Combining CD47 blockade with trastuzumab eliminates HER2-positive breast cancer cells and overcomes trastuzumab tolerance

In this study, we sought to determine the contribution of macrophage-mediated ADCP to trastuzumab efficacy, separate from NK-mediated ADCC. To that end, we first took HER2⁺ breast cancer cells that are sensitive to both ADCP and ADCC and generated cell lines that are tolerant to ADCC. ADCC-tolerant breast cancer cell lines were derived from two parental HER2⁺ breast cancer cell lines, SKBR3 and BT474, through repeated exposure (10 rounds of selection) to opsonizing concentrations of trastuzumab in the presence of human peripheral blood mononuclear cells (PBMCs) with an effector:target ratio of 100:1. This selection and enrichment process yielded sublines that, unlike the parental lines, demonstrated relative (though not absolute) resistance to trastuzumab-mediated ADCC. To test whether trastuzumab could still bind to the cells, we stained the parental as well as the resistant sublines with trastuzumab and found comparable levels (SI Appendix, Fig. S1). This confirmed that the mechanism of ADCC resistance is not necessarily the loss of the epitope recognized by trastuzumab, thereby suggesting that ADCC and ADCP are two independent mechanisms impacting trastuzumab efficacy. To confirm that relative resistance to ADCC was stable upon passaging, the cell lines were subjected again to an ADCC assay. Parental and ADCC-selected SKBR3/BT474 breast cancer cells were stained with the fluorescent dye 5-(and 6)-Carboxyfluorescein diacetate succinimidyl ester (CFSE) and preopsonized with either trastuzumab or PBS. Next, the breast cancer cells were cocultured with primary human PBMCs or with purified primary human NKs at various effector:target cell ratios. Cells were then stained with DAPI and subjected to flow cytometry analysis to determine the percentage of DAPI⁺ and CFSE⁺ cells remaining in the PBS or trastuzumab-treated conditions. As can be seen in Fig. 1, the parental lines BT474 and SKBR3 were killed more efficiently by both human PBMCs and NK cells compared to ADCC-tolerant sublines. The tolerance to ADCC was most pronounced when purified NK cells were used as effector cells.

After confirming that the SKBR3 and BT474 ADCC-selected sublines demonstrated relative resistance to trastuzumab-mediated ADCC, we next asked whether these cell lines retained their sensitivity to ADCP. SKBR3 and BT474 parental and the ADCC-selected sublines derived from them were treated with either PBS, IgG4, trastuzumab, Hu5F9-G4 (anti-CD47 monoclonal antibody), or trastuzumab and Hu5F9-G4 (combination). We conducted the in vitro phagocytosis assays using macrophages from two different sources: 1) human peripheral blood monocyte-derived macrophages (Fig. 2 C–E); and 2) nonobese diabetic scid-γ (NSG) bone marrow-derived macrophages (BMDMs) (Fig. 2 A and B). We chose NSG BMDMs for our in vitro phagocytosis assays to be consistent with our subsequent use of NSG mice in our in vivo xenograft assays. We observed similar results using both of these macrophage preparations. Breast cancer cells were labeled with CFSE and incubated with differentiated macrophages for 2 h. After the incubation, mouse BMDM were stained with a fluorescent F4/80 antibody and cells were subjected to flow cytometry. Phagocytosis was determined by the percentage of cells that were both F4/80⁺ and CFSE⁺. SKBR3 and BT474 parental and ADCC-tolerant lines demonstrated increased susceptibility to ADCP when opsonized with combination treatment of trastuzumab and Hu5F9-G4 compared to treatment with trastuzumab or Hu5F9-G4 alone (Fig. 2 A and B). Similarly, phagocytosis assays with SKBR3 and BT474 parental and ADCC-tolerant cell lines as target cells were incubated with human PBMC-derived macrophages for 2 h. Target cells were labeled with CFSE and human macrophages were stained with fluorescent anti-CD45. Phagocytosis was determined by the percentage of cells that were both CD45⁺ and CFSE⁺. Breast cancer cells that were treated with the combination treatment demonstrated a significant increase in their susceptibility to macrophage-mediated phagocytosis compared to those treated with a single intervention (Fig. 2 C and D). To assess the levels of surface CD47 across all cell lines used, we stained cells with anti-CD47 and used flow cytometry to compare the two parental lines to the ADCC-tolerant sublines derived from them; we found no significant differences in CD47 levels (SI Appendix, Fig. S1).

