KDM4B/DHX9 promotes chemoresistance in small-cell lung cancer through the MYCN-driven signaling pathway

Abstract

Small cell lung cancer (SCLC) is prone to developing chemoresistance, which is associated with epigenetic reprogramming. While LSD1-mediated histone demethylation has been reported, a systematic investigation into the role of histone demethylases in SCLC chemoresistance is lacking. In this study, we established nine chemoresistant cell lines from parental cells via gradual dose escalation and conducted RNA sequencing, followed by linear regression and random effects meta-analysis to identify genes associated with chemoresistance among 23 histone demethylases. We subsequently validated our findings using two neuroendocrine SCLC cell lines and their corresponding chemoresistant counterparts. We identified KDM4B as a key driver of chemoresistance, with immunohistochemical analysis revealing its elevated expression in chemoresistant tissues from SCLC patients, although its association with neuroendocrine subtypes warrants further investigation. KDM4B promoted chemoresistance through the Hedgehog pathway by enhancing cell proliferation and stemness. Mechanistically, KDM4B interacts with DHX9 and corecruits to the MYCN promoter to promote its transcription and activate the Hedgehog signaling pathway. An inhibitor of DHX9 had synergistic antitumor effects with cisplatin and etoposide, and effectively rescued the chemosensitivity of SCLC both in vitro and in vivo. These findings provide valuable insights for future studies aimed at developing therapeutic strategies to overcome chemoresistance in SCLC.

Introduction

Small cell lung cancer (SCLC) accounts for approximately 15% of all lung malignancies, with an estimated 250,000 new cases diagnosed annually worldwide [1]. SCLC is the most aggressive subtype of lung cancer, and more than 70% of patients present with extensive-stage disease at diagnosis and a 5-year survival rate of less than 7% [2]. Although the initial overall response rate (ORR) for the etoposide and cisplatin (EP) regimen ranges from 50 to 80%, most patients experience relapse within six months because of acquired chemoresistance [3, 4]. However, despite recent advancements in immunotherapy and targeted therapy that have shown promise in treating SCLC [5,6,7], progress remains slow, and the EP regimen remains a critical component of systemic treatment. Therefore, understanding the mechanisms underlying chemoresistance and developing effective therapeutic strategies remain urgent priorities.

Histone methylations are involved in transcriptional repression and activation, DNA damage and repair, and heterochromatin formation, and these modifications are associated with chemoresistance in SCLC. As a polycomb repressive complex 2 (PRC2) component, the histone methyltransferase EZH2 is responsible for adding methyl groups to lysine 27 of histone H3 (H3K27me3), a modification that mediates transcriptional repression [8, 9]. Importantly, EZH2 inhibitors reverse chemoresistance in SCLC by promoting the expression of SLFN11 and CDKN1C, genes that are specifically involved in DNA damage repair and cell cycle arrest [10, 11]. Histone demethylation, a reversible process of histone methylation, can promote tumor progression by activating oncogenes or by inhibiting the expression of tumor suppressor genes, thereby affecting genome stability and other signaling pathways [12]. Recent studies have shown that inhibitors of lysine-specific histone demethylase 1 A (LSD1) induce complete and durable tumor regression by activating the NOTCH signaling pathway, thereby suppressing the expression of the neuroendocrine cell lineage-specific transcription factor ASCL1 and related genes [13]. However, the clinical trial of the LSD1 inhibitor GSK2879552 in SCLC was terminated because of associated severe adverse events (including encephalopathy in four patients) and limited efficacy [14]. Currently, few systematic studies have explored the role of the histone demethylase family in chemoresistance in SCLC.

Previous studies have linked KDM4B to tumor development and treatment resistance across various cancers. KDM4B/JMJD2B, a JmjC domain-containing histone lysine demethylase, comprises JmjC and JmjN domains, two PHD domains, and two Tudor domains. The catalytic JmjC domain binds bivalent iron and, together with cofactors O₂ and α-KG, specifically demethylates H3K9me2/3 and H3K36me2/3, thereby modulating chromatin accessibility and gene expression patterns essential for stem cell biology and organogenesis [15, 16]. KDM4B promotes human embryonic stem cell self-renewal and differentiation while regulating adipocyte and osteocyte development [17,18,19]. However, its dysregulation contributes to oncogenesis in multiple cancers, including prostate, neuroblastoma, gastric, colorectal, and liver cancers [20]. Mechanistically, KDM4B drives tumorigenesis by activating oncogenes or silencing tumor suppressors. In gastric cancer, KDM4B promotes epithelial-mesenchymal transition through the formation of transcriptional complexes with c-Jun or β-catenin [21]. In renal carcinoma, KDM4B modulates cell cycle progression [22], whereas in colorectal cancer, it maintains autophagy-derived amino acid pools under glucose deprivation [23]. KDM4B also contributes to drug resistance through multiple mechanisms: by regulating topoisomerase II accessibility in breast cancer cells, thus affecting anthracycline sensitivity [24]; by enhancing AR-dependent c-Myc transcription in prostate cancer, thereby contributing to resistance to androgen deprivation therapy [25, 26]; and by interacting with eIF2α to inhibit its phosphorylation, thereby causing PI3K-AKT inhibitor resistance in PTEN-deficient breast cancer [27]. Nevertheless, the role of KDM4B in the development and relapse of SCLC remains unexplored.

While KDM4B is implicated in tumor progression and resistance, no specific inhibitors are currently available for clinical targeting of KDM4B, making the investigation of transcriptional regulatory complexes a promising alternative approach. Existing KDM4B inhibitors can be categorized into three classes on the basis of their mechanisms of action: KDM4B cofactor inhibitors and antagonists, the JmjC domain inhibitors, and substrate inhibitors [16, 28]. However, since histone demethylases containing the Jumonji C domain share similar structures and functions, targeting KDM4B may inadvertently lead to widespread inhibition of histone demethylation, potentially causing unwanted off-target effects. Alternatively, histone methylation is crucial for regulating gene transcription and requires the assistance of transcriptional regulatory complexes. For instance, the histone methyltransferase MLL1 catalyses H3K4 methylation and features a C-terminal SET domain that is activated by a protein complex known as WRAD [29]. The WRAD complex consists of two β-propeller proteins, namely, WDR5 and RbBP5, the SPRY/DNA-binding domain protein Ash2L, and Dpy30. WDR5, RbBP5, and Ash2L are essential for catalytic activation and stabilization of the active conformation of the SET domain [30]. WDR5 acts as an epigenetic scaffold, and identification of its binding domains has led to early advancements in the development of WDR5 degraders and functional inhibitors [31]. Similarly, KDM4B requires cofactors to regulate gene transcription. Therefore, we aimed to investigate key cofactors that interact with KDM4B to specifically regulate gene expression and explore whether inhibitors targeting these cofactors can increase chemosensitivity in SCLC.

In this study, we systematically investigated demethylated proteins associated with chemoresistance in SCLC to elucidate the underlying mechanisms and identify potential reversal strategies. First, we generated chemoresistant counterparts from nine chemosensitive cell lines through long-term dose escalation. Then, using RNA-seq data from six paired cell lines, we applied linear regression and random-effects meta-analysis to identify genes associated with treatment response among the 23 histone demethylases. Next, we selected two neuroendocrine cell lines and corresponding chemoresistant cells for subsequent experimental verification and mechanism exploration. Finally, we used inhibitors targeting molecules that interact with demethylated proteins in both in vitro and in vivo models to assess their efficacy in reversing SCLC chemoresistance.

