CD47 blockade augmentation of Trastuzumab antitumor efficacy dependent upon -dependent-cellular-

Li-Chung Tsao, … , Herbert Kim Lyerly, Zachary C. Hartman

JCI Insight. 2019. https://doi.org/10.1172/jci.insight.131882.

Research In-Press Preview Immunology Oncology

The HER2-specific (mAb), Trastuzumab, has been the mainstay of therapy for HER2+ breast cancers (BC) for ~20 years. However, its therapeutic mechanism of action (MOA) remains unclear, with antitumor responses to Trastuzumab remaining heterologous and metastatic HER2+ BC remaining incurable. Consequently, understanding its MOA could enable rational strategies to enhance its efficacy. Using both novel murine and human versions of Trastuzumab, we found its antitumor activity dependent on Fcγ-Receptor stimulation of tumor-associated- (TAM) and Antibody-Dependent-Cellular-Phagocytosis (ADCP), but not cytotoxicity (ADCC). Trastuzumab also stimulated TAM activation and expansion, but did not require adaptive immunity, natural killer cells, and/or . Moreover, inhibition of the innate immune ADCP checkpoint, CD47, significantly enhanced Trastuzumab- mediated ADCP, TAM expansion and activation, resulting in the emergence of a unique hyper-phagocytic population, improved antitumor responses and prolonged survival. In addition, we found tumor-associated CD47 expression was inversely associated with survival in HER2+ BC patients and that human HER2+ BC xenografts treated with Trastuzumab+CD47 inhibition underwent complete tumor regression. Collectively, our study identifies Trastuzumab- mediated ADCP as a significant antitumor MOA that may be clinically enabled by CD47 blockade to augment therapeutic efficacy.

Find the latest version: https://jci.me/131882/pdf 1 CD47 blockade augmentation of Trastuzumab antitumor efficacy dependent upon 2 antibody-dependent-cellular-phagocytosis 3 4 Li-Chung Tsao1, Erika J. Crosby1, Timothy N. Trotter1, Pankaj Agarwal1, Bin-Jin Hwang1, 5 Chaitanya Acharya1, Casey W. Shuptrine1, Tao Wang1, Junping Wei1, Xiao Yang1, Gangjun 6 Lei1, Cong-Xiao Liu1, Christopher A. Rabiola1, Lewis A. Chodosh2, William J. Muller3, Herbert 7 Kim Lyerly1,4,5, Zachary C. Hartman1,5

8 1Department of Surgery, Duke University, Durham NC, USA

9 2Department of Cancer Biology, University of Pennsylvania, Philadelphia PA, USA

10 3Department of Biochemistry, McGill University, Montreal Quebec, Canada

11 4Department of Immunology, Duke University, Durham NC, USA

12 5Department of Pathology, Duke University, Durham NC, USA

13

14 Keywords: Breast cancer, HER2, Trastuzumab, antibody therapeutics, ADCP, Macrophages, 15 CD47, checkpoint blockade,

16

17 Corresponding Author:

18 Zachary C. Hartman

19 Department of Surgery and Pathology

20 Division of Surgical Sciences, Center for Applied Therapeutics

21 Duke University Medical Center

22 MSRBI Room 414 Box 2606, 203 Research Drive, Durham NC 27710

23 Phone- 919-613-9110

24 [email protected]

25

26 27 COI Statement:

28 The authors declare no potential conflicts of interest. An invention disclosure at Duke 29 University’s Office of Licensing and Ventures has been generated for “identifying phagocytic 30 macrophage as a biomarker” by the following authors: LT, ZH and HKL. Patent # T-006575. 31 Abstract

32 The HER2-specific monoclonal antibody (mAb), Trastuzumab, has been the mainstay of

33 therapy for HER2+ breast cancers (BC) for ~20 years. However, its therapeutic mechanism of

34 action (MOA) remains unclear, with antitumor responses to Trastuzumab remaining

35 heterologous and metastatic HER2+ BC remaining incurable. Consequently, understanding its

36 MOA could enable rational strategies to enhance its efficacy. Using both novel murine and

37 human versions of Trastuzumab, we found its antitumor activity dependent on Fcγ-Receptor

38 stimulation of tumor-associated-macrophages (TAM) and Antibody-Dependent-Cellular-

39 Phagocytosis (ADCP), but not cytotoxicity (ADCC). Trastuzumab also stimulated TAM activation

40 and expansion, but did not require adaptive immunity, natural killer cells, and/or neutrophils.

41 Moreover, inhibition of the innate immune ADCP checkpoint, CD47, significantly enhanced

42 Trastuzumab-mediated ADCP, TAM expansion and activation, resulting in the emergence of a

43 unique hyper-phagocytic macrophage population, improved antitumor responses and prolonged

44 survival. In addition, we found tumor-associated CD47 expression was inversely associated with

45 survival in HER2+ BC patients and that human HER2+ BC xenografts treated with

46 Trastuzumab+CD47 inhibition underwent complete tumor regression. Collectively, our study

47 identifies Trastuzumab-mediated ADCP as a significant antitumor MOA that may be clinically

48 enabled by CD47 blockade to augment therapeutic efficacy.

49

50

51

52

53 54 Introduction:

55 Approximately 20% of Breast Cancer (BC) overexpress HER2, recognized as an

56 oncogenic driver of an aggressive cancer phenotype with a poor prognosis (1, 2). Monoclonal

57 (mAbs) targeting HER2 were developed in the 1980s to inhibit HER2 oncogenic

58 signaling, leading to the clinical development and regulatory approval of Trastuzumab in 1998

59 for metastatic HER2 overexpressed BC, followed by clinical trials of Trastuzumab use in the

60 adjuvant setting. Following its approval, additional HER2 targeting mAbs have also been

61 generated to improve outcomes (3, 4). However, the clinical benefit associated with HER2 mAb

62 therapies in patients with HER2 overexpressing BC remains heterologous and metastatic

63 HER2+ BC remains incurable (5, 6). Consequently, mechanistic studies of the antitumor

64 mechanism(s) of action (MOA) of Trastuzumab and its resistance remain critical, not only to

65 improve outcomes in patients with HER2+ BC, but also to gain insight into mechanisms that

66 would extend mAb therapies to other types of cancers.

67 While suppression of HER2 signaling was a primary focus of early mechanistic studies,

68 subsequent studies also focused on the role of immunity in mediating the antitumor effects of

69 Trastuzumab (7). In particular, studies have shown the interaction of anti-HER2 antibodies with

70 Fcγ-receptors (FCGR) expressed on innate immune cells such as macrophages, ,

71 natural killer (NK) cells and dendritic cells may be involved in its therapeutic activity (8, 9). The

72 consequences of crosstalk with FCGR-bearing immune cells (8-10) are supported by the clinical

73 observation that some host FCGR polymorphisms are associated with improved clinical

74 outcome in HER2+ BC patients treated with Trastuzumab (11). Specifically, several studies

75 have suggested the importance of these receptors in mediating Antibody-Dependent-Cellular-

76 Cytotoxicity (ADCC), through NK cells or neutrophils for Trastuzumab efficacy (8, 9, 12-14).

77 However, other studies have suggested the importance of adaptive immunity in mediating

78 Trastuzumab efficacy, indicating that T cells may be critical for its antitumor MOA (8, 15). 79 While multiple MOAs involving either innate or adaptive immunity are possible, an

80 underexplored mechanism is through mAb engagement of FCGRs to stimulate macrophage

81 mediated Antibody- Dependent-Cellular-Phagocytosis (ADCP). Inconsistent reports about the

82 role of ADCP exist, with a recent study demonstrating the ability of Trastuzumab to elicit ADCP

83 (16), while another study suggests that Trastuzumab-mediated ADCP triggers macrophage

84 immunosuppression in HER2+ BC (17). These disparate results may be partially attributed to

85 the use of a wide range of tumor models (many not specifically driven by active HER2-

86 signaling), as well as the use of different HER2-specific mAb clones of varied isotypes, which

87 can elicit a range of different responses from various FCGRs (18, 19). Thus, the immunologic

88 basis for the activity of Trastuzumab remains inconclusive, but could be effectively investigated

89 through the development and use of appropriate HER2 targeting mAbs and model systems.

90 In this study, we developed and utilized fully murinized Trastuzumab mAbs (clone 4D5)

91 with isotypes of different activating-to-inhibitory ratio (A/I ratio, calculated by dividing the affinity

92 of a specific IgG isotype for an activating receptor by the affinity for the inhibitory receptor) (19),

93 as well as clinical-grade Trastuzumab, to determine the MOA for Trastuzumab antitumor

94 efficacy. These mAbs were tested in multiple settings to interrogate ADCC and ADCP, as well

95 as the impact on HER2 signaling and complement-dependent cytotoxicity (CDC). To determine

96 the antitumor efficacy of these HER2 mAbs, we employed orthotopic implantation of HER2+

97 murine BC cells (transformed using a constitutively active isoform of human HER2) in

98 immunocompetent models, as well as Fcgr-/-, immune-deficient backgrounds, and human

99 HER2+ BC xenograft models. In addition, we utilized a novel transgenic HER2+ BC model

100 driven by an oncogenic isoform of human HER2 to simulate an endogenous mammary tumor

101 immune microenvironment (20, 21). Collectively, these studies revealed an essential role for

102 tumor-associated macrophages (TAMs) in mediating the therapeutic activity of Trastuzumab

103 through promoting ADCP of HER2+ tumor cells without evidence for significant induction of 104 adaptive responses against HER2. We also observed that this effect was subverted by

105 innate mechanisms of immunosuppression in the that limit

106 macrophage ADCP.

107 Previous studies have demonstrated that ADCP is principally regulated by anti-

108 phagocytic “don’t eat me” signals that are amplified in many cancers (22, 23). Chief among

109 these is CD47, which has been shown to be highly expressed in different cancers and functions

110 to suppress phagocytosis through binding to and triggering signaling of macrophage SIRPα (23,

111 24). Notably, CD47 expression is also upregulated in BC (25). As a potential means to subvert

112 innate immune regulation and enhance ADCP and possibly alter the macrophage phenotype in

113 HER2+ BC, we also targeted the CD47-SIRPα innate immune checkpoint. In this study, we

114 demonstrate that TAM ADCP can be significantly enhanced by blocking the CD47-SIRPα

115 checkpoint to enable Trastuzumab-mediated macrophage phagocytosis of HER2+ tumor cells.

116 Collectively, these findings support the importance of the ADCP MOA, as well as suggest the

117 therapeutic potential of utilizing CD47-SIRPα checkpoint blockade in combination with

118 Trastuzumab in HER2+ BC and potentially in other resistant HER2+ cancers (i.e. gastric,

119 bladder, etc) (26).

120

121 Results:

122 Generation of murine Trastuzumab (4D5) and its antitumor dependence on ADCP by

123 tumor-associated macrophages (TAMs).

124 Trastuzumab was based on a HER2-specific mouse IgG1 monoclonal antibody (4D5-IgG1, low

125 A/I ratio), which was subsequently ‘humanized’ to a human IgG1 isotype (high A/I ratio) that

126 allows for superior activation of Fc receptors (27). Thus, to accurately study the function of

127 Trastuzumab in an immunocompetent mouse model, we constructed a murine 4D5 monoclonal 128 antibody, but using the IgG2A isotype (4D5-IgG2A, high A/I ratio, Figure 1A) to better

129 approximate an Fc-receptor activating ‘murine version' of Trastuzumab (18, 19, 28).

130 Unsurprisingly, we found that 4D5-IgG2A HER2 binding is equivalent to Trastuzumab (Figure

131 S1A). This allowed us to interrogate the importance of the HER2-antibody Fc region as well as

132 minimize the humoral immune responses against Trastuzumab, a human antibody, when

133 administered into a murine host (Figure S1B).

