CD40 Enhances Type I Interferon Responses Downstream of CD47 Blockade, Bridging Innate and Adaptive Immunity.

Disrupting the binding of CD47 to SIRPα has emerged as a promising immunotherapeutic strategy for advanced cancers by potentiating antibody dependent cellular phagocytosis (ADCP) of targeted antibodies. Pre-clinically, CD47/SIRPα blockade induces antitumor activity by increasing the phagocytosis of tumor cells by macrophages and enhancing the cross-presentation of tumor antigens to CD8+ T cells by dendritic cells; both of these processes are potentiated by CD40 signaling. Here we generated a novel, two-sided fusion protein incorporating the extra cellular domains of SIRPα and CD40L, adjoined by a central Fc domain, termed SIRPα-Fc-CD40L. SIRPα-Fc-CD40L bound CD47 and CD40 with high affinity and activated CD40 signaling in the absence of Fc receptor cross-linking. No evidence of hemolysis, hemagglutination or thrombocytopenia was observed in vitro or in cynomolgus macaques. Murine SIRPα-Fc-CD40L outperformed CD47 blocking and CD40 agonist antibodies in murine CT26 tumor models and synergized with immune checkpoint blockade of PD1 and CTLA4. SIRPα-Fc-CD40L activated a type I interferon response in macrophages and potentiated the activity of ADCP-competent targeted antibodies both in vitro and in vivo. These data illustrated that whereas CD47/SIRPα inhibition could potentiate tumor cell phagocytosis, CD40-mediated activation of a type I interferon response provided a bridge between macrophage and T cell mediated immunity that significantly enhanced durable tumor control and rejection.


Introduction
Tumor immunogenicity results from the processing, and subsequent presentation of tumor antigens released from dead or dying tumor cells on MHC molecules expressed on antigen-presenting cells (APC). The CD47/SIRPa pathway is an endogenous pathway that facilitates clearance of aged or dying cells (1). CD47 is expressed by many somatic and hematopoietic tissues and prevents red blood cell (RBC) and platelet destruction by macrophages and splenic CD4 þ dendritic cells (DC; refs. [2][3][4][5][6]. The antiphagocytic activity of CD47/ SIRPa serves as a macrophage checkpoint or "do not eat me" signal (1). The prophagocytic "eat me" signal that triggers RBC destruction by splenic macrophages is dependent upon a second "activating" signal, which includes CD18-containing integrins and "labeling" of susceptible cells by calreticulin (5,7).
Many tumors have increased expression of CD47 suggesting a coopted protective mechanism against immune-mediated destruction. Early studies hypothesized that the role of CD47 as a macrophage "do not eat me" signal might explain the observed antitumor benefit following treatment with CD47 blocking antibodies or SIRPa-Fc fusion proteins (8)(9)(10). Other studies extended these observations to show that SIRPa signaling in CD8a þ DCs enhances cGAS/STINGmediated sensing of phagocytosed tumor mitochondrial DNA, pro-moting a type I IFN response that facilitates cross-presentation of tumor antigens to CD8 þ T cells (11,12). Increased priming of CD8 þ T cells in the presence of CD47/SIRPa inhibition dramatically enhanced tumor rejection in multiple preclinical tumor models (11,12), demonstrating that the CD47/SIRPa axis is capable of bridging innate and adaptive immunity.
CD8a þ DCs expressing the transcription factor BATF3 are essential for antitumor immunity (13) and have the unique ability to crosspresent exogenous tumor antigens, which gain entry to the cytosol of DCs following endocytosis, to CD8 þ T cells. CD40 ligation by CD40 ligand (CD40L), expressed by resting CD8a þ (but not CD8a À ) DCs, enhances the antigen cross-presenting activity of DCs (14)(15)(16)(17)(18). Interestingly, whereas CD40 ligation facilitates a type I IFN response via STING activation, STING activation is not essential for the antitumor immune response to CD40 stimulation (19,20). However, antitumor immunity to CD40 agonists is dependent on BATF3 þ DCs and CD8 þ T cells (19). These data indicate that, like CD47/SIRPa, the CD40/CD40L axis is capable of bridging innate and adaptive immunity; however, the two pathways appear to have distinct dependence upon a type I IFN response.
The nonoverlapping roles of CD47/SIRPa and CD40/CD40L in bridging innate and adaptive cancer immunotherapy suggests that these two pathways could be complimentary or synergistic in combination. SIRPa-Fc fusion proteins, CD40L-Fc fusion proteins, and CD40 agonist antibodies have been successfully used as therapeutics (21)(22)(23)(24)(25)(26)(27)(28). The extracellular domains (ECD) of PD-1 and OX40L can be adjoined via an Fc domain (PD-1-Fc-OX40L), leading to superior therapeutic properties compared with anti-PD-1 and anti-OX40 combinations (29). As a result, we sought to develop a SIRPa-Fc-CD40L agonist redirected checkpoint (ARC) fusion protein to approach these distinct pathways with a single biologic. Given the costimulatory role of CD40 in augmenting the antigen cross-presenting ability of DCs and also the phagocytic activity of macrophages, we hypothesized that pairing SIRPa with CD40L may enhance antitumor immunity (30).
Here, we developed and characterized a SIRPa-Fc-CD40L for both humans and mice. Human SIRPa-Fc-CD40L engaged CD47 and CD40 simultaneously and with high affinity, causing activation of primary human T cells in vitro by increasing cytokine production, proliferation, and activation of NFkB and NIK signaling. In mouse models, murine SIRPa-Fc-CD40L provided higher rates of tumor rejection and long-term immunity compared with CD47 blocking antibodies, CD40 agonist antibodies, or the combination of the two. The mechanism of action of SIRPa-Fc-CD40L was attributed to enhanced activation of DCs in vivo, upregulation of a type I IFN response, and potentiation of macrophage-mediated phagocytosis in vitro. Taken together, CD40 synergistically enhanced the antitumor activity of CD47/SIRPa blockade by upregulating the type I IFN response in macrophages.

Construct generation and protein purification
The coding sequences of both human (h) and mouse (m) SIRPa-Fc-CD40L were codon-optimized and directionally cloned into the pcDNA3.4 TOPO TA (Thermo Fisher Scientific, catalog no. A14697; Supplementary Table S1) mammalian expression vector and nucleotide sequences were verified by Sanger sequencing methodology (performed off site by GENEWIZ). For transient transfection-based protein production runs, human or mouse SIRPa-Fc-CD40Lexpressing vectors were transfected into Expi293 cells using the ExpiFectamine 293 transfection kit (Thermo Fisher Scientific, catalog no. A14524) and cell culture media containing SIRPa-Fc-CD40L was harvested on day 6 posttransfection. In addition, the hSIRPa-Fc-CD40L vector was stably transfected into CHO cells and a final clone expressing high levels of hSIRPa-Fc-CD40L was selected for large scale protein production and purification (Selexis SA). Briefly, SIRPa-Fc-CD40L containing cell culture media was harvested by centrifugation at 5,000 rpm for 10 minutes followed by filtration through a 0.2-mm filter. The clarified supernatant was bound to a HighTrap Protein A HP column, washed and eluted under standard conditions. The eluted protein was dialyzed into 1Â PBS and the concentration was determined by absorption at 280 nm using a NanoDrop spectrophotometer (Thermo Fisher Scientific).

Electron microscopy
Samples were diluted in PBS (30-to 140-fold) and imaged over a layer of continuous carbon supported by nitrocellulose on a 400-mesh copper grid. Electron microscopy was performed using an FEI Tecnai T12 electron microscope operating at 120 keV equipped with an FEI Eagle 4k Â 4k CCD camera (NanoImaging Services). Negative stain grids were transferred into the electron microscope using a room temperature stage. Images of each grid were acquired at multiple scales to assess the overall distribution of the specimen. After identifying suitable target areas for imaging, high magnification images were acquired at nominal magnifications of 110,000Â and 67,000Â. The images were acquired at a nominal underfocus of À1.6 mmol/L to À0.9 mmol/L and electron doses of approximately 25 e À /A 2 . Two-dimensional averaging analysis was performed using automated picking protocols followed by reference-free alignment based on the XMIPP processing package (31,32).

