HDAC1 Upregulation by NANOG Promotes Multidrug Resistance and a Stem-like Phenotype in Immune Edited Tumor Cells

The phenotypic and functional heterogeneity among cancer cells within tumors is well documented (1). These features of cancer cells have the potential to limit the effectiveness of radio- and chemotherapy as well as immunotherapy. For example, conventional therapies may eliminate the bulk of the tumor but spare highly aggressive cancer cells that have a remarkable capacity to survive, self-renew, and advance the malignancy (2, 3). These residual tumor cells have been found to possess key stem-like properties and increased tumor-initiating capacities (4). We recently demonstrated that immune selection drives the evolution of tumor cells toward an immune-resistant and stem-like phenotype (5, 6), which is consistent with what has been reported for other types of conventional cancer treatment, such as chemotherapy or radiotherapy (7–9). In the process, transcription factor NANOG links the emergence of a stem-like state in the tumor and immune escape (5). Although it is clear that NANOG acts as a transelement to activate gene expression, recent data have demonstrated the role of NANOG in gene repression to regulate embryonic development (10). Many reports provide clues about the importance of epigenetic reprogramming in NANOG-mediated gene silencing (11–13). However, the underlying mechanisms of treatment resistance in cancer remain largely unknown.

Substantial efforts to elucidate the molecular basis of these stem-like properties and the associated treatment resistance revealed that many of these molecular mechanisms have been linked to an epigenetic alteration of tumor cells (14). Of the various epigenetic modifications, histone acetylation is an important determinant of gene expression and is generally associated with elevated transcription, whereas histone deacetylation is often associated with gene repression (15). Histone deacetylases (HDAC) enzymatically remove the acetyl group from histones and play an important role in regulating cell proliferation and differentiation (16). Moreover, these HDACs, especially HDAC1, were further increased in relapsed tumor cells after treatments, while inhibition of HDACs enhanced the antitumor effect of the treatment (17, 18). Despite the crucial roles played by HDAC1 in tumorigenesis as well as the development of resistance against cancer therapy, molecular mechanisms in the regulation of HDAC1 expression have not yet been extensively studied.

In this study, we demonstrated a crucial role of HDAC1 at the crossroads between NANOG and epigenetic states in immunoedited tumor cells by identifying HDAC1 as a novel NANOG transcriptional target. Therefore, we have provided the proof of the principle in a preclinical model that HDAC1 inhibition is an effective strategy to control human cancer, particularly in the context of immune-based therapy.

Six- to 8-week-old female NOD/SCID mice were purchased from Central Lab. Animal Inc. All mice were maintained and handled under the protocol approved by the Korea University Institutional Animal Care and Use Committee (KUIACUC-2014-175). All animal procedures were performed in accordance with recommendations for the proper use and care of laboratory animals.

CaSki, MDA-MB231, and HEK293 cell lines were purchased from ATCC. All cell lines were obtained between 2010 and 2014 and tested for mycoplasma using Mycoplasma Detection Kit (Thermo Fisher Scientific). The identities of cell lines were confirmed by short tandem repeat profiling by IDEXX Laboratories, Inc. and used within 6 months for testing. Generation of the immunoresistant CaSki P3 cell line has been described previously (19). For generation of CaSki-NANOG cells, pMSCV-NANOG plasmids were first transfected along with viral packaging plasmids (VSVG and Gag-pol) into HEK293FT cells. Three days after transfection, the viral supernatant was filtered through a 0.45-μm filter and infected into CaSki cells. Infected cells were then selected with 1 μg/mL puromycin. For the generation of the MDA-MB231 P3 tumor line, NOD/SCID mice were inoculated subcutaneously with 1 × 10⁶ MDA-MB231 P3 cells per mouse. Seven days following tumor challenge, mice received adoptive transfer with 2 × 10⁶ MART-1–specific CTLs and 3,000 U of IL2 (Novartis). This treatment regimen was repeated for three cycles. All cells were grown at 37°C in a 5% CO₂ incubator/humidified chamber.

The following chemical reagents were used in this study: FK228 (Selleckchem), sodium butyrate (NaB, Selleckchem), 5-azacytidine (5-AzaC, Sigma), cisplatin (Selleckchem), and 5-fluorouracil (5-FU, Selleckchem).

The pMSCV-NANOG plasmids have been described previously (5). The promoter region of the HDAC1 gene was isolated by PCR from genomic DNA extracted from CaSki cells using a primer set, 5′-AGCTCGAGGAGCAATGTTTGGCACACA-3′ (forward) and 5′-AGAAGCTTTCTGCGCCATCTTGCTCG-3′ (reverse). The PCR products were digested with XhoI and HindIII and subcloned into the XhoI/HindIII restriction sites of the pGL3-Basic vector (Promega).

Site-directed mutagenesis was performed using a QuickChange XL Site-Directed Mutagenesis Kit (Stratagene) according to the manufacturer's instructions. To create mutations in the NANOG-binding site of HDAC1, the following primers were used: 5′-ATCTGGTGGAGTGGTTCTTCCTGGTAGAGTTGGGGGCAAT-3′ (forward) and 5′-ATTGCCCCCAACTCTACCAGGAAGAACCACTCCACCAGAT-3′ (reverse). Mutations were verified by DNA sequencing.

Total RNA was isolated using RNeasy Micro Kit (Qiagen), and the cDNAs were synthesized by reverse transcriptase (RT) using iScript cDNA Synthesis Kit (Bio-Rad), according to the manufacturer's recommended protocol. qRT-PCR was performed using iQ SYBR Green super mix (Bio-Rad) with the specific primers on a CFX96 real-time PCR detection system. All qRT-PCR experiments were performed in triplicate and quantification cycle (Cq) values were determined using Bio-Rad CFX96 Manager 3.0 software. Relative quantifications of the mRNA levels were performed using the comparative C_t method with β-actin as the reference gene. Fold change was calculated relative to the expression level of mRNA in control cells. Predesigned QPCR primers for INK4D (#P318368), CLOCK (#P243182), NCOA2 (#P111758), NCOA3 (#P240721), EP300 (#P202425), and CREBP (#P173646) were purchased from Bioneer. The sequences of primers used for real-time PCR experiments are shown in Supplementary Table S1.

