Recombinant mouse IFN-alpha15/A (Cat# 12100-9) and mouse IFN-beta (Cat#12410) were purchased from PBL Biosciences. Anti-mouse XCR1 mAb (clone MARX10), anti-human AGP3 peptide mAb (clone 4D2), anti-mouse CD11c mAb (clone N418) and Ab-IFN-alpha15/A (Ab-IFN) fusions were made on a mIgG1-SEFL1 N297A backbone with a G3G4S linker. Molecules were cloned into pTT5 expression vector and produced similar to a process described previously (27). The fusion proteins were purified from conditioned medium using MabSelect SuRe (Cytiva) and Size Exclusion Chromatography/Superdex 200 Hiload 26/600 (GE Healthcare), with a final formulation of HBSS (pH 7.6).

RAW cells that stably express XCR1 or CD11c were generated using retroviral transduction. Briefly, GP2-293 cells (Clontech #631458) were co-transfected with a murine stem cell virus (MSCV)-based retroviral vector encoding either mouse XCR1 (NM_011798.4) or mouse CD11c (NM_0213343.3) and a vector encoding the vesicular stomatitis virus (VSV) G viral envelope protein using Lipofectamine 3000 (Thermo Fisher Scientific #L3000015) according to the manufacturer’s specifications. Viral supernatant was collected 48 h post-transfection, filtered through a 0.45 µm filter, and then directly added to low passage target cells (RAW-Lucia™ ISG Cells, Invivogen) after the addition of polybrene (10 µg/mL final concentration, EMD Millipore #TR-1003-G). Transduced cells were “spinfected” for 90 min (1200g at 32°C) and allowed to recover for 48 h. RAW cells were enriched for purity and expression of the transduced gene by fluorescence-activated cell sorting (FACS) to > 95% in the final sorted population using a FACS ARIA sorter (BD Biosciences) after staining with Abs that detect mouse XCR1 (mouse IgG2βκ anti-mouse XCR1; Biolegend 148204; clone ZET) or mouse CD11c (Armenian hamster IgG anti-mouse CD11c; Biolegend 117330; clone N418) using 5 µl of labeled Ab in 100 µl of FACS staining buffer for 30 minutes on ice.

Aliquots of single cell suspensions of human dissociated ovarian tumors (serous carcinoma) were purchased from Conversant Biologics (Discovery Life Sciences) and pre-screened for cell viability (> 50%) and adequate detection of major immune cell subsets to avoid evaluating tumors that lack expression of surface markers due to a processing artifact. 2 ovarian samples were selected: OVAR1 (BTC1000-J6110002946111116MS, Stage IV) and OVAR2 (BTC1000-J6110002936091316MS, Stage III-C).

RAW cells were plated at 1x10⁵ cells per well in a non-TC treated U-bottom plate (Falcon 351177) and kept at 4°C. Cells were washed with Dulbecco’s phosphate-buffered saline (D-PBS) (Corning 21-031-CV) and stained with 100 μL of Fixable Viability Dye eFluor™ 780 (1:1000 dilution, Invitrogen 65-0865-18) for 30 min. Cells were again washed once with D-PBS and then blocked for 20 minutes with 100 μL of purified rat anti-mouse CD16/CD32 Mouse Fc Block (BD 553142; clone 2.4G2) at a concentration of 8.3 μg/mL. Cells were washed with staining media (PBS + 2% heat inactivated FBS + 0.05% bovine serum albumin). Cells were then stained for 30 min with 50 μl of Ab-IFN fusion proteins (starting concentration of 100 nM and 2-fold serial dilutions) or mAbs, which served as positive controls. Cells were then washed with staining media and then stained for 30 minutes with 100 uL of allophycocyanin conjugated anti-mouse IgG secondary Ab (1:500 dilution, Jackson ImmunoRes 715-136-151). Cells were then washed once with staining media and fixed in 100 μL of 4% PFA for 15 minutes at room temperature. Cells were then resuspended in staining media and analyzed on BD FACSymphony™.

RAW cell lines were used to measure activity of Ab-IFN fusion proteins using the IFN-inducible secreted luciferase reporter or flow cytometry to measure endogenous markers of myeloid activation (PD-L1, CD86). Cells were cultured in RPMI-1640 + GlutaMAX™ Medium (Gibco 61870036) supplemented with 10% (v/v) heat inactivated FBS (Gibco 10082147) and Penicillin-Streptomycin (Gibco 15140122). Cells were manually lifted and seeded in 100 μL in cell culture media at 30,000 cells per well in 96-well Clear Flat Bottom TC-treated Culture Microplate (Falcon 353072) and incubated (37°C, 5% CO2) for 16-24 h. Post incubation, cells were then stimulated with a dose titration of Ab-IFN fusion proteins. Fusion proteins were diluted in cell culture media with a starting concentration of 8000 nM (2-fold concentration) followed by 14 point 5-fold serial dilutions. Initial culture media from cells were aspirated, and 100 uL of cell culture media was added along with 100 μL of stimulation media. Cells were stimulated for 16-24 h in cell incubator (37°C, 5% CO2). Post stimulation, supernatant was moved to a 96-well V-bottom plate and spun down at 500 g, 4°C, for 5 min. 20 μL of supernatant was moved to a black 96-well clear bottom plate and 50 uL of reconstituted QUANTI-Luc™ (Invivogen rep-qlc1) was added. Plate was gently tapped to mix and immediately read for luminescence (Envision^®, Perkin Elmer).

