Immunologic checkpoint blockade has been proven effective in a variety of malignancies. However, high rates of resistance have substantially hindered its clinical use. Understanding the underlying mechanisms may lead to new strategies for improving therapeutic efficacy. Although a number of signaling pathways have been shown to be associated with tumor cell–mediated resistance to immunotherapy, T cell–intrinsic resistant mechanisms remain elusive. Here, we demonstrated that diacylglycerol kinase alpha (Dgka) mediated T-cell dysfunction during anti–PD-1 therapy by exacerbating the exhaustion of reinvigorated tumor-specific T cells. Pharmacologic ablation of Dgka postponed T-cell exhaustion and delayed development of resistance to PD-1 blockade. Dgka inhibition also enhanced the efficacy of anti–PD-1 therapy. We further found that the expression of DGKA in cancer cells promoted tumor growth via the AKT signaling pathway, suggesting that DGKA might be a target in tumor cells as well. Together, these findings unveiled a molecular pathway mediating resistance to PD-1 blockade and provide a potential therapeutic strategy with combination immunotherapy.
Immunologic checkpoint blockade therapies have changed the paradigm of cancer treatment, leading to significant clinical benefit and durable response in a variety of cancer types (1–6). Anti–PD-1 therapy significantly prolongs progression-free survival and overall survival in patients with advanced melanoma and lung cancer (7–9). Its clinical benefit is further enhanced by combination with anti–CTLA-4 therapy (10, 11). However, the clinical response rate to anti–PD-1 treatment remains variable from 10% to 40% in different cancers (12, 13), limiting its utility in cancer therapy (14). Approximately 60% of patients who receive PD-1 antibody display primary resistance (15), and a substantial number of patients who initially respond to immunotherapy develop acquired resistance over time (16). A study shows that approximately 25% of melanoma patients develop acquired resistance, defined as disease progression at a median follow-up of 21 months (17). The primary and acquired resistance severely constrains the clinical benefits of checkpoint blockade therapies. Few effective therapeutic options are available for such patients, suggesting the urgency of improving our understanding of the underlying resistant mechanisms and overcoming it through a combination target of the involved pathways (18).
Studies have identified several signaling pathways involved in tumor cell–mediated resistance. For instance, antigen presentation inactivation in tumor cells results from B2M mutation and the loss of heterozygosity in the human leukocyte antigen (HLA) region are observed in some patients with acquired resistance (19–21). Inactivating mutations for genes in the interferon-γ (IFNγ) pathway in tumor cells are shown to dramatically reduce the efficacy of anti–PD-1 therapy (22), and sustained type I interferon signaling contributes to the resistance of PD-1 blockade (23–25). These findings reveal that tumor cells still escape from immunotherapy by reducing their sensitivity to immune recognition and attack. Small-molecule inhibitors or antibodies targeting these pathways may alleviate the resistance and provide new options to enhance the effect of checkpoint blockade therapy (26, 27).
T-cell exhaustion is a major obstacle eliciting antitumor immunity, which is characterized by the stepwise and progressive loss of T-cell function and culminates in the physical deletion of responding cells (28–31). PD-1 blockade induces a durable antitumor immune response by reinvigorating exhausted tumor-specific T cells (32, 33). However, studies find that exhausted T cells in chronic infection display a stable epigenetic profile distinct from effector or memory T cells and is minimally remodeled after PD-1 blockade. The restored T cells are likely to be exhausted and lose function again in the adverse tumor microenvironment (TME), thus leading to acquired resistance to checkpoint blockade (34, 35). Therefore, understanding the mechanisms of dysfunction in these reinvigorated T cells may provide a promising strategy to extend their antitumor activity and enhance the efficacy of cancer immunotherapy. However, very little is known regarding the regulatory pathways of T-cell dysfunction through which resistance to anti–PD-1 therapy occurs (36).
Impaired T-cell receptor (TCR) signaling has been shown to promote exhaustion in T cells (37–39). PD-1–induced exhaustion also starts with the inhibition of TCR signaling, which is mediated by tyrosine phosphatases SHP1/2. TCR signaling determines T-cell fate through a number of downstream pathways regulating cell survival, proliferation, differentiation, and cytokine production. Activated phospholipase C γ1 (PLCγ1) after TCR engagement cleaves phosphatidylinositol 4,5-biphosphate (PIP2) into second messengers, diacylglycerol (DAG) and inositol trisphosphate (IP3; refs. 40–43). Although IP3 triggers the release of Ca2+ from the endoplasmic reticulum (ER), DAG activates the PKCθ and MAPK/ERK pathways. DGK downregulates TCR signaling by phosphorylating DAG into phosphatidic acid (PA; refs. 44–48). In a previous study, we identified Dgka as a candidate gene repressing T-cell proliferation and function in the TME using an in vivo–pooled shRNA screen strategy (49). We and others have also shown that the genetic targeting of Dgka by shRNA or CRISPR/Cas9 can enhance the antitumor activity of T cells (50, 51). These findings suggest that DGK might participate in the exhaustion of T cells by suppressing TCR signaling.
Here, we report that Dgka accelerated the exhaustion and dysfunction of tumor-specific T cells by downregulating the DAG/ERK pathway and promoted tumor cell growth through the activation of the PA/AKT pathway, indicating it is a dual target for cancer treatment. Inhibition of Dgka activity using small molecules extended the antitumor activity of tumor-specific T cells and expanded the therapeutic efficacy of PD-1 blockade. Analysis of colorectal cancer specimens showed that DGKA expression in T cells and tumor cells were both associated with poor prognosis of the disease. Together, these findings provide a potential combination strategy of DGKA blockade with anti–PD-1 for cancer immunotherapy.
Materials and Methods
Cells and mice
B16, LLC1, 4T1, CT26, FHC, DLD1, HT29, HCT116, RKO, SW480, SW620, and HEK 293T cell lines were originally obtained from the ATCC in 2015. B16-OVA and MC38 cells were a gift from Dr. Pan Zheng (University of Maryland, Baltimore, MD) in 2017. All the cells were proved Mycoplasma free and were cultured in Dulbecco's minimum essential medium (DMEM) supplemented with 10% fetal bovine serum (FBS), penicillin (100 μg/mL), streptomycin (100 μg/mL; Invitrogen) and maintained in an incubator with a humidified atmosphere of 5% CO2 at 37°C. All the cells were passaged less than five generations before being used in experiments.
