Small Cell Lung Cancer Classification: Unraveling Heterogeneity to Enable Personalized Treatments.

Introduction
Small cell lung cancer (SCLC), constituting approximately 15% of all lung cancer cases, is highly associated with smoking and is characterized by rapid proliferation, increased vascularization, and aggressive behavior (1–3). Due to early metastasis, most patients are diagnosed at extensive-stage (ES) disease, excluding surgical resection (4–6).
Systemic doublet chemotherapy (CHT) consisting of a platinum-based agent with the topoisomerase inhibitor etoposide has been the standard of care in advanced disease stages for decades (7). Unfortunately, initial treatment efficacy is followed by disease relapse through acquired resistance, which reflects dismal overall survival (OS; refs. 1, 8, 9). The addition of immunotherapy (e.g., atezolizumab and durvalumab) to first-line ES-SCLC treatment has led to modest improvements in survival in unselected populations. Meanwhile, durvalumab administered after chemoradiotherapy recently demonstrated a clinically relevant survival benefit in limited-stage disease (10–12).
Non-SCLCs (NSCLC) have increasingly been categorized into therapeutically relevant subgroups based on their mutational profiles. Targeted agents against gain-of-function alterations in ALK, KRAS, EGFR, BRAF, MET, and ROS1 oncogenes have produced significant improvements for patients with NSCLC (13, 14). In contrast, the genomic landscape of SCLC is uniform, making it difficult to define therapeutically meaningful subclasses. Therefore, SCLC is still regarded and treated as a single, monolithic entity.
The low number of surgical resections has limited SCLC whole tissue sample availability. Accordingly, SCLC research relies on preclinical models, including cell lines, patient‐derived xenografts (PDX), or genetically engineered mouse models (GEMM). However, the implementation of these models, together with emerging profiling studies, has been important in understanding SCLC biology (1, 15, 16). Although the same therapeutic strategies apply to all patients with SCLC in clinical settings, SCLC displays a heterogeneous underlying molecular profile beyond its mutational landscape and cellular plasticity (17, 18). Exploring these molecular subtypes has gained relevance because they likely influence metastatic patterns (19), the efficacy of treatment modalities (20, 21), and the development of therapy resistance (17, 22).
Several attempts to classify SCLC into different subgroups have been noted in the last few years, yielding varying descriptions of the molecular landscape. We recently reviewed major clinical and preclinical advancements in SCLC research (18, 23). Here, we first provide an overview of relevant SCLC subgrouping attempts from the past 10 years and list the main mechanisms influencing cellular subtype composition. Furthermore, we describe developments in tackling SCLC’s dynamic profile and include select therapeutic target discoveries per subgroup. Lastly, we synthesize existing SCLC classification terminology by highlighting the links between the proposed classifications and offer suggestions for potential future directions within SCLC research.