Materials and methods

Cells and cell culture

A total of 9 human SCLC cell lines were used in this study, namely, H69, H209, H1688, H146, H446, H82, H526, H1339, and DMS114. We obtained gene expression data for 72 SCLC cell lines from the Genomics of Drug Sensitivity in Cancer (GDSC) database [32] and classified the cell lines into different subtypes on the basis of the expression of the transcription factors ASCL1, NeuroD1, YAP1 and POU2F3 [33]. DMS114 cells and their corresponding chemoresistant cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM), while all the other cells were cultured in RPMI-1640 medium (Nanjing Keygen Biotech), both supplemented with 10% foetal bovine serum (FBS) and 1% penicillin/streptomycin, and all the cells were incubated at 37 °C with 5% CO₂. When the cells reached 80–90% confluence (approximately 0.8–1.0 × 10⁵ cells/cm²), they were passaged or seeded accordingly. Except for the multidrug-resistant cell line H69AR (derived from H69 by doxorubicin induction), which was obtained from the American Type Culture Collection (ATCC, USA) [34], all the other chemoresistant cell lines were established by culturing the parental cells in complete medium supplemented with increasing concentrations of cisplatin or etoposide over up to 12 months, with concentrations ranging from 0.2 to 0.6 μg/mL. All the cell lines used in this study were confirmed to be mycoplasma-free and were routinely monitored through morphological observations and growth kinetic analyses. Experiments were conducted using cells at passages 3–20 after thawing to ensure consistent cellular characteristics.

Cell transfection and infection

Cells were transfected with siRNAs targeting KDM4B or DHX9 and the corresponding negative control siRNA (Tsingke, China) using PepMute™ siRNA Transfection Reagent (SignaGen Laboratories) according to the manufacturer’s protocol to achieve transient knockdown of the target genes. The human wild-type KDM4B (KDM4B-WT) and mutant KDM4BΔJmjC (G494C, G545C) genes were inserted into the pcDNA3.1-3×Flag vector to generate recombinant plasmids [35, 36], which were then transfected into cells using Genjet™ Plus DNA In Vitro Transfection Reagent (SignaGen Laboratories). Additionally, lentiviral vectors harbouring shKDM4B or the KDM4B sequence were used to infect cells to generate stable transfectants (GeneCopoeia, Inc., Rockville, MD, USA). After 2–3 days of stable proliferation, puromycin (2 μg/mL) was used to select successfully transduced cells. Cells with confirmed knockdown or overexpression of KDM4B, as verified by qPCR and immunoblotting, were expanded and utilized for subsequent experiments. The sequences of all the primers used for the siRNAs and shRNAs are provided in Supplementary Table S1.

SCLC cohorts

A total of four public SCLC cohorts, GSE149507, GSE60052, GSE30219 and TU-SCLC, were included in this study. GSE149507 includes microarray expression data from 18 pairs of SCLC and adjacent lung tissue obtained from surgical resection [37]. GSE60052 includes RNA sequencing data from 79 cancer and 7 normal tissues from patients with SCLC [38]. GSE30219 includes expression profiles of 293 lung tumors (including lung adenocarcinoma, lung squamous cell carcinoma, large cell neuroendocrine carcinoma, and SCLC) and 14 normal lung tissues [39].

The TU-SCLC cohort from Shanghai Pulmonary Hospital, Tongji University, includes 112 treatment-naive primary SCLC tumors and paired normal adjacent tissues, and includes transcriptome and proteome data from these tissues, as well as clinical information for the corresponding patients [40]. The psych package was used to perform Pearson correlation analysis, and the survival and survminer packages were used to perform univariate and multivariate Cox regression analyses of the data, where the optimal cut-off value for grouping was defined by the surv_cutpoint function.

RNA extraction and sequencing

Following the confirmation of the responsiveness of chemosensitive cells and their induced chemoresistant cells to cisplatin and/or etoposide through cell viability assays, the cells were harvested for RNA sequencing. The detailed procedure was as follows: (1) Cell collection and lysis: For each cell type, three biological replicates were prepared. Cells in the logarithmic growth phase were collected, thoroughly lysed in TRIzol reagent (Invitrogen), and stored at −80 °C. (2) RNA extraction and quality control: Total RNA was extracted using a magnetic bead-based universal RNA extraction kit (Beijing Meijimei Biotechnology Co., R6623-02). RNA quality was assessed using a NanoDrop 2000 spectrophotometer (for purity and concentration) and a LabChip GX Touch system (for integrity, as measured by the RNA integrity number, RIN). Only samples with an RNA concentration ≥ 30 ng/μL, total amount ≥ 0.5 μg, and RIN ≥ 4 were used for library construction. (3) Library preparation: Sequencing libraries were constructed using the VAHTS Universal V6 RNA-seq Library Prep Kit for Illumina (Vazyme, NR604) and VAHTS mRNA Capture Beads 2.0 (Vazyme, N403-01). (4) Sequencing: The libraries were sequenced on an Illumina platform to generate 150 bp paired-end reads. The sequencing depth exceeded 6 Gb per sample, with >85% of bases achieving a Q30 quality score. (5) Bioinformatic processing: Raw reads were quality-checked with FastQC (v0.11.9) and adapter-trimmed using Trim Galore (v0.6.7). High-quality reads were aligned to the GRCh38 human reference genome (Ensembl release 107) using HISAT2 (v2.2.1). Gene-level read counts were obtained using featureCounts (Subread v2.0.1). Counts were normalized to TPM (transcripts per million), and differential expression analysis was performed using the limma package (v3.54.2) in R.

RNA extraction and quantitative real-time PCR

The RNA of cells under experimental conditions was extracted using the TRIzol method. RNA purity and concentration were assessed by measuring the optical density at 260 nm, 280 nm, and 230 nm using a spectrophotometer. Only samples with A₂₆₀/A₂₈₀ ratios between 1.8 and 2.1 and A₂₆₀/A₂₃₀ ratios ≥2.0 were used for downstream analysis. For reverse transcription, 1000 ng of total RNA was reverse transcribed into cDNA in a 20 μL reaction volume using 2x EZ Color SYBR Green QPCR Master Mix (EZBioscience, USA). Quantitative real-time PCR (qRT–PCR) was performed using a Color All-in-one Reverse Transcription Kit (EZBioscience, USA) on a Bio-Rad CFX Real-Time PCR System. The primer sequences of all the genes are provided in Supplementary Table S2.

Immunoblot analysis

Cells were lysed using RIPA buffer containing protease and phosphatase inhibitors, and protein concentrations were determined using a BCA assay (Nanjing Keygen Biotech). A protein concentration of 1 μg/μL was prepared with loading buffer containing SDS and reducing agents. The proteins were separated by 12% SDS-polyacrylamide gel electrophoresis and transferred onto PVDF membranes (Millipore, USA). Following transfer, the membranes were blocked with 5% nonfat milk in Tris-buffered saline containing 0.1% Tween 20 (TBST) for 2 h at room temperature. The membranes were then incubated overnight at 4 °C with the appropriate primary antibodies, followed by incubation with horseradish peroxidase (HRP)-conjugated secondary antibodies for 1 h at room temperature. Finally, the protein bands were visualized using enhanced chemiluminescence (Yazyme, China) and imaged with a NiceAlliance Q9 system. The band intensities were quantified using ImageJ software and normalized to that of GAPDH as the loading control. All primary antibodies used in this study are listed in Supplementary Table S3.

Immunofluorescence staining

The cells were fixed with 4% paraformaldehyde for 30 min at room temperature, followed by permeabilization with 0.3% Triton X-100 in PBS for 30 min. After blocking with 5% goat serum for 1 h at room temperature, the cells were incubated with primary antibodies overnight at 4 °C. After three washes with PBS, the cells were incubated with fluorophore-conjugated secondary antibodies for 1 h at room temperature in the dark. The nuclei were counterstained with DAPI for 5 min. After the final washes, the cells were visualized and imaged using a confocal laser scanning microscope (Nikon, Japan) with a 100× oil immersion objective lens.