All mice were maintained under specific pathogen–free conditions and in accordance with the animal experimental guidelines of Sun Yat-sen University. All animal procedures were approved by the Institutional Animal Care and Use Committee of Sun Yat-sen University (L102012017001P). C57BL/6, BALB/c, and NSG mice were purchased from Beijing Vital River Laboratory Animal Technology. OT-I TCR transgenic mice were purchased from The Jackson Laboratory and maintained at Guangdong Medical Lab Animal Center.
Stable silencing of DGKA in tumor cells
shRNAs were cloned into the pLKO.1 lentiviral vector (10878, Addgene) with the miR30 backbone downstream a U6 promoter. 5 × 106 HEK 293T cells were seeded on a 10-cm culture plate 1 day prior to transfection in DMEM with 10% FBS. When the confluence reached approximately 70% on the next day, 11 μg shRNA lenti-vector, 10 μg pMD2.G (Addgene, 12259), and 3.5 μg psPAX2 (Addgene, 12260) plasmids were added into 2 mL 0.25 M CaCl2, and then 2 mL 2× Hepes-buffered saline (2× HBSS) was dropped into the CaCl2–DNA mixture and vortexed to get calcium phosphate–DNA complexes. The complexes were incubated at room temperature for 20 minutes and then added into the 293T cell culture. Transfection media were replaced with DMEM supplemented with 20% FBS 6 hours later. The lentivirus supernatant was harvested 48 hours after transfection and filtered by 0.22-μm strainer (Millipore). Tumor cells (1 × 105) seeded in a 6-well plate were transduced with 1 mL lentiviral supernatant with 4 μg/mL polybrene per well and incubated at 37°C for 24 hours, and then replaced with DMEM supplemented with 10% FBS. Stable cell lines were selected by 5 mg/mL puromycin for three days, then maintained with 2 mg/mL puromycin. Five different shRNAs were tested, and studies were performed using the two shRNAs showing the best knockdown of protein expression.
The shRNA sequences (5′–3′):
T-cell culture and stimulation in vitro
C57BL/6 or OT-I TCR transgenic mice (ages 6–10 weeks) were sacrificed, and spleens and mesenteric lymph nodes were collected and grinded on a 70-μm cell strainer (BD Biosciences). Red blood cells were removed using ACK lysis buffer (155 mmol/L ammonium chloride, 10 mmol/L potassium bicarbonate, 1 mmol/L EDTA; Sigma). CD8+ T cells were isolated by a Dynabeads mouse CD8+ negative selection kit (Invitrogen). Cells with high purity (>90%) were washed by PBS and cultured in complete RPMI-1640 with 10% FBS, 20 mmol/L HEPES, 1 mmol/L sodium pyruvate, 0.05 mmol/L 2-mercaptoethanol, 2 mmol/L L-glutamine, 100 μg/mL streptomycin, and 100 μg/mL penicillin (Invitrogen).
For in vitro stimulation, OT-I T cells prepared from OT-I TCR transgenic mice were resuspended at 2 × 106 cells/mL and stimulated with the SIINFEKL (OVA257–264) peptide (GenScript) at 1 μg/mL for 3 days. The cells were used for adoptive cell transfer therapy in the OT-I/B16-OVA model as described below.
For experiments of Dgka inhibition in vitro, CD8+ T cells prepared from C57BL/6 mice were resuspended at 2 × 106 cells/mL and pretreated with different DGKA inhibitors (ritanserin, R59022, R59949, 100 nmol/L; Sigma) or ketanserin (100 nmol/L; Sigma) individually for 24 hours, and then stimulated with 5 μg/mL anti-CD3 and 5 μg/mL anti-CD28 (Invitrogen) for 48 hours. The cells were treated with Brefeldin A (5 μg/mL; eBioscience) for 4 hours before harvest, and intracellular staining was performed using a Fixation/Permeabilization Solution Kit (BD Biosciences) according to the manufacturer's protocol as described below.
Purification of PD-1 antibody from hybridoma
The G4 hybridoma of PD-1 antibody was provided by Dr. Lieping Chen (Yale University, New Haven, CT) in 2016. PD-1 antibody was generated and purified as described previously (52). In brief, hybridoma G4 was grown in RPMI-1640 supplemented with 10% low-IgG FBS (Invitrogen) and 25 mmol/L HEPES (Invitrogen). Supernatant was harvested, concentrated using a tangential flow miniplate concentrator (Millipore Corp), and purified using a 5-mL HiTrap protein G–Sepharose column (Amersham Pharmacia Biotech). Purified monoclonal antibodies were dialyzed against PBS and concentrated using a Centriprep Centrifugal Filter Device (Millipore Corp).
Tumor models of resistance to anti–PD-1 therapy
MC38, LLC1, or B16 tumor cells (2 × 105) were inoculated subcutaneously (day 0) in the inguinal area of C57BL/6 mice (ages 6–8 weeks). On day 12, mice bearing tumors of similar size were randomly divided into two groups and injected intraperitoneally with PD-1 antibody at 200 μg/mouse or PBS control every 3 days, until the experimental endpoint when tumor diameter reached 20 mm. The tumor size was measured every 3 days using a vernier caliper, and the animal survival rate was recorded every day. The mice were sacrificed at the indicated time points, and subcutaneous tumors were isolated. Tumors at early stage during anti–PD-1 therapy were harvested on day 17 in the three tumor models. Late-stage tumors, defined as the regrowth of the tumor, were collected on day 27 for MC38 tumors, day 24 for LLC1 tumors, and day 21 for B16 tumors. The function of CD8+ tumor-infiltrating lymphocytes (TIL) were isolated and analyzed by FACS as described below.
In the OT-I/B16-OVA model, B16-OVA (2 × 105) were inoculated subcutaneously (day 0) in the inguinal area of C57BL/6 mice (ages 6–8 weeks). OT-I T cells expanded as indicated in the T-cell stimulation section were transferred (3 × 106) into mice by intravenous injection on day 12. The mice were randomly divided into 2 groups, then PD-1 antibody was injected intraperitoneally from day 15 at 200 μg/mouse every 3 days until the experimental endpoint, and PBS was used as control. OT-I T cells were sorted from tumors on day 19 (early stage) and day 27 (late stage) for further analysis.
