CRISPR Screen Identifies HDAC3 as a Novel Radiosensitizing Target in Small Cell Lung Cancer

Abstract

Small cell lung cancer (SCLC) is an aggressive malignancy, with most patients presenting with prognostically poor extensive-stage disease. Limited progress in standard care stresses the urgent need for novel therapies. Radiotherapy offers some survival benefit for selected SCLC patients but could be enhanced with radiosensitizers. Here, we identify HDAC3 as a novel radiosensitizing target in SCLC using a CRISPR knockout screen and demonstrate its efficacy and mechanism. SBC5 cells were transduced with a custom EpiDrug sgRNA library and treated with ionizing radiation (IR) to identify radiosensitizing genes. HDAC3 emerged as a candidate and was validated through genetic knockdown (KD) and pharmacologic inhibition (RGFP966) in multiple SCLC cell lines. Both approaches enhanced radiosensitivity, as shown by cell viability (dose-modification factor [DMF]₁₀ = 1.14–1.69) and clonogenic assays (DMF₁₀ = 1.16–1.41). We assessed changes in chromatin accessibility by ATAC-seq, and IR-induced DNA damage and repair using γH2AX foci detection, DSB repair assays and immunoblotting of repair proteins. HDAC3-deficient cells exhibited increased chromatin accessibility, greater IR-induced DSBs, and impaired repair capacity, resulting in persistent DNA damage. This repair defect sensitized cells to PARP inhibitors, where combining RGFP966 with Olaparib or Talazoparib produced additive to synergistic effects. In SCLC xenograft models, HDAC3 KD or RGFP966, combined with IR, achieved significant tumor growth inhibition. Collectively, we identified HDAC3 as a novel radiosensitizing target in SCLC. Its functional loss increased the generation and persistence of IR-induced DNA DSBs, effectively sensitizing SCLC cell lines and xenografts to IR, providing a potential radiosensitization strategy to treat SCLC.

Introduction

Lung cancer is the leading cause of cancer-related deaths world-wide with small cell lung cancer (SCLC), the most aggressive form, accounting for approximately 15% of all diagnosed cases ¹. SCLC is characterized by a short doubling-time, high growth fraction, and early metastasis, and patient prognosis remains exceptionally poor with a dismal 5-year overall survival (OS) of <7% ¹. Two-thirds of patients diagnosed with SCLC present with extensive-stage (ES) disease, defined by extrathoracic metastases or locoregional disease that is too large to be encompassed by a tolerable radiation port ^2,3. Despite this, the standard of care for ES-SCLC patients has minimally changed for nearly three decades, consisting of cisplatin/carboplatin and etoposide chemotherapy and recently approved immunotherapy ^2,4,5. While the initial response rate is high, rapid relapse is inevitable due to acquired treatment resistance ¹. In this dire setting, there is an urgent need for novel therapeutic strategies to improve tumor control, especially for ES-SCLC.

The high proliferation rate of SCLC makes it particularly susceptible to DNA-damaging therapies; therefore, enhancing the efficacy of an existing DNA-damaging therapeutic, such as radiotherapy (RT), serves one potential treatment strategy. Currently, RT in SCLC is primarily given concomitantly with chemotherapy in the first-line setting for patients with limited-stage (LS) disease ^1,2. The use of RT in ES-SCLC has traditionally been reserved for the palliation of symptoms; however, given that up to 75% of these patients have persisting intrathoracic disease following chemotherapy and subsequent disease progression, consolidative thoracic radiotherapy (TRT) has been considered for ES-SCLC patients ¹. A phase III trial investigating the use of consolidative TRT in patients with ES-SCLC found an improvement in their 2-year OS compared to patients who did not receive TRT (13% versus 3%, p = 0.004), with no severe toxic effects ⁶; however, more than 40% of patients still experienced intrathoracic disease recurrence, highlighting the need for greater tumor control. One approach is to combine RT with radiation sensitizers (radiosensitizers), compounds that enhance the lethal effects of radiation ^7,8, to improve RT outcomes for patients with ES-SCLC.

Pharmacological inhibition of key enzymes involved in DNA damage response, such as ATR, and DNA repair pathways, such as PARP, has shown preclinical and clinical efficacy as radiosensitizers in SCLC ^9–12. More recently, our group showed effective radiosensitization using AZD1390, a selective inhibitor of the DNA damage response master ATM, in SCLC ¹³. To identify novel radiosensitizers of SCLC, we conducted a targeted CRISPR dropout screen using a custom Epi-Drug library comprising of druggable targets and numerous epigenetic regulators ^13,14 and identified HDAC3 as a novel gene candidate mediating SCLC radiosensitivity. HDAC3 is a class I histone deacetylase with functions reported in chromatin remodeling, transcriptional regulation, DNA replication and repair, cell cycle progression, and apoptosis ^15–18. It is upregulated in various cancers including gastric ^19,20, ovarian ²¹, fusion-positive rhabdomyosarcoma (FP-RMS) ²² and lung adenocarcinoma ²³, and has been linked to poor patient outcomes ^22,24. Recently, HDAC3 functional loss has been found to radiosensitize FP-RMS ²², strengthening its potential applicability as a radiosensitizing target in other cancer types. However, there is still a paucity of studies examining HDAC3’s radiosensitizing role in cancer

Here, we demonstrate that the functional loss of HDAC3 radiosensitizes SCLC models in vitro and in vivo. We further investigate the mechanism of HDAC3 functional loss-mediated radiosensitivity, with a focus on chromatin accessibility and DNA damage and repair.

Materials and Methods

Cell lines and reagents

U2OS cells (RRID:CVCL_1453), H1048 cells (RRID:CVCL_1679), and SHP77 cells (RRID:CVCL_0063) were purchased from the American Type Culture Collection (ATCC). HEK-293T cells (RRID:CVCL0042) were generously provided by Dr. Robert Bristow. U2OS cells (RRID:CVCL_1453) expressing EJ/DR reporters were generously provided by Dr. Ranjit Bindra and Dr. Simon Powell. U2OS cells (RRID:CVCL_1453) expressing the EJ5-GFP reporter were generously provided by Dr. Jeremy M. Stark. SBC-5 cells (RRID:CVCL1_1562) were generously provided by Dr. Kazuhiro Yasufuku. NCI-H446 cells (RRID:CVCL_1693) were generously provided by Dr. Charles Rudin. Detailed cell line culture conditions are available in Supplementary Table S1. All cell lines were verified by Short Tandem Repeat fingerprinting (The Centre for Applied Genomics) and tested negative for mycoplasma within 6 months of use. RGFP966 (Cederlane Labs #S7229) was reconstituted in dimethyl sulfoxide (DMSO) at 10mM and stored at −80°C for in vitro work. For in vivo administration, RGFP966 was dissolved in DMSO and diluted in a vehicle of 40% (v/v) polyethylene glycol 300 (TCI Chemicals # H0543), 5% (v/v) TWEEN 80 (Sigma-Aldrich # P1754), and 45% (v/v) saline (Baxter # JF7123). The final DMSO concentration was 10% (v/v).

