CDK2 Inhibition Enhances Antitumor Immunity by Increasing IFN Response to Endogenous Retroviruses

Cell-cycle disorders are well-known associated factors in human cancers (1–5), either directly or indirectly mediated by the misregulation of cyclin-dependent kinases (CDKs; ref. 6). In the “classical” cell-cycle model, cyclin E and CDK2 regulate the G₁–S transition through retinoblastoma protein (RB) phosphorylation, leading to E2F-regulated gene expression and cell proliferation (7, 8). CDK inhibitors reduce the kinase activity of the cyclin/CDK complex and block the transition from G₁- to S-phase (7). CDK4/6 inhibitors such as ribociclib, abemaciclib, and palbociclib have been approved for clinical treatment of cancers, and CDK2 may be a potential target for reversing CDK4/6 inhibitor resistance in estrogen receptor–positive breast cancer (9). Many CDK2 inhibitors, such as PF-07104091 and CYC065, are currently in clinical trials (https://clinicaltrials.gov). In addition to competitive inhibitors, proteolysis-targeting chimeras (PROTAC) containing both CDK2 ligand and E3 ligase ligand have been developed for targeted degradation of CDK2 as an alternative approach to improve the efficacy of CDK2 inhibition (10, 11).

Additional studies have uncovered the role of the CDK family in other biological processes besides the cell cycle. For example, selective CDK4/6 inhibitors not only induce tumor cell-cycle arrest but also promote antitumor immunity through increased type III IFNs (12); cyclinD-CDK4 kinase destabilizes programmed death-ligand 1 (PD-L1) via Cul3^SPOP to control cancer immune surveillance (13); CDK1 phosphorylation of ULK1-ATG13 in mitosis connects autophagy to cell cycle (14). Although numerous studies have focused on the role of CDK2 inhibitors in treating tumors through the cell-cycle arrest, induction of apoptosis or differentiation (1, 2, 10, 15–17), little is known on whether CDK2 has other functions involved in antitumor immunity.

Approximately 10% of mammalian genome sequences correspond to endogenous viral elements, including endogenous retroviruses (ERV), which are thought to be derived from ancient viral infections of germ cells (18). Transcripts of ERVs are found in most tissues, and these transcripts have the potential to generate double-stranded RNA (dsRNA) by bidirectional transcription. Like the pathogen-associated molecular patterns of exogenous viruses, ERV transcripts may be detected by pattern recognition receptors, leading to activation of the innate immune response, including the production of pro-inflammatory cytokines, chemokines, and IFN-I (19). This endogenous viral element-derived immunity may play an important role in maintaining the basal antiviral state in host cells (20).

DNA methylation is the covalent addition of a methyl group to a DNA nucleotide. In most animal studies, DNA methylation only occurs on cytosines, particularly at CpG dinucleotide sites, established by DNA methyltransferases. DNMT3A, 3B, and 3L are members of the DNMT3 family; DNMT3A and DNMT3B are thought to perform all de novo methylation of DNA during development (21) whereas DNMT3L is catalytically inactive but stimulates the activity of the other DNMT3 enzymes. In contrast, DNA methyltransferase 1 (DNMT1) is thought to be the main enzyme responsible for maintenance of DNA methylation after replication (21). Proper regulation of DNNMT1 is necessary to maintain DNA methylation at imprinted genes. Cytosine methylation plays a key role in regulating gene expression, including expression of transposable elements such as ERVs (22). ERVs are generally fully methylated in all cells (23). Although some evidence points to ERV expression as a simple byproduct of epigenetic reprogramming, other studies indicate ERVs could contribute to multipotency (23). DNMT1 inhibitors 5-azacytidine and 5-aza-2′-deoxycytidine can increase the expression of multiple DNA hypermethylated ERVs, induce the IFN response, and regulate PD-L1 (24–26). By analyzing the tumor growth curves and gene expression profiles of MCA205 wild-type (WT) or Cdk2^−/− tumors in C57/BL6N and NU/NU mice, we discovered that CDK2-deficient cancer cells have enhanced antitumor immunity responses in the tumor microenvironment. We further identified a novel mechanism by which CDK2 inhibition triggered IFN-I response through increasing dsRNA from ERVs.

MCA205 fibrosarcoma cell line was generated and provided by Dr. S. A. Rosenberg (NCI, Bethesda, MD) in 2015. TC-1 lung cancer cell line was generated and provided by Dr T.C. Wu (John Hopkins University, Baltimore, MD) in 2015. H22, 293T, A549, and Huh7 cell lines were purchased from Cell Resource Center, Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences (Beijing, P.R. China) in 2018. Authentication of tumor cell lines (MCA205, TC-1, H22, 293T, A549, Huh7) with short tandem repeat DNA profile were performed every year with Procell Life Science & Technology Co., Ltd. (Wuhan, P.R. China). Mycoplasma contamination was tested by PCR with supernatant of cell culture at the first passage of cell recovery (Beyotime, catalog no. C0301S). Only cell cultures which were negative for Mycoplasma testing were used in this study. Normocin (InvivoGen, catalog no. ant-nr-1) and Plasmocure (InvivoGen, catalog no. ant-pc) were used to remove the contamination when Mycoplasma was detected. Only cell lines free of Mycoplasma after the withdrawal of Normocin and Plasmocure were used in all experiments and kept in the lab. TC-1, MCA205, 293T, A549, and Huh7 cells were cultured in DMEM (Thermo Fisher Scientific, catalog no. 11995040) and H22 cells were cultured in RPMI1640 (GIBCO, catalog no. C11875500BT). All of the media were supplemented with 10% FBS (Thermo Fisher Scientific, catalog no. 10099141) and 1% penicillin and streptomycin (Thermo Fisher Scientific, catalog no. 15140122). For subculturing adherent mammalian cells (TC-1, MCA205, 293T, A549, and Huh7 cells): removed and discarded the spent cell culture media from the culture vessel, washed cells using PBS (Gibco, catalog no. C1001050013T), removed and discarded the wash solution from the culture vessel, added the prewarmed trypsin (Gibco, catalog no. 25200-072) to the flask, incubated the culture vessel at room temperature for approximately 2 minutes, added prewarmed complete growth medium to the flask, transferred the cells to a 15 mL conical tube and centrifuged then at 1,100 rpm for 3 minutes, resuspended the cell pellet in a minimal volume of prewarmed complete growth medium and removed a sample for counting, diluted cell suspension to the seeding density, and pipet the appropriate volume into new cell culture vessels, and return the cells to the incubator. For passaging H22 cells grown in suspension culture: centrifuged the cell suspension at 1,100 rpm for 2 minutes, and resuspended the cell pellet in fresh growth medium and removed a sample for counting. Finally, seeded into new cell culture vessels. Plated 2 × 10⁵ cells into 6-well plates for overexpression and deletion experiments and 8 × 10⁵ cells for Western blot analysis and qPCR experiments. All cells were incubated at 37°C in a humidified air atmosphere with 5% CO₂. Cells underwent approximately four passages in culture before experiments. The procedure of cell harvesting was similar to the culture method, after adding complete medium and centrifugation, washed with PBS again, and added corresponding lysis reagents.