Synthetic siRNAs specific for GFP, NANOG, HDAC1, HDAC2, and HDAC3 were purchased from Bioneer; nonspecific GFP, 5′-GCAUCAAGGUGAACUUCAA-3′ (sense), 5′-UUGAAGUUCACCUUGAUGC-3′ (antisense); NANOG, 5′-GCAACCAGACCUGGAACAA-3′ (sense), 5′-UUGUUCCAGGUCUGGUUGC-3′ (antisense); HDAC1, 5′- GAGUCAAAACAGAGGAUGA-3′ (sense), 5′-UCAUCCUCUGUUUUGACUC-3′ (antisense); HDAC2, 5′- GACGGAAACUGAGCUCAGU-3′ (sense), 5′-ACUGAGCUCAGUUUCCGUC-3′ (antisense); HDAC3, 5′-GAGCUUCAAUAUCCCUCUA-3′ (sense), 5′-UAGAGGGAUAUUGAAGCUC-3′ (antisense). CDKN2D, 5′-CACCUAAACGGUUCAGUUU-3′ (sense), 5′-AAACUGAACCGUUUAGGUG-3′ (antisense); CDKN1B, 5′-CGACGAUUCUUCUACUCAA-3′ (sense), 5′-UUGAGUAGAAGAAUCGUCG-3′ (antisense); TRIM17, 5′-CAGAGUUCCCGGACAGAUU-3′ (sense), 5′-AAUCUGUCCGGGAACUCUG-3′ (antisense); CUL1, 5′-GACGAAGGACGAAAAGGAA-3′ (sense), 5′-UUCCUUUUCGUCCUUCGUC-3′ (antisense); RBX1, 5′-GAAGCGCUUUGAAGUGAAA-3′ (sense), 5′-UUUCACUUCAAAGCGCUUC-3′ (antisense); NOXA, 5′-GUUAUACUCAGUGUUGAUU-3′ (sense), 5′-AAUCAACACUGAGUAUAAC-3′ (antisense); CLOCK, 5′-CAGACUUUACAGAGUACAU-3′ (sense), 5′-AUGUACUCUGUAAAGUCUG-3′ (antisense); NCOA2, 5′-CUCAUCCGUUCUCAGACUA-3′ (sense), 5′-UAGUCUGAGAACGGAUGAG-3′ (antisense); NCOA3, 5′-GAGACUUGGAUAAUCUAGA-3′ (sense), 5′-UCUAGAUUAUCCAAGUCUC-3′ (antisense); EP300, 5′-GAUGAAUGCGGGCAUGAAU-3′ (sense), 5′-AUUCAUGCCCGCAUUCAUC-3′ (antisense); CBP, 5′-CAGUGAAUCGCAUGCAAGU-3′ (sense), 5′-ACUUGCAUGCGAUUCACUG-3′ (antisense). Cells were transfected with 100 pmol of synthesized siRNAs using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions.

For in vitro CTL assays, tumor cells were harvested by trypsinization, washed once with DMEM containing 0.1% FBS, resuspended, and labeled in 1 mL 0.1% FBS containing DMEM and 10 μmol/L carboxyfluorescein diacetate succinimidyl ester (CFSE; Molecular Probes). The suspended cells were incubated for 10 minutes in a 37°C incubator with 5% CO₂. After collection, the CFSE-labeled cells were resuspended in 10 μg/mL MART-1 peptide containing 1 mL DMEM. After 1 hour, CFSE-labeled cells were incubated for 4 hours with a MART-1–specific CD8⁺ T-cell line at an E/T ratio of 1:1. After incubation for 4 hours at 37°C, the frequency of apoptotic cells was determined by staining with anti-active caspase-3 antibody and performing flow cytometry as described previously (20).

Cells were plated at 1 × 10³ cells/well in 6-well, super-low adherence vessels (Corning) containing serum-free DMEM-F12 (Thermo Fisher Scientific) supplemented with EGF (20 ng/mL), basic FGF (20 ng/mL), and 1× B27. Medium was replaced every 3 days to replenish nutrients. Colonies more than 50 μm in diameter were counted under a microscope.

Cells were harvested by trypsin treatment and then washed and resuspended in Opti-MEM. NOD/SCID mice were subcutaneously injected with 10⁴ or 10⁵ cells. Tumor formation was monitored every 2 days. After 12 days, tumor tissue was excised and weighed.

HDAC activity was measured in whole-cell lysates with HDAC activity Colorimetric Assay Kits (BioVision), according to the manufacturer's recommended protocol. Briefly, 100 μg of the sample protein was incubated with a colorimetric acetylated lysine–containing substrate in the absence or presence of TSA for 3 hours at 37°C. Lysine developer was then added, followed by incubation for an additional 30 minutes. HDAC activity was measured with an uQuant microplate reader (BioTek) at 405 nm wavelength. Then, 50 μg of HeLa nuclear extract without and with TSA treatment was used as positive control and negative control, respectively.

Total histone proteins were isolated by acid extraction methods (21). Briefly, cells were harvested by trypsin treatment and then resuspended in 1 mL of hypotonic lysis buffer (10 mmol/L Tris-Cl pH8.0, 1 mmol/L KCl, 1.5 mmol/L MgCl₂, and 1 mmol/L DTT). After incubation for 30 minutes on a rotator at 4°C, intact nuclei were isolated by centrifugation (10,000 × g, 10 minutes, 4°C) and resuspended in 400 μL of 0.4 N H₂SO₄ and they were incubated overnight. Nuclear debris was removed by centrifugation (16,000 × g, 10 minutes, 4°C), and the supernatant containing histones was isolated by TCA precipitation.