CRISPR Dropout Screen

SBC5 cells (RRID:CVCL_1679) were transduced with a custom EpiDrug sgRNA library ¹⁴ and the treatment group given an optimized radiation schedule at clinically relevant fractions ^2,6 of 1.5 Gy on day 4 (T4), T13, T25, T28 and 3 Gy at T34. Screen results were analyzed using the DrugZ algorithm to identify radiosensitizing candidates. Detailed methodologies for the CRISPR dropout screen are provided in Supplementary Methods and have been previously described by Ran et al. (2024)¹³.

HDAC3 Knockdown (KD) Cell Line Generation Using small-hairpin RNA (shRNA)

Lentivirus for transduction was generated following standard protocols (Supplementary Methods). Three independent shRNA constructs against HDAC3 were used to generate KD models in SCLC cell lines. In addition, a non-targeting shRNA (shSCR) was used as a control to account for potential off-target effects by HDAC3-targeting shRNA. Sequence details for shRNAs used in this study are available in Supplementary Table S2 and detailed transduction methods are available in Supplementary Methods.

Immunoblotting

Immunoblotting was conducted as previously described ²⁵ and detailed methodologies are provided in Supplementary Methods. Antibodies utilized in this study include: rabbit monoclonal HDAC3 antibody (Cell Signaling Technology #85057S; 1:1000), mouse monoclonal HDAC1 antibody (Cell Signaling Technology #5356S; 1:1000), mouse monoclonal GAPDH antibody (ThermoFisher Scientific #AM4300; 1:5000), mouse monoclonal Histone H3 antibody (Cell Signaling Technology #3638S; 1:1000;), rabbit polyclonal Acetyl-Histone H3 (Lys9/Lys14) antibody (Cell Signaling Technology #9677S; 1:1000), rabbit monoclonal ATM (phospho S1981) antibody (Epitomics #2152-1; 1:1000), rabbit monoclonal ATM antibody (Epitomics #1549-1; 1:1000), rabbit monoclonal DNA-PKcs (phospho S2056) antibody (Abcam #ab124918, RRID:AB_11001004; 1:2000), mouse monoclonal DNA-PKcs antibody (Abcam # ab44815, RRID:AB_731982; 1:1000), mouse monoclonal Ku70 antibody (Santa Cruz #sc-5309, RRID:AB_628453; 1:1000), rabbit polyclonal Rad51 antibody (Santa Cruz #sc-8349, RRID:AB_2253533; 1:1000), IRDye 680RD Donkey anti-Rabbit antibody (LI-COR BioSciences #LIC-926-68073, RRID:AB_10954442; 1:3000), and IRDye 800CW Donkey anti-Mouse antibody (LI-COR BioSciences #LIC-925-32212, RRID:AB_2716622; 1:3000).

Cell Radiosensitization Assays

Long-term Viability Assays

Cell lines were treated with various doses of RGFP966 or DMSO control for 96 hours prior to plating in 96-well plates with fresh media containing inhibitor or DMSO. For HDAC3 KD cell lines, cells were plated one day prior to irradiation at optimal seeding densities. On the following day, plates were exposed to single doses of radiation at a central dose rate of approximately 0.74 Gy/min with a Cesium-137 source using the Gammacell-40 Extractor (Nordion). After 10 days of growth, cell viability was assessed using AlamarBlue^™ Cell Viability Reagent (ThermoFisher Scientific #DAL1025) and a compatible microplate reader.

Clonogenic Survival Assays (CSAs)

Cell plating numbers for all CSAs were optimized such that 30–60 colonies were formed at each dose of radiation and drug. Cell lines were treated with various doses of RGFP966 or DMSO control for 72–96 hours prior to plating in 6-well plates in technical triplicates with fresh media containing inhibitor or DMSO. For HDAC3 KD cell lines, cells were seeded in 6-well plates in technical triplicates one day prior to irradiation. On the following day, irradiation was administered as above. After 72 hours, the media was changed to treatment-free media and cells were cultured for 1 to 3 weeks until colonies, defined as 50 cells or more, formed. Cells were fixed and stained with crystal violet solution (0.5 g crystal violet, 0.85 g NaCl, 5 mL 10% formal saline, 50 mL absolute ethanol, 45 mL water) and colonies were counted manually.

HDAC3 Rescue Experiment

SBC5 cells (RRID:CVCL_1679) were transduced with the pLKO.1 puro plasmid (Addgene #8453, RRID:Addgene_8453) carrying shRNA targeting the 3’UTR region of HDAC3. Positive cells were collected after selection with 5 μg/ml of puromycin for 72 hours and successful knockdown was validated by immunoblotting for HDAC3. The SBC5 HDAC3 KD cells were subsequently transduced with the pLenti-CMV Blast empty (w263-1) plasmid (Addgene #17486, RRID:Addgene_17486) carrying either the full-length Wildtype HDAC3 or HDAC3 K25A mutant construct. The transduced cells were selected for 7 days with blasticidine and validated by immunoblotting for HDAC3. Subsequent CSAs were performed as per the methodology described in the previous section.

ATAC-seq

ATAC-Sequencing

ATAC-seq was performed following the Omni-ATAC protocol previously described ²⁶ with modifications. Detailed methodologies are available in Supplementary Methods.

Data Processing and Analysis

Reads were aligned to human genome hg38 using Bowtie2 (v. 2.4.5, RRID:SCR_016368) and filtered to remove multimapping reads using Sambamba (v. 0.7.0, RRID:SCR_024328). Next, Samtools (v. 1.14, RRID:SCR_002105) was used to shift the read start sites (forward reads +4 bases, reverse reads −5 bases) to adjust for overhangs generated by the Tn5 transposon. Duplicate reads were flagged and removed using Picard (RRID:SCR_006525) MarkDuplicates (v. 2.10.9). From the final reads, peaks were called using MACS2 (v. 2.2.7.1, RRID:SCR_013291) using a q-value cutoff of 0.01. Raw reads output and cleanup are summarized in Supplementary Table S3. Differential accessibility (DA) analysis was performed with DiffBind (v. 3.6.1, RRID:SCR_012918) to identify consensus peaks that show significant changes in accessibility between conditions. Consensus and DA peaks were annotated using Bioconductor package ChIPseeker (v. 1.32.0, RRID:SCR_021322) and TxDb.Hsapiens.UCSC.hg38.knownGene and org.Hs.eg.db. For metagene analysis, bigwig files were generated using deeptools (RRID:SCR_016366) bamCoverage (v. 3.5.1) normalizing to reads per kilobase per million (RPKM) and a bin size of 10 bases. Accessibility signal from each sample was calculated for gene bodies (+/− 3kb) using computeMatrix scale-regions or gene transcription start sites (TSS; +/− 3kb) using computeMatrix reference-point. Metagene profiles and ATAC signals were plotted using a custom script and ggplot2 in R (v. 4.2.0, RRID:SCR_014601).