134 To test the antitumor efficacy of 4D5-IgG2A, we began by interrogating its impact on

135 oncogenic HER2 signaling. As HER2 is weakly transformative in most cell lines, we employed a

136 highly oncogenic isoform of human HER2 (HER2Δ16) that constitutively dimerizes to create a

137 transformed BALB/c mammary cell line dependent upon HER2 signaling (21). In studies using

138 HER2Δ16, we observed that both 4D5-IgG2A and Trastuzumab could suppress HER2 signaling

139 (although not as potent as Lapatinib (Figure S1C-D), but not significantly enough to prevent

140 tumor in vitro (Figure S1E). This is in line with several recent studies, suggesting that

141 the impact of Trastuzumab is mediated through immune based mechanisms (29, 30). Using

142 transformed MM3MG-HER2Δ16 as a model for HER2-driven BC growth in vivo, we next

143 implanted these cells in the mammary fat pad of immunocompetent BALB/c mice. Tumor

144 bearing mice were treated weekly with 4D5-IgG2A or clinical-grade Trastuzumab to determine if

145 they could suppress tumor growth in an immunocompetent context. We found that both 4D5-

146 IgG2A and Trastuzumab significantly suppressed HER2+ BC growth demonstrating that murine

147 IgG2A was capable of significant antitumor activity (Figure 1B). Notably, we observed that 4D5-

148 IgG2A and Trastuzumab significantly increased the levels of tumor-associated macrophages

149 (TAMs) (Figure 1C), but did not increase other immune infiltrates such as NK cells and T cells

150 (Figure S2A). Furthermore, using IFN-γ ELISPOT assays we found 4D5-IgG2A and

151 Trastuzumab treatment had no effect on systemic adaptive T cells responses against human

152 HER2 (Figure S2B-C). In agreement with published reports (12), we observed NK cell- 153 mediated ADCC was increased by 4D5 or Trastuzumab treatment in co-culture systems (Figure

154 S2D). To determine if NK cells and/or adaptive immune cells mediate antitumor immunity in

155 vivo, we next tested HER2 mAb ability to suppress HER2+ BC growth in T cell, , and NK

156 cell deficient SCID-Beige mice. Contrary to published reports (8), we found surprisingly no

157 change in its antitumor efficacy (Figure 1D), suggesting the roles of adaptive immune and NK

158 cells are minimal in Trastuzumab/4D5 action in our in vivo model system. As levels

159 (LY6G+ CD11b+) were suppressed (Figure S2A) and previous studies have also implicated

160 neutrophils in Trastuzumab-mediated immunity (14), we next depleted neutrophils using anti-

161 LY6G in SCID-Beige studies (Figure S3D-E), but did not observe any difference in antitumor

162 efficacy (Figure 1E). To investigate the possible role of complement dependent cytotoxicity

163 (CDC), we performed CDC assays in vitro and found that neither 4D5-IgG2A nor Trastuzumab

164 were able to induce CDC in comparison to polyclonal HER2 Abs, in line with other studies of

165 Trastuzumab (31) (Figure S2E).

166 The increase of TAMs levels after treatment suggested a functional role in Trastuzumab

167 antitumor immunity. We therefore implemented several strategies to deplete macrophages in

168 our SCID-beige HER2+ BC model. Using a prolonged anti-CSF1R antibody injection strategy

169 (32), we achieved significant reduction of TAMs and which also limited TAM increase in 4D5-

170 treated tumors (Figure 1F). Importantly, the reduction of TAM levels resulted in a significant

171 decrease of HER2 mAb therapeutic efficacy (Figure 1G). We also utilized clodronate liposome

172 injection to deplete macrophages in this model, but found we could only readily deplete

173 macrophages in systemic circulation and not those in the tumor (Figure S3A). Interestingly, this

174 depletion had no effect on HER2 mAb therapy (Figure S3B), suggesting that macrophages in

175 the mammary tumor are the major antitumor effectors. To explore the efficacy of macrophage-

176 mediated HER2-specific antitumor activity, we established a BMDM co-culture system to

177 investigate the relative ADCC and ADCP activity mediated by 4D5-IgG2A and Trastuzumab 178 (33). Using Latrunculin A, an inhibitor of actin polymerization and therefore blocking

179 phagocytosis of immune complexes (34), we revealed the dominant antitumor activity of HER2

180 mAbs mediated by macrophages is through ADCP (Figure 1H). Concanamycin A, an V-ATPase

181 inhibitor reported to also inhibit perforin and cytotoxicity (35), had no effect on HER2 mAb

182 activities. Collectively, these results suggested that Trastuzumab therapy modifies the tumor

183 microenvironment by promoting TAM expansion, and that the dominant mechanism of action by

184 Trastuzumab is mediated by ADCP of HER2+ tumor cells by macrophages.

185

186 The ADCP activity of 4D5 requires the engagement with Fcγ-Receptors and is isotype

187 dependent

188 To further validate the mechanism of ADCP by 4D5-IgG2A treatment, we utilized Fcer1g-/-

189 animals to test the requirement for Fcγ-receptor (FCGR) engagement on phagocytic immune

190 cells. Using macrophages cultured from Fcer1g-/- and control mice, in vitro ADCP assays

191 revealed that FCGRs on macrophages are critical for 4D5-induced ADCP of HER2+ BC (Figure

192 2A). Accordingly, we found the in vivo antitumor efficacy of 4D5-IgG2A therapy are mostly

193 ablated in Fcer1g-/- mice (Figure 2B). Importantly, FCGR expression was also required for

194 macrophage expansion by 4D5-IgG2A in the tumor microenvironment (Figure 2C).

195 These data demonstrated that HER2 mAb engagement with macrophage FCGRs is required for

196 ADCP activity. Among the four mouse FCGRs, FCGR4 is the predominant FCGR mediating

197 macrophage ADCP, plays a central role for mouse IgG2A activity, and has also been shown to

198 exhibit the strongest binding affinity for Trastuzumab (16, 36-38). To determine the impact of

199 HER2-mAb isotype on FCGR4 engagement and antitumor efficacy, we compared the efficacy of

200 4D5-IgG1 (low A/I ratio) and compared its antitumor efficacy with 4D5-IgG2A (Figure 1A). We

201 found that unlike 4D5-IgG2A which elicited significant antitumor effects in vivo and ADCP in 202 vitro, 4D5-IgG1 has no effect against HER2+ BC in vivo (Figure 2D) and was inferior in

203 promoting tumor ADCP by BMDM (Figure 2E). To determine their impact on FCGR4 and other

204 activating FCGRs directly, we developed a mouse FCGR activation and signaling to NFAT-

205 luciferase reporter system based on published methods (39). In agreement with established

206 literatures on mouse IgG subclasses and FCGR biology (18, 19, 40), we found that 4D5-IgG2A

207 engages with all three activating FCGRs, whereas 4D5-IgG1 only weakly activates FCGR3

208 (Figure 2F-2H). Additionally, mouse FCGR1 and FCGR4 have strong human-murine cross-

209 reactivity with clinical grade human Trastuzumab (human IgG1 isotype) as reported before (40),

210 thus potentially explaining its in vivo efficacy in mice. Collectively, these results illustrate that

211 HER2 mAb’s antitumor activity requires the successful engagement and activation of Fcγ-

212 receptors on macrophages to induce ADCP.

213

214 CD47 Blockade increased therapeutic efficacy of 4D5 and augments tumor-associated

215 macrophage expansion and phagocytosis.

216 Our findings strongly supported an ADCP MOA for Trastuzumab antitumor efficacy, which

217 suggests strategies to enhance ADCP may be synergistic with Trastuzumab therapies. As

218 previous studies have demonstrated that blockade of CD47-SIRPα can enhance mAb

219 therapeutic efficacy, we investigated if targeting this ADCP-specific axis would enhance HER2

220 mAb ADCP without affecting ADCC activity. To begin our investigation, we documented the

221 elevated expression of Cd47 in our MM3MG-HER2Δ16 tumors and generated CD47-KO cells

222 (Figure S4A) to determine the contribution of this axis to ADCP and ADCC in vitro. We observed

223 that CD47-KO tumor cells exhibited generally enhanced ADCP that was significantly enhanced

224 by HER2 mAbs, but had no effect on ADCC (Figure 3A). Additionally, we found that 4D5-

225 mediated ADCP of CD47-KO tumors elicited the expression of pro-inflammatory cytokines and

226 chemokines by macrophages (e.g. IL6, TNFα, CCL3, CCL4 etc.), presumably due to enhanced 227 ADCP activity (Figure 3B and Figure S5). This demonstrates that 4D5-IgG2A alone triggers

228 ADCP but was insufficient to stimulate significant pro-inflammatory activation within

229 macrophages. However, upon blockade of the CD47 negative regulatory axis, ADCP and an

230 associated pro-inflammatory phenotype was significantly enhanced in macrophages.

231 As CD47 directly altered ADCP and macrophage activation in vitro, we next evaluated the

232 impact of CD47-KO expression on tumor growth and HER2 mAb therapy in vivo. We found that

233 CD47-KO HER2+ BC cells showed a delayed growth when implanted into mice, and were

234 significantly more susceptible to 4D5-IgG2A inhibition (Figure 3C). Furthermore, we found

235 significantly elevated TAM levels in CD47-KO tumors compared to the control tumors after 4D5-

236 IgG2A treatment (Figure 3D). In a reciprocal approach, we overexpressed Cd47 in the tumor

237 cells (Figure S4B) and found this increased tumor resistance to 4D5-IgG2A therapy (Figure 3E)

238 and prevented TAMs increase (Figure 3F). These two genetic approaches validated the role of

239 CD47 in suppressing Trastuzumab ADCP-mediated antitumor activity, and suggest blockade of

240 CD47 could unleash the full potential of Trastuzumab therapeutic efficacy by altering

241 macrophage activation and expansion.

242 As recent studies have suggested CD47 blockade antibodies can elicit clinical responses

243 (41), we next wanted to determine if CD47 blockade may enhance Trastuzumab efficacy. Thus,

244 we combined 4D5-IgG2A mAb with CD47 blockade antibody MIAP410 in immunocompetent

245 mice bearing the MM3MG-HER2Δ16 tumors. While 4D5-IgG2A and CD47 blockade

246 monotherapies both showed therapeutic efficacy, their combination significantly suppressed

247 tumor growth more effectively than either 4D5-IgG2A or CD47 alone and also further increased

248 TAM levels (Figure 4A and 4B and S7). In contrast, we observed that levels of other infiltrating

249 immune cell types, except for regulatory T cells, were not significantly increased by weekly

250 treatment of 4D5-IgG2A with CD47 blockade (Figure S6A). As regulatory T cells were altered,

251 we speculated that adaptive immune responses could also play a role in these enhanced 252 responses. To explore the impact of adaptive immunity in the context of CD47 blockade, we

253 repeated our in vivo experiments in adaptive immune-deficient SCID-Beige mice (Figure 4C). As

254 before, we observed a strong combinatorial effect between HER2 mAb and CD47 blockade,

255 suggesting adaptive immunity and NK cells were not essential to the enhanced response with

256 this combination therapy. Also as before, we found that CD47 blockade with 4D5-IgG2A further

257 increased TAMs levels (Figure 4D), suggesting that relieving the CD47 checkpoint specifically

258 promotes macrophage expansion and phagocytosis in tumors.

259 In order to directly demonstrate tumor ADCP by endogenous macrophages in the tumor

260 microenvironment, we labeled MM3MG-HER2Δ16 tumor cells with DiD dye (a carbocyanine

261 membrane-binding probe) prior to implantation, a strategy to detect phagocytosis of labeled

262 target cells in vivo (42). When the tumors reached a volume of ~1000 mm3, we treated the

263 animals with 4D5-IgG2A antibody or in combination with CD47 blockade (Figure 4E). FACS

264 analysis showed increased phagocytosis of labeled tumor cells by TAMs in 4D5-IgG2A treated

265 animals (Figure 4F), directly demonstrating 4D5-IgG2A treatment promotes ADCP of HER2+

266 tumor cells in vivo. Furthermore, we found the addition of CD47 blockade further increased

267 ADCP of labeled tumor cells by TAMs (Figure 4F). As expected, this therapeutic mechanism

268 requires the engagement with FCGRs on macrophages, since 4D5+αCD47 induced ADCP of

269 tumor cells in vivo was completely abolished in Fcer1g-KO mice (Figure 4G). In sum, these

270 studies demonstrate that HER2 mAb stimulates ADCP from endogenous TAMs against HER2+

271 BC, which can be boosted via combination with CD47 blockade therapy.

272

273 CD47 blockade synergizes 4D5 therapeutic activity in a transgenic HER2+ breast cancer

274 mouse model 275 Having demonstrated efficacy in an orthotopic model of HER2+ BC, we wanted to extend our

276 study using a spontaneous model of HER2+ BC that approximates a late stage HER2+ BC

277 (where HER2 mAbs are not highly effective) (43). Analogous to a clinical trial (Figure 5A), the

278 individual animals with palpable breast tumors (~200mm3) were enrolled in a specific treatment

279 group. We found that mice in the 4D5-IgG2A monotherapy treatment group had a significant

280 increase in survival time and delayed tumor growth, whereas CD47 blockade monotherapy had

281 no significant effect compared to the control group (Figure 5B and 5C). Strikingly, combination

282 therapy of 4D5-IgG2A with CD47 blockade resulted in a further prolonged survival rate and

283 delayed tumor growth compared to 4D5 monotherapy, suggesting that this combination may be

284 efficacious in advanced HER2+ BCs. To determine if these therapies again alter the immune

285 infiltrates, we analyzed the composition of the tumor microenvironment by flow cytometry. As

286 before, we found an increase in TAMs within the 4D5-IgG2A monotherapy group, whereas the

287 combination therapy group showed an even higher increase (Figure 5D). Additionally, we also

288 observed a slight reduction of T cell infiltration and neutrophil levels (Figure S6B).

289

290 Single-cell transcriptome analysis of TAMs in HER2+ BC after 4D5 with CD47 blockade

291 combination therapy

292 To further determine if macrophages were differentially activated, we performed single-cell RNA

293 sequencing on dissociated tumors from the HER2 transgenic mice. These studies confirmed the

294 increase of macrophages upon 4D5-IgG2A plus αCD47 treatment and also revealed the

295 emergence of a distinct group of macrophages (that we termed “Phag MΦ” cluster) that are

296 phenotypically distinct from the resident macrophage clusters (i.e. “M1 MΦ” and “M2 MΦ”

297 clusters) (Figure 6A and 6B). Notably, we found that this Phag MΦ cluster contained large

298 quantities of human HER2 RNA and other tumor specific transcripts (such as Epcam and Cyto-

299 keratins), indicating that they have actively phagocytosed tumor cells. This cluster was 300 expanded by 4D5-IgG2A treatment and increased further by combination 4D5+CD47 mAbs

301 treatment (Figure 6B and Table 1). In agreement with our FACS analysis, the level of total

302 macrophages were significantly increased while T cell and neutrophil levels were reduced after

303 4D5 or combination therapy (Figure 6B and Table 1). Interestingly, the frequency of cytotoxic

304 expression (Ifng and Gzmb) among CD8+ T cells were increased following treatments

305 (Figure 6 and Table S1).