Tissue microarrays (TMA) containing four 1.0-mm cores from 479 formalin-fixed, paraffin-embedded cervical neoplasia and matched nonadjacent normal cervical epithelial tissue specimens were described previously. Tissue specimens were prospectively collected from patients who were enrolled in Gangnam Severance Hospital (Seoul, South Korea) and the Korea Gynecologic Cancer Bank between 1996 and 2010. Some of the paraffin blocks were obtained from the Korea Gynecologic Cancer Bank through Bio & Medical Technology Development Program of the Ministry of Education, Science and Technology, Korea (NRF-2012M3A9B8021800).

All procedures were conducted in accordance with the Declaration of Helsinki. The tissue specimens and medical records were obtained with informed content of all patients and approval of the Institutional Review Board of Gangnam Severance Hospital (approval no. #3-2010-0030; Seoul, South Korea). This study was additionally approved by the Office of Human Subjects Research at the NIH. Cervical cancer was staged according to the International Federation of Gynecology and Obstetrics staging system and graded according to the World Health Organization grading system. TMAs were produced from formalin-fixed, paraffin-embedded tissues, and representative areas were meticulously selected from hematoxylin and eosin–stained slides. Tissue cylinders of 1.0-mm diameter were extracted from selected areas of donor blocks and transplanted into recipient blocks using a tissue arrayer (Beecher Instruments, Inc.).

For control and test RNAs, rRNA was removed using Ribo-Zero Magnetic Kit (Epicentre, Inc.) from each 5 μg of total RNA. The construction of library was performed using SENSE Total RNA-Seq Library Prep Kit (Lexogen, Inc., Vienna, Austria) according to the manufacturer's instructions. Library production is initiated by the random hybridization of starter/stopper heterodimers to the remaining RNA. These starter/stopper heterodimers contain Illumina-compatible linker sequences. A single-tube reverse transcription and ligation reaction extends the starter to the next hybridized heterodimer, where the newly synthesized cDNA insert is ligated to the stopper. Second-strand synthesis is performed to release the library from the beads, and the library is then amplified. Barcodes were introduced when the library is amplified. High-throughput sequencing was performed as paired-end 100 sequencing using HiSeq 2000 (Illumina, Inc.). RNA sequencing (RNA-seq) reads were mapped using TopHat software tool to obtain the alignment file. The alignment file was used for assembling transcripts, estimating their abundance and detecting differential expression of genes or isoforms using cufflinks. The accession number for the sequencing data reported in this article is GEO: GSE88965.

NOD/SCID mice were inoculated subcutaneously with 2 × 10⁶ MDA-MB231 P3 cells per mouse. Seven days following tumor challenge, FK228 (0.05 mg/kg) or PBS was administered via the intraperitoneal route. The following day after FK228 treatment, mice received adoptive transfer with 2 × 10⁶ MART-1–specific CTLs. This treatment regimen was repeated for three cycles. Mice were monitored for tumor burden and survival for 23 and 38 days after challenge, respectively.

All data are representative of at least three separate experiments. Individual data points were compared by two-tailed Student t test. For IHC data, statistical analysis was performed using R software version 3.1.2. The Mann–Whitney U test was used to compare the protein levels between each group. The χ² test was used to assess associations between molecular markers. Survival distributions were estimated using the Kaplan–Meier method with the log-rank test. A Cox proportional hazards model was created to identify independent predictors of survival. In all cases, P < 0.05 was considered statistically significant.

Previously, we established a highly immunoresistant cervical tumor cell line, CaSki P3, which was generated from its immune susceptible parental cell line CaSki P0 through three rounds of in vitro selection by mixing CaSki P0 cells pulsed with the MART-1 peptide together with MART-1–specific CTLs (19). Consistent with our previous results from other immunoresistant models of TC-1 P3 and CaSki/D^b P3, the immune-edited CaSki P3 cells showed a stem-like (Supplementary Figs. S1A–S1C) and immune-resistant phenotype (19). In addition, immune selection has also driven tumor cells to become refractory against chemotherapeutic agents, such as cisplatin and 5-FU, indicating multimodality resistance of the immunoedited CaSki P3 cells (Fig. 1A). Epigenetic therapies have increasingly been recognized as new approaches to relapsed cancer after therapy (14). To investigate the potential therapeutic application of epigenetic therapy for the immunoresistant cancer cells, we compared sensitivity of CaSki P0 and P3 cells against three epigenetic targeted agents, 5-AzaC, sodium butyrate (NaB), and romidepsin (FK228). Interestingly, CaSki P3 cells were more sensitive to two different HDAC inhibitors (NaB and FK228) than CaSki P0 cells, and these cells were especially sensitive to class I HDAC inhibitor FK228, whereas 5-AzaC resulted in no difference in cell viability between P0 and P3 cells (Fig. 1A). On the basis of the above observation, we reasoned that difference in susceptibility of CaSki P0 and P3 cells to HDAC inhibitors may be due to their dissimilarity in HDAC activity or HDAC protein level. Indeed, CaSki P3 cells had more total HDAC activity and increased protein levels of class I HDACs (HDAC 1, 2, and 3), except for HDAC8 and HDAC4 as compared with CaSki P0 cells (Figs. 1B and C). Moreover, we did not observe increased HDAC activity or HDAC upregulation in the negative control CaSki N3 cells, which were generated through serial selection by mixing CaSki P0 cells pulsed with MART-1 peptide together with irrelevant NY-ESO1–specific T cells (Fig. 1B and C). These results suggest that HDAC activity and expression increased by immune selection with antigen-specific T cells may confer susceptibility to HDAC inhibitors in tumor cells. We then tested whether these increased HDACs are related to various aggressive phenotypes of CaSki P3. CaSki-P3 cells transfected with siRNAs targeting HDAC1, 2, or 3 were more susceptible to apoptosis induced by antigen-specific CTLs or granzyme B, a key component in CTL-mediated apoptosis (Supplementary Fig. S1D; Fig. 1D and E). Moreover, delivery of siHDAC1, 2, or 3 increased the susceptibility to cisplatin and 5-FU (Fig. 1F), and it also reduced the sphere-forming capacity when cells were cultured under suspension conditions, compared with delivery of siGFP (Fig. 1G). However, CaSki N3 cells transfected with siHDAC1, 2, and 3 did not significantly alter the susceptibility to CTL, granzyme B, cisplatin, and 5-FU as well as sphere-forming capacity (Fig. 1D–G). Taken together, our data indicate that HDACs, HDAC1, 2, and 3 are increased during immunoediting, and they contribute to immunoresistance, chemoresistance, and stem-like phenotype of immunoedited tumor cells. Among HDACs, HDAC1 has dominant effects on the aggressive phenotypes.

