Eradication of Canine Diffuse Large B-Cell Lymphoma in a Murine Xenograft Model with CD47 Blockade and Anti-CD20

Because of a high demand for cancer treatments by pet owners and a large clinical need, immunotherapeutic approaches are now being applied to veterinary settings (1). In the United States alone, the incidence of canine cancer is estimated to be over 4 million new cases per year, and cancer is believed to be the leading cause of disease-related mortality in dogs (2). Lymphoma is among the most common types of canine cancer, with incidence rates of approximately 20–107 cases per 100,000 dogs at risk (3). Humans exhibit a high degree of genomic similarity with dogs, and dogs share environments and exposures with their human owners (4). Many canine and human cancers appear to be conserved at the molecular level and in their clinical behavior, particularly lymphoma (2, 3). Veterinary studies represent an ethical use of animals in research, enabling the study of naturally occurring disease and providing accurate representations of tumor heterogeneity, the tumor microenvironment, and the course of illness progression. Thus, veterinary studies may be valuable “stepping stones” in conceptual advancement toward understanding and treating human disease as well as ameliorating disease in valued companions (5).

A number of passive immunotherapies have become the standard of care for human cancers, including CD20 antibodies for patients with non-Hodgkin lymphoma and other B-cell malignancies. The CD20 mAb rituximab was the first mAb approved for the treatment of cancer and is responsible for increasing the 10-year survival of patients with non-Hodgkin lymphoma by over 15% (6). Expression of CD20 in dogs on normal and malignant B cells is conserved, supporting the promise of CD20 as a therapeutic target in canine lymphoma (7, 8). Recently, a canine-specific CD20 mAb was developed with canine Fc regions (1E4-cIgGB) that effectively depletes normal B cells in healthy dogs and is undergoing investigation for canine B-cell lymphoma (9).

Multiple mechanisms of action have been described for CD20 mAbs, but studies indicate a major in vivo mechanism is antibody-dependent phagocytosis by macrophages (10–14). The CD47/signal regulatory protein alpha (SIRPα) axis is an immune checkpoint that limits the macrophage response to tumor-specific antibodies (11, 14–16). By binding to SIRPα, an inhibitory receptor on macrophages and other myeloid cells, CD47 transduces inhibitory signals that allow tumor cells to evade macrophage-mediated destruction (10, 11, 15, 17–21). As such, the combination of CD47-blocking agents and tumor-binding antibodies that bind to macrophage Fc receptors is highly effective in preclinical models of human lymphoma (10, 11). Many cancers express high CD47, and multiple CD47-blocking reagents are now under investigation in clinical trials for both solid and hematologic malignancies (clinicaltrials.gov identifiers NCT02216409, NCT02367196, NCT02663518, NCT02678338).

In this study, we investigated whether immunotherapeutic targeting of CD47 and CD20 could be applied to the canine system. We first characterized the canine CD47 and SIRPα homologs. Next, we identified a lead candidate that potently blocks canine CD47, induces macrophage phagocytosis of canine lymphoma cells, and eliminates canine lymphoma in xenotransplantation models. Finally, we confirmed that CD47-blocking therapies augment the therapeutic response produced by anti-CD20 against canine lymphoma.

The CLBL-1 canine diffuse large B-cell lymphoma (DLBCL) cell line (22) was obtained from Dr. Barbara Rütgen (University of Vienna, Vienna, Austria) in 2009 and was authenticated in 2015 by the Modiano laboratory using short tandem repeat (STR) testing (DDC Medical). A GFP-luciferase⁺ CLBL-1 variant was generated by transfection with a Sleeping Beauty transposon system as described previously (23). Briefly, 1 × 10⁶ CLBL-1 cells were transfected using a Nucleofector system, program DN-100 (Lonza) with 1 μg of transposon-expressing pDNA vector along with 2 μg of the GFP/luc vector pKT2/CLP-Luc-ZOG in 100 μL of nucleofector solution SF (Lonza). CLBL-1 cells were grown in Iscove's modified Dulbecco's medium (IMDM) plus GlutaMAX (Invitrogen) supplemented with 20% FBS (Omega Scientific or Atlas Biologicals), 100 U/mL penicillin, and 100 μg/mL streptomycin (Invitrogen). J774 cells were obtained from ATCC in 2012 and 2015 and authenticated in 2015 by the Modiano laboratory using STR testing (DDC Medical). J774 cells were cultured in DMEM (Thermo Fisher Scientific) with 10% FBS (Atlas Biologicals).

