CAMK1D Triggers Immune Resistance of Human Tumor Cells Refractory to Anti–PD-L1 Treatment

Endogenous T-cell responses against tumor antigens occur frequently in a broad variety of cancer types (1–3). Although these T-cell responses correlate to improved patient prognoses (2, 4, 5), they often do not rescue patients from tumor progression. A major reason, lies in the capacity of tumor cells to regulate T-cell activity through expression of the immune-inhibitory ligand PD-L1. The latter stimulates the inhibitory receptor PD-1 expressed on effector T cells and reduces T-cell receptor signaling (6). PD-L1 expression in healthy and tumor tissues can be induced by inflammatory cytokines such as IFNγ by effector T cells (7–9) and serves as a mechanism to prevent autoimmune diseases (10). Consequently, blockade of PD-L1/PD-1 interactions by therapeutic antibodies has resulted in stunning immune rejection of tumors in many patients (11–14). Still, a significant proportion of patients with cancer lack responses to anti–PD-L1/PD-1 therapies (15–17) possibly due to impaired IFNγ responsiveness resulting in reduced PD-L1 expression, severe, and irreversible T-cell exhaustion, or PD-1-induced blockade of T-cell differentiation (18). However, because functional tumor-reactive T cells are found in many patients refractory to anti–PD-L1/PD-1 treatment (3, 5), these mechanisms may only explain immune response resistance in a minor fraction of cases. Additional immune regulatory interactions may impose protection against immune destruction. Several immune-inhibitory receptors such as TIM3 or VISTA, triggered by ligands expressed in tumors, are characterized (19, 20), but immune resistance is likely caused by more than immune regulatory ligands controlling T-cell activity such as tumor cell–intrinsic resistance mechanisms.

Multiple myeloma is a rarely curable B-cell malignancy characterized by the accumulation of malignant plasma cell clones in the bone marrow (21). In multiple myeloma, spontaneous cytotoxic T-cell responses against myeloma-associated antigens occur (1). Immune checkpoint molecules are expressed by myeloma cells and induce tumor-related immune suppression (22–24). PD-L1 is commonly expressed on malignant plasma cells (9) and high expression of PD-L1 associates with disease progression and is upregulated at relapse or in the refractory stage (25). Nevertheless, results of a phase I trial with PD-1–blocking antibodies reported no objective responses amongst the 27 treated patients with multiple myeloma (26). There is thus rationale to assume that other immune checkpoint molecules may play a role in tumor escape mechanisms. Various immunotherapeutic treatments are being tested in multiple myeloma, including antibodies against CD38 (e.g., daratumumab, isatuximab), SLAMF7 (elotuzumab), BCMA-CAR-T–based treatments or BCMA-T-cell bispecific antibodies (27–31).

Here, we performed a systematic search for genes that regulated immune responsiveness in tumor cells, using multiple myeloma as anti–PD-L1/PD-1 treatment unresponsive tumor model (26).

To identify genes that inhibit tumor immune destruction by CTL, we applied a high-throughput (HTP) genetic screen allowing the silencing of a multitude of genes and subsequently assessed tumor lysis by patient-derived marrow-infiltrating lymphocytes (MIL). We identified 90 genes that regulated immune responsiveness after cytotoxic T-cell attack. Among them, calcium/calmodulin-dependent protein kinase 1D (CAMK1D) was coexpressed with PD-L1 and protected against T-cell–induced tumor cell killing in multiple myeloma and other PD-L1–refractory human cancers.

Patients with previously untreated multiple myeloma (n = 332) or monoclonal gammopathy of unknown significance (MGUS; n = 22) at the University Hospitals of Heidelberg and Montpellier as well as 10 healthy normal donors were included in this study, which was approved by the ethics committee (#229/2003 and S-152/2010) after written informed consent. Patients were diagnosed, staged, and response to treatment assessed according to standard criteria (32–34).

Normal bone marrow plasma cells and myeloma cells from the aforementioned patients were purified using anti-CD138 microbeads (Miltenyi Biotec, Germany #130-051-301) from bone marrow aspirates published previously (31, 35). Peripheral CD27⁺ memory B cells (n = 11) were FACS sorted as described previously (36). The human myeloma cell lines U266, RPMI-8226, LP-1, OPM-2, SK-MM-2, AMO-1, JJN-3, NCI-H929, KMS-12-BM, KMS-11, KMS-12-PE, KMS-18, MM1.S, JIM3, KARPAS-620, L363, and ANBL6 were purchased from the German Collection of Microorganisms and Cell Cultures and the ATCC, the XG lines were generated at INSERM U1040 (37). KMM-1 cells were obtained from the National Institutes of Biomedical Innovation, Health, and Nutrition. Cell line identity was regularly assessed by DNA fingerprinting and compared with the initial sample. Cell lines were grown from initial of first passage aliquots on a regular basis. Mycoplasma contamination excluded by PCR-based assays, and EBV infection status by clinical routine PCR-based diagnostics. If not otherwise stated, cell lines used for expression profiling were assessed from initial or early passage aliquots. Polyclonal plasmablastic cells (n = 10) were generated as published previously (35, 38, 39). The human uveal melanoma cell line Mel270 was established, characterized, and provided by Prof. Griewank (University Hospital Essen, Essen, Germany; ref. 40). KMM-1-luc cells were generated after transfection with a pEGFP-luc plasmid (provided by Dr. Rudolf Haase, LMU Munich, Germany) and selected for the G418-resistant gene. Lipofectamine LTX with Plus reagent (Thermo Scientific #15338100) were used as transfection reagents according to the manufacturer's instructions. Transfected cells were selected for 14 days with G418-containing medium (0.6 mg/mL). KMM-1-luc cells were sorted twice for the expression of GFP by flow cytometry (with 87% and 100% purity, respectively) and cultured in the presence of 0.6 mg/mL G418. Cell sorting was conducted in collaboration with the DKFZ sorting core facility, using the FACSARIA II cell sorter (BD Biosciences) and data were analyzed using FlowJo (Tree Star). KMM-1, U266, and Mel270 were cultured under standard conditions in RPMI media supplemented with 10% FCS, 100 U/mL penicillin G, and 100 μg/mL streptomycin at 37°C in a humidified atmosphere under 5% CO₂.

Peripheral blood mononuclear cells (PBMC) were isolated from buffy coats of healthy donors via Biocoll Density Gradient Centrifugation (Biochrome). Briefly, buffy coats were diluted 1:10 in RPMI and added to 50 mL conical centrifuge tubes, containing 15 mL of biocoll solution. Density-gradient centrifugation was performed at 2,000 rpm for 20 minutes at room temperature using low brake. Afterwards, PBMCs were collected, washed twice with RPMI, and frozen in aliquots of 5 × 10⁷ cells per vial using freezing media A-B (1:1; freezing medium A: 60% AB serum + 40% RPMI; freezing medium B: 80% AB serum + 20% DMSO).

MILs were isolated from the bone marrow of a patient with multiple myeloma. Briefly, T cells were isolated from the negative fraction of CD138-sorted bone marrow cells using Untouched Human T cells Dynabeads (Invitrogen #11344D) following manufacturer's instructions. Cells were stained for anti-CD3 [pacific blue anti-human CD3 (Clone OKT3), BioLegend], anti-CD4 [APC/Cy7 mouse anti-human CD4 (clone RPA-T4), BD Biosciences] and anti-CD8 [pacific blue mouse anti-human CD8 (clone RPA-T8), BD Biosciences], tested for HLA-A2 positivity [APC mouse anti-human HLA-A2 Clone BB7.2 (RUO), BD Biosciences] via flow cytometry and subsequently expanded using the rapid expansion protocol as described below.

