De Novo and Histologically Transformed Small-Cell Lung Cancer Is Sensitive to Lurbinectedin Treatment Through the Modulation of EMT and NOTCH Signaling Pathways

Small-cell lung cancer (SCLC) is a high-grade neuroendocrine cancer that accounts for 15% of lung cancers and is typically metastatic at diagnosis (1). Annually, SCLC causes more than 200,000 deaths worldwide—approximately 30,000 deaths in the United States, and has a dismal median overall survival (OS) of approximately 1 year. The 2-year survival rate is below 5% for advanced disease (2). SCLC is characterized by a ubiquitous loss-of-function mutation in two major tumor suppressors, TP53 and RB1 (1). SCLCs can be divided into four primary, molecularly defined subtypes that differ substantially in biology and response to therapy. These subtypes are classified by the relative expression of four transcriptional factors ASCL-1 (SCLC-A), NEUROD1 (SCLC-N), YAP1 (SCLC-Y), and POU2F3 (SCLC-P; ref. 3). Although mutationally quite similar, work from our group and others has highlighted the intra- and intertumoral heterogeneity of SCLC.

The addition of programmed death-ligand 1 (PD-L1) antibody to first-line chemotherapy in extensive-stage (ES) SCLC provides durable responses in only a minority of patients, improving median progression-free survival (PFS) by 2 months. Few treatment options exist after relapse from first-line chemo-immunotherapy in the second line and beyond (4). Lurbinectedin, a synthetic analogue of a natural marine-based tetrahydroisoquinoline, is approved by the FDA as a second-line treatment for recurrent metastatic or advanced-stage SCLC (5). Lurbinectedin binds to the DNA minor groove in the “GG rich” region (6) and blocks RNA Pol II–mediated transcription. The phase II trial with lurbinectedin as second-line therapy in extensive-stage SCLC showed partial response in 35.2% of patients with a median duration of the response of 5.3 months (7). However, in a subsequent phase III trial (ATLANTIS), lurbinectedin in combination with doxorubicin failed to show superiority against the standard-of-care chemotherapy, cyclophosphamide/doxorubicin/vincristine or topotecan in patients with SCLC who have progressed from one line of platinum therapy (8).

Very little is known about the sensitivity of different molecular subtypes of SCLC to lurbinectedin treatment and biomarkers of lurbinectedin response. Therefore, defining the subset of patients with SCLC who are most likely to benefit from lurbinectedin treatment is a critical clinical need. Although a previous study proposed biomarkers for lurbinectedin sensitivity, this was based on analysis of human cell lines tested in vitro and in vivo (9). Further studies in cancer-relevant patient-derived xenograft (PDX) models representing the major SCLC subtypes and improved understanding of lurbinectedin-mediated transcriptomic and proteomic changes in SCLC are needed.

In addition to de novo SCLC, previous reports by our group and others have established lineage plasticity as an important mechanism of resistance to targeted therapies in lung cancer. In EGFR-mutant lung adenocarcinomas, up to 14% of acquired resistance to first-line osimertinib is attributable to the histologic transformation of SCLC (10, 11). Cases of EGFR wild-type (WT) lung adenocarcinoma (LUAD) transforming into SCLC have also been previously reported (12). Clinical outcomes of patients after SCLC transformation treated with SCLC-directed therapies are worse than de novo SCLC, with a high but transient response to platinum-based chemotherapy (13). The largest retrospective review of EGFR-mutant SCLC included 67 patients and reported a response rate of 54% to doublet platinum-etoposide, compared with the 60% to 70% response rates reported for de novo extensive-stage SCLC. Moreover, there were zero out of 17 responses observed among EGFR-mutant SCLC treated with later-line immunotherapy (14). Therefore, there remains an unmet need for effective therapies for patients with transformed EGFR-mutant SCLC that addresses their unique biology.

In this study, we investigated the preclinical efficacy of lurbinectedin in a panel of human and PDX-derived cell lines and de novo SCLC PDX models, representing the major subtypes. This study also demonstrates the antitumor effect of lurbinectedin with or without osimertinib in multiple PDX models of SCLC transformation. We further analyzed transcriptomic and proteomic changes pre- and postlurbinectedin treatment in vitro and in vivo to improve mechanistic understanding of lurbinectedin response in SCLC.

Human SCLC cell lines were obtained from the ATCC and the European Collection of Authenticated Cell Cultures (ECACC). PDX cell lines were derived in-house. Cell lines were maintained in RPMI media supplemented with 10% FBS, penicillin (100 U/mL), streptomycin (50 g/mL), and 2 mmol/L l-glutamine and incubated at 37°C with 5% CO₂. Cell lines were tested and authenticated by short tandem repeat profiling (DNA fingerprinting) and routinely tested for Mycoplasma species before any experiments were performed. Generally, the cells were used within the first three to five passages for any experiments.

Lurbinectedin was purchased from Selleckchem (catalog No. S9603) and also provided by Jazz Pharmaceuticals. Osimertinib was purchased from Selleckchem (catalog No. S7297).

SCLC cell lines were plated in 96-well plates (2,000 cells/well) in triplicate for each concentration. Cells were kept in complete RPMI media containing 10% FBS, penicillin (100 U/mL), and streptomycin (50 g/mL) and 2 mmol/L l-glutamine overnight and then treated with DMSO for control or different concentrations of lurbinectedin for 24 hours. The cell viability was assessed with CellTiter-Glo luminescent cell viability assay reagent (Promega) according to the manufacturer's protocol. Half-maximal inhibitory concentrations (IC₅₀) were calculated using GraphPad Prism Ver. 9.0 (GraphPad Software, Inc.; RRID:SCR_002798).