Osteosarcoma lines OSCA-40 and OSCA-78 were derived in the Modiano laboratory in 2004 and 2008, respectively. They were authenticated in 2015 by the Modiano laboratory using STR testing (DNA Diagnostic Center) and cultured as described previously (24). Hemangiosarcoma cell line COSB was rederived by the Modiano laboratory in 2007 by xenograft passage of parental line SB. It was authenticated in 2015 by the Modiano laboratory using STR testing (DNA Diagnostic Center) with 1 of 20 alleles differing from the parental line. Hemangiosarcoma cell line Emma was derived by the Modiano laboratory in 2008 and authenticated in 2015 by the Modiano laboratory using STR testing (DNA Diagnostic Center). COSB and Emma were cultured as described previously (25–27). Canine melanoma cell lines TLM1, CMGD2, and CMGD5 were derived by the Modiano laboratory in 1996, 2001, and 2001, respectively. They were authenticated in 2015 by the Modiano laboratory using STR testing (DNA Diagnostic Center). Canine melanoma cell lines were cultured as described previously (28, 29). Canine glioma cell lines Mac and Candy were provided by Dr. John Ohlfest [University of Minnesota (UMN), Minneapolis, MN] in 2009. They were cultured in neurobasal medium supplemented with N2 and B27 (Invitrogen), 10% FBS, l-glutamine, and Primocin (Invitrogen). They were passaged less than 20 times in the Modiano laboratory. Raji cells were obtained from ATCC between 2012 and 2015 and authenticated by the repository. Raji cells were cultured in RPMI1640 plus GlutaMAX (Invitrogen) supplemented with 10% FBS (Omega Scientific), 100 U/mL penicillin, and 100 μg/mL streptomycin (Invitrogen).

High-affinity SIRPα protein CV1 was produced as described previously (11). CV1-hIgG4 (11) and Hu5F9-G4 (30) were produced as described previously by expression in FreeStyle 293-F cells (Invitrogen) or CHO-K1 cells (Lonza). The native mouse antibody to dog CD20, 1E4, and chimeric mouse antibodies to CD20 on the canine IgGB background (IgG1 equivalent) Rtx-cIgGB and 1E4-cIgGB, were produced as described previously (9) and provided by Elanco Animal Health. For 1E4-cIgGB and Rtx-cIgGB ELISAs, binding to immobilized cCD20(ECD2)-hFc fusion protein was detected using a rabbit anti-canine IgG-HRP (US Biologicals #I1904-25Z) and Sure Blue TMB substrate (KPL) with absorbance measured at 650 nm in a Spectra Max (Molecular Devices).

Sequence accession numbers for CD47 sequences included canine CD47 (NP_001074190.1), human CD47 (NP_001768.1), and mouse CD47 (BAA25401.1). Sequence accession numbers for SIRPα sequences included canine SIRPα (XP_005634938.1), human SIRPα variant 1 (PDB sequence ID 4CMM|A), human SIRPα variant 2 (PDB sequence ID 2UV3|A), C57BL/6 mouse SIRPα (XM_006498982.2), and NOD mouse SIRPα (31, 32). Multiple sequence alignment and phylogenetic analysis were performed using Clustal Omega (33). Sequence identity and similarity between CD47 and SIRPα variants were assessed by NCBI pBLAST alignment (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The human SIRPα secondary structure was described previously (32).

Canine CD47-Fc was constructed by inserting the canine CD47 extracellular domain (amino acids 1–116 of mature protein) into a modified pFUSE-hIgG4-Fc vector (Invivogen) with an IL2 signal sequence and engineered Ser228Pro mutation (34). A C15G was introduced to the CD47 domain to facilitate purification as described for human CD47 (11, 35). Canine CD47-Fc was expressed in FreeStyle 293-F cells (Invitrogen) and purified using HiTrap Protein A columns (GE Healthcare). Canine CD47-Fc was eluted using citrate-based (pH 3.0) or arginine-based (pH 4.0) buffers and immediately neutralized using 1.0 mol/L Tris-HCl (pH 8.0).

Canine SIRPα tetramers were produced as described for mouse SIRPα (11). Briefly, the canine N-terminal domain of SIRPα (amino acids 1–118 of mature protein) was expressed in E. coli with an N-terminal maltose binding protein (MBP) tag followed by a rhinovirus 3C protease cleavage site and C-terminal biotin acceptor peptide (BAP) and 8× histidine tag. The MBP-fusion protein was purified using Ni-NTA chromatography, and eluted products were digested with 3C protease to remove MBP, purified by an additional Ni-NTA chromatography step, and subjected to size-exclusion chromatography with a Superdex-S75 column. The protein was biotinylated with BirA ligase. Tetramers were formed by incubation with Alexa Fluor 647–conjugated streptavidin. See Supplementary Materials for engineered protein sequence information.

The extracellular domain of canine CD47, with the C15G mutation, was displayed on the surface of S. cerevisiae strain EBY100 as an N-terminal fusion to Aga2, leaving a free N-terminus as previously reported for human CD47 (36). The N-terminal domains of canine, C57BL/6, and NOD SIRPα were displayed on yeast by fusion to the C-terminus of Aga2 using the pCT302 vector as described for human and mouse SIRPα variants (11).

Binding of canine CD47-Fc to canine and mouse SIRPα was assessed using yeast expressing canine or mouse SIRPα variants on their surface. Binding of canine CD47-Fc, Hu5F9-G4, and CV1-hIgG4 was detected using an Alexa Fluor 647–conjugated anti-human IgG (H+L) secondary antibody (Invitrogen). Blockade of canine SIRPα was assessed in a competition assay using varying concentrations of Hu5F9-G4 and CV1-hIgG4 combined with 50 nmol/L Alexa Fluor 647–conjugated streptavidin canine SIRPα tetramers and incubated for 30 minutes on ice. Analysis was performed using an LSRFortessa with high-throughput sampler (BD Biosciences). Geometric mean fluorescence intensity was measured using FlowJo v9.4.10 (Tree Star), and data were fit to sigmoidal dose–response curves using Prism 5 (GraphPad).