Immunohistochemical staining

The fixed and embedded tumor tissue was cut into 4 μm-thick sections for target protein staining. For staining, the sections were first incubated with the primary antibody overnight at 4 °C, followed by a 2-h incubation with the corresponding secondary antibody. Signal amplification was achieved using horseradish peroxidase-labeled avidin, followed by the application of a DAB-H2O2 substrate for color development. Finally, nuclear staining, dehydration, and sealing were performed to complete the preparation.

For SCLC molecular subtyping, the dominant subtypes were determined using the histoscore (H-score) quantification method. The H-score was calculated as H-score = Σ(i × P_i), where i represents staining intensity (1 = weak, 2 = moderate, 3 = strong) and P_i represents the percentage of positive tumor cells (0–100%) at each intensity level, yielding scores ranging from 0 to 300 [41]. Tumors were classified according to the transcription factor with the highest H-score, with samples demonstrating comparable expression levels of multiple markers designated composite subtypes.

Co-immunoprecipitation (Co-IP)

Co-IP was performed following the manufacturer’s instructions of the Pierce™ Crosslink Magnetic IP/Co-IP Kit (Thermo Scientific, USA). Approximately 6 × 10⁶ cells were collected and gently lysed on ice with 500 μL of IP lysis buffer by slow rotation for 5 min to ensure efficient protein extraction. The protein concentration was determined using the BCA assay and adjusted to 1000 μg/μL for subsequent experiments. 20 μL of the lysate was reserved as the input control. Next, 40 μL of Protein A/G magnetic beads were washed three times with wash buffer, resuspended, and incubated with 5 μg of target protein antibody in a total volume of 100 μL at room temperature for 15 min to form bead–antibody complexes. Afterwards, 500 μL of the prepared protein lysate was added and incubated overnight at 4 °C with gentle agitation to capture antibody-bound proteins. The bead-bound immune complexes were then washed three times with wash buffer to remove nonspecific proteins. The immunocomplexes were eluted using 100 μL of elution buffer, mixed with 25 μg of 5× loading buffer, and denatured at 98 °C for 10 min. These prepared samples were ultimately analysed by mass spectrometry and immunoblotting to assess protein interactions.

Chromatin immunoprecipitation (ChIP)

ChIP was performed using the Chromatin Immunoprecipitation (ChIP) Kit, Protein A/G Magnetic Beads (FI8903, Fitgene, China) according to the manufacturer’s protocol. Briefly, 3 × 10⁷ cells were collected and crosslinked with 1% formaldehyde in PBS at room temperature for 10 min, then quenched with 1.375 M glycine for 5 min. After being washed with cold PBS, the cells were lysed in 500 μL of lysis buffer containing protease inhibitors and sonicated to shear the chromatin into 100–1000 bp fragments. The lysate was cleared by centrifugation at 4 °C, with 30 μL of the supernatant saved as the protein input and another 30 μL as the DNA input control. The remaining supernatant was incubated overnight at 4 °C with 5 μg of bait protein antibody or normal IgG. The next day, 40 μL of protein A/G magnetic beads was added and incubated at 4 °C for 2 h. Bead-antibody-chromatin complexes were washed three times with wash buffer, then eluted with elution buffer. The cross-links were reversed at 65 °C for 6 h, followed by RNase A and proteinase K digestion. DNA was purified by phenol–chloroform extraction and ethanol precipitation, then dissolved in nuclease-free water for downstream analysis, including qPCR or sequencing. The relevant primer sequences are presented in Supplementary Table S4.

Cell viability

Cell viability was assessed using the CCK-8 reagent (EZBioscience, USA). The cells were seeded at a density of 1×10⁴ cells per well. The next day, treatments with or without drugs were applied, and the cells were cultured continuously. On the third day, the medium was replaced with fresh medium containing 10% CCK-8 reagent, and the cells were incubated for 1 to 4 h. The absorbance was measured at 450 nm using a BioTek Synergy H1 microplate reader (Agilent, USA). The half-maximal inhibitory concentration (IC50) was calculated relative to that of the untreated control group.

Colony formation assay

For adherent cells, colony formation assays were performed, whereas for suspended cells, soft agar colony formation assays were used. Cells were seeded onto 6-well plates at the specified density and cultured for 10–14 days in complete medium supplemented with 10% serum. During this period, appropriate drug treatments were applied according to the experimental objectives. Clone sizes were monitored under a microscope, and samples were collected once the clones grew to more than 50 cells. Each well was fixed with paraformaldehyde for 30 min, followed by incubation with crystal violet dye at room temperature for 30 min. Finally, the plates were rinsed with water, air-dried, and photographed, and the number of clones was quantified using ImageJ software.

Sphere formation assays

Sphere formation assays were conducted using serum-free medium supplemented with 0.25% methylcellulose solution with multiple growth factors. The medium contained 1x B27 supplement, 6.4 μg/mL insulin, 10 ng/mL basic fibroblast growth factor (bFGF), and 20 ng/mL epidermal growth factor (EGF). Cells were seeded into low-adsorption 24-well plates at a density of 1000 cells per well with 300 μL of medium, and cultured for 10–14 days. Fresh medium (30–50 μL) was gently added every other day to maintain nutrient levels. The sphere formation efficiency (SFE) was calculated by the formula SFE = (number of spheres with a diameter greater than 75 μm in each well/number of cells planted) × 100%.

SCLC cell line-derived xenografts and allografts

The animal studies were approved by the Animal Care and Use Committee of Southern Medical University (LAEC-2022-186FS and LAEC-2024-258). For the xenograft models, 5 × 10⁶ SCLC cells were injected subcutaneously into 4-week-old female BALB/c nude mice, and when the tumor volume reached approximately 100 mm³, the mice were randomly assigned to treatment groups with five mice per group; the tumor size was measured every 4 days using calipers and the tumor volume was calculated as (length × width²)/2. For the allograft model, 1 × 10⁶ RP-luc SCLC cells—stably transfected with a luciferase reporter gene and derived from spontaneous murine SCLC induced by intratracheal instillation of replication-deficient adenoviruses expressing Cre-recombinase in Rb1^flox/flox and Trp53^flox/flox mice—were orthotopically implanted into 5-week-old female C57BL/6 mice via direct visualization thoracic surgery, with luciferase expression enabling in vivo fluorescent imaging; once stable lung fluorescence signals were detected, the mice were randomized into groups (n = 6) receiving weekly treatments of either chemotherapy (cisplatin 2 mg/kg intraperitoneally on day 1 and etoposide 4 mg/kg intraperitoneally on days 1–3) or DHX9 inhibitors (2 mg/kg intraperitoneally on days 1–5), with tumor progression monitored weekly by fluorescence imaging.

Hierarchical clustering dendrogram

The amino acid sequences of 23 demethylation-modified proteins were obtained from the UniProtKB database, after which multiple sequence alignment was performed to assess their similarity. A phylogenetic tree was subsequently constructed based on the calculated alignment distances, and the ggtree R package was used to visualize the hierarchical clustering dendrogram.

Differential expression analysis

The limma package was used to analyse differences in gene expression between chemosensitive and chemoresistant cells. Genes were classified as resistance-related if they exhibited an absolute log fold change (logFC) greater than 0.1 and a p-value less than 0.05. Here, the fold change (FC) refers to the ratio of gene expression in resistant cells to that in sensitive cells. Additionally, linear regression analyses were conducted to compare different cell types with their respective resistant counterparts. A random effects model was then applied for meta-analysis to integrate the results across multiple cell lines, enabling the identification of genes that were consistently dysregulated. The false discovery rate (FDR) method was used to correct for multiple testing.