Fresh cell pellet or frozen tumor samples were shipped to GENEWIZ for RNA isolation, library preparation, and RNA sequencing (RNA-seq). Total RNA was extracted using the RNeasy Plus Universal Mini Kit following the manufacturer's instructions (Qiagen). RNA-seq libraries were prepared using the NEBNext Ultra II RNA Library Prep Kit for Illumina following the manufacturer's instructions (NEB). RNA samples and sequencing libraries were validated on the Agilent TapeStation (Agilent Technologies) and quantified by using Qubit 2.0 Fluorometer (Invitrogen) as well as by quantitative PCR (KAPA Biosystems). Sequencing libraries were clustered on one flow cell lane and then loaded on the Illumina HiSeq instrument (4000 or equivalent; RRID: SCR_020127) according to the manufacturer's instructions. Samples were sequenced using a 2- × 150-bp paired-end (PE) configuration. Image analysis and base calling were conducted by HiSeq Control Software (HCS). Raw sequence data (.bclfiles) were converted into fastq files and demultiplexed using Illumina's bcl2fastq 2.17 software (RRID: SCR_015058). One mismatch was allowed for index sequence identification.

The RNA-seq reads were quantified with Salmon v1.1.0 (RRID:SCR_017036) running on raw reads being mapped to the mm-10 genome using 25-mer indexing. Mapping validation (–validatemappings), bootstrapping with 30 resamplings (–numBootstraps), sequence-specific biases (–seqBias), coverage biases (–posBias), and GC bias corrections (–gcBias) were enabled apart from default settings. Sleuth v0.30.0 in gene mode (RRID:SCR_016883) was used for differential gene expression and principal component analysis (PCA) along with transcript per million normalization. The transcript to gene map was based on Ensembl 92 (RRID:SCR_002344). Wald test was used to identify DEGS, and significant genes were marked using FDRs as per Benjamini–Hochberg method and Sleuth-based estimation of log₂ fold change (FC) as specified. Across the sets of data of differential gene expression, the gene set enrichment analysis (GSEA; RRID: SCR_003199) was done, and genes were ranked on P value scores computed as −log₁₀(P) × (sign of beta). GSEA annotations of Hallmark genes were taken from Molecular Signatures Database (MSigDB v7.0.1; RRID: SCR_016863) gene-set enrichment analysis for functional enrichment were performed using the clusterProfiler R package (v3.16.0; RRID:SCR_016884)

A total of 61 RNA-seq libraries from two different experiments (de novo and transformed systems) were processed using the same pipeline for compatibility. The de novo experiment consisted of 30 samples from three cell lines namely Lx110, Lx33, and Lx1322, in which each cell line consists of five control replicates and five lurbinectedin treatment replicates. From this experiment, 16 samples were removed due to high variation between replicates, identified from PCA graphs (three from the Lx33 control group, two from the Lx33 treatment, three from the Lx110 control group, two from the Lx110 treatment group, three from the Lx1322 control group, and two from the Lx1322 treatment group). The transformed experiment consisted of 31 samples from two cell lines Lx831b and Lx1042, each with four conditions: lurbinectedin treatment, osimertinib treatment, combination treatment, and a vehicle control. All conditions had four replicates, except for the Lx1042 combination treatment, which had three replicates. From the transformed experiment, 12 samples were removed due to poor PCA clustering from high variation between replicates, one of which also had a unique alignment rate under 5% from the STAR aligner (RRID:SCR_004463; one from the Lx831b lurbinectedin group, one from the Lx831b osimertinib group, two from the Lx831b vehicle group, two from the Lx831b combination condition, two from the Lx1042 lurbinectedin condition, two from the Lx1042 vehicle group, one from the Lx1042 osimertinib group, and one from the Lx1042 combination group). Quality control of raw reads was performed using FastQC (v0.11.8; RRID: SCR_014583). Trim Galore! (version 0.6.5) was used to trim the adapter sequences with a quality threshold of 20 (RRID : SCR_011847). The human genome and transcriptome reference used was GRCh38.p13 genome assembly from GENCODE release 36 (RRID: SCR_014966). The alignment was performed by using STAR aligner (v2.7.5b; RRID:SCR_004463). Gene-level read counts were obtained by using Salmon (v1.2.1; RRID: SCR_017036) for all libraries using mapping mode. All remaining samples have passed the quality control requirements with >50% of reads uniquely mapping (>10M uniquely mapped reads for each library) using STAR aligner (RRID: SCR_004463).

Differential expression analysis was performed using the gene-level read counts and the DESeq2 (v1.28.1) R package (RRID: SCR_015687). Genes with less than five reads in total across all samples are filtered as inactive genes. A gene is considered differentially expressed if the P_adj value is less than 0.05 and the absolute log₂ FC is greater than 1. The over representation and gene set enrichment analysis for functional enrichment are both performed using the clusterProfiler R package (v3.16.0; RRID:SCR_016884). The gene sets used for functional analysis are obtained from The Molecular Signatures Database (MSigDB; RRID: SCR_016863). The Fisher test was used to determine whether the overlap between the DEG and the genes in the term is statistically significant (P < 0.05). The bold terms with an asterisk in front are the terms that are significantly enriched (P_adj < 0.05).

We performed the between-sample normalization using the variance stabilizing transformation of the DESeq2 package. The 1,000 most variable genes are used to perform PCA as well as calculating the Euclidean distances between each sample. Gene expression heat maps show the z-scores of DESeq2 VST normalized gene-level read counts. All visualizations were generated using plotly R package (v4.9.2.1; RRID: SCR_013991) except for heat maps and volcano plots. The heat maps and volcano plots were generated using heatmaply (v1.1.0; ref. 15) and Glimma (v1.16.0; RRID: SCR_017389) R packages, respectively. R (v.4.0.3; RRID: SCR_001905) was used to perform all bioinformatics analysis.

