanoparticles combined with low-dose RT function as radiosensitisers by enhancing ROS generation while simultaneously releasing Mn²⁺ ions that activate cGAS–STING signalling, thereby promoting cytotoxic T-lymphocyte infiltration [15]. Similarly, ZnO-Au@mSiO₂ nanoparticles synergise with low-dose RT to trigger ICD, inhibit PD-1/PD-L1, and activate cGAS-STING, collectively reversing T-cell exhaustion [64]. Together, these findings suggest combining optimised (preferably moderate-to-low) RT doses with nanotechnology to activate the cGAS-STING pathway is a promising approach for enhancing immunotherapeutic efficacy (Fig. 2).
In summary, RT can induce ICD and initiate systemic antitumour immunity, however, its immunogenic potential is highly dependent on radiation dose and fractionation. While moderate RT doses promote DAMPs release and cGAS–STING–mediated immune activation, excessive doses may instead enhance immunosuppressive signalling through PD-L1 upregulation and Trex1-mediated inhibition of innate immune sensing. Therefore, optimising RT parameters, rather than simple dose escalation, is critical for maximising ICD induction and achieving effective synergy with immunotherapeutic strategies.

Molecular targeted therapy-induced ICD
Recent studies have revealed that various molecular targeted agents, such as protein tyrosine kinase inhibitors (TKIs), exert antitumour effects in NSCLC, which are associated with the induction of ICD-related immunogenic signalling, and their molecular mechanisms share common features but also exhibit distinct differences. Preclinical studies, primarily based on NSCLC cell lines and murine models, suggest that ALK inhibitors (e.g. crizotinib and ceritinib), EGFR inhibitors (osimertinib), AXL inhibitors (bemcentinib and marsdenia tenacissima extract), and vascular endothelial growth factor receptor inhibitors (anlotinib) have been shown to induce key features of ICD, including CALR exposure, ATP release and HMGB1 secretion. This represents a convergent immunogenic signalling axis contributing to antitumour immune activation [17, 73–77]. However, the mechanistic emphasis varies. Crizotinib-induced ICD relies on T lymphocytes and the IFN-γ pathway, while also upregulating PD-L1 and strongly sensitising NSCLC to immunotherapy with PD-1 antibodies [73]. Osimertinib may increase plasma-soluble CALR levels in patients by upregulating the phosphorylation of eIF2α [17]. Bemcentinib has been reported to exert only modest effect on MHC-I expression [75]. In preclinical models the marsdenia tenacissima extract induces ER stress by inhibiting AXL phosphorylation, leading to mitochondrial depolarisation and ROS accumulation [76]. Anlotinib inhibits AKT/mTORC1 and Pparδ signalling to block M2 polarisation of tumour-associated macrophages during ICD [77] (Fig. 3).
In parallel with their ICD-associated immunomodulatory properties, these TKIs have demonstrated substantial clinical efficacy across multiple stages of NSCLC (see Table 2 for more clinical data). In patients with advanced ALK-positive NSCLC, crizotinib significantly prolonged PFS (HR = 0.402) and improved the ORR (87.5% vs. 45.6%) compared to chemotherapy [78]. Ceritinib has been shown to provide superior OS (HR = 0.59) and PFS (median 13.8 vs. 8.3 months) when used as initial therapy, compared to crizotinib [79]. Furthermore, it yields a better PFS than chemotherapy (median 5.4 vs. 1.6 months, HR = 0.49) in patients who have progressed following prior treatment with crizotinib [80]. For EGFR-mutant NSCLC, osimertinib has been shown to be effective at some stages of treatment: it achieved an ORR of 52% in the neoadjuvant setting [81]; when used as adjuvant therapy, it increased the 4-year disease-free survival rate to 70% (compared to 29% in the control group) and reduced the risk of central nervous system recurrence by 76% (HR = 0.24) in patients with stage IB-IIIA disease [82]; and when used as maintenance therapy for unresectable stage III NSCLC, it increased median PFS to 39.1 months (HR = 0.16) [83]. In AXL-positive advanced NSCLC, bemcentinib combined with DOC resulted in a median PFS of 3.1 months (95% CI: 1.6–5.7) and a median OS of 12.3 months (95% CI: 4.9–16.9) [84], while a marsdenia tenacissima extract-based injection plus chemotherapy improved effective rate (RR = 1.32, 95% CI: 1.14–1.54) and quality of life improvement rate (RR = 2.04, 95% CI: 1.69–2.47) [85]. Furthermore, the combination of anlotinib and a PD-1 inhibitor in patients who had progressed following prior immunotherapy treatment was associated with a disease control rate of 82.8%, a median PFS of 5.0 months (90% CI: 4.2–7.3), and a median OS of 15.1 months (90% CI: 4.2–7.3) [86]. Taken together, these data establish the clinical efficacy of TKIs in NSCLC and suggest that immune modulation involving ICD-associated mechanisms, may contribute to therapeutic outcomes. However, direct clinical evidence linking ICD induction to patient benefit remains limited, underscoring the need for biomarker-driven validation.