306 Using differential analysis, we first assessed the impact of our treatments on

307 the M1-like and M2-like macrophage clusters in comparison to control (Figure 7A and 7B). Of

308 note, these two macrophage clusters do not demonstrate evidence for hyper-phagocytosis of

309 tumor cells at this time point of analysis, as evidenced by their lack of tumor marker uptake

310 (Figure 6B). Gene expression data revealed our treatments promoted macrophage polarization

311 into a pro-inflammatory antitumor phenotype, as evidenced by an increase in involved in

312 interferon, inflammatory cytokines, chemokines and TLR pathways (Figure 7A and 7B).

313 Accordingly, these changes were the most significant with combination therapy, and also more

314 strongly observed in the M1-like MΦ cluster.

315 In contrast, the Phag MΦ cluster (predominantly presence in the combination treatment group)

316 have surprisingly increased expression of gene signatures for wound-repair (e.g.

317 and Tenascins), ECM remodeling (e.g. Collagens and MMPs), growth factors

318 (e.g. Igf1, Tgfb and Egf) and anti-inflammatory genes (e.g. IL4, IL13, IL1r) compared to the

319 other two MΦ clusters (Figure 7C). This is also accompanied by decreased expression of genes

320 for pro-inflammatory cytokines/chemokines, phagocytosis/opsonization, and

321 presentation (Figure 7C).

322 These scRNAseq analyses revealed that while Trastuzumab with CD47 blockade polarizes

323 macrophages into an antitumor phenotype and greatly increases tumor phagocytosis, prolonged

324 treatment and continuous tumor hyper-phagocytosis may also trigger a transcriptional switch in 325 TAMs for repair of ADCP-induced tissue damage. Thus, while these studies demonstrate the

326 antitumor efficacy of Trastuzumab+CD47 blockade, it also suggest that prolonging this process

327 can trigger a wound healing response in macrophages that could have pro-tumor and/or

328 immunosuppressive functions (44-46).

329

330 Human CD47 gene expression is a prognostic factor in HER2+ breast cancer and limits

331 the therapeutic activity of Trastuzumab.

332 As all of our investigations had been performed on different murine HER2+ BC models, we also

333 wanted to determine if ADCP activity of Trastuzumab can be seen in human HER2+ BC and if

334 CD47 could likewise limit its antitumor efficacy. Based on our findings, we hypothesized that

335 CD47 expression may allow for resistance and reduced survival of HER2+ BC patients

336 undergoing Trastuzumab therapies. To investigate this hypothesis, we utilized the METABRIC

337 (Molecular Taxonomy of Breast Cancer International Consortium) gene expression dataset (47)

338 and stratified breast cancer patients of different molecular subtypes into “CD47 high” and “CD47

339 low” groups based on optimum threshold. This analysis revealed that CD47 gene expression

340 associates with lower patient overall survival (Figure 8A) and was most significant in the HER2+

341 molecular subtype compared to TNBC or ER+ subtypes (Figure 8B). This suggests that CD47

342 signaling may be an important resistance mechanism for HER2+ breast cancer and

343 Trastuzumab therapy.

344 We next investigated whether human CD47 limits the ADCP effect of Trastuzumab against

345 amplified HER2+ human BC cells. To address this in vitro, we first generated CD47-KO KPL-4

346 (HER2+ BC) cells (Figure S4C) and compared them to controls after Trastuzumab treatment in

347 ADCP experiments using human PBMC derived macrophages. As in mouse studies, we found

348 loss of human CD47 in tumor cells increased their susceptibility to ADCP elicited by 349 Trastuzumab (Figure 8C). To determine if this antitumor effect also occurs in vivo against

350 human HER2+ BC cells, we implanted KPL-4 control and CD47-KO cell lines into SCID-beige

351 mice (which contain a mouse SIRPα that can bind to human CD47 (48)) and treated with clinical

352 grade Trastuzumab. As before, we saw a strong effect from Trastuzumab treatment that was

353 significantly enhanced with CD47-KO, resulting in tumors being completely eliminated (Figure

354 8D). In Trastuzumab-treated mice, we again found a significant increase of TAMs (Figure 8E)

355 and an upregulation of pro-inflammatory genes (Figure 8F) as seen in the murine tumor model.

356 Unfortunately, the complete regression of CD47KO+Trastuzumab tumors precluded any further

357 analysis of these tumors. Collectively, these studies suggest that the dominant antitumor

358 mechanism of Trastuzumab therapy is through ADCP of HER2+ tumor cells, which can be

359 substantially impaired through the CD47-SIRPα axis. This suggests that combinatorial therapy

360 with CD47 blockade could be beneficial in patients with Trastuzumab resistance.

361

362 Discussion:

363 Even since the demonstration of clinical benefit provided by therapeutic HER2 specific mAbs

364 to patients with HER2 overexpressing BC, the mechanism of action for the therapeutic HER2

365 mAb, Trastuzumab has been the subject of numerous studies. Some reports suggest that

366 Trastuzumab may both block oncogenic HER2 signaling as well as inducing ADCC (7, 49, 50).

367 Using reflective murine versions of clinically approved HER2 specific mAb Trastuzumab, our in

368 vitro studies confirmed these reported MOAs, specifically blockade of HER2 signaling and

369 Trastuzumab-mediated ADCC by NK cells. In contrast, the in vivo antitumor mechanisms of

370 Trastuzumab/4D5 remain less conclusive, with early studies suggesting the importance of signal

371 blockade (51, 52), and subsequent studies demonstrating the direct involvement of ADCC

372 eliciting FcR-expressing cells (10) (such as neutrophils and NK cells), and more recent studies

373 highlighting the importance of adaptive immunity) (8, 15). Notably, few studies have examined 374 Trastuzumab-mediated ADCP with a single study documenting the ability of Trastuzumab to

375 elicit ADCP in vivo (16), while another study suggested that Trastuzumab-mediated ADCP from

376 tumor-associated macrophages (TAMs) is immunosuppressive (17). Consequently, our novel

377 models and agents provided a reliable platform and opportunity to interrogate the in vivo

378 antitumor mechanism of HER2 specific mAbs against HER2 driven BC.

379 In this study using multiple models of human HER2 expressing BC, i.e. MM3MG-HER2Δ16,

380 KPL-4 and an endogenous transgenic HER2+ BC model that is tolerant to human HER2, and

381 using the murine version of Trastuzumab with the functionally equivalent mouse isotype (4D5-

382 IgG2A), we demonstrate that macrophages are the major effectors carrying out the antitumor

383 immunity of Trastuzumab therapy through antibody-dependent-cellular-phagocytosis (ADCP).

384 Although TAMs have been shown to promote tumor progression, it is known that they also

385 retain their Fc-dependent antitumor function when induced by targeted therapies (i.e.

386 monoclonal antibodies) (53, 54). Our conclusion about the therapeutic impact of TAMs is

387 supported by the following findings: (1) the therapeutic effect of Trastuzumab is equivalent in

388 wild type and in SCID-beige mice and does not alter systemic HER2-specific adaptive immunity

389 and T cell/NK cell infiltration in tumors, indicating adaptive immunity and NK cells are not

390 necessary immune cells to mediate antitumor effects; (2) The depletion of macrophages but not

391 neutrophils had a significant negative effect on Trastuzumab efficacy, (3) Trastuzumab

392 treatment greatly and consistently increased TAMs frequency; (4) Trastuzumab treatment

393 induced ADCP of HER2+ tumor cells in vitro and in vivo in a Fc-receptor dependent fashion; (5)

394 Blocking of the innate immune ADCP CD47-SIRP1α regulatory axis significantly enhanced

395 Trastuzumab therapeutic outcomes and also increased ADCP of tumor cells; (6) Trastuzumab

396 combination with CD47 blockade induced TAMs into a highly phagocytic, immune-stimulatory

397 and antitumor phenotype but also produced a wound-healing, immune-regulatory group of

398 TAMs after prolonged tumor phagocytosis. 399 Our study provides insight on the potential of utilizing TAMs as a potent mediator of innate

400 antitumor immunity that can be further exploited. It was initially believed that macrophages were

401 present in high numbers in solid tumors as a mechanism of rejection. However, it soon became

402 clear that TAMs are typically unable to induce an effective antitumor response in the

403 immunosuppressive tumor microenvironment (55). Furthermore, high TAMs infiltration levels are

404 often associated with poor patient prognosis in breast, lung, prostate, liver, thyroid, pancreas,

405 kidney and many other solid cancer malignancies (56). Indeed, studies have shown that

406 immunosuppressive TAMs can support tumor development by promoting , tissue

407 invasion, and suppressing tumor attack by NK and CTL cells (57). In contrast, TAMs

408 in colorectal cancer have a more activated, immune-stimulatory phenotype and interestingly,

409 high TAM density in colorectal cancer correlates with increased patient survival, (54, 58).

410 Nonetheless, TAMs in multiple histologic types of tumors retain their expression of Fcγ-

411 receptors and increasing evidence suggests mAbs can phenotypically modify

412 immunosuppressive TAMs towards an antitumor phenotype (53, 54, 59). As such, the

413 manipulation of TAMs, potentially through a tumor targeting mAb (e.g. Trastuzumab) or

414 targeting of regulatory axis receptors (e.g. CD47/SIRPα), are promising therapeutic approaches

415 for multiple types of cancer.

416 While previous studies (8, 9) have documented the involvement of T cell immunity in

417 mediating HER2 mAbs efficacy, we were unable to detect a significant enhancement of adaptive

418 T cell responses with Trastuzumab monotherapy in either our orthotopic or HER2-tolerant

419 endogenous models of HER2+ BC. This may be due to the nature of our tumor models, the

420 timing of our analysis, or the specific mAb utilized. In our immunocompetent in vivo studies, we

421 utilized both murine and human HER2 mAbs similar to Trastuzumab (isotypes with a high A/I

422 ratio), as well as both human HER2 transformed cells and an endogenous mouse model of

423 HER2+ BC. Previous studies (8, 9, 17) have utilized rat neu expressing ErbB2 models, non- 424 HER2 transformed cells, and/or alternate Ab isotypes (mouse IgG1 with a low A/I ratio), which

425 may account for a lack of ADCP activity and alteration of immunogenicity. Of note, a recent

426 study using 4D5 antibody containing mouse IgG1 isotype reported that HER2 mAb elicited

427 macrophage ADCP is an immunosuppressive mechanism (17). Given that the mouse IgG1

428 subclass strongly activates inhibitory FCGR signaling on effector cells (low A/I ratio) and

429 therefore being very different from Trastuzumab (human IgG1, high A/I ratio) (18, 19, 40), this

430 emphasizes the need of using functionally equivalent mouse isotypes in translational studies to

431 accurately model human antibody therapy. Nevertheless, clinical studies have demonstrated

432 significant associations between adaptive immune responses and Trastuzumab +

433 chemotherapy efficacy (60). Phagocytosis of tumor cells by macrophages has been

434 documented to boost the priming of tumor specific adaptive CD4+ and CD8+ T cells (36, 61),

435 while different types of chemotherapy have been documented to enhance phagocytosis and

436 augment immunogenic tumor cell death (62). Taken together, we believe that the clinical use of

437 immunogenic chemotherapy combinations could stimulate adaptive immunity that would be

438 potentially enhanced by Trastuzumab-mediated ADCP. However, in the absence of strong

439 immune-stimulation (potentially through chemotherapy or immunogenic cell lines), Trastuzumab

440 does not appear highly effective at eliciting adaptive immunity and functions mainly through the

441 stimulation of ADCP.