Protein extraction was done as described previously (16). Briefly, cells were harvested and washed with ice-cold PBS followed by resuspending the pelleted cells and in ice-cold RIPA buffer (Thermo Fisher Scientific, No. 89901) supplemented with protease and phosphatase inhibitors (Thermo Fisher Scientific, No. 78446). After incubating for 1 hour on ice suspension, cells were centrifuged at 14,000 rpm for 10 minutes in a refrigerated benchtop centrifuge (Eppendorf, No. 5340 R) to prepare the cell-free protein extracts. Protein lysates were quantified using a micro-BCA Protein Assay Kit (Pierce, No. 23235). Protein samples were prepared and run on SDS-PAGE followed by wet-transferring proteins to 0.45-μm Immobilon-FL polyvinylidene difluoride membrane (Millipore, No. IPFL00010). Membranes were blocked with Pierce Starting Block (PBS) Blocking Buffer (Thermo Fisher Scientific) at room temperature for 1 hour and then incubated overnight with primary antibodies (1:1,000) at 4°C. Incubation with horseradish peroxidase-linked secondary antibody was done at room temperature at a concentration of 1:10,000 and the bands were detected using iBright Western Blot Imaging Systems (Thermo Fisher Scientific). The list of antibodies is provided in Supplementary Table S1.

Apoptosis was detected by Annexin-V – PI assay according to the manufacturer's protocol in the FITC Annexin V Apoptosis Detection Kit I (BD Biosciences). Briefly, in 6-well plates, 2 to 4 × 10⁵ cells were seeded and treated with or without respective IC₅₀ concentrations of lurbinectedin for each cell line for 24 and 48 hours. After the treatment, harvested cells were stained with annexin-V and propidium iodide (PI) and analyzed using the Cytek Aurora (Cytek) flow cytometer with FlowJo software (version 10.6, BD Biosciences).

For this study, 6-week-old female nude mice weighing approximately 22 to 24 grams were obtained from ENVIGO. The mice were injected subcutaneously into the right flanks with 2 × 10⁶ cells mixed in a 1:1 mixture of PBS and Matrigel (No. CB40234, Thermo Fisher Scientific). At a tumor volume of approximately 100 to 120 mm³, mice were randomized and treated with either vehicle (5% glucose + 0.5% HPMC, i.v.) or lurbinectedin (0.2 mg/kg, once a week, i.v.). The treatment schedule continued throughout the experiment until the mice were sacrificed. Tumor volumes were measured using calipers and calculated as width² × 0.5 × length. Tumor volume along with body weights were monitored twice a week. To determine statistical significance among the different treatment groups, ANOVA followed by Student t test were performed using GraphPad Prism software, Prism (version 9.0). The animal study was approved by the Institutional Animal Care and Use Committee (IACUC). The identifying number of the approved protocol is 13–07–007.

For this study, 6-week-old female nude mice were obtained from ENVIGO. The mice were injected subcutaneously into the right flanks with 2 × 10⁶ cells mixed in a 1:1 mixture of PBS and Matrigel (Thermo Fisher Scientific, No. CB40234). At a tumor volume of approximately 120–1,500 mm³, mice were randomized and treated with either vehicle (5% glucose + 0.5% HPMC, i.v.) or osimertinib (25 mg/kg, orally, 5 days a week), lurbinectedin (0.2 mg/kg, i.v., once a week) or a combination of osimertinib and lurbinectedin (osimertinib 25 mg/kg, orally, 5 days a week + lurbinectedin 0.2 mg/kg, i.v., once a week). The treatment schedule continued throughout the experiment until the mice were sacrificed. Tumor volumes were measured using calipers and calculated as width² × 0.5 × length. Tumor volume along with body weights were monitored twice a week. To determine statistical significance among the different treatment groups, ANOVA followed by Student t test were performed using GraphPad Prism software Prism (version 9.0; RRID: SCR_002798). The animal study was approved by the IACUC. The identifying number of the approved protocol is 13–07–007.

Cell viability and flow cytometry data were expressed as means ± SD, and tumor volume data in in vivo studies were expressed as means ± SE. GraphPad Prism (version 9.0) software (RRID: SCR_002798) was used to determine statistical significance among the biologically distinct groups/cells and P value less than 0.05 was considered to be statistically significant (ns, P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001).

The raw and processed RNA-seq data have been deposited to Gene Expression Omnibus under accession number GSE223372.

To investigate the subtype-specific effect of lurbinectedin in vitro, a panel of 17 human SCLC cell lines, including three PDX-derived cell lines (representing all four molecular subtypes of SCLC) were treated with lurbinectedin (drug concentration ranging from 0.3 nmol/L to 30 nmol/L). After 24 hours of treatment, all SCLC lines showed sensitivity to lurbinectedin at a nanomolar range with IC₅₀ between 1.905 and 30 nmol/L (Fig. 1A). Interestingly, SCLC-P (H211, H526, and CORL-311) and SCLC-N (H82, H524, and H2171) cell lines showed relatively enhanced sensitivity to lurbinectedin treatment; and SCLC-A (H69, H146, and H720) cell lines showed relative resistance to lurbinectedin treatment (Fig. 1A). The cell lines were then treated for 24, 48, and 72 hours and showed a time-dependent sensitivity to lurbinectedin (dose–response curves of all cell lines is depicted in Supplementary Fig. S1). The IC₅₀ range is lower than the peak plasma concentration of lurbinectedin reported in a previous clinical trial (7).

Although lurbinectedin inhibited SCLC growth in most SCLC models, a subset of SCLC cell lines demonstrated relatively greater resistance to lurbinectedin treatment. To further investigate the potential SCLC subtype–specific effect of lurbinectedin, the IC₅₀ concentrations of the cell lines in our cohort were compared with the baseline gene expression of the cell lines (17). Interestingly, a statistically significant difference in sensitivity between SCLC-A and SCLC-P subtypes (P = 0.044) was observed with SCLC-P cells being more sensitive to lurbinectedin as compared with the SCLC-A subtype (Supplementary Fig. S2A). Previous reports have clearly demonstrated that preclinical models belonging to SCLC-P and SCLC-N subtypes had higher expression of MYC (3). MYC expression (though statistically not significant) may predict sensitivity to lurbinectedin (Supplementary Fig. S2B; P = 0.065).