Immunofluorescence Staining

Immunofluorescence staining was performed as per the protocol described by the Centre for Flow Cytometry & Scanning Microscopy (CCSM; Sunnybrook Research Institute) ²⁷ with some modifications. Detailed methodologies are available in Supplementary Methods. Nuclear foci were visualized with LSM 700 confocal microscope (Zeiss) and quantified using ImageJ (RRID:SCR_003070). Antibodies utilized for staining include mouse monoclonal Anti-phospho-Histone H2A.X (Ser139) antibody (Sigma-Aldrich #05-636, RRID:AB_309864; 1:1000) and Alexa Fluor^® 555 Donkey anti-Mouse antibody (ThermoFisher Scientific #A-31570; 1:1000).

HR and NHEJ Reporter Assays

U2OS cells stably expressing EJ5-GFP (Addgene #44026, RRID:Addgene_44026) ²⁸, EJ-RFP ²⁹, or DR-GFP (Addgene #26475, RRID:Addgene_26475) ²⁸ reporter systems were used to assess DSB repair by canonical non-homologous end-joining (c-NHEJ), non-canonical NHEJ (referred to as mutagenic NHEJ or m-NHEJ throughout the manuscript) ²⁹, and homologous recombination (HR), respectively. Cells were plated in 6-well plates one day prior to transfection with 4 μg of pCBASce I (Addgene #26477, RRID: Addgene_26477) to induce DBSs. Transfections were performed using Lipofectamine 3000 Transfection Reagent (ThermoFisher Scientific #L3000015) according to the manufacturer’s recommendations. DSB repair activity was assessed after 72 hours by quantification of the percentages of GFP+ and DsRed+ cells using the FACSymophony A3 Cell Analyzer (BD Biosciences). For RGFP966, cells were pre-treated with the inhibitor for 96 hours prior to plating and the inhibitor was maintained throughout the assay. Data was analyzed using FlowJo 10 (RRID:SCR_008520).

PARPi Synergy

Cell lines were seeded at optimal seeding densities in 384-well plates. On the following day, cells were treated with predetermined dose ranges of RGFP966, in combination with either PARP inhibitors, Olaparib or Talazoparib. Staurosporine and DMSO were used as controls and each condition was conducted in technical triplicates. After 6 days of growth, cell viability was assessed using CellTiter-Glo^® luminescence as the readout. SynergyFinder+ (RRID:SCR_026127) was used to calculate synergy scores using the Zero Interaction Potency (ZIP) model. Most Synergistic Area Score (MSAS) is the average synergy score in the more synergistic 3×3 region.

SCLC Cell Line-Derived Xenografts (ClDX) and In Vivo Anti-Tumor Efficacy

The Institutional Animal Care and Use Committee approved all animal protocols for this study. Experiments were performed in NOD/SCID gamma (NSG; NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ, RRID:BCBC_4142; The Jackson Laboratory or Cancer Stemcell Colony) mice (male or female) that were 6 to 8 weeks old at the time of tumor cell engraftment. Each mouse was subcutaneously injected with two million SCLC cells in the right flank. For tumor irradiation, mice were anesthetized using 5% isoflurane and restrained in a custom-built lead shielding device, designed by the Princess Margaret Cancer Centre Machine Shop, to reduce dose to normal mouse tissue. Tumor volumes (TV) were calculated from digital caliper measurements using the formula: TV = (xy²)/2. Mice weights and TV were measured three times weekly until mouse euthanasia at a TV of 1000–1500 mm³.

SBC5 HDAC3 isogenic xenografts

Mice were engrafted with SBC5 shSCR or HDAC3 KD (shHDAC3 #2, shHDAC3 #3) cells in the right flank. At a TV of approximately 150 mm³, tumors were irradiated in daily 2 Gy fractions with an X-Ray irradiator (XRAD 320, Precision X-Ray) on days 1–4.

RGFP966 Drug Efficacy Study

Mice were randomized at a TV of 60–80 mm³ and RGFP966 (40 mg/kg) or vehicle was administered intraperitonially from Monday to Friday (daily), for 14 total doses. Tumors were irradiated in daily 3 Gy fractions for 3 days with the XRAD 320 on days 4–6 after randomization. Radiation was administered after drug dosing.

Statistical Analysis

Clonogenic survival values were normalized to non-irradiated controls to account for cell death resulting from HDAC3 loss alone. Survival curves were fitted using the linear-quadratic model, S = exp(αD + βD²), where D = radiation dose. For each HDAC3 KD cell line and drug dose, the dose modification factor (DMF), defined as a ratio of the radiation dose needed to achieve an equivalent survival fraction between the radiation-alone and radiation plus KD or drug combination groups, was calculated at 10% survival. A DMF value >1.0 confers radiosensitization. Chromatin accessibility between groups was compared using the Wilcoxon rank-sum test. Comparisons of DNA damage and repair between groups were made using a one-way ANOVA test for multiple comparisons and an unpaired student’s t-test, respectively. For in vivo tumor growth delay, time to reach endpoint was compared using Kaplan-Meier analysis and the log-rank test. Tumor growth inhibition (TGI) was calculated by comparing the average TV of the treatment group with the control group at the specified timepoint. All analyses were performed using GraphPad Prism 10 (GraphPad Software, RRID:SCR_002798).

Data Availability Statement

The data generated in this study are available upon request from the corresponding author.

Results

HDAC3 functional loss Radiosensitizes SCLC Cell Lines

We performed a targeted CRISPR knockout screen with ionizing radiation (IR) and used a false discovery rate (FDR) threshold of 0.1 to identify single gene knockouts (KO) mediating SCLC radiosensitivity at sequenced timepoints (T13, T25, T40). Radiation was delivered in 1.5 Gy fractions on days 4, 13, 25, and 28 of the screen, to apply sustained selective pressure while minimizing toxicity-induced dropout. A total of six genes were identified at all three timepoints (ATM, BRCA2, HDAC3, POLE2, BRCC3, PSMB1; Fig. 1A; Supplementary Fig. 1A), several of which are well-known regulators of DNA damage response (ATM, BRCA2, POLE2, BRCC3) whose loss is expected to show radiosensitization and thus, confirmed the robustness of our screen to uncover novel radiosensitizers in SCLC. ATM was the top candidate at all three timepoints (Supplementary Fig. 1A). Indeed, ATM inhibition has previously been shown as a radiosensitizer in SCLC by our group ¹³. HDAC3 KO was a novel candidate with increasing rank over the course of the screen (Supplementary Fig. 1A) and strikingly, HDAC3 was the sole member of the HDAC family identified with an FDR < 0.1 in the screen (Fig. 1B; Supplementary Fig. 1B).

Figure 1. In vitro CRISPR dropout screen in SBC5 cells identifies HDAC3 as a novel radiosensitizing candidate in SCLC.