Functional ELISA
For characterization of the h/m SIRPa-Fc-CD40L ARCs, highbinding ELISA plates (Corning) were coated overnight at 4 C with 5 mg/mL of either anti-hFc, anti-mFc, recombinant hCD47-Fc, hCD40-His, mCD47-Fc or mCD40-His, in PBS (reagents were obtained from Jackson ImmunoResearch Laboratories, Sino Biologics, Inc., Acro-Biosystems, and R&D Systems). Plates were then blocked with casein buffer for 1 hour at room temperature and then probed with serial dilutions of the h/m SIRPa-Fc-CD40L, along with the appropriate standards (human and mouse; IgG, SIRPa-Fc, and Fc-CD40L) for 1 hour at room temperature. Plates were washed with TBS-T (Trisbuffered saline containing 0.1% Tween 20) and then detection antibody was added for 1 hour at room temperature in the dark. Detection antibodies included anti-hIgG-HRP, anti-mIgG-HRP, anti-hCD40L, anti-mCD40L, and then anti-Goat-HRP (all antibodies were obtained from Jackson ImmunoResearch Laboratories or R&D Systems). Plates were washed again and SureBlue TMB Microwell Peroxidase Substrate (KPL; purchased from VWR, catalog no. 95059-282) was added to each well and allowed to incubate at room temperature in the dark. To stop the reaction, 1N sulfuric acid was added to each well and absorbance at 450 nm was read immediately on a BioTek plate reader. Samples were run at a minimum in triplicate and at multiple dilutions.
For SIRPa-blocking ELISA, hCD47-Fc (Sino Biological Inc.) was used to coat a high binding ELISA plate as described above. The following day, binding was detected using a recombinant hSIRPa-Biotin protein (Acro Biosystems) in combination with increasing concentrations of either hSIRPa-Fc-CD40L (to compete with SIRPa-Biotin binding for CD47) or an anti-CD47 blocking antibody (clone CC2C6, BioLegend), which served as a positive control. Following this incubation, plates were washed as described above and probed with an avidin-HRP detection antibody (BioLegend) for 1 hour at room temperature in the dark. Plates were then washed and analyzed as above.

Surface plasmon resonance
Direct binding of h/m SIRPa-Fc-CD40L fusion protein to recombinant protein targets was performed using a Bio-Rad ProteOn XPR36 protein interaction array instrument. To determine the on-rates (K a ), off-rates (K d ), and binding affinities (KD) of hSIRPa-Fc-CD40L to its intended binding targets (referred to as "Ligands"), histidine-tagged versions of the human recombinant targets-CD47 (AcroBiosystems), CD40 (AcroBiosystems), FcgR1A (Sino Biological Inc.), FcgR2B (Sino Biological Inc.), FcgR3B (Sino Biological Inc.), and FcRn (R&D Systems) were immobilized to a Ni-sulfate activated ProteOn HTG sensor chip (Bio-Rad). Increasing concentrations of the hSIRPa-Fc-CD40L fusion protein (referred to as "Analyte") diluted in PBS/Tween (0.005%) buffer pH 7.0 were injected for 3 minutes followed by a dissociation phase of 5 minutes at a flow rate of 80 mL/minute. hSIRPa-Fc and hCD40L-Fc (AcroBiosystems) recombinant proteins were used as positive control analytes for binding to their partners (CD47 and CD40, respectively) and human IgG was used as a positive control for binding to Fcg receptors and FcRn. To assess analyte binding to FcRn ligand, the pH of the buffer was reduced to pH 5.5.

Cell culture
CHO-K1, CT26, A20, WEHI3, Toledo, Raji, Ramos, A431, HCC827, K562, HCC1954, and MCF7 cells were obtained from ATCC (between 2017 and 2019) and cultured according to their guidelines; maintained at 37 C in 5% CO 2 . All parental cell lines in active culture are tested monthly using the Venor GeM Mycoplasma Detection Kit (Sigma). All transfected cell lines are tested an additional two times, separated by at least 2 weeks, posttransfection, and confirmed to remain negative for Mycoplasma. Research cell banks (RCB) of all purchased cell lines were generated within 5 passages of initial cell line thawing. All cells used for experimentation were used within 10 passages of vial thawing. Expression from receptor/ligand overexpressing cells lines was verified by flow cytometry prior to generating the RCB, and then periodically checked again when new vials were thawed and put into culture.