In addition to supernatant collection, cells were dislodged for Ab staining for analysis by flow cytometry. 50 μL of 0.25% Trypsin-EDTA (Gibco, 25200056) was added to the stimulated cells and incubated at 37C, 5% CO2 for 3 minutes to dislodge cells. Post incubation of Trypsin-EDTA, 150 μL of cell culture media was added and gently pipetted to dislodge adherent cells. Cells were then moved to a 96-well non-TC treated U-bottom plate to stain with Ab for flow cytometry. Cells were washed with D-PBS (Corning 21-031-CV) and stained with 100 μL Fixable Viability Dye eFluor™ 780 (1:1000 dilution, Invitrogen 65-0865-18) in D-PBS. Cells were washed once with D-PBS and Fc blocked for 20 min with 50 μL of purified rat anti-mouse CD16/CD32 Mouse Fc Block (BD 553142) at a concentration of 16.6 μg/mL diluted in staining media (D-PBS, 2% HI FBS, 0.05% BSA). Anti-mouse CD274 (PD-L1) mAb (clone MIH5), PerCP-eFluor 710 conjugated (eBioscience 46-5982-80) was diluted in staining media at a dilution of 1:200 (2x) and 50 uL of staining mixture was added onto the cells in blocking solution for a final dilution of 1:400 (1x). Cells were stained for 30 min and subsequently washed once with staining media. Cells were then fixed in 100 μL of 4% PFA (room temperature) for 15 min. Cells were then resuspended in staining media and analyzed on BD FACSymphony™.

6 – 8 week old female C57BL/6 and BALB/c mice were purchased from Charles River Laboratories (Hollister, CA, United States). IFNAR1 knockout mice (B6(Cg)-Ifnar1^tm1.2Ees/J) from Jackson Laboratories (Bar Harbor, ME, United States) and BALB/c mice (CRL) were backcrossed at least 10 times to generate BALB/c-Ifnar1^tm1.2Ees mice. RAG2 knockout (129S6/SvEvTac-Rag2^tm1Fwa ) mice were purchased from Taconic. All experimental studies were conducted under protocols approved by the Institutional Animal Care and Use Committee of Amgen (IACUC). Animals were housed at Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International-accredited facilities (at Amgen) in ventilated micro-isolator housing on corncob bedding. All mice were maintained in pathogen-free conditions in a temperature-controlled environment with 12/12 hour light/dark cycles and received sterile pellet food and reverse osmosis-purified water ad libitum.

Pharmacokinetic characterization of recombinant IFN and Ab-IFN fusion proteins were tested following a single dose intravenous bolus administration in female C57BL/6 healthy mice. Blood for plasma preparation was collected at specified timepoints via submandibular vein puncture for each serial timepoint when applicable. The collected plasma specimens were stored at approximately -70°C until transferred for subsequent analysis.

Commercial or in-house developed immunoassays were used to measure the drug concentrations of the Ab-IFN fusion proteins, along with recombinant mouse IFN proteins (IFN-beta and IFN-alpha). To quantitate IFN-alpha and IFN-beta recombinant proteins (all from PBL), VeriKine™ Mouse IFN Alpha ELISA Kit (PBL) and VeriKine™ Mouse IFN Beta ELISA Kit (PBL) were used respectively. For mIgG1 (SEFL1) Ab-IFN-alpha15/A fusions (Ab-IFN^WT and Ab-IFN^A169D), anti-IFN-alpha mAb (clone RMMA-1, PBL) was used as a capture reagent and donkey anti-mouse IgG light chain mAb (20B07-02A09, Jackson ImmunoResearch) was used as a detection reagent. Of note, suitable immunoassays could not be developed to measure drug concentrations of Ab-IFN^L53A fusion proteins or the XCR1 parental mAb due to lack of adequate capture with the anti-IFN-alpha mAb or the mouse XCL1 ligand (data not shown).

Eight-week-old mice were injected subcutaneously on the lower right flank with cancer cell lines. 3e5 MC38 (original source unknown, not authenticated), CT26 (CRL-2639, ATCC), B16.F10 (CRL-6475, ATCC) cells were injected with a 30G insulin syringe, in 100-μL suspension, on the shaved flank of briefly sedated mice. On day 10, mice were randomized into treatment groups with tumor volumes averaging 100 mm³ or euthanized to harvest spleens and tumors to characterize Clec9a and XCR1expression in untreated mice. Tumor volumes were measured at least 2 times a week with a caliber using the formula, tumor volume = l (length) x w (width) x h (height).

