Characterization of BAY 1905254, an Immune Checkpoint Inhibitor Targeting the Immunoglobulin-Like Domain Containing Receptor 2 (ILDR2) Free

Immune checkpoint inhibitors (ICI) have become an indispensable approach in cancer therapy. Immune checkpoints, expressed by various cell subsets in the tumor microenvironment (TME) and in lymphoid organs, that is, by immune, tumor, or stromal cells (1), play a role in regulating the immune system. Immune checkpoints can be exploited by cancer cells, which can thereby shut down the antitumor immune response (2). The PD-1/PD-L1 axis, being one of the first identified immune checkpoints, has been intensively investigated. Its blockade shows significant clinical activity and benefit in the treatment of various cancer types (3). However, there is a distinct number of patients who either do not respond to PD-1/PD-L1 inhibitors or develop acquired resistance during treatment (4). Additional treatment approaches are clearly needed, and immune checkpoints independent of cancer cell expression are of high value in this regard.

The immunoglobulin-like domain containing receptor 2 (ILDR2) is a type I transmembrane protein that has been identified as a new member of the B7 family of immunomodulatory receptors (5). ILDR2 has been previously implicated in the development of type II diabetes and the formation of tricellular junctions (6–8), but its exact physiological role remains unclear. ILDR2 has been described to be expressed in several nonlymphoid tissue types, that is, kidney, testis, and liver (7, 9). In particular, ILDR2 expression has been reported in kidney podocytes, and siRNA-mediated ILDR2 deficiency is found to be associated with cytoskeletal injury (9). It has also been suggested that ILDR2, together with its homolog ILDR1, regulates alternative pre-mRNA splicing through binding to certain splicing factors (10). ZNF70, a member of the Kruppel C2H2-type zinc finger protein family, has been reported to bind to ILDR2 resulting in HES1 gene expression changes (11). However, these studies have been performed only in ILDR2-transfected cells in vitro.

ILDR2 was reported to exhibit immunosuppressive functions, namely the inhibition of T-cell activity (5). The immunomodulatory function of ILDR2 and its role in antigen-specific immune tolerance has been demonstrated in several models of autoimmune diseases, including mouse models of rheumatoid arthritis, multiple sclerosis, and type I diabetes (5, 12). ILDR2 has also been reported to be upregulated in monocyte-derived dendritic cells (DC) differentiated toward the DC2 phenotype (i.e., DCs driving the differentiation of TH2 cells; ref. 13). ILDR2 has been shown to be upregulated in aged mouse cortical microglia, but the role of ILDR2 in the cerebral cortex remains to be established (14).

Herein, we report the preclinical characterization of BAY 1905254, a humanized anti-ILDR2 hIgG2 function–blocking antibody (anti-ILDR2) generated to block the immunosuppressive effects of ILDR2 on T-cell activation. This mAb is mouse/human cross-reactive, as ILDR2 has a 98% sequence identity between the extracellular domains of human and murine orthologs. According to detailed preclinical characterization, BAY 1905254 blocked the immunosuppressive effects of ILDR2 and promoted antigen-specific T-cell responses in vivo. BAY 1905254 also demonstrated potent antitumor efficacy in several syngeneic tumor models, both as a monotherapy as well as in combination with an anti–PD-L1, docetaxel, or vaccination. We additionally showed that ILDR2 is expressed in a stromal cell subset in the lymph nodes, where it could affect the process of T-cell priming.

Mouse B16F10 melanoma, CT26 colon carcinoma, and 3C9-D11-H11 myeloma cells, as well as human H9 T cells and HEK-293T kidney cells, were obtained from ATCC (B16F10 and CT26 obtained in 2010 and 3C9-D11-H11 in 2015). Cell lines were authenticated using short-tandem repeat (STR) DNA fingerprinting by the DSMZ (German Collection of Microorganisms and Cell Cultures GmbH) before using them in the experiments to rule out potential cross-contamination among cell lines. In the B16F10-OVA cell line (prepared by Bayer AG), B16F10 melanoma cells stably expressed a foreign antigen (OVA, full-length chicken ovalbumin) to boost the immunogenic system in the B16F10 syngeneic mouse melanoma model (generated in 2015). In the 293T-huILDR2 cells (prepared by Compugen Ltd), HEK-293T cells overexpressed human ILDR2. Mouse MBT-2 bladder cancer cells were purchased from RIKEN BRC (obtained in 2015). The human mammary gland cell line DU4475 was obtained from ATCC in 2015. All cell lines were tested to be free from Mycoplasma contamination using MycoAlert (Lonza) at least every 4 passages and before generating a cell line substock.

B16F10 cells were cultured in DMEM/Ham F12 medium (Biochrom, #FG 4815) containing 10% FCS (Biochrom, #S 0615). CT26, MC-38, MBT-2, and B16F10-OVA cells were cultured in RPMI-1640 medium (Biochrom, #FG 1215) containing 10% FCS. 3C9-D11-H11 cells were cultured in DMEM (Biochrom, #FG 0445) containing 5% horse serum (Biochrom, #S 9135) and 2.5% FCS. H9 cells were cultured in RPMI-1640 (Thermo Fisher Scientific, #11875085) containing 10% FBS (Thermo Fisher Scientific, #10082147), and HEK-293T and 293T-huILDR2 cells were cultured in FreeStyle 293 Expression Medium (Thermo Fisher Scientific, #12338018) supplemented with puromycin (5 μg/mL; Thermo Fisher Scientific, #A1113802). Cells from passages 2–10 were used in the experiments, and for transplantation, cells were harvested in a subconfluent (80%) state.

Human anti–PD-L1 (Bayer AG) and docetaxel (Taxotere, Sanofi) were used in combinatory treatments and CTX-R1G11 hIgG2 antibody (Bayer AG) was used as isotype control. Ovalbumin-derived peptides 257–264 SIINFEKL (OVA_257-264) and a longer 30-mer peptide (SMLVLLPDEVSGLEQLESIINFEKLTEWTS) were synthesized by Peptide Specialty Laboratories GmbH. The animals were immunized with the 30-mer OVA peptide, and cytosine-phosphate-guanine oligodeoxynucleotide (CpG, InvivoGen, #tlrl-1826-1) was used as vaccine adjuvant.

ILDR2-deficient (knockout, KO) mice were custom-made by Taconic Biosciences. For the generation of C57BL/6NTac-Ildr2^em5099Tac mice, a 17.8 kb genomic region containing ILDR2 exons 1 and 2 (including the putative promoter region) was deleted using CRISPR/Cas9-mediated gene editing to ablate ILDR2 expression. The position of the cutting sites was 1.7 kb upstream of exon 1 and 0.6 kb downstream of exon 2, respectively. A pair of crRNA:tracrRNA complexes (Dharmacon) was incubated with Cas9 protein (New England Biolabs) and microinjected into the pronucleus of C5BL/6NTac one cell-stage embryos. After recovery, the injected one cell-stage embryos were transferred to one of the oviducts of 0.5 days postcoitum pseudopregnant NMRI female mice (Taconic Biosciences).

The studies on human samples were conducted according to the ethical principles of research with human participants (including the revised Helsinki Declaration and the General Assembly of the World Medical Association). The study protocols were approved by the local ethics committees with written informed consent obtained from each volunteer.

For analysis of ILDR2 RNA expression of human colorectal cancer samples, lymph nodes from patients (n = 6) with gastrointestinal disorders (including colorectal and gastric cancer), representing the class of mesenteric lymph nodes except for one sample that derived from the greater omentum, were collected between January and May 2018. Lymph node stromal cells were isolated freshly for subsequent culture of the cells followed by PrimeFlow analysis at the CRO Natural and Medical Sciences Institute (NMI), the University of Tübingen (Tübingen, Germany).

The expression of ILDR2 mRNA was characterized in 72 human normal tissues from 3 individuals using the BIOIVT XpressWay profile technology. Formalin-fixed paraffin-embedded (FFPE) samples from commercial provider were used for analysis of ILDR2 mRNA expression in various human primary cancers (Supplementary Table S1) by RNAscope. Blood from healthy donors was provided by Clinical Research Services Berlin GmbH and collected in citrate vacutainers (Sarstedt, #02.1067.001). Fresh blood was subsequently used for isolation of peripheral blood mononuclear cells (PBMC).

A human phage display panning approach was designed to isolate and characterize antibodies that bind to ILDR2 with high affinity. Panning reactions were carried out using streptavidin-coated magnetic beads (M-280 Streptavidin Dynabeads, Thermo Fisher Scientific) to capture the biotinylated antigens. Beads were recovered using a magnetic rack (Promega). All phage panning experiments were performed using the XOMA031 human fab antibody phage display library, also referred to as the XOMA phagemid (XOMA Corporation). All blocking steps used skim milk in PBS at a final concentration of 5%.

