MODULATING PHAGOCYTOSIS: THERAPEUTIC TARGETING OF THE ANTI- PHAGOCYTIC SIGNAL CD47 IN MALIGNANCIES

A DISSERTATION SUBMITTED TO THE DEPARTMENT OF CANCER BIOLOGY AND COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

Mark Ping Chao November, 2010

© 2011 by Mark Ping Chao. All Rights Reserved. Re-distributed by Stanford University under license with the author.

This work is licensed under a Creative Commons Attribution- Noncommercial 3.0 United States License. http://creativecommons.org/licenses/by-nc/3.0/us/

This dissertation is online at: http://purl.stanford.edu/jk771dx5817

ii I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Irving Weissman, Primary Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Ravindra Majeti, Co-Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Michael Clarke

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Ronald Levy

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Beverly Mitchell

Approved for the Stanford University Committee on Graduate Studies. Patricia J. Gumport, Vice Provost Graduate Education

This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file in University Archives.

iii

iv ABSTRACT

Growing evidence indicates that cancers are composed of functionally heterogeneous cells that are organized by a cellular hierarchy and are maintained by a -like population. These cancer stem cells (CSCs) are defined by their ability to differentiate and replicate the heterogeneity of the tumor and by their ability to self-renew. This growing body of research has led to important implications in cancer therapy design postulating that therapies not effectively targeting the CSC are unlikely to eradicate disease. Thus, the development of CSC targeted therapies is necessary to improve current clinical outcomes. This body of work identifies CSC targeted therapies in acute myeloid (AML) through identifying ligands specifically expressed on leukemia stem cells (LSC). Using expression screening approaches, CD47 was identified as a ligand highly expressed on LSC, which inhibits phagocytosis by binding its ligand SIRPα on phagocytes. CD47 was up-regulated on several human with over-expression contributing to leukemia evasion of phagocytosis and correlating with a worse clinical prognosis in several AML and acute lymphoblastic leukemia (ALL) patient cohorts. Blockade of the CD47-SIRPα interaction with a monoclonal blocking against CD47 enabled phagocytosis of AML and ALL cells and eliminated disease in leukemia- engrafted mice. CD47 is ubiquitously expressed at low levels in several normal tissue types, indicating that administration of anti-CD47 antibody to human patients could lead to significant off-target toxicity. The toxicity profile of anti-CD47 antibody was investigated, demonstrating that anti-CD47 antibody treatment specifically targeted tumor but not normal cells, resulting in minimal toxicity in vivo. The selective targeting of tumor but not normal cells by anti-CD47 antibody was found to be due to the selective expression of the pro-phagocytic signal calreticulin on tumor but not normal cells. Lastly, the ability of anti-CD47 antibody efficacy to be augmented in a combination antibody approach was investigated. While anti-CD47 antibody enables tumor elimination through blockade of an inhibitory phagocytic signal (CD47-SIRPα ligation), it was investigated whether this antibody could be combined with a second antibody that works through classic delivery of a positive phagocytic stimulus by activating Fc (FcR) on immune effector cells. Utilizing non-Hodgkin lymphoma (NHL) as a model disease, anti-CD47 antibody synergized with rituximab (a well studied FcR- activating antibody) to eliminate tumor cells by phagocytosis while combination antibody therapy eliminated disease and achieved high cure rates in vivo in contrast to single agent therapy. In summary, this body of work identifies CD47 as a therapeutic target of several human cancers including AML, ALL, and NHL and demonstrates that a monoclonal blocking antibody against CD47 has pre-clinical efficacy in eliminating these tumors through the promotion of phagocytosis. The results of this work provide the pre-clinical rationale for the development of an anti-CD47 antibody therapy in hematologic malignancies as well as in other cancers, leading to a filing of an Investigational New Drug Application and Phase I clinical trial.

v ACKNOWLEDGEMENTS

Throughout the course of pursuing a PhD, I have enjoyed the support of countless people, all whom this would not be possible without their support. From my colleagues, collaborators, mentors, family, and friends, I am truly indebted to all of them. During my time at Stanford University I have been fortunate to be able to interact and collaborate with many phenomenal scientists both inside and apart from the Weissman and Majeti labs. I especially want to thank my collaborators and all members of the Weissman and Majeti labs for being so supportive. The unique collaborative culture, friendly environment, and eagerness to support each other both academically and personally have been invaluable to me. I am specifically indebted to Ash Alizadeh, Siddhartha Jaiswal, Christopher Park, and Max Jan for enriching my scientific training as well as being key resources during my PhD work. I especially want to thank the lab managers, Libuse Jerabek, Theresa Storm, and Feifei Zhao, all whom do more than just manage the lab and truly empower all lab members in their work. I want to specifically thank my two PhD advisors, Ravi Majeti and Irv Weissman, both of whom have impacted my development as a scientist in profound ways. They both have been invaluable to my development as a scientist and future clinician. Their passion for bench to bedside research, commitment to the scientific process, and support in the development of others are all qualities I have cherished. My mentorship by these two individuals has been one of the defining qualities of my PhD experience. Throughout my time at Stanford I have also had relationships that have fostered my career development as a future clinician and hematologist-oncologist. Specifically, the Stanford Division of Hematology, including Jason Gotlib and Linda Boxer, and my academic advisor, Susan Knox, have been instrumental in guiding my interests in this field. I would like to also thank my PhD dissertation committee members Ron Levy, Michael Clarke, Beverley Mitchell, and Maximillian Diehn for their time and efforts in guiding my research. I also would like to thank my funding sources during this time including the Stanford Medical Scholars Program, Howard Hughes Medical Institute, and Cancer Biology Program, all whom provided opportunities to allow me to pursue my research. Lastly, I would like to thank my family and friends, of whom none of this would be possible. My friends have been the support, guidance, and enjoyment that have fueled me during my successes and failures. And my family and fiancee, whom my love cannot be greater, have influenced me in all phases of my life and have developed my passions, motivations, and personal character.

vi TABLE OF CONTENTS

CHAPTER 1 Introduction to Cancer Stem Cells: New Implications for Cancer Therapy…………………………….1 SUMMARY……………………………………………………………………………………………..2 The Cancer Stem Cell Hypothesis………………………………………………………………………2 Techniques to Identify Cancer Stem Cells……………………………………………………………....5 Identifying Cancer Stem Cells in Solid Tumors………………………………………………...... 6 Cancer Stem Cells in Hematologic Malignancies………………………………………………...…….9 Clinical Utility of Cancer Stem Cells…………………………………………………………………...9 CONCLUSION…………………………………………………………………………………………12

CHAPTER 2 CD47 is an adverse prognostic factor and therapeutic antibody target on human acute myeloid leukemia stem cells……………………………………………………………………………………...13 SUMMARY…………...... …………...... …………...... 14 INTRODUCTION…………...... …………...... …………...... 14 EXPERIMENTAL PROCEDURES……………...... ……………...... 15 RESULTS……………………………………...... 18 CD47 is More Highly Expressed on AML LSC Than Their Normal Counterparts and is Associated with the FLT3-ITD Mutation……………………………………………...... 18 Identification and Separation of Normal HSC From Leukemia Cells in the Same Patient Based On Differential CD47 Expression……………………………………………………...21 Increased CD47 Expression in Human AML is Associated with Poor Outcomes……...... 24 Monoclonal Directed Against Human CD47 Preferentially Enable Phagocytosis of AML LSC by Human Macrophages…………………………….…………..27 Monoclonal Antibodies Directed Against Human CD47 or SIRPα Enable Phagocytosis of AML LSC by Mouse Macrophages…………………………………….…...29 A Monoclonal Antibody Directed Against CD47 Does Not Induce of AML LSC……………………………………………………………………………………………30 A Monoclonal Antibody Directed Against Human CD47 Inhibits AML LSC Engraftment and Depletes AML In Vivo……………………………………………………..32 A Monoclonal Antibody Directed Against Human CD47 Enables Phagocytosis of AML In Vivo and Targets AML LSC……………………………………………………………….35 DISCUSSION………………………………………………………………………….…………..38

CHAPTER 3 A monoclonal antibody against CD47 eliminates human acute lymphoblastic leukemia………………42 SUMMARY…………...... …………...... …………...... 43 INTRODUCTION…………...... …………...... …………...... 43 EXPERIMENTAL PROCEDURES……………...... ……………...... 44 RESULTS……………………………………...... 47 CD47 Expression is Increased on a Subset of Human ALL Cells Compared to Normal Bone Marrow……..……………………………………………………………………………47 CD47 Expression is an Independent Prognostic Predictor in Mixed and High-Risk ALL..…..47 Blocking Monoclonal Antibodies Against CD47 Enable Phagocytosis of ALL Cells ………………….……………………………………………………………………….49 Ex Vivo Coating of ALL Cells with an Anti-CD47 Antibody Inhibits Leukemic Engraftment…………………………………………...... 51 Anti-CD47 Antibody Eliminates ALL Engraftment in the Peripheral Blood and Bone Marrow…………………………..…………………………………………………………….53

vii Anti-CD47 Antibody Eliminates ALL Engraftment in the Spleen and Liver………………..53 DISCUSSION…………….……………………………………………………………………….55

CHAPTER 4 Toxicity studies of anti-CD47 antibody: the pro-phagocytic signal calreticulin is required for anti-CD47 efficacy………………………………………………………………………………………58 SUMMARY…………...... …………...... …………...... 59 INTRODUCTION…………...... …………...... …………...... 59 EXPERIMENTAL PROCEDURES……………...... ……………...... 61 RESULTS……………………………………...... 62 A Monoclonal Antibody Directed Against Mouse CD47 Enables Phagocytosis of Mouse AML and Does Not Deplete Normal HSC In Vivo…………………………………...62 Cell surface calreticulin is expressed on cancer, but not normal, stem and progenitor cells….64 Increased CD47 expression on cancer cells protects them from calreticulin- mediated phagocytosis………………………….……………………………………………..66 Calreticulin is the dominant pro-phagocytic signal on several human cancers and is required for anti-CD47 antibody-mediated phagocytosis………………….………………….69 DISCUSSION………………………………………………………………………………..……..73

CHAPTER 5 Augmenting anti-CD47 antibody therapy: a combination antibody approach with rituximab in non-Hodgkin lymphoma………………………………………………………………………………...79 SUMMARY…………...... …………...... …………...... 80 INTRODUCTION…………...... …………...... …………...... 80 EXPERIMENTAL PROCEDURES……………...... ……………...... 81 RESULTS……………………………………...... 87 CD47 Expression is Increased on NHL Cells Compared to Normal B Cells……….………...87 Increased CD47 Expression Correlates with a Worse Clinical Prognosis and Adverse Molecular Features in Multiple NHL Subtypes ………………..……………………87 Blocking Antibodies Against CD47 Enable Phagocytosis of NHL Cells by Macrophages and Synergize with Rituximab in Vitro…………………………...……………90 Ex Vivo Coating of NHL Cells with an Anti-CD47 Antibody Inhibits Tumor Engraftment…………………….……………………………………………………………..93 Combination Therapy with Anti-CD47 Antibody and Rituximab Eliminates Lymphoma in Both Disseminated and Localized Human NHL Xenotransplant Models..…...93 Combination Therapy with Anti-CD47 Antibody and Rituximab Eliminates Lymphoma in Primary Human NHL Xenotransplant Mouse Models…………….………….97 Synergy Between Anti-CD47 Antibody and Rituximab Does Not Occur Through NK Cells or Complement………….…………………………………………………………103 Anti-CD47 Antibody Synergizes with Rituximab Through FcR-Independent and FcR-Dependent Mechanisms…………….…………………………………………………..107 DISCUSSION……………...………………………………………………………………………….109

CHAPTER 6 Therapeutic antibody targeting of CD47: Current Implications and Future Directions……………….117 SUMMARY……………………………………………………………………………………….118 DISCUSSION……………………………………………………………………………………..118 Targeting CD47 to Promote Macrophage Killing……...……………………….…………....118 Antibody targeting of CD47 on Normal Cells…………………………………….…………118 Future Directions: Augmenting macrophage effectors in anti-CD47 antibody therapy……..118 Anti-CD47 antibody: from pre-clinical therapy to clinical translation………………………121

viii

REFERENCES………………………………………………………………………………………..122

ix LIST OF FIGURES

Chapter 1: Figure 1: The cancer stem cell model…………………………………………………………………...4 Figure 2: Evaluation of cancer expression in a primary human tumor……………….7 Figure 3: Proposed mechanism of CSC-targeted therapy……………………………………………….10

Chapter 2: Figure 1: CD47 is more highly expressed on AML LSC compared to their normal counterparts……...19 Figure 2: CD47 mRNA expression correlates with cell surface expression…………………………….20 Figure 3: CD47 expression across FAB and cytogenetic subgroups of AML…………………………..21 Figure 4: Identification and separation of normal HSC from leukemia cells in the same patient based on differential CD47 expression………………………………………………………………….23 Figure 5: Derivation and distribution of CD47 mRNA expression of AML samples in low CD47 and high CD47 expression groups………………………………………………………………………25 Figure 6: Increased CD47 expression in human AML is associated with poor clinical outcomes……..26 Figure 7: Monoclonal antibodies directed against CD47 preferentially enable phagocytosis of human AML LSC by human and mouse macrophages in vitro………………………………………...28 Figure 8: Monoclonal antibodies directed against human CD47 enable phagocytosis of human AML LSC by human and mouse macrophages…………………………………………………………30 Figure 9: Immobilized anti-CD47 antibody does not stimulate apoptosis of human AML LSC……….31 Figure 10: A monoclonal antibody directed against human CD47 inhibits AML LSC engraftment in vivo……………………………………………………………………………………..32 Figure 11: A monoclonal antibody directed against human CD47 depletes AML in vivo……………..33 Figure 12: Summary of AML engraftment in anti-CD47 antibody and control IgG antibody treated mice……………………………………………………………………………………………..34 Figure 13: A monoclonal antibody directed against human CD47 enables phagocytosis of AML in vivo and targets LSC…………………………………………………………………………………36 Figure 14: Clodronate treatment depletes mouse phagocytes, but not human AML in vivo…………...37 Figure 15: A monoclonal antibody directed against human CD47 targets AML LSC in vivo………....38 Figure 16: Increased CD47 expression on AML LSC inhibits phagocytosis and can be targeted by a blocking monoclonal antibody to eliminate these cells……………………………………………39

Chapter 3: Figure 1: CD47 expression is increased on a subset of human ALL cells compared to normal bone marrow…………………………………………………………………………………….48 Figure 2: CD47 expression is an independent prognostic predictor in mixed and high-risk ALL……...50 Figure 3: Blocking monoclonal antibodies enable phagocytosis of ALL cells by human and mouse macrophages in vitro…………………………………………………………………………….51 Figure 4: Ex vivo coating of ALL cells with an anti-CD47 antibody inhibits leukemic engraftment….52 Figure 5: Anti-CD47 antibody eliminates ALL engraftment in the peripheral blood and bone marrow…………………………………………………………………………………………………..54 Figure 6: Anti-CD47 antibody eliminates ALL engraftment in the spleen and liver…………………...55

Chapter 4: Figure 1: An anti-mouse CD47 antibody enables phagocytosis of mouse leukemia in vitro and improves the mortality of leukemia-engrafted mice…………………………………………………....63 Figure 2: Anti-mouse CD47 antibody does not deplete normal HSC in vivo………………………….64 Figure 3: Cell surface calreticulin is expressed on cancer, but not normal, stem and progenitor cells …………………………………………………………………………………………………….67 Figure 4: Live calreticulin positive cells from normal human tissue cells have higher levels of CD47 compared to calreticulin negative cells………………………………………………………68 Figure 5: Increased CD47 expression on cancer cells protects them from calreticulin- mediated phagocytosis…………………………………………………………………………………70

x Figure 6: Calreticulin expression is unaffected by CD47 shRNA knockdown in Raji cells………….71 Figure 7: Cell surface calreticulin is the dominant pro-phagocytic signal on several human cancers and is required for anti-CD47 antibody-mediated phagocytosis………………………………72 Figure 8: CD47 is expressed on normal human cells…………………………………………………..73 Figure 9: Abrogation of anti-CD47 antibody-mediated phagocytosis is dose dependent on calreticulin blockade……………………………………………………………………………………74 Figure 10: Model for the integration of pro (CRT)- and anti (CD47)-phagocytic signals on normal and tumor cells at steady state and during anti-CD47 antibody therapy……………………….76

Chapter 5: Figure 1: CD47 expression is increased on NHL cells compared to normal B cells, confers a worse clinical prognosis, and correlates with adverse molecular features in multiple NHL subtypes…………………………………………………………………………………88 Figure 2: CD47 is an independent adverse prognostic factor and is associated with adverse molecular features of DLBCL and CLL………………………………………………………………..91 Figure 3: Blocking antibodies against CD47 enable phagocytosis of NHL cells by macrophages and synergize with rituximab in vitro…………………………….…………………..….94 Figure 4: Correlation of CD47 expression with anti-CD47 antibody-mediated phagocytosis, anti-CD47 antibody-mediated effect on apoptosis, and CD20/CD47 expression on primary NHL cells and cell lines…………………………………………….………………………………...... 96 Figure 5: Ex vivo coating of NHL cells with an anti-CD47 antibody inhibits tumor engraftment….....98 Figure 6: Ex vivo coating of NHL cells with rituximab inhibits tumor engraftment……………...……99 Figure 7: Combination therapy with anti-CD47 antibody and rituximab eliminates lymphoma in both disseminated and localized human NHL xenotransplant mouse models…………………...... 100 Figure 8: Combination therapy with anti-CD47 antibody and rituximab eliminates NHL in a disseminated lymphoma xenotransplant model in SCID and NSG mice…………………………...…101 Figure 9: Combination therapy with anti-CD47 antibody and rituximab eliminates lymphoma in primary human NHL xenotransplant mouse models…………………………………………………..103 Figure 10: Clinical characteristics and morphology of primary NHL patient samples (SUNHL7 and SUNHL31) used for xenotransplantation studies……………………………………………...….106 Figure 11: Synergy between anti-CD47 antibody and rituximab does not occur through NK cells or complement……………………………………...………………………………………………….108 Figure 12: Therapeutic effects of anti-CD47 antibody and rituximab in inducing apoptosis, ADCC, or CDC………………………………………………..……………………………………….110 Figure 13: Anti-CD47 antibody synergizes with rituximab through FcR-independent and FcR- dependent mechanisms……………………………………..………………………………………….112 -/- Figure 14: SIRPα expression on Fcγ macrophages and generation of F(ab’)2 fragments of anti- CD47 antibody and rituximab………………………………………………………………………….114

xi LIST OF TABLES

Chapter 1 Table 1: Cancer Stem Cells (CSCs) Identified in Primary Tumor Isolates……………………………6

Chapter 2 Table 1: Clinical and molecular characteristics of AML samples from the validation cohort and comparison between low CD47 and high CD47 expression groups………………………22 Table 2: Multivariable analysis of prognostic factors in a validation cohort of normal karyotype AML samples………………………………………………………………………………27

Chapter 4 Table 1: Treatment of wild type mice with a blocking anti-mouse CD47 monoclonal antibody causes isolated neutropenia………………………………………………………………………….....65

Chapter 5 Table 1: Summary of expression datasets of NHL patients…………………………………………89-90 Table 2: Kaplan-Meier survival analysis for antibody treatment of Raji-engrafted mice……………..113 Table 3: Kaplan-Meier survival analysis for antibody treatment of primary human NHL- engrafted mice………………………………………………………………………………...... 113

xii

CHAPTER 1

Introduction to Cancer Stem Cells: New Implications for Cancer Therapy

Portions of this chapter were published in the following article: Chao MP, Weissman IL, Park CY. Cancer Stem Cells: On the Verge of Clinical Translation. Lab Med. 2008; 39:679-686.

1 SUMMARY

Researchers have long studied malignant tumors using experimental techniques that assume that cancers are composed of identical cells; however, growing evidence indicates that cancers are composed of functionally heterogeneous cells maintained by a stem cell-like population. Identification of tumor-initiating, or cancer stem cells, has been largely guided by principles established for normal stem cells. Cancer stem cells must be capable of initiating tumors when transplanted into immunodeficient mice, and the resulting tumors must replicate the heterogeneity of the original tumor. Demonstration of the ability to self-renew, a key feature of stem cells, is achieved through serial passage of tumors in xenohosts. Given their potential roles in tumor initiation, maintenance, and , the impact of cancer stem cells on the practice of medicine is likely to be profound.

INTRODUCTION

It is widely accepted that cancer develops from normal cells through the accumulation of genetic and epigenetic alterations (1). As cancers are frequently monoclonal in nature when analyzed at the chromosomal level, it is not surprising that many have interpreted this finding to signify that malignant tumors also are composed of cells that are homogeneous with respect to their malignant properties. Interestingly, though, this view of cancer was at odds with observations made over half a century ago, when several researchers observed that in some and mouse cancers, only a small proportion of cells could give rise to new tumors (2-5). Data generated over the last 15 years now indicate that many cancers are not composed of functionally identical cells, but instead are similar to their normal tissue counterparts, being organized hierarchically and maintained by a stem cell-like population, now referred to as cancer stem cells (CSC) (Figure 1).

The Cancer Stem Cell Hypothesis Cohnheim is typically credited as the first to propose that tumors originate from stem cells when he asserted in 1875 that adult tissues contain a population of embryonic remnants “lost” during developmental organogenesis that ultimately give rise to malignancies later in life (6). Although different versions of the modern-day CSC hypothesis have been promulgated in the century after Cohnheim’s declaration, it was not until John Dick and colleagues described the presence of a unique population of human acute myeloid leukemia (AML) cells that CSCs were experimentally demonstrated. Specifically, this leukemia-initiating, or leukemia stem cell (LSC) population showed the ability to initiate leukemias in immunodeficient mice. Identification of this tumorigenic population also allowed elucidation of the hierarchical nature of leukemia, beginning with the LSC and ending with downstream differentiated components (7, 8). While it remains controversial whether or not

2 cancers arise directly from tissue stem cells, the term “cancer stem cell” has arisen to reflect the shared properties of normal stem cells and tumorigenic, or tumor-initiating, cancer cells: 1) the ability to self- renew; and 2) the ability to differentiate. While it is important to remember that the term “CSC” does not make a specific ascription as to the cell of origin for cancers, the concept that CSCs arise from transformed tissue stem cells is an attractive hypothesis (9-11). Normal stem cells, by virtue of their long life and their ability to divide to produce additional stem cells, would be able to accumulate the multiple mutations required for cancer development. More differentiated progeny, while also susceptible to mutagenic stimuli, are short-lived and cannot self-renew, and hence their acquired mutations are less likely to be propagated. Because malignant tumors frequently contain cells present at various stages of differentiation like normal tissue, it is easier to envision that cancers recapitulate normal developmental pathways than “dedifferentiate” from more mature cell types. The identification of the leukemia-initiating cell, or leukemia stem cell, in human AML was possible largely due to the technological advances that allowed researchers to prospectively isolate normal hematopoietic stem cells (HSC). Prior to the identification of the LSC, the most primitive human hematopoietic cells had been shown to lack mature blood lineage markers (Lin-), express CD34, and lack CD38 (Lin-CD34+CD38-), with Lin-CD34+CD38-CD90+ cells being highly enriched for HSC activity (12). Recognizing that AML frequently expresses CD34, similar to normal hematopoietic stem and progenitor cells, Dick and colleagues fractionated CD34+ AML blasts on the basis of CD38 expression and demonstrated that CD34+CD38- blasts were capable of initiating leukemias when transplanted into immunodeficient NOD/SCID mice (7, 8). Such leukemia-initiating activity was confined to the CD34+CD38- cell fraction, and neither CD34+CD38+ nor CD34- cells could transfer disease. Moreover, engrafted blasts could be serially transplanted into secondary recipients, formally providing functional evidence that the CD34+CD38- fraction could self-renew. Because of the immunophenotypic similarities between normal HSCs and LSCs, it has been postulated that LSCs arise from HSCs; however, studies by Blair and colleagues showed that LSCs do not express CD90 (Thy-1), a marker that is present on human HSCs (13). These observations have led to two theories addressing the origin of LSCs: 1) LSCs arise from HSCs and the loss of CD90 occurs as the result of leukemic transformation; or 2) LSCs do not arise from HSCs, with transformation to the LSC occurring in a downstream progenitor cell. In order to address this question, we recently showed that normal primitive hematopoietic cells exhibiting a phenotype similar to LSC (CD34+CD38-CD90- CD45RA-Lin-) are multipotent progenitors, suggesting that AML stem cells originate from this normally non-self-renewing population (14). Additional evidence indicates that while HSC may harbor mutations associated with leukemia, the presence of mutations is not synonymous with leukemic transformation. For example, analysis of AML t(8;21) patients in remission revealed the presence of AML1/ETO mRNA in both purified HSCs and mature hematopoietic cells (15). Studies of blast crisis of chronic myeloid leukemia (CML) suggest that aberrant acquisition of self-renewal may be a late 3 event in leukemia development, evidenced by the activation of the Wnt/β-catenin self-renewal pathway in the granulocyte-macrophage progenitor, a cell not typically capable of self-renewal (16). Taken together, these studies indicate that pre-leukemic genetic changes can take place within HSCs, but ultimate transformation to the LSC may require additional mutations that take place not in the HSC, but in a downstream progenitor population (Figure 1). Based on studies of AML LSCs and subsequently identified solid tumor CSCs, the current version of the CSC hypothesis holds that tumors are maintained by a subset of tumor cells, a.k.a CSCs, that have the ability to self-renew as well as differentiate into all the cell types that comprise the heterogeneity of a tumor (Figure 1). The remaining cancer cells, although actively proliferating and differentiating, cannot self-renew and therefore are ultimately destined to die (17). In normal tissues, stem cell growth is subject to tight homeostatic controls that control the balance between tissue maintenance and excessive growth. CSCs are distinguished from their normal tissue stem cell counterparts by the loss of these homeostatic controls. Indeed, molecular pathways regulating tissue homeostasis and stem cell self-renewal including Bmi-1, sonic hedgehog, PTEN, and Wnt/β-catenin are either aberrantly expressed or deregulated in many cancers compared to their normal tissue counterparts (18, 19).

Figure 1: The cancer stem cell model (A) Tissue stem cells can give rise to functionally identical daughter stem cells through the process of self-renewal, or they can give rise to multipotent progenitors that cannot self-renew but can differentiate into all the mature cell types in that tissue. Multipotent progenitors, in turn, give rise to more committed progenitors that eventually differentiate into mature cell types. Normal stem cells are capable of propagating genetic and epigenetic mutations because they can self-renew. (B) Cancers are organized in a similar manner to their normal tissue counterparts. Cancer stem cells can self-renew to form additional cancer stem cells, or they can divide into a daughter stem cell or a clonogenic progenitor that ultimately gives rise to more differentiated cell types that are similar in appearance

4

to the tissue of origin. Although the origin of cancer stem cells has not been directly demonstrated, it has been hypothesized that they may arise directly from a normal stem cell or alternatively may result from the aberrant acquisition of self-renewal properties in normal multipotent or committed progenitors (arrows).

Techniques to Identify Cancer Stem Cells Experimental isolation of CSCs has relied primarily on two basic laboratory methods: 1) Isolation of candidate populations based on the presence of cell surface markers - following antibody labeling, cells are separated by fluorescence-activated cell sorting (FACS) or immunomagnetic bead selection; 2) Transplantation of purified candidate CSC into immunodeficient mice - CSCs must demonstrate the capacity to initiate tumors that recapitulate the morphologic and immunophenotypic heterogeneity of the original tumor following transplantation. Self-renewal is demonstrated by the ability of engrafted tumors to serially transplant into secondary recipients. This experimental approach of xenotransplantation followed by serial transplantation has become the accepted method for defining CSCs, with this view recently confirmed by the AACR Workshop on Cancer Stem Cells (17). Several groups have utilized other methods to identify CSCs based on presumed similarities to normal stem cells including: 1) cellular exclusion of dyes by active efflux transporters frequently present in normal stem cell populations (i.e. “side-population cells”); and 2) cell-cycle status - stem cells are enriched for in slow-dividing populations. Such efforts have been met with variable success (20). Many groups have utilized these same techniques using tumor cell lines, with some reporting enrichment of engraftment activity by these methods (20); however, the relevance of these findings to primary human tumors is unclear since culture-adapted cell lines may be altered compared to their primary parental tumor cell isolates. Numerous investigators have also explored the use of mouse tumor models to study CSC biology in both hematopoietic and solid tumors (21-24). The relevance of these models will need to be verified for primary human tumors. Although definitive demonstration of CSC activity requires xenotransplantation of candidate populations directly isolated from primary tumors, some laboratories have employed techniques to expand primary tumor samples prior to purification of candidate CSC because of difficulties in obtaining large quantities of primary tissue. Such techniques include purifying CSCs from xenografts initiated by bulk tumor cells or expanding tumors in vitro under defined culture conditions. Both techniques have led to promising results. For example, CD44+/α2β1hi/CD133+ tumor-initiating cells from primary prostate cancers can be maintained for over sixty days in culture and give rise to more differentiated cells (25). In addition, distinct cell populations generate tumors in vivo after cell culture of bulk primary human (26) and human chronic myeloid leukemia cells (27). Taken together, these studies suggest that stem cell-like behavior can be demonstrated in primary tumors that have been expanded in vivo or in vitro. It will be interesting to see whether the CSC populations identified in these studies can be confirmed using primary tumor isolates.

5

Table 1: Cancer Stem Cells (CSCs) Identified in Primary Tumor Isolates *CD45+ sorted cells were used for injection † Lineage marker positive cells were also excluded ‡Tertiary transplants also performed §Transplants were performed using CSC purified from primary xenografts 1° (primary); Tx (transplant); 2° (secondary); Min. (minimum); Max (maximum); n/a (not available)

Identifying Cancer Stem Cells in Solid Tumors Prospective isolation of CSCs in solid tumors has been a significant challenge due to technical issues (e.g. the requirement of tissue dissociation to isolate cells, finding permissive sites for engraftment) and the absence of well-defined developmental hierarchies to guide the selection of putative CSC markers. Nevertheless, investigators have successfully identified CSCs in a number of solid tumors using markers also expressed on their normal primitive counterparts (Table 1). Al-Hajj et al. were the first to characterize a solid tumor CSC when they identified a population of breast cancer cells that exhibits tumor-initiating ability and recapitulates tumor

6 heterogeneity when injected into the mammary pads of immunocompromised (NOD/SCID) mice (28). Representing 11-30% of total breast cancer cells, this tumorigenic cell population was FACS purified on the basis of its CD44+CD24-/low cell surface phenotype. CD44+CD24-low engrafted tumors could be serially transplanted and recapitulate the original tumor heterogeneity, giving rise to CD44+CD24+, CD44+CD24- and CD44- cells. Moreover, 100-fold more cells from the CD44+CD24+ or CD44- cell populations lacked tumorigenic potential. Although the relationship between normal and malignant human breast stem cells remains unclear, CD44 is expressed in the basal layer of a variety of epithelia where the normal epithelial stem cell population resides (28, 29). This co-localization of CD44 expression and the normal epithelial stem cell pool is consistent with the notion of breast cancer originating from a less differentiated tissue progenitor. Using CD44 as a marker, tumorigenic populations have also been isolated from pancreatic and squamous cell carcinomas. CD44+CD24+ESA+ pancreatic adenocarcinoma cells were uniquely capable of engrafting NOD/SCID mice and exhibited increased activity of the stem-cell associated sonic hedgehog pathway (30). Head and neck squamous cell carcinomas (HNSCC) also contained a minority population of CD44+ cells that recapitulated the original tumor morphologic heterogeneity and successfully engrafted secondary tumors when injected into either NOD/SCID or Rag2/IL-2 receptor common γ –chain double knockout mice, indicating that other immunodeficient mouse strains in addition to NOD/SCID may serve as hosts for tumor xenografts (29). Similar to both normal and cancerous breast tissue, CD44+ expression was largely confined to the basal layer in HNSCC (Figure 2).

Figure 2: Evaluation of cancer stem cell marker expression in a primary human tumor Immunohistochemical staining of a well-differentiated head and neck squamous cell carcinoma (HNSCC) with an anti-CD44 antibody reveals a basal layer staining pattern (dashed lines), similar to primitive normal tissue (image courtesy of L. Ailles).

7 In addition to CD44, another cell surface , CD133, has been used to identify CSCs in human brain and colon cancer. CD133 is expressed on normal primitive neural, hematopoietic, endothelial, and epithelial cells, thus it is not surprising that it is a CSC marker on multiple tumor types. Studies by Singh and colleagues (31) showed that CD133 is expressed in a subset of cells from multiple brain tumor types that are able to initiate clonally derived neurospheres, which are aggregates of neural stem and progenitor cells grown in vitro. These neurospheres could self-renew in vitro and differentiate into all of the neuronal lineages, both in vitro and in vivo. In glioblastomas and medulloblastomas, CD133+ cells (representing 6-29% of total tumor cells) could initiate tumors when injected into the brains of NOD/SCID mice, while 100 fold more CD133- cells could not (32). In addition, these CD133+ tumor cells could be serially transplanted and gave rise to tumors with both CD133+ and CD133- cells, thus recapitulating original tumor cell heterogeneity. Similar to the neurosphere assay for primitive glial cells, CD133+ colon spheres from dissociated colon cancers could be repetitively passaged in liquid cultures while CD133- colon cancer cells invariably died under the same conditions (33). In addition, CD133+ colorectal cancer cells and CD133+ colon spheres cultured in vitro were uniquely capable of forming tumors when injected subcutaneously into the flanks of SCID mice; only CD133+ cells could be serially transplanted into secondary recipients. O’Brien and colleagues independently demonstrated the tumorigenic activity of CD133+ colon cancer cells by transplanting them into the renal capsule of NOD/SCID mice (34). Limiting dilution analysis showed the frequency of colon cancer-initiating cells as 1 in 54,000 total tumor cells and 1 in 262 CD133+ tumor cells. Thus, while this study showed that CD133 enriches for CSC activity 200-fold, it also suggested that incorporating additional cell surface markers to identify CSCs might yield more highly purified CSC populations. Dalerba et al. confirmed the utility of this approach when they demonstrated a more robust colon CSC phenotype using a combination of antibodies to isolate tumorigenic activity. CD44+/EpCAMhigh (EpCAM is also know as ESA) tumor cells initiated tumors at a higher frequency than CD133+ cancer cells. Moreover, a subset of colon cancers lacked CD133 expression. Nevertheless, using the CD44+/EpCAMhigh phenotype, tumor- initiating cells could be isolated from these tumors, and addition of CD166/ALCAM to the marker profile could further enrich for CSC activity (35). Although CD44 and CD133 have been the most common CSC markers identified in solid tumors to date, other markers have been identified as well. Schatton and colleagues used a genetics approach to identify a CSC population in melanoma. From gene micrroarrays generated from stages of melanoma progression, they identified ABCB5, a known chemoresistance mediator in melanoma, as a molecular marker of neoplastic progression. They then showed that ABCB5+ cells from primary uniquely formed new tumors when transplanted into NOD/SCID mice (36). In addition, transplanted ABCB5+ cells recapitulated original tumor heterogeneity. They also exhibited a hierarchical organization, with ABCB5+ cells giving rise to both ABCB5+ and ABCB5- tumor cells while ABCB5- cells only giving rise to ABCB5- cells. 8

Cancer Stem Cells in Hematologic Malignancies Considerable advances have been made in the identification of CSCs in solid tumors, and the search for CSCs in hematologic neoplasms has been no less earnest (Table 1). LSCs have been identified in both precursor B- and T-cell lymphoblastic leukemia/lymphoma (B-ALL, T-ALL). In BCR-ABL positive B-ALL, the LSC is present in the CD34+CD38- fraction and expresses the B- lineage marker CD19 (37, 38). This LSC phenotype also extends to B-ALL positive for ETV6- RUNX1, a common chromosomal rearrangement abnormality in B-ALL (38). These data are seemingly at odds with a prior study indicating that both CD34+CD10- and CD34+CD19- B-ALL blasts initiate leukemias in NOD/SCID mice (39). While additional study is warranted to clarify these potential inconsistencies, these data also raise the possibility that LSC populations may be distinct in different B-ALL subgroups. In T-ALL, both CD34+/CD4- and CD34+/CD7- blasts were capable of NOD/SCID engraftment, and acquisition of CD4 or CD7 was associated with absence of engraftment activity (40). Both the B-ALL and T-ALL data suggest that LSC activity in ALL is retained in hematopoietic populations more primitive than normal committed lymphoid cells. Furthermore, similar to LSCs in AML, ALL is organized hierarchically with LSC giving rise to more differentiated progeny. Myeloma stem cells were identified based on the differential expression of CD138/syndecan, an antigen normally present on terminally differentiated plasma cells. CD138-CD34- cells isolated from the bone marrow of multiple myeloma patients were able to form colonies in vitro as well as initiate myeloma-like disease when transplanted into NOD/SCID mice (41). More recent studies have refined this phenotype, demonstrating that CD19+CD27+ cells isolated from the peripheral blood of myeloma patients can engraft NOD/SCID mice. Notably, this phenotype is consistent with memory B cells, suggesting that malignant B lineage CSCs give rise to terminally differentiated myeloma cells (42).

Clinical Utility of Cancer Stem Cells Although there is extensive support for the CSC hypothesis in the laboratory, what is the relevance of CSCs to tumor clinical behavior and patient outcomes? The CSC hypothesis makes a number of predictions including: 1) CSCs are relatively resistant to conventional therapies; 2) CSC burden at the time of treatment is likely to predict clinical response and overall survival; and 3) therapies that do not effectively target the CSC are unlikely to eradicate disease. A number of recent studies suggest that these predictions of the CSC hypothesis are correct. One significant prediction of the CSC hypothesis is that CSCs are relatively resistant to standard cytotoxic therapies. Like their normal stem cell counterparts, CSCs infrequently enter the cell cycle and express higher levels of drug-efflux , making them more likely to be resistant to cytotoxic therapies that target actively dividing cells (43-45). The increased resistance of LSCs to chemotherapy has been shown experimentally, as two standard agents, cytarabine and adriamycin, preferentially kill leukemic blasts, whereas LSCs are relatively resistant (46). Relative resistance to 9 radiotherapy has been shown for glioblastoma, where CD133+ CSCs exhibit increased resistance to radiation-induced death compared to CD133- glioblastoma tumor cells (47). The importance of eradicating CSC is probably best illustrated by the example of CML. While experimental validation of the LSC population has not been achieved using xenotransplantation models, it is known that the Lin-CD34+CD38-CD45RA-CD71- fraction of bone marrow cells in CML patients contains quiescent primitive cells and that these cells exhibit insensitivity to imatinib, the standard CML therapy originally designed to target the characteristic BCR-ABL fusion of CML (48, 49). Based on mathematical modeling of the kinetics of reduction of BCR-ABL positive cells in response to imatinib, Michor et al. determined that during, and after stoppage of therapy, the number of BCR-ABL positive transcripts were consistent with the presence of an LSC population that had not been eradicated during treatment and were now reconstituting disease (50, 51). The previous examples suggest that resistance of CSCs to standard therapies could represent a major factor in overall cancer resistance to chemotherapy and radiation. Thus, therapeutic approaches that fail to eradicate the CSC population are unlikely to eliminate disease (figure 2). Eradication of the non-cancer stem cell population might initially decrease tumor burden, but such treatments would ultimately fail due to the inability to eradicate the tumor-forming population. Indeed, this pattern of disease relapse after cytotoxic treatment is seen in many tumor types.

