DNA-PKcs as an upstream mediator of OCT4-induced MYC activation in small cell lung cancer

Abstract

Small cell lung cancer (SCLC) is a neuroendocrine tumor noted for the rapid development of both metastases and resistance to chemotherapy. High mutation burden, ubiquitous loss of TP53 and RB1, and a mutually exclusive amplification of MYC gene family members contribute to genomic instability and make the development of new targeted agents a challenge. Previously, we reported a novel OCT4-induced MYC transcriptional activation pathway involving c-MYC, pOCT4^S111, and MAPKAPK2 in progressive neuroblastoma, also a neuroendocrine tumor. Using tumor microarray analysis of clinical samples and preclinical models, we now report a correlation in expression between these proteins in SCLC. In correlating c-MYC protein expression with genomic amplification, we determined that some SCLC cell lines exhibited high c-MYC without genomic amplification, implying amplification-independent MYC activation. We then confirmed direct interaction between OCT4 and DNA-PKcs and identified specific OCT4 and DNA-PKcs binding sites. Knock-down of both POU5F1 (encoding OCT4) and PRKDC (encoding DNA-PKcs) resulted in decreased c-MYC expression. Further, we confirmed binding of OCT4 to the promoter/enhancer region of MYC. Together, these data establish the presence of a DNA-PKcs/OCT4/c-MYC pathway in SCLCs. We then disruptively targeted this pathway and demonstrated anticancer activity in SCLC cell lines and xenografts using both DNA-PKcs inhibitors and a protein-protein interaction inhibitor of DNA-PKcs and OCT4. In conclusion, we demonstrate here that DNA-PKcs can mediate high c-MYC expression in SCLCs, and that this pathway may represent a new therapeutic target for SCLCs with high c-MYC expression.

1. Introduction

Small cell lung cancer (SCLC), the most common neuroendocrine tumor of the lung, accounts for 15% of all lung cancer cases. SCLC is a disease found almost exclusively in smokers and is a highly aggressive cancer with extremely poor outcomes. Initially, most patients diagnosed with SCLC respond to platinum + etoposide (PE)-based therapy, but fatal relapse occurs in nearly all patients [1]. Unlike non-small cell lung cancer, extensive genomic profiling of SCLC tumors has proven difficult, as surgical resection of tumors is uncommon. Although numerous different molecular targets and pathways are dysregulated, ubiquitous loss of RB1 and TP53 function is characteristic of SCLC [2–4]. Amplification and/or overexpression of the MYC family of oncogenes (MYC, MYCL, and MYCN) occurs in a subset of SCLC that is associated with tumor progression, treatment resistance, and an overall poor outcome [5, 6]. Genomic amplification of MYC occurs in 25% of primary tumors [2], and up to 50% of SCLC cell lines [6]. Amplification of members of the MYC family of oncogenes is mutually exclusive in that only one family member drives a given cancer [5–7]. Transcriptional dysregulation also plays a key role, and c-MYC can be overexpressed in SCLC in the absence of genomic amplification, although the frequency is yet to be determined [7]. High MYC expression dramatically accelerated tumorigenesis and metastases in an Rb1/Trp53 null SCLC GEMM model, and a Rb1^fl/fl Trp53^fl/fl Myc^LSL/LSL (RPM) mice, a MYC expressing variant SCLC model, show specific sensitivity to inhibitors of aurora A/B kinases [5] and CHK1 [8].

Therapies aimed at post-transcriptionally affecting c-MYC’s stability can be subdivided into two main categories: 1) molecules that affect the phosphorylation of Ser62, a residue critical for c-MYC activation [9–15], and 2) those that increase c-MYC ubiquitination to accelerate degradation [16] [17]. Although these approaches have shown some pre-clinical promise, neither strategy has yet been deemed robust enough to justify clinical trials. Similarly, Aurora kinase A inhibitors upregulate ubiquitin-mediated degradation of c-MYC and demonstrated preclinical efficacy in a high c-MYC expressing SCLC cancers models [18] but modest clinical activity has been reported to date.

The challenge of directly targeting c-MYC, given its nuclear localization in the nucleus and lack of drug-binding pockets [19] has led to an extensive examination of upstream mediators of MYC transcription as potential targets [20]. Bromodomain and extra-terminal motif (BET) inhibitors, decreased MYC expression in models of multiple myeloma [21] and medulloblastoma [22], and they are in clinical trial for several cancers (NCT01943851, NCT03266159) [23]. In the present report, we describe that DNA-PKcs and OCT4 are key mediators of high c-MYC expression in SCLC, another neuroendocrine cancer.

2. Materials and Methods

2.1. Materials and reagents

Recombinant human His-OCT4 protein from ProteinONE; Recombinant human OCT4-mycDDK protein from OriGene; Recombinant human DNA-PK protein kinase from Creative BioMart; DNA-PK inhibitor NU7441 (KU57788) and ATM kinase inhibitor KU60019 from Tocris Bioscience; CC-115, a dual inhibitor of DNA-PK and mTOR, from Selleckchem; M3814 (nedisertib) from Chemgood; ABT-199 (venetoclax) and ABT-737 obtained from Abbott under MTA; ABT-263 (navitoclax) from MedChemExpress; All oligonucleotides were from Integrated DNA Technologies (IDT). Antibodies and their sources are included in Supplementary Methods.

2.2. Cell lines and reagents

Human small cell lung cancer cell lines were kindly provided by Drs. John Minna and Adi Gazdar in the UT Southwestern Medical Center. The genetic characteristics and the amplification status of MYC family of genes are included in Table 1. All SCLC cell lines were maintained in HyClone RPMI-1640 medium, except HEK293FT which was maintained in DMEM, with 10% fetal bovine serum at 37°C in a humidified 20% O₂ and 5% CO incubator. All cell lines used in this study were mycoplasma-free as examined by MycoAlert Mycoplasma Detection Kit (Lonza) and routinely checked for cell line identification using short tandem repeat genotyping by GenePrint^® 10 System (Promega) as compared with the original primary sample material within the CCcells database: www.CCcells.org.