Gene set enrichment analysis

Gene set enrichment analysis (GSEA) was performed on the lists of differentially expressed genes ranked by log2FC using the clusterProfiler R package to identify enriched signaling pathways in each group. The Kyoto Encyclopedia of Genes and Genomes (KEGG) gene set was obtained from the MSigDB molecular signature database for GSEA.

Statistical analysis

The results were analysed and visualized using R or GraphPad Prism 9.5. Each experiment was conducted in triplicate, and the results are presented as the mean ± standard deviation (SD). For comparisons between two groups, a two-tailed t-test was used when the data were normally distributed; otherwise, the nonparametric rank-sum test was applied. For comparisons involving three or more groups, one-way ANOVA was used for normally distributed data, followed by Tukey’s post hoc test for pairwise comparisons. Asterisks denote statistical significance: * for p < 0.05, ** for p < 0.01, and *** for p < 0.001.

Results

KDM4B is associated with chemoresistance in SCLC

The expression profiles of four transcription factors in 72 SCLC cell lines obtained from the GDSC database enabled the classification of nine SCLC cell lines into four major subtypes: SCLC-A (H69, H209, H1688, and H146), SCLC-N (H446 and H82), SCLC-Y (H1339 and DMS114), and SCLC-P (H526), highlighted in red font in Supplementary Fig. S1A. To identify key genes associated with acquired chemoresistance, we subjected 9 parental SCLC cell lines to long-term exposure to low concentrations of chemotherapy (Fig. S1B). For example, H446 cells treated with cisplatin (DDP) were labeled with H446DDP, while those treated with etoposide (VP-16) were designated H446VP. Exceptionally, H69AR, which is resistant to multiple drugs, was sourced from the American Type Culture Collection and is the chemoresistant counterpart to H69.

A total of 23 histone demethylation proteins were included to explore their potential role in chemoresistance systematically (Fig. 1A). By analyzing six established pairs of chemosensitive parental cell lines and their corresponding chemoresistant counterparts, we identified five genes: KDM1A, KDM1B, KDM3A, KDM4B, and KDM6B, all of which satisfied the criteria of logFC > 0.1 and p < 0.05 in at least four paired cells (Fig. 1B). Subsequent linear regression analysis and random-effects meta-analysis further revealed that the expression levels of KDM4B (β = 0.295, OR = 1.34, p < 0.001) and KDM3A (β = 0.542, OR = 1.72, p = 0.009) were significantly associated with chemoresistance (Figs. 1C and S1C), whereas the expression level of KDM6B was marginally associated (β = 0.262, OR = 1.30, p = 0.061). After adjustment for multiple testing by FDR correction, only KDM4B remained statistically significant (FDR-p = 0.003). Notably, KDM4B expression was consistently upregulated across all six pairs of cell lines, in contrast to the variable expression pattern of KDM3A (Fig. S1D).

Fig. 1: KDM4B is highly expressed in chemoresistant cells and tissues of SCLC.

A Phylogenetic tree of histone demethylation proteins, illustrating the similarities between different proteins. B The forest plot summarizes the summary results of genes highly expressed in chemoresistant cells across 6 pairs of cells. C The forest plot shows linear regression analysis results of KDM4B expression in 6 cell pairs, along with summary estimates derived from a random effects model. RE Model, random effects model; OR, odds ratio. D, E qPCR results showing the expression levels of KDM4A and KDM4B in 9 pairs of chemosensitive and corresponding chemoresistant SCLC cell lines. F, G Immunoblot analyses demonstrating KDM4B protein expression in 9 pairs of chemosensitive and corresponding chemoresistant SCLC cell lines. S, Sensitive; R, Resistant. H Representative IHC staining of KDM4B in SCLC tissues, with statistical quantification presented. ***, p < 0.001; **, p < 0.01; *, p < 0.05.

Given the structural similarity between KDM4A and KDM4B, we investigated whether their expression differed between chemosensitive and corresponding chemoresistant SCLC cells. Quantitative PCR and immunoblot analyses revealed that KDM4B expression was significantly upregulated across multiple chemoresistant SCLC cell lines representing the SCLC-A, SCLC-N, and SCLC-Y subtypes. In contrast, KDM4A expression was significantly increased only in chemoresistant cells derived from H209 and DMS114 (Figs. 1D–G and S2). Immunofluorescence staining confirmed elevated KDM4B levels in resistant cells compared with their parental counterparts, with predominant nuclear localization (Fig. S3). Furthermore, immunohistochemical analysis of tumor tissues from SCLC patients revealed significantly increased KDM4B expression in chemoresistant samples (Fig. 1H). However, owing to the limited clinical sample size, a definitive correlation between KDM4B expression and specific SCLC molecular subtypes could not be established in this cohort (Fig. S4A and Supplementary Table S5). Notably, analysis of the TU-SCLC cohort revealed significant positive Pearson correlations between KDM4B expression and the levels of ASCL1 (r = 0.22, p = 0.036) and NEUROD1 (r = 0.3, p = 0.004), whereas no such significant correlation was observed in the GSE60052 dataset (ASCL1: r = 0.14, p = 0.21; NEUROD1: r = 0.10, p = 0.39. Fig. S4B, C). Collectively, these results indicate that KDM4B is highly expressed in chemoresistant tissues and cells and may be associated with SCLC chemosensitivity; however, its relationship with molecular subtypes remains unclear.

KDM4B is highly expressed in SCLC and is associated with shorter overall survival

Next, the expression of KDM4B was investigated using sequencing data from paired or unpaired SCLC and adjacent lung tissues. Based on the structural similarity of KDM4A, KDM4B, and KDM4C (Figs. 1A and 2A), we examined the expression levels of all three molecules in tumor and normal tissues. In the paired normal and SCLC tissues of GSE149507, GSE60052, and TU-SCLC, both KDM4A and KDM4B were significantly overexpressed in SCLC (p < 0.001) (Figs. 2B, C, and S5A–D), and similar trends were confirmed at the protein level (Fig. 2D), while KDM4C was not significant. In addition, in the GSE30219 dataset containing normal and various lung tumor tissues, KDM4B was significantly overexpressed specifically in SCLC (p < 0.001), while no significant increase was observed in other lung tumor types (Fig. 2E). Taken together, these results indicate that KDM4A and KDM4B are elevated in SCLC and may be involved in the occurrence and development of SCLC.

Fig. 2: KDM4B is highly expressed in SCLC and is associated with shorter overall survival.

A Schematic illustration of the domain structure of the KDM4 family. B Gene expression of KDM4B in paired SCLC and normal lung tissues from the GSE149507 and GSE60052 datasets. C KDM4B gene expression in paired normal and tumor tissue from the TU-SCLC cohort. D KDM4B protein expression in paired normal and tumor tissue from the TU-SCLC cohort. E Gene expression of KDM4B across normal lung tissues and various lung cancer subtypes in the GSE30219 dataset. NTL non-tumoral lung, ADC adenocarcinoma, SQC squamous cell carcinoma, LCNE large cell neuroendocrine tumors, SCLC small cell lung cancer. F, G Univariate Cox regression analysis results showing the associations between KDM4B gene and protein expression and overall survival in SCLC patients. H, I Multivariate Cox regression analysis results incorporating KDM4B gene or protein expression together with clinical factors, including sex, smoking, TNM stage, pleural invasion, vascular invasion, and neural invasion. ***, p < 0. 001; **, p < 0. 01; *, p < 0. 05.