Defining Subgroups of SCLC

Summary of impactful classifications
Historically, only a few research groups have aspired to classify SCLC in the past century. Gazdar and colleagues (24) subdivided 50 SCLC cell lines into “classic” and “variant” phenotypes. However, the worldwide resurgence of SCLC profiling studies has influenced classification attempts within the last few years. Figure 1A highlights these attempts, focusing on the relationships between proposed categories. Supplementary Table S1 summarizes the SCLC subgrouping events within the past 10 years.
George and colleagues (25) performed whole-genome sequencing on 110 SCLC tumor samples to gain a comprehensive overview of the SCLC genetic landscape. Interestingly, SCLC tumor progression necessitated almost universal, biallelic inactivation of the tumor suppressor genes TP53 and RB1. Although a small fraction of SCLC tumors portrays chromothripsis as well as wild-type RB1 and TP53 status, dual inactivation of these genes is now known as a characteristic feature of SCLC tumorigenesis (25–27). Although TP53 and RB1 expression could not define mutually exclusive subgroups, unsupervised hierarchical clustering of RNA sequencing (RNA-seq) data on primary SCLC tumor samples (n = 81) elucidated two categories with varying neuroendocrine (NE) profiles. Group II portrayed elevated NE marker expression [e.g., chromogranin A (CHGA), gastrin-releasing peptide] compared with group I (17, 25). Meanwhile, low NOTCH activity and subsequent tumor progression in group II were reflected in higher levels of achaete-scute homolog 1 (ASCL1) and protein delta homologue 1 (DLK1), two genes involved in NE differentiation (17, 25). In this study and a subsequent study, NOTCH receptors were identified as tumor suppressors in SCLC and inhibitors of the NE state (25, 28).
Considering NE characteristics within a tumor population, Calbo and colleagues (29) revealed cellular subgroups marked by alternate morphology and marker expression. This in vivo study observed the parallel existence of non-NE cells with larger cell sizes and mesenchymal marker expression (e.g., vimentin or CD44) among small NE cells in the tumor microenvironment (TME). Moreover, oncogenic Ras expression in mouse cell lines conferred transitions from NE to non-NE phenotypes, propagating tumor progression and metastasis (29).
Subdividing the molecular tumor landscape by NE marker positivity has been a reoccurring trend throughout SCLC classification efforts. Borromeo and colleagues (30) investigated human cell lines in 2016 and discovered that the transcription factors (TF) ASCL1 and neurogenic differentiation factor 1 (NEUROD1) enable a distinction between NE (ASCL1- and/or NEUROD1-high) and non-NE (ASCL1- and NEUROD1-low) SCLC subgroups. Mollaoglu and colleagues (31) were likewise able to highlight differences in NE marker expression through GEMMs. ASCL1-high SCLCs reflect the “classic” SCLC phenotype defined by Gazdar and colleagues (24). These are characterized by floating cell aggregates, NE marker expression such as L-dopa decarboxylase or neuron-specific enolase, as well as bombesin-like immunoreactivity, whereas NEUROD1-high SCLCs assigned to the “variant” subtype represent lower NE marker levels (17, 24). ASCL1, a TF essential for normal pulmonary NE development, is required for tumorigenesis (30). Distinct gene expression profiles and differing genetic functions have been attributed to the transcriptional activators ASCL1 and NEUROD1. Downstream targets uniquely bound by ASCL1 include L-MYC, RET, NFIB, and BCL2, whereas MYC and the TF OTX2 are characteristic targets of NEUROD1 (30).
Investigation of the oncogene MYC in a TP53−/RB−/MYC+ (RPM) mouse model revealed its function as a driver of NEUROD1-positive, NE-low SCLC. Mollaoglu and colleagues (31) highlighted shifting TF expression from ASCL1- to NEUROD1-high SCLCs through increased MYC amplification in the latter group, yielding aggressive tumors with metastatic potential. Subsequent studies have underlined this temporal evolution between SCLC subsets, with increased MYC expression driving transitions from ASCL1-high SCLCs to NEUROD1-high, NE-low profiles and transformation into non-NE phenotypes as a resistance mechanism to therapy (32, 33). Baine and colleagues (34) demonstrated that SCLC tumor specimens could harbor ASCL1-positive cell populations and other populations positive for NEUROD1 within the same tumor. Others, including our group, have since observed dual TF expression (e.g., termed SCLC-AN) in surgically resected tumor samples through immunohistochemistry (IHC; ref. 20). Expression of multiple TFs in the same lesions underscores tumor plasticity and highlights its transitional ability, as delineated in the following sections.
McColl and colleagues (35) identified insulinoma-associated protein 1 (INSM1) as being highly expressed in the majority of SCLCs with “classic” phenotypes when investigating previously performed gene expression profiling of SCLC cell lines and validating it by analyzing data on human SCLC tumors (35). INSM1 codes for a NE TF and is associated with high ASCL1 and NEUROD1 expression (6, 35). In contrast, a subset of SCLCs with “variant” features demonstrated gene expression of the TF yes-associated protein 1 (YAP1), a downstream target of the Hippo pathway. Surprisingly, a few cell lines had RB1 wildtype and YAP1 positivity, a rare event in SCLC. YAP1 expression is downstream of RB1 signaling, and YAP1-positive SCLCs have been found to be less sensitive to CHT (35). In this study, mutually exclusive expression of INSM1 and YAP1 defined NE and non-NE subgroups. Consensus clustering, weighted gene coexpression network analysis of RNA-seq data of 50 SCLC cell lines from the Cancer Cell Line Encyclopedia (CCLE), and validation in mouse and human tumors enabled a further categorization of ASCL1-high SCLCs according to differential hairy and enhancer of split-1 (HES1) expression—with HES1 as a marker of NOTCH activity (36). Although Wooten and colleagues identified ASCL1-high (NE), NEUROD1-expressing (NE-v1), and YAP1-positive (non-NE) clusters, the authors proposed a fourth subgroup marked by the dual expression of ASCL1 and HES1 (NE-v2). This variant NE subgroup was not separable from others using only a few markers and portrayed the least drug sensitivity (36). Besides INSM1, the NE TF atonal bHLH TF 1 (ATOH1) defines a small subset of SCLCs, as seen in a study by Simpson and colleagues (37) that analyzed the global SCLC transcriptome through circulating tumor cell–derived explant models. The biological rationale of this small subgroup remains to be elucidated.
Other factors beyond NE marker expression have been explored to comprehend the heterogeneous SCLC landscape. Pulmonary NE cells (PNEC) have been described as progenitor cells for SCLC (18). In addition to subsets defined by high ASCL1 and NEUROD1 expression, domain-focused CRISPR screening on human SCLC cell lines revealed a fraction of SCLCs marked by Pou class 2 homeobox 3 (POU2F3) gene expression (38). As POU2F3 is a TF essential for the development of chemosensory cells located in the respiratory and gastrointestinal tracts, the “tuft cell” has been postulated to be a cell of origin for some SCLC tumors (38). Similarly to “variant” SCLCs, POU2F3-high “tuft cell derivatives” in this study by Huang and colleagues (38) did not portray NE markers (e.g., synaptophysin or CHGA) and rather carried tuft cell–related features. Recent work suggests that POU2F3-high SCLCs may also arise from basal cells (39).
In 2019, Rudin and colleagues (17) stratified SCLC’s molecular profile according to TF positivity of ASCL1 (SCLC-A), NEUROD1 (SCLC-N), POU2F3 (SCLC-P), and YAP1 (SCLC-Y). Borrowing this terminology and applying curated NE scores, SCLC-A and SCLC-N could be assigned to George’s NE GROUP II on the basis of varying NE marker expression (17, 25, 39, 40). Conversely, the SCLC-Y and SCLC-P phenotypes encompass more non-NE features, with low or absent NE marker positivity. Numerous subsequent preclinical and clinical validation studies ensued after this milestone article in SCLC. Although overlapping data about subgroups defined by the TFs ASCL1, NEUROD1, or POU2F3 have been noted, discordant results for the fourth subgroup marked by YAP1 expression exist (41–43).
In our study from 2022, we could not validate a YAP1-dominant subgroup through IHC analysis of surgically resected tissue samples from 386 patients with SCLC (20). Instead, we proposed a “quadruple negative” subtype (SCLC-QN) with low expression of the 4 TFs (20). Other research groups have likewise failed to define an exclusive YAP1-positive subgroup in validation studies (21, 34, 37, 44–46). A recent cell line analysis questioned YAP1 as a reliable subtype-defining marker as YAP1-expressing SCLC cell lines bearing SMARCA4 mutations resembled SMARCA4-deficient undifferentiated tumors more than SCLC (47). Although there is no evidence for a YAP1-defined subtype in which most SCLC cells express high YAP1 levels, the presence of a fraction of YAP1-positive cells in SCLC tumors has been noted in GEMMs and human samples (48, 49).
In an integrative analysis including cell lines and tumor specimens of NE carcinomas, Wang and colleagues (49) described five subgroups characterized by the TFs ASCL1, NEUROD1, HNF4A, POU2F3, and YAP1 (subtypes SCLC-A, SCLC-N, SCLC-H, SCLC-P, and SCLC-Y, respectively). Although these subtypes were validated through IHC, tumors were largely assigned to subtypes SCLC-A and SCLC-N, with subtype SCLC-H constituting only 3% of all SCLC tumors (total n = 216; ref. 49).
Besides TF expression, increased immune system involvement and ensuing inflammatory characteristics in a subset of SCLCs led Gay and colleagues (21) to propose the “SCLC-inflamed (SCLC-I)” subgroup as a non-NE SCLC subtype. Application of non-negative matrix factorization (NMF) to the George and colleagues’ transcriptomic dataset on 81 SCLC tumor samples revealed four groups. This inflamed subgroup portrayed lower expression of the TFs ASCL1, NEUROD1, and POU2F3 (21). As in SCLC-P, NE markers were not elevated in SCLC-I. However, expression of the TF RE1 silencing transcription (REST), an NE gene repressor, was increased. This subset also showed elevated gene expression of immune-related genes (e.g., IFNγ activation, immune checkpoints, or human leukocyte antigens). Furthermore, important immune checkpoint molecules, for example, CD274 (PD-L1), PDCD1 (PD-1), or CD86 (CTLA4), were also expressed at higher levels in this subgroup. Applying these four categories to the analysis of IMpower133 data revealed that SCLC-I benefitted most from immunotherapy (21).

