Blocking CD47 boosts how immune cells attack kidney cancer by nibbling tumor cells

To evaluate whether CD47 could be a potential therapeutic target for patients with RCC, we evaluated mRNA expression levels of CD47 and PD-L1 and their prognostic value in RCC using TCGA datasets. The transcriptional level of CD47 was higher than that of PD-L1 (Fig. 1a). The CD47 mRNA expression level was the highest in clear cell RCC, which had a worse prognosis than other histologic subtypes of RCC (i.e., papillary RCC, chromophobe RCC; Supplementary Fig. 1). High CD47 expression levels showed a correlation with poor overall survival, whereas PD-L1 expression did not (Fig. 1b). The same patterns in overall survival rates were consistently observed in SIRPα and PD-1, whose ligands are CD47 and PD-L1, respectively. But the opposite result was shown in THBS1 which is encoding TPS-1 (Supplementary Fig. 2). We then used the CCLE datasets to examine expression patterns of CD47 and PD-L1 in human RCC cell lines. Consistent with the results of TCGA data analysis, high CD47 expression compared to PD-L1 was detected in human RCC cell lines (Fig. 1c). Based on these transcriptional profiles from public databases, we assessed the surface expression levels of CD47 and PD-L1 in several human RCC cell lines using flow cytometry. Similar to the results from TCGA and CCLE datasets, all RCC cell lines expressed CD47 but not PD-L1 except for SNU1272 (Fig. 1d,e). CD47 was expressed at varying levels in different RCC cell lines.

Next, we evaluated whether interference with the CD47-SIRPα interaction could increase the phagocytic activity of macrophages against RCC cells using the CD47-blocking antibody IMC-002 (Fig. 2a). The phagocytosis assay was performed in vitro using monocyte-derived macrophages from human peripheral blood and analyzed using flow cytometry. An increase in phagocytosis caused by CD47 blockade was observed in all RCC cell lines expressing CD47 (Fig. 2b,c). High CD47 expression in RCC cells was correlated with an increase in phagocytosis due to IMC-002 (Fig. 2d). However, unlike CD47, no increase in phagocytosis was observed when PD-L1 was blocked with an anti-human PD-L1 antibody (Supplementary Fig. 3).

To understand the mechanism by which CD47 blockade affects phagocytosis by macrophages, we visualized phagocytosis and cell-to-cell interactions between macrophages and RCC cells using a confocal microscope. Unexpectedly, trogocytosis, in which macrophages “bite off” a part of a cell, occurred more frequently than phagocytosis, in which macrophages eat whole cells (Fig. 2e). This was observed even when the target cells were not opsonized by antibodies, that is, in the presence of the isotype control (IgG4). As a result of CD47 blockade, we detected a substantial increase in trogocytosis but only a marginal increase in phagocytosis (Fig. 2e,f).

In addition, visualization at the single-cell level confirmed that contact between macrophages and RCC cells was increased by blocking CD47 with IMC-002 (Supplementary Video 1; Fig. 3a). The frequency and duration of contact between these two cells increased significantly (Fig. 3b,c). An increase in trogocytosis under the CD47-blocking condition was observed, consistent with other in vitro assays presented (Fig. 3d). These data demonstrate that macrophages eliminate RCC cells through trogocytosis, which is augmented by sustained intercellular interactions after CD47 blockade.

We attempted to find drugs that could synergistically promote the antitumor effects of macrophages after CD47 blockade. We screened the direct cytotoxic effects of drugs on tumor cells and their effects on trogocytosis using four VEGFR TKIs that have been approved by the FDA for RCC treatment⁶. Although the direct cytotoxic effects of all VEGFR TKIs on tumor cells were low (Supplementary Fig. 4), some of the VEGFR TKIs altered the susceptibility of RCC cells to macrophages. Sunitinib (average 0.83-fold) and lenvatinib (average 1.06-fold) suppressed or did not affect trogocytosis in most RCC cell lines, except SNU4600 (Lenvatinib, average 1.73-fold). By contrast, cabozantinib (average 1.74-fold) and axitinib (average 1.67-fold) elevated trogocytosis in most RCC cell lines except SNU-1272 (average 1.08-fold, 1.26-fold; Fig. 4a). In Caki-2, trogocytosis was not increased by any VEGFR TKIs (Fold change, 0.43–1.27). To elucidate the effects of VEGFR TKIs on susceptibility to macrophages in RCC cells, we screened several molecules that act on phagocytic signaling, phosphatidylserine, calreticulin, and CD47. However, no significant differences were found (Supplementary Fig. 5).