For studies using CV1-hIgG4, cells were incubated with CD47 antagonists, washed, and then binding was detected with an Alexa Fluor 647–conjugated anti-human IgG (H+L) antibody (Invitrogen). Wild-type human SIRPα allele 2 was produced as a fusion to human IgG4 (WT-hIgG4) as described previously (11) and was used as a staining control. Dead cells were excluded from the analysis using DAPI (Sigma Aldrich). Samples were analyzed on an LSRFortessa. For studies using directly conjugated Hu5F9-G4, 1 × 10⁶ cells of each line were incubated with normal canine IgG and mouse IgG1 for 10 minutes on ice. Cells were subsequently labeled using Hu5F9-G4 conjugated to Alexa Fluor 488 or an isotype control antibody (human IgG4 conjugated to Alexa Fluor 488) and incubated on ice for 30 minutes. Analysis was performed using a LSRII (BD Biosciences), and geometric mean fluorescence intensity was measured using FlowJo.

Primary mouse macrophages derived from NSG mice were differentiated ex vivo in the presence of 10 ng/mL murine M-CSF (Peprotech) (11). Canine macrophages were derived from blood obtained from healthy companion dogs with approval from the UMN Institutional Animal Care and Use Committee (UMN IACUC, protocol #1304-30546A). Peripheral blood mononuclear cells were enriched by density gradient centrifugation over Ficoll-Paque Premium (GE Healthcare). Monocytes were purified on an autoMACS Pro Separator (Miltenyi Biotec) using anti-CD14 microbeads (Miltenyi Biotec) and differentiated to macrophages by culture for 7–10 days in IMDM plus GlutaMax (Invitrogen) supplemented with 10% autologous serum for each individual donor and 100 U/mL penicillin and 100 μg/mL streptomycin.

Macrophage phagocytosis was evaluated as described previously (11). Briefly, 50,000 macrophages were cocultured with 100,000 cancer cells labeled with either calcein AM or CFSE (Invitrogen). Phagocytosis in response to therapeutic treatments was evaluated after 2 hours of coculture and analyzed by flow cytometry. NSG macrophages were identified by staining with PE/Cy7 or APC-anti-mouse F4/80 (eBioscience). Canine macrophages were stained with Tri-color–conjugated or PE-conjugated anti-CD14 clone TuK4 (Thermo Fisher, Miltenyi Biotec). Phagocytosis was evaluated as the percentage of macrophages engulfing fluorescent tumor cells using an LSRFortessa with high-throughput sampler or a LSR II and analyzed using FlowJo v9.4.10. Microscopic evaluation of phagocytosis was performed using an ImageStream MKII Flow Cytometer (Amnis).

Antibody-mediated direct cytotoxicity was evaluated in response to Rtx-cIgGB or 1E4-cIgGB. Cells were cultured with antibodies for 72 hours, at which time viability was analyzed by MTS assay (Promega) according to the manufacturer's protocol. Complement-dependent cytotoxicity was evaluated by incubation of cells with antibodies against CD20 and rabbit complement (ImmunO, MP Biomedicals,) or heat-inactivated rabbit complement (56°C for 1 hour).

Mice were maintained in a barrier facility and handled according to protocols approved by the Stanford University Administrative Panel on Laboratory Animal Care (protocol #26270) and UMN IACUC (protocol #1207A17291). Nod.Cg-Prkdc^scid IL2rg^tm1Wjl/SzJ (NSG) mice (37) were used for all in vivo experiments. Mice were engrafted with CLBL-1 lymphoma cells at approximately 6–12 weeks of age, and experiments were performed with age and sex-matched cohorts of 7–10 mice.

For a model of disseminated lymphoma, NSG mice were engrafted with 100,000 CLBL-1 cells or GFP-luciferase⁺ CLBL-1 cells intravenously via retro-orbital injections. Fluorescent dissection microscopy and bioluminescence imaging was performed as described previously (11). Treatment with PBS or 250 μg of CV1-hIgG4 was initiated 24 hours or 6 days postengraftment and continued daily for 1 to 2 weeks, as indicated. Orbital tumor scores were assessed on a three-point scale and scored as follows for visible tumor infiltration: 1, undetectable or mild; 2, moderate; 3, high. Aggressive intra-abdominal lymphoma was modeled by injection of 50,000 cells intraperitoneally. In this model, systemic tumor dissemination leads to gross organomegaly and weight gain; therefore, tumor burden was assessed by weight gain over time. Treatment with PBS or antibodies against CD20 was initiated 3 days postengraftment and continued twice per week until day 32 post-tumor injection. For a localized model of canine lymphoma, 40,000 CLBL-1 cells were engrafted subcutaneously. Treatment was initiated 24 hours postengraftment with anti-CD20 (250 μg/mouse, 3 days/week, continued until day 36) and CV1 (200 μg/mouse/day, continued until day 24). Tumor dimensions were measured to calculate volumes according to the ellipsoid formula (π/6 × length × width²).