De novo NMF by Nabet and colleagues uncovered a previously unknown heterogeneity within the SCLC-I subtype. Investigation of tumor samples from 271 patients with ES-SCLC included in the IMpower133 clinical trial yielded four clusters with respective distributions, with alternative descriptions in brackets: 31.4% (n = 85) NMF 1 (or SCLC-N), 32.5% (n = 88) NMF 2 (or SCLC-A), 14.4% (n = 39) NMF 3 (or SCLC-I–NE), and 21.8% (n = 59) NMF 4 (or SCLC-I–non-NE; ref. 44). NMF 1 to NMF 3 portrayed dominant NE features, leaving NMF 4 as the only non-NE SCLC category. Importantly, POU2F3-expressing SCLCs were assigned to the SCLC-I–non-NE subgroup but were not a separate class in this analysis. The two “immune-hot” SCLC-I categories (NMF 3–4) were differentiated by levels of immune-suppressive tumor-associated macrophages (TAM), with high TAM levels in SCLC-I–non-NE tumors (44). These authors not only described ASCL1-high, inflamed tumors (NMF 3) not previously reported in the SCLC literature but also highlighted diverse immune cell involvement as a critical agent in SCLC profiles.
Proteome profiling of 112 tumors from patients with SCLC by Liu and colleagues (46) likewise delineated subgroups with increased immune cell infiltration as a further classification besides NE marker expression. In line with former results, these authors detected upregulation of immune-related pathways (e.g., IFNγ signaling or antigen presentation machinery) in their “immune-hot” subgroups (NMF 3–4). However, higher immune involvement was significantly negatively correlated with NE features. Additionally, SCLC subgroups with greater immune system involvement seemed to be compatible with inflamed and POU2F3-driven subgroups (46).