We then assessed whether the VEGFR TKIs synergistically boosted CD47 blockade-induced trogocytosis (Fig. 4b). Cabozantinib (average 1.68-fold) and axitinib (average 1.65-fold), which increased trogocytosis as monotherapy, further elevated trogocytosis when used in combination with anti-CD47 blocking antibody in most RCC cell lines. Lenvatinib (average 1.81-fold), which did not affect trogocytosis as a monotherapy, enhanced trogocytosis when used in combination with anti-CD47 blocking antibody in most RCC cell lines except for Caki-1. Notably, the combined treatment of cabozantinib and anti-CD47 blocking antibody improved trogocytosis against Caki-2 cells, which was not affected by monotherapy (average fold change, CD47 blockade, 1.18-fold; cabozantinib, 1.21-fold; combination; 1.93-fold). However, sunitinib did not increase macrophage susceptibilities of RCC cell lines under any conditions and even abolished the trogocytic effects of CD47 blockade.

We observed variation in the increased level of trogocytosis by CD47 blockade according to the donors of the macrophages (Fig. 5a). To identify the factor causing this variation, we revisited the trogocytosis assay data and determined the relevance between the altered level of trogocytosis and several factors in the macrophages. As a result, we found a correlation between CD11b expression and trogocytosis induced by CD47 blockade. High CD11b expression in macrophages resulted in a greater increase in CD47 blocking–induced trogocytosis (Fig. 5b). This was reaffirmed by confirming that the trogocytosis increased by CD47 blockade was completely abrogated by anti-CD11b blocking antibody. However, trogocytosis increased by axitinib, one of the VEGFR TKIs, was not abolished (Fig. 5c). Collectively, these results imply that CD11b contributes CD47 blockade-induced trogocytosis of macrophages.

We next examined the correlation between altered trogocytosis by CD47 blockade and SIRPα, a receptor of CD47. Unlike CD11b, the expression level of SIRPα and the increased level of trogocytosis induced by CD47 blockade were not related (Supplementary Fig. 6). Macrophages used in this study were classified as variant 1 or variant 2/7 by genotyping of SIRPA^35,36. We re-analyzed the trogocytosis changes induced by CD47 blockade according to the variant types of SIRPα and found no association (Supplementary Fig. 6).

The Cancer Genome Atlas (TCGA) kidney cancer RNA sequencing datasets were obtained from UCSC Xena (https://xenabrowser.net↗)⁴⁶. Fragments per kilobase of transcript per million (FPKM) data were downloaded from the KIRC (n = 607), KIRP (n = 321), and KICH (n = 89) datasets, converted to log₂(FPKM + 1), and analyzed. To analyze the correlation between overall survival and gene expression level, patients with RCC were assigned to either gene-high or gene-low groups based on the median value of gene expression. Additionally, survival analysis was performed with the survival data of each patient. The kidney cancer cell line RNA sequencing datasets were obtained from the Cancer Cell Line Encyclopedia (CCLE)⁴⁷. Transcripts per million (TPM) data obtained from CCLE were converted to log₂(TPM + 1) and analyzed.

The human RCC cell lines used in this study, ACHN, A498, A704, Caki-1, Caki-2, SN12C, SNU-1272, and SNU4600, were purchased from the Korean Cell Line Bank (Seoul, Korea). ACHN, A498, A704, Caki-1, Caki-2, and SN12C cell lines were cultured in Dulbecco’s modified Eagle medium supplemented with 10% heat-inactivated fetal bovine serum (FBS; Gibco, Waltham, MA, USA) and 10 µg/ml gentamicin (Gibco). SNU-1272 and SNU-4600 cell lines were cultured in RPMI 1640 medium supplemented with 10% heat-inactivated FBS and 10 µg/ml gentamicin.