In the sorting experiment, n = 15 to 20 mice were pooled in each group. Lymphocytes were enriched by density gradient centrifugation and stained with BUV737-conjugated anti-CD8α (53–6.7) and FITC-conjugated anti-CD45.1 (A20; BioLegend). Primary and secondary sorting was performed to exclude tumor contamination. The purity of OT-I T cells after the secondary sorting was more than 98%. Finally, about 0.5 million OT-I T cells were used for RNA or protein extraction, respectively, as described below.
In vivo treatment of ritanserin and PD-1 antibody
In the OT-I/B16-OVA model, mice bearing tumors of similar size were adoptively transferred with OT-I T cells (3 × 106) as indicated in the T-cell stimulation section on day 12, and then randomly divided into 4 groups. From day 15, each group was treated with DMSO, ritanserin, PD-1 antibody, or PD-1 antibody plus ritanserin. Ritanserin was injected intraperitoneally at 5 mg/kg every day for 7 successive days, for seven administrations in total. PD-1 antibody was injected intraperitoneally at 200 μg/mouse every 3 days, for two administrations in total. Tumor size was measured every 3 days using a vernier caliper, and the animal survival rate was recorded until the experimental endpoint when tumor diameter reached 20 mm.
In the MC38, LLC1, and B16 models, mice bearing tumors of similar size were randomly divided into 4 groups and were treated with DMSO, ritanserin, PD-1 antibody, or PD-1 antibody plus ritanserin at the same dosages as described above, respectively. Tumor size was measured every 3 days using a vernier caliper, and the animal survival rate was recorded until the experimental endpoint when tumor diameter reached 20 mm.
In the SW480 and CT26 models, SW480, SW480-shLacZ, SW480-shDGKA or CT26, CT26-shLacZ, CT26-shDgka cells (5 × 106) were inoculated subcutaneously into NSG mice (ages 6–8 weeks). CT26, CT26-shLacZ, and CT26-shDgka cells (5 × 106) were also inoculated subcutaneously into BALB/c mice (ages 6–8 weeks). In the ritanserin treatment group, mice were intraperitoneally injected with ritanserin (5 mg/kg) every day for 7 successive days from day 9, and DMSO was used as control. PD-1 antibody was injected intraperitoneally at 200 μg/mouse every 3 days, for two administrations in total. In the shRNA groups, no ritanserin or PD-1 antibody was administrated. Tumor size was measured every 3 days using a vernier caliper, and the animal survival rate was recorded until the experimental endpoint when the tumor diameter reached 20 mm.
T-cell isolation from tumors
In the MC38, LLC1, B16, and B16-OVA models, mice were sacrificed at indicated time points, and tumors were cut into small pieces in petri dishes with 5 mL of PBS and washed once with PBS. Tumors were digested in 15 mL RPMI supplemented with 2% FBS, Collagenase Type IV (50 U/mL; Invitrogen) and DNase (20 U/mL; Roche) at 37°C for 2 hours. The digested tissues were transferred into Miltenyi C tubes and run program “m_tumor_01” on the gentleMACS Dissociator (Miltenyi Biotech). Suspensions were washed three times with cold PBS, and then passed through a 70-μm strainer. 50 mL centrifugation tubes containing 15 mL of mouse percoll (Sigma) were filled with 15 mL cell suspension gently, and centrifuged at 450 × g for 30 minutes. Pipet the nebulous lymphocytes layer after density gradient centrifugation, stained with antibodies, and either analyzed or sorted by flow cytometry as indicated.
T-cell staining and flow cytometry
For membrane staining, T cells were stained with BUV737-conjugated anti-CD8 (53-6.7), FITC-conjugated anti-CD45.1 (A20), PE-Cy7 conjugated anti-Tigit (1G9), and PE-conjugated anti-Tim3 (B8.2C12; BioLegend) at 4°C for 30 minutes, and then washed twice with PBS. For intracellular cytokine analysis, T cells enriched from tumors were restimulated by precoated anti-CD3 and anti-CD28 antibodies with Brefeldin A for 4 hours in vitro. Cells underwent membrane staining first, and then fixed in the fixation solution at room temperature for 1 hour, followed by staining with BV421-conjugated anti-IFNγ (XMG1.2) and APC-conjugated anti-TNFα (MP6-XT22; BioLegend) diluted in permeablization solution at 4°C for 45 minutes. The fluorescence minus one staining was used as controls. Cells were analyzed by BD FACSAria II (BD Biosciences). Data analysis was performed using FlowJo software (FlowJo). Antibodies were purchased from BioLegend.
RNA-sequencing and data analysis
For OT-I T cells, we harvested naïve and in vitro–activated effector (day 3) OT-I T cells, and OT-I T cells from early- or late-stage B16-OVA tumors. For tumor cells, SW480 and CT26 cells stably transduced with shLacZ, shDGKA, or shDgka were collected. Total RNA was extracted using RNeasy columns (Qiagen) and quantified using a Qubit RNA high-sensitivity assay kit (Life Technologies). RNA integrity number was assessed by the Agilent 2100 Bioanalyzer (Agilent) using the LabChip GX RNA Assay Reagent Kit (PerkinElmer). TruSeq Stranded mRNA Library Preparation Kit (Illumina) was used to measure transcript abundance following the manufacturer's protocol. Briefly, 1 μg of total RNA was purified by capturing mRNA with polyA+ before fragmentation and reverse-transcribed to cDNA. The cDNA was then ligated with adaptors and amplified. All final libraries were quantified using KAPA library quantification kits (KAPA Biosystems) on the StepOnePlus Real-Time PCR system (Applied Biosystems) according to the manufacturer's protocol. The quality and average fragment size of the final libraries were also determined using LabChip GX DNA High Sensitivity Reagent Kit (PerkinElmer). Libraries with unique indexes were pooled and sequenced on a NextSeq 500 benchtop sequencer (Illumina) using NextSeq 500 High Output v2 kit and paired-end 150-bp sequencing chemistry.