We previously identified that NANOG is one of the key molecules that drive immunoresistant and stem-like phenotypes of cancer cells (5). It has also been reported that NANOG confers resistance to chemotherapeutic agents (22, 23). Consistently, the CaSki P3 cells showed increased NANOG expression compared with CaSki P0 cells (Supplementary Fig. S1E). We questioned whether NANOG is required for resistance to chemotherapeutic agents and susceptibility to HDAC inhibitors in CaSki P3 cells. As expected, silencing of NANOG increased the sensitivity of CaSki P3 cells to cisplatin (Fig. 2A). Notably, silencing of NANOG did in fact decrease the sensitivity of CaSki P3 cells to FK228 (Fig. 2A). To determine the relationship between NANOG and HDACs, we transfected siRNA targeting NANOG into CaSki P3 cells and then analyzed the protein levels of HDACs. When we silenced NANOG, HDAC1 protein was significantly decreased, while the levels of HDAC2 and HDAC3 protein were unaffected (Fig. 2B). Conversely, delivery of NANOG to CaSki P0 cells reduced the sensitivity to cisplatin, but it increased the sensitivity to FK228 (Fig. 2C). When NANOG was overexpressed in CaSki P0 cells, only HDAC1 protein was increased (Fig. 2D). These results suggest that NANOG is a key mediator that determines susceptibility to FK228 as well as resistance to cisplatin through regulation of HDAC1 expression. We then asked whether HDAC1 potentiates NANOG-mediated phenotypes. Despite NANOG overexpression, HDAC1 knockdown cells were more susceptible to apoptosis induced by granzyme B (Fig. 2E) or to cisplatin (Fig. 2F). Moreover, siHDAC1-treated CaSki-NANOG cells had reduced in vitro sphere-forming capacity and in vivo tumorigenicity (Fig. 2G–I). In contrast with CaSki-NANOG cells, CaSki-no insert cells transfected with siHDAC1 did not significantly alter the susceptibility to granzyme B and cisplatin, and sphere-forming capacity (Fig. 2E–G). Thus, our findings demonstrate that HDAC1 plays a crucial role in various NANOG-mediated aggressive phenotypes.

As NANOG regulates multiple gene expression programs that are critical for self-renewal of stem cells, we asked whether NANOG regulates HDAC1 expression through its transcriptional function. To address this, we used a mutant form of NANOG (NANOG Mut) that was previously found to have weak transcriptional activity (5). When we transfected HEK293 and CaSki cells with wild-type NANOG (NANOG Wt), HDAC1 protein and mRNA expression levels were profoundly increased, while transfection of NANOG Mut had no significant impact on HDAC1 expression (Fig. 3A and B), indicating that NANOG-mediated HDAC1 regulation is dependent on the transcriptional activity of NANOG. We were further encouraged by the presence of NANOG-binding elements in the HDAC1 promoter region, suggesting the possibility that NANOG is a direct transcriptional activator of HDAC1 (Fig. 3C). Luciferase assays showed a significant increase in HDAC1 promoter activity upon cotransfection with NANOG Wt but not upon cotransfection with NANOG Mut (Fig. 3D). Moreover, mutation of the NANOG-binding site in the HDAC1 promoter region eliminated the HDAC1 promoter activation by NANOG Wt (Fig. 3D). Chromatin immunoprecipitation (ChIP) assays confirmed direct binding of NANOG to the HDAC1-regulatory region (Fig. 3E). This was also validated in the CaSki P0 and P3 cells, where we noted more NANOG occupancy in P3 cells, relative to P0 cells (Fig. 3F). These results demonstrate that NANOG directly regulates HDAC1 transcription by binding to the promoter region of the HDAC1 gene.

HDAC1 contributes to gene regulation by removing acetyl groups from lysine residues in histone proteins (16); hence, we asked whether HDAC1 expression mediated by NANOG can lead to a decrease in histone acetylation. To assess this, we measured the levels of histone acetylation in CaSki-no insert-siGFP, CaSki-NANOG-siGFP, and CaSki-NANOG-siHDAC1 cells. Although addition of NANOG failed to markedly decrease global acetylation of histone H3 and H4, interestingly, acetylation of specific lysine residues 14 and 27 in histone H3 was significantly decreased, and these decreased levels of AcH3K14 and AcH3K27 induced by NANOG were reversed upon HDAC1 silencing (Fig. 3G and H; Supplementary Fig. S2A and S2B). Altogether, these findings demonstrate that NANOG causes a decrease in AcH3K14 and AcH3K27 through transcriptional activation of HDAC1, suggesting a link between NANOG and epigenetic change.