Proteins required for phage display were biotinylated using a Sulfo-NHS-LC-Biotin kit (Pierce). Free biotin was removed from the reactions by dialysis against PBS. The biotin-labeled proteins included mouse IgG2a Fc region (ILDR2-HM) and the extracellular domain of a control antigen (CGEN-150XX) fused to the same mouse IgG2a sequence (CGEN-1500XX-M). The control protein was used for depletion steps in the panning experiments. Unwanted binders to streptavidin beads and the mouse IgG2a Fc domain were removed during the panning process by coupling Dynal Streptavidin Magnetic Beads (Dynal) with control antigen CGEN-150XX-M (100 pmol). A 3-mL aliquot of the XOMA031 phage library stock was mixed with 3 mL of 10% skim milk in PBS. A 300-μL aliquot of the CGEN-150XX-M coupled beads was added to this mixture and incubated for 30 minutes at 22°C on a tube rotator. The depletion beads were then removed with a magnetic test tube rack and discarded. For the selection of specific binders to ILDR2-HM, the blocked and depleted phage library was mixed with 300 μL Dynal Streptavidin Magnetic beads coupled to biotinylated ILDR2-HM (100 pmol). Reactions were incubated at 22°C for 1–2 hours, and nonspecific phages were removed by washing with PBS-Tween (0.05%) followed by PBS. Bound phages were then eluted by incubation with 500 μL of 100 mmol/L triethylamine (EMD), and the eluate was neutralized by mixing with 500 μL Tris-HCl, pH 8.0. The entire neutralized phage eluate was infected into 10 mL of a TG1 Escherichia coli (Lucigen) culture grown to an optical density (OD) of 0.5 T at 600 nm. The resulting XOMA031 phagemid-transformed cells were then spread on selective agar plates containing carbenicillin (100 μg/mL; Novagen, Millipore Sigma) and incubated overnight at 37°C.

The resulting E. coli lawns were scraped and resuspended in 2-YT Broth (Teknova). Cells were inoculated into a 100-mL culture of 2-YT supplemented with carbenicillin (100 μg/mL) and grown at 37°C until the OD at 600 nm reached 0.5. The culture was infected with M13K07 helper phage [3 × 10⁹ plaque-forming units (PFU)/mL; New England Biolabs]. The selection antibiotic for M13K07, kanamycin (Calbiochem, Millipore Sigma), was added at 50 μg/mL. The culture was then maintained at 25°C to allow phage packaging. A 1-mL aliquot of the culture supernatant was carried over for either a subsequent round of panning or Fab-binding screens. Second and third rounds were conducted in the same way, except that the rescued phage supernatant from the previous round was used instead of the phage library.

The isolated Fab fragments were then screened by flow cytometry for their ability to bind to ILDR2 expressed on the cell surface. The XOMA031 library is based on phagemid constructs that also function as isopropylthiogalactoside (IPTG)-inducible Fab expression vectors. Eluted phage pools from the third panning round were diluted in 2YT media, infected into TG1 E. coli cells, and spread on an agar plate to generate single colonies (i.e., clones). Individual clones were inoculated into 1-mL cultures of 2YT supplemented with carbenicillin (100 μg/mL) contained in VWR Deep Well Plates. Protein expression was induced by adding IPTG (final concentration 1.2 mmol/L; CalBiochem, Millipore Sigma). Expression cultures were incubated overnight at 25°C. Each plate included duplicate “blank” sample wells as negative controls. These were created from noninoculated aliquots of 2YT/carbenicillin processed the same way as the Fab expression cultures.

Fab proteins secreted into the E. coli periplasm were prepared as periplasmic extracts (PPE). To prepare PPEs, TG1 E. coli cells were first pelleted from the 1-mL expression cultures by centrifugation. Cell pellets were then suspended in 75-μL aliquots of ice-cold PPB buffer (Teknova) supplemented with Complete-Mini Protease Inhibitor (Roche) according to manufacturer's instructions. Cell suspensions were then incubated at 4°C for 10 minutes with shaking. Subsequently, 225 μL ice-cold water supplemented with Complete-Mini Protease Inhibitor was added per well. The cell suspensions were again incubated at 4°C for 10 minutes with shaking. The periplasmic extracts were separated from cell debris by centrifugation and transferred to microtiter plates for flow cytometry analysis. Each plate included duplicate “blank” PPE wells as negative controls. These were created from noninoculated cultures processed the same way as the Fab PPEs.

Fabs with affinity for ILDR2 were identified by flow cytometry using an Intellicyt iQue high-throughput flow cytometer (Sartorius). Individual Fab PPEs were tested for binding to HEK-293T cells overexpressing human ILDR2 (293T-huILDR2 cells). All analyses included negative control HEK-293T cells mock-transfected with an “empty vector” that did not contain the ILDR2 gene (293T-EV cells). Reagent preparation and wash steps were carried out in flow cytometry buffer (PBS with 1% bovine serum albumin; BSA). A 30-μL sample from each Fab or blank PPE was mixed with 30 μL of 293T-huILDR2 cells (2 × 10⁶ cells/mL), incubated at 4°C for 1 hour, and washed with flow cytometry buffer. Cells were mixed with an anti–C-myc primary antibody (final concentration 10 μg/mL; Roche). The cells were stained with an anti-mouse IgG Fc AlexaFluor-647 (final concentration 7.5 μg/mL; Jackson ImmunoResearch). The cells were then fixed in 4% paraformaldehyde in flow cytometry buffer. Samples were read on a HTFC screening system (Intellicyt). Data was analyzed using FCS Express or FloJo software (both obtained from De Novo Software). Results were expressed as the ratio of mean fluorescence intensity (MFI) signals from 293T-huILDR2 cells to 293T-EV control cells. Positive hits were identified as those giving an MFI ratio of 2 or higher. Hits from the high-throughput screening (HTS) were reformatted to human IgG1 antibodies and subsequently tested to characterize their biophysical properties and functional activity. The therapeutic drug candidate BAY 1905254 was reformatted as human IgG2 isotype. Human isotypes were used in all experiments except for the in vitro bead-based assays, where an early development candidate of BAY 1905254 in a mouse IgG2a Fc backbone was used.

Confirmed hits from the panning campaign were tested by both surface plasmon resonance (SPR) and flow cytometry to assess their binding affinity to ILDR2. All SPR experiments were performed at 22°C using a Biacore 3000 instrument (GE Healthcare). Anti-mouse IgG capture surfaces were prepared over all four flow channels of a CM5 Biacore chip (GE Healthcare BR100012) using standard amine coupling. The activation step using 100 μL of 100 mmol/L 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC; GE Healthcare BR100050) mixed with 100 μL of 390 mmol/L N-hydroxysuccinimide (NHS; GE Healthcare BR100050) was 7 minutes, followed by ten minutes of immobilizing the anti-mouse IgG antibody reagent [GE proprietary reagent, binds all IgG subclasses (IgG1, IgG2a, IgG2b, and IgG3), IgA, and IgM, GE Healthcare BR100838] at approximately 5 μg/mL in pH 5.0 acetate buffer, followed by blocking of all remaining activated carboxyl groups with ethanolamine for 7 minutes. Final immobilization levels ranged from approximately 11,600 RU to approximately 11,900 RU. For each cycle, a mouse Fc fusion construct of human ILDR2 (prepared in-house) diluted to 0.75 μg/mL in running buffer (HBS-P from GE Healthcare containing 0.01 mol/L Hepes, 0.15 mol/L NaCl, and 0.005% surfactant P-20 with 100 μg/mL filtered BSA) was captured for 1 minute over flow cells 1 and 2, and 2 minutes over flow cell 4, at a flow rate of 5 μL/minutes. Flow cell 3 was left without captured ILDR2 to provide a reference surface. Capture levels of the ILDR2 averaged approximately 210 RU over flow cell 1, approximately 180 RU over flow cell 2, and approximately 320 RU over flow cell 4. After capturing the ILDR2, running buffer was flowed over all flow cells for 2 minutes at a flow rate of 50 μL/minute to stabilize the capture surfaces. Kinetics of the binding of purified Fab fragment of BAY 1905254 to the captured ILDR2 were measured by injecting the BAY 1905254 Fab at concentrations of 295 nmol/L, 98 nmol/L, and 33 nmol/L over all four flow cells for 2 minutes, followed by 10 minutes of dissociation at a flow rate of 50 μL/minute. After each cycle, the mouse antibody capture surfaces were regenerated with two 30-second pulses of 10 mmol/L glycine-HCl (pH 1.7). Blank injections of running buffer identical to the BAY 1905254 Fab injections were included for double referencing. All samples were prepared in the running buffer described above. All sensorgram data were zeroed, aligned, and double-referenced using Scrubber2 software (BioLogic Ltd). The three sensorgrams generated for all three flow cells were globally fit with a simple bimolecular kinetic model that included a term for mass transport using Scrubber2 to estimate the association rate constant k_a and the dissociation rate constant k_d. The equilibrium dissociation constant K_D was calculated with the ratio k_d/k_a.