The VEGFR TKIs used in this study (sunitinib malate, cabozantinib, lenvatinib, and axitinib) were purchased from Selleck Chemicals (Boston, MA, USA). IMC-002, an IgG4 monoclonal antibody targeting human CD47, was kindly provided by ImmuneOncia Therapeutics Inc. (Yongin, Korea). Anti-human CD11b blocking antibody, IgG4 isotype control antibody, and IgG2b isotype control antibody were purchased from BioLegend (San Diego, CA, USA).

Cell surface molecules on tumor cell lines and human macrophages were measured by flow cytometry using the following fluorochrome-conjugated antibodies: CD47-PE, CD47-APC, PD-L1-PE, and CD11b-PE (BD Biosciences, Franklin Lakes, NJ, USA). CD172a-PerCP-eF710 was purchased from eBioscience (Thermo Fisher Scientific, Waltham, MA). Flow cytometry analysis was performed using a FACS Calibur or FACS Canto II (BD Biosciences), and the data were analyzed using FlowJo v10.7 software (FlowJo LLC, Ashland, OR, USA, https://www.flowjo.com/↗).

Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll density gradient separation from healthy donor peripheral blood collected using a leukocyte reduction system chamber. Monocytes were isolated from PBMCs by negative selection using a pan monocyte isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer’s recommendations. Freshly isolated monocytes were cultured in RPMI 1640 medium supplemented with 10% heat-inactivated FBS, 1% penicillin/streptomycin, 2 mmol/L L-glutamine (Gibco), and 25 ng/ml recombinant human macrophage colony-stimulating factor (Gibco). The medium and cytokines were replaced every 2–3 days, and differentiated macrophages were harvested on day 7 to be used for the functional assay.

Target cells were labeled with cell proliferation dye eFlour 670 (Invitrogen, Carlsbad, CA, USA) or CellTrace CFSE (Invitrogen) and preincubated with 5 µg/ml IMC-002 or Ultra-LEAF purified human IgG4 isotype control recombinant antibody (BioLegend), respectively. For an E:T ratio of 1:1, 5 × 10⁴ target cells and human macrophages were cocultured in a u-bottom 96-well plate. After 2 h of incubation, whole cells were collected in round-bottom test tubes and stained with CD11b-PE (BD Biosciences) or CD11b-APC (BD Biosciences). The stained samples were analyzed using a FACS Canto II (BD Biosciences) on the same day after staining. Then, the data were analyzed using FlowJo v10.7 software (FlowJo LLC). To confirm the effect of VEGFR TKIs on phagocytosis, target cells were plated at 2 × 10⁵ cells/well in 6-well plates. After allowing the cells to adhere overnight (16–18 h), they were treated with VEGFR TKI (10 µM) for 24 h. Thereafter, the phagocytosis assay proceeded in the same manner as described above.

The cells were labeled with 1 µM CellTrace Far Red or CellTrace CFSE at 37 °C for 10 min. After washing once with complete media, the cells were labeled with 3 µM Hoechst to stain the nuclei at 37 °C for 15 min. For an E:T ratio of 1:1, 5 × 10⁵ labeled macrophages and CaKi-1 cells were cocultured in FACS tubes. After 4 h of incubation, the cells were collected and placed in a confocal dish for imaging. A confocal laser scanning microscope (FV 1200, Olympus) with a 40 × objective lens (UPLSAPO 40 × 2; Numerical Aperture = 0.95) was used. The acquired images were processed and analyzed using ImageJ (https://imagej.nih.gov/ij/↗).

A modified Olympus IX83 epifluorescence microscope with a 40× objective lens (UPlanFLN; NA = 1.30) and an ANDOR Zyla 4.2 sCOMS camera were used for imaging experiments. For fluorescence imaging, a U-LH75XEAPO Xenon lamp (75 W, Olympus) and yellow (EX BP 530/30, BS 550, EM BP 575/40) and Cy5 (EX BP 620/60, BS 660, EM BP 770/75) filters were used. The microscope was automatically controlled by Micro-Manager, and the acquired images were analyzed and processed with ImageJ.

