Emerging advances in defining the molecular and therapeutic landscape of small-cell lung cancer

Abstract

Small-cell lung cancer (SCLC) has traditionally been considered a recalcitrant cancer with a dismal prognosis, with only modest advances in therapeutic strategies over the past several decades. Comprehensive genomic assessments of SCLC have revealed that most of these tumours harbour deletions of the tumour-suppressor genes TP53 and RB1 but, in contrast to non-small-cell lung cancer, have failed to identify targetable alterations. The expression status of four transcription factors with key roles in SCLC pathogenesis defines distinct molecular subtypes of the disease, potentially enabling specific therapeutic approaches. Overexpression and amplification of MYC paralogues also affect the biology and therapeutic vulnerabilities of SCLC. Several other attractive targets have emerged in the past few years, including inhibitors of DNA-damage-response pathways, epigenetic modifiers, antibody–drug conjugates and chimeric antigen receptor T cells. However, the rapid development of therapeutic resistance and lack of biomarkers for effective selection of patients with SCLC are ongoing challenges. Emerging single-cell RNA sequencing data are providing insights into the plasticity and intratumoural and intertumoural heterogeneity of SCLC that might be associated with therapeutic resistance. In this Review, we provide a comprehensive overview of the latest advances in genomic and transcriptomic characterization of SCLC with a particular focus on opportunities for translation into new therapeutic approaches to improve patient outcomes.

Introduction

Small-cell lung cancer (SCLC) accounts for approximately 15% of lung cancers, is characterized by a high degree of neuroendocrine differentiation and arises predominantly in current or former smokers^1,2. In clinical practice, SCLC has been classified as limited-stage SCLC (LS-SCLC) and extensive-stage SCLC (ES-SCLC) according to the US Veterans Administration classification system. LS-SCLC constitutes SCLC that is confined to one hemithorax where the primary tumour and regional lymph nodes can be adequately encompassed within one radiation field, with no extra thoracic metastasis. On the contrary, ES-SCLC includes disease that cannot be classified as limited stage including pleural, pericardial or distant metastasis. In comparison with non-SCLC (NSCLC), ES-SCLC has a worse prognosis, with a 5-year overall survival (OS) generally <5%^3–5. However, with the advent of immune checkpoint inhibitors (ICIs), updated clinical trial results indicate that the 3-year and 5-year OS of patients with ES-SCLC might be approaching 15% and 12%, respectively^6,7. Approximately 250,000 patients are diagnosed with SCLC each year globally, of whom approximately 200,000 die from the disease⁸. Few patients with SCLC undergo surgical resection because >70% of them are diagnosed with ES-SCLC, for which systemic therapy is usually the only treatment option⁹.

Multiple preclinical studies have suggested that neuropeptides have a regulatory role in SCLC by promoting cancer cell proliferation and survival. Similarly, acetylcholine, a neurotransmitter that is known to stimulate cell growth through either the nicotinic or muscarinic cholinergic signalling pathway, is synthesized by neuroendocrine SCLC cells and acts as an autocrine growth factor¹⁰. The cholinergic signalling axis also seems to have pro-invasive and pro-migratory activity in SCLC¹¹. Owing to the secretion of, and/or autoimmune responses to, specific molecules, some patients with SCLC present with a wide range of paraneoplastic syndromes with neuromuscular manifestations, which further complicates patient management¹².

First-line therapy for SCLC has traditionally involved platinum-based chemotherapy as a ‘one-size-fits-all’ approach, given the perceived homogeneous pathological morphology of this malignancy and high objective response rates (ORRs) to DNA-damaging agents in patients with this disease. However, these responses are not durable, with most patients having disease recurrence or progression within 6 months^13,14. Despite the favourable results of studies testing the addition of ICIs to chemotherapy in patients with ES-SCLC, which led to FDA approvals in 2019 and 2020, the improvement in median OS with these combinations compared with chemotherapy alone is, at best, modest¹⁵. These results contrast with those in NSCLC, in which trials testing ICIs have translated into meaningful improvements in patient outcomes.

Obtaining adequate tumour samples from patients with SCLC is difficult, especially at relapse, owing to rapid disease progression and patient co-morbidities; therefore, diagnosis of SCLC usually relies on small biopsy samples or fine-needle aspiration cytology. As a result, limited data are available on the genomic landscape, mutational complexity and disease biology of SCLC at progression because this tumour type has traditionally not been included in large-scale sequencing initiatives, such as the Pan-Cancer Analysis of Whole Genomes (PCAWG) or The Cancer Genome Atlas (TCGA). In the past decade, researchers have focused on identifying transcription factor-defined subtypes in SCLC, which has accelerated our understanding of the disease biology and enabled an improved characterization of tumour heterogeneity and plasticity. In this Review, we provide an overview of advances in the multiomic characterization of SCLC over the past decade and how they could translate into novel diagnostic and therapeutic approaches.

Key genomic aberrations in SCLC

SCLC is strongly associated with smoking, and a majority of these tumours (~97.5%) arise in current or former smokers^1,2,16. In addition to the de novo SCLCs diagnosed in non-smokers (<2%)¹, some SCLCs seem to occur as a result of selective pressure and transformation of lung adenocarcinomas (LUADs) harbouring EGFR mutations or ALK alterations (~14%)^17–19. SCLC is characterized by a high tumour mutational burden (TMB; ~9.9 mutations per megabase), with C:G>A:T transversions being the most common base substitutions caused by DNA damage mechanisms resulting from smoking²⁰. Although a high TMB is known to be associated with an immunologically active tumour microenvironment (TME) owing to efficient neoantigen processing and presentation, ICIs have demonstrated limited efficacy in patients with SCLC²¹. Unlike other tumour types, TMB does not conclusively predict response to ICIs in patients with SCLC^22,23. This observation can be partly explained by the transcriptional repression of MHC molecules on the surface of SCLC cells, resulting in limited antigen presentation²⁴.

In a landmark study, investigators performed comprehensive genomic profiling of 110 SCLCs²⁵. However, these were mainly surgically resected tumours, which do not necessarily reflect typical SCLCs with disseminated presentation at diagnosis. In this study, a majority of SCLCs (~95%) harboured biallelic inactivation of the tumour-suppressor genes TP53 and RB1²⁵ (Fig. 1a). This observation supported the long-standing hypothesis of loss of tumour-suppressor activity as a sentinel rate-limiting event for SCLC initiation. This concept was subsequently validated in the RP model, a genetically engineered mouse model (GEMM) in which the deletion of these two genes leads to the development of tumours resembling SCLCs in patients²⁶. The loss of RB1 and/or TP53 is a common feature of SCLC but is not necessarily a universal event; indeed, ~5% of SCLCs harbour wild-type RB1 and/or TP53 (ref. 27). A large-scale genomic analysis of real-world samples (>3,000 SCLCs) with results published in 2023 revealed that SCLCs with wild-type RB1 and/or TP53 tend to have lower TMB and are more prevalent in younger individuals and those of non-white ancestry²⁸.

Fig. 1 ∣. Genetic and transcriptomic landscape of SCLC.

a, Mutational landscape of small-cell lung cancer (SCLC). SCLC is characterized by an almost ubiquitous loss of TP53 and RB1. A substantial percentage of SCLCs also have amplification of MYC family members (MYC, MYCN and MYCL), and a smaller percentage harbour mutations in other genes, as shown. b, Evolution of SCLC transcriptional subtypes over time. The relative expression of key transcription factors is used to define subtypes within SCLC. This figure summarizes the chronology of preclinical studies that demonstrated the complexity of SCLC biology and evolution of these subtypes. NE, neuroendocrine.

Additional genes recurrently altered in SCLC include other members of the retinoblastoma family (RBL1 and RBL2), tumour suppressors (NF1 and PTEN), components of the Notch signalling pathway (NOTCH1, NOTCH2, NOTCH3 and NOTCH4), and regulators of chromatin integrity (ARID1, CREBBP, EP300 and KMT2D)²⁸ (Fig. 1a). Overall, these genomic characteristics were similar between treatment-naive patients and patients with relapsed and/or recurrent (RR) SCLC after platinum-based chemotherapy, suggesting that these events are truncal and occur clonally early in SCLC tumorigenesis²⁹. The aforementioned real-world study also identified recurrent alterations in SCLC, including mutations in STK11 (1.7%), KEAP1 (~3%) and PI3K pathway-related genes (PTEN (9.9%), PIK3CA (5.6%) and RICTOR (5.6%))²⁸. Interestingly, these investigators identified a small subset of SCLCs that are positive for the human papilloma virus (HPV). Further analysis revealed that 12.7% of tumours with wild-type TP53 and/or RB1 are HPV⁺ compared with only 1.8% of those harbouring TP53 and RB1 mutations. The mechanistic relationship between HPV status and TP53 and RB1 mutations in SCLC needs to be addressed in future studies. Interestingly, RICTOR amplification in patients with SCLC was found in another study, further suggesting the potential of testing mTORC1/2 inhibitors in this subgroup of patients³⁰. Most importantly, no recurrent targetable genomic alterations have been validated in clinical trials involving patients with SCLC. Even if multiple molecularly targeted therapies have been tested in clinical trials involving patients with SCLC, none has provided durable benefit³¹.

Analyses of somatic copy number variants (CNVs) have consistently identified recurrent losses of 3p (encompassing FHIT and RASSF1), 10q (PTEN), 13q (RB1) and 17p (TP53), chromosomal lesions and amplifications in 1p (MYCL), 2p (MYCN), 8q (MYC), 5p and 3q^25,28. In contrast to other cancer types examined extensively in TCGA, these somatic CNV profiles do not define any subgroups of SCLC³².