The affinity of BAY 1905254 binding to human ILDR2 expressed on HEK-293T cells was assessed by flow cytometry. In a 96-well plate, BAY 1905254 was 2-fold serially diluted across 16 wells in flow cytometry buffer (PBS with 1% BSA and 0.1% NaN₃). 293T-huILDR2 cell suspension was added to each well at 1 × 10⁵ cells/well so that the final BAY 1905254–binding site concentration ranged from 3.2 to 209 nmol/L. An additional well containing only cells and an identical titration of BAY 1905254 into 293T-EV cells were used as controls. The cells were equilibrated with the BAY 1905254 concentrations by gently shaking the plate for approximately 4 hours at 4°C. Excess Alexa-647–tagged goat anti-human IgG was added (64 nmol/L; Jackson ImmunoResearch) in flow cytometry buffer to all wells and incubated for 30 minutes at 4°C. After several wash and resuspending of the cell pellets in flow cytometry buffer, the MFI after collection of 10,000 events was performed using an IntelliCyt HTFC flow cytometer. The average MFI as a function of each BAY 1905254–binding site concentration was nonlinearly fit to a 4-parameter equation that relates flow cytometry fluorescent signals to cell-bound ligand (15) using GraphPad Prism software (GraphPad Software) to estimate the K_D of BAY 1905254 binding to cell-surface–expressed human ILDR2.

To obtain a comprehensive expression profile of ILDR2 mRNA across healthy human tissues and cell types, public microarray mRNA expression data was gathered from Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/) and ArrayExpress (https://www.ebi.ac.uk/arrayexpress/). Microarray samples from a total of 510 studies profiled with Affymetrix Human Genome U133 Plus 2.0 Array were included in the analysis (Supplementary Table S2). Samples were grouped into 3,297 tissue categories based on organ annotation, type of origin, or disease state. To obtain comparable expression intensity estimates across all samples, all microarray samples were normalized together using the R package RefPlus from Bioconductor (https://www.bioconductor.org/; ref. 16). For delineating background noise from relevant mRNA expression, MAS5 detection calls were applied (17). ILDR2 expression was deemed significant in each tissue if the median MAS5 detection call P value for all samples in the corresponding set of microarrays was below 0.05. Normalized linear expression values for ILDR2 as derived from the RefPlus algorithm were visualized using boxplots generated via basic plotting commands from the R language (18).

Similarly, to identify tissues with ILDR2 expression in mouse, the 115 public microarray studies were downloaded from the GEO database, and all samples measured with mouse-specific Affymetrix Mouse Gene 1.0 ST Array were processed together in one batch using the RefPlus algorithm (16). Mouse samples were grouped into 704 different tissues and cell types. Visualization was done as for the human samples using basic plotting functions from the R language (18). The MAS5 algorithm could not be applied to this dataset as this Affymetrix chip does not contain mismatch probes required for computing the MAS5 P values.

The expression of ILDR2 mRNA was further characterized for 72 independent human normal tissues (each from 3 individual donors) using the BioIVT XpressWay profile technology (https://www.bioivt.com/other-human-xpressway-profiles/), which is based on TaqMan RT-PCR mRNA expression profiling, and a primer set to detect the extracellular domain of ILDR2 (Applied Biosystems assay Hs01025494_m1). Template amount ensuring a minimum copy number of control gene mRNA transcripts was determined using quantitative gene expression techniques (β-actin, amplicon length 295 bp, >3,800 copies/100 ng total RNA; GAPDH, amplicon length 71 bp, >10,000 copies/100 ng total RNA). Cycle threshold (C_t) values were used to determine the initial mRNA copy numbers of both target and GAPDH genes by interpolation from a global standard curve generated for XpressWay profile technology.

The expression of ILDR2 mRNA in various human primary cancers was analyzed using the RNAscope Reagent Kit (Advanced Cell Diagnostics) according to the manufacturer's instructions. Species-specific ILDR2/PPIB and Dapβ were used as probes and Mayer Hematoxylin (Dako, #S3309) was used for counterstaining.

For the characterization of ILDR2 expression in mouse stromal cell subpopulations, 1 × 10⁵ MC-38 mouse colon cancer cells were injected subcutaneously to female C57BL/6N mice (Charles River). Around 8 days later, tumor-draining and non-draining lymph nodes (axial, brachial, inguinal) and spleens were enzymatically digested to enrich the stromal cell fraction, as described previously (19). In brief, lymph nodes and spleens were dissected with forceps and incubated in RPMI-1640/enzyme mix containing dispase II (0.8 mg/mL; Sigma, #D4693-1G), collagenase P (0.2 mg/mL; Roche, #11213865001), and DNase I (0.1 mg/mL; Sigma, #11284932001). The samples were incubated at 37°C, mixed and dissociated by pipetting until the tissue pieces were completely dissolved. To enrich for stromal cells, CD45⁺ cells were depleted using magnetic separation with mouse CD45 MicroBeads (Miltenyi Biotec) according to manufacturer's instructions. For subsequent FACS, cells were stained with the following antibodies: anti-CD45-FITC (30-F11; 1:200), anti-Ter-119-BV421 (Ter-119; 1 1:200), anti-podoplanin-PE-Cy7 (8.1.1; 1:200), and anti-CD31-PE (MEC13.3; 1:200; all BioLegend). Dead cells were excluded by staining with Sytox Blue (Thermo Fisher Scientific). Stromal cell subsets were differentiated on the basis of the presence or absence of the stromal cell markers CD45, CD31, and podoplanin using a BD FACS Aria II (BD Biosciences). Fibroblastic reticular cells (FRC) were defined as CD45^–podoplanin⁺CD31^–; lymphatic endothelial cells (LEC) as CD45^–podoplanin⁺CD31⁺; and blood endothelial cells (BEC) as CD45^–podoplanin⁻CD31⁺.

For ILDR2 mRNA analysis by TaqMan RT-PCR (Thermo Fisher Scientific, Applied Biosystems), RNA was extracted from the sorted stromal cell populations using the RNeasy Micro Kit (Qiagen) following the manufacturer's instructions. RNA from lymph nodes and spleen of 4 different mice were pooled as the number of lymphoid stromal cells from one mouse was too low. Reverse transcription was performed using the SuperScript VILO cDNA Synthesis Kit (Thermo Fisher Scientific, #11754-250) following the manufacturer's instructions. TaqMan PCR was performed in a 384-well format using the epMotion 5075 (Eppendorf AG) automated pipetting system. The TaqMan PCR reaction was prepared in 10 μL using the 2× TaqMan Fast Advanced Master Mix (Thermo Fisher Scientific, #4444557) following the manufacturer's instructions. Mouse GAPDH and ILDR2 (both from Thermo Fisher Scientific) were used as TaqMan PCR primers, and the expression of ILDR2 mRNA was normalized to GAPDH expression.

ILDR2 expression was analyzed in stromal cells obtained from whole lymph nodes and the spleen from naïve wild-type (WT) C57BL/6N and ILDR2 KO mice (both from Taconic Biosciences), in all stromal cells obtained from lymph nodes derived from patients with colorectal cancer, and commercially available cryopreserved human lymph node single-cell preparations from melanoma patients (Folio Conversant) using the PrimeFlow RNA Assay technology (Invitrogen). Following enzymatic digestion performed as described above, the cells were flushed through a 70-μm or 100-μm cell strainer and centrifuged. For the spleen cells, a red blood cell lysis was performed using Red Blood Cell Lysing Buffer Hybri-Max (Sigma, #R7757) according to the manufacturer's instructions. The resulting lysates were filtered and centrifuged at 300 × g for 7 minutes. The lysated spleen pellets were washed and resuspended in 2 mL of FACS buffer. Cell viability was assessed using LIVE/DEAD Fixable Yellow Dead Cell Stain Kit (Thermo Fisher Scientific, #L34959) for gating on single live cells. Cells were washed, and anti-CD16/CD32 was added for blocking Fc gamma receptors (1:100 in PBS + 2% FCS; eBioscience, #14-0161-82) for 15 minutes at 4–8°C. Subsequently, the antibody mixture, including anti-CD45-APC-Cy7 (clone 30-F11; BioLegend, #103115; 1:200), anti-podoplanin-PE-Cy7 (clone 8.1.1; BioLegend, #127412; 1:200), anti-CD31-BV421 (clone 390; BioLegend, #102423; 1:200) was added, and cells were incubated for 20 minutes at 4–8°C. Cells were then washed with PBS + 2% FCS, and the PrimeFlow protocol was performed in 96-well V-bottom plates using the PrimeFlow RNA Assay Kit (Thermo Fisher Scientific, #88-18005-210) with ILDR2-specific probe sets (Thermo Fisher Scientific, #VB1-3055309-PF for murine ILDR2, #VA1-3017626-PF for human ILDR2) according to the manufacturer's instructions. As a positive control, RPL13A-targeted probe sets (Thermo Fisher Scientific, #VB4-16154 and #VA4-13187) were used. The flow cytometric analysis was performed using the LSRFortessa (BD Biosciences). Data were analyzed using the FlowJo 10.2 software (BD Biosciences).