In summary, molecular targeted agents, including TKIs, polo-like kinase inhibitors [87], and serine/threonine kinase activators [88] (see Table 3 for a list of polo-like kinase inhibitors and serine/threonine kinase activators that could induce ICD in NSCLC), could reshape the tumour immune microenvironment by inducing ICD and synergising with ICIs. Future efforts should focus on elucidating the specific molecular pathways of ICD induction and validating efficacy through optimised combination strategies and prospective clinical trials. This will advance the precise application of targeted therapies in NSCLC treatment.

Photodynamic therapy (PDT)-induced ICD
The induction of ICD by PDT relies on the preferential accumulation of photosensitisers (PSs) in tumour tissues, intersystem crossing-mediated singlet/triplet excitation, and subsequent generation of Type II ROS (primarily singlet oxygen, ¹O₂) via energy transfer, or Type I ROS (e.g. superoxide anion O₂•−, hydrogen peroxide H₂O₂, hydroxyl radical •OH) via electron transfer [89]. These ROS induce tumour cell death through both direct cytotoxicity and vascular damage, while also initiating intracellular stress responses that are critical for ICD induction. [90].
The immunogenic profile of PDT-induced cell death is strongly influenced by the subcellular localisation of PSs, particularly those targeting the ER or mitochondria. Valančiūtė et al. demonstrated that methylene blue-PDT (MB-PDT) triggers ICD in NSCLC by inducing ER stress, promoting CALR membrane translocation, and elevating intercellular cell adhesion molecule-1 expression. MB-PDT inhibits NSCLC proliferation, downregulates Bcl-2 and activates Caspase 3/7. It also enhances CD8⁺ T cell-mediated tumour lysis via granzyme B in cell lines and organoids [91]. Our research group has developed a perylene monoamide-based, mitochondrial-targeted ROS supergenerator, PMIC-NC. This supergenerator not only induces hypoxia-enhanced Type I ROS bursts with the aid of proton transients but also triggers the generation of type I/II ROS through electron or energy transfer under near-infra-red radiation (NIR) irradiation, releasing ATP and HMGB1 with exposing CALR, eliciting a robust ICD effect (Fig. 4). Its mitochondrial/lung-specific localisation increases therapeutic efficacy. By integrating NIR-PDT with hypoxia-responsive chemotherapy and ICD induction, this platform enables immunogenic photochemical therapy in preclinical models of primary, distant, and metastatic tumours [16].
Novel PS systems (e.g. nano drug delivery systems and metal-based PSs) optimise NSCLC therapy. The ICG@MnO₂@Exo-anti-PD-L1 nanoplatform developed by Guo et al. reversed immunosuppression via PD-L1 blockade and generated O₂ by catalysing H₂O₂ decomposition to release Mn²⁺ and activate macrophage STING. It also triggered NF-κB p65-mediated M1 polarisation and pro-inflammatory cytokine secretion alongside HMGB1 release to induce ICD in LLC cells [92]. Ai et al.’s NIR-activatable BPQDs@Lipo-YSA targeted NSCLC by inducing the PRKN/AKT1 pathway-mediated mitophagy process. Under NIR, it triggered potent ICD (CALR exposure and HMGB1 release), enhancing in vitro cytotoxicity and inhibiting in vivo tumour growth [93]. Metal-based PSs, such as specific iridium (III) complexes, have been shown to induce potent ICD in non-NSCLC models [89, 94], whereas some metal phthalocyanines [95, 96] and certain iridium (III) complexes [97, 98] have exhibited photodynamic efficacy in NSCLC models without a confirmed link to ICD highlighting the need for rigorous ICD-specific validation. Future research should assess the balance of safety and efficacy of metal-based PSs for inducing ICD in NSCLC.
