Graft failure in the transplantation of hematopoietic stem cells occurs despite donor-host genetic identity of human leukocyte antigens, suggesting that additional factors modulate engraftment. With the nobese diabetic (NOD)-severe combined immunodeficiency (SCID) xenotransplantation model, we found that the NOD background allowed better hematopoietic engraftment than did other strains with equivalent immunodeficiency-related mutations. We used positional genetics to characterize the molecular basis for this strain specificity and found that the NOD Sirpa allele conferred support for human hematopoiesis. NOD SIRP-alpha showed enhanced binding to the human CD47 ligand, and its expression on mouse macrophages was required for support of human hematopoiesis. Thus, we have identified Sirpa polymorphism as a potent genetic determinant of the engraftment of human hematopoietic stem cells.

Inhibition of macrophage SIRPα–CD47 interactions mediates phagocytosis and clearance of acute myeloid leukemia stem cells.

Transplantation of human hematopoietic stem cells (HSC) has been the most important clinical application of stem cell biology. A key discovery enabling successful HSC transplantation was the recognition that HSC engraftment is controlled by human leukocyte antigen (HLA) polymorphism. Although HLA disparity plays a key role in graft rejection, graft failure can occur even in patients receiving an HLA-identical transplant, suggesting that additional, as yet uncharacterized, factors modulate engraftment. We have identified a new genetic determinant that controls the outcome of human HSC transplantation. Xenotransplantation in the non-obese diabetic/severe combine immune-deficient (NOD.SCID) mouse is the best functional assay for human HSC, based on their ability to repopulate recipient animals. We show that NOD.SCID mice provide significantly better support for human hematopoietic grafts than mice of different strain backgrounds carrying equivalent immunodeficiency mutations. To identify the molecular basis for this strain-specificity, we used positional genetics combined with in vitro and in vivo assays of human hematopoiesis. We generated reciprocal congenic strains between NOD and the related non-obese diabetes resistant (NOR) strain which is 88% identical to NOD, and found that support of human hematopoiesis is conferred by variation in a single gene, signal regulatory protein α (Sirpα) on chromosome 2. Human HSC were unable to engraft NOD.SCID mice carrying the NOR allele of Sirpα. NOD SIRPα displays 24 amino acid variations compared to NOR and C57BL/6, concentrated in the immunoglobulin IgV-like domain of this Ig-superfamily member. SIRPα is expressed on myeloid cells and mediates signals that modulate diverse macrophage functions. Indeed, lentiviral gene transfer of the NOD Sirpα variant into NOR macrophages restored their ability to support human hematopoiesis in long-term chimeric stroma-based assays. CD47, the only known cellular ligand of SIRPα, is ubiquitously expressed and modulates multiple cellular actions on hematopoietic cells including platelet activation and adhesion, and leukocyte adhesion and cytokine production. NOD SIRPα demonstrates enhanced binding to human CD47 compared to the NOR protein, suggesting that differential interaction with CD47 underlies the strong effect of Sirpα variation on support of human hematopoiesis in vitro and in vivo. Our findings reveal a novel SIRPα-dependent, macrophage-mediated mechanism critical in HSC transplantation. Future analysis of Sirpα variants within human populations could identify alleles associated with hematopoietic support in BM failure syndromes or correlated with outcomes in clinical transplantation.