Pathologic analysis of tumor specimens was performed by UMN Comparative Pathology Shared Resource by staining with hematoxylin–eosin, anti-CD20 [primary antibody: rabbit anti-human CD20 polyclonal antibody (Thermo Scientific)], detected with a rabbit EnVision+ Kit (Dako), anti-CD3 [primary antibody: rat anti-human CD3 mAb (Serotec), detected with a rat-on-mouse HRP polymer detection system (Biocare Medical)], and anti-CD79a [primary antibody (mouse anti-human CD79A mAb (Biocare), detected with a mouse EnVision+ Kit].

A homolog of canine CD20 has been identified, and a speciated antibody to canine CD20 has been developed (7–9). Therefore, we began our investigation by examining the canine CD47–SIRPα interaction. The existence of a canine CD47 homolog was supported by a prior serologic study that examined antibody cross-reactivity (38). In addition, a predicted homolog of canine CD47 was annotated in the canine genome (NM_001080721.1). However, no biochemical or functional analysis has been performed to validate this sequence. Alignment of the extracellular domain of the putative canine CD47 sequence with that of human and mouse showed the gene product was more closely related to human CD47 with 66% amino acid identity and 79% amino acid similarity (Fig. 1A and B; Supplementary Table S1). A putative canine SIRPα homolog was predicted from the canine genome based on sequence similarity (XP_005634938.1), although no molecular validation has been performed. The CD47-binding domain of this gene product aligns with human and mouse SIRPα variants (Fig. 1C). As with canine CD47, the putative canine SIRPα homolog shares more similarity with human SIRPα variants than those of mouse (Fig. 1D). The CD47-binding domain of human SIRPα is highly polymorphic (31). Interestingly, the predicted canine SIRPα gene product shares 84%–85% similarity with human SIRPα variants, whereas the similarity between human SIRPα variants 1 and 2 is 90% (Supplementary Table S2).

To define these predicted gene products as functional CD47 and SIRPα homologs, we assessed whether an interaction exists between the two proteins at the biochemical level. We cloned the N-terminal extracellular domain of the putative canine SIRPα protein and expressed it on the surface of yeast as a fusion to Aga2, a method previously used to evaluate the functions of human and mouse SIRPα (Fig. 2A; ref. 11). Next, we produced the extracellular domain of canine CD47 as an Fc fusion protein (canine CD47-Fc). As hypothesized, we found that canine CD47-Fc binds to canine SIRPα with a dose-dependent relationship (Fig. 2B and C). This finding establishes the predicted gene products as canine CD47 and SIRPα and confirms a biochemical interaction between the extracellular domains of the two proteins.

To validate the potential of CD47 blockade in xenograft models of canine lymphoma (22, 39), we evaluated whether canine CD47 could bind to mouse SIRPα. Allelic variants of mouse SIRPα have been described with different binding affinities for human CD47. In particular, the SIRPα allele from NOD mice interacts with human CD47, whereas there is minimal binding by the SIRPα allele from C57BL/6 mice (31). The preservation of this interaction across species is a major determinant for xenotransplantation of human tissues into immunocompromised mice, and it enables in vivo evaluation of therapeutic responses to CD47-blocking therapies. Therefore, we investigated whether cross-reactivity exists between canine CD47 and mouse SIRPα. We expressed the mouse NOD and C57BL/6 SIRPα alleles on the surface of yeast and used canine CD47-Fc to assess binding. We found that, similar to human CD47 (31), canine CD47 exhibited much stronger binding to the SIRPα allele expressed by NOD mice versus that expressed by C57BL/6 mice (Fig. 2D and E). As with canine SIRPα, canine CD47-Fc binds to NOD SIRPα in a dose-dependent manner. Therefore, this finding identifies the NOD strain as the most appropriate background for evaluation of the canine CD47–SIRPα interaction.

By cloning the canine CD47 and SIRPα genes, we were able to generate tools to investigate the ability of therapeutic candidates to bind and block canine CD47. As with the extracellular domain of canine SIRPα, we generated a construct that contains the extracellular domain of canine CD47 for display on the surface of yeast and tested the ability of two CD47-blocking therapeutics to bind and block canine CD47. The first was Hu5F9-G4, a humanized antibody to CD47. The second was CV1-hIgG4, a high-affinity SIRPα-Fc fusion protein (11). CV1-hIgG4 blocks both human and mouse CD47 (11); thus, we hypothesized CV1-hIgG4 may also be a functional antagonist of canine CD47. To evaluate their binding to canine CD47, we incubated yeast cells expressing canine CD47 with varying concentrations of Hu5F9-G4 or CV1-hIgG4. We found that the agents differed in their ability to bind canine CD47 (Fig. 3A). CV1-hIgG4 demonstrated potent binding, with an EC₅₀ of approximately 741 pmol/L (11). On the other hand, Hu5F9-G4 exhibited less potent binding by orders of magnitude and did not fully saturate the cells at the maximum concentration tested (10 μg/mL).

To evaluate the interaction between canine CD47 and SIRPα further, we expressed and purified the N-terminal domain of canine SIRPα from E. coli (Supplementary Fig. S1A and S1B). We produced fluorophore-labeled canine SIRPα tetramers and confirmed that they could bind to canine CD47 expressed on the surface of the yeast. Using this system, we generated an assay to investigate whether Hu5F9-G4 or CV1-hIgG4 could act as competitive antagonists to canine CD47. We incubated yeast-expressing canine CD47 with varying concentrations of the CD47-blocking reagents in combination with the fluorophore-labeled canine SIRPα tetramers. In this competition assay, CV1-hIgG4 demonstrated potent antagonism of canine CD47 with an IC₅₀ of 2.92 nmol/L (Fig. 3B; ref. 11). Because of its weaker affinity for canine CD47, Hu5F9-G4 was unable to antagonize canine CD47 even at the highest concentration tested. These biochemical studies suggested high-affinity SIRPα reagents, such as CV1 or bivalent CV1-hIgG4, could act as therapeutic candidates for blockade of canine CD47.