Chromosomal translocations, insertions and deletions resulting in gene fusions frequently drive oncogenesis in solid tumours³³, although their role in SCLC remains to be fully determined. Among several intrachromosomal rearrangements identified in SCLC-derived cell lines and tumours, a recurrent RLF–MYCL fusion is the most commonly described (in ~5–10% of models)^34,35. An Rlf–Mycl-driven mouse model of SCLC was established using a CRISPR–Cas9 somatic gene editing approach, in which this fusion accelerated the transformation, proliferation and dissemination of SCLC cells to distinct organs, mirroring features of SCLC typically observed in the clinic³⁶. Further studies characterizing functionally active fusions with clinical relevance in SCLC and their therapeutic vulnerabilities are warranted.

The amplification or overexpression of certain genes could affect the outcomes of patients with SCLC. For example, the amplification or overexpression of MYC paralogues could guide several therapeutic strategies. The MYC family comprises a set of regulator genes that have a crucial role in cell cycle progression, apoptosis and cell transformation. MYC amplifications can lead to overexpression of MYC protein, which can upregulate genes involved in cell cycle progression thereby driving enhanced cell proliferation and metabolism. MYC induces transcription of AURKA, which encodes Aurora A kinase, a regulator of mitosis and SCLC tumorigenesis³⁷. Indeed, constitutive activation of MYC paralogues is sufficient to promote cell cycle entry and sustain replication in different cellular contexts³⁸. The inhibition of this interaction results in MYC degradation and cell death³⁹. A proteomic analysis identified overexpression of MYC as the top predictive biomarker of sensitivity to the CHK1 inhibitor prexasertib⁴⁰. Although prexasertib failed to significantly improve ORR in a phase II trial involving patients with RR SCLC⁴¹, patients with tumours with amplification or overexpression of MYC might derive benefit from CHK1 inhibition. In a subgroup analysis of another phase II study, patients with SCLC overexpressing MYC who received the Aurora A kinase inhibitor alisertib plus paclitaxel had a significant improvement of progression-free survival (PFS; 7.2 months versus 4.5 months with placebo and paclitaxel; HR 0.43, 95% CI 0.26–0.70; P < 0.001)⁴². In a GEMM of SCLC⁴³, overexpression of MYCN was associated with resistance to platinum-based agents. Moreover, genome-scale CRISPR–Cas9 screening identified deubiquitinase USP7 as an MYCN-associated synthetic vulnerability. Pharmacological inhibition of USP7 resensitized SCLCs to chemotherapy⁴³. In the aforementioned large-scale real-world analysis, amplification of 4q12 and CCNE1 was associated with improved and worse OS, respectively²⁸. These data suggest that these MYC-related candidate genes influence the outcomes of patients with SCLC and, thus, could represent attractive targets for drug development.

Given that brain metastases are present in 10–20% of individuals with SCLC at diagnosis and generally portend an unfavourable prognosis^44,45, identifying novel systemic targeted therapies with enhanced intracranial penetrance is an unmet need. In the past few years, studies have found a higher prevalence of alterations in PTEN and genes regulating mTOR activity in samples derived from SCLC brain metastases relative to other metastatic sites²⁸. The potential of targeting these vulnerabilities with combination therapies remains to be explored.

Hereditary pathogenic germline variants (PGVs) in the DNA damage repair (DDR) pathway across solid tumours could imply therapeutic vulnerabilities, in particular, synthetic lethality of poly(ADP-ribose) polymerase (PARP) inhibitors in patients harbouring pathogenic germline mutations in BRCA1 or BRCA2 (ref. 46). Moreover, certain non-BRCA PGVs can inform targeted therapies (such as inhibitors of ATM or CHEK2) and improve clinical trial eligibility of patients across various solid tumours, including SCLC⁴⁷. In an observational study⁴⁸, whole-exome sequencing (WES) was performed in peripheral blood mononuclear cells from patients with SCLC. Among 77 patients, 34 (44.2%) had pathogenic or likely PGVs. Nine (11.7%) of those mutations were included in the list of the American College of Medical Genetics and Genomics, which lists 24 high-penetrance autosomal dominant or recessive genes requiring counselling with a geneticist⁴⁹ (Fig. 1). Most of the identified PGVs included genes involved in the DDR pathways. Germline carriers were significantly enriched among first-degree relatives with cancer or lung cancer (OR 1.82 and 2.60, respectively) and had prolonged recurrence-free survival after platinum-based chemotherapy compared with non-carriers⁴⁸. The difference in OS was not statistically significant, suggesting that PGVs have limited prognostic value. A patient with SCLC and a PGV in BRIP1, which encodes a helicase that interacts with BRCA1, received the topoisomerase 1 inhibitor camptothecin and the PARP inhibitor olaparib and had a response that was ongoing 2 months after starting treatment⁴⁸. These data suggest that germline predisposition can guide therapeutic decisions in patients with SCLC, especially in selecting inhibitors targeting the DDR pathways, potentially with favourable outcomes. Germline testing can help in identifying additional treatment options and thus, should be encouraged in patients with SCLC, particularly in never-smokers or light smokers. Furthermore, a positive result could lead to cascade testing and influence screening recommendations in first-degree family members.

Molecular subtypes in SCLC

Transcription factor-defined subtypes

The first sign indicative of discrete SCLC subtypes came from the observation of morphological differences among SCLC cell lines in the 1980s: SCLC cells from the ‘classic’ subtype grow predominantly as spherical aggregates of floating cells with or without central necrosis, whereas those from the ‘variant’ subtype grow either as loosely adherent aggregates or as a more tightly adherent monolayer⁵⁰. These findings ignited multiple omics analyses that further defined the currently accepted four distinct molecular subtypes of SCLC. These integrated approaches resulted in the identification of a framework that differentiates biologically distinct SCLC subtypes based on the expression levels of four transcription factors: ASCL1 (SCLC-A subtype), NEUROD1 (SCLC-N), POU2F3 (SCLC-P) and YAP1 (SCLC-Y)⁵¹ (Fig. 1). The A and N subtypes are generally characterized by higher expression levels of neuroendocrine markers, whereas the P and Y subtypes have lesser neuroendocrine differentiation. In a real-world multiomic assessment of 437 SCLCs that were predominantly metastatic, 35.7%, 17.6%, 6.4%, 21.1% and 19.2% of samples were from the A, N, P and Y subgroups, or mixed, respectively⁵². Efforts directed at testing feasibility and validation of the four transcription factor-defined subtypes using immunohistochemistry (IHC) in primary SCLC tumours led to their identification in some studies, with substantial intratumoural heterogeneity, again suggesting the inherent plasticity of SCLC⁵³ (Fig. 2). In the past few years, investigators have presented an additional perspective about the YAP1-overexpressing subtype^54,55, although further IHC analysis did not confirm an exclusive YAP1 subtype and rather found YAP1 expressed at low levels primarily in combined SCLC⁵⁶. Indeed, whether YAP1 is an actual truncal driver of the SCLC-Y subtype has been a topic of debate because subsequent studies of SCLCs and circulating tumour cell (CTC)-derived xenografts (CDXs) could not define a subgroup based on YAP1 expression alone^57,58. Other effectors from the Hippo signalling pathway, such as TAZ, might be drivers of this subtype⁵⁹. The role of other effectors from the Hippo signalling pathway, such as TAZ, in this subtype remains to be established.

Fig. 2 ∣. Heterogeneity, plasticity and evolution of SCLC.

a, Histological transformation of lung adenocarcinoma (LUAD) to small-cell lung cancer (SCLC) as a mechanism of acquired resistance to targeted therapy. LUADs harbouring a loss-of-function mutation in TP53 and/or RB1 can transform to a more aggressive neuroendocrine (NE) subtype of lung cancer resembling SCLC. This dynamic and complex process was first identified in patients with EGFR-mutant LUADs receiving tyrosine-kinase inhibitors (TKIs) and is accompanied by transcriptional and epigenetic reprogramming. b, Heterogeneity, plasticity and evolution of de novo SCLC. Studies from the past few years have demonstrated that SCLCs transform from one subtype to another (that is, from ASCL1-driven to NEUROD1-driven and YAP1-driven subtypes)⁵⁴. MYC amplification and alterations affecting the Notch signalling pathway have a major role in subtype switching, transforming these tumours from NE to non-NE phenotypes. NE phenotypes tend to be less immunogenic and often tagged as immune ‘cold’, whereas YAP1-high SCLCs have better antigen presentation and response to immune checkpoint inhibitors.

In a US National Cancer Institute (NCI)-driven initiative, an unsupervised analysis of 118 cell lines derived from patients with SCLC revealed the existence of two distinct clusters based on sensitivity to various molecularly targeted agents⁶⁰. Cells from the A, N and Y subtypes were distributed in both clusters, whereas those from the P subtype tended to be sensitive to treatment. Further analysis revealed that ribosomal and EIF2-dependent signalling pathways were selectively activated in the sensitive cell lines. The screening further established that SCLC-A-derived cells are selectively sensitive to the BCL-2 inhibitor ABT-737, whereas SCLC-P-derived or SCLC-Y-derived cell lines are more sensitive to the multitarget tyrosine kinase inhibitor (TKI) inhibitor dasatinib and the BCR–ABL1 inhibitor ponatinib. Another NCI-driven screen revealed differential sensitivities to 103 agents approved by the FDA and 423 investigational agents across 63 cell lines derived from patients with SCLC⁶¹. Interestingly, in this study sensitivity or resistance to etoposide was correlated with sensitivity or resistance to most of the other drugs⁶¹, indicating common resistance mechanisms. Etoposide-sensitive cell lines often expressed high levels of miR-200c-3p and low levels of miR-140-5p and miR-9-5p. BCL-2/BCL-X_L inhibitors resulted in similar microRNA expression patterns, whereas agents targeting nuclear proteins regulating mitosis (such as inhibitors of polo-like kinase and aurora kinases) showed a different pattern in terms of microRNA signatures⁶¹. Finally, work from our group and others has shown that the SCLC-Y subtype is characterized by high expression of signatures related to a T cell-inflamed phenotype and the STING pathway^52,55. Together, these findings provide preclinical evidence of the potential link between transcriptional subtypes or biomarkers (such as certain microRNAs) and therapeutic vulnerabilities in SCLC.