Flow cytometry
Briefly, isolated cells were washed one time with 1Â PBS, followed by centrifugation at 400 Â g for 5 minutes. Cells were then stained with antibodies for 30 minutes on ice in the dark. Indicated antibodies were purchased from Sony, BioLegend, or Abcam, and used at their recommended concentrations, and diluted in FACS buffer [1Â PBS buffer containing 1% BSA, 0.02% sodium azide, and 2 mmol/L EDTA]. The AH1-tetramer reagent was purchased from MBL International (catalog no. TB-M521-2), and was incubated with cells for 1 hour on ice in the dark before adding the rest of the antibody cocktail. Following this incubation period, stained cells were washed by adding 0.5 mL of FACS buffer, and then centrifuged at 400 Â g for 5 minutes. The supernatant above the pellet was then aspirated and the resulting cells were resuspended in FACS buffer and flow cytometry was performed on a BD LSRII Fortessa according to manufacturer's recommendations, and data were analyzed by FlowJo.
In vitro functional assays Phagocytosis assay Frozen vials of peripheral blood mononuclear cells (PBMC) from healthy human donors were purchased from STEMCELL Technologies and monocytes were isolated using a commercial kit according to the manufacturer's protocol (STEMCELL, catalog no. 19059) and were confirmed to be CD14 þ by flow cytometry postisolation using an APCconjugated anti-human CD14 (BioLegend, catalog no. 367118) and the appropriate isotype control antibody. Following the isolation of monocytes, a live cell count was performed using Trypan blue staining 4 Â 10 5 live and monocytes were seeded in Iscove modified Dulbecco media (IMDM) containing 10% FBS and 50 ng/mL of human macrophage colony-stimulating factor (M-CSF; BioLegend) in 24-well plates (Corning). Cells were incubated in 37 C/5% CO 2 for 7 days to differentiate into macrophages. Media were changed on day 4 and supplemented with M-CSF. On day 7, macrophages were polarized to an M1 state by replacing the media with fresh media containing 10 ng/mL lipopolysaccharide (LPS-EB; InvivoGen) and 50 ng/mL human IFNg (STEMCELL Technologies). Cells were incubated for another 48 hours and M1 polarization status was confirmed by flow cytometry for CD11b, HLA-DR, CD80, and CD206 (BioLegend). Macrophages were confirmed to be M1 polarized if they stained positive for CD11b, HLA-DR, and CD80 and negative for CD206. On day 9, the media containing LPS and IFNg was replaced with fresh media (IMDM þ 10% FBS), and macrophage:tumor cocultures were initiated.
Flow cytometry-based analysis: Tumor cells were harvested and incubated with SIRPa-Fc-CD40L either alone at 1 mmol/L test concentration or in combination with select antibody-dependent cellular phagocytosis (ADCP)-competent antibodies (based on the type of tumor cell used in the assay) at 0.06 mmol/L test concentration for 37 C for 1 hour. Following the incubation, cells were washed with Dulbecco PBS (DPBS) and stained with a CellTracker Green CMFDA dye (Thermo Fisher Scientific/Invitrogen, catalog no. C2925) according to the manufacturer guidelines. Cells were washed twice with DPBS and resuspended in serum-free media. The stained tumor cells were cocultured with M1-polarized macrophages for 2 hours at 37 C at a tumor:macrophage ratio of 5:1. When appropriate, Fc receptors and CD40 receptors were blocked on macrophages using commercial antibodies (20 mg/mL; BioLegend and R&D Systems, respectively) and calreticulin was blocked on tumor cells using a calreticulin blocking peptide (10 mg/mL; MBL International Corporation) for 1 hour at 37 C prior to coculture initiation. After the coculture incubation period, unbound/unengulfed tumor cells were removed and the macrophages were rinsed in DPBS and harvested using a nonenzymatic cell stripping solution (Thermo Fisher Scientific/ Corning, catalog no. 25-056-C1). Harvested macrophage Fc receptors were blocked using a commercial human Fc blocking reagent (Bio-Legend). Thereafter, the Fc-receptor blocked macrophages were stained with an anti-human CD11b antibody conjugated to PE/Cy7 (BioLegend), washed, and analyzed on an LSRII Fortessa flow cytometer (BD Biosciences) to detect macrophages that stained positive for both CD11b and the CellTracker Green CMFDA dye following a phagocytic event. Cells were pregated on CD11b þ macrophages. Samples were analyzed on a LSRII Fortessa and flow cytometry standard (FCS) files were analyzed using the FlowJo software version 10. The percentage of CD11b þ Green dye þ macrophages were plotted in GraphPad Prism 8. A phagocytosis index was calculated by setting the maximum response across the treatment groups to 1 and calculating the fold change for each of the treatment groups relative to the maximum response. A similar flow-based methodology was used to study phagocytosis of mouse tumor cells (WEHI or A20) by bone marrow-derived macrophages (BMDM) with mouse SIRPa-Fc-CD40L or mouse anti-CD47 (BioXCell, clone MIAP301). BMDMs were isolated and grown in IMDM with 10% FBS supplemented with mouse M-CSF (50 ng/mL) for 7 days, and activated with LPS (10 ng/mL) and mIFNg (50 ng/mL) for 2 days.
Fluorescence microscopy-based analysis: Human monocytes (CD14 þ ) were plated in 16-well chamber glass slides (Thermo Fisher Scientific) to differentiate into macrophages and were polarized to an M1 state using LPS and IFNg as described above. On the day of the experiment, tumor cells (Toledo) were prepared and cocultured with the M1 macrophages in the chamber slides in the presence of SIRPa-Fc-CD40L either alone or in combination with select ADCPcompetent antibodies for 2 hours at 37 C. After the incubation period the unbound/unengulfed tumor cells were removed from the wells and the macrophages were rinsed four times in 1Â DPBS. Subsequent blocking and staining steps were performed in the chamber slide. The Fc receptors on macrophages were blocked as described previously and subsequently stained with an anti-human CD11b antibody conjugated to Alexa Fluor 594 (BioLegend) overnight at 4 C protected from light. Macrophages were rinsed twice in DPBS to remove unbound stain and coverslips were mounted in ProLong Diamond anti-fade mountant containing DAPI (Invitrogen) and allowed to dry overnight at room temperature protected from light. Once dry, the edges of the coverslips were sealed with clear nail polish and fluorescent imaging was performed on a Zeiss 800 confocal microscope. Four representative images were acquired per well in the chamber slide under 10Â and 60Â magnifications and ImageJ software (NIH, Bethesda, MD) was used to process the raw images and quantify the number of engulfed tumor cells (labeled green) per total macrophages (labeled red) based on the pixel density. The compiled data included one representative image per condition and the quantification data incorporated at least four images per well and two wells per condition.
Time-lapse microscopy-based analysis: Human monocytes (CD14 þ ) were plated in a 96-well plate (Millipore Sigma), differentiated into macrophages, and polarized to an M1 state using LPS and IFNg as described above. On the day of the experiment, Toledo tumor cells were labeled with the IncuCyte phRodo Red cell labeling kit (Sartorius, catalog no. 4649) according to the manufacturer's guidelines and then were cocultured with the M1 macrophages at a 5:1 ratio in the presence of SIRPa-Fc-CD40L (1 mmol/L), anti-CD20 (ritux-imab; 1 mg/mL), anti-CD47 (BioLegend clone CC2C6; 33 mg/mL); or the combination of anti-CD20 with either SIRPa-Fc-CD40L or anti-CD47. The 96-well plate was placed in the IncuyCyte S3 and time-lapse imaging was performed over the course of 5 hours at 37 C. Five images per well, in duplicate, using two distinct donors were acquired under 10Â magnification for each treatment group and the red object integrated intensity per image was calculated using the IncuCyte S3 software. The phagocytosis index was calculated by taking the maximum signal observed from each experiment and setting that as the 100% point. Each data point from a given experiment was normalized to the maximum value for that experiment (i.e., fluorescence or luminescence). These calculations were repeated for each experimental replicate, and then all combined data was plotted using GraphPad Prism 8.
In vivo DC activation assay BALB/C mice were obtained from Jackson Laboratories, and treated with a single intravenous injection of 1 Â 10 7 PBS washed sheep red blood cells (RBC diluted in PBS; Rockland Immunochemicals), anti-CD47 (clone MIAP301), anti-SIRPa (clone P84), or mSIRPa-Fc-CD40L. Antibodies were given at a total dose of 100 mg and the ARC at 300 mg; all in a volume of 100 mL diluted in PBS. After 6 or 24 hours, mice were humanely euthanized and splenocytes were isolated for flow cytometry-based immune profiling of DCs, by removing the spleens from treated animals, mechanically dissociating them using the flat ends of two razors followed by repeated pipetting to break up larger pieces, and then passing the cells through a 100-mm cell strainer. RBCs were lysed with RBC lysis buffer according to manufacturer's recommendations (BioLegend, catalog no. 420301) and the remaining mononuclear cells were stained with fluorescent-conjugated antibodies to CD11c, DC1R2, I-A b (MHC-II), and either CD4 or CD8. After 30 minutes on ice in the dark, cells were washed, and then analyzed on an LSR II Fortessa flow cytometer. CD4 þ or CD8 þ DCs were pregated on CD11c þ /DC1R2 þ populations, and activated DCs were also positive for I-A b . Additional pharmacodynamic activity was assessed in the peripheral blood, 24 hours following single intraperitoneal injections of a dose titration of mSIRPa-Fc-CD40L. A small amount of peripheral blood was collected from the tail and RBCs were lysed as described above. Populations of CD20 þ cells were assessed by flow cytometry.
NFkB-Luciferase reporter assay CHOK1/mCD40/NFkB-luciferase cells were treated with a dose titration of mSIRPa-Fc-CD40L, recombinant Fc-mCD40L protein (Sino Biologics), or anti-mCD40 (clone FGK4.5; BioXcell). CHOK1/hCD40/NFkB-luciferase cells, were treated with a dose titration of SIRPa-Fc-CD40L and recombinant hCD40L-His protein (Sino Biologics). Bright-Glo luciferase reagent and a GloMax Navigator luminometer were used to assess activation of NFkB signaling. The expression vector, luciferase reagent (catalog no. E2650), and luminometer (catalog no. GM2000) are all from Promega and were used according to manufacturer suggestions. Briefly, 10,000 CHO-K1/ CD40/NFkB-luciferase cells were plated in each well of a 96-well plate AE the ARC or other test agents. Plates were incubated at 37 C/5% CO 2 for 6 hours, then the Bright-Glo reagent was added and luminescence was assessed on the luminometer.