Although KDM4A and KDM4B exhibit similar expression patterns in SCLC, our previous findings indicate that KDM4B is more likely to be involved in chemoresistance. Univariate Cox regression analysis of KDM4B gene expression in the TU-SCLC cohort revealed that higher KDM4B expression was significantly associated with shorter overall survival (OS) in patients (HR = 2.34, 95% CI [1.33–4.13], p = 0.0032) (Fig. 2F). Multivariate Cox regression analysis adjusted for potential confounding factors including sex, smoking, TNM stage, pleural invasion, vascular invasion, and neural invasion, confirmed that high KDM4B gene expression remained an independent risk factor for poor OS (HR = 2.301, 95% CI [1.226–4.320], p = 0.009) (Fig. 2H). A similar analysis using KDM4B protein expression revealed a trend consistent with the gene-level results, although the p value was not statistically significant (univariate Cox analysis, HR = 2.83, 95% CI [0.88–9.12], p = 0.08; multivariate Cox analysis, HR = 3.28, 95% CI [0.91–11.85], p = 0.07) (Fig. 2G, I), indicating that future studies with larger SCLC patient cohorts are warranted to further validate the prognostic value of KDM4B protein levels. Additionally, correlation analysis between protein and gene expression of KDM4B revealed a moderate positive correlation (Spearman’s r = 0.31, p = 0.0019), suggesting that protein levels may be modulated by factors beyond gene transcription, such as protein translation, post-translational modifications, and degradation processes (Fig. S5E).

KDM4B promotes chemoresistance in SCLC

The SCLC-A and SCLC-N subtypes account for approximately 70% of SCLC cases. The chemoresistant H69AR and H446DDP cells, corresponding to H69 and H446, respectively, are resistant to both cisplatin and etoposide and were selected for subsequent experiments. KDM4B knockdown cell models were established using siRNA and shRNA, and the efficiency of gene downregulation was verified by qPCR and immunoblotting (Figs. 3A, B and S6A, B). Cell viability assays demonstrated that KDM4B downregulation significantly increased sensitivity to cisplatin and etoposide in H69AR and H446DDP cells (Fig. 3C, D). Conversely, KDM4B was overexpressed in the chemosensitive parental H69 and H446 cells (Figs. 3E and S6C), and subsequent viability assays revealed that increased KDM4B expression significantly enhanced resistance to both chemotherapeutic agents (Fig. 3F). Furthermore, KDM4B promoted chemoresistance in another SCLC-N subtype cell line, H82, and its chemoresistant counterpart H82DDP (Fig. S7).

Fig. 3: KDM4B promotes chemoresistance in SCLC.

A, B Immunoblotting showing the knockdown efficiency of KDM4B in cell lines with transient and stable knockdown. C, D Cell viability assays showing the IC50 values of cisplatin and etoposide in H69AR and H446DDP cells and their respective KDM4B knockdown derivatives. IC50, the half-maximal inhibitory concentration. E Immunoblotting showing the overexpression efficiency of KDM4B in H69 and H446 cells. F Cell viability assays showing the IC50 values of cisplatin and etoposide in H69 and H446 cells and their KDM4B-overexpressing counterparts. G Growth curves of subcutaneous xenografts in the groups in which KDM4B expression was downregulated and the control group. H, I Anatomical images and weights of subcutaneous xenografts from the KDM4B downregulation group and the control group at the experimental endpoint. J Growth curves of subcutaneous xenografts in the KDM4B overexpression and control groups. K, L Anatomical images and weights of subcutaneous xenografts in the KDM4B overexpression group and the control group at the experimental endpoint. ***, p < 0. 001; **, p < 0. 01; *, p < 0. 05; ns, not significant.

To further validate the role of KDM4B in vivo, we used subcutaneous xenograft tumor models. We observed that tumors derived from H69AR cells with KDM4B knockdown exhibited significantly slower growth than tumors derived from control cells following chemotherapy treatment (Fig. 3G). However, there was no significant difference in the tumor growth between the KDM4B knockdown and control groups when chemotherapy was not administered. At the end of the experiment, the tumors were excised and weighed, revealing enhanced chemosensitivity in KDM4B-knockdown tumors (Fig. 3H, I). Additionally, we established subcutaneous xenografts using parental H69 cells and H69 cells overexpressing KDM4B. After chemotherapy, tumors formed by KDM4B-overexpressing cells grew significantly faster than those formed by control H69 cells (Fig. 3J). Correspondingly, the volume and weight of tumors harvested from KDM4B-overexpressing cells were substantially greater than those from control cells (Fig. 3K, L). Furthermore, IHC staining confirmed KDM4B expression in the xenograft tumors (Fig. S8). Collectively, these findings support that KDM4B contributes to chemoresistance in SCLC.

KDM4B promotes the Hedgehog signaling pathway

To elucidate the mechanism by which KDM4B contributes to chemoresistance, we performed RNA sequencing on KDM4B-downregulated and negative-control cells. Notably, pathway analysis revealed significant downregulation of both the Hedgehog and NOTCH signaling pathways following KDM4B knockdown (Fig. 4A). A detailed examination revealed that the genes in the pathway, the ligands (DHH), receptors (PTCH1 and PTCH2) and coreceptor GAS1 in the Hedgehog pathway, and the ligands (JAG1), receptors (NOTCH2, NOTCH3, NOTCH4) and effector molecules (HEY1, HES1, HES5) in the NOTCH pathway were downregulated (Fig. 4B). Furthermore, the expression of ligands (WNT5A and WNT5B) in the WNT signaling pathway was also reduced after KDM4B knockdown. Most of these transcriptomic changes were validated by qPCR (Figs. 4C, D and S9), supporting the hypothesis that KDM4B may contribute to chemoresistance through cross-talk among these critical signaling pathways. Despite the significant downregulation of Hedgehog signaling, its precise role in SCLC chemoresistance remains to be fully defined.

Fig. 4: KDM4B regulates the Hedgehog signaling pathway.

A Summary of signaling pathways significantly associated with KDM4B, including upregulated and downregulated pathways. B Heatmap showing transcriptomic changes in the Hedgehog and NOTCH pathways after KDM4B knockdown. C, D The results of qPCR show the expression of PTCH1/2 and GLI1 after KDM4B knockdown. E The immunoblotting results show the expression of PTCH1 and GLI1 following KDM4B knockdown. F The results of the clonogenic assays in H69AR/H446DDP and corresponding KDM4B knockdown cells. G The results of the sphere formation assay in H69AR/H446DDP and corresponding KDM4B knockdown cells. ***, p < 0. 001; **, p < 0. 01; *, p < 0. 05.

In the KDM4B-knockdown H69AR and H446DDP cells, immunoblotting revealed that the expression of genes involved in the Hedgehog signaling pathway, including the receptors PTCH1 and PTCH2 and the effector transcription factor GLI1, was significantly downregulated (Fig. 4E). Given that activation of the Hedgehog pathway is known to promote cell proliferation and maintain stemness characteristics, we conducted clonogenic assays and sphere formation experiments to further evaluate the role of KDM4B. The results revealed a significant reduction in both the number of colonies and spheres formed following KDM4B knockdown in H69AR and H446DDP cells compared with those in the control groups (Fig. 4F, G). The above results indicate that KDM4B regulates the Hedgehog signaling pathway and promotes the proliferation and stemness of SCLC cells.

KDM4B promotes chemoresistance via the Hedgehog signaling pathway

We next utilized two SMO inhibitors of the Hedgehog pathway, cyclopamine (Cyclo) and sonidegib (Soni), to investigate whether KDM4B promotes chemoresistance through this pathway. At concentrations of 10 μM for Cyclo and 20 μM for Soni, the survival rates of H69/H69AR and H446/H446DDP cells remained above 90% (Fig. 5A, B). Treatment of H69AR and H446DDP cells with these inhibitors for 24 h effectively inhibited the pathway, as demonstrated by the downregulation of PTCH1, PTCH2, and GLI1 expression, whereas KDM4B expression remained unaffected (Figs. 5C, D and S10).

