Antibody therapy targeting the CD47 protein is effective in a model of aggressive metastatic leiomyosarcoma

We evaluated the level of expression of mRNA for CD47 and TAM markers CD68 and CD163 in a group of 51 cases of LMS and 19 benign leiomyomas. CD47, CD68, and CD163 expression were all significantly up-regulated on LMS cases versus benign leiomyoma cases (Fig. 1A). We next analyzed CD47 protein expression on 16 LMS, three leiomyoma, and three normal muscle samples using frozen section immunohistochemistry. Consistent with our gene-expression profiling results, CD47 expression was high on 14 of the 16 LMS samples, but low or absent on five of the six leiomyoma and normal muscle samples (Fig. 1B, Fig. S1, and Table S1).

In experimental models of leukemia, lymphoma, and bladder cancer, blocking CD47-mediated SIRPα signaling with anti-CD47 mAbs induced phagocytosis of tumor cells by human and mouse macrophages (13–17). Given the prognostic association of TAMs in LMS and the confirmation of CD47 expression in LMS patient samples, we developed an in vitro assay to study the interaction between LMS cells and macrophages with the hypothesis that anti-CD47 mAbs may enable phagocytosis of LMS cells by macrophages.

CD47 expression on human LMS cell lines LMS04 and LMS05 was confirmed using flow cytometry (Fig. 2A). Peripheral blood mononuclear cells (PBMCs) were purified from human whole blood and macrophages were enriched by plastic adherence for 1 wk. Nonadherent cells were removed and macrophage purity was confirmed using immunohistochemistry (IHC) for a panel of immune cell markers; PBMC-derived macrophages expressed CD68 and CD163 and only rare contaminating lymphocytes were seen, as evidenced by the absence of expression of CD3, CD4, CD8, and CD20 (Fig. 2B).

LMS04 and LMS05 were green fluorescently labeled and subsequently incubated with blocking (clone B6H12) or nonblocking (clone 2D3) anti-CD47 mAbs, or IgG control antibodies, before being cocultured with the PBMC-derived macrophages. B6H12-treated LMS04 and LMS05 cells were both readily phagocytosed by the human macrophages; in contrast, 2D3- or IgG-treated LMS04 and LMS05 cells were not phagocytosed by macrophages. Representative images of phagocytosis are shown for LMS05 in Fig. 2C.

To visualize this interaction in real time using live-cell imaging, red fluorescent protein (RFP)-positive macrophages were derived from the bone marrow of C57BL/K_a Rosa26-mRFP1 mice and were incubated with green fluorescently labeled LMS05 cells, which had been treated with anti-CD47 (B6H12) or IgG antibodies. Consistent with the results observed with human macrophages, the red fluorescent mouse macrophages were able to efficiently phagocytose anti-CD47–treated cells, in contrast to IgG-treated cells, which were not phagocytosed by the macrophages (Fig. 3 and Movies S1, S2, and S3).

We evaluated whether anti-CD47 mAbs could inhibit tumor growth of xenotransplanted LMS tumors in NOD/SCID/ IL-2Rγ^null (NSG) mice, which lack B, T, and NK cells but retain functional macrophages capable of phagocytosis (18). For the evaluation, 100,000 LMS04 or LMS05 cells were engrafted subcutaneously on the backs of NSG mice and, as soon as palpable tumors appeared (2 wk), mice were treated with a fully humanized anti-CD47 mAb (clone B6H12) or human IgG by intraperitoneal injections three times per week (full details of treatment protocol are in Table S2). At the conclusion of the experiment, animals were killed and tumor presence was confirmed by H&E staining and Vimentin IHC. We found that treatment with anti-CD47 mAbs dramatically inhibited tumor growth of both LMS04 and LMS05 xenotransplanted tumors, with LMS05 showing a greater than fivefold decrease in average tumor mass (t test, P = 0.0001) and LMS04 showing a greater than 30-fold decrease in average tumor mass (t test, P = 0.0188) (Fig. 4 A–D).

Histological analysis of LMS04- and LMS05-engrafted mice was performed to evaluate the presence of distant metastases. In LMS05-engrafted mice, lungs, livers, kidneys, spleens, hearts, and brains showed no evidence of metastatic disease, as evaluated by H&E staining and Vimentin IHC. In contrast, metastases developed in the in the lungs of LMS04-engrafted mice. In the IgG control-treated mice, a total of 409 lung metastases were observed in the five mice, whereas only six metastatic growths were observed in the lungs of the five mice receiving anti-CD47 mAbs, representing a nearly 70-fold decrease in the metastatic potential of LMS04 xenotransplanted tumors upon anti-CD47 mAb treatment (t test, P = 0.03) (Fig. 4E). Notably, the six pulmonary metastases observed in the anti-CD47 mAb-treated mice were all single-cell growths, whereas IgG-control–treated mice showed significant numbers of well-established multicell metastatic clusters (Fig. 4 F and G).

Distant metastases and tumor recurrences after primary tumor resections are the principal factors underlying mortality in LMS patients and represent significant challenges to the clinical management of these tumors (2, 3). We therefore sought to recapitulate the progression of the human disease by developing a metastatic xenotransplantation model of LMS. One hundred thousand LMS04 cells were engrafted subcutaneously on the backs of 14 NSG mice and individual animals were killed each week to check for the presence of lung metastases (Fig. 5A). At early time points, mice showed no evidence of lung metastases, indicating that the metastatic growths observed at later time points in the IgG-treated LMS04-engrafted mice were not resulting from the initial tumor injection procedure, but rather were originating from the primary tumors (Fig. 5B). Six weeks after subcutaneous injection of tumor cells on the back, LMS04 tumor cells started to appear as single cells in the lungs, as determined by H&E staining and Vimentin IHC. At this time the primary tumors in the remaining eight mice were resected and individual mice were killed each week to assess the progression of the metastatic lung growths. Postresection, the number and size of metastases in the lungs increased over time (Fig. 5B) and mice also developed secondary axillary lymph-node tumors, which were confirmed as originating from the LMS04 cells by H&E analysis and by Vimentin IHC (Fig. 5C).

