Signal regulatory protein alpha blockade potentiates tumoricidal effects of macrophages on gastroenterological neoplastic cells in syngeneic immunocompetent mice

Abstract Aim Immunotherapies blocking the CD47‐SIRPα pathway by targeting CD47 enhance macrophage phagocytosis of neoplastic cells in mouse models. As SIRPα exhibits relatively restricted tissue expression, SIRPα antagonists may be better tolerated than agents targeting CD47, which is ubiquitously expressed in many tissues. Here, we investigated the therapeutic impact of monoclonal antibodies (mAbs) against CD47 and/or SIRPα on gastroenterological tumors in syngeneic immunocompetent mouse models. Methods We used in vitro and in vivo phagocytosis assays in C57BL/6J (B6) mice to investigate anti‐CD47/SIRPα mAb effects on Hepa1‐6 and CMT93 originating from B6 mice. The influence of these mAbs on macrophage transmigration was also assessed. To investigate anti‐SIRPα mAb therapy‐induced inhibitory effects on sporadic colon cancer growth, we used a CDX2P9.5‐NLS Cre;APC + /FLOX (CPC‐APC) mouse model. Results Systemic anti‐SIRPα mAb administration significantly increased Hepa1‐6 and CMT93 cell susceptibility to macrophage phagocytosis, both in vitro and in vivo. Conversely, similarly administered anti‐CD47 mAb did not promote macrophage phagocytosis of target cells, whereas cells incubated with anti‐CD47 mAb prior to inoculation were more susceptible to macrophage phagocytosis. In vitro cell migration assays revealed that binding with anti‐CD47 mAb inhibited macrophage transmigration. Anti‐SIRPα mAb treatment inhibited tumor progression in CPC‐APC mice and significantly improved overall survival. Anti‐CD47 mAb administration in vivo eliminated the phagocytosis‐promoting CD47 blockade effect, probably by inhibiting macrophage transmigration/chemotaxis. In contrast, anti‐SIRPα mAb exhibited enhanced macrophage phagocytic activity and marked anti‐tumor effects against gastroenterological malignancies. Conclusion SIRPα mAb augmentation of macrophage phagocytic activity may represent an effective treatment strategy for human gastrointestinal tumors.

cytosolic protein tyrosine phosphatase SHP-1 and/or SHP-2 recruitment and activation. [1][2][3][4] For certain cells (i.e., erythrocytes, platelets, or leukocytes), surface CD47 can protect against macrophagemediated phagocytosis by binding to the inhibitory macrophage receptor SIRPα. [5][6][7] The interaction of SIRPα on macrophages with CD47 on leukemia and even solid tumor cells also prevents phagocytosis of such neoplastic cells. Immunotherapies intended to block the CD47-SIRPα pathway have been proven to augment macrophage phagocytosis of several different types of neoplastic cells, primarily in mouse xenogeneic and even syngeneic transplantation models 8-13 ; however, their preclinical activities in further clinically relevant models remain to be elucidated.
We have previously shown that the interspecies incompatibility of CD47 is responsible for human macrophage-mediated phagocytosis of xenogeneic porcine cells. Porcine CD47 does not induce SIRPα tyrosine phosphorylation in human macrophages, and porcine cell manipulation for expression of human CD47 markedly reduces their susceptibility to human macrophage-mediated phagocytosis. 14 Beyond such interspecies incompatibility, the non-obese diabetic (NOD) mouse strain provides a better background for human leukocyte and cancer cell engraftment than those of other strains with equivalent immunodeficiency-related mutations owing to the stronger engagement of the SIRPα inhibitory receptor with human CD47, preventing engulfment of human grafts. 15 By using NOD mice with multiple immunodeficient phenotypes, anti-human CD47 monoclonal antibodies (mAbs), as inhibitors of the CD47-SIRPα interaction, have been shown to display anti-tumor activity against various human cancers including leukemia and solid tumors. 9,13 In these xenograft models, anti-human CD47 mAb targets human CD47 on inoculated human cancer cells but does not recognize mouse CD47 ubiquitously expressed on the host mouse cells. This has raised concerns that effects caused by CD47 binding to multiple other ligands, such as integrins and thrombospondin, which govern a number of processes in normal tissues, might be overlooked. Although examined in mouse models with locally and mainly subcutaneously transplanted syngeneic tumors, [8][9][10][11][12][13] this concern remains to be investigated following systemic tumor inoculation. In the present study, we investigated the anti-tumor effects of mAbs targeting CD47 or SIRPα, which exhibits relatively restricted tissue expression, by use of a mouse model in which syngeneic gastrointestinal cancer cells are intraperitoneally inoculated and a gene-targeting mouse model in which colon cancers grow sporadically.