Given the differing susceptibilities of SKBR3 and BT474 ADCC-tolerant sublines to ADCP versus ADCC, we next asked whether the increased ADCP was due to an Fc-dependent mechanism. To test this, we generated a trastuzumab F(ab′)₂ fragment and found that the Fc portion of trastuzumab is necessary to generate the enhanced ADCP in the combinatorial treatment. In the absence of trastuzumab’s Fc domain, the combination of Hu5F9-G4 and trastuzumab F(ab′)₂ fragment produced significantly less ADCP than the combination of Hu5F9-G4 and trastuzumab with its Fc portion intact (Fig. 2E). Taken together, these results demonstrate that the susceptibility of BT474/SKBR3 ADCC-tolerant sublines to ADCP depends, at least in part, on engagement of Fc-receptors on macrophages.

To determine the impact of anti-CD47 combination with trastuzumab in vivo, we chose to use NSG mice that are devoid of T and B lymphocytes, have no NK cells, and cannot generate adaptive responses—but are macrophage-replete. This experimental xenograft model allowed us to evaluate the effect of macrophage innate immunity on human tumor cells growing in an animal that only has innate effector cells. Importantly, although CD47–SIRPα interactions are generally highly species-specific, we know based on previous studies that the NSG mouse version of SIRPα binds human CD47 with high affinity, so we had an opportunity to test human-relevant treatment efficacy with human HER2⁺ breast cancer xenografts. We first sought to establish how the parental SKBR3/BT474 breast cancer cell lines respond to the different treatment combinations. GFP-Luciferase–labeled BT474 parental cells were engrafted into the left mammary fat pads of female NSG mice. Baseline tumor burden was established using in vivo bioluminescence imaging (“starting range”). Twenty-five days after engraftment, mice were treated with the aforementioned single treatments or combinatorial treatment and tumor growth was monitored for a period of 17 wk. At week 7, treatments were stopped to determine whether the tumors would progress in the absence of treatment. Treatment with Hu5F9-G4 maintained tumor burden within the starting range that further progressed upon stopping treatment (Fig. 3A). Trastuzumab treatment resulted in lower tumor burden at starting range but also demonstrated progression once treatment was stopped. In the combinatorial treatment however, tumors were significantly below the starting range and did not show signs of tumor progression within the 10-wk nontreatment period following the last dose.

Using a similar approach, we next tested the susceptibility of BT474 trastuzumab ADCC-tolerant cell line xenografts (Res 1 and Res 2) in vivo. Treatment with Hu5F9-G4 alone resulted in maintenance of tumor burden within the starting range and treatment with trastuzumab alone had significant efficacy that was enhanced by the addition of the anti-CD47 Hu5F9-G4 (Fig. 3 B and C). Notably, the cancer cell line derivatives were selected in the presence of trastuzumab and PMBCs. They are therefore relatively resistant to trastuzumab-mediated ADCC as seen in Fig. 1, but remain sensitive to trastuzumab-mediated ADCP as evidenced by the in vitro phagocytosis assays as well as the in vivo treatment.

Taken together, these results indicate that the combination treatment inhibits tumor growth more significantly than either treatment alone in the parental BT474 cell lines. And, the combination treatment’s inhibition of tumorigenesis is maintained well after the combination treatment is stopped. Although the combination treatment’s benefit was less pronounced in the ADCC-selected BT474 Res 1 and Res 2 lines, Fig. 3 D and E demonstrates that the combination treatment provided a survival benefit over the single-agent antibody-alone control treatments. The greater combined in vivo efficacy with anti-CD47 plus trastuzumab is therefore consistent with our hypothesis—based upon our in vitro models—that ADCP in the combination-treated group is significantly greater than in the mice receiving either PBS or single-agent anti-CD47 or anti-HER2 antibody treatments. Given the combination treatment’s ability to inhibit tumorigenesis as well as increase survival rates, these results support the scientific rationale for using such combination therapy in future translational studies.