Neuroendocrine-defined subtypes

Preclinical studies using the RPM GEMM (combining the genetic alterations from RP mice with oncogenic Myc) have demonstrated that the transition from neuroendocrine to non-neuroendocrine SCLC is linked to overexpression of MYC in the setting of intact Notch signalling⁵⁴. Also in RPM mice, the dynamic transcriptional evolution of SCLC subtypes from A to Y is driven by MYC in a conserved trajectory, with the A subtype being inversely correlated with MYC overexpression⁵⁶. Activation of MAPK signalling also contributes to the transition from a high to low neuroendocrine phenotype in SCLC⁶². A similar attempt to profile canonical SCLC subtypes and create an atlas using single-cell RNA-sequencing (scRNA-seq) data from 21 tumours revealed the presence of substantial intratumoural heterogeneity⁶³. These data again indicate that SCLC subtypes have inherent plasticity (Fig. 2) and the canonical subtypes can leverage diverse transcriptional programmes to generate tumour heterogeneity. Thus, understanding the factors that facilitate SCLC subtype transitions will be crucial to identify therapeutic vulnerabilities.

Distinct biological features further characterize SCLCs with high versus low neuroendocrine differentiation. At the transcriptomic level, neuroendocrine differentiation of SCLC has been described using approaches including several emerging computational tools for scRNA-seq, suggesting changes in plasticity from one phenotype to another^57,63–65 (Fig. 2). High-neuroendocrine SCLC is characterized by the upregulation of genes involved in the cell cycle and DDR. By contrast, low-neuroendocrine tumours are associated with high immunogenicity, increased cell adhesions, activation of epithelial-to-mesenchymal transition (EMT) programmes and alterations in metabolism, including auxotrophy of arginine-regulated pathways^19,65–68. Analysis of cell death pathways in SCLC revealed that high-neuroendocrine tumours acquire addiction to the thioredoxin antioxidant pathway, whereas low-neuroendocrine SCLC is vulnerable to ferroptosis, suggesting potential therapeutic approaches for targeting these subtypes⁶⁹.

The diverse mechanisms associated with neuroendocrine differentiation might predict distinct therapeutic vulnerabilities and guide novel treatment approaches. Owing to upregulation of DDR pathways, high-neuroendocrine SCLC might be vulnerable to DNA-damaging agents (such as platinum agents and topoisomerase inhibitors), DDR inhibitors (for example, PARP inhibitors) and inhibitors of cell cycle checkpoints (such as CHK1 and ATR). Exploratory analyses of samples from patients with SCLC involved in a phase II trial testing topotecan and the ATR inhibitor berzosertib demonstrated that all patients with an objective response had high-neuroendocrine SCLCs with upregulation of pathways associated with replication stress⁶⁷. Conversely, those with low-neuroendocrine SCLCs could potentially benefit from ICIs.

SCLC subtypes and response to ICIs

To better understand the predictive and prognostic value of the transcription factor-defined subtypes in the context of treatment with ICIs, researchers analysed tumour samples from patients involved in IMpower133. This randomized phase III trial tested the efficacy of adding the anti-PD-L1 antibody atezolizumab to conventional platinum-based chemotherapy in patients with treatment-naive ES-SCLC. Using gene expression data from >200 participants, four subgroups were identified: three of them were consistent with the previously described A, N and P subtypes, and the other was characterized by high expression of genes associated with inflammatory and immunogenic signalling, and referred to as the SCLC-I subtype (Fig. 1b).

Independent studies^55,57 have suggested that SCLC-I or YAP-high subtype has higher expression of HLA genes and genes encoding immune checkpoints. In this subtype, the expression of a set of 18 IFNγ-related genes in T cells, which can predict responsiveness to ICIs independent of TMB, was upregulated along with that of CD274 and PDCD1 (which encode PD-L1 and PD-1, respectively), CTLA4 and the genes that encode its ligands (CD86 and CD80), and TIGIT, ICOS, LAG3 and genes encoding cytokines and chemokines, such as CCL5 and CXCL10. Interestingly, increases in the number of natural killer (NK) cells, tumour-associated macrophages (TAMs) and T cells, including CD8⁺ T cells, were also observed, indicating better T cell infiltration and response to ICIs relative to the other subtypes. Indeed, retrospective analysis has shown that patients with SCLC-I derived a greater clinical benefit from atezolizumab (with a hazard ratio for OS of 0.57 (95% CI 0.321–0.998) relative to the other subtypes)⁵⁷. Moreover, RNA expression profiling of a subset of tumours from patients involved in IMpower133 indicated that the percentage of long-term survivors defined as patients who survived for ≥18 months after randomization, was similar in the atezolizumab and placebo arms (28% versus 32%). However, within the SCLC-I subset the percentage of patients who were long-term survivors was higher in the atezolizumab group (55% versus 30% in the placebo group)⁷⁰.

Another exploratory study, this one involving patients with RR SCLC who received durvalumab plus olaparib in a phase II trial, demonstrated that those with low-neuroendocrine tumours derive greater clinical benefit from ICIs than those with high-neuroendocrine tumours in the context of intact Notch signalling (that is, activated pathway promoting neuroendocrine differentiation)⁷¹. Moreover, bulk RNA-seq analyses of samples from patients with SCLC involved in IMpower133 revealed the presence of a neuroendocrine subset and a non-neuroendocrine subset (with different levels of expression of POU2F3) in the SCLC-I subtype. Differential expression experiments demonstrated that non-neuroendocrine SCLC-I is enriched in immunosuppressive TAMs relative to neuroendocrine SCLC-I despite having similar levels of effector T cells⁷². This study identified that, unlike patients with non-neuroendocrine SCLC-I, those with neuroendocrine SCLC-I can derive benefit from atezolizumab and also have higher levels of delta-like ligand 3 (DLL3, a repressor of Notch signalling highly expressed in SCLC and with lower expression in non-malignant tissues)⁷³, suggesting that they could derive benefit from DLL3-targeted T cell engagers. Together, these data highlight that novel immunotherapy combinations might have greater activity in patients with SCLC-I given the unique TME and proposed presence of distinct subsets with immune-inflamed phenotype⁷⁰. While these associative observations are biologically intriguing, the inclusion of the SCLC-I subtype as a predictive biomarker will require further clinical investigation and validation in independent datasets. Moreover, since the SCLC-I subtype cannot be assessed by IHC, the uniform adaptation of this subtype as a clinically relevant biomarker remains challenging.

SCLC subtypes and response to targeted therapies

Transcription factor-defined subgroups of SCLC are characterized by distinct biological features that might guide treatment with molecularly targeted agents. Promising agents, including T cell engagers targeting DLL3, are currently being tested in clinical trials (Supplementary Table 1).

DLL3-targeting agents.

The SCLC-A subtype has high expression of MYCL and DLL3 (ref. 73). The antibody–drug conjugate (ADC) rovalpituzumab tesirine was developed based on these findings; however, a randomized phase III study failed to demonstrate an OS benefit with rovalpituzumab tesirine over topotecan in patients with RR SCLC⁷⁴.

In a preclinical study, a DLL3-targeted IL-18-secreting chimeric antigen receptor T cell product showed antitumour activity by activating both genetically engineered and endogenous T cells, as well as by reprogramming the TME by increasing PD-L1 expression on macrophages and dendritic cells⁷⁵.

T cell engagers are multispecific antibodies that simultaneously bind to a tumour-associated antigen and the CD3 complex (expressed on T cells) and thereby not only target the cell surface protein on cancer cells but also engage cytotoxic T cells independently of MHC class I⁷⁶. Tarlatamab is a DLL3-targeted bispecific T cell engager that can induce T cell proliferation and activation and cytokine production, thereby inducing release of performin and granzyme B. Tarlatamab caused effective tumour regression in a mouse model driven by orthotopic SHP-77 SCLC cells and cleared liver metastases induced by H82 SCLC cells in another mouse model⁷⁷. Encouragingly, six clinical trials, five of which involve patients with SCLC, are currently evaluating the safety and efficacy of tarlatamab. A phase I trial with results published in 2023 tested tarlatamab in patients with RR SCLC who had received a median of two prior lines of therapy⁷⁸. The ORR was 23.4% and the median duration of response was 12.3 months, suggesting this agent to be a promising option with an acceptable safety profile⁷⁸. The DeLLphi-301 trial evaluated two doses of tarlatamab (10 mg and 100 mg) in patients with SCLC who had received a median of two prior lines of therapy⁷⁹. The ORR was 40% and 32% in the 10 mg and 100 mg cohort, respectively. The median PFS was 4.9 months and 3.9 months, respectively. Based on these promising results, in May 2024 the FDA granted Accelerated Approval to tarlatamab (10 mg) for patients with ES-SCLC and disease progression on or after platinum-based chemotherapy⁸⁰. Importantly, the frequency of grade ≥3 adverse events was ≥50% in both cohorts of DeLLPhi-301, including grade five events in 3% and 6% of patients receiving 10 mg and 100 mg, respectively, suggesting that developing more tolerable treatments in the second line of treatment and beyond is an area of unmet need. Other DLL3-targeted bispecific T cell engagers, such as BI-764532 and HPN328, enhance CD4⁺ and CD8⁺ T cell activity against DLL3-expressing SCLCs in mice^81,82.

BCL-2 inhibitors.