MIL cultures were ex vivo expanded using a modified version of the Rapid Expansion Protocol (REP; refs. 41, 42). A total of 2 × 10⁶ of freshly isolated MILs were diluted to 6 × 10⁵ cell/mL in CLM supplemented with 3,000 U/mL rHuIL2 (Novartis Pharma). Cells were incubated in 25 cm² tissue culture flask for 48 hours at 37°C and 5% CO₂. An excess of irradiated allogeneic PBMCs from healthy donors were added as “feeder cells” to support the activation and propagation of T cells early during the REP (43). Thus, PBMCs from three different buffy coats (at a ratio of 1:1:1) were irradiated with 60 Gy (Gammacell 1000) and used as feeder cells to support MILs expansion. A total of 2 × 10⁶ MILs were coincubated with 2 × 10⁸ feeder cells (in a ratio 1:100) in 400 mL of MIL expansion medium (CLM/AIM-V) with 30 ng/mL OKT3 antibody (Thermo Scientific) and 3,000 IU/mL IL2 for 5 days in a G-Rex 100 cell culture flask. Afterwards, 250 mL of supernatant was replaced with 150 mL of fresh media and IL2 was replenished to keep the concentration at 3,000 IU/mL. On day 7, MILs were divided into three G-Rex 100 flasks in a final volume of 250 mL medium each and media was again replenished on day 11. On day 14 of the expansion, MILs were counted and frozen in aliquots of 4 × 10⁷ cells/mL in freezing media A (60% AB serum and 40% RPMI1640) and B (80% AB serum and 20% DMSO).

For the generation of flu-specific CD8⁺ T cells (flu TC), PBMCs from HLA-A2 healthy donors were isolated as described above. Total CD8⁺ T cells were sorted from PBMCs by magnetic separation (Miltenyi Biotec, #130-096-495; day 0) according to the manufacturer's instructions and expanded in the presence of A2-matched flu peptide (GILGFVFTL) for 14 days.

The autologous and peptide-loaded CD8-negative fraction was irradiated with 60 Gray (Gy; IBL 437C Blood Irradiator) and used as feeder cells for 1 week. Afterwards, these cells were substituted with irradiated (60 Gy; IBL 437C Blood Irradiator) T2 cells and used as fresh feeder cells. On day 1 and day 8, 100 IU/mL IL2 (Novartis Pharma) and 5 ng/μL IL15 (R&D Systems) were added to the expansion. The percentage of flu-antigen–specific T cells was determined by pentamer staining (GILGFVFTL-APC, ProImmune #P007-0A-E) on day 7 and 14 via flow cytometry analysis according to the manufacturer's instructions. After antigen-specific expansion, flu TCs were sorted by FACS and expanded further for 14 days by using the rapid expansion protocol.

Gene expression was measured using end-point PCR. Briefly, total RNA was isolated from cell pellets using the RNeasy Mini Kit (Qiagen #74106) according to the manufacturer's guidelines. RNA quality and concentration were analyzed using the Scan Drop (AnalytikJena). One microgram of RNA was reverse transcribed to complementary DNA (cDNA) using the QuantiTect Reverse Transcription Kit (Qiagen #205313) according to the manufacturer's protocol. Synthesized cDNA was amplified using conventional PCR. PCR samples were set up in a 25 μL volume using 2× MyTaq HS Red Mix (Bioline #BIO-25044), 500 nmol/L of gene-specific primer mix (Supplementary Table S1), and 100 ng of template cDNA. Water was added to the reaction mix instead of cDNA for contamination controls. The PCR program was set as the following: 95°C for 3 minutes, 35 cycles of three repetitive steps of denaturation (95°C for 30 seconds), annealing (60°C for 30 seconds) and extension (72°C for 30 seconds), and a final step at 72°C for 5 minutes. PCR products were run on a 2% agarose gel in Tris-acetate-EDTA (TAE) buffer (Thermo Fisher Scientific #B49) using a gel electrophoresis system (Thermo Scientific) and DNA bands were visualized using UV light of myECL Imager (Thermo Scientific).

Knockdown efficiency of siRNA sequences was measured by quantitative PCR (qPCR). For qPCR, 10 ng of template cDNA, 2× QuantiFast SYBR Green PCR Mix (Qiagen #204056) and 300 nmol/L of gene-specific primer mix (Supplementary Table S1) was used per 20 μL reaction and each sample was prepared in triplicates. Reactions were run using the QuantStudio 3 (Applied Biosystems). Gene expression was normalized to β-actin and results were shown as fold change. The analysis was performed using comparative C_t method.

Gene expression profiling was performed using U133 2.0 plus arrays (Affymetrix) as published previously (31, 44, 45). Expression data are deposited in ArrayExpress under accession numbers E-MTAB-317.

Transcriptomic gene expression (RNASeqV2, RSEM) and clinical data from all available tumor entities was downloaded from The Cancer Genome Atlas [TCGA; using getTCGA function of TCGA2STAT (version 1.2) package for R; ref. 46]. Log₂-normalized expression values for uveal melanoma (TCGA-UVM, 80 patients), ovarian cancer (TCGA-OV, 303 patients), stomach adenocarcinoma (TCGA-STAD), and stomach and esophageal carcinoma (TCGA-STES, 599 patients) were correlated (Pearson r) using the ggpubr package for R. Survival curves were generated using survminer package for R. FAS expression was cut at the median to generate Fas-high and -low sets. Similarly, CAMK1D expression was cut at the median for the Kaplan–Meier survival curves. Significance was calculated using the log-rank test.

Gene knockdown in tumor cells was induced using reverse siRNA transfection with Lipofectamine RNAiMAX (Thermo Scientific #13778-150). Briefly, 200 μL of 250 nmol/L siRNA solution (Supplementary Table S1) was added to each well of a 6-well plate. Four microliters of RNAiMAX transfection reagent was diluted in 196 μL of RPMI (Sigma-Aldrich) and incubated for 10 minutes at room temperature. Four-hundred microliters of additional RPMI was added and 600 μL of RNAiMAX mix was given to the siRNA-coated wells and incubated for 30 minutes at room temperature. A total of 3.5 × 10⁵ KMM-1 (WT or luc) cells were resuspended in 1.2 mL of antibiotic-free RPMI culture medium supplemented with 10% FCS, seeded in the siRNA-RNAiMAX–containing wells, and incubated for 48 hours at 37°C, 5% CO₂. Final siRNA concentration was 25 nmol/L in all cases.

To isolate phosphorylated proteins from cells, tumor cells were pelleted at 0.5 × g for 5 minutes and washed once with PBS at 4°C. The cell pellets were lysed with one pellet volume of Phosphoplex Lysis Buffer (Merck Millipore #43-040) containing protease inhibitor cocktail (Cabliochem #539134, 1:100) and phosphatase inhibitor cocktail (Sigma-Aldrich #P0044, 1:100) at 4°C for 15 minutes on a rotator. Samples were centrifuged at 17,000 × g at 4°C for 15 minutes. Supernatants containing the protein lysates were collected into fresh tubes and quantified using the Pierce BCA Protein Assay Kit (Thermo Scientific #23225) according to the manufacturer's protocol.

Briefly, supernatants were diluted 1:5 in water and pipetted together with BCA standards into a 96-well plate. BCA solution A and B were mixed 50:1 and 200 μL of this mix was added to each well. After 30-minute incubation at 37°C, the absorbance at 562 nm was measured with the TECAN reader and the protein concentration of the samples was calculated using the standard curve. Proteins were stored at −20°C.

Thirty micrograms of protein lysates were denaturated in 4× NuPAGE LDS Sample Buffer (Thermo Scientific #NP0008) containing 10% ß-mercaptoethanol (Sigma-Aldrich #M6250-100ML) at 70°C for 10 minutes. Samples were spun down and separated on NuPAGE 4%–12% Bis-Tris Gels (Thermo Scientific NP0321BOX) along with PageRuler Prestained Protein Ladder (Thermo Scientific #26616) and run at 115–150 V for 90 minutes using 1× MES running buffer (Life Technologies #NP0002).