Fig. 5: KDM4B promotes chemoresistance via the Hedgehog signaling pathway.

A, B Inhibition rates of various concentrations of SMO inhibitors on H69/H69AR and H446/H446DDP cells. C, D Effects of SMO inhibitors on the expression of receptors and effectors in the Hedgehog pathway. E Effect of SMO inhibitors on cell proliferation in H69AR and H446DDP cells. F Effects of SMO inhibitors on cell proliferation in rescue experiments using H69 and H446 cells. G Influence of SMO inhibitors on the sphere formation efficiency of H69AR and H446DDP cells. H Effects of SMO inhibitors on sphere formation efficiency in rescue experiments using H69 and H446 cells. I Effect of SMO inhibitors on the chemosensitivity of H69 and H446, as well as corresponding KDM4B-overexpressing cells. IC50, the half-maximal inhibitory concentration. ***, p < 0. 001; **, p < 0. 01; *, p < 0. 05.

We found that treatment with Cyclo or Soni significantly decreased the number of clones formed by H69AR and H446DDP cells (Fig. 5E). In H69 and H446 cells, Cyclo or Soni significantly alleviated the increase in clone number induced by KDM4B overexpression (Fig. 5F). In addition, Cyclo or Soni significantly suppressed the spheroidization rate of H69AR and H446DDP cells (Fig. 5G). In rescue experiments using H69 and H446 cells, Cyclo or Soni effectively mitigated the increase in the spheroidization rate caused by KDM4B overexpression (Fig. 5H). Importantly, in H69 and H446 cells overexpressing KDM4B, Cyclo or Soni significantly attenuated the chemoresistance mediated by KDM4B overexpression (Fig. 5I). These results collectively suggest that KDM4B promotes chemoresistance by activating the Hedgehog signaling pathway, thereby enhancing cell proliferation and stemness. However, whether Hedgehog signaling pathway inhibitors can increase the chemosensitivity of SCLC in vivo remains to be further validated.

KDM4B binds to the MYCN promoter and promotes MYCN expression

To investigate the mechanisms by which KDM4B specifically regulates the Hedgehog signaling pathway, we performed ChIP-seq analysis targeting the KDM4B protein to identify downstream molecules directly regulated by it. Our analysis revealed that KDM4B predominantly binds to intronic regions (44%), followed by intergenic regions (24%) and the 2k-base pair region upstream of gene transcription start sites (17%) (Fig. 6A). By integrating KDM4B binding sites with gene expression data, we identified 8 genes that were potentially upregulated and 11 genes that were potentially downregulated by direct binding to KDM4B (Fig. 6B). We then selected several protein-coding genes for validation and found that knockdown of KDM4B significantly increased SMARCD1 expression and decreased GALNT1, MYCN, TUBD1, and SLC35D1 expression (Fig. 6C, D). Examination of ChIP-seq peaks revealed enrichment of KDM4B at the promoters of MYCN, GALNT1, and SLC35D1, whereas enrichment of gene bodies was detected mainly for TUBD1 and SMARCD1 (Figs. 6E and S11A–C). In support of this, RNA sequencing data showed significant downregulation of MYCN, GALNT1, and SLC35D1 expression following KDM4B knockdown (Fig. S11D).

Fig. 6: KDM4B binds to the MYCN promoter and promotes MYCN expression.

A The results of ChIP-seq targeting KDM4B reveal gene regions directly bound by KDM4B. B Venn diagram illustrating the overlap between genes directly bound by KDM4B and those that are differentially expressed following KDM4B knockdown. The results of qPCR confirmed that protein-coding genes were directly bound by KDM4B and differentially expressed after KDM4B knockdown, including genes whose expression increased (C) or decreased (D), after KDM4B knockdown. E Snapshot from the IGV program displaying the KDM4B peaks at the MYCN genomic loci (bottom), from which 3 pairs of specific primers were designed (top). F, G ChIP–PCR assay in H69AR and H446DDP cells demonstrated that a specific antibody against KDM4B, but not isotype IgG, could capture the fragment containing the KDM4B binding site in the MYCN promoter, which was amplified by specific primers using qPCR. H Immunoblot illustrating the relationship between KDM4B and MYCN expression in H69 and H446 cells. I Immunoblot showing the relationship between KDM4B downregulation and MYCN expression in H69AR and H446DDP cells. J Immunoblotting was performed to evaluate the levels of Flag and MYCN after transduction of wild-type or mutant KDM4B in H69 and H446 cells. ***, p < 0.001; **, p < 0.01. *, p < 0.05.

The amplification of genes in the MYC family (including MYC, MYCL, and MYCN) is closely associated with SCLC [42]. Among these genes, MYCN encodes the proto-oncogene N-Myc, and multiple studies have identified its overexpression as a driver of chemoresistance in SCLC [43,44,45]. Analysis of the TU-SCLC and GSE560052 cohorts revealed a significant positive correlation between MYCN and NEUROD1 expression (Fig. S4B, C). Previous work has demonstrated that the KDM4 family promotes the expression of MYCN by regulating H3K9me3 levels and facilitates adrenergic state transition in neuroblastoma [46], which is consistent with our findings that KDM4B transcriptionally regulates MYCN. Based on the ChIP-seq peak of KDM4B at the MYCN promoter, we designed three pairs of primers to verify KDM4B binding and identify the precise binding site (Fig. 6E). In both H69AR and H446DDP cells, ChIP–PCR targeting KDM4B suggested that KDM4B bound 200–400 bp upstream of the MYCN transcription start site (Fig. 6F, G). We further investigated the effect of KDM4B on MYCN protein levels and observed that MYCN expression increased upon KDM4B overexpression in H69 and H446 cells, whereas MYCN protein levels decreased following KDM4B knockdown in H69AR and H446DDP (Fig. 6H, I). Moreover, immunoblot analysis showed that overexpression of wild-type KDM4B (KDM4B-WT), but not the enzymatically inactive mutant KDM4BΔJmjC, significantly upregulated MYCN in both cell lines, indicating that the catalytic activity of the JmjC domain is essential for KDM4B-mediated regulation of MYCN (Fig. 6J).

KDM4B/DHX9 promotes MYCN-driven Hedgehog signaling pathway

KDM4B functions as a histone demethylase, removing methylation marks from H3K9me3 and H3K36me3, thereby modulating chromatin openness and participating in transcriptional regulation alongside other transcriptional cofactors. To elucidate the regulatory mechanism underlying MYCN expression, we performed a DNA pull-down assay using a biotin-labeled MYCN promoter fragment incubated with cellular proteins, followed by streptavidin magnetic bead affinity purification and mass spectrometry identification of MYCN promoter-binding proteins. After structural proteins such as histone H4 and histone H2B, as well as cellular ribosomal components, including 60S and 40S ribosomal proteins, were excluded, the top 10 proteins were selected and sorted by significance (Table 1). Among these, the RNA helicases DDX5 and DHX9 have been previously reported to facilitate gene transcription and can be selectively inhibited by specific compounds. Mass spectrometry identified peptides corresponding to DDX5, DHX9, and KDM4B (Fig. 7A). Subsequent DNA pull-down followed by immunoblotting in H69AR and H446DDP cells confirmed that DDX5, DHX9, and KDM4B physically associate with the MYCN promoter (Fig. 7B). Coimmunoprecipitation experiments targeting KDM4B further revealed that KDM4B interacts specifically with DHX9 but not with DDX5 (Fig. 7C). Analysis of RNA sequencing data revealed a decrease in DHX9 mRNA expression following KDM4B knockdown (Fig. 7D). Immunoblotting confirmed that the DHX9 protein level decreased upon KDM4B knockdown but increased with increasing KDM4B expression (Fig. 7E). DHX9 can be associated with the activation of signaling pathways, such as DHX9 and NF-κB and type I interferon pathways, by promoting transcription. Collectively, these results indicated that KDM4B physically interacts with DHX9 to directly regulate MYCN expression.