We next investigated whether canine lymphoma cells expressed CD47 on their surface. We found that CV1-hIgG4 binds to CLBL-1 cells with high affinity (Supplementary Fig. S2A). CD47 was similarly expressed on the surface of primary canine DLBCL samples and nonmalignant leukocytes (Supplementary Fig. 2B). We also confirmed CD47 expression in each of 24 primary canine DLBCL samples using an available RNA-sequencing dataset (refs. 40, 41; Supplementary Fig. 2C). As CD47 is highly expressed by human cancer cells (10, 11, 19–21), we also explored whether other canine cancer samples expressed CD47 on their surface. Additional samples included osteosarcoma, melanoma, and glioblastoma cell lines (Supplementary Fig. S2D and S2E). A range of CD47 expression was observed; yet, as with human cancer specimens, all samples tested expressed detectable surface CD47 (Supplementary Fig. S2D and S2E). This finding suggests that CD47 may allow canine cancer cells to evade the detection by the immune system and confirms the presence of the therapeutic target on multiple types of canine cancer.

To evaluate the functional efficacy of our CD47-blocking reagents against canine cancer, we tested the reagents in phagocytosis assays using CLBL-1 canine B-cell lymphoma cells (28). CLBL-1 cells were labeled with fluorescent dye calcein AM and cocultured with primary NOD-scid-γ_c^−/− (NSG; ref. 37) macrophages in the presence or absence of CD47-blocking therapies. Phagocytosis was measured by flow cytometry and evaluated as the percentage of calcein AM–positive macrophages, a population that exhibited phagocytic engulfment by microscopy (Fig. 3C; Supplementary Fig. S3). Compared with control treatment, Hu5F9-G4 was only able to induce moderate phagocytosis. However, treatment with CV1-hIgG4 produced maximal phagocytosis, corresponding with its greater ability to bind and block canine CD47 (Fig. 3C and D).

We then evaluated the ability of canine macrophages to respond to CD47-blocking therapies. Canine macrophages were differentiated ex vivo from purified monocytes in the presence of autologous serum. After 7–10 days, the cells were firmly adherent to tissue culture plates, exhibited a macrophage morphology, and expressed the macrophage marker CD14 (Supplementary Fig. S4A and S4B). Canine macrophages from three donors were cocultured with fluorescently labeled CLBL-1 cells and CD47-blocking therapies. Similar trends were observed as with NSG macrophages: CV1-hIgG4 produced a significant increase in phagocytosis compared with no treatment or Hu5F9-G4 (Fig. 3E). Both therapies were comparable at inducing phagocytosis by canine macrophages when tested against human lymphoma cells (Supplementary Fig. S4C). Together, these in vitro findings led us to select CV1-hIgG4 as a lead therapeutic candidate for evaluation as monotherapy in vivo.

To investigate the therapeutic efficacy of CD47 blockade in vivo against canine cancer, we established a xenotransplantation model of canine lymphoma (Fig. 4A). We injected 100,000 canine GFP-luciferase⁺ CLBL-1 lymphoma cells into the retro-orbital venous sinuses of NSG mice (37). The lymphoma cells efficiently engrafted and disseminated throughout the body. One day after engraftment, mice were randomized into two cohorts, and treatment was initiated with either vehicle control or 250 μg CV1-hIgG4. Treatment was administered daily for 2 weeks, and the mice were monitored for disease progression. Three weeks postengraftment, nearly all mice in the control group developed early signs of morbidity, with evidence of disease burden around the retro-orbital injection site (Fig. 4B and C). In comparison, none of the mice treated with CV1-hIgG4 exhibited retro-orbital disease at this time. Mice were monitored for survival, and all mice in the control group developed rapid progression of their disease requiring humane euthanasia (Fig. 4D). In contrast, 50% of the mice treated with CV1-hIgG4 exhibited delayed progression of their disease, while the remaining 50% had no evidence of disease even after 200 days (P = 0.002).

As a model of more advanced disease, mice were engrafted with 100,000 GFP-luciferase⁺ CLBL-1 cells, and treatment was initiated 6 days postengraftment, when bioluminescence imaging confirmed lymphoma cells had disseminated throughout the body of the animals (Fig. 4E). Mice were then randomized into treatment with either vehicle control or 250 μg CV1-hIgG4 daily. Mice were treated for 1 week, then bioluminescence imaging was performed to assess tumor burden. Even after this short course of intervention, mice treated with CV1-hIgG4 exhibited significantly less tumor burden than animals treated in the control group (Fig. 4F). However, treatment with CV1-hIgG4 did not significantly enhance survival in this model (data not shown).