For the analysis of ILDR2 protein expression in tissue and cell lysates, quantification of ILDR2 peptides using a targeted proteomics mass spectrometry assay was established and performed by Evotec (Evotec GmbH). For sample preparation, human embryonic kidney cells (HEK cells) overexpressing human ILDR2 (293T-huILDR2 cells), human mammary gland cells (DU4475 cells, ATCC, #HTB-123), as well as monkey (Macaca fascicularis) testis and kidney tissue, were subjected to sodium deoxycholate (SDC) buffer–based lysis and digest. Samples were homogenized and lysed in SDC lysis buffer [1% (w/v) SDC, 10 mmol/L TCEP (tris(2-carboxyethyl)phosphine), 40 mmol/L CAA, 100 mmol/L Tris pH 8.5] using a GentleMACS dissociator. Proteins were reduced and alkylated while shaking at 750 rpm at 99°C for 10 minutes. Samples were sonicated in an ultrasonic bath, and protein concentrations were determined at 660 nm wavelength. Aliquots of 200 μg of each sample were prepared for the subsequent digest. After 1:2 dilution with H₂O, LysC/Trypsin mix (Promega) was added at an enzyme-to-substrate ratio of 1:200 for each enzyme. On the next day, the digest was stopped by adding 99% ethylacetate/1% TFA (trifluoroacetic acid). Samples were desalted using SDB-RP cartridges and subsequently extracted using 100 mg C18 Sep-Pak cartridges (Waters). Peptide samples were fractionated by high pH reverse-phase chromatography. Briefly, lyophilisates were reconstituted in 20 mmol/L ammonium formate (pH 10, buffer “A”), loaded onto an XBridge C₁₈ 250 × 3.0 mm analytic column (Waters) operated with the ÄKTA Explorer system (GE Healthcare), and separated by applying a gradient of increasing buffer “B” concentration (20 mmol/L ammonium formate, 80% ACN, pH 10). For each sample, the collected 36 fractions of eluting peptides were combined in a nonlinear way to generate 12 fractions. After desalting, all sample fractions were snap-frozen, lyophilized, and reconstituted in 0.1% formic acid for MS analysis.

For the mass spectrometric analysis, samples were loaded onto a reverse-phase analytic column (packed in-house with C18 beads), resolved by an acetonitrile gradient using a nanoflow UPLC (ultra-performance liquid chromatography) system, and directly electrosprayed via a nanoelectrospray ion source into a Q Exactive HF-X mass spectrometer (Thermo Fisher Scientific). The Q Exactive HF-X mass spectrometer was operated in a data-dependent acquisition mode to automatically switch between full scans and the acquisition of higher-energy collisional dissociation fragmentation spectra (MS/MS mode) of the fifteen most abundant peptide ions in the orbitrap mass analyzer (resolution R = 60,000 for MS and R = 15,000 for MS/MS scans).

For data processing, all raw files acquired were processed with the MaxQuant software suite (version 1.5.7.13) for peptide and protein identification and quantification (20) using a human database from Swiss-Prot (version from 02/2017) and a M. fascicularis database from UniProt (version from May 2017), respectively. Carbamidomethylation of cysteine was set as a fixed modification and oxidation of methionine and N-terminal acetylation were set as variable modifications. The minimum required peptide length was seven amino acids and up to two missed cleavages were allowed. A false discovery rate (FDR) of 0.01 was selected for both protein and peptide identifications.

The binding of ILDR2-Fc to human T cells and the inhibitory effect of a human IgG1 variant of BAY 1905254 on the ILDR2-Fc binding was evaluated using flow cytometry. For analysis of human T cells, PBMCs were isolated from human whole blood from healthy donors by density gradient centrifugation. Histoplaque was added per one Leucosep tube and centrifuged at 800 × g at 22°C for 20 minutes. The interphase was transferred into a new tube and washed three times with ice-cold PBS without Ca²⁺ and Mg²⁺ containing 2% FCS (Biochrom AG). Subsequently, T cells were isolated by magnetic-activated cell sorting (MACS) using Pan-T Isolation Kit (Miltenyi Biotec, #130-096-535). The obtained T cells were either activated or directly stained with ILDR2-Fc for flow cytometry.

For T-cell activation, wells were coated at 4°C overnight with anti-CD3 (R&D Systems, #MAB100) and anti-CD28 (R&D Systems, #MAB342) at 1 μg/mL in PBS. Wells were then washed twice, and 1 × 10⁵ freshly isolated T cells resuspended in complete RPMI medium (Invitrogen, #61870-044) containing 10% FCS (Biochrom AG, #S0615) and penicillin/streptomycin (50 U/mL; Biochrom AG, #A2213) were added. T cells were stimulated for 48–72 hours prior to analyzing for activation and ILDR2-Fc binding by flow cytometry.

After incubation with the anti-CD16/CD32 Fc block (BioLegend, #422302; 1:100 in PBS + 2% FCS) for 5 minutes at 22°C, cells were incubated with PE-labeled ILDR2-Fc fusion protein (5 μg/mL; comprised of the extracellular domain of human ILDR2 fused to a mouse IgG2a Fc fragment) or PE-labeled mIgG2a-isotype control. ILDR2-Fc and the isotype control (Bioxcell, #BE0085) were incubated for 30 minutes at 4–8°C in an antibody mix containing fluorescently labeled antibodies against CD3 (BD Biosciences, #555634), CD4 (BD Biosciences, #560650), CD8α (BD Biosciences, #557832), and CD25 (BD Biosciences, #555434). Cells were washed twice with PBS + 2% FCS and analyzed by flow cytometry. Dead cells were excluded by staining with Sytox Blue (Thermo Fisher Scientific) prior to the analysis. Activation of T cells was confirmed by upregulation of CD25 expression. The flow cytometric analysis was performed using the BD LSR II. Data was analyzed using the FlowJo 10.2 software (BD Biosciences).

For evaluating the antibody blockade of ILDR2-Fc binding, human H9 cells were grown to a density of 0.6–1.5 × 10⁶ cells/mL. Cells were harvested and resuspended at a concentration of 2 × 10⁶ cells/mL in a blocker buffer containing PBS (Life Technologies) supplemented with 2% human serum (Life Technologies) and incubated for 30 minutes at 4°C prior to addition of the fusion protein.

ILDR2 recombinant proteins comprised of the ILDR2 extracellular domain fused to either mouse IgG2a or human IgG1 Fc domains were generated (for sequences, see Supplementary Information). For demonstration of dose-dependent binding of ILDR2-Fc to cells, ILDR2-Fc or an Fc-control was diluted in a series of 3-fold dilutions in blocker buffer, starting at a concentration of 200 μg/mL. Thirty microliters of each Fc fusion dilution was incubated with 30 μL of cells at 4°C for 45 minutes. Cells were pelleted and washed once with flow cytometry buffer (dPBS with 1% BSA and 0.1% NaN₃) at 4°C. Biotin-labeled Goat anti-mouse IgG2a specific antibody (50 μL; Jackson ImmunoResearch, #115- 065-206) in blocker was added, and cells were incubated for 30 minutes (1:250 in blocker) at 4°C. Cells were again pelleted and washed once with flow cytometry buffer, followed by addition of SA-647 (50 μL; Jackson ImmunoResearch, #016-600- 084) in blocker and incubated for 30 minutes at 4°C. Cells were then washed twice with flow cytometry buffer, resuspended in 80 μL of flow cytometry buffer, and read on an Intellicyt HTFC screening system.

ILDR2-Fc self-competition assays were performed as above, with the exception that ILDR2-Fc was directly labeled with Alexa FluorTM 647 (Thermo Fisher Scientific) for detection of binding to H9 cells. Directly labeled ILDR2-Fc (110 nmol/L) was added to cells at the same time as unlabeled material (ILDR2-Fc or Fc-control at various concentrations), and incubated for 45 minutes at 4°C. Cells were then washed twice with flow cytometry buffer, resuspended, and read on an Intellicyt HTFC screening system.

To test antibody blockade of ILDR2-Fc binding to H9 cells, 110 nmol/L of ILDR2-Fc was incubated with varying amounts (0–100 μg/mL) of a human IgG1 variant of BAY 1905254 for 30 minutes at 4°C. The protein/antibody complexes were then added to the H9 cells, and the assay processed as describe above. Human hIgG1 (Synagis, #NDC 60574-4114-1) and mouse mIgG2a (clone MOPC-173; BioLegend) were used as isotype controls. Data were analyzed using Prism software (GraphPad).