Having explored the molecular mechanism by which the NANOG–HDAC1 axis confers tumor-aggressive phenotypes, we aimed to determine the clinical relevance of protein levels in human cancer. We compared NANOG, HDAC1, AcH3K14, and AcH3K27 levels by IHC in cervical tissue from patients with cervical intraepithelial neoplasia (CIN) or invasive cervical carcinoma. HDAC1 level was increased during tumor progression from low-grade CIN (LGCIN) to high-grade CIN (HGCIN), while AcH3K14 level was decreased during tumor progression (Fig. 4A and B). Furthermore, AcH3K27 level was lower in cervical cancer tissues than in LGCIN or normal tissue (Figs. 4A and B). The correlation between the levels of NANOG, HDAC1, AcH3K14, and AcH3K27 was assessed in CIN and cervical cancer specimens. HDAC1 was positively correlated with NANOG, whereas it was negatively correlated with both AcH3K14 and AcH3K27. AcH3K14 and AcH3K27 were positively correlated with each other, and they were negatively correlated with NANOG (Fig. 4C). We next examined the relationship of each protein level or acetylation level of histone H3 with patient survival outcomes. Kaplan–Meier plots demonstrated that patients with high levels of HDAC1 and low levels of AcH3K14 and AcH3K27 displayed worse 10-year overall survival (Fig. 4D). Furthermore, patients with combined NANOG⁺/HDAC1⁺, NANOG⁺/AcH3K14⁻, NANOG⁺/AcH3K27⁻, HDAC1⁺/AcH3K14⁻, and HDAC1⁺/AcH3K27⁻ showed significantly worse overall survival and disease-free survival than patients with NANOG⁻/HDAC1⁻, NANOG⁻/AcH3K14⁺, NANOG⁻/AcH3K27⁺, HDAC1⁻/AcH3K14⁺, and HDAC1⁻/AcH3K27⁺ (Fig. 4D). Combination of all three parameters showed a similar result: the patients with combined NANOG⁺/HDAC1⁺/AcH3K14⁻ and NANOG⁺/HDAC1⁺/AcH3K27⁻ showed significantly worse overall survival than patients with NANOG⁻/HDAC1⁻/AcH3K14⁺ and NANOG⁻/HDAC1⁻/AcH3K27⁺ (Fig. 4D). Thus, our data clearly indicate that the NANOG/HDAC1/AcH3K14/AcH3K27 axis serves as an important prognostic factor in human cervical neoplasia.

To gain insight into the role of HDAC1 in NANOG-mediated transcriptional regulation, we performed genome-wide total RNA-seq analysis in CaSki-no insert-siGFP, CaSki-NANOG-siGFP, and CaSki-NANOG-siHDAC1 cells. Firstly, to explore the downstream target genes of NANOG, we compared gene expression in CaSki-NANOG-siGFP cells relative to that in CaSki-no insert-siGFP cells, and then, we identified differentially expressed genes (DEG) that included 2,166 genes upregulated by NANOG and 2,151 genes downregulated by NANOG (Fig. 5A). Consistent with our experimental results, this analysis revealed that only HDAC1 was upregulated by NANOG, suggesting a good internal validation of the RNA-seq results (Supplementary Fig. S3A). Next, we compared the DEGs in CaSki-NANOG-siGFP cells and CaSki-NANOG-siHDAC1 cells to explore the HDAC1-dependent NANOG target genes. Among the 2,166 upregulated DEGs, 863 genes were reversed upon HDAC1 knockdown, suggesting HDAC1 dependence (Group 1), while 1,303 genes were not affected by HDAC1 knockdown (Group 2; Fig. 5A). Among the 2,151 downregulated DEGs, expression of 1,213 genes was reversed upon HDAC1 knockdown, suggesting HDAC1 dependence (Group 3), and 938 genes were not affected by HDAC1 knockdown (Group 4; Fig. 5A). The cellular process affected by the NANOG–HDAC1 axis was then examined by performing functional enrichment analysis of the DEGs using DAVID software (24, 25) with gene sets of Group 1 to 4. This analysis showed that genes upregulated by NANOG were mainly related to the homeostatic process, wound response, defense response, and cell proliferation, whereas genes downregulated by NANOG were involved in transcriptional regulation, apoptosis, metabolism, and development (Fig. 5B). Interestingly, genes downregulated through HDAC1 dependence (Group 3) were significantly enriched for transcriptional regulation and apoptosis (Fig. 5B). Taken together, our data not only indicate that a respectable amount of NANOG-responsive genes were affected by HDAC1, but they also indicate that the function of NANOG target genes was split according to HDAC1 dependence, suggesting the crucial role played by HDAC1 in NANOG-mediated transcriptional regulation.