To generate Cdk2-deficient MCA205 and TC-1 cells, single-guide RNA (sgRNA) sequences target Cdk2 were synthesized and cloned into the LentiCRISPR V2 vector (a gift from Feng Zhang, Addgene plasmid #52961). To generate Cdk2^−/−Rb^−/− and Cdk2^−/−Mavs^−/− double-knockout MCA205 cells, sgRNA sequences target Rb or Mavs were synthesized and cloned into the LentiCRISPR v2 vector, respectively (synonym for Rb1 is Rb). All of the sgRNA sequences were selected from a sgRNA library (a gift from Dr. Feng Zhang) and synthesized by GENEWIZ in Suzhou. Lentivirus was produced in 293T cells by cotransfection of pMD2.G (Addgene plasmid #12259), psPAX2 (Addgene plasmid #12260) and Cdk2, Rb, or Mavs sgRNA plasmid or control vector plasmid. Virus supernatant was collected at 48 hours after transfection, passed through a 0.45 µmol/L filter, and stored at −80°C or used immediately. At 48 hours after infection, cells were cultured in puromycin (4 µg/mL for cells; InvivoGen, catalog no. ant-pr-1) selection medium for at least 7 days. For double-knockout cell lines, Cdk2^−/− cells were infected with Rb Lentiviruses for 2 days without puromycin selection. Monoclonal cells cultured in 96-well plate were acquired by FACSAria III cell sorter (Becton Dickinson). The sequences of the primers and sgRNAs used in the current study were listed in Supplementary Table S1.

MCA205 WT cells were treated with 5 µmol/L roscovitine (Selleck, catalog no. S1153) for 24 hours or with 1 µmol/L CPS2 (a gift from Yu Rao) for 24 or 96 hours. Cells were treated or not as described. After total cellular RNA was extracted by using NucleoZol (MNG, catalog no. 740404.200), RNA concentration was determined using the Nanodrop machine and software (Thermo Fisher Scientific). A total of 4 μg RNA was used to generate cDNA with PrimeScript II first-strand cDNA Synthesis Kit for qRT-PCR (TaKaRa, catalog no. 6110A). Ifnb1, Statl, Tlr3, Ddx58, Ccl5, Ccl7, Taq1, Mx1, Gapdh, Dnmt1, and MMERVK10C relative copy number were determined by calculating the fold-change difference in the gene of interest relative to Gapdh. qRT-PCR (Biomake, catalog no. B21202) was performed on the Applied Biosystems 7500 machine. All primer sequences used in the current study were listed in Supplementary Table S1.

DNA extraction and purification was performed with a kit (TIANGEN, catalog no. DP304), and DNA bisulfite conversion was performed with a kit (EpiArt DNA Methylation Bisulfite Kit, Vazyme, catalog no. EM101-01) according to supplier's recommendations. PCR amplification was conducted with the Zymo Taq DNA Polymerase (Zymo, catalog no. E2001). All PCR products were electrophoresed on 1.5% agarose gels and visualized using 4S Red Plus Nucleic Acid Stain (Sangon Biotech, catalog no. A606695-0100). The E.Z.N.A. Gel Extraction Kit (Omega, catalog no. D2500-02) was used to purify PCR products. The pEASY-T1 Simple Cloning Kit (TRANSGEN, catalog no. CT111) was used to clone PCR-purified products into pEASY-T1 Cloning Vector, which eliminates the multicloning sites and is designed for cloning and sequencing Taq-amplified PCR products. Products were sequenced by GENEWIZ. All primer sequences used in the current study are listed in Supplementary Table S1.

Components from the PrimeScript II first-strand cDNA Synthesis Kit for RT-PCR (Takara, catalog no. 6110A) were adapted to perform reverse transcription with RNA from Cdk2^−/− MCA205 cells as described previously (25, 27). Briefly, for first-strand cDNA synthesis, 50 ng RNA was used for actin control and 200 ng RNA for MMERVK10C cells. A total of 1 µmol/L of a gene-specific primer ligated to a TAG sequence not specific for the mouse genome [GSP sense/antisense (RT) TAG] was used in the reaction. RNA and primer were then preheated at 65°C for 5 minutes and placed on ice immediately. Then for the total reaction: 5×PrimeScript II Buffer (4 µL), RNase Inhibitor (40 U/µL) [0.5 µL (20 U)], PrimeScript II RTase (200 U/µL) [1 µL (200 U)], 250 ng Actinomycin D (Sigma, catalog no. A4262), and the above reaction solution was preheated and ice treated (10 μL). Actinomycin D was added to prevent second-strand cDNA RT resulting in antisense artifacts. We performed cDNA synthesis at 50°C for 50 minutes and ended at 95°C for 5 minutes. After cDNA synthesis, 2U recombinant RNase H (Beyotime, catalog no. D7089) was added to each reaction and incubated for 60 minutes at 37°C and ended with 0.5 mol/L EDTA (PH = 8; Thermo Fisher Scientific, catalog no. AM9260G). Finally, the first-strand cDNA mix was ethanol precipitated and resuspended in 10 µL sterile water. Gene- and strand-specific PCR were performed with the 2×EasyTaq PCR Super Mix (TRANSGEN, catalog no. AS111). To amplify sense cDNA and antisense cDNA, a TAG-primer and GSP sense (PCR) and a TAG primer and GSP antisense (PCR) were used, respectively. We performed sense and antisense specific PCR using both sense and antisense cDNA of β-actin as an internal negative control. All cDNA products were electrophoresed on 1% agarose gels, visualized using 4S Red Plus Nucleic Acid Stain (Sangon Biotech, catalog no. A606695-0100). All primer sequences used in the current study are listed in Supplementary Table S1.

For overexpression of DNMT1, the dCAS9-VP64_GFP plasmid (Addgene, catalog no. 61422) was digested with BamHI (NEB, catalog no. R0136S) and NheI (NEB, catalog no. R0131), and replaced the VP64 sequence with the cDNA sequence of Dnmt1 (Dnmt1 cDNA sequence was obtained through gene synthesis by GENEWIZ). 293T cells were transfected with the dCAS9-VP64_GFP plasmid, packaging plasmids psPAX2 (Addgene, plasmid catalog no. 12260) and envelope pMD2.G (Addgene, plasmid catalog no. 12259) using Lipofectamine 3000 (Invitrogen, catalog no. L3000-015) according to the manufacturer's instructions. Lentivirus was harvested from cell culture medium (DMEM + 10% FBS) after 48 hours in 6-well plate. Lentiviral particles were then collected by centrifugation (5,000 rpm/10 minutes) followed by passing through a 0.45 µmol/L filter and stored at −80°C or used immediately. Cdk2^−/− MCA205 cells were then infected with the DNMT1 lentivirus for 48 hours. Transfected cells were isolated by FACS. For overexpression of CDK2, we first constructed a plasmid dCAS9-CDK2_GFP with the help of GENEWIZ Company (Cdk2 cDNA sequence was obtained through gene synthesis). The base sequence of Cdk2 sgRNA in dCAS9-CDK2_GFP plasmid was changed but the encoded amino acid has not changed to avoid cutting by CRISPR CAS9 enzyme:

• Original: 5′-TGTGGCGCTTAAGAAGATCCGGCTCGACACT-3′

• Modified: 5′-TAGTCGCCCTAAAAAAAATAAGATTGGATACGG-3′

For overexpression of mutant CDK2 (T160A), we constructed a plasmid dCAS9-CDK2^T160A_GFP with the help of GENEWIZ Company. The base sequence which CDK2 sgRNA in dCAS9-CDK2^T160A_GFP plasmid was changed but the encoded amino acid has not changed, as above. We constructed two cell lines, WT-OVA and Cdk2^−/−-OVA MCA205 cells, stably expressing OVA according to the above method. For OVA plasmid, the VP64 sequence was replaced with the cDNA sequence of OVA by digesting with BamHI and NheI restriction sites in dCAS9-VP64_GFP plasmid (Addgene, catalog no. 61422).