The capacity of an early development candidate of BAY 1905254 (in a mouse IgG2a Fc backbone) to block the inhibitory effect of the ILDR2-Fc fusion protein was evaluated in a bead-based assay. Tosyl-activated beads (Invitrogen) were coated with anti-hCD3 (UCHT-1; BD Biosciences; 50 μg/mL) at 500 × 10⁶ beads/mL in 0.1 mol/L phosphate buffer rotated in mixer (Hula Mixer; Invitrogen) at 37°C overnight. Beads were washed twice with 0.1 mol/L sodium phosphate buffer, then coated with 160 μg/mL of control Ig (hIgG1-FC, R&D Systems), rhPDL1-Fc (R&D Systems) or ILDR2-Fc (extracellular domain fused to hIgG1-Fc backbone) at 400 × 10⁶ beads/mL for 5 hours at 37°C. Then, beads were washed twice with 0.1 mol/L sodium phosphate buffer and resuspended in PBS containing 0.05% BSA for 1 hour at 22°C. A total of 1 × 10⁵ control Ig or ILDR2-Fc-coated beads were incubated with anti-mouse IgG (Alexa 647-anti-mouse IgG; Jackson) in flow cytometry buffer (1x PBS, 0.5% BSA) 30 minutes at 4°C. Beads were then washed and analyzed by flow cytometry (MACSQuant Analyzer 9; Miltenyi Biotec) to detect the levels of mouse anti-hCD3 coated on the bead. Anti-CD3 and ILDR2-Fc or PD-L1-Fc coated beads were cocultured with human CD4⁺ T cells (10⁵ cells/well) with 1:1 cells to beads ratio in the presence of isotype controls (20 μg/mL; hIgG1, Synagis, NDC 60574-4114-1 or mouse IgG2a control clone MOPC-173, BioLegend) or BAY 1905254 in a mIgG2a Fc backbone for 5 days. PD-L1-Fc–coated beads served as positive control and were also cocultured with CD4⁺ T cells at 1:1 cells to bead ratio in the presence of PD-L1 blocking antibody (YW243.55S). Cells were cultured in complete IMDM (Gibco) supplemented with 10% AB human serum (Gibco), Glutamax (Gibco), sodium pyruvate (Gibco), MEM non-essential Amino Acids Solution (Gibco), and 2-mercaptoethanol (Gibco). At the end of 5-day culture, cells were stained with anti-CD25 (eBioscience, #17-0251-81), anti-CD4 (BD Biosciences, #345770), and fixable live/dead dye (BD Horizon Fixable Viability Stain 450, BD Biosciences) to determine total CD4⁺CD25⁺ T cells for each bead type by flow cytometry. Data acquisition was performed with MACSQuant Analyzer 9 (Miltenyi Biotec) and data analyzed using MACSQuantify, Microsoft Excel or GraphPad Prism 4 software. Supernatants were collected and assayed for IFNγ secretion by ELISA according to the manufacturer's instructions (Human IFNγ, R&D Systems, # DY285B).

To further analyze the immunosuppressive effect of ILDR2 on antigen-specific T-cell responses in vitro, lymph nodes from 8- to 12-week-old C57BL/6 WT or ILDR2-deficient (KO) mice (custom-made by Taconic Biosciences) were digested enzymatically to include the stromal cell fraction, as described by Fletcher and colleagues (19). In brief, lymph nodes were dissected and incubated in RPMI-1640 (Biochrom, #FG1215) containing 5% FCS (Biochrom, #S0615), dispase II (0.8 mg/mL; Sigma, #D4693-1G), collagenase P (0.2 mg/mL; Roche, #11213865001), and DNase I (0.05 mg/mL; Sigma, #11284932001). The samples were incubated at 37°C, mixed, and suspended by pipetting until the tissue pieces were completely dissolved. Cells were flushed through a 70-μm cell strainer and 1 × 10⁶ cells were seeded in complete DMEM (Biochrom, #FG0445) in 96-well round bottom microplates. For subsequent stimulation of the OT-I transgenic cells, 1 × 10⁶ lymph node cells from WT or ILDR2 KO mice were pulsed with OVA_257-264 (0.025 or 0.05 ng/mL).

Spleens were isolated from OT-I transgenic RAG1 KO (Taconic) and enzymatically digested with collagenase type IV (1 mg/mL; Gibco, #17104-019) in 5% FCS (Biochrom, #S0615) RPMI-1640 (Biochrom, #FG1215) and incubated for 30 minutes at 37°C. Cells were flushed through a cell strainer, and red blood cells were lysed by incubating cells with ACK lysis buffer (Thermo Fisher Scientific) for 1 minute at 22°C. PBS was added for dilution and cells were centrifuged. To track cell proliferation, 2 × 10⁵ OT-I transgenic cells were labeled with 2 μmol/L of the cell proliferation dye (CFSE; Life Technologies, #C34554) following the manufacturer's instructions prior to cultivation. A total of 2 × 10⁵ OT-I cells in complete DMEM were subsequently added to 1 × 10⁶ lymph node cells. Supernatants were collected after 24 and 48 hours for subsequent cytokine determination by MSD (Meso Scale Discovery, V-PLEX Plus Proinflammatory Panel1 Mouse Kit), and T-cell proliferation was analyzed after 96 or 120 hours by flow cytometry using the FACSCanto II or LSRFortessa (BD Biosciences). For determination of absolute cell numbers, counting beads (123count eBeads Counting Beads, Thermo Fisher Scientific) were added to the cells prior to the analysis. Data were analyzed using the FlowJo 10.2 software (BD Biosciences).

To analyze the effect of anti-ILDR2 blockade on antigen-specific T-cell proliferation in vitro, lymph node cells from 8- to 12-week-old C57BL/6 mice (Taconic) were pulsed with the OVA_257-264 peptide (0.01 or 0.05 ng/mL). Cells were incubated with BAY 1905254 or with the CTX-R1G11 hIgG2 isotype control at different antibody concentrations (5, 25, 100, or 500 ng/mL). Cells were incubated for another hour at 37°C before addition of CFSE-labeled OT-I transgenic cells, as described above. Analysis of T-cell activation by flow cytometry and MSD was performed as described above for the OT-I T-cell stimulation with lymph node cells from WT and ILDR2 KO mice.

All animal experiments were conducted in accordance with the German Animal Welfare Law and approved by local authorities or with the regulations of the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) and by the Institutional Animal Care and Use Committee (IACUC) of CrownBio or The National Council for Animal Experimentation (Israel).

For antigen-specific T-cell proliferation and cytotoxicity analyses, CD8⁺ OVA-specific T cells were isolated from OT-I CD45.2 mice (Taconic M&B A/S) by negative MACS using a CD8a⁺ T-Cell Isolation Kit (Miltenyi Biotec, #130-104-075). For analysis of T-cell proliferation, OT-I transgenic T cells were labeled using CellTrace CFSE Cell Proliferation Kit (Life Technologies, #C34554) following the manufacturers' instructions. The CFSE-labeled OT-I transgenic T cells (2 × 10⁶ cells in 100 μL of PBS) were adoptively transferred into 8- to 12-week-old CD45.1-expressing C57BL/6 WT mice (Taconic), followed by subcutaneous (s.c.) vaccination with 0.5 μg OVA and 1 μg CpG. The mice were intraperitoneally (i.p.) treated with BAY 1905254 and anti–PD-L1 (both at 10 mg/kg) as monotherapies or in combination (n = 8 mice/group) on days 1 and 4. OT-I T-cell proliferation was analyzed by flow cytometry on day 5 using the FACSCanto II or the LSRFortessa (BD Biosciences) and CountBright absolute counting beads (Life Technologies, #MP 36950). Data were analyzed using the FlowJo 10.2 software (BD Biosciences).

For in vivo OT-I T-cell cytotoxicity analysis, OT-I transgenic T cells (2 × 10⁶ cells in 100 μL of PBS) were adoptively transferred into 8- to 12-week-old CD45.1-expressing C57BL/6 WT mice (Taconic), followed by s.c. vaccination with 10 μg OVA and 10 μg CpG on day 1. The mice were treated with BAY 1905254 and anti–PD-L1 (both at 10 mg/kg) as monotherapies or in combination (n = 10 mice/group) on days 1, 4, and 7. On day 7, 1 × 10⁷ CFSE-labeled WT spleen cells were pulsed with OVA_257-264 or an irrelevant peptide (p53 short peptide: AIYKKSQHM, PSL GmbH) at 1 μg/mL for 30 minutes at 37 °C. The CFSE-labeled peptide-pulsed WT splenocytes (4 × 10⁶ cells in 100 μL of PBS) were injected intravenously (i.v.). On the next day, the number of OVA-positive cells in the spleen and in the inguinal lymph nodes was determined by flow cytometry as described above.

For the B16F10 and B16F10-OVA models, 1 × 10⁴ mouse melanoma cells were injected subcutaneously to female C57BL/6N mice (Charles River). For the combination studies with docetaxel or with adoptive OT-I T-cell transfer, 1 × 10⁴ B16F10-OVA cells were injected subcutaneously to female JAX C57BL/6J (Charles River) or B6.SJL-Ptprc^a Pepc^b/BoyCrl mice (Charles River), respectively. For ILDR2 KO mice with adoptive OT-I T-cell transfer, 1 × 10⁴ B16F10-OVA cells were injected subcutaneously to female C57BL/6N Tac-Ildr2^em5099Tac mice (Taconic). For the MBT-2 mouse bladder cancer model, 4 × 10⁵ cancer cells were injected subcutaneously to female C3H mice (Vital River Laboratory Animal Technology Co) at CrownBio. For the CT26 mouse colon carcinoma model, 5 × 10⁵ cancer cells were injected subcutaneously to female C57BL/6JOlaHsd (Envigo) or BALB/cAnN mice (Charles River). For the 3C9-D11-H11 mouse myeloma model, 1 × 10⁴ cancer cells were injected subcutaneously to female BALB/cAnN mice (Charles River). Cells were inoculated in 50% Matrigel (Basement Membrane Matrix, BD Biosciences) or PBS (MBT-2 model).