CDX2P9.5-NLS Cre and APC LOX/FLOX/FLOX mouse embryos were
obtained from the University of Michigan. These embryos were transferred to pseudopregnant B6 mice, and those carrying the CDX2P9.5-NLS Cre recombinase transgene and a loxP-treated APC allele (CPC-APC mice) primarily developed colorectal adenocarcinomas from 9 weeks. 16

| Anti-SIRPα and anti-CD47 mAbs
Anti-SIRPα mAb was prepared from My-1 hybridoma cells as previously reported. 17 Hybridoma cells were grown in hypoxanthine-aminopterin-thymidine medium supplemented with IL-6, and culture supernatants were screened for Abs reactive to SIRPα-expressing leukocytes by flow cytometry (FCM). Anti-SIRPα mAb was prepared in ascitic fluid from ICR nu/nu mice, determined to be of the IgG type, and purified using protein G-affinity chromatography. 17 Miap301 hybridoma cells producing anti-CD47 mAb were kindly donated by P.

| Lentiviral-encoded small hairpin RNA (shRNA) knockdown of Hepa1-6 cells
ShRNA constructs targeting knockdown of mouse CD47 or a GFP control were transduced into Hepa1-6 and CMT93 cells as follows.
Cells were seeded into 48-well plates (BD Falcon, San Diego, CA, USA) and incubated at 37°C for 18-20 h in a humidified atmosphere with 5% CO 2 . Hexadimethrine bromide (Sigma-Aldrich, St. Louis, MO, USA) was then added to each well. An appropriate amount of lentiviral particles at a suitable multiplicity of infection was also added to appropriate wells. Cells were incubated with the viral particle mixture at 37°C overnight. CD47 protein level knockdown was determined by staining with anti-CD47 mAb (clone miap301) with fold knockdown calculated by determining the reduction of mean fluorescence intensity normalized over isotype-control antibody. The following oligonucleotides were used to knockdown

| In vivo mAb treatment in CPC-APC mice
CPC-APC mice originated from Apc F/wt mice harboring a CDX2-Cre transgene in which colorectal tumorigenesis is driven by APC allelic loss. These mice were administered intraperitoneal injections of either 400 μg/mouse of rat anti-mouse IgG control Ab (Jackson ImmunoResearch, West Grove, PA, USA) or anti-SIRPα mAb (My-1) once weekly from the eighth week until the day of harvest. For tumor size evaluation, mouse colonoscopy was performed using a grading system according to tumor circumference: grade 1 (very small but detectable tumors) and grades 2-5 (tumors occupying up to one-eighth-[grade 2]; a quarter-[grade 3]; half-[grade 4]; or more than half [grade 5] of the colonic circumference). We previously reported that the colonoscopy evaluation procedure tumor detection specificity was 1.00 and the sensitivity was 0.98. 18 All mice were killed at week 20 to assess colorectal cancer development by histological analysis. To evaluate anti-tumor effects, mice were euthanized at week 20 to assess colorectal cancer development via histological analysis. This experiment was independent from the survival experiment. The experimental schema is shown in supporting information (supplemental material and methods).

| Flow cytometry
Cell suspensions were pre-incubated with anti-CD16/32 (2.4G2) mAb to block Fcγ II/III receptors and stained for 15 min with the following fluorochrome-conjugated mAbs in a six-color staining combination. Jose, CA, USA). Data were analyzed using FlowJo 7.6.5 (Tree Star Inc., Ashland, OR, USA). For human in vitro assay, APC-labeled anti-CD14 (61D3) was used.