We next aimed to elucidate the mechanism by which HDAC1 promotes NANOG-mediated stem-like property. It was demonstrated that cyclin-dependent kinase (CDK) inhibitor genes are important for HDAC1-mediated tumorigenesis in cancer as well as proliferation of stem cells (26–28). Among the 10 CDK inhibitor genes regulated by NANOG from the RNA-seq data, only CDKN2D (encoding INK4D) and CDKN1B (encoding p27) were downregulated in CaSki-NANOG-siGFP cells, and they were reversed in CaSki-NANOG-siHDAC1 cells, indicating HDAC1-dependent downregulation by NANOG (Fig. 5C). We validated NANOG-HDAC1 dependency of these genes by qRT-PCR (Fig. 5D). Because NANOG caused a decrease in AcH3K14 and AcH3K27 in an HDAC1-dependent manner, we reasoned that decreased expression of CDKN2D and CDKN1B may be due to HDAC1-mediated epigenetic silencing. ChIP-qPCR showed that NANOG expression caused loss of AcH3K14 and AcH3K27 occupancy on the promoter region of CDKN2D and CDKN1B, and these histone modification events were reversed by HDAC1 knockdown (Fig. 5E). Consistent with these observations, HDAC1 was more enriched in the CDKN2D and CDKN1B promoters in CaSki-NANOG-siGFP cells compared with the promoters in CaSki-no insert-siGFP cells (Fig. 5E). These results suggest that NANOG downregulates the expression of CDKN2D and CDKN1B genes through HDAC1-mediated epigenetic repression. To assess the potential role of HDAC1-induced CDKN2D and CDKN1B repression in NANOG-mediated tumorigenesis, siRNA-mediated knockdown of CDKN2D or CDKN1B was performed in CaSki-NANOG-siGFP or CaSki-NANOG-siHDAC1 cells. In CaSki-NANOG cells, diminished tumor sphere-forming capacity induced by siHDAC1 treatment was partially reversed upon knockdown of either CDKN2D or CDKN1B, and it was completely reversed upon knockdown of both CDKN2D and CDKN1B (Fig. 5F). These results suggest that loss of CDKN2D and CDKN1B caused by HDAC1 contributes to the stem-like property of NANOG cells.

Previously, it was demonstrated that antiapoptotic protein MCL-1 is a key element in NANOG-mediated immunoresistance (5). In addition, the role of MCL-1 also applied to NANOG-mediated chemoresistance, as evidenced by knockdown of MCL-1 causing the increase of susceptibility to cisplatin and 5-FU, compared with control (Fig. 6A). Interestingly, knockdown of HDAC1 in CaSki-NANOG cells diminished the levels of MCL-1 protein but not the levels of MCL-1 transcripts (Fig. 6B and C), thus establishing a link between HDAC1 and MCL-1. Our previous report suggested that NANOG-mediated MCL-1 upregulation is closely associated with an activation of AKT signaling pathway (5). However, it seems that downregulation of MCL-1 protein by NANOG–HDAC1 axis is AKT pathway independent, as evidenced by failure to cause a change in AKT phosphorylation upon HDAC1 knockdown (Fig. 6B).

We next aimed to elucidate the molecular mechanism(s) by which HDAC1 regulates NANOG-dependent MCL-1 upregulation. It has been reported that the MCL-1 protein level is strictly controlled by proteasomal degradation through its ubiquitination by various E3 ubiquitin ligases and their cofactors (29). We reasoned that the NANOG-induced accumulation of MCL-1 protein may be due to repressive effects of NANOG on the expression of genes involved in MCL-1 degradation. Therefore, we compared the expression of genes involved in negative regulation of MCL-1 protein among NANOG target genes from the RNA-seq data. Four genes, including TRIM17, CUL1, RBX1, and NOXA genes (30–32), were downregulated in CaSki-NANOG-siGFP cells, and they were reversed in CaSki-NANOG-siHDAC1 cells, indicating HDAC1-dependent downregulation by NANOG (Supplementary Fig. S3B). We validated NANOG–HDAC1 dependency of these genes by qRT-PCR (Fig. 6D). Among them, only TRIM17 and NOXA were found to act as negative regulators of MCL-1 downstream of the NANOG–HDAC1 axis, as evidenced by knockdown of TRIM17 or NOXA causing reversal of the decrease in MCL-1 upon treatment with siHDAC1 in CaSki-NANOG cells (Fig. 6E). Furthermore, ChIP-qPCR showed that NANOG expression caused loss of AcH3K14 and AcH3K27 occupancy in the promoter region of TRIM17 and NOXA genes, and these histone modification events were reversed by HDAC1 knockdown (Fig. 6F). Consistent with these observations, HDAC1 was more enriched in the TRIM17 and NOXA promoters in CaSki-NANOG-siGFP cells compared with the promoters in CaSki-no insert-siGFP cells (Fig. 6F). This suggests that NANOG downregulates the expression of TRIM17 and NOXA genes through HDAC1-mediated epigenetic repression, resulting in an increase in MCL-1 protein. To assess the potential role of HDAC1-mediated repression of TRIM17 and NOXA in NANOG-mediated immunoresistance and chemoresistance, siRNA-mediated knockdown of TRIM17 or NOXA was performed in CaSki-NANOG-siGFP or CaSki-NANOG-siHDAC1 cells. In CaSki-NANOG cells, increased susceptibility to granzyme B, cisplatin, and 5-FU after siHDAC1 treatment was reversed upon TRIM17 or NOXA knockdown (Fig. 6G and H). These results indicate that loss of TRIM17 and NOXA caused by HDAC1 induces accumulation of MCL-1 protein and thus contributes to NANOG-mediated immunoresistance and chemoresistance.

We next examined whether the NANOG–HDAC1 axis is conserved across multiple cancer types. First, we profiled levels of NANOG and HDAC1 protein in a variety of human cancer cells. In all of these cells, we observed a positive correlation between NANOG and HDAC1 protein levels (Supplementary Fig. S4A and S4B). To verify the phenotypic effects of NANOG and HDAC1 in different types of cancer, we selected three representative lines, 526mel, H1299 and HCT116. Knockdown of NANOG robustly dampened the levels of HDAC1 and MCL-1 protein, whereas it increased the levels of INK4D and p27 protein and levels of AcH3K14 and AcH3K27 across all tested lines (Supplementary Fig. S4C). Notably, knockdown of HDAC1 resulted in identical effects on the level of these molecules compared with treatment with siNANOG, although it did not affect NANOG protein level (Supplementary Fig. S4C). Furthermore, both siNANOG- and siHDAC1-treated tumor cells were more susceptible to anticancer drugs (cisplatin and 5-FU) and to immune-mediated apoptosis, and they also had diminished sphere-forming capacity, compared with siGFP-treated tumor cells (Supplementary Fig. S4D–S4G). These results demonstrate that the biochemical and functional properties of the NANOG–HDAC1 axis are conserved across multiple types of cancer cells and that HDAC1 is a key molecule downstream of NANOG, capable of phenocopying NANOG.