Current research on PDT-induced ICD mechanisms has shifted from the cytotoxic effects to the construction of immunoregulatory networks. Future priorities include quantifying the correlations between subcellular localisation of PSs and ICD pattern, the impact of tumour heterogeneity on ICD efficiency, and the long-term biosafety of metal-based PSs. Advances in nanotechnology and mechanistic understanding are expected to facilitate the rational design of PDT-based strategies that integrate ICD induction with immunotherapy, thereby improving the precision and translational potential of PDT in NSCLC.

Other emerging induction strategies and regulatory factors
Beyond conventional ICD-inducing therapies, emerging modalities such as oncolytic virotherapy and tumour-treating fields have been reported to induce ICD-associated features in preclinical NSCLC models, including DAMPs exposure ect (see Table 3) [99–103]. Recent studies further highlight the role of the gut microbiota in shaping ICD-associated immune responses through the gut–lung axis. Specific beneficial bacteria (e.g. Lactobacillus rhamnosus, Akkermansia muciniphila, Turicibacter and Peptococcus) can enhance antitumour immunity by promoting DCs maturation, T-cell activation and effector cell infiltration, thereby facilitating immune priming downstream of ICD. Such effects may reshape the tumour immune microenvironment to favour ICD-associated immune responses. Conversely, other gut microbiota (e.g. Enterorhabdus and Desulfovibrio) may promote an immunosuppressive microenvironment by increasing myeloid-derived suppressor cells infiltration, raising pro-tumour cytokine levels and inhibiting NK cell and M1 macrophage activation. This can potentially impair ICD-triggered immune responses [104–106]. Collectively, these findings underscore the multifaceted role of the gut microbiota in modulating ICD-associated immunity and suggest that microbiota-targeted interventions may serve as a potential adjuvant strategy to enhance ICD-based therapies in NSCLC, pending further mechanistic and clinical validation.

Conclusion and perspectives
In the treatment of NSCLC, inducing ICD represents an important mechanism that links direct tumour cell killing with the activation of systemic antitumour immunity. Conventional therapies, including chemotherapy, RT, molecular targeted therapy and PDT, can trigger ICD via distinct mechanisms: chemotherapy primarily through ER stress; RT through DNA damage; targeted agents by disrupting specific signalling pathways; and PDT through ROS generation (see Fig. 5 for a summary) [48, 73, 107, 108]. Combining RT and targeted therapies with ICIs can enhance ICD-associated immune priming while counteracting therapy-induced immune suppression, such as PD-L1 upregulation [67, 69, 73]. The mechanistic overlaps and differences among these modalities provide a basis for designing sequential and synergistic combination regimens.
Although extensive research related to the immunomodulatory effects of ICD has been conducted, there are several challenges to its clinical translation. Firstly, the context-dependent and heterogeneous biological effects of distinct ICD inducers across diverse tumour microenvironments require deeper mechanistic investigation in preclinical models [109]. Secondly, future clinical trials should incorporate validated ICD-related biomarkers to monitor dynamic changes in the tumour immune microenvironment of patients with NSCLC. Thirdly, personalised therapeutic strategies are required to enhance antitumour immune activation while minimising systemic toxicity.
In summary, ICD serves as a critical bridge between tumour cell death and the activation of adaptive antitumour immunity, offering a promising strategy to overcome the immunosuppressive tumour microenvironment in NSCLC and showing translational potential in clinical applications. However, its successful clinical translation requires a deeper understanding of its mechanisms, the identification of reliable biomarkers and rigorous and scientific clinical trials. Ultimately, these advances are essential for establishing a refined immunotherapeutic paradigm for NSCLC.