Unless noted otherwise, Ab-IFN fusion proteins and/or mAbs were administered via intraperitoneal (IP) injection at a dose of 10 mg/kg (or 200 μg/mouse) starting the day of randomization. For pharmacodynamic studies to evaluate Ab-IFN fusion protein activity ex vivo, mice were dosed with either a single injection treatment and evaluated approximately 24 h later or a multi-dose treatment (3x per week on M, TH, M) and harvested 24 h following last treatment. For efficacy studies to evaluate anti-tumor activity and/or survival, mice were dosed every 3 – 4 days (2 times per week) until the end of study (tumor volume reached ≥1500 mm³).

Tumors and spleens were harvested, and single cell suspensions were prepared. In brief, tumors were mechanically minced into small fragments and placed into an enzymatical digestion buffer containing 0.2 mg/ml Liberase TL (Roche) and 20 μg/ml DNase I (Ambion), unless otherwise noted. Tumor samples were then homogenized using a gentle MACS Dissociator (Miltenyi Biotech) and incubated at 37°C for 15 minutes on a rotator at 200 rpm. Homogenized cell preparations were passed through a 70-µm cell strainer and washed with FACS wash buffer (PBS plus 2% FBS and 0.05% BSA). Spleens were manually dissociated and filtered through a 40-µm cell strainer. Red blood cells (RBCs) in spleen samples were lysed using 1 ml of RBC lysis buffer (eBioscience) at 4°C for 5 min. 1e6 single cells were loaded into 96-well plates, washed with D-PBS and kept at 4°C. Next, cells were stained with Live/Dead™ Fixable Blue Dead Cell Stain Kit (Invitrogen L23105, 1:1,000 dilution) for 30 min, followed by a wash in D-PBS, and then blocked with purified rat anti-mouse CD16/CD32 (Clone 2.4G2, BD 553142, 1:30 dilution in BD Horizon Brilliant Stain Buffer, BD 566349) for 20 min. Cell surface staining was done for 30 minutes at 4°C with an Ab cocktail (diluted in BD Horizon Brilliant Stain Buffer, BD 566349) washed twice in FACS wash buffer, fixed for 10 min at room temperature in 4% PFA, and then stored at 4°C before evaluating on the flow cytometer. Data acquisition was performed using an LSRII or FACSymphony™ flow cytometer (BD Biosciences) and analyzed using FlowJo software (Treestar). Immune cell subsets within the samples were evaluated by first generating scatter plots to select single cells (FSC-H vs FSCH-A and SSC-H vs SSCA-A), then live cells based on negative staining for the live/dead marker and finally positive staining for CD45. Within the CD45⁺ population, several cell-specific surface markers (Abs utilized are listed in Supplementary Methods, Table I ) were then used to evaluate immune cell subsets and immune activation.

The presence of anti-drug antibodies (ADA) was determined against anti-XCR1 mAb and Ab-IFN fusion proteins using a homogeneous bridging MSD assay. Test molecules were labeled with biotin (Thermo Fisher) and ruthenium (MSD) to be used as capture and detection reagents respectively in the MSD assay. To analyze study samples, they were combined with a mixture of the prepared capture and detection reagents, and then capture-ADA-detection complexes were formed and captured on a streptavidin-coated MSD plate (MSD). The plate was then read for electrochemiluminescence (ECL). Pretreated study samples and Group 1 samples were included in the assay as negative controls and plasma samples spiked with a goat anti-mIgG1 Ab (Jackson Immuno Research) at 1 μg/mL was included as a positive control. The presence of ADA was compared between treatment groups by measuring assay responses (ECL values).

Data analysis was performed using GraphPad Prism version 9.5.1 (GraphPad Software, San Diego, CA, United States). EC50 values in this study were determined after log10 transformation of molecule concentrations using a 4-parameters nonlinear fit analysis. Group/replicate sizes are indicated in figure captions.

To determine the surface molecule that is optimal to target cDC1s, we compared expression of Clec9a and XCR1, 2 receptors uniquely expressed on cDC1s, in human myeloid cell subsets from healthy donor peripheral mononuclear cells (PBMCs) and a small cohort of ovarian cancer tumors. Both XCR1 and Clec9a were expressed in the BDCA3⁺ cDC1 subset in PBMCs ( Figure 1A ) and did not show expression in any other immune cell subsets in whole blood ( Supplementary Figures 1A, B ). As expected, Clec9a was also not expressed in human pDCs ( Supplementary Figure 1C ). Of note, XCR1 was found to be expressed in a subset (~50%) of BDCA3⁺ cDC1s, while Clec9a was uniformly expressed in this population ( Figures 1A, B ). In ovarian tumors, XCR1 was expressed in the majority (~80%) of BDCA3⁺ cDC1s ( Figures 1C, D ), suggesting enrichment of XCR1⁺ cDC1s in tumors and possibly, an enrichment of a more mature cDC1 DC population since Clec9a and BDCA3 are reported to also be expressed in pre-DCs, while XCR1 was not (28). In contrast to XCR1, surface expression of Clec9a was undetected in the human dissociated tumor samples ( Figure 1D ), likely due to Clec9a sensitivity to clipping from enzymatic digestion based on loss of Clec9a expression in PBMCs treated with Collagenase ( Supplementary Figures 2A, B ).