Introduction CD47 binds to SIRPα on the surface of macrophages and delivers a “do not eat” signal that suppresses phagocytosis. There is increasing evidence that acute myeloid leukemia (AML) stem cells exploit the CD47-SIRPα pathway to escape macrophage-mediated destruction. Blockade of CD47 using a soluble SIRPα-Fc fusion protein (SIRPαFc) has emerged as a promising strategy to neutralize the suppressive effects of CD47 and promote the eradication of AML cells. However, little information is available regarding the optimal structure of SIRPαFc. In particular, the influence of the Fc region, which can mediate antibody-dependent cellular cytotoxicity and complement activation, on anti-leukemic activity and toxicity has not been explored. Results We have generated three unique human SIRPαFc fusion proteins that vary in their Fc regions: SIRPα-G1, which contains the Fc region from human IgG1 with full effector activity; SIRPα-G4, bearing the Fc region from human IgG4, which has low effector activity; and SIRPα-G4m, which possesses a mutated human IgG4 Fc region that is devoid of any effector activity. These three fusion proteins were tested for their ability to promote macrophage-mediated phagocytosis of patient-derived AML cells in vitro. Although all three proteins were able to stimulate tumor cell destruction, SIRPα-G4m was clearly the least potent, while SIRPα-G1 and SIRPα-G4 showed similar activity. Next, the anti-leukemic activity of the fusion proteins was assessed in an AML xenograft model in NOD.SCID mice. SIRPα-G1 induced a profound anti-leukemic effect and was superior to both SIRPα-G4 and SIRPα-G4m, particularly with respect to eradicating tumor cells within the transplanted femur. Thus, while only a low level of Fc activity was required for maximal pro-phagocytic activity in vitro, full effector activity (human IgG1) provided superior anti-leukemic activity in vivo. The strong anti-tumor activity of this fusion protein presumably results from the simultaneous delivery of a positive macrophage activating signal (through Fc receptors) and blockade of the negative “do not eat” signal from CD47. Increased Fc effector activity could also carry the risk of increased toxicity. Since human SIRPα has no measurable binding to mouse CD47, to assess tolerability in mice we generated a surrogate fusion protein consisting of NOD mouse SIRPα linked to a mouse IgG2a Fc region with full effector function (mSIRPα-G2a). Repeat administration of high dose mSIRPα-G2a to mice (50 mg/kg IP twice per week for 8 weeks) produced no adverse clinical effects. No abnormalities were observed in hematological parameters, (including erythrocyte, platelet and leukocyte counts) or bone marrow CD150+CD48- LSK hematopoietic stem cells, nor were gross or microscopic changes noted in any tissue. Furthermore, taking advantage of a fortuitous cross-reactivity between NOD SIRPα and human CD47, we conducted a xenograft study with patient-derived AML cells using the mSIRPα-G2a fusion protein. Compared to control Fc, mSIRPα-G2a profoundly reduced leukemic burden in both the injected femur and non-injected bone marrow at doses significantly below the 50 mg/kg used in the tolerability studies. Thus, a mouse surrogate fusion that can bind both human CD47 on xenograft AML cells and endogenous CD47 on host tissue is both safe and effective. A pilot repeat-dose toxicity study using various human SIRPαFc proteins is currently underway in non-human primates. Conclusions These results demonstrate that SIRPαFc fusion proteins that combine Fc activity with CD47 blockade lead to effective AML destruction in vitro and in vivo, and are well tolerated in mice. Thus the therapeutic window in a homologous model system appears to be sufficiently wide to proceed with formal IND-enabling studies. On the basis of these findings we are moving forward with the development of a SIRPαFc therapeutic for the treatment of AML. Disclosures: Uger: Trillium Therapeutics/Stem Cell Therapeutics: Employment. Pang:Trillium Therapeutics/Stem Cell Therapeutics: Employment. Wong:Trillium Therapeutics/Stem Cell Therapeutics: Employment. House:Trillium Therapeutics/Stem Cell Therapeutics: Employment. Dodge:Trillium Therapeutics/Stem Cell Therapeutics: Employment. Viau:Trillium Therapeutics/Stem Cell Therapeutics: Employment. Vigo:Trillium Therapeutics/Stem Cell Therapeutics: Employment. Tam:Trillium Therapeutics/Stem Cell Therapeutics: Employment. Truong:Trillium Therapeutics/Stem Cell Therapeutics: Employment. Jin:Trillium Therapeutics/Stem Cell Therapeutics: Research Funding. Malko:Trillium Therapeutics/Stem Cell Therapeutics: Research Funding. Ho:Trillium Therapeutics/Stem Cell Therapeutics: Research Funding. Prasolava:Trillium Therapeutics/Stem Cell Therapeutics: Research Funding. Danska:Trillium Therapeutics/Stem Cell Therapeutics: Research Funding. Wang:Trillium Therapeutics/Stem Cell Therapeutics: Research Funding. Petrova:Trillium Therapeutics/Stem Cell Therapeutics: Employment.