442 In identifying ADCP as a critical mechanism for Trastuzumab efficacy, we also explored if it

443 could be further enhanced through the blockade of the CD47 innate immune checkpoint. CD47

444 is highly expressed in BC and functions to suppress phagocytosis through binding with SIRPα

445 on macrophages (23, 24). Interestingly, we found CD47 gene expression is a negative

446 prognostic factor in human BC, most significantly in HER2+ BC. As treatment of HER2

447 overexpressing tumors with Trastuzumab has been available for many years, this observation

448 suggests that CD47 may be functioning in Trastuzumab-treated patients to mediate 449 ADCP/therapeutic resistance. This conclusion is supported by the enhanced effects observed

450 between Trastuzumab and CD47 blockade in augmenting ADCP and antitumor effects in our

451 study. Moreover, single cell transcriptome analysis of the tumor microenvironment demonstrates

452 that Trastuzumab therapy stimulates TAMs into a pro-inflammatory antitumor phenotype, which

453 is further boosted by CD47 blockade (Figure 7A and 7B). Such changes in macrophage

454 phenotypes were also observed in co-cultured ADCP experiments. This suggests combination

455 of targeted mAbs therapy with CD47-SIRPα blockade could be beneficial in HER2+ BC and

456 potentially other solid tumors. Proof-of-concept studies using tumor-targeting mAbs and CD47

457 blockade have been demonstrated in preclinical lymphoma models, as well as a recent phase I

458 study of anti-CD20 mAbs (Rituximab) and CD47 blockade, in Rituximab-refractory Non-

459 Hodgkins Lymphoma patients (41, 63).

460 Additionally by implementing different methods, such as multi-color FACS analysis and

461 single-cell transcriptome analysis, we are the first to demonstrate in vivo tumor phagocytosis by

462 macrophages upon combination of Trastuzumab with CD47 blockade therapies. Moreover, we

463 were able to identify a distinct cluster of hyper-phagocytic TAMs within the TME. The

464 identification of this population of TAMs may also serve as a predictive biomarker of this form of

465 therapy. Gene expression analysis suggested that after profound phagocytosis of tumor cells,

466 these macrophages switched to a tissue repair phenotype, as evidenced by their upregulation of

467 gene signatures for wound-healing, growth factors, ECM remodeling, and anti-inflammatory

468 markers compared to resident macrophages (Figure 7C). Indeed, several studies have

469 demonstrated that cellular phagocytosis over time influences macrophage phenotype, causing a

470 switch from pro-inflammatory to a growth promoting, reparative phenotype (44-46). Interestingly,

471 while the total number of CD8+ TILs were reduced by prolonged combination therapy, the

472 relative percentage of cytotoxic T cells was greatly increased (Figure S7), possibly suggesting a

473 boost in overall tumor-specific T cells frequency. In this manner, this combination therapy may 474 allow for enhanced tumor antigen presentation at the earlier time points of treatment through

475 increasing tumor phagocytosis and antigen uptake, while prolonged treatment limits general T

476 cell infiltration after progression to a wound-healing TAM phenotype. Future experiments using

477 Trastuzumab+αCD47 mAbs analyzing multiple treatment time points, reducing the length of

478 treatment, or combining with other immune checkpoint blockades could potentially improve the

479 infiltration of tumor-specific CTLs.

480 While this is an area in need of additional study, our results suggest that strategies to

481 specifically enhance ADCP activity may be critical in overcoming resistance to HER2 mAb

482 therapies by inhibiting tumor growth and potentially enhancing antigen presentation. While only

483 a single clinical trial using combination of a therapeutic mAb (anti-CD20) and CD47 mAbs has

484 been reported, this study demonstrated a ~50% response rate (11 of 22 patients) and ~36%

485 complete response rate (8 of 22 patients) in resistant/refractory non-Hodgkins’ lymphoma (41).

486 These clinical findings, in conjunction with our recent preclinical studies, strongly suggest

487 combination therapy approach of Trastuzumab with CD47-SIPRα checkpoint blockade could

488 potentially show more benefits and insights of Trastuzumab therapy in HER2+ BC patients.

489 However, the transcriptional switch seen in macrophages after prolonged ADCP also requires

490 attention in future studies that utilize CD47 blockade in combination with targeted mAbs.

491 In sum, our study suggests that the dominant therapeutic MOA for Trastuzumab is through its

492 elicitation of TAM mediated ADCP, which can be enhanced by strategies to specifically augment

493 ADCP. This has potential implications for the use of Trastuzumab in HER2+ cancers, as well as

494 the utilization of other targeted therapies (such as EGFR, CD20, etc.), where efforts to enhance

495 and control ADCP have not been prioritized.

496

497 Methods: 498 Cell lines and genetic modifications strategies.

499 Mouse mammary gland cell lines MM3MG and EPH4 were obtained from ATCC and cultured as

500 described by ATCC protocol. The cDNA of a naturally occurring splice variant of human HER2

501 (HER2Δ16), ), or wild type HER2, were transduced into MM3MG and NMUMG cells using

502 lentiviral transduction. Human HER2+ breast cancer cell line KPL4 was a kind gift from Dr.

503 Kurebayashi (University of Kawasaki Medical School, Kurashiki, Japan) (64) and SKBR3 were

504 purchased from ATCC and cultured as described by ATCC protocol. Jurkat-NFAT-LUC line

505 were obtained from Invivogen (jktl-nfat). CRISPR-Cas9 approached were used to knockout

506 mouse Cd47 in MM3MG-HER2Δ16 cells or human CD47 in KPL4 cells. Gene targeting of

507 mouse Cd47, human CD47 and control gene GFP by CRISPR/Cas9 was accomplished through

508 the use of pLentiCRISPRv2 (Addgene plasmid # 52961) using published protocols (65). Genes

509 were targeted using the guide sequences (cccttgcatcgtccgtaatg and ggataagcgcgatgccatgg) for

510 mouse Cd47, (atcgagctaaaatatcgtgt and ctactgaagtatacgtaaag) for human CD47, and

511 (GGGCGAGGAGCTGTTCACCG) for the GFP control. Successful targeting of CD47 was

512 determined by flow cytometry screening after single cell clonal selection. The overexpression

513 vector of mouse Cd47 was generated by synthesizing the Cd47 gene and cloning it into

514 pENTR1a (using NEB Gibson Isothermal Assembly Mix) and then using L/R clonase to

515 generate expression lentiviruses (pLenti-CMV-Puro) and cells were selected using puromycin.

516

517 Mice

518 Female Balb/c (Jackson Labs, Bar Harbor, MA), SCID-beige (C.B-Igh-1b/GbmsTac-Prkdcscid-

519 Lystbg N7; Taconic Biosciences, Model# CBSCBG), Fcer1g-/- (C.129P2(B6)-Fcer1gtm1Rav N12;

520 Taconic Biosciences, Model# 584) mice between the ages of 6 and 10 weeks old were used for

521 all experiments. The HER2Δ16 transgenic model was generated by crossing MMTV-rtTA strain 522 (a kind gift by Dr. Lewis Chodosh, UPenn, Philadelphia, USA) with TetO-HER2d16-IRES-EGFP

523 strain (a kind gift by Dr. William Muller, McGill University, Montreal, Canada) (20). 6-weeks old

524 mice were put on doxycycline diet and enrolled for experiments when they develop palpable

525 breast tumor (usually in 4-6 weeks post dox diet).

526

527 Therapeutic antibodies and other experimental reagents

528 Clinical Grade Trastuzumab (human IgG1) were obtained from Duke Medical Center. 4D5, the

529 murine version of Trastuzumab (with the IgG2A and IgG1 mouse isotypes) were produced by

530 GenScript through special request. CD47 Blockade antibody MIAP410 (BE0283) and control

531 mouse IgG2A (BE0085) were purchased from BIOXCELL. Neutrophil depletion anti-LY6G

532 antibody (IA8, BP0075-1) and macrophage depletion antibody anti-CSF1R (AS598, BE0213)

533 were purchased from BIOXCELL. Clodronate liposomes were purchased from

534 www.clodronateliposomes.org

535

536 Orthotopic implanted HER2+ breast cancer mouse models and therapeutic antibody

537 treatments

538 MM3MG cells expressing human HER2Δ16 were implanted into their mammary fat pads (1x106

539 cells) of Balb/c mice. For the human xenograft model, KPL-4 cells (1x106 cells) were implanted

540 into mammary fat pads (MFP) of SCID-Beige Balb/c mice. Tumor growth were measured with

541 caliper-based tumor volume measurement (length x width x depth) over time. For therapeutic

542 treatments, Trastuzumab or 4D5 were administered weekly (200 µg per mice intraperitoneally)

543 around 4-5 days post tumor implantation. CD47 blockade (MIAP410) were administered weekly

544 when indicated (300 µg per mice intraperitoneally) around 4-5 weeks post tumor implantation.

545 For macrophage depletion, anti-CSF1R antibody were administered triweekly (300 µg per mice 546 intraperitoneally), starting at two weeks before tumor implantation and with treatment

547 maintained over the course of the experiment. Clodronate liposomes were administered

548 biweekly (100 µL per mice, intraperitoneally). For neutrophil depletion, anti-LY6G antibody were

549 administered biweekly (300 µg per mice intraperitoneally) for the first two weeks post tumor

550 implantation.

551

552 Transgenic HER2Δ16 mouse model and therapeutic antibody treatments

553 The HER2Δ16 transgenic mouse model was generated by crossing two strains of mice, TetO-

554 HER2Δ16-IRES-EGFP and MMTV-rtTA. This system was described previously (20), but utilizes

555 a TET-ON system (with MTV-rtTA) to drive expression of HER2Δ16 to generate HER2+ BC. For

556 experiments, one-month old mice were put on Doxycycline diet (200mg/kg, Bio-Serv,

557 Flemington, NJ) to induce spontaneous HER2-driven breast cancer. Individual animals were

558 randomly enrolled into a specific treatment group as soon as palpable breast tumors were

559 detected (~200mm3) in any of the eight mammary fat pads. Control and 4D5-IgG2A antibodies

560 were treated 200 µg weekly, whereas MIAP410 were treated 300 µg weekly intraperitoneally.

561 Animals were terminated once their total tumor volume reached >2000 mm3.

562

563 Flow Cytometry Analysis of tumor infiltrating immune cells

564 When tumor growth reached humane end point size (>1000 mm3), whole tumors from mice

565 were harvested and cut into <1 mm small pieces, and incubated for 1 hour in digestion buffer

566 (DMEM + 100 µg/mL collagenase + 0.2 U/mL DNAse + 1 µg/mL hyalurodinase). Single cell

567 suspensions were spin down through a 70 µm filter and washed with medium. Approximately 5

568 million cells were used for staining and flow cytometry analysis. The following panel of immune

569 cell markers (Biolegend) were used: CD45 BV650, CD11b PE-Cy7, LY6G APC, LY6C BV410, 570 F4/80 PerCP-CY5.5, CD8B APC-CY7, CD4 PE-TR, CD49b FITC and viability dye (Aqua or

571 Red). Tumor-associated macrophages (TAM) were identified by F4/80+ LY6G- LY6C- CD11b+

572 CD45+ gating. LY6G+ neutrophils were identified by LY6G+ CD11b+ CD45+ gating, whereas

573 LY6C+ monocytes were gated on LY6C+ CD11b+ CD45+ cells.

574

575 In vivo ADCP assay

576 MM3MG-HER2Δ16 cells were labeled with Vybrant DiD labeling solution (Thermo V22887)

577 according to manufacturer’s protocol, and labeled cells were implanted (1x106) into MFP of

578 Balb/c mice. Once tumor reaches around 1000 mm3 in sizes, mice were treated with either

579 control antibody (200 µg), 4D5 (200 µg), or 4D5 in combination with MIAP410 (300 µg) per day

580 for two consecutive days. Tumor associated macrophages were analyzed by FACS (CD11b+,

581 F4/80+, LY6G-, LY6C-) and the percentage of TAMs that have taken up DiD-labeled tumor cells

582 were quantified for in vivo ADCP analysis.

583

584 In vitro ADCP and ADCC assays

585 ADCP and ADCC by macrophages – Bone marrow derived macrophages (BMDM) were

586 generated from mouse tibia, differentiated for 10 days with 50 ng/mL mouse MCSF (Peprotech

587 315-02). Briefly, 50 million bone marrow cells were plated in 10 cm2 tissue culture dish with

588 MCSF on day 0. Unattached cells in supernatant were removed and fresh media + MCSF were

589 added on day 3, day 6 and day 9. BMDM were used for ADCP/ADCC assays on day 10. Tumor

590 cells MM3MG-HER2Δ16 were labeled with Brilliant Violet 450 Dye (BD 562158) according to

591 manufacturer’s protocol, and incubated with control or anti-HER2 antibodies (10 µg/mL) in 96-

592 wells (100,000 cells/well) for 30 minutes at 37 0C. BMDM were then added for co-culture at a

593 3:1 ratio of Tumor vs BMDM. After 2 hours co-culture, phagocytosis of BV450-labeled tumor 594 cells by BMDM were analyzed by FACS with CD45-APC staining and Live-death (Red) staining.

595 When indicated, ADCP inhibitor Latrunculin A (120 nM, Thermo L12370) and ADCC inhibitor

596 Concanamycin A (1 µM, Sigma C9705) were added as assay controls. For human

597 macrophages ADCP assay, human monocytes-derived macrophages (hMDM) were generated

598 from three donors’ PBMCs. hMDM were generated with 50 ng/mL human MCSF (Peprotech

599 300-25) and 50 ng/mL human GM-CSF (Peprotech 300-03). KPL-4 cells were used as human

600 HER2+ tumor targets and labeled and co-cultured similarly as with mouse ADCP assay.

601

602 FCGR binding/activation assay

603 expressing mouse Fcgr1, Fcgr2b, Fcgr3 or Fcgr4 with NFAT-Luciferase reporter

604 were generated with lentiviral transduction and selected with puromycin (validated in Figure

605 S4D-F). For the assay, MM3MG breast cancer lines expressing HER2 were first plated and

606 treated with Trastuzumab or 4D5 antibodies or control IgG for 1 hour. Jurkat-FCGR-NFAT-LUC

607 effector cells were added and co-cultured for 4 hours. FCGR signaling activation were assessed

608 by luciferase activity quantification.