Table 1.

Genetic characteristics of small cell lung cancer cell lines used in the current study.

Cell line	Calculated copy number			ASCL1	NEUROD1	YAP1	POU2F3	Primary or Metastasis
	MYC	MYCN	MYCL
NCI-H1876	N	N	7	high	low	medium	low	Met
NCI-H69	N	78	N	high	low	low	low	Met
NCI-H209	N	N	11	high	low	low	low	Met
NCI-H526	N	37	11	low	low	low	high	Met
NCI-H1963	N	N	223	high	low	low	low	Prim
NCI-H345	5	N	N	high	low	low	low	Met
NCI-H510	5	N	117	high	low	low	low	Met
NCI-H1048	4	N	N	low	low	medium	high	Met
NCI-H417	93	N	5	-	-	-	-	Prim
NCI-H510A	5	N	188	high	low	low	low	Met
NCI-H1870*	37	N	4	-	-	-	-	-
NCI-H1607	N	N	N	-	-	-	-	Met
NCI-H146	N	N	N	high	low	low	low	Met
HCC33	N	N	99	low	low	low	low	Met
NCI-H889	N	N	25	high	low	medium	low	Met
NCI-H82	151	N	14	low	high	low	low	Met
NCI-H847	59	N	7	-	-	low	low	Met
NCI-H524	111	N	6	low	high	low	low	Met
NCI-H2171	14	N	N	low	high	low	low	Met

^*NCI-H1304 and NCI-H1870 have been shown to be identical. Our STR data showed that NCI-H1870 does not match with other cell lines used in this manuscript.

The levels of ASCL1, NEUROD1, YAP1, and POU2F3 expression are determined based on the data from Cancer Cell Line Encyclopedia (CCLE), Broad Institute and Dr. Adi Gazdar, The University of Texas Southwestern.

N: not amplified (calculated copy number ≤3). “-“: not available.

2.3. In vitro cytotoxicity assays

Small cell lung cancer cells (3–5,000 cells/well depending on doubling time) were seeded and incubated in 96-well plates for 24 h before addition of inhibitors (to 1 nM – 10 µM, in 3-fold increments). Six replicates of each drug concentration were used. After 96 h of drug incubation, cellular viability was measured using the DIMSCAN assay, as outlined in previous studies [24, 25].

2.4. Immunoblotting and immunoprecipitation

The cell lines were prepared for immunoprecipitation and immunoblotting as indicated previously [26, 27]. See Supplementary Methods for the details.

2.5. Immunohistochemistry and semi-quantitation of tumor microarray (TMA)

The TMA slides of small cell lung cancer and other lung cancer clinical samples were purchased from BioMax. One set of TMA slides included tumor samples from 80 SCLC clinical samples; a second TMA set included tumors of varying histology, along with matched cancer adjacent lung tissue and normal adjacent lung tissue, from 25 lung cancers (BioMax, Cat.# LC751). Immunohistochemistry staining (IHC) for OCT4, pOCT4^S93, DNA-PKcs, and c-MYC was performed as described [28]. After staining, slides were scanned using a Zeiss Axioscan 7 Microscope Slide Scanner supported by ZEN (Zeiss). Semi-quantitative analysis of IHC images used ImageJ Fiji software as previously reported [29].

2.6. Construction of shRNAs of POU5F1 and PRKDC

The forward and reverse oligo primers were mixed together, denatured at 95°C for 5 min, and re-annealed at 37°C for 1 h to form a duplex DNA that contains the shRNA target sequence. The shRNA for POU5F1 and PRKDC inserts were then ligated into the Age1/EcoR1 restriction enzyme sites of the pLKO.1 puro and Tet-inducible lentiviral vectors, Tet-pLKO puro, respectively, with puro antibiotic marker for selection of stable clones in eukaryotic cells. The pLKO.1 puro eGFP and Tet-pLKO puro Scrambled shRNA sequence were used as a non-targeting control (pLKO.1 puro NT and Scramble shRNAs). The 5’-end DNA sequencing primer (Table S1) was used to confirm the POU5F1 and PRKDC shRNA sequences in pLKO.1 puro and Tet-pLKO puro vector.

2.7. In vivo xenograft activity experiments

Briefly, 6-to-8-week-old nu/nu mice (Envigo) were subcutaneously injected with cell line-derived xenografts (CDXs) at 10–20 million cells (100 µL) per mouse as described previously [30]. Tumor volume was measured by ½ length × width × height. Mice were randomized to treatment groups when progressively growing tumors reached 150 to 250 mm³. Formulation of treatment agents and dosing schedules are described in Fig. 6c. Mice were sacrificed once tumor volume exceeded 1500 mm³. All animal use was conducted in the TTUHSC Laboratory Animal Resources Center (LARC) under protocols approved by the Institutional Animal Care and Use Committee (IACUC).

Fig. 6. Effect of DNA-PKcs inhibition on molecular targets and cancer proliferation using in vitro and xenograft SCLC models.

a. Pharmacological inhibition of the OCT4 phosphorylation at Ser93 can down-regulate c-MYC. Human MYC-amplified and overexpressing SCLC cell lines including NCI-H82, NCI-H417, NCI-H847, NCI-H1048, and NCI-H1870 were treated separately with a variety of concentrations of DNA-PK inhibitors, M3814, and narciclasine (narcic) for 24 h. The protein total lysates were analyzed by SDS-PAGE and then immunoblotted with the antibodies as indicated. GAPDH was conducted as a positive control showing an equal loading.

b. In vitro cytotoxicity of DNA-PKcs inhibitors in SCLC cell lines. Inhibitory concentrations of 50% (IC₅₀) were assessed in 15 SCLC cell lines. IC₅₀ values of M3814 and narciclasine are presented. IC₅₀ values of other DNA-PKcs inhibitors, NU7441, CC-115, and AZD7648, are shown (Fig. S7a). Relative cytotoxicity (Panel IC₅₀/Cell line IC₅₀) was calculated to determine if c-MYC protein level affected sensitivity to DNA-PKcs inhibitors. Black bar: c-MYC amplification, gray bar: high c-MYC expression without genomic amplification, white bar: low c-MYC expression without genomic amplification. The c-MYC protein expression status was determined in comparison to the cell line with lowest c-MYC expression without MYC genome amplification. MYC genomic amplification status of the models is reported in Table 1.