To identify how lurbinectedin decreased viability in SCLC, we next investigated lurbinectedin-mediated apoptosis induction in SCLC. Annexin-V/PI-based flow cytometry was performed in six SCLC cell lines including SCLC-P (H211 and H526), SCLC-N (H82 and H446), and SCLC-A (H720 and H69) with or without lurbinectedin treatment for 24 and 48 hours. In agreement with the viability data, the highest apoptosis induction was observed in H526 and H211 (SCLC-P) cells (about 50%–55% apoptotic population in 48 hours post lurbinectedin treatment; Fig. 1B). The percentage of apoptotic cells ranged between 15% and 30% in the cell lines of the SCLC-N subtype (Fig. 1B). In agreement with viability data, cell lines of SCLC-A subtype had the lowest apoptotic induction posttreatment ranging from 5% to 12% (Fig. 1B).

Therefore, SCLC cell lines are sensitive to lurbinectedin treatment with enhanced sensitivity, particularly in the SCLC-P subtype.

The genes associated with response and resistance to lurbinectedin are understudied. To identify novel genes that predict response to lurbinectedin, we next correlated the baseline gene expression of the SCLC cell lines (17) to the IC₅₀ of lurbinectedin after 24 hours of treatment (Fig. 1C). Genes involved in the Wnt/β-catenin pathway including WD repeat domain 74 (WDR 74) and RNF43 as well as a DNA damage response (DDR) regulator, FAM117A were top genes associated with lurbinectedin sensitivity. These genes have not been studied in SCLC, and the functional role of these novel genes in lurbinectedin response needs to be further investigated in a larger cohort or preclinical models and clinical samples.

Next, transcriptomic analysis of sensitive (H526; SCLC-P) and resistant (H69; SCLC-A) SCLC cell lines pre- and post-lurbinectedin treatment was used to determine treatment-induced changes in genes involved in SCLC tumorigenesis.

More than 70% of SCLC tumors express a high neuroendocrine (NE) phenotype and expression of NE genes. Interestingly, in both models, the predominant transcriptional regulator of the corresponding cell lines, that is, POU2F3 for H526 and ASCL1 for H69 was significantly suppressed post-lurbinectedin treatment (Fig. 2A and B). NEUROD1 has previously been reported to regulate the migration and survival of NE lung carcinoma by regulating NCAM (18). Interestingly, in H69, NEUROD1 and NCAM1 were significantly enriched post-lurbinectedin treatment. NEUROD1, but not NCAM1, was also upregulated in H526 cells. CHGA and SYP were significantly upregulated in the H526 (SCLC-P) cell line. Furthermore, expression of the non-NE marker YAP1 decreased in H526 cells but increased in H69 (SCLC-A) cells indicating a change in neuroendocrine characteristics of both the cell lines upon lurbinectedin treatment. POU3F2 (BRN2), a transcription factor crucial for cell lineage determination and expression of NE markers including ASCL1, ND1, NCAM1, SYP, and CHGA (19), was also elevated (Fig. 2A and B) post-lurbinectedin treatment. The functional implications of these shifts in gene expression of factors previously implicated in SCLC biology should be investigated in future studies. Western blot analysis of the key transcription factors showed a decrease in the expression of ASCL1, NEUROD1, and YAP1 16 hours post-lurbinectedin treatment (Supplementary Fig. S2C).

Next, the expression of multiple genes associated with NE differentiation, tumor progression, and therapeutic resistance in SCLC and other cancers were investigated. Interestingly, lurbinectedin treatment enhanced the expression of genes associated with tumor progression like POU class 5 homeobox 1 (POU5F1), insulin-like growth factor binding protein 2 (IGFBP2), RNA polymerase II, I, and III subunit L (POLR2L), and NANOG (Fig. 2C and D) in both cell lines. Collectively, these data demonstrate that lurbinectedin treatment modulates NE genes and genes associated with tumor progression in in vitro SCLC models.

Gene-set enrichment analysis (GSEA) pre- versus. post-lurbinectedin treatment showed a uniquely upregulated pathway in the more sensitive H526 (SCLC-P) cell line (Fig. 3A and C) On the other hand, significant enrichment of MYC targets, epithelial–mesenchymal transition (EMT) pathways, downregulation of G₂–M checkpoint, and mTORC1 pathways were observed post-lurbinectedin treatment in the more resistant H69 (SCLC-A) cells (Fig. 3C and D). Adducts between DNA and lurbinectedin have been previously shown to activate apoptosis-mediated cell death and to inhibit tumor growth by triggering double-strand breaks and S-phase accumulation of cells in different cancer cell lines (20). Both MYC activation (21) and EMT (22) have been implicated as drivers of tumor proliferation and chemoresistance. Taken together, the involvement of these pathways highlights potential novel mechanisms of lurbinectedin resistance in SCLC.

Interestingly, the oncogenic PI3K/AKT/mTOR and the TGFβ pathway in the sensitive H526 SCLC model showed downregulation post-lurbinectedin treatment (Fig. 3A and C). The PI3K/AKT/mTOR signaling pathway is a major promoter of cell growth, survival, and invasion in multiple cancer types and is also known to be constitutively active in some subsets of SCLC (22). Dysregulated activation of the TGFβ signaling promotes chemoresistance, EMT, and proliferation in several cancers.

Further analyses of top differentially expressed genes (DEG) highlighted apoptosis signaling (including HAP1, SAT1, ATF3, and KLF10 genes) and immune pathway (including ATF3, JUNB) genes to be upregulated post-lurbinectedin treatment in the sensitive H526 model (Fig. 3B). In contrast, upregulation of genes involved in regulating EMT and metastasis (FTL), NOTCH signaling (NRARP), NF-κB (PIDD1, JUNB), and WNT-β catenin signaling (LGR4) were observed in the more resistant H69 model post-lurbinectedin treatment, nominating them as potential markers of resistance (Fig. 3E).