Fig. 7: KDM4B/DHX9 promotes MYCN-driven Hedgehog signaling pathway.

A The sequences of the DDX5, DHX9 and KDM4B peptides detected in the DNA pull-down products identified the MYCN-binding protein of H69AR. B The KDM4B, DDX5 and DHX9 proteins in the pull-down products of H69AR and H446DDP cells were verified by immunoblotting. C Co-IP and immunoblot experiments were performed to identify the endogenous interactions between DDX5, DHX9, and KDM4B. D KDM4B and DHX9 expression in H69AR and the corresponding KDM4B knockdown cells. E Immunoblot showing the regulatory effect of KDM4B on DHX9 expression. F MYCN, PTCH1 and GLI1 expression in H69AR and H446DDP cells after 24 h of treatment with DHX9 inhibitor. G The expression of MYCN, PTCH1 and GLI1 in H69 and H446 cells after 24 h of treatment with DHX9 inhibitor in the rescue experiment. H Representative IHC staining of DHX9 in SCLC tissues. The statistical graph is on the right side of the images. I Spearman correlation analysis results for KDM4B and DHX9. ***, p < 0.001; **, p < 0.01.

Table 1 Mass spectrometry results of pull-down products from H69AR cells, including KDM4B and the top 10 proteins ranked by significance.

DHX9 is a highly conserved DExD/H-box RNA helicase that is localized to both the nucleus and the cytoplasm and plays pivotal roles in diverse cellular processes, including transcriptional activation, RNA editing, and microRNA biogenesis. To investigate the role of DHX9 in chemoresistance, cell viability assays were performed, and the results demonstrated that DHX9 knockdown significantly increased the sensitivity of H69AR and H446DDP cells to cisplatin and etoposide (Fig. S12). To further elucidate whether the enzymatic activity of DHX9 modulates MYCN expression and Hedgehog pathway signaling, we used DHX9-IN-2, a selective inhibitor of DHX9 enzymatic activity. Treatment with 3 μM DHX9-IN-2 maintained approximately 90% of the viability of H69AR and H446DDP cells after 24 h (Fig. S13A, B). Under these conditions, DHX9 protein levels remained unchanged, whereas MYCN, PTCH1 and GLI1 expression was downregulated (Fig. 7F). Furthermore, KDM4B overexpression in chemosensitive cells increased the expression of DHX9, MYCN, PTCH1 and GLI1, which was partially reversed by DHX9 inhibition (Fig. 7G). These findings indicate that KDM4B regulates MYCN and Hedgehog signaling in cooperation with DHX9, and that inhibiting DHX9 suppresses its transcriptional activation of MYCN. The precise mechanism by which KDM4B upregulates DHX9 remains to be investigated.

In the TU-SCLC cohort, both univariate and multivariate Cox regression analyses revealed that higher DHX9 expression was linked to shorter OS in SCLC patients at both the gene (Fig. S14A, B) and protein levels (Fig. S14C, D). Additionally, DHX9 levels were significantly elevated in chemoresistant SCLC tissues (Fig. 7H) and strongly positively correlated with KDM4B expression (Spearman’s r = 0.93, p = 0.007) (Fig. 7I).

DHX9 inhibitors increase the chemosensitivity of SCLC

Combination therapy not only boosts treatment efficacy but also helps prevent the development of drug resistance that is frequently associated with monotherapy. The results of the combined analysis of a DHX9 inhibitor with cisplatin or etoposide indicated that DHX9-IN-2 may enhance the effects of chemotherapy, particularly with etoposide (H69AR, HSA = 11.99; H446DDP, HSA = 11.96) (Figs. 8A and S15A). Combining 3 μM DHX9-IN-2 significantly reduced the resistance of H69AR cells to cisplatin and etoposide, as well as the resistance of H446DDP cells to these drugs (Figs. 8B and S15B). To further assess the impact of DHX9 inhibition on chemoresistance in vivo, subcutaneous tumor models were established using H69AR cells. The results demonstrated that weekly administration of chemotherapy in combination with the DHX9 inhibitor significantly enhanced tumor suppression compared to chemotherapy alone (Fig.8C–E).

Fig. 8: DHX9 inhibitors increase the chemosensitivity of SCLC.

A Inhibitory effects of different concentrations of DHX9-IN-2 combined with cisplatin and etoposide on H69AR cells. B The results of the CCK8 assay showing the IC50 of cisplatin and etoposide in H69AR cells when they are combined with 3 μM DHX9-IN-2. IC50, the half-maximal inhibitory concentration. C Growth curves of tumors derived from H69AR cells and treated with chemotherapy and/or DHX9 inhibitors (n = 5). D Images of subcutaneous tumors derived from H69AR cells after chemotherapy with or without DHX9 inhibitors are shown. E Comparison of tumor weight between the groups treated with chemotherapy with or without DHX9 inhibitors. F, G Bioluminescence imaging signals from mice with orthotopic SCLC tumors were measured on Days 7, 14, 21, and 28 (n = 6). The top panel of F shows the scheme of the therapeutic strategy for the SCLC orthotopic tumor. H Weight curve of the mice. HSA, The Highest Single Agent model; DDP, Cisplatin; VP-16, Etoposide; ***, p < 0. 001; **, p < 0. 01; *, p < 0.05; ns, not significant.

To better recapitulate the natural progression of SCLC, we established orthotopic tumor models by injecting mouse-derived RP cells expressing luciferase into the lungs of C57BL/6 mice. After stable tumor establishment was confirmed via bioluminescence imaging (BLI), the mice were randomized into two treatment groups: those treated with chemotherapy alone or those treated with chemotherapy combined with the DHX9 inhibitor. On Day 21 after treatment, compared with those treated with chemotherapy alone, mice receiving the combination therapy exhibited significantly reduced BLI signal intensity, demonstrating slower tumor progression with the addition of DHX9 inhibition (Fig. 8F, G). Importantly, body weight measurements indicated that compared with chemotherapy alone, the combined treatment did not increase toxicity (Fig. 8H). Collectively, these results indicate that DHX9 inhibitors can increase the sensitivity of SCLC to chemotherapy, suggesting a promising strategy for overcoming chemoresistance.

Discussion

In this study, we identified KDM4B as a critical gene associated with chemoresistance in SCLC. Our experiments demonstrated that KDM4B enhances cell stemness and promotes cell proliferation by activating the Hedgehog signaling pathway, ultimately contributing to chemoresistance. Further mechanistic investigations revealed that KDM4B interacts with DHX9, corecruiting it to the MYCN promoter to facilitate MYCN transcription and activate the Hedgehog signaling pathway. Notably, the use of DHX9 inhibitors indicated that MYCN transcription is dependent on DHX9 enzyme activity. Both in vivo and in vitro experiments confirmed that the combination of DHX9 inhibitors significantly increased the sensitivity of SCLC to chemotherapy.