As previous studies have demonstrated synergy between CD47-blocking therapies and tumor-specific antibodies (10, 11), we next examined whether these principles might also apply to canine lymphoma. The chimeric antibody to canine CD20, 1E4-cIgGB, consists of the variable regions of mouse mAb 1E4, which binds canine CD20 extracellular domain 2 with high affinity (Supplementary Fig. S5A), and the canine IgGB Fc region (9). 1E4-cIgGB binds canine B cells in vitro (Supplementary Fig. S5B) and can deplete B cells in healthy dogs (9). Various independent mechanisms of action mediate B-cell depletion and the therapeutic effect of anti-CD20, including direct cytotoxicity, complement-dependent cytotoxicity (CDC), and antibody-dependent phagocytosis (ADP; refs. 12, 42, 43). We examined the capacity of unmodified mAb 1E4 (mouse IgG2a) and chimeric 1E4-cIgGB to mediate these effects in vitro. Neither 1E4 nor 1E4-cIgGB exhibited direct cytotoxicity against CLBL-1 canine lymphoma cells (Fig. 5A), but both induced significant complement-mediated lysis of the target cells (Fig. 5B). In addition, 1E4 significantly increased phagocytosis of CLBL-1 cells in an ADP assay with IFNγ-primed mouse J774 cells as effectors (Fig. 5C).

The anti-canine CD20, 1E4-cIgGB, depletes B cells in healthy dogs, but it has not previously been investigated in dogs with naturally occurring lymphoma. Therefore, we evaluated the therapeutic efficacy of 1E4-cIgGB in vivo using intraperitoneal CLBL-1 xenografts in NSG mice (39, 44). In this aggressive model, CLBL-1 forms disseminated intra-abdominal tumors and grows systemically in visceral organs, leading to progressive weight gain until the tumor endpoint at 25–30 days. Mice were treated twice weekly with 1E4-cIgGB or a control antibody, Rtx-cIgGB, which consists of the variable domains of rituximab with a canine IgGB Fc region. Rituximab is specific for human CD20 and does not cross-react with the canine CD20 homolog (ref. 7; Supplementary Fig. S5A). Weight gain was significantly decreased in the 1E4-cIgGB treatment group as compared with the Rtx-cIgGB treatment and the no treatment groups, indicating a therapeutic response (Fig. 6A). Engraftment of high-grade canine B-cell lymphoma was observed in the spleens of all mice (Fig. 6B and C). Nevertheless, histologic analysis demonstrated an increase in the tumor necrosis scores from mice in the 1E4-cIgGB treatment group as compared with the Rtx-cIgGB treatment and the no treatment groups (Fig. 6D and E; Supplementary Table S3), suggesting that 1E4-cIgGB increased tumor cell death in vivo.

We next tested the hypothesis that CD47 blockade would enhance the effects of antibodies against CD20 to ablate canine lymphoma. We examined the capacity of this combination strategy to enhance phagocytosis of CLBL-1 canine lymphoma cells by primary NSG macrophages in vitro. CD47 was targeted using Hu5F9-G4 or monomeric CV1, which lacks a prophagocytic Fc region and therefore acts as a pure CD47 antagonist (11, 45–47). CV1 is not sufficient to induce phagocytosis by itself, but it augments phagocytosis when combined with a tumor-opsonizing antibody (11, 45–47). CD20 was targeted using the unmodified 1E4 mouse antibody or chimeric 1E4-cIgGB. CD47 blockade with monomeric CV1 or 1E4 alone were unable to elicit significant phagocytosis (Fig. 7A). However, the combination of CV1 and either 1E4-cIgGb or 1E4-mIgG2a elicited significant enhancement of phagocytosis that was greater than either therapy as a single agent. CV1 had no significant effect on proliferation of CLBL-1 cells (Supplementary Fig. S6), indicating CV1 did not cause growth inhibition via direct effects on tumor cells.

Subsequently, we examined this combination strategy in mice bearing subcutaneous CLBL-1 xenografts. Treatment was initiated 24 hours after tumor engraftment using anti-CD20 (250 μg of Rtx-cIgGB or 1E4-cIgGB) three times per week and 200 μg of monomeric CV1 daily per mouse. Treatment was continued for 24 days (CV1) or 36 days (anti-CD20). Subcutaneous tumors were visible within 2–3 weeks, and there was progressive tumor growth in mice receiving vehicle control treatment or Rtx-cIgGB. CD47 blockade and anti-CD20 therapy were well tolerated when administered alone or in combination, with no overt toxicity (Supplementary Fig. S7A and S7B). Both of these therapies significantly delayed tumor growth when administered alone, and remarkably no tumors were grossly visible by day 21 in the group treated with the combination of CV1 and 1E4-cIgGB (Fig. 7B; Supplementary Fig. S7B). Consistent with this observation, CV1 and 1E4-cIgGB monotherapy prolonged survival of tumor-bearing mice (Fig. 7C), but 17 of 20 mice in these groups were sacrificed due to tumor progression by 60 days postengraftment. In contrast, 100% (10/10) of the mice treated with monomeric CV1 in combination with 1E4-cIgGB survived until the experiment was terminated at day 68. All of the mice in the control groups and the 1E4-cIgGB monotherapy treatment group and 8 of 10 mice in the CV1 treatment group had microscopic evidence of tumor in the subcutis and in lymphoid and visceral organs (Supplementary Fig. S7C). The derivation of the tumors from canine B-cell lymphoma cells was confirmed by IHC (Fig. 7D). Infiltrating granulocytes and macrophages were repeatedly seen in the tumors and may have been associated with necrosis. The robust therapeutic activity of the CV1 and 1E4-cIgGB combination was underscored by the fact that none of the mice in this treatment group had any microscopic evidence of lymphoma in the subcutis or in any lymphoid or visceral organs examined (Supplementary Fig. S7C), suggesting complete eradication of disease by this combination strategy.