Taken together the data show that lurbinectedin treatment induces apoptotic signaling pathways and downregulation of oncogenic PI3K/AKT and TGFβ pathways in a sensitive model. On the other hand, an upregulation of EMT and MYC targets were observed post-lurbinectedin treatment in resistant models that may nominate these pathways as potential mechanisms of resistance to the drug in SCLC. Future validation studies would reveal the individual contribution of these pathways in SCLC.

Previous studies have shown the effect of lurbinectedin on DDR in SCLC models. In this study, RNA-seq analysis showed increased expression of both H2AX and PARP1 upon lurbinectedin treatment (Supplementary Fig. S3A), which confirms the modulation of the DDR pathway. Western blot analysis of SCLC-P (H526) and SCLC-A (H146) cells at 8 and 16 hours post-lurbinectedin treatment demonstrated an upregulation of γH2AX, a marker for double-stranded DNA (dsDNA) damage in both cell lines (Supplementary Fig. S3B). Significant upregulation of pRPA32, another marker of DNA damage, was observed in H526 cells post-lurbinectedin treatment. Confirming the apoptosis analysis, cleavage of both PARP and caspase-3 was observed in H526 and H146 cell lines post-lurbinectedin treatment (Supplementary Fig. S3B). Interestingly, lurbinectedin treatment led to an upregulation of phospho-ATM in H526 (SCLC-P subtype) but not in H146 (SCLC-A subtype) cells. Lurbinectedin treatment enhanced pCHK1 in both H526 and H146 cells. Therefore, these results suggest subtype-specific induction of apoptosis and DDR post-lurbinectedin treatment in SCLC in vitro models (Supplementary Fig. S3B).

Previous studies have reported the effect of lurbinectedin in vitro; however, an antitumor effect of lurbinectedin in cancer-relevant PDX models of SCLC is lacking. In vitro effect in cell lines often does not translate to the antitumor effects in vivo due to the inherent complexity of in vivo models and the presence of tumor vasculature. Therefore, we next investigated the antitumor effect of lurbinectedin in multiple unique PDX models of SCLC representing the three major subtypes. This study included three PDX models that have been previously characterized by genomic, transcriptomic, and proteomic assays with correlative clinical data (23). The three PDX models belong to SCLC-A (LX110), SCLC-N (LX33), and SCLC-P (LX1322) subtypes. Subcutaneous tumor-bearing mice from each model were treated with lurbinectedin (0.2 mg/kg, i.v., once a week) or vehicle. Lurbinectedin treatment significantly delayed tumor growth in an LX110 model (P < 0.0001; Fig. 4A). In LX110, tumors in the vehicle group reached a maximum tumor volume of 1,484 mm³ at day 46, whereas the tumor in the lurbinectedin-treated group was only 240 mm³ at day 46 indicating an 84% reduction in tumor growth relative to control (Fig. 4A; tumor volume summarized in Supplementary Table S2). In the group treated with lurbinectedin, the tumors reached their largest size by day 80, leading to a notable increase in the survival of the mice, whereas lurbinectedin treatment caused minimal antitumor effects in LX33 and LX1322 (Fig. 4A; tumor volume summarized in Supplementary Table S2). Although the in vitro results indicated SCLC-A to be more resistant to lurbinectedin treatment, the PDX in vivo model indicated the SCLC-A subtype to be more sensitive than SCLC-P or SCLC-N subtype. This discrepancy highlights the limitations of both systems and the necessity of complementary in vitro and in vivo studies to confirm the antitumor effect of therapies.

Bulk RNA-seq of LX110 vehicle versus lurbinectedin-treated tumors showed upregulation of TRIM72, which suppresses tumor progression and modulates the cellular energy state by modulating the AKT–AMPK–mTORC1 pathway (ref. 24; Fig. 4B) and HSPB7, another PI3K/AKT signaling modulating gene. Upregulation of RYR1 was also observed, which is a predictive marker of resistance to etoposide treatment in SCLC (25) and modulates WNT and the PI3K/AKT/mTOR pathway. WNT16 and ATF3, which are known to be activated upon DNA damage to promote apoptosis and interact with NF-κB, were downregulated upon lurbinectedin treatment (Fig. 4B). Interestingly, a downregulation of RSC1A1 (regulator of solute carriers 1) and other solute carrier genes SLC2A1, SLC16A3, SLC34A2, and SLC6A17 was also observed. Many solute carrier family members have been previously linked with protumorigenic activity and ferroptosis. The GSEA of the LX110 model indicated significant suppression of MTORC1, EMT, and glycolysis pathways, from the MSigDB Hallmark databases, post-lurbinectedin treatment in the sensitive LX110 model (Fig. 4C).

In the LX1322 PDX model, treatment with lurbinectedin led to the upregulation of IFITM3, which promotes tumor growth by enhancing EMT and influencing the WNT/β-catenin pathway. Additionally, LY6E, another molecule identified, acts as a tumor promoter by modulating the PI3K–AKT pathway. Interestingly, the SFRP2 gene, which is a negative regulator of WNT/ β-catenin signaling and the tumor suppression gene ANGPTL2 (angioarrestin), which promotes apoptosis and inhibits EMT post-lurbinectedin treatment was downregulated (Fig. 4B).

In LX33, representing the SCLC-N subtype, matrix metalloprotease 2 (MMP2) was downregulated, which is essential for metastasis and endo-mesenchymal transition. SFRP2, a modulator of WNT/β-catenin pathway, was also suppressed post-lurbinectedin treatment (Fig. 4B). Furthermore, REL, an important member of NF-κB signaling, and lung cancer–associated transcript 1 (LUCAT1) was upregulated post-lurbinectedin treatment (Fig. 4B). GSEA showed suppression of the EMT pathway and activation of glycolysis and mTORC1 signaling pathways from the MSigDB Hallmark database (Fig. 4C).

In summary, lurbinectedin treatment caused a significant delay in tumor growth in an SCLC-A PDX model, and transcriptomic analysis suggested a role of WNT/β-catenin, AKT–PI3K–mTORC1, and EMT signaling in the subtype-specific antitumor effect of lurbinectedin treatment in SCLC models.