Proteins were transferred from the gel to a polyvinylidene difluoride (PVDF) membrane (Millipore #ISEQ00010) using a semi-dry Western blot method. Briefly, the PVDF blotting membrane was activated in 100% methanol (Millipore) for 1 minutes and afterwards placed in Pierce 1-Step Transfer Buffer (Thermo Science #84731 × 5) until use. Blots were assembled from anode to cathode into the Pierce Power Blot cassette (Thermo Scientific) and run at 24 V for 10 minutes. Membranes were washed in 1× TBS and then placed in blocking solution (5% BSA/0.05% TBST) for 2 hours. Primary antibodies (anti-CAMK1D (Abcam #ab172618) 1:20,000, anti–caspase-3 (Abcam #ab32351) 1:750, anti–caspase-6 (Abcam #ab108335) 1:2,000, anti–caspase-7 (Thermo Scientific MA5-15159) 1:1,000, anti–caspase-3 (phospho S150; Abcam #ab59425) 1:850, anti–caspase-6 (phospho S257; Abcam #ab135543) 1:250, and sodium potassium ATPase (Abcam #ab76020; 1:20,000) were diluted in 5% BSA/0.05% TBST and kept on the membrane overnight at 4°C on a rotator. Membranes were then washed three times for 10 minutes with 1 % BSA/0.05% TBST. Afterwards, horseradish peroxidase (HRP)–conjugated secondary antibodies [anti-rabbit 1:4,000, Cell Signaling Technology (#7074) or anti-mouse 1:4,000, Cell Signaling Technology (#7076)] were added to 1% BSA/TBST and kept on the membrane at room temperature for 1 hour on a shaker. Thereafter, the membranes were washed for 10 minutes with 1% BSA/TBST, then TBST and finally with TBS. The blots were incubated with the Amersham ECL Western Blotting Detection Reagent (Reagent A and Reagent B, 1:1, GE Healthcare # RPN2109) for 4 minutes and the chemiluminescence was detected with myECL Imager (Thermo Scientific). Images were analyzed using the ImageJ software (Wayane Rasband).

For detection of direct protein–protein interaction, coimmunoprecipitation was performed. Briefly, 10 million tumor cells (10 × 10⁶) were seeded in 10 cm² petri dishes. The next day, cells were stimulated for 4 hours with 100 ng/mL rHuFasL (BioLegend #589404). Unstimulated cells were used as negative control. Afterwards, tumor cells were detached, resuspended in ice-cold TBS, and centrifuged at 400 × g for 6 minutes at 4°C. Supernatant was discarded, cell pellet was resuspended in 1.5 mL TBS, and centrifuged at 500 × g for 8 minutes at 4°C. Cell pellet was lysed with 1.5 mL lysis buffer (50 mmol/L Tris-HCl, 150 mmol/L NaCl, 0.5% NP40, or Triton-X) containing protease inhibitor (Millipore #539134-1 ML) and kept on a rotator for 1 hour at 4°C. Afterwards, cells were centrifuged for 20 minutes at 20,000 × g at 4°C. Supernatant was collected and centrifuged for further 5 minutes at 20,000 × g at 4°C. Protein-G agarose (Sigma-Aldrich) was washed with 1 mL TBS and centrifuged for 1 minute at 12,000 × g. One milliliter of cell supernatant containing cytoplasmatic proteins was added to 60 μL protein-G agarose (Sigma-Aldrich #11719416001), incubated with caspase-3 (1:50; Cell Signaling Technology, #9662), caspase-6 (1:50; Abcam #ab108335) or caspase-7 (1:100; Cell Signaling Technology, #9491) antibodies and incubated overnight on a rotator at 4°C. Ninety microliters of cell lysates were frozen at −20°C. The next day, the immunoprecipitated samples were centrifuged at 12,000 × g at 4°C for 1 minute. Supernatant was discarded and protein-G agarose was washed three times with lyses buffer and centrifuged at 12,000 × g at 4°C for 1 minute. 2× LDS containing 10% β-mercaptoethanol was added to the immunoprecipitated samples, while 4× LDS containing 10% β-mercaptoethanol was added to the lysates. Samples were denaturated for 10 minutes at 95°C on a thermocycler. Samples were spun down and separated on NuPAGE 4%–12% Bis-Tris Gels (Thermo Scientific #NP0335BOX) along with PageRuler Prestained Protein Ladder (Thermo Scientific #26616) and run at 115–150 V for 90 minutes. After electrophoresis, proteins were transferred on a PVDF membrane (Millipore). CAMK1D antibody (1:10,000) was diluted in 5% BSA/0.05% TBST and kept on the membrane overnight at 4°C on a rotator. Membranes were then washed three times for 10 minutes with 1% BSA/0.05% TBST. Afterwards, HRP-conjugated secondary antibodies (anti-rabbit 1:3,000) was added to 1% BSA/TBST and kept on the membrane at room temperature for 1 hour on a shaker. The membrane was washed. The blot was incubated with the ECL Detection Reagent (Reagent A and Reagent B, 1:1, GE Healthcare) for 4 minutes and the chemiluminescence was detected with myECL Imager (Thermo Scientific). Images were analyzed using the ImageJ software (Wayane Rasband).

To generate KMM-1-luc cells, 3.5 × 10⁵ KMM-1 WT cells were seeded in a 6-well plate and incubated at 37°C overnight. Fifteen microliters Lipofectamine LTX reagent were diluted in 150 μL Opti-MEM medium (Gibco). Simultaneously, 3.5 μg of pEGFP-Luc plasmid was diluted in 175 μL Opti-MEM medium and 3.5 μL of Plus Reagent was added. 150 μL of diluted DNA was added to 150 μL diluted Lipofectamine LTX (Life Technologies) reagent and incubated for 5 minutes at room temperature. DNA–lipid complex was then added to the growth medium of the myeloma cells. Cells were incubated at 37°C for 48 hours before investigation of transfection efficacy by flow cytometry.

KMM-1-luc cells were reverse transfected with the desired siRNA sequences (Supplementary Table S1) in white 96-well plate (Perkin Elmer) and incubated for 48 hours at 37°C, 5% CO₂. On the same day of transfection, MILs were thawed and treated with benzonase (100 IU/mL; Merck). Cell density was adjusted to 0.6 × 10⁶ cells/mL in CLM supplemented with 3,000 IU/mL rhuIL2 (Novartis) for 48 hours. IL2 was depleted 24 hours before the coculture. Briefly, cells were collected, centrifuged at 1,400 rpm for 10 minutes, and resuspended in CLM at a concentration of 0.6 × 10⁶ cells/mL. Flu TC were thawed 6 hours before coculture. For the cytotoxicity assays, MILs, flu TC, the supernatant of activated MILs, or rHuFasL were added to transfected tumor cells at desired effector-to-target (E:T) ratio/concentration, and incubated for 20 hours at 37°C, 5% CO₂. For the viability setting, only CLM was added to the tumor cells. After coculture, supernatant was removed, remaining tumor cells were lysed using 40 μL/well of cell lysis buffer for 10 minutes. After tumor cell lysis, 60 μL/well of luciferase assay buffer was added and luciferase intensity was measured by using the Spark 20M plate reader (Tecan) with a counting time of 100 msec. Luciferase activities (relative luminescence units = RLUs) were either represented as raw luciferase values or as normalized data to scramble or unstimulated controls.

Target genes in KMM-1 or U266 tumor cells were knocked down with reverse siRNA transfection for 48 hours. The reverse siRNA transfection was performed using transparent 96-well microplates (TPP). In parallel, MILs were thawed and prepared as described previously in section MILs expansion. After 48 hours, MILs (E:T 10:1) or rHuFasL (100 ng/mL) were added to the target cells in CLM with YOYO-1 (final concentration 1:5,000) and cocultured at 37°C. For viability controls, the according amount of CLM with YOYO-1 (final concentration 1:5,000) was added. MILs or rHuFasL-mediated tumor lysis was imaged on the green channel using an IncuCyte ZOOM Live Cell Imager (ESSEN BioScience) for the indicated time points at a 10× magnification. Data were analyzed with the Incucyte ZOOM 2016A software by creating a top-hat filter-based mask for the calculation of the area of YOYO-1–incorporating cells (indicating dead cells).