The growth inhibition we observed in our in vivo xenograft assays could be due to macrophage-mediated cell removal. Alternatively, it could be caused by reduced cellular proliferation. To tease out these possibilities, we conducted an in vivo innate clearance assay (depicted in Fig. 4A). Notably, we designed this assay to be performed within 4 h. This timeframe is long enough to allow us to test whether the combination treatment would lead to increased cellular removal of parental and ADCC-tolerant cells in an in vivo setting where both ADCC and ADCP were possible. This timeframe, however, is not long enough to allow for any meaningful cellular proliferation. This assay thus allowed us to focus on the former inquiry relating to whether treatment-induced tumor growth inhibition resulted from cellular removal, rather than the latter question concerning cellular proliferation. Briefly, the BT474 parental, BT474 Res 1, BT474 Res 2, and BT474 Res 3 were stained with CellTrace yellow, CFSE, CellTrace Far red, and CellTrace violet, respectively. The cell lines were then opsonized with IgG1+IgG4 (a control), Hu5F9-G4, trastuzmab, or the combination treatment. The parental and three ADCC-tolerant cell lines were then combined (based on their respective antibody treatments), intraperitoneally injected in equal starting numbers into immunocompetent C57BL/6 mouse, and incubated for 4 h. The cells were retrieved using an intraperitoneal lavage technique and analyzed with flow cytometry to determine the fraction of cells recovered from each of the cell lines.

As can be seen in Fig. 4B, the combination treatment led to a statistically significant decrease in the percentage of recovered cells versus either Hu5F9-G4 or trastuzumab treatment alone. This was true across all cell lines, parental and ADCC-tolerant (except for BT474 Res 2). This short-term in vivo Winn assay thus demonstrates that the combination of trastuzumab and magrolimab leads to rapid clearance of intraperitoneally injected cancer cells, thereby suggesting that the enhanced efficacy is due to cell removal—likely by peritoneal macrophages. The combination treatment leads to indiscriminate clearance of parental and ADCC-tolerant cell types alike, despite the relative resistance of the sublines to ADCC and other effects of trastuzumab. These results further underscore the potential utility of using the combination therapy in clinical trials, especially in the context of advanced-stage HER2⁺ breast cancer patients whose tumors have become clinically resistant to trastuzumab. Our results suggest the combination of anti-CD47 with anti-HER2 antibody trastuzumab treatment may help resensitize refractory HER2⁺ breast cancers to trastuzumab treatment.

BT474 and SKBR3 parental cell lines were purchased from ATCC. BT474 and SKBR3 ADCC-resistant lines 1, 2, and 3 were generated in the laboratory of M.D.P. Briefly, cells were separately cultured in the presence of a clinically therapeutic dose (100 μg/mL) of trastuzumab. Next, these trastuzumab-treated breast cancer cell lines were cocultured with human PBMCs. Upon coculture with human PBMCs, the majority of SKBR3/BT474 breast cancer cells were eliminated. Those breast cancer cells that survived were expanded to cellular confluency. Afterward, this process of trastuzumab treatment followed by human PBMC coculture was repeated 10 times.

NSG and C57BL/6 mice were bred and maintained at the Stanford University Research Animal Facility in accordance with the Administrative Panel on Laboratory Animal Care.

Trastuzumab was obtained from Stanford University Hospital (Genentech). Trastuzumab F(ab′)₂ fragments were generated using standard protocols as previously described. IgG1was obtained from BioXCell. IgG4 was obtained from BioLegend. Hu5F9-G4 was obtained from Lonza. Data were acquired using a BD LSR Fortessa Analyzer and analyzed using FlowJo software.

Human PBMCs were isolated using a Ficoll-Paque gradient (GE Healthcare Bio-Sciences) from human blood obtained from the Stanford University Medical Center. Human NK cells were isolated using an EasySep Human NK Cell Isolation Kit (Stem Cell Technologies).

Whole bone marrow cells were isolated from NSG mice. Monocytes were terminally differentiated into macrophages by incubating whole bone marrow in Iscove’s Modified Dulbecco’s Medium (IMDM) containing 10% FBS supplemented with 10 ng/mL recombinant murine macrophage colony stimulating factor (M-CSF; Peprotech) for 7 to 10 d and harvesting the adherent fraction.

To prepare human macrophages from PBMCs, normal human blood was obtained from the Stanford University Medical Center and PBMCs were isolated therefrom using a Ficoll-Paque gradient (GE Healthcare Biosciences). Monocytes were further isolated from the PBMC fraction by incubating PBMCs in cell culture plates for 1 h at 37 °C. Nonadherent cells were discarded by washing the cell culture plates with PBS. Adherent cells were incubated for 7 to 10 d in IMDM containing 10% human serum AB (Gemini Bio-Products) to allow for terminal differentiation of monocytes to macrophages.