Next, we evaluated XCR1 and Clec9a expression in mouse tissues using only mechanical digestion since mouse Clec9a was also found to be sensitive to enzymatic clipping ( Supplementary Figures 2C, D ), although XCR1 was not ( Supplementary Figures 2E, F ). In spleen, we evaluated XCR1 and Clec9a expression in most major immune cell subsets (B, T, and NK cells, neutrophils, monocytes, pDCs and CD11c⁺ DCs). As expected, XCR1 was only expressed in CD8⁺ cDC1s, while Clec9a was expressed in CD8⁺ cDC1s and pDCs ( Figure 2A ). In mouse tumors, XCR1 was undetected in the B/T/NK cells, neutrophils, or monocytes, but, importantly, was detected in CD103⁺ cDC1s ( Figure 2B ), establishing XCR1 as a good target for cDC1s in both human and mouse tumors. Unfortunately, Clec9a was not detected in any immune cell subsets in tumors, including CD103⁺ cDC1s ( Figure 2B ). While the mouse tumors were not enzymatically digested, we could not rule out loss of Clec9a detection due to an artifact of tissue processing since Clec9a is an endocytic receptor (29) and downregulation could also occur with release of β-actin from dying cells during tissue processing. Of interest, tumor cDC1s were approximately 90% positive for XCR1 ( Figure 2B ) and XCR1 appeared to have increased expression in cDC1s (by MFI) in tumors compared to spleen ( Figure 2C ) in both B16 and MC38 tumor-bearing mice ( Figure 2D ). Thus, based on these findings and the enrichment of XCR1 expression on tumor DCs compared to non-tumor DCs in humans and mice, XCR1 was selected as the cDC1 target.

To target type I IFN to XCR1⁺ cells, we generated anti-mouse XCR1 Ab fusions to mouse IFN-alpha isoform 15/A (referred to as XCR1Ab-IFN^WT fusion), which has medium to low activity compared to the other mouse IFN isoforms (30). In addition, XCR1 Ab fusions were generated with two IFN muteins, A169D and L53A (XCR1Ab-IFN^A169D and XCR1Ab-IFN^L53A) ( Figure 3A ), which were selected based on conserved amino acids between mouse and human of IFN-alpha and the low activity of these human IFN mutein counterparts (31). Untargeted IFN fusion proteins (referred to as isoAb-IFN^WT, isoAb-IFN^A169D, and isoAb-IFN^L53A) were also generated using a control Ab (anti-human AGP3 peptide Ab, Clone 4D2), which lacks binding in mice and served as the isotype control ( Figure 3A ). To screen the IFN activity of the Ab-IFN fusion proteins, RAW264.7 myeloid cells expressing an IFN-inducible luciferase reporter were generated to express murine XCR1, denoted RAW-XCR1 cells ( Figures 3B, C ). All Ab-IFN fusion proteins evaluated did not bind to RAW parental cells ( Figure 3D ), but XCR1Ab-IFN fusions and the XCR1 parental mAb bound to RAW-XCR1 cells with similar binding affinity ( Figure 3E ). In contrast, isoAb-IFN fusions lacked binding to RAW-XCR1 cells ( Figure 3E ).

Next, IFN activity of the fusion proteins was assessed using the luciferase reporter and cell surface expression of PD-L1, which is directly induced by type-I IFN signaling (32). Since RAW parental cells endogenously express the type I IFN-alpha heterodimer receptor (IFNAR1/2), the XCR1Ab-IFN^WT and the XCR1Ab-IFN^mut fusions showed similar activity to their matching non-targeted isoAb-IFN fusion proteins ( Figures 3F, H ), validating the IFN proteins had the same activity in the XCR1Ab-IFN and the isoAb-IFN fusion proteins. Of interest, both Ab-IFN^mut fusions (IFN^A169D and IFN^L53A) showed a clear reduction in IFN activity, both by EC50 and maximum (Emax) activity, compared to the Ab-IFN^WT fusions in RAW parental cells ( Figures 3F, H ), confirming the single point mutations in mouse IFN-alpha15 reduced IFN activity. In contrast, XCR1Ab-IFN fusions showed a clear gain in activity compared to their untargeted isoAb-IFN fusions in RAW-XCR1 cells ( Figures 3G, I ). The EC50s of XCR1Ab-IFN^WT, XCR1Ab-IFN^A169D and XCR1Ab-IFN^L53A fusion proteins were approximately 20-, 200- and 200-fold more active in RAW-XCR1 cells compared to RAW parental cells in the luciferase assay ( Figures 3G, J ) and approximately 30-, 5,000-, and 3,000-fold greater based on PD-L1 expression ( Figures 3I, J ). Notably, XCR1Ab-IFN^mut, A169D and L53A, showed a greater gain in potency (200 to 5,000-fold) in RAW-XCR1 cells compared to RAW parental cells, while the XCR1Ab-IFN^WT only showed a slight gain (20 to 30-fold) ( Figures 3G, I, J ), validating the potential to increase the therapeutic window using an Ab-targeted mutein approach.