Unless otherwise indicated, the mice (n = 8–13 per group) were randomized according to primary tumor size prior to treatment. In a therapeutic setting, BAY 1905254, anti–PD-L1, and the isotype control (all 10 mg/kg, i.p.) were administered every 3 days when tumors had reached a mean size of 55–140 mm³ (5–15 days after tumor inoculation, depending on the mouse model). Docetaxel (20 mg/kg, i.v.) was given once when tumors reached a mean size of 90 mm³ (9 days after tumor inoculation). In a preventive setting in the B16F10 mouse melanoma model, treatment with BAY 1905254 was started on the day of tumor cell inoculation. Tumor volume [(length × width²)/2] was measured by caliper at least twice weekly. Treatment-to-control ratios (T/C) were calculated based on the mean tumor volumes at the end of each study.

For assessing IFNγ protein concentrations, the snap-frozen tumor samples were lysed with MSD Tris Lysis Buffer (Meso Scale Discovery, #R60TX-2) and stainless steel beads (5 mm in diameter; Qiagen, #69989) using TissueLyser II (Qiagen, #85220). The lysates were centrifuged at 25,000 × g at 4°C for 20 minutes, and 2 μL of the supernatants were diluted 1:50 with distilled water and analyzed for IFNγ concentration by ELISA using V-PLEX Proinflammatory Panel 1 Mouse Kit (Meso Scale Discovery, #N05048A-1).

For the analysis of immune parameter expression, snap-frozen tumor samples (20–30 mg, n = 5 tumor samples/group) were lysed with TissueLyser II (Qiagen, #85220), followed by total RNA extraction with RNeasy Plus Mini Kit (Qiagen, #74134) and cDNA synthesis with Superscript III First-Strand Synthesis (Thermo Fisher Scientific, #18080051) following the manufacturers' instructions. Gene expression analysis was performed by real-time quantitative PCR (qRT-PCR) in a 384-well plate (MicroAmp Optical 384-Well Reaction Plate, Applied Biosystems, #4309849) with 7900HT Fast RT-PCR System (Applied Biosystems). The TaqMan PCR reaction was prepared in 10 μL containing 15 ng cDNA and the probe of interest (Supplementary Table S3) using 2× TaqMan Fast Advanced Master Mix (Thermo Fisher Scientific, #4444557) following the manufacturer's instructions. The mRNA expression was calculated using the 2^−ΔΔ method and normalized to the endogenous GAPDH expression.

For assessing intratumoral CD45 protein expression or CD8α⁺ DCs, the harvested tumor tissues were dissociated with gentleMACS Octo Dissociator (Miltenyi Biotec) and Tumor Dissociation Kit, mouse (Miltenyi Biotec, #130–096-730) following the manufacturer's instructions. The generated single-cell suspensions were transferred through a 70-μm strainer, the filtrates were centrifuged at 300 × g for 7 minutes, and the pellets, including tumor-infiltrating lymphocytes, were resuspended in FACS buffer (3% FCS in PBS). The LIVE/DEAD Fixable Yellow Dead Cell Stain Kit (Thermo Fisher Scientific #L34959) was used to stain and detect dead cells. The Fc-blocking reagent (eBioscience, #14-0161-82) was then added. The cells were stained with anti-murine CD45 (BD Pharmingen, #5577235) or anti-murine CD8α (BD Pharmingen, #553033), anti-murine CD11b (eBioscience, 48-0112-82), and anti-murine MHCII (eBioscience, #17-5321-81) and the number of intratumoral CD45⁺ cells or CD8α⁺ DCs was determined by flow cytometry using FACSCanto II or the LSRFortessa (BD Biosciences). Absolute cell numbers were calculated by correlating the cell events to the respective tumor weight. For analysis of endogenous OVA-specific CD8⁺ T cells, processing of the tumor tissue and the subsequent cell staining for flow cytometry analysis was the same as described above. In addition, the PE-labeled OVA_257-264 peptide-MHC-pentamer (ProImmune, #F093-2A-E) was added to the antibody mix (1:20) and incubated together with the tumor cells.

Statistical analyses were performed using R (version 3.3.2 or newer; ref. 18) or the GraphPad Prism 7 software. If not otherwise stated, log or square root–transformed tumor volume data were assessed using ANOVA followed by Dunnett test for pairwise comparisons or Tukey method for multiple comparisons. For the MBT-2 model, statistical analysis of difference in tumor volume between the treatment groups was conducted using an independent-sample t test, and analyzed with SPSS (Statistical Product and Service Solutions, version 18.0; IBM). The in vivo combination data were analyzed using a linear model estimated with generalized least squares that included separate variance parameters for each study group. Mean comparisons between the treatment and control groups were performed using the estimated linear model. Statistical analyses of the cytokine concentrations were performed by two-way ANOVA analysis. For immune parameters, intratumoral immune cell populations, and cytotoxicity, ANOVA and Tukey method for multiple comparisons were used. P < 0.05 were considered significant. Data are expressed as means ± SD, unless otherwise indicated.

To generate an ILDR2-specific antibody, ILDR2-binding Fab sequences were identified using a human phage display panning approach. The antibody BAY 1905254 was selected for further characterization because of its high binding affinity and selectivity to ILDR2, as analyzed by SPR and flow cytometry. BAY 1905254 was cross-reactive to mouse, rat, and cynomolgus monkey ILDR2. In binding studies with ILDR2-transfected cells, the equilibrium dissociation constant was in a single-digit nanomolar range (K_D = 2.0 ± 0.4 nmol/L with 95% confidence interval of the fit), whereas no binding on mock-transfected cells was observed (Supplementary Fig. S1A). No binding to the closely related family members ILDR1 and ILDR3 was detected by SPR (Supplementary Fig. S1B). The affinity of purified BAY 1905254 Fab fragment binding to the ILDR2-Fc fusion protein (K_D = 2.8 nmol/L, k_a = 1.2 × 10⁵ L/mol/sec, k_d = 3.3 × 10^–4/sec) estimated from SPR studies closely resembled the K_D estimated from the flow cytometry binding study.

ILDR2 has been reported to be expressed in several nonlymphoid tissues in mice (7, 9). In our analyses, we found that the kidney and testis in mice, rats, and monkeys, as well as testis in humans, have detectable levels of ILDR2 mRNA and ILDR2 protein as demonstrable by targeted proteomics analysis (Supplementary Table S4). Compiled statistical analyses using publicly available microarray data indicated that ILDR2 is also expressed in mouse and human lymphatic tissues (Supplementary Fig. S2). In line with this, TaqMan PCR analyses by the BioIVT XpressWay profile technology indicated that in humans, ILDR2 is expressed in the spleen and tonsils (Supplementary Fig. S3). Likewise, previous reports demonstrate high ILDR2 expression in murine lymph node stroma (21). Therefore, to get a better understanding of the ILDR2 expression in lymphoid tissues, various stromal cell subsets present in mouse lymph nodes and spleen were isolated for TaqMan PCR detection. ILDR2 mRNA was detected particularly in the podoplanin⁺ and CD31^– FRCs isolated from lymph nodes (Fig. 1A). PrimeFlow RNA analysis of lymph node cells from ILDR2 WT and KO mice confirmed that ILDR2 was selectively expressed by CD45^– FRCs in mice (Fig. 1B). In humans, PrimeFlow analysis of cryopreserved lymph nodes from patients with melanoma indicated ILDR2 expression in FRCs, as well as in podoplanin/CD31 double-negative stromal cells (Fig. 1C). A portion of CD45⁺ cells, many of which were CD11c⁺ and, thus, likely represented DCs, expressed ILDR2 mRNA. Also, low ILDR2 expression was detected in FRC-like cells, obtained by culturing of lymphoid stromal cells from patients with colorectal cancer (Fig. 1D). The assessment of ILDR2 and podoplanin expression in a published single-cell RNA-seq dataset covering 19,081 splenic fibroblastic reticular cells (22) confirmed specific coexpression of these two genes across all identified cell-type clusters (Supplementary Fig. S4). Taken together, these data demonstrated that ILDR2 is expressed by lymphoid stromal cells, particularly by the FRC subpopulation of lymphatic organs. The syngeneic tumor cells used in this study and various human tumor tissues were also tested for ILDR2 expression by in situ hybridization using the RNAscope technology. However, all samples were negative for ILDR2 expression or only showed very weak signals. Taken together, data suggest that ILDR2 has an immunosuppressive effect in lymph nodes where T-cell priming occurs and not in the TME.