Proteins from the BCL-2 family regulate cell apoptosis and have been successfully targeted in patients with cancer, especially haematological malignancies, with several BCL-2 inhibitors approved worldwide⁸³. BCL-2 is a direct transcriptional target of ASCL1 and is, therefore, overexpressed in SCLC-A. In addition, anti-apoptotic BCL-2 family members, such as BCL-X_L, are usually overexpressed in SCLC and thus, might also be actionable targets in this malignancy. In preclinical models of SCLC, combined inhibition of either BCL-2 or BCL-X_L and MCL1 (a protein promoting cell survival that is upregulated as a mechanism of acquired resistance to BCL-2/BCL-X_L inhibition) had promising synergistic activity^84,85. However, similar to many other therapeutic approaches for SCLC, despite robust preclinical activity, BCL-2 inhibition has failed to improve upon the clinical efficacy of standard-of-care chemotherapy^86,87, raising the question of how to best translate these preclinical findings in the context of SCLC.

Targeted approaches for SCLC-N.

The SCLC-N subtype is associated with overexpression of MYC. Although approaches directly targeting MYC have not been evaluated in patients with SCLC, patients with SCLC-N might benefit from the previously discussed therapeutic approaches proposed for any SCLCs overexpressing MYC, such as CHK1 inhibitors⁴⁰, Aurora A kinase inhibitors⁴² and pegylated arginine deiminase⁸⁸. Notably, several direct inhibitors of MYC have been developed and tested preclinically^89,90. Finally, an ongoing trial is testing MRT-2359, a potent orally bioavailable degrader of GSPT1 that indirectly targets MYC⁹¹, in patients with SCLC harbouring MYCN or MYCL alterations (NCT05546268). MYC driven tumours are more reliant on protein translation. MRT-2359 targets GSPT1, a translational termination factor, by promoting complex formation between GSPT1 and CRBN (an E3 ligase), allowing degradation of GSPT1. This leads to preferential antiproliferative activity against MYC-driven lung cancers.

Targeted approaches for SCLC-P.

The transcription factor POU2F3 is typically expressed in tuft cells, a rare chemosensory cell type in the pulmonary epithelium. A molecular profiling study with results published in 2024 revealed that the SCLC-P subtype has a characteristic global hypomethylation phenotype⁹². In a preclinical study, SCLC cells overexpressing POU2F3 were dependent on the lineage transcription factors SOX9 and ASCL2, as well as IGF1R, indicating potential vulnerability to TKIs targeting this receptor⁹³. In a subgroup analysis of patients involved in IMpower133, those with tumours of the SCLC-P subtype had the least favourable OS⁵⁷. Additional analysis of drug–response data from other trials identified that SCLC-P cells are sensitive to PARP inhibitors⁵⁷. Although neither IGF-1R inhibitors^94,95 nor PARP inhibitors^96,97 have demonstrated clinical benefit in unselected patients with SCLC, testing these agents in those with SCLC-P might be worthwhile.

Targeted approaches for SCLC-Y.

The SCLC-Y subtype was initially identified by transcriptomic subgrouping of SCLC cell lines from the Cancer Cell Line Encyclopedia (CCLE) and is characterized by lower expression of INSM1 and reciprocal high expression of YAP1, a key mediator of the Hippo signalling pathway. These SCLC-Y cell lines were resistant to irinotecan and BCL-2 inhibitors. In addition, RB1 expression was intact in these cell lines, implying potential sensitivity to inhibition of cyclin-dependent kinase (CDK) 4 and 6 (ref. 98). Another preclinical study showed that cells from the SCLC-Y subtype are sensitive to mTOR and PLK inhibitors⁹⁹. Several preclinical and clinical studies have found that SCLC-Y is enriched in gene signatures related to cytotoxic T cells, NK cells and stimulators of interferon-dependent signalling, suggesting potential benefit from ICIs in this subtype^52,55.

Other transcription factor-defined subtypes

Several studies have suggested the existence of other transcription factor-defined subtypes of SCLC. Gene expression analysis of cells from CDXs revealed that a subset of SCLCs is defined by high expression of ATOH1 and other genes associated with hair cell differentiation in the mammalian auditory epithelium⁵⁸. This subtype was also found in clinical tumours, although it is rarer (~1%) than other transcription-factor-defined subtypes^58,68. Bioinformatic analysis of transcriptomic data of SCLC cell lines from the CCLE also defined a new SCLC subtype characterized by high expression of ASCL1, HES1 (a downstream factor of Notch signalling) and genes associated with drug metabolism pathways (such as xenobiotic and drug transporter pathways). This subtype was named Nev2 or SCLC-A2, and its characterization will be important given its lower sensitivity to all tested drugs relative to other SCLC cell lines¹⁰⁰.

Epigenetic landscape

Clinically actionable genomic alterations are rare in SCLC. By contrast, genetic alterations in histone-modifying genes, such as those encoding chromatin modifiers (CREBBP and EP300) and histone methyltransferases (KMT2A and KMT2D), are a hallmark of this cancer type¹⁰¹. Epigenetic changes regulate most of the transcriptional heterogeneity not attributable to genomic alterations and, thus, targeting these changes is an emerging therapeutic approach in SCLC. In fact, analyses of DNA methylation patterns in tumour-derived samples and circulating tumour DNA (ctDNA) have been identified as an approach to classify SCLC subtypes as well as to understand phenotypic evolution during disease progression. In a study with results published in 2024, features consistently detected in SCLC suggested a role for the TME in shaping global methylation patterns⁹².

Genome-wide mapping of enhancers has emerged as a powerful tool to characterize relevant transcription factors. Cellular identity and function are governed primarily by a group of transcription factors referred to as ‘master’ transcription factors, which bind to active enhancers and preferentially occupy large enhancer domains termed super-enhancers that regulate genes required for establishing cell identity and fate^102,103. Identifying super-enhancers has been an important avenue in epigenomic research because they could be therapeutic targets. In a study integrating super-enhancer profiling, transcription factor chromatin immunoprecipitation sequencing and proteomics, the transcription factors ASCL1, NKX2-1 and PROX1 were found to be associated with super-enhancers in the SCLC-A subtype¹⁰⁴. FOXA2, an essential regulator of neuroendocrine identity, was found to be a downstream target of the ASCL1–Nkx-2.1–PROX1 transcription factor network. Analysis of cell line data from the DepMap database of cancer dependencies¹⁰⁵ showed that ASCL1 and FOXA2 are SCLC-A-specific dependencies¹⁰⁴. As more studies map the epigenome of the SCLC subtypes, further therapeutic targets of interest could be identified.

EZH2, the catalytic subunit of the polycomb repressive complex 2 (PRC2) that promotes trimethylation of histone H3 lysine 27 (H3K27me3) and, as a consequence, cell self-renewal, is an epigenetic target that has attracted immense interest in SCLC. EZH2 has pleiotropic functions that contribute to immune evasion and suppression of intratumoural antigen presentation¹⁰⁶. Thus, EZH2 hyperactivation enables SCLC to evade recognition and targeting by innate and adaptive immunity. Pharmacological inhibition of EZH2 in a SCLC patient-derived xenograft (PDX) model led to growth inhibition ex vivo and in vivo, providing preclinical evidence for EZH2 as a therapeutic vulnerability¹⁰⁷. The antigen-presentation machinery is a core element in the adaptive immune system recognition of cancer cells. NK cells leverage the ligand NKG2DL to target cancer cells. In terms of innate immunity, the chemokine CCL2 recruits monocytes that eventually differentiate into macrophages. Hyperactivity of PRC2 through EZH2 leads to transcriptional silencing of HLA, KLRK1 (which encodes NKG2DL) and CCL2 (ref. 108). This equips SCLCs with a compromised antigen-presentation machinery, leading to defective NK cell cytotoxicity and disrupting macrophage recruitment. In preclinical models, pharmacological inhibition of EZH2 restores T cell-mediated killing of cancer cells and increases the expression of MHC class I^109,110. This observation suggests that in clinical settings, pharmacological inhibition of EZH2 could improve immunogenicity and potentially augment responses to ICIs in patients with SCLC. No such trials are underway, although a phase I/II trial investigating the combination of the EZH1/2 inhibitor valemetostat and irinotecan in patients with recurrent SCLC has been completed and results are pending (NCT03879798). Preclinical studies also showed that inhibition of EZH2 promotes loss of a neuroendocrine phenotype, resulting in upregulation of SLFN11, an inducer of lethal replication blockade in response to DNA-damaging agents^109,110 (Fig. 2).

LSD1 has been proposed as another epigenetic target in SCLC. This enzyme demethylates the monomethylated and dimethylated forms of histones¹¹¹. Investigators demonstrated that inhibition of LSD1 activates Notch signalling, resulting in suppression of ASCL1 and tumour growth in a GEMM model of SCLC¹¹². Although LSD1 inhibition has shown promising preclinical results¹¹³, a phase I trial testing the LSD1 inhibitor GSK2879552 failed to show tolerability and meaningful clinical activity in patients with RR SCLC¹¹⁴ (Supplementary Table 1). In preclinical models, combination of LSD1 pharmacological inhibitors with anti-PD-1 antibodies promotes antigen presentation via upregulation of MHC I and enhances CD8⁺ T cell activity, which in turn provides a strong rationale for ICI-based combination approaches in patients with non-neuroendocrine SCLC¹¹⁵.

Preclinical studies also showed that inhibition of EZH2 promotes loss of a neuroendocrine phenotype, resulting in upregulation of SLFN11, an inducer of lethal replication blockade in response to DNA-damaging agents^109,110 (Fig. 2). ASXL3 is key for maintaining the viability of pluripotent respiratory epithelial cells and human SCLC cells in vitro and in vivo¹¹⁶. Another preclinical study showed that the lineage-specific transcription factor PAX9 interacts with the nucleosome remodelling and deacetylase (NuRD) complex and functions as a transcriptional repressor that inhibits enhancer activity, leading to repression of genes involved in neural differentiation and tumour-suppressor genes¹¹⁷. Taken together, these observations suggest that targeting epigenetic modifications modulates lineage plasticity and can reprogramme cancer stem cells to a more differentiated or non-malignant-like phenotype¹¹⁸.