After validating IFN targeting to XCR1+ cells was enhanced with the Ab-IFN fusions in vitro, we performed in vivo studies to measure pharmacokinetic/pharmacodynamic (PK/PD) indices. Of note, we selected the XCR1Ab-IFN^L53A mutein fusion and the corresponding isotype control as our proof-of-concept mutein fusions because these fusion proteins showed the most attenuated in vitro IFN activity (ie. based on EC50 and Emax activity), which we reasoned would result in the lowest untargeted IFN activity and toxicity in vivo. Immunoassays were developed to measure the drug concentration of the Ab-IFN^WT fusions and recombinant mouse IFN proteins (IFN-beta and IFN-alpha15), which revealed improved kinetics by increasing the half-life from 1.6 h for the recombinant IFN-alpha15 to approximate 60 h with the Ab fused IFN-alpha15 (Ab-IFN^WT) ( Figure 4A ). Because of insufficient binding of the capturing anti-IFN Ab to the Ab-IFN^L53A mutein fusions, we were unable to directly measure drug concentration for this mutein.

Subsequently, we developed ex vivo flow cytometry PD assays to measure receptor occupancy (RO) and IFN activity of XCR1Ab-IFN fusions in cDC1s in the spleen and tumors. Mice were given a single injection of 20 μg or 200 μg (1 or 10 mg/kg) Ab-IFN fusion proteins and tissues were harvested 24 h later ( Figure 4B ). A gating strategy that utilized the restricted expression of XCR1 in CD8⁺ cDC1s in the spleen and CD103⁺ cDC1s in the tumor was adopted to evaluate XCR1 RO and IFN activity in cDC1s following treatment ( Figure 4C ). Due to the lack in availability of a non-competing anti-mouse XCR1 mAb, a fluorescently conjugated competing XCR1 Ab was used to measure the proportion of unbound or “free” XCR1 receptors on cDC1s, such that loss of XCR1 detection reflected RO of the XCR1Ab-IFN^L53A fusion ( Figure 4D ). In the spleen, a significant loss of XCR1 detection in cDC1s was observed from mice treated with the XCR1Ab-IFN^L53A fusion compared to the untargeted isotype control ( Figure 4D ) and resulted in approximately 90% and 30% RO following a 10 mg/kg and 1 mg/kg dose ( Figure 4E ). Importantly, XCR1 RO correlated to a significant increase in IFN activity in cDC1s in the spleen, measured by PD-L1 and CD86 expression, from mice treated with the XCR1Ab-IFN fusion compared to isotype control ( Figure 4F ), suggesting efficient delivery of IFN muteins to XCR1⁺ cDC1s. Moreover, a dose of 10 mg/kg isotype control fusion had slightly less CD86 expression than a dose of 1 mg/kg XCR1Ab-IFN^L53A fusion suggesting an approximate 10-fold gain in IFN activity in cDC1s with XCR1-targeting in vivo. In the tumor, XCR1 detection was also decreased in mice treated with the high dose of the XCR1Ab-IFN^L53A fusion (10 mg/kg) compared to the isotype control ( Figures 4D, E ), which correlated with a significant gain in IFN activity in cDC1s, based on greater PD-L1 and CD86 expression ( Figure 4F ). As expected, XCR1-negative, Ly6G⁺ neutrophils did not show a significant gain in IFN activity with the XCR1Ab-IFN^L53A fusion compared to the isotype control ( Figure 4G ), validating the specificity of the XCR1Ab-IFN^mut to increase IFN activity in cDC1s. Of note, although XCR1 RO was observed in the tumor and correlated with increased targeted IFN activity in tumor cDC1s compared to isotype control, XCR1 RO and IFN activity induced by the isotype control was much more prominent in the spleen compared to the tumor, likely due to increased drug exposure in the spleen compared to tumor.