NIK/NFkB reporter assay
U2OS/NIK/NFkB reporter cells expressing CD40 were purchased from Eurofins/DiscoverX (catalog no. 93-1059C3) and cultured according to their recommendations. On the day of the assay, 10,000 U2OS/NIK/NFkB reporter cells were plated into each well of a 96-well plate with a dose titration of either SIRPa-Fc-CD40L, recombinant Fc-hCD40L protein (Sino Biologics), or an anti-CD40 agonist antibody (clone HB14; BioLegend). After 6 hours in culture, luminescence activity was assessed on a luminometer (Promega) as described above.

AIMV activation assay
Proliferation: PBMCs were isolated from 50 healthy donor buffy coats and CD8 þ T cells were depleted using CD8 þ RosetteSep (STEMCELL Technologies). Cell depletion was confirmed by flow cytometry and was >90% for all donors. Cells were plated at a density of 4-6 million cells per mL in AIM-V medium (Gibco). On days 5, 6, and 7, cells were gently resuspended in 3 Â 100 mL aliquots and transferred to each well of a round-bottom 96-well plate. Cultures were pulsed with 0.75 mCi [3H]-Thymidine (PerkinElmer) in 100 mL AIM-V culture medium, and incubated for a further 18 hours, before harvesting onto filter mats (PerkinElmer), using a TomTec Mach III cell harvester. Counts per minute (cpm) for each well were determined by Meltilex (PerkinElmer) scintillation counting on a 1450 Microbeta Wallac Trilux Liquid Scintillation Counter (PerkinElmer), with low background counting.
IL2 ELISpot: PBMCs were isolated, CD8 cells depleted, and cultured in AIM-V as described above, and on day 8, cells were assessed. Briefly, ELISpot plates (Millipore) were prewetted and coated overnight with 100 mL/well IL2 capture antibody (R&D Systems) in PBS. Cells were plated at a density of 4-6 million cells/mL in a volume of 100 mL per well, in sextuplicate. After 8 days, ELISpot plates were developed by sequential washing in dH 2 O and PBS (Â3), prior to the addition of 100 mL filtered biotinylated detection antibody (R&D Systems). Following incubation at 37 C for 1.5 hours, plates were further washed with PBS and 1% BSA, and incubated with 100 mL filtered streptavidin-AP (R&D Systems) for 1 hour; and then washed again. One-hundred microliters BCIP/NBT substrate (R&D Systems) was added to each well for 30 minutes at room temperature. Spot development was stopped by washing wells three times with dH 2 O. Dried plates were scanned on an Immunoscan Analyzer and spots per well (SPW) were determined using Immunoscan Version 5 software. Samples included human SIRPa-Fc-CD40L (at 0.3, 3, 30, and 300 nmol/L), the neoantigen KLH (Keyhole limpet hemocyanin; 300 nmol/L) as a positive control, and exenatide (Bydureon, 20 mmol/L) as a clinical benchmark control; representing a negative control in this assay. For ELISpot, a mitogen-positive control (PHA at 8 mg/mL) was included on each plate as an internal test for ELISpot function and cell viability.

Type I IFN response Gene expression analysis
Human macrophage:Toledo/Raji tumor cocultures were set up as described above. After 2 hours of coculture and treatment with either SIRPa-Fc-CD40L (1 mmol/L), anti-human CD20/rituximab (0.06 mmol/L), or the combination of both agents; cells were collected from the plate using a nonenzymatic cell stripping solution (Thermo Fisher Scientific/Corning, catalog no. 25-056-C1), washed with PBS, and stained with a fluorescently conjugated antibody for CD11b, after Fc blocking. After a 30-minute incubation on ice in the dark, cells were washed and CD11b þ populations were sorted using a FACS Melody (BD Biosciences). Sorted macrophages were lysed with RLT lysis buffer (Qiagen RNeasy Micro Kit, catalog no. 74004) containing 5% 2-mercaptoethanol and processed according to manufacturer's direc-tions; including on-column DNase I digestion. RNA was quantitated using a Nanodrop and 250 ng of resulting RNA was reverse transcribed (Origene First-strand cDNA Synthesis kit) according to manufacturer's recommendations. The resulting cDNA was diluted with nuclease-free water and the equivalent of 10 ng of starting RNA served as the template for each qPCR reaction. Gene expression was assessed using SYBR Green and the CFX96 Touch Real-Time PCR Detection System (Bio-Rad). Validated qPCR primers for IFNa1, IFNb1, CD80, CD86 and b-Actin (ACTB) as a housekeeping control, were purchased from Origene. Fold change in gene expression was determined using the DDC t method compared with untreated samples. Data generated were from a minimum of three experimental replicates all run in at least triplicate. Error bars represent SEM and significance was determined using one-way ANOVA.

Tumor model systems
For CT26, A20, and WEHI3 studies, BALB/C mice were subcutaneously implanted with 5 Â 10 5 (CT26 and A20) or 1 Â 10 6 (WEHI3) tumor cells in 100 mL of PBS, into the rear flank, respectively, on day 0. On treatment days (treatment schematics shown in figures and described in figure legends), tumor bearing mice were randomized and either untreated or treated via intraperitoneal injection with the mSIRPa-Fc-CD40L ARC, or anti-CD40 (clone FGK4.5), anti-CD47 (clone MIAP301), anti-CD20 (clone AISB12), anti-PD-1 (clone RMP1-14), or anti-CTLA4 (clone 9D9). All test agents were diluted in PBS and injected in volumes between 100 and 200 mL. All therapeutic antibodies are from BioXcell. Tumor volume (mm 3 ) and overall survival was assessed throughout the time-course. Survival criteria included total tumor volume less than 1,700 mm 3 with no sign of tumor ulceration. Complete responders, in which tumors established and were subsequently rejected, are listed in the appropriate figures. Cohorts of CT26 experimental mice were euthanized between days 11 and 13 for immune profiling in splenocytes and tumor tissue using flow cytometry. Tumors were excised from these mice and dissociated using a tumor dissociation kit (Miltenyi Biotec, catalog no. 130-096-730). Tumors were excised, combined with the dissociation reagent, and then minced with a razor. The resulting slurry was transferred to a 1.5 mL Eppendorf tube and placed in a 37 C shaker for 30 minutes. During this period, the slurry was pipetted up and down every 10 minutes to break up larger pieces. Dissociated cells were homogenized through a 100-mm strainer to isolate tumor cells and infiltrating immune cells. Experimental group sizes are described in each figure legend and come from a minimum of two independent experiments.
For CD4, CD8, and IFNAR1 depletion experiments, mice were treated via intraperitoneal injection of 100 mg of anti-CD4 (clone GK1.5), 100 mg of anti-CD8 (clone 2.43), or 500 mg of anti-IFNAR1 (clone MAR1-5A3) on the schedules described in the appropriate figures (all antibodies are from BioXcell). Antibodies were diluted in de Silva et al. PBS and injected in volumes of 100 mL. CD4, CD8, and IFNAR1 populations in the peripheral blood were assessed at several time points to verify depletion and were normalized to untreated mice. Peripheral blood from WEHI3-bearing mice treated on days 7, 9, and 11 with 300 mg SIRPa-Fc-CD40L AE anti-IFNAR1 was collected, the RBCs lysed, and the resulting mononuclear cells (MNC) assessed by flow cytometry with fluorescently conjugated antibodies to CD45, CD11c, CD20, CD4, and CD8 (antibodies from BioLegend).

Safety studies Nonhuman primate studies
Na€ ve cynomolgus macaques (2-4 years of age) were given SIRPa-Fc-CD40L by intravenous infusion every week, for 5 consecutive weeks, at doses of 0.1, 1, 10, and 40 mg/kg. Hematology and clinical chemistry parameters were collected by venipuncture before and after each dose. CD47 expression on circulating RBCs was assessed by flow cytometry. MNCs and serum were isolated from whole blood using Ficoll gradient separation. The resulting erythroid pellet was stained with anti-CD47 and pregated using forward/side scatter to isolate a majority RBC population. All studies were conducted at Charles River Laboratories in accordance with Institutional Animal Care and Use Committee (IACUC) guidelines.