Other than identifying cellular dependencies, preclinical research over the past few years has shown that epigenetic modifications lead to transcriptional reprogramming and alter cellular plasticity in SCLC. In response to EGFR TKIs, EGFR-driven LUADs can retain molecular drivers, acquire an EMT phenotype and transform to SCLC¹¹⁹. This process is thought to be mediated by epigenetic changes, although the exact mechanisms remain to be elucidated. Moreover, SCLCs can leverage lineage plasticity and transition from a neuroendocrine state to a non-neuroendocrine state via epigenetic reprogramming (Fig. 2). Epigenetic changes of cells grown in culture poorly recapitulate the epigenetic landscape of primary human tissue^120,121. Fortunately, multiple GEMMs and CDX models¹²² are available that enable more accurate mapping of the SCLC epigenome. Future research studies should aim to apply epigenetic advances in human SCLC primary tissue to further identify therapeutic targets.

Metabolomic landscape

Among its multiple roles, MYC functions as a master regulator of several metabolic pathways, including glycolysis, glutamine metabolism and nucleotide biosynthesis¹²³. Metabolomic profiling of a panel of 29 SCLC-derived cell lines and 47 primary SCLCs revealed that several purine nucleotides and purine synthesis genes, such as IMPDH1 and IMPDH2, are substantially elevated in ASCL1^lowMYC^high SCLC¹²⁴. A clinical trial explored the efficacy of gemcitabine, an agent that represses nucleotide metabolism, in patients with SCLC resistant to chemotherapy, with an ORR of 13%¹²⁵. In a preclinical study, MYC-overexpressing SCLC-derived cell lines and patient samples had increased expression of genes related to glycolysis; using in vivo and in vitro models, researchers identified that glycolysis in MYC^high cells can be targeted by inhibiting the enzyme PFKFB3 (ref. 126). Steadystate metabolomic analyses of RPM mice demonstrated that tumours from this SCLC model are dependent on arginine-regulated pathways, including polyamine biosynthesis and the mTOR pathway¹²⁷. Arginine depletion with pegylated arginine deiminase suppressed tumour growth and promoted the survival in MYC-overexpressing GEMM and PDX models⁸⁸.

Nuclear magnetic resonance-based analyses of blood samples from a small cohort of patients with newly diagnosed SCLC (n = 30) identified a shift in amino acid, energy and lipid metabolism pathways relative to healthy individuals (n = 25)¹²⁸. Notably, patients with SCLC had substantially higher glutamic acid and lower glutamine levels. Additionally, patients with SCLCs had high levels of 3-hydroxybutyric acid, acetoacetic acid, acetone, glycerol and LDL triglycerides, and lower levels of LDL4 (ref. 128). This is a promising area of research, although the clinical relevance of these findings remains to be established.

Biology of refractory SCLC

Chemotherapy-refractory SCLC progresses rapidly and, as a result, patients typically undergo a rapid clinical deterioration. Despite years of research, the mechanisms underpinning therapeutic resistance remain elusive, mainly as a result of the limited tissue availability after treatment. In the past decade, studies using PDX and CDX models have shed light on the biology of refractory SCLC.

WES and transcriptome sequencing of five clinical samples collected after relapse during autopsy identified multiple significantly mutated genes, including LRP1B, RYR2, USH2A and TP53 (ref. 129). Analysis of CNVs revealed monoallelic loss of regions in chromosome 17p (containing TP53), chromosome 5q (containing APC) and chromosome 13q (containing RB1) and allele-specific gains of regions of chromosome 3q (containing SOX2) and 8q (containing MYC) in 60% of patients¹²⁹. In another study, WES of SCLCs showed decreased expression of ASCL1 in samples obtained after relapse from patients on platinum-based doublet therapy compared with those from treatment-naive patients²⁹. These samples also had somatic CNVs, including gains in ABCC1 (encoding a glycoprotein associated with the multidrug-resistant phenotype) and deletions in MYCL, MSH2 and MSH6 (with the latter two being involved in DDR)²⁹. Furthermore, RNA-seq analysis revealed that in ASCL1^low chemosensitive SCLC cell lines, activation of WNT signalling induces chemoresistance²⁹. A study in PDX models showed that EZH2-mediated transcriptional silencing of SLFN11 promotes acquired chemoresistance and, thus, inhibition of EZH2 could re-sensitize these tumours to chemotherapy¹³⁰. Another study in PDX models showed that REST-mediated activation of Notch signalling can promote neuroendocrine to non-neuroendocrine transition in SCLC, increasing intratumour heterogeneity in 10–50% of PDXs¹³¹. Of note, REST is considered a non-neuroendocrine marker in SCLC. This transition promoted chemoresistance in preclinical models, in which co-administration of Notch inhibitors with chemotherapy led to tumour regression and delayed relapse, making Notch signalling a potential therapeutic target in SCLC¹³¹.

Intratumour heterogeneity and lineage plasticity

Lineage plasticity is the ability of a cell to modify its identity to resemble a distinct lineage phenotypically different from its original form. Multiple research groups have reported the transformation of EGFR-mutant LUAD into SCLC as a mechanism of resistance to EGFR TKIs¹⁹. A genomic study of 3,600 real-world SCLCs confirmed the presence of EGFR mutations in 107 patients and revealed alterations in other NSCLC driver genes such as ALK (n = 5), RET (n = 5), ROS1 (n = 3) and NTRK1 (n = 1), indicating additional potential mechanisms of transformation²⁸.

Analysis of samples derived from patients with SCLC showed a differential transcriptome profile between those believed to arise de novo and those believed to arise via transformation from LUAD. Detailed analysis identified that de novo SCLC has enhanced expression of genes involved in the PRC2 complex, AKT–mTOR pathway and inhibition of Notch signalling, all of which favour the neuroendocrine phenotype¹⁹. Single-cell profiling of 54,523 SCLC cells derived from patients showed that cancer cells can be in an intermediate state between SCLC-A and SCLC-N, providing further proof of the heterogeneity and plasticity in SCLC⁶³ (Fig. 2). Single-cell analysis of preclinical SCLC models described MYC as a crucial driver of plasticity between SCLC subtypes. In these models, MYC induced activation of Notch signalling and promoted a shift between ASCL1⁺ to NEUROD1⁺ to YAP1⁺ states, providing further insight into the heterogeneity and biological complexity of SCLC⁵⁴ (Fig. 2). Although intratumoural plasticity in SCLC has only been proposed in the past decade, it has been demonstrated as a mechanism of therapeutic resistance. Activation of Notch signalling can promote a transition from a neuroendocrine to a non-neuroendocrine phenotype, making cells more chemoresistant^54,132. Activation of DLL3 has been shown to promote HES1-mediated inhibition of ASCL1, increasing intratumour heterogeneity¹³³. These data imply the complexity of the transitional states of SCLC subtypes that can occur as a continuum across time and space, making SCLC a therapeutically challenging tumour type to target owing to the multifaceted nature of resistance mechanisms.

Opportunities and challenges with ICIs

The immunotherapy strategies currently available are primarily aimed at increasing T cell-mediated cytotoxicity against cancer cells, which can be achieved using cancer vaccines, adoptive transfer of antigen-specific T cells, ICIs targeting CTLA4, PD-1 or PD-L1, among other approaches^134,135. Assessing PD-L1 expression in clinical settings is challenging, although studies in SCLC have revealed that only ~1.9% these tumours have amplification of CD274 (ref. 136). Depending on the PD-L1 antibody used for IHC, studies have found a wide distribution (1–70%) of PD-L1 positivity in metastatic and extensive-stage disease^137,138. A strong correlation between PD-L1 levels and clinical benefit in patients with ES-SCLC has not been consistently demonstrated across trials of ICIs^139,140. Similarly, defining a therapeutically relevant cut-off for TMB, which is generally high in SCLC owing to the clonally high mutational load secondary to the mutagenic effects of smoking, has been another challenge. Although some previous studies have demonstrated that patients with SCLCs in the highest quartile for tissue TMB tend to have the most favourable response to ICIs²³, similar studies have demonstrated that the benefit of chemotherapy–ICI combination regimens in ES-SCLC, such as durvalumab–tremelimumab with chemotherapy^140,141 or pembrolizumab–chemotherapy¹⁴², is somewhat inconsistent in tumours with a TMB cut-off, for reasons that remain to be elucidated.

In a randomized phase II trial testing the concurrent and sequential addition of the anti-CTLA4 antibody ipilimumab to carboplatin–paclitaxel, those receiving sequential administration of the ICI had a marginal yet statistically insignificant improvement in PFS and OS¹⁴³. Similarly, in a randomized phase III trial, the addition of ipilimumab to etoposide plus either cisplatin or carboplatin also failed to show a significant improvement in OS¹⁴⁴.

Although approaches using anti-PD-1 antibody-based therapies have demonstrated antitumour efficacy and led to newer FDA approvals in ES-SCLC, these have not necessarily translated into long-term clinically meaningful improvements in median PFS or OS in a majority of patients with ES-SCLC. In the phase III IMpower133 trial, the addition of atezolizumab to carboplatin–etoposide was tested in patients with ES-SCLC (n = 403). The trial showed a statistically significant improvement in OS (12.3 months versus 10.3 months; HR 0.76, 95% CI 0.60–0.95; P = 0.015) and PFS (5.2 months versus 4.3 months; HR 0.77, 95% CI 0.62–0.96; P = 0.02)^5,22. In CASPIAN, another phase III trial, 537 patients were randomly assigned to receive the anti-PD-L1 antibody durvalumab plus chemotherapy versus chemotherapy alone. The addition of durvalumab improved median OS duration (12.9 months versus 10.5 months; HR 0.81, 95% CI 0.67–0.97; P = 0.02)^6,145. These results led to the FDA approval of atezolizumab and durvalumab in combination with platinum-based chemotherapy, which has become a standard-of-care first-line treatment option for ES-SCLC.