BT474 and SKBR3 cells were separately labeled with 5 µM CellTrace CFSE according to the manufacturer’s protocol (Thermo Fisher Scientific). Flow cytometry-based cytotoxicity assay was predicated on Edri-Brami et al. (55). For the ADCC assay, 2.5 × 10³ CFSE-labeled SKBR3/BT474 cells were plated in individual wells of a 96-well ultralow attachment plate (Corning) in RPMI containing 10% heat-inactivated FBS (heat inactivation was achieved by incubating FBS at 56 °C for 30 min).

BT474 and SKBR3 cells were separately labeled with 5 µM CellTrace CFSE according to the manufacturer’s protocol (Thermo Fisher Scientific). For the phagocytosis assay, 1 × 10⁵ CFSE-labeled SKBR3/BT474 cells were then plated in each well of a 96-well ultralow attachment plate (Corning) in serum-free IMDM. Isotype antibody control, trastuzumab, anti-CD47 monoclonal antibody (Hu5F9-G4), and trastuzumab plus Hu5F9-G4 combination treatment were added to the CFSE-labeled SKBR3/BT474 cells at a concentration of 10 µg/mL. The cells and antibody treatments were incubated at 37 °C for 30 min. During the incubation period of the cells and antibody treatments, NSG and human-derived macrophages were harvested via incubation in TrypLE (ThermoFisher Scientific) for 5 min followed by gentle scraping. After the cells and antibody treatments had incubated for 30 min, 5 × 10⁴ macrophages were added each experimental well. The macrophages and SKBR3/BT474 cells were cocultured at 37 °C for 2 h. Following the 2-h coculture, macrophages were stained with either APC-conjugated F4/80 (for NSG mouse macrophages; BioLegend) or APC-conjugated CD45 (for human macrophages; BioLegend). The SKBR3/BT474 plus macrophage cocultures were subsequently analyzed using a BD LSR Fortessa Analyzer. The percent phagocytosis was calculated as the percentage of GFP⁺ SKBR3/BT474 cells contained within APC⁺ macrophages (i.e., double-positive cells).

Lentiviral production of a pCDH-CMV-EF1-puro construct (Systems Bioscience) containing a ubiquitin promoter driving the expression of a fusion protein containing the Luc2 (pgl4) luciferase gene (Promega) and the eGFP gene (Becton Dickenson) was carried out using standard protocols. BT474 parental and ADCC-resistant cells were transduced with lentivirus. GFP⁺/Luciferase⁺ BT474 cells were suspended in RPMI containing 25% Matrigel (BD Biosciences) and 1 × 10⁵ cells were implanted into the left mammary fat pad of 4- to 8-wk-old NSG female mice. Bioluminescent activity was visualized in vivo after D-luciferin injection (Biosynth) on an IVIS Spectrum (Caliper Life Sciences) instrument and quantified using Image 4.0 software. Total flux (photons/second) values were obtained from the mammary fat pad region. Mice were matched based on total flux 25 d after cell engraftment and subsequently treated weekly via intraperitoneal injections of the following treatments: trastuzumab was administered via intraperitoneal injection at 100 μg, once a week; Hu5F9-G4 was administered via intraperitoneal injection at 250 μg, every other day; PBS was administered via intraperitoneal injection at 100 μL, once a week. Treatments were stopped after 7 wk. Mice were imaged every week and differences in tumor growth were assessed using Student’s t test.

BT474 cells were separately labeled with the following according to the manufacturer’s protocol (Thermo Fisher Scientific): BT474 parental cells were labeled with 5 µM CellTrace yellow, BT474 Res 1 cells were labeled with 5 µM CellTrace CFSE, BT474 Res 2 cells were labeled with 1 µM CellTrace Far red, BT474 Res 3 cells were labeled with 5 µM CellTrace violet. The four differently labeled cell lines were subsequently mixed in a 1:1 ratio and incubated with the following antibody treatments for 20 min at 37 °C: PBS, IgG1+IgG4 isotype control (10 µg/mL, each), trastuzumab (10 µg/mL), Hu-5F9-G4 (10 µg/mL), trastuzumab plus Hu-5F9-G4 (10 µg/mL, each). The labeled cells were subsequently injected into the intraperitoneal space of C57BL/6 female mice. All injections were performed in 0.5 mL in PBS, 1 × 10⁶ of each BT474 cell line per mouse. Four hours after the injections, mice were killed and the peritoneums were washed out with PBS. The peritoneal lavage was subsequently analyzed by flow cytometry.