609

610 Multiplex cytokine and chemokine assay.

611 BMDM were co-cultured with MM3MG-HER2Δ16 cells for 24 hours, and supernatants were

612 harvested for analysis of cytokines/chemokines levels. The 26-Plex Mouse ProcartaPlex™

613 Panel1 (Thermo) was used and analyzed using the Luminex MAGPIX system.

614

615 METABRIC analysis of CD47 expression in breast cancer patients 616 Pre-processing METABRIC data: Previously normalized gene expression and clinical data were

617 obtained from the European Genome-Phenome Archive (EGA) under the accession id

618 EGAS00000000098 after appropriate permissions from the authors (47). The discovery dataset

619 was composed of 997 primary breast tumors and a second validation set was composed of 995

620 primary breast tumors. The expression data were arrayed on Illumina HT12 Bead Chip

621 composed of 48,803 transcripts. Multiple exon-level probe sets from a transcript cluster

622 grouping were aggregated to a single gene-level probe set using maximum values across all the

623 probes for a given gene. The resulting gene expression matrix consists of 28,503 genes.

624 In order to assess the prognostic significance of CD47 in METABRIC data we generated

625 Kaplan-Meier survival curves on patients stratified by the average expression of CD47 (in to low

626 and high groups) using R package ‘survminer’ (version 0.4.3). Distributions of Kaplan-Meier

627 survival curves for progression-free and overall survival were compared using log- test, and

628 a log-rank test p- value ≤ 0.05 is considered to be statistically significant.

629

630 Single-Cell RNA-Seq Analysis

631 Fastq files from 10X library sequencing were processed using the CellRanger pipeline available

632 from 10X genomics. As part of the processing the assembled sequencing reads were mapped

633 to the mm10 mouse genome. In order to obtain the transcript counts of human ERBB2 (HER2)

634 the sequencing reads were separately aligned to the current version of the ,

635 GRCh38.

636 The gene expression files consisting of raw counts at the gene level for each cell which was

637 analyzed using version 2.3.4 of the Seurat package. The human ERBB2 counts were combined

638 with the mm10 based counts into once expression matrix for each sample. Briefly, the data

639 analysis steps using Seurat consisted of combining the gene counts for all the cells in the 640 different conditions into one matrix, filtering low quality cells, normalizing, and adjusting for cell

641 cycle and batch effects. Unsupervised clustering was done to separate the cell types and

642 markers for the cell types were identified using differential gene expression. These markers

643 were then used for identifying the cell subpopulations within the tumor microenvironment,

644 namely the Immune cells, Tumor cells and Fibroblasts. The normalized gene counts were used

645 to generate tSNE maps for visualization of the cell types and heatmaps for the cell type specific

646 gene expression. Expression of predefined gene sets representing pathways of interest where

647 obtained from previous publications and summarized in Table S2. The data discussed in this

648 publication have been deposited in NCBI's Gene Expression Omnibus and are accessible

649 through GEO Series accession number GSE139492

650 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE139492)

651

652 Statistical methods

653 All statistical analysis of tumor growth comparisons and tumor immune infiltrates were

654 performed with GraphPad Prism (v8) using two-way ANOVA or one-way ANOVA test with

655 Tukey’s multiple comparisons. Unless otherwise indicated in the figure, tests results were

656 shown between treatment vs control group. Group sizes for animal tumor growth experiments

657 were determined based on preliminary datasets. All subjects in animal experiments were

658 randomized into a treatment or control group. For in vitro experiments, i.e. ADCP/ADCC/CDC

659 assays, ELISPOT assays, FCGR signaling assay and cytokine assays, all data were statistically

660 analyzed by one-way ANOVA test with Tukey’s multiple comparisons, and performed with at

661 least four biological replicates per experiment and repeated at least two times. RT-qPCR data

662 were analyzed by two-sided Student’s t test for each target gene. 95% confidence interval was

663 considered for statistics and p<0.05 was considered significant. 664

665 Study Approval:

666 All animals were maintained, bred, in accordance with Duke Institutional Animal Care and Use

667 Committee–approved protocol (A198-18-08), and supervised by Division of Laboratory Animal

668 Resources (DLAR).

669

670 Author Contributions:

671 Li-Chung Tsao, H. Kim Lyerly and Zachary C. Hartman conceived the project and wrote the

672 manuscript. Li-Chung Tsao and Zachary C. Hartman designed the experiments, analyzed and

673 interpreted the data. Li-Chung Tsao performed the animal experiments, with assistance from

674 Erika J. Crosby, Bin-Jin Hwang, Xiao Yi Yang, Cong-Xiao Liu and Christopher A. Rabiola. Tao

675 Wang provided technical support and suggestions for flow cytometry. Zachary C. Hartman

676 generated various cell lines and CRISPR knock outs with assistance from Junping Wei and

677 Gangjun Lei. Li-Chung Tsao performed the in vitro cell culture assays. Li-Chung Tsao designed

678 single-cell RNA sequencing experiment and generated 10X Genomics cDNA libraries. Pankaj

679 Agawal performed the clustering and statistical analysis and illustration of single-cell RNA-Seq

680 results. Chaitanya Acharya conducted the stratification and statistical analysis of METABRIC

681 datasets. Erika J. Crosby, Timothy N. Trotter and Casey W. Shuptrine provided intellectual input

682 and discussions. The MMTV-rtTA mouse strain was provided by Dr. Lewis Chodosh (UPenn)

683 and TetO-HER2d16-IRES-EGFP mouse by Dr. William Muller (McGill University).

684

685 Acknowledgements: This research was supported by grants from the National Institutes of

686 Health (NIH) (5K12CA100639-09 to ZCH, 1R01CA238217-01A1 to ZCH) and (T32CA009111 to

687 LCT), Department of Defense (DOD) (BC113107 to HKL), Merck MISP Program (LKR160831 to 688 ZCH), and Susan G. Komen (CCR14299200 to ZCH). We would also like to acknowledge all of

689 the members of the Applied Therapeutics Center for their generous technical assistance, help

690 and insights into this project.