Previously, we showed that NANOG inhibition in vivo restored the success of adoptive immune therapy, suggesting that blockade of the NANOG pathway could be a promising approach for immune-based cancer therapy (5). However, pharmacologic inhibitors of NANOG are yet to be developed. Because HDAC1 expression appears to be a common feature of human cancer that confers immune resistance, and HDAC1 inhibition phenocopies NANOG knockdown, we reasoned that inhibition of HDAC1 may serve as an interventional strategy for reversing the resistance of immune-edited tumor cells to CTL killing for a clinical application. To evaluate this idea, we used human MART-1–specific T cells (clone KKM; ref. 33). As CaSki cells rarely express MART-1 antigen, we chose the MDA-MB231 cells, which highly express the MART-1 antigen and established immunoresistant MDA-MB231 P3 cells from parental MDA-MB231 cells (hereafter MDA-MB231 P0) by selection with MART-1–specific CTL in vivo (Supplementary Fig. S5A). The phenotypic and molecular properties of MDA-MB231 P3 cells were consistent with those of CaSki P3 cells (Supplementary Fig. S5B–5E) and MDA-MB231 P3 were more susceptible to FK228 than P0 cells (Supplementary Fig. S5F). Moreover, the biochemical and functional properties of the NANOG–HDAC1 axis were conserved in the MDA-MB231 P3 cells (Supplementary Fig. S4C–S4G).

To demonstrate the therapeutic value of inhibiting HDAC1 and its downstream molecular axis, the efficacy of antigen-specific CTLs recognizing MART-1 was tested in NOD/SCID mice bearing MDA-MB231 P3 tumors. Mice received MART-1–specific CTL together with FK228, according to the schedule described in Fig. 7A. Tumors excised on day 23 were substantially smaller in terms of size and weight among mice receiving both MART-1–specific CTL and FK228 compared with mice receiving either treatment alone (Fig. 7B and C). Importantly, 100% of mice that received both adoptive transfer and FK228 survived, even at 38 days after tumor challenge; in contrast, all animals in the other groups had died by then (Fig. 7D). Western blot analysis of ex vivo isolated tumors at day 23 after challenge demonstrated a decreased MCL-1 level and a concurrent increase in INK4D and p27 proteins and AcH3K14 and AcH3K27 levels among mice receiving FK228 treatment (Fig. 7E), demonstrating achievement of successful in vivo delivery of FK228 to the tumor and modulation of the NANOG–HDAC1 molecular axis. In addition, the FK228-treated tumors contained fewer proliferating cells than the PBS-treated tumors, as measured by Ki67 staining, and this was unaffected by adoptive transfer of CTL (Fig. 7F). We assessed antigen-specific CTL infiltration into the tumor by labeling the T cells with CFSE prior to adoptive transfer and then counting the frequency of CFSE⁺ cells inside the tumor following transfer. We observed a slight, but not statistically significant, decrease in the frequency of antigen-specific CTLs in the tumors of FK228-treated mice compared with tumors of PBS-treated mice (Fig. 7G). However, the overall cytotoxic effect of these CTLs was greater after delivery of FK228 relative to that after delivery of PBS, as indicated by the percentage of apoptotic tumor cells (Fig. 7H). Interestingly, there was no difference in the percentage of apoptotic tumor cells in PBS-administered mice with or without adoptive transfer of CTL, suggesting that the MDA-MB231 P3 cells could be resistant to immune-mediated control due to the presence of the NANOG–HDAC1 molecular axis (Fig. 7H). Taken together, our data show that inhibition of HDAC1 represents an attractive, widely applicable strategy for the control of NANOG^high immunorefractory human cancer, either as a sole modality or synergistically, as part of an immune-based therapy.

Here, we identified that HDAC1 is a key component of several NANOG-dependent aggressive phenotypes, including immunoresistance, chemoresistance, and stem-like property of cancer cells. Our transcriptome profiling analysis demonstrated that HDAC1, which is directly regulated by NANOG, is functionally involved in the NANOG-mediated transcriptional network. In this study, we focused on the role of HDAC1 in the NANOG-mediated transcriptional repression process to determine the molecular mechanisms by which NANOG confers multiaggressive phenotypes in an HDAC1-dependent manner. In doing so, we discovered that HDAC1 contributes to NANOG-mediated CDKN2D and CDKN1B gene repression for a stem-like property, and it is required for NANOG-mediated TRIM17 and NOXA gene repression, leading to upregulation of antiapoptotic MCL-1, which confers immunoresistance and chemoresistance.

Cancer immunotherapy has been reasonably successful in generating tumor-specific immune responses, leading to significant antitumor effects. However, in some cases, the clinical response is still limited. Accumulating evidence indicates that tumors evade immune responses through extrinsic mechanisms associated with upregulation of immunosuppressive cytokines that create an immunosuppressive tumor microenvironment, or intrinsic mechanisms that generate resistance to immune eradication by downregulating MHC molecules and tumor antigens (34). In addition, we recently suggested yet another mechanism of immune escape that confers tumor cell resistance to CD8⁺ T-cell–mediated apoptosis (5, 33). In an effort to elucidate the molecular mechanisms underlying tumor immunoresistance, we found that immune pressure enforced through vaccination incites gain of NANOG, a master transcription factor that mediates the emergence of a stem-like cancer cell state and immune evasion (5). Thus, strategies that impede the NANOG signaling pathway may not only conquer the problem of immune escape but also that of the stem-like state in cancer. Here, we have found that NANOG confers immunoresistance, chemoresistance, and stem-like phenotypes to tumor cells through transcriptional induction of HDAC1. Becasue HDAC1 has been implicated as a central channel in the development of these aggressive phenotypes by NANOG, we believe that gain of HDAC1 may also underlie the failure of wide-spectrum clinical interventions. Recent reports have demonstrated synergistic effects of HDAC inhibitors and cancer immunotherapy (35–38). Our finding that the expression of class I HDACs, notably that of HDAC1, is increased in immunoedited tumor cells provides a strong rationale for targeting of immunorefractory tumors. Among various HDAC inhibitors, FK228 is more specific to HDAC1 and 2, and it has been approved for clinical use by FDA (39). It was reported that FK228 treatment at 0.1 to 1 mg/kg significantly prolongs the survival of mice bearing U-937 lymphoma (40). However, our data obtained from in vivo experiments showed that low-dose (0.05 mg/kg) treatment with FK228 only induces a slight decrease in proliferation of tumor cells, without any apoptotic effect on tumor cells. However, a combination of low-dose FK228 and adoptive transfer of CTL led to apoptosis of tumor cells. In this way, our results encourage the application of rational, mechanism-based combinations of selected epigenetic drugs and immune-based therapy for the treatment of relapsed tumor.