RNA sequencing (RNA-seq) was performed by Novogene (Hangzhou, China). All data analyses were based on high-quality clean data. Gene-expression values were represented as RNA-seq by Expectation Maximization data normalized within each sample to the upper quartile of total reads. EdgeR provides statistical routines to determine differential expression in digital gene-expression data using a model based on a negative binomial distribution. The Benjamini–Hochberg method was used to control the false discovery rate to adjust the obtained P value. Gene Ontology (GO) enrichment analysis of differentially expressed genes was performed by the cluster Profiler R software package. GO terms with a corrected P value less than 0.05 were considered to be significantly enriched in differentially expressed genes.
Gene set enrichment analysis (GSEA) was performed to analyze involved biological functions and pathways of DGKA. We used the Molecular Signatures Database (MSigDB) H (hallmark gene sets) collection of chemical and genetic perturbations and KEGG subsets of canonical pathways and cancer modules. GSEA results were shown using normalized enrichment scores, accounting for the size and degree to which a gene set is overrepresented at the top or bottom of the ranked list of genes (nominal P value < 0.05 and FDR ≤0.25).
The raw RNA-seq data files of SW480 cells have been deposited in the Genome Sequence Archive for Human (GSA-Human) with the accession number HRA000655 (https://bigd.big.ac.cn/gsa-human/browse/HRA000655). The raw RNA-seq data files of OT-I T cells have been deposited in the Genome Sequence Archive (GSA) with the accession number CRA003876 (https://bigd.big.ac.cn/gsa/browse/CRA003876).
Quantitative real-time PCR
Total RNA was extracted with TRIzol (Invitrogen) from OT-I T cells. RNA (500 ng) was used for reverse transcription to generate cDNA using the two-step reverse transcription kit (Takara). From 100 ng of cDNA template, quantitative real-time PCR (qRT-PCR) analysis for Dgka and Actb was performed by the CFX96 real-time PCR detection system (Bio-Rad) using their specific forward primers and reverse primers according to the manufacturer's protocol (TransGen Biotech). The results were normalized to the housekeeping gene, Actb, and fold changes were analyzed based on ΔΔCt using the CFX manager (Bio-Rad).
Gene-specific primers (5′–3′):
Western blot analysis
For Western blot, anti-DGKA (ab197249) was purchased from Abcam. Anti-GAPDH (#5174), anti–β-Actin (#4970), anti–pAKT-Ser473 (#4070), anti–pAKT-Thr308 (#13038), anti–pERK1/2-Thr202/Tyr204 (#4370), anti-AKT (#4691), anti-ERK1/2 (#4695), anti-CDK4 (#12790), anti-CDK6 (#13331), anti-Cyclin D1 (#2978), anti-Cyclin D2 (#3741), horseradish peroxidase (HRP)–linked anti-mouse IgG (#7076), and anti-rabbit IgG (#7074) were purchased from Cell Signaling Technology.
Cells were lysed in RIPA lysis buffer containing 1 mmol/L phenylmethylsulfonyl fluoride (Beyotime). Lysates were clarified by centrifugation at 10,000 × g for 10 minutes at 4°C. Protein concentration was measured using the BCA Assay (Thermo Scientific) and diluted in sample loading buffer (Life Technologies). Protein (15 μg) was loaded per well and separated by electrophoresis (Bio-Rad) in a 10% sodium dodecyl sulfate-polyacrylamide gel (SDS-PAGE). Protein was electrophoretically transferred to a PVDF membrane (Amersham Pharmacia Biotech), incubated with a blocking solution of 5% BSA at room temperature for 1 hour, and incubated with primary antibodies prepared in 2% BSA at 4°C overnight. On day 2, the PVDF membrane was washed in 0.1% TBST buffer and incubated with HRP-linked secondary antibodies. The protein bands were detected by enhanced chemiluminescence.
Cell viability assay
Cell viability was determined using the CCK8 assay (Roche). Briefly, SW480 or CT26 cells were seeded at 4 × 103 cells/well in 96-well plates. Cells were cultured overnight and then changed to fresh medium containing various concentrations (5, 10, 20, 40, 80 μmol/L) of ritanserin, R59022, R59949, or ketanserin dissolved in DMSO. On day 2, 10 μL CCK8 solution was added into each well; after 4 hours, absorbance at 450 nm was measured by microplate reader (Tecan Infinite 200 Pro). The effect on cell viability was assessed as the percentage of cell viability compared with DMSO-treated control cells.
Cell cycle was measured by flow cytometry. In brief, SW480 cells treated with ritanserin in 6-well plates were collected at 48 hours, washed once with cold PBS, and fixed with precold 70% ethanol overnight in 4°C, then subsequently incubated in a 37°C incubator with propidium iodide (PI; Invitrogen) working solution for 30 minutes shielded from light. Cells were analyzed by BD FACSAria II (BD Biosciences). Data analysis was performed using FlowJo software (FlowJo).
Colorectal cancer samples
This study was approved by the Ethics Committee of Sun Yat-sen University Cancer Center (GZR2017-216). Written informed consent was obtained from patients or their guardians. All samples were obtained at Sun Yat-sen University Cancer Center (Guangzhou, China) between May 1, 2010, and May 1, 2015. A total of 192 patients diagnosed with pT3N0M0 colorectal cancer who underwent total mesorectal excision (TME) were identified. These patients had not previously received any antitumor therapy when diagnosed. The pathologic diagnosis was confirmed by two pathologists independently. Formalin-fixed paraffin-embedded surgery samples were used for IHC.
IHC experiments were performed according to the standard streptavidin–biotin–peroxidase complex method. Tumor samples from colorectal cancer patients were sectioned into 4- to 5-μm-thick slides. Serial sections were used for the detection of DGKA and CD8. The slides were baked in a 60°C oven for 2 hours, and deparaffinized in dimethylbenzene for 10 minutes twice, then rehydrated serially in 100%, 95%, 90%, 80%, and 70% graded ethanol for 10 minutes, respectively, and washed by running distilled water. High-pressure pretreatment in citrate buffer (pH 6.0; Zsbio) was used for antigen retrieval for 2 minutes, and 0.3% H2O2 (Zsbio) was used to block endogenous peroxidase activity at room temperature for 30 minutes. The slides were incubated in blocking solution (5% horse serum, 3% BSA, 0.1% Tween20 in PBS; Zsbio) for 1 hour at room temperature, and then incubated with rabbit anti-human DGKA (ab197249) or anti-CD8 (ab93278; Abcam) at 4°C overnight in a humidified chamber. On day 2, the slides were washed three times with PBS and incubated with HRP-labeled goat anti-rabbit IgG (Zsbio) 1 hour at room temperature and washed with PBS two times, then stained with 3,3′-diaminobenzidine tetrahydrochloride (DAB; Zsbio) and counterstained with hematoxylin (Beyotime).