The efficacy of the anti-PD-1 antibody pembrolizumab as monotherapy was tested in the phase Ib KEYNOTE-028 trial, in which the ORR was 35%, with responses lasting for >16 weeks¹⁴⁶. These results failed to translate into a significant improvement in OS in the phase III KEYNOTE-604 trial testing the addition of pembrolizumab to chemotherapy¹⁴⁷. Another study with somewhat disappointing results is the phase III SKYSCRAPER-02 trial, which failed to demonstrate an OS benefit from the addition of the anti-TIGIT antibody tiragolumab to atezolizumab–chemotherapy¹⁴⁸. Thus, despite the encouraging success of ICIs in some solid tumour types, in SCLC meaningful clinical improvement seems limited to a small percentage of patients. In CASPIAN, only a subset of patients (17.6%) were alive at 36 months of follow-up indicating that, indeed, most patients do not derive meaningful benefit from ICIs⁶.

The ongoing phase III ADRIATIC trial is evaluating the role of consolidation durvalumab with or without the anti-CTLA4 antibody tremelimumab in LS-SCLC who did not have disease progression after completion of concurrent chemoradiotherapy¹⁴⁹. Results of the interim analysis of durvalumab (n = 264) versus placebo (n = 266) were presented in 2024 (ref. 194). Durvalumab significantly improved median OS (55.9 months versus 33.4 months with placebo; HR 0.73, 95% CI 0.57–0.93; P = 0.01) and median PFS (16.6 months versus 9.2 months; HR 0.76, 95% CI 0.61–0.95; P = 0.016)¹⁵⁰. This approach is an extrapolation of that tested in patients with NSCLC in the PACIFIC trial, which showed both a PFS and an OS benefit¹⁵¹. A trial performed in Asia is evaluating whether any practical difference in efficacy and toxicity exists between adding an anti-PD-1 antibody (serplulimab) and adding an anti-PD-L1 antibody (atezolizumab) to chemotherapy in previously untreated patients with ES-SCLC (NCT05468489). Over the past year, emerging data from trials testing novel approaches using ICIs in combinations with anti-VEGF agents for ES-SCLC have shown promising PFS and OS benefits that will hopefully result in major advances in the treatment of SCLC^152,153.

A major challenge to the efficacy of ICIs is the repression of the MHC I-dependent antigen presentation machinery, which is crucial for T cell engagement. Reports suggest that low-neuroendocrine SCLCs have higher expression of HLA relative to high-neuroendocrine SCLC and, thus, are associated with longer responses to ICIs¹⁰⁹. These data suggest that de-repression of HLA, a strategy that remains under investigation, might help to achieve durable responses to ICIs in SCLC.

The loss of cell cycle checkpoint controls owing to the inactivation of RB1 and TP53 increases the susceptibility of SCLC to DNA damage. Several studies have shown that targeting proteins involved in DDR, such as PARP, CHK1 (ref. 154) and WEE1 (refs. 155–157), leads to activation of the innate immune cGAS–STING pathway, increases in expression of chemokines (for example, CCL5 and CXCL10) and recruitment of cytotoxic T cells. Consequently, in preclinical models of SCLC, a combination of inhibitors of PARP, CHK1 and/or WEE1 augmented the effect of anti-PD-L1 antibodies^158,159. Several clinical trials are now testing these combinations^41,160 (NCT04538378, NCT04728230, NCT05718323, NCT05815160).

SCLC cells typically have elevated levels of CD47, a membrane protein that promotes immune evasion by interacting with an inhibitory macrophage receptor SIRPα, thus evading macrophage-mediated phagocytosis¹⁶¹. RRx-001, a small-molecule MYC inhibitor and downregulator of CD47 that facilitates repolarization of macrophages from M2 to M1 by blocking the SIRPα–CD47 interaction, is under clinical investigation in patients with SCLC¹⁶². Moreover, a study has provided evidence that administration of antibody targeting ganglioside GM2, a TAM marker overexpressed in SCLC, extends survival of severe combined immunodeficient mice by inhibiting multiorgan metastasis and promoting apoptosis¹⁶³. A study with results published in 2024 showed that CD38 expression is increased after chemotherapy and ICIs in SCLC preclinical models and clinical samples, suggesting that CD38 blockade could be of value as a combination strategy for chemo-immunotherapy in SCLC¹⁶⁴.

Retrospective transcriptional analysis of SCLC samples from patients enrolled in IMpower133 showed that those with tumours of the ‘inflamed’ subtype (SCLC-I) had improved responses to chemotherapy plus ICIs⁵⁷. However, these findings have not been universally validated in other trials of chemotherapy–ICI combinations. Moreover, prospective trials are required to confirm the clinical significance of this subtype. Investigators performed a retrospective analysis of samples from the phase I/II CheckMate 032 trial to identify predictors of response to ICIs. This trial assessed the efficacy of nivolumab alone or combined with ipilimumab in patients with advanced-stage solid tumours including SCLC¹⁶⁵. No statistically significant correlation was found between the four SCLC subtypes and OS or PFS in either treatment group. Interestingly, OS was improved in patients receiving ICIs with higher intratumoural infiltration of CD8⁺ T cells. Similarly, in the KEYNOTE-604 trial of pembrolizumab plus chemotherapy, SCLC subtypes were not associated with differential OS benefit but the presence of a T cell inflamed gene expression profile was associated with improved OS in both the experimental and standard treatment arms¹⁴². This finding underscores the need to further interrogate the TME in SCLC to better understand potential biomarkers of response to ICIs. Some emerging data indicate that the functional status of RB1 in SCLC could also serve as a predictive biomarker of ICI response. SCLCs with a low RB1 loss-of-function score or wild-type RB1, a status that seems to be common in non-neuroendocrine SCLCs, and those with high levels of YAP1, tend to derive more benefit from ICIs^98,166. These data suggest that RB1 mutational status in SCLC can affect antitumour immunity and, ultimately, response to ICIs.

Single-cell analyses

Thus far, analyses of SCLC preclinical models and clinical samples have primarily focused on bulk transcriptomic assessment. These studies have shed light on the immune microenvironment, metastatic ability, rare cell populations, intratumoural and intertumoural heterogeneity, and phenotypic plasticity. Improvements in scRNA-seq technologies over the past few years have provided further insight on these aspects.

In a study that defined a single-cell atlas of SCLC⁶³, researchers analysed the transcriptome of clinical samples of 21 SCLCs and 24 LUADs, as well as four tumour-adjacent non-malignant lung tissue samples. This unique cohort included both treated and untreated patients. As expected, the study found higher intertumoural heterogeneity in SCLC relative to LUAD. The SCLC samples included were from the A (n = 14), N (n = 6) and P (n = 1) subtypes. Interestingly, this study identified non-canonical phenotypes in SCLC, including a population of SCLC cells expressing both ASCL1 and NEUROD1, suggesting a mixed subtype or plasticity of SCLC subtypes. SCLC-N tumours were enriched in signatures associated with metastasis, stem cell-related features, neuronal differentiation and immune evasion.

In another study⁵⁴, time-series scRNA-seq of mouse-derived and human-derived SCLCs revealed that MYC drives the loss of neuroendocrine characteristics and enrichment of non-neuroendocrine signalling pathways. The investigators further showed that ASCL1 and NEUROD1-driven cell clusters are present in the early stages of carcinogenesis and largely absent at later stages⁵⁴. Pseudotime trajectory analysis, which has been widely used to study dynamic gene regulatory programmes along continuous biological processes, showed a temporal shift of SCLC subtypes from the A to N and then Y subtype. In-depth analyses indicated that MYC drives activation of Notch signalling to dedifferentiate SCLCs, promoting plasticity. A similar study¹⁶⁷ found a subtle difference in clustering between early and late stages of tumour development in preclinical in vivo models. Loss of neuroendocrine characteristics and suppression of cell cycle regulatory gene signatures were found in the later stage of tumour development. Interestingly, detailed analysis revealed that AP-1 transcription factors are upregulated at the later stages in most tumour cells. In another study¹⁶⁸, researchers used modified single-cell tagged reverse transcription sequencing to characterize the SCLC TME and heterogeneity. Analysing samples from a small cohort of patients with SCLC (n = 11), they found five neuroendocrine subtypes of cells with high expression of ASCL1 and variable expression of NEUROD1. This finding further strengthens the concept of mixed lineage phenotypes and plasticity. Moreover, this study is consistent with previous reports that showed the gradual decrease in MYC expression from a non-neuroendocrine to neuroendocrine SCLC phenotype. Single-cell analysis of SCLC PDXs and clinical samples¹⁶⁹ revealed a higher degree of intratumoural heterogeneity in chemoresistant models relative to those with sensitivity to chemotherapy. Chemoresistant models had higher expression of components of signalling pathways previously linked to chemoresistance, such as MYC, mTOR and WNT, and associated with EMT. The results confirmed the existence of multiple cisplatin resistance pathways, at the intratumoural, intrapatient and interpatient levels.

scRNA-seq analysis of SCLCs showed that SCLC-N has lower infiltration of CD8⁺ T cells, higher infiltration of regulatory T cells and preferentially an immune cold phenotype relative to the other subtypes⁶³. This study revealed that the N subtype correlated with exhausted CD8⁺ T cells and profibrotic myeloid cells. Characterization of cells from primary tumours (n = 3,365), tumour-adjacent non-malignant tissue (n = 1,274) and cells from RR tumours (n = 272) revealed decreased myeloid cell infiltration and an enriched lymphocyte fraction in primary tumour relative to adjacent non-malignant tissue¹⁶⁸. Interestingly, both non-malignant tissue and the TME had increased infiltration of CD8⁺ T cells expressing cytotoxic markers. Detailed classification revealed variation in the expression of T cell exhaustion markers and class II HLA genes among patients, indicating intertumoural heterogeneity in the TME. In conclusion, the advances in single-cell analyses of SCLC is providing valuable insight into the biology and complexity of these tumours, which could help to improve future clinical trial design.