Osimertinib shows initial activity in EGFR-mutant lung cancer; however, all patients develop resistance, often through SCLC transformation. There is a critical unmet need for effective later-line therapies for patients with transformed EGFR-mutant SCLC that addresses their unique biology. We hypothesized that combining EGFR- and SCLC-directed therapies may maximize disease response for this unique patient population. Therefore, to test the effect of lurbinectedin with or without osimertinib in transformed SCLC, three unique and novel PDX models of transformed SCLC were used: LX1042, LX151, and LX831b. After an initial tumor size of 120 to 150 mm³, mice from all three models were treated with (i) lurbinectedin (0.2 mg/kg, i.v., once a week), (ii) osimertinib (25 mg/kg, orally, 5 d/wk), or (iii) a combination of lurbinectedin and osimertinib.

In LX1042, at day 47 after tumor initiation, the control group reached the maximum tumor burden of 1,808 mm³ (Fig. 5A; tumor volume summarized in Supplementary Table S3). As expected, single-agent osimertinib had no appreciable antitumor effect in this transformed SCLC model and the average tumor volume was 1,369 mm³. Interestingly, in the lurbinectedin treatment group on day 47, the tumor volume reached only 859 mm³ showing an approximately 52% inhibition in tumor growth compared with the control group thus confirming the efficacy of lurbinectedin treatment alone in transformed SCLC (P = 0.0026; Fig. 5A). Significant tumor reduction was observed in the osimertinib + lurbinecetdin combination group. The average tumor volume was only 293 mm³ at day 47 showing more than 80% reduction of tumor burden (P < 0.0001). Most encouragingly at day 50, four of 10 mice had complete tumor regression. The complete tumor clearance continued until day 65 (Fig. 5A). The study continued with four mice until day 92 when the average tumor volume was only 619 mm³, indicating that lurbinectedin in combination with osimertinib caused significant tumor regression and survival benefit of transformed SCLC PDX model (Fig. 5A).

In LX151, at day 56, the average tumor volume of the control group was 1,934 mm³ and in the osimertinib-treated group, it was 1,226 mm³, showing no significant tumor regression compared with the control (Fig. 5B; tumor volume summarized in Supplementary Table S3). Interestingly, single-agent lurbinectedin treatment resulted in an average tumor volume of 108 mm³ at day 56 (Fig. 5B) showing a 95% inhibition of tumor growth (P = 0.0003). In the lurbinectedin + osimertinib combination group, there was significant tumor regression with an average tumor volume of only 47 mm³ at day 56 (P = 0.0002). Most encouragingly, 50% of mice had total tumor regression by day 21 in the lurbinectedin and osimertinib combination group (Fig. 5B).

Finally, for LX831b, the vehicle group reached an average tumor volume of 1,116.62 mm³ at day 55 (Fig. 5C; tumor volume summarized in Supplementary Table S3). The single-agent osimertinib treatment group reached an average tumor volume of 1,531.75 mm³ at day 80 (Fig. 5C). In the lurbinectedin-treated group, mice reached a tumor volume of 1,613.20 mm³ at day 77 (Fig. 5C). Finally, at day 27 in the osimertinib + lurbinectedin group, the average tumor volume was only 457.13 mm³ (P = 0.0046). Three mice in the combination group reached a tumor volume of 1,023.23 mm³ at day 107 showing a delay in tumor growth and survival benefit.

Therefore, in all three unique SCLC transformation PDX models used in this study, single-agent lurbinectedin treatment was significantly more effective than osimertinib treatment alone, and the combination treatment of lurbinectedin and osimertinib caused significant tumor regression opening much-needed avenues for therapeutic intervention in this very aggressive lung cancer subtype.

Next, transcriptomic analysis from available tumors in LX1042- and LX831b-transformed SCLC–PDX models was performed.

In LX1042, combination treatment led to significant upregulation of NOTCH2 (an important member of NOTCH signaling pathway), SFRP2 (regulator of WNT/β-catenin signaling), nerve growth factor (which promotes tumor invasiveness by activating PI3K/AKT/GSK3β pathway), and PLK2 (gene linked to drug sensitivity by interacting with both NOTCH family of proteins; Fig. 6A) as compared with the vehicle control. On the other hand, downregulation of HES1 (a NOTCH target gene that has also been reported to interact with NOTCH ligands like DLL1), death-associated protein kinase 2 (DAPK2), and SPRY4 (which have been linked to tumor proliferation and apoptosis) along with PHLDA2, which has been known to promote EMT by modulating PI3K/AKT signaling pathway in colorectal cancer were observed (Fig. 6A).

GSEA in the LX1042 model showed suppression of glycolysis, mTORC1, MYC targets, and NOTCH signaling pathways from the MSigDB Hallmark database (Fig. 6B), in the combination treatment as compared with the vehicle, which confirmed the importance of these signaling pathways in SCLC progression.

In LX831b, analysis of top DEG showed increased expression of JAG1 (a NOTCH ligand) and PLK2 in treated samples compared with untreated controls (Fig. 6C). Like LX1042, even in LX831b, a downregulation of HES1 and PHLDA2 that promotes EMT by modulating PI3K/AKT signaling pathway in colorectal cancer were observed (26). PKP1 (plakophilin 1), a tumor suppressor in lung cancer (27) and involved in a feedforward loop with MYC in squamous cell lung cancer (ref. 28; Fig. 6C), was also downregulated upon combination treatment. Interestingly, GSEA of the MSigDB Hallmark database indicated activation of the G₂–M checkpoint (Fig. 6D) pathway.

Together, these results suggest the involvement of MYC and NOTCH signaling in the appreciable antitumor response of lurbinectedin + osimertinib in transformed SCLC models. The exact functional role of these pathways needs to be investigated in future investigations.