691 692 References: 693 694 1. Slamon DJ, Clark GM, Wong SG, Levin WJ, Ullrich A, and McGuire WL. Human breast cancer: 695 correlation of relapse and survival with amplification of the HER-2/neu oncogene. Science (New 696 York, NY). 1987;235(4785):177-82. 697 2. Di Fiore PP, Pierce JH, Kraus MH, Segatto O, King CR, and Aaronson SA. erbB-2 is a potent 698 oncogene when overexpressed in NIH/3T3 cells. Science (New York, NY). 1987;237(4811):178- 699 82. 700 3. Hudis CA. Trastuzumab--mechanism of action and use in clinical practice. The New England 701 journal of medicine. 2007;357(1):39-51. 702 4. Untch M, Rezai M, Loibl S, Fasching PA, Huober J, Tesch H, et al. Neoadjuvant treatment with 703 trastuzumab in HER2-positive breast cancer: results from the GeparQuattro study. Journal of 704 clinical oncology : official journal of the American Society of Clinical Oncology. 2010;28(12):2024- 705 31. 706 5. Gianni L, Pienkowski T, Im Y-H, Tseng L-M, Liu M-C, Lluch A, et al. 5-year analysis of neoadjuvant 707 pertuzumab and trastuzumab in patients with locally advanced, inflammatory, or early-stage 708 HER2-positive breast cancer (NeoSphere): a multicentre, open-label, phase 2 randomised trial. 709 The Lancet Oncology. 2016;17(6):791-800. 710 6. de Azambuja E, Holmes AP, Piccart-Gebhart M, Holmes E, Di Cosimo S, Swaby RF, et al. Lapatinib 711 with trastuzumab for HER2-positive early breast cancer (NeoALTTO): survival outcomes of a 712 randomised, open-label, multicentre, phase 3 trial and their association with pathological 713 complete response. The Lancet Oncology. 2014;15(10):1137-46. 714 7. Bianchini G, and Gianni L. The immune system and response to HER2-targeted treatment in 715 breast cancer. The Lancet Oncology. 2014;15(2):e58-e68. 716 8. Park S, Jiang Z, Mortenson ED, Deng L, Radkevich-brown O, Yang X, et al. The Therapeutic Effect 717 of Anti-HER2 / neu Antibody Depends on Both Innate and Adaptive Immunity. Cancer Cell. 718 2010;18(2):160-70. 719 9. Stagg J, Loi S, Divisekera U, Foong S, Duret H, and Yagita H. Anti – ErbB-2 mAb therapy requires 720 type I and II interferons and synergizes with anti – PD-1 or anti-CD137 mAb therapy. 2011:10-5. 721 10. Clynes RA, Towers TL, Presta LG, and Ravetch JV. Inhibitory Fc receptors modulate in vivo 722 cytoxicity against tumor targets. Nature Medicine. 2000;6:443. 723 11. Musolino A, Naldi N, Dieci MV, Zanoni D, Rimanti A, Boggiani D, et al. 724 fragment C receptor polymorphisms and efficacy of preoperative chemotherapy plus 725 trastuzumab and lapatinib in HER2-positive breast cancer. The pharmacogenomics journal. 726 2016;16(5):472-7. 727 12. Muntasell A, Cabo M, Servitja S, Tusquets I, Martinez-Garcia M, Rovira A, et al. Interplay 728 between Natural Killer Cells and Anti-HER2 Antibodies: Perspectives for Breast Cancer 729 Immunotherapy. Front Immunol. 2017;8:1544. 730 13. Muntasell A, Rojo F, Servitja S, Rubio-Perez C, Cabo M, Tamborero D, et al. NK Cell Infiltrates and 731 HLA Class I Expression in Primary HER2(+) Breast Cancer Predict and Uncouple Pathological 732 Response and Disease-free Survival. Clinical cancer research : an official journal of the American 733 Association for Cancer Research. 2019;25(5):1535-45. 734 14. Albanesi M, Mancardi DA, Jonsson F, Iannascoli B, Fiette L, Di Santo JP, et al. Neutrophils 735 mediate antibody-induced antitumor effects in mice. Blood. 2013;122(18):3160-4. 736 15. Mortenson ED, Park S, Jiang Z, Wang S, and Fu YX. Effective anti-neu-initiated antitumor 737 responses require the complex role of CD4+ T cells. Clinical cancer research : an official journal 738 of the American Association for Cancer Research. 2013;19(6):1476-86. 739 16. Shi Y, Fan X, Deng H, Brezski RJ, Rycyzyn M, Jordan RE, et al. Trastuzumab triggers phagocytic 740 killing of high HER2 cancer cells in vitro and in vivo by interaction with Fcgamma receptors on 741 macrophages. J Immunol. 2015;194(9):4379-86. 742 17. Su S, Zhao J, Xing Y, Zhang X, Liu J, Ouyang Q, et al. Immune Checkpoint Inhibition Overcomes 743 ADCP-Induced Immunosuppression by Macrophages. Cell. 2018;175(2):442-57 e23. 744 18. Bruhns P, and Jonsson F. Mouse and human FcR effector functions. Immunol Rev. 745 2015;268(1):25-51. 746 19. Nimmerjahn F, and Ravetch JV. Divergent Immunoglobulin G Subclass Activity Through Selective 747 Fc Receptor Binding. Science. 2005;310(5753):1510. 748 20. Turpin J, Ling C, Crosby EJ, Hartman ZC, Simond AM, Chodosh LA, et al. The ErbB2ΔEx16 splice 749 variant is a major oncogenic driver in breast cancer that promotes a pro-metastatic tumor 750 microenvironment. Oncogene. 2016;35(47):6053-64. 751 21. Crosby EJ, Wei J, Yang XY, Lei G, Wang T, Liu CX, et al. Complimentary mechanisms of dual 752 checkpoint blockade expand unique T-cell repertoires and activate adaptive anti-tumor 753 immunity in triple-negative breast tumors. Oncoimmunology. 2018;7(5):e1421891. 754 22. Jaiswal S, Chao MP, Majeti R, and Weissman IL. Macrophages as mediators of tumor 755 immunosurveillance. Trends in immunology. 2010;31(6):212-9. 756 23. Mantovani A, and Longo DL. Macrophage Checkpoint Blockade in Cancer — Back to the Future. 757 New England Journal of Medicine. 2018;379(18):1777-9. 758 24. Huang Y, Ma Y, Gao P, and Yao Z. Targeting CD47: The achievements and concerns of current 759 studies on cancer immunotherapy. Journal of Thoracic Disease. 2017;9(2):E168-E74. 760 25. Zhang H, Lu H, Xiang L, Bullen JW, Zhang C, Samanta D, et al. HIF-1 regulates CD47 expression in 761 breast cancer cells to promote evasion of phagocytosis and maintenance of cancer stem cells. 762 Proceedings of the National Academy of Sciences. 2015;112(45):E6215. 763 26. Yan M, Schwaederle M, Arguello D, Millis SZ, Gatalica Z, and Kurzrock R. HER2 expression status 764 in diverse cancers: review of results from 37,992 patients. Cancer metastasis reviews. 765 2015;34(1):157-64. 766 27. Carter P, Presta L, Gorman CM, Ridgway JB, Henner D, Wong WL, et al. Humanization of an anti- 767 p185HER2 antibody for human cancer therapy. Proceedings of the National Academy of 768 Sciences. 1992;89(10):4285. 769 28. Stewart R, Hammond SA, Oberst M, and Wilkinson RW. The role of Fc gamma receptors in the 770 activity of immunomodulatory antibodies for cancer. Journal for ImmunoTherapy of Cancer. 771 2014;2(1):1-10. 772 29. Moasser MM. Two dimensions in targeting HER2. Journal of clinical oncology : official journal of 773 the American Society of Clinical Oncology. 2014;32(19):2074-7. 774 30. Moasser MM, and Krop IE. The Evolving Landscape of HER2 Targeting in Breast Cancer. JAMA 775 Oncol. 2015;1(8):1154-61. 776 31. Wang Y, Yang YJ, Wang Z, Liao J, Liu M, Zhong XR, et al. CD55 and CD59 expression protects 777 HER2-overexpressing breast cancer cells from trastuzumab-induced complement-dependent 778 cytotoxicity. Oncol Lett. 2017;14(3):2961-9. 779 32. Gordon SR, Maute RL, Dulken BW, Hutter G, George BM, McCracken MN, et al. PD-1 expression 780 by tumour-associated macrophages inhibits phagocytosis and tumour immunity. Nature. 781 2017;545(7655):495-9. 782 33. Willingham SB, Volkmer JP, Gentles AJ, Sahoo D, Dalerba P, Mitra SS, et al. The CD47-signal 783 regulatory alpha (SIRPa) interaction is a therapeutic target for human solid tumors. 784 Proceedings of the National Academy of Sciences. 2012;109(17):6662-7. 785 34. Oliveira CA, Kashman Y Fau - Mantovani B, and Mantovani B. Effects of latrunculin A on 786 immunological phagocytosis and macrophage spreading-associated changes in the F-actin/G- 787 actin content of the cells. (0009-2797 (Print)). 788 35. Toda H, Araki K, Moritomo T, and Nakanishi T. Perforin-dependent cytotoxic mechanism in 789 killing by CD8 positive T cells in ginbuna crucian carp, Carassius auratus langsdorfii. 790 Developmental & Comparative Immunology. 2011;35(1):88-93. 791 36. Guilliams M, Bruhns P, Saeys Y, Hammad H, and Lambrecht BN. The function of Fcgamma 792 receptors in dendritic cells and macrophages. Nat Rev Immunol. 2014;14(2):94-108. 793 37. Bournazos S, Wang TT, and Ravetch JV. The Role and Function of Fcgamma Receptors on 794 Myeloid Cells. Microbiol Spectr. 2016;4(6). 795 38. Nimmerjahn F, Lux A, Albert H, Woigk M, Lehmann C, Dudziak D, et al. FcγRIV deletion reveals its 796 central role for IgG2a and IgG2b activity in vivo. Proceedings of the National Academy of 797 Sciences. 2010;107(45):19396. 798 39. Parekh BS, Berger E, Sibley S, Cahya S, Xiao L, LaCerte MA, et al. Development and validation of 799 an antibody-dependent cell-mediated cytotoxicity-reporter gene assay. MAbs. 2012;4(3):310-8. 800 40. Dekkers G, Bentlage AEH, Stegmann TC, Howie HL, Lissenberg-Thunnissen S, Zimring J, et al. 801 Affinity of human IgG subclasses to mouse Fc gamma receptors. MAbs. 2017;9(5):767-73. 802 41. Advani R, Flinn I, Popplewell L, Forero A, Bartlett NL, Ghosh N, et al. CD47 Blockade by Hu5F9-G4 803 and Rituximab in Non-Hodgkin's Lymphoma. N Engl J Med. 2018;379(18):1711-21. 804 42. Lawson MA, McDonald MM, Kovacic N, Hua Khoo W, Terry RL, Down J, et al. Osteoclasts control 805 reactivation of dormant myeloma cells by remodelling the endosteal niche. Nature 806 Communications. 2015;6:8983. 807 43. Turpin J, Ling C, Crosby EJ, Hartman ZC, Simond AM, Chodosh LA, et al. The ErbB2DeltaEx16 808 splice variant is a major oncogenic driver in breast cancer that promotes a pro-metastatic tumor 809 microenvironment. Oncogene. 2016;35(47):6053-64. 810 44. Fadok VA, Bratton DL, Konowal A, Freed PW, Westcott JY, and Henson PM. Macrophages that 811 have ingested apoptotic cells in vitro inhibit proinflammatory cytokine production through 812 autocrine/paracrine mechanisms involving TGF-beta, PGE2, and PAF. The Journal of clinical 813 investigation. 1998;101(4):890-8. 814 45. Lech M, and Anders H-J. Macrophages and : How resident and infiltrating mononuclear 815 phagocytes orchestrate all phases of tissue injury and repair. Biochimica et Biophysica Acta 816 (BBA) - Molecular Basis of Disease. 2013;1832(7):989-97. 817 46. Minutti CM, Knipper JA, Allen JE, and Zaiss DMW. Tissue-specific contribution of macrophages to 818 wound healing. Seminars in Cell & Developmental Biology. 2017;61:3-11. 819 47. Curtis C, Shah SP, Chin S-F, Turashvili G, Rueda OM, Dunning MJ, et al. The genomic and 820 transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature. 821 2012;486:346. 822 48. Kwong LS, Brown MH, Barclay AN, and Hatherley D. Signal-regulatory protein α from the NOD 823 mouse binds human CD47 with an exceptionally high affinity-- implications for engraftment of 824 human cells. Immunology. 2014;143(1):61-7. 825 49. Hurvitz SA, Betting DJ, Stern HM, Quinaux E, Stinson J, Seshagiri S, et al. Analysis of Fcgamma 826 receptor IIIa and IIa polymorphisms: lack of correlation with outcome in trastuzumab-treated 827 breast cancer patients. Clinical cancer research : an official journal of the American Association 828 for Cancer Research. 2012;18(12):3478-86. 829 50. Barok M, Isola J, Pályi-Krekk Z, Nagy P, Juhász I, Vereb G, et al. Trastuzumab causes antibody- 830 dependent cellular cytotoxicity–mediated growth inhibition of submacroscopic JIMT-1 breast 831 cancer xenografts despite intrinsic drug resistance. Molecular Cancer Therapeutics. 832 2007;6(7):2065. 833 51. Ritter CA, Perez-Torres M, Rinehart C, Guix M, Dugger T, Engelman JA, et al. Human breast 834 cancer cells selected for resistance to trastuzumab in vivo overexpress epidermal growth factor 835 receptor and ErbB ligands and remain dependent on the ErbB receptor network. Clinical cancer 836 research : an official journal of the American Association for Cancer Research. 2007;13(16):4909- 837 19. 838 52. Yakes FM, Chinratanalab W, Ritter CA, King W, Seelig S, and Arteaga CL. Herceptin-induced 839 inhibition of phosphatidylinositol-3 kinase and Akt Is required for antibody-mediated effects on 840 p27, cyclin D1, and antitumor action. Cancer research. 2002;62(14):4132-41. 841 53. Grugan KD, McCabe FL, Kinder M, Greenplate AR, Harman BC, Ekert JE, et al. Tumor-Associated 842 Macrophages Promote Invasion while Retaining Fc-Dependent Anti-Tumor Function. The Journal 843 of Immunology. 2012;189(11):5457. 844 54. Gul N, and van Egmond M. Antibody-Dependent Phagocytosis of Tumor Cells by Macrophages: A 845 Potent Effector Mechanism of Monoclonal Antibody Therapy of Cancer. Cancer Res. 846 2015;75(23):5008-13. 847 55. Brown JM, Recht L, and Strober S. The Promise of Targeting Macrophages in Cancer Therapy. 848 Clinical cancer research : an official journal of the American Association for Cancer Research. 849 2017;23(13):3241-50. 850 56. Ruffell B, Affara NI, and Coussens LM. Differential macrophage programming in the tumor 851 microenvironment. Trends Immunol. 2012;33(3):119-26. 852 57. Noy R, and Pollard JW. Tumor-associated macrophages: from mechanisms to therapy. Immunity. 853 2014;41(1):49-61. 854 58. Zhang QW, Liu L, Gong CY, Shi HS, Zeng YH, Wang XZ, et al. Prognostic significance of tumor- 855 associated macrophages in solid tumor: a meta-analysis of the literature. PLoS One. 856 2012;7(12):e50946. 857 59. Mantovani A, Marchesi F, Malesci A, Laghi L, and Allavena P. Tumour-associated macrophages as 858 treatment targets in oncology. Nature reviews Clinical oncology. 2017;14(7):399-416. 859 60. Taylor C, Hershman D Fau - Shah N, Shah N Fau - Suciu-Foca N, Suciu-Foca N Fau - Petrylak DP, 860 Petrylak Dp Fau - Taub R, Taub R Fau - Vahdat L, et al. Augmented HER-2 specific immunity 861 during treatment with trastuzumab and chemotherapy. (1078-0432 (Print)). 862 61. Abès R, Gélizé E, Fridman WH, and Teillaud J-L. Long-lasting antitumor protection by anti-CD20 863 antibody through cellular . Blood. 2010;116(6):926. 864 62. Pfirschke C, Engblom C, Rickelt S, Cortez-Retamozo V, Garris C, Pucci F, et al. Immunogenic 865 Chemotherapy Sensitizes Tumors to Checkpoint Blockade Therapy. Immunity. 2016;44(2):343- 866 54. 867 63. Chao MP, Alizadeh AA, Tang C, Myklebust JH, Varghese B, Gill S, et al. Anti-CD47 antibody 868 synergizes with rituximab to promote phagocytosis and eradicate non-Hodgkin lymphoma. Cell. 869 2010;142(5):699-713. 870 64. Kurebayashi J, Otsuki T, Tang CK, Kurosumi M, Yamamoto S, Tanaka K, et al. Isolation and 871 characterization of a new human breast cancer cell line, KPL-4, expressing the Erb B family 872 receptors and interleukin-6. BrJCancer. 1999;79(5-6):707-17. 873 65. Shalem O, Sanjana NE, Hartenian E, Shi X, Scott DA, Mikkelsen TS, et al. Genome-Scale CRISPR- 874 Cas9 Knockout Screening in Human Cells. Science. 2014;343(6166):84. 875 876 877 878 Figure and figure legends:

A B MM3MG-HER2Δ16 grow th in Balb/c mice

2000 Control IgG Trastuzumab

3 1500 4D5-IgG2A

m

m

e z

i 1000

s

r

o m

u **** T 500 ****

0 0 5 10 15 20 25 30 Days post tumor injection C Control IgG 4D5-IgG2A Trastuzumab Tumor-associated macrophages (TAMs)

80 ***

- ***

C 6

Y 60

L

/

-

s

l

l

G

e

6

c

Y

+ L

/ 40

5

+ 4

13.1% 0

D

8 C 82.6% 4

75.8%

f

F

/ LY6G

o 20

+

b

1

1 D

C 0 F4/80 % Control IgG 4D5-IgG2A Trastuzumab Gated on liv e CD45+ CD11b+ Ly 6C- tumor inf iltrates

D SCID-Beige mice E F Neutrophil depletion Depletion of TAMs G Macrophage depletion Control IgG s 1200 Control IgG

1500 s

n 80 * l

Control IgG l o

Trastuzumab i

1000 e t

) Control IgG + CSF1R

c

3

a

l

** ) 1000 )

Control IgG + LY6G - 3

4D5-IgG2A u

3 4D5-IgG2A

C p

m 60 n.s

m

6 m

750 o 4D5-IgG2A + CSF1R m

Y 800 m 4D5-IgG2A p

m 1000 (

n.s

(

L

(

+

e

-

e

e

5

z z

i ***

4D5-IgG2A + LY6G G

 i z 4 600

i 40

s

s

6

s

500 D ***

r

r

Y

r

C

o

o

L

o

400

+

m m

500 +

m

b

u u

0 20 u

**** 1

T T

250 8

/ T

1 200 4

**** n.s D

F

C

%

n 0 0 0 0 i x R A R 0 5 10 15 20 25 0 5 10 15 20 0 5 10 15 20 25 R 1 2 1 l F G F ro S g S t C -I C Days post tumor injection Days post tumor injection Days post tumor injection n  5  o D C 4 + A 2 G g -I 5 D 4