Figure 3: Proposed mechanism of CSC-targeted therapy (A) Conventional cancer therapies preferentially target actively proliferating cells but otherwise eliminate cancer cells in a nonspecific manner; however, relatively resistant CSCs remain. Because the tumorigenic CSC population is not eradicated, regrowth of the tumor occurs and the patient relapses. (B) CSC-targeted therapies eliminate the tumorigenic potential of the cancer and result in a cure because the remaining cancer cells exhibit limited self- renewal capacity and eventually die.

Several studies have established the association of LSC-related parameters and clinical course. Consistent with the notion that LSCs drive clinical responses to therapy, the percentage of CD34+

10 blasts in AML shows no correlation with treatment outcome or overall survival (52-54). Rather, the frequency of LSCs at the time of diagnosis correlates with both higher levels of residual disease and decreased likelihood of achieving remission following induction therapy. Specifically, an LSC frequency greater than 7.5% correlates with poorer survival (55). In addition, undifferentiated AML (FAB subtype M0) has a higher frequency of CD34+CD38- blasts and a poorer prognosis than other FAB types (56, 57). Finally, some studies have shown a correlation between the ability of AML to engraft NOD/SCID mice and worse responses to therapy and poorer overall survival, suggesting that leukemic engraftment in the xenotransplant setting may reflect biologic parameters in (55, 58) Attempts to apply breast CSC parameters to determine prognosis have been met with mixed results. While a high percentage of CD44+CD24low tumor cells were found in primary tumors with distant metastasis, no correlation was found between the frequency of CD44+CD24low tumor cells and tumor progression or overall survival (59). Does this mean that evaluation of CSC frequency in breast cancer lacks prognostic value? Not necessarily. Immunohistochemical quantitation, which was employed in this study, may be very difficult since evaluation of limited numbers of tumor tissue sections may not accurately reflect CSC frequency for the entire tumor. Additionally, relying on the absence of a marker to identify CSCs can create difficult interpretation issues. Hopefully, in the future additional markers present on breast CSC will be identified, and such markers could be used in combination with CD44. Another method to measure the “stemness” of an entire tumor involves a analysis approach. When breast CSC gene expression was compared to gene signatures derived from bulk breast tumors, those patients with tumors exhibiting a more CSC-like gene signature had a higher risk of metastasis and relapse (60). These data may simply reflect a higher percentage of CSCs in the tumor or they may reflect increased activity of “stemness” pathways in these breast tumors. Data from gliomas also indicates a correlation between expression of CD133, the glioma CSC marker, and prognosis. Frequencies of CD133 expression greater than 1% were associated with shorter overall survival. Furthermore, CD133 expression was an independent risk factor for tumor re-growth as well as time to malignant progression for WHO grade 2 and 3 tumors (61). Consistent with this observation, CD133+ cell frequency also increased with tumor grade. While the majority of WHO grade 2 tumors analyzed contained less than 1% CD133+ cells, on average glioblastomas (WHO grade 4) contained greater than 25% CD133+ cells, consistent with their poor differentiation. The pattern of CD133 expression also correlated with tumor grade, as increasing numbers of CD133+ cells were present in clusters in higher-grade lesions. Finally, higher CD133+ expression also correlated with more frequent malignant progression of grade 2 and 3 tumors to grade 4. Data from other solid tumors also support the association between CSCs and prognosis. HNSCC patients with greater than 15% CD44+ tumor cells experience significantly lower survival rates than those patients with less than 15% CD44+ cells (L. Ailles, personal communication). In melanoma, higher frequencies of ABCB5+ cells are present in more advanced lesions suggesting a correlation between CSC burden and prognosis (36). 11 CONCLUSION

Despite the rapid advances in CSC biology in the past decade, several key questions remain. Are CSCs present in all tumor types, and at what frequency? How pure are the defined CSC populations, and will additional markers further refine these populations? How, and at what development stage, do CSCs arise from their normal counterparts? Will elucidation of the molecular mechanisms regulating CSC function translate to cures for previously intractable cancers? These questions have mobilized researchers to identify CSCs in numerous human cancers, and the number of CSCs identified will only continue to grow over the coming years. Although there have been important concerns raised against the CSC hypothesis, emerging data confirms that CSCs are critical to tumor behavior and clinical responses to therapy. With the continued development of techniques to isolate CSCs and to study their function in physiologic models, and the will of investigators to translate their findings into therapies, we now have the opportunity to put the CSC hypothesis to the test in the most stringent of environments— the clinic. The impact of such therapies is likely to be profound.

12

CHAPTER 2

CD47 is an adverse prognostic factor and therapeutic antibody target on acute myeloid leukemia stem cells

Portions of this chapter were published in the following article: Majeti R*, Chao MP*, Alizadeh AA, Pang WW, Jaiswal S, Weissman IL. CD47 is an adverse prognostic factor and therapeutic antibody target on human myeloid leukemia stem cells. Cell. 2009; 138:268-299.*co-first author

13 SUMMARY

Acute myelogenous leukemia (AML) is organized as a cellular hierarchy initiated and maintained by a subset of self-renewing leukemia stem cells (LSC). We hypothesized that increased CD47 expression on human AML LSC contributes to pathogenesis by inhibiting their phagocytosis through the interaction of CD47 with an inhibitory receptor on phagocytes. We found that CD47 was more highly expressed on AML LSC than their normal counterparts, and that increased CD47 expression predicted worse overall survival in 3 independent cohorts of adult AML patients. Furthermore, blocking monoclonal antibodies directed against CD47 preferentially enabled phagocytosis of AML LSC and inhibited their engraftment in vivo. Finally, treatment of human AML LSC-engrafted mice with anti- CD47 antibody depleted AML and targeted AML LSC. In summary, increased CD47 expression is an independent poor prognostic factor that can be targeted on human AML stem cells with blocking monoclonal antibodies capable of enabling phagocytosis of LSC.

INTRODUCTION

According to the cancer stem cell model, tumors are organized as a cellular hierarchy maintained by a small pool of self-renewing cancer stem cells which must be eliminated in order to eradicate the tumor (62, 63). For the development of cancer stem cell-targeted therapies, it is necessary to identify molecules and pathways that are preferentially expressed in these cancer stem cells and that are critical for pathogenesis. To date, human acute myeloid leukemia (AML) stem cells (LSC) are the most well studied cancer stem cell population (64). AML is an aggressive malignancy with five year overall survival between 30-40%, and much lower for those over age 65 (65, 66). Cytogenetic abnormalities are prognostic in adults with AML, however, up to 50% have a normal karyotype (67, 68). In these patients, the presence of specific molecular mutations can provide prognostic information, particularly internal tandem duplications within the fms-related tyrosine kinase 3 gene (FLT3-ITD) (69, 70). In published reports assaying a variety of subtypes of AML, LSC were found to be negative for expression of lineage markers (Lin-), positive for expression of CD34, and negative for expression of CD38 (8, 64). We have recently shown that the Lin-CD34+CD38-CD90- fraction of human cord blood contains a non-HSC multipotent progenitor (MPP), and have hypothesized that this MPP is the cell of origin for human AML (71). Consistent with this hypothesis, we have shown that pre-leukemic mutations occur in a clonal HSC population, eventually leading to the development of LSC at the MPP stage in AML or the granulocyte-macrophage progenitor (GMP) stage in myeloid blast crisis CML (10, 15, 16). We report here the identification of higher expression of CD47 on AML LSC compared to their normal counterparts, HSC and MPP, a finding corroborated by microarray gene expression

14 analysis (72). CD47 is a widely expressed transmembrane protein (73). CD47 serves as the ligand for signal regulatory protein alpha (SIRPα), which is expressed on phagocytic cells including macrophages and dendritic cells, that when activated initiates a signal transduction cascade resulting in inhibition of phagocytosis (74-78). In our own studies, we have found that expression of mouse CD47 in a human AML cell line inhibits phagocytosis and facilitates engraftment in immunodeficient mice, and that CD47 expression on mouse HSC and progenitors increases upon mobilization and is required for engraftment upon transplantation. We hypothesize that increased expression of CD47 on human AML contributes to pathogenesis by inhibiting phagocytosis of these cells through the interaction of CD47 with SIRPα.

EXPERIMENTAL PROCEDURES

Human Samples Normal human bone marrow mononuclear cells were purchased from AllCells Inc. (Emeryville, CA, USA). Human AML samples (Figure 1B) were obtained from patients at the Stanford Medical Center with informed consent, according to an IRB-approved protocol (Stanford IRB# 76935 and 6453). Human CD34-positive cells were enriched with magnetic beads (Miltenyi Biotech, Auburn, CA, USA).

Flow Cytometry Analysis and Cell Sorting A panel of antibodies was used for analysis and sorting of AML LSC (Lin-CD34+CD38-CD90-, where lineage included CD3, CD19, and CD20), HSC(Lin-CD34+CD38-CD90+), and MPP (Lin_CD34+CD38-CD90-CD45RA-). Analysis of CD47 expression was performed with an anti-human CD47 PE antibody (clone B6H12, BD Biosciences, San Jose CA, USA). For analysis of mouse bone marrow, the following antibodies were used: Sca1 PB, cKit Alexa 750, Flk2 PE, CD34 FITC, Lineage (CD3, CD4, CD5, CD8, B220, Mac1) PeCy5 (Ebiosciences, San Diego, CA, USA).

Mouse CD47 and Anti-Mouse SIRPα Antibodies Monoclonal mouse anti-human CD47 antibodies included the following:BRIC126, IgG2b (AbD Serotec), 2D3, IgG1 (Ebiosciences), and B6H12.2, IgG1. Monoclonal rat anti-mouse CD47 antibody used was MIAP301, IgG2a. The B6H12.2 and MIAP301 hybridomas were obtained from the American Type Culture Collection (Rockville, MD, USA). Antibody was either purified from hybridoma supernatant using protein G affinity chromatography according to standard procedures or obtained from BioXCell (Lebanon, NH, USA). Monoclonal rat anti-mouse SIRPα, P84, IgG1 was purchased from BD PharMingen (San Jose, CA, USA). Isotype controls included mouse IgG1 and rat IgG2a antibodies (Ebiosciences). 15

Methylcellulose Colony Assay Methylcellulose colony formation was assayed by plating sorted cells into a 6-well plate, each well containing 1 ml of complete methylcellulose (Methocult GF+ H4435, Stem Cell Technologies). Plates were incubated for 14 days at 37°C, then scored based on morphology.

Genomic DNA Preparation and Analysis of FLT3-ITD by PCR Genomic DNA was isolated from cell pellets using the Gentra Puregene Kit according to the manufacturer’s protocol (Gentra Systems, Minneapolis, MN). FLT3-ITD status was screened by PCR using primers that generated a wild-type product of 329 bp and ITD products of variable larger sizes. All reactions were performed in a volume of 50 ul containing 5 ul of 10x PCR buffer (50mM

KCL/10nM Tris/2mM MgCl2/0.01% ), 1 ul of 10mM dNTPs, 2 units of Taq polymerase (Invitrogen), 1 ul of 10uM forward primer 11F (5’-GCAATTTAGGTATGAAAGCCAGC-3’) and reverse primer 12R (5’-CTTTCAGCATTTTGACGGCAACC-3’), and 10-50 ng of genomic DNA. PCR conditions for amplification of the FLT3 gene were 40 cycles of denaturation (30 sec at 95°C) annealing (30 sec at 62°C), and extension (30 sec at 72°C).

In Vitro Phagocytosis Assays Human AML LSC or normal bone marrow CD34+ cells were CFSE-labeled and incubated with either mouse or human macrophages in the presence of 7 mg/ml IgG1 isotype control, anti-CD45 IgG1, anti- CD47 (clones B6H12.2, BRIC126, or 2D3), or anti-mouse SIRPa antibody for 2 hr. Mouse GFP- positive leukemia cells were incubated with mouse macrophages in the presence of 10 mg/ml of rat IgG2a isotype control or anti-mouse CD47 (MIAP301) for 2 hr. Cells were then analyzed by fluorescence microscopy to determine the phagocytic index (number of cells ingested per 100 macrophages). In some experiments, cells were then harvested and stained with either a mouse or human macrophage marker and phagocytosed cells were identified by as macrophage+CFSE+. Statistical analysis using Student’s t test was performed with GraphPad Prism.

Annexin V Apoptosis Assays 105 AML LSC or Jurkat cells were incubated with serum free IMDM media for 2, 6, or 24 hours at 37°C in the presence of media alone, 5nM staurosporine (Sigma) as a positive control, or the following soluble antibodies at 7µg/ml: mouse IgG1 isotype control, anti-human CD45 IgG1, anti-human CD47 (clone B6H12.2), or anti-mouse SIRPα. In parallel, the antibodies were first immobilized on the plate surface as previously described (79), and 1x105 LSC or Jurkat cells were added and incubated at 37°C for 2, 6, or 24 hours, with both media alone or staurosporine as controls. Cells were then washed and the presence of apoptotic cells was determined by annexin V and 7-AAD staining using the Annexin V-

16

PE Apoptosis Detection Kit I (BD Biosciences) according to the manufacturer’s protocol. Assays performed with Jurkat cells were run in triplicate for each experimental condition.

In Vivo Pre-Coating Engraftment Assay LSC isolated from AML specimens were incubated with 28 ug/mL of IgG1 isotype control, anti-CD45 IgG1, or anti-CD47 IgG1 (B6H12.2) antibody at 4°C for 30 minutes. A small aliquot of cells was then stained with donkey anti-mouse PE secondary antibody (Ebioscience) and analyzed by flow cytometry to assess coating. Approximately 105 coated LSC were then transplanted into each irradiated newborn NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ (NOG) mouse as described (71). Mice were sacrificed 13 weeks post- transplantation and bone marrow was analyzed for human leukemia engraftment (hCD45+hCD33+) by flow cytometry (71). The presence of human leukemia was confirmed by Wright-Giemsa staining of hCD45+ cells and FLT3-ITD PCR (data not shown). Statistical analysis using Student’s t-test was performed with GraphPad Prism (San Diego, CA).

Generation of Mouse Model of AML A murine HoxA9-IRES-Meis1 expression cassette (H9M) was cloned into a retroviral MSCV-pgk-GFP vector and Phoenix ecotropic cells were used to produce retroviral particles. 5-10-week-old C57BL/6 mice were injected intraperitoneally with 150 mg/kg 5-Fluorouracil (Abraxis). Mice were sacrificed four days later, and the marrow cultured for two days with murine SCF (100 ng/ml), IL-3 (10 ng/ml), and human IL-6 (10 ng/ml) (Peprotech). Cells were transduced with retrovirus on retronectin coated plates (Takara) for six-hours with 4 ug/ml polybrene for two days. After two additional days, 105 GFP+ progenitors were transplanted via retroorbital injection. Mice were sacrificed when moribund and cell suspensions of spleen and marrow were used for secondary transplantation or cryopreserved. Secondary transplanted mice also developed AML with cryopreservation of these cells as well.

In Vivo Antibody Treatment of Human AML LSC Engrafted Mice 1–2.5 x 105 FACS-purified LSC were transplanted into NOG pups. Eight to twelve weeks later, human AML engraftment (hCD45+CD33+ cells) was assessed in the peripheral blood and bone marrow by tail bleed and aspiration of the femur, respectively. Engrafted mice were then treated with daily intraperitoneal injections of 100 mg of anti-CD47 antibody or IgG control for 14 days. On day 15 mice were sacrificed and the peripheral blood and bone marrow were analyzed for AML.

In Vivo Human AML Phagocytosis Assay A GFP encoding lentivirus was prepared from the pCDH-CMV construct (System Biosciences, Mountain View, CA, USA) using standard techniques. AML LSC from sample SU028 were transduced overnight and transplanted into newborn NOG pups as described. Twelve weeks later human 17 CD45+CD33+GFP+ leukemia engraftment was assessed in the peripheral blood, and GFP+ human leukemia-engrafted mice were injected intraperitoneally with a single 100 mg dose of either anti-CD47 antibody (clone B6H12.2) or IgG control. Four hours later, mice were sacrificed and bone marrow, spleen, and liver were analyzed by flow cytometry for the presence of GFP+ leukemia cells within F4/80-positive mouse phagocytes. The presence of human CD45_GFP+ mouse F4/80+ events identified mouse phagocytes with ingested human leukemia cells.

In Vivo Macrophage Depletion Liposomal clodronate and control liposomes were prepared as described (80). Macrophages were depleted in AML LSC-engrafted NOG mice with the following treatment schedule: 200 ml of either clodronate or liposomal control was injected intravenously via the retro-orbital sinus 2 days prior to treatment of these mice with anti-CD47 antibody for IgG control. One hundred microliters of either clodronate or liposomal control was then injected in the same manner on days 2, 6, and 10 after initiation of daily antibody treatment. Mice were then sacrificed on day 14 to assess human leukemic engraftment as described.

Secondary Transplantation AML LSC-engrafted mice treated with daily injections of either IgG control or anti-CD47 antibody were sacrificed at the end of 14 days of treatment. 5x105 whole bone marrow cells were transplanted into newborn NOG mice. Twelve weeks later peripheral blood and bone marrow was harvested and analyzed for human CD45+CD33+ leukemia engraftment as described.

AML Patients, Microarray Gene Expression Data, and Statistical Analysis Gene expression and clinical data were analyzed for three previously described cohorts of adult AML patients: (1) a training dataset of 285 patients with diverse cytogenetic and molecular abnormalities described by Valk et al.(81), (2) a test dataset of 242 patients with normal karyotypes described by Metzeler et al.(82), and (3) a validation dataset of 137 patients with normal karyotypes described by Bullinger et al.(83) The clinical end points analyzed included overall and event-free survival, with events defined as the interval between study enrollment and removal from the study owing to a lack of complete remission, relapse, or death from any cause, with data censored for patients who did not have an event at the last follow-up visit.

RESULTS

CD47 is More Highly Expressed on AML LSC Than Their Normal Counterparts and is Associated with the FLT3-ITD Mutation

18

In our investigation of several mouse models of myeloid leukemia, we identified increased expression of CD47 on mouse leukemia cells compared to normal bone marrow (80). This prompted investigation of CD47 expression on human AML LSC and their normal counterparts. Using flow cytometry, CD47 was more highly expressed on multiple specimens of AML LSC than normal bone marrow HSC and MPP (Figure 1). This increased expression extended to the bulk leukemia cells, which expressed CD47 similarly to the LSC-enriched fraction. A

B

Figure 1: CD47 Is More Highly Expressed on AML LSC Compared to Their Normal Counterparts (A) Relative CD47 expression on normal bone marrow HSC (Lin-CD34+CD38-CD90+) and MPP (Lin_CD34+CD38-CD90-CD45RA-), as well as LSC (Lin-CD34+CD38-CD90-) and bulk leukemia cells from human AML samples, was determined by flow cytometry. Mean fluorescence intensity was normalized for cell size and against lineage-positive cells to account for analysis on different days. The same sample of normal bone marrow (red, n = 3) orAML (blue, n = 13) is indicated by the samesymbol in the different populations. Normalizedmean expression (and range) for each populationwere as follows:HSC30.6 (28.8–33.4),MPP 31.8 (30.0–33.4), LSC 59.8 (21.6–104.7), and bulk AML 56.3 (22.1–85.1). The differences between the mean expression of HSC with LSC (p = 0.003), HSC with bulk leukemia (p = 0.001),MPP with LSC (p = 0.004), andMPP with bulk leukemia (p = 0.002) were statistically significant using a two-sided Student’s t test. The differenc between the mean expression of AML LSC compared to bulk AML was not statistically significant with p = 0.50 using a paired two-sided Student’s t test. (B) Clinical and molecular characteristics of primary human AML samples manipulated in vitro and/or in vivo.

19 Examination of a subset of these samples indicated that CD47 surface expression correlated with CD47 mRNA expression (Figure 2). To investigate CD47 expression across morphologic, cytogenetic, and molecular subgroups of AML, gene expression data from a previously described cohort of 285 adult patients were analyzed (81). No significant difference in CD47 expression among FAB (French-American-British) subtypes was found (Figure 3A). In most cytogenetic subgroups, CD47 was expressed at similar levels, except for cases harboring t(8;21)(q22;q22), a favorable risk group which had a statistically significant lower CD47 expression (Figure 3B). In molecularly characterized AML subgroups, no significant association was found between CD47 expression and mutations in the tyrosine kinase domain of FLT3 (FLT3-TKD), over-expression of EVI1, or mutations in CEBPA, NRAS, or KRAS. However, higher CD47 expression was strongly correlated with the presence of FLT3-ITD (p<0.001), which is observed in nearly one third of AML with normal karyotypes and is associated with worse overall survival (69, 70). This finding was separately confirmed in two independent datasets of 214 and 137 AML patients (Table 1) (83, 84).

Figure 2: CD47 mRNA Expression Correlates With Cell Surface Expression CD47 mRNA and cell surface expression were evaluated for 8 samples of primary human AML. Increasing mRNA expression was associated with increasing cell surface expression, as determined by normalized mean fluorescence intensity from flow cytometry (p=0.046).

20

A B

Figure 3: CD47 Expression Across FAB and Cytogenetic Subgroups of AM Gene expression microarray data from 285 patients was analyzed for the expression of CD47 across (A) FAB subgroups and (B) cytogenetic subgroups of human AML. Box plots capture the distribution of CD47 expression within the corresponding groups, wherein each box spans the interquartile range and the whiskers span 1.5 times the interquartile distance or to the highest or lowest point, whichever is shorter; any data beyond these whiskers are shown as points. No difference in expression was detected among FAB subgroups as measured by analysis of variance. Among cytogenetic subgroups, the only statistically significant difference detected was a lower expression of CD47 in t(8;21) AML compared to the others.

Identification and Separation of Normal HSC From Leukemia Cells in the Same Patient Based On Differential CD47 Expression In the LSC-enriched Lin-CD34+CD38- fraction of specimen SU008, a rare population of CD47lo-expressing cells was detected, in addition to the majority CD47hi-expressing cells (Figure 4A). These populations were isolated by fluorescence-activated cell sorting (FACS) to >98% purity and either transplanted into newborn NOG mice or plated into complete methylcellulose. The CD47hi cells failed to engraft in vivo or form any colonies in vitro, as can be observed with some AML specimens (85). However, the CD47lo cells engrafted with normal myelo-lymphoid hematopoiesis in vivo and formed numerous morphologically normal myeloid colonies in vitro (Figure 4B,C). This specimen harbored the FLT3-ITD mutation, which was detected in the bulk leukemia cells (Figure 4D). The purified CD47hi cells contained the FLT3-ITD mutation, and therefore, were part of the leukemic clone, while the CD47lo cells did not. Human cells isolated from mice engrafted with the CD47lo cells contained only wild type FLT3, indicating that the CD47lo cells contained normal hematopoietic progenitors.

21

Clinical Feature* Overall Low CD47 High CD47 P† n=137 n=95 n=37 Age, yrs. 0.26 Median 46 47 46 Range 16-60 24-60 16-60 WBC, x109/L <0.01 Median 24 17 35 Range 1-238 1-178 1-238 Centrally reviewed FAB 0.29 Classification, no. (%) M0 11 (8) 9 (9) 2 (5) M1 28 (20) 16 (17) 2 (32) M2 36 (26) 22 (23) 11 (30) M4 33 (24) 25 (26) 8 (22) M5 19 (14) 16 (17) 3 (8) M6 2 (1) 2 (2) 0 (0) Unclassified 6 (4) 4 (4) 0 (0) FLT3-ITD, no. (%) <0.05 Negative 84 (61) 63 (66) 17 (46) Positive 53 (39) 32 (34) 20 (54) FLT3-TKD, no. (%) 0.24 Negative 109 (87) 78 (89) 27 (79) Positive 17 (13) 10 (11) 7 (21) NPM1, no. (%) 0.10 Wild-Type 55 (45) 41 (49) 10 (30) Mutated 66 (55) 43 (51) 23 (70) CEBPA, no. (%) 1 Wild-Type 100 (86) 70 (86) 27 (87) Mutated 16 (14) 11 (14) 4 (13) MLL-PTD, no. (%) 1 Negative 121 (93) 83 (92) 34 (94) Positive 9 (7) 7 (8) 2 (6) Event-free survival 0.004 Median, mos. 14 17.1 6.8 Disease-free at 3 yrs, % (95% CI) 36 (27-44) 41 (31-52) 22 (8-36) Overall survival 0.002 Median, mos. 18.5 22.1 9.1 Alive at 3 yrs, % (95% CI) 39 (31-48) 44 (33-55) 26 (12-41) Complete remission rate, no. (%) CR after 1st Induction, no. (%) 60 (46%) 46 (48%) 14 (38%) 0.33 CR after 2nd Induction, no. (%) 84 (74%) 64 (75%) 20 (69%) 0.63 Randomization to 2ndary consolidative therapy Allogeneic-HSCT, no. (%) 29 (22%) 25 (26%) 4 (11%) 0.09 Autologous-HSCT, no. (%) 23 (17%) 17 (18%) 6 (16%) 0.98

Table 1: Clinical and Molecular Characteristics of AML Samples from the Validation Cohort and Comparison Between Low CD47 and High CD47 Expression Groups

22

A B

C

D

Figure 4: Identification and Separation of Normal HSC From Leukemia Cells in the Same Patient Based on Differential CD47 Expression (A) CD47 expression on the Lin-CD34+CD38- LSC-enriched fraction of specimen SU008 was determined by flow cytometry. CD47hi- and CD47lo-expressing cells were identified and purified using FACS. The left panels are gated on lineage-negative cells, while the right panels are gated on Lin-CD34+CD38- cells. (B) Lin-CD34+CD38-CD47lo and Lin-CD34+CD38-CD47hi cells were plated into complete methylcellulose, capable of supporting the growth of all myeloid colonies. Fourteen days later, myeloid colony formation was determined by morphologic assessment. Representative CFU-G/M (left) and BFU-E (right) are presented. (C) Lin-CD34+CD38-CD47lo cells were transplanted into two newborn NOG mice. Twelve weeks later, the mice were sacrificed and the bone marrow was analyzed by flow cytometry for the presence of human CD45+CD33+ myeloid cells and human CD45+CD19+ lymphoid cells. (D) Normal bone marrow HSC, bulk SU008 leukemia cells, Lin-CD34+CD38-CD47hi cells, Lin_CD34+CD38-CD47lo cells, or human CD45+ cells purified from the bone marrow of mice engrafted with Lin-CD34+CD38-CD47lo cells were assessed by PCR for the presence of the FLT3-ITD mutation. The wild-type (WT) FLT3 and the FLT3-ITD products are indicated.

23 Increased CD47 Expression in Human AML is Associated with Poor Clinical Outcomes We hypothesized that increased CD47 expression on human AML contributes to pathogenesis, and predicted that AML with higher expression of CD47 would be associated with worse clinical outcomes. Consistent with this hypothesis, analysis of a previously described group of 285 adult AML patients with diverse cytogenetic and molecular abnormalities (81) revealed that a dichotomous stratification of patients into low CD47 and high CD47 expression groups was associated with a significantly increased risk of death in the high expressing group (p=0.03, Figure 5A-C). The association of overall survival with this dichotomous stratification of CD47 expression was validated in a second test cohort of 242 adult patients (82) with normal karyotypes (NK-AML) (p=0.01, Figure 5A,D). Applying this stratification to a distinct validation cohort of 137 adult patients with normal karyotypes (83), we confirmed the prognostic value of CD47 expression for both overall and event-free survival (Figure 6). Analysis of clinical characteristics of the low and high CD47 expression groups in this cross-validation cohort also identified statistically significant differences in white blood cell (WBC) count and FLT3-ITD status, and no differences in rates of complete remission and type of consolidative therapy including allogeneic transplantation (Table 1). Kaplan-Meier analysis demonstrated that high CD47 expression at diagnosis was significantly associated with worse event-free and overall survival (Figure 6A,B). Patients in the low CD47 expression group had a median event-free survival of 17.1 months compared to 6.8 months in the high CD47 expression group, corresponding to a hazard ratio of 1.94 (95% confidence interval 1.30 to 3.77, p=0.004). For overall survival, patients in the low CD47 expression group had a median of 22.1 months compared to 9.1 months in the high CD47 expression group, corresponding to a hazard ratio of 2.02 (95% confidence interval 1.37 to 4.03, p=0.002). When CD47 expression was considered as a continuous variable, increased expression was also associated with a worse event-free (p=0.02) and overall survival (p=0.02).

24

Figure 5: Derivation and Distribution of CD47 mRNA Expression of AML Samples in Low CD47 and High CD47 Expression Groups (A) Datasets employed for the derivation and validation of a CD47 dichotomous discriminator predictive of survival in AML. (B) Left: frequency distribution of mRNA expression levels for the training dataset of 285 AML patients (described by Valk et al.(81)), with the corresponding threshold for stratification of patients into CD47 Low (lowest 72%) and CD47 High (highest 28%) groups demarcated by the dashed vertical line. Right: the cross- validation dataset of 137 patients with AML and normal karyotypes (described by Bullinger et al.(83)) is similarly stratified using this threshold into CD47 Low and CD47 High groups. (C and D) Overall survival of patients within the (C) training (n=285) and (D) test (n=242) datasets when stratified according to the threshold defined above.

25

Figure 6: Increased CD47 Expression in Human AML Is Associated with Poor Clinical Outcomes Event-free (A and C) and overall (B and D) survival of 132 AML patients with normal cytogenetics (A and B) and the subset of 74 patients without the FLT3-ITD mutation (C and D). Patients were stratified into low CD47 and high CD47 expression groups based on an optimal threshold (28% high, 72% low) determined by microarray analysis from an independent training data set. The significance measures are based on log-likelihood estimates of the p value, when treating the model with CD47 expression as a binary classification.

Despite the association with FLT3-ITD (Table 1), increased CD47 expression at diagnosis was significantly associated with worse event-free and overall survival in the subgroup of 74 patients without FLT3-ITD, when considered either as a binary classification (Figure 6C,D) or as a continuous variable (p=0.02 for both event-free and overall survival). In multivariable analysis considering age, FLT3-ITD status, and CD47 expression as a continuous variable, increased CD47 expression remained associated with worse event-free survival with a hazard ratio of 1.33 (95% confidence interval 1.03 to 1.73, p=0.03) and overall survival with a hazard ratio of 1.31 (95% confidence interval 1.00 to 1.71, p=0.05) (Table 2).

26

Outcome Measure/Variables Considered HR 95% CI P Event-free survival CD47 expression, continuous, per 2-fold increase 1.33 1.03-1.73 0.03 FLT3-ITD, positive vs. negative 2.21 1.39-3.53 <0.001 Age, per year 1.03 1.00-1.06 0.03

Overall survival CD47 expression, continuous, per 2-fold increase 1.31 1.00-1.71 0.05 FLT3-ITD, positive vs. negative 2.29 1.42-3.68 <0.001 Age, per year 1.03 1.01-1.06 0.01

Table 2: Multivariable Analysis of Prognostic Factors in a Validation Cohort of Normal Karyotype AML Samples Multivariable analysis considering age, FLT3-ITD status, and CD47 expression as a continuous variable in association with event-free and overall survival in a validation cohort of AML patients with normal cytogenetics.

Monoclonal Antibodies Directed Against Human CD47 Preferentially Enable Phagocytosis of AML LSC by Human Macrophages We hypothesized that increased CD47 expression on human AML contributes to pathogenesis by inhibiting phagocytosis of leukemia cells, leading us to predict that disruption of the CD47-SIRPα interaction with a monoclonal antibody directed against CD47 will preferentially enable the phagocytosis of AML LSC. Several anti-human CD47 monoclonal antibodies have been generated including some capable of blocking the CD47-SIRPα interaction (B6H12.2 and BRIC126) and others unable to do so (2D3) (86). The ability of these antibodies to enable phagocytosis of AML LSC, or normal human bone marrow CD34+ cells, by human macrophages in vitro was tested. Incubation of AML LSC with human macrophages in the presence of IgG1 isotype control antibody or mouse anti- human CD45 IgG1 monoclonal antibody did not result in significant phagocytosis as determined by either immunofluorescence microscopy (Figure 7A) or flow cytometry (Figure 8). However, addition of the blocking anti-CD47 antibodies B6H12.2 and BRIC126, but not the non-blocking anti-CD47 antibody 2D3, enabled phagocytosis of AML LSC (Figure 7A,C). No phagocytosis of normal CD34+ cells was observed with any of the antibodies (Figure 7C).

27 A

B

C

D

28

Figure 7: Monoclonal Antibodies Directed against Human CD47 Preferentially Enable Phagocytosis of Human AML LSC by Human and Mouse Macrophages In Vitro (A and B) CFSE-labeled AML LSC were incubated with human peripheral blood-derived macrophages (A) or mouse bone marrow-derived macrophages (B) in the presence of IgG1 isotype control, anti-CD45 IgG1, or anti- CD47 (B6H12.2) IgG1 antibody. These cells were assessed by immunofluorescence microscopy for the presence of fluorescently labeled LSC within the macrophages (indicated by arrows). (C) CFSE-labeled AML LSC or normal bone marrow CD34+ cells were incubated with human (left) or mouse (right) macrophages in the presence of the indicated antibodies and then assessed for phagocytosis by immunofluorescence microscopy. The phagocytic index was determined for each condition by calculating the number of ingested cells per 100 macrophages. For AML LSC, the differences between isotype or anti-CD45 antibody with blocking anti-CD47 antibody treatment (B6H12.2 and BRIC126) were statistically significant with p < 0.001 for all pairwise comparisons with human and mouse macrophages. For human macrophages, the differences between AML LSC and normal CD34+ cells were statistically significant for B6H12.2 (p < 0.001) and BRIC126 (p = 0.002). For mouse macrophages, the difference between isotype control and anti-SIRPa antibody was statistically significant (p = 0.02). (D) AML LSC were incubated in the presence of the indicated antibodies or the staurosporine-positive control as described above, but in the absence of macrophages. At the end of the incubation, apoptotic cells were identified by Annexin V staining as determined by flow cytometry. No statistically significant increase in apoptosis was detected with any of the antibodies.

Monoclonal Antibodies Directed Against Human CD47 or Mouse SIRPα Enable Phagocytosis of AML LSC by Mouse Macrophages The CD47-SIRPα interaction has been implicated as a critical regulator of xenotransplantation rejection in several cross species transplants; however, there are conflicting reports of the ability of CD47 from one species to bind and stimulate SIRPα of a different species (86-88). In order to directly assess the effect of inhibiting the interaction of human CD47 with mouse SIRPα, the in vitro phagocytosis assays described above were conducted with mouse macrophages. Incubation of AML LSC with mouse macrophages in the presence of IgG1 isotype control antibody or mouse anti-human CD45 IgG1 monoclonal antibody did not result in significant phagocytosis as determined by either immunofluorescence microscopy (Figure 7B) or flow cytometry (Figure 8). However, addition of the blocking anti-CD47 antibodies B6H12.2 and BRIC126, but not the non-blocking 2D3, enabled phagocytosis of AML LSC (Figure 7B,C). The CD47-SIRPα interaction was alternatively disrupted by a monoclonal antibody directed against mouse SIRPα, which also enabled phagocytosis of AML LSC (Figure 7C).

29 Figure 8: Monoclonal Antibodies Directed Against Human CD47 Enable Phagocytosis of Human AML LSC by Human and Mouse Macrophages CFSE-labeled AML LSC were incubated with human peripheral blood-derived macrophages (A,B) or mouse bone marrow-derived macrophages (C,D) in the presence of IgG1 isotype control, anti-CD45 IgG1, or anti-CD47 (B6H12.2) IgG1 antibody. The macrophages were harvested, stained with a fluorescently labeled anti-human CD14 antibody (A,B) or anti-mouse macrophage antibody (C,D), and analyzed by flow cytometry. CD14+CFSE+ or mMac+CFSE+ double-positive events identify macrophages that have phagocytosed CFSE-labeled LSC. The percentage of macrophages with ingested CFSE-labeled LSC as determined by flow cytometry is indicated. For human macrophages, the differences between isotype and anti-CD45 antibody with anti-CD47 antibody-treated cells approached statistical significance with p=0.11 and p=0.15 respectively. For mouse macrophages, the differences between isotype and anti-CD45 antibody with anti-CD47 antibody-treated cells were statistically significant with p=0.05 and p=0.04 respectively.

A Monoclonal Antibody Directed Against CD47 Does Not Induce Apoptosis of AML LSC Antibodies directed against CD47 have been shown to directly induce apoptosis of several malignant hematopoietic cell lines, as well as primary human chronic lymphocytic leukemia B cells only when immobilized or cross-linked (79, 89-91). These prior reports raise the alternative hypothesis that anti-CD47 antibodies induce apoptosis of AML LSC, which are then recognized by macrophages and phagocytosed. In order to assess the ability of the blocking anti-CD47 antibody B6H12.2 to directly induce apoptosis of primary human AML LSC, these cells were incubated in vitro with antibodies as described above, but in the absence of macrophages, and expression of Annexin V was determined by flow cytometry. No increase in Annexin V-positive apoptotic cells was detected with the anti-CD47 antibody compared to controls over the time period tested (Figure 7D). Even when plate-bound, the anti-CD47 antibody did not induce apoptosis of AML LSC (Figure 9D). Furthermore, phagocytosis of

30

AML LSC was detected as early as 15 minutes after incubation with blocking anti-CD47 antibody, while no apoptosis was detected at 2 hours (Figure 9E). These results indicate that the blocking anti- CD47 antibody B6H12.2 does not directly induce apoptosis of human AML LSC.

Figure 9: Immobilized Anti-CD47 Antibody Does Not Stimulate Apoptosis of Human AML LSC (A) Representative flow cytometry plots illustrating Annexin V/7-AAD staining in Jurkat cells treated with media, the indicated soluble antibodies, or staurosporine as a positive control. (B and C) Jurkat cells were treated with media, the indicated soluble antibodies (B), plate-bound immobilized antibodies (C), or staurosporine for either 2 hours or 24 hours. No apoptosis was detected with soluble antibodies (B). Immobilized anti-CD47 stimulated a statistically significant increase in Annexin V-positive apoptotic Jurkat cells at both 2 hours (p=0.01) and 24 hours (p=0.05). (D) Primary human AML LSC were treated with media, the indicated plate-bound immobilized antibodies, or staurosporine for either 2 hours or 6 hours. Immobilized anti-CD47 did not stimulate a statistically significant increase in Annexin V-positive apoptotic cells at either 2 hours or 6 hours. (E) A time course for phagocytosis of AML LSC SU028 cells by mouse macrophages was determined by measuring the phagocytic index as described above. Phagocytosis of AML LSC is detected as early as 15 minutes after addition of the anti- CD47 antibody B6H12.2.