Patients with extensive-stage SCLC typically have robust responses to first-line platinum-based chemotherapy, but most will experience chemoresistant relapse within the first year. Advances in therapeutic targeting of SCLC have been extremely limited and have resulted in only modest improvement of patients' dismal prognosis. Combination of atezolizumab and chemotherapy in the first line resulted in a modest increase in PFS of 1 month and OS of 2 months in extensive-stage SCLC when compared with chemotherapy alone (29). Similarly, in the CASPIAN study, the durvalumab plus platinum–etoposide group led to only a modest improvement in OS (13.0 months vs. 10.3 months) to the platinum–etoposide group (30). The complex genomic landscape, heterogeneity of these tumors, early metastasis of the disease, and almost inevitable drug resistance have made it difficult to identify effective and durable targets for this disease. Lurbinectedin is FDA approved as a second-line treatment for metastatic SCLC. However, the effect of lurbinectedin in SCLC subtype–specific gene expression changes or the efficacy of lurbinectedin in transformed SCLC has not been investigated. Our study explored the preclinical efficacy of lurbinectedin in a large and diverse panel of in vitro and in vivo models of de novo SCLC. We further investigated the lurbinectedin-mediated gene expression changes in SCLC models. Finally, this study is the first to show the effect of lurbinectedin in transformed SCLC preclinical models with or without osimertinib treatment.

SCLC is characterized by a ubiquitous loss-of-function mutation in TP53 and RB1, and amplification/overexpression of MYC family members in a subset of SCLC preclinical models and clinical samples (31). Beyond genomic alterations, SCLC is usually divided into four major subtypes mainly on the basis of the expression of four transcription factors, ASCL1. NEURO-D1, POU2F3, and YAP1. These subtypes have distinct biological characteristics and therapeutic vulnerabilities. A trend toward subtype-specific sensitivity to lurbinectedin was found, with the SCLC-P subtype being the most sensitive, followed by the SCLC-N subtype and SCLC-A subtype showing relatively less sensitivity in the cell line panel tested in this study.

Interestingly, MYC expression was correlated with lurbinectedin sensitivity (3). The correlation was not statistically significant, which was probably due to the limited sample size. MYC overexpression has been linked to increased aggressiveness in SCLC and is a predictive biomarker for DDR inhibitors like CHK1 and ATR (32, 33). This result suggests that MYC expression could be a potential predictive biomarker for lurbinectedin response, which needs to be further validated in clinical datasets.

In addition, genes in the NF-κB signaling pathway (like PLAA, SRGN, and TNFAIP8) were the top predictive biomarkers of resistance to lurbinectedin treatment. On the other hand, genes involved in DDR and WNT/β-catenin pathway predicted sensitivity to lurbinectedin. Aberrant NF-κB and Wnt/β-catenin signaling has been shown to regulate cell proliferation, cell survival, and immune modulation in multiple cancer types. These pathways need to be further validated in clinical datasets.

There is no information about lurbinectedin-mediated gene expression changes in different SCLC subtypes. Transcriptomic and proteomic analysis of models representing the most sensitive (SCLC-P) and most resistant (SCLC-A) subtypes, pre- and post-lurbinectedin treatment, showed increased expression of DNA damage markers in SCLC models post-lurbinectedin treatment.

Interestingly, activation of apoptosis pathway and suppression of TGFβ and PI3K–AKT–mTOR signaling pathways were observed exclusively in the sensitive SCLC-P model. On the other hand, in the relatively resistant SCLC-A model, lurbinectedin treatment led to an activation of the EMT pathway that has a known role in tumor progression and proliferation. Furthermore, in the SCLC-A model, we found activation of a negative regulator of the NOTCH pathway (NRARP) along with genes involved in NF-κB activation and regulation of Wnt/β-catenin pathway. Notch signaling has been known to play a tumor-suppressive role in SCLC and to counteract protumorigenic pathways like WNT/β-catenin and NF-κB in various cancers. Therefore, this study shows modulation of the apoptosis, PI3K–AKT–mTOR and TGFβ pathway to be a notable contributor toward lurbinectedin sensitivity and EMT pathway enrichment and NOTCH signaling modulation to be a contributor toward lurbinectedin resistance in SCLC models. Further studies are required to delineate the involvement of these pathways in acquired resistant models.

Lurbinectedin treatment significantly delayed tumor growth in LX110 (SCLC-A) as compared with LX1322 (SCLC-P) or LX33 (SCLC-N) models. The discrepancy of the subtype-specific effect of lurbinectedin in in vitro and in vivo models can be accounted for by (i) inherent differences between cell lines and complex PDX models that is acquired during cell culture and repeated passaging, (ii) distinct mutational profiles of the models, and (iii) heterogeneity and plasticity of SCLC patient tumors, which is not fully represented in cell lines.

GSEA analysis of pre-/posttreatment tumors indicated significant suppression of glycolytic, mTORC1, and EMT pathways in the sensitive LX110 model. Further analysis of top DEG revealed upregulation of tumor-suppressive genes of PI3K–AKT–mTOR and WNT–β-catenin pathways in LX110 and downregulation of WNT/β-catenin pathway antagonist SFRP2 gene in the less responsive models (LX33 and LX1322). Interestingly, TRIM72 and HSPB7 were exclusively upregulated in LX110 when compared with the other less responsive PDX models of LX33 and LX1322, highlighting their role in lurbinectedin response. Taken together, the in vivo de novo SCLC PDX model data suggested the crucial role of WNT/β-catenin, PI3K–AKT–mTOR, and EMT pathway suppression in inducing tumor regression, which also corroborated with our in vitro study findings.