(A) Venn diagram showing overlap in radiosensitizing genes analyzed by DrugZ at the three sequenced timepoints: T13, T25, and T40. The top six gene candidates appearing at all three timepoints are highlighted. (B) Gene rank of HDAC3 and all other HDACs, grouped by class, according to −log (adjusted P value). LacZ is a negative control. (C) Western blot and quantification of the SBC5 cell line showing HDAC3 and H3K9K14ac expression in shSCR control and HDAC3 KD cells using three independent shRNAs. HDAC3 bands were normalized to GAPDH and H3K9K14ac bands were normalized to Histone H3. (D) Clonogenic assays in HDAC3 KD SBC5 and H1048 cell lines with DMF₁₀ values shown in brackets. (E) Clonogenic rescue assay in HDAC3 KD SBC5 cells with DMF₁₀ values shown in brackets. Data represents a single experiment (n=1). (F) Western blot and quantification of the SBC5 cell line treated with different doses of RGFP966 for 72 hours, showing H3K9K14ac expression. H3K9K14ac bands were normalized to Histone H3. (G) Clonogenic assays in SBC5 and H1048 cell lines treated with RGFP966 with DMF₁₀ values shown in brackets. For C-G, western blot quantifications and DMF values for each condition are normalized to shSCR control for C-D, ‘HDAC3 KD + HDAC3 WT’ for E, and 0μM RGFP966 for F-G. For C-D and F-G, data are presented as the mean ± SD of three replicates. All comparisons were calculated with a one-way ANOVA Dunnett’s multiple comparisons test. (*, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001).

To validate radiosensitization mediated by HDAC3 KO, we generated shRNA-mediated HDAC3 knockdown (KD) models in multiple SCLC cell lines (Fig. 1C; Supplementary Fig. 2A–2B). We functionally confirmed HDAC3 KD in our cell lines by probing for and observing upregulation of histone 3 lysine 9 and lysine 14 acetylation (H3K9K14ac; Fig. 1C; Supplementary Fig. 2A–2B), which holds functions in gene expression activation ^30,31 and is expected with HDAC3 functional loss ^15–17. HDAC3 KD increased radiosensitivity in the H446 cell line and to a lesser extent in the SHP77 cell line (DMF₁₀ range = 1.14–1.69; Supplementary Fig. 2C), shown by long-term cell viability assays. We further confirmed this radiosensitizing effect in the SBC5 and H1048 cell lines with clonogenic survival assays (CSA) and, consistent with the cell viability data, we observed increased radiosensitivity in HDAC3 KD cells compared to the control (SBC5: DMF₁₀ range = 1.16–1.32, p = 0.044 for shHDAC3 #2; H1048: DMF₁₀ range = 1.21–1.24, p < 0.01 for shHDAC3 #1, p = 0.024 for shHDAC3 #2; Fig. 1D). To corroborate whether this enhanced radiosensitivity was through loss of HDAC3’s deacetylase activity, we exogenously overexpressed either wildtype (WT) HDAC3 or the deacetylase-dead K25A mutant HDAC3 constructs in HDAC3 KD SBC5 cells ³². We found that overexpressing WT HDAC3 but not the catalytic-dead mutant rescued the loss of colony forming ability in irradiated HDAC3 KD cells, lending support to the idea that increased radiosensitivity in HDAC3-deficient cells is through loss of its enzymatic function (Fig. 1E).

Next, we assessed radiosensitization using an HDAC3-selective inhibitor (HDAC3i), RGFP966 ³³. We treated SCLC cell lines with different doses of RGFP966 in combination with IR. We observed a dose-dependent increase in H3K9K14ac with RGFP966 treatment (Fig. 1F; Supplementary Fig. 2D). Consistent with HDAC3 KD, RGFP966 treatment demonstrated a dose-dependent increase in radiosensitivity compared to DMSO-treated cells in the SBC5 and H1048 cell lines (DMF₁₀ range = 1.20–1.38 for 1 μM, p < 0.01 for both; Fig. 1G), shown by CSA, and in the H446 cell line (DMF₁₀ = 1.38 for 1 μM, p < 0.01; Supplementary Fig. 2E), shown by long-term cell viability assay, using two different doses of RGFP966.

HDAC3 functional loss Increases Chromatin Accessibility and Radiation-Induced DNA Double-Strand Break (DSB) Generation

Having confirmed radiosensitization in multiple SCLC cell lines with HDAC3 functional loss, we then sought to determine the mechanism underlying this radiosensitizing effect. Histone acetylation is an epigenetic mark known to facilitate a more relaxed or open chromatin structure ³⁴, which has been associated with greater DNA damage responses ³⁵. We therefore explored how HDAC3 functional loss affected chromatin accessibility in SCLC cells and influenced radiation-induced DNA DSB generation.

To investigate potential changes in chromatin accessibility, we performed the Assay for Transposase-Accessible Chromatin Sequencing (ATAC-seq) using SBC5 cells following genetic or pharmacologic perturbation against HDAC3. Negative controls included WT and shSCR cells, and conditions with HDAC3 functional loss were HDAC3 KD and RGFP966-treated (at 2 μM and 5 μM) cells. Metagene analysis comparing RPKM-normalized accessibility around ATAC-seq consensus peaks demonstrated greater chromatin accessibility with HDAC3 functional loss than the controls (shHDAC3, p < 0.001, Fig. 2A, left; RGFP966, p < 0.0001, Fig. 2B, left). We also compared accessibility around protein coding genomic regions (consisting of transcription start and end sites), and consistent with ATAC-seq consensus peaks, the accessibility peak was greater in samples with HDAC3 functional loss than the controls (shHDAC3, p < 0.01, Fig. 2A, right; RGFP966, p < 0.0001, Fig. 2B, right).

Figure 2. HDAC3 functional loss increases chromatin accessibility.

(A, B) Metagene plots of ATAC-seq consensus peaks (left) and protein-coding genes (right) in (A) HDAC3 KD and (B) RGFP966-treated samples compared to the control. Peak values are the mean RPKM-normalized accessibility in each sample. Samples were compared using a Benjamini-Hochberg (BH) corrected Wilcoxon Rank Sum Test. (C, D) Volcano plot (left) and genomic feature annotations (right) of differentially accessible consensus peaks in (C) HDAC3 KD and (D) RGFP966-treated samples compared to the control. Volcano plots: Red dots are more accessible, and blue dots are less accessible peaks in HDAC3 KD and RGFP966-treated samples compared to the control. Genomic feature annotations: “Up” refers to peaks with increased accessibility and “Down” refers to peaks with decreased accessibility in HDAC3 KD and RGFP966-treated samples.