c. Antitumor activity of DNA-PKcs inhibitors in combination with Bcl-2 inhibitors in SCLC xenograft models. Models were MYC-high/MYC-amplified NCI-H847 (20 × 10⁶), MYC-high/MYC-unamplified NCI-H1048 (5 × 10⁶), and MYC-low HCC-33 (5 × 10⁶) cells, injected subcutaneously between the scapula. Tumor-bearing mice randomized to control and treatment groups when tumor volume reached 150–300 mm³ were treated with: M3814 (100 mg/kg, p.o., Days 1–5 per week for six weeks, prepared in 0.5% Methocel, 0.25% Tween-20, 300 mmol/L sodium citrate buffer, pH = 2.5), ABT-263 (100 mg/kg, p.o., Days 1–5 per week for six weeks, prepared in 60% phosal 50 propylene glycol, 30% polyethylene glycol 400, and 10% ethanol), narciclasine (1 mg/kg, p.o., Days 1–5 per week for six weeks, prepared in 10% (2-hydroxyproply)-β-cyclodextrin), or ABT-199 (75mg/kg, p.o., Days 1–5 per week for six weeks). Event-free survival (EFS) of mice engrafted with NCI-H847m (top) and NCI-H1048 (middle & bottom) SCCL treated with vehicle (black), a Bcl-2 inhibitor (blue), a DNA-PKcs inhibitor (red), and the combination (purple). For the NCI-H847m, the NCI-H847 cell line was injected subcutaneously into mice, tumors were grown to 1,500 mm³ and harvested for expansion in experimental mice. * p<0.05, ** p<0.01, ***p<0.001.

2.8. Statistical analysis

Pearson coefficient correlation was used to determine the significance of the different expression in proteins using GraphPad Prism. For in vitro experiments, combination indices (CIs) were calculated using Calcusyn (Biosoft, Cambridge, United Kingdom) [31, 32]. Mouse EFS was graphically presented by Kaplan-Meier analysis, and survival curves were compared by a log-rank test where P values less than 0.05 were considered significant. All the experiments were performed in triplicate except and were consistently repeatable; for simplicity, one representative experiment for each condition is shown.

3. Results

3.1. OCT4, pOCT4^S93, and DNA-PKcs levels were increased in SCLC with high c-MYC

Mass spectrometry analysis of co-immunoprecipitating proteins from an OCT4-overexpressing SCLC cell line (NCI-H82) identified DNA-PKcs as one of the binding partners of OCT4 (Fig. S1a–b), and post-translational modification (PTM) mass spectrometry analysis identified amino acid residue Ser93 as a phosphorylation site of OCT4 (Fig. S1c–d). PhosphoMotif Finder predicted DNA-PKcs as a candidate kinase to phosphorylate the Ser93 amino acid residue of OCT4 [27]. Using custom antibodies against pOCT4^S93 and pOCT4^S111 (the specificity of anti-pOCT4^S9 antibody shown in Fig. S1e–f), we showed that DNA-PK inhibition reduced c-MYC and pOCT4^S93, but not pOCT4^S111 (Fig. S1g), demonstrating that DNA-PKcs phosphorylates the Ser93 residue.

In an initial tumor microarray (TMA) analysis of 80 small cell lung carcinomas, semi-quantitative immunohistochemical analysis of c-MYC, OCT4, and DNA-PKcs proteins showed positive correlations between c-MYC and OCT4, pOCT4^S93, and DNA-PKcs (Fig. 1a, p<0.0001 for OCT4 and pOCT4^S93, p=0.003 for DNA-PKcs; representative staining results are shown in Fig. S2a). Relative protein levels of c-MYC did not differ by tumor stage (I-III) at diagnosis (Fig. 1b, p=0.78). Also, stages of tumor did not affect pOCT4^S93, OCT4 and DNA-PKcs protein expression in SCLC tumors (Fig. S2b). In tumors with or without extensive necrosis, no significant difference in c-MYC protein level was observed between necrotic and non-necrotic tumors (Fig. 1c, p=0.56). In a second TMA set of 25 lung cancers of various histology with matched tumor-adjacent lung and adjacent normal lung tissues, protein levels of c-MYC (p=0.007), pOCT4^S93 (p<0.0001, but not OCT4, p=0.18), and DNA-PKcs (p<0.0001) were significantly higher in tumors (as a class) relative to matched tumor-adjacent lung or adjacent normal lung tissue (Fig. 1d). Exemplary staining results are shown in Fig. S2c.

Fig. 1. Positive correlations between c-MYC and OCT4, pOCT4^S93, or DNA-PKcs, in SCLC and other lung cancers.

a. Protein levels of c-MYC, OCT4, pOCT4^S93, or DNA-PKcs in SCLC cancers. Protein expression was assessed using semi-quantitative imaging analysis of tumor microarray (TMA) of SCLC patients (n = 80) (BioMax, cat# LC818c). Representative immunohistochemical (IHC) staining shown in Fig. S2a. Three SCLC were excluded from analysis due to inconsistent sample size between slides. Each dot represents an individual SCLC.

b. c-MYC protein expression by SCLC tumor stage at diagnosis. The TMA slides (BioMax, cat# from SCLC (BioCase number per stage, n = 21 (stage I), n = 29 (stage II), and n = 27 (stage III). No difference in c-MYC protein levels was observed between stages (P=0.78).

c. Comparison of c-MYC expression between necrotic (n=18) and non-necrotic (n=59) SCLC tumor samples. c-MYC expression in the TMA of 80 SCLC clinical samples were grouped by necrotic status of the tumors to compare the expression of c-MYC between necrotic tumors and non-necrotic tumors. Tumor necrosis status was defined by the commercial vendor.