Experimental animal guidelines
All animal studies have been conducted in accordance with, and with the approval of an IACUC and reviewed and approved by a licensed veterinarian. Experimental mice were monitored daily and euthanized by CO 2 asphyxiation and cervical dislocation prior to any signs of distress.

Statistical analysis
GraphPad Prism was used to plot and generate all graphs throughout; as well as automatically calculate error and significance. Experimental replicates (N) are shown in figures and figure legends. Unless noted otherwise, values plotted represent the mean from a minimum of three distinct experiments and error is SEM. Statistical significance (P value) was determined using unpaired t tests or one-way ANOVA with multiple comparisons. Significant P values are labeled with one or more " Ã ", denoting Ã , P < 0.05; ÃÃ , P < 0.01; ÃÃÃ , P < 0.001; and ÃÃÃÃ , P < 0.0001. Mantel-Cox statistical tests were used to determine the significance between the survival curves. P values are noted in the legends to these figures.

Production and characterization of SIRPa-Fc-CD40L
The ECDs of SIRPa and CD40L were fused via an antibody Fc domain for both human and mouse to generate SIRPa ECD -Fc-CD40L ECD ; hereafter referred to as SIRPa-Fc-CD40L. In silico structural modeling predicted that each individual domain of the adjoined construct would fold in accordance with the native molecules, suggesting preservation of both binding functions (Fig. 1A). Purified SIRPa-Fc-CD40L was then analyzed for the presence of each individual domain by Western blotting using anti-SIRPa, anti-Fc, and anti-CD40L (Fig. 1B), revealing a glycosylated protein that formed a dimer under nonreducing conditions at the predicted monomeric molecular weight of 88.1 kDa. To further characterize the native state of SIRPa-Fc-CD40L in the absence of detergents, electron microscopy was performed, demonstrating that the major peak fraction contained a hexameric species (!60% for both mouse and human), consistent to what had been previously described for TNF ligand fusion proteins ( Fig. 1C; ref. 33). A minor peak fraction ($30%-35%) was also present, which comprised a tetrameric protein complex that had equivalent activity in the dual-binding ELISA (hexamer EC 50 ¼ 24.21 nmol/L, tetramer EC 50 ¼ 33.3 nmol/L, reference ¼ 18.1 nmol/L). A dualbinding ELISA assay was developed to quantitatively demonstrate simultaneous binding of SIRPa to recombinant CD47 and CD40L to recombinant CD40 (Fig. 1D). Individual ELISAs also confirmed binding to recombinant human CD47, Fc, and CD40 ( Supplementary  Fig. S1A). Binding affinity studies using surface plasmon resonance (SPR; Supplementary Table S2) indicated that human SIRPa-Fc-CD40L bound with 0.628 nmol/L affinity to recombinant human CD47, 4.74 nmol/L affinity to recombinant human CD40, had undetectable binding to FcgR1a, -2b and -3b, while having preserved 2.33 nmol/L binding affinity to FcRn (Fig. 1E). To justify testing of human SIRPa-Fc-CD40L in nonhuman primates, high-affinity binding to recombinant cynomolgus macaque CD40 (3.24 nmol/L) and CD47 (1.7 nmol/L) were also confirmed. Finally, to confirm that SIRPa-Fc-CD40L interacted with native CD47 and CD40 in a similar manner to recombinant CD47 and CD40, CHOK1-hCD47, and CHOK1-hCD40 reporter cell lines were developed (Supplementary Fig. S1B and S1C). Flow cytometry studies using these reporter cell lines confirmed that SIRPa-Fc-CD40L bound to CHOK1-CD47 cells (31.85 nmol/L EC 50 ) and CHOK1-CD40 cells (22.48 nmol/L EC 50 ), but not to parental CHOK1 cells ( Fig. 1E; Supplementary Fig. S1B and S1C). A functional ELISA demonstrated that SIRPa-Fc-CD40L outcompeted a commercially available single-sided SIRPa-Fc control for binding to recombinant CD47, generating an EC 50 of 22 nmol/L, comparable with the 14 nmol/L EC 50 produced by a commercial CD47-blocking antibody (Fig. 1F).
The mouse surrogate of SIRPa-Fc-CD40L (referred to as mSIRPa-Fc-CD40L) was generated to assess activity in vivo, and was first characterized similarly to the human ARC; including by Western blot analysis ( Supplementary Fig. S1D) and dual-binding ELISA to both mouse CD47 and mouse CD40 simultaneously ( Supplementary  Fig. S1E).

SIRPa-Fc-CD40L functional activity-SIRPa domain
Because of the bifunctionality of SIRPa-Fc-CD40L, distinct functional assays were utilized to independently characterize the activity of both the SIRPa and CD40L domains. A common in vitro assay used to characterize the SIRPa/CD47 axis analyzes the ability of purified macrophages to phagocytose tumor cells (2,34), particularly in the presence of ADCP-competent therapeutic antibodies (35). Accordingly, in vitro tumor cell phagocytosis assays were established to determine whether SIRPa-Fc-CD40L enhanced macrophagemediated phagocytosis of various tumor cell lines both alone and in combination with targeted antibodies. Initially, several CD20 þ lymphoma (Toledo, Raji, Ramos) cell lines were cocultured with human monocyte-derived macrophages to identify a suitable platform for assessing phagocytosis; using three analogous approaches; (i) immunofluorescence (IF); to visualize the overlap in signal between labeled macrophages and labeled tumor cells, (ii) flow cytometry; to quantitate distinct populations of dually-positive tumors and macrophages, and (iii) live-cell imaging using the IncuCyte platform; to both visualize and quantitate macrophage-mediated phagocytosis of tumor cellsusing tumor cells that were precoated with a pH-sensitive marker (pHrodo) that only fluoresces when a tumor cell is engulfed by a macrophage and enters a low pH (pH 4.5-5.5) phagosome ( Fig. 2A-E;  Supplementary Fig. S2A). All three approaches demonstrate that SIRPa-Fc-CD40L was capable of inducing phagocytosis as a monotherapy, and that this activity was significantly enhanced in CD20 þ lymphoma cells lines when combined with the ADCP-competent anti-CD20 agent, rituximab ( Fig. 2A-D; Supplementary Fig. S2A). The phagocytosis-stimulating activity of SIRPa-Fc-CD40L/rituximab combinations was partially inhibited when Fc receptors were blocked on macrophages; however, activity was not impacted when macrophages were pretreated with a CD40-blocking antibody indicating that binding of ARC to CD40 does not contribute to phagocytosis of tumor cells.
Calreticulin on tumor cells serves as a prophagocytic signal, facilitating tumor cell phagocytosis following blockade of the CD47/SIRPa pathway (36). We demonstrated that both calreticulin and Fc receptor engagement was required for efficient phagocytosis of CD20 þ B-cell lymphoma cells by the combination of SIRPa-Fc-CD40L and rituximab using a calreticulin-blocking peptide, confirming the importance of Fc interactions for Fc-competent targeting antibodies and providing evidence that the initiation of phagocytosis by SIRPa-Fc-  CD40L was driven by the SIRPa domain (Fig. 2C). We observed similar monotherapy phagocytosis activity of SIRPa-Fc-CD40L to the CC2C6 clone of anti-CD47 (Fig. 2D). When both agents were combined with rituximab, the SIRPa-Fc-CD40L/rituximab combination stimulated significantly greater phagocytosis activity than the anti-CD47/rituximab combination suggesting competition by two separate therapeutics that require Fc receptor engagement. We examined the phagocytic activity in a range of human tumor cell lines using several ADCP-targeted antibodies; EGFR þ melanoma (A431 cells) and lung (HCC827 cells), EGFR À chronic myelogenous leuke-mia (CML; K562 cells), and HER2 þ breast (HCC1954 HER2 HI and MCF7 HER2 LOW cells) were used to facilitate combinations with EGFR-(cetuximab) and HER2-(trastuzumab) targeted antibodies. Consistent with the lymphoma cell lines, monotherapy SIRPa-Fc-CD40Lstimulated macrophage phagocytosis, which was enhanced in combination with targeted antibodies (Fig. 2E; Supplementary Fig. S2B). Trastuzumab did not induce phagocytosis in the HER2 LOW cell line MCF7; however, SIRPa-Fc-CD40L demonstrated modest monotherapy activity. We did not observe phagocytic activity of EGFR-negative K562 cells with monotherapy or combinations of SIRPa-Fc-CD40L  with cetuximab (Fig. 2E). Similar phagocytic activity was observed using the mouse surrogate mSIRPa-Fc-CD40L to treat cocultures of BMDMs and either A20 or WEHI3 cells (Supplementary Fig. S1F). Interestingly, a higher phagocytosis index was generated using mSIRPa-Fc-CD40L, as compared with a mouse CD47-blocking antibody.
Finally, using an in vivo mouse assay that examines the activation status of splenic DCs in response to SIRPa/CD47 inhibitors or sheep RBCs (5), we observed that intravenous administration of sheep RBCs, CD47 blocking antibodies, or SIRPa blocking antibodies all stimulated upregulation of both activated CD4 þ and CD8a þ DCs that were positive for MHC-II within 6 hours ( Fig. 2F; Supplementary Fig. S2D-S2F). Similarly, administration of mouse SIRPa-Fc-CD40L also upregulated surface expression of MHC-II, CD80, and CD86 on splenic CD4 þ and CD8a þ in a higher proportion of overall splenic DCs than was observed in the antibody-treated groups.