Animal experimental protocols were approved by the Institutional Animal Care and Use Committee of Suzhou Institute of Systems Medicine (Suzhou, Jiangsu, P.R. China). C57BL/6N, athymic nude BALB/c mice (Nu/Nu), Ifnar^−/− C57BL/6 mice (female, weights between 18–20 g, ages between 6–8 weeks) were purchased from Model Animal Research Center of Nanjing University (Nanjing, P.R. China). B6.129(Cg)-Foxp3^{tm4(YFP/icre)Ayr}/J were generated from Rudensky's lab and obtained through Jax Lab (RRID:IMSR_JAX:016959). Mice were randomly divided into indicated groups (5–10 mice/group) before inoculation. WT, WT-V2 (transfected with CRISPR/CAS9 vector), WT-EV, Cdk2^−/−-EV (transfected with dCAS9_GFP plasmid) or corresponding Cdk2-deficient tumor cells (1 × 10⁶ cells in 100 µL PBS per mouse) were subcutaneously implanted in C57BL/6N mice. Chemotherapy was administered when tumor size was up to 30 to 40 mm² (normally 7 days after tumor cell injection), by intraperitoneal injection of mitoxantrone (MTX; 1 mmol/L, 200 µL; Sigma-Aldrich, catalog no. M6545) with three injections (at 7, 9, and 11 days). Roscovitine (15; Seliciclib/CYC202, catalog no. S1153, Selleck) was first dissolved in DMSO (1 volume; Solarbio, catalog no. D8371). A carrier solution was produced by using a diluent containing 10% Tween 80 (Sigma-Aldrich, lot no. BCBT1238), 20% N-N dimethylacetamide (Aladdin, lot no. G1911162), and 70% polyethylene glycol 400 (Sigma-Aldrich, lot no. BCBV0194). Mice were randomized into two groups (5 mice/group) before treatment. Roscovitine was administered intraperitoneally at 5, 6, 8, 9, 10, 11, 13, 15, 18, and 20 days. First treatment dose was 50 mg/kg, and for other treatments dose was 12.5 mg/kg. The control group received intraperitoneal injections of PBS following identical schedules. CPS2 was administered when tumor size was 30 to 40 mm² (normally 7 days after tumor cell injection) by intertumoral injection (2 µmol/L/mouse in 50 µL) at 7, 9, 11, 14, 17, and 20 days. The control group received intertumoral injections of DMSO following identical schedules. When appropriate, mice were treated with blocking antibodies (50 μg/20 g per injection) by tail vein injection against IFNAR1 (InVivoMAb, Clone: MAR1-5A3, BioXcell) on days 0, 3, 6, 9, 12, 14, and 18 after subcutaneous injection with cancer cells. In some experiments, CD8⁺ T cells were depleted by injecting 200 μg/mouse anti-CD8 antibodies (BE0004-1, BioXCell) intravenously at the indicated timepoints. The control group received same amount of PBS following identical schedules. All mice were sacrificed by asphyxiation with CO₂. Tumor growth was monitored by periodic measurement using calipers, multiplying tumor length and width. Tumors were harvested on day 10 postimplantation for flow cytometry analysis, immunofluorescence staining, enzyme-linked immune spot (ELISpot) assay, and RNA sequencing (RNA-seq). Animals were sacrificed when the volume of tumors reached 300 mm². Tumor progression was monitored 2 to 3 times per week and depicted as error bars of mean ± SEM at each timepoint.

Syngenic solid tumors from C57BL/6N mice injected with WT and Cdk2^−/− MCA205 cells were shredded by mechanical disruption, then chemically digested in dissociation buffer (DMEM containing 2 mg/mL collagenase type IV (Worthington Biochemical), 0.02 mg/mL DNase I (Sigma-Aldrich), and 5% FBS and PenStrep) with agitation at 37°C for 45 minutes. RBC lysis was required if RBC clumps were visible, otherwise no lysis was required (PharmLyse, BD Biosciences). Single-cell suspensions were incubated with appropriate antibodies for 30 minutes on ice. For detecting IFNγ⁺CD8⁺ cells, phorbol 12-myristate 13-acetate (Beyotime, catalog no. S1819) and Ionomycin (Beyotime, catalog no. S1672) were used to stimulate cells. Antibodies were diluted in PBS (Hyclone) plus 2% FBS. The Vivid Yellow stain kit (Invitrogen, catalog no. L34959) was used to exclude dead cells from further analyses. dsRNA was detected by the J2 antibody [English and Scientific Consulting Kft (SCICONS)] after cells were fixed with 75% ice-cold ethanol (Aladdin, catalog no. E111963) diluted in PBS for 60 minutes at −20°C and permeabilized with 0.1% Triton X-100 (Solarbio, catalog no.T8200) in PBS for 15 minutes at room temperature. Flow cytometry was performed on a Attune NxT Flow Cytometer (Thermo Fisher Scientific) and data was analyzed using FlowJo (Tree Star). All of the antibodies used in this study were listed in Supplementary Table S2.

Tumors were harvested at 10 days after subcutaneous transplantation and fixed in 4% paraformaldehyde (Beyotime, catalog no. P0099) for 1 day at 4°C, transferred to 30% sucrose (Sigma-Aldrich, catalog no. SIAL16104) solution for dehydration for 2 days at room temperature, embedded in OCT (Solarbio, catalog no. 4583) at room temperature, and store in a refrigerator at −20°C. Tissue sections were cut using a Leica cryostat (Leica CM1950) to a thickness of 5 µmol/L.

For immunofluorescence staining, frozen slices were left at room temperature for 20 minutes. After washing three times with PBS, slides were incubated with 4% paraformaldehyde for 15 minutes. After washing, slides were incubated with 0.2% Triton X-100 and with 10% FBS to block nonspecific sites of antibody adsorption for 10 minutes and subsequently incubated with primary antibodies at 4°C overnight. Then, the slices were washed three times with PBS and incubated with corresponding secondary antibody: goat anti-rabbit IgG conjugated with Alexa Fluor 488 (for cleaved caspase-3), goat anti-rabbit IgG conjugated with Alexa Fluor 568 (for CD11b), goat anti-mouse IgG conjugated with Alexa Fluor 568 (for CD8a; Thermo Fisher Scientific) or Cy3-labeled Goat Anti-Mouse IgG (H+L) (Beyotime, catalog no. A0521). After washing, slices were incubated at room temperature for 5 to 10 minutes in the dark with Hoechst (Thermo Fisher Scientific, catalog no. H3569) working solution (1:5,000) with one drop of antiquenching agent (SouthernBiotech, catalog no. 0100-01), then covered with a glass seal after the edges dried. Images were captured and processed with confocal microscope (Leica TCS SP6, installed at Suzhou Institute of Systems Medicine, P.R. China) and analyzed by ImageJ software. If the samples were subjected to quantitative analysis, the same microscope instrument settings were applied. All of the primary and secondary antibodies used in this study were listed in Supplementary Table S2.