31 A Monoclonal Antibody Directed Against Human CD47 Inhibits AML LSC Engraftment and Depletes AML In Vivo The ability of the blocking anti-CD47 antibody B6H12.2 to target AML LSC in vivo was tested. First, a pre-coating strategy was utilized in which AML LSC were purified by FACS and incubated with IgG1 isotype control, anti-human CD45, or anti-human CD47 antibody. An aliquot of the cells was analyzed for coating by staining with a secondary antibody demonstrating that both anti- CD45 and anti-CD47 antibody bound the cells (Figure 10A). The remaining cells were transplanted into newborn NOG mice that were analyzed for leukemic engraftment 13 weeks later. In all but one mouse, the isotype control and anti-CD45 antibody coated cells exhibited long-term leukemic engraftment; however, most mice transplanted with cells coated with anti-CD47 antibody had no detectable leukemia engraftment (Figure 10B).

Figure 10: A Monoclonal Antibody Directed Against Human CD47 Inhibits AML LSC Engraftment In Vivo (A and B) Three primary human AML samples were incubated with IgG1 isotype control, anti-CD45 IgG1, or anti- CD47 IgG1 antibody (B6H12.2) prior to transplantation into newborn NOG mice. A portion of the cells was analyzed for coating by staining with a secondary anti-mouse IgG antibody and analyzed by flow cytometry (A). 13 weeks later, mice were sacrificed and the bone marrow was analyzed for the percentage of human CD45+CD33+ myeloid leukemia cells by flow cytometry (B). The difference in engraftment between anti-CD47-coated cells and both isotype (p=0.003) and anti-CD45 (p=0.04) coated cells was statistically significant using Fisher’s exact test.

Next, a treatment strategy was utilized in which mice were first engrafted with human AML LSC and then administered daily intraperitoneal injections of 100 micrograms of either mouse IgG or anti-CD47 antibody for 14 days, with leukemic engraftment determined pre- and post-treatment. Analysis of the peripheral blood showed near complete elimination of circulating leukemia in mice treated with anti-CD47 antibody, often after a single dose, with no response in control mice (Figure 11A,B). Similarly, there was a significant reduction in leukemic engraftment in the bone marrow of mice treated with anti-CD47 antibody, while leukemic involvement increased in control IgG-treated mice (Figure 11C,D and Figure 12A). Histologic analysis of the bone marrow identified monomorphic leukemic blasts in control IgG-treated mice (Figure 11E, panels 1,2) and cleared hypocellular areas in anti-CD47 antibody-treated mice (Figure 11E, panels 4,5). In the bone marrow of some anti-CD47

32 antibody-treated mice that contained residual leukemia, macrophages were detected containing phagocytosed pyknotic cells (Figure 11E, panels 3,6).

A B

C D

E

Figure 11: A Monoclonal Antibody Directed against Human CD47 Depletes AML In Vivo (A–D) Newborn NOG mice were transplanted with AML LSC, and 8–12 weeks later, peripheral blood (A and B) 33 and bone marrow (C and D) were analyzed for baseline engraftment prior to treatment with anti-CD47 (B6H12.2) or control IgG antibody (Day 0). Mice were treated with daily 100 mg intraperitoneal injections for 14 days, at the end of which they were sacrificed and peripheral blood and bone marrow were analyzed for the percentage of human CD45+CD33+ leukemia. (A) Pre- and post-treatment human leukemic chimerism in the peripheral blood from representative anti-CD47 antibody and control IgG-treated mice as determined by flow cytometry. (B) Summary of human leukemic chimerism in the peripheral blood assessed on multiple days during the course of treatment demonstrated elimination of leukemia in anti-CD47 antibody-treated mice compared to control IgG treatment (p = 0.007). (C) Pre- and post-treatment human leukemic chimerism in the bone marrow from representative anti-CD47 antibody or control IgG-treated mice as determined by flow cytometry. (D) Summary of human leukemic chimerism in the bone marrow on day 14 relative to day 0 demonstrated a dramatic reduction in leukemic burden in anti-CD47 antibody-treated mice compared to control IgG treatment (p = 0.006). (E) H&E sections of representative mouse bone marrow cavities from mice engrafted with SU004 AML LSC post-treatment with either control IgG (panels 1 and 2) or anti-CD47 antibody (panels 4 and 5). IgG-treated marrows were packed with monomorphic leukemic blasts, while anti-CD47-treated marrows were hypocellular, demonstrating elimination of the human leukemia. In some anti-CD47 antibody-treated mice that contained residual leukemia, macrophages were detected containing phagocytosed pyknotic cells (panels 3 and 6, arrows).

Figure 12: Summary of AML Engraftment in Anti-CD47 and Control IgG Antibody Treated Mice Newborn NOG mice were transplanted with AML LSC. 8-12 weeks later, bone marrow was analyzed for baseline engraftment prior to treatment with anti-CD47 (B6H12.2) or control IgG antibody. Mice were treated for 14 days

34

with daily 100 microgram intraperitoneal injections, at the end of which, they were sacrificed and bone marrow was analyzed for the level of human CD45+CD33+ leukemia cells (A) or human CD34+ LSC-enriched cells (B). (A) Treatment with anti-CD47 antibody dramatically reduced bone marrow leukemic engraftment (p=0.004), while no effect was seen in IgG treated mice (p=0.17). Pre-treatment engraftment levels for IgG and anti-CD47 treated mice were not different (p=0.86). (B) Treatment with anti-CD47 antibody dramatically reduced bone marrow CD34+ LSC-enriched cell engraftment (p=0.066), while no effect was seen in IgG treated mice (p=0.33). Pre- treatment engraftment levels for IgG and anti-CD47 treated mice were not different (p=0.11).

A Monoclonal Antibody Directed Against Human CD47 Enables Phagocytosis of AML In Vivo and Targets AML LSC The in vivo mechanism of the anti-human CD47 antibody was investigated using two approaches to determine if the blocking B6H12.2 anti-CD47 antibody eliminates human AML in vivo by enabling phagocytosis of these cells. First, primary human AML LSC were transduced with a lentivirus expressing GFP and transplanted into NOG mice. Engrafted mice were treated with a single dose of mouse IgG or anti-CD47 antibody and 4 hours later bone marrow, spleen, and liver were examined for the presence of GFP-positive human leukemia cells within F4/80 positive mouse phagocytes by flow cytometry. Unlike IgG control-treated mice, human GFP+ AML cells were detected within phagocytes from all three tissues in anti-CD47-treated mice (Figure 13A,B). In the second experiment, mouse phagocytes were depleted in human AML LSC-engrafted mice prior to treatment with anti-CD47 antibody by administering liposomal clodronate, which accumulates in lysosomes resulting in death of phagocytes (Figure 14A). Depletion of phagocytes inhibited the ability of anti- CD47 antibody to eliminate human AML from both the peripheral blood and bone marrow in vivo (Figure 13C). Finally, in vivo targeting of AML LSC was investigated. First, the percentage of CD34+ LSC- enriched human leukemia cells present in the bone marrow after treatment was determined by flow cytometry. Treatment with anti-CD47 antibody resulted in a statistically significant decrease in the percentage of human CD34+ leukemia cells remaining in the bone marrow after treatment (Figure 13D and Figure 12B). Next, targeting of AML LSC was functionally assessed by secondary transplantation of bone marrow from IgG control or anti-CD47 antibody-treated mice. Secondary mice transplanted from IgG treated mice engrafted human leukemia in the peripheral blood (Figure 15) and bone marrow (Figure 13E). However, secondary mice transplanted from anti-CD47 treated mice developed no engraftment in the blood or marrow, which could be the result of in vivo antibody coating with anti- CD47. Regardless, the lack of secondary engraftment clearly indicates that treatment with anti-CD47 antibody targeted AML LSC in vivo.

35 A B

C

D E

Figure 13: A Monoclonal Antibody Directed against Human CD47 Enables Phagocytosis of AML In Vivo and Targets LSC (A and B) Flow cytometry plots (A) and quantitation (B) from NOG mice engrafted with lentivirally transduced GFP-positive SU028 AML LSC, 4 hr after treatment with a single 100 mg intraperitoneal dose of anti-CD47 antibody (B6H12.2) or control IgG (n = 2 for each). Cell suspensions from bone marrow, spleen, and liver were stained for human CD45 and mouse F4/80, which recognizes phagocytes. All plots are gated on human CD45-

36

negative cells. Double-positive events represent GFP-positive leukemia cells within mouse phagocytes. Error bars indicate the standard deviation of duplicate measurements. (C) NOG mice engrafted with the indicated AML LSC were treated with liposomal clodronate to deplete phagocytes and administered daily intraperitoneal injections of anti-CD47 antibody for 14 days. The percentage of residual human leukemia cells in the peripheral blood (left) and bone marrow (right) was determined as described above. Depletion of phagocytes resulted in a statistically significant inhibition of the ability of anti-CD47 antibody to eliminate human AML from both the peripheral blood (p = 0.03) and bone marrow (p = 0.04). Clodronate treatment by itself had no effect on leukemic engraftment (Figures S12A and S12B). (D) The percentage of human CD34+ LSC-enriched human leukemia cells remaining in the bone marrow after treatment with either IgG control or anti-CD47 antibody was determined by flow cytometry. Treatment with anti-CD47 antibody resulted in a statistically significant decrease (p < 0.001) compared to the control. (E) 500,000 whole bone marrow cells from IgG control (n = 12) or anti-CD47 (n = 9) antibody-treated mice were secondarily transplanted into NOG mice. Twelve weeks later, secondary mice were sacrificed and analyzed for human leukemia engraftment in the peripheral blood (Figure 15) and bone marrow as described above. Secondary mice transplanted from IgG-treated mice engrafted human leukemia in the bone marrow, while secondary mice transplanted from anti-CD47-treated mice developed no engraftment (p < 0.001). Statistical significance was determined using Fisher’s exact test.

Figure 14: Clodronate Treatment Depletes Mouse Phagocytes, but not Human AML In Vivo (A) The percentage of mouse F4/80-positive phagocytes in the bone marrow after treatement with liposomal clodronate (n=3) according to the Experimental Procedures is markedly reduced compared to controls (n=3, p=0.01). (B) NOG mice engrafted with the indicated AML LSC, were treated with liposomal clodronate to deplete phagocytes, and then treated with daily intraperitoneal injections of control IgG antibody for 14 days. The percentage of residual human leukemia cells in the peripheral blood (left) and bone marrow (right) was determined as described above. Depletion of phagocytes resulted in no statistically significant elimination of human AML from either the peripheral blood or bone marrow.

37

Figure 15: A Monoclonal Antibody Directed Against Human CD47 Targets AML LSC In Vivo 500,000 whole bone marrow cells from IgG control (n=12) or anti-CD47 (n=9) antibody-treated mice were secondarily transplanted into NOG mice. 12 weeks later, secondary mice were sacrificed and analyzed for human leukemia engraftment in the peripheral blood as described. Secondary mice transplanted from IgG treated mice engrafted human leukemia in the peripheral blood, while secondary mice transplanted from anti-CD47 treated mice developed no engraftment in the blood (p<0.001). Statistical significance was determined using Fisher’s exact test.

DISCUSSION

We report here the identification of higher expression of CD47 on AML LSC compared to their normal counterparts and hypothesize that increased expression of CD47 on human AML contributes to pathogenesis by inhibiting phagocytosis of these cells through the interaction of CD47 with SIRPα (Figure 16A). Consistent with this hypothesis, we demonstrate that increased expression of CD47 in human AML is associated with decreased overall survival. We also demonstrate that disruption of the CD47-SIRPα interaction with monoclonal antibodies directed against CD47 preferentially enables phagocytosis of AML LSC by macrophages, inhibits engraftment, and targets AML LSC in vivo. Together, these results establish the rationale for considering the use of an anti- CD47 monoclonal antibody as a novel therapy for human AML.

38

Figure 16: Increased CD47 Expression on AML LSC Inhibits Phagocytosis and Can Be Targeted by a Blocking Monoclonal Antibody to Eliminate These Cells (A) Increased expression of CD47 on human AML LSC contributes to pathogenesis by inhibiting phagocytosis of these cells through the binding of CD47 to SIRPα which delivers a strong inhibitory signal able to overcome an as yet uncharacterized positive stimulus for phagocytosis. (B) A blocking monoclonal antibody directed against CD47 would treat human AML by disrupting the dominant inhibitory SIRPα signal in phagocytes, thereby unmasking the positive stimulus and enabling phagocytosis and elimination of AML LSC. (C) The combination of one antibody able to bind an LSC-specific molecule and engage Fc receptors on phagocytes, thereby delivering a strong positive signal for phagocytosis, with a second blocking anti-CD47 antibody may result in a synergistic stimulus for phagocytosis and specific elimination of AML LSC.

The enabling of phagocytosis by blocking monoclonal antibodies directed against CD47 is a novel mechanism of action for a therapeutic monoclonal antibody in the treatment of cancer. Currently approved antibody therapies are believed to act via stimulation of antibody-dependent cellular 39 cytotoxicity (ADCC), disruption of critical receptor-ligand interactions, or through unknown mechanisms (92). Blocking anti-CD47 monoclonal antibodies would treat human AML by enabling phagocytosis and elimination of AML LSC (Figure 16B). In support of this mechanism of action in vivo, we show that treatment of human AML LSC-engrafted mice with anti-CD47 antibody results in rapid phagocytosis of AML cells (Figure 13A,B), and that depletion of phagocytes with clodronate abrogates this effect (Figure 13C). As demonstrated here, CD47 contributes to pathogenesis by conferring a survival advantage to LSC and progeny blasts through evasion of phagocytosis by the innate immune system. Moreover, some dendritic cells express SIRPα (93-96), and we propose that increased CD47 expression on AML LSC also serves to prevent the activation of adaptive immune responses. AML LSC are enriched in the Lin-CD34+CD38- fraction, which in normal bone marrow contains HSC and MPP. The identification of cell surface molecules that can distinguish between leukemic and normal stem cells is essential for flow cytometry-based assessment of minimal residual disease (MRD) and for the development of prospective separation strategies for use in cellular therapies. Several candidate molecules have recently been identified, including CD123 (97), CD44 (98), CD96 (99), CLL-1 (100), and now CD47. We demonstrate that not only is CD47 more highly expressed on AML LSC compared to normal HSC and MPP, but also that this differential expression can be used to separate normal HSC/MPP from leukemia cells. This is the first demonstration of the prospective separation of normal HSC from leukemia cells in the same patient sample, and offers the possibility of leukemia-depleted autologous HSC transplantation therapies.

Targeting of CD47 on AML LSC with Therapeutic Monoclonal Antibodies Cell surface molecules preferentially expressed on AML LSC compared to their normal counterparts are candidates for targeting with therapeutic monoclonal antibodies. Thus far, several molecules have been targeted on AML including CD33 (92), CD44 (98), CD123 (101), and now CD47. Here we report that a monoclonal antibody directed against CD47 targets AML LSC in vivo, as shown by direct reduction in the percentage of human CD34+ LSC-enriched leukemia cells in the bone marrow, and complete elimination of engraftment in secondary transplants (Figure 13D,E). Targeting of leukemia cells and cell lines with anti-CD47 antibodies has previously been reported to directly induce apoptosis. Treatment of primary human B-CLL cells was shown to induce caspase-independent cell death (79), while a different anti-CD47 antibody was shown to induce apoptosis of several hematopoietic cell lines (89-91). These reports raise the alternative hypothesis that anti-CD47 antibodies induce apoptosis of AML LSC, which are then recognized by macrophages and phagocytosed. However, several caveats must be considered when comparing these prior reports to our current study. First, the report on B-CLL involved a mature lymphocytic neoplasm, which is very biologically different from immature aggressive AML, and demonstrated apoptosis not with soluble antibody, but only with cross-linking of antibody, which can result in different effects. Second, the 40 additional reports utilized cell lines and not primary leukemia cells, which are very biologically distinct regarding both proliferation and cell death. Ultimately, we feel that it is not possible to extrapolate the effect of anti-CD47 antibodies from these reports to primary human AML cells. Several lines of evidence suggest that targeting of CD47 with a monoclonal antibody acts by disrupting the CD47-SIRPα interaction, thereby preventing a phagocytic inhibitory signal, rather than by acting through induction of apoptosis, ADCC, or other mechanisms. First, two blocking anti-CD47 antibodies enabled AML LSC phagocytosis, while one non-blocking antibody did not, even though all three bind the cells similarly (Figure 7C). Second, an anti-mouse SIRPα antibody also enabled phagocytosis of human AML LSC by mouse macrophages, demonstrating phagocytosis without direct binding of antibody to AML LSC (Figure 7C). Third, in the case of the B6H12.2 antibody used for most of our experiments, no direct induction of apoptosis of primary AML LSC was detected when added either as a soluble antibody or as an immobilized plate-bound antibody (Figure 7D and Figure 9D). Fourth, phagocytosis of AML LSC was detected as early as 15 minutes after addition of anti-CD47 antibody, while no apoptosis was detected at 2 hours (Figure 7D and Figure 9E). In fact, only minimal Annexin V-positive staining was detected on Jurkat cells 2 hours after incubation with immobilized plate-bound anti-CD47 antibody (Figure 9C). Fifth, if phagocytosis were occurring secondary to apoptosis, then depletion of phagocytes with clodronate should not inhibit the effect of the antibody, which would still directly kill the leukemia cells. However, clodronate did inhibit the ability of anti- CD47 antibody to deplete human AML (Figure 13C), indicating that enabling of phagocytosis is the most likely mechanism. Finally, the isotype-matched anti-CD45 antibody, which also binds LSC, failed to produce the same effects, making ADCC less likely (Figure 7). In fact, the B6H12.2 antibody is mouse isotype IgG1, which is less effective at engaging mouse Fc receptors than antibodies of isotype IgG2a or IgG2b (102). For human clinical therapies, blocking CD47 on AML LSC with humanized monoclonal antibodies should promote LSC phagocytosis through a similar mechanism, as indicated by the human macrophage-mediated in vitro phagocytosis (Figure 7A,C). The experimental evidence presented here provides the rationale for anti-CD47 monoclonal antibodies as monotherapy for AML. However, such antibodies may be equally, if not more effective as part of a combination strategy. The combination of a blocking anti-CD47 antibody with a second antibody able to bind an LSC-specific molecule (for example CD96) and engage Fc receptors on phagocytes may result in a synergistic stimulus for phagocytosis and specific elimination of AML LSC (Figure 16C). Furthermore, combinations of monoclonal antibodies to AML LSC that include blocking anti-CD47 and human IgG1 antibodies directed against two other cell surface antigens will be more likely to eliminate leukemia cells with pre-existing epitope variants or antigen loss that are likely to recur in patients treated with a single antibody.

41 CHAPTER 3

A monoclonal antibody against CD47 eliminates human acute lymphoblastic leukemia

42

SUMMARY

Acute lymphoblastic leukemia (ALL) is the most common pediatric malignancy, and represents 15% of adult leukemias. Although overall prognosis for pediatric ALL is favorable, high-risk pediatric patients, and most adult patients, have significantly worse outcomes. Multi-agent chemotherapy is standard therapy for both pediatric and adult ALL, but is associated with systemic toxicity and long- term side effects, and is relatively ineffective against certain ALL subtypes. Recent efforts have focused on the development of targeted therapies for ALL including monoclonal antibodies. Here we report the identification of CD47, a protein that inhibits phagocytosis, as an antibody target in standard and high-risk ALL. CD47 was more highly expressed on a subset of human ALL patient samples compared to normal cell counterparts and was an independent predictor of survival and relapse in several ALL patient cohorts. Furthermore, a blocking monoclonal antibody against CD47 enabled phagocytosis of ALL cells by macrophages in vitro. Coating of ALL cells with anti-CD47 antibody ex vivo inhibited tumor engraftment in vivo. Lastly, anti-CD47 antibody eliminated ALL in the peripheral blood, bone marrow, spleen, and liver of mice engrafted with primary ALL. These data provide pre- clinical support for the development of an anti-CD47 antibody therapy for treatment of human ALL.

INTRODUCTION

Acute lymphoblastic leukemia (ALL), a clonal malignancy of lymphocyte precursors, is the most common malignancy in children, comprising nearly one third of all pediatric cancers. Conversely, ALL in adults is less common, comprising 15% of all de novo leukemias. More than 80% of children diagnosed with ALL can achieve cure with multi-agent treatment regimens.(103) In contrast, the prognosis for adults is significantly worse, with a five-year event-free survival (EFS) around 40% (103). Within both pediatric and adult ALL cohorts, subsets of patients have significantly worse outcomes with stratification into high-risk categories based upon several criteria including age, initial white blood cell count, presence of extramedullary disease at diagnosis, minimal residual disease, cytogenetic and karyotype analysis, and others. (104, 105) In pediatric cases, high-risk patients have relatively poor prognoses with an estimated four-year EFS of 46% compared to 91% for standard-risk patients.(105) Recently, cytogenetic abnormalities and karyotype analysis have been incorporated into standard clinical prognostics. Specifically, the presence of BCR-ABL (Ph+) or MLL rearrangements are associated with an unfavorable prognosis, while the TEL-AML1 rearrangement or trisomy of 4, 10, or 17 are more favorable(105) In addition, patients with hyperdiploid karyotypes experience better outcomes, while hypodiploidy has been associated with high risk of adverse outcomes.(106) Although standard multi-agent chemotherapy cures a significant number of patients with standard-risk pediatric ALL, these same therapies are significantly less effective in both the high-risk

43 pediatric population and in all adults with ALL. Thus, additional therapies are necessary to more effectively treat these patient subsets. As an alternative to chemotherapy, monoclonal antibodies have recently emerged as an attractive therapeutic modality due to the ability to selectively target leukemia cells, thereby minimizing systemic toxicity. Indeed, several monoclonal antibodies are currently in clinical trials for the treatment of ALL (reviewed in (107)). In our previous investigation, we identified CD47 as a therapeutic antibody target in acute myeloid leukemia (AML)(108), and hypothesize that a monoclonal antibody against CD47 could be similarly effective in ALL. CD47 is a widely expressed transmembrane protein that has been implicated in activation, cell migration and adhesion, and multiple immunologic and neurologic processes (reviewed in (109)). In addition, CD47 functions as an inhibitor of phagocytosis by binding its ligand, signal regulatory protein alpha (SIRPα), on phagocytes, which in turn initiates a signal transduction cascade leading to inhibition of phagocytosis.(74-78) While this function is partly attributed to self-recognition in normal physiologic conditions, many cancers appear to upregulate CD47 as a mechanism of immune evasion. This idea is supported by our recent finding that CD47 plays a role in evasion of AML from the innate immune system. Specifically, the ability of a human leukemia cell line to engraft and disseminate in immunodeficient mice was dependent on CD47- mediated evasion of phagocytosis.(80) We further demonstrated that this mechanism could be therapeutically targeted in human cancers by a monoclonal blocking anti-CD47 antibody that could eliminate human AML in vivo(108) and bladder cancer in vitro.(110) In this study, we investigated whether a blocking monoclonal antibody against CD47 could eliminate primary human ALL in vitro and in vivo, in order to determine the pre-clinical feasibility of an anti-CD47 antibody as a therapy in standard and high-risk ALL.

EXPERIMENTAL PROCEDURES

Human Samples Normal human bone marrow cells were purchased from AllCells Inc. (Emeryville, CA, USA). Human ALL samples were obtained from patients at the Stanford University Medical Center, with informed consent, according to an IRB-approved protocol (Stanford IRB# 11177).

Flow Cytometry Analysis The following antibodies were used for analysis of ALL and NBM cells: CD3 APC-Cy7 and CD19 APC (BD Biosciences, San Jose, CA, USA). Analysis of CD47 expression was performed with an anti- human CD47 FITC antibody (clone B6H12.2, BD Biosciences). For human engraftment analysis in mice, the following antibodies were used: mouse Ter119 PeCy5, mouse CD45.1 PeCy7, human CD45 PB, human CD19 APC, and human CD3 APC-Cy7 (Ebiosciences, San Diego, CA, USA). 44

ALL microarray gene expression data and statistical analysis We used previously described methods for the univariate and multivariate statistical analyses of CD47 gene expression data and its relationship to clinical and pathological variables.(108) Briefly, gene expression and clinical data were analyzed for three previously described cohorts of ALL patients: 1) a dataset of 360 pediatric ALL patients with B- and T-ALL subtypes, diverse risk profiles and corresponding therapies (16) including a subset (n=205) with available data on disease free survival.(111) Microarray and clinical data were obtained from St. Jude Children’s Research Hospital at http://www.stjuderesearch.org/data/ALL1 and http://www.stjuderesearch.org/data/ALL6/data_request.txt, respectively; 2) a dataset of 207 pediatric B- precursor ALL patients with high-risk features uniformly treated through the Children’s Oncoloy Group Clinical Trial P9906(112) obtained from NCBI through the Gene Expression Omnibus (GSE11877); and 3) 254 pediatric ALL patients registered to Pediatric Oncology Group trials stratified for the presence of recurrent cytogenetic abnormalities and remission versus failure within each cytogenetic group(113) with data obtained from the National Cancer Institute caArray (https://array.nci.nih.gov/caarray/project/willm-00090). Affymetrix probeset summaries were derived from the corresponding microarray raw CEL data files using a Custom Chip Definition File derived from NCBI Reference Sequences (version 12)(114), and then normalized using MAS 5.0 using BioConductor (115). For survival analyses, NM_198793_at was selected as the probeset to represent CD47 mRNA based on it demonstrating highest expression among the 3 probesets for CD47 on the microarrays, and based on its exonic structure capturing the CD47 splice variant expressed in hematopoietic tissues. We assessed the relationship of CD47 mRNA expression and outcomes as continuous variables using univariate Cox proportional-hazards analysis, with disease free or overall survival as the dependent variable (116). Using the coxph function in the R statistical package, the Wald test was used to assess the significance of each covariate, represented by the base-2 logarithms of CD47 mRNA expression. For dichotomous stratification of CD47 expression, we used maximally selected chi-square statistics as implemented within X-tile to define an optimal threshold (117). To guard against erroneous overestimation of p-values through multiple hypothesis testing, we corrected the log- Kaplan-Meier p-values using the Miller-Siegmund method(118) well as sub-sampling (n=1000) based internal cross-validation (117).

Therapeutic antibodies Anti-human CD47 antibodies (B6H12.2, BRIC126, 2D3), anti-SIRPα antibody, IgG control, and anti- CD45 antibodies were used as in (108). The anti-CD47 antibody clone BRIC126 was obtained from AbD Serotec (Raleigh, NC, USA).

45 Generation of mouse and human macrophages Isolation of mouse and human macrophages were performed as previously described.(108) Briefly, femurs and tibias from wild-type Balb/C mice were harvested into a single cell suspension and incubated for 7-10 days in IMDM 10% fetal calf serum with 10ng/ml murine M-CSF (Peprotech, Rocky Hill, NJ, USA) at 37°C. Cells were then washed and adherent cells trypsinized and plated for in vitro phagocytosis assays. For human macrophages, mononuclear cells were isolated from human peripheral blood by ficoll density gradient centrifugation and plated onto 10cm petri dishes at 37°C for one hour. Non-adherent cells were then washed off and the remaining adherent cells were incubated for 7-10 days in IMDM with 10% human AB serum. Cells were then trypsinized and plated for in vitro phagocytosis assays.

In vitro phagocytosis assays Phagocytosis assays were performed as described in.(108) Briefly, bulk ALL cells were CFSE-labeled and incubated with either mouse or human macrophages in the presence of 10µg/ml of the indicated antibodies at a target:effector cell ratio of 4:1 (2x105:5x104). Incubation occurred at 37°C for 2 hours and then analyzed by fluorescent microscopy for phagocytosis using the phagocytic index: number of cells ingested per 100 macrophages.

Ex vivo antibody coating of ALL cells Human ALL cells were incubated with 30µg/ml of either IgG1 isotype control, anti-CD45, or anti- CD47 antibody for 30 minutes at 4°C. Cells were washed and then 1-4x106 cells were transplanted into sublethally-irradiated NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ (NSG) adults or pups and analyzed for ALL engraftment in the peripheral blood and bone marrow 6-10 weeks later. Antibody coating of ALL cells was confirmed by flow cytometry with a secondary antibody prior to transplantation into mice. Sublethal irradiation was 230rads and 100rads for NSG adults and pups, respectively.

In vivo treatment of human ALL engrafted mice 1-4x106 bulk human ALL cells were transplanted intravenously via the retro-orbital sinus into sublethally-irradiated (230 rads) adult NSG mice. Alternatively, ALL cells were transplanted into the facial vein of 2-4 day old sublethally-irradiated (100 rads) NSG pups. Six to ten weeks later peripheral blood and bone marrow ALL engraftment (B-ALL: hCD45+CD19+; T-ALL: hCD45+CD3+) was assessed by tail bleed and aspiration of the femur, respectively. Engrafted mice were treated for 14 days with daily 100µg intraperitoneal injections of either IgG control or anti-CD47 antibody (clone B6H12.2). On day 15, mice were sacrificed and analyzed for ALL engraftment in the peripheral blood, bone marrow, spleen, and liver.

46

Bone marrow tissue section preparation and staining Mouse tibias from antibody-treated NSG mice were harvested and preserved in formalin. Hematoxylin and eosin staining and immunohistochemistry of human CD45+ cells were performed by Comparative Biosciences Inc. (Sunnyvale, CA, USA).

RESULTS

CD47 Expression is Increased on a Subset of Human ALL Cells Compared to Normal Bone Marrow To determine whether CD47 may be involved in the pathogenesis of ALL, we first investigated CD47 cell surface expression on primary human ALL and normal bone marrow cells by flow cytometry. We surveyed 23 diverse patients with ALL that included both precursor B and T lineage subtypes, as well as several patients with cytogenetically high-risk Ph+ B-ALL (Figure 1C). Compared to normal mononuclear bone marrow cells, CD47 was more highly expressed on human ALL samples, approximately 2-fold when considering all samples, with similar expression between B and T subtypes (Figure 1A,C). However, assessing CD47 mRNA expression in a large cohort of ALL patients(113), we found that T-ALL patients expressed significantly higher levels compared to B-ALL patients (Figure 1B). Additionally, we found that cytogenetically high-risk Ph+ B-ALL had lower CD47 expression compared to cytogenetically normal B-ALL, and were similar to normal bone marrow (Figure 1A).

CD47 Expression is an Independent Prognostic Predictor in Mixed and High-Risk ALL Since CD47 expression was increased on ALL samples, and given the observed heterogeneity in CD47 expression across ALL subtypes, we investigated whether the level of CD47 expression correlated with clinical prognosis. First, CD47 expression was investigated as a prognostic predictor in pediatric ALL patients with mixed risk and treatment utilizing gene expression data from a previously described patient cohort(119). 360 patients were stratified into high and low CD47-expressing groups based on an optimal cutpoint (see methods) and clinical outcomes were determined. Among the subset of this cohort with available outcome data (n=205)(111), patients expressing higher levels of CD47 had worse outcomes, whether CD47 expression was tested as a continuous variable (p=0.03; HR 1.78 per 2- fold change in CD47 expression; 95% CI 1.05-3.03), or as a dichotomous variable relative to an internally validated optimal threshold (uncorrected p=0.0005, corrected (120) p=0.01; HR 3.05; 95% CI 1.49- 6.26) (Figure 2A).

47

Figure 1: CD47 expression is increased on a subset of human ALL cells compared to normal bone marrow (A) Relative CD47 protein expression was determined on normal human bone marrow cells and human ALL cells by flow cytometry. CD47 expression was increased on bulk ALL cells compared to normal bone marrow cells (p=0.006). Within all samples, normalized mean expression (and range) was determined as follows: NBM 575.3 (486.6-680.9), bulk ALL 954.7 (501.3 – 1794.5), B-ALL 980.5 (501.3 – 1628.4), T-ALL 1034.9 (626.2 – 1794.5), Ph+ ALL 676.2 (628.2 – 719.7). Differences between mean expression of NBM and T-ALL (p=0.03), NBM and B-ALL (p=0.006) were statistically significant. The difference between mean expression of NBM and Ph+ ALL was not statistically significant (p=0.21, student’s t-test). (B) Relative CD47 mRNA expression was determined on a large cohort of ALL patients (n=254) from,(113) and subdivided into B or T-ALL subtypes. In three different probe sets, CD47 expression was higher in T-ALL patients compared to B-ALL patients (p<0.0001 for all three probe sets). (C) Clinical characteristics of ALL patient samples used for in vitro and in vivo experiments. POG9900=vincristine, dexamethasone, PEG-asparaginase, and intrathecal methotrexate. CALGB8811=induction therapy including cyclophosphamide, daunorubicine, vincristine, prednisone, and L-asparaginase. CNS2=WBC count <5cells/µl of CSF and blasts present on cytocentrifuge examination. CNS3=WBC count > 5 cells/µl of CSF with blasts present on cytocentrifuge examination. Hypodiploid= <45 chromosomes per leukemia cell. n/a=not available.

48

Second, to investigate the prognostic power of CD47 expression in high-risk ALL patients, clinical outcome in a uniformly treated previously described cohort of 207 high-risk pediatric ALL patients was investigated.(112) For this cohort, high-risk was defined by age>10 years, presenting WBC count>50,000/µl, hypodiploidy, BCR-ABL positive disease, and central nervous system or testicular involvement. In these high-risk ALL patients, higher CD47 expression correlated with a worse overall survival when CD47 expression was again considered as either a continuous variable (p=0.0009, HR 3.59 per 2-fold change in CD47 expression; 95% CI 1.70 to 7.61), or as a dichotomous one relative to an internally validated optimal threshold (uncorrected p=0.001, corrected p=0.01; HR 2.80; 95% CI 1.21 to 6.50) (Figure 2B). In multivariate analysis, CD47 expression remained a significant prognostic factor when age at diagnosis, gender, WBC count, CNS involvement, and minimal residual disease were considered as covariates (Figure 2C). Lastly, we utilized a third independent gene expression dataset to investigate whether CD47 expression could predict disease relapse.(113) Indeed, CD47 expression was higher in patients failing to achieve a complete remission (CR) compared to those that did achieve a CR (Figure 2D). Taken together, these separate observations among distinct and diverse cohorts establish that higher expression of CD47 is an independent predictor of adverse outcomes in pediatric patients with standard- and high- risk ALL, including induction failure, relapse, and death.

Blocking Monoclonal Antibodies Against CD47 Enable Phagocytosis of ALL Cells Next, we investigated whether ALL cells could be eliminated by macrophage phagocytosis enabled through blockade of the CD47-SIRPα interaction with a blocking anti-CD47 antibody. First, we incubated human macrophages with fluorescently-labeled ALL cells in the presence of an IgG1 isotype control, anti-CD45 isotype-matched, or anti-CD47 antibody and measured phagocytosis by immunofluorescence microscopy (Figure 3A). Two different blocking anti-CD47 antibodies (B6H12.2 and BRIC126) enabled phagocytosis of ALL cells compared to IgG1 isotype and anti-CD45 control antibodies as measured by significant increases in the phagocytic index (Figure 3B). In addition, anti- CD47 antibodies enabled phagocytosis of all ALL subtypes profiled, including those with lower CD47 expression (Ph+ALL), as well as those with cytogenetically high-risk (Ph+ALL and MLL+ALL). Since several studies report that CD47-SIRPα signaling may be species-specific,(86, 88) the ability of anti- CD47 antibody-coated human cells to be phagocytosed by mouse macrophages was determined before proceeding with in vivo antibody treatment experiments in mouse xenotransplants. Similar to human macrophages, two blocking anti-CD47 antibodies (B6H12.2 and BRIC126) enabled phagocytosis of ALL cells by mouse macrophage effectors compared to IgG1 isotype and anti-CD45 antibody controls (Figure 3C). Furthermore, no phagocytosis was observed with a non-blocking anti-CD47 antibody (2D3). Lastly, blockade of SIRPα with an anti-mouse SIRPα antibody also resulted in increased

49 phagocytosis, thus supporting the proposed mechanism of increased phagocytosis resulting from disruption of the CD47-SIRPα interaction (Figure 3C).

Figure 2: CD47 expression is an independent prognostic predictor in mixed and high-risk ALL (A) Pediatric ALL patients (n=360) (119) were stratified into CD47 high- and low-expressing groups based on an optimal cut point. Disease-free survival (DFS) was determined by Kaplan-Meier analysis. CD47 high-expressing patients had a worse DFS compared to CD47 low-expressing patients when CD47 expression was considered as a continuous variable. (B) Pediatric ALL patients (n=207) with high-risk (as defined by age>10 years, presenting WBC count>50,000/µl, hypodiploidy, and BCR-ABL positive disease) and uniform treatment(112) were stratified into CD47 high- and low-expressing groups using a similar approach as in A. CD47 high-expressing patients had a worse overall survival compared to CD47 low-expressing patients (p=0.0009). (C) Multivariate analysis of prognostic covariates was performed from patients analyzed for CD47 expression in (B).(112) When incorporated into this multivariate analysis, CD47 expression still remained prognostic (p=0.035). (D) ALL patients from(113) were stratified into groups either achieving a complete remission (CR) or not achieving a CR (no CR). CD47 expression was higher in patients failing to receive a CR compared to those who did (p=0.0056).

50

Figure 3: Blocking monoclonal antibodies enable phagocytosis of ALL cells by human and mouse macrophages in vitro (A) Primary human ALL cells from various ALL subtypes were fluorescently-labeled and incubated with human macrophages in the presence of the indicated antibodies for 2 hours, after which they were examined by immunofluorescence microscopy (Leica) with images acquired by Image-Pro Plus software. Representative photomicrographs are shown with ALL cells (green) and macrophages. Arrows represent macrophages containing phagocytosed ALL cells. (B) The phagocytic index (number of target cells ingested per 100 macrophages) was determined for the indicated antibodies with human macrophage effector cells. The anti-CD47 antibodies B6H12.2 and BRIC126 enabled significant levels of phagocytosis compared to IgG1 isotype or anti-CD45 antibody controls (p<0.0001). (C) Using mouse macrophages as effector cells, the blocking anti-CD47 antibodies B6H12.2 and BRIC126 enabled phagocytosis of ALL cells compared to IgG1 isotype and anti-CD45 antibody controls (p<0.0001). Anti-mouse SIRPα antibody also enabled phagocytosis compared to IgG1 isotype control (p=0.002). In contrast, the non-blocking anti-CD47 antibody, 2D3, did not enable phagocytosis compared to IgG1 isotype control (p=0.17). Each data point/symbol represents a distinct ALL patient sample as labeled in Figure 1C. All statistical comparisons were conducted with a two-sided student t-test.