Tumor cells were transfected with the indicated siRNAs (Supplementary Table S1) in a 96-well plate. Afterwards, T cells were added at the indicated E:T ratio for 20 hours and 100 μL of supernatants were harvested for the detection of IFNγ (Human IFNγ ELISA Set; BD OptEIA #555142), IL2 (Human IL2 ELISA Set; BD OptEIA #555190), Granzyme B (Human Granzyme B ELISA development kit; Mabtech #3485-1H-20), and TNF (Human TNF ELISA Set; BD OptEIA #555212). Experiments were performed according to the manufacturer's instructions. Polyclonal stimulation (Dynabeads Human T-Activator CD3/CD28, Invitrogen #11131D) for 20 hours was used as positive control. Absorbance was measured at λ = 450 nm, taking λ = 570 nm as reference wavelength using the Spark microplate reader (TECAN).

Flow cytometry was used for the detection of proteins expressed on the plasma membrane of tumor and T cells. Intracellular staining was performed for the detection of caspase-3 (FITC Active Caspase-3 Apoptosis Kit, BD Biosciences, #550480) according to manufacturer's instructions. Tumor cells were detached from plates using PBS-EDTA, centrifuged at 500 × g for 5 minutes, and resuspended in FACS buffer (5 × 10⁵ cells/tube). Live T cells and tumor cells were distinguished by using Live/Dead Fixable Yellow Dead Cell Stain (Thermo Scientific #L34959) according to manufacturer's instructions followed by blocking with kiovig (human plasma-derived immunoglobulin, Baxter) at a concentration of 100 μg/mL in FACS buffer (PBS, 2% FCS) for 15 minutes in the dark on ice. Samples were washed two times in FACS buffer and incubated with either fluorophore-conjugated primary antibodies or isotype control [APC anti-human CD274 (PD-L1; clone 29E.2A3), BioLegend #329707; Alexa Fluor 647 Mouse anti-human CCR9 (clone 112509 (RUO), BD Biosciences #557975; Brilliant Violet 421 anti-human CD95 (Fas; clone DX2), BioLegend #305623; PE anti-human CD95 (Fas; clone DX2), BD Biosciences #555674; APC anti-human CD261 (DR4, TRAIL-R1; clone DJR1), BioLegend #307207; PE anti-human CD262 (DR5, TRAIL-R2; clone DJR2), BioLegend #307405; Biotin anti-human CD120a (TNFR1; clone W15099A), BioLegend #369908; PE/Cy7 anti-human CD120b (TNFR2; clone 3G7A02), BioLegend #358411; PE/Cy7 anti-human CD279 (PD-1) antibody, BioLegend #329918; APC mouse anti-human CD178 (clone NOK-1), BD Biosciences #564262; PE anti-human CD253 (TRAIL; clone RIK2), BioLegend #308206; APC anti-human TNFα (clone Mab11), BioLegend #502912] for 20 minutes on ice in the dark. Afterwards, cells were washed twice, they were acquired with the FACSCanto II cell analyzer machine (BD Biosciences) or FACSLyrics Flow cytometer and data were analyzed using FlowJo (Tree Star).

KMM-1 cells grown on coverslips were washed with Ringer solution (118 mmol/L NaCl, 5 mmol/L KCl, 1.2 mmol/L MgCl₂, 1.2 mmol/L Na₂HPO₄, 2 mmol/L NaH₂PO₄, 1.8 mmol/L CaCl₂, 5 mmol/L glucose, 9.1 mmol/L HEPES, pH 7.4, with NaOH) and loaded with Fura-2-AM ester (Thermo Fisher Scientific) for 45 minutes. After 15 minutes, MILs or rHuFasL (50 ng/mL) was added to scr siRNA–transfected cells and recording of the intracellular free Ca²⁺ was continued for further 30 minutes. Experiments were performed using a ZEISS live cell imaging setup based on an inverse microscope (Axio Observer Z.1) equipped with Fluor 40×/1.3 objective lens (ZEISS). Fura 2-AM–loaded KMM-1 cells were illuminated with light of 340 nm or 380 nm (BP 340/30 HE, BP 387/15 HE) using a fast wavelength switching and excitation device (Lambda DG-4, Sutter Instrument), and fluorescence was detected at 510 nm (BP 510/90 HE and FT 409) using an AxioCam MRm LCD camera (ZEISS). Data were recorded and analyzed with ZEN 2012 software (ZEISS).

For the generation of the supernatant of polyclonally activated MILs, 1 × 10⁶ MILs were suspended in 1 mL of CLM collected in a 15 mL tube and stimulated with 25 μL of Dynabeads Human T-Activator CD3/CD28 (Thermo Scientific) for 20 hours. Afterwards, only the supernatant (100 μL/well) of activated T cells was added to knocked down tumor cells and incubated overnight at 37°C, 5% CO₂. Luciferase-based cytotoxicity assay was performed as described above. Alternatively, MILs were stimulated with tumor cells at an E:T ratio of 10:1. After 20-hour coculture, plates were centrifuged at 450 × g for 5 minutes and 100 μL/well of the supernatant was collected for cytokines detection (ELISA).

For the functional neutralization experiment, anti-FasL (clone NOK-1, BioLegend #306409) or isotype control (Clone MOPC-21, BioLegend #400153) were preincubated with MILs for 1 hour at 37°C, 5% CO₂. As negative control, antibodies were cultivated in the absence of T cells. Afterwards, antibody-containing supernatants were used to stimulate KMM-1-luc cells, which were reverse transfected with the indicated siRNAs (Supplementary Table S1). The final concentration of the neutralizing antibodies was 100 ng/mL for anti-FasL and isotype control. As positive control recombinant FasL protein (100 ng/mL, BioLegend #589404) was added to the tumor cells instead of T cells. Twenty hours after coculture, luciferase intensity was measured as described above.

For the experiments using the anti-Calmodulin (W-7 hydrochloride; Tocris #0369) inhibitor, 1 × 10⁴ KMM-1-luc (scr or CAMK1D-transfected) cells/well were seeded in white 96-well plates (Perkin Elmer) in 100 μL of RPMI 10% FCS. The small-molecule inhibitor was added at the indicated concentrations for 1 hour at 37°C, before 100 ng/mL rHuFasL or medium control was added. DMSO treatment served as negative control. After 20-hour stimulation, luciferase-based cytotoxicity assay was performed. For CAMK1D inhibition, 1 × 10⁴ KMM-1-luc or 1 × 10⁴ Mel270 cells/well were incubated overnight in a 96-well plate. QPP-A inhibitor (Merck Millipore; CAS 404828-08-6) was added at the indicated concentrations 1 hour before rHuFasL stimulation (100 ng/mL) or medium control. DMSO treatment served as negative control. After 20-hour stimulation, luciferase-based cytotoxicity assay was performed.

Tumor cells were stimulated with rHuFasL (100 ng/mL) for 15 minutes, 30 minutes, 1, 2, 4, and 8 hours. Unstimulated cells served as controls. For the detection of intracellular phosphorylated analytes, a general pathway (MILLIPLEX MAP Multi-Pathway Magnetic Bead 9-Plex kit, Merck Millipore #48-680MAG) was used following manufacturer's instructions. For the detection of proteins involved in the activation of apoptosis the MILLIPLEX MAP Early Phase Apoptosis 7-plex-kit (Merck Millipore #48-669MAG) together with active caspase-3 Magnetic Bead MAPmate (Merck Millipore #46-604MAG) were used following manufacturer's instructions. Beads specific for GAPDH (MILLIPLEX MAP GAPDH Total Magnetic Bead MAPmate; Merck Millipore #46-667MAG) served as normalization control. Twenty micrograms of protein lysates were used for the detection of ERK/MAP kinase 1/2 (Thr185/Tyr187), Akt (Ser473), STAT3 (Ser727), JNK (Thr183/Tyr185), p70 S6 kinase (Thr412), NFκB (Ser536), STAT5A/B (Tyr694/699), CREB (Ser133), and p38 (Thr180/Tyr182) phosphorylated Akt (Ser473), JNK (Thr183/Tyr185), Bad (Ser112), Bcl-2 (Ser70), p53 (Ser46), cleaved caspase-8 (Asp384), cleaved caspase-9 (Asp315), and active caspase-3 (Asp175). The assay was performed according to the manufacturer's instructions and samples were measured using the MAGPIX Luminex instrument (Merck Millipore).