SIRPa-Fc-CD40L functional activity-CD40L domain
On the basis of the reported role of NIK signaling for CD40dependent cross-priming, functionality of the CD40L domain of SIRPa-Fc-CD40L was evaluated using two different CD40-dependent NFkB/NIK reporter systems ( Fig. 3A and B; refs. 37,38). These data indicated that SIRPa-Fc-CD40L had similar activity to a single-sided CD40L fusion protein in both reporter systems. SIRPa-Fc-CD40L was present in a soluble form in both assays, and no Fc receptor or other cross-linking agents were present. Along these lines, a CD40 agonist antibody was unable to stimulate NIK/NFkB activity in the same system in the absence of an accessory cell that can provide Fc receptor engagement (Fig. 3B). These data indicated that SIRPa-Fc-CD40L can stimulate CD40 signaling in the absence of cross-linking, likely due to its inherent hexameric configuration. Similar observations were made using mSIRPa-Fc-CD40L in comparison with a mouse CD40 agonist antibody ( Supplementary Fig. S1G).
The observation that SIRPa-Fc-CD40L stimulated CD40 signaling prompted investigation of other cellular functions that depend on CD40 signaling. CD40 stimulates proliferation of B cells and CD4 þ T cells from human PBMCs in the presence of cross-linked anti-CD40 antibodies or CD40L (39,40). To investigate this readout, CD8 þ T-cell-depleted PBMCs were isolated from a total of 33-50 different human donors and cultured in the presence of a dose titration of SIRPa-Fc-CD40L ( Fig. 3C and D; Supplementary Fig. S3A). As compared with a media-only negative control and a KLH-positive control, soluble SIRPa-Fc-CD40L stimulated dose-dependent proliferation of human PBMCs over a 7-day culture (Fig. 3C), and a dosedependent increase in the number of IL2-secreting PBMC on day 8 of the culture (Fig. 3D), previously reported as a downstream event of CD40 activation (41).
The activation status of macrophages from SIRPa-Fc-CD40Ltreated macrophage:Toledo lymphoma cell cocultures was assessed by qRT-PCR for expression of the type I IFN-regulatory genes IFNa1 and IFNb1, and the macrophage activation markers CD80 and CD86 (Fig. 3E). Monotherapy SIRPa-Fc-CD40L and rituximab (anti-CD20) induced macrophage activation and the expression of type I IFN genes in the isolated macrophages, which was enhanced when the two agents were combined. Similar results were observed in macrophages isolated after coculture with Raji cells in the presence of SIRPa-Fc-CD40L, rituximab, or the combination (Supplementary Fig. S3B).
A macrophage reporter system (RAW264.7 ISG) was utilized to determine activation of a type I IFN response by mSIRPa-Fc-CD40L. When RAW264.7 ISG cells were cocultured with A20 lymphoma cells, mSIRPa-Fc-CD40L stimulated an increase in IFN gene-driven luciferase activity (Fig. 3F). Commercially available recombinant murine Fc-CD40L was also able to stimulate IFN production; however, no significant signal was observed using a recombinant single-sided SIRPa-Fc protein indicating that type I IFN-activation acted downstream and independently from tumor cell phagocytosis, through CD40 engagement (Fig. 3F). A murine rituximab surrogate (anti-CD20) induced some monotherapy IFN response and significantly amplified the monotherapy signal seen with mSIRPa-Fc-CD40L (Fig. 3F) providing additional rationale for the combination of SIRPa-Fc-CD40L with targeted ADCP-competent antibodies. Such combinations may have had the ability to increase tumor cell phagocytosis and initiate pathways capable of activating APC and enhancing antigen processing/presentation. These data indicated that the magnitude of the type I IFN response could be enhanced when SIRPa and CD40L were physically linked to one another.