Ex Vivo Coating of ALL Cells with an Anti-CD47 Antibody Inhibits Leukemic Engraftment The ability of a blocking anti-CD47 antibody to eliminate ALL in vivo was investigated by two independent methods. First, the ability of anti-CD47 antibody to inhibit ALL engraftment was determined using an antibody pre-coating assay. ALL cells were coated ex vivo with either IgG1 isotype control, anti-CD45, or anti-CD47 antibody (B6H12.2), transplanted into sublethally-irradiated immunodeficient NOD/SCID/Il2γr null (NSG) mice, and measured for ALL engraftment in the peripheral blood and bone marrow 6-10 weeks later. Prior to transplantation, coating of ALL cells with

51 antibody was verified by flow cytometry (Figure 4A). Antibody pre-coating experiments were performed with both primary B- and T-ALL samples to include the two major disease subtypes. Anti- CD47 antibody significantly inhibited leukemic engraftment of both B- and T-ALL cells in the peripheral blood (Figure 4B) and bone marrow (Figure 4C) compared to IgG1 isotype or anti-CD45 antibody controls. Interestingly, pre-coating of T-ALL cells (but not B-ALL cells) with anti-CD45 antibody reduced tumor engraftment. Anti-CD45-mediated inhibition of T-ALL engraftment was unlikely due to antibody-opsonization, given that B-ALL cells coated with anti-CD45 antibody engrafted similarly to uncoated cells incubated with IgG1 isotype control antibody. Rather the effect observed in T-ALL may be due to modulation of CD45-dependent functions important to engraftment for T-ALL but not B-ALL. Regardless, pre-coating with the anti-CD47 antibody nearly completely eliminated ALL engraftment in vivo.

Figure 4: Ex vivo coating of ALL cells with an anti-CD47 antibody inhibits leukemic engraftment (A) ALL cells were incubated with the indicated antibodies in vitro, and positive cell coating was detected by staining a portion of the cells with a fluorescently-labeled secondary antibody. A flow cytometry plot of a representative ALL sample is shown. (B-C) Pre-coated ALL cells were then transplanted into NSG mice, and human ALL chimerism was assessed 6-10 weeks later in the peripheral blood (B) or bone marrow (C). Ex vivo coating of ALL cells (ALL4 and ALL8) with anti-CD47 antibody inhibited engraftment in the peripheral blood compared to IgG1 isotype control (p=0.02). Ex vivo coating of ALL cells with anti-CD47 antibody inhibited bone marrow engraftment compared to IgG1 isotype control (p=0.02), while no difference in engraftment levels were detected between anti-CD45 antibody and IgG1 isotype control (p=0.67, considering both B and T-ALL samples). Each symbol represents a different primary ALL sample, with each point representing a different mouse. p-values were calculated using the Fisher’s exact test. Red diamond=ALL4, blue diamond=ALL4. 52

Anti-CD47 Antibody Eliminates ALL Engraftment in the Peripheral Blood and Bone Marrow In the second method of investigating the in vivo efficacy of an anti-CD47 antibody against human ALL, mice were first stably engrafted with ALL cells and then treated with antibody. 1-4x106 B- or T-ALL cells were transplanted into sublethally-irradiated NSG newborn pups or adults. Six to ten weeks later, ALL engraftment was measured in the peripheral blood and bone marrow by flow cytometry. Those mice that had significant levels of ALL engraftment were then selected for in vivo antibody therapy (Figure 5A), as determined by greater than 10% human chimerism in the peripheral blood and/or bone marrow with engraftment ranging from 10-97%. ALL engrafted mice were treated with daily intraperitoneal injections of 100µg IgG control or anti-CD47 antibody (B6H12.2) for 14 days. This dosing regimen was selected based on our prior study demonstrating elimination of AML in mouse xenotransplants.(108) Tumor burden was then measured post-treatment in the peripheral blood and bone marrow by flow cytometry. Compared to IgG control, anti-CD47 antibody therapy reduced the level of circulating leukemia, and in most cases eliminated ALL from the peripheral blood (Figure 5A,B). This effect was observed for mice transplanted with both B- and T-ALL cells. Similarly, anti- CD47 antibody reduced or eliminated ALL engraftment in the bone marrow, while ALL disease burden increased with IgG control treatment (Figure 5C). Bone marrow histology of antibody-treated mice revealed infiltration of monomorphic leukemic blasts in control IgG-treated mice (Figure 5D). Anti- CD47 antibody-treated bone marrow demonstrated normal mouse hematopoietic cells with cleared hypocellular areas. Immunohistochemistry of mouse marrows confirmed near complete invasion of human CD45-positive leukemic blasts in IgG-treated marrow compared to few human CD45-positive leukemia cells observed in anti-CD47 antibody-treated marrow (Figure 5D).

Anti-CD47 Antibody Eliminates ALL Engraftment in the Spleen and Liver Hepatomegaly and splenomegaly can cause clinical complications and are a common finding in ALL, being observed in up to 69% of patients at diagnosis.(121, 122) To determine whether anti- CD47 antibody could potentially treat hepatosplenomegaly in ALL, we investigated the ability of anti- CD47 antibody to eliminate ALL engrafted in the spleen and liver. Of the ALL samples utilized for in vivo treatment studies, we identified three B-ALL patient samples (ALL8, ALL21, and ALL22) that gave rise to disease in the spleen and/or liver, with associated splenomegaly, when transplanted into NSG mice. These mice were treated for 14 days with the identical regimen of either IgG or anti-CD47 antibody as in Figure 5 with spleen weights and ALL chimerism in the spleen and liver measured post- treatment. Control IgG-treated B-ALL-engrafted mice exhibited significant splenomegaly compared to untransplanted NSG mice (Figure 6A,B). In contrast, anti-CD47 antibody treatment reduced ALL- induced splenomegaly to spleen sizes similar to untransplanted NSG mice (Figures 6A,B). To determine whether this effect was due to direct elimination of ALL cells in the spleen, the spleens of B- ALL-engrafted mice treated with IgG control or anti-CD47 antibody were analyzed for ALL disease burden. Compared to IgG-treated mice, anti-CD47 antibody significantly eliminated B-ALL 53 engraftment in the spleen (Figure 6C). Similarly, anti-CD47 antibody significantly eliminated ALL in the liver compared to the extensive leukemic inflitration observed with control IgG treatment (Figure 6D). These results indicate that anti-CD47 antibody is highly effective in eliminating ALL in the spleen and liver, in addition to the peripheral blood and bone marrow.

Figure 5: Anti-CD47 antibody eliminates ALL engraftment in the peripheral blood and bone marrow (A) NSG mice engrafted with primary B and T-ALL patient samples were treated for 14 days with daily intraperitoneal injections of 100µg IgG control or anti-CD47 antibody. Peripheral blood human ALL chimerism (huCD45+CD19/CD3+) pre- and post-treatment were measured by flow cytometry. Peripheral blood chimerism is shown from representative treatment mice. (B) Anti-CD47 antibody treatment reduced the level of circulating leukemia compared to IgG control (p=0.0002). (C) Anti-CD47 antibody treatment also reduced ALL engraftment in the bone marrow compared to IgG control (p=0.0004). Each symbol represents a different patient sample, with each data point representing a different mouse. (D) (Top) Hematoxylin and eosin bone marrow sections from representative mice engrafted with B-ALL post-treatment. IgG-treated marrows were primarily packed with monomorphic leukemic blasts, while anti-CD47 antibody-treated marrows demonstrated areas of normal mouse hematopoiesis. (Bottom) Leukemic infiltration was confirmed by immunohistochemical analysis of human CD45 demonstrating robust human leukemia infiltration in IgG-treated bone marrow compared to anti-CD47 antibody- treated marrow.

54

Figure 6: Anti-CD47 antibody eliminates ALL engraftment in the spleen and liver (A) NSG mice engrafted with primary B-ALL cells from sample ALL8, ALL21, or ALL22 were treated for 14 days with daily injections of IgG control or anti-CD47 antibody. Spleens were then harvested, with representative spleens from IgG control or anti-CD47 antibody treatment shown. (B) Spleen weights were determined from mice treated with anti-CD47 antibody demonstrating a reduction in spleen size compared to control IgG-treated mice (p=0.04) to sizes similar to that of normal spleens (p=0.09). Control IgG-treated mice demonstrate splenomegaly compared to normal mice (p=0.0002, student t-test). (C-D) Levels of ALL engraftment were determined at the end of antibody treatment in the spleen (C) and liver (D). Compared to IgG control, treatment with anti-CD47 antibody eliminated ALL disease in the spleen (p<0.0001) and liver (p<0.0001, student t-test).

DISCUSSION

We report here that CD47 is expressed at high levels on a large subset of human ALL subtypes, that cell surface CD47 is a monoclonal antibody target for eliminating ALL blasts by enhancing innate immune system recognition of leukemic blasts by macrophage-mediated phagocytosis, and that CD47 itself is an independent prognostic variable in ALL that can predict disease free survival, overall survival, and relapse in both mixed and high-risk ALL patients. Together, these data suggest that ALL pathogenesis relies on mechanisms to evade innate immune recognition and that modulation of the innate immune recognition of tumor cells may be a viable treatment modality.

55 Although most cases of ALL blast clearance by anti-CD47 antibody therapy correlates with elevated levels of CD47 expression on leukemic cells, increased levels of cell surface CD47 expression is not required for the anti-leukemia effect of CD47 blockade. Specifically, while CD47 expression was increased overall on ALL samples compared to normal bone marrow, Ph+ B-ALL samples did not exhibit statistically significant increases in CD47 expression (Figure 1A). Despite lower CD47 expression, Ph+ B-ALL samples were efficiently phagocytosed in the presence of anti-CD47 antibodies (Figures 3B,C). Although the nature of lower CD47 expression on Ph+ samples may simply reflect intrinsic differences in CD47 regulation in this subtype, this also suggests that CD47 regulation is not required for the development of this subset of ALL patients. Thus, anti-CD47 antibody-mediated phagocytosis is likely not entirely due to CD47 expression level. Alternatively, the phagocytosis stimulated by anti-CD47 antibody may be due to an imbalance of pro- and anti-phagocytic signals on ALL blasts with enhancement of a yet unknown positive stimulus for phagocytosis present on ALL cells but not normal cells. Such candidate stimuli include phosphatidylserine,(123) annexin-1,(124) and calreticulin(125) which are targets under current active investigation. We demonstrate here that CD47 is a new antibody target for the treatment of ALL. Within the last few years, several cell surface proteins have been identified as candidate targets, and some monoclonal antibodies have proceeded into early and late phase clinical trials. Most therapeutic antibodies in clinical development have been focused on B-ALL. One such candidate is CD20, as its expression is observed in approximately 40 to 50% of B-ALL cases (reviewed in (126)). Rituximab, an anti-CD20 antibody, initially approved for treatment of B cell lymphoma, has demonstrated a significant survival advantage when added to standard chemotherapy in some ALL clinical trials, particularly the Burkitt’s subtype.(127, 128) Although effective in adult CD20+ B-ALL, there is a paucity of clinical data on the efficacy of rituximab in pediatric ALL. In contrast to CD20, CD22 is expressed in a larger percentage of B-ALL cases and is present on greater than 90% of B-ALL patients. Epratuzumab, a humanized monoclonal anti-CD22 antibody, is currently being investigated in clinical trials. Although early clinical studies with epratuzumab as a single agent in relapsed ALL showed limited effect,(129) anti-CD22 antibody-immunotoxin conjugates may improve the efficacy of epratuzumab, since CD22 is reported to be rapidly internalized upon antibody binding.(130) Several immunoconjugates directed against CD22 are currently being explored in Phase I trials.(131, 132) In addition, antibodies and immunotoxins to other antigens including CD19 are currently being explored,(133) (reviewed in (134)). Perhaps the best success of targeted therapy has been observed in Ph+ B-ALL. Since its demonstration of efficacy in chronic myeloid leukemia, imatinib, an ABL tyrosine kinase inhibitor, has been utilized in Ph+ B-ALL with preliminary success. As a single agent, imatinib can produce response rates of 20-30%; however, these response durations are short.(135) The combination of imatinib with chemotherapy has been promising, with three-year overall survival rates of 55% in patients treated with imatinib+hyperCVAD compared to 15% for patients receiving hyperCVAD alone.(136) Although 56 nearly all patients with newly diagnosed Ph+ ALL initially respond to imatinib, a substantial portion of patients experience disease relapse during or after treatment. Notably, neither imatinib nor subsequent second generation kinase inhibitors are curative on their own, indicating the need for additional targeted therapies. Although several therapeutic antibodies are in clinical development for B-ALL, there are relatively few antibody therapies for treatment of T-ALL. The most prominent antibody for T-ALL, alemtuzumab, is targeted at CD52, as it is expressed on greater than 95% of normal lymphocytes and at higher levels in T compared to B lymphoblasts.(137) Although pre-clinical data suggest potential efficacy of alemtuzumab, early phase clinical trials do not report a significant benefit as a single agent or in combination with chemotherapy for the treatment of relapsed T-ALL.(138) In contrast to the targeted therapies developed for B-ALL and T-ALL, our data strongly suggest that an anti-CD47 antibody can be effective in eliminating both B- and T-ALL and thus could increase the number of therapeutic options for both. Because anti-CD47 antibody treatment may be able to eliminate ALL blasts with limited toxicity(108) and is equally effective in targeting low, standard, and high-risk ALL, these results provide a strong pre-clinical rationale for development of an anti-CD47 antibody for the treatment of ALL patients.

57 CHAPTER 4

Toxicity studies of an anti-CD47 antibody: the pro-phagocytic signal calreticulin is required for anti-CD47 efficacy

Portions of this chapter were published in the following article: Majeti R*, Chao MP*, Alizadeh AA, Pang WW, Jaiswal S, Weissman IL. CD47 is an adverse prognostic factor and therapeutic antibody target on human myeloid leukemia stem cells. Cell. 2009; 138:268-299.*Co-first author

58

SUMMARY

In normal physiologic conditions, cellular homeostasis is partly regulated by the balance of pro- and anti-phagocytic signals. Recently, we identified increased expression of the anti-phagocytic protein CD47 on several human cancers, and postulated that this allows cancer cells to evade phagocytosis by the innate immune system. Consistent with this hypothesis, we showed that blockade of CD47 with an anti-CD47 monoclonal antibody enabled phagocytosis of cancer cells leading to in vivo tumor elimination. Importantly, even though most normal cells express CD47, albeit at lower levels, the anti- CD47 antibody did not enable their phagocytosis. In order for target cells to be phagocytosed upon blockade of an anti-phagocytic signal, the cells must also display a potent pro-phagocytic signal. We identified calreticulin as a pro-phagocytic signal highly expressed on the surface of several human cancers, but minimally expressed on normal cell counterparts. The level of calreticulin expression correlated with CD47 expression, and increased CD47 expression on cancer cells was necessary for protection from calreticulin-mediated phagocytosis. Significantly, anti-CD47 antibody-mediated phagocytosis required the interaction of target cell-expressed calreticulin with its receptor LRP on phagocytic cells, as blockade of the calreticulin/LRP interaction prevented anti-CD47 antibody mediated phagocytosis. These findings identify calreticulin as the dominant pro-phagocytic signal on several human cancers, provide an explanation for the selective targeting of tumor cells by an anti- CD47 antibody, and highlight a novel role for pro- and anti-phagocytic signals in the immune evasion of cancer.

INTRODUCTION

Malignant cellular transformation is initiated by genetic mutations and epigenetic reprogramming that activate oncogenes and inactivate tumor suppressor pathways leading to inheritance of several hallmarks shared by most cancer cells such as self-sufficiency in growth signals, insensitivity to anti-growth signals, tissue invasion and metastasis, limitless replicative potential, sustained angiogenesis, and evasion of apoptosis (139). In addition to these cell intrinsic properties, recent evidence suggests that many cancers are also able to evade the immune system (140-142). Such mechanisms include downregulation or loss of HLA class I expression, loss of tumor antigen expression, shedding of NKG2D ligands, production of immunosuppressive cytokines, defective death receptor signaling, lack of immune co-stimulation, and others (reviewed in (141-143)). Recently, we have shown that evasion of phagocytosis is another mechanism by which tumor cells escape immunosurveillance through upregulation of the anti-phagocytic signal, CD47 (80, 108, 110, 144). CD47 is a pentaspanin cell surface protein that serves as a “don’t eat me” signal through ligation of its receptor SIRPα, causing inhibition of phagocytosis (74, 145, 146). In normal physiology,

59 CD47 is required to prevent macrophage phagocytosis of self, however in cancer, CD47 permits evasion of host macrophage phagocytosis (reviewed in (144)). We detected increased CD47 expression on several tumor types including mouse (80) and human (108) acute myeloid leukemias (AML) as well as human bladder cancers (110), and demonstrated that CD47 is required for tumorigenicity (80). Importantly, we demonstrated that disruption of the CD47-SIRPα interaction could be a therapeutic modality for eliminating tumors. Specifically, a monoclonal blocking antibody against CD47 enabled phagocytosis of leukemia and bladder cancer cells in vitro and elimination of AML in vivo (108, 110). Given that CD47 is expressed semi-ubiquitously across normal tissues, an anti-CD47 antibody could enable significant toxicity. In this study we investigated the toxicity profile of an anti-CD47 antibody and found that anti-CD47 antibody enabled phagocytosis of tumor cells but not normal cells. This selective targeting was likely due to the notion that in order for target cells to be phagocytosed upon blockade of an anti-phagocytic signal, the cells must also display a potent pro-phagocytic signal. We hypothesized that the selective targeting of tumor cells was due to the presence of an unknown pro- phagocytic stimulus present on tumor cells but not on normal cells that could be unmasked after CD47 blockade. We identify cell surface calreticulin as this pro-phagocytic stimulus that explains the lack of anti-CD47 antibody-mediated toxicity against most normal cells. Calreticulin is principally localized to the endoplasmic reticulum (ER) where it serves as a calcium buffering chaperone involved in intracellular calcium homeostasis and ER calcium capacity (147). Calreticulin is also expressed on the cell surface in conditions of DNA damage (148, 149), likely from cell surface mobilization of intracellular ER stores, where it functions mainly in the clearance of apoptotic cells (125, 150). During apoptosis, cell surface calreticulin serves as a pro-phagocytic signal by binding to its macrophage ligand, low density lipoprotein-related protein (LRP), which leads to engulfment of the target cell (125, 151). CD47 has been implicated in the regulation of phagocytosis of apoptotic cells, as these cells become phagocytosed due to an upregulation of cell surface calreticulin and coordinate loss of CD47 expression (125). Thus, in normal cells, the ultimate cue for phagocytosis depends in part on the imbalance of CD47 expression in favor of calreticulin expression, which occurs in damaged or apoptotic cells. Although several pro-phagocytic signals have been identified on normal dying cells (152, 153), such signals have not previously been identified on cancer cells. Given the likely presence of pro- phagocytic signals on cancer cells revealed by the activity of anti-CD47 antibodies (80, 108, 110), calreticulin is attractive candidate “eat me” signal in cancers given its relationship to CD47 on apoptotic cells. In cancer, little is known about the steady state expression of cell surface calreticulin and its possible relationship to anti-phagocytic signals such as CD47. In this report, we demonstrate that cell surface calreticulin is highly expressed on tumor cells but not normal cells, functions as the dominant pro-phagocytic signal, and is required for anti-CD47 antibody-mediated phagocytosis.

60

EXPERIMENTAL PROCEDURES

Cell Lines and Human Samples MOLT4 and Daudi cell lines were a gift from the lab of Ronald Levy. 639V was obtained from the DSMZ. All other cell lines were obtained from the American Type Culture Association (ATCC). Normal human bone marrow mononuclear cells were purchased from AllCells Inc. (Emeryville, CA, USA). Normal peripheral blood and human cancer samples were obtained from patients at the Stanford Medical Center with informed consent according to IRB-approved protocols: AML, ALL, and NHL human samples from Stanford IRB# 76935, 6453, and 13500, bladder cancer samples from Stanford IRB #1512, glioblastoma samples from Stanford IRB# 9363, and ovarian cancer samples from Stanford IRB #13939. Normal fetal bladder and brain cells were purchased from ScienCell Research Laboratories (Carlsbad, CA, USA).

Flow Cytometry Analysis For analysis of normal peripheral blood cells, normal bone marrow cells, AML, CML, ALL, bladder cancer, ovarian cancer, and brain cancer, the following antibodies were used: CD34, CD38, CD90, CD45, CD31, CD3, CD4, CD7, CD11b, CD14, CD19, CD20, CD56, (Invitrogen, Carlsbad, CA and BD Biosciences, San Jose, CA, USA). Lineage negative (Lin-) was defined as CD3- CD19-CD20- for AML LSC and CD45-CD31- for GBM and bladder cancer CSC. Lin- was defined as CD3-CD4-CD7-CD8-CD11b-CD14-CD19-CD20-CD56-Glycophorin A- for NBM HSC, chronic phase CML GMP, CML CMP, and CML LSC. Analysis of CD47 expression was performed using an anti- human CD47 FITC antibody (clone B6H12.2, BD Biosciences). Analysis of human cell surface calreticulin expression was performed using mouse anti-human calreticulin conjugated to PE or FITC (clone FMC 75, Abcam, Cambridge, MA, USA).

In vitro phagocytosis assay Generation of human macrophages and in vitro phagocytosis assays were performed as previously described (108). Briefly, primary human samples or cell lines were incubated with 10µg/ml IgG1 isotype control (Ebiosciences, San Diego, CA, USA), 10µg/ml anti-CD47 antibody (clone B6H12.2, ATCC), 4µg/ml calreticulin blocking peptide (MBL International Coorporation, Woburn, MA), or 10ug/ml RAP (Fitzgerald Industries International, Concord, MA, USA). Cells were then analyzed by fluorescence microscopy to determine the phagocytic index (number of cells ingested per 100 macrophages). shRNA knockdown of Raji cells shRNA constructs targeting knockdown of human CD47 or a GAPD control packaged in the SMARTvector 2.0 lentiviral vector containing a turbo GFP reporter were purchased from Dharmacon,

61 Inc. (Lafayette, CO). Viral titers for each shRNA construct were greater than 10>8 TU/ml. Raji cells were transduced with these lentiviral constructs, analyzed and sorted for GFP expression, expanded, and sorted again for GFP expression for stable propagation of lentivirally-transduced cells. Knockdown of CD47 protein levels was assessed by flow cytometry with anti-CD47 antibody (B6H12.2) with fold knockdown calculated by reduction in MFI normalized over isotype control.

RESULTS

A Monoclonal Antibody Directed Against Mouse CD47 Enables Phagocytosis of Mouse AML and Does Not Deplete Normal HSC In Vivo CD47 is expressed at low levels on most normal tissues, including HSC. In order to investigate the viability of targeting CD47 as a therapeutic strategy, we utilized a mouse model of AML and a blocking anti-mouse CD47 monoclonal antibody (MIAP301) (78). A serially transplantable mouse model of AML was generated by transduction of 5-fluoruracil-treated wild type bone marrow with a retrovirus encoding HoxA9 and Meis1, as well as GFP (154). These mouse leukemia cells exhibited a 3-5 fold increase in CD47 surface expression compared to normal bone marrow (data not shown), similar to that observed with human AML. We first investigated the ability of the blocking anti-mouse CD47 monoclonal antibody to enable phagocytosis of the mouse leukemia cells and found that unlike an isotype-matched control, anti-mouse CD47 antibody enabled phagocytosis of GFP-positive leukemia cells by mouse macrophages in vitro (Figure 1A,B). Next, wild type mice were administered daily intraperitoneal injections of 200 micrograms of anti- mouse CD47 antibody for 14 days. This dose resulted in antibody coating of 100 percent of total bone marrow cells (Figure 2A). The mice appeared grossly normal and were sacrificed at the end of the treatment course. Analysis of the bone marrow showed no difference in overall cellularity (data not shown), percentage of Lin-Kit+Sca+ (KLS) cells (Figure 2B), or percentage of HSC (Figure 2C). Complete blood counts showed no evidence of anemia, but did indicate isolated neutropenia in the anti- CD47 antibody treated mice (Table 1A). Metabolic panels showed no serological evidence of hepatic or renal damage (Table 1B). Finally, in a pilot experiment, we found that treatment of mouse leukemia engrafted mice with the anti-mouse CD47 antibody resulted in a statistically significant increased survival compared to control IgG (Figure 1C). These results suggest that targeting CD47 with a blocking monoclonal antibody yields no unacceptable toxicity and is a viable therapeutic strategy.

62

Figure 1: An anti-mouse CD47 antibody enables phagocytosis of mouse leukemia in vitro and improves the mortality of leukemia-engrafted mice (A) GFP+ mouse AML cells were incubated with mouse bone marrow-derived macrophages in vitro in the presence of 10 mg/ml of rat IgG2a isotype control or anti-mouse CD47 antibody for 2 hours. Phagocytosis of GFP+ leukemia cells was observed by fluorescence microscopy (arrows). (B) Quantitative analysis of phagocytosis was determined by calculating the phagocytic index in triplicate assays. Anti-MsCD47 antibody enabled a statistically significant increase in phagocytosis of mouse leukemia cells compared to isotype control (p < 0.001). Error bars indicate the standard deviation of triplicate measurements. (C) 106 mouse acute myeloid leukemia cells were transplanted via retro-orbital sinus injection into sublethally-irradiated syngeneic wild-type mice resulting in peripheral blood and bone marrow engraftment by day 14. On day 21, these mice were administered daily 200 microgram intraperitoneal injections of either rat IgG control (n=7) or anti-mouse CD47 antibody (n=8). Survival of each cohort is reported in a Kaplan-Meier analysis, showing a statistically significant increased survival in mice treated with the antimsCD47 antibody.

63

Figure 2: Anti-mouse CD47 antibody does not deplete normal HSC in vivo (A) An anti-mouse CD47 antibody (MIAP301) was injected with daily 200µg intraperitoneal injections of anti- mouse CD47 antibody or rat IgG control. On day 2 or day 14, bulk peripheral blood (PB) or bone marrow (BM) cells, respectively, were stained with a rat IgG secondary antibody to assess antibody coating. (B and C) C57BL/6 wild-type mice were treated for 14 days with daily 200µg intraperitoneal injections of either anti-msCD47 or rat IgG control antibody. Bone marrow from these mice was aspirated pre- and post-treatment and indicated no effect of either treatment on the frequency of Lin_Kit+Sca+ (KLS) cells (C) or Lin_Kit+Sca+Flk2_CD34_ HSC (D) in the bone marrow. Representative flow cytometry plots are shown in (C). No differences in the percentage of HSC pre- and post-treatment were observed with either control IgG (p = 0.09) or anti-msCD47 (p = 0.81).

64

Table 1: Treatment of Wild Type Mice With a Blocking Anti-Mouse CD47 Monoclonal Antibody Causes Isolated Neutropenia (A-B) C57BL/6 wild type mice were treated for 14 days with daily 200 microgram intraperitoneal injections of either rat IgG control (n=5) or anti-msCD47 antibody (n=5). Peripheral blood was drawn post-treatment and hematology (A) and chemistry (B) lab analysis was performed. Results shown are the average of 5 mice in each treatment group. WBC=white blood cell, RBC=red blood cell, HGB=hemoglobin, HCT=hematocrit, PLTS=platelets, neuts=neutrophils, lymphs=lymphocytes, LDH=lactate dehydrogenase. Platelet frequencies were reported as adequate (+) or inadequate (-). p-values of IgG compared to anti-msCD47 treated mice are shown for each lab parameter.

Cell surface calreticulin is expressed on cancer, but not normal, stem and progenitor cells We hypothesized that the selective targeting of tumor cells but not normal cells was due to the presence of an unknown pro-phagocytic stimulus present on tumor cells but not on normal cells that could be unmasked after CD47 blockade. We investigated whether cell surface calreticulin could represent this pro-phagocytic signal. Cell surface calreticulin expression was determined on a variety of primary human cancer cells and their normal cell counterparts by flow cytometry. In hematologic malignancies, cell surface calreticulin was expressed on a greater percentage of bulk cells in acute myeloid leukemia (AML, average=23.9%), acute lymphocytic leukemia (ALL, 17.6%), chronic phase chronic myeloid leukemia (CML, 47.6%), and non-Hodgkin’s lymphoma (NHL, 18.3%) when compared to normal bone marrow (2.6%) and normal peripheral blood cells (2.6%) (Figure 3A). In solid tumors, cell surface calreticulin was also expressed on a greater percentage of bulk cells in ovarian cancer (average=20.5%), glioblastoma (31.7%), and bladder cancer (23.7%) when compared to normal fetal (0.3%), astrocytes, (2.5%) and normal fetal bladder cells (1.41%) (Figure 3B). In this analysis, annexin V-positive cells were excluded, indicating that calreticulin-positive cancer cells were not apoptotic. Given that primary human tumors are heterogeneous and contain a subpopulation of 65 tumor-initiating cells (reviewed in (62)), we next investigated whether cell surface calreticulin was present on the cancer stem cell (CSC) population of each tumor type in which the immunophenotype of functional CSC is known. In AML and chronic phase CML, cell surface calreticulin was expressed on CD34+CD38-CD90-Lin- AML (7, 13) and CD34+CD38-CD90+ chronic phase CML(16) leukemia stem cells (LSC) as well as downstream progenitor populations, while normal bone marrow hematopoietic stem and progenitor populations expressed minimal cell surface calreticulin (Figure 3C). For AML, similar levels of cell surface calreticulin expression were observed for LSC compared to other cellular subsets. In contrast, CML LSC expressed higher levels of cell surface calreticulin compared to downstream CMP and GMP populations. Cell surface calreticulin was also expressed on CSC of solid tumors including CD44+Lin- bladder CSC (110) and CD133+Lin- glioblastoma CSC (31, 32) (Figure 3D). We next determined whether there was a correlation between calreticulin (CRT) and CD47 expression in human tissues, postulating that a balance between pro- (CRT) and anti- (CD47) phagocytic signals may be maintained as a homeostatic mechanism. CRT and CD47 cell surface expression were profiled in a variety of human cancer cell lines, primary cancers, and normal cells. CD47 expression correlated with CRT expression in a variety of hematologic and solid tumor cell lines as well as in primary human AML, CML, and ALL patient samples (Figure 3E). Notably, normal cells expressed minimal levels of both CRT and CD47 (Figure 3E, top panels). In normal human bone marrow and fetal bladder, those cells that were CRT positive expressed higher levels of CD47 compared to CRT negative cellular counterparts (Figure 4). Thus, in both normal and cancer cells, there is a strong positive correlation between CRT and CD47 expression.

Increased CD47 expression on cancer cells protects them from calreticulin-mediated phagocytosis We observed increased expression of cell surface calreticulin and CD47 on human cancer cells leading us to hypothesize that increased CD47 expression protects these cells from calreticulin- mediated phagocytosis. To investigate this hypothesis, we performed in vitro phagocytosis assays on two different CRT-expressing cancer cell lines: one expressing high CD47 levels (Raji) and one deficient in CD47 expression (MOLM13). First, Raji cells, a Burkitt’s NHL cell line that expresses high levels of CD47 and calreticulin (Figure 5A and Figure 6), were incubated with human macrophages under conditions where CD47 expression was knocked down to various levels by lentiviral transduction of shRNAs (Figure 5A,B). Cell surface calreticulin expression was unaffected by shRNA-mediated CD47 knockdown (Figure 6). Upon incubation with human macrophages, Raji cells with approximately 2 fold knockdown of CD47 expression (shCD47-1 and shCD47-2) were robustly phagocytosed by human macrophages compared to the minimal phagocytosis observed in wild type and GAPD control transduced Raji cells (Figure 5C). Phagocytosis of shCD47-1 and shCD47-2 Raji cells was dependent on the calreticulin-LRP interaction as the observed phagocytosis was completely abrogated in the presence of a CRT blocking peptide (Figure 5C). In the second 66 experiment, MOLM13 cells, a human AML cell line that is deficient in CD47 expression (80) but expresses calreticulin (Figure 6), was incubated with human macrophages. As expected, MOLM13 cells were robustly phagocytosed at baseline, while phagocytosis was significantly reduced when the CRT- LRP interaction was blocked (Figure 5D). These findings demonstrate that overexpression of CD47 in cancers counterbalances calreticulin-mediated phagocytosis.

67 Figure 3: Cell Surface calreticulin is expressed on cancer, but not normal, stem and progenitor cells (A) Cell surface calreticulin expression was determined by flow cytometry on a panel of primary human patient samples from several hematologic cancer types and normal cell counterparts including normal bone marrow (NBM, n=9), normal peripheral blood (NPB, n=3), acute myeloid leukemia (AML, n=8), acute lymphoblastic leukemia (ALL, n=21), chronic myeloid leukemia (CML, n=13), and non-Hodgkin’s lymphoma (NHL, n=7). (B) The same analysis as in A was performed for solid tumors (glioblastoma, n=9; transitional cell bladder carcinoma, n=8; serous papillary ovarian carcinoma, n=9) and normal human fetal tissues (neurons, n=3; astrocytes, n=6, bladder cells, n=6). ESA+ urothelium was analyzed for normal fetal bladder. Primary human bladder cancer patient samples and samples that had been passaged once in mice were used for profiling. (C and D) Cell surface calreticulin expression was determined on normal stem and progenitor cells, lymphocytes, and cancer stem and progenitor cells. Each symbol represents a different patient sample. HSC=CD34+CD38-CD90+Lin-, LSC=CD34+CD38-CD90-Lin-, GMP=CD34+CD38+IL3rα+CD45RA+, CMP=CD34+CD38+IL3rα+CD45RA-. (D) Calreticulin expression did not differ between bulk and cancer stem cell populations for either bladder cancer (p=0.54) or glioblastoma (p=0.14). Bladder cancer CSC=CD44+Lin-(110), glioblastoma CSC=CD133+Lin-(31, 32). (E) The correlation between cell surface calreticulin and CD47 expression was determined for normal human and cancer cell lines (top left) and primary human normal and cancer samples (top right, bottom panels). Calreticulin and CD47 expression were calculated as mean fluorescence intensity normalized over isotype control and for cell size. Pearson correlation (r) and p-value is shown for each linear correlation. For top left panel: blue solid circle=HL60, blue open circle=Kasumi1, blue open inverted triangle=MOLM13, blue open diamond=KG-1, red triangle=Jurkat, red solid square=CCRF-CEM, red open square=CCRF-HSB2, red diamond=MOLT4, black star=Raji, black open diamond=SUDHL6, black open triangle=Daudi, black x=U937, green plus=639V, green open diamond=HT1197, green inverted triangle=UMUC3. *p<0.05, **p<0.005 (2-tailed student’s t-test). Annexin V positive cells were excluded in the analysis of all samples.

Figure 4: Live calreticulin positive cells from normal human tissue cells have higher levels of CD47 compared to calreticulin negative cells (A,B) Left panel: bulk normal human bone marrow cells (A) or normal human fetal bladder (ESA positive) urothelial cells (B) were profiled for calreticulin cell surface expression by flow cytometry. Right panel: calreticulin cell surface negative and positive cells were profiled for CD47 expression, demonstrating higher CD47 expression on calreticulin positive cells. Annexin V positive cells were excluded from the analysis of both bulk normal human bone marrow cells and fetal bladder cells. Data is representative of a pool of samples.

68

Calreticulin is the dominant pro-phagocytic signal on several human cancers and is required for anti- CD47 antibody-mediated phagocytosis In prior studies, we demonstrated that in several human cancers overexpression of CD47 contributes to evasion of macrophage phagocytosis, and furthermore that monoclonal antibody- mediated blockade of CD47 can enable phagocytosis and elimination of tumors in vitro and in mouse xenografts (80, 108, 110). Although anti-CD47 antibody is effective in enabling phagocytic elimination of tumors and is an attractive therapeutic anti-cancer agent, its potential off-target effects represent a potential concern given that CD47 is expressed at low levels ubiquitously in normal tissues (155). We show here that normal hematopoietic progenitor cells, which express CD47, were not phagocytosed when coated with anti-CD47 antibody. Additionally, administration of a blocking anti-mouse CD47 antibody to wild type mice caused minimal tissue toxicity. The lack of antibody toxicity is not likely exclusively due to overexpression of CD47 on cancer cells compared to normal counterparts given that both normal and cancer cells are coated with anti-CD47 antibody at therapeutic doses. Instead, it is likely due to the fact that in order for target cells to be phagocytosed upon blockade of an anti- phagocytic signal (CD47), the cells must also display a potent pro-phagocytic signal, which is absent on normal cells. Given the known role of CRT as a pro-phagocytic signal, its correlation with CD47 expression (Figure 3E), and its ability to be counterbalanced by CD47 (Figure 5), we investigated whether the expression of cell surface calreticulin on cancer but not normal cells could explain the selective targeting of tumor cells by a blocking anti-CD47 antibody. In vitro phagocytosis assays were performed incubating primary normal human cells or cancer cells with human macrophages in the presence of anti- CD47 antibody. CD47 was expressed on all normal and cancer cells profiled (Figure 3E and Figure 4,8), but expression of calreticulin was primarily restricted to normal cells (Figure 3A,B). No phagocytosis of cells from a variety of normal human tissue types was observed with anti-CD47 antibody (Figure 7B), while primary cancer cells from a variety of tumor types were robustly phagocytosed (Figure 7A,C). Significantly, anti-CD47 antibody-mediated phagocytosis of cancer cells was completely abrogated when cells were simultaneously incubated with peptides that inhibited the CRT-LRP interaction including a calreticulin blocking peptide and RAP, an inhibitor of LRP (125) (Figure 7C).

69

Figure 5: Increased CD47 Expression on Cancer Cells Protects Them from Calreticulin-Mediated Phagocytosis (A) CD47 protein expression was determined by flow cytometry on Raji cells transduced with lentiviruses encoding shRNA CD47-knockdown constructs (shCD47) or controls. (B) Relative CD47 expression levels were quantified by comparing MFI to wild type Raji cells. (C) Raji cell clones were incubated with human macrophages in media alone or CRT blocking peptide for 2 hours, after which phagocytosis was analyzed by fluorescence microscopy. Knockdown of CD47 in Raji cells (shCD47-1,-2) resulted in increased phagocytosis compared to untransduced Raji cells. No difference in phagocytosis was observed between untransduced and GAPD control- transduced Raji cells (p=0.45) Blockade of calreticulin on CD47-knockdown Raji cells completely abrogated phagocytosis. (D) MOLM-13 cells, a CD47-deficient human AML cell line, were incubated with human macrophages for two hours with the indicated peptides and monitored for phagocytosis as above. Significant phagocytosis was observed with IgG1 isotype control, while blockade of calreticulin or LRP reduced levels of phagocytosis (p=0.03 and p=0.01, respectively). *p<0.05, **p<0.005, ***p<0.0005.

70

Figure 6: Calreticulin expression is unaffected by CD47 shRNA knockdown in Raji cells Raji cells were transduced with lentiviral constructs encoding shRNA directed against CD47 (Raji shCD47-1, shCD47-2) or a GAPD control (Raji GAPD). Cell surface calreticulin expression was determined by flow cytometry and demonstrated no difference on wild type, untransduced Raji cells compared to Raji cells transduced with either GAPD control, shCD47-1, or shCD47-2 lentivirus. Cell surface calreticulin expression for MOLM13 cells is also shown.

Abrogation of anti-CD47 antibody-mediated phagocytosis was dose-dependent on calreticulin- LRP blockade (Figure 9). Notably, additional blockade of other pro-phagocytic signals was not required to abolish anti-CD47 antibody-mediated phagocytosis as cells incubated with anti-CD47 antibody under CRT-LRP blockade were phagocytosed at baseline control levels (Figure 7C). Blockade of the calreticulin-LRP interaction alone had no effect on phagocytosis when compared to IgG control (Figure 7C). These results demonstrate that anti-CD47 antibody-mediated phagocytosis requires the presence of cell surface calreticulin.