d. Higher c-MYC, pOCT4^S93, and DNA-PKcs in lung cancers. Lung cancer (n = 25) tumor tissue (L) of varying histology with matched cancer-adjacent lung tissue (AT) and adjacent normal lung tissue (N) (BioMax) were semi-quantitatively evaluated for the expression of c-MYC, OCT4, pOCT4^S93, and DNA-PKcs. Tumor protein levels of c-MYC, pOCT4^S93, and DNA-PKcs, but not OCT4, were higher than matched non-tumor tissues. One adenocarcinoma was excluded from analysis for DNA-PKcs and OCT4 due to inconsistent sample size between slides, and one adenocarcinoma was similarly excluded from pOCT4^S93 analysis. Each dot represent represents a single tumor or matched tissues. P-values shown are comparison of T vs. AT or N.

In 19 SCLC cell lines, we found that c-MYC positively correlated with DNA-PKcs (p<0.001) and pOCT4^S93 (p<0.01), but OCT4 expression did not correlate with c-MYC levels (p=0.24, Fig. 2a–b). Higher c-MYC protein was not always due to genomic amplification. Of the 19 SCLC cell lines, four showed relatively higher expression of c-MYC protein without genomic amplification (Fig. 2a). POU5F1 (encoding OCT4) mRNA levels did not correlate with the mRNA levels of MYC family of genes (Fig. S3a–S3b). In contrast, the expression of c-MYC protein did positively correlated with MYC mRNA expression (Fig. S3c, p=0.01), strongly suggesting that higher c-MYC is due to higher gene transcription.

Fig. 2. Correlation between c-MYC and pOCT4^S93 or DNA-PKcs protein SCLC cell lines and direct binding of OCT4 to DNA-PKcs.

a. Gene amplification and protein expression status in SCLC cell lines (n = 19). MYC family of genes (MYCN, MYC, and MYCL); Blue: MYCL amplification, Red: MYC amplification, Brown: MYCN amplification (see Table1). Protein levels assessed by immunoblotting (IB). *higher expression of c-MYC protein without genomic amplification. mRNA expression levels of MYC, MYCL, MYCN, and POU5F1 (encoding OCT4) are shown (Fig. S3).

b. Correlations between c-MYC and pOCT4^S93 or DNA-PKcs protein levels. Protein levels of c-MYC, OCT4, pOCT4^S93, and DNA-PKcs (Fig. 2a) were quantified using ImageJ by two-way normalization using β-Actin and NCI-H1876, a cell line with lower c-MYC expression. Relative protein levels were compared using Pearson correlation.

c. Workflow schema of OCT4 recombinant fusion protein expression and purification. HTBH-DDK (Vector) or OCT4-HTBH-DDK (OCT4) recombinant protein was induced with vehicle (−) or DOX (1 μg/mL) (+) for 48 h in NCI-H82 cells stably transduced with either pCW57.1-HTBH-DDK or pCW57.1-OCT4-HTBH-DDK. Subcellular fractions (C: cytosolic, N: nuclear) were collected, and recombinant proteins were isolated by two-step purification using Ni-NTA and SA beads, in-gel digested with NH₂-terminal His-tagged AcTEV protease, eluted, concentrated, separated by SDS-PAGE and then stained with colloidal blue or immunoblotted. NCI-H82 host cells have relatively high c-MYC expression with a lower OCT4 level (Fig 2a).

d. Confirmation of OCT4 pulldown assay. Proteins of subcellular fractions were incubated with OCT4 antibody after inducing with doxycycline (DOX) for 48h. Immunoblotting confirmed the purification of OCT4. C: cytosolic, N: nuclear fraction.

e. Identification of OCT4 protein partners. Subcellular protein fractions were subjected to pulldown assay using antibodies to OCT4, DNA-PKcs (XRCC7), and the end-joining factors Ku70 (XRCC6), Ku80 (XRCC5), LIG4, XRCC4, XLF (NHEJ1), Artemis, and XLS (PAXX). PC (Positive Control): protein lysate (20 μg) of DOX-treated NCI-H82/pCW57.1-OCT4-HTBH-DDK. Input for OCT4 was the TEV-cleavage nuclear fraction (2 μg) isolated from NCI-H82/pLenti-OCT4-HTBH-DDK.

f. Endogenous OCT4 binds to endogenous DNA-PKcs. Protein lysate (500 μg) from NCI-H82 host cells was immunoprecipitated with normal rabbit IgG, anti-DNA-PKcs antibody (Left), or anti-OCT4 antibody (Right), followed by immunoblotting with anti-OCT4 or anti-DNA-PKcs antibody, respectively.

g. OCT4 binds to DNA-PKcs in a cell-free system. (Left) Purified His-OCT4 recombinant protein (0.5 μg) was incubated with different amounts of purified DNA-PKcs and pulled down by Ni-NTA beads followed by immunoblotting with anti-DNA-PKcs or anti-OCT4 antibodies. (Right) Purified OCT4-mycDDK was reacted with different amounts of purified DNA-PKcs, immunoprecipitated using anti-DNA-PKcs antibody, followed by immunoblotting with anti-OCT4 antibody.

3.2. Direct interaction of OCT4 to DNA-PKcs is independent of DNA-damage responses

To investigate OCT4-induced MYC activation via DNA-PKcs, we first further studied the characteristics of the interaction between OCT4 and DNA-PKcs. By exogenously expressing a recombinant fusion protein between OCT4 and protein tag, HTBH-DDK, using a lentiviral vector (pSW57.1) followed by two-step purification (Fig. 2c), DNA-PKcs was identified as an OCT4 binding protein in the nuclear fraction of NCI-H82 SCLC cells (Fig. 2d–e). Of note, interaction between end-joining factors and OCT 4 was not detected (Fig. 2e), likely due to the cells being in the status without DNA-damage. These data imply a non-canonical role of DNA-PKcs that is independent of DNA-damage response activation. Direct interaction of OCT4 and DNA-PKcs was demonstrated in parental NCI-H82 SCLC cells (Fig. 2f). In a cell-free system, DNA-PKcs binding to OCT4 was proportionate to the amount DNA-PKcs reacted (Fig. 2g left). In the samples that OCT4-mycDDK reacted with 0.5 or 1 µg of DNA-PKcs and immunoprecipitated with anti-DNA-PKcs antibody, OCT4 (using anti-FLAG antibody) was detected in a dose-proportionate manner (Fig. 2g right). These experiments demonstrated that OCT4 and DNA-PKcs directly interacted in the nucleus of cells in the absence of stimuli to induce DNA-damage.