The primary RNAi screening was conducted using a sublibrary of the genome-wide siRNA library siGENOME (Dhamacon, GE Healthcare), which comprised 2,887 genes (1,288 genes for GPCR/kinase and 1,599 genes for custom library). The library was prepared in Prof. Boutros's group (DKFZ) as described in ref. 47. Each well contained a pool of four nonoverlapping siRNAs (SMARTpool) targeting the same gene. This arrayed screening approach was performed in duplicates and was adopted from Khandelwal and colleagues (48). The siRNA sequences of the genome-wide library were distributed in the 384-well plates and positive and negative siRNA controls were added in empty wells. The final concentration of all siRNA sequences was 25 nmol/L. Reverse transfection was performed as described in section reverse siRNA transfection. The readout was performed using Mithras LB 940 microplate reader with a counting time of 100 msec. 40 × 384-well plates were subjected to the luciferase-based screening assay performed on KMM-1-luc cells. 20 × 384-well plates were subjected to the luciferase-independent CellTiter-Glo (CTG) screening performed on luciferase-negative KMM-1 cells without the addition of MILs to exclude genes affecting cell viability. Briefly, for the readout, supernatant was removed in each well containing siRNA-transfected tumor cells, and 20 μL of the CTG reagent (prediluted 1:4 in RPMI) were added. After 15-minute incubation in the dark, plates were measured using the Mithras reader as described above.

For the screening analysis, the raw RLUs from the primary screening were processed using the cellHTS2 package in R/Bioconductor (49). Values from both conditions were quantile normalized against each other using the aroma.light package in R. Pearson correlation (r²) between the two replicate values was calculated for each setting. Differential scores (cytotoxicity vs. viability) were calculated using the LOESS local regression method. To identify candidate hits, the following thresholds were applied on the z-scores of the samples: for the viability setting, genes showing a z > +2.0 or z <− 2.0 were excluded. For the cytotoxicity setting, CCR9 was used as threshold score. In addition, genes having a z-score >+ 0.5 or <− 0.5 in the CTG-based viability screening were filtered out from the candidate list.

For the secondary screening, a customized library (G-CUSTOM-223794) containing 128 genes from the primary screening was distributed in several 96-well plates along with positive and negative siRNA controls. Reverse transfection was performed. For the cytotoxicity setting MILs (10:1 ratio) or supernatant of anti-CD3/anti-CD28 magnetic beads activated MILs were added to knockdown tumor cells (1 × 10⁴ cells/well). Instead, CLM medium was added to the viability plates. After 20 hours, luciferase-based readout was performed. RLUs were normalized to Mock control. Cytotoxicity/viability ratios were calculated according to the formula:

Cytotoxicity/viability ratio = (Norm. RLU cytotoxicity setting/Norm. RLU viability setting).

The hit-list was generated by including only hits with improved T-cell–mediated cytotoxicity over mock transfection, (Cytotoxicity/viability ratio < 1). Pearson's correlation was calculated with Microsoft Excel.

Camk1d knockout MC38 murine colorectal cells were generated using the CRISPR/Cas9 technique. In vivo experiments were performed in two cohorts of mice: C57BL6 (n = 12) and NOD/SCID gamma chain (NSG) mice (n = 12) were subcutaneously injected with 1 × 10⁵ MC38 Camk1d KO (g3 clone 11) or 1 × 10⁵ MC38 NTS (clone 12) cells each into the right and left flank of one mouse, respectively. Tumor growth was measured twice a week with a caliper and the volume was determined using the following formula: Tumor volume (mm³) = (Width² × Length) × (π/6). Mice were sacrificed when tumors exceeded 1.5 cm in diameter.

For statistical analysis, GraphPad Prism software v6.0 (GraphPad Software) was used. If not stated, statistical differences between the control and the test groups were determined by using two-tailed unpaired Student t test. In all statistical tests, a P value ≤0.05 was considered significant with *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; and ****, P ≤ 0.0001.

To identify novel genes involved in immune escape mechanisms of PD-L1 unresponsive cancer cells, an HTP screening approach (48) was adapted. The HLA-A2–positive human multiple myeloma cell line KMM-1, expressing high levels of PD-L1 and lower levels of CCR9 (48), was used as a tumor model in this study. As a reporter system for tumor cell survival, we stably transfected KMM-1 cells with e-GFP-firefly luciferase, allowing for luminescence imaging to detect immune-mediated tumor cell destruction in an HTP format (Fig. 1A).

As a source of tumor-reactive T cells, we used marrow-infiltrating, PD-1–positive T cells (MILs) from an HLA-A2–matched patient (Fig. 1A; Supplementary Fig. S1A). These MILs were not terminally exhausted as they displayed strong IFNγ secretion after polyclonal stimulation (Fig. 1B), which exceeded that of tumor antigen–specific CD8⁺ cytotoxic Survivin T cells (48). MILs recognized KMM-1 tumor cells, despite high PD-L1 expression (Fig. 1B). However, their limited killing capacity (Supplementary Fig. S1B) suggested the presence of resistance mechanisms against T-cell–induced death.

Silencing of firefly-luciferase (siFLuc), ubiquitin C (siUBC), a gene essential for cell survival, or transfection with a mixture of siRNAs inducing cell death (siCD) resulted in strong reduction of luciferase expression, indicating appropriate gene silencing and sensitivity of the luciferase-based readout. This was also maintained upon coculture of siRNA-treated KMM-1 cells with MILs (Fig. 1C).

We next studied the effect of PD-L1 and CCR9 on KMM-1 cells (Fig. 1D). The knockdown of PD-L1 did not result in increased KMM-1 killing by MILs, despite high expression of PD-L1 on the tumors and of PD-1 on the MILs (Fig. 1E). In contrast, CCR9 silencing significantly improved tumor cell rejection (Fig. 1E), suggesting that PD-L1 did not play a decisive role in immune resistance of KMM-1 cells. We therefore used CCR9 as a positive control within the screen.

To this end, KMM-1 cells were transfected in a multiwell format with a siRNA library consisting of a pool of four nonoverlapping siRNAs per target per well, targeting a total of 2,887 genes (Supplementary Table S2) covering a broad spectrum of all gene families. The screening approach comprised a viability setup, in which we assessed the intrinsic viability effect of each gene knockdown, and a cytotoxicity setup, where siRNA-transfected tumor cells were cocultured with MILs (Supplementary Fig. S2).

Negative (scramble siRNA sequences, scr1 and scr2) and positive controls (siRNAs targeting luciferase and essential viability genes) were included as a reference to calculate the effect of gene knockdown on cell viability. Overall, the distribution of values across test replicates and setups was highly concordant showing no viability or cytotoxicity effect of scr siRNAs but robust signal reduction after FLuc and UBC knockdown (Fig. 2A). Calculated z-scores for the impact on cell viability and T-cell cytotoxicity (Fig. 2B) and for the relative impact on T-cell–mediated tumor cell lysis (LOESS score) of each gene, revealed 128 genes whose silencing improved tumor cell lysis by T cells to a higher degree than CCR9 (Fig. 2C). Among them, we found several genes with described immune regulatory function in multiple myeloma such as CD5 (50), FES (51), and PAK3 (52). PD-L1 did not have any effect on T-cell–mediated killing of multiple myeloma cells (Fig. 2B and C). The identification of these validated immune checkpoints in combination with good immune checkpoint control performance supported the robustness and sensitivity of the screen.

For further validation, we subjected the 128 candidate hits to a secondary screening procedure using the same setup as for the HTP screen. Silencing of 90 candidates again increased T-cell–mediated killing of tumor cells, and only had little effect on intrinsic tumor cell viability, thus confirming their immune regulatory role in KMM-1 cells (Fig. 2D). The highest immune modulatory effect was elicited by the serine/threonine calcium/calmodulin-dependent protein kinase 1D (CAMK1D; Fig. 2B–D).