Next, we performed a Differential Accessibility (DA) analysis on ATAC-seq consensus peaks to better understand the direction and magnitude of change in accessibility for each consensus peak in samples with HDAC3 functional loss compared to the controls. Our analysis yielded both more accessible and less accessible DA regions in samples with HDAC3 functional loss (Fig. 2C, left; Fig. 2D, left). Consistent with the overall increase in chromatin accessibility, there were far more regions with increased accessibility than decreased accessibility in cells with HDAC3 functional loss (Fig. 2C, left; Fig. 2D, left). To further examine the distribution of genomic features in more accessible (Up) and less accessible (Down) DA peaks, we annotated DA peaks and all consensus peaks to genomic features. Interestingly, the majority of DA peaks that were more accessible in samples with HDAC3 functional loss were primarily within 1 kilobase of transcription start sites (Promoter; Fig. 2C, right; Fig. 2D, right). Altogether, these data demonstrate that the functional loss of HDAC3 results in a more open or accessible chromatin structure, particularly at gene promoter regions, indicating changes in transcriptional activity at these gene loci.

Accessible chromatin has been associated with greater DNA damage propensity ³⁵; therefore, we hypothesized that increased chromatin accessibility in cells with HDAC3 functional loss would result in an increase in the generation of radiation-induced DNA DSBs. To investigate this, we quantified DNA DSBs with γH2AX sub-nuclear foci in SBC5 and H1048 cells, fixed at 0.5 hours after 4 Gy IR exposure. A 4 Gy dose was chosen to induce measurable DNA damage while limiting excessive cytotoxicity. The average number of γH2AX foci per nucleus was greater in cells with HDAC3 functional loss (SBC5: 23.90 ± 7.023 with p<0.01 for shHDAC3 #1, 26.47 ± 10.25 with p <0.0001 for shHDAC3 #2, 25.12 ± 8.351 with p<0.0001 for 5 μM RGFP966, Fig. 3A, left; H1048: 14.46 ± 5.513 with p <0.001 for shHDAC3 #1, 14.14 ± 4.861 with p=0.016 for 5 μM RGFP966, Fig. 3B, left) than the control cells (SBC5: 22.70 ± 4.255; H1048: 12.67 ± 5.624). We also observed a similar trend, though not statistically significant, in shHDAC3 #2 in the H1048 cell line (Supplementary Fig. 3B, left).

Figure 3. Cells with HDAC3 functional loss have increased generation and persistence of DNA DSBs and impaired DSB repair function.

(A, B) Left, upper: quantification of γH2AX immunostaining in SBC5 (A) and H1048 (B) cells at 0.5 hours after 4 Gy radiation. Left, lower: representative DNA DSBs detected by γH2AX sub-nuclear foci immunostaining in SBC5 (A) and H1048 (B) cells at 0.5 hours after irradiation. Middle, upper: quantification of γH2AX immunostaining in SBC5 (A) and H1048 (B) cells at 48 hours after 4 Gy radiation. Middle, lower: representative DNA DSBs detected by γH2AX sub-nuclear foci immunostaining in SBC5 (A) and H1048 (B) cells at 48 hours after irradiation. Right: average number of γH2AX foci per nucleus, over a time course of 48 hours from 4 Gy IR exposure, in SBC5 (A) and H1048 (B) cells. Cells were pre-treated with RGFP966 for 5 days prior to IR exposure, followed by drug removal. Data are presented as mean ± SD of three replicates. (C-E) Chromatin-based repair assay to assess canonical NHEJ (C), mutagenic NHEJ (D), and HR (E) function in U2OS cells treated with 5 μM RGFP966. Left: representative flow cytometry gates used to define GFP+ or RFP+ cells in the control (top) and RGFP966-treated conditions (bottom). Right: quantification of %GFP+ or %RFP+ cells for each condition, normalized to the control. Data are presented as mean ± SD of ≥3 replicates. (F) Western blot depicting the effect of IR (4 Gy), HDAC3 inhibition by RGFP966 and their combination on DNA damage. Cells were processed 6 hours post IR. pATM (Ser 1981), total ATM, p-DNA-PKcs (Ser 2056), total DNA-PKcs, RAD51, and Ku70 protein levels were detected. All protein levels were normalized to GAPDH. Data represents a single experiment (n=1). (G, H) Synergy distribution in pairwise combinations of RGFP966 with Olaparib (G) or Talazoparib (H) in SBC5 (left) and H1048 (right) cells. Data represents mean ± SD of three replicates. For A and B, comparisons were calculated with a one-way ANOVA Dunnett’s multiple comparisons test. For C-E, comparisons were calculated with an unpaired Student’s T-Test. (*, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001).

Cells with HDAC3 functional loss Demonstrate Persisting DNA Damage and Impaired DNA DSB Repair Function

We next sought to determine how HDAC3 functional loss in cells affected DNA DSB resolution following irradiation. We stained for γH2AX foci in SBC5 and H1048 cells fixed at 6, 24, and 48 hours after 4 Gy IR exposure. Compared to cells in the control group, which demonstrated complete resolution of γH2AX foci after 48 hours of recovery, cells with HDAC3 functional loss demonstrated persisting γH2AX foci, consistent with persisting DNA DSBs. The average number of γH2AX foci per nucleus at 48 hours was greater in cells with HDAC3 functional loss (SBC5: 6.879 ± 4.963 with p <0.0001 for shHDAC3 #1, 9.130 ± 4.867 with p<0.0001 for shHDAC3 #2, 11.82 ± 8.181 with p<0.0001 for 5 μM RGFP966, Fig. 3A, middle; H1048: 4.837 ± 3.837 with p=0.0103 for shHDAC3 #1, 7.308 ± 5.893 with p<0.0001 for 5 μM RGFP966, Fig. 3B, middle) than the control cells (SBC5: 3.968 ± 3.060; H1048: 3.721 ± 2.974). A similar trend, though not statistically significant, was also observed in shHDAC3 #2 in the H1048 cell line (Supplementary Fig. 3B, right). When comparing the average number of γH2AX foci per nucleus over a time course of 48 hours, the control cells demonstrated a rapid decrease in the number of γH2AX foci at 6, 24, and 48 hours after IR (Fig. 3A, right; Fig. 3B, right). In contrast, cells with HDAC3 functional loss demonstrated a slower decrease in the number of γH2AX foci at each timepoint and a plateau-like effect after 24 hours, suggesting suppressed DNA DSB repair function (Fig. 3A, right; Fig. 3B, right; Supplementary Fig. 3A and 3C).

To investigate the effects of HDAC3 functional loss on DSB repair, we functionally evaluated canonical non-homologous end-joining (c-NHEJ), mutagenic NHEJ (m-NHEJ) ²⁹, and homologous recombination (HR) repair pathways of DNA DSB repair in U2OS cells stably expressing EJ5-GFP, EJ-RFP, or DR-GFP reporters, respectively. Cells were transfected with the I-SceI endonuclease to generate a DSB in the reporter sequence, where repair by c-NHEJ and HR allows for GFP expression and repair by m-NHEJ allows for RFP expression. Cells treated with RGFP966 demonstrated a 57.9% reduction in c-NHEJ (p<0.001), 57.3% reduction in m-NHEJ (p<0.01), and a 71.1% reduction in HR (p<0.0001) repair function compared to the control (Fig. 3C–E), suggesting that HDAC3 function may be essential for efficient NHEJ- and HR-mediated DSB repair.