| In vitro phagocytosis assay
Target cells were labeled with 5 μmol/L carboxyfluorescein succinimidyl ester (CFSE) (Molecular Probes, Eugene, OR, USA), gently mixed, and incubated for 15 minutes at 37°C in a CO 2 incubator protected from light. 19 To prepare peritoneal macrophages, peritoneal cells were harvested from B6 mice after intraperitoneal phosphate buffered saline (PBS) injection, plated in a 24-well plate (BD Biosciences), and cultured at 37°C for 2 h. Macrophages were used after non-adherent cells were washed away. 20   Target cells were re-suspended in M199 medium (Sigma-Aldrich) and then injected into the peritoneal cavity. After 3 h, peritoneal cells were harvested and macrophages that had phagocytosed target cells could be identified as F4/80 + CFSE + populations by FCM analysis.

| Cell migration assay
A cell migration assay using a Transwell culture system was performed as described previously. 21 Briefly, B6 peritoneal macrophages were labeled with PKH26 using PKH26 Fluorescent Cell Linker Kits

| Immunohistochemical analysis
Dissected colon fragments were immediately immersed in Tissue Tek OCT compound (Sakura Finetel, Torrance, CA, USA) and cryopreserved in liquid nitrogen. Immunohistochemistry was performed on 6-μm-thick sections incubated with primary biotinylated Abs specific for F4/80, CD47 (EPR4150(2), Abcam) followed by incubation with streptavidin-HRP (Vector Laboratories, Burlingame, CA, USA) and detection using 3-amino-9-ethylcarbazole (Sigma-Aldrich) as a substrate. Immunoreactive cells were counted at 400× magnification and normalized against the colon tissue surface area. The sections were stained with hematoxylin and eosin (HE).

| In vitro phagocytosis assay using human reticuloendothelial macrophages
Human reticuloendothelial macrophages were isolated as described previously. 14 In brief, the mononuclear cells were isolated from the perfusion effluents of liver allografts for clinical liver transplantation by gradient centrifugation with Separate-L (Muto Pure Chemicals Co., Ltd, Tokyo, Japan). To prepare macrophages, cells were plated in a 24-well plate (BD Biosciences) and cultured at 37°Ϲ for 2 h.
Macrophages were used after non-adherent cells were washed away.
The purity of CD14 + macrophages was confirmed by FCM analysis immunofluorescence using FACS Calibur (Becton Dickinson, Mountain View, CA, USA). More than 95% of the cells demonstrated positivity for the CD14 antigen. Without any pre-culture, freshly isolated human RE macrophages were immediately subjected to the phagocytosis assays. Anti-human SIRPα mAb (4C7) and anti-human CD47 mAb (B6H12) were used for human phagocytosis assay using CFSElabeled human hepatoma cell line (Huh7) as targets. Huh7 cells were purchased from The Japanese Cancer Research Resources Bank and were maintained in 10% RPMI. Ethical approval for this study was obtained from the Ethics Committee at The Hiroshima University Hospital. Informed consent was obtained from all donors for participation in this study.