Influences on SCLC Subtype Composition
Plasticity, the capacity of cancer cells to exhibit multiple phenotypes in response to internal (e.g., transcriptional or epigenetic changes) or external cues (e.g., inflammation or treatment), represents an additional hallmark of cancer also reflected within SCLC biology (16, 50, 51). In contrast to the relatively static, oncogene-defined subgroups of NSCLC, SCLC subtypes are dynamic cell states arising through the interplay of (epi)genetic regulation, cellular origin, or the TME (Fig. 1B; refs. 16, 52). The following section will highlight relevant drivers of cell state transition.

Epigenetic drivers and TF gene expression
DNA methylation levels affect chromatin accessibility, and dysregulated patterns have been observed in cancerous cells compared with their healthy counterparts (53, 54). Although tumor suppressor genes within tumor cells feature hypermethylated promoter cytosine preceding guanine (CpG) sites hampering gene transcription, the converse is true for oncogenes (53). SCLC tumor promoters display significantly elevated DNA methylation compared with other cancers (54). Aberrant expression of chromatin modifiers directly affects downstream gene transcription, propagating progression and subtype plasticity (55).
Enhancer of zeste homolog 2 (EZH2), a protein within the PRC2 complex enabling trimethylation of histone H3 lysine 27 in gene enhancers and promoters, is significantly dysregulated in SCLC cells compared with normal lung tissue or NSCLC samples (54–56). Increased methyltransferase activity decreases chromatin availability and gene transcription through elevated methylation levels at CpG-promoter islands. High EZH2 expression and concurrent downregulation of Schlafen-11 (SLFN11) are associated with acquired resistance to CHT and worse prognosis (54, 55, 57, 58). Interestingly, inhibiting EZH2 enabled renewed SLFN11 expression (58). EZH2 overactivity also significantly affects plasticity and subtype composition by influencing TF-expression reprogramming (55, 59). Exemplarily, drug-induced blockage of EZH2 greatly decreased ASCL1 protein expression in SCLC cells and promoted a transition from NE-high to non-NE cell states with increased immunogenic gene signatures and repaired TAP1 expression, a protein facilitating MHC I antigen presentation ability to CD8+ T cells (59).
Binding of the histone demethylase KDM6A is increased at enhancers of NE genes and demonstrates preferences for ASCL1 gene transcripts (60). Loss of KDM6A activity loosens the chromatin structure at NEUROD1 promoters of SCLC tumors and stimulates transitions from ASCL1- to NEUROD1-positive SCLCs (60). Additionally, histone acetyltransferase activity (e.g., CREBBP or EP300) relaxes chromatin structure, facilitating transcription. CREBBP has been described as a strong tumor suppressor and is significantly mutated in SCLC, resulting in loss of function (61, 62).
Next, lysine-specific demethylase 1A (LSD1/KDM1A) is significantly expressed in SCLC, influencing NE differentiation through NE gene expression regulation (63–65). Inhibition of LSD1 through ORY-1001 in PDXs suppressed ASCL1 and downstream NE-signature genes, leading to a transition to a non-NE phenotype and upregulation of Notch signaling pathway members (e.g., NOTCH1 or HES1; ref. 64). REST, the downstream target of NOTCH1, was also upregulated following LSD1 inhibition (64). Importantly, loss of LSD1 activity promotes MHC I gene expression and induces an immunogenic, non-NE phenotype with increased T-cell presence and potential sensitivity to immunotherapy (65). LSD1 represses the mRNA-binding protein ZFP36L1, which modulates NE genes such as INSM1 and SOX2; restoration of ZFP36L1 expression through LSD1 inhibition drives the emergence of SCLC cells with immunogenic features (66).
Epigenetic modifiers seem to bear preferences for specific TFs; specifically, LSD1 inhibitors only demonstrate adequate efficacy in NE subtypes (60, 63). Combination therapy of agents targeting epigenetic modifiers (e.g., LSD1 and EZH2) might therefore aid in locking cells into less resistant cell states and abate SCLC progression (50).
Similarly, the mammalian SWI/SNF (mSWI/SNF) complex is solely relevant for POU2F3-positive SCLC in chromatin remodeling and gene expression (67, 68). It represses TAZ of the oncogenic YAP/TAZ complex, inhibiting its tumor suppressive role, and is an important regulator of SCLC progression (69). Cooperative binding of the coactivators POU2AF2 (i.e., C11orf53 or OCA-T1) and POU2AF3 (i.e., COLCA2 or OCA-T2) results in increased chromatin accessibility and gene expression of POU2F3-specific signatures (68, 70–72).
Conversely, downregulation of either POU2F3 or POU2AF2 expression and the inhibition of SWI/SNF activity in this subtype restructure chromatin with decreased expression of POU2F3-related genes and tuft cell–like features, leading to altered subtype identity (67, 68, 70).
Gopal and colleagues (73) detected SCLC cells’ capability of bivalent cellular transition, whereas environmental influences determine transition rates between state compositions. Targeted inhibition of the epigenome with JIB-04 (pan-histone demethylase inhibitor) reduced the number of cells expressing ASCL1 compared with NEUROD1- or YAP1-positive cells, whereas iBET-726 (bromodomain and extra terminal protein inhibitor) revealed greater ASCL1 positivity and increased conversion from NEUROD1- to ASCL1-positive states. Therapy-induced epigenetic regulation, therefore, directly affected subtype composition and transition. Cell state identity greatly influenced therapy response and tumor progression, with greater progression tendencies in ASCL1-high cells (73). The question arises whether SCLC subtypes are rather transient cell states with an internal capacity for change when under pressure, instigating altered intrinsic tumor features (e.g., growth advantages, differential therapy response, or immune desert phenotypes).