The relationship between KDM4B and the neuroendocrine phenotype of SCLC warrants further investigation. In our exploratory analysis, KDM4B expression positively correlated with the neuroendocrine lineage-specific transcription factors ASCL1 and NEUROD1 in the TU-SCLC cohort, although this association was not significant in GSE60052. Given the limited sample size of our patient-derived SCLC tumor tissue samples, independent validation was not feasible. Nevertheless, emerging evidence from the study of KDM4A, another KDM4 family member, provides support. KDM4A has been identified as a critical epigenetic driver of neuroendocrine prostate cancer (NEPC), promoting tumor progression via direct MYC regulation [47]. Moreover, KDM4A knockdown reduces the expression of neuroendocrine markers (INSM1, secretogranin-3, and NEUROD1) and inhibits SCLC proliferation in vitro and in vivo [48]. These findings suggest that KDM4 family members may regulate neuroendocrine differentiation across tumor types. However, given the inconsistent correlations and limited sample size, the relationship between KDM4B expression and SCLC molecular subtypes, particularly neuroendocrine phenotypes, remains exploratory and requires validation in larger, well-characterized patient cohorts with comprehensive molecular profiling.

This study demonstrated that KDM4B contributes to chemoresistance in SCLC through the Hedgehog signaling pathway, although other mechanisms may also be involved. Preclinical studies have indicated that the Hedgehog signaling pathway is upregulated in SCLC and promotes disease progression and chemotherapy relapse [49, 50]. However, a multicenter trial in patients with extensive-stage SCLC revealed that the combination of vismodegib (a SMO inhibitor) with etoposide and cisplatin did not significantly improve progression-free survival (PFS) or OS [51]. In our study, inhibition of the Hedgehog signaling pathway increased chemosensitivity in vitro. This discrepancy may be due to the focus on tumor cells alone, while the in vivo environment includes tumor cells, immunocytes, blood vessels, and stroma, along with their crosstalk. Moreover, RNA sequencing and qPCR validation revealed that KDM4B not only regulates the Hedgehog signaling pathway but may also influence other pathways, such as the NOTCH and WNT pathways, which have been reported to be related to chemoresistance in SCLC [52, 53]. These findings highlight the importance of investigating upstream regulators of chemoresistance in SCLC.

KDM4B modulates chromatin accessibility by demethylating H3K9me3 or H3K36me3, thus participating in the regulation of gene expression. Integrated analysis of ChIP-seq data targeting KDM4B, combined with RNA-seq data before and after KDM4B knockdown, revealed MYCN, GALNT1, and SLC35D1 as potential direct downstream targets of KDM4B. MYCN proto-oncogene (MYCN) is involved in various cancers and has been linked to the Hedgehog signaling pathway. For instance, activation of the Sonic Hedgehog (SHH) pathway enhances MYCN expression in SHH medulloblastoma mouse models [54]. Additionally, MYCN overexpression has been shown to generate humanized SHH medulloblastoma models derived from various human neural progenitor cells, characterized by the activation of the SHH pathway [55]. In our study, we observed enrichment of KDM4B at the MYCN promoter in two chemoresistant SCLC cell lines, suggesting that KDM4B may directly promote MYCN expression, which is consistent with the findings of recent studies. Notably, Abu-Zaid et al. described a positive feedback loop between MYCN and KDM4 [46], wherein MYCN induces the expression of KDM4 family members to drive neuroblastoma cells from a mesenchymal state to an adrenergic state, while KDM4 proteins sustain adrenergic identity via MYCN amplification. Further mechanistic studies revealed that KDM4 regulates the adrenergic transcriptional profile through epigenetic mechanisms. KDM4B requires cooperation with additional transcriptional complexes to effectively regulate gene transcription. We revealed that KDM4B interacts with RNA helicase A (DHX9), and together, these genes are recruited to the MYCN promoter, thereby promoting MYCN expression and activating the Hedgehog signaling pathway. DHX9 acts as a multifunctional coactivator by bridging transcription factors, the coactivator CBP/p300, and RNA polymerase II, playing a broad role in transcriptional regulation [56, 57]. Previous studies have emphasized the importance of the functional domains of DHX9 in various transcriptional programs, including the regulation of oncogenic MYC signaling in non-small cell lung cancer [58, 59]. In our study, inhibition of DHX9 enzymatic activity significantly downregulated MYCN expression and suppressed Hedgehog signaling without altering DHX9 protein levels, indicating that the catalytic activity of DHX9 is essential for this regulatory axis. Furthermore, wild-type KDM4B but not the mutant KDM4B (KDM4BΔJmjC) was capable of modulating MYCN transcription. Collectively, these findings suggest that KDM4B regulates MYCN expression through a dual mechanism involving chromatin remodeling via histone demethylation and the facilitation of transcriptional complex assembly through protein–protein interactions with DHX9. This hybrid model offers novel mechanistic insights into transcriptional regulation in SCLC and identifies DHX9 as a potential therapeutic target to enhance chemosensitivity.

This study, together with prior research, suggests that DHX9 can be targeted not only to restore chemosensitivity but also to enhance the overall efficacy of chemotherapy in SCLC. Research has shown that the deletion of DHX9 leads to an accumulation of cytoplasmic double-stranded RNA and R-loops, enhancing immunogenicity and resulting in reduced tumor growth in SCLC [60]. Notably, ATX968, also known as DHX9-IN-2, was identified as a potent and selective DHX9 inhibitor via multiple complementary approaches, including enzymatic assays (IC50 = 8 nM), biophysical binding analysis (KD = 1.3 nM), mechanistic characterization (non-ATP competitive), and selectivity profiling against related helicases and 97 kinases [57]. Despite these promising findings, several potential side effects must be considered for its application in SCLC treatment. Inhibiting DHX9 may impact genomic stability, raising concerns about potential long-term effects on rapidly dividing tissues, such as the bone marrow and gastrointestinal tract. Additionally, the possibility of cross-inhibition with other homologous proteins remains to be fully addressed, and the concentration-dependent effects of DHX9-IN-2 on both targeted and off-target interactions require further investigation.

There are several strengths in our study. First, we performed a comprehensive analysis that revealed that KDM4B is correlated with chemoresistance in SCLC, and validated these findings through both in vivo and in vitro experiments. Second, we employed multiple techniques, such as ChIP-seq, DNA pull-down and CoIP, to reveal that KDM4B physically interacts with DHX9 and corecruits to the MYCN promoter, directly promoting its transcription. Third, our research revealed that DHX9 inhibitors can enhance SCLC sensitivity to chemotherapy, and notably, a phase I/II clinical trial in which a DHX9 helicase inhibitor was orally administered was initiated in January 2025 (NCT06625515), underscoring the translational potential of targeting DHX9. However, this study has several limitations. First, although the sequencing data suggested that KDM4B modulates multiple signaling pathways to promote chemoresistance in SCLC, only the most significantly enriched Hedgehog pathway was thoroughly validated. The involvement of the NOTCH and WNT pathways was confirmed only by qPCR, and possible cross-talk among these pathways was not explored, which warrants further investigation. Moreover, considering the predominance of the NE subtype in SCLC, this study was conducted primarily using representative NE-type cell lines. Therefore, the generalizability of these findings to non-NE subtypes remains uncertain and should be evaluated in more diverse preclinical models. Additionally, although specific knockdown of DHX9 reversed chemoresistance, the potential off-target effects as well as the on-target specificity of DHX9 inhibitors require further experimental validation. Finally, while the DHX9 inhibitor has shown preclinical efficacy in tumors characterized by BRCA deficiency, MSI-H, and/or dMMR, additional studies are necessary to evaluate its therapeutic applicability in SCLC.

In summary, we demonstrated that KDM4B interacts with DHX9 and together they are recruited to the MYCN promoter to directly promote its transcription, thereby promoting the involvement of the Hedgehog signaling pathway in chemoresistance. Furthermore, the application of DHX9 inhibitors enhanced the chemosensitivity of SCLC. Future studies with more extensive validation may provide new therapeutic targets to overcome chemoresistance in SCLC.