To better understand how HDAC3 depletion impairs DSB repair, we performed western blots on key DNA damage response (pATM and ATM) and DSB repair (RAD51, Ku70, DNA-PKcs, and p-DNA-PKcs) proteins in SBC5 and U2OS cells treated with 5 μM RGFP966 and 4 Gy IR. Cells treated with the combination of HDAC3i and IR demonstrated increased ATM activation through elevated pATM levels compared to IR treatment alone, suggestive of increased DNA damage response signaling (Fig. 3F). A similar trend was also observed in SBC5 and U2OS HDAC3 KD cells, where pATM levels were elevated in the shHDAC3 + IR condition, compared to IR only (Supplementary Fig. 3D). Moreover, cells treated with the combination of RGFP966 and IR also demonstrated decreased p-DNA-PKcs and RAD51 expression, proteins involved in NHEJ and HR, respectively, indicating decreased DSB repair function (Fig. 3F). Interestingly, we did not observe any discernable effect on Ku70 expression, a key player that acts upstream of DNA-PKcs in the NHEJ repair pathway (Fig. 3F).

Given the compromised DSB repair pathways in HDAC3-defective cells, we examined whether these cells exhibited increased sensitivity to PARP inhibitors, the combination of which could promote DNA damage through an impaired repair system, leading to enhanced cell death. We used the Zero interaction potency (ZIP) model to determine the degree of synergy in the combination of HDAC3i, RGFP966, with either PARP inhibitors, Olaparib or Talazoparib (a synergy score of < −10 indicates antagonism, −10 to 10 indicates additive effect, >10 indicates synergy). We found that the combination of RGFP966 with either PARP inhibitor had a strong additive to synergistic effect in the SBC5 and H1048 cell lines. SBC5 cells treated with RGFP966 and Olaparib had a mean ZIP synergy score of 3.24 (p<0.01) suggesting an overall additive effect; however, the MSAS was higher at 8.30 with the maximum score at 17.31, suggesting synergism (Fig 3G, left). Similarly, when treated with Talazoparib, the MSAS and maximum ZIP scores at 17.92 and 20.60, respectively, suggested synergy, with a mean ZIP score of 3.16 (Fig. 3H, left). Likewise, the H1048 cell line had mean ZIP scores suggesting an additive effect (Olaparib = 3.68 with p<0.0001; Talazoparib = 2.44 with p<0.001), although the maximum scores (Olaparib: max = 14.21; Talazoparib: max = 13.60) suggested synergism (Fig. 3G, right; Fig. 3H, right). Collectively, the data suggests that rather than being broadly synergistic, RGFP966 has synergistic interactions at specific PARP inhibitor concentrations. Moreover, these findings demonstrate that SCLC cells with HDAC3 functional loss are more vulnerable to PARP inhibitor treatment, potentially through promoting DSB formation via impaired DSB repair, leading to cell death.

Taken together, our results demonstrated that the functional loss of HDAC3 increased chromatin accessibility, associated with greater DNA damage propensity and therefore, increased generation of radiation-induced DNA DSBs. In parallel, cells with HDAC3 functional loss showed reduced NHEJ and HR repair functions, suggesting its role in deficient DNA repair and DSB resolution, yielding persisting DNA damage following irradiation and increased sensitivity to PARP inhibition.

HDAC3 functional loss Enhances Anti-Tumor Efficacy of Radiation in SCLC Xenografts

To assess the radiosensitizing effect of HDAC3 functional loss in vivo, we first evaluated the anti-tumor activity of HDAC3 KD with radiation in the SBC5 xenograft model of SCLC. The SBC5 cell line was chosen based on its higher DMF in clonogenic assays with HDAC3 KD—particularly with the shHDAC3 #2 construct—and its use in our CRISPR screen to identify HDAC3 as a target for radiosensitization. Mice were engrafted with SBC5 shSCR or shHDAC3 cells and given localized IR of 8 Gy in four daily fractions at the tumor site on days 1–4, once tumors reached an approximate tumor volume (TV) of 150 mm³. This radiation schedule minimized toxicity while reflecting clinically relevant fraction sizes of 2 Gy or 3 Gy. HDAC3 KD alone achieved 50% tumor growth inhibition (TGI) on day 13 (TV = 468.49 mm³ with p<0.01 for shHDAC3 #2; TV = 448.801 mm³ with p=0.012 for shHDAC3 #3; Supplementary Fig. 4A) relative to shSCR (TV = 934.35 mm³) and significantly increased the median time to the volumetric endpoint (shSCR = 15 days vs shHDAC3 #2 = 32 days and shHDAC3 #3 = 25 days, p <0.01 for both; Supplementary Fig. 4A). The combination of HDAC3 KD with IR demonstrated significant TGI on day 39 (TV = 414.46 mm³ with p<0.0001 for shHDAC3 #2; TV = 719.28 mm³ with p<0.01 for shHDAC3 #3; Fig. 4A, left) compared to shSCR with IR (TV = 1062.81 mm³), and significantly increased the median time to the volumetric endpoint (shSCR+IR = 27 days vs shHDAC3 #2+IR = 65 days and shHDAC3 #3+IR = 44 days, p <0.01 for both; Fig. 4A, right).

Figure 4. The combination of HDAC3 functional loss and IR improves tumor control in SCLC xenograft models.

(A, B) Tumor growth curves (left) and Kaplan-Meier plots (right) of (A) SBC5 and (B) H1048 cell-line derived xenograft models. (A) shSCR control and HDAC3 KD xenografts were treated with 8 Gy radiation in four daily fractions (days 1–4). (B) H1048 xenografts were treated with either 14 doses of 40 mg/kg RGFP966, 9 Gy radiation in three daily fractions (days 4–6), or the combination. (C) Western blot and quantification of H3K9K14ac expression in H1048 tumors. H3K9K14ac bands were normalized to Histone 3. Each lane represents individual tumor samples isolated from individual mice. (D) Illustration of HDAC3 functional loss-mediated radiosensitization. In cells with functional loss of HDAC3, histone acetylation is retained, leading to decreased chromatin compaction. Accessible or open chromatin is more inclined to radiation-induced DNA damage, generating more DSBs. In parallel, HDAC3 functional loss suppresses DSB repair by HR and NHEJ, resulting in unrepaired DSBs that persist longer in the cell. Created in BioRender.com. For A-B, a Kaplan-Meier event was defined as tumor volume over 1000 mm³. Comparisons were calculated with a one-way ANOVA Tukey’s multiple comparisons test for tumor growth curves and the Log Rank Test for the Kaplan-Meier plots. (*, p<0.05; **, p<0.01; ****, p<0.0001)