MYC and the Notch signaling pathway
SCLC evolves over time because of its recalcitrant nature and dynamic dominant gene expression regulated through driver genes (e.g., MYC amplification; refs. 74, 75). MYC overexpression is a known promoter of subtype transition, with ASCL1 positivity as a necessary precursor for the development of NEUROD1-high cells (32, 52). MYCL, MYC, and MYCN comprise the MYC oncogene family. About 70% of ASCL1-positive SCLCs demonstrate MYCL expression, whereas elevated MYC and MYCN levels in SCLC also exist (23, 52, 76). MYC engages with the SWI/SNF complex through the binding of SMARCB1, and BRG1 (i.e., SMARCA4) within the SWI/SNF complex drives gene expression of MYC and its downstream targets (77). Besides propagating subtype transition away from NE-high states, MYC amplification is also an important acquired resistance mechanism in SCLC (22, 75).
ASCL1 loss during subtype evolution is concurrent with the upregulation of signaling pathways of non-NE phenotypes, including the Hippo or Notch pathway (52). The latter also regulates cell fate determination and organogenesis during development (78). Downstream Notch pathway targets include the inhibitory delta-like ligands DLL1, DLL3, and DLL4, whereas DLL3 modulates NOTCH expression (78, 79). NOTCH dysregulation induces REST overexpression and maintenance of cancer cell stemness, resulting in loss of NE differentiation and a drive toward progression (28, 48, 52, 78). The NE-inhibitory effects induced by Notch signaling activation (and ensuing REST upregulation) seem to be irreversible (28, 51).

Cell of origin
As mentioned above, PNECs, lung epithelial cells (e.g., basal, club, or AT2 cells), or tuft cells have been proposed as progenitors for SCLC (Fig. 1B; refs. 23, 28, 39). In vivo studies employing adeno-Cre viruses on TP53−/ RB1−murine models observed more effective tumorigenesis in Ad5-CGRP-Cre–infected cells (PNEC marker) compared with AT2 cells expressing surfactant protein C, suggesting the PNEC as a potential cell of origin (80, 81). Olsen and colleagues (52) highlighted that SCLC tumors originating from club cells in the lung portrayed elevated ASCL1 expression compared with YAP1, underscoring that this cell of origin does not have the same tendency toward progression as others (52). Histological transformation ofEGFR-mutated NSCLC to SCLC has also been observed (18). A recent GEMM study reported that ASCL1+ PNECs can undergo histologic transformation driven by MYC upregulation (82). In contrast, AT2 cells required activation of the Akt signaling pathway to acquire tolerance to MYC expression, which enabled their transition into a basal-like intermediate state. This transition represents a potential step in the progression from EGFR-mutated NSCLC toward SCLC (82). Although tuft-like SCLC cells have been linked to the TF POU2F3, Ireland and colleagues demonstrated that aberrant MYC regulation can result in POU2F3 positivity, but only from a basal cell of origin (38, 39). Defective ASCL1 or PTEN expression and transgenic MYCT58A expression in their organoid model propagated POU2F3 positivity within basal-derived cells, shifting the molecular landscape (39). Basal cells yielded heterogeneous SCLC phenotypes (e.g., SCLC-A, SCLC-N, SCLC-P) in immunocompetent GEMMs (39). Thus far, no other direct associations between PNECs, transformed NSCLC, or other putative progenitor cells (AT2, club cells) and specific SCLC subtypes have been observed. Both the originating cell type and subsequent genetic aberrations (e.g., PTEN loss) play key roles in tumor cells’ capacity to differentiate into other subtype-dominant states under selective pressures (83).

TME
Although SCLC demonstrates a deficient antigen presentation machinery, exogenous IFN signaling can reimplement MHC I expression, reengaging the immune system (84). A small fraction of SCLCs (i.e., SCLC-I) bear greater tumor–immune cell interplay (SCLC-I), and the regulation of the epigenome, among other factors, affects this interaction (21, 44, 46, 85). Exemplarily, drug-induced inhibition of key epigenetic modifiers in the PRC2 complex (e.g., EZH2) enabled SCLCs to regain MHC I presentation ability (84, 86). Immunosuppressive cancer-associated fibroblasts within the TME have been described as important mediators of plasticity between SCLC phenotypes, and the lack of tumor-infiltrating T lymphocytes has been linked with defective immune cell engagement in the TME (87, 88). Besides the inflamed subgroup, immune profiling studies have revealed that SCLCs with elevated NE marker expression (e.g., SCLC-A or SCLC-N) portray an immune desert phenotype with low expression of the immune markers poliovirus receptor (PVR), indoleamine 2,3-dioxygenase (IDO), MHC II, and TIM3, as well as poor immune cell infiltration (89).
Taken together, SCLC subtype composition depends on numerous factors that intricately affect SCLC biology. The next sections delineate how insights into SCLC’s heterogeneous landscape have been applied in translational settings.