476 Introduction: A large body of work has shown that acute myeloid leukemia (AML) clones are hierarchically organized and maintained by leukemia initiating cells (AML-LSC). However, little is known about molecular regulators that govern AML-LSC fate. Using the non-obese diabetic (NOD)–severe combined immunodeficiency (SCID) xenotransplantation model, our group recently found that CD47-SIRPα protein interaction is essential for repopulation of normal hematopoietic stem cells (HSC) in mice (Takenaka et al, Nat Immunol 2007). The NOD background conferred the best support for human engraftment, whereas mice with other polymorphisms of Sirpa could not be engrafted (i.e. NOD.NOR-Idd13.SCID). CD47 on human AML contributes to pathogenesis by inhibiting phagocytosis of leukemia cells through CD47-SIRPα interaction (Jaiswal et al, Cell 2009). CD47 is increased on human AML LSCs compared to normal HSCs, and high levels of CD47 are associated with poor patient outcome (Majeti et al, Cell 2009). This indicates that CD47 may function as an important molecular regulator of AML-LSC fate through communication with SIRPα protein expressed on cells of the innate immune system. Results: Consistent with published results, we observed increased CD47 expression in primary AML cells compared to normal cord blood cells. We next investigated the functional relevance of CD47 for AML-LSCs by xenografting primary human AMLs into NOD.SCID and NOD.NOR-Idd13.SCID (Idd) mice. Following intravenous (i.v.) transplantation, none of three primary human AML samples could engraft Idd mice while robust engraftment in NOD.SCID mice was observed, consistent with our previous data using normal human HSCs. When the same samples were transplanted intrafemorally (i.f.), 2 of 6 Idd mice showed engraftment in the injected femur but no engraftment in other bones or the spleen; evidence of the latter is linked to stem cell function. To assess the role of innate immunity, we pre-treated mice with antibody against murine CD122 which depletes host natural killer (NK) cells and macrophages. Engraftment in the injected femur was observed in NOD.SCID mice (43/43) for all 10 AML samples tested, and interestingly in 31 of 42 Idd mice (8/10 AML samples tested), with similar engraftment levels (Idd 37.4±6% vs NOD.SCID 53.8±5%, p=0.33). As expected, NOD.SCID mice supported migration to non-injected bones (38/43 mice, 10/10 AML samples tested). In contrast to results obtained in the absence of anti-CD122 pre-treatment, engraftment in non-injected bones was now detectable in 8 of 42 Idd mice (2/10 AML samples tested), pointing to improved migration of AML-LSCs following depletion of NK cells and macrophages. However, the engraftment level in non-injected bones was significantly lower in Idd compared to NOD.SCID mice (2.4±4% vs 63.2±6%, p=0.001). Moreover, AML-LSCs were unable to repopulate the spleens of Idd mice. Homing assays revealed decreased homing to BM and spleen 16 hours following i.v. injection in Idd compared to NOD.SCID mice. Conclusion: Our results support the hypothesis that CD47-SIRPα interaction is critical for engraftment of AML-LSCs. Attenuation of CD47-SIRPα interaction (as in Idd mice) leads to decreased engraftment ability of AML-LSCs. Enhancement of engraftment in the injected femur and in some cases migration to other bones in Idd mice following anti-CD122 treatment is likely mediated through interference with cells of the innate immune system. Interruption of CD47-SIRPα signaling through targeting of either CD47 or SIRPα provides a potential therapeutic approach for eradication of AML-LSCs. Disclosures: Dick: Roche: Research Funding; CSL Ltd: Research Funding.