To determine whether the observed tumor cell killing was mediated by cytokines or other soluble proteins released by activated MILs, an additional setting was included. MILs were polyclonally stimulated with anti-CD3/anti-CD28 magnetic beads and only their cell culture supernatant was added to the tumor cells. In this setup, silencing of few genes had an impact on tumor cell lysis indicating a role in resistance toward T-cell–secreted cytotoxic cytokines (Fig. 2E). Most of the identified candidate genes, including CAMK1D, regulated tumor cell killing only upon direct interaction with T cells. Taken together, these results provided an indication that multiple myeloma cells express multiple immune regulatory genes, among them CAMK1D, that confer immune resistance after T-cell engagement.

On the basis of the immune resistance phenotype associated with CAMK1D expression in our screens, we validated and characterized the immune regulatory role of CAMK1D. We first deconvoluted the pool of siCAMK1D to exclude potential dominant off-target effects of single siRNAs within the pool. Three of four siRNAs (s1, s2, and s3) and the pool of all siRNAs increased T-cell–mediated cytotoxicity, whereas no viability impact was detected (Fig. 3A). All siRNAs significantly reduced CAMK1D mRNA and protein expression (Fig. 3B and C). In a luciferase-independent assay, employing live-cell imaging, we confirmed an increase MIL-induced apoptosis of CAMK1D-deficient KMM-1 cells (Fig. 3D). This could be inhibited with MHC-I–blocking antibodies, indicating that tumor cell apoptosis was induced by MHC-I–restricted CD8⁺ MILs in a T-cell receptor–dependent manner (Fig. 3E). To corroborate this finding, we pulsed KMM-1 cells with an HLA-A2–restricted influenza (flu) peptide and cocultured them with PD-1–positive, flu peptide–specific CD8⁺ cytotoxic T cells (flu TC; Supplementary Fig. S3A). Again, siCAMK1D, but not PD-L1 silencing, resulted in a significant increase of T-cell–mediated tumor cell lysis (Fig. 3F; Supplementary Fig. S3B). These data demonstrated that CAMK1D-mediated resistance of KMM-1 cells toward antigen-specific T cells, independent of the T-cell source. CAMK1D-mediated immune protection also occurred in the PD-L1⁺, HLA-A2⁺ multiple myeloma cell line, U266 (Fig. 3G–I; Supplementary Fig. S3C). We therefore studied CAMK1D expression in a large cohort of CD138-purfied malignant plasma cells from multiple myeloma patients with monoclonal gammopathy of unknown significance (MGUS), human myeloma cell lines (HMCL), memory B cells (MBC), plasmablasts (PPC), and normal bone marrow plasma cells (BMPC). CAMK1D expression was highest in MBC, but was also expressed in all multiple myeloma, MGUS, PPC, and in 30 of 32 HMCL samples, with higher expression than normal bone marrow plasma cells (BMPC; Fig. 3J). Thus, these data indicated that CAMK1D was consistently expressed in human multiple myelomas and conferred resistance against cytotoxic T-cell attack.

As classical immune checkpoint molecules expressed by tumor cells regulate T-cell activity mostly through engagement of inhibitory receptors (53), we wondered whether CAMK1D, being an intracellular kinase, might indirectly regulate T-cell activity. We therefore studied parameters of T-cell effector function upon contact with CAMK1D-proficient or -deficient KMM-1 cells, including the secretion of INFγ, Granzyme B, IL2, or TNFα.

Although we consistently detected increased T-cell–mediated tumor cell killing after CAMK1D knockdown in KMM-1 cells, functional analysis of T cells did not reveal any increased T-cell function after interaction with CAMK1D-deficient compared with wt tumor cells (Supplementary Fig. S3D). Therefore, we concluded that CAMK1D did not affect type 1 effector T-cell function and hypothesized that it may instead have regulated the sensitivity of tumor cells toward cytotoxic T-cell attack. Thus, we exposed KMM-1 cells to the cytotoxic agents FasL (rHuFasL), TRAIL (rHuTRAIL), or TNF (rHuTNF) commonly used by T cells to kill their target cells. The respective cell death–mediating receptors for FasL and TRAIL, Fas, DR4, and DR5 were highly expressed on KMM-1 cells while the TNF receptors TNFR1 and TNFR2 were not (Fig. 4A). Whereas CAMK1D-proficient KMM-1 cells were resistant against all tested cytotoxic agents, CAMK1D-deficient tumor cells were dramatically reduced after exposure to FasL and TRAIL (Fig. 4B). We also detected FasL on 28.2% and 16.1% of CD4⁺ and CD8⁺ MILs, respectively (Fig. 4C) and on 12.7% of flu TC (Supplementary Fig. S3E). TRAIL expression was detected only on 12.5% and 5.3% of CD4⁺ and CD8⁺ MILs, while membrane-bound TNF was hardly detectable (Fig. 4C). Neutralization of FasL by mAbs completely abrogated CAMK1D-induced protection against the cytotoxic activity of MILs (Fig. 4D). Thus, CAMK1D-mediated intrinsic tumor resistance against activated T cells by interfering with Fas-mediated death signaling. In line with this, U266 cells highly expressed Fas (Fig. 4E) and were protected by CAMK1D expression against Fas-mediated cell death similar to KMM-1 cells (Fig. 4F).

Because Fas–FasL interactions represent a major cytotoxic mechanism, we tested whether CAMK1D also protected solid tumors against immune rejection. In the human cancer cell lines PANC-1 and MCF7, Fas expression was low. However, we found high Fas and CAMK1D expression in Mel270, a PD-L1⁺ human uveal melanoma cell line (Fig. 4G and H; Supplementary Fig. S4A). Uveal melanoma is a highly treatment-refractory and anti–PD-1–resistant subtype of melanoma (54). Silencing of CAMK1D significantly increased the cytolytic response of Mel270 toward FasL exposure (Fig. 4I), indicating that uveal melanomas exploited CAMK1D for resistance against T-cell attack. In contrast, CAMK1D silencing in the Fas-negative tumor cell lines PANC-1 and MCF-7 did not sensitize these cells toward T-cell killing (Supplementary Fig. S4B and S4C). These data provided rationale for CAMK1D inhibition only in the context of Fas-positive tumors to achieve significant antitumor immune response. We hypothesized that CAMK1D expression in uveal melanoma might protect those tumors with strong Fas expression against immune rejection. We therefore stratified patients in the TCGA database cohort according to CAMK1D and Fas. Kaplan–Meier analyses showed that overexpression of CAMK1D in Fas receptor^high, but not in Fas receptor^low tumors correlated with poor patient prognosis (Fig. 4J). This suggested that CAMK1D exerted a tumor-protective effect only in the context of Fas activation during an immune response. Overexpression of CAMK1D and PD-L1 were tightly coregulated in uveal melanomas (Fig. 4K). Consequently, our study with PD-L1 expressing yet refractory tumor models shows that CAMK1D represented another level of immune resistance superseding the PD-L1 axis in mediating immune suppression.

Using the TCGA database, we studied CAMK1D and PD-L1 coregulation in other tumor entities that are largely unresponsive to anti–PD-1 treatment, specifically ovarian, stomach, and esophageal carcinomas (55, 56). As observed in uveal melanoma, CAMK1D, and PD-L1 were coexpressed and we detected significant correlations of CAMK1D and Fas receptor expression with poor outcomes (Supplementary Fig. S5A–S5F). Hence, CAMK1D is coregulated with PD-L1 and controls tumor rejection after Fas receptor activation in several anti–PD-1 treatment–refractory tumors.

FasL binding to Fas receptor results in complex signaling events leading to a caspase cascade that initiates apoptosis (57); this binding also stimulates Ca²⁺ influx into the cytoplasm, which ultimately triggers CAMK1D activation (58). We speculated that CAMK1D might interfere with the apoptotic cascade to mediate its tumor-protective effect. Thus, we assessed the impact of CAMK1D expression on tumor cell killing in the absence of effector caspases. Silencing of each downstream effector caspase-3, -6, and -7 completely abrogated the increased lysis of CAMK1D-deficient tumor cells after FasL exposure (Fig. 5A and B). Thus, CAMK1D selectively regulated cellular sensitivity toward apoptotic cell death. These results demonstrated the necessity of simultaneous activity of all three effector caspases for efficient induction of apoptotic cell death after Fas activation.

