The basic characteristics of the patients who participated in this study have been described in our previous study.1213 Briefly, subjects were patients diagnosed with SCLC and received chemotherapy from 1997 to 2017 at Kyungpook National University Hospital (KNUH). To determine the effect on chemotherapy, patients who received more than two cycles of either etoposide–cisplatin or irinotecan–cisplatin chemotherapy, which were then the primary treatment, were included in this study. Patients who received concurrent chemoradiation were excluded to avoid the effect of radiation on the chemotherapy response, but patients who received radiation sequentially after chemotherapy were included. A total of 261 patients were enrolled. Genomic DNA samples were extracted from the patients’ peripheral blood lymphocytes before starting first-line chemotherapy. The chemotherapy response was evaluated for every two cycles of chemotherapy and analyzed by dividing into responders (complete or partial response) and nonresponders (stable or progressive disease).

We performed ChIP-seq to select putative functional SNPs. Experiments were conducted on H82 and H524 cell lines with high NeuroD1 expression among four SCLC cell lines (Supplementary Fig. 1). SNPs in H3K4me3, as one of the histone modification markers, related to promoter regions were also evaluated. Through ChiP-seq of NeuroD1 and H3K4me3, SNPs in promoter regions with high peaks in both NeuroD1 and H3K4me3 were selected as putative functional SNPs. In the H82 and H524 cell lines, 435 and 1,302 SNPs were collected, respectively, in the regions where the NeuroD1 peak and H3K4me3 peak coincide. Among 1,737 SNPs, 230 were prepared for genotyping, except for SNPs with low minor allele frequencies (≤ 0.2 by HapMap-JPT data) and those in linkage disequilibrium (r² > 0.8) assessed using TagSNP in the SNPinfo web server (https://snpinfo.niehs.nih.gov/). After excluding 13 SNPs with genotyping failure and 46 SNPs with a deviation from the Hardy–Weinberg equilibrium (P < 0.05) or a low call rate (< 95%), 171 SNPs were evaluated for the association with the clinical outcomes of patients with SCLC (Supplementary Table 1). Genotyping was performed using an iPLEX Assay and MassARRAY System (Agena Bioscience, San Diego, CA, USA).

The H82 and H524 cells were obtained from the American Type Culture Collection. Cells were cultured at 37°C in a humidified atmosphere with 5% CO₂ in RPMI medium (Corning Inc., Corning, NY, USA) supplemented with 10% FBS (Corning Inc.), 100 U/mL penicillin, and 100 mg/mL streptomycin. Anti-NeuroD1 antibody (4373) from Cell Signaling Technology (Danvers, MA, USA) and H3K4me3 antibody (ab8580) from Abcam (Cambridge, UK) were used in this study.

ChIP-seq was essentially performed in the H82 and H524 cells as previously described17 using anti-NeuroD1 and H3K4me3 antibodies. ChIP-seq library was prepared using a TruSeq ChIP Library Preparation Kit (Illumina, San Diego, CA, USA). Sequencing was performed on an Illumina NovaSeq 6000 system. Sequence reads for each sample were aligned to the human genome (hg19) using Bowtie (version 1.1.2 parameter set -n 2 -m 1 -X 500),18 allowing up to two nucleotide mismatches to the reference genome per seed and returning only uniquely mapped reads. Mapped data (SAM file format) were sorted and indexed using SAMtools (version 1.9).19 Then, duplicate reads were excluded using a MarkDuplicates module in Picard (version 1.118). Peaks were called in the aligned sequence data using a model-based analysis of ChIP-seq (MACS2 version 2.1.1).20 ChIPseeker (version 1.6.6),21 a bioconductor package within the statistical programming environment R (version 3.5.1) to facilitate batch annotation of enriched peaks, was used to identify nearby genes and transcripts from the peaks obtained from MACS2. For variant calling, we identified regions where the peaks of the samples for each combination overlap. In the BAM file for each sample, reads aligned to the region were used in the downstream analysis. The aligned reads that can be used in the analysis were created through Split ‘N’ Trim, mapping quality reassignment, indel realignment, and base recalibration process. The aligned reads are used for variant calling with HaplotypeCaller. Variant filtering for each sample was performed based on Fisher Strand values (> 30.0) and Qual By Depth values (< 2.0) in the VariantFilteration module of GATK v2015.1-3.4.0-1-ga5ca3fc.222324 The datasets for ChIP-seq (NeuroD1 and H3K4me3) have been deposited and are available at Gene Expression Omnibus (accession no. GSE211475) (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE211475).

Chromatin from H82 and H524 cells was immunoprecipitated using the Pierce Magnetic ChIP kit (Thermo Fisher Scientific, Waltham, MA, USA) using 10 µg of anti-NeuroD1 and anti-H3K4me3 antibodies and 2 µg of normal rabbit IgG antibodies per immunoprecipitation reaction. ChIP DNA was subjected to real-time qPCR using SYBR Green PCR Master Mix (Qiagen, Hilden, Germany) with the following primers: ring finger protein 145 (RNF145) forward, 5′-GGCTCGCGAAGAAAAACGCG-3′; RNF145 reverse, 5′-CCCTTCTGTCTTCTCCCTCC-3′; cyclin-dependent kinase 2 interacting protein (CINP) forward, 5′-CCTCACTCCGCTCCAGTTTC-3′; and CINP reverse, 5′-CAGGGGCTCCCCTTTACT-3′. ChIP–qPCR enrichment analysis was performed using the comparative Ct method. Each sample was normalized to the input, and the fold difference between the sample and IgG was calculated using 2(−ΔΔCt).25

