Macrophages show higher levels of engulfment after disruption of cis interactions between CD47 and the checkpoint receptor SIRPα

Immune cells come in frequent contact with cells that either originate from the same organism, and are thus ‘self’, or else are ‘foreign’, such as microbes. ‘Self ’ recognition occurs in part through interactions that ultimately inhibit immune activation. An important clinical example in anti-cancer therapy is the inhibitory T-cell receptor PD-1 (also known as PDCD1) that interacts with PD-L1 (also known as CD274) on ‘self’ cells in parallel with T-cell receptor interactions; if the latter activates a T-cell to attack, then PD-1 engagement effectively pacifies the T-cell. Blocking this ‘checkpoint’ by systemic injection of inhibitory antibodies (e.g. anti-PD-1 antibody) leads to elimination of tumors in a fraction of otherwise untreatable patients, with the best responses occurring against tumors that are highly mutated (i.e. ‘foreign’) (Mandal et al., 2019). Activation-dominated-by-inhibition applies also to macrophages; these phagocytes express the inhibitory receptor SIRPα (signal regulatory protein-α, CD172A) that engages in trans with the ‘marker of self’ CD47, which is expressed on all cells (Oldenborg et al., 2000). CD47 is also present on the macrophage, which raises the possibility of cis binding of SIRPα to CD47 within the same membrane, as has been postulated for other receptors on other immune cells (Doucey et al., 2004).

When a macrophage engages a ‘self’ cell, a phagocytic synapse forms in which trans binding of SIRPα to CD47 leads to local accumulation of SIRPα and phosphatase-mediated signaling that opposes engulfment (‘eating’) of ‘self’ cells including tumor cells (Fig. 1Ai, upper) (Oldenborg et al., 2000; Tsai and Discher, 2008; Weiskopf et al., 2013; Chowdhury et al., 2019). Phagocytosis-activating interactions occur in parallel, with the clearest pathway involving Fc-receptors (FcRs) that bind and organize IgG-type antibodies on a target cell (Lopes et al., 2017; Bakalar et al., 2018). IgG-opsonized targets activate the actomyosin tension that makes phagocytic internalization a highly efficient process – especially when CD47 is absent from a phagocytic target or else blocked with anti-CD47, including Fc-deficient F(ab′)₂ (Tsai and Discher, 2008; Sosale et al., 2015). However, any effect of anti-CD47 binding to macrophage CD47 remains unclear.

One humanized anti-CD47 IgG has been infused recently at high doses into lymphoma patients in combination with a second IgG (i.e. anti-CD20, Rituximab) that opsonizes the cancer cells to activate macrophages. 32% of all patients were found to exhibit a complete response (Fig. 1Ai, lower) (Advani et al., 2018), similar to that achieved when blocking PD-1 on other tumor types (Borghaei et al., 2015). Safety concerns with intravenous anti-CD47 include loss of blood cells, especially loss of red blood cells (RBCs), which is seen across various clinical trials (Sikic et al., 2018; summarized in Andrechak et al., 2019) (Fig. 1Aii). Anti-CD47 binds CD47 on all cells, and injection of anti-mouse CD47 antibody in common mouse strains also causes anemia (Ingram et al., 2017), as does CD47 knockout in at least one strain of mouse (Oldenborg et al., 2002). Splenic macrophages engulf CD47-deficient RBCs when injected in normal mice (Oldenborg et al., 2000), consistent with anti-CD47 blockade (Ingram et al., 2017), but no clinical or mouse study has determined the factors that activate blood cell engulfment. We hypothesized that CD47 is highly abundant on macrophages and that it modulates phagocytosis through a cis interaction with SIRPα. The following results show that either blocking or depleting CD47 on macrophages causes these cells to engulf more targets in phagocytosis assays, and depletion also suppresses a basal level of inhibitory signaling by SIRPα. Studies of binding and of a SIRPα-expressing cancer line additionally prove consistent with molecularly detailed simulations of a cis interaction between CD47 and SIRPα on the same cell surface.

To first determine relative levels of CD47 and SIRPα on diverse cell types, primary antibodies labeled with fluorescein were used in flow cytometry quantification (Fig. 1B; Fig. S1A). Relative to CD47 levels on human monocytic THP-1 cells, CD47 levels were within 2–3-fold for most other hematopoietic and non-hematopoietic cell types. RBCs are an exception, but ∼10-fold higher levels are consistent with an 80–90% excess of CD47 relative to levels needed to inhibit engulfment by macrophages (Tsai and Discher, 2008). SIRPα expression was expectedly restricted among hematopoietic cells to phagocytes, including human THP-1 macrophages, cytokine (granulocyte colony stimulating factor; G-CSF)-differentiated human stem cells (HSCs+G-CSF), and peripheral blood mononuclear cells (PBMCs) (Fig. 1B); the SIRPα levels found on human mesenchymal stem cells (MSCs) and A549 lung cancer cells agrees with those found in past studies (Vogel et al., 2003; Sosale et al., 2016). Transcriptome analyses further showed that SIRPα was expressed in some mouse cancer cell types as well as primary mouse macrophages, which have an expression profile similar to THP-1 macrophages (Fig. S1B). Nonetheless, the protein analyses showed that CD47 exceeds SIRPα by 3:1 or greater for all cell types, and so all of these cell types should robustly signal ‘self’ to macrophages.