Antitumor activity of murine SIRPa-Fc-CD40L
The syngeneic CT26 colon tumor model was used to provide an initial assessment of the antitumor activity of murine SIRPa-Fc-CD40L in comparison with CD40 agonist and CD47-blocking antibodies. Implanted CT26 tumors were grown to approximately 30 mm 3 before treatment was initiated with a fixed regimen of two doses of either CD40 agonist antibody (clone FGK4.5), CD47-blocking antibody (clone MIAP301), a combination of both antibodies or murine SIRPa-Fc-CD40L. As compared with vehicle controls, both CD40 agonist and CD47-blocking antibodies provided moderate extensions in tumor growth, with 25% of mice completely rejecting primary tumors in the CD40 agonist monotherapy group (Fig. 4A). Mice treated with a combination of CD40 and CD47 antibodies were observed to have a longer delay in tumor outgrowth and 33% of mice rejected tumors. In comparison with the antibody groups, complete tumor rejection was observed in 50% of the mice treated with mSIRPa-Fc-CD40L, along with significant tumor growth delay and prolonged survival (Mantel-Cox test, P ¼ 0.0364 vs. anti-CD40/anti-CD47 combination group) in the remaining mice. A majority of the mice that rejected the primary tumor rejected a secondary tumor challenge in the absence of additional treatment (60%; Fig. 4A). In both the antibody combination and mSIRPa-Fc-CD40L-treated groups, there was an increase in the proportion of AH1-tetramer (specific for the MHC H-2L d -restricted immunodominant epitope of gp70 expressed by the CT26 tumor cell line) CD8 þ T cells in both the tumor and spleen ( Fig. 4B; Supplementary Fig. S4A). To determine whether these T-cell responses contributed to therapeutic efficacy, we repeated these studies in CD4 þ and CD8 þ T-cell antibody-depleted mice ( Fig. 4C; Supplementary Figs. S3A and  S4B-S4D). CD4 þ T cells were partially required for therapeutic efficacy, whereas the loss of CD8 þ cells completely eliminated the therapeutic benefit of mSIRPa-Fc-CD40L, similar to CD47blocking antibodies (11,12). CD8 depletion following the initiation of mSIRPa-Fc-CD40L therapy partially abrogated the observed antitumor efficacy (Supplementary Fig. S4D).
CD47-blocking antibodies plus rituximab potentiate macrophagemediated phagocytosis, correlating to the combination's promising clinical efficacy in patients with late-stage diffuse large B-cell lymphoma (42,43). Because SIRPa-Fc-CD40L potentiated the activity of rituximab, we investigated the combination of SIRPa-Fc-CD40L with a murine surrogate for rituximab (anti-mouse CD20; clone AISB12) in two CD20 þ mouse tumor models, WEHI3 and A20. In both tumor models, we observed similar control of established tumor growth when anti-CD20 antibodies or mSIRPa-Fc-CD40L were tested as (1 mmol/L), or the combination of both agents. F, RAW 264.7-Lucia ISG cells were cocultured with A20 lymphoma cells in the presence of 50 mg/mL of mSIRPa-Fc-CD40L, recombinant Fc-mCD40L, mSIRPa-Fc, or their combination or 1 mg/mL anti-mCD20, or the combination of mSIRPa-Fc-CD40L and anti-mCD20. After 24 hours, type I IFN-induced luminescence was assessed using a luminometer. The maximum signal of luminescence across all experiments was set to 1, and all other values were normalized accordingly. monotherapies ( Fig. 4D and E). We observed additive control of tumor growth when anti-CD20 antibodies and mSIRPa-Fc-CD40L were combined as compared with anti-CD20 monotherapy in A20 (P ¼ 0.0302) and WEHI (P ¼ 0.0166) or mSIRPa-Fc-CD40L monotherapy in A20 (P ¼ 0.0017) and WEHI (0.0095). Next, we sought to explore our previous in vitro findings implicating SIRPa-Fc-CD40L in type I IFN activation in an in vivo setting and assessed the impact of an IFNa receptor 1 (IFNAR1) blocking antibody on antitumor efficacy. Antibody-mediated blockade of IFNAR1 significantly reduced the efficacy of mSIRPa-Fc-CD40L both alone and in combination with anti-CD20 in mice bearing established WEHI3 tumors ( Fig. 4D and E;  Supplementary Fig. S4C and S4E). IFNAR1 blockade most significantly impacted mSIRPa-Fc-CD40L-treated groups, with less effect on anti-CD20 monotherapy-treated mice. Consistent with this observation, tumor control was similar between anti-CD20 monotherapy and in combination with mSIRPa-Fc-CD40L; indicating that a majority of the combinatorial benefit required a functional type I IFN response. When depletion of IFNAR1 was initiated following initial treatment with mSIRPa-Fc-CD40L, there was only marginal (WEHI3) or no (A20) acceleration in tumor growth, indicating that Antitumor efficacy of the murine SIRPa-Fc-CD40L surrogate. A, CT26 tumor growth curves of mice treated with two doses of vehicle (PBS; n ¼ 21), anti-CD40 (clone FGK4.5; n ¼ 8), anti-CD47 (clone MIAP301; n ¼ 8), the combination of both antibodies (n ¼ 9; 100 mg per antibody, per dose), or mSIRPa-Fc-CD40L (150-300 mg per dose; n ¼ 10). STV stands for "starting tumor volume" on the day that treatment began. B, The percentage of AH1 antigen-specific CD8 þ T cells in the spleens and tumors using tetramer reagents. C, CT26 tumor growth curves in mice predepleted of CD4, CD8, or both CD4 and CD8 cells prior to the initiation of treatment with vehicle (PBS; n ¼ 10) or mSIRPa-Fc-CD40L ARC (n ¼ 10; 300 mg per dose) on days 7, 9, and 11. Tumor growth curves of WEHI3 (D) or A20 tumors (E) in mice treated with vehicle (PBS; n ¼ 11 mice in each tumor model), anti-CD20 (100 mg per dose; n ¼ 9 in A20 and n ¼ 10 in WEHI3), mSIRPa-Fc-CD40L (300 mg per dose; n ¼ 11 in A20 and n ¼ 10 in WEHI3), or the combination of mSIRPa-Fc-CD40L and anti-CD20 (n ¼ 10 in A20 and n ¼ 9 in WEHI3) on days 7, 9, and 11 (WEHI3) or days 10, 12, and 14 (A20); when tumors were established and reached approximately 65 to 70 mm 3 . Also shown are tumor growth curves in mice predepleted of IFNAR1 cells prior to the initiation of treatment. Groups included anti-CD20 (n ¼ 10 in both tumor models), mSIRPa-Fc-CD40L (n ¼ 12 in A20 and n ¼ 11 in WEHI3), or the combination of the two agents (n ¼ 10 in A20 and n ¼ 8 in WEHI3).  Fig. S4E).
As the efficacy of SIRPa-Fc-CD40L was immune-mediated, we hypothesized that CTLA4-and PD-1-blocking antibodies may improve the antitumor effects of SIRPa-Fc-CD40L. To study these combinations, CT26 tumors grown to approximately 89 mm 3 were treated with a fixed regimen of three doses of mSIRPa-Fc-CD40L combined with three doses of CTLA4-or PD-1-blocking antibodies in different sequences (Fig. 5A and B). The larger starting tumor volumes were selected so that the monotherapy activity of both treatments would be reduced to observe additive or synergistic effects of the combination. mSIRPa-Fc-CD40L had similar antitumor efficacy to anti-CTLA4 or anti-PD-1 antibodies ( Fig. 5A and B); however SIRPa-Fc-CD40L antitumor efficacy was significantly improved when given together with anti-CTLA4 (57% rejection), together with anti-PD-1 (50% rejection), following anti-CTLA4 (58.8% rejection), or following anti-PD-1 (25% rejection; Fig. 5A and B; Supplementary Fig. S5A-S5D). Nearly all of the mice that rejected the primary CT26 tumor, rejected a secondary tumor challenge compared with na€ ve mice ( Supplementary Fig. S5A-S5D).
Checkpoint blockade of PD-1 and CTLA4 can enhance the antitumor activity of immunotherapeutics with immune-priming activity, including CD40 agonists (44). To understand the mechanistic basis for synergy between PD-1/CTLA4 blockade and mSIRPa-Fc-CD40L, we excised CT26 tumors from anti-PD-1-or anti-CTLA4-treated mice 11 days after inoculation and performed phenotypic analysis of the tumor-infiltrating lymphocytes. Both agents expanded CD40 þ dendritic cells/B cells and CD3 þ T cells, and induced the upregulation of MHC-I and MHC-II (Fig. 5C). Thus, initial treatment with anti-PD-1 or anti-CTLA4 stimulated expansion of CD40-expressing immune cells potentially explaining the improved responsiveness with mSIRPa-Fc-CD40L. Checkpoint inhibitor blockade did not affect the tumor surface expression of CD47 ( Fig. 5C; Supplementary Fig. S5E), suggesting that checkpoint combination synergy functions independently of phagocytosis activity.