To evaluate the IHC staining, three independent pathologists were blinded to the clinicopathologic information performed scoring using a light microscope (CKX41, Olympus). Generally, DGKA signal was detected in the cytoplasm. The staining intensity of DGKA was stratified into four classes in colorectal carcinoma tissues, namely, 0, 1, 2, and 3, which were designated as absent, weak, moderate, and strong signals, respectively. The percentage of stained cells was categorized as 0, 1, 2, 3, and 4 to indicate 0% to 25%, 26% to 50%, 51% to 75%, and 76% to 100% stained cells, respectively. The score for each tissue was calculated by multiplying the staining value by the percentage category value, and the average score from the three pathologists was used as the final score. The IHC cutoff for high or low expression of DGKA was determined through the ROC curve analysis. Patients with DGKA scores above the obtained cutoff value were considered to have “high” expression and vice versa.
Analysis of online databases
The association between expression of DGKA and survival in different cohorts of patients was analyzed by multiple online tools. A cohort of colorectal cancer patients with the microarray data from the NCBI Gene Expression Omnibus (GEO) database (GSE17536, n = 107) was analyzed using the online website http://www.genomicscape.com/microarray/survival.php. The RNA-seq data of colorectal cancer patients from The Cancer Genome Atlas (TCGA) database (n = 263) were analyzed using the website http://bioinfo.henu.edu.cn/. Different types of tumors (adrenocortical carcinoma, uveal melanoma, kidney renal clear cell carcinoma, pancreatic ductal carcinoma, gastric cancer, and lung cancer) with RNA-seq data from TCGA database were analyzed using the online websites http://gepia.cancer-pku.cn and http://kmplot.com/analysis/.
All sample sizes were large enough to ensure proper statistical analyses that were performed using GraphPad Prism 7.0 (GraphPad Software) and the SPSS software (standard version 23.0; SPSS). Statistical significance was determined as indicated in the figure legends. P < 0.05 was considered significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001. All t test analyses were two-tailed unpaired t tests. Kaplan–Meier plots with the log-rank test were used to analyze survival data.
T-cell exhaustion is associated with resistance to PD-1 blockade
In chronic infection, T cells reinvigorated by PD-1 blockade eventually become exhausted and lose function if antigen stimulation is persistent (34, 53). We suspected that a similar process might happen in tumors, which then leads to resistance to PD-1 blockade. Three tumor models with diverse immunogenicity, namely, MC38 colon carcinoma, LLC1 lung carcinoma, and B16 melanoma were treated with anti–PD-1 until disease progression. Different responses to anti–PD-1 therapy were observed (Fig. 1A–C). Whereas MC38 tumors showed a long-lasting response to PD-1 blockade, limited effect was observed in B16 tumors, and the growth of treated LLC1 tumors was delayed 4 to 5 days compared with control tumors. All tumors eventually restarted to grow, indicating that resistance to PD-1 blockade was common in various cancer types. We then compared the phenotypes of TILs in tumors at early and late stages of anti–PD-1 treatment. CD8+ T cells from early-stage tumors displayed a phenotype similar to effector T cells, indicating the successful reinvigoration after PD-1 blockade (Fig. 1D–I). However, when resistance was established at late stage, T cells acquired an exhausted phenotype characterized with increased expression of coinhibitory receptors Tigit and Tim3 (Fig. 1D–F; Supplementary Fig. S1A–S1E). Decreased production of effector cytokines IFNγ and TNFα was also observed in these T cells when they were restimulated by CD3/CD28 antibodies in vitro (Fig. 1G–I). These data suggested that T-cell exhaustion contributed to the development of resistance to PD-1 blockade.
Next, we focused on the phenotypes of tumor-specific T cells during anti–PD-1 therapy. We used the OT-I/B16-OVA mouse model in which OT-I TCR transgenic T cells specifically recognize the surrogate tumor antigen ovalbumin (OVA) expressed in B16 melanoma cells (54, 55). Mice bearing B16-OVA melanoma received OT-I T cells by tail-vein injection and were then treated with PD-1 antibody every 3 days until disease progression (Fig. 2A). Similar to our previous report, we found that the adoptive transfer of OT-I T cells only slightly delayed the growth of the B16-OVA tumor compared with PBS controls (Fig. 2B; ref. 49). PD-1 blockade improved the therapeutic activity of these transferred T cells. Nevertheless, most of the tumors restarted to grow two weeks after treatment, indicating the evolution of resistance to anti–PD-1 therapy (Fig. 2B). Similarly, OT-I T cells from late-stage tumors displayed an exhausted phenotype with increased expression of coinhibitory receptors Tigit and Tim3 (Fig. 2C), and reduced production of effector cytokines IFNγ and TNFα (Fig. 2D), although they were reinvigorated by PD-1 blockade at the beginning. These data demonstrated that resistance to anti–PD-1 therapy was accompanied by the exhaustion of tumor-specific T cells.