Next, we tested the radiosensitizing effect of RGFP966 in the H1048 xenograft model of SCLC. Our rationale for selecting the H1048 cell line to evaluate the combination of RGFP966 with radiation in vivo was based on the clonogenic assays from Fig. 1G, where the same dose of RGFP966 (1.0 μM) achieved greater radiosensitization in H1048 cells than in SBC5 cells. Mice were injected with 40 mg/kg RGFP966 daily for a total of 14 doses, and IR of 9 Gy in three daily fractions was administered to the tumor site on days 4–6 from the start of drug treatment. This radiation schedule allowed for a shortened treatment duration to reduce cumulative toxicity associated with combination treatment, while still maintaining consistency with the SBC5 xenograft model. RGFP966 monotherapy did not achieve significant TGI compared to the vehicle arm (Fig. 4B, left; Supplementary Fig. 4B). However, the combination of RGFP966 with IR led to significant TGI on day 34 (TV = 389.34 mm³, p=0.005; Fig. 4B, left) compared to the IR arm (TV = 666.37 mm³), and increased the median time to the volumetric endpoint (IR = 41 days vs RGFP966+IR = 48 days, p = 0.019; Fig. 4B, right).We observed initial weight loss during the treatment period in both RGFP966 monotherapy and combination therapy arms; however, mice body weights stabilized following treatment cessation (Supplementary Figure 4C). Abdominal distension was also noted in both arms, which improved after treatment but did not fully resolve. HDAC3 inhibition by RGFP966 was confirmed in tumors by an increase in H3K9K14ac levels in the drug monotherapy and RGFP966+IR combination arms, compared to the vehicle and IR arms (Fig. 4C). Each lane represents individual tumor samples isolated from individual mice.

Discussion

Human HDAC enzymes consist of 18 genes divided into four main classes, where HDAC3 is a member of class I HDACs alongside HDAC1, HDAC2, and HDAC8 ³⁶. In this study, using a custom Epi-Drug CRISPR screen, we found that HDAC3 appeared as the sole member of the HDAC family that ranked below the FDR cutoff of 0.1 across all screen timepoints. Despite functional and structural similarities shared by class I HDACs, the statistical significance of HDAC3 suggests its unique role in mediating SCLC radiosensitization.

Subsequent validation confirmed increased radiosensitivity using genetic perturbation through shRNA-mediated KD or pharmacological inhibition with the HDAC3i, RGPF966, in both SCLC cell lines and xenograft models. Mechanistically, we found that the loss of HDAC3 function increased chromatin accessibility, which was associated with significantly higher levels of radiation-induced DNA DSBs (Fig. 4D). These DSBs also persisted longer in cells with HDAC3 functional loss, which was associated with impaired DSB repair by NHEJ and HR repair pathways (Fig. 4D). To our knowledge, we report the first study on targeting HDAC3 as a radiosensitizing therapeutic strategy in SCLC.

HDAC3 upregulation has been observed in several cancers, such as human hepatocellular carcinoma, liver, colorectal, and breast cancers ^22,37–40, where its aberrant expression has been associated with advanced disease and poor patient outcomes ^22,24. The role of HDAC3 in cancer and its therapeutic potential has been preclinically studied in some cancers. For example, studies in cholangiocarcinoma and pancreatic cancers have demonstrated that the overexpression of HDAC3 promoted tumor cell proliferation, migration, and invasion ^41,42, while HDAC3 KD or inhibition with an HDAC2/3 inhibitor, MI192, decreased cell-line and xenograft tumor growth in cholangiocarcinoma ⁴¹. However, the potential of targeting HDAC3 as a radiosensitizer in cancer treatment has not been explored outside of Cassandri et al.’s 2024 study that found HDAC3 to be a unique radiosensitizer among class I HDACs to FP-RMS in vitro and in vivo ²². As such, much remains to be uncovered about HDAC3’s radiosensitizing role in different cancer types, highlighting the novelty and significance of the present study.

Histone acetylation is a well-known epigenetic modification that promotes an open chromatin structure, allowing transcription machinery better access to promoter regions and transcriptional activation ³⁴. With respect to HDAC3, Bhaskara et al. (2010) demonstrated that HDAC3 deletion in hepatocytes in vivo greatly reduced chromatin compaction and heterochromatin content by transmission electron microscopy (TEM) ¹⁵. In the present study, we demonstrated that the functional loss of HDAC3 increased H3K9K14ac levels (by immunoblotting) and genome-wide chromatin accessibility by ATAC-seq. The advantage of using ATAC-seq, as compared to TEM, is the ability to identify differentially accessible chromatin regions between the control and HDAC3-perturbed conditions, and greater quantitative assessment of the direction and magnitude of the difference. In our analysis, more than 50% of differentially accessible peaks that were more accessible in cells with HDAC3 functional loss than the control, were annotated to promoter regions. This suggests an increase in the transcriptional activity of genes regulated by HDAC3 function, influencing downstream cellular processes. For example, a 2007 study demonstrated that HDAC3 relocalization to the cytoplasm and proteolytic cleavage increased histone acetylation and transcriptional activation of a pro-apoptotic HDAC3-target gene, the Fas-encoding gene, allowing for efficient apoptosis induction ⁴³. It is possible that HDAC3 functional loss mediates an increase in chromatin accessibility resulting in the transcriptional activation of pro-apoptotic HDAC3-target genes that increase radiation-induced apoptosis. The various roles of HDAC3 in the cell encourages further exploration of other mechanisms contributing to HDAC3-mediated radiosensitization, such as its effects on the cell cycle and apoptosis. Separately, we observed an interesting difference in total histone H3 levels between RGFP966-treated and HDAC3 KD cells. Cells treated with RGFP966 showed a dose-dependent increase in total H3 alongside elevated H3K9K14ac, whereas HDAC3 KD did not alter total H3 levels, despite a similar increase in acetylation. This distinction may reflect fundamental differences in how cells respond to acute enzymatic inhibition versus sustained protein loss. As a rapid and selective HDAC3 inhibitor ³³, RGFP966 may transiently affect chromatin accessibility, histone biosynthesis, or cell cycle progression ¹⁵, while the slower kinetics of shRNA-mediated KD may allow for compensatory regulation of histone gene expression ⁴⁴. Future studies examining HDAC3’s impact on histone dynamics, including synthesis, turnover, and incorporation during cell cycle progression, may help further clarify these differential effects.