71

Figure 7: Cell Surface Calreticulin is the Dominant Pro-phagocytic Signal on Several Human Cancers and is Required for Anti-CD47 Antibody-Mediated Phagocytosis (A) Primary human AML cells were fluorescently-labeled with CFSE and incubated with human macrophages in the presence of the indicated antibodies/peptides for 2 hours, after which phagocytosis was analyzed by fluorescence microscopy. Arrows indicate phagocytosis. (B) Normal adult human bone marrow (NBM), adult peripheral blood (NPB), fetal neurons, fetal astrocytes, and fetal bladder cells were incubated with human macrophages in the presence of the indicated antibodies and monitored for phagocytosis as above. No difference in phagocytosis was detected between IgG1 isotype control and anti-CD47 antibody incubation (p=0.77). (C) Primary human cancer cells were incubated with human macrophages in the presence of the indicated antibodies/peptides for two hours and monitored for phagocytosis as above. Each data point represents a different patient sample. Compared to IgG1 isotype control, incubation with anti-CD47 antibody enabled phagocytosis of cancer cells (p<0.0001) while incubation with calreticulin blocking peptide (p=0.37) or RAP, an LRP inhibitor (p=0.67,) did not enable phagocytosis. In the presence of anti-CD47 antibody, incubation of cancer cells with either calreticulin blocking peptide or RAP completely abrogated anti-CD47 antibody-mediated phagocytosis (p=0.77 and p=0.16, respectively compared to IgG1 isotype control). *****p<0.00001.

72

Figure 8: CD47 is expressed on normal human cells (A-B) CD47 expression was determined by flow cytometry on normal human hematopoietic cells (A) and fetal tissue cells (B) demonstrating expression on all normal cells profiled. Flow cytometry plots are from a representative sample of each normal tissue cell type.

DISCUSSION

In this report, we determine that an anti-CD47 antibody selectively targets tumor cells with minimal toxicity to normal cells and identify calreticulin as a pro-phagocytic signal highly expressed on the surface of several human cancers, but minimally expressed on normal cell counterparts. The level of calreticulin correlated with expression of CD47, an anti-phagocytic signal, and increased CD47 expression on cancer cells was necessary for protection from calreticulin-mediated phagocytosis. Significantly, anti-CD47 antibody-mediated phagocytosis required the interaction of target cell- expressed calreticulin with its receptor LRP on phagocytic cells. These findings identify calreticulin as the dominant pro-phagocytic signal on several human cancers, provide an explanation for the selective targeting of tumor cells by an anti-CD47 antibody, and highlight a novel role for pro- and anti- phagocytic signals in the immune evasion of cancer.

73

Figure 9: Abrogation of anti-CD47 antibody-mediated phagocytosis is dose dependent on calreticulin blockade (A) Cell surface CRT and CD47 expression was determined by flow cytometry on Jurkat cells, a T cell leukemia cell line. (B) Jurkat cells were incubated with human macrophages in the presence of the indicated antibodies and blocking peptides and phagocytosis was determined by fluorescence microscopy. Anti-CD47 antibody was used at 10µg/ml. CRT blocking peptide concentrations are shown as µg/ml. Each condition was performed in duplicate.

Anti-CD47 antibody preferentially eliminates tumor cells with minimal toxicity due to differential expression of cell surface calreticulin We recently demonstrated that several cancers overexpress CD47 and that a blocking anti- CD47 monoclonal antibody can eliminate tumor cells in vitro and in vivo (108, 110). These pre-clinical findings provide a strong rationale for the use of an anti-CD47 antibody in the treatment of human cancers. However, given the broad low level expression of CD47 on both hematopoietic and most other normal tissues, antibody toxicity could be a significant barrier to clinical translation. To investigate this issue, we injected a blocking anti-mouse CD47 antibody into wild type mice at a dose that coated >98%

74 of bone marrow cells and observed no overt toxicity, with the exception of isolated neutropenia. Moreover, a recent report demonstrated that inhibition of CD47 with either an antibody or morpholino could confer radioprotective effects to normal tissues (156). We also demonstrate that despite low level CD47 expression, normal human cells from several tissue types are not phagocytosed by human macrophages when coated with anti-CD47 antibody (Figure 7B). We speculate that the selective phagocytosis of tumor cells is not simply dictated by CD47 expression level, but more importantly due to the presence of the pro-phagocytic signal calreticulin on tumor cells and not on normal cells. Several lines of evidence support this hypothesis. First, normal cells that express CD47 but not calreticulin are not phagocytosed with an anti-CD47 antibody despite antibody coating (Figure 7B). Second, tumor cells that express CD47 and calreticulin are phagocytosed when coated with anti-CD47 antibody (Figure 7A,C). Third, phagocytosis of tumor cells with anti-CD47 antibody is completely abrogated when the pro-phagocytic calreticulin-LRP interaction is blocked (Figure 7A,C). Collectively, these findings demonstrate that calreticulin expression is necessary for anti-CD47 antibody-mediated phagocytosis, and this expression is restricted to tumor cells. This study indicates that the therapeutic widow for anti-CD47 antibody therapy is not just due to target cell CD47 expression level, but that it is also dependent on the expression of pro-phagocytic calreticulin on the target cell. Based on our findings, we propose a model in which the overall contribution of pro (CRT)- and anti (CD47)-phagocytic signals determines whether normal and tumor cells are phagocytosed at steady state, and in the context of anti-CD47 antibody therapy (Figure 10). At steady state, tumor cells express calreticulin, but evade phagocytosis through overexpression of CD47 leading to dominance of the ”don’t eat me” anti-phagocytic signal (Figure 10A,B). Normal cells express low levels of CD47, and avoid phagocytosis because of a lack of CRT expression. In contrast, cells undergoing DNA damage or apoptosis express calreticulin on their cell surface (125, 148), which is dominant over low CD47 expression and leads to phagocytosis. In the context of anti-CD47 antibody therapy, the “don’t eat me” signal (CD47) is blocked, unmasking the “eat me” signal (CRT) on tumor cells, leading to phagocytosis (Figure 10C,D). In contrast, blockade of CD47 on normal cells does not lead to phagocytosis since the “eat me” signal (CRT) is absent. In this way, selective phagocytosis of tumor cells induced by anti-CD47 antibody is dependent on the expression levels of pro- and anti- phagocytic signals. Although calreticulin appears to be primarily expressed on the surface of apoptotic or malignant cells, prior reports detected surface calreticulin on some human normal cells including activated peripheral blood T cells (157) and circulating neutrophils (158). In addition, a blocking monoclonal anti-CD47 antibody has been shown to enhance phagocytosis of apoptotic neutrophils (159, 160). Interestingly, in our mouse toxicity studies, administration of a blocking anti-mouse CD47 antibody led to selective depletion of neutrophils, while other hematopoietic cells were unaffected (108). Similar to tumor cells, this selective neutropenic toxicity is likely due to unmasking of

75 calreticulin expression on neutrophils when the “don’t eat me” signal (CD47) is blocked by anti-CD47 antibody.

Figure 10: Model for the integration of pro (CRT)- and anti (CD47)-phagocytic signals on normal and tumor cells at steady state and during anti-CD47 antibody therapy (A,B) At steady state, tumor, normal, and damaged cells express varying levels of cell surface CD47 and CRT, and it is the integration of both signals that determines whether the target cell will be phagocytosed. Tumor cells express CRT, but also higher levels of CD47 that delivers a dominant negative phagocytic signal (minus sign), leading to evasion of phagocytosis. In contrast, normal cells express lower levels of CD47, but do not express CRT, and thus no phagocytosis occurs. Lastly, damaged or apoptotic cells exhibit high levels of CRT expression, and this positive phagocytic signal (plus sign) dominates over the low CD47 expression level, leading to phagocytosis (dashed arrow). (C,D) During anti-CD47 antibody therapy, the negative phagocytic stimulus (CD47) is blocked. In tumor cells, this unmasks the positive phagocytic signal (CRT), leading to phagocytosis. In contrast, normal cells are not phagocytosed since the positive phagocytic stimulus (CRT) is absent.

76

Calreticulin is the dominant pro-phagocytic signal on several human cancers We demonstrate that several human cancers, including both hematopoietic and solid tumor malignancies, broadly express the pro-phagocytic signal calreticulin. Expression of known physiologic pro-phagocytic signals has previously been identified in several cancers including phosphatidylserine (161-165) and annexin-1 (reviewed in (153)). However, most of these studies were not performed on primary human patient samples as in this study. Additionally, ligand expression appears to be mixed across tumor types (153) with the functional role of these ligands in cancer unknown. We demonstrate that steady state expression of cell surface calreticulin represents a consistent and novel pro-phagocytic signal present on many human cancers. In contrast to other potential pro-phagocytic signals, calreticulin is the dominant “eat me” signal supported by the key observation that anti-CD47 antibody- mediated phagocytosis was completely abolished by calreticulin-LRP blockade alone (Figure 7C). Based on these findings, a complete survey of human tumors for cell surface calreticulin expression will be required to determine whether the regulation of the CD47-CRT phagocytic axis is a universal trait of cancers. One key question is raised by these studies: Why do cancers express cell surface calreticulin, a pro-phagocytic signal? We have demonstrated that certain cancers evade the innate immune system by upregulating anti-phagocytic signals, specifically CD47 (80, 108, 110). From a cancer selection viewpoint, one might expect cancers to simultaneously downregulate pro-phagocytic signals to further increase their ability to evade macrophage phagocytosis. We propose two possible explanations. First, expression of cell surface calreticulin may be an unwanted consequence of cellular stress, whereby CD47 expression is upregulated to compensate and enable phagocytic evasion. In normal physiology, cell surface calreticulin is induced on cells undergoing DNA damage (125, 148, 149), marking these damaged cells for homeostatic phagocytosis. It is possible that a small fraction of these cells may upregulate CD47 and are able to compensate and avoid phagocytic clearance, which allows these damaged cells to survive and acquire additional mutations, eventually transforming into malignant cells. Several lines of evidence support this hypothesis. First, CD47 and CRT expression are highly correlated in several human tumors (Figure 3E). Second, the small percentage of live cells that are calreticulin positive in some normal human tissue types (bone marrow and bladder) express higher CD47 levels than their calreticulin negative counterparts (Figure 4). Third, this increase in CD47 expression appears to protect against calreticulin-mediated phagocytosis as knockdown of CD47 to 50% of wild type levels enabled calreticulin-dependent phagocytosis compared to the phagocytosis- resistant wild type cell line (Figure 5). In a second hypothesis, expression of cell surface calreticulin may confer some unknown pro- tumorigenic phenotype to cancer cells that is independent of phagocytosis. One possibility is that cell surface calreticulin may confer a more invasive and angiogenic phenotype as its ligand, LRP, is expressed on several vascular cell types (reviewed in (166)). In two reports, overexpression of calreticulin or calreticulin fragments in tumor cell lines enhanced in vitro migration and invasion (167, 77 168), however, other studies have reported opposing roles for calreticulin (169, 170). In all of these studies the function of cell surface calreticulin was not distinguished from its intracellular roles. Other possible tumorigenic roles include cell adhesion (171) and immune escape through reduction of MHC class I antigen presentation (172). The potential role of cell surface calreticulin in these contexts has not yet been explored. In summary, we have identified cell surface calreticulin as the dominant pro-phagocytic signal on several human cancers that is absent on normal cell counterparts and is required for anti-CD47 antibody-mediated phagocytosis. These findings have a major impact on the development of an anti- CD47 antibody therapy for human malignancies and highlight the novel dynamic relationship between pro- and anti-phagocytic signals in the pathogenesis of human cancer.

78

CHAPTER 5

Improving an anti-CD47 antibody therapy: a combination antibody approach with rituximab in non-Hodgkin lymphoma

Portions of this chapter were published in the following article: Chao MP*, Alizadeh AA*, Tang CZ, Myklebust JH, Varghese B,Gill S, Jan M, Cha AC, Chan CK, Tan BT, Park CY, Zhao F, Kohrt HE, Malumbres R, Briones J, Gascoyne RD, Lossos IS, Levy R, Weissman IL, Majeti R. Anti-CD47 antibody synergizes with rituximab to promote phagocytosis and eradicate non-Hodgkin lymphoma. Cell. 142: 699-713.

79 SUMMARY

Monoclonal antibodies are standard therapeutics for several cancers including the anti-CD20 antibody rituximab for B cell non-Hodgkin lymphoma (NHL). Rituximab and other antibodies are not curative, and must be combined with cytotoxic chemotherapy for clinical benefit. Here we report the eradication of human NHL solely with a monoclonal antibody therapy combining rituximab with a blocking anti- CD47 antibody. We identified increased expression of CD47 on human NHL cells, and determined that higher CD47 expression independently predicted adverse clinical outcomes in multiple NHL subtypes. Blocking anti-CD47 antibodies preferentially enabled phagocytosis of NHL cells and synergized with rituximab. Treatment of human NHL-engrafted mice with anti-CD47 antibody reduced lymphoma burden and improved survival, while combination treatment with rituximab led to elimination of lymphoma and cure. These antibodies synergized through a novel mechanism combining Fc receptor (FcR)-dependent and FcR-independent stimulation of phagocytosis that might be applicable to many other cancers.

INTRODUCTION

Emerging evidence has demonstrated that monoclonal antibodies (mAbs) either alone or in combination are an effective modality for cancer treatment (92). Although therapies combining a monoclonal antibody with chemotherapeutic agents are effective in several human cancers, antibodies alone are not curative. Antibodies effective against cancer are believed to function by several mechanisms including: antibody-dependent cellular cytotoxicity (ADCC), stimulation of complement- dependent cytotoxicity (CDC), inhibition of signal transduction, or direct induction of apoptosis (173). Non-Hodgkin lymphoma (NHL) is the fifth most common cancer in the United States consisting of indolent and aggressive subtypes with a five-year overall survival ranging from 25-75% (174). The anti-CD20 antibody, rituximab (Rituxan), is part of standard therapy for many CD20- positive B cell lymphomas, and significantly improves long-term survival in combination with conventional chemotherapy (173). As a single agent or in combination with chemotherapy, rituximab is not curative in the majority of B cell NHL patients and rituximab resistance has been observed (reviewed in (173). Numerous lines of evidence indicate that rituximab acts at least in part by engaging Fc receptors (FcRs) on immune effector cells, such as NK cells and macrophages, and stimulating effector functions such as ADCC (102, 175). Although resistance has been reported to occur through several mechanisms including epitope loss, modulation of complement activity, or diminished ADCC (176), the exact mechanisms have not been fully elucidated, and there has been limited development of novel agents that can overcome this resistance.

80

Immune effector cells, including NK cells and phagocytes, are critical to the efficacy of many anti-cancer antibodies. Phagocytic cells, including macrophages and dendritic cells, express signal regulatory protein alpha (SIRPα), which binds CD47, a widely expressed transmembrane protein (73). CD47-mediated activation of SIRPα initiates a signal transduction cascade resulting in inhibition of phagocytosis (74-78). In identifying a role for CD47 in cancer pathogenesis, we previously demonstrated that forced expression of mouse CD47 on a human leukemia cell line facilitated tumor engraftment in immunodeficient mice through the evasion of phagocytosis (80). We further demonstrated that this mechanism could be targeted therapeutically in human acute myeloid leukemia (AML) with a blocking anti-CD47 antibody that enabled phagocytosis and eliminated AML stem cells (108). Based on this antibody mechanism, we hypothesized that the combination of a blocking anti- CD47 antibody with a second FcR-activating antibody would both prevent an inhibitory signal and deliver a positive stimulus resulting in the synergistic phagocytosis and elimination of target cells. Here we report that human NHL cells overexpress CD47 and that antibody blockade of the CD47-SIRPα interaction enables phagocytosis of NHL cells. Given that rituximab binds FcRs, activating immune effector functions including phagocytosis (102), we tested the antibody synergy hypothesis by investigating the combination of a blocking anti-CD47 mAb with rituximab against human NHL.

EXPERIMENTAL PROCEDURES

Cell lines A Burkitt’s lymphoma cell line (Raji) and a DLBCL cell line (SUDHL4) were obtained from the American Type Culture Collection or generated in the lab. The NHL17* cell line was generated from a patient with DLBCL by culturing bulk cells in vitro with IMDM supplemented with 10% human AB serum for 1.5 months.

Human Samples Normal human peripheral blood and human NHL samples were obtained from discarded specimens or from patients at the Stanford University Medical Center, Stanford, CA with informed consent, according to an IRB-approved protocol (Stanford IRB# 13500) or with informed consent from the Norwegian Radium Hospital, Oslo, Norway according to a Regional Ethic Committee (REK)-approved protocol (REK# 2.2007.2949). Normal tonsils for germinal center B cell analysis were obtained from discarded tonsillectomy specimens from consented pediatric patients at Stanford University Medical Center according to an IRB-approved protocol (Stanford IRB# 13500).

81 Flow Cytometry Analysis For analysis of normal peripheral blood cells, germinal center B cells, and primary NHL cells, the following antibodies were used: CD19 APC, CD20 PECy5, CD20 PE, CD3 QD605, CD3 PB, CD10 PECy7, CD10 APC, CD45 PB, CD5 PECy7, CD38 FITC (Invitrogen, Carlsbad, CA and BD Biosciences, San Jose, CA). Analysis of CD47 expression was performed with an anti-human CD47 FITC antibody (clone B6H12.2, BD Biosciences). Cell staining and flow cytometry analysis was performed as previously described (108). All primary NHL patient samples were analyzed from cryopreserved specimens.

Evaluation of Prognostic Value of CD47 in NHL Gene expression and clinical data were analyzed for 8 previously described cohorts of adult NHL patients, including 4 studies of patients with DLBCL, 3 with B-CLL, and 1 with MCL (177-183). Table 1 details the main microarray datasets analyzed herein, including the training, test, and validation DLBCL cohorts, as well as the relationship of each dataset to figures presented here. The clinical end points analyzed included overall (OS), progression free (PFS), and event-free survival (EFS), with events defined as the interval between study enrollment and need for therapy or death from any cause, with data censored for patients who did not have an event at the last follow-up visit.

Therapeutic Antibodies Rituximab (anti-CD20, human IgG1) was obtained from the Stanford University Medical Center, mouse anti-human CD20, IgG2a from Beckman Coulter (Miami, FL), and anti-CD47 antibody BRIC126, IgG2b from AbD Serotec (Raleigh, NC). Other anti-CD47 antibodies were used as in (108). All in vivo antibody experiments were performed using the anti-CD47 B6H12.2 antibody.

In Vitro Isobologram Studies In vitro phagocytosis assays were conducted with NHL cells incubated with anti-CD47 antibody (B6H12.2), anti-CD20 IgG2a, or rituximab either alone or in combination at concentrations from 1µg/ml to 10µg/ml. The concentration of each antibody required to produce a defined single-agent effect (phagocytic index) was determined for each cell type. Concentrations of the two antibodies combined to achieve this same phagocytic index were then determined and plotted on an isobologram and the combination index determined. The combination index (CI) was calculated from the formula CI=(d1/D1) + (d2/D2), whereby d1=dose of drug 1 in combination to achieve the phagocytic index, d2=dose of drug 2 in combination to achieve the phagocytic index, D1=dose of drug 1 alone to achieve the phagocytic index, D2=dose of drug 2 alone to achieve the phagocytic index. A CI of less than, equal to, and greater than 1 indicates synergy, additivity, and antagonism, respectively.

Annexin V Apoptosis Assays 82

Assays were performed as previously described (108). Percentage dead cells (% Annexin V+ and/or 7- AAD+) were reported and normalized over values from incubation of cells with media alone.

Preparation of Mouse and Human Macrophages Mouse and human macrophages were prepared as previously described (108). BALB/c or Fcγr-/- (Taconic, Hudson, NY) mouse bone marrow mononuclear cells were used to differentiate into mouse macrophages.

In Vitro Phagocytosis Assays Human primary NHL cells or NHL cell lines were CFSE-labeled and incubated with either mouse or human macrophages (in a 4:1 target:effector cell ratio) in the presence of 10µg/ml IgG1 isotype control, anti-CD45 IgG1, anti-CD47 (clones B6H12.2, BRIC126, or 2D3), rituximab, or anti-SIRPα for 2 hours as previously described (108). Cells were then analyzed by fluorescence microscopy to determine the phagocytic index (number of cells ingested per 100 macrophages). Statistical analysis using Student’s t-test was performed with GraphPad Prism (San Diego, CA).

Preparation of mouse NK cells Purified mouse NK cells were generated as previously described (184). In brief, Balb/C mouse spleens were dissociated into a single cell suspension, followed by red blood cell lysis and anti-16/32 blockade. Cells were then stained with anti-DX5 microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) and positively selected by manual MACS column enrichment (Miltenyi Biotec). The positive fraction was then sorted for DX5+CD3-PI- cells on a fluorescence-activated cell sorter Aria (BD) to >95% purity. Purified NK cells were cultured at 3 x 106 cells/mL in RPMI with 10% FCS, 2mM L-glutamine, 100U/mL penicillin, 100ug/mL streptomycin (all from Invitrogen), and 5ug/mL 2-mercaptoethanol (Sigma-Aldrich, St Louis, MO) with 1500U/mL recombinant human interleukin (IL)-2 (Chiron, Emeryville, CA) for 48 hours.

Preparation of human NK cells Human buffy coats were obtained from the Stanford Blood Center under an IRB-approved research protocol. Peripheral blood mononuclear cells were separated using Ficoll-gradient centrifugation, blocked with anti-CD16/32 and then stained with anti-CD56 microbeads (Miltenyi Biotec), followed by manual MACS column enrichment. The positive fraction was then sorted for CD56+CD3-PI- cells on a fluorescence-activated cell sorter Aria to >95% purity. NK cells were then cultured overnight at 3 x 106 cells/mL in RPMI media with 10% FCS, 2mM L-glutamine, 100U/mL penicillin, 100ug/mL streptomycin, and 5ug/ml 2-mercaptoethanol (Sigma-Aldrich, St Louis, MO) supplemented with 1000U/mL recombinant human interleukin (IL)-2 (Chiron, Emeryville, CA).

83 51Chromium release assay Raji and SUDHL4 cells were resuspended at 1 x 106 cells in 400µL complete RPMI (RPMI, FCS, L- glutamine, penicillin, streptomycin, 2-ME, as described above) and labeled with 300uCi 51Cr (Perkin- Elmer) for 3 hours. Cells were then washed twice with complete RPMI, resuspended at 1 x 105 cells/ml, and incubated for 30 minutes with anti-CD47 antibody (B6H12.2) alone and/or with rituximab, as well as with B6H F(ab')2 fragments alone and/or with rituximab, or isotype control, generated as described above. Target cells were then plated in triplicate and incubated with mouse or human NK cells at a o 17.5:1 effector-target ratios for 4 hours at 37 C and 5% CO2. At the completion of the assay, supernatant was collected and quantified using a gamma counter (Cobra/AII, Packard Bioscience, Meridien, CT). Specific 51Cr release was calculated using the equation: % specific lysis = 100 x (test release) - (spontaneous release) / (maximal release) - (spontaneous release) (185).

Complement Dependent Cytotoxicity Assay Experiments were performed as previously described (186). Briefly, 1x106/ml SUDHL4 or NHL17* cells were incubated with 50% media and 50% human or BALB/c-derived mouse complement- preserved serum (Innovative Research, Novi, Michigan) for 1 hour at 37°C in the presence of IgG1 isotype control, anti-CD47 antibody (B6H12.2), rituximab, or the combination of anti-CD47 antibody and rituximab. Percent cell lysis was calculated by percentage of propidium iodide positive cells normalized to a baseline with cells incubated with 50% complement and media alone.

In vivo depletion of immune effector cells Depletion of immune effector cells in SCID mice were performed as previously described for macrophages (108), NK cells (187), and complement (188). Detection of macrophage depletion was determined by F4/80+ cell frequency, NK cell depletion by DX5+ cell frequency, andcomplement depletion by presence of C3 in mouse sera by ELISA against anti-mouse complement component C3 (clone: RmC11H9) (Cedarlane Laboratories, Ontario, Canada).

Preparation of F(ab')2 fragments

Rituximab (human IgG1 isotype) F(ab')2 fragments were produced by pepsin cleavage of IgG using a

F(ab')2 preparation (Thermo Scientific, Rockford, IL) following the manufacturer’s recommendations. Similarly, anti-human CD47 (mouse IgG1 isotype) F(ab')2 fragments were produced by ficin cleavage using a mouse IgG1 Fab and F(ab’)2 Preparation Kit (Thermo Scientific). After digestion, the reaction mixture was applied to a protein A column to remove Fc fragments and undigested IgG. The generation of F(ab')2 fragments was confirmed by SDS-PAGE and the binding of fragments to CD20 or CD47 on the surface of Raji cells was determined by flow cytometry (Figure S7). Absence of the Fc portion on the fragments was further confirmed by flow cytometry using Alexa Fluor 488-conjugated Protein A beads (Invitrogen) (Figure 14). 84

Generation of Luciferase-Positive Cell Lines and Luciferase Imaging Analysis The dual reporter gene L2G (Luc-2A-eGFP) containing a modified firefly luciferase gene joined to eGFP at the 3’ end (189) was obtained as a gift from Dr. Christopher Contag (Stanford University). The L2G construct was digested with NheI and NotI (New England BioLabs, Ipswich, MA), ligated into the pCDH-CMV-MCS lentiviral cDNA expression vector (System Biosciences, Mountain View, CA) with a rapid DNA ligation kit (Roche Applied Science, Indianapolis, IN), and transformed into OneShot OmniMax 2-T1 competent cells (Invitrogen). Lentiviral production of the L2G- pCDH-CMV- MCS construct was performed using standard VSV.G and packaging vectors in 293TF cells (Invitrogen). Transduction of L2G containing lentivirus was performed in Raji and SUDHL4 cells with polybrene (Sigma, St Louis, MO) and centrifuged for 2 hours at 2000 x g and 37°C. Transduced cells were resuspended in standard growth medium and GFP positive cells were sorted to purity after one week and reanalyzed three weeks later for stable luciferase-GFP expression. Mice were imaged at the Stanford Center for Innovation in In-Vivo Imaging (Stanford, CA). Mice were anesthetized and given an intraperitoneal injection of firefly D-luciferin (Biosynth, Staad, Switzerland) at a dose of 375mg luciferin per 1kg of mouse body weight (190). Luciferase imaging of mice was then performed using the Xenogen In Vivo Imaging System (IVIS) Spectrum (Caliper LifeSciences, Hopkinton, MA) with a 1-60 second exposure time, medium binning, and 16 f/stop. Multiple images were taken until luciferase signal intensity reached a plateau and image parameters were kept identical across duration of experiments. Luciferase image analysis was performed using Living Image 3.0 (Caliper LifeSciences). Luciferase units were quantified in average radiance (photons/second/cm2/sr).

In Vivo Pre-Coating Engraftment Assay Assay were performed as previously described (108). Luciferase positive Raji cells or human NHL cells were incubated with 30 µg/mL of IgG1 isotype control, anti-CD45 IgG1, anti-CD47 IgG1 (B6H12.2) antibody, or rituximab. and transplanted intravenously into SCID mice or sublethally-irradiated (200 rads) NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ (NSG) mice via intravenous injection into the retro-orbital plexus. Mice were analyzed two weeks later for engraftment by luciferase imaging or human lymphoma bone marrow chimerism. p-values were generated from the Fisher’s exact test. All experiments involving mice were performed according to Stanford University institutional animal guidelines.

In Vivo Antibody Treatment in a Disseminated Lymphoma Xenograft Model 1.5x106 luciferase-labeled Raji cells were injected intravenously into the retro-orbital sinus of 6-10 week old SCID or NSG mice. Those mice with luciferase-positive lymphoma were randomized into four treatment groups with daily intraperitoneal injections of 200µg mouse IgG control, anti-CD47,

85 rituximab, or 200µg anti-CD47 + 200µg rituximab for three weeks. Antibody treatment was then stopped and mice were followed for survival analysis. A complete remission (CR) was defined as no evidence of lymphoma by bioluminescence at the end of treatment. A relapse was defined as evidence of lymphoma by bioluminescence after the end of treatment in a mouse with a prior CR.

In Vivo Antibody Treatment in a Localized Lymphoma Xenograft Model 3x106 luciferase-labeled Raji cells were injected subcutaneously into the right flank of 6-10 week old NSG mice. Those mice with luciferase-positive lymphoma were randomized into four treatment groups treated with daily intraperitoneal injections of 400µg mouse IgG control, 400µg anti-CD47, 200µg rituximab, or 400µg anti-CD47 + 200µg rituximab for four weeks. Tumor volume was measured every 3-4 days using the formula (length*width)/2. Antibody treatment was then stopped and mice were followed for survival analysis.

In Vivo Antibody Treatment of Primary NHL Engrafted Mice 2x106 NHL cells were transplanted intravenously via retro-orbital plexus into sublethally-irradiated (200 rads) NSG mice. Two weeks later, bone marrow was aspirated from these mice and those mice with evidence of human lymphoma engraftment (hCD45+CD19/CD10+ bone marrow cells) were then treated with daily intraperitoneal injections of either 200µg IgG, 200µg anti-CD47 antibody (clone B6H12.2), 200µg rituximab, or 200µg anti-CD47 antibody + 200µg rituximab. After fourteen days of treatment, bone marrow cells from these mice were aspirated and antibody treatment was stopped. Mice were then followed for survival analysis. A complete remission (CR) was defined as no evidence of lymphoma in the BM at end of treatment. A relapse was defined as evidence of lymphoma in the BM after end of treatment in a mouse with a prior CR.

In vivo depletion of immune effector cells Depletion of immune effector cells in SCID mice were performed as previously described for macrophages (108), NK cells (187), and complement (188). Detection of macrophage depletion was determined by F4/80+ cell frequency, NK cell depletion by DX5+ cell frequency, andcomplement depletion by presence of C3 in mouse sera by ELISA against anti-mouse complement component C3 (clone: RmC11H9) (Cedarlane Laboratories, Ontario, Canada).

In vivo phagocytosis In vivo phagocytosis was performed as previously described (108) analyzing mice transplanted with GFP+Raji cells into adult NSG mice. Mice were given a single dose of antibody and analyzed 4 hours later for in vivo phagocytosis.

86

RESULTS

CD47 Expression is Increased on NHL Cells Compared to Normal B Cells We examined CD47 protein expression on primary human NHL samples and normal B cells by flow cytometry. Compared to both normal peripheral blood B cells and normal germinal center B cells, CD47 was more highly expressed on a large subset of primary patient samples from multiple B cell NHL subtypes (Figure 1A and 2A), including Diffuse Large B Cell Lymphoma (DLBCL), B cell Chronic Lymphocytic Leukemia (B-CLL), Mantle Cell Lymphoma (MCL), Follicular Lymphoma (FL), Marginal Zone Lymphoma (MZL) and pre-B acute lymphoblastic leukemia (pre-B ALL). Across NHL subtypes, we found differing levels of CD47 expression that also varied within each NHL subtype (Figure 1B).

Increased CD47 Expression Correlates with a Worse Clinical Prognosis and Adverse Molecular Features in Multiple NHL Subtypes Having previously shown a correlation between CD47 mRNA and protein expression (108), we assessed CD47 mRNA expression across NHL subtypes for associations with morphologic and molecular subgroups using gene expression data from previously described patient cohorts (Table 1; (179, 181-183)), and investigated the prognostic implications of increased CD47 expression in disease outcome. Higher CD47 expression was associated with adverse prognosis in DLBCL, B-CLL, and MCL (Figure 1C-E). In patients with DLBCL, whether treated with or without rituximab-based combination chemotherapy (Figures 1C and 2B), higher CD47 expression was significantly associated with risk of death. This increased risk was largely due to disease progression (Figure 2C), a finding validated in an independent cohort of patients using quantitative RT-PCR on fixed archival specimens (Figure 2D). We next investigated whether increased CD47 expression was associated with known adverse molecular features in NHL, focusing on DLBCL, CLL, and MCL. In DLBCL, global gene expression analysis identified two prognostically distinct subgroups based on “cell-of-origin” of the B cell malignant clone: normal germinal center B cells (GCB-like), which is associated with a favorable clinical outcome, or activated blood memory B cells (ABC-like), which is associated with a poor clinical outcome (181, 191, 192). CD47 expression was significantly higher in ABC-like DLBCL (Figures 1F and 2E). CD47 expression was not found to have independent prognostic value within GCB and ABC subtypes, suggesting a strong association with the cell-of-origin classification of DLBCL. Higher CD47 expression was also associated with unmutated immunoglobulin heavy chain variable regions (IgVH) in CLL (Figures 1G and 2F) and significantly correlated with the proliferative index in MCL (Figure 1H), both known adverse prognostic factors (182, 193, 194). In multivariate analyses, CD47 expression remained prognostic of disease progression independent from the international prognostic index in DLBCL (174), and two major prognostic factors in CLL: IgVH

87 mutation status and ZAP-70 status (Figure 2G). Within the small MCL cohort, a multivariate model did not find independent prognostic value for CD47 when considering the proliferation index (data not shown).

Figure 1: CD47 Expression is Increased on NHL Cells Compared to Normal B Cells, Confers a Worse Clinical Prognosis, and Correlates with Adverse Molecular Features in Multiple NHL Subtypes (A) CD47 expression on normal peripheral blood (PB) B cells (CD19+), normal germinal center (GC) B cells (CD3-CD5-CD20+CD10+CD38+), and NHL B cells (CD19+) was determined by flow cytometry, and

88

mean fluorescence intensity was normalized for cell size. Each point represents a different patient sample: DLBCL=2, CLL=15, MCL=4, FL=6, MZL=2, and pre-B ALL=1 (****p<0.0001). Normalized mean expression (and range) for each population were: normal PB B cells 420.9 (267.3-654.0), normal GC B cells 482.5 (441.1- 519.9), and NHL 888.5 (270.1-1553). (B) CD47 expression across NHL subtypes including DLBCL (DL, n=15), MCL (n=34), FL (n=28), and B-CLL (n=14) was determined as in A. Normalized mean expression (and range) for each population were: DL 725.7 (261.2 – 1344), MCL 1055 (444.2-2196), FL 825.1 (283.6-1546), CLL 713.6 (432.8-1086), (*p<0.05). (B to D) Prognostic influence of CD47 mRNA expression is shown on overall (C and E) and event-free (D) survival of patients with DLBCL, B-CLL, and MCL. For DLBCL and CLL, stratification into low and high CD47 expression groups was based on an optimal threshold determined by microarray analysis; this cut point was internally validated for both DLBCL and CLL, and also externally validated in an independent DLBCL cohort. For MCL, stratification relative to the median was employed as an optimal cut point could not be defined. Significance measures are based on log-likelihood estimates of the p-value, when treating CD47 expression as a continuous variable, with corresponding dichotomous indices also provided in Table 1. (F to H) CD47 mRNA expression is shown in relation to cell-of-origin classification for DLBCL (F), immunoglobulin heavy chain mutation status (IgVH) for CLL (G), and proliferation index for MCL (H). Analyses for C-H employed publicly available datasets for NHL patients (Table 1). NGC=normal germinal center, ABC=activated B cell-like, GCB=germinal center B cell-like.

89

Table 1: Summary of Expression Datasets of NHL Patients Characteristics of gene expression datasets of NHL patients used for CD47 expression and prognostic studies. LN=lymph node, LLMPP=leukemia/lymphoma molecular profiling project, IPI=International Prognostic Index, COO=cell of origin, PFS=progression-free survival, PBMC=peripheral blood mononuclear cells, EFS=event-free survival.

Blocking Antibodies Against CD47 Enable Phagocytosis of NHL Cells by Macrophages and Synergize with Rituximab in Vitro Given the increased expression of CD47 on NHL cells, and its known function as an inhibitor of phagocytosis, we tested the ability of mouse anti-human CD47 antibodies to enable phagocytosis of human NHL cell lines, primary NHL cells, and normal peripheral blood (NPB) cells by human macrophages in vitro. Incubation of NHL cells with human macrophages in the presence of IgG1 isotype control or anti-CD45 IgG1 antibody did not result in significant phagocytosis; however, two different blocking anti-CD47 antibodies (B6H12.2 and BRIC126) enabled phagocytosis of NHL cells, but not NPB cells (Figure 3A,B). Prior to conducting in vivo mouse xenotransplantation experiments, we repeated the in vitro phagocytosis assays described above with mouse macrophages. Incubation of NHL cells with mouse macrophages in the presence of IgG1 isotype control or anti-CD45 IgG1 antibody did not result in significant phagocytosis; however, phagocytosis of NHL cells was observed with blocking antibodies to CD47 (B6H12.2 and BRIC126), while no phagocytosis was observed with a non-blocking antibody (2D3) (Figure 3B). In addition, disruption of the CD47-SIRPα interaction with an anti-mouse SIRPα antibody also resulted in significant phagocytosis (Figure 3B).

90

Figure 2: CD47 is an Independent Adverse Prognostic Factor and is Associated with Adverse Molecular Features of DLBCL and CLL (A) Clinical features of primary human NHL samples used in vitro and/or in vivo are shown. Stage is reported according to Ann Arbor staging for all NHL except CLL, where Rai score was reported. n/a=not available or not applicable; PB=peripheral blood; LN=lymph node; CVP=cyclophosphamide, vincristine, prednisone; R- CHOP=rituximab, cyclophosphamide, adriamycin, vincristine, prednisone. (B) Patients were stratified into low CD47 and high CD47 expression groups as in Figure 1C with overall survival (OS) of patients obtained in (181). CD47 mRNA expression was validated by Taqman RT-PCR from data obtained in (182) with corresponding progression-free survival (PFS) (C) and OS (D) shown of low and high CD47 expression groups. P-values are 91 reported for CD47 considered as a dichotomous variable (Table S1). (E) CD47 expression correlates with “cell-of- origin” subtype in DLBCL. CD47 mRNA expression is shown of DLBCL patients treated with either CHOP or R- CHOP from (179). (F) Histograms are shown for CD47 gene expression levels between patients with mutated IgVH or unmutated IgVH. Each data point represents a single patient. Data is presented for 2 independent gene expression datasets in B-CLL (left, (183); right, (178)). n=sample size. (G) CD47 expression as a continuous variable independently predicted overall survival in a multivariate analysis for DLBCL when international prognostic index was considered as a covariate, and independently predicted event-free survival in CLL when ZAP-70 expression status and IgVH mutation status were considered as covariates. Analysis was performed in data obtained from (179) for DLBCL and (183) for CLL. ABC=Activated B Cell-Like, GCB=Germinal Center B cell- Like, Type 3=subgroup not expressing characteristic of ABC or GCB. HR=hazard ratio, P=p-value.