Expression of Dgka during T-cell exhaustion
To further explore the molecular mechanisms of T-cell exhaustion during anti–PD-1 therapy, we performed RNA-seq on OT-I T cells isolated from B16-OVA tumors at early or late stage during anti–PD-1 therapy (Supplementary Fig. S1F and S1G). Naïve and in vitro–activated effector OT-I T cells were used as controls. The transcriptional profiles of OT-I T cells from early-stage tumors were similar to effector T cells and expressed effector molecules such as Gzmb, Ifng, and Il2, suggesting a successful reinvigoration by anti–PD-1. In contrast, OT-I T cells from late-stage tumors displayed exhausted transcriptional profiles, characterized by high expression of coinhibitory receptors, such as Havcr2, Lag3, and Tigit, and missing expression of effector cytokines, consistent with the flow cytometry findings (Fig. 2E; Supplementary Table S1). Dgka, a negative regulator of TCR signaling, was one of the most highly upregulated transcripts in OT-I T cells from late-stage tumors (Fig. 2E). We confirmed the upregulation of Dgka at mRNA and protein level by qRT-PCR and Western blot, respectively (Fig. 2F). Increased expression of Dgka was also detected in endogenous CD8+ TILs from late-stage B16 tumors treated with anti–PD-1 (Supplementary Fig. S2A and S2B). These data suggested that Dgka might be involved in the process of T-cell dysfunction during anti–PD-1 treatment.
We then evaluated Dgka expression during the primary exhaustion of T cells. Owing to the suppressive TME, adoptively transferred OT-I T cells quickly underwent exhaustion in B16-OVA tumors and lost expression of effector cytokines within a week (Supplementary Fig. S3A). Dgka expression was also upregulated in these exhausted OT-I T cells (Supplementary Fig. S3B and S3C), suggesting that it might promote the dysfunction of T cells in the TME.
Dgka can prevent T-cell activity by inhibiting TCR signaling, as reported in previous studies (56–58). We therefore evaluated its expression during T-cell activation. OT-I T cells were stimulated with the SIINFEKL (OVA257-264) peptide antigen in vitro. Consistent with previous reports (48, 56), Dgka was abundantly expressed in naïve OT-I T cells but rapidly downregulated after activation (Fig. 2G). These findings suggested that Dgka might expedite the exhaustion of T cells by interfering with TCR signaling.
Dgka inhibition promotes TCR signaling
We and others have shown that targeting Dgka using shRNA or CRISPR/Cas9 can enhance the proliferation and function of T cells in tumors. However, these genetic editing approaches are not appropriate to establish combination therapies with anti–PD-1. We then utilized small-molecule inhibitors to explore the role of Dgka in T-cell dysfunction and whether this further alleviates resistance to PD-1 blockade. Currently, there are three available DGKA inhibitors: R59022, R59949, and ritanserin. These compounds share the same lipophilic structure, suggesting that they might display similar selectivity among DGK isozymes (Fig. 3A). R59022 and R59949 inhibit DGKA and moderately attenuate some other DGK isotypes such as DGKE and DGKD, but not DGKZ (59). Ritanserin is also an antagonist of serotonin receptor and has been tested in several phase II clinical trials treating mental disorders (60–62). In order to find the compound with the best efficacy on T cells, we first analyzed the effects of these inhibitors on T-cell activation in vitro. To exclude the possibility that ritanserin functions in T cells through the serotonin receptor, ketanserin, a selective serotonin receptor antagonist, was also evaluated. CD8+ T cells were individually pretreated with inhibitors for 24 hours and then stimulated with anti-CD3/anti-CD28 for 48 hours. Compared with the DMSO control group, all of the three DGKA inhibitors, but not the serotonin receptor antagonist ketanserin, increased expression of IFNγ and TNFα, demonstrating that the inhibition of Dgka promoted T-cell activity. Overall, ritanserin treatment exhibited the highest expression of tested cytokines, indicating its superior effect on Dgka inhibition compared with the other two inhibitors (Fig. 3B). Ritanserin was then used in following experiments.
DGKA negatively regulates TCR signaling by transforming the second messenger DAG into PA (58). We then examined the DAG downstream ERK pathway activity after DGKA inhibition. Mouse CD8+ T cells were pretreated with DMSO or ritanserin for 24 hours, and then stimulated with anti-CD3/anti-CD28. Treatment with ritanserin significantly increased phosphorylated ERK1/2 proteins in T cells during activation, whereas the total ERK1/2 protein did not change (Fig. 3C). These data showed that small-molecule inhibition of Dgka significantly enhanced TCR downstream signaling.
Dgka inhibition reduces T-cell exhaustion and delays anti–PD-1 resistance
We next investigated whether Dgka inhibition could overcome resistance to anti–PD-1 therapy. Mice bearing B16-OVA melanomas of similar sizes were intravenously injected with OT-I T cells on day 12 and then treated with ritanserin, PD-1 antibody, or the combination on day 15, respectively. Anti–PD-1 was administrated every 3 days, and ritanserin was given for 7 successive days. We found that treatment with ritanserin alone showed better therapeutic effect than the DMSO control group (Fig. 3D and E). OT-I T cells displayed effector features, with low expression of coinhibitory receptors Tigit and Tim3 and increased production of effector cytokines IFNγ and TNFα in tumors treated with ritanserin, although not to the extent as the combination therapy with anti–PD-1 (Fig. 3F and G). The best therapeutic effect was observed in the combination group, which displayed significantly prolonged survival time (Fig. 3D–F). Ritanserin plus anti–PD-1 improved the activity of OT-I T cells in tumors and delayed resistance to PD-1 blockade. Two weeks after the first treatment, although reinvigorated OT-I T cells had become exhausted in tumors treated with anti–PD-1 only, OT-I T cells maintained an effector phenotype in the combination group (Fig. 3F and G). Combination therapy substantially increased the number of effector OT-I cells in the tumors (Fig. 3H and I). These data demonstrated that Dgka inhibition reduced T-cell exhaustion and significantly delayed resistance to PD-1 blockade.
Dgka inhibition enhances the efficacy of anti–PD-1 therapy
Next, we assessed the therapeutic effect of ritanserin in B16, LLC1, and MC38 tumors. The models demonstrated different responses, and all developed resistance to PD-1 blockade. Tumors treated with ritanserin exhibited comparable therapeutic effects to anti–PD-1 administration, indicating that both drugs restored the function of endogenous tumor-specific T cells (Fig. 4A–C). The combination therapy generated sustained antitumor responses and significantly prolonged the survival time in all three models (Fig. 4D–F). These results demonstrated that Dgka inhibition could be a promising approach to enhance the therapeutic efficacy of PD-1 blockade.