IFNγ ELISpot assays kits (catalog no. 551881) were purchased from BD Bioscience and used to detect IFNγ-producing cells in the tumor microenvironment and in cocultivation experiments of WT or Cdk2^−/−_OVA MCA205 cells with OT1 cells. As for detecting IFNγ-producing cells in cocultivation experiments, briefly, CD8⁺ T cells were isolated from spleens and lymph nodes of OT-I mice by using PE antimouse CD8a Antibody (BioLegend, catalog no. 100707), anti-PE magnetic beads (Miltenyi Biotec) and an autoMACS pro separator. OT-1 cells (1 × 10⁵ cells/well) were cocultured with WT-OVA or Cdk2^−/−-OVA MCA205 cells (2 × 10⁴ cells/well) at a ratio of 5:1 for 16 to 18 hours. As for detecting IFNγ-producing cells in the tumor microenvironment, tumor single-cell suspensions were acquired as above. Multi-Screen 8 well strip Assay plates (EMD Millipore) were first coated with purified anti-mouse IFNγ antibody diluted in PBS (100 µL/well, BD, catalog no. 51-2525KZ) at 4°C. After 16 to 18 hours coculture with tumor single-cell suspensions (1 × 10⁶ cells/well) at 37°C in a 5% CO₂ and humidified incubator, the plate was incubated with biotiny-lated anti-mouse IFNγ antibody (BD, catalog no. 511818KA) and streptavidin–HRP (BD, catalog no. 557630), respectively. Final substrate solution was added to stop the reaction, and the plate was scanned and counted using CTL ImmunoSpot S6 Analyzers (LLC) after completely drying.

MCA205 WT cells were treated with 5 µmol/L roscovitine or 1 µmol/L CPS2 for 24 hours. Cells were treated or not as described and lysed in lysis buffer (Beyotime, catalog no. P0013). Protein concentration was determined using BCA Protein Assay Kit (Beyotime, catalog no. P0011). Equal amounts of total protein were separated by SDS-PAGE gels (10% or 12.5%, Epizyme, catalog no. PG112) and transferred to 0.45 µmol/L polyvinylidene difluoride membranes (Millipore, catalog no. IPVH00010). Membranes were blocked with 5% nonfat milk in Tris-buffer saline containing 0.05% Tween 20 (TBST). Subsequently, the membranes were incubated with specific primary antibodies. The membranes were scanned with the ChemiDoc XRS+ system (Bio-Rad). All primary and secondary antibodies used in this study were listed in Supplementary Table S2.

For total RNA-seq analysis, total RNA was extracted from cell lysates using RNeasy Mini Kit (QIAGEN, catalog no. 74104) according to the manufacture. Then, NEBNext rRNA Depletion Kit (NEB, catalog no. E6310) was used to prepare RNA for library construction. For CD45⁺ cell mRNA-seq analysis, after staining with anti-CD45.2, CD45⁺ cells were collected by FASC and RNA was extracted from cell lysates using RNeasy Micro Kit (QIAGEN, catalog no. 74004). Then, NEBNext Poly (A) mRNA Magnetic Isolation Module (NEB, catalog no. E7490) and NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (NEB, catalog no. E7760L) were used to prepare RNA-seq library according to the manufacturer's instructions. Briefly, after the rRNA depletion or poly-A selection steps, RNA first- and second-strand synthesis was performed, followed by NEBNext End Prep, Adaptor Ligation, Size Selection of Adaptor-ligated DNA, PCR Enrichment of Adaptor-ligated DNA and Cleanup of PCR Reaction steps. Final Library size distribution was 320 bp and sequenced on an Illumina Novaseq 6000 (Hiseq-PE150 sequencing strategy and paired-end 2*150 bp reads length). An average of 8 G bp raw bases and Q30>90% were generated for each sample. For sequencing alignment, the raw reads were filtered using CLC Genomics Workbench 12 (QIAGEN Bioinformatics) to remove low-quality and adaptor bases. In CLC Trim tool, quality scores limit is 0.05 and the maximum number of ambiguities nucleotides are 2. Filtered reads were mapped to Mus_musculus_ensembl_v80_sequence and GEVE reference sequences (http://geve.med.u-tokai.ac.jp/annotation-datasheet/) on CLC. We did not perform mitochondrial filtering.

Differential gene expression was determined using DESeq2 and genes were ranked according to the formulation [metric = −log (P,10)/log (FC,2)] for gene set enrichment analysis (GSEA).

For GSEA, we used the GSEA Preranked tool, with the above ranked gene list, using MSigDB v7.4 Hallmarks gene sets or immunologic signature gene sets collection and the “classic” method for calculating enrichment scores. We performed heatmap analysis using BMKCloud http://www.biocloud.net/, and ImmuCC http://218.4.234.74:3200/immune/ was used to estimate the relative immune cell compositions in mouse tissue from the transcriptome data profiled on RNA-seq platforms. The sequencing results are shown in Supplementary Table S3 for tumors samples and Supplementary Table S4 for cancer cells samples. The raw data are available in the NCBI BioProject database with the access number PRJNA738675.

IFNβ ELISA was performed with VeriKine Mouse IFNβ ELISA Kit (PBL, catalog no. 42400-1) according to supplier's recommendations. Data were collected from SpectraMax (Molecular Devices).

WT or Cdk2^−/− cancer cells were seeded (5×10⁵/well) in 6-well plate. Twelve hours later, MCA205 WT and Cdk2^−/− cells were infected with VSV-GFP virus (multiplicity of infection, MOI = 0.01) and FACS sorted for GFP after 24 hours infection. VSV-GFP virus was a gift from Dr Feng Ma (Suzhou Institute of Systems Medicine, Jiangsu, P.R. China). VSV-GFP virus (MOI = 0.1) were inoculated into Vero cells for proliferation, incubated until the cytopathic effect reached about 75%, and the supernatant was collected for purification by 30%, 45%, and 60% sucrose gradient centrifugation (26,000 g/2 hours).

Cell proliferation was determined manually using the cell counting method in chamber slides. Cell count and viable cell rate after 1 µmol/L CPS2 treatment was calculated for indicated time. Cells were plated at 1 × 10⁵ cells/mL and cultured for 1, 2, 3, or 4 days, then the cell number was counted with trypan blue to determine the dead cells to analyze cell viability. Cell supernatant was replaced with fresh medium every 2 days. The cell count number and viability were further analyzed and plotted using GraphPad Prism v.6. Cell proliferation was also analyzed with Cell Counting Kit-8 (Beyotime, catalog no. C0038) according to supplier's recommendations.