Lymphoma is responsible for significant morbidity and mortality in pet dogs, bearing remarkable similarity to human non-Hodgkin lymphoma and providing clinically realistic opportunities to explore combinatorial strategies and dosing regimens that may translate to human disease. In this study, we explored the relationship between the CD47/SIRPα axis and anti-CD20 immunotherapy in canine B-cell lymphoma. Agents targeting CD47, an immune checkpoint, show great potential for multiple types of human cancer in preclinical models and are now under investigation in human clinical trials (10, 11, 19, 20, 30). We identified high-affinity SIRPα variants (11) as a class of therapeutics that potently antagonize canine CD47. CV1-hIgG4 stimulated phagocytosis of canine lymphoma cells in vitro and produced cures in a significant proportion of animals when used as a single agent in a xenograft mouse model of canine lymphoma. Therefore, our data support the evolutionary conservation of structure and function of the CD47/SIRPα axis in the canine system.

Passive immunotherapies with tumor-specific antibodies have become the standard of care for many human cancers, particularly the use of antibodies to CD20 for B-cell lymphoma. Antibody therapies are now emerging for canine cancer (1). The CD47/SIRPα axis limits macrophage responses to therapeutic antibodies, and the combination of CD47-blocking therapies with anti-CD20 shows remarkable synergy for human non-Hodgkin lymphoma (10, 11). Therefore, we investigated the combination of CD47-blocking therapies with 1E4-cIgGB, a canine-specific anti-CD20 under development for canine lymphoma (9). We found that 1E4-cIgGB can act via multiple effector mechanisms, including CDC and ADP, suggesting favorable properties for passive immunotherapy of canine B-cell lymphoma. We observed a therapeutic effect using 1E4-cIgGB as a single agent in a model of aggressive, intra-abdominal canine lymphoma. Furthermore, when combined with CD47-blocking therapies, we observed synergy in vitro and in vivo. We found that the combination of CV1 monomer and 1E4-cIgGB elicited cures in 100% of mice bearing canine lymphoma, whereas nearly all mice in the single-agent groups exhibited evidence of disease. Therefore, our findings suggest that the combination with CD47-blocking therapies could augment the potential benefits of 1E4-cIgGB for canine lymphoma.

The CD47/SIRPα axis has been best characterized in the context of macrophage activation. We do not exclude the possibility that other SIRPα⁺ myeloid cells might contribute to the efficacy observed in our models, as granulocytes and myeloid dendritic cells respond to CD47-blocking therapies (15, 48). Moreover, CD47 blockade can initiate antigen presentation that stimulates adaptive immune responses against tumors (48–50) and targeting CD47 might directly activate T cells or NK cells (51–53). Although lymphocytes are not necessary for a therapeutic response in xenograft models, they could yield greater efficacy in immunocompetent settings and will be important to evaluate in future veterinary studies.

We did not observe significant differences between CD47 expression on healthy canine leukocytes and primary lymphoma cells. However, CD47 expression alone is not sufficient to predict therapeutic efficacy or toxicity (46). In immunocompetent mice, a therapeutic window for targeting CD47 on tumors exists despite CD47 expression on normal cells (20, 48, 54). In nonhuman primates, therapeutic dosing of CD47-blocking agents can be achieved while limiting side effects (11, 30). IL10 appears to prevent macrophages from attacking healthy cells when the CD47/SIRPα interaction is disrupted (55); other factors that regulate the response to treatment likely depend on a balance between prophagocytic and other inhibitory signals, and the identification of these signals remains an active area of investigation.

Nevertheless, the strategy of combining CV1 with an opsonizing antibody is designed to maximize the therapeutic window for targeting CD47. In contrast to antibodies to CD47 or CV1-hIgG4, CV1 blocks CD47 without contributing a prophagocytic stimulus. CV1 does not stimulate phagocytosis by itself, but instead lowers the threshold for phagocytosis in the presence of an opsonizing antibody (11, 45–47). The combination of CV1 with 1E4-cIgGB enables targeting of CD47 without toxicity, as the anti-CD20 agent confers specificity.

The main toxicity from the use of CD47-blocking reagents is reduced red blood cell mass. Moderate anemia has been observed in mice and cynomolgus monkeys treated with anti-CD47 and high-affinity SIRPα-Fc fusion proteins (11, 20, 30). However, the anemia is transient and dosing strategies have been developed to reduce this toxicity (30). To reduce possible immunogenicity, chimeric SIRPα-Fc fusion proteins could be engineered by substituting the human Fc for an active or inactive canine Fc region (9). Nonetheless, immunogenicity may be reduced when used in combination with an agent such as 1E4-cIgGB that causes B-cell depletion (9).