DGKA inhibition prevents tumor cell growth
Because a previous study shows that tumor cells also express DGKA (63), we then investigated the role of DGKA in cancer cells. Expression of DGKA was detected in a number of human colorectal cancer cell lines and murine tumor cell lines, though it was barely detected in normal enterocytes (Fig. 5A and B). We first evaluated the impact of DGKA inhibitors on tumor cell growth by the CCK8 assay. SW480 cells were treated with the three available DGKA inhibitors (ritanserin, R59022, or R59949), and DMSO and ketanserin were used as controls. Although all the DGKA inhibitors exhibited growth suppression on tumor cells in a concentration-dependent manner, a stronger effect was observed in ritanserin treated cells. No growth inhibition was observed in cells treated with ketanserin, indicating that ritanserin functioned via the inhibition of DGKA rather than the serotonin receptor (Fig. 5C). In order to examine the effect of DGKA inhibition on tumor growth in vivo, we transplanted SW480 cells into immune-deficient NSG mice and treated them with ritanserin. Consistent with the in vitro results, DGKA blockade by ritanserin delayed tumor growth (Fig. 5D).
We also analyzed the effect of DGKA knockdown on tumor cell proliferation by shRNA. SW480 cells were infected with shRNAs targeting DGKA or control LacZ shRNA. The two DGKA shRNAs significantly reduced DGKA protein expression in infected cells (Supplementary Fig. S4A). DGKA knockdown inhibited SW480 cell proliferation in vitro (Supplementary Fig. S4B). Consistent with the in vitro results, DGKA knockdown delayed tumor growth in immune-deficient NSG mice (Fig. 5E). Consistent with the results in human cancer cells, Dgka knockdown exhibited growth suppression in murine CT26 colon carcinoma cells, which expressed high Dgka (Fig. 5F and G). Ritanserin suppressed CT26 tumor cell proliferation in a concentration-dependent manner, and the effect was abolished in Dgka-knockdown CT26 cells (Fig. 5H), demonstrating that ritanserin functioned through Dgka blockade. Dgka knockdown also delayed CT26 tumor growth in immune-deficient NSG mice (Fig. 5I).
Ritanserin targets T cells and tumor cells simultaneously
To further evaluate the contributions of Dgka inhibition in T cells and tumor cells to the therapeutic effect of ritanserin, immune-competent BALB/c mice were used to examine growth of CT26 tumors in vivo. Similar to the results in B16, MC38, and LLC1 tumors, the combination with ritanserin significantly enhanced therapeutic efficacy of anti–PD-1 in tumors infected with control LacZ shRNA (Fig. 5J). In order to evaluate the contribution of T cells, we further used Dgka-knockdown CT26 cells, which excluded the effect of ritanserin on tumor cells. We found that ritanserin treatment delayed the growth of CT26-shDgka tumors and showed a combinatory effect with anti–PD-1, although efficacy was lower than in the control CT26-shLacZ tumors (Fig. 5K), indicating that blockade of Dgka in T cells contributed to the therapeutic effect. Compared with the DMSO control group, CD8+ T cells in the CT26-shDgka tumors showed lower expression of Tigit and Tim3 and increased production of effector cytokines IFNγ and TNFα in the ritanserin-treated group, and the combination group displayed the lowest expression of Tigit and Tim3 and the highest production of IFNγ and TNFα in CD8+ T cells (Supplementary Fig. S5A and S5B), suggesting that blockade of Dgka in T cells enhanced antitumor immune responses. These findings indicated that Dgka inhibition in T cells and tumor cells both contributed to the therapeutic effect of ritanserin.
DGKA blockade inhibits the AKT pathway in tumor cells
In order to explore the underlying mechanisms by which DGKA promoted tumor progression, we performed RNA-seq using SW480 cell lines transfected with DGKA shRNA or control LacZ shRNA. Consistent with the role of DGKA in promoting tumor cell growth, GO category analyses identified cell division as the most differentiated biological process (Fig. 6A). We, therefore, analyzed cell-cycle distribution in tumor cells by PI staining. Substantial growth arrest in the G1–S phase was observed in DGKA-knockdown groups (Fig. 6B and C). GSEA indicated inhibition of the AKT signaling pathway in DGKA shRNA cells (Fig. 6D). The AKT pathway has been reported to closely interact with PA, the product transformed from DAG by DGKA (64, 65). Accordingly, ritanserin decreased phosphorylated AKT proteins in a time-dependent manner, whereas total AKT protein remained unchanged (Fig. 6E). To further elucidate the downstream signaling of AKT, we detected the expression of proteins essential for G1–S transition in the cell cycle. DGKA shRNA or ritanserin downregulated Cyclin D1, Cyclin D2, CDK4, and CDK6 proteins (Fig. 6F and G), which are well-known molecules downstream the AKT pathway (66). We thus elucidated that DGKA inhibition could prevent tumor cell growth by downregulating the AKT/CDK pathway. These data demonstrated that the role of DGKA in promoting tumor growth was dependent on signaling pathways downstream to its phosphorylated product PA.
DGKA expression in TILs or tumor cells correlates with poor survival
After revealing the critical roles of DGKA in T-cell dysfunction and tumor cell growth, we speculated that its expression might promote disease progression in patients with cancer. Only a minority of patients with colorectal cancer have demonstrated significant responses to PD-1 blockade, suggesting the presence of immune resistance (15). We collected samples from 192 pT3N0M0 colorectal cancer patients who underwent TME and examined the expression of DGKA by IHC (Supplementary Tables S2 and S3). DGKA was not only expressed in T cells but was abundantly accumulated in tumor cells of some patients. TILs and tumor cells were then assessed for the expression of DGKA by three independent pathologists who categorized them as high or low (Fig. 7A and B). In concordance with its role in promoting T-cell dysfunction, high expression of DGKA in CD8+ TILs and tumor cells correlated with unfavorable prognosis (Fig. 7C and D). Patients had shortest overall survival time when their TILs and tumor cells simultaneously expressed a high DGKA (Fig. 7E and F). Univariate and multivariate analyses, performed to identify clinical factors associated with prognosis, identified DGKA expression in both TILs and tumor cells as independent prognostic factors of overall survival. The clinicopathologic variables are summarized in Supplementary Tables S2–S4. These results indicated that DGKA promoted the progression of malignancy by affecting both T cells and tumor cells.