A subset of EGFR-mutant LUAD transforms to SCLC as a mechanism of therapeutic resistance. The FLAURA study indicates SCLC transformation as a mechanism of resistance to osimertinib in patients with T790M-positive non–small cell lung cancer (NSCLC; ref. 34). Patient outcomes following transformation to SCLC are extremely poor. Clinical outcomes of patients after SCLC transformation treated with SCLC-directed therapies are similar to de novo SCLC, with high but transient response to platinum-based chemotherapy (14). As EGFR inhibitors have increased in potency, lineage plasticity is seen more frequently, as a mechanism of adaption that frees the cancer cell from EGFR signaling dependence (35). There are limited data on effective therapies for transformed SCLC with EGFR mutations in the later-line setting. In a recent study with 47 transformed patients with SCLC with EGFR mutations, a combination of chemotherapy ± bevacizumab and PD-L1 inhibitor is found to be beneficial and potentially a safe option for SCLC-transformed patients harboring the EGFR L858R mutation (36). Beyond RB1 and TP53 loss, very little is known about the biomarker landscape of transformed SCLC and their therapeutic vulnerability, thus making it difficult to treat the subset of patients with transformed SCLC that seems to be more aggressive in nature. We are the first to investigate the effect of lurbinectedin as a single agent or in combination with the EGFR inhibitor osimertinib in a transformed SCLC PDX model. In all three models, single-agent lurbinectedin treatment showed significantly more antitumor response than osimertinib treatment alone. Furthermore, combining lurbinectedin with osimertinib led to a significant delay in tumor growth, and we did not observe any measurable tumor volume in an appreciable subset of mice in two out of three models.

Gene expression analysis showed upregulation of tumor-suppressive genes of WNT/β-catenin and NOTCH signaling in transformed SCLC PDX models. Moreover, genes that are either targets or promoters of tumor-promoting pathways like NOTCH, NF-κB, or PI3K/AKT were suppressed post lurbinectedin + osimertinib treatment. It is also interesting to note that upon combination treatment in both PDX models of transformed SCLC, the PLK2 gene was upregulated, and we observed downregulation of HES1 and PHLDA2 genes opening up scope for future studies on these genes as probable biomarkers in this subset of patients.

Several reports indicate the complex molecular cross talk of NF-κB, Wnt/β-catenin, PI3K/AKT, and NOTCH signaling, which plays a crucial role in tumorigenesis. β-catenin physically interacts with and inhibits NF-κB in breast and colon cancer by forming a complex with RelA and p50 (37). In addition, NOTCH1 signaling has been reported to activate NF-κB in breast cancer in an AKT-dependent manner (38). It is interesting to note that in NSCLC, β-catenin signaling has been reported to be activated in a NOTCH3-dependent manner as a result of EGFR inhibition facilitating survival of a subset of cancer cells (39). In addition, the PI3K–AKT–mTOR pathway has been reported to confer chemo-resistance and phenotypic transition in SCLC (40). This indicates that there is an intricate cross talk among these pathways that plays a crucial role in tumorigenesis in several cancer types. In this study, significant modulation of Wnt/β-catenin, PI3K/AKT, NOTCH, and NF-κB pathway genes was observed from the DEG and pathway analysis. Future studies combining inhibitors of these pathways (i.e., gedatolisib for PI3K, Rova-T for NOTCH pathway, and sulforaphane for Wnt pathway, etc.) that are in the clinical trial with lurbinectedin may open up therapeutic avenues to increase the efficacy of lurbinectedin in less responsive patients with SCLC.

In conclusion, our data highlight the role of MYC as a predictive biomarker of lurbinectedin response. Tumor-promoting pathways like PI3K–AKT–mTOR, and EMT may play a role in lurbinectedin resistance in de novo SCLC models. This study also provides the first evidence that lurbinectedin, either as a single agent or in combination with osimertinib, causes significant tumor regression in transformed SCLC models. NOTCH plays a crucial role in the osimertinib + lurbinectedin–mediated antitumor effect.

The clinical implications of these findings are substantial, as future clinical trials of lurbinectedin, in combination with the inhibitor targeting these pathways should be investigated to overcome resistance to lurbinectedin. Moreover, we establish lurbinectedin as an effective therapeutic target in transformed SCLC, a very aggressive and recalcitrant cancer type. Because lurbinectedin is already FDA approved, the inclusion of lurbinectedin in the therapeutic regimen for EGFR-mutant LUAD cases that develop resistance to standard-of-care EGFR-targeted therapies can be readily implemented.

C.M. Rudin reports personal fees from AbbVie, Amgen, AstraZeneca, D2G, Daiichi Sankyo, Epizyme, Genentech/Roche, Ipsen, Jazz, Kowa, Lilly, Merck, Syros, and Bridge Medicines; other support from Auron, DISCO, and Earli; and personal fees from Harpoon Therapeutics outside the submitted work. T. Sen reports grants from Jazz Pharmaceuticals during the conduct of the study. No disclosures were reported by the other authors.

S. Chakraborty: Data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review, and editing. C. Coleman: Data curation, software, formal analysis, and validation. P. Manoj: Investigation. D. Demircioglu: Data curation, software, formal analysis, supervision, investigation, writing–review, and editing. N. Shah: Investigation. E. de Stanchina: Investigation. C.M. Rudin: Resources. D. Hasson: Data curation, software, formal analysis, validation, visualization, methodology, writing–review, and editing. T. Sen: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, visualization, methodology, writing–original draft, project administration, writing–review, and editing.

This work is supported by NIH/NCI R01 CA258784 (to T. Sen), Congressionally Directed Medical Research Programs (DOD-IITRA) LC190161 (to T. Sen), LCFA-BMS/ILC Foundation Young Investigator Research Awards in Translational Immuno-oncology (to T. Sen), Jazz Pharmaceuticals (to T. Sen), and NCI R01 CA197936 and U24 CA213274 (to C.M. Rudin). The development of the Bioinformatics for Next Generation Sequencing (BiNGS) shared resource facility is partially supported by the NCI P30 (P30CA196521) Cancer Center support grant, the ISMMS Skin Biology and Disease Resource-based Center NIAMS P30 support grant (AR079200), and the Black Family Stem Cell Institute. The work was also supported by the computational resources and staff expertise provided by Scientific Computing at the Icahn School of Medicine at Mount Sinai, the Office of Research Infrastructure of the NIH under award number S10OD026880, and the ISMMS Genomics Technology Facility.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.