Next, we adapted a multiple dose regimen over the course of one week ( Figure 5A ) to induce continued IFN signaling in cDC1s with the goal of priming cDC1s and eliciting robust anti-tumor T cell responses. While our goal was to assess T cell infiltration using this regimen, we first repeated our ex vivo PD assays to confirm XCR1 RO and IFN activity was maintained. Unexpected, we did not observe significant XCR1 RO ( Figure 5B ), or a gain in IFN activity in cDC1-targeted cells in the spleen or tumor ( Figure 5C ), following 3 injections with the XCR1Ab-IFN^L53A fusion. In addition, in stark contrast to our results with a single treatment, IFN activity in untargeted Ly6G⁺ cells in the spleen was significantly higher with the isotype control compared to the XCR1Ab-IFN^L53A fusion ( Figure 5D ), suggesting a loss of drug activity and/or exposure occurs with the XCR1Ab-IFN^L53A fusion. Assessment of tumor volumes ( Supplementary Figures 3A, B ) at the time of tissue harvest from this PD study and from a parallel in vivo study designed to evaluate anti-tumor efficacy and survival ( Supplementary Figures 3C–F ) revealed no significance between the XCR1Ab-IFN^L53A and isoAb-IFN^L53A fusion treatments, further suggesting multiple injections with the XCR1Ab-IFN^L53A fusion was not maintaining a gain in cDC1 activation.

One common explanation for loss of drug exposure is immunogenicity and the generation of anti-drug antibodies (ADA). We first evaluated whether the XCR1 mAb itself may be immunogenic and repeated our single and multi-treatment regimen in WT mice to evaluate RO ( Figure 6A ). Unlike the XCR1Ab-IFN^L53A fusion, the XCR1 mAb showed similar RO in the spleen following either a single or multiple treatments in WT mice from both Balb/c and C57Bl/6 backgrounds ( Figure 6B ), eliminating any strain-specific immunogenicity to the XCR1 parental Ab. Next, we hypothesized that perhaps the IFN^L53A mutein itself may have weak immunogenicity and targeting to cDC1s directly might enhance immunogenicity, leading to induced ADA production that was specific to this mutein. To test this hypothesis, we devised an in vivo experiment to assess whether other XCR1Ab-IFN fusion proteins, XCR1Ab-IFN^A169D and XCR1Ab-IFN^WT, may also result in loss of drug exposure following multiple dosing and used the XCR1 mAb as the control ( Figure 6C ). Again, we used our XCR1 RO assay to measure drug coverage. Similar to the XCR1Ab-IFN^L53A fusion, a clear loss in RO of the XCR1Ab-IFN^WT and XCR1Ab-IFN^A169D was observed following multiple injections compared to a single injection, yet the XCR1 mAb maintained similar RO with both treatments ( Figure 6D ). These data suggest that the XCR1Ab-IFN^WT and XCR1Ab-IFN^A169D fusions were also losing drug exposure following multiple injections while the XCR1 mAb was not. Of note, the XCR1Ab-IFN^WT appeared to show a more significant loss in RO (ie. more free XCR1 was detected) than the XCR1Ab-IFN^A169D ( Figure 6D ), suggesting IFN activity might be driving the loss of drug exposure. To explore whether IFN activity was driving the loss of drug exposure of the XCR1Ab-IFN fusions, a similar in vivo experiment was performed using IFNAR1 knockout (IFNAR1-ko) mice, this time with the addition of the isotype control fusion ( Figure 6E ). In contrast to WT mice, we did not observe a loss of XCR1 RO with the XCR1Ab-IFN^A169D with multiple injections compared to a single injection in IFNAR1-ko mice ( Figure 6F ). Hence, the enhanced IFN activity in cDC1s with the XCR1Ab-IFN fusions likely plays a key role in the specific loss of drug exposure of these fusions. In summary, we found XCR1Ab-IFN fusions, but not isoAb-IFN fusions or the XCR1 mAb, were driving a robust loss of drug exposure following multiple injections in just one week.

Next, we sought to determine whether the loss of drug activity of the XCR1Ab-IFN fusion proteins was the result of an immunogenic response and the development of ADA. In addition, we added a new treatment group in our study to assess whether the combination of the XCR1 mAb and the untargeted isoAb-IFN fusion could also drive this response or if it was a unique response to the XCR1Ab-IFN fusion proteins ( Figure 7A ). We found that multiple injections with the XCR1Ab-IFN^L53A fusion ( Figure 7B ), but not any other fusion proteins or combinations ( Figures 7C, D ), resulted in ADA production. Thus, the robust loss of drug activity, loss of drug detection, and ADA response appeared unique to treatment with the XCR1Ab-IFN fusion proteins and suggest IFN activation of cDC1s is driving the ADA response.

To explore whether targeting IFN to a broader group of DCs, may also induce a robust ADA response or provide an alternative approach to target IFN to cDC1s, we generated a panel of CD11cAb-IFN fusion proteins, similar to the XCR1Ab-IFN fusions, and validated their ability to target and enhance IFN activity in vitro in CD11c⁺ RAW (RAW-CD11c) ( Supplementary Figures 4A–G ). For CD11c-targeted IFN studies, we selected the CD11cAb-IFN^A169D fusion, not CD11cAb-IFN^L53A, to move forward in vivo because it showed the greatest gain in activity in CD11c⁺ RAW cells in vitro.