Macrophages w ith dye-labeled HER2+ BC H ADCP ADCC Macrophages with 4D5-IgG2A **** 4 20 unlabeled HER2+ BC Control IgG 4D5-IgG2A + ADCP inhibitor **** **** ****

h 3

t a

6.6% 18% 4.8% 15 e

s

d

i

l

s

l

o e

MΦ t y

c 2

c

r o

23% 10 o

g

a

m

h

u

t

p %

% 1

Tumor 5 CD45 76% 0 0 r r r r 5 b r r r r G 5 b G a o o o o D a to to to to Ig D it it it it Ig i i i i l 4 m ib ib ib ib l 4 m ib ib ib ib o u h h h h o u h h h h tr z n n n n tr z n n n n n tu i i i i n tu i i i i Brilliant Violet 450+ (Tumor label dy e) o s P P C C o s P P C C C ra C C C C C ra C C C C T D D D D T D D D D A A A A A A A A + + + + + + + + G 5 G 5 5 5 Ig D Ig D G G l 4 l 4 Ig D Ig D o o l 4 l 4 tr tr o o n n tr tr o o n n C C o o C C 879 880 Figure 1. Generation of murine Trastuzumab and its antitumor dependence on Antibody- 881 Dependent-Cellular-Phagocytosis (ADCP) by tumor-associated macrophages (TAMs). (A) 882 Cartoon presentation of Trastuzumab and 4D5 antibodies used in this study. (B) MM3MG cells 883 expressing human HER2Δ16 were implanted into the mammary fat pads (1x106 cells) of Balb/c 884 mice. Trastuzumab (human IgG1) or 4D5 (mouse IgG2A) were administered weekly (200 µg per 885 mice). n = 8-10. (C) Tumors (>1000 mm3 volume) were processed into single cell suspensions, 886 and TAMs (%CD11b+ F4/80+ LY6G- LY6C- of CD45+ cells) were analyzed by FACS. n = 8-10. 887 (D) Experiment as in Figure 1B was repeated in SCID-Beige animals. n = 8. (E) Experiment in 888 SCID-Beige was repeated using neutrophils-depleting anti-LY6G antibodies (clone IA8, 300 µg 889 per mice biweekly). (F-G) To deplete macrophages, SCID-Beige mice were pre-treated with 890 anti-CSF1R antibody (clone AFS98, 300µg, 3 times per week) for two weeks. (F) Macrophage 891 depletion were verified by FACS. (G) 4D5-IgG2A injection were performed, with anti-CSF1R 892 treatment maintained throughout the experiment. n = 8. (H) Trastuzumab/4D5 induced ADCP of 893 HER2+ BC cells by Bone-marrow-derived-macrophages (BMDM). MM3MG-HER2Δ16 cells 894 were labeled with Brilliant Violet 450 Dye, and co-cultured with BMDM (3:1 ratio) with control or 895 anti-HER2 antibodies (10 µg/mL). ADCP rates were measured by percentage of BMDM uptake 896 of labeled tumor cells (CD45+ and BV450+), and Antibody-dependent-cellular-cytotoxicity 897 (ADCC) rates were measured by percentage of dying free tumor cells (CD45- and LIVE/DEAD 898 stain+). ADCP inhibitor (Latrunculin A) or ADCC inhibitor (Concanamycin A) were added as 899 assay controls. n = 3, experiment has been repeated three separate times. (B, D, E and G) 900 Tumor growth were measured with caliper-based tumor measurement over time. Two-way 901 ANOVA test with Tukey’s multiple comparisons (C, F and H) One-way ANOVA test with Tukey’s 902 multiple comparisons. All data represent mean ±SEM, **P<0.01, ***P < 0.001, ****P<0.0001. 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922

ADCP B A Wild type Balb/c Fcer1g-/- Balb/c *** Control IgG 800 Control IgG 1500 100 4D5-IgG2A Control IgG

) 4D5-IgG2A 3 ) 4D5-IgG2A

600 3 m

s 80

i

m

s

m

) (

o 1000

m

x t

( n.s

e

a

y z

60 e i

c 400

M

z

s

i

o

r

s

g

%

o

r

a (

40 o m

h **** 500

u m

P 200

T u

20 T 0 0 0 0 5 10 15 20 WT BMDM Fcer1g-/- BMDM 0 5 10 15 20 days post tumor injection days post tumor injection

C Tumor-associated macrophages D MM3MG-HER2Δ16 grow th E Control IgG ADCP

**** 1400 **** s

l 80

l 4D5-IgG1

e )

s 100 c

3 1200

n

- o

m 4D5-IgG2A

i C t 60

6 1000

m a

l 80

(

Y

u

s

i

L e

p n.s

s )

- 800

z

o

x

o

i

t p a 60

G *

s y

40

c

6

M

+

r

600 o

Y

b

o

g

%

L

1

( a

40

m

1

h +

u 400 ****

p D

0 20

T

8

C /

20 4

n 200

i F

% 0 0 0 G A G A 0 5 10 15 20 25 Control IgG 4D5-IgG1 4D5-IgG2A Ig 2 Ig 2 l G l G o g o g tr -I tr -I n 5 n 5 Days post tumor injection o D o D C 4 C 4 + + + + T T O O W W -K -K g g r1 r1 e e c c F F

F Mouse FCGR1 activation G Mouse FCGR3 activation H Mouse FCGR4 activation 100000 10000 100000 *** ****

**** y ****

y t

t Control IgG ****

i

y

i

t

v i

v **** Control IgG **** i

i **** Control IgG

t

v

t

i

c t

c 4D5-IgG2A

c a a 10000 10000

4D5-IgG2A **** 4D5-IgG2A

**** a

g

g )

) 4D5-IgG1

)

g

n

n

e

e

i

i

l

l

e n

4D5-IgG1 s s 4D5-IgG1

i 1000

l

s a

a ** a

a Trastuzumab

r

r

a

n

a n

r Trastuzumab

e e

n Trastuzumab g

g **

f

f

i

i

e

i

i

g

f

i

s

s

i

c

c

s

c

u

u 4

4 1000 1000

u

L

4

L

R

R

-

-

L

R

-

T

T

G

G

T

G

c

c

A

A

c

A

F

F

F

F

F

F

m

N

N

m

( ( N

m 100

(

e

e

e

v

v

i

i

v t

t 100 100

i

t

a

a

l

l

a

l

e

e

e

R R R

10 10 10 0.01 0.1 1 10 100 0.01 0.1 1 10 100 0.01 0.1 1 10 100 Ab (ug/mL) Ab (ug/mL) 923 Ab (ug/mL) 924 Figure 2. The Antibody-dependent-cellular-phagocytosis (ADCP) activity of mouse 925 Trastuzumab (4D5) requires the engagement with Fcγ-receptors (FCGR) and is IgG2A 926 isotype dependent. (A) Fcγ-receptors are required for 4D5-induced ADCP of HER2+ BC cells 927 by Bone-marrow-derived-macrophages (BMDM) in vitro. BMDM were generated from wild type -/- 928 and Fcer1g mice, and ADCP experiment were performed with the conditions described in 929 Figure 1E. n = 3. (B-C) FCGR is required for the antitumor activity of 4D5 therapy. (B) Wild type -/- 930 or Fcer1g Balb/c mice were implanted with MM3MG-HER2Δ16 cells as before (Figure 1B). 931 4D5-IgG2A or control antibodies were administered weekly (200 µg per mice intraperitoneally) 932 and tumor growth were measured. n = 5. (C) Tumor-associated macrophages (TAMs) from 933 tumors in Figure 2B were analyzed by FACS. n = 4-5. (D-F) The ADCP activity of 4D5 is IgG2A 934 isotype dependent. (D) MM3MG-HER2Δ16 tumor growth in mice were repeated using 4D5 935 antibodies containing the mouse IgG1 as comparison to previous IgG2A isotype. n = 8-10. (E) 936 ADCP experiments with BMDM cultures were performed using 4D5-IgG1 versus 4D5-IgG2A 937 antibody isotypes. n = 4. (F-H) Mouse FCGR signaling activation assay. MM3MG breast cancer 938 cells expressing HER2 were plated and treated with indicated antibodies concentrations for 1 939 hour. Jurkat cells containing NFAT-luciferase reporter and expressing mouse FCGR1 (F), 940 FCGR3 (G) or FCGR4 (H) were added to the target cells containing antibodies and co-cultured 941 for 4 hours. FCGR signaling activation were assessed by luciferase activity quantification. n = 4. 942 (A, C, and E) One-way ANOVA with Tukey’s multiple comparisons. (B, D, F, G and H) Two-way 943 ANOVA test with Tukey’s multiple comparisons to control IgG group. All data represent mean 944 ±SEM, *P < 0.05, ***P < 0.001, ****P < 0.0001. 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970

A Control IgG 4D5-IgG2A 6.69% 16.5% ADCP ADCC GFP-KO 40 Control IgG 3 Control IgG 4D5-IgG2A 4D5-IgG2A ***

30

h

t

s i a 2

s ***

e o

15.7% d

28.6% t

l

y

l

c e

o 20

*** c

g

r a

CD47-KO o

h

m p

u 1

t %

10 %

0 0 CD45 GFP-KO CD47-KO GFP-KO CD47-KO Brilliant Violet 450+ (Tumor label dye)

B IL6 TNFα CCL3 CCL4 * 800 Control 600 Control *** Control *** 800 1500 Control *** 4D5-IgG2A 4D5-IgG2A 4D5-IgG2A 4D5-IgG2A 600 600

400 1000

L

L

L

L

m

m

m

m /

400 / /

/ 400

g

g

g

g

p

p

p p 200 500 200 200

0 0 0 0 GFP-KO CD47-KO GFP-KO CD47-KO GFP-KO CD47-KO GFP-KO CD47-KO

C MM3MG-HER2Δ16 grow th D TAMs GFP-KO, Control Treatment 80 Control IgG

GFP-KO, 4D5-IgG2A Treatment s

s 4D5-IgG2A

l *** n

1000 CD47-KO, Control Treatment l

o

e

i t

c 60

a

CD47-KO, 4D5-IgG2A Treatment l

-

) u

3 800

G

p

6

m

o

Y m

p *

(

L

600 40

+

e

+

z

b

i

0

s

1

8

r 1

400 /

o 4

D 20

m

F

C

u

T %

200 n i 0 0 0 5 10 15 20 25 30 35 GFP-KO CD47-KO days post tumor development

E Parental MM3MG-HER2Δ16 MM3MG-HER2Δ16 CD47-OE F TAMs 2000 Control IgG 1500

Control IgG 80 *** Control IgG

s

)

s l

4D5-IgG2A n

3 l

) 4D5-IgG2A o

3 4D5-IgG2A

1500 e

i

t c

m 60

a

l

- m

m 1000 u

( 1.5x

G

m

p

( 6

e 4.4x

o z

1000 Y e

i **

p

z s

L 40

i

+

r

s

+

b

o

r

0

1

o

m 8

500 1

*** / u

500 m

4

D u

T 20

F

C

T

%

n i 0 0 0 0 5 10 15 20 25 0 5 10 15 20 25 Parental tumor CD47-OE tumor days post tumor injection days post tumor injection 971 972 Figure 3. CD47 suppresses the anti-tumor activity of mouse Trastuzumab (4D5). (A) CD47 973 knockout cells were generated from MM3MG-HER2Δ16 cells using CRISPR-Cas9 technology. 974 A control GFP knockout line was generated in parallel. Control and CD47-KO MM3MG- 975 HER2Δ16 cells were labeled with Brilliant Violet 450 Dye, and incubated with Bone-marrow- 976 derived-macrophages (BMDM) at 3:1 ratio with control or 4D5 antibodies (10 µg/mL). Antibody- 977 dependent-cellular-phagocytosis (ADCP) and cytotoxicity (ADCC) activity were measured by as 978 described in Figure 1H. n = 3. Experiment has been repeated two separate times using CD47- 979 KO clones containing a different guide RNA. (B) Secreted cytokines and chemokines by 980 macrophages from co-culture experiment with HER2+ BC were analyzed using the Luminex 981 platform. Additional cytokines detected can be found in Figure S5. n = 3. (C-D) Control and 982 CD47-KO MM3MG-HER2Δ16 cells were implanted into mouse mammary fat pads and treated 983 with 4D5-IgG2A or control antibodies as described before. TAMs were analyzed by FACS after 984 tumor volume reached >1000mm3. n = 5. (E-F) Cd47 overexpressing cells (CD47-OE) were 985 generated in MM3MG-HER2Δ16 cells after transduction with Cd47 cDNA under control of the 986 EF1s promoter. CD47-OE tumor cell growth were compared to parental MM3MG-HER2Δ16 987 cells in mice treated with control antibody or 4D5-IgG2A. TAMs were analyzed by FACS. n = 5. 988 (A, B, D and F) One-way ANOVA with Tukey’s multiple comparisons test. (C and E) Two-way 989 ANOVA test with Tukey’s multiple comparisons. All data represent mean ±SEM, *P < 0.05, ***P 990 < 0.001, ****P < 0.0001. 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010

MM3MG-HER2Δ16 grow th MM3MG-HER2Δ16 grow th A B C In SCID-Beige D 800 TAMs TAMs Control IgG Control-IgG

100 1400 s L

** I 100 ****

T

) s

) CD47

 s

3 CD47

l n

3 1200 +

600 l

o

b e

i 80

m

t

m 1

c 80

a 4D5-IgG2A 1

m 4D5-IgG2A l

- 1000

m

(

D

u (

*

G

p

C e

6 60

e **** o 60

z 4D5-IgG2A +CD47 n

4D5-IgG2A + CD47 Y

i i

p 800

400 z

i

L

s

-

+

s

+

r

G b

40 r 6

0 40 o

1 600

o

8

Y

1

/

m

L

4

D

* m

u +

200 F C

20 u 400 20

T

**** 0

%

T

n

8

i /

200 4 0 F 0 0 % 7 5 7 7 5 7 G 4 4 G 4 4 0 Ig D Ig D l D 4 D 0 5 10 15 20 l D 4 D o C C o C C tr   tr   0 5 10 15 20 25 n n o + o + 5 Days post tumor injection 5 C D C D 4 4 Days post tumor injection E