3.3. OCT4 binding to MYC promoter/enhancer region to increase c-MYC protein level

In two SCLC cell lines, NCI-H82 and NCI-H847, shRNA knock-down of MYC resulted in reduced expression of cyclin A, one of the downstream targets of c-MYC (Fig. S4a left) and in decreased cell proliferation (Fig. S4a right). To establish OCT4 as the upstream of c-MYC, POU5F1 was knocked-down using shRNA, and it resulted in reduced expression of pOCT4^S93, c-MYC, and cyclin A (Fig. 3a) and slower proliferation relative to non-targeting-shRNA (NT-shRNA) transfected cells in NCI-H82 (Fig. 3b, p=0.03 for NCI-H82 and p=0.14 for NCI-H847).

Fig. 3. Effects of POU5F1 knock-down on pOCT4, pOCT4^S93, c-MYC, and Cyclin A.

a. Constitutive knockdown of the POU5F1 reduced protein levels of c-MYC and Cyclin A. pLKO.1-Puro NT shRNA (mock), and pLKO.1-Puro POU5F1–1 and POU5F1–2 shRNAs were transduced into NCI-H82 and NCI-H847 cells and selected with 1 µg/mL of puromycin for at least 2 weeks to obtain stable clones. Protein lysates (20 μg) were analyzed by SDS-PAGE and immunoblotting for POU5F1 knockdown efficiency on OCT4 and pOCT4^S93 and effects on levels of c-MYC and its downstream target, Cyclin A. GAPDH was used as a loading control. The specificity of the anti-pOCT4^S93 antibody used is shown (Fig. S1e–f).

b. Constitutive knockdown of POU5F1 reduced growth rate of MYC-overexpressing human SCLC cell lines in vitro. Growth curves were generated for NCI-H82 and NCI-H847 cells stably expressing mock (NT) or POU5F1 (Supplementary Methods). Briefly, cells (2 × 10⁵) were plated in 6-well plates and whole medium changed every 3–4 days for 10 days. Cell number was determined by trypan blue exclusion counting in triplicate.

c. Phosphorylation of OCT4 at Ser93 is critical for binding to the OCT4 binding sites of the MYC promoter/enhancer region. OCT4 binding sites 1 & 2 on the MYC promoter/enhancer region are marked in read.

Left: Structure of biotin-labeled MYC probe (Biotin-MYC-1209/−1140 dsDNA sequence). MYC promoter/enhancer sequence with OCT4 binding sites used for EMSA experiment (right) is shown. Right: EMSA experiment demonstrating that S93A mutant of OCT4 attenuates the binding of OCT4 on the MYC promoter/enhancer region. The protease AcTEV-cleaved purified recombinant proteins including OCT4 or OCT4^S93A (0.5 μg each) were incubated with 20 fmol of 70-bp biotin-labeled MYC probe, Biotin-MYC^{−1209/−1140} dsDNA for 30 min at 15°C and the protein/DNA binding capacity was determined by EMSA using 1:3000 streptavidin-HRP conjugate and ECL for band signal detection.

d. Assessment of DNA-binding ability of wild-type OCT4 and mutant OCT4S93A using Luciferase reporter assay (left) and MYC^{−1209/−1140}/MYC reporter assay (right). Empty vector, POU5F1-mycDDK, and POU5F1^S93A-mycDDK (4 µg each) were separately co-transfected with reporter gene MYC^{−1209/−1140}/DDK-MYC-mER^TM (4 µg) in HEK293FT cells. After 48 h, the equal amount of protein lysates (20 µg) was run and analyzed by SDS/PAGE and immunoblotting using specific antibodies, as indicated. Anti-ERα: DDK-c-MYC-mERTM expression, anti-DDK (FLAG): expression of exogenous mycDDK-tagged wild-type OCT4 and its mutant.

e. Mutant OCT4^S93A overexpression reduced levels of endogenous c-MYC protein and downstream target, Cyclin A. pCW57.1-MCS1–2A-MCS2 (Vector), pCW57.1-POU5F1-mycDDK (tagged wild type OCT4^S93), and pCW57-POU5F1^S93A-mycDDK (tagged mutant OCT4^S93A) were transduced into NCI-H82 cells and stable clones selected. Cells were treated with vehicle or DOX for 48 h for protein induction. Protein lysates (20 μg) were separated by SDS-PAGE and immunoblotted with the indicated antibodies. GAPDH used as loading control. While overexpression of wildtype OCT4^S93 (DOX = 0.1 and 1 μg/ml) increased c-MYC protein over endogenous levels (DOX = 0) (Left), overexpression of serine 93 mutant, OCT4^S93A (DOX = 0.1 and 1 μg/ml), reduced endogenously expressed levels of c-MYC (DOX = 0)(Middle). NOTE: Image of c-MYC and Cyclin A levels in wildtype OCT4^S93 overexpressing cells (Left) is underexposed compared to empty vector (Right) and mutant OCT4S^93A (Middle) to demonstrate increase of c-MYC and Cyclin A protein levels over endogenous production (i.e. c-MYC and Cyclin A levels when DOX = 0).