To determine the optimal dose for the CD11cAb-IFN^A169D fusion to enhance IFN activity in cDC1s in vivo, we first performed a single injection dose response experiment ( Figure 8A ). Notably, CD11cAb-IFN^A169D was able to bind with specificity to cDC1s in spleen and tumor resulting in 100% CD11c RO in the spleen and ~70% CD11c RO in the tumor following a single injection with just 10 μg (0.5 mg/kg) ( Figure 8B ). In addition, CD11cAb-IFN^A169D induced a potent gain in IFN activity in cDC1s compared to isotype control in the spleen and the tumor at the same dose of 10 μg ( Figure 8C ). Of interest, CD11cAb-IFN^A169D also induced a potent gain in IFN activity in cDC1s in the spleen at the lower dose of 1 μg, which did not correlate to CD11c RO suggesting the IFN activity assay may be more sensitive and/or little CD11c RO is required to induce IFN activity in CD11c⁺ cells. As expected, CD11cAb-IFN^A169D and the isotype control fusion showed similar activity in untargeted Ly6G⁺ neutrophils ( Figure 8D ).

Next, we utilized our multiple injection regimen to determine whether targeting IFN to the less specific cDC1 marker, CD11c, would also result in loss of drug activity ( Figure 8E ). Of note, we also increased the concentration of the doses to anchor data to results from previous studies with the XCR1Ab-IFN fusion proteins. Multiple injections of the CD11cAb-IFN^A169D fusion resulted in robust loss of drug exposure as measured by lack of CD11c RO at the 20 μg dose ( Figure 8F ). In addition, loss of CD11c-targeted IFN activity in cDC1s and untargeted IFN activity in Ly6G⁺ neutrophils compared to isotype control fusions was observed at the 20 and 200 μg doses ( Figures 8G, H ). Of interest, CD11c RO was not lost at the highest dose of 200 μg, suggesting perhaps loss of drug activity could be rescued by increasing the dose; however, direct assessment of IFN activity in cells showed activity was dramatically lost even though drug binding was still observed ( Figures 8F–H ). In contrast to the CD11cAb-IFN^A169D fusion, the untargeted isoAb-IFN^A169D fusion showed a clear dose response increase in IFN activity in cDC1s in both the spleen and tumor. Likewise, assessment of tumor volumes ( Supplementary Figures 5A, B ) from this PD study and from a parallel in vivo efficacy study ( Supplementary Figures 5C–F ) revealed treatment with the isoAb-IFN^A169D fusion also led to a significant dose response decrease in tumor volumes and a significant increase in survival. Notably, treatment with the isoAb-IFN^A169D fusion, which lacked loss of drug exposure with the multiple injection regimen, showed superior IFN activity in cDC1s, anti-tumor activity and survival compared to the CD11cAb-IFN^A169D fusion ( Supplementary Figures 5E, F ). Thus, these results clearly illustrate the significant negative impact on in vivo activity due to the loss of drug exposure of the CD11cAb-IFN^A169D fusion.

To determine whether the loss in activity of the CD11cAb-IFN fusions correlated to any significant changes in PK of our fusion proteins and to further validate the role of ADA, we designed an in vivo study to compare drug concentrations of the Ab-IFN fusions in WT mice ( Figure 8I ) and RAG2 knockout mice that lacked Ab production ( Figure 8K ). In WT mice, only the drug concentration of the isoAb-IFN fusion was detected ( Figure 8J ). In contrast, both the CD11c-targeted IFN fusions and the isotype control IFN fusions were detected in RAG2 knockout mice ( Figure 8L ), illustrating the unique ability of type I IFN targeted to cDC1s to drive a robust humoral response and an undesired ADA response.

Here we report a role of cDC1s in the induction of ADA. While cDC1s are well established to be efficient at cross-presenting and priming CD8⁺ T cells against cell-associated antigens, cDC1s DCs are generally regarded to have low CD4⁺ T cell activation compared to cDC2 DCs (33). However, a study using an anti-Clec9a Ab fusion with OVA (Clec9a-Ab-OVA) revealed cDC1s can be potent inducers of both CD8⁺ and CD4⁺ T cell expansion leading to a robust humoral response against both the OVA antigen and the Clec9a Ab (34). In this study, specific targeting of cDC1s with the Clec9a-Ab-OVA fusion led to a persistent in vivo expansion of OVA-specific CD8⁺ T cells, CD4⁺ T cells and CD4⁺ T follicular helper cells, revealing that targeting cDC1s can induce both a CD8⁺ T cell mediated response and a CD4⁺ T cell mediated humoral response. Furthermore, systemic addition of CpG as an adjuvant enhanced the cDC1-mediated CD8⁺ T cell expansion and the humoral response to the molecules. Thus, this study suggests cDC1-targeted Abs can induce an ADA response due to their biological role in stimulating CD4⁺ T cells and the humoral response. In addition, in the context of tumor immunity, a couple recent studies have also shown cDC1s to have an essential dual role in enhancing both CD8⁺ and CD4⁺ T cell activation and suggested the cDC1 activation of CD8⁺ T cells can not be separated from the dual role in activating CD4⁺ T cells (35, 36). Of interest, Ferris et al. (35) reveals the important role of cDC1s in anti-tumor immunity is their unique function in antigen processing and priming both CD4⁺ and CD8⁺ T cells and in orchestrating their cross-talk. Thus, shedding a new light on the role of cDC1s in tumor immunity and providing evidence that cDC1-targeted Abs may enhance ADA induction due to their dual role in stimulating both CD8⁺ and CD4⁺ T cells.