Implications of SCLC Subtyping

Diagnostic considerations
Considering their translational potential, aspects of SCLC heterogeneity have been explored (Fig. 2A). Genomic analyses revealed that other mutations besides the nearly universal inactivation of the genes TP53 and RB1 might exist in SCLC. Specifically, mutations in the genes KIAA1211, COL22A1, and FMN2 may arise in up to 20% (25). As these mutations are not directly targetable with existing therapies and the mutational landscape of SCLC is rather unsuitable for subtyping, other traits such as TF positivity defined by IHC, a pivotal tool in routine diagnostics, have gained relevance. These exploratory analyses suggest that elevated ASCL1 expression is an independent negative prognosticator in surgically treated individuals (20, 90, 91). Surgically resected tissue samples are valuable for profiling studies to enable the comprehensive analysis of representative tumor areas. However, as surgery only plays a minor role in SCLC therapy, assembling larger surgical tissue specimen collections poses significant challenges. Representative small biopsies retrieved through bronchoscopy-based technologies, transthoracic needle biopsy, or other similar modalities are often used (18). Unfortunately, SCLC biopsy specimens frequently display crush artifacts and necrotic areas and generally fail to mirror the entire tumor expression profile (26). Collecting blood samples for liquid biopsy assays has gained global interest due to their easy retrieval, minimally invasive nature, and cost-effectiveness (92). Furthermore, liquid biopsy for disease detection and surveillance is promising in SCLC, considering the high tumor load in the bloodstream, even at earlier stages (93). Two independent cell-free tumor DNA (cfDNA)–based classifiers were reported to distinguish SCLC subtypes through methylation profiling (94, 95). Chemi and colleagues (94) utilized preclinical models and plasma cfDNA from patients with SCLC for genome-wide DNA methylation profiling. Here, tumor-specific methylation patterns detected in cfDNA aided in discriminating between survival rates and were diagnostic for TF-defined molecular subtypes (ASCL1-defined, NEUROD1-defined, or double-negative; ref. 94). Heeke and colleagues (95) built a custom DNA methylation-based classifier to investigate differences in methylation patterns among molecular subgroups (SCLC-A, SCLC-N, SCLC-P, and SCLC-I) in tumor tissue and plasma cfDNA. Significant differences in global methylation levels were observed between subtypes, with SCLC-P and SCLC-N exhibiting genome-wide hypomethylation and hypermethylation, respectively (95). Alternate DNA methylation patterns detectable in PDXs, tumor tissue, or blood plasma not only can discriminate between malignant and healthy cells but also can decipher subtype heterogeneity (54, 94, 95). Preliminary results highlight the potential of exploiting variances in epigenetic regulation for subtype discrimination. Further validation is needed before these profiling techniques can be implemented for subtyping and disease monitoring in translational settings.