Statistical analyses were performed with software GraphPad Prism 6. The continuous variables are presented as mean ± SEM. Data with normal distribution were analyzed by one-way ANOVA or unpaired two-tailed Student t tests, survival curve comparison was analyzed by log-rank (Mantel–Cox) test and tumor growth curves were compared by the Mann–Whitney U test. P values were indicated by *, P < 0.05; **, P < 0.01; ***, P < 0.001 and ****, P < 0.0001.

The accession number for the RNA-seq data reported in this article is Sequence Read Archive PRJNA738675. The datasets and materials used or analyzed during the current study are available from the corresponding authors on reasonable request.

The success of CDK4/6 inhibitors for breast cancer (12, 28, 29) increased interest in therapeutically targeting other CDKs. We found that CDK2 was highly expressed in a variety of tumors in the TIMER2.0 database (Supplementary Fig. S1A), which was consistent with other reports (30–32). We also analyzed the survival rate of patients with high or low expression of CDK2 by using both GEPIA2 and Kaplan–Meier Plotter online databases. As shown in Supplementary Figs. S1B–S1D, the disease-free survival rate in lung adenocarcinoma and the overall survival rate in sarcoma were higher in patients with low expression of CDK2. The above results indicate CDK2 is frequently overexpressed in tumors and is associated with worse patient outcome.

We assessed the impact of CDK2 inhibition on in vivo tumor growth of subcutaneous MCA205 mouse fibrosarcoma cells in syngeneic immunocompetent C57BL/6N WT mice and in immunocompromised nude mice. The CDK2 inhibitor roscovitine (Seliciclib) administered intraperitoneally caused slower growth of bulky tumors, evidenced by approximately 60% reduction in tumor volume, with one of five achieving complete clearance, in C57BL/6N mice at the day 27 endpoint (Fig. 1A). There was no significant difference in the tumor curves of nude mice treated with PBS or roscovitine (Fig. 1B). These results suggest that the roscovitine-mediated tumor regression may depend on the intact immunity in the recipient mice.

To further determine the potential relationship between CDK2 in cancer cells and antitumor immunity in the recipient mice, we used CRISPR/Cas9 to establish several independent Cdk2^−/− clones of MCA205 fibrosarcoma and TC-1 lung carcinoma cells. The CDK2 expression of these Cdk2^−/− MCA205 cells, Cdk2^−/− TC-1 cells, and their parental WT cells were verified by Western blot analysis (Supplementary Fig. S1E–S1G). Considering the role of CDK2 in the cell cycle, we compared the in vitro growth of WT and Cdk2^−/− MCA205 cells and there was no significant difference between growth rates (Supplementary Fig. S1H and S1I). These results indicate that there is no significant difference between WT and Cdk2^−/− cells in vitro cell proliferation.

After transplantation of WT and Cdk2^−/− MCA205 or TC-1 cells into syngeneic immunocompetent C57BL/6N mice by subcutaneous injection, both Cdk2^−/− MCA205 and Cdk2^−/− TC-1 tumors exhibited a significant reduction in tumor growth as compared with their WT parental cells (Fig. 1C and D; Supplementary Fig. S1J). Besides the tumor sizes, we also observed that the survival duration of mice transplanted with Cdk2^−/− cells was greater than WT cells (Supplementary Fig. S1K). To rule out potential effects of CRISPR/Cas9on tumor growth, WT and V2-vector MCA205 cells were also implanted into C57BL/6N mice, resulting in no significant difference between their tumor sizes (Supplementary Fig. S1L). Both Cdk2^−/− MCA205 and TC-1 cells grew into tumors of similar sizes as their corresponding WT cells in NU/NU mice (Fig. 1E and F). To determine whether the reduced tumor growth of Cdk2^−/− MCA205 cells depends on CD8⁺ T cells, CD8-targeting antibodies were administered by intravenous injection. Cdk2^−/− MCA205 tumors grew faster in mice injected with anti-CD8 antibodies as compared with Cdk2^−/− MCA205 tumors in mice injected with PBS (Fig. 1G). These results together suggest that CDK2 deficiency may affect tumor growth in an immune-dependent fashion, instead of directly affecting cancer cell proliferation in vitro or in vivo.

Combining chemotherapies with targeted therapies has been effective in reducing the rate of insufficient response. The efficacy of the combination of CDK2 inhibition and MTX in treating fibrosarcoma remains unknown. We therefore compared the effects of MTX on suppressing tumor growth of WT and Cdk2^−/− MCA205 fibrosarcoma implanted in C57BL/6N mice and NU/NU mice. Cdk2^−/− MCA205 tumors were more sensitive to anthracycline chemotherapy compared with WT tumors (Fig. 1H). This combinatorial effect depended on the intact immunity in the recipient mice, as MCA205 fibrosarcoma implanted in NU/NU mice failed to reduce their growth in response to MTX (Fig. 1I).

Our observations of immune-mediated tumor growth differences between WT and Cdk2^−/− cancer cells prompted us to explore the WT and Cdk2^−/− tumor microenvironment. At 10 days after subcutaneous transplantation, a considerably higher percentage of CD8⁺ T cells and CD4⁺ T cells were observed in Cdk2^−/− tumors than in WT tumors (Fig. 2A; Supplementary Fig. S2A), suggesting enhanced antitumor immune responses in the Cdk2^−/− tumor microenvironment. An optimal and efficient antitumor immune program also requires cooperation between dendritic cells (DC) and T cells. We sought to characterize whether Cdk2 depletion would impact the DC population (MHCII⁺CD11c⁺). Compared with WT tumors, a significant increase in percentages of tumor-resident DCs was observed in Cdk2^−/− tumors (Fig. 2B; Supplementary Fig. S2B). ELISpot assays also demonstrated that knockout Cdk2 in cancer cells significantly increased the secretion of IFNγ in the tumor microenvironment (Fig. 2C and D).

We also performed RNA-seq analysis on RNA extracted from CD45⁺ cells in WT and Cdk2^−/− tumors (Supplementary Table S3). The ImmuCC analysis on the relative immune cell compositions based on the expression of cell type–specific biomarkers (http://wap-lab.org:3200/immune/; ref. 33) indicated higher numbers of natural killer (NK) cells, DC cells, CD8⁺ T cells, and CD4⁺ T cells in Cdk2^−/− tumors than WT tumors (Fig. 2E).Genes involved in the IFNα and IFNγ pathways (Fig. 2F and G) as well as several immune signature pathways (Supplementary Fig. S2C), such as “effector versus exhausted CD8 T cell_up” and “CD8 T cell versus NK cell_up,” were enriched in Cdk2^−/− tumors compared with WT tumors based on GSEA.

Cdk2^−/− tumors exhibited an increase in the expression of genes linked to IFN response, immune response, cytokine secretion, and T-cell activation (Fig. 2H), which included IFN genes Ifnb1 and Ifng themselves consistent with the result of the IFNγ ELISpot (Fig. 2C). The transcription factors Irf7, Irf1, and Irf8, antigen presentation-related genes B2m, Tap2, and Tap1, chemokine-encoding genes Cxcl10, Cxcl5, Ccl7, Ccl2, Il4, and Il7, as well as the activated CD8⁺ marker genes Gzmb and Gzma were also upregulated in Cdk2^−/− tumors (Fig. 2H).