Because ILDR2 showed high expression in lymph nodes, we tested whether it exerted an immunosuppressive effect on T-cell priming. To investigate the effect of ILDR2 on T cells in vitro, human CD4⁺ and CD8⁺ T cells were isolated from human PBMCs. First, binding of the ILDR2-Fc fusion protein to unstimulated or anti-CD3/CD28–activated T cells was analyzed by flow cytometry. Weak binding of the labeled ILDR2-Fc fusion protein was observed on CD4⁺ and CD8⁺ unstimulated T cells (Supplementary Fig. S5A). The binding was increased in activated T cells obtained through stimulation with anti-CD3 and anti-CD28 (Fig. 2A). These data indicated that the putative receptor for ILDR2 is expressed on human T cells and is upregulated upon activation of CD4⁺ and CD8⁺ T cells. Screening assays performed using a human IgG1 variant of BAY 1905254 showed reproducible, dose-dependent inhibition of ILDR2-Fc binding to human T cells (Supplementary Fig. S5B and S5C), demonstrating the ability of BAY 1905254 to disrupt the binding of ILDR2 to its putative receptor on T cells. Next, the immunomodulatory activity of BAY 1905254 was analyzed in prestimulated human T cells using anti-CD3 and ILDR2-Fc–coated beads (Fig. 2B). The beads coated with the ILDR2-Fc fusion protein showed suppressed T-cell receptor (TCR)-stimulated IFNγ secretion (Fig. 2C) and T-cell proliferation (Fig. 2D) compared with beads coated with a control Fc protein. A comparable reduction of T-cell activity was observed when using the PD-L1-Fc–coated beads. The immunosuppressive effect of ILDR2-Fc, indicated by suppressed IFNγ secretion by T cells and T-cell proliferation, could be partially reversed by treatment with BAY 1905254. Correspondingly, the immunosuppressive effect caused by the PD-L1-Fc was partially reversed by treatment with anti–PD-L1. Thus, BAY 1905254 was capable of restoring the immunosuppressive effect of ILDR2 on T-cell priming, comparably with PD-L1 inhibition.

To further investigate the role of ILDR2 in T-cell responses, ILDR2-deficient C57BL/6 mice were generated by CRISPR/Cas9-mediated gene editing. No obvious clinical, macroscopic, or microscopic differences were observed between the ILDR2 KO and WT mice (Supplementary Table S5). To address the immunosuppressive effect of ILDR2 in vitro, lymph node cells from ILDR2 WT and KO mice were pulsed with the OVA_257-264 peptide and cocultured with OT-I transgenic T cells. Overall, there was a trend for increased cytokine secretion when the OT-I transgenic T cells were cultured with ILDR2-deficient lymph node cells (Fig. 3A–D). IL2 was significantly increased after 24 hours (P < 0.05), and IFNγ and IL6 were increased after 48 hours (P < 0.01 and P < 0.05, respectively). For TNFα, the difference between WT and ILDR2-deficient cells remained nonsignificant. These data were in accordance with the results obtained using the human ILDR2-Fc fusion protein and indicated that ILDR2 had an immunosuppressive effect on antigen-specific T-cell responses in mouse lymph nodes in vitro.

To investigate the effect of BAY 1905254-mediated ILDR2 blocking on T-cell activity in vivo, two OVA vaccination models were generated. In these models, ovalbumin-specific, TCR transgenic OT-I T cells were adoptively transferred and subsequently analyzed for proliferation and OVA-specific cell lysis. Mice treated with BAY 1905254 in combination with anti–PD-L1 showed significantly increased antigen-specific T-cell proliferation and cytotoxicity (both P < 0.01; Fig. 3E and F), as indicated by an increased number of proliferated OT-I transgenic T cells and increased lysis of OVA-pulsed spleen cells, respectively. Endogenous OVA-specific T cells were analyzed using pentamers, but no significant difference was observed. These data suggest that the blocking of ILDR2 by BAY 1905254 together with PD-L1 inhibition promotes antigen-specific T-cell activity in vivo.

Next, the in vivo efficacy and tolerability of the anti-ILDR2 BAY 1905254 was assessed in several syngeneic mouse models, using a preventive or a therapeutic setting (Supplementary Table S6). In the preventive setting, treatment with BAY 1905254 was initiated on the day of subcutaneous cancer cell inoculation, whereas in the therapeutic setting, treatment was initiated when a predetermined tumor size was achieved. In a weakly immunogenic B16F10 mouse melanoma model, intraperitoneal administration of BAY 1905254 at 10 mg/kg (every 3 days) using the preventive setting reduced tumor growth in comparison with the isotype control group with a T/C_volume ratio of 0.63 (P < 0.05; Fig. 4A). In the more immunogenic MBT-2 bladder cancer and CT26 colon cancer models, treatment with BAY 1905254 at 10 mg/kg (every 3 days) using the therapeutic setting resulted in reduced tumor volumes with T/C ratios of 0.38 and 0.50 (both P < 0.05; Fig. 4B and C). In all syngeneic models, BAY 1905254 was well tolerated without treatment-related body weight losses (Supplementary Fig. S6.)

The antitumor efficacy of BAY 1905254 was further investigated in combination with other antitumor therapies, namely an anti–PD-L1, docetaxel, or tumor antigen immunization. The combination potential of BAY 1905254 with anti–PD-L1 was studied in various syngeneic mouse models, including the CT26 mouse colon carcinoma, the B16F10 and B16F10-OVA mouse melanoma, as well as the 3C9-D11-H11 mouse myeloma, models. In the CT26 model, BAY 1905254 administered as a single agent at 3 mg/kg had no antitumor efficacy (Fig. 5A). However, when combining BAY 1905254 with anti–PD-L1, an additive antitumor efficacy was observed (P < 0.001). In the B16F10 and B16F10-OVA models, treatment with BAY 1905254 in combination with anti–PD-L1 revealed potent antitumor efficacy in comparison with the isotype control (B16F10, P = 0.004; B16F10-OVA, P < 0.001; Fig. 5B and C). In the B16F10 model, the combination treatment was synergistic. In the B16F10-OVA mouse melanoma model, BAY 1905254 combined with OT-I T-cell transfer resulted in significant tumor growth inhibition with a T/C ratio of 0.30 (P < 0.001; Supplementary Fig. S7A). In contrast, when using a similar setting in ILDR2-deficient mice, BAY 1905254 showed no antitumor efficacy when administered alone (P = 0.664) or in combination with anti–PD-L1 (P = 0.785 in comparison with anti–PD-L1 monotherapy, Supplementary Fig. S7B), indicating that the effect of BAY 1905254 is target-specific. In the 3C9-D11-H11 model, although monotherapy with BAY 1905254 showed no efficacy, treatment with BAY 1905254 in combination with anti–PD-L1 resulted in synergistic inhibition of tumor growth with improved efficacy in comparison to the isotype control (T/C ratio of 0.07, P < 0.001) or either of the agents alone (both P > 0.001; Supplementary Fig. S7C). A complete tumor eradication was observed in 58% (7/12) of mice in the combination treatment group at the end of the study.

The combination potential of BAY 1905254 with the chemotherapeutic drug docetaxel was assessed in the B16F10-OVA model. In this model, intraperitoneal administration of BAY 1905254 at 10 mg/kg had no efficacy as a monotherapy (T/C ratio of 0.85). However, treatment with BAY 1905254 in combination with docetaxel resulted in a significant reduction of tumor growth with a T/C ratio of 0.39 (P < 0.001; Fig. 5D).

To demonstrate in vivo that the activation of innate and adaptive immunity by immunization can be further enhanced by ILDR2 blockade, BAY 1905254 treatment was combined with an initial immunization with OVA and CpG in the B16F10-OVA model. In this study, a clear antitumor efficacy was observed with T/C ratios of 0.66 (P = 0.02), 0.50 (P < 0.001), and 0.38 (P < 0.001) for BAY 1905254 as a monotherapy, initial immunization only, and BAY 1905254 in combination with immunization, respectively (Fig. 5E). All combination treatments were well tolerated as indicated by an increase or only a minor loss (<5%) in mean body weight in all models tested (Supplementary Fig. S6).

To further investigate the mode of action of BAY 1905254, the effect of ILDR2 blockade on immune activation within the TME was analyzed ex vivo. In several cancer models, a trend for higher intratumoral IFNγ was observed in the tumors of BAY 1905254–treated mice compared with isotype control–treated mice (Supplementary Fig. S8). The effect was even more pronounced in the tumors of mice treated with BAY 1905254 in combination with anti–PD-L1. The combination treatment with BAY 1905254 and anti–PD-L1 resulted in immune activation in the B16F10 model, as indicated by the increased expression of immune-relevant genes, that is, encoding immune checkpoints or transcription factors involved in Th cell and macrophage differentiation (Fig. 6A). The number of CD8α⁺ DCs, specialized in cross-presenting exogenous antigens to CD8⁺ T cells (23, 24), was increased in the tumors treated with BAY 1905254 in combination with anti–PD-L1, indicating a shift to improved cross-presentation of antigens in the TME (Fig. 6B and C).