We focused on the CLBL-1 cell line as a model as it is the only known canine cell line that faithfully represents DLBCL, reproducibly forms tumors, and retains its phenotype in the xenotransplantation setting (39). Although primary canine DLBCL xenografts have been described (44), these tumors only form when implanted intraperitoneally into conditioned NSG mice (44) and their variability in growth makes therapeutic evaluation challenging. The CLBL-1 model demonstrated proof-of-concept for combined targeting of CD47 and CD20 at the biological level. Therefore, we believe the most informative and relevant next step of this research endeavor will be clinical studies in dogs with spontaneous disease.

We found that CD47 is widely expressed on multiple types of canine cancer. Therefore, as with the human CD47–SIRPα interaction, the principles established by this study may be generalizable to other types of canine cancer and other canine antibodies to tumors. A popular and successful strategy in immuno-oncology is to combine treatments; CD47-blocking therapies may thus complement emerging treatment modalities using passive immunotherapy for canine diseases (56). Furthermore, anti-CD20 is used for inflammatory or autoimmune disorders, and similar spectrums of disease exist in canine models that could also benefit from B-cell depletion in combination with CD47-blocking therapies. Overall, our findings indicate that CD47 blockade alone and in combination with anti-CD20 may be an effective strategy for treating canine lymphoma. These findings justify further evaluation in the veterinary translational setting.

K. Weiskopf has ownership interest (including patents) in Forty Seven Inc and Alexo Therapeutics and is a consultant/advisory board member for Alexo Therapeutics. A.M. Ring has ownership interest (including patents) as equity in Forty Seven Inc and Alexo and patent inventorship and is a consultant/advisory board member for Alexo Therapeutics and Forty Seven Inc. I.L. Weissman is a founder and director at Forty Seven Inc, has ownership interest (including patents) in Forty Seven Inc, is a consultant/advisory board member for Forty Seven Inc, and has provided expert testimony for Forty Seven Inc. S. Prohaska and K.M. McKenna are employees of Forty Seven Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. Weiskopf, D. Ito, A.M. Ring, K. Bloink, S. Rue, D.E. Lowery, S. Prohaska, I.L. Weissman, J.F. Modiano

Development of methodology: K. Weiskopf, D. Ito, P. Schnorr, A.M. Ring, K. Bloink, J. Efe, D.E. Lowery, I.L. Weissman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Weiskopf, K.L. Anderson, D. Ito, P. Schnorr, H. Tomiyasu, A.M. Ring, J. Efe, A. Barkal, K.M. McKenna, I. Cornax, T. O'Brien, M.G. O'Sullivan, J.F. Modiano

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Weiskopf, K.L. Anderson, D. Ito, H. Tomiyasu, A.M. Ring, K. Bloink, J. Efe, A. Barkal, K.M. McKenna, I. Cornax, T. O'Brien, M.G. O'Sullivan, I.L. Weissman, J.F. Modiano

Writing, review, and/or revision of the manuscript: K. Weiskopf, K.L. Anderson, D. Ito, P. Schnorr, H. Tomiyasu, A.M. Ring, K. Bloink, J. Efe, I. Cornax, S. Prohaska, T. O'Brien, M.G. O'Sullivan, I.L. Weissman, J.F. Modiano

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.L. Anderson, S. Rue, I. Cornax, S. Prohaska

Study supervision: K. Weiskopf, I.L. Weissman, J.F. Modiano

Other (provided input into study design, developed the anti-CD20 antibody described in the study, and provided antibody materials used in study): S. Rue

Other (obtained funding): J.F. Modiano

Other (participated in discussions, planning the study, managed/oversaw manufacture of Hu5F9-G4): S. Prohaska

We thank T. DePauw, R. Majeti, J. Volkmer, J. Liu, T. Storm, L. Jerabek, A. McCarty, S. Willingham, P. Ho, M. Frank, O. Weiskopf, C. Schnorr, members of the Weissman and Modiano laboratories, and the CD47 Disease Team for technical assistance, discussions, and reagents. The authors gratefully acknowledge funding from donors to the Animal Cancer Care and Research Program of the University of Minnesota that helped support this project.

Research reported in this article was supported by grants F30 CA168059 (to K. Weiskopf), F30 CA195973 (to K.L. Anderson), P01 CA139490 (to I.L. Weissman), P30 CA077598 (Comprehensive Cancer Center Support Grant to the Masonic Cancer Center, University of Minnesota), and the Stanford Medical Scientist Training Program T32 GM07365 (to K. Weiskopf, A.M. Ring, and A. Barkal) from the NIH, grants D13CA-033 (to J.F. Modiano) and D12CA-302 (to D. Ito) from Morris Animal Foundation, the Stanford University SPARK Program (to K. Weiskopf and A.M. Ring), the Joseph & Laurie Lacob Gynecologic/Ovarian Cancer Fund (to I.L. Weissman), the Virginia and DK Ludwig Fund for Cancer Research (to I.L. Weissman), an Anonymous Donors Fund (to I.L. Weissman), the Siebel Stem Cell Institute (to I.L. Weissman), the Thomas and Stacey Siebel Foundation (to I.L. Weissman), and the Skippy Frank Fund for Life Sciences and Translational Research (to I.L. Weissman and J.F. Modiano). J.F. Modiano is supported by the Alvin and June Perlman Chair in Animal Oncology, University of Minnesota College of Veterinary Medicine.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.