We then used immunofluorescence followed by microscopic quantitation to determine immune cell infiltration in the tumor microenvironment. Increased numbers of infiltrating CD8⁺ cells and CD11b⁺ cells, as well as apoptotic cancer cells marked by activated caspase-3, were observed in Cdk2^−/− tumors as compared with WT tumors. Altogether, these results further confirmed that Cdk2 depletion can promote T-cell and DC recruitment into tumors and enhance antitumor immune gene programs.

The above results showing an enhanced IFN response, a rise in chemokines such as Cxcl10, and an increased infiltration of tumor-infiltrating lymphocytes (TIL) in the Cdk2^−/− tumor microenvironment suggest a possible link between IFN-mediated tumor cell chemokine expression and increased TIL infiltration, which may be responsible for the enhanced antitumor immune response in Cdk2^−/− tumors. To test this hypothesis, RNA-seq was performed with RNA extracted from WT and Cdk2^−/− MCA205 cells (Supplementary Table S4). MCA205 Cdk2^−/− cells exhibited an increased expression of genes linked to the type I IFN response and enriched in the “IFNα response” pathway (Fig. 3A and B). By qRT-PCR analysis, we have further confirmed some of these genes, including the type I IFN gene Ifnb1 itself, the transcription factor Stat1, the antiviral gene Mx2, the pattern recognition receptor genes Tlr3 and Ddx58, the antigen presentation related gene Tap1, as well as the chemokine-encoding genes Ccl5 and Ccl7 (Fig. 3C). We speculated that increased IFN-stimulated gene (ISG) expression might be caused by enhanced IFN production in MCA205 Cdk2^−/− cells. Notably, mRNA and protein abundance of IFNβ was significantly increased in Cdk2^−/− cells by qRT-PCR (Fig. 3C) and ELISA (Fig. 3D), respectively. Collectively, these data suggest that CDK2 deficiency in MCA205 cells may drive ISG expression by increasing IFN production, including IFNβ.

To confirm the biological effects of the enhanced type I IFN response in Cdk2^−/− cancer cells, we tested whether they exhibited increased antiviral activity. WT and Cdk2^−/− MCA205 cells were infected with VSV-GFP (MOI = 0.01), and the GFP fluorescence intensity at 24 hours postinfection was measured by flow cytometry. As expected, the GFP fluorescence of MCA205 Cdk2^−/− cells was weaker compared with that in WT cells, suggesting that Cdk2^−/− cancer cells had stronger antiviral activity than WT cells (Supplementary Fig. S3A and S3B).

Beside antiviral ISGs, many ISGs may play important roles in antitumor immunity. The high expression of chemokine-encoding genes Cxcl10, Ccl5, and Ccl7 in Cdk2^−/− cancer cells was correlated with the increased immune cell infiltration in Cdk2^−/− tumors (Fig. 2). To determine the functional consequences of increased antigen presentation gene expression, we constructed ovalbumin (OVA) stably expressing cancer cell lines (WT-OVA and Cdk2^−/−-OVA MCA205). Co-culture of Cdk2^−/− cancer cells with MHC class I-restricted OVA-specific CD8⁺ T cells (OT-I cells) significantly increased IFNγ production in OT-I cells, based on ELISpot analysis (Fig. 3E and F). These results suggest that an increased IFN-I response may contribute to intensified antigen presentation and increased immune cell infiltration and activation in Cdk2^−/− tumors.

To further address whether the enhanced IFN-I pathway of Cdk2^−/− cancer cells was involved in the antitumor immune response in vivo, we treated WT C57 BL/6N mice carrying WT or Cdk2^−/− MCA205 tumors with neutralizing antibodies against the heterodimeric type I IFN receptor (IFNAR1). The IFNAR1-blocking antibodies significantly increased the tumor growth of Cdk2^−/− cells while having little effect on the tumor growth of WT cells (Fig. 3G). We also transplanted WT or Cdk2^−/− MCA205 cancer cells to Ifnar1^−/− mice and found that WT and Cdk2^−/− grew tumors at similar rates in Ifnar1^−/− mice (Fig. 3H). These results suggest that the elevated IFN-I response of Cdk2^−/− tumors may improve antitumor immune response and reduce the growth of Cdk2^−/− tumors. To further determine the contribution of mitochondrial antiviral-signaling protein (MAVS), a critical adaptor in RNA-mediated IFN-I induction pathway, we generated Cdk2^−/−Mavs^−/− MCA205 cells. The MAVS expression of these Cdk2^−/−, Cdk2^−/− Mavs^−/− MCA205 cells and their parental WT cells were verified by Western blot analysis (Supplementary Fig. S3C). Cdk2^−/−Mavs^−/− tumors grew faster than Cdk2^−/− tumors, suggesting that the RNA-mediated IFN-I response may be responsible for the durable antitumor effect of mice bearing Cdk2^−/− cells (Fig. 3I). Thus, these data provide in vivo evidence supporting a model in which type I IFN secreted from CDK2-deficient tumor cells stimulate the host type I IFN signaling response and establish robust antitumor immunity. Administration of an IFNAR1 blocking in this system to inhibit both tumor and nontumor IFNAR responses completely prevented tumor regression. These results suggest that onset of antitumor immunity is dependent on the effects of tumor-derived type I IFN on Ifnar^+/+ host cells.

Increased IFN-I response via IFNβ production in tumor cells has been shown to occur in response to DNA demethylation caused by 5-azacytidine, which inhibits the activity of DNMT (24–26). We reanalyzed our RNA-seq data and found that although other Dnmt genes, including Dnmt3a, were upregulated, Dnmt1 was downregulated in MCA205 Cdk2^−/− cells (Fig. 4A; Supplementary Table S4). Although DNMT3s were upregulated, DNMT1 was downregulated to a greater degree. We also used the GEPIA2 database to run a correlation analysis of CDK2 and DNMT1 expression in publicly available sarcoma gene expression profiles and found that they were positively correlated (R = 0.54; Supplementary Fig. S4A). Dnmt1 mRNA expression dropped markedly in roscovitine-treated MCA205 cells line (Supplementary Fig. S4B), and both Dnmt1 mRNA and protein were reduced in Cdk2^−/− as compared with WT MCA205 cells (Fig. 4B and C). These data suggest the critical role of CDK2 in maintaining the expression of Dnmt1.

In accordance with previous studies (25, 26) showing that Dnmt1 maintains DNA methylation at ERV 3′ long-terminal repeat sequences, bisulfite-sequencing showed that DNA methylation at the 3′ LTR of the beta-like ERV MMERVK10C was reduced in Cdk2^−/− MCA205 cells (Supplementary Fig. S4C). More importantly, the expression of MMERVK10C (Fig. 4D) and numerous other ERVs (Supplementary Fig. S4D) were upregulated in MCA205 Cdk2^−/− cells, based on our qPCR and RNA-seq analyses, respectively. Further linking ERVs with a dsRNA-triggered IFN response, analysis by the TAG-aided sense/antisense transcript detection technique (27) detected bidirectional transcription producing sense and antisense transcripts of MMERVK10C non-repeat sequences, but not β-actin, in Cdk2^−/− MCA205 cells (Supplementary Fig. S4E). Quantification of dsRNA performed by flow cytometry or immunofluorescence using the dsRNA-specific J2 antibody showed a significantly higher abundance of dsRNA within Cdk2^−/− MCA205 cells than those within WT MCA205 cells (Fig. 4E–G). In addition, the mRNA abundance of the dsRNA pattern recognition receptors, Ddx58 and Tlr3, was significantly increased in Cdk2^−/− MCA205 cells (Fig. 3C).