Given that CD47 expression on primary NHL samples was variable, we investigated whether the level of CD47 expression correlated with the degree of anti-CD47 antibody-mediated phagocytosis by two independent methods. First, lentiviral shRNA vectors were used to knockdown expression of CD47 in Raji cells, which express high levels of CD47. Several Raji clones were generated with a range of CD47-knockdown, verified by flow cytometry (Figure 4A,B). Those clones with a greater than 50% reduction in CD47 expression (shCD47-1 and shCD47-2) demonstrated a significant reduction in anti- CD47 antibody-mediated phagocytosis (Figure 4C). In the second approach, a statistical analysis demonstrated a positive correlation between CD47 expression level and degree of anti-CD47 antibody- mediated phagocytosis with both mouse and human macrophages effector cells (Figure 4D). It has been reported that immobilized or cross-linked antibodies against CD47 induce apoptosis of primary human B-CLL cells, as well as several malignant lymphoid cell lines (79, 89-91). These studies raise the alternative possibility that anti-CD47 antibodies directly induce apoptosis of NHL cells that are then recognized by macrophages and phagocytosed. To test the ability of a blocking anti-CD47 antibody to induce apoptosis, NHL cells were incubated with anti-CD47 antibody, or controls in the absence of macrophages. Anti-CD47 antibody did not induce apoptosis of NHL cells when incubated in suspension for either two hours (Figure 2B, right) or eight hours (Figure 4E,F). Incubation of NHL cells with immobilized anti-CD47 antibody resulted in increased apoptosis compared to controls (Figure S2G,H), consistent with prior reports (79). Since phagocytosis of NHL cells occurs in the presence of soluble anti-CD47 mAbs, it is unlikely that these mAbs induce apoptosis of NHL cells that are then secondarily phagocytosed. Next, we investigated the ability of a blocking anti-CD47 mAb to synergize with the anti- CD20 mAb rituximab in the phagocytosis and elimination of NHL cells. We examined CD20 expression on NHL cells and found no difference between normal B cells and NHL cells (Figure 4I,J). Incubation of NHL cells with rituximab in the presence of mouse or human macrophages resulted in significant phagocytosis (Figure 2D,E). We then tested the synergy hypothesis by isobologram analysis, which is a standard method for measuring synergy between two therapeutic agents (195, 196). Using Raji, SUDHL4, and NHL17 cells, which express varying levels of both CD47 and CD20 (Figure 4K), anti-CD47 antibody synergized with rituximab or anti-human CD20 (mouse IgG2a) antibody, as indicated by concave isobologram curves and combination indices less than 1 (Figure 3C). In a second approach, in vitro phagocytosis assays were conducted with primary NHL cells incubated with either 92 anti-CD47 antibody or rituximab alone, or both in combination at half of the single agent dose. NHL cells exhibited a significant increase in phagocytosis when incubated with the combination compared to either antibody alone when using mouse (Figure 3D) or human (Figure 3E) macrophage effectors. No phagocytosis of NPB cells was observed with either antibody alone or in combination with human macrophages (Figure 3E).

Ex Vivo Coating of NHL Cells with an Anti-CD47 Antibody Inhibits Tumor Engraftment Next, the ability of blocking anti-CD47 antibody to eliminate NHL in vivo either alone or in combination with rituximab was explored by two separate treatment strategies. First, the effect of ex vivo anti-CD47 antibody coating on the engraftment of human NHL cells was tested. Luciferase- expressing Raji and SUDHL4 cell lines were pre-coated ex vivo with anti-CD47, IgG1 isotype control, or anti-CD45 antibody and transplanted intravenously into SCID mice. Ex vivo coating with anti-CD47 antibody prevented engraftment of both cell lines as demonstrated by bioluminescent imaging (Figure 5A-F). Ex vivo coating of Raji cells with rituximab also inhibited engraftment when transplanted into SCID mice (Figure 6). In addition to these cell lines, we identified a primary NHL patient specimen that engrafted in NSG mice in the bone marrow upon intravenous transplantation (Figure 10A,B). As with the cell lines, ex vivo coating of these primary cells with anti-CD47 antibody, but not controls (Figure 5G), resulted in complete inhibition of bone marrow engraftment (Figure 5H).

Combination Therapy with Anti-CD47 Antibody and Rituximab Eliminates Lymphoma in Both Disseminated and Localized Human NHL Xenotransplant Models In the second treatment strategy, mice were first engrafted with NHL and then administered single antibody or combination antibody therapy. Human NHL is marked by lymphadenopathy, extra- nodal masses, and often dissemination of disease into the circulation. To best study this disease, we established novel disseminated and localized NHL xenotransplantation models in NSG mice. These mice are deficient in T, B, and NK cells (197), allowing us to determine if phagocyte effector cells, including macrophages and neutrophils, were able to mediate a therapeutic effect. In the disseminated model, luciferase-expressing Raji cells were transplanted intravenously into adult NSG mice. Two weeks later, these mice were administered daily injections of either control mouse IgG, anti-CD47 antibody, rituximab, or anti-CD47 antibody and rituximab. Anti-CD47 antibody treatment decreased the lymphoma burden in these mice (Figure 7A,B), and significantly prolonged survival compared to control IgG, although all mice eventually died (Figure 7C and Table 2). Similar results were seen with rituximab, and were not statistically different compared to anti-CD47 antibody (Figure 7A-C and Table 2). In contrast, combination therapy with anti-CD47 antibody and rituximab eliminated lymphoma in 60% of mice as indicated by long-term survival (Figure 7C) and the absence of luciferase-positive lymphoma (data not shown) more than four months after the end of treatment. In humans, rituximab efficacy is thought to be primarily mediated by both macrophages and NK cells (102, 198). Given that

93 NSG mice lack NK cells, we conducted a similar experiment in NK cell-containing SCID mice, and observed similar therapeutic responses as in NSG mice (Figure 8A,B).

94

Figure 3: Blocking Antibodies Against CD47 Enable Phagocytosis of NHL Cells by Macrophages and Synergize with Rituximab in Vitro (A) CFSE-labeled NHL cells were incubated with human macrophages and the indicated antibodies and examined by immunofluorescence microscopy to detect phagocytosis (arrows). Photomicrographs from a representative NHL sample are shown. (B) Phagocytic indices of primary human NHL cells (blue), normal peripheral blood (NPB) cells (red), and NHL cell lines (purple, orange, and green) were determined using human (left) and mouse (middle) macrophages. Antibody-induced apoptosis (right panel) was tested by incubating NHL cells with the indicated antibodies or staurosporine without macrophages, and assessing the percentage of apoptotic and dead cells (% annexin V and/or PI positive). (C) Synergistic phagocytosis by anti-CD47 antibody (B6H12.2) and anti-CD20 mAbs was examined by isobologram analysis and determination of combination indices (CI). CIobs indicates observed results, and the dashed line indicates the expected results if antibodies were additive. (D,E) Synergy between anti-CD47 antibody and rituximab in the phagocytosis of NHL and NPB cells was assessed by determining the phagocytic index when incubated with a combination of both antibodies compared to either antibody alone at twice the dose, with either mouse (D) or human (E) macrophages. NHL17*: cell line derived from primary sample NHL17. ***p<0.001, ****p<0.0001, *****p<0.00001. Figure 2B p-values represent comparison against IgG1 isotype control antibody.

In the localized NHL model, luciferase-expressing Raji cells were transplanted subcutaneously in the right flank of NSG mice. Once tumors were palpable (approximately 2 weeks), mice were treated with either control mouse IgG, anti-CD47 antibody, rituximab, or the combination of anti-CD47 antibody and rituximab. Mice treated with anti-CD47 antibody or rituximab demonstrated a decreased rate of lymphoma growth compared to control IgG-treated mice as measured by both luciferase signal and tumor volume (Figure 7D-F). In contrast, mice treated with the combination of anti-CD47 antibody and rituximab demonstrated complete elimination of lymphoma, with 86% of treated mice having no measurable mass or luciferase-positive lymphoma 4 weeks after the end of treatment (Figures 7D-F and 8C-E). Mice treated with either IgG, anti-CD47 antibody, or rituximab eventually had to be sacrificed due to enlarged tumor growth in compliance with institutional animal protocols. In contrast, all mice treated with the combination of anti-CD47 antibody and rituximab showed no evidence of tumor growth, remained relapse free, and were alive at over 197 days after tumor engraftment. Out of six mice achieving a complete remission, five remained relapse free while one mouse died of non-tumor related causes (Figure 7E). For both disseminated and localized xenograft models, phenotypic expression of CD47 and CD20 in transplanted Raji cells did not differ from Raji cells in culture (Figure 8F).

95

Figure 4: Correlation of CD47 Expression with Anti-CD47 Antibody-Mediated Phagocytosis, Anti-CD47 Antibody-Mediated Effect on Apoptosis, and CD20/CD47 Expression on Primary NHL Cells and Cell Lines (A) CD47 protein expression was determined by flow cytometry for Raji cells transduced with lentiviruses encoding shRNA CD47-knockdown constructs (shCD47) or controls. (B) Relative CD47 expression levels were quantified by comparing MFI to wild type Raji cells. (C) Wild type Raji cells and transduced CD47-knockdown clones were incubated with mouse macrophages for two hours in the presence of the indicated anti-CD47 antibody concentrations and the phagocytic index was determined. The fold change is reported as the phagocytic index of anti-CD47 antibody divided by phagocytic index of the IgG1 isotype control. Each condition was performed in

96

triplicate. *p<0.05, *****p<0.00001. (D) A positive correlation between CD47 protein expression and phagocytic index with anti-CD47 antibody is shown for primary NHL samples (solid symbols). Normal peripheral blood samples are shown (open symbols). Symbols refer to sample notation from Figure 1A. Correlations were calculated by the Pearson correlation coefficient. (E) Representative flow cytometry plots illustrating Annexin V/7-AAD staining in NHL7 cells treated for eight hours with media, the indicated soluble antibodies, or staurosporine as a positive control. No apoptosis was detected with soluble antibodies. (F) NHL cells were treated for 8 hours with the indicated soluble antibodies and the percentage dead cells (%Annexin V+ and/or 7-AAD+) for each antibody normalized to cells treated with media alone is reported. (G and H) NHL cells were also incubated with plate-immobilized antibodies for 2 hours (G) and 8 hours (H) and analyzed for apoptosis as described above. In panels F and H, Raji and SUDHL4 cells were incubated for 18 hours. Symbol annotations for NHL samples are those used in Figure 1A. (I) Relative CD20 protein expression of normal peripheral blood B cells (CD19+) and NHL B cells (CD19+) was determined by flow cytometry and calculated in a similar manner and displayed as in Figure 1A. No difference in CD20 expression was observed between normal B cells and NHL cells (p=0.35). (J) Scatter plot is shown of CD20 expression in lymphoma cells from individual lymphoma patient samples determined by flow cytometry, DL: diffuse large B cell lymphoma (n=15), MCL: mantle cell lymphoma (n=34), FL: follicular lymphoma (n=28) and CLL: chronic lymphocytic leukemia (n=14). (K) CD47 and CD20 expression is shown for two NHL cell lines and one NHL sample expanded in culture (NHL17*). *p<0.02 P-values were determined compared to IgG1 isotype. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001

Combination Therapy with Anti-CD47 Antibody and Rituximab Eliminates Lymphoma in Primary Human NHL Xenotransplant Mouse Models Although the majority of pre-clinical evaluation of therapeutics for lymphoma has been performed using cell lines, such cell lines are not ideal for studying human lymphoma as they cannot recapitulate the heterogeneity of the primary disease and have been subjected to artificial in vitro selection pressures. We report here novel mouse xenograft models for NHL in which intravenous transplantation of cells from a DLBCL patient (NHL7/SUNHL7) and a FL patient (NHL31/SUNHL31) give rise to robust lymphoma engraftment in the bone marrow and/or peripheral blood (Figures 2A and 10A). Primary DLBCL cells were transplanted intravenously into sublethally-irradiated NSG mice. Two weeks later, these mice were treated with daily injections of either mouse IgG, anti-CD47 antibody, rituximab, or anti-CD47 antibody combined with rituximab for 14 days. Treatment with anti- CD47 antibody either alone or in combination with rituximab eliminated human lymphoma in the bone marrow, while treatment with rituximab resulted in a reduction of disease in 60% of mice (Figure 9A,B). Treatment was then discontinued, and all mice were monitored for survival. Mice treated with either anti-CD47 antibody or rituximab alone had a significantly longer survival compared to mice treated with control IgG, but all eventually died secondary to disease (Figure 9C). Most significantly, 8 out of 9 mice (89%) administered combination antibody treatment were cured of lymphoma, as indicated by long-term disease-free survival more than four months after the end of treatment (Figure 9C and Table 3) with no detectable lymphoma in the bone marrow (data not shown). In a second primary NHL engraftment model, primary FL cells were transplanted intravenously in a similar manner. CD20+CD10+ lymphoma engraftment in the peripheral blood and bone marrow was detected after 8 weeks. At this time, mice were treated with a single intraperitoneal injection of either 100µg control IgG, 100µg anti-CD47 antibody, 200µg rituximab, or the combination of 100µg anti-CD47 antibody and 200µg rituximab, and then followed for disease progression. A single dose of anti-CD47 antibody alone or in combination with rituximab eliminated lymphoma both in the peripheral blood (Figure 9D)

97 and bone marrow (Figure 9E). In contrast, a single dose of rituximab enabled a partial reduction in tumor burden that rebounded back to baseline levels in the peripheral blood with no tumor reduction observed in the bone marrow. Notably, treatment with anti-CD47 antibody, rituximab, or the combination of the two did not cause a significant change in the numbers of mouse CD45+ cells in the peripheral blood or bone marrow of untransplanted mice (Figure 10E).

Figure 5: Ex Vivo Coating of NHL Cells with an Anti-CD47 Antibody Inhibits Tumor Engraftment (A-F) Luciferase-expressing Raji (A) and SUDHL4 (C) cells were incubated ex vivo with 30 µg/ml of indicated antibody and assessed for antibody coating by flow cytometry. Coated cells were injected intravenously into adult

98

SCID mice, and two weeks later mice transplanted with Raji (B) and SUDHL4 (D) were subject to bioluminescent imaging (1-IgG1 isotype control, 2-anti-CD45, 3 and 4-anti-CD47, 5-luciferase negative control). Bioluminescence for Raji (E) and SUDHL4 (F) engrafted mice was quantified (n=3 per antibody condition). No tumor engraftment was observed in mice transplanted with anti-CD47-coated cells compared to IgG-coated cells (p<0.05) for both Raji and SUDHL4, as assessed by bioluminescent imaging. Data are represented as mean +/- SD. (G) Bulk lymphoma cells from a human NHL patient were incubated ex vivo with 30 µg/ml of indicated antibody and analyzed by flow cytometry for coating. (H) Coated cells were injected intravenously into sublethally-irradiated adult NSG mice, and two weeks later the percentage of human lymphoma in the bone marrow was measured. Compared to IgG1 isotype control, anti-CD47 antibody pre-treatment inhibited engraftment of NHL cells (p<0.0001) while anti-CD45 coated cells engrafted similarly to controls (p=0.54). All p-values were determined using Fisher’s exact test.

Figure 6: Ex vivo Coating of NHL Cells with Rituximab Inhibits Tumor Engraftment (A) Luciferase-expressing Raji cells were incubated ex vivo with 30 µg/ml of human IgG1 isotype control antibody or rituximab and assessed for coating by flow cytometry. (B) Coated cells were injected intravenously into adult SCID mice, and two weeks later were subject to bioluminescent imaging. (C) Bioluminescence was quantified (n=2 per antibody condition). No tumor engraftment was observed in mice transplanted with rituximab-coated cells compared to IgG1 isotype-coated cells (p=0.04), as assessed by bioluminescent imaging.

99

Figure 7: Combination Therapy with anti-CD47 Antibody and Rituximab Eliminates Lymphoma in both Disseminated and Localized Human NHL Xenotransplant Mouse Models (A) NSG mice transplanted intravenously with luciferase-expressing Raji cells were treated two weeks later daily for 21 days with either 100µg IgG, 100µg anti-CD47 antibody, 200µg rituximab, or 100µg anti-CD47 antibody and 200µg rituximab (n=8 per treatment group). Luciferase imaging of representative mice from pre- and 10 days post- treatment are shown. (B) Quantified bioluminescence was determined and averaged for all mice in each treatment group. (C) Kaplan-Meier survival analysis was performed (Table 2). p-values compare IgG control to combination

100

antibody treatment or anti-CD47 antibody/rituximab single antibody to combination. Arrows indicate start (day 14) and stop (day 35) of treatment. (D) Luciferase-expressing Raji cells were transplanted subcutaneously in the flank of NSG mice. When palpable tumors (~0.1cm3) formed, treatment began with daily injections of either 400µg IgG, 400µg anti-CD47 antibody, 200µg rituximab, or 400µg anti-CD47 antibody and 200µg rituximab. Luciferase imaging of representative mice from each treatment group is shown before (day 0) and during (day 14) treatment. (E) Quantified bioluminescence was determined and averaged for all mice in each treatment group (n=7). (F) Tumor volume was measured with average volume shown. p-values were derived from a two-way ANOVA and compared to IgG treatment. *p<0.05, ***p<0.001, ****p<0.0001. Complete remission was defined as the number of mice with no evidence of tumor at the indicated date. Relapse was defined as the number of mice achieving a complete remission that later developed recurrence of tumor growth. For panel E, one mouse that achieved a complete remission died of non-tumor related causes. Data presented in B, E, and F are mean values +/- SD.

Figure 8: Combination Therapy with Anti-CD47 Antibody and Rituximab Eliminates NHL in a Disseminated Lymphoma Xenotransplant Model in SCID and NSG Mice (A) Raji cells were injected intravenously via retro-orbital plexus into adult SCID mice. Mice were then treated with 200µg daily intraperitoneal injections of indicated antibody (n=7 per group) and analyzed for survival by Kaplan-Meier analysis. Arrows represent start (day 7) and stop (day 28) of treatment. (B) Statistical results for Kaplan-Meier analysis are shown with p-values generated by the Mantel-Cox test and compared to control IgG treatment. (C) Luciferase-labeled Raji cells were engrafted subcutaneously into the flank of NSG mice and then 101 treated with daily injections of 400µg mouse IgG, 400µg anti-CD47 antibody, 200µg rituximab, or 400µg anti- CD47 antibody and 200µg rituximab when tumors were palpable. Gross pictures of representative mice from each treatment group are shown at day 25 of antibody treatment. (D and E) Luciferase signal (D) and tumor volume (E) is shown for each individual mouse treated with the combination of anti-CD47 antibody and rituximab. Arrow represents end of antibody treatment (day 35). (F) CD47 and CD20 expression were determined by flow cytometry on Raji cells in culture (primary), engrafted intravenously (IV), and subcutaneously (SQ) into NSG mice. Mean fluorescence intensity was normalized over isotype.

To assess the ability of these two primary NHL xenotransplant models to model the primary disease, we compared histological sections of the primary NHL specimen and the transplanted tumor. Similar DLBCL and FL morphology was observed for NHL7 and NHL31, respectively (Figure 10B). We next determined whether the percentage of macrophages infiltrating the tumor differed between the primary patient and the xenografted tumor. For NHL31, the percentage of infiltrating human macrophages (CD68+) in the primary lymph node was similar to the percentage of infiltrating mouse macrophages (F4/80+) in bone marrow of transplanted mice (Figure S5C). Analyzing infiltrating macrophage frequency by flow cytometry, no difference was observed between the primary specimen and xenograft for either NHL7 or NHL31 (Figure 10D).

102

Figure 9: Combination Therapy with Anti-CD47 Antibody and Rituximab Eliminates Lymphoma in Primary Human NHL Xenotransplant Mouse Models (A,B) Sublethally-irradiated NSG mice were transplanted intravenously with cells from a primary DLBCL patient (NHL7). Two weeks later, bone marrow was aspirated and analyzed for engraftment of human CD45+CD19+ lymphoma. Engrafted mice were treated with daily injections of either 100µg control IgG, 100µg anti-CD47 antibody, 200µg rituximab, or 100µg anti-CD47 antibody and 200µg rituximab for 14 days. Flow cytometry of human lymphoma engraftment in the bone marrow of two representative mice is shown pre- and 14 days post- treatment in (A). Engraftment data from all mice is indicted in (B). **p<0.01, comparing pre- and post-treatment values for each respective antibody treatment. (C) Kaplan-Meier survival analysis (Table S3) of mice from each antibody treatment cohort is shown (n≥10 per antibody group), with p-values calculated comparing control IgG to combination antibody treatment or anti-CD47 antibody/rituximab single antibody to combination treatment.

103 Arrows indicate start (day14) and stop (day 28) of treatment. (D-E) Mice engrafted intravenously with a primary FL patient sample (NHL31) were treated with a single dose of control IgG, anti-CD47 antibody, rituximab, or anti- CD47 antibody and rituximab with disease burden followed in the peripheral blood and bone marrow. Compared to IgG control and rituximab, anti-CD47 antibody alone or in combination with rituximab eliminated tumor burden in the peripheral blood (p=0.04, 2-way ANOVA), and bone marrow (p<0.001, t-test). (E) Lyphoma engraftment in the bone marrow was determined 14 days post-treatment. Each antibody treatment group consisted of 3 mice. For mice reported in panels D and E, human lymphoma chimerism was between 5-25% in the peripheral blood and bone marrow.

Synergy Between Anti-CD47 Antibody and Rituximab Does Not Occur Through NK Cells or Complement Rituximab has been demonstrated to eliminate malignant cells via apoptosis, NK cell-mediated ADCC, and CDC (reviewed in (199)). However, it is not known whether anti-CD47 antibody enables ADCC or CDC in addition to phagocytosis. Therefore, we investigated whether anti-CD47 antibody alone could induce anti-tumor effects by macrophage-independent mechanisms, and whether synergy with rituximab could occur through these modalities. First, to investigate possible synergy in direct apoptosis, NHL cells were incubated with either anti-CD47 antibody/rituximab alone or in combination without macrophages, and cell death was measured. No synergistic induction of apoptosis was observed when NHL cells were incubated with soluble (Figures 11A and 12A) or immobilized (Figure 12B,C) antibodies. Furthermore, cross-linking of soluble anti-CD47 antibody alone or in combination with rituximab by macrophages did not induce increased apoptosis of non-phagocytosed NHL cells compared to IgG1 isotype control, while rituximab induced a slight increase in apoptosis (Figure 12D). No synergistic apoptosis was observed in this context (Figure 12D). The small increase in apoptosis upon antibody treatment was not FcR-dependent given that results were similar with macrophages lacking the Fcγ subunit (200) (Figure 12D). Second, we investigated whether NK cells could play a role in mediating tumor elimination by anti-CD47 antibody alone or in synergy with rituximab. A prior report observed increased NK-cell cytotoxicity of cancer cell lines with an anti-CD47 antibody in vitro, however the mechanism of cell killing was not elucidated (201). We first determined whether human or mouse NK cells expressed SIRPα, which might suggest that anti-CD47 antibody could induce cytotoxicity by blockade of the CD47-SIRPα interaction, similar to macrophage-mediated phagocytosis. Both human NK cells, CD3-CD56+CD7+ (202), as well as mouse NK cells, CD3-DX5+, expressed minimal to no SIRPα (Figure 11B). Next, the ability of anti-CD47 antibody to induce NK cell-mediated ADCC through FcRs or by CD47-SIRPα blockade was investigated. Utilizing human NK cells as effectors, anti-CD47 antibody did not induce increased ADCC of Raji or SUDHL4 cells compared to IgG1 isotype control (Figure 11C). While rituximab did enable ADCC of these two NHL cell lines, no synergistic effect between anti-CD47 antibody and rituximab was observed (p=0.77, Figure 11C). Since anti-CD47 antibody (B6H12.2) is a mouse IgG1 isotype, we repeated these assays with mouse NK cells. Anti-CD47 antibody caused increased ADCC of these two NHL cell lines compared to isotype control, while rituximab induced ADCC to a lesser degree (Figure 11D). To

104

investigate whether anti-CD47 antibody-mediated ADCC was Fc-dependent, we generated a F(ab’)2 fragment of the anti-CD47 antibody (Figure 14B-I). The F(ab’)2 fragment did not enable ADCC, indicating that the increased ADCC was likely mediated through FcRs (Figure 11D). The combination of anti-CD47 antibody or F(ab’)2 fragment with rituximab did not induce a statistically significant increase in ADCC compared to single agent therapy, indicating no synergistic effect. The difference in the effect of anti-CD47 antibody and rituximab with human compared to mouse NK cells is likely due to species-specific interactions of antibody Fc and effector FcR. Third, we investigated the role of complement in anti-CD47 antibody-mediated cytotoxicity. Using either human (Figure 11E) or mouse (Figure 11F) complement, anti-CD47 antibody did not induce CDC of either an NHL cell line or a primary NHL sample, while rituximab did induce significant CDC against both of these samples. Moreover, the combination of anti-CD47 antibody and rituximab did not induce increased CDC compared to rituximab alone. Fourth, we investigated the relative contribution of major components of the innate immune system (macrophages, NK cells, and complement) in mediating the therapeutic effects of anti-CD47 antibody and rituximab in vivo. Luciferase-labeled Raji cells were engrafted intravenously into SCID mice, which have functional macrophages, NK cells, and complement. Mice were then separated into 4 cohorts receiving selective depletion of either macrophages by clodronate, NK cells by anti-asialoGM1 antibody, complement by cobra venom factor, or a vehicle control. These four cohorts were then treated with combination anti-CD47 antibody and rituximab therapy for 3 days, and tumor burden was measured by bioluminescent imaging pre- and post-treatment. Compared to vehicle control, NK cell and complement depletion had no effect on tumor elimination by combination antibody therapy (Figure 12E). In contrast, macrophage depletion significantly abrogated the therapeutic effect, indicating that macrophages, and not NK cells or complement, are required for combination anti-CD47 antibody and rituximab-mediated elimination of NHL in vivo.

105

Figure 10: Clinical Characteristics and Morphology of Primary NHL Patient Samples (SUNHL7 and SUNHL31) Used for Xenotransplantation Studies (A) Detailed treatment and immunophenotypic information on SUNHL7/NHL7 and SUNHL31/NHL31. R-CHOP: rituximab, cyclophosphamide, doxorubicin, vincristine, prednisone. R-EPOCH: rituximab, etoposide, doxorubicin, vincristine, cyclophosphamide, prednisone. (B) Hematoxylin and eosin (H&E) stain of sections from a cutaneous lymphoma of DLBCL patient NHL7/SUNHL7 and from a lymph node of FL patient NHL31/SUNHL31. H&E stains of mice xenotransplants with these patient samples are also shown. Cellular morphology was similar between the primary and transplanted tumor for both NHL7 and NHL31. (C) Immunohistochemistry (left) is

106

shown of a primary lymph node specimen from patient NHL31, and immunofluorescence (right) of mouse bone marrow engrafted with NHL31, demonstrating similar macrophage frequencies. (D) The percentage of human CD68+ macrophages and mouse F4/80+ macrophages was determined by flow cytometry from the primary patient samples NHL7 (peripheral blood) and NHL31 (lymph node) and from subsequent xenografted tumors in the peripheral blood (NHL7) and bone marrow (NHL31), respectively. Triplicate mice were analyzed for both xenografts. (E) NSG mice were treated for 14 days with daily injections of either control IgG, anti-CD47 antibody, rituximab, or the combination of anti-CD47 and rituximab in the same manner as in Figure 5A-C. The number of mouse CD45+ cells was then analyzed post-treatment in the bone marrow with each treatment done in duplicate.

Anti-CD47 Antibody Synergizes with Rituximab Through FcR-Independent and FcR-Dependent Mechanisms We hypothesize that the observed synergy between an anti-CD47 antibody and rituximab occurs through the combination of two separate mechanisms for stimulating phagocytosis: (1) FcR- independent through blockade of an inhibitory phagocytic signal by anti-CD47 antibody, and (2) FcR- dependent through delivery of a positive phagocytic signal by rituximab. We utilized four independent methods to investigate this hypothesis. First, NHL cells were incubated with anti-mouse SIRPα antibody, rituximab, or a combination of both in the presence of mouse macrophages. Synergistic phagocytosis was observed with the combination of anti-SIRPα antibody and rituximab by isobologram analysis (Figure 13A), and with a large panel of primary NHL samples (Figure 13B). Second, mouse macrophages lacking the Fcγ receptor, thereby incapable of enabling FcR-dependent phagocytosis (200), but still expressing SIRPα (Figure 14A), were utilized as effector cells for phagocytosis of NHL cells incubated with either anti-CD47 antibody, anti-SIRPα antibody, rituximab, or anti-CD47/anti- SIRPα antibody in combination with rituximab. As predicted, anti-CD47 antibody and anti-SIRPα antibody, but not rituximab, enabled phagocytosis of NHL cells, and no synergistic phagocytosis was observed (Figure 13C). Third, F(ab’)2 fragments of both anti-CD47 antibody and rituximab were generated (Figure 14B-I) and utilized in phagocytosis assays with NHL cells and wild type macrophages. Anti-CD47 F(ab’)2 antibody synergized with rituximab as demonstrated by isobologram analysis (Figure 13D). Additionally, anti-CD47 F(ab’)2, but not rituximab F(ab’)2, enabled phagocytosis of NHL cells (Figure 13E). Consistent with our proposed mechanism, synergistic phagocytosis was observed with the combination of either full length anti-CD47 or anti-CD47 F(ab’)2 with full length rituximab, but not with any combination involving rituximab F(ab’)2 (Figure 13E). Fourth, synergistic phagocytosis was investigated in vivo using GFP+ Raji cells engrafted in NSG mice. As single agents, anti-CD47 antibody and rituximab enabled phagocytosis of Raji cells engrafted in the liver as evidenced by an increased percentage of mouse macrophages containing phagocytosed GFP+Raji cells, enumerated by detection of hCD45-GFP+F4/80+ cells by flow cytometry (Figure 13F). Most significantly, combination anti-CD47 antibody and rituximab treatment enabled significantly increased phagocytosis compared to either single agent demonstrating that synergistic phagocytosis occurred in vivo (Figure 13F).

107

Figure 11: Synergy Between Anti-CD47 Antibody and Rituximab Does Not Occur Through NK Cells or Complement (A) NHL cells were incubated with the indicated soluble antibodies for 2 hours and the percentage of dead cells was calculated (% Annexin V+ and/or 7-AAD+). No statistically significant difference in % dead cells was observed with the combination of anti-CD47 antibody and rituximab compared to either anti-CD47 antibody alone (p=0.24) or rituximab alone (p=0.95). (B) SIRPα expression is shown for both human and mouse NK cells as

108

determined by flow cytometry. (C,D) Chromium release assays measuring ADCC were performed in triplicate with human (C) and mouse (D) at an effector:target ratio of 17.5:1 and percent specific lysis is reported. Antibodies were incubated at 10µg/ml except anti-CD47 full length or F(ab’)2 antibody+rituximab (5µg/ml). (E) CDC assay with human complement was performed in duplicate. Compared to IgG1 isotype control, anti-CD47 antibody did not enable CDC (p>0.2), while rituximab did (p<0.001) by 2-way ANOVA for both SUDHL4 and NHL17*. Combination treatment with anti-CD47 antibody and rituximab did not enable greater levels of CDC compared to rituximab (p=0.78). (F) CDC assay with mouse complement was performed in duplicate. Compared to IgG1 isotype control, anti-CD47 antibody did not enable CDC (p>0.25) while rituximab did (p=0.03, p=0.08, respectively) for both SUDHL4 and NHL17*. No difference in CDC between CD47 antibody+rituximab and rituximab alone was observed (p>0.13) for both SUDHL4 and NHL17*. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. NHL17*=Primary NHL17 cells expanded in culture.

DISCUSSION

In this report, we demonstrate a novel mechanism of synergy between monoclonal antibodies in cancer therapy leading to cures in the absence of chemotherapy. Specifically, we utilized a blocking anti-CD47 antibody in combination with the clinical therapeutic anti-CD20 antibody rituximab to eradicate human NHL. Previously, we demonstrated that CD47 regulates AML pathogenesis by inhibiting phagocytosis, and that a blocking anti-CD47 antibody enabled phagocytosis and elimination of AML stem cells (80, 108). Here we report that CD47 is also highly expressed on human NHL, and that increased expression is associated both with poor clinical outcomes and with known adverse molecular features in multiple NHL subtypes. As with AML, blocking anti-CD47 antibodies enabled phagocytosis and elimination of tumor cells from diverse NHL subtypes. Most importantly, we found that the synergistic combination of anti-CD47 and rituximab cured 60-90% of NHL-engrafted mice in three xenotransplantation models, while a single dose of anti-CD47 antibody therapy alone eliminated tumor burden in a fourth model. Finally, we demonstrate that the mechanism of synergy involves FcR- independent enabling of phagocytosis by anti-CD47 antibody and FcR-dependent stimulation of phagocytosis by rituximab. These results establish the basis for a combination antibody therapy for NHL, and possibly other human cancers.

109

Figure 12: Therapeutic Effects of Anti-CD47 Antibody and Rituximab in Inducing Apoptosis, ADCC, or CDC, Related to Figure 6 (A) NHL cells were incubated with the indicated soluble antibodies for 8 hours and the percentage of dead cells was calculated (%Annexin V+ and/or 7-AAD+). No statistically significant difference in % dead cells was observed with the combination anti-CD47 antibody and rituximab treatment compared to either antibody alone (p>0.5). (B and C) NHL cells were incubated with the indicated immobilized antibodies for 2 hours (B) and 8 hours (C). No statistically significant differences in % dead cells were observed between combination antibody 110

treatment and either antibody alone at either 2 hours (p>0.2) or 8 hours (p>0.58). (D) Raji cells were incubated with mouse macrophages in the presence of the indicated antibodies for 2 hours. The non-phagocytosed Raji cells were then analyzed for apoptosis. Anti-CD47 antibody did not enable increased cell death compared to IgG1 isotype control (p=0.1), whereas rituximab enabled a minimal increase (p=0.005). Combination treatment with anti- CD47 antibody and rituximab did not enable increased apoptosis compared to either single agent alone (p=0.11). Experiments were performed in triplicate. (E) Luciferase-labeled Raji cells were injected intravenously into adult SCID mice. Two weeks later these mice were selectively depleted of either macrophages, NK cells, complement, or not depleted (vehicle control). Mice were then treated with the combination of anti-CD47 antibody and rituximab for 3 days with engraftment measured by luciferase signal pre and post-treatment. N=2-4 mice per depletion group. Compared to vehicle control, macrophage depletion abrogated anti-CD47+rituximab therapy with the response approaching statistical significance (p=0.16), while NK cell or complement depletion had no effect (p=0.27 and p=0.38, respectively).

We identified increased CD47 expression on a large subset of NHL cells compared to normal B cells and observed that antibody-mediated blockade of the CD47-SIRPα interaction enabled phagocytosis of NHL cells but not normal counterparts. Additionally, susceptibility of anti-CD47 antibody-mediated phagocytosis on NHL cells depended in part on CD47 expression. Indeed, the level of phagocytosis enabled by anti-CD47 antibody increased with higher CD47 expression on primary NHL samples (Figure 4D). Additionally, when CD47 expression was knocked-down by shRNA, a corresponding reduction in phagocytosis with anti-CD47 antibody was observed (Figures 4A-C). Therefore, the level of CD47 expression on NHL cells is a major determinant of susceptibility to anti- CD47 antibody therapy. The importance of CD47 expression to NHL disease pathogenesis also has significant clinical prognostic implications, specifically that higher levels of CD47 expression conferred a worse overall and event-free survival of NHL patients from several subtypes that have distinct molecular, pathologic, and clinical features. In fact, higher CD47 expression was associated with aggressive and adverse molecular features unique to each NHL subtype, and independently predicted prognosis apart from these molecular associations. On this basis, CD47 expression could be incorporated into standard clinical prognostic considerations across multiple subtypes of NHL, and may be useful in risk-adapted therapy decision-making.

111

Figure 13: Anti-CD47 Antibody Synergizes with Rituximab Through FcR-Independent and FcR-Dependent Mechanisms (A) Isobologram analysis of phagocytosis induced by anti-SIRPα antibody and rituximab is shown for Raji cells and mouse macrophages. (B,C) NHL cells were incubated in vitro with the indicated antibodies in the presence of 112

wild type (B) or Fcγr-/- (C) mouse macrophages, and the phagocytic index was determined. (D) Isobologram analysis of phagocytosis induced by anti-CD47 F(ab’)2 antibody and rituximab is shown for Raji cells and mouse macrophages. (E) NHL cells were incubated with wild type mouse macrophages in the presence of the indicated full length or F(ab’)2 antibodies (single antibodies at 10µg/ml, combination antibodies at 5µg/ml each) and the phagocytic index was determined. (F) The level of in vivo phagocytosis measured as the percentage of mouse macrophages containing phagocytosed GFP+ Raji cells (hCD45-GFP+F4/80+) was determined by flow cytometry of livers from mice engrafted with GFP+ Raji cells and then treated with the indicated antibodies (see methods), with each treatment group performed in duplicate. Compared to IgG control, anti-CD47 antibody and rituximab enabled increased levels of phagocytosis. Compared to anti-CD47 antibody alone, combination anti-CD47 antibody and rituximab enabled higher levels of phagocytosis. *p<0.05, **p<0.01, ****p<0.0001.

Table 2: Kaplan-Meier Survival Analysis for Antibody Treatment of Raji-Engrafted Mice Statistical analysis of NSG mice engrafted with disseminated Raji cells treated with either control IgG, anti-CD47 antibody, rituximab, or anti-CD47 antibody and rituximab. 95% confidence intervals were calculated for the hazard ratio, and p-values calculated using the Mantel-Cox test. All statistics are calculated compared to IgG- treated mice.

Table 3: Kaplan-Meier Survival Analysis for Antibody Treatment of Primary Human NHL-Engrafted Mice Statistical analysis of NSG mice engrafted with primary human NHL cells (NHL7) treated with either control IgG, anti-CD47 antibody, rituximab, or anti-CD47 antibody and rituximab. 95% confidence intervals were calculated for the hazard ratio, and p-values calculated using the Mantel-Cox test. All statistics are calculated compared to IgG-treated mice. *90% of mice are still alive at current date of manuscript.