Previous studies have suggested that stemness factors, including NANOG, may contribute to epigenetic reprogramming in embryonic stem cells and adult stem cells (11, 12). However, the relationship between NANOG and epigenetic modification in cancer is largely unknown. In this study, we report that NANOG is capable of inducing epigenetic alterations by directly upregulating HDAC1. In this regard, we found that NANOG causes a decrease in AcH3K14 and AcH3K27 levels but does not alter other histone marks, such as AcH3K9. This HDAC1-dependent manner indicates that deacetylation of H3K14 and H3K27 can be potential epigenetic markers responsible for the NANOG–HDAC1 axis. Interestingly, although NANOG overexpression leads to a significant decrease in acetylation at H3K14 and K27, loss of HDAC1 in NANOG-overexpressing cells induces a relatively modest increase in acetylation levels at most sites of histone acetylation. These results suggest that NANOG-induced deacetylation of H3K14 and K27 may also be affected by other factors, such as histone acetyltransferases (HAT) in addition to HDAC1. Indeed, NANOG repressed the expression of HAT genes, including CLOCK, NCOA2, NCOA3, EP300, and CBP, in an HDAC1-dependent manner (Supplementary Fig. S6A and S6B). Of these, it was demonstrated that EP300 and CBP are specifically required for acetylation of H3K27 (41), and CLOCK showed a preference for acetylation of H3K14 than that at the other sites (42). It is not known how NANOG confers site-specific acetylation of H3, but this may partly be explained by findings that the NANOG–HDAC1 transcriptional program has been shown to decrease the expression of these HATs responsible for acetylation of H3K14 or H3K27. Moreover, decrease of these HAT genes, such as CLOCK, EP300, and CBP, may confer residue specificity on histone deacetylation by NANOG–HDAC1 axis (Supplementary Fig. S6C). Thus, our results provide insight into the link between the stemness factor NANOG and epigenetic reprogramming in tumor cells.

Indeed, we found that HDAC1 expression, in addition to NANOG (5), was elevated in multiple types of NANOG^high human tumor cells. Moreover, depletion of HDAC1 increased the susceptibility of tumor cells to chemotherapeutic agents, reduced the stem-like character of tumor cells, and rendered them more vulnerable to apoptosis by CTLs. These results were entirely consistent with those of NANOG depletion. This underscores the value of HDAC1 expression as a prognostic factor in cervical neoplasia, and it strongly suggests that the functions of HDAC1 in tumor cells that were observed in vitro hold true for cancer patients as well. Importantly, HDAC1 expression in cervical cancer tissue was positively correlated with NANOG expression, and it was negatively correlated with levels of AcH3K14 and AcH3K27, thus validating the proposed biochemical pathway. The NANOG–HDAC1–AcH3K14 and AcH3K27 molecular axis may also be critically related to progression of cervical neoplasia, as we have supporting data showing that the degree of the molecular axis is correlated with advanced disease stage and worse prognosis in patients.

Altogether, we propose that NANOG⁺ cancer cells enriched by immune selection drive (43) preferential expression of HDAC1 via transcriptional regulation and undergo HDAC1-mediated epigenetic reprogramming (Supplementary Fig. S7). In the process, HDAC1 promotes a stem-like phenotype as well as immune- and drug resistance mediated by NANOG. Despite its important role in enhancing NANOG-mediated aggressive features, HDAC1 also serves as a vulnerability factor by acting as the Achilles' heel, potentially leading to the decline of tumors if it is selectively targeted. Therefore, these findings emphasize that blockade of HDAC1 may be a promising therapeutic approach for cancer, especially if there is high NANOG expression.

T.-C. Wu is a consultant/advisory board member for Papivax LLC and Papivax Biotech Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.-H. Song, K.-M. Lee, T.W. Kim

Development of methodology: K.-H. Song, K.H. Noh

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.-H. Song, C.H. Choi, H.-J. Lee, S.J. Oh, S.R. Woo, S.-O. Hong, K.H. Noh, H. Cho, J.-H. Kim, J.-Y. Chung, S.M. Hewitt

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.-H. Song, C.H. Choi, S.-O. Hong, H. Cho, E.J. Chung, J.-Y. Chung, S.M. Hewitt, C. Yee, T.C. Wu

Writing, review, and/or revision of the manuscript: K.-H. Song, C.H. Choi, H. Cho, E.J. Chung, J.-Y. Chung, S.M. Hewitt, C. Yee, M. Son, C.-P. Mao, T.C. Wu, T.W. Kim

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.J. Chung, S. Baek

Study supervision: T.W. Kim

This work was funded by the National Research Foundation of Korea (NRF-2014R1A2A1A10054205 and NRF-2013M3A9D3045881).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.