To further determine whether reduced expression of Dnmt1 is responsible for the increased IFN-I response and antitumor immunity in Cdk2^−/− MCA205 cells, we used dCAS9-VP64_GFP plasmid to generate cell lines overexpressing DNMT1 in Cdk2^−/− cancer cells (Supplementary Fig. S4F). We found that overexpression of DNMT1 attenuated the abundance of dsRNA (Fig. 4H and I), MMERVK10C (Fig. 4J), dsRNA recognition receptor genes, Ifnb1 itself and ISGs (Fig. 4K). Overexpression of DNMT1 significantly increased the tumor growth of the Cdk2^−/− cells to similar rate as the WT cells (Fig. 4L). To rule out potential differential growth effects of the VP64 vector, WT and VP64-vector MCA205 cells were implanted into C57BL/6N mice, resulting in no differences among their tumor sizes (Supplementary Fig. S4G). These results indicate that downregulation of Dnmt1 expression in Cdk2^−/− cancer cells results in decreased methylation of ERVs and elevated ERV dsRNA transcripts, which trigger the activation of IFN-I response and antitumor immunity.

According to published studies, CDK2initiated phosphorylation of the RB protein results in its inactivation and disassociation from E2Fs, allowing E2F to activate gene transcription (34, 35). We observed that the phosphorylation level of RB (S807/811 sites) was lower in Cdk2^−/− cells than that in WT cells (Fig. 5A). Because mammalian Dnmt1 is a bona fide E2F target gene (36), we sought to determine whether RB is involved in the reduced expression of DNMT1 in Cdk2^−/− cells. We generated Cdk2^−/−Rb^−/− MCA205 cell lines using CRISPR/Cas9, and showed that knockout of RB increased both mRNA and protein abundance of DNMT1 in Cdk2^−/− MCA205 cells (Fig. 5B and C). On the other hand, the elevated expression of MMERVK10C (Fig. 5D), Ifnb1, Stat1, Tlr3, and ISGs observed in Cdk2^−/− MCA205 cells (Fig. 5E) were reduced in Cdk2^−/−Rb^−/− cells to similar quantities as WT cells. Both WT and Cdk2^−/−Rb^−/− MCA205 cells rapidly developed tumors at higher rates than Cdk2^−/− cells after transplantation into syngeneic immunocompetent C57BL/6N mice (Fig. 5F). The above results supported the idea that the CDK2-RB-E2F-DNMT1 axis may mediate the IFN-I response through regulating ERV expression.

To further determine whether CDK2 kinase activity is required for regulating ERV expression, IFN-I response and antitumor immunity, we reconstituted Cdk2^−/− cancer cells with WT Cdk2 or a kinase-dead (T160A) mutant of Cdk2 (ref. 37; Fig. 5G). Accordingly, we found restoration of tumor growth after transplantation of Cdk2^−/− cancer cells reconstituted with WT Cdk2 but not with Cdk2^T160A (Fig. 5H). We also found that dsRNA formation and the IFN-I response were inhibited in Cdk2^−/− cells reconstituted with WT Cdk2 but not with Cdk2^T160A (Fig. 5I and J).

Overall, our studies suggest a novel CDK2-mediated immune regulation pathway that proceeds through the RB-E2F-DNMT1 axis to control ERV transcription and subsequent IFN-I responses, which we believe plays an important role in regulating antitumor immune responses, such as antigen presentation and immune cell recruitment in tumor microenvironment (Fig. 5K).

Compared with competitive kinase inhibitors, PROTACs demonstrate superior selectivity and higher potency (10, 11, 38, 39). CPS2, a CDK2 PROTAC, has been shown to specifically degrade CDK2 without affecting other proteins (10).

To determine the efficacy of CPS2-induced CDK2 degradation, we treated MCA205 cells with different concentrations of CPS2 for 24 hours and found that CPS2 at the concentration of 250 nmol/L can effectively degrade CDK2 (Fig. 6A). To understand whether short-term inhibition of CDK2 affected the cell cycle, we performed cell proliferation and cell viability tests. The results demonstrated that CPS2 did not inhibit the proliferation nor affect the viability of MCA205 cells (Fig. 6B and C). Both the mRNA and protein of DNMT1 were reduced after CPS2 treatment (Fig. 6D and E). MMERVK10C transcription (Fig. 6F) and the formation of dsRNA (Fig. 6G and H) were increased after CPS2 treatment. In addition, the mRNA expression of ISGs such as Ddx58, Tlr3, Ifnb1, Stat1, Ccl5, and Ccl7 was also increased after CPS2 treatment (Fig. 6I). Moreover, more CD8⁺ T cells, CD8⁺ IFNγ⁺ T cells, DCs, monocytes, neutrophils, and macrophages, but fewer regulatory T cells, infiltrated into the tumor microenvironment after in vivo treatment with CPS2, as compared with the DMSO-treated group (Fig. 6J and K; Supplementary Fig. S5). CPS2 treatment also significantly reduced MCA205 and H22 hepatocellular carcinoma tumor growth in C57/BL6N and BALB/c mice, respectively (Fig. 6L and M), but not in NU/NU mice (Fig. 6N). We also found decreased expression of DNMT1, P-RB, and CDK2 in H22 cells after CPS2 treatment, which was consistent with the results of MCA205 cells (Supplementary Fig. S6A). Finally, the tumor-bearing mice treated with CPS2 survived longer than those treated with DMSO treatment (Supplementary Fig. S6B and S6C).

For translational relevance, we assessed human A549 lung carcinoma and Huh7 hepatocellular carcinoma cell lines treated in vitro with CPS2 to ensure that the ERV and IFN induction by CDK2 inhibition was conserved from mouse to human. As expected, we observed the same phenomenon in the human cell lines (Supplementary Fig. S6D and S6E). These results indicate that ablation or inhibition of CDK2 can have the same effect on mouse and human cell lines, and both can cause an increase in endogenous retrovirus and IFNβ expression.

Although there are many reports about CDK2 inhibitors suppressing tumor growth, most of them clarify the mechanism as arresting cell cycle (7, 16, 17), inducing apoptosis (17, 40) and inducing differentiation (10). Other functions of CDK2, including antitumor immunity, are not very clear. Through CDK2 inhibitors, such as roscovitine and PROTAC-CPS2, or gene knockout by CRISPR/Cas9, we showed that CDK2 is dispensable for the proliferation of MCA205 cells or TC-1 cells in vitro. Although some studies indicate that CDK2 is required for cell proliferation (3, 7, 15), many other reports suggest that CDK2 is nonessential (6, 41–44).