CAMK1D activation depends on Ca²⁺/calmodulin (CaM) binding, allowing the CAMK kinase (CAMKK) to phosphorylate and fully activate CAMK1D (58, 59). We speculated that FasL-expressing MILs might trigger Ca²⁺ release and therefore compared intracellular Ca²⁺ in KMM-1 cells on single-cell level after exposure to MILs or rHuFasL. Shortly after treatment, both conditions induced an increase of intracellular Ca²⁺ (Fig. 5C and D).

W-7 hydrochloride inhibits Ca²⁺/calmodulin complexes (60) consequently impacting CAMK1D activation. Treatment with 5 μmol/L W-7 hydrochloride was not toxic to KMM-1 cells (Fig. 5E) and sharply recapitulated the effect of CAMK1D silencing on FasL-induced tumor cell apoptosis, suggesting that CAMK1D was the decisive target of calmodulin for mediating FasL resistance (Fig. 5F). Because both CAMK1D silencing and W-7 hydrochloride treatment only incompletely blocked CAMK1D, we explored whether their combination reduced cell viability after FasL exposure. This combinatorial treatment resulted in a 3-fold further increase of tumor cell killing (Fig. 5F). To corroborate these findings, we applied CAMK1D inhibitor (QPP-A) to multiple myeloma and uveal melanoma cell lines. The additional treatment with rHuFasL induced a significant loss of tumor cell viability, confirming that CAMK1D played a substantial role in conferring resistance toward apoptosis (Fig. 5G).

These results demonstrated that CAMK1D activation in cancer cells was (i) triggered by CTL via FasL-induced Ca²⁺ release and (ii) was required to control Fas-induced tumor cell apoptosis. To confirm the immune-resistant role of CAMK1D in vivo, we knocked out Camk1d in the murine colorectal cell line MC38 (Supplementary Fig. S6A). In vitro analysis of Camk1d-deficient tumor cells revealed their increased sensitivity toward FasL as well as TRAIL-mediated apoptosis (Supplementary Fig. S6B). Thus, we injected MC38-Camk1d KO and -NTS (nontargeting sequence) cells into the left and right flank of the same mouse of both immunodeficient NSG and immunocompetent C57BL6 mice. Camk1d KO and NTS tumors developed rapidly in a similar manner in NSG mice, whereas a significant difference was observed in C57BL6 mice where Camk1d-deficient tumors were significantly reduced (Fig. 5H). These data demonstrated that the immune system in immunocompetent mice was not able to reduce tumor outgrowth due to Camk1d expression.

To elucidate CAMK1D involvement in the Fas signaling cascade, we studied activation of caspase-8 and -9, the prototypic initiator caspases of the extrinsic and intrinsic apoptotic pathway, respectively (61). FasL-induced activation of both caspases was comparably effective in CAMK1D-proficient and -deficient KMM-1 cells (Fig. 6A and B). Consequently, we hypothesized that CAMK1D regulated the activity of downstream effector caspases. To this end, we first studied the activation of the central executioner caspase-3. We observed an increase in caspase-3 activation in CAMK1D-deficient KMM-1 cells after FasL treatment (Fig. 6C–E). In addition, we also detected increased cleavage of the effector caspase-6 and -7 in CAMK1D-deficient tumor cells (Fig. 6F; Supplementary Fig. S7A). The phosphorylation and activation of the transcription factor cAMP response element-binding protein (CREB) was increased in CAMK1D-proficient cells, which was responsible for the transcription of the antiapoptotic molecule Bcl-2 (Supplementary Fig. S7B). We also observed that at early time points (15 minutes, 30 minutes, and 1 hour) after rHuFasL stimulation, the phosphorylation of ERK1/2 was enhanced in wild-type cells, but not in CAMK1D knockdown cells (Supplementary Fig. S7B). The altered activation of the presented proteins implied that CAMK1D not only controlled activation and activity of effector caspases, but also induced the expression of antiapoptotic and mitogenic proteins leading to tumor cell resistance toward FasL-positive T cells. CAMK1D has thus far not been established as a regulator of effector caspase activity. In silico analysis predicted a binding motif for CAMK1D on caspase-3 and caspase-6 (Supplementary Fig. S7D). Notably, CAMK1D coimmunoprecipitated with caspase-3, caspase-6, and caspase-7 and the interaction increased upon rHuFasL treatment (Fig. 6G and H; Supplementary Fig. S7C). A direct CAMK1D/effector caspase interaction could result in stoichiometric inhibition of caspase cleavage by initiator caspases. Alternatively, the effector caspases may also serve as targets of CAMK1D kinase activity. Phosphorylation of inhibitory serine residues impedes caspases activation, proteolytic activity, and ultimately hampers apoptosis induction (62).

The inhibitory phosphorylation sites of caspase-3 (Ser150) and caspase-6 (Ser257; refs. 63, 64) were located in the kinase-function critical distance of up to 4 amino acids apart from the predicted binding site for CAMK1D (Supplementary Fig. S7D). We therefore wondered whether CAMK1D was able to phosphorylate Ser150 and Ser257 of caspase-3 and -6, respectively. CAMK1D-deficient KMM-1 cells had reduced phosphorylation of inhibitory serine residues on both caspase-3 and -6 already at steady–state conditions (Fig. 6I–L). In KMM-1 wt cells, phosphorylation transiently decreased 15–30 minutes after FasL treatment (which is attributed to transient stimulation of phosphatases; ref. 65), but recovered to prestimulation expression within 1 (caspase-3) to 4 hours (caspase-6). In contrast, caspase-3 and -6 phosphorylation was persistently low in CAMK1D-deficient KMM-1 cells, resulting in overall much lower caspase inactivation compared with CAMK1D wt cells. This demonstrated that CAMK1D was required for steady-state inactivation of effector caspases through phosphorylation and for the rapid restoration of caspase-3 and -6 phosphorylation after FasL stimulation.

CAMK1D, upon its activation through FasL, regulated activation and activity of all effector caspases after cytotoxic T-cell encounter. These results suggested that this effect was at least partially achieved by the inhibitory phosphorylation of the effector caspases.

Despite noteworthy improvements in the field of immunotherapy, where immune checkpoint blockade (ICB) has broad clinical success (11, 13, 14, 66) a significant proportion of patients with cancer do not respond to ICB (15, 16, 67). Unknown immune checkpoint molecules might be employed by tumor cells to escape the antitumor immune response. Here, we used a multiple myeloma cell line to conduct a systematic search for genes controlling immune rejection in PD-L1–refractory human tumors. Among the identified genes, CAMK1D was chosen for further validation and mode of action analysis. CAMK1D expression is elevated in invasive carcinomas compared with carcinoma in situ and overexpression of CAMK1D in nontumorigenic breast epithelial cells increased proliferation and epithelial-to-mesenchymal transition (68). We reported a different role of CAMK1D in controlling the resistance of PD-L1⁺ tumor cells against apoptosis triggered by cytotoxic T cells. Tumor cell killing in CAMK1D-deficient cells was independent of the T-cell source, as both MILs and flu-specific CD8⁺ T cells were able to reproduce the same effect. Tumor cells can evade the immune system either by intrinsically increasing tumor cell resistance (69) or by hampering immune cell activation (53). Our data demonstrated that CAMK1D-deficient tumor cells did not enhance T-cell function. On the other hand, CAMK1D acted as central mediator of intrinsic tumor resistance toward CTL. Cytotoxic T cells eliminate tumor cells through the extrinsic apoptosis pathway, initiated by death receptor signaling, activating pro–caspase-8 (57, 70, 71), or by triggering the intrinsic pathway through the release of cytotoxic granules. This induces mitochondrial damage, apoptosome formation, and subsequent activation of pro–caspase-9 (72). Both initiator caspases activate the common executioner caspases-3, -6, and -7, which in turn cleave key intracellular substrates including endonucleases, thus irreversibly triggering the apoptotic cell death (73, 74). We observed that the initiator caspase-8 and -9 were not differentially affected in CAMK1D-proficient and -deficient tumor cells. Caspase-8 is inactivated upon phosphorylation of tyrosine-380, which leads to increased resistance to CD95-induced apoptosis (75). However, CAMK1D is a serine/threonine protein kinase and in silico analysis revealed no binding site between CAMK1D and caspase-8. Coexpression of death receptor ligands and cytotoxic granules enabled CTL to simultaneously trigger both apoptosis pathways, which differed from other common mechanisms of cell apoptosis, which generally trigger either one or the other pathway (76). Therefore, efficient resistance mechanisms against T-cell–induced apoptosis may target the common end route of both pathways. Consequently, inhibition of effector caspase activity may represent a hallmark of tumor immune resistance.