Using biotin-labeled MYC DNA sequence inclusive of the OCT4 binding regions reported previously (Fig. 3c, left), we determined that mutation of Ser93 prevented OCT4 binding to the MYC promoter region (Fig. 3c, right). Mutation of Ser93 also reduced MYC promoter activity in a HEK293FT cell reporter assay (Fig. 3d). Further, in NCI-H82 cells transfected with either DOX-inducible OCT4 wild-type, OCT4^S93A, or vector control, mutation of Ser93 abrogated an increase of c-MYC, indicating that Ser93 residue of OCT4 is critical for increasing MYC expression (Fig. 3e).

3.4. POUs domain of OCT4 is critical for the interaction with DNA-PKcs

To further identify the domain of OCT4 that interacts with DNA-PKcs, we created POU5F1 constructs with wild-type sequence or various deletion mutants (Fig. 4a). Immunoprecipitation after transient transfection into HEK293FT cells revealed that an intact POUs domain is necessary for DNA-PKcs to bind to OCT4 while the POU_HD domain is not (Fig. 4b).

Fig. 4. DNA-PKcs binds to POUs domain of OCT4 and phosphorylates OCT4 at Ser93.

a. Structure diagrams of OCT4 wild-type and mutants. 1: pCMV6-entry-mycDDK (empty vector); 2: OCT4^aa1−137-mycDDK; 3: OCT4^aa1−216-mycDDK; 4: OCT4^aa1−293-mycDDK; 5: OCT4^aa1−360-mycDDK; 6: OCT4^aa138−293-mycDDK; 7: OCT4∆^aa138−212-mycDDK; 8: OCT4∆^aa231−289-mycDDK; 9: OCT4∆^{aa138−212/231−289}-mycDDK.

b. POUs domain of OCT4 is required for DNA-PKcs binding. DNA constructs as listed in Fig. 4a, were transiently transfected into HEK293FT cells using lipofectamine^® and PLUS™ reagent for 48 h, and whole cell lysate (500 μg) was immunopreciptated with EZview™ Red anti-FLAG-M2 affinity gel (40 μL) followed by immunoblotting detection with anti-DNA-PKcs or anti-FLAG-M2 antibody for OCT4 wild type and mutants. DNA-PKcs failed to bind constructs lacking the OCT4 POUs domain. DNA-PKcs and GAPDH were used as a loading control (bottom).

c. Genetic knockdown of PRKDC (DNA-PKcs) reduced endogenous c-MYC protein expression, but not that of MYCN or MYCL. Tet-pLKO-Puro Scrambled shRNA-(negative control) and Tet-pLKO-Puro PRKDC shRNA were transduced into SCLC cell lines that were MYC-amplified (NCI-H82, NCI-H417), non-MYC-amplified (NCI-H526 (MYCN-amplified), NCI-H510A (MYCL-amplified)), non-MYC-amplified (NCI-H889 (MYCL-amplified), NCI-H1048, Fig. S4) and protein was induced with vehicle or DOX (1 μg/mL) for 48 h. Protein lysates (20 μg) were analyzed by immunoblotting detection with the indicated antibodies. GAPDH was used as a loading control. Knockdown of PRKDC (DNA-PKcs) only decreased protein levels of c-MYC and not that of other MYC family members.

d. A DNA-PK inhibitor, but not an ATM inhibitor, decreased protein levels of pOCT4^S93 and c-MYC. pCW57.1-POU5F1-mycDDK was transduced into NCI-H82 cells and protein was induced with 1 μg/mL of DOX for 24 h prior to addition of DNA-PKcs inhibitor, NU7441, or ATM inhibitor, KU60019, at the final concentrations indicted, for 24 h. Protein lysates (20 μg) were separated in 4–12% gradient SDS/PAGE gels and immunoblotted with the antibodies indicated. Treatment with DNA-PKi, NU7441, decreased levels of pOCT4^S93 and c-MYC in concert.

In four SCLC cell lines, two with and two without MYC-amplification, shRNA knock down of PRKDC (encoding DNA-PKcs) resulted in decreased expression of pOCT4^S93 and c-MYC (Fig. 4c and Fig. S4) without affecting levels of L-MYC and MYCN. To determine if phosphorylation of OCT4 at Ser93 residue is specific to DNA-PKcs in its effect on c-MYC increase, NCI-H82 cells were treated with a DNA-PKcs inhibitor (NU7441) or KU-60019, an inhibitor of ATM, another DNA-damage response kinase. While the DNA-PK inhibitor reduced pOCT4^S93 at concentrations that also reduced c-MYC levels, the effect of the ATM inhibitor on pOCT4^S93 was minimal (Fig. 4d), demonstrating that DNA-PKcs has role in OCT4 activation that distinguishes it from other DNA damage response kinases. These data demonstrate the effect of DNA-PKcs on c-MYC via Ser93 of OCT4 is distinguished from another serine/threonine kinase (ATM) that is activated by DNA double-strand breaks.

3.5. OCT4 binding motif of DNA-PKcs is identified

The molecular size of DNA-PKcs (469 kDa) makes exogenous expression challenging for protein binding experiments [33]. To overcome this issue, constructs of six DNA-PKcs fragments were prepared (Fig. 5a) and co-transfected with POU5F1 tagged with mycDDK. The constructs and their expression were confirmed in HEK293FT cells (Fig. S5a–g). Of the six, only Fragment 6, which included the kinase domain (KD), directly bound OCT4 (Fig. 5b). Fragment 6 was further dissected (Fig. 5c, Fragments A-G) and similarly co-transfected with POU5F1-mycDDK. Results showed that amino acids (a.a.) 3719 to 3784 of the kinase domain of DNA-PKcs are necessary for DNA-PKcs/OCT4 binding (Fig. 5c). A deletion mutant (amino acids (aa) 3735 to 3767) in the DNA-PKcs KD domain failed to bind OCT4 (Fig. 5d). The delimited human OCT4-binding motif (residues 3719 to 3784) does not include the catalytic site (CS: ³⁹¹⁹GIGDRHLHN³⁹²⁷) of human DNA-PKcs. Deletion of the evolutionarily conserved catalytic site (CS) in the KD domain of human DNA-PKcs resulted in loss of phosphorylation OCT4 at Ser93 (Fig. 5e). These experiments identified the binding sites of DNA-PKcs to OCT4.