We further analyzed data from public databases to evaluate the prognostic effect of DGKA in different cancer types. High expression of DGKA was associated with poor survival in two cohorts of colorectal cancer patients (GSE17536 and TCGA; Supplementary Fig. S6A and S6B). Similar results were found in adrenocortical carcinoma, uveal melanoma, kidney renal clear cell carcinoma, pancreatic ductal carcinoma, gastric cancer, and lung cancer from TCGA (Supplementary Fig. S6C–S6H).
In the present study, we found that the development of acquired resistance to PD-1 blockade was accompanied by the exhaustion and dysfunction of tumor-specific T cells, which was independent of tumor types, demonstrating a pivotal role of T-cell fitness in anticancer immunity. T-cell priming under suboptimal conditions, i.e., suppressive TME or engagement of PD-1 with its ligand PD-L1, generates defective downstream signaling, which directs T cells into a dysfunctional status and exhaustion. We hypothesized that the inhibition of TCR signaling could be involved in the resistance to PD-1 blockade. Indeed, we found that Dgka, a negative regulator of TCR signaling, was highly expressed in exhausted T cells from resistant tumors. The expression of DGKA was also observed in tumor cells and correlated with poor survival in colorectal cancer patients. We clarified that the DGKA-catalyzed conversion of DAG into PA generated distinct outcomes in T cells and tumor cells. Although reduced DAG in T cells exacerbated exhaustion by downregulating TCR signaling, increased PA in tumor cells promoted their growth by activating their downstream AKT pathway. The inhibition of DGKA could thus display dual effects on T-cell invigoration as well as tumor cell suppression. Indeed, small-molecule inhibition of DGKA significantly delayed resistance and enhanced the therapeutic efficacy of PD-1 blockade. Taken together, we demonstrated that Dgka-mediated T-cell dysfunction is an important mechanism of resistance to anti–PD-1 therapy and combinational targeting of DGKA and PD-1 can be a potent therapy for cancer treatment.
In addition to checkpoint blockade, adoptive T-cell transfer (ACT) using TILs, engineered CAR- or TCR-T cells have become another breakthrough advance in cancer treatment. However, a similar process of exhaustion happens in adoptive transferred T cells after infiltration into the suppressive TME, which has become a major challenge of ACT in solid tumor treatment (67). Compared with these genetic approaches, which require complicated manipulation on cells before infusion, pharmacologic inhibition of DGKA using small-molecule inhibitors is more convenient to combine with ACT and might expand their utility in clinics.
Two Dgk isozymes, Dgkz and Dgkz, are dominantly expressed in T cells (68–70). We and others have shown that targeting Dgka and/or Dgkz enhances the anticancer activity of adoptive transferred T cells, indicating that Dgkz is involved in T-cell exhaustion as well (49–51). In this study, increased expression of Dgkz was also observed in exhausted OT-I T cells from tumors resistant to PD-1 treatment. Therefore, simultaneous inhibition of Dgka and Dgkz in T cells might be a better strategy to combine with PD-1 blockade. It is worth identifying DGKZ-specific inhibitors and testing this strategy in the future.
Among the three DGKA inhibitors, ritanserin displayed the optimal effect on both T-cell activation and tumor cell suppression. It is a dual inhibitor of DGKA and serotonin receptor. In several phase II clinical trials treating mental disorders, including schizophrenia, alcoholism, and insomnia, ritanserin has been shown to be safe in humans, although the therapeutic effect on these mental diseases was modest (60–62, 71–73). Therefore, the combination of ritanserin and anti–PD-1 has potential in clinical practice. Nevertheless, questions remain about the possible neurologic effect of ritanserin on patients with cancer. We observed that mice were less active after ritanserin treatment. Therefore, the effects of ritanserin on neurologic health must be further evaluated in patients with cancer. The selectivity of ritanserin among DGK isozymes needs to be clarified as well.
DGKA inhibition produced different effects in T cells and tumor cells. In multicellular organisms, it is common that different types of cells respond differently to the same signal molecule due to the variation of signal transduction pathways. DGKA catalyzes the conversion of DAG into PA. Our data suggested that T cells are more sensitive to DAG inhibition, whereas tumor cells preferentially depend on PA for proliferation. We speculate that T cells and tumor cells might have different contexts of signaling pathways, which lead to diverse sensitivities to the two compounds. It is of interest to clarify the detailed mechanisms of the difference. Although Dgka inhibition postponed the exhaustion of tumor-specific T cells and enhanced the efficacy of anti–PD-1 therapy, most of the treated mice still succumbed to their malignancy, indicating the existence of additional unknown mechanisms of T-cell exhaustion and dysfunction. Further elucidation of these mechanisms is warranted to successfully overcome resistance to PD-1 blockade.
P. Zhou reports a patent for 201911382073.0 pending to Sun Yat-sen University Cancer Center. No disclosures were reported by the other authors.
L. Fu: Data curation, formal analysis, investigation, methodology, writing–original draft, project administration. S. Li: Data curation. W. Xiao: Resources, formal analysis. K. Yu: Formal analysis. S. Li: Methodology. S. Yuan: Methodology. J. Shen: Resources. X. Dong: Formal analysis. Z. Fang: Methodology. J. Zhang: Methodology. S. Chen: Methodology. W. Li: Methodology. H. You: Data curation. X. Xia: Writing–review and editing. T. Kang: Writing–review and editing. J. Tan: Writing–review and editing. G. Chen: Resources. A.-K. Yang: Writing–review and editing. Y. Gao: Resources. P. Zhou: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, methodology, project administration, writing–review and editing.
The authors thank Dr. Shijun Wen for discussion on the structures of DGKA inhibitors, and Dr. Lieping Chen for providing the hybridoma of PD-1 antibody. This study is supported by grants from the National Key Research and Development Program of China (number 2016YFA0500304), the National Natural Science Foundation in China (81773052, 81572806, and 81802853), the Guangzhou Science Technology and Innovation Commission (201607020038), the Science and Technology Projects of Guangdong Province (2016A020215086), the Guangdong Innovative and Entrepreneurial Research Team Program (2016ZT06S638), the Medical Scientific Research Foundation of Guangdong Province (2016111214046444), and the leading talents of Guangdong Province program.
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.