To observe an interaction between a macrophage and cancer cell, microwell arrays were fabricated using double replicas of a silicon master fabricated by standard photolithography. In addition, 85 µm × 50 µm square microwell arrays were fabricated on a silicon wafer with negative photoresist SU-8 50 and developer. The silicon master was treated with trichloro(1H,1H,2H,2H-perfluorooctyl) silane. Then, we poured the polydimethylsiloxane precursor mixture (Sylgard 184; base:curing agent = 10:1) on the silicone master and cured it at 70 °C for 4 h to replicate. The replicated polydimethylsiloxane mold was placed on a glass surface functionalized with an acrylate group, perfused with the precursor solution poly(ethylene glycol) dimethacrylate, and cured with a UV light. Hydrogel microwells were treated with air plasma (100 W) for 1 min and coated with 2 µM fibronectin solution at 37 °C for an hour. Macrophages and cancer cells were labeled with 1 µM CellTrace Far Red or CellTrace Yellow at 37 °C for 10 min, respectively. These labeled cells were loaded on the fibronectin-coated microwell array (1 × 10⁵ cells/ml). Time-lapse imaging was started 1 h from cell addition to wait for the cells to spread to the surface.

Cell viability assays were performed as in the previous study⁴⁸. The RCC cell lines were plated at 3,000 cells per well in 96-well plates. The cells were exposed to various VEGFR TKIs (sunitinib, cabozantinib, lenvatinib, and axitinib) for 72 h in eight conditions from 0 to 10 μM in 1/2 serial dilutions. Dimethyl sulfoxide solution (Gibco) was used as a control. Cell viability was analyzed using a CellTiter-Glo Luminescent cell viability assay kit (Promega, Madison, WI, USA). The luminescent signal was measured using a GloMax Discover microplate reader (Promega).

Genomic DNA was extracted from healthy donor PBMCs using a GeneAll Exgene Cell SV kit (GeneAll Biotechnology, Seoul, Korea). Exon 3 of SIRPA was amplified using EconoTaq PLUS GREEN 2X master mix (Lucigen Corporation, Middleton, WI, USA). The PCR thermal cycle was conducted on a Veriti 96-well thermal cycler (Applied Biosystems, Waltham, MA, USA) as follows: initial denaturation at 94 °C for 2 min; 35 cycles of 94 °C for 30 s, 58 °C for 30 s, and 72 °C for 1 min 40 s; final extension at 72 °C for 7 min. Amplicons were purified using Wizard SV gel and a PCR clean-up system (Promega), and the sequences were confirmed by Sanger sequencing. The primers used for PCR amplification and Sanger sequencing are as follows: PCR, (forward) 5′-AAACACACTGGCACGAGTCTA-3′, (reverse) 5′-GCCTTAACGGAGGAACCCAA-3′; Sanger sequencing, (forward) 5′-AGAATACAGGCTCATGTTGCAGGT-3′, (reverse) 5′-GCCTTCAGCAAATAGCATGACGT-3′.

Statistical significance was analyzed by GraphPad Prism v.7.0 (GraphPad Software, San Diego, CA, https://www.graphpad.com/↗). Data are presented as the means ± SD. Student two-tailed t test were used to compare two groups. P values < 0.05 were considered significant. *P < 0.05; **P < 0.005; ***P < 0.001; ****P < 0.0001.

As mentioned in the Methods, the main datasets used and analyzed in this study are available in the USCS Xena, (https://xenabrowser.net/datapages/↗) GDC TCGA Kidney cancer datasets. Gene expression data of cell lines (CCLE dataset) were obtained from the depmap portal, (https://sites.broadinstitute.org/ccle/datasets↗) DepMap Public 22Q2 Primary Files, CCLE_expression.csv. All other data used in this study are available from the corresponding authors for reasonable purposes.