Fig. 5. The OCT4-binding motif (OBM) of DNA-PKcs is required for the interaction with OCT4.

a. Structures of PRKDC and of constructs of six gene fragments used for binding experiments. A full-length PRKDC gene was divided into 6 fragments and subcloned into the empty vector pCMV6-AN-HA. Gene fragment constructs 1 – 6 are: 1, pCMV6-AN-HA-PRKDC^nt1−2742 (with HEAT1 repeat); 2, pCMV6-AN-HA-PRKDC^{nt2734−3756} (with HEAT2 repeat); 3, pCMV6-AN-HA-PRKDC^{nt3742−6726} (with Lzip and NUC motifs); 4, pCMV6-AN-HA-PRKDC^{nt6715−7599} (with KIP domain); 5, pCMV6-AN-HA-PRKDC^{nt7588−10464} (with KIP and FAT domains); 6, pCMV6-AN-HA-PRKDC^{nt10459−12384} (with KD and FATC domains). HEAT: Huntington, EF3, PP2A, and TOR1 kinase; Lzip: leucine-zipper; NUC: NUC194 region; KIP: DNA-PKcs kinase-interacting protein domain; FAT: FRAP/mTOR, ATM, and TRRAP; KD: PI3K/PI4K kinase domain; FATC: FAT COOH-terminal domain; nt, nucleotide.

b. DNA-PKcs^{nt10459−12384} is required for binding to OCT4. The PRKDC DNA fragments 1 – 6 were co-transfected individually with pCMV6-POU5F1-mycDDK into HEK293FT cells. After 48 h transfection, the total protein lysate (500 μg) was IP with EZview™ Red anti-FLAG-M2 affinity gel (40 μL) and immunoblotting with anti-HA or anti-FLAG-M2 antibody. Only Fragment 6, DNA-PKcs^{nt10459−12384}, containing the KD domain, bound to OCT4.

c. DNA-PKcs^{aa3719−3784} is required for binding to OCT4. (Top) Diagram of 7 subconstructs (A-G) of PRKDC Fragment 6 (PRKDC^{nt10459−12384}) from Fig 5b, expressed as the corresponding DNA-PKcs amino acids. HEK293FT cells were co-transfected PRKDC Fragment 6 subconstructs A-G along with pCMV6-POU5F1-mycDDK for 48 h. Protein lysate was IP with EZview™ Red anti-FLAG-M2 affinity gel and immunoblotting with anti-HA or anti-FLAG-M2. Analysis indicts that the DNA-PKcs sequence necessary for binding to OCT4 is within amino acids 3719 – 3784. There is a low band in Lane F of weaker signal strength than other binding subconstructs, therefore it cannot be excluded that amino acids below 3719 may partially contribute binding. Binding restriction around amino acids 3719 – 3785 is better evidenced.

d. Deletion of a randomly selected residues (aa 3735–3767) within OCT4-binding motif (OBM: 3719 to 3784) of DNA-PKcs^{aa3619−4043} disrupted OCT4 binding. HEK293FT cells were co-transfected pCMV6-POU5F1-mycDDK along with pCMV6-AN-HA (empty vector), pCMV6-AN-HA-PRKDC^{nt10855−12129}, and pCMV6-AN-HA-PRKDC^{nt10855−12129∆OBM}. Protein lysates were immunoblotted as above. OCT4-mycDDK acted as a positive control.

e. (top) Structures of the truncated forms, HA-PRKDC^{nt10855−12129} and HA-PRKDC^{nt10855−12129∆CS}, and multiple sequence alignments of the catalytic site of PRKDC from different species. The numbers from the NH₂- to COOH-terminus represent the amino acid sequence that corresponds to that of the catalytic site (CS: ³⁹¹⁹GIGDRHLHN³⁹²⁷) of human PRKDC. (bottom) Deletion of the catalytic site of DNA-PKcs prevented phosphorylation of OCT4 at Ser93. The pCMV6-POU5F1-mycDDK was co-transfected along with the pCMV6-AN-HA (empty vector), pCMV6-AN-HA-PRKDC^{nt10855−12129}, or pCMV6-AN-HA-PRKDC^{nt10855−12129∆CS} into 293FT cells. After 48 h, protein lysates (500 μg) were immunoprecipitated using EZview™ Red anti-HA affinity gel (40 μl), separated using a 4–12% Bis-Tris NuPAGE^® polyacrylamide gradient gel, and followed by immunoblotting with the indicated antibodies. OCT4-mycDDK detected by anti-FLAG M2 was used for protein expression control.

3.6. In vitro and in vivo activity of DNA-PKcs inhibitors and a protein-protein interaction inhibitor in combination with Bcl-2 inhibitors

Given the results above, commercially available DNA-PKcs inhibitors (M3814, NU7441, and AZD7648), and a compound that we previously identified as a protein-protein interaction inhibitor of OCT4 and DNA-PKcs, narciclasine [34], were tested for targeted inhibitory effects on pOCT4^S93 and c-MYC expression in four SCLC cell lines. Narciclasine, M3814, and NU7441 reduced the expression of c-MYC and pOCT4^S93 (Fig. 6a & Fig. S6a), and narciclasine exposure results in cleaved pro-apoptotic markers, including PARP and caspase 3 (Fig. S6b). In a panel of 15 SCLC cell lines, the in vitro IC₅₀ values at +96 h of M3814 and narciclasine ranged from 0.05 – 4.0 μM and 19.5 – 109 nM, respectively (Fig. 6b). Significant differences in cytotoxicity were not observed between high vs. low c-MYC groups (NU7441: p=0.84, CC-115: p=0.4, AZD7648: p=0.13, Fig. S7a), which may be due to relatively small number of cell lines per group or lack of specificity of the inhibitors. The IC₅₀ ranges were 0.06 – 9.5 μM for NU7441, 0.04 – 5.7 μM for CC-115, and 0.03 – 69 μM for AZD7648 (Fig. S7a); the IC₅₀ value of each cell line relative to the panel as a whole was compared (Fig S7b). Next, the in vitro effect of M3814 or narciclasine, in combination with Bcl-2 inhibitors ABT-737 or ABT-199 (venetoclax, an approved agent in high Bcl-2 expressing SCLC), on death effector proteins was assayed (Fig. S8a). Bcl-2 inhibitors was selected as a combining agent to assess the effects of DNA-PKcs inhibition without directly inducing DNA-damage response elements. In NCI-H847 (MYC-amplified) and NCI-H1048 (high c-MYC without genomic amplification), combination treatment more potently reduced protein levels of c-MYC and pOCT4^S93 and increased apoptotic proteins than treatment with single agents alone. In NCI-H847 (high c-MYC), NCI-H1048 (high c-MYC), and NCI-H69 (low c-MYC), the cytotoxicity of M3814 increased synergistically in combination with ABT-737, while narciclasine displayed synergy with ABT-737 and ABT-199 in NCI-H847 and NCI-H1048 cells, but not in NCI-H69 cells (Fig. S8b).