4D5-IgG2A Control IgG 4D5-IgG2A 4D5-IgG2A + αCD47 Control IgG + αCD47 G

F umor

t

/C

in

Φ

Balb M

WT WT 58.2%

unlabeled unlabeled 31.6% -

9.5% 17.3% 28.0% /

-

in

Φ Fcerg

M 15.8% 13.3% labeled labeled tumor

In vivo ADCP by TAMs In vivo ADCP by TAMs 50

*** *** Control IgG )

) 60

s

s

l

l

l

l

s

e e

40 * s 4D5-IgG2A + CD47

c

e

c

e

g

g

+

+

a

a

0 0

h *

h

8 8

p n.s /

30 / 40

p

o

4

4

o

r

r

F

F

c

c

-

-

a

a

1

1

r m

20 r

m

G

G

+

+ 20

+

D

+

i

D

b

i

b D

1 10

1

D

1

%

1

%

D

D

C

C ( 0 ( 0 WT mice Fcer1g-KO G A 7 g 2 4 l I G o g D tr -I C n 5 a o D + C 4 A 2 G Ig 5- 4D 1011 1012 Figure 4. CD47 Blockade increased therapeutic efficacy of mouse Trastuzumab and 1013 augments tumor-associated macrophage (TAMs) expansion and phagocytosis. (A) Tumor 1014 growth experiment (as in Figure 1B) were repeated using CD47 blockade antibody (MIAP410, 1015 300 µg per mice) alone or in combination with 4D5-IgG2a. (B) TAM populations were analyzed 1016 by FACS after tumor volume reached >1000mm3. Analysis of additional immune cell types are 1017 shown in Figure S4D. Mean ±SEM, n = 8-10. (C) Repeat of similar tumor growth experiment 1018 and treatments in SCID-Beige mice. (D) TAM populations from SCID-Beige experiment were 1019 analyzed by FACS. n = 10. (E) Schematic representation of in vivo Antibody-dependent-cellular- 1020 phagocytosis (ADCP) experiment. MM3MG-HER2Δ16 cells were labeled with Vybrant DiD dye 1021 and implanted (1x106 cells) into mammary fat pads of Balb/c mice. Once tumor volume reaches 1022 ~1000 mm3, mice were treated with either control antibody, 4D5-IgG2A (200 µg), or in 1023 combination with MIAP410 (300 µg). On the next day, tumors were harvested and tumor- 1024 phagocytic macrophages were quantified by FACS. (F) Representative FACS plots and 1025 graphical summary showing frequency of macrophages (CD11b+, F4/80+, LY6G-, LY6C-) that 1026 have phagocytosed DiD-labeled tumor cells. n = 6. (G) Similar in vivo ADCP experiment were 1027 repeated in Fcer1g-/- mice. n = 8. (A and C) Two-way ANOVA test with Tukey’s multiple 1028 comparisons. (B, D, G and G) One-way ANOVA test with Tukey’s multiple comparisons. All data 1029 represent mean ±SEM *P < 0.05, **P < 0.01, ***P < 0.001, ****P<0.0001 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 A

B Transgenic HER2Δ16 BC mouse model C Transgenic HER2Δ16 BC mouse model Survival Tumor grow th Control 100 Control IgG

aCD47 3000 CD47

l 3

a 80

m 4D5-IgG2A v

i 4D5-IgG2A

m

v

r e u 4D5-IgG2A + aCD47 4D5-IgG2A + CD47

s 60 2000

m

t

u

l

n

o

e

v

c

40 r

r o

e ****

**** m P 1000

n.s ## u t

20

l

a

t o

0 T 0 10 20 30 40 50 0 0 2 4 6 days post tumor development weeks post tumor development

Tumor-associated macrophages

D 40 * -

s ***

C

n

6

o

i t

Y 30 ***

a

L

l

-

u

p

G

o

6

p Y

20

L

+

b

+

1

0

1

8 /

D 10

4

C

F

n

i % 0 x 7 A 7 R 4 2 4 l D G D ro C g C t  -I  n 5 o D + C 4 A 2 G g -I 5 D 4 1053 1054 Figure 5. CD47 blockade synergizes with mouse Trastuzumab therapeutic activity in a 1055 transgenic human HER2+ breast cancer (BC) mouse model. (A) Schematic representation 1056 of experiment using the endogenous human HER2 transgenic mouse model. Spontaneous 1057 breast tumors in the transgenic animals were induced with doxycycline diet. Four treatment 1058 arms were set up: Control IgG (200 µg weekly, n= 15), CD47 blockade (MIAP410, 300 µg 1059 weekly, n= 14), 4D5-IgG2A (200 µg weekly, n=16) and 4D5-IgG2A combined with MIAP410 1060 (n=16). Individual animals were consecutively enrolled into a specific treatment arm as soon as 1061 palpable breast tumors were detected (~200mm3). (B) Survival of mice in each treatment arm, 1062 time of start is on the day of palpable tumor detection and treatment enrollment. Log-rank 1063 (Mantel-Cox) test for survival analysis, ****P < 0.0001 of treatment vs control group, ## P < 0.01 1064 significant difference observed between “4D5” group vs “4D5+αCD47” group. (C) Tumor burden 1065 in animals from each treatment arm were measured over time after enrollment in treatment arm. 1066 Each individual animal develops 1 to 4 total tumors in their mammary fat pads. The total tumor 1067 burden per mice is shown. Animals were terminated when their total tumor volume 1068 reached >2000 mm3 (D) Tumors in the transgenic mice were harvested, processed into single 1069 cell suspensions, and analyzed by FACS. Each individual tumor were treated as an individual 1070 measurement. Mean ±SEM, Control IgG n=23, αCD47 n=27, 4D5 n=38, 4D5+ αCD47 n=32, 1071 One-way ANOVA with Tukey’s multiple comparisons test, *P < 0.05, **P < 0.01, ***P < 0.001. 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094

Φ M

A B Tumor Control IgG Phag Gene Category M1-like MΦ M2-like MΦ Neutr CD8/NK CD4 M1 MΦ APC

M2 MΦ Immune markers T/NK APC Tumor markers Myeloid markers

Macrophage markers Neutrophils

Phag_MΦ Neutrophil markers

Control IgG Phagocytosis genes

APC genes

M1 MΦ Inflammatory genes

M2 MΦ T/NK Complements APC B cell markers DC markers

Neutrophils T/NK cell markers Phag_MΦ

Cytotoxic genes αCD47 LY6C genes M2 markers

Wound healing Collagens Matrix M1 MΦ C Metalloproteases4D5-IgG2A + αCD47 M2 MΦ

T/NK M1-like MΦ M2-like MΦ Tumor Phagocytic MΦ CD8/CD4/NK APC Gene Category Neutrophils APC Immune markers

Tumor markers Neutrophils Myeloid markers Phag_MΦ Macrophage markers

4D5 Neutrophil markers

Phagocytosis genes

APC genes M1 MΦ

M2 MΦ T/NK Inflammatory APC genes

Complements

Neutrophils Phag_MΦ B cell markers DC markers

T/NK cell markers 4D5 + αCD47 Cytotoxic genes LY6C genes M2 markers

1095 Wound healing Collagens 1096 Figure 6. Single-cell transcriptome analysisMatrix of immune clusters within HER2+ breast Metalloproteases 1097 cancer after Trastuzumab with CD47 blockade therapy. HER2+ tumors from HER2Δ16 1098 transgenic animals were isolated for Single-Cell RNA-Sequencing using 10X Genomics 1099 platform. Data from all tumors were pooled for clustering and gene expression analysis. (A) 1100 tSNE plots showing distinct clusters of immune cells in tumors from four treatment groups: 1101 control IgG, αCD47, 4D5-IgG2A or combination. (B-C) Heat map of relevant gene markers 1102 confirmed the various immune cell clusters in control tumors (B), and the expansion of 1103 macrophage clusters in the combination therapy treated tumors (C). Macrophages that contains 1104 tumor specific transcripts (e.g. hERBB2, Epcam, Krt8) were labeled as tumor phagocytic 1105 macrophages (Phag MΦ, predominantly found in combination treatment group). 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 B C A Activation of Activation of MΦ comparisons M1-like TAM M2-like TAM

Wound

healing Interferon Interferon M2 markers ECM remodeling Growth factors

Anti-

inflammatory

inflammatory

inflammatory

-

- Pro Pro

Pro- Chemotaxis

Chemotaxis inflammatory

kB

kB

-

-

NF

NF

TLR/MyD88 TLR/MyD88

Antigen

CD47

CD47

CD47

CD47

IgG2A α

IgG2A presentation

α

-

α

α

-

4D5

4D5

Control Control IgG

Control IgGControl

IgG2A +

IgG2A + -

- Phagocytosis 4D5

4D5 & opsonization

Φ

Φ

Φ

M

M1 M M1 M2 M M2 1133 Phag 1134 Figure 7. Differential gene expression analysis of TAM clusters in HER2+ BC after 1135 Trastuzumab with CD47 blockade therapy. (A-B) Differential gene expression analysis of 1136 gene signatures for IFN, pro-inflammation, chemotaxis and TLR/MyD88/NFkb pathways in M1- 1137 like MΦ clusters (A) and M2-like MΦ clusters (B) revealed how they were affected by the 1138 treatment regimens. (C) Differential gene expression analysis of immuno-regulatory gene 1139 signatures (wound-healing, ECM remodeling, growth factors, anti-inflammation) versus immuno- 1140 stimulatory gene signatures (pro-inflammation, chemotaxis, antigen presentation, 1141 phagocytosis/opsonization) among the three distinct macrophage clusters in the combined 1142 dataset. 1143 1144 1145 1146 1147 1148 1149 CD47 High CD47 Low A CD47 High CD47 Low B subtype: ER+ subtype: HER2+ subtype: TNBC

p = 0.0038 p = 0.035 p = 0.012 p = 0.6

0 2500 5000 Time (Days) 0 2500 5000 0 2500 5000 0 2500 5000 Time (Days)

Human MDM phagocytosis of KPL-4 cells KPL-4 grow th in SCID-Beige Tumor-associated macrophages C Control IgG D Control IgG E Trastuzumab

50 *** 1500 Trastuzumab 15 ****

s l

CD47-KO + Control IgG l e

40 )

c

s

3

i -

s CD47-KO + Trastuzumab

+

m

1

r

o 5

t 1000

m 10

4

( y

30 G

c

D e

*** +

o

z

C

i

0

g

s

8

a

+

20 /

r

h

b

4

o

p 1

500 F 5

m

1

%

u %

10 D

T ****

C

n 0 i 0 0 Parental KPL4 CD47-KO KPL4 0 20 40 60 80 b G a G Ig Ig Days post tumor injection l m l o u o tr z tr n tu n o s o C ra C Sorted TAMs T + O -K F n 15 gene expression

7 o i Control Rx 4 s D

s * C e

r Trastuzumab p

x 10

e

e n

e ** *** * g *** *

e 5 *

v

i

t

a

l

e R 0 IL-6 TNFA IRF1 Cxcl2 Ccl3 TLR4 iNOS IL-10 IL-4 IRF4 CXCR4 1150 Inflammatory Anti-inflammatory 1151 Figure 8. Human CD47 gene expression is a prognostic factor in HER2+ breast cancer 1152 and limits the therapeutic activity of Trastuzumab. (A-B) Kaplan-Meier survival curve for 1153 breast cancer (BC) patients METABRIC Dataset. (A) Stratified into low and high groups based 1154 on average expression of CD47 in all patients. (B) The same patient stratification based on 1155 disease subtype (ER+, HER2+ and TNBC). (C) CD47 knockout in human HER2+ BC line KPL-4 1156 was generated using CRISPR-Cas9 approach. Control and CD47-KO KPL-4 cells were labeled 1157 with Brilliant Violet 450 Dye, and incubated with human monocytes-derived-macrophages 1158 (hMDM) at a 3:1 ratio, in the presence of control or Trastuzumab (10 µg/mL). Antibody- 1159 dependent-cellular-phagocytosis (ADCP) activity were measured by percentage of hMDM 1160 uptake of labeled KPL-4 cells (CD45+ and BV450+). Mean ±SEM, biological replicates n = 4. 1161 Experiment has been repeated using hMDMs generated from three healthy PBMC donors. (D) 1162 Control or CD47-KO KPL-4 cells were implanted into mammary fat pads of SCID-Beige Balb/c 1163 mice (5x105 cells). Trastuzumab (50 µg) or control human IgG1 were administered weekly and 1164 tumor volume were measured. Two-way ANOVA test with Tukey’s multiple comparisons, 1165 ****P<0.0001. (E) Tumor infiltrating macrophages (F4/80+ Gr1- CD11b+) populations were 1166 analyzed by FACS, except for “CD47-KO + Trastuzumab” group as no tumor growths have 1167 occurred. Mean ±SEM, n = 7. (F) Tumor-associated macrophages from control treated and 1168 trastuzumab treated tumors were sorted by FACS (F4/80+ Gr1- CD11b+CD45+) and analyzed 1169 with RT-qPCR for the expression of pro- and anti-inflammatory genes. Mean ±SEM , n=7. 1170 Multiple two-sided t-test. (C and E) One-way ANOVA test with Tukey’s multiple comparisons, * 1171 P <0.05, ** P <0.01, ***P < 0.001 1172 1173 1174

1175

1176

1177

1178

1179

1180

1181

1182

1183

1184

1185

1186

1187

1188

1189

1190

1191

1192

1193

1194 1195 Table 1

Treatment # of tumors Average length on MΦ cluster size Tumor Phagocytic MΦ analyzed treatment regimen (% MΦ among (% MΦ containing

immune cells) hERBB2 transcripts) Control 3 28 days 2354/4527 (52 %) 4.22 % αCD47 2 36 days 1228/2154 (57 %) 3.95 %

4D5-IgG2A 4 45 days 2938/3815 (77 %) 9.07 %

4D5-IgG2A + αCD47 4 56 days 4673/5079 (92 %) 48.44 %

Table 1. Single-Cell Transcriptome Analysis of macrophage cluster sizes and frequency of tumor phagocytic macrophages. Numbers represent mean of replicates in each treatment group.