Biomarker-driven clinical trials

Here we discuss studies from the past few years indicating the potential to perform prospective trials of novel agents in biomarker-selected cohorts of patients with SCLC. Although some of the approaches that use integral biomarker-guided therapies for patient selection continue to evolve, these trials are a step towards stratification of SCLC into several categories and delivering tailored therapies, similar to NSCLC (Fig. 3).

Fig. 3 ∣. Translational roadmap for SCLC.

Integration of bedside-to-bench-to-bedside platforms to improve therapy and biomarker discovery in small-cell lung cancer (SCLC). New patients are diagnosed using pathology and imaging, and biospecimens (tissue and blood) are collected. The collected samples are then used to derive preclinical models that can be used in ‘wet-lab’ experiments to understand the biology of SCLC and identify novel therapies and biomarkers. These experiments are complemented with cutting edge computational analyses and artificial intelligence and/or machine learning approaches for ‘n-of-1’ studies and treatment selection. Patients then receive new drugs, and the response in overall survival, progression-free survival and other outcomes is monitored. CDX, circulating tumour cell-derived xenograft; PDX, patient-derived xenograft.

A study found that patients with ES-SCLC harbouring ‘disruptive’ mutations in TP53 have better responses to chemotherapy, whereas those harbouring wild-type RB1 have poor responses¹⁷⁰. In another study, RB1-proficient SCLCs were found to be sensitive to CDK4/6 inhibition. Given that ~14% of SCLCs express RB1 protein and have high levels of YAP1 expression, investigating the role of YAP1 as a potential predictive biomarker of response to CDK4/6 inhibitors in a subset of patients with SCLC will be important¹⁷¹.

SLFN11 encodes a DNA/RNA helicase that destabilizes stalled replication forks and has been identified as a relevant predictive biomarker of sensitivity to DNA-damaging agents, especially PARP inhibitor monotherapy, in SCLC^172–174 in several retrospective analyses of clinical trials. In a randomized phase II trial, patients with ES-SCLC and SLFN11-positive status receiving temozolomide and the PARP inhibitor veliparib had a median OS of 12.2 months versus 7.5 months in those with SLFN11-negative disease (P = 0.014)⁹⁵. Subsequently, a prospective study that included patients with SLFN11-positive ES-SCLC who were randomly allocated to either maintenance atezolizumab (standard of care) versus atezolizumab in combination with talazoparib showed a marginal, yet statistically significant improvement in median PFS (2.8 months versus 4.2 months; HR 0.70, 80% CI 0.52–0.94)¹⁷⁵. The latter study mainly emphasizes the feasibility of prospectively using predictive biomarkers to select and stratify patients with SCLC.

DLL3 is another predictive biomarker with a high prevalence in patients with SCLC (~80% of samples from patients with SCLC)¹⁷⁶. In a first-in-human study of BI-764532, a novel DLL3-targeted T cell engager in patients with DLL3-positive SCLC or neuroendocrine carcinoma, the ORR was 26% and the durable response rate was 51% in those with SCLC¹⁷⁷. This trial provides another example of the potential to perform prospective studies in a molecularly selected population of patients with SCLC. HPN328 is a trispecific DLL3-targeting T cell engager with an albumin binding moiety for half-life extension that was tested in a phase I/IIa study involving patients with neuroendocrine malignancies, and was associated with an ORR of 25% in those with SCLC (n = 9)¹⁷⁸.

Transcriptomic analysis of xenograft models identified SEZ6 as an antigen commonly expressed on the surface of SCLC cells, resulting in the development of the anti-SEZ6 ADC ABBV-011 (ref. 179). In a phase I study evaluating ABBV-011 alone or in combination with the anti-PD-1 antibody budigalimab in patients with SCLC, monotherapy with ABBV-011 was associated with an ORR of 25% and a clinical benefit rate of 65%¹⁸⁰. Overexpression of somatostatin receptor 2 occurs in the majority of neuroendocrine neoplasms, including SCLC (48%), with a significant association with poorer outcomes. Somatostatin receptor 2 activates downstream MAPK and AKT signalling. In preclinical models, downregulation of somatostatin receptor has an antiproliferative role in tumours by inducing apoptosis¹⁸¹. In this regard, key additional novel targeted therapies that are currently being investigated in early phase trials including patients with SCLC include ²²⁵Ac-DOTATATE, a radiopharmaceutical tested in patients with SCLC overexpressing somatostatin receptor 2 (NCT05595460) (Supplementary Tables 1,2).

Tumour-associated calcium signal transducer 2 (commonly referred to as TROP2), a transmembrane glycoprotein, is another promising target in SCLC (Supplementary Table 2). High expression of TROP2 has been noted in 10% of patients with SCLC. Sacituzumab govitecan, a first-in-class TROP2-targeted ADC, consists of an anti-TROP2 monoclonal antibody linked to the topoisomerase 1 inhibitor SN-38. A phase I/II multicentre trial (NCT01631552) in a cohort including 62 patients with pretreated SCLC found an ORR of 18% and a median OS of 7.1 months. An ongoing phase II study evaluating sacituzumab govitecan in patients with metastatic or locally advanced solid tumours including SCLC found a preliminary ORR of 29% and clinical benefit rate of 36%¹⁸². Addition of ICIs to first-line standard chemotherapy for SCLC has been a remarkable step forward, although, owing to the high frequency of primary or acquired resistance to anti-PD-L1 antibodies, novel promising targets for immunotherapeutic approaches are being investigated, one of which is CD276 (commonly referred to as B7-H3) (Supplementary Table 2). Expression of B7-H3 has been found to be correlated with a reduction of activation of T cell-related and IFNγ-related signatures. Ifinatamab deruxtecan, an ADC comprising an anti-B7-H3 antibody conjugated to an exatecan derivative payload, has demonstrated antitumour activity in preclinical models of SCLC. In a phase I/II trial, confirmed responses were achieved by 52.4% of patients, including complete responses in 4.8%¹⁸³. The median PFS was 5.8 months and median OS was 12.2 months. Moreover, ifinatamab deruxtecan was deemed tolerable, with manageable toxicity. Grade ≥3 treatment-related adverse events were observed in 36.4% of patients¹⁸³.

Liquid biopsies

Technological advances have improved the detection, enumeration and characterization of somatic mutations in cell-free DNA (cfDNA) and CTCs. As a result, cfDNA next-generation sequencing (NGS) and CTC counts have become prognostic biomarkers in several solid tumour types, including SCLC^184,185. Within the context of SCLC, genomic evolution remains relatively under-studied, primarily owing to the challenges associated with obtaining adequate tumour specimens for analysis. This limitation is exacerbated by the aggressive nature of SCLC, which often leads to difficulties in obtaining timely and sufficient biopsy-derived samples, particularly during disease recurrence. Additionally, the inherent constraints of traditional tumour biopsy samples, characterized by their small size and often suboptimal quality, further underscore the necessity for alternative approaches such as liquid biopsies to better understand the molecular evolution and resistance mechanisms of SCLC, and select personalized treatment approaches for patients with this disease. As SCLC typically spreads haematogenously¹⁸⁶, cfDNA and CTC sampling are promising avenues to longitudinally characterize SCLC.

Despite a paucity of studies evaluating CTCs in patients with SCLC, the available emerging data have revealed promising insights. These data have shown a high prevalence of CTCs in patients with LS-SCLC and ES-SCLC, with their presence being significantly associated with inferior PFS and OS^185,187,188. Furthermore, changes in CTC numbers following chemotherapy are an independent prognostic factor, highlighting their potential utility as dynamic biomarkers for treatment response and disease monitoring¹⁸⁹.

In parallel, plasma cfDNA analysis using NGS has emerged as a powerful tool for profiling the genomic landscape of SCLC. Targeted sequencing of SCLC-associated genes has demonstrated the presence of non-synonymous variants in cfDNA in a vast majority of patients with SCLC¹⁹⁰ (94% in a cohort of 62 patients), offering insight into the mutational landscape and potential therapeutic vulnerabilities of the disease. Moreover, longitudinal cfDNA sampling can help to detect targetable alterations weeks prior to radiographic progression, as well as enrichment in variants of genes involved in DDR (such as BRCA1) and Notch signalling at relapse^191–195, underscoring the potential of cfDNA analysis for early detection of treatment resistance and disease relapse. Additional emerging data also suggest that ctDNA NGS provides a better snapshot of genomic heterogeneity of SCLC than tumour NGS in individual patients¹⁹².

Studies from the past few years have highlighted the utility of cfDNA methylation profiling as a complementary approach for SCLC subtyping and outcome prediction. This method is based on the principle that tumour cells acquire recurrent aberrant DNA methylation distinct from blood and other non-malignant tissue cells. By analysing the aberrant DNA methylation patterns present in cfDNA, researchers have been able to accurately classify SCLC subtypes, track the evolution of SCLC subtypes over time and correlate epigenetic signatures with clinical outcomes^91,196. In a study with results published in 2024, subtype switch detection in cfDNA seemed to be associated with promoter methylation of immune-related genes, highlighting insights that can be gained from the analyses of promoter methylation using liquid biopsies⁹¹.

Other than plasma-based liquid biopsies, non-blood bodily fluids present opportunities in SCLC. Pleural fluid is an ultrafiltrate of blood enriched with non-haematopoietic cfDNA owing to the lack of peripheral blood cells¹⁹⁷. Malignant pleural effusions include tumour cells that infiltrate into the pleura and thus are in close proximity to the tumour. Thus, cfDNA from pleural fluid might better correlate with tumour tissue than plasma cfDNA, as shown in the setting of NSCLC¹⁹⁸. In comparison to plasma cfDNA, pleural fluid cfDNA might enable more sensitive detection of SCLC. Another advantage of pleural fluid cfDNA sampling is that it overcomes the limitation of clonal haematopoiesis of indeterminate potential (CHIP)¹⁹⁹. CHIP can lead to false-positive results in cfDNA sampling, especially related to the fact that TP53 variants are common both in SCLC and CHIP, making it hard to discern the origin of TP53 variants without performing additional sequencing of peripheral blood mononuclear cells¹⁹⁷. Nonetheless, pleural fluid-based cfDNA sampling has the downside that aspiration is more invasive than blood sampling. Further research is needed to assess the feasibility and clinical utility of pleural fluid compared to plasma sampling methods.