The effects of narciclasine and ABT-199 were evaluated in xenograft models of the NCI-H847m (MYC-amplified) and NCI-H1048 (MYC-nonamplified but high c-MYC) cell lines (Fig 6C). In the NCI-H847m model (Fig 6c top), ABT-199 (alone) significantly extended EFS relative to control (p<0.001). Narciclasine (alone) also significantly extended EFS relative to control (p=0.0027). Narciclasine + ABT-199 demonstrated the greatest antitumor activity (p<0.001). In the NCI-H1048 xenograft model (Fig 6c middle), ABT-199 (alone) extended EFS relative to both control and narciclasine (alone)(p<0.001). Narciclasine (alone) did not extend survival relative to controls, contrary to expectations from in vitro results. Narciclasine + ABT-199 significantly extended EFS relative to controls and narciclasine (alone)(p<0.001) and relative to ABT-199 (alone) (p <0.001). We also evaluated M3814 in combination with ABT-263, an orally available analog of ABT-737, in the NCI-H1048 model (Fig. 6c bottom). M3814 (alone) did not extend survival; ABT-263 (alone) increased EFS (p=0.01), while the drug combination produced the greatest increase in EFS compared to controls (p<0.001). Change of tumor volumes and body weights of the individual animals during these experiments are shown in Fig. S9. Overall, the drug combinations were well-tolerated.

4. Discussion

The current study demonstrates that OCT4 mediated MYC activation increases MYC transcriptional activation in preclinical models and clinical samples of SCLC. The expression of the proteins in the c-MYC/OCT4/DNA-PKcs axis did not differ depending on stages of the disease, likely due to the molecular heterogeneity resulting from complex genetic and epigenetic mechanisms. This necessitates mechanistic investigations on molecular changes. In the current project, we highlight the important role of DNA-PKcs in increasing c-MYC expression in SCLC independent of its canonical role in DNA-damage repair and identifies DNA-PKcs as a potential therapeutic target for reducing MYC overactivation in a clinically important subset of SCLC. Several recent studies reported the role of epigenetic alterations in DNA-PK expression that may impact cancer progression and metastasis [35–37]. The studies focus the investigation on the roles of DNA-PKcs in DNA repair mechanisms. The paucity of data on DNA-PK’s roles that are not directly related to DNA repair prompts warrants investigation on the impact of epigenetic changes in DNA-PKcs on the OCT4 or c-MYC. On the other hand, OCT4 is reported to play an important role in DNA methylation and histone modification. The effect of such changes may also be one of the mechanisms of c-MYC regulation[38].

c-MYC protein itself has long been considered to be undruggable due to the lack of readily accessible binding sites for inhibitory therapeutic compounds [39, 40]. One option for targeting this novel pathway would be direct inhibition of DNA-PKcs, which is the upstream of the MYC transcriptional activation pathway. However, while we did observe a degree of SCLC antitumor activity using currently available DNA-PKcs inhibitors, overall, our data suggests that targeting DNA-PKcs itself may not be the most effective approach to suppress c-MYC overactivation. There are also other issues in developing kinase inhibitors that directly bind the kinase, such as potential for developing resistance to the inhibitor, downstream target effect, etc. [41–43]. Therefore, as we had identified the protein regions necessary for the DNA-PKcs and OCT4 binding interaction (with resultant OCT4 Ser93 phosphorylation and subsequent increase of MYC expression), and while recognizing the limitations of targeting protein-protein interactions [44], we utilized the compounds that are identified as the inhibitor (narciclasine) of the binding of DNA-PKcs to OCT4 [34] for the preclinical activity in SCLC models.

In preclinical activity studies, narciclasine demonstrated in vitro activity at low nanomolar concentrations, but had in vivo activity as a single agent against high c-MYC expressing SCLC xenografts that was less than anticipated - although narciclasine did significantly extend SCLC xenograft survival when combined with a Bcl-2 inhibitor. Narciclasine has shown anticancer activity in vitro in cell lines of several different cancer types, but reported preclinical in vivo activity to date is limited to glioblastoma models. The discrepancy between the in vitro and in vivo activities of narciclasine observed in this study and by others warrants further investigation by pharmacokinetic, pharmacodynamic, and structure-activity relations approaches.

In summary, we have identified a novel pathway of OCT4-induced MYC transcriptional activation that is dependent on DNA-PKcs phosphorylation of Ser93 of OCT4 in the absence of DNA damage pathway activation and have demonstrated that this pathway can be disruptively targeted by inhibiting binding of DNA-PKcs to OCT4. We identified one such inhibitor, narciclasine, which was well-tolerated in preclinical SCLC xenograft models and which prolonged survival in combination with clinical stage Bcl-2 inhibitors. Such minimally-toxic therapies might find rapid clinical translation as interim or maintenance therapies for high MYC-expressing SCLC cancers. We propose future studies to include identification and optimization of compounds that inhibit the novel pathway.