113

-/- Figure 14: SIRPα Expression on Fcγ Macrophages and Generation of F(ab’)2 Fragments of Anti-CD47 Antibody and Rituximab (A) BM cells from Fcγr-/- mice were differentiated into macrophages as described and stained for the mouse macrophage marker F4/80 and SIRPα. Flow cytometry plot of SIRPα expression on F4/80+ macrophages is shown. (B) F(ab’)2 fragments of anti-CD47 antibody (mouse IgG1, clone B6H12.2) were generated by ficin cleavage and run on a SDS-PAGE gel to confirm digestion. No F(ab’)2 fragment was generated in the absence of the enzyme catalyst cysteine. (C) After digestion, the reaction mixture was purified by running over a protein A column to remove Fc fragments and undigested IgG and then reanalyzed by SDS-PAGE. (D) Binding of anti-

114

CD47 F(ab’)2 to CD47-positive Raji cells was confirmed by flow cytometry with a fluorophore-conjugated secondary antibody bound to the kappa light chain. (E) Raji cells with bound anti-CD47 F(ab’)2 antibody did not bind a protein A secondary, demonstrating complete Fc cleavage. (F to I) Rituximab F(ab’)2 fragments were generated by pepsin cleavage with SDS PAGE gels shown after enzyme digestion (F) and after protein A purification (G). No detectable Fc fragments were eluted given the near complete efficiency of pepsin digestion. Rituximab F(ab’)2 binding activity to CD20-positive Raji cells was confirmed (H) and a lack of Fc domain confirmed by an absence of protein A secondary staining (I). RTX: rituximab. MFI: mean fluorescence intensity

Macrophages are the Primary Mediators of Therapy with Combination Anti-CD47 Antibody and Rituximab Treatment While it is thought that many therapeutic monoclonal antibodies for human malignancies, including rituximab, function primarily through NK cell-mediated ADCC, several lines of evidence indicate that the therapeutic effect of anti-CD47 antibody alone or in combination with rituximab is mediated primarily through macrophage phagocytosis. First, synergistic macrophages phagocytosis was observed with combination anti-CD47 antibody and rituximab in vitro, while no synergy was observed for direct apoptosis, ADCC, or CDC (Figures 4C-E, 12, 11A, and 11C-F). Second, phagocytosis of NHL cells in vivo was observed with either anti-CD47 antibody or rituximab alone, and most importantly, significantly increased with combination therapy (Figure 13F). Third, the therapeutic effect of combination antibody treatment was similar in an NHL xenotransplant model using complement and NK cell-deficient NSG mice (Figure 4C) as in complement and NK cell-competent SCID mice (Figure 8A,B), suggesting that macrophages alone are sufficient to mediate the therapeutic effect. Fourth, depletion of macrophages, but not complement or NK cells abrogated the synergistic effect of anti-CD47 antibody in combination with rituximab (Figure 12E). Although the role of macrophages in mediating the therapeutic effects of monoclonal antibodies has been underappreciated, these studies highlight the importance of macrophages as effectors of anti-CD47 antibody therapy in human NHL.

Implications of Combination Therapy with Anti-CD47 and Rituximab in Human Malignancies This study describes a novel mechanism of antibody synergy in the elimination of NHL, supporting the feasibility of a combination antibody clinical therapy in the absence of chemotherapy. Combination therapy with two or more monoclonal antibodies possesses several advantages compared to monotherapies in NHL or other malignancies. First, therapy solely with monoclonal antibodies targeting cancer-specific antigens would result in decreased off-target toxicity compared to current therapeutic regimens that utilize combination chemotherapy. Second, synergy between two distinct antibody effector mechanisms, FcR-independent and FcR-dependent as shown here, would result in increased therapeutic efficacy. Third, antibody targeting of two distinct cell-surface antigens would be more likely to eliminate cancer cells with pre-existing epitope variants or epitope loss, such as those reported in rituximab-refractory/resistant NHL patients (203-205). Fourth, a bispecific FcR-engaging antibody with one arm binding and blocking CD47 and the other binding to a validated cancer antibody

115 target (CD20) could reduce potential antibody toxicity, while retaining the synergy effect. Given that CD47 is expressed in multiple normal tissue types, such a bispecific antibody could reduce potential toxicity by binding with higher affinity to cancer cells expressing both CD47 and the target antigen (CD20) compared to normal cells principally expressing a single antigen. Although we demonstrated that an anti-mouse CD47 antibody is relatively non-toxic to wild type mice (108), a clinical anti-human CD47 antibody may have a different human toxicity profile that could be overcome by a bispecific antibody. In addition to its application in NHL, the reported novel mechanism of antibody synergy provides the first proof-of-principle that a blocking mAb directed against CD47 can synergize with an FcR-activating antibody to provide superior therapeutic efficacy than either agent alone. This finding raises the possibility of potential synergy between an anti-CD47 antibody and other clinically approved therapeutic antibodies that may activate FcRs on immune effector cells for the treatment of diverse human malignancies including: trastuzumab (Herceptin) for HER2-positive breast carcinomas, cetuximab (Erbitux) for colorectal carcinomas and head and neck squamous cell carcinomas, alemtuzumab (Campath) for CLL and T-cell lymphoma, and others in clinical development (206). To date, we have demonstrated effective anti-CD47 antibody targeting of several human cancers including AML (108), bladder cancer (110), and now NHL, leading us to speculate that CD47 targeting will be effective against a wide range of human cancers.

116

CHAPTER 6

Therapeutic antibody targeting of CD47: Current Implications and Future Directions

Portions of this chapter were published in the following article: Jaiswal S, Chao MP, Majeti R, Weissman IL. Macrophages as mediators of tumor immunosurveillance. Trends Immunol. 2010: 31: 212-19.

117 SUMMARY

Tumor immunosurvelliance is a well-established mechanism for regulation of tumor growth. In addition to T- and NK-cells, macrophages play a major role in the recognition and clearance of foreign, aged, and damaged cells. Macrophage phagocytosis is negatively regulated via the receptor SIRPα upon binding to CD47, a ubiquitously expressed protein. We recently showed that CD47 is upregulated in several human tumors and demonstrate that increased CD47 expression results in the ability of these tumors to evade phagocytosis by the immune system. This mechanism of immune evasion can be therapeutically exploited through blockade of the CD47-SIRPα interaction with a monoclonal blocking antibody targeting CD47. Indeed, an anti-CD47 antibody eliminated several human tumor types in vitro and when engrafted into mouse xenotransplant models through enabling of phagocytosis. These findings support the rationale for development of an anti-CD47 antibody therapy in human malignancies and also may lead to several possibilities to augment antibody therapy.

DISCUSSION

Targeting CD47 to promote macrophage killing We showed that upregulation of CD47 in human myeloid leukemias and their tumor-initiating leukemia stem cells contributes to pathogenesis through the engagement of CD47 with SIRPα, leading to inhibition and evasion of phagocytosis by the host innate immune system (80) and (108). We have further demonstrated that this effect is extended to other human tumors as well. These findings have therapeutic implications in that blockade of the CD47-SIRPα interaction by a blocking monoclonal antibody against CD47 could enable phagocytosis of leukemia and other cancer cells. Anti-CD47 antibody was effective at enabling phagocytosis of human AML, ALL, NHL, CML, bladder, ovarian, and brain cancers in vitro as well as mediate tumor elimination in vivo in AML, ALL, and NHL. There is accumulating evidence that anti-CD47 antibodies may be effective in other tumors as well. Blocking antibodies directed against CD47 have been shown to eliminate a multiple myeloma cell line in vivo (89). In solid tumors, anti-CD47 antibodies have been shown to eliminate human head and neck squamous cell carcinoma cell lines (201) as well as primary human bladder cancers in vitro (110). Such reports demonstrate that antibody-mediated blockade of CD47 signaling may be therapeutic in multiple hematopoietic and solid tumor malignancies. We are currently characterizing CD47 expression in every human tumor type as well as investigating the pre-clinical therapeutic potential of an anti-CD47 antibody in these tumors. What is the mechanism behind anti-CD47 antibody-mediated elimination of tumor cells? In our own studies, we have shown that the mechanism of antibody action is mediated by macrophages, 118 and is caused by enabling phagocytosis via blockade of the CD47-SIRPα interaction, that is independent from Fc-receptor (FcR)-mediated antibody killing. Several lines of evidence support this finding. First, this therapeutic effect was mediated primarily by phagocytes, as depletion of macrophages in AML-engrafted immunocompromised mice with liposomal clondronate abrogated the anti-CD47 therapeutic response. Second, two distinct anti-CD47 antibodies that block the CD47-SIRPα interaction enabled phagocytosis, whereas a non-blocking anti-CD47 antibody did not, even though all three antibodies bound to target cells similarly. Third, an anti-SIRPα antibody also enabled phagocytosis of AML LSC, bulk ALL, and bulk NHL cells. Fourth, an isotype-matched anti-CD45 antibody did not enable phagocytosis of cancer cells, although the antibody bound to the cells. Fifth, an

F(ab)’2 fragment of the anti-CD47 antibody enabled phagocytosis of cancer cells even though the Fc receptor was absent. Sixth, anti-CD47 antibody did not enable ADCC or CDC using human NK cells and complement, respectively. Seventh, in mice containing functional macrophages, NK cells, and complement, anti-CD47 antibody-mediated tumor elimination in vivo required macrophages, but not NK cells or complement. Thus, these findings support a novel FcR-independent mechanism of antibody action for an anti-CD47 antibody.

Antibody targeting of CD47 on normal cells CD47 is a widely expressed protein on many cells of the hematopoietic system, as well as other tissues (155). Although much evidence supports the pre-clinical efficacy of anti-CD47 antibody- mediated elimination of tumor cells, given the widespread expression of CD47 in the hematopoietic system and other normal tissues, toxicity could be a barrier to translating the pre-clinical efficacy of an anti-CD47 antibody into a clinical therapy. To address this, the effect of an anti-CD47 antibody on normal cells was investigated on both human and mouse normal cell counterparts. Relatively minimal toxicity was observed. First, normal human CD34+ cells as well as normal peripheral blood cells were not phagocytosed in vitro with an anti-CD47 antibody despite phagocytosis of human leukemia cells. Second, when an anti-mouse CD47 antibody capable of enabling phagocytosis was administered to wild-type mice in vivo, minimal toxicity was observed, with no reduction in the HSC pool. If CD47 is expressed on multiple cell types in the hematopoietic system and in several organ tissues, why is significant toxicity with antibody-targeting of CD47 not observed? We demonstrated that the therapeutic selectivity is caused by the presence of the pro-phagocytic signal, calreticulin, on cancer cells that is not present on normal cell counterparts, as discussed in chapter 5. Anti-CD47 antibody therapy appears to operate in a therapeutic window that is reliant on the presence or absence of pro-phagocytic signals in combination with the anti-phagocytic signal, CD47.

Future Directions: Augmenting macrophage effectors in anti-CD47 antibody therapy That macrophages play an important role in tumor immunosurveillance and clearance in myeloid leukemias as well as other cancers is underappreciated. However, the notion that macrophages

119 are major mediators of an anti-CD47 antibody therapeutic effect has important implications in translation of such a therapy to human leukemia patients, other hematologic malignancies, as well as other cancers. For both hematologic and solid tumor malignancies, chemotherapy is the standard therapeutic option for treatment. However, chemotherapy regimens in many cancer types are far from being curative therapies, as exemplified by therapy in AML. Standard therapy for human AML involves cytostatic drugs such as anthracycline-based combination chemotherapy, that leads to long- term survival rates ranging from 10–75%, arguing for a need for improved therapy. Chemotherapy has been shown to induce an inflammatory response that attracts infiltrating macrophages into tumors, which in multiple tumors is associated with a better clinical prognosis (207-209). Thus, an anti-CD47 antibody therapy administered in the post-chemotherapy setting may result in increased efficacy by utilizing the infiltrating macrophages as immune effector cells at the site of disease. However, if chemotherapy led to long-term expression of pro-phagocytic signals (such as calreticulin, as proposed in chapter 5) in normal tissue cells, the anti-CD47 therapy could have significant added toxicity. Determining the conditions, if any, that induce cell surface calreticulin expression on normal cells is currently being investigated for this purpose. In addition to modulation by chemotherapy, increasing macrophage cell numbers may also augment the efficacy of an anti-CD47 antibody therapy by expanding the available pool of effector phagocytic cells. In the multiple tumor models tested including AML, ALL, and NHL, the therapeutic efficacy of anti-CD47 antibody appears partially dependent on tumor burden. For example, in AML and ALL-engrafted mice with severe tumor burden (bone marrow disease greater than 50%), anti-CD47 antibody therapy was unable to eliminate bone marrow disease compared to mice engrafted with lower levels of disease (bone marrow disease less than 50%) (Chapter 2, Figure 12). In these cases, the variable efficacy of anti-CD47 antibody in low versus high tumor-bearing mice is likely due to the number of macrophage effector cells available to mediate tumor elimination by phagocytosis. Accordingly, in leukemic mice with higher levels of bone marrow disease, less numbers of macrophages are available for leukemic clearance, thus resulting in a diminished therapeutic response compared to mice with low tumor burden and thus higher levels of available macrophages. Given that the level of leukemic disease will vary greatly in human patients, the ability to maximize the therapeutic effect of anti-CD47 antibody will be critical to its success. As a potential solution, anti-CD47 antibody therapy may be further improved by increasing the number of macrophages available in the tumor environment. Such macrophage manipulation could be achieved by increasing endogenous numbers of macrophages through administration of cytokines, including macrophage-colony stimulating factor (M- CSF) or granulocyte-macrophage-colony stimulating factor (GM-CSF). The combination of such cytokines with tumor-specific antibodies has been shown to be therapeutically effective in tumor models (210, 211). In addition, GM-CSF is clinically approved for use in AML patients, and is generally well-tolerated (reviewed in (212)). Therefore, combination therapy with GM-CSF and anti-

120

CD47 antibody could be a viable therapy that could be translated clinically. In addition to increasing endogenous macrophage production, exogenous macrophages could be increased through ex vivo transplantation of macrophage progenitors, including chronic myeloid progenitor (CMP) and granulocyte-macrophage progenitor (GMP) cells, that could be followed by anti-CD47 antibody (213) or fully allogeneic (214) mouse CMP/GMP into mice challenged with bacterial or fungal pathogens can increase myeloid effector cell numbers, and prevent fulminant infection and death. Currently, we are investigating these therapeutic combinations in pre-clinical mouse models.

Anti-CD47 antibody: from pre-clinical therapy to clinical translation Three years ago, our laboratory identified the anti-phagocytic signal CD47 as playing an important role in the pathogenesis of AML through evasion of phagocytosis. Within the past three years, we have identified CD47 as a therapeutic target and developed an anti-CD47 antibody that is effective at reducing or eliminating disease in several human cancers including AML, bladder cancer, ALL, Non-Hodgkin Lymphoma and several others. Additionally, we have demonstrated that anti- CD47 antibody possesses a minimal pre-clinical toxicity profile and investigated in depth the therapeutic mechanisms of anti-CD47 antibody and its ability to selectively target tumor but not normal cells. Currently, we are rapidly developing an anti-CD47 antibody for the treatment of human leukemias, Non-Hodgkin Lymphoma, and other cancers in human patients. Our current efforts are involved in the generation of humanized and chimerized anti-CD47 antibodies for the treatment of human patients and are studying the pharmacokinetic and toxicity profiles of these antibody formats. We hope to begin a Phase I trial with these anti-CD47 antibodies for the treatment of human cancer patients within the next several years and are excited about its therapeutic potential.

121 REFERENCES

1. Vogelstein, B. and Kinzler, K. W. Cancer genes and the pathways they control. Nat Med, 10: 789-799, 2004. 2. Makino, S. Further evidence favoring the concept of the stem cell in ascites tumors of . Ann N Y Acad Sci, 63: 818-830, 1956. 3. Bruce, W. R. and Van Der Gaag, H. A Quantitative Assay for the Number of Murine Lymphoma Cells Capable of Proliferation in Vivo. Nature, 199: 79-80, 1963. 4. Kleinsmith, L. J. and Pierce, G. B., Jr. Multipotentiality of Single Embryonal Carcinoma Cells. Cancer Res, 24: 1544-1551, 1964. 5. Park, C. H., Bergsagel, D. E., and McCulloch, E. A. Mouse myeloma tumor stem cells: a primary cell culture assay. J Natl Cancer Inst, 46: 411-422, 1971. 6. Cohnheim, V. Congenitales, quergestreiftes muskelsarkom der nieren. . Virchows Arch Pathol Anat Physiol Klin Med, 65: 64-69, 1875. 7. Lapidot, T., Sirard, C., Vormoor, J., Murdoch, B., Hoang, T., Caceres-Cortes, J., Minden, M., Paterson, B., Caligiuri, M. A., and Dick, J. E. A cell initiating human acute myeloid leukaemia after transplantation into SCID mice. Nature, 367: 645-648, 1994. 8. Bonnet, D. and Dick, J. E. Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nat Med, 3: 730-737, 1997. 9. Passegue, E., Jamieson, C. H., Ailles, L. E., and Weissman, I. L. Normal and leukemic hematopoiesis: are leukemias a stem cell disorder or a reacquisition of stem cell characteristics? Proc Natl Acad Sci U S A, 100 Suppl 1: 11842-11849, 2003. 10. Weissman, I. Stem cell research: paths to cancer therapies and regenerative medicine. Jama, 294: 1359-1366, 2005. 11. Clarke, M. F. and Fuller, M. Stem cells and cancer: two faces of eve. Cell, 124: 1111-1115, 2006. 12. Baum, C. M., Weissman, I. L., Tsukamoto, A. S., Buckle, A. M., and Peault, B. Isolation of a candidate human hematopoietic stem-cell population. Proc Natl Acad Sci U S A, 89: 2804- 2808, 1992. 13. Blair, A., Hogge, D. E., Ailles, L. E., Lansdorp, P. M., and Sutherland, H. J. Lack of expression of Thy-1 (CD90) on acute myeloid leukemia cells with long-term proliferative ability in vitro and in vivo. Blood, 89: 3104-3112, 1997. 14. Majeti, R., Park, C., and Weissman, I. Identification of a Hierarchy of Multipotent Hematopoietic Progenitors in Human Cord Blood. Cell Stem Cell, 1: 635-645, 2007. 15. Miyamoto, T., Weissman, I. L., and Akashi, K. AML1/ETO-expressing nonleukemic stem cells in acute myelogenous leukemia with 8;21 chromosomal translocation. Proc Natl Acad Sci U S A, 97: 7521-7526, 2000. 16. Jamieson, C. H., Ailles, L. E., Dylla, S. J., Muijtjens, M., Jones, C., Zehnder, J. L., Gotlib, J., Li, K., Manz, M. G., Keating, A., Sawyers, C. L., and Weissman, I. L. Granulocyte- macrophage progenitors as candidate leukemic stem cells in blast-crisis CML. N Engl J Med, 351: 657-667, 2004. 17. Clarke, M. F., Dick, J. E., Dirks, P. B., Eaves, C. J., Jamieson, C. H., Jones, D. L., Visvader, J., Weissman, I. L., and Wahl, G. M. Cancer stem cells--perspectives on current status and future directions: AACR Workshop on cancer stem cells. Cancer Res, 66: 9339-9344, 2006. 18. Lobo, N. A., Shimono, Y., Qian, D., and Clarke, M. F. The Biology of Cancer Stem Cells. Annu Rev Cell Dev Biol, 23: 675-699, 2007. 19. Dalerba, P., Cho, R. W., and Clarke, M. F. Cancer stem cells: models and concepts. Annu Rev Med, 58: 267-284, 2007. 20. Ward, R. J. and Dirks, P. B. Cancer Stem Cells: At the Headwaters of Tumor Development. Annu Rev Pathol, 2: 175-189, 2007. 21. Cho, R. W., Wang, X., Diehn, M., Shedden, K., Chen, G. Y., Sherlock, G., Gurney, A., Lewicki, J., and Clarke, M. F. Isolation and molecular characterization of cancer stem cells in MMTV-Wnt-1 murine breast tumors. Stem Cells, 26: 364-371, 2008.

122

22. Kennedy, J. A., Barabe, F., Poeppl, A. G., Wang, J. C., and Dick, J. E. Comment on "Tumor growth need not be driven by rare cancer stem cells". Science, 318: 1722; author reply 1722, 2007. 23. Xin, L., Lawson, D. A., and Witte, O. N. The Sca-1 cell surface marker enriches for a prostate- regenerating cell subpopulation that can initiate prostate tumorigenesis. Proc Natl Acad Sci U S A, 102: 6942-6947, 2005. 24. Monzani, E., Facchetti, F., Galmozzi, E., Corsini, E., Benetti, A., Cavazzin, C., Gritti, A., Piccinini, A., Porro, D., Santinami, M., Invernici, G., Parati, E., Alessandri, G., and La Porta, C. A. Melanoma contains CD133 and ABCG2 positive cells with enhanced tumourigenic potential. Eur J Cancer, 43: 935-946, 2007. 25. Collins, A. T., Berry, P. A., Hyde, C., Stower, M. J., and Maitland, N. J. Prospective identification of tumorigenic prostate cancer stem cells. Cancer Res, 65: 10946-10951, 2005. 26. Fang, D., Nguyen, T. K., Leishear, K., Finko, R., Kulp, A. N., Hotz, S., Van Belle, P. A., Xu, X., Elder, D. E., and Herlyn, M. A tumorigenic subpopulation with stem cell properties in melanomas. Cancer Res, 65: 9328-9337, 2005. 27. Eisterer, W., Jiang, X., Christ, O., Glimm, H., Lee, K. H., Pang, E., Lambie, K., Shaw, G., Holyoake, T. L., Petzer, A. L., Auewarakul, C., Barnett, M. J., Eaves, C. J., and Eaves, A. C. Different subsets of primary chronic myeloid leukemia stem cells engraft immunodeficient mice and produce a model of the human disease. Leukemia, 19: 435-441, 2005. 28. Al-Hajj, M., Wicha, M. S., Benito-Hernandez, A., Morrison, S. J., and Clarke, M. F. Prospective identification of tumorigenic breast cancer cells. Proc Natl Acad Sci U S A, 100: 3983-3988, 2003. 29. Prince, M. E., Sivanandan, R., Kaczorowski, A., Wolf, G. T., Kaplan, M. J., Dalerba, P., Weissman, I. L., Clarke, M. F., and Ailles, L. E. Identification of a subpopulation of cells with cancer stem cell properties in head and neck squamous cell carcinoma. Proc Natl Acad Sci U S A, 104: 973-978, 2007. 30. Li, C., Heidt, D. G., Dalerba, P., Burant, C. F., Zhang, L., Adsay, V., Wicha, M., Clarke, M. F., and Simeone, D. M. Identification of pancreatic cancer stem cells. Cancer Res, 67: 1030- 1037, 2007. 31. Singh, S. K., Clarke, I. D., Terasaki, M., Bonn, V. E., Hawkins, C., Squire, J., and Dirks, P. B. Identification of a cancer stem cell in human brain tumors. Cancer Res, 63: 5821-5828, 2003. 32. Singh, S. K., Hawkins, C., Clarke, I. D., Squire, J. A., Bayani, J., Hide, T., Henkelman, R. M., Cusimano, M. D., and Dirks, P. B. Identification of human brain tumour initiating cells. Nature, 432: 396-401, 2004. 33. Ricci-Vitiani, L., Lombardi, D. G., Pilozzi, E., Biffoni, M., Todaro, M., Peschle, C., and De Maria, R. Identification and expansion of human colon-cancer-initiating cells. Nature, 445: 111-115, 2007. 34. O'Brien, C. A., Pollett, A., Gallinger, S., and Dick, J. E. A human colon cancer cell capable of initiating tumour growth in immunodeficient mice. Nature, 445: 106-110, 2007. 35. Dalerba, P., Dylla, S. J., Park, I. K., Liu, R., Wang, X., Cho, R. W., Hoey, T., Gurney, A., Huang, E. H., Simeone, D. M., Shelton, A. A., Parmiani, G., Castelli, C., and Clarke, M. F. Phenotypic characterization of human colorectal cancer stem cells. Proc Natl Acad Sci U S A, 104: 10158-10163, 2007. 36. Schatton, T., Murphy, G. F., Frank, N. Y., Yamaura, K., Waaga-Gasser, A. M., Gasser, M., Zhan, Q., Jordan, S., Duncan, L. M., Weishaupt, C., Fuhlbrigge, R. C., Kupper, T. S., Sayegh, M. H., and Frank, M. H. Identification of cells initiating human melanomas. Nature, 451: 345- 349, 2008. 37. Cobaleda, C., Gutierrez-Cianca, N., Perez-Losada, J., Flores, T., Garcia-Sanz, R., Gonzalez, M., and Sanchez-Garcia, I. A primitive hematopoietic cell is the target for the leukemic transformation in human philadelphia-positive acute lymphoblastic leukemia. Blood, 95: 1007- 1013, 2000. 38. Castor, A., Nilsson, L., Astrand-Grundstrom, I., Buitenhuis, M., Ramirez, C., Anderson, K., Strombeck, B., Garwicz, S., Bekassy, A. N., Schmiegelow, K., Lausen, B., Hokland, P., Lehmann, S., Juliusson, G., Johansson, B., and Jacobsen, S. E. Distinct patterns of

123 involvement in acute lymphoblastic leukemia. Nat Med, 11: 630-637, 2005. 39. Cox, C. V., Evely, R. S., Oakhill, A., Pamphilon, D. H., Goulden, N. J., and Blair, A. Characterization of acute lymphoblastic leukemia progenitor cells. Blood, 104: 2919-2925, 2004. 40. Cox, C. V., Martin, H. M., Kearns, P. R., Virgo, P., Evely, R. S., and Blair, A. Characterization of a progenitor cell population in childhood T-cell acute lymphoblastic leukemia. Blood, 109: 674-682, 2007. 41. Matsui, W., Huff, C. A., Wang, Q., Malehorn, M. T., Barber, J., Tanhehco, Y., Smith, B. D., Civin, C. I., and Jones, R. J. Characterization of clonogenic multiple myeloma cells. Blood, 103: 2332-2336, 2004. 42. Matsui, W., Wang, Q., Barber, J. P., Brennan, S., Smith, B. D., Borrello, I., McNiece, I., Lin, L., Ambinder, R. F., Peacock, C., Watkins, D. N., Huff, C. A., and Jones, R. J. Clonogenic multiple myeloma progenitors, stem cell properties, and drug resistance. Cancer Res, 68: 190- 197, 2008. 43. Morrison, S. J. and Weissman, I. L. The long-term repopulating subset of hematopoietic stem cells is deterministic and isolatable by phenotype. Immunity, 1: 661-673, 1994. 44. Guan, Y., Gerhard, B., and Hogge, D. E. Detection, isolation, and stimulation of quiescent primitive leukemic progenitor cells from patients with acute myeloid leukemia (AML). Blood, 101: 3142-3149, 2003. 45. Dean, M., Fojo, T., and Bates, S. Tumour stem cells and drug resistance. Nat Rev Cancer, 5: 275-284, 2005. 46. Costello, R. T., Mallet, F., Gaugler, B., Sainty, D., Arnoulet, C., Gastaut, J. A., and Olive, D. Human acute myeloid leukemia CD34+/CD38- progenitor cells have decreased sensitivity to chemotherapy and Fas-induced apoptosis, reduced immunogenicity, and impaired dendritic cell transformation capacities. Cancer Res, 60: 4403-4411, 2000. 47. Bao, S., Wu, Q., McLendon, R. E., Hao, Y., Shi, Q., Hjelmeland, A. B., Dewhirst, M. W., Bigner, D. D., and Rich, J. N. Glioma stem cells promote radioresistance by preferential activation of the DNA damage response. Nature, 444: 756-760, 2006. 48. Holyoake, T., Jiang, X., Eaves, C., and Eaves, A. Isolation of a highly quiescent subpopulation of primitive leukemic cells in chronic myeloid leukemia. Blood, 94: 2056-2064, 1999. 49. Druker, B. J., Tamura, S., Buchdunger, E., Ohno, S., Segal, G. M., Fanning, S., Zimmermann, J., and Lydon, N. B. Effects of a selective inhibitor of the Abl tyrosine kinase on the growth of Bcr-Abl positive cells. Nat Med, 2: 561-566, 1996. 50. Michor, F. Quantitative approaches to analyzing imatinib-treated chronic myeloid leukemia. Trends Pharmacol Sci, 28: 197-199, 2007. 51. Michor, F. Chronic myeloid leukemia blast crisis arises from progenitors. Stem Cells, 25: 1114-1118, 2007. 52. Selleri, C., Notaro, R., Catalano, L., Fontana, R., Del Vecchio, L., and Rotoli, B. Prognostic irrelevance of CD34 in acute myeloid leukaemia. Br J Haematol, 82: 479-481, 1992. 53. Kyoda, K., Nakamura, S., Hattori, N., Takeshima, M., Nakamura, K., Kaya, H., Matano, S., Okumura, H., Kanno, M., Ohtake, S., and Matsuda, T. Lack of prognostic significance of CD34 expression in adult AML when FAB M0 and M3 are excluded. Am J Hematol, 57: 265- 266, 1998. 54. Bradstock, K., Matthews, J., Benson, E., Page, F., and Bishop, J. Prognostic value of immunophenotyping in acute myeloid leukemia. Australian Leukaemia Study Group. Blood, 84: 1220-1225, 1994. 55. van Rhenen, A., Feller, N., Kelder, A., Westra, A. H., Rombouts, E., Zweegman, S., van der Pol, M. A., Waisfisz, Q., Ossenkoppele, G. J., and Schuurhuis, G. J. High stem cell frequency in acute myeloid leukemia at diagnosis predicts high minimal residual disease and poor survival. Clin Cancer Res, 11: 6520-6527, 2005. 56. Haferlach, T., Schoch, C., Loffler, H., Gassmann, W., Kern, W., Schnittger, S., Fonatsch, C., Ludwig, W. D., Wuchter, C., Schlegelberger, B., Staib, P., Reichle, A., Kubica, U., Eimermacher, H., Balleisen, L., Gruneisen, A., Haase, D., Aul, C., Karow, J., Lengfelder, E., Wormann, B., Heinecke, A., Sauerland, M. C., Buchner, T., and Hiddemann, W. Morphologic 124

dysplasia in de novo acute myeloid leukemia (AML) is related to unfavorable cytogenetics but has no independent prognostic relevance under the conditions of intensive induction therapy: results of a multiparameter analysis from the German AML Cooperative Group studies. J Clin Oncol, 21: 256-265, 2003. 57. Costello, R., Mallet, F., Chambost, H., Sainty, D., Arnoulet, C., Gastaut, J. A., and Olive, D. The immunophenotype of minimally differentiated acute myeloid leukemia (AML-M0): reduced immunogenicity and high frequency of CD34+/CD38- leukemic progenitors. Leukemia, 13: 1513-1518, 1999. 58. Pearce, D. J., Taussig, D., Zibara, K., Smith, L. L., Ridler, C. M., Preudhomme, C., Young, B. D., Rohatiner, A. Z., Lister, T. A., and Bonnet, D. AML engraftment in the NOD/SCID assay reflects the outcome of AML: implications for our understanding of the heterogeneity of AML. Blood, 107: 1166-1173, 2006. 59. Abraham, B. K., Fritz, P., McClellan, M., Hauptvogel, P., Athelogou, M., and Brauch, H. Prevalence of CD44+/CD24-/low cells in breast cancer may not be associated with clinical outcome but may favor distant metastasis. Clin Cancer Res, 11: 1154-1159, 2005. 60. Liu, R., Wang, X., Chen, G. Y., Dalerba, P., Gurney, A., Hoey, T., Sherlock, G., Lewicki, J., Shedden, K., and Clarke, M. F. The prognostic role of a gene signature from tumorigenic breast-cancer cells. N Engl J Med, 356: 217-226, 2007. 61. Zeppernick, F., Ahmadi, R., Campos, B., Dictus, C., Helmke, B. M., Becker, N., Lichter, P., Unterberg, A., Radlwimmer, B., and Herold-Mende, C. C. Stem cell marker CD133 affects clinical outcome in glioma patients. Clin Cancer Res, 14: 123-129, 2008. 62. Reya, T., Morrison, S. J., Clarke, M. F., and Weissman, I. L. Stem cells, cancer, and cancer stem cells. Nature, 414: 105-111, 2001. 63. Jordan, C. T., Guzman, M. L., and Noble, M. Cancer stem cells. N Engl J Med, 355: 1253- 1261, 2006. 64. Wang, J. C. and Dick, J. E. Cancer stem cells: lessons from leukemia. Trends Cell Biol, 15: 494-501, 2005. 65. Estey, E. and Dohner, H. Acute myeloid leukaemia. Lancet, 368: 1894-1907, 2006. 66. Lowenberg, B., Downing, J. R., and Burnett, A. Acute myeloid leukemia. N Engl J Med, 341: 1051-1062, 1999. 67. Byrd, J. C., Mrozek, K., Dodge, R. K., Carroll, A. J., Edwards, C. G., Arthur, D. C., Pettenati, M. J., Patil, S. R., Rao, K. W., Watson, M. S., Koduru, P. R., Moore, J. O., Stone, R. M., Mayer, R. J., Feldman, E. J., Davey, F. R., Schiffer, C. A., Larson, R. A., and Bloomfield, C. D. Pretreatment cytogenetic abnormalities are predictive of induction success, cumulative incidence of relapse, and overall survival in adult patients with de novo acute myeloid leukemia: results from Cancer and Leukemia Group B (CALGB 8461). Blood, 100: 4325- 4336, 2002. 68. Grimwade, D., Walker, H., Oliver, F., Wheatley, K., Harrison, C., Harrison, G., Rees, J., Hann, I., Stevens, R., Burnett, A., and Goldstone, A. The importance of diagnostic cytogenetics on outcome in AML: analysis of 1,612 patients entered into the MRC AML 10 trial. The Medical Research Council Adult and Children's Leukaemia Working Parties. Blood, 92: 2322-2333, 1998. 69. Mrozek, K., Marcucci, G., Paschka, P., Whitman, S. P., and Bloomfield, C. D. Clinical relevance of mutations and gene-expression changes in adult acute myeloid leukemia with normal cytogenetics: are we ready for a prognostically prioritized molecular classification? Blood, 109: 431-448, 2007. 70. Schlenk, R. F., Dohner, K., Krauter, J., Frohling, S., Corbacioglu, A., Bullinger, L., Habdank, M., Spath, D., Morgan, M., Benner, A., Schlegelberger, B., Heil, G., Ganser, A., and Dohner, H. Mutations and treatment outcome in cytogenetically normal acute myeloid leukemia. N Engl J Med, 358: 1909-1918, 2008. 71. Majeti, R., Park, C. Y., and Weissman, I. L. Identification of a hierarchy of multipotent hematopoietic progenitors in human cord blood. Cell Stem Cell, 1: 635-645, 2007. 72. Majeti, R., Becker, M. W., Tian, Q., Lee, T. L., Yan, X., Liu, R., Chiang, J. H., Hood, L., Clarke, M. F., and Weissman, I. L. Dysregulated gene expression networks in human acute myelogenous leukemia stem cells. Proc Natl Acad Sci U S A, 2009. 125 73. Brown, E. J. and Frazier, W. A. Integrin-associated protein (CD47) and its ligands. Trends Cell Biol, 11: 130-135, 2001. 74. Barclay, A. N. and Brown, M. H. The SIRP family of receptors and immune regulation. Nat Rev Immunol, 6: 457-464, 2006. 75. Oldenborg, P. A., Zheleznyak, A., Fang, Y. F., Lagenaur, C. F., Gresham, H. D., and Lindberg, F. P. Role of CD47 as a marker of self on red blood cells. Science, 288: 2051-2054, 2000. 76. Blazar, B. R., Lindberg, F. P., Ingulli, E., Panoskaltsis-Mortari, A., Oldenborg, P. A., Iizuka, K., Yokoyama, W. M., and Taylor, P. A. CD47 (integrin-associated protein) engagement of dendritic cell and macrophage counterreceptors is required to prevent the clearance of donor lymphohematopoietic cells. J Exp Med, 194: 541-549, 2001. 77. Okazawa, H., Motegi, S., Ohyama, N., Ohnishi, H., Tomizawa, T., Kaneko, Y., Oldenborg, P. A., Ishikawa, O., and Matozaki, T. Negative regulation of phagocytosis in macrophages by the CD47-SHPS-1 system. J Immunol, 174: 2004-2011, 2005. 78. Oldenborg, P. A., Gresham, H. D., and Lindberg, F. P. CD47-signal regulatory protein alpha (SIRPalpha) regulates Fcgamma and complement receptor-mediated phagocytosis. J Exp Med, 193: 855-862, 2001. 79. Mateo, V., Lagneaux, L., Bron, D., Biron, G., Armant, M., Delespesse, G., and Sarfati, M. CD47 ligation induces caspase-independent cell death in chronic lymphocytic leukemia. Nat Med, 5: 1277-1284, 1999. 80. Jaiswal, S., Jamieson, C. H., Pang, W. W., Park, C. Y., Chao, M. P., Majeti, R., Traver, D., van Rooijen, N., and Weissman, I. L. CD47 is upregulated on circulating hematopoietic stem cells and leukemia cells to avoid phagocytosis. Cell, 138: 271-285, 2009. 81. Valk, P. J., Verhaak, R. G., Beijen, M. A., Erpelinck, C. A., Barjesteh van Waalwijk van Doorn-Khosrovani, S., Boer, J. M., Beverloo, H. B., Moorhouse, M. J., van der Spek, P. J., Lowenberg, B., and Delwel, R. Prognostically useful gene-expression profiles in acute myeloid leukemia. N Engl J Med, 350: 1617-1628, 2004. 82. Metzeler, K. H., Hummel, M., Bloomfield, C. D., Spiekermann, K., Braess, J., Sauerland, M. C., Heinecke, A., Radmacher, M., Marcucci, G., Whitman, S. P., Maharry, K., Paschka, P., Larson, R. A., Berdel, W. E., Buchner, T., Wormann, B., Mansmann, U., Hiddemann, W., Bohlander, S. K., and Buske, C. An 86-probe-set gene-expression signature predicts survival in cytogenetically normal acute myeloid leukemia. Blood, 112: 4193-4201, 2008. 83. Bullinger, L., Dohner, K., Kranz, R., Stirner, C., Frohling, S., Scholl, C., Kim, Y. H., Schlenk, R. F., Tibshirani, R., Dohner, H., and Pollack, J. R. An FLT3 gene-expression signature predicts clinical outcome in normal karyotype AML. Blood, 111: 4490-4495, 2008. 84. Jongen-Lavrencic, M., Sun, S. M., Dijkstra, M. K., Valk, P. J., and Lowenberg, B. MicroRNA expression profiling in relation to the genetic heterogeneity of acute myeloid leukemia. Blood, 111: 5078-5085, 2008. 85. Ailles, L. E., Gerhard, B., and Hogge, D. E. Detection and characterization of primitive malignant and normal progenitors in patients with acute myelogenous leukemia using long- term coculture with supportive feeder layers and cytokines. Blood, 90: 2555-2564, 1997. 86. Subramanian, S., Parthasarathy, R., Sen, S., Boder, E. T., and Discher, D. E. Species- and cell type-specific interactions between CD47 and human SIRPalpha. Blood, 107: 2548-2556, 2006. 87. Ide, K., Wang, H., Tahara, H., Liu, J., Wang, X., Asahara, T., Sykes, M., Yang, Y. G., and Ohdan, H. Role for CD47-SIRPalpha signaling in xenograft rejection by macrophages. Proc Natl Acad Sci U S A, 104: 5062-5066, 2007. 88. Takenaka, K., Prasolava, T. K., Wang, J. C., Mortin-Toth, S. M., Khalouei, S., Gan, O. I., Dick, J. E., and Danska, J. S. Polymorphism in Sirpa modulates engraftment of human hematopoietic stem cells. Nat Immunol, 8: 1313-1323, 2007. 89. Kikuchi, Y., Uno, S., Kinoshita, Y., Yoshimura, Y., Iida, S., Wakahara, Y., Tsuchiya, M., Yamada-Okabe, H., and Fukushima, N. Apoptosis inducing bivalent single-chain antibody fragments against CD47 showed antitumor potency for multiple myeloma. Leuk Res, 29: 445- 450, 2005. 90. Kikuchi, Y., Uno, S., Yoshimura, Y., Otabe, K., Iida, S., Oheda, M., Fukushima, N., and Tsuchiya, M. A bivalent single-chain Fv fragment against CD47 induces apoptosis for leukemic cells. Biochem Biophys Res Commun, 315: 912-918, 2004. 126