Liquid biopsies are leading to changes in early detection and disease monitoring across many solid tumour types, although these assays might not yet be ready for prime time in SCLC. Comparisons across different methodologies (including plasma versus non-plasma and methylation versus genomic profiling) are essential to determine the optimal liquid biopsy assay. However, emerging data collectively show the robustness and sensitivity of liquid biopsy in monitoring, detecting and profiling SCLC subtypes²⁰⁰, which will help to guide patient stratification and treatment personalization. Validation and implementation of liquid biopsies in clinical trials and prospective studies of patients with SCLC with the ultimate goal of improving patient outcomes in this aggressive malignancy is warranted.

Future directions

Over the past several decades, SCLC research has produced only a few selected treatment approaches that improve patient outcomes in a clinically meaningful way. On the basis of the rapid advances in multi-omic approaches and promising emerging data in SCLC, characterizing the unique molecular subtypes at the genomic, transcriptomic and epigenetic levels is key to better understand therapeutic vulnerabilities and implications for treatment resistance.

As SCLC is a highly aggressive disease, it is often detected in an advanced or metastatic stage and has a dismal prognosis. In addition, many emerging clinical trials of novel therapeutic agents, including ICIs, have yet to be successful in showing durable and meaningful improvements in OS. This challenge is further complicated by the lack of integral biomarkers to select patients for standard-of-care precision medicine approaches. In contrast to NSCLC, thus far the genomic characterization of human SCLC cell lines and clinical tumour samples has not yet identified targetable genomic alterations that offer drastically improved patient outcomes^{25,29,35,60,61}. Several reasons might explain this failure. Although the number of sequenced SCLC samples has increased over time, this remains far from the case for other cancer types included in sizeable genomic databases, such as the PCAWG and TCGA. Moreover, the majority of sequenced clinical tumours were LS-SCLCs surgically resected in the early-stage setting, which are biologically and molecularly different from ES-SCLC owing to selective pressures that often arise in the metastatic setting. Obtaining clinical samples sufficient to conduct multiomic studies has been a constant challenge in SCLC. Furthermore, publications from the past few years about subtype switching and the intertumoural and intratumoural heterogeneity of SCLC, especially RR disease, highlight the crucial need for serial monitoring of patient samples. Less-invasive and easier approaches, such as ctDNA sequencing, might alleviate some of these limitations. Several emerging results of ctDNA sequencing have been reported, although these studies could not identify targetable genomic alterations, which might have occurred because these studies used targeted gene panels^190,191,201. Comprehensive genomic analyses of SCLC ctDNA using WES of a large number of samples might identify potential targetable genomic alterations.

Autopsy-derived samples can also be adequate for research studies. So-called rapid autopsy programmes, some of which involve patients with SCLC, have been launched in many institutions²⁰², and preliminary results have been presented¹²⁹. The paucity of representative tissue samples has been one of the major roadblocks in conducting large profiling studies in SCLC, as small biopsy specimens often do not represent intratumour heterogeneity or the molecular landscape of the entire tumour. This hurdle can be countered using rapid research autopsy samples²⁰³. A study using rapid research autopsy samples from individuals with RR metastatic SCLC¹²⁹ detected the heterogeneous immune and neuroendocrine landscape as well as the clonal evolution of SCLC, indicating the utility of this approach in SCLC translational research^129,203,204. PDXs and human cell lines can also be a useful research tool; however, results from studies using these models must be interpreted carefully because promising preclinical results¹⁵⁹ often do not translate into clinical efficacy²⁰⁵. Overall, the genomic and transcriptomic concordance between PDXs and their original tumours has been validated²⁰⁶. Nevertheless, PDX models consist of a more differentiated neuroendocrine SCLC phenotype than clinical samples^58,68. Moreover, sampling lesions to derive SCLC cell lines does not fully recapitulate the clinical distribution of SCLC metastases.

As discussed, expression levels of transcription factors rather than different genomic profiles define SCLC subtypes with therapeutic implications. Testing these hypotheses in biomarker-driven clinical trials is warranted (Fig. 3). Several questions need to be resolved before launching such trials. What validated assays are required to define these transcriptional subtypes: gene expression, protein expression or IHC staining? What cut-off must be used? Biomarker-driven clinical trials can limit patient recruitment and be hindered by turnaround time, especially in the case of an aggressive tumour type such as SCLC, for which urgent initiation of systemic treatment is needed.

Tumour heterogeneity and plasticity promote resistance to therapy and accelerate the metastatic potential of SCLC. Epigenetic modulators such as LSD1 and EZH2 have been proposed to have a role in the plasticity of neuroendocrine differentiation^110,112. Epigenetic reprogramming might regulate changes in plasticity, which can reduce tumour heterogeneity and therapeutic resistance. Several epigenetic modifiers, including inhibitors of LSD1 or EZH2, such as benzenesulfonate, tazemetostat, PF-06821497 and tulmimetostat, are currently being tested in clinical trials involving patients with SCLC (Supplementary Table 1).

Future approaches should focus on early detection, interception and identification of individuals at risk of developing SCLC, continued development of biomarkers predictive of benefit from chemotherapy and ICIs, optimization of existing biomarkers and treatment options, identification of the mechanisms and clinical implications of subtype switching and lineage plasticity, development of novel therapies, and increasing ethnic diversity in clinical trials (Supplementary Box 1).

Conclusions

The treatment landscape for SCLC remained largely stagnant for several decades (Fig. 4). In 2012, the NCI categorized SCLC as a recalcitrant cancer because of its poor prognosis and lack of effective therapies. Numerous preclinical studies have revealed promising novel therapeutic approaches for SCLC, which have been tested in clinical trials. Unfortunately, most of these have failed to show an OS benefit — for example, the addition of ICIs to first-line chemotherapy moderately extends OS (by only 2–3 months)^22,145. Data emerging from preclinical and clinical studies support a focus on SCLC molecular subtypes and other biomarkers in future prospective clinical trials with patient stratification guided by validated precision oncology-based assays.

Fig. 4 ∣. Current management strategy for SCLC.

Proposed strategy for the management of patients with small-cell lung cancer (SCLC) as of 2024. ^aLimited-stage SCLC (LS-SCLC) includes stage I–III disease (T any, N any, M0) that can be safely treated with definitive radiation doses. Excludes T3–4 disease owing to the presence of multiple lung nodules that are too extensive or have tumour and/or nodal volume too large to be encompassed in a tolerable radiation plan. ^bExtensive-stage (ES-SCLC) includes stage IV (T any, N any, M 1a/b/c) and T3–4 disease. ^cDefined according to the Eighth Edition of TNM Staging of Lung Cancer²⁰⁷. ^dIn patients with medically inoperable LS-SCLC who can be considered for stereotactic ablative radiotherapy or concurrent radiotherapy. ^eBased on the results from phase III ADRIATIC trial¹⁵⁰, not yet approved by the FDA.

Targeted therapies have emerged as a revolutionary treatment approach for patients with NSCLC but their efficacy in SCLC and other high-grade neuroendocrine tumours remains limited. Drug development strategies for SCLC are pairing several novel agents with predictive biomarkers, many of which could be tested in prospective trials including the patient subgroups most likely to benefit from these experimental therapies. The results of these studies, which should address the challenges we have described (Supplementary Box 1), are awaited and might provide new, much needed treatment options for patients with SCLC. In 2024, the presentation of results from the phase III ADRIATIC trial¹⁵⁰ testing durvalumab after concurrent chemoradiation which has demonstrated a statistically significant improvement in OS and PFS in patients with LS-SCLC, and the accelerated approval of tarlatamab by the FDA for ES-SCLC⁸⁰, provide meaningful advances in the treatment of SCLC (Fig. 4). Therefore, the identification of novel, clinically applicable biomarkers and actionable therapeutic targets, understanding of mechanisms of resistance to current standard therapies, and continued improvement of the outcomes in patients with early-stage SCLC are the major future focus areas in SCLC clinical research.

Key points.

• Continuing to improve the biological understanding and therapeutic strategies for early-stage small-cell lung cancer (SCLC) remains a critical need to improve the overall prognosis of SCLC.

• Immune checkpoint inhibitors improve the overall survival in a minority of patients with extensive-stage SCLC (ES-SCLC). Repression of MHC I-dependent antigen-presentation machinery and low infiltration of cytotoxic T lymphocytes remain major challenges in improving outcome in a majority of patients with ES-SCLC.

• The major transcriptional subtypes of SCLC have distinct biology and therapeutic vulnerabilities. However, application of SCLC subtypes as clinically actionable biomarkers continues to be a challenge.

• Intratumoural heterogeneity and plasticity after treatment highlight the complexity of SCLC and present potential hurdles in treatment strategies for relapsed SCLC. Improvement in single-cell analysis methods, liquid biopsy platforms and rapid research autopsy programmes will potentially facilitate the understanding of SCLC biology, capturing the intratumoural heterogeneity and plasticity in treatment-resistant SCLC.

• Novel cell surface targets (such as DLL3, SEZ6, TROP2 and B7-H3) and MHC-independent immunotherapy strategies are paving the path for promising clinical trials.

• Results of the phase III ADRIATIC trial in limited-stage SCLC and accelerated FDA approval of tarlatamab for the treatment of ES-SCLC demonstrate significant advances in the treatment of SCLC. Future work will focus on understanding treatment resistance and potential combinatorial strategies to improve the therapeutic landscape in SCLC.