We determined the effect of rs2043268A>G or rs762105A>G on the activity of the promoter of RNF145 or CINP using a luciferase reporter assay. A 393-bp fragment including rs2043268A>G and a 535-bp fragment including rs762105A>G were synthesized by PCR using human genomic DNA with the following primers: RNF145 forward, 5′-CCGCTCGAGTCCCTTCGTCTTACCCTGCT-3′; RNF145 reverse, 5′-CCCAAGCTTTATCTCGCCAAGGGTGCTGA-3′; CINP forward, 5′-CCGCTCGAGCAGGCCTTGACACTCATCCA-3′; and CINP reverse, 5′-CCCAAGCTTCTGGTATCCCACATTCTGCC-3′. The PCR products were cloned into the XhoI/HindIII site of the pGL3-basic vector (Promega, Madison, WI, USA). The correct sequences of all clones were confirmed by DNA sequencing. All constructs were confirmed by direct sequencing before use. The SCLC cell line H446 was transfected with each reported construct and pRL-SV40 Vector (Promega) using a lipofectamine 2000 transfection reagent (Thermo Fisher Scientific) according to the manufacturer’s protocol. The cells were collected 48 hours after transfection, and lysates were prepared using the Dual-Luciferase Reporter Assay System (Promega). The luciferase activity was measured using a Synergy HTX Multi-Mode Microplate Reader (BioTek Instruments). Firefly luciferase activity measurements were normalized with respect to pRL-SV40 Renilla luciferase activity to correct for variations in transfection efficiency.

The response to chemotherapy was analyzed as the ratio of responders and nonresponders according to clinical variables and genotypes. The adjusted odds ratio (aOR) and 95% confidence intervals (CIs) were calculated using logistic regression analysis. For survival assessment, the period from the first day of chemotherapy to the date of patient death or last follow-up was calculated and defined as the overall survival (OS). The log-rank test and Kaplan-Meier method were used to analyze the estimated OS according to the clinical variables and genotypes. To determine the association between SNPs and OS, the adjusted hazard ratio (aHR) and 95% CIs were calculated using multivariate statistical models (Cox proportional hazards models). The following clinical factors were adjusted: age, sex, smoking status, clinical stage, Eastern Cooperative Oncology Group performance status, weight loss, chemotherapy regimen, second-line chemotherapy, neuron-specific enolase, and radiation to the tumor. All statistical analyses were conducted using the Statistical Analysis System version 9.4 for Windows (SAS Institute, Cary, NC, USA).

This study was approved by the Institutional Review Board of Kyungpook National University Hospital (Approval No. KNUCH 2020-03-040). Blood samples for genotyping were provided by the National Medical Center-KNUH Biobank with support from the Ministry of Health, Welfare and Family Affairs. Informed consent was submitted by all subjects when they were enrolled. This study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines.

The chemotherapy response and OS according to the clinical variables of 261 patients are presented in Supplementary Table 2. A total of 190 patients had a complete or partial response after first-line chemotherapy, with a response rate of 72.8% (95% CI, 67.4–78.2). Among the two chemotherapy regimens, the irinotecan–cisplatin regimen showed a higher response rate than the etoposide–cisplatin regimen, but there was no difference in OS. The median survival time was 10.5 months (95% CI, 9.3–11.4). Age, clinical stage, Eastern Cooperative Oncology Group performance status, weight loss, neuron-specific enolase, second-line chemotherapy, and radiation to tumor were the variables that affected OS (Supplementary Table 2). These variables were adjusted in the association analysis between SNPs and clinical outcomes.

Among the examined SNPs, two were associated with both chemotherapy response and OS. RNF145 rs2043268A>G was significantly associated with worse chemotherapy response and OS (under a recessive model, aOR, 0.50, 95% CI, 0.26–0.94, P = 0.031, and aHR, 1.88, 95% CI, 1.38–2.57, P < 0.001; Table 1, Fig. 1). CINP rs762105A>G was also significantly associated with worse chemotherapy response and OS (under a dominant model, aOR, 0.47, 95% CI, 0.23–0.99, P = 0.046, and aHR, 2.03, 95% CI, 1.47–2.82, P < 0.001; Table 1, Fig. 1). After stratification by stage, RNF145 rs2043268 was associated with chemotherapy response and OS in both limited disease (LD) and ED (Table 2). However, CINP rs762105 was associated with chemotherapy response and OS only in ED and not in LD (Table 2).

To confirm whether the two SNPs were located in the functional promoter or enhancer sites, we conducted ChIP–qPCR assays using antibodies against NeuroD1 and H3K4me3. The rs2043268-containing region demonstrated a much stronger fold enrichment of NeuroD1 and H3K4me3, which mark the active promoter, in both H82 and H524 cell lines (Fig. 2A and B). The rs762105-containing region also demonstrated a much stronger fold enrichment of NeuroD1 and H3K4me3 than IgG (Fig. 2C and D).

The results of ChIP–qPCR assays suggested that rs2043268A>G and rs762105A>G are regulatory SNPs that could affect the promoter activity of each gene. We conducted in vitro luciferase assays to determine whether the SNPs affect the promoter activity of each gene. The promoter activity of RNF145 was significantly lower in the rs2043268 G allele than in the rs2043268 A allele (P = 0.006, Fig. 3A). The promoter activity of CINP was significantly higher in the rs762105 G allele than in the rs762105 A allele (P = 0.026, Fig. 3B).