Existing studies ^16,45 have shown that cells with HDAC3 loss exhibited delays in progressing through the S phase of the cell cycle, suggesting difficulties in completing DNA replication compared to control cells. While cells in S phase are less sensitive to radiation ⁴⁶, it is possible that insufficient repair occurs after HDAC3-deficient cells caught in the S phase are irradiated, leading to cell death. Our data supports this idea as we found that cells with HDAC3 functional loss generated more radiation-induced DNA DSBs that persisted longer, compared to control cells. Bhaskara et al. (2008) reported similar findings in mouse embryonic fibroblasts (MEF) where HDAC3^−/− cells displayed more γH2AX foci immediately after IR treatment, and 10% of HDAC3^−/− MEFs displayed persisting markers of DNA damage 24 hours post-irradiation ¹⁶. Our γH2AX IF observations were further supported by our western blots of DSB repair proteins which demonstrated a greater activation of ATM (as shown by pATM/ATM) in cells treated with the combination of HDAC3 functional loss and IR, compared to cells treated with IR alone. Since activated ATM is one of the earliest responders to DSBs and is involved in phosphorylating H2AX into γH2AX, the greater levels of pATM are suggestive of elevated DNA damage signaling in cells treated with combination treatment. In a separate study, Bhaskara et al. (2010) demonstrated a 50–60% reduction in NHEJ and HR repair function in HEK293 cells depleted of HDAC3 ¹⁵. The present study assessed m-NHEJ repair function, in addition to c-NHEJ and HR function, highlighting another aspect of novelty in this study. We showed that HDAC3 inhibition with RGFP966 in U2OS cells, resulted in a 60–70% reduction in repair function for all three repair pathways. m-NHEJ or non-canonical NHEJ pathways function as a backup for when the canonical DSB repair pathways, that is c-NHEJ and HR, are compromised ^29,47. Therefore, our results suggested that HDAC3 inhibition suppressed both major and backup DSB repair pathways, resulting in persisting DNA DSBs that can lead to genomic instability and cancer cell death.

Studies evaluating class I and II HDAC inhibitors, such as sodium butyrate (NaB) and trichostatin A (TSA), in combination with radiation in human melanoma and non-small cell lung cancer cells, respectively, have demonstrated that cells treated with these inhibitors have reduced expression of DNA repair proteins Ku70, Ku80, and DNA-PKcs, leading to impaired DNA repair capacity and persisting γH2AX foci ^48,49. Likewise, we show that HDAC3-deficient cells exposed to IR had reduced phospho-DNA-PKcs and RAD51 levels, pointing towards impaired DSB repair. Interestingly, similar to Cassandri et al.’s findings ²², we did not see reduced expression of Ku70 levels, another crucial protein found more upstream to DNA-PKcs in the NHEJ repair pathway. This suggests that the NHEJ deficiencies in HDAC3-depleted, IR-exposed cells result from reduced DNA-PKcs activation and its subsequent downstream effects, rather than from defects earlier in the pathway. PARPi have been recognized to work well in cancers deficient in DNA repair mechanisms, particularly DSB repair ⁵⁰. We observed increased sensitivity to PARP inhibitors in HDAC3-deficient SCLC cells, consistent with HDAC3’s role in DSB repair and suggesting a promising, though exploratory, therapeutic avenue beyond the scope of this study. The present study provides an important foundation to mechanistically explain HDAC3 functional loss mediated radiosensitization.

From a translational perspective, the present study is the first to investigate HDAC3 functional loss as a radiosensitizing strategy in SCLC, with both in vitro and in vivo validation. We evaluated the potential of targeting HDAC3 to enhance radiosensitivity in two SCLC xenograft models (SBC5 and H1048) and assessed the combined anti-tumor efficacy of HDAC3 loss and radiation. In the SBC5 xenograft model, HDAC3 KD tumors treated with radiation showed a significant radiosensitizing effect, with tumor growth significantly reduced compared to radiation alone. We acknowledge that the growth-inhibitory effects of HDAC3 loss ^41,42 may confound interpretation of our radiosensitization results, representing a limitation of the study. However, therapeutically, HDAC3 inhibition may offer dual benefits by both impairing tumor growth and enhancing radiation efficacy. Future studies are needed to separate the relative contributions of HDAC3 loss mediated growth inhibition versus radiosensitization. In the H1048 xenograft model, RGFP966 also yielded radiosensitization; however, this model is less confounded by proliferation effects, as HDAC3 function remained intact during tumor establishment and early growth, and pharmacologic inhibition was only introduced after tumors were measurable.

Separately, in the H1048 xenograft model, abdominal distension and weight loss were noted in mice receiving RGFP966. While body weights stabilized after treatment completion, abdominal distension persisted to endpoint. Future work should focus on dose-ranging studies to identify effective and better-tolerated doses of RGFP966, as well as explore alternative HDAC3 inhibitors with improved tolerability. HDAC3 inhibition as a single agent has been studied in different cancers, where slowed cell proliferation and tumor growth were observed ^51–53. However, it has not been studied in combination with radiation in SCLC, a novel aspect of the present study. Altogether, our in vivo findings not only demonstrate effective radiosensitization through HDAC3 functional loss but also position HDAC3 as a promising and previously unrecognized therapeutic target in SCLC.

While we assessed the role of HDAC3 in several SCLC-xenograft models, future work would leverage the use of an allograft syngeneic model to evaluate the interactions between HDAC3-deficiency, radiosensitization and the immune system. Current literature points to HDAC3 playing an important role in the regulation of T cells’ development and activity ^53–55. It is possible that HDAC3 functional loss would result in enhanced tumor control in syngeneic mouse models through increased expression of chemokines that improve T cell recruitment to the tumor microenvironment ^53,54 and potentially through supporting T differentiation into a more potent state as cytotoxic effector cells ⁵⁵. Regardless, our findings establish a strong basis for continued investigation of HDAC3’s role in cancer care.

An emerging area of interest in SCLC research is the identification of molecular subtypes, which include ASCL1 (A), NEUROD1 (N), POU2F3 (P), and YAP1/inflamed (Y/I) ^56,57. These subtypes display distinct transcriptional programs and may exhibit differential therapeutic vulnerabilities ^56,57. Although subtype-specific responses were not the primary focus of our study, we observed a potential trend suggesting that neuroendocrine-low subtypes (P and Y/I) may be more susceptible to the combination of HDAC3 functional loss and radiation, than neuroendocrine-high subtypes (N and A). Our study included four representative cell lines: SBC5 (Y/I), H1048 (P), H446 (N), and SHP77 (A). Clonogenic assays and xenograft studies using the SBC5 (Y/I) and H1048 (P) cell lines both demonstrated significant radiosensitization following HDAC3 functional loss. In contrast, our in vitro data for the H446 (N) cell line showed a trend toward radiosensitization, but did not reach statistical significance, and for the SHP77 (A) cell line, we did not observe any radiosensitization. Although limited by the number of cell lines, these findings warrant further investigation into subtype-specific responses to HDAC3-targeted radiosensitization. Future studies with a broader panel of SCLC cell lines representing each subtype are necessary to determine whether molecular subtyping can guide the clinical utility of HDAC3-targeted radiosensitization strategies.

In summary, we identified HDAC3 as a novel radiosensitizing target in SCLC cell lines and xenografts. We functionally demonstrated that the loss of HDAC3 function increased chromatin accessibility and suppressed DSB repair, which led to significantly higher levels of radiation-induced DNA damage that persisted longer, ultimately leading to its radiosensitizing effects. The present study lays the foundation for HDAC3 functional loss-mediated radiosensitization as a potential strategy in the treatment of SCLC.