| Statistical analysis
Data are presented as the means ± SD and were analyzed by the Mann-Whitney U-test. Survival curves were generated using the Kaplan-Meier method and compared between different groups by log-rank tests. A P-value of <0.05 was considered statistically significant. All statistical analyses were performed using SPSS software (version 22; IBM Corp., Armonk, NY).
CD47 mRNA levels of CD47KD#1 and CD47KD#2 Hepa1-6 cell lines were decreased when compared with naïve Hepa1-6 cells (Figure 1B). The decrease of CD47 was also confirmed by fluorescent immunostaining ( Figure S2A). CFSE-labeled tumor cells were used as targets, and peritoneal macrophages isolated from B6 mice were used as effectors in an in vitro phagocytosis assay in which the macrophages engulfing target cells could be identified by FCM. Both CD47KD#1 and CD47KD#2 Hepa1-6 cells were significantly more susceptible to macrophage-mediated phagocytosis compared to scrambled Hepa1-6 cells (P = 0.001 for both) ( Figure 1C,D). These results were consistent with those of the in vivo phagocytosis assay, wherein CFSE-labeled CD47KD#1 and CD47KD#2 Hepa1-6 cells were more susceptible to phagocytosis than scrambled Hepa1-6 cells following intraperitoneal injection into B6 mice (P = 0.001 for both) ( Figure 1E,F). FCM analysis revealed that CD47KD#1 CMT93 cell CD47 expression levels were reduced to >10% that of control vector-transfected CMT93 cells (scrambled CMT93 cells), whereas CD47KD#2 CMT93 cells sustained 50% of control CD47 expression levels ( Figure S1A). CD47 mRNA levels of CD47KD#1 and CD47KD#2 CMT93 cell lines were decreased when compared with naïve CMT93 cells ( Figure S1B). The decrease of CD47 was also confirmed by fluorescent immunostaining ( Figure S2B). In vivo phagocytosis assays also revealed that CD47KD CMT93 cells were susceptible to macrophage phagocytosis in a CD47 expression leveldependent manner (P = 0.034) ( Figure S1C and D).

| Treatment with anti-SIRPα mAb increases gastroenterological tumor cell susceptibility to macrophage-mediated phagocytosis
We next investigated the therapeutic impact of mAbs blocking As neither Hepa1-6 nor CMT93 cells express surface SIRPα (Figure S2A,B), anti-SIRPα mAb exclusively targeted peritoneal macrophages in these models. Notably, the similarly administered anti-CD47 mAb did not promote Hepa1-6 and CMT93 cell macrophage phagocytosis in the in vivo phagocytosis assay, and combined anti-SIRPα and anti-CD47 mAb administration by i.p. did not lead to any synergistic macrophage phagocytosis effects ( Figure 3A,B). Since IgG2 subclass of anti-CD47 mAb was used in this study, biding anti-CD47 mAb to CD47 on inoculated cancer cells theoretically whereas anti-SIRPα mAb did not affect transmigration through the membrane, suggesting that signaling through CD47, rather than SIRPα, may be involved in macrophage transmigration/chemotaxis ( Figure 4A,B). Hence, it is likely that in vivo systemic anti-CD47 mAb administration will lead to less frequent contact with macrophages and tumor cells, limiting its effect on macrophage phagocytosis.
We further evaluated the effects of anti-CD47 and anti-SIRPα mAbs by inoculating Hepa1-6 and CMT93 via the portal vein in B6 mice as a long-term in vivo study. Those cancer cells were not able to be constantly engrafted in untreated syngeneic B6 mice, probably reflecting their vigorous innate immune responses. We found that depletion of NK and NKT cells by using NK1.1 mAb in mice achieved liver metastasis of colorectal cancer cell CMT93 but could not do so with HCC Hepa1-6. In the colorectal liver metastasis model with CMT93, the treatment with anti-SIRPα mAb significantly prolonged the survival of mice inoculated with CMT93, but that with anti-CD47 mAb did not ( Figure 5).

| Anti-SIRPα mAb therapy inhibits tumor progression in a sporadic colon cancer mouse model with conditional mutations in APC
To investigate anti-SIRPα mAb therapy inhibitory effects on sporadic solid neoplastic tumor growth, we used a conditional APC knockout CPC-APC mouse model, in which colon-preferential gene targeted mice recapitulate human colorectal cancer. 16 We assessed anti-SIRPα

| Anti-SIRPα mAb treatment enhances phagocytic activity of human reticuloendothelial macrophages against human hepatoma cells
We carried out an additional study of in vitro phagocytosis assay using human hepatoma cell line (Huh7), and human reticuloendothelial macrophages obtained from liver allograft perfusate collected at liver transplantation, as a further clinically relevant model. Consistent with the results from mouse studies, the treatment with anti-SIRPα mAb significantly enhanced phagocytic activity of human macrophages against human hepatoma cell line, compared with that with isotype-matched control Ab, whereas that with nti-CD47mAb rather reduced (Figure 8).