Caspases can be regulated by posttranslational modifications such as phosphorylation and ubiquitylation that can block caspases activation and activity (62). Indeed, phosphorylation of caspase-3 by p38 at Ser150, directly inhibits caspase-3, hindering Fas-induced apoptosis in neutrophils (63). Likewise, in the colon cancer cell line SW480, caspase-6 is inhibited by ARK5-phosphorylation, leading to the evasion of Fas-induced apoptosis (64). Caspase-7 can be inhibited posttranslationally by PAK2-mediated phosphorylation at Ser30, Thr173 and Ser239, which negatively regulates caspase-7 activity (77). We proposed a model where FasL stimulation increases calcium release from the ER, thereby binding and activating calmodulin, the upstream activator of CAMK1D. The binding of calmodulin to CAMK1D allows CAMKK to phosphorylate and fully activate CAMK1D. As a consequence, CAMK1D bound to the effector caspases inhibiting their activation acting as a direct stoichiometric inhibitor and by phosphorylation CAMK1D subsequently reduced the activity of the effector caspases. Moreover, activated CAMK1D translocated into the nucleus where it phosphorylated and activated CREB, leading to the transcription of Bcl-2. Thus, CAMK1D is an immune checkpoint molecule that interferes with tumor cell death, sustaining antiapoptotic pathways.

As CAMK1D is ubiquitously expressed, the pharmacologic inhibition may increase tumor susceptibility toward T-cell attack, but also impair T-cell activity. In line with this concern, blockade of the tyrosine kinase JAK2 sensitized multiple myeloma tumor cells to NK-cell attack (78); however, the function of NK and T cells was impaired in human myeloproliferative neoplasms (79–81). More studies must be conducted to clarify the impact of CAMK1D-targeted therapy on T cells. Nonetheless, CAMK1D remains a potential target for cancer immunotherapy, in particular, for those patients who experience relapse or demonstrate unresponsiveness to conventional therapies.

Our studies confirm the role of CAMK1D in vivo as a novel immune checkpoint molecule conferring resistance toward immune attack. It is conceivable that tumor cells exploit Fas signaling imposed by cytotoxic T cells to activate an apoptosis resistance mechanism targeting the final effector expression of both intrinsic and extrinsic apoptotic pathways resulting in an increased resistance to immune cell attack. In T-cell–infiltrated tumors, this mechanism may impact the treatment resistance of tumor cells, as CAMK1D may reduce the efficacy of antitumor treatments that directly or indirectly exploit the intrinsic apoptotic signaling pathways to trigger cancer cell death.

V. Volpin reports grants from Deutsche Forschungsgemeinschaft (DFG, German Research Foundation; project number 324392634 - TRR 221) during the conduct of the study, personal fees from iOmx Therapeutics AG (advisor) outside the submitted work, as well as a patent for EP20175030 (intracellular kinase associated with resistance against T-cell–mediated cytotoxicity and uses thereof pending).

T. Michels reports personal fees from iOmx Therapeutics AG (advisor) outside the submitted work and a patent for EP20175030 (intracellular kinase associated with resistance against T-cell–mediated cytotoxicity and uses thereof pending).

A. Sorrentino reports personal fees from iOmx Therapeutics AG (advisor) outside the submitted work and a patent for EP20175030 (intracellular kinase associated with resistance against T-cell–mediated cytotoxicity and uses thereof pending).

M. Boutros reports grants from Cellzome/GlaxoSmithKline outside the submitted work and a patent for EP20175030 (intracellular kinase associated with resistance against T-cell–mediated cytotoxicity and uses thereof pending).

S. Haferkamp reports personal fees from Bristol-Myers Squibb and MSD and grants from DFG (HA8418) during the conduct of the study.

H. Goldschmidt reports the following potential conflicts of interest: grants, research support, nonfinancial support, and personal fees from Amgen, Bristol-Myers Squibb/Celgene, Janssen/Cilag, and Sanofi; grants, research support, and personal fees from Chugai; research support and nonfinancial support from Takeda; research support and personal fees from Novartis; and grants from the Dietmar-Hopp-Foundation and the Johns Hopkins University.

H. Goldschmidt reports the following potential conflicts of interest outside the submitted work: research support from Incyte, Molecular Partners, Merck Sharp and Dohme (MSD), and Mundipharma GmbH; nonfinancial support from Adaptive Biotechnology; and personal fees from Beupdated Helbig Consulting & Research AG Schweiz, Chop GmbH, Congress Culture Concept München; ConnectMedia Warschau/Polen; Tumorzentrum Munchen; FomF Gmbh; GlaxoSmithKline; GWT Forschung und Innovation Dresden; Kompetenznetz Maligne Lymphome (KML); MedConcept GmbH; Medical Communication GmbH; Münchner Leukämie Labor; New Concept Oncology; Omnia Med Deutschland; Onko Internetportal dkg-web GmbH; STIL Forschungs GmbH; Veranstaltungskonzept Gesundheit Metternich; Institut für Versorgungsforschung in der Onkologie GbR; and ArtTempi.

G. Vereb reports grants from National Research, Development and Innovation Office, Hungary (OTKA K119690) during the conduct of the study.

N. Khandelwal reports a patent for EP20175030 (intracellular kinase associated with resistance against T-cell–mediated cytotoxicity and uses thereof pending).

P. Beckhove reports grants from Deutsche Forschungsgemeinschaft (DFG, German Research Foundation; project number 324392634 - TRR 221) and other from iOmx Therapeutics AG (supply of small molecules) during the conduct of the study, other from iOmx Therapeutics AG (shareholder) outside the submitted work, as well as a patent for EP20175030 (intracellular kinase associated with resistance against T-cell–mediated cytotoxicity and uses thereof pending).

No potential conflicts of interest were disclosed by the other authors.

V. Volpin: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, visualization, writing–original draft, project administration, writing–review and editing. T. Michels: Conceptualization, software, formal analysis. A. Sorrentino: Conceptualization, investigation. A.N. Menevse: Resources, investigation. G. Knoll: Investigation. M. Ditz: Investigation. V.M. Milenkovic: Investigation. C.-Y. Chen: Investigation. A. Rathinasamy: Resources. K. Griewank: Resources. M. Boutros: Resources. S. Haferkamp: Resources. M. Berneburg: Resources. C.H. Wetzel: Resources. A. Seckinger: Formal analysis. D. Hose: Resources, formal analysis. H. Goldschmidt: Conceptualization. M. Ehrenschwender: Resources. M. Witzens-Harig: Resources. A. Szoor: Investigation. G. Vereb: Resources. N. Khandelwal: Conceptualization, resources. P. Beckhove: Conceptualization, supervision, funding acquisition, visualization, writing–original draft, project administration, writing–review and editing.

The authors thank Dr. Haase (LMU, Munich, Germany) for the pEGFP-Luc plasmid, Prof. Moldenhauer (DKFZ) for the MHC-I antibody, Martins Freire (iOmx Therapeutics AG) for the bulk MC38-Camk1d KO cells, and Soelch (RCI) for technical support of QPP-A testing. This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation; project number 324392634 - TRR 221) and the National Research, Development and Innovation Office, Hungary (OTKA K119690).

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.