Safety and activity of SIRPa-Fc-CD40L in nonhuman primates
Enthusiasm surrounding the clinical utility of SIRPa/CD47 inhibition is somewhat tempered by expression of CD47 on erythrocytes and platelets, and the associated risk of hemolysis and thrombocytopenia (21,42,45). The Fc domain of hSIRPa-Fc-CD40L does not bind effector Fc receptors (Supplementary Table S2), and in vitro studies did not reveal evidence of hemolysis in human or cynomolgus macaque erythrocytes ( Supplementary Fig. S6B); however, the in vitro systems used to test this question had significant limitations, including a complete lack of macrophages. Thus, a more accurate measure of  hSIRPa-Fc-CD40L safety in cynomolgus macaques and overall mechanism of action. A, Cynomolgus macaques were treated with vehicle or 0.1, 1, 10, and 40 mg/kg of SIRPa-Fc-CD40L. Blood chemistry analysis assessed peripheral erythrocyte counts, hemoglobin levels, and hematocrit. B, CD47 receptor occupancy was assessed following the first dose by flow cytometry. Plotted is the inverse of the Free-CD47 flow cytometry signal, since the loss in detectable CD47 signal with increasing dose is directly proportional to the percentage of cells whose CD47 receptor is already bound by hSIRPa-Fc-CD40L. C, The fold change in peripheral lymphocyte count from predose to 24 hours postdose is shown. D-G, The proposed mechanism of action of SIRPa-Fc-CD40L. D, Tumorexpressed CD47 can provide a "do not eat me" signal to APCs through the binding of SIRPa. E, SIRPa-Fc-CD40L can relieve this inhibitory signal while simultaneously providing an "eat me" signal via costimulation of CD40 by CD40L. This enhances tumor phagocytosis, APC activation, increased antigen processing/presentation, and induction of an antitumor antigen-specific CD8 þ T-cell response. F and G, Combining SIRPa-Fc-CD40L with targeted ADCP-competent antibodies potentiates their phagocytosis activity. Cynomolgus macaques are a relevant species for evaluating SIRPa/CD47-related toxicity due to high homology of CD47 between human and cynomolgus macaque (98.69% identity) and the development of the priming dose strategy for the CD47-specific Hu5F9-G4 antibody (46). We confirmed species cross-reactivity of hSIRPa-Fc-CD40L with recombinant cynomolgus CD47 (1.7 nmol/L binding affinity) and CD40 (3.24 nmol/L binding affinity). We sought to test the safety and activity of hSIRPa-Fc-CD40L following repeat doses in cynomolgus macaques. There was no evidence of hemolysis or thrombocytopenia following repeated infusion with hSIRPa-Fc-CD40L at doses up to 40 mg/kg over the course of the study (Fig. 6A). Mild declines in hematology parameters were noted; however, these declines were also noted in the vehicle control group and were most likely related to procedural effects and repeated blood collections. Dose-dependent receptor occupancy was observed across the dosing groups which peaked 4 hours postinfusion at 80.96% AE 2.6%, and was roughly equivalent between the 10 and 40 mg/kg dose groups ( Fig. 6B; Supplementary Fig. S6A). CD47 occupancy was stable and remained at 62.78% AE 2.3% occupancy on RBC CD47 when evaluated 168 hours postinfusion (Fig. 6B). We also observed dose-dependent episodic fluctuation in the total number of lymphocytes before and after each dose, which were of lower magnitude than the postdose reductions observed for circulating CD40 þ lymphocytes (Fig. 6C). This peripheral decrease in CD40 þ B cells was consistent with similar observations seen in the blood of mice treated with mSIRPa-Fc-CD40L ( Supplementary Fig. S6C-S6F). In mice, the decrease in B cells was dose dependent, plateaued at a single intraperitoneal dose of 150 mg, and was accompanied by a significant increase in CD8 þ DCs ( Supplementary Fig. S6C-S6F). Finally, we observed dosedependent increases in multiple serum cytokines/chemokines in cynomolgus macaques following each infusion of hSIRPa-Fc-CD40L, including CCL2, CXCL9, CXCL10, IFNa, IL6, IL15, and IL23 (Supplementary Table S3), suggestive of an on-target pharmacodynamic biology. The overall mechanism through which SIRPa-Fc-CD40L is proposed to bridge macrophage-mediated tumor cell phagocytosis to APC activation and antigen presentation is outlined in Fig. 6D-G.

Discussion
The description of CD47 as a "do not eat me" signal in a broad range of cancers stimulated exploration of what combinations of "eat me" signals may enhance antitumor immunity in the setting of CD47 blockade (10,47,48). Preclinical combinations of CD47 blockade and ADCP-competent antibodies, including rituximab and trastuzumab, enhance tumor phagocytosis (25,35,36,42,49). At least 50% of patients with relapsed or refractory diffuse large B-cell lymphoma or follicular lymphoma treated with Hu5F9-G4, a humanized, IgG4 isotype, CD47-blocking mAb, in combination with rituximab demonstrate objective responses (42). CD47 blockade enhances antigen presentation in immune-neglected tumors (50), yet only sporadic clinical responses have been observed using CD47/SIRPa blocking therapeutics as monotherapy or in combination with PD-1/L1-blocking antibodies. Thus, we sought to determine whether the addition of CD40L could enhance APC activation, antigen presentation, and antitumor immunity downstream of CD47/SIRPa blockade using a single fusion protein containing both SIRPa and CD40L.
Given the oligomeric nature of SIRPa-Fc-CD40L, it was important to independently characterize the functionality of both the SIRPa and CD40L domains. Our data demonstrated that the SIRPa domain functioned as expected by binding to CD47 with high affinity and potentiating macrophage-mediated phagocytosis alone and in the presence of multiple ADCP-competent antibodies. Similarly, the CD40L domain activated both canonical and noncanonical NFkB signaling, and stimulated human PBMC proliferation and IL2 secretion in vitro. CD47 blockade induces a type I IFN response following uptake of tumor mitochondrial DNA, leading to effective antitumor immunity (12). Similarly, SIRPa-Fc-CD40L directly stimulated a type I IFN response, effects not observed with SIRPa-Fc fusion proteins or CD47-blocking antibodies. Such findings could obviate the need for additional type I IFN or STING agonists to enhance antitumor immunity.
While in vitro assays tend to favor the biology of either the SIRPa or CD40L domains (i.e., NFkB reporter assays primarily inform on CD40 activation, whereas macrophage phagocytosis assays primarily inform on SIRPa/CD47 blockade), in vivo studies provided a more complete view of the overall functionality of the construct. SIRPa-Fc-CD40L significantly improved rejection of both primary and secondary tumors as compared with individual antibodies targeting CD40 and CD47 used alone or in combination, similar to PD-1-Fc-OX40L (29). The observation of enhanced antitumor immunity cannot be fully explained by the AH1-tetramer-positive CD8 T-cell population; however, it is possible that other clonotypes were activated that were not detected by the AH1 tetramer, which is consistent with the observation that CD8 À depletion eliminated antitumor immunity. It is also possible that the cytolytic activity of individual tumor specific CD8 þ T-cell clones was enhanced, which is under continued investigation.
Additional evidence for SIRPa-Fc-CD40L bridging an innate and adaptive immune response was provided by the observation that antitumor immunity was dependent both on type I IFNs and T cells in vivo. The T-cell-mediated immune responses downstream of CD47/SIRPa blockade may have been restrained by immune checkpoints. Accordingly, the sequencing of these antibodies with SIRPa-Fc-CD40L had a dramatic influence on the control and rejection of established tumors. Pretreatment with anti-CTLA4 or anti-PD-1 increased the proportion of CD40 þ DCs and B cells within tumorinfiltrating leukocytes, providing possible mechanistic insight that could explain the antitumor benefit when CD40L was also present. Further exploration of SIRPa-Fc-CD40L in combination with checkpoint inhibitors is thus warranted.
These data suggest that while CD47 blockade is an effective strategy to enhance macrophage mediated tumor cell phagocytosis, enhancing a type I IFN response via CD40 stimulation in a coordinated fashion with CD47/SIRPa blockade powerfully enhances antitumor immunity. The observation that SIRPa-Fc-CD40L stimulated dose-dependent elevation in multiple serum cytokines and CD40 þ B-cell margination in cynomolgus macaques, without causing hemolysis or thrombocytopenia, provides justification to further explore this strategy in human patients with cancer.