ADA development is a major hurdle for cancer immunotherapy (37, 38). An evaluation of the FDA’s clinical pharmacology data for 121 biological products (approved prior to 2015) reported 89% incidence of ADA, 60% impact on safety, and 49% indicated an impact on efficacy (39). There are several reasons why a biotherapeutic can induce ADA such as drug properties (e.g., non-human sequence, glycosylation, impurities, aggregation), drug pharmacokinetics, drug mechanism of action and individual patient characteristics (e.g., disease type, genetic factors, concomitant immunomodulators; probably the hardest to discern) (40). Of note, while earlier studies suggested ADA induction was due to inclusion of murine, non-human, sequences, which led the Ab field to focus on developing fully human Abs, it’s now clear that ADA induction is much more complex and many fully humanized Abs also induce ADA (41).

Our results suggest biotherapeutics that stimulate cDC1 (and/or CD11c⁺ DC) activation and function may be prone to induction and/or enhancement of ADA given the biology of cDC1s. Currently, there are no approved DC-targeted Ab therapies that directly stimulate DC antigen presentation to determine the clinical translation of our studies; however, two Abs (GSK2618966 and bococizumab) that failed clinical development due to ADA were recently found to have an unanticipated role in DC activation and maturation (42, 43). GSK2618966, a humanized Fc-disabled immunoglobulin G1 (IgG1) blocking Ab to the interleukin-7 receptor (IL-7R) was developed to block IL-7 signaling in T cells for the treatment of numerous autoimmune diseases (RA, type I diabetes, MS, SLE and primary Sjogren syndrome), but was discontinued during Phase I clinical studies due to its high immunogenicity (92% of subjects) and the high incidence of neutralizing Abs (42). The IL-7R, a heterodimer consisting of the IL-7Rα chain and the common γ chain, is a potent inducer of T cell proliferation, activation, and survival (44) and its expression is restricted to T cells. However, the IL-7Rα chain can also form a heterodimer with thymic stromal lymphoprotein receptor (TSLPR), which functions on myeloid cells and, importantly, plays a role in DC maturation and activation (45). Due to the enhanced immunogenicity of GSK2618966 and the role of TSLPR in activating DCs (45), GSK2618966 binding and activity in DCs was recently explored. GSK2618966 was found to bind monocyte-derived DCs and potently activate DC maturation and T cell proliferation, revealing an unexpected target-mediated immunogenicity due to the biological role of DCs (42). Another striking example is the clinical development of bococizumab, a humanized blocking Ab to proprotein convertase subtilisin/kexin type 9 (PCSK9), designed to lower low density lipoprotein cholesterol (LDL-C) for the treatment of atherosclerotic cardiovascular disease. Bococizumab, after 6 clinical trials, showed an unexpected attenuation of clinical efficacy over time and high immunogenicity, 48% developed ADA and 29% developed neutralizing ADA (46). While the mechanism for the immunogenicity of bococizumab is not well understood, recent studies suggest that the MOA of PCSK9 inhibitors is more complex than their role in lowering LDL-C and that a second, immune-mediated, mechanism may also be in play that is dependent on PCSK9 signaling in DCs (43, 47, 48). Of interest, bococizumab was found to activate CD4⁺ T cells in a PBMC assay developed to assess Ab immunogenicity and, importantly, addition of a blocking pan HLA-2 Ab resulted in loss of CD4⁺ T cell activation, suggesting bococizumab’s immunogenicity is dependent on antigen presentation by DCs (49). Thus, providing additional evidence that biotherapeutics that stimulate DC maturation (intentionally or unintentionally) may have a higher likelihood to induce ADA given their biological role in antigen presentation.

Given the significant impact in patients when ADA develop and the high cost of clinical trials, there is a need to develop ADA assays early during preclinical development and, of utmost significance, to continue monitoring ADA and effect on PK over time during clinical trials and post-marketing. For bococizumab, the phase II study only identified about 7% of ADA incidence, rather than the 48% in phase III, and if the ADA issue was properly identified in phase II, the time and cost of running 8 phase III trials might have been avoided (50). One of the reasons for the disconnect between the phase II and phase III studies was a lack of standard assays across studies, which made it difficult to summarize ADA levels. The recent development of in vitro assays to assess activation of CD4⁺ T cells and DCs to predict Ab immunogenicity using established controls (49, 51, 52) may provide the ability to standardize assays and prevent the development of highly immunogenic therapies from moving into the clinic.