91. Uno, S., Kinoshita, Y., Azuma, Y., Tsunenari, T., Yoshimura, Y., Iida, S., Kikuchi, Y., Yamada-Okabe, H., and Fukushima, N. Antitumor activity of a monoclonal antibody against CD47 in xenograft models of human leukemia. Oncol Rep, 17: 1189-1194, 2007. 92. Adams, G. P. and Weiner, L. M. Monoclonal antibody therapy of cancer. Nat Biotechnol, 23: 1147-1157, 2005. 93. Braun, D., Galibert, L., Nakajima, T., Saito, H., Quang, V. V., Rubio, M., and Sarfati, M. Semimature stage: a checkpoint in a dendritic cell maturation program that allows for functional reversion after signal-regulatory protein-alpha ligation and maturation signals. J Immunol, 177: 8550-8559, 2006. 94. Latour, S., Tanaka, H., Demeure, C., Mateo, V., Rubio, M., Brown, E. J., Maliszewski, C., Lindberg, F. P., Oldenborg, A., Ullrich, A., Delespesse, G., and Sarfati, M. Bidirectional negative regulation of human T and dendritic cells by CD47 and its cognate receptor signal- regulator protein-alpha: down-regulation of IL-12 responsiveness and inhibition of dendritic cell activation. J Immunol, 167: 2547-2554, 2001. 95. Sarfati, M., Fortin, G., Raymond, M., and Susin, S. CD47 in the immune response: role of and SIRP-alpha reverse signaling. Curr Drug Targets, 9: 842-850, 2008. 96. Seiffert, M., Cant, C., Chen, Z., Rappold, I., Brugger, W., Kanz, L., Brown, E. J., Ullrich, A., and Buhring, H. J. Human signal-regulatory protein is expressed on normal, but not on subsets of leukemic myeloid cells and mediates cellular adhesion involving its counterreceptor CD47. Blood, 94: 3633-3643, 1999. 97. Jordan, C. T., Upchurch, D., Szilvassy, S. J., Guzman, M. L., Howard, D. S., Pettigrew, A. L., Meyerrose, T., Rossi, R., Grimes, B., Rizzieri, D. A., Luger, S. M., and Phillips, G. L. The interleukin-3 receptor alpha chain is a unique marker for human acute myelogenous leukemia stem cells. Leukemia, 14: 1777-1784, 2000. 98. Jin, L., Hope, K. J., Zhai, Q., Smadja-Joffe, F., and Dick, J. E. Targeting of CD44 eradicates human acute myeloid leukemic stem cells. Nat Med, 12: 1167-1174, 2006. 99. Hosen, N., Park, C. Y., Tatsumi, N., Oji, Y., Sugiyama, H., Gramatzki, M., Krensky, A. M., and Weissman, I. L. CD96 is a leukemic stem cell-specific marker in human acute myeloid leukemia. Proc Natl Acad Sci U S A, 104: 11008-11013, 2007. 100. van Rhenen, A., van Dongen, G. A., Kelder, A., Rombouts, E. J., Feller, N., Moshaver, B., Stigter-van Walsum, M., Zweegman, S., Ossenkoppele, G. J., and Jan Schuurhuis, G. The novel AML stem cell associated antigen CLL-1 aids in discrimination between normal and leukemic stem cells. Blood, 110: 2659-2666, 2007. 101. Lock, R. B., Jin, L., Lee, E. M., Ramshaw, H. S., Busfield, S., Peoppl, A. G., Wilkinson, L., Guthridge, M. A., Boyd, A., Gearing, D., Vairo, G., Lopez, A., and Dick, J. E. CD123 (IL-3 Receptor {alpha} Chain) Neutralization by a Monoclonal Antibody Selectively Eliminates Human Acute Myeloid Leukemic Stem Cells. ASH Annual Meeting Abstracts, 110: 161-, 2007. 102. Nimmerjahn, F. and Ravetch, J. V. Antibodies, Fc receptors and cancer. Curr Opin Immunol, 19: 239-245, 2007. 103. Pui, C. H., Relling, M. V., and Downing, J. R. Acute lymphoblastic leukemia. N Engl J Med, 350: 1535-1548, 2004. 104. Smith, M., Arthur, D., Camitta, B., Carroll, A. J., Crist, W., Gaynon, P., Gelber, R., Heerema, N., Korn, E. L., Link, M., Murphy, S., Pui, C. H., Pullen, J., Reamon, G., Sallan, S. E., Sather, H., Shuster, J., Simon, R., Trigg, M., Tubergen, D., Uckun, F., and Ungerleider, R. Uniform approach to risk classification and treatment assignment for children with acute lymphoblastic leukemia. J Clin Oncol, 14: 18-24, 1996. 105. Schultz, K. R., Pullen, D. J., Sather, H. N., Shuster, J. J., Devidas, M., Borowitz, M. J., Carroll, A. J., Heerema, N. A., Rubnitz, J. E., Loh, M. L., Raetz, E. A., Winick, N. J., Hunger, S. P., Carroll, W. L., Gaynon, P. S., and Camitta, B. M. Risk- and response-based classification of childhood B-precursor acute lymphoblastic leukemia: a combined analysis of prognostic markers from the Pediatric Oncology Group (POG) and Children's Cancer Group (CCG). Blood, 109: 926-935, 2007. 106. Pui, C. H., Campana, D., and Evans, W. E. Childhood acute lymphoblastic leukaemia--current status and future perspectives. Lancet Oncol, 2: 597-607, 2001. 127 107. Pui, C. H. and Jeha, S. New therapeutic strategies for the treatment of acute lymphoblastic leukaemia. Nat Rev Drug Discov, 6: 149-165, 2007. 108. Majeti, R., Chao, M. P., Alizadeh, A. A., Pang, W. W., Jaiswal, S., Gibbs, K. D., Jr., van Rooijen, N., and Weissman, I. L. CD47 is an adverse prognostic factor and therapeutic antibody target on human acute myeloid leukemia stem cells. Cell, 138: 286-299, 2009. 109. Matozaki, T., Murata, Y., Okazawa, H., and Ohnishi, H. Functions and molecular mechanisms of the CD47-SIRPalpha signalling pathway. Trends Cell Biol, 19: 72-80, 2009. 110. Chan, K. S., Espinosa, I., Chao, M., Wong, D., Ailles, L., Diehn, M., Gill, H., Presti, J., Jr., Chang, H. Y., van de Rijn, M., Shortliffe, L., and Weissman, I. L. Identification, molecular characterization, clinical prognosis, and therapeutic targeting of human bladder tumor- initiating cells. Proc Natl Acad Sci U S A, 106: 14016-14021, 2009. 111. Holleman, A., den Boer, M. L., Cheok, M. H., Kazemier, K. M., Pei, D., Downing, J. R., Janka-Schaub, G. E., Gobel, U., Graubner, U. B., Pui, C.-H., Evans, W. E., and Pieters, R. Expression of the outcome predictor in acute leukemia 1 (OPAL1) gene is not an independent prognostic factor in patients treated according to COALL or St Jude protocols. Blood, 108: 1984-1990, 2006. 112. Kang, H., Chen, I. M., Wilson, C. S., Bedrick, E. J., Harvey, R. C., Atlas, S. R., Devidas, M., Mullighan, C. G., Wang, X., Murphy, M., Ar, K., Wharton, W., Borowitz, M. J., Bowman, W. P., Bhojwani, D., Carroll, W. L., Camitta, B. M., Reaman, G. H., Smith, M. A., Downing, J. R., Hunger, S. P., and Willman, C. L. Gene expression classifiers for relapse-free survival and minimal residual disease improve risk classification and outcome prediction in pediatric B- precursor acute lymphoblastic leukemia. Blood, 115: 1394-1405. 113. Willman, C. L. Discovery of novel molecular classification schemes and genes predictive of outcome in leukemia. Hematol J, 5 Suppl 3: S138-143, 2004. 114. Dai, M., Wang, P., Boyd, A. D., Kostov, G., Athey, B., Jones, E. G., Bunney, W. E., Myers, R. M., Speed, T. P., Akil, H., Watson, S. J., and Meng, F. Evolving gene/transcript definitions significantly alter the interpretation of GeneChip data. Nucleic Acids Res, 33: e175, 2005. 115. Bolstad, B. M., Collin, F., Simpson, K. M., Irizarry, R. A., and Speed, T. P. Experimental design and low-level analysis of microarray data. Int Rev Neurobiol, 60: 25-58, 2004. 116. Cox, D. R. Regression models and life tables. J R Stat Soc, B: 187-220, 1972. 117. Camp, R. L., Dolled-Filhart, M., and Rimm, D. L. X-tile: a new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization. Clin Cancer Res, 10: 7252- 7259, 2004. 118. Siegmund, M. a. Maximally selected chi square statistics. Biometrics, 38: 1011-1016, 1982. 119. Yeoh, E. J., Ross, M. E., Shurtleff, S. A., Williams, W. K., Patel, D., Mahfouz, R., Behm, F. G., Raimondi, S. C., Relling, M. V., Patel, A., Cheng, C., Campana, D., Wilkins, D., Zhou, X., Li, J., Liu, H., Pui, C. H., Evans, W. E., Naeve, C., Wong, L., and Downing, J. R. Classification, subtype discovery, and prediction of outcome in pediatric acute lymphoblastic leukemia by gene expression profiling. Cancer Cell, 1: 133-143, 2002. 120. Miller, R. and Siegmund, D. Maximally selected chi square statistics. Biometrics, 38: 1011- 1016, 1982. 121. Till, M. M., Hardisty, R. M., and Pike, M. C. Long survivals in acute leukaemia. Lancet, 1: 534-538, 1973. 122. Onciu, M., Lai, R., Vega, F., Bueso-Ramos, C., and Medeiros, L. J. Precursor T-cell acute lymphoblastic leukemia in adults: age-related immunophenotypic, cytogenetic, and molecular subsets. Am J Clin Pathol, 117: 252-258, 2002. 123. Fadok, V. A., Voelker, D. R., Campbell, P. A., Cohen, J. J., Bratton, D. L., and Henson, P. M. Exposure of phosphatidylserine on the surface of apoptotic lymphocytes triggers specific recognition and removal by macrophages. J Immunol, 148: 2207-2216, 1992. 124. Scannell, M., Flanagan, M. B., deStefani, A., Wynne, K. J., Cagney, G., Godson, C., and Maderna, P. Annexin-1 and peptide derivatives are released by apoptotic cells and stimulate phagocytosis of apoptotic neutrophils by macrophages. J Immunol, 178: 4595-4605, 2007. 125. Gardai, S. J., McPhillips, K. A., Frasch, S. C., Janssen, W. J., Starefeldt, A., Murphy-Ullrich, J. E., Bratton, D. L., Oldenborg, P. A., Michalak, M., and Henson, P. M. Cell-surface

128

calreticulin initiates clearance of viable or apoptotic cells through trans-activation of LRP on the phagocyte. Cell, 123: 321-334, 2005. 126. Jeha, S. New therapeutic strategies in acute lymphoblastic leukemia. Semin Hematol, 46: 76- 88, 2009. 127. Thomas, D. A., Faderl, S., O'Brien, S., Bueso-Ramos, C., Cortes, J., Garcia-Manero, G., Giles, F. J., Verstovsek, S., Wierda, W. G., Pierce, S. A., Shan, J., Brandt, M., Hagemeister, F. B., Keating, M. J., Cabanillas, F., and Kantarjian, H. Chemoimmunotherapy with hyper-CVAD plus rituximab for the treatment of adult Burkitt and Burkitt-type lymphoma or acute lymphoblastic leukemia. Cancer, 106: 1569-1580, 2006. 128. Gokbuget, N. and Hoelzer, D. Novel antibody-based therapy for acute lymphoblastic leukaemia. Best Pract Res Clin Haematol, 19: 701-713, 2006. 129. Raetz, E. A., Cairo, M. S., Borowitz, M. J., Blaney, S. M., Krailo, M. D., Leil, T. A., Reid, J. M., Goldenberg, D. M., Wegener, W. A., Carroll, W. L., and Adamson, P. C. Chemoimmunotherapy reinduction with epratuzumab in children with acute lymphoblastic leukemia in marrow relapse: a Children's Oncology Group Pilot Study. J Clin Oncol, 26: 3756- 3762, 2008. 130. Du, X., Beers, R., Fitzgerald, D. J., and Pastan, I. Differential cellular internalization of anti- CD19 and -CD22 immunotoxins results in different cytotoxic activity. Cancer Res, 68: 6300- 6305, 2008. 131. Herrera, L., Bostrom, B., Gore, L., Sandler, E., Lew, G., Schlegel, P. G., Aquino, V., Ghetie, V., Vitetta, E. S., and Schindler, J. A phase 1 study of Combotox in pediatric patients with refractory B-lineage acute lymphoblastic leukemia. J Pediatr Hematol Oncol, 31: 936-941, 2009. 132. Wayne, A. S., Kreitman, R. J., Findley, H. W., Lew, G., Delbrook, C., Steinberg, S. M., Stetler-Stevenson, M., Fitzgerald, D. J., and Pastan, I. Anti-CD22 immunotoxin RFB4(dsFv)- PE38 (BL22) for CD22-positive hematologic malignancies of childhood: preclinical studies and phase I clinical trial. Clin Cancer Res, 16: 1894-1903. 133. Uckun, F. M., Messinger, Y., Chen, C. L., O'Neill, K., Myers, D. E., Goldman, F., Hurvitz, C., Casper, J. T., and Levine, A. Treatment of therapy-refractory B-lineage acute lymphoblastic leukemia with an apoptosis-inducing CD19-directed tyrosine kinase inhibitor. Clin Cancer Res, 5: 3906-3913, 1999. 134. Gokbuget, N. and Hoelzer, D. Treatment with monoclonal antibodies in acute lymphoblastic leukemia: current knowledge and future prospects. Ann Hematol, 83: 201-205, 2004. 135. Ottmann, O. G. and Wassmann, B. Imatinib in the treatment of Philadelphia - positive acute lymphoblastic leukaemia: current status and evolving concepts. Best Pract Res Clin Haematol, 15: 757-769, 2002. 136. Thomas, D. A., Faderl, S., Cortes, J., O'Brien, S., Giles, F. J., Kornblau, S. M., Garcia-Manero, G., Keating, M. J., Andreeff, M., Jeha, S., Beran, M., Verstovsek, S., Pierce, S., Letvak, L., Salvado, A., Champlin, R., Talpaz, M., and Kantarjian, H. Treatment of Philadelphia chromosome-positive acute lymphocytic leukemia with hyper-CVAD and imatinib mesylate. Blood, 103: 4396-4407, 2004. 137. Alinari, L., Lapalombella, R., Andritsos, L., Baiocchi, R. A., Lin, T. S., and Byrd, J. C. Alemtuzumab (Campath-1H) in the treatment of chronic lymphocytic leukemia. Oncogene, 26: 3644-3653, 2007. 138. Angiolillo, A. L., Yu, A. L., Reaman, G., Ingle, A. M., Secola, R., and Adamson, P. C. A phase II study of Campath-1H in children with relapsed or refractory acute lymphoblastic leukemia: a Children's Oncology Group report. Pediatr Blood Cancer, 53: 978-983, 2009. 139. Hanahan, D. and Weinberg, R. A. The hallmarks of cancer. Cell, 100: 57-70, 2000. 140. Stagg, J., Johnstone, R. W., and Smyth, M. J. From cancer immunosurveillance to cancer immunotherapy. Immunol Rev, 220: 82-101, 2007. 141. Swann, J. B. and Smyth, M. J. Immune surveillance of tumors. J Clin Invest, 117: 1137-1146, 2007. 142. Zitvogel, L., Tesniere, A., and Kroemer, G. Cancer despite immunosurveillance: immunoselection and immunosubversion. Nat Rev Immunol, 6: 715-727, 2006.

129 143. Khong, H. T. and Restifo, N. P. Natural selection of tumor variants in the generation of "tumor escape" phenotypes. Nat Immunol, 3: 999-1005, 2002. 144. Jaiswal, S., Chao, M. P., Majeti, R., and Weissman, I. L. Macrophages as mediators of tumor immunosurveillance. Trends Immunol. 145. Tsai, R. K., Rodriguez, P. L., and Discher, D. E. Self inhibition of phagocytosis: The affinity of 'marker of self' CD47 for SIRPalpha dictates potency of inhibition but only at low expression levels. Blood Cells Mol Dis. 146. Jiang, P., Lagenaur, C. F., and Narayanan, V. Integrin-associated protein is a ligand for the P84 neural adhesion molecule. J Biol Chem, 274: 559-562, 1999. 147. Michalak, M., Groenendyk, J., Szabo, E., Gold, L. I., and Opas, M. Calreticulin, a multi- process calcium-buffering chaperone of the endoplasmic reticulum. Biochem J, 417: 651-666, 2009. 148. Obeid, M., Panaretakis, T., Joza, N., Tufi, R., Tesniere, A., van Endert, P., Zitvogel, L., and Kroemer, G. Calreticulin exposure is required for the immunogenicity of gamma-irradiation and UVC light-induced apoptosis. Cell Death Differ, 14: 1848-1850, 2007. 149. Obeid, M., Tesniere, A., Ghiringhelli, F., Fimia, G. M., Apetoh, L., Perfettini, J. L., Castedo, M., Mignot, G., Panaretakis, T., Casares, N., Metivier, D., Larochette, N., van Endert, P., Ciccosanti, F., Piacentini, M., Zitvogel, L., and Kroemer, G. Calreticulin exposure dictates the immunogenicity of cancer cell death. Nat Med, 13: 54-61, 2007. 150. Ogden, C. A., deCathelineau, A., Hoffmann, P. R., Bratton, D., Ghebrehiwet, B., Fadok, V. A., and Henson, P. M. C1q and mannose binding lectin engagement of cell surface calreticulin and CD91 initiates macropinocytosis and uptake of apoptotic cells. J Exp Med, 194: 781-795, 2001. 151. Orr, A. W., Pedraza, C. E., Pallero, M. A., Elzie, C. A., Goicoechea, S., Strickland, D. K., and Murphy-Ullrich, J. E. Low density lipoprotein receptor-related protein is a calreticulin coreceptor that signals focal adhesion disassembly. J Cell Biol, 161: 1179-1189, 2003. 152. Gardai, S. J., Bratton, D. L., Ogden, C. A., and Henson, P. M. Recognition ligands on apoptotic cells: a perspective. J Leukoc Biol, 79: 896-903, 2006. 153. Lim, L. H. and Pervaiz, S. Annexin 1: the new face of an old molecule. Faseb J, 21: 968-975, 2007. 154. Lessard, J. and Sauvageau, G. Bmi-1 determines the proliferative capacity of normal and leukaemic stem cells. Nature, 423: 255-260, 2003. 155. Reinhold, M. I., Lindberg, F. P., Plas, D., Reynolds, S., Peters, M. G., and Brown, E. J. In vivo expression of alternatively spliced forms of integrin-associated protein (CD47). J Cell Sci, 108 ( Pt 11): 3419-3425, 1995. 156. Maxhimer, J. B., Soto-Pantoja, D. R., Ridnour, L. A., Shih, H. B., Degraff, W. G., Tsokos, M., Wink, D. A., Isenberg, J. S., and Roberts, D. D. Radioprotection in normal tissue and delayed tumor growth by blockade of CD47 signaling. Sci Transl Med, 1: 3ra7, 2009. 157. Arosa, F. A., de Jesus, O., Porto, G., Carmo, A. M., and de Sousa, M. Calreticulin is expressed on the cell surface of activated human peripheral blood T lymphocytes in association with major histocompatibility complex class I molecules. J Biol Chem, 274: 16917-16922, 1999. 158. Ghiran, I., Klickstein, L. B., and Nicholson-Weller, A. Calreticulin is at the surface of circulating neutrophils and uses CD59 as an adaptor molecule. J Biol Chem, 278: 21024- 21031, 2003. 159. Park, Y. J., Liu, G., Lorne, E. F., Zhao, X., Wang, J., Tsuruta, Y., Zmijewski, J., and Abraham, E. PAI-1 inhibits neutrophil efferocytosis. Proc Natl Acad Sci U S A, 105: 11784-11789, 2008. 160. Lawrence, D. W., King, S. B., Frazier, W. A., and Koenig, J. M. Decreased CD47 expression during spontaneous apoptosis targets neutrophils for phagocytosis by -derived macrophages. Early Hum Dev, 85: 659-663, 2009. 161. Connor, J., Bucana, C., Fidler, I. J., and Schroit, A. J. Differentiation-dependent expression of phosphatidylserine in mammalian plasma membranes: quantitative assessment of outer-leaflet lipid by prothrombinase complex formation. Proc Natl Acad Sci U S A, 86: 3184-3188, 1989.

130

162. Utsugi, T., Schroit, A. J., Connor, J., Bucana, C. D., and Fidler, I. J. Elevated expression of phosphatidylserine in the outer membrane leaflet of human tumor cells and recognition by activated human blood . Cancer Res, 51: 3062-3066, 1991. 163. Rao, L. V., Tait, J. F., and Hoang, A. D. Binding of annexin V to a human ovarian carcinoma cell line (OC-2008). Contrasting effects on cell surface factor VIIa/ activity and prothrombinase activity. Thromb Res, 67: 517-531, 1992. 164. Woehlecke, H., Pohl, A., Alder-Baerens, N., Lage, H., and Herrmann, A. Enhanced exposure of phosphatidylserine in human gastric carcinoma cells overexpressing the half-size ABC transporter BCRP (ABCG2). Biochem J, 376: 489-495, 2003. 165. Zwaal, R. F., Comfurius, P., and Bevers, E. M. Surface exposure of phosphatidylserine in pathological cells. Cell Mol Life Sci, 62: 971-988, 2005. 166. Strickland, D. K., Kounnas, M. Z., and Argraves, W. S. LDL receptor-related protein: a multiligand receptor for lipoprotein and proteinase catabolism. Faseb J, 9: 890-898, 1995. 167. Liu, M., Imam, H., Oberg, K., and Zhou, Y. Gene transfer of vasostatin, a calreticulin fragment, into neuroendocrine tumor cells results in enhanced malignant behavior. Neuroendocrinology, 82: 1-10, 2005. 168. Chen, C. N., Chang, C. C., Su, T. E., Hsu, W. M., Jeng, Y. M., Ho, M. C., Hsieh, F. J., Lee, P. H., Kuo, M. L., Lee, H., and Chang, K. J. Identification of calreticulin as a prognosis marker and angiogenic regulator in human gastric cancer. Ann Surg Oncol, 16: 524-533, 2009. 169. Pike, S. E., Yao, L., Setsuda, J., Jones, K. D., Cherney, B., Appella, E., Sakaguchi, K., Nakhasi, H., Atreya, C. D., Teruya-Feldstein, J., Wirth, P., Gupta, G., and Tosato, G. Calreticulin and calreticulin fragments are endothelial cell inhibitors that suppress tumor growth. Blood, 94: 2461-2468, 1999. 170. Pike, S. E., Yao, L., Jones, K. D., Cherney, B., Appella, E., Sakaguchi, K., Nakhasi, H., Teruya-Feldstein, J., Wirth, P., Gupta, G., and Tosato, G. Vasostatin, a calreticulin fragment, inhibits angiogenesis and suppresses tumor growth. J Exp Med, 188: 2349-2356, 1998. 171. Coppolino, M. G., Woodside, M. J., Demaurex, N., Grinstein, S., St-Arnaud, R., and Dedhar, S. Calreticulin is essential for integrin-mediated calcium signalling and cell adhesion. Nature, 386: 843-847, 1997. 172. Howe, C., Garstka, M., Al-Balushi, M., Ghanem, E., Antoniou, A. N., Fritzsche, S., Jankevicius, G., Kontouli, N., Schneeweiss, C., Williams, A., Elliott, T., and Springer, S. Calreticulin-dependent recycling in the early secretory pathway mediates optimal peptide loading of MHC class I molecules. Embo J, 28: 3730-3744, 2009. 173. Cheson, B. D. and Leonard, J. P. Monoclonal antibody therapy for B-cell non-Hodgkin's lymphoma. N Engl J Med, 359: 613-626, 2008. 174. A predictive model for aggressive non-Hodgkin's lymphoma. The International Non-Hodgkin's Lymphoma Prognostic Factors Project. N Engl J Med, 329: 987-994, 1993. 175. Glennie, M. J., French, R. R., Cragg, M. S., and Taylor, R. P. Mechanisms of killing by anti- CD20 monoclonal antibodies. Mol Immunol, 44: 3823-3837, 2007. 176. Cartron, G., Watier, H., Golay, J., and Solal-Celigny, P. From the bench to the bedside: ways to improve rituximab efficacy. Blood, 104: 2635-2642, 2004. 177. Fabris, S., Mosca, L., Todoerti, K., Cutrona, G., Lionetti, M., Intini, D., Matis, S., Colombo, M., Agnelli, L., Gentile, M., Spriano, M., Callea, V., Festini, G., Molica, S., Lambertenghi Deliliers, G., Morabito, F., Ferrarini, M., and Neri, A. Molecular and transcriptional characterization of 17p loss in B-cell chronic lymphocytic leukemia. Genes Chromosomes Cancer, 47: 781-793, 2008. 178. Haslinger, C., Schweifer, N., Stilgenbauer, S., Dohner, H., Lichter, P., Kraut, N., Stratowa, C., and Abseher, R. Microarray gene expression profiling of B-cell chronic lymphocytic leukemia subgroups defined by genomic aberrations and VH mutation status. J Clin Oncol, 22: 3937- 3949, 2004. 179. Lenz, G., Wright, G., Dave, S. S., Xiao, W., Powell, J., Zhao, H., Xu, W., Tan, B., Goldschmidt, N., Iqbal, J., Vose, J., Bast, M., Fu, K., Weisenburger, D. D., Greiner, T. C., Armitage, J. O., Kyle, A., May, L., Gascoyne, R. D., Connors, J. M., Troen, G., Holte, H., Kvaloy, S., Dierickx, D., Verhoef, G., Delabie, J., Smeland, E. B., Jares, P., Martinez, A., Lopez-Guillermo, A., Montserrat, E., Campo, E., Braziel, R. M., Miller, T. P., Rimsza, L. M., 131 Cook, J. R., Pohlman, B., Sweetenham, J., Tubbs, R. R., Fisher, R. I., Hartmann, E., Rosenwald, A., Ott, G., Muller-Hermelink, H. K., Wrench, D., Lister, T. A., Jaffe, E. S., Wilson, W. H., Chan, W. C., and Staudt, L. M. Stromal gene signatures in large-B-cell lymphomas. N Engl J Med, 359: 2313-2323, 2008. 180. Malumbres, R., Chen, J., Tibshirani, R., Johnson, N. A., Sehn, L. H., Natkunam, Y., Briones, J., Advani, R., Connors, J. M., Byrne, G. E., Levy, R., Gascoyne, R. D., and Lossos, I. S. Paraffin-based 6-gene model predicts outcome in diffuse large B-cell lymphoma patients treated with R-CHOP. Blood, 111: 5509-5514, 2008. 181. Rosenwald, A., Wright, G., Chan, W. C., Connors, J. M., Campo, E., Fisher, R. I., Gascoyne, R. D., Muller-Hermelink, H. K., Smeland, E. B., Giltnane, J. M., Hurt, E. M., Zhao, H., Averett, L., Yang, L., Wilson, W. H., Jaffe, E. S., Simon, R., Klausner, R. D., Powell, J., Duffey, P. L., Longo, D. L., Greiner, T. C., Weisenburger, D. D., Sanger, W. G., Dave, B. J., Lynch, J. C., Vose, J., Armitage, J. O., Montserrat, E., Lopez-Guillermo, A., Grogan, T. M., Miller, T. P., LeBlanc, M., Ott, G., Kvaloy, S., Delabie, J., Holte, H., Krajci, P., Stokke, T., and Staudt, L. M. The use of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med, 346: 1937-1947, 2002. 182. Rosenwald, A., Wright, G., Wiestner, A., Chan, W. C., Connors, J. M., Campo, E., Gascoyne, R. D., Grogan, T. M., Muller-Hermelink, H. K., Smeland, E. B., Chiorazzi, M., Giltnane, J. M., Hurt, E. M., Zhao, H., Averett, L., Henrickson, S., Yang, L., Powell, J., Wilson, W. H., Jaffe, E. S., Simon, R., Klausner, R. D., Montserrat, E., Bosch, F., Greiner, T. C., Weisenburger, D. D., Sanger, W. G., Dave, B. J., Lynch, J. C., Vose, J., Armitage, J. O., Fisher, R. I., Miller, T. P., LeBlanc, M., Ott, G., Kvaloy, S., Holte, H., Delabie, J., and Staudt, L. M. The proliferation gene expression signature is a quantitative integrator of oncogenic events that predicts survival in mantle cell lymphoma. Cancer Cell, 3: 185-197, 2003. 183. Wiestner, A., Rosenwald, A., Barry, T. S., Wright, G., Davis, R. E., Henrickson, S. E., Zhao, H., Ibbotson, R. E., Orchard, J. A., Davis, Z., Stetler-Stevenson, M., Raffeld, M., Arthur, D. C., Marti, G. E., Wilson, W. H., Hamblin, T. J., Oscier, D. G., and Staudt, L. M. ZAP-70 expression identifies a chronic lymphocytic leukemia subtype with unmutated immunoglobulin genes, inferior clinical outcome, and distinct gene expression profile. Blood, 101: 4944-4951, 2003. 184. Olson, J. A., Leveson-Gower, D. B., Gill, S., Baker, J., Beilhack, A., and Negrin, R. S. NK cells mediate reduction of GVHD by inhibiting activated, alloreactive T cells while retaining GVT effects. Blood. 185. Verneris, M. R., Karami, M., Baker, J., Jayaswal, A., and Negrin, R. S. Role of NKG2D signaling in the cytotoxicity of activated and expanded CD8+ T cells. Blood, 103: 3065-3072, 2004. 186. Tobin, E., Denardo, G., Zhang, N., Epstein, A. L., Liu, C., and Denardo, S. Combination immunotherapy with anti-CD20 and anti-HLA-DR monoclonal antibodies induces synergistic anti-lymphoma effects in human lymphoma cell lines. Leuk Lymphoma, 48: 944-956, 2007. 187. Lacerda, J. F., Ladanyi, M., Jagiello, C., and O'Reilly, R. J. Administration of rabbit anti-asialo GM1 antiserum facilitates the development of human Epstein-Barr virus-induced lymphoproliferations in xenografted C.B-17 scid/scid mice. Transplantation, 61: 492-497, 1996. 188. Cragg, M. S. and Glennie, M. J. Antibody specificity controls in vivo effector mechanisms of anti-CD20 reagents. Blood, 103: 2738-2743, 2004. 189. O'Connell-Rodwell, C. E., Mackanos, M. A., Simanovskii, D., Cao, Y. A., Bachmann, M. H., Schwettman, H. A., and Contag, C. H. In vivo analysis of heat-shock-protein-70 induction following pulsed laser irradiation in a transgenic reporter mouse. J Biomed Opt, 13: 030501, 2008. 190. Wu, J. C., Sundaresan, G., Iyer, M., and Gambhir, S. S. Noninvasive optical imaging of firefly luciferase reporter gene expression in skeletal muscles of living mice. Mol Ther, 4: 297-306, 2001. 191. Alizadeh, A. A., Eisen, M. B., Davis, R. E., Ma, C., Lossos, I. S., Rosenwald, A., Boldrick, J. C., Sabet, H., Tran, T., Yu, X., Powell, J. I., Yang, L., Marti, G. E., Moore, T., Hudson, J., Jr., Lu, L., Lewis, D. B., Tibshirani, R., Sherlock, G., Chan, W. C., Greiner, T. C., Weisenburger, 132

D. D., Armitage, J. O., Warnke, R., Levy, R., Wilson, W., Grever, M. R., Byrd, J. C., Botstein, D., Brown, P. O., and Staudt, L. M. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature, 403: 503-511, 2000. 192. Shipp, M. A., Ross, K. N., Tamayo, P., Weng, A. P., Kutok, J. L., Aguiar, R. C., Gaasenbeek, M., Angelo, M., Reich, M., Pinkus, G. S., Ray, T. S., Koval, M. A., Last, K. W., Norton, A., Lister, T. A., Mesirov, J., Neuberg, D. S., Lander, E. S., Aster, J. C., and Golub, T. R. Diffuse large B-cell lymphoma outcome prediction by gene-expression profiling and supervised machine learning. Nat Med, 8: 68-74, 2002. 193. Oscier, D. G., Gardiner, A. C., Mould, S. J., Glide, S., Davis, Z. A., Ibbotson, R. E., Corcoran, M. M., Chapman, R. M., Thomas, P. W., Copplestone, J. A., Orchard, J. A., and Hamblin, T. J. Multivariate analysis of prognostic factors in CLL: clinical stage, IGVH gene mutational status, and loss or mutation of the p53 gene are independent prognostic factors. Blood, 100: 1177-1184, 2002. 194. Katzenberger, T., Petzoldt, C., Holler, S., Mader, U., Kalla, J., Adam, P., Ott, M. M., Muller- Hermelink, H. K., Rosenwald, A., and Ott, G. The Ki67 proliferation index is a quantitative indicator of clinical risk in mantle cell lymphoma. Blood, 107: 3407, 2006. 195. Chou, T. C. Preclinical versus clinical drug combination studies. Leuk Lymphoma, 49: 2059- 2080, 2008. 196. Tallarida, R. J. An overview of drug combination analysis with isobolograms. J Pharmacol Exp Ther, 319: 1-7, 2006. 197. Shultz, L. D., Lyons, B. L., Burzenski, L. M., Gott, B., Chen, X., Chaleff, S., Kotb, M., Gillies, S. D., King, M., Mangada, J., Greiner, D. L., and Handgretinger, R. Human lymphoid and myeloid cell development in NOD/LtSz-scid IL2R gamma null mice engrafted with mobilized human hemopoietic stem cells. J Immunol, 174: 6477-6489, 2005. 198. Taylor, R. P. and Lindorfer, M. A. Immunotherapeutic mechanisms of anti-CD20 monoclonal antibodies. Curr Opin Immunol, 20: 444-449, 2008. 199. Smith, M. R. Rituximab (monoclonal anti-CD20 antibody): mechanisms of action and resistance. Oncogene, 22: 7359-7368, 2003. 200. Takai, T., Li, M., Sylvestre, D., Clynes, R., and Ravetch, J. V. FcR gamma chain deletion results in pleiotrophic effector cell defects. Cell, 76: 519-529, 1994. 201. Kim, M. J., Lee, J. C., Lee, J. J., Kim, S., Lee, S. G., Park, S. W., Sung, M. W., and Heo, D. S. Association of CD47 with -mediated cytotoxicity of head-and-neck squamous cell carcinoma lines. Tumour Biol, 29: 28-34, 2008. 202. Milush, J. M., Long, B. R., Snyder-Cappione, J. E., Cappione, A. J., 3rd, York, V. A., Ndhlovu, L. C., Lanier, L. L., Michaelsson, J., and Nixon, D. F. Functionally distinct subsets of human NK cells and monocyte/DC-like cells identified by coexpression of CD56, CD7, and CD4. Blood, 114: 4823-4831, 2009. 203. Foran, J. M., Norton, A. J., Micallef, I. N., Taussig, D. C., Amess, J. A., Rohatiner, A. Z., and Lister, T. A. Loss of CD20 expression following treatment with rituximab (chimaeric monoclonal anti-CD20): a retrospective cohort analysis. Br J Haematol, 114: 881-883, 2001. 204. Kennedy, A. D., Beum, P. V., Solga, M. D., DiLillo, D. J., Lindorfer, M. A., Hess, C. E., Densmore, J. J., Williams, M. E., and Taylor, R. P. Rituximab infusion promotes rapid complement depletion and acute CD20 loss in chronic lymphocytic leukemia. J Immunol, 172: 3280-3288, 2004. 205. Hiraga, J., Tomita, A., Sugimoto, T., Shimada, K., Ito, M., Nakamura, S., Kiyoi, H., Kinoshita, T., and Naoe, T. Down-regulation of CD20 expression in B-cell lymphoma cells after treatment with rituximab-containing combination chemotherapies: its prevalence and clinical significance. Blood, 113: 4885-4893, 2009. 206. Finn, O. J. Cancer . N Engl J Med, 358: 2704-2715, 2008. 207. Le, T., Williams, K., Senterman, M., Hopkins, L., Faught, W., and Fung-Kee-Fung, M. Histopathologic assessment of chemotherapy effects in epithelial ovarian cancer patients treated with neoadjuvant chemotherapy and delayed primary surgical debulking. Gynecol Oncol, 106: 160-163, 2007.

133 208. Welsh, T. J., Green, R. H., Richardson, D., Waller, D. A., O'Byrne, K. J., and Bradding, P. Macrophage and mast-cell invasion of tumor cell islets confers a marked survival advantage in non-small-cell . J Clin Oncol, 23: 8959-8967, 2005. 209. Funada, Y., Noguchi, T., Kikuchi, R., Takeno, S., Uchida, Y., and Gabbert, H. E. Prognostic significance of CD8+ T cell and macrophage peritumoral infiltration in colorectal cancer. Oncol Rep, 10: 309-313, 2003. 210. Charak, B. S., Agah, R., and Mazumder, A. Granulocyte-macrophage colony-stimulating factor-induced antibody-dependent cellular cytotoxicity in bone marrow macrophages: application in bone marrow transplantation. Blood, 81: 3474-3479, 1993. 211. Hernandez-Ilizaliturri, F. J., Jupudy, V., Reising, S., Repasky, E. A., and Czuczman, M. S. Concurrent administration of granulocyte colony-stimulating factor or granulocyte-monocyte colony-stimulating factor enhances the biological activity of rituximab in a severe combined immunodeficiency mouse lymphoma model. Leuk Lymphoma, 46: 1775-1784, 2005. 212. Dale, D. C. Colony-stimulating factors for the management of neutropenia in cancer patients. Drugs, 62 Suppl 1: 1-15, 2002. 213. BitMansour, A., Burns, S. M., Traver, D., Akashi, K., Contag, C. H., Weissman, I. L., and Brown, J. M. Myeloid progenitors protect against invasive aspergillosis and Pseudomonas aeruginosa infection following hematopoietic stem cell transplantation. Blood, 100: 4660- 4667, 2002. 214. Arber, C., Bitmansour, A., Shashidhar, S., Wang, S., Tseng, B., and Brown, J. M. Protection against lethal Aspergillus fumigatus infection in mice by allogeneic myeloid progenitors is not major histocompatibility complex restricted. J Infect Dis, 192: 1666-1671, 2005.

134