We analyzed the growth and microenvironment of WT MCA205 tumors in the presence or absence of CDK2 inhibitors, as well as Cdk2^−/− MCA205 cells in syngeneic immunocompetent C57BL/6N mice with and without anti-CD8 treatment and immunodeficient NU/NU mice. Our results demonstrated a novel mechanism responsible for the antitumor activity of CDK2 inhibition, through enhancing the antitumor immune response in the tumor microenvironment. Our study has provided evidence showing that CDK2 plays an important role in controlling the abundance of cellular dsRNA from endogenous retroviral transcripts through DNMT1-mediated DNA methylation. We found numerous ERV transcripts and their associated dsRNAs upregulated, while the expression of DNMT1 was downregulated in Cdk2^−/− cells, as compared with the corresponding WT MCA205 cells. We also found that overexpression of DNMT1 in Cdk2^−/− cells reduced ERV transcript and dsRNA abundance to similar amounts as WT cells, suggesting a link from CDK2 to DNMT1 to ERV dsRNAs. The relationship between DNMT1, ERV, and dsRNA has been reported in many studies (12, 19, 24–26). DNMT1 is the known target gene of E2Fs regulated by E2F-RB (36, 45), and RB can be phosphorylated and inactivated by CDK2 (2, 34, 35, 41). We have shown that the high abundance of ERV transcripts and dsRNAs found in Cdk2^−/− cells was reduced back to WT-like quantities in Cdk2^−/−Rb^−/−cells but not in Cdk2^−/− cells reconstituted with T160A-mutant CDK2 that cannot phosphorylate RB. These observations together suggest that CDK2 controls ERV transcription through RB phosphorylation via the CDK2-RB-E2F-DNMT1-ERV pathway.

Our results have also demonstrated that Cdk2^−/− cells have hyper-activated IFN-I responses, which include upregulation of dsRNA receptor genes such as Tlr3 and Ddx58, IFN-I genes such as IFNβ, IFN signaling molecules such as Stat1, antiviral ISGs such as Mx2, chemokines such as Ccl5 and Ccl7, as well as antigen presentation molecules such as Tap1. In addition to Cdk2^−/− cells, Cdk2^−/− cells reconstituted with T160A-mutant CDK2 also have high abundance of dsRNAs and hyperactivation of the IFN-I response, whereas Cdk2^−/−Rb^−/−cells and DNMT1-overexpressed Cdk2^−/− cells have similar basal quantities of dsRNAs and IFN response gene transcripts as WT MCA205 cells. These data suggest that accumulation of dsRNAs through the CDK2-RB-E2F-DNMT1-ERV pathway is responsible for the hyperactivation of IFN-I responses in Cdk2^−/− cells.

Our data suggest that the durable antitumor effects of mice bearing Cdk2^−/− cells are mediated by IFN-I–dependent activation of antigen presentation and immune cell recruitment in the tumor microenvironment. Although Cdk2^−/− MCA205 cells proliferated in vitro and grew tumors in immunodeficient nude mice at similar rates as WT MCA205, Cdk2^−/− tumors were much smaller in immunocompetent C57BL/6N mice than WT tumors. Furthermore, the antitumor effect of CDK2 ablation was ablated with anti-CD8 treatment, suggesting the reduced growth of Cdk2^−/− tumors is due to enhanced antitumor immune responses in the tumor microenvironment provided by the C57BL/6N recipients. Indeed, we observed increased antigen presentation and infiltration of immune cells, including CD8⁺ T cells and myeloid cells such as DCs in Cdk2^−/− tumors. This enhanced antitumor activity likely depends on increased IFN-I responses, such as induction of antigen presentation genes and immune cell recruitment chemokines. The differences in WT and Cdk2^−/− tumor growth rates were diminished when C57BL/6N mice were treated with blocking antibodies against IFNAR1 or when implanted into IFNAR1-deficient mice. Overall, our studies provide a working model that can explain the enhanced antitumor immunity associated with CDK2 inhibition as shown in Fig. 5K. CDK2 as a checkpoint kinase phosphorylates RB and activates E2F to upregulate DNMT1, which methylates ERV genes and suppresses the basal quantity of ERV transcripts, dsRNAs, and the IFN-I response. When CDK2 activity is inhibited in Cdk2^−/− cells or by CDK2 inhibitors, RB is underphosphorylated and E2F-mediated DNMT1 is reduced, which leads to increased ERV transcription, dsRNA abundance, and IFN-I response. The activated IFN-I response in cancer cells may trigger antitumor immunity in the tumor microenvironment by enhancing antigen presentation and recruitment of immune cells. Further studies are needed to determine how CDK-deficient cancer cells interact with the neighboring immune cells to trigger antitumor immunity in the tumor microenvironment.

Since the 2015 FDA approval of Palbociclib, the first CDK4/6 inhibitor, numerous CDK inhibitors, including CDK2-selective inhibitors, have begun development. In addition to the traditional competitive inhibitors, PROTAC technology, as an effective way of targeting and degrading protein molecules, provides a more promising method by reducing off-target effects or inhibiting multiple targets (10, 11, 38, 39). Our results show that the CDK2 PROTAC drug CPS2 had no significant effect on cell proliferation and could strongly reduce DNMT1 expression, increase ERV dsRNA and ISG expression, and promote infiltration of immune cells, including CD8⁺ T cells, CD8⁺IFNγ⁺ T cells, and myeloid cells such as DCs, neutrophils, monocytes, and macrophages following CPS2 treatment. We noticed that tumors grew slower in the roscovitine-treated group than the untreated group in NU/NU mice (Fig. 1E), whereas the tumor sizes in NU/NU mice administered with CPS2 or DMSO were almost the same (Fig. 6L). Further studies are needed to determine whether these differences are due to different administration methods or specificities between roscovitine and CPS2, as roscovitine inhibits CDK1/2/5/7/9 kinase activity whereas CPS2 is more selective for CDK2. We also assessed two human cell lines treated with CPS2 to ensure that CDK2 inhibition was conserved from mouse to human. Our studies have provided insight into a novel mechanism by which CDK2 inhibitors suppress tumor growth through enhancing ERV-mediated antitumor immunity.

Y. Rao reports grants from NSFC during the conduct of the study. No disclosures were reported by the other authors.

Y. Chen: Investigation, methodology, writing–original draft. Q. Cai: Investigation. C. Pan: Investigation. W. Liu: Investigation. L. Li: Funding acquisition, investigation. J. Liu: Investigation. M. Gao: Investigation. X. Li: Investigation. L. Wang: Investigation. Y. Rao: Investigation. H. Yang: Supervision, funding acquisition, investigation. G. Cheng: Supervision, methodology, writing–review and editing.

This project is financially supported by Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences (2021-I2M-1-047 and 2019-I2M-1-003), National Natural Science foundation of China (82073181 and 81802870), the Non-profit Central Research Institute Fund of Chinese Academy of Medical Sciences (2020-PT310-006, 2019XK310002, and 2018TX31001), the Key Project of Jiangsu Provincial Health Commission (K2019021), U.S. NIH funds (AI069120, AI056154, AI140718, and AI028697). H. Yang is supported by Natural Science foundation of Jiangsu Province (BK20211554 and BK20170407) and the Innovative and Entrepreneurial Team grant (2018-2021) from Jiangsu Province. L. Li is supported by the Chinese Postdoctoral Science Foundation (2019M650564) and Innovative and Entrepreneurial Doctor grant (2020-2022) from Jiangsu Province. We appreciate the help of animal facility and Immunology Platform at Suzhou Institute of systems Medicine.

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.