Therapeutic implications
The distinct vulnerability profiles of SCLC subtypes may offer a future framework for optimal personalized therapy (Fig. 2B). As the inhibitory ligand DLL3 is aberrantly expressed on the surface of most ASCL1-positive tumor cells (91, 96), applying DLL3-directed therapies might be a promising strategy in SCLC-A. Tarlatamab, a bispecific T-cell engager (BiTE) binding DLL3 on cancer cells and CD3 on T cells to promote T cell–mediated tumor lysis, received accelerated FDA approval in May 2024 for recurrent ES-SCLC. As tarlatamab was not administered in a subtype-specific manner in previous trials (97, 98), higher response rates might be expected in preselected patients with SCLC-A tumors. Additionally, other DLL3-targeted BiTEs (BI 764532 and QLS31904), trispecific T-cell activating constructs (HPN328), and chimeric antigen receptor T cells targeting DLL3 might become SCLC-A subtype–specific therapies in the future (99–102).
Furthermore, although rovalpituzumab tesirine, the first DLL3-directed antibody–drug conjugate (ADC), demonstrated only modest antitumor activity in late-phase clinical trials (103, 104), next-generation ADCs may also emerge as viable subtype-specific therapies for ASCL1-defined SCLCs (105). Although some studies suggest that DLL3 is also highly expressed in SCLC-N (34), most research, including a recent analysis of 1,721 patients with SCLC, still indicates the highest DLL3 expression in SCLC-A (106). Beyond DLL3, seizure-related 6 homolog (SEZ6) is distinctly upregulated in SCLC compared with other cancers, with a particular association with NE-high subtypes, especially SCLC-A (106, 107). A recent phase I trial investigating the effects of the SEZ6-directed ADC ABBV-011 in heavily pretreated patients with relapsed/refractory SCLC showed manageable safety profiles and encouraging antitumor activity (108). Besides cell surface targets, epigenetic inhibitors against LSD1 or histone deacetylases (HDAC) have also been linked with potentially higher efficacy in SCLC-A given the elevated INSM1 levels and CREBBP inactivation in this subtype (17, 62, 109, 110). LSD1 inhibitors disrupt the interaction between LSD1 and INSM1, thereby inhibiting tumor growth and NE-associated gene expression (e.g., ASCL1; ref. 110). Likewise, the efficacy of HDAC inhibition is most pronounced in CREBBP-deleted tumors by restoring histone acetylation and the transcription of cellular adhesion genes resulting from CREBBP loss (62). Lastly, as the antiapoptotic protein BCL-2 is frequently overexpressed in SCLC-A, BCL-2 inhibitors (e.g., venetoclax and navitoclax) might represent subtype-specific therapeutic agents in SCLC-A (17, 111, 112).
SCLC-N is often associated with MYC amplification and Aurora kinase A (AURKA) activity; therefore, targeting these tumors with MYC and AURKA inhibitors is reasonable (6, 77). Developing a clinical inhibitor for MYC has been difficult because of its disordered structure and the lack of a distinct binding pocket (113). However, recent advancements with the pan-MYC inhibitor Omomyc have provided early breakthroughs in MYC-driven cancers. Omomyc demonstrated cell-penetrating properties and therapeutic activity in different experimental models of NSCLC (114). Previous studies, including AURKA inhibitors (e.g., alisertib and danusertib), have revealed promising efficacy, both as monotherapies and in combination with other treatments including immunotherapy. A phase II study evaluating the safety and efficacy of alisertib in relapsed/refractory SCLC demonstrated potent antitumor efficacy (115). Additionally, it has been reported that MYC expression and mutations in cell-cycle regulators could serve as potential predictive biomarkers for alisertib efficacy, further justifying AURKA inhibitors as fitting therapeutic agents for SCLC-N (115). Due to SCLC-N tumors’ increased arginine biosynthesis and the selective tropism of the oncolytic Seneca Valley virus (SVV) for this subtype, NEUROD1‐driven tumors may be particularly sensitive to arginine depletion caused by pegylated arginine deaminase (ADI‐PEG 20) and to intravenous SVV-001 therapy (116, 117).
As SCLC-P cells tend to depend on insulin-like growth factor 1 receptor (IGF1R) expression (38), IGF1R inhibitors (e.g., dalotuzumab and figitumumab) may have therapeutic implications in POU2F3-defined SCLCs. Recent preclinical studies showed that high mRNA expression of POU2F3 in SCLC cell lines predicts the effect of lurbinectedin, an RNA polymerase II inhibitor that received FDA approval in 2020 for the second-line treatment of metastatic SCLC (23, 118, 119). Although the subsequent phase III trial failed to show a clear survival benefit for patients with lurbinectedin-treated SCLC, individualized clinical trials for SCLC-P should be considered (120). SCLC-P is more sensitive to PARP inhibitors than other subtypes due to its intrinsic defects in DNA repair, reliance on PARP-dependent pathways, and increased replication stress, yielding synthetic lethality upon PARP inhibition (121). In a recent study, SCLC-P models exhibited significantly greater sensitivity to PARP inhibitors, highlighting their potential as subtype-specific therapeutic agents (21). Lastly, two recent preclinical studies elucidated the mSWI/SNF complex as a potential therapeutic vulnerability in POU2F3-driven cells. Targeting subunits of this complex, such as the ATPase SMARCA2/4 (with the inhibitor FHD-286 or PROTAC degrader AU-24118) or the bromodomain-containing 9 (with the degrader FHD-609), was efficacious in this subtype (67, 122).
SCLC-I is the only subtype for which retrospective analysis of large-scale clinical trial data has confirmed the subtype-specific therapeutic efficacy of a particular agent (21). Retrospective analysis of IMpower133, the first randomized immunotherapy-based phase III clinical trial (10), revealed that patients with SCLC-I benefit the most from the addition of immune checkpoint blockade to standard-of-care CHT (21). The median OS of patients with SCLC-I treated with chemoimmunotherapy (atezolizumab + carboplatin and etoposide) was >18 months compared with just over 10 months in patients receiving CHT alone (21). In comparison, the OS difference between the atezolizumab and placebo groups was only 2 months in unselected patients (10). This impressive efficacy in SCLC-I is attributed to the tumors’ T cell-rich, immune oasis phenotype accompanied by an inflamed gene signature (21). As both mTOR and CDK inhibitors have modulatory effects on the tumor immune microenvironment and may enhance immunotherapy efficacy when used in combination, these agents could be promising in this subtype (123, 124).
SCLC-I–non-NE is not as susceptible to immunotherapy as anticipated by its inflamed profile. Nevertheless, B7H3 inhibition might be a promising therapeutic option in these cases as studies investigating its expression in SCLC cell lines and tumors found that this cell surface marker was more highly expressed in non-NE SCLC subgroups (107, 125). Of note, the FDA has granted orphan drug designation to the B7H3-directed ADC 7MW3711 based on the encouraging results of preclinical studies demonstrating its significant tumor-killing effects, good safety profile, and pharmacokinetic properties (126). With respect to SCLC-I–NE tumors, the treatment of choice should be immunotherapy; however, they are also presumably susceptible to DLL3-directed therapies, given the direct association between DLL3 expression and NE differentiation (127).