We tested whether anti-CD47 mAb treatment could show efficacy in a model for neoadjuvant therapy using the metastatic xenotransplantation model of LMS04: 100,000 LMS04 cells were transplanted subcutaneously on the backs of NSG mice and were allowed to grow for 6 wk, the time point that we previously determined to correspond to the onset of lung metastases. Mice were then randomized based on tumor size, treated for 1 wk with anti-CD47 mAb or IgG control antibodies, and the primary tumors were surgically resected (Fig. 6A and Table S3). After resection, the antibody treatments continued; the mice were monitored daily and the presence of metastatic tumor deposits in the axillary region was recorded. All primary tumor grafts that grew on the backs of mice were lateral to the spine and metastatic deposits always occurred in the axillary area on the same side of the mouse. We therefore interpret these deposits as lymphatic metastases to axillary lymph nodes. Anti-CD47 mAb treatment increased recurrence-free survival after primary tumor resections, as measured by time to presence of axillary lymph-node tumors (Fig. S2). In addition to delaying the onset of these lymph-node tumors, anti-CD47 mAb treatment led to a nearly threefold decrease in average volume of the lymph-node tumors (t test, P = 0.05) and diminished the total number of lymph-node tumors observed. IgG treatment resulted in a total of 10 lymph-node tumors distributed over six mice and anti-CD47 mAb treatment resulting in a total of five lymph-node tumors distributed over six mice (Fig. 6 B and C).

Importantly, anti-CD47 mAb treatment almost completely inhibited the formation of lung metastases after resection of the primary tumor grafts. A computational morphometric approach was used to quantify the number of Vimentin-positive stained cells in whole sections of mouse lungs. Five control mice that did not carry LMS04 xenografts were used to determine the background staining and the six mice from each group (IgG or anti-CD47–treated) were scored. The results are shown in Table S4 and show a significant difference between the two groups. Histologic examination of the lung sections confirmed the quantitative measurement and, furthermore, showed that two types of metastatic deposits could be identified: those consisting of single cells and those that contained two or more cells in a tight cluster. Single-cell metastatic deposits were found in all 12 mice, whereas multicell metastatic clusters were found almost exclusively in the IgG control group. Two anti-CD47–treated mice (#650 and #550) showed an exception to this rule and had four and one multicell foci containing more than four cells, respectively; these foci were small (< 10 cells) and were associated with inflammatory cells and fibrin depositions, morphologic features that suggest degradation of the cells in the metastatic growths. (Fig. 6F). One of the six anti-CD47–treated mice (#540) did not show a single metastatic cell in the whole-lung sections, and the remaining five mice had scattered single cells throughout the lung parenchyma, explaining the variable degree of Vimentin reactivity measured in Table S4. Fig. 6G further illustrates the differences in lung metastases between the two groups, showing that the anti-CD47–treated mouse with the highest lung metastasis score (#548) had only scattered individual cells in the lungs, whereas the IgG control mouse with the lowest score (#2798) had four foci of multicell clusters. Together, these results indicate that anti-CD47 mAbs may be effective as neoadjuvant therapeutic treatment options with the potential to address the significant clinical problem of distant metastases after primary LMS tumor resection.

The clinicopathologic features of the 51 LMS tumor samples and of the 19 leiomyoma samples used in this study have been described previously (21). Gene-expression profiling was performed using 44 K spotted cDNA microarrays and expression values of CD47, CD68, and CD163 were compared between LMS and leiomyoma samples using Student t tests. Gene-expression data can be accessed from the Stanford Microarray Database (http://smd.stanford.edu) (22). Gene-expression studies were performed with the approval of the Stanford University Institutional Review Board.

CD47 protein expression was evaluated by fluorescent IHC on OCT-embedded, fresh-frozen tissue, as described in SI Materials and Methods. Colorimetric IHC on formalin-fixed, paraffin-embedded cell pellets from PBMC-derived human macrophages and on full cross-sections of formalin-fixed, paraffin-embedded mouse organs are also described in SI Materials and Methods.

LMS04 and LMS05 cells were cultured as described in SI Materials and Methods.

LMS04 and LMS05 cell suspensions were stained with phycoerythrin-conjugated anti-CD47 (BD PharmMingen) or isotype control antibodies. CD47 expression was analyzed by flow cytometry on a BD FACSAria instrument (Becton Dickinson).

Human macrophages were prepared from PBMCs, as described previously (14), and in vitro phagocytosis assays were performed as described in SI Materials and Methods.

RFP-positive mouse macrophages were prepared from C57BL/K_a Rosa26-mRFP1 transgenic mice (23) and live-cell imaging of macrophage phagocytosis was performed as described in SI Materials and Methods.

Primary tumor growth experiments and treatment protocols are described in SI Materials and Methods and Table S2. The development of the metastatic tumor model, and experiments evaluating the effects anti-CD47 treatment on metastatic disease, are described in SI Materials and Methods and Table S3. All procedures followed protocols approved by the Stanford Committee on Animal Research.

To assess metastatic tumor presence in the lungs, we performed IHC for human Vimentin, which does not react with murine tissue, and then performed morphometric analyses as described in SI Materials and Methods.