ILDR2 has been described as a novel member of the B7 family of immunomodulatory receptors exhibiting clear T-cell inhibitory activity (5, 12). We showed that an ILDR2-Fc fusion protein bind to activated, but not naïve, human T cells. However, the respective ligand partner of ILDR2 on T cells has not yet been identified. In our study, we generated a human/mouse/monkey cross-reactive hIgG2 antibody, BAY 1905254, to block the immunosuppressive activity of ILDR2. BAY 1905254 is a function blocking antibody and specifically binds to ILDR2, but not to its homologues ILDR1 or ILDR3.

Our results indicated that, besides its expression in nonimmune tissue types, ILDR2 is also expressed in lymph node FRCs, a specialized stromal cell subset located in the T-cell zone of the lymph node where T-cell priming and initiation of local T-cell responses occur (25, 26). FRCs create a three-dimensional network where T cells and DCs interact with each other to present antigens for initiation of T-cell responses (27, 28). FRCs have been described to be essential for homing, that is, recruitment of T cells and DCs into the lymph node (29, 30). FRCs have also been shown to exhibit immunoregulatory properties (31–33) and proposed to have a crucial role in immune cell homeostasis and adaptive immunity (34). For instance, FRCs are demonstrated to have an immunosuppressive effect on T-cell responses by production of nitric oxide (35). FRCs express PD-L1 and can induce antigen-specific T-cell tolerance under steady-state conditions (36). Consistently, human FRCs and podoplanin/CD31 double-negative stromal cells are described to secrete important cytokines and chemokines, which change upon inflammatory stimuli (37). On the basis of their position in the lymph node and their described immunoregulatory function, ILDR2-expressing FRCs most likely have an effect on the initiation of antitumor immune responses. Accordingly, it has previously been reported that a tumor induces a significant change of gene expression in stromal cells, especially in FRCs, in the tumor-draining lymph node in mice (38).

Although ILDR2 mRNA was found in both nonlymphoid and lymphoid tissues in several species in our expression analyses, ILDR2 protein could not be reliably and consistently detected by IHC or Western blotting, partially due to antibody incompatibility upon denaturing conditions. On the basis of targeted proteomics analysis, ILDR2 peptides were detected in the kidney and testis across species (human, mouse, and cynomolgus monkey), whereas the signal in lymphatic tissues and human PBMCs was inconsistent. This data is concordant with public IHC data from The Human Protein Atlas (39), where ILDR2 shows solely moderate nuclear staining in the kidney and testis, but no staining in the lymphatic tissues. Taken together, the data indicated that ILDR2 is a low-abundant transcript with little expression on the protein level in the lymphatic system.

Antigen-specific stimulation of OT-I transgenic T cells in the presence of lymph node cells of ILDR2-deficient mice led to increased cytokine release compared with cocultivation with WT lymph node cells in our study. These data confirm that ILDR2 has a suppressive effect on enhanced T-cell activation in vitro. We further demonstrated that an ILDR2-Fc fusion protein has immunosuppressive activity on human T-cell activation in a bead-based assay. In accordance with these data showing an immunosuppressive effect of ILDR2, ILDR2 blockade by BAY 1905254 enhanced antigen-specific T-cell proliferation and cytotoxicity in vivo in ovalbumin vaccination models with OT-I transgenic T cells when administered in combination with an anti–PD-L1. This is also consistent with previously published data on the immunosuppressive function of ILDR2 in homeostasis and in several autoimmune disease models (5, 12). ILDR2 could therefore represent a therapeutic target in immuno-oncology, and BAY 1905254 an immune checkpoint inhibitor for cancer immunotherapy.

For preclinical characterization, we analyzed the antitumor efficacy of BAY 1905254 in various syngeneic mouse models in vivo. BAY 1905254 exhibited antitumor efficacy which correlated with increasing mutational load in the studied cancer cell lines (40). Additive or even synergistic antitumor effects were observed in several syngeneic tumor models when BAY 1905254 was administered in combination with an anti–PD-L1, presumably due to the distinct mechanisms of action of ILDR2 and PD-1/PD-L1. These data provide a rationale for the therapeutic combination of BAY 1905254 with an anti–PD-L1. In the B16F10 mouse melanoma model, where a synergistic antitumor effect was observed, anti–PD-L1 showed no efficacy as a monotherapy. Future studies will reveal if treatment with BAY 1905254, alone or in combination with an anti–PD-1/PD-L1, is useful in patients resistant to PD-1/PD-L1 blockade. Combination with an immunogenic cell death–inducing chemotherapeutic (docetaxel) or with tumor antigen immunization showed additive antitumor efficacy. Thus, an increased number of antigens or an enhanced initial activation of innate immune cells can further promote the antitumor potency of BAY 1905254.

Ex vivo analyses of tumors from syngeneic mouse models revealed that the combination of BAY 1905254 with an anti–PD-L1 promoted immune activation and could result in increased intratumoral IFNγ, as well as enhanced infiltration of CD8α⁺ DCs. The combination treatment led to upregulation of other immune checkpoints such as CTLA4, LAG3, and PD-1 as a compensatory mechanism, consistent with previous publications (41–43). Interestingly, this DC population is known to play a crucial role in cross-presentation and thus, in the initiation of CD8⁺ T-cell responses (23, 24). These data suggest that the blocking of ILDR2 not only has a direct effect on the initiation of T-cell responses in the lymph node but could also affect the ratio of intratumoral CD8α⁺ DCs. Interestingly, it has previously been reported that DCs can directly interact with FRCs through the binding of CLEC-2 to podoplanin, which could facilitate interactions of ILDR2 with receptors on DCs (44). On the basis of our data, our hypothesis on the mode of action of BAY 1905254 is that the blockade of the immunosuppressive function of ILDR2 in the lymph node results in enhanced T-cell priming. Together with anti–PD-L1, ILDR2 blockade further led to an increased number of CD8α⁺ DCs in the TME, which likely promoted antigen cross-presentation and thereby the subsequent CD8⁺ T-cell responses. ILDR2 blockade can drive immune activation and higher production of intratumoral IFNγ. Our hypothesis on the mode of action is depicted in Fig. 7.

In summary, BAY 1905254 blocked the immunosuppressive function of ILDR2 and could increase T-cell activation. BAY 1905254 exhibited antitumor efficacy in vivo both as a monotherapy as well as in combination with various other therapies currently in use. Preclinical in vivo data support the rationale of combining BAY 1905254 with an anti–PD1/PD-L1 for improved antitumor efficacy. Thus, our results suggest ILDR2 to be an immune checkpoint inhibitor and BAY 1905254 a candidate for cancer immunotherapy.

J. Huetter is a research scientist at and has ownership interest (including patents) in Bayer AG. I. Gutcher is a senior scientist at Bayer AG. W.-D. Doecke is a biomarker strategist at and has ownership interest (including patents) in Bayer AG. M.V. Luetke-Eversloh is a research scientist at Bayer AG. S. Golfier is a research scientist at Bayer AG. H.G. Roider is an employee at Bayer AG. J. Hunter is Chief Scientific Officer at and has ownership interest (including patents) in Compugen Ltd. Z. Levine is a senior vice president at Compugen Ltd. I. Barbiro is a scientist at Compugen Ltd. G. Cojocaru is Director, R&D Research at Compugen Ltd. I. Vaknin is Associate Director, Immune-Oncology Bioassays at Compugen Ltd. B. Kreft is Head of Immuno-Oncology Research at Bayer AG. L. Roese is a senior scientist at Bayer AG. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Huetter, U. Gritzan, J. Hunter, Z. Levine, O. Levy, I. Vaknin, B. Kreft, L. Roese

Development of methodology: J. Huetter, M.V. Luetke-Eversloh, A. Pow, O. Levy, L. Roese

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Huetter, U. Gritzan, I. Gutcher, W.-D. Doecke, M.V. Luetke-Eversloh, S. Golfier, A.-L. Frisk, A. Pow, A. Drake, M. Azulay, I. Barbiro, I. Vaknin, L. Roese

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Huetter, U. Gritzan, I. Gutcher, M.V. Luetke-Eversloh, S. Golfier, H.G. Roider, A. Pow, A. Drake, O. Levy, G. Cojocaru, L. Roese

Writing, review, and/or revision of the manuscript: J. Huetter, U. Gritzan, W.-D. Doecke, H.G. Roider, A. Drake, O. Levy, B. Kreft, L. Roese

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Huetter, L. Roese

Study supervision: J. Hunter, Z. Levine, I. Vaknin, L. Roese

The authors would like to gratefully thank Petra Maulwurf, Tanja Lehmann, Kathleen Busch, Natascha Manovski, and Daniela Eisenbeiser for outstanding technical assistance. The authors further acknowledge the Natural and Medical Sciences Institute (NMI) at the University of Tübingen for experimental support on the PrimeFlow analyses of human lymph node cells. Aurexel Life Sciences Ltd. (www.aurexel.com) is acknowledged for editorial support funded by Bayer AG.

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.