Synopsis and Future Directions
The uncovered heterogeneity of SCLC reflected in varying classification attempts, partly hampers a concordant description of this disease but also highlights the many angles needed to view the SCLC landscape in its entirety. Diverse immunologic profiles of SCLCs highlight the importance of investigating the tumor immune microenvironment. Although Gay and colleagues (21) showed that patients with SCLC-I tumors benefit most from immunotherapy, Nabet and colleagues (44) delineated which subpopulation (SCLC-I–NE) likely demonstrates the best response to immune checkpoint blockade. Owonikoko and colleagues (128) described a YAP1-driven SCLC subtype with an inflamed phenotype showing upregulation of specific immune-related genes. Whether and to what extent the inflamed subtype overlaps with alternative classifications (i.e., SCLC-QN or SCLC-Y) and which tumor–immune system interactions are responsible for the inflamed cellular state in this small population of SCLC tumors still need to be addressed.
Many of these studies have described NE marker positivity as a classifier for SCLC subgroups. The NE TFs ASCL1 and NEUROD1 seem to represent important transcriptional activators in SCLC tumor biology. POU2F3-positive, non-NE SCLCs have also been repeatedly observed although not all studies agree on POU2F3 gene expression as an exclusive subtype-defining criterion. The variances in immune profiles seen in groups of patients with SCLC denote the potential relevance of not only categorizing SCLC tumors according to their molecular subtypes and NE differentiation but also by their immune phenotype (e.g., immune oasis or immune desert). Although classifications through subtyping or immune profiling often overlap, using both can provide a deeper understanding of the SCLC landscape.
Besides tissue-based IHC or RNA-seq, liquid biopsies hold tremendous promise for tumor state interrogation and molecular subtyping (94, 95). Therefore, blood-based assays such as cfDNA methylation profilers or the use of circulating tumor cells might aid in monitoring disease progression among patient subsets, facilitate earlier recurrence detection through epigenetic changes, or reveal links to molecular subtypes (94, 95, 129). These methods remain experimental and require large-scale validation before clinical implementation.
A preclinical study exploring SCLC’s metabolomic state demonstrated ASCL1high and ASCL1low subgroups, with elevated guanosine nucleotide synthesis in the latter group (130). Inosine monophosphate dehydrogenase, a downstream target of MYC, was identified as a potential druggable target for ASCL1low SCLCs (130). Although results are preliminary, future investigation of cancer metabolomics may yield promising subtype-differentiating markers.
Comprehensively characterizing SCLC’s dynamic landscape will additionally necessitate reassessment of selected in vitro or in vivo models. PDXs have shown high congruency with the original patient tumor and have been found to more adequately capture tumor heterogeneity as well as TF gene expression compared with cell lines (54, 73, 131). Exogenous factors such as immune system involvement need to be accounted for in model systems (e.g., through the use of allograft assays or organoid models to better study tumor–immune cell interplay; ref. 39).
Furthermore, the applicability of SCLC TFs (ASCL1, NEUROD1, or POU2F3) for subtyping and biomarker use has been explored in other cancer entities, including prostate, breast, small cell bladder cancer, and olfactory neuroblastoma (132–137). Interestingly, NE differentiation in prostate cancer also seems to depend on ASCL1 expression, and loss of this lineage-defining TF precedes a transition to a more basal-like cell state (132). An olfactory neuroblastoma GEMM model (RB1−/TP53−/MYC+) revealed similarities with SCLC, with ASCL1 loss necessary for cell transition to non-NE states, significant plasticity between TF-defined subtypes, and susceptibility to similar therapeutic targets (135). Although these TFs have been useful in some tumor entities, not all are reliable markers for subtype composition or cellular proliferation. Exemplarily, INSM1 positivity in IHC stainings of tumor samples proved to be a reliable marker of NE differentiation in breast cancer, whereas ASCL1 and POU2F3 were not frequently detected (136). Continued exploration of the underlying molecular landscape within SCLC will not only further unveil SCLC biology but also strengthen avenues of exchange with other NE cancers.
An exact characterization of SCLC subtype states will enable research advances in preclinical settings and precede biomarker discovery, with the potential for translational implementation. Despite SCLC’s heterogeneous properties, it will be useful to delineate which cell states are most advantageous for treatment. Therefore, combination treatments targeting subtype-specific vulnerabilities (e.g., LSD1 or MYC) might capture sensitive cell states and facilitate constraining plasticity. Moving forward, it will be important to address questions such as the mechanistic links of treatment-related resistance or possible organotropism during SCLC progression.

Conclusions
We have assembled a comprehensive overview of major SCLC subclassification attempts from the past 10 years and summarized the key factors influencing SCLC’s dynamic behavior. Besides molecular characteristics, epigenetic regulation and the tumor immune landscape significantly affect subtype composition and the underlying tumor biology. Looking forward, synchronizing the varying subtyping terminology while further elucidating SCLC’s heterogeneous landscape will aid in discovering additional subtype-directed biomarkers to enable personalized treatment.

Supplementary Material
Table S1Supplementary Table 1