| DISCUSSION
CD47 functions as a phagocytosis inhibitor through ligation of SIRPα expressed on phagocytes, leading to tyrosine phosphatase activation and inhibition of myosin accumulation at the phagocytic synapse submembrane assembly site. 23 CD47 loss leads to homeostatic phagocytosis of aged or damaged cells. 5,7,24 We have demonstrated that interspecies CD47 incompatibility in xenotransplantation leads to SIRPα-mediated "don't eat me" signal absence, resulting in the rejection of xenogeneic cells by human macrophages. 14,20 Whereas macrophages can be activated by pro-phagocytic signaling pathways through activating receptors such as Fcγ and complement receptors, their phagocytic activity is also controlled by immune inhibitory receptor the signal strength. 25 Additionally, lectin-mediated carbohydrate binding may provide activating signals to macrophages without opsonization. 26 We have also demonstrated that a CD47-SIRPα inhibitory signal induced by genomic manipulation overrides such an activating signal delivered to macrophages by xeno-antigens.
Increased cell surface heterophilic carbohydrate antigen expression can be targeted by macrophages, such as Thomsen-Friedenreich antigen (Galß1-3GalNAcα-), a common feature in malignant and premalignant epithelia. 27 Hence, it is likely that CD47-SIRPα signaling may also override the activating signals delivered to macrophages by cancer antigens. Emerging evidence indicates that in various neoplastic cells including solid cancers, CD47 expression is required to avoid innate immune surveillance and elimination by phagocytosis.
CD47 is expressed on multiple human tumor types including acute and chronic myeloid leukemia, acute lymphoblastic leukemia, non-Hodgkin's lymphoma, multiple myeloma, bladder cancer, and other solid tumors. [8][9][10]12,30,31 Whereas CD47 is ubiquitously expressed at low levels on normal cells, multiple tumors express increased CD47 levels compared to their normal cell counterparts. 10,29 Although it has been recently demonstrated that hypoxiainducible factor 1 directly activates CD47 gene transcription in hypoxic solid cancer cells, 32 the molecular mechanisms regulating CD47 expression in cancer cells have not been fully determined.
Nevertheless, numerous studies have demonstrated that targeting of CD47 produces noticeable effects on tumor growth inhibition and metastasis prevention of various human cancer types in immunodeficient mouse xenotransplantation models. [9][10][11]33 Recently, studies using syngeneic immunocompetent mouse models have shown that Consistent with these previous findings, which suggest that CD47 targeting represents a potentially effective therapeutic strategy, intraperitoneally administered anti-CD47 mAb did not promote macrophage phagocytosis of gastrointestinal cancer cells in our syngeneic immunocompetent mouse model, although cancer cells incubated with anti-CD47 mAb prior to inoculation were susceptible to macrophage-mediated phagocytosis. As CD47 also serves as a membrane-associated glycoprotein that suppresses immune cell function, anti-CD47 mAb binding with host immune cells may influence their intrinsic functions. 21,22 Notably, murine dendritic cell chemotaxis is significantly reduced by anti-CD47 mAb treatment. 38 44 Their findings suggest that there is a fine-tuned collaborative action between SIRPα expression on macrophages and tumor progression.
SIRPα is tyrosine phosphorylated and sequestrates SHP2 from IKKß to PI3K regulatory subunit PI3Kp85, resulting in affecting PI3K-Akt and NF-κB pathways in the tumor microenvironment. Therefore, it is likely that administration of anti-CD47 and/or anti-SIRPα mAb interferes Akt and NF-κB activation in macrophages, influencing macrophages migration, survival, cytokines production, and/or angiogenesis in tumor sites. 44 We have demonstrated that anti-SIRPα mAb promotes phagocytic activity of macrophages not only against hepatoma but also against colon carcinoma cells, but anti-CD47 mAb does not.
It remains to be elucidated whether anti-SIRPα mAb causes functional alteration of tumoricidal macrophages or not.
In conclusion, we have demonstrated that anti-SIRPα mAb exhi-