To study anti-CD47 blockade effects on phagocytosis of human RBCs (hRBCs) by human macrophages, fresh hRBCs were pre-incubated with opsonizing anti-hRBC at various levels, with or without saturating levels of anti-CD47, and then added to THP-1 macrophages (Fig. 1C; Fig. S2A–D). Engulfment of opsonized hRBCs increases with anti-CD47 treatment, and saturation binding curves reveal a ∼2-fold higher half-max activity and >10% higher saturation (Δ_sat) (Fig. 1C). However, for hRBCs without any opsonization (i.e. anti-hRBC=0), anti-CD47 does not promote phagocytosis (Fig. 1C; Fig. S2D). This is surprising because if freshly isolated RBCs display an endogenous opsonizing signal of relevance to the effects of systemic blockade, then anti-CD47 should increase in vitro engulfment. The anemia and loss of other blood cells in clinical blockade of CD47 (Fig. 1Aii) thus motivate a search for other mechanism(s).

Anti-CD47 infused intravenously should bind CD47 on macrophages in the spleen, among other accessible tissues. THP-1 macrophages were therefore pre-incubated with anti-CD47, and excess antibody was removed before assaying phagocytosis. Engulfment of hRBCs increased even with non-opsonized hRBCs (Fig. 1D, left; Fig. S2A,B), and addition of anti-hRBC to hRBCs also increased engulfment, with maximal levels seen for anti-CD47 pre-bound to both THP-1s and hRBCs (Fig. 1D, right; Fig. S2A). The increased percentage of macrophages engulfing hRBCs was also accompanied by more hRBCs being engulfed per macrophage (Fig. S2A, top); THP-1s only phagocytosed ∼1 hRBC in the absence of any added antibodies, whereas ∼2 hRBCs were often engulfed per macrophage under conditions of maximal engulfment. Regardless, pre-binding of anti-CD47 to THP-1 macrophages tended to increase the percentage of engulfment (i.e. percentage of cells with internalized particles) by an amount similar to that achieved with anti-CD47 blockade on fully opsonized hRBCs (Fig. 1D, Fig. S2C). The typical increase (∼Δ_sat in Fig. 1C,D) underscores the relatively large effect of blocking CD47 on the macrophage. Therefore, hyper-phagocytosis of opsonized targets is a result of blockade of both the target and the macrophage, based potentially on inhibition of the cis CD47-SIRPα interaction.

Anti-CD47 on a macrophage could conceivably dissociate and bind an RBC target and vice versa – although symmetric saturation of both cells will eliminate exchange and was seen to maximize engulfment (Fig. 1D). To achieve a more-stable asymmetric blockade effect, CD47 was suppressed in THP-1 macrophages using shRNA, without affecting SIRPα (Fig. S3A,B). Phagocytosis was assayed (again per macrophage as in Fig. S2A, top) with IgG-opsonized sheep red blood cells (ShRBCs) as targets for knockdown (KD) and wild-type (WT) THP-1 macrophages; it is already known that sheep CD47 does not bind human SIRPα (Tsai and Discher, 2008). ShRBCs were engulfed ∼30% more per KD macrophage compared to WT and to KD controls (Fig. 2A; Fig. S3C). Compared to the ShRBCs, opsonized human-RBCs show a decrease in ‘self’-inhibited background engulfment by both WT and KD THP-1 macrophages, following the expected trends (Fig. S3D). IgG-opsonized microbeads were likewise engulfed at higher levels per KD macrophage (Fig. 2A). Macrophages are thus hyper-phagocytic if their CD47 is either blocked with an antibody (Fig. 1D) or knocked down, especially when CD47 on the target is lacking.

Phosphorylation of tyrosine residues in the cytoplasmic immunoreceptor tyrosine-based inhibiting motif (ITIM) of SIRPα increases when CD47 on a phagocytic target binds SIRPα in trans, with phospho-SIRPα activating a downstream phosphatase (Veillette et al., 1998) that turns off actomyosin (Fig. 2A, images) and thereby turns off efficient engulfment (Tsai and Discher, 2008). Basal phosphorylation of SIRPα in the absence of phagocytic targets remains unexplained (Oldenborg et al., 2000; Ide et al., 2007; Tsai and Discher, 2008), but we observe lower levels of phospho-SIRPα in CD47 KD THP-1-macrophages than in WT, with a linear increase from zero as a function of CD47 levels (Fig. 2B). Note that the deepest KD essentially inverts the CD47:SIRPα stoichiometry from ∼3:1 excess CD47 to >2:1 excess SIRPα (Fig. 1B; Fig. S3B). Decreased basal signaling with CD47 depletion and the increased phagocytosis led us to hypothesize that CD47 on a WT macrophage binds in cis on the same macrophage membrane to SIRPα.

To quantify cis binding in competition with trans binding, binding curves were generated for soluble fluorescent CD47 added to WT and KD THP-1 macrophages. Flow cytometry reveals a saturable, moderate affinity (1.6 μM) interaction for SIRPα on WT cells (Fig. 2Ci). However, a substantial knockdown of CD47 (to ∼13%) strongly increases this apparent binding affinity to 0.26 μM (Fig. 2Ci), and intermediate knockdown shows that the apparent dissociation constants (K_d) increase linearly with CD47 levels on the macrophage (Fig. 2Cii). Our previous measures of soluble SIRPα binding to CD47 on hRBCs, which lack any cis interactions (Fig. 1Ai), gave K_d=0.2 μM (Sosale et al., 2015); such a value matches the intercept of the THP-1 results. The same studies of hRBCs also showed anti-CD47 (B6H12 clone) binds as expected, with much higher affinity (40 nM).

Numerous cell types express both CD47 and SIRPα, including human A549 lung cancer cells (Fig. 1B) and mouse B16 melanoma cells (Figs S1B and S4A). We hypothesized that with such cells cis CD47–SIRPα interactions regulate their phagocytosis by modulating trans presentation of ‘self’ (Fig. 3A), as is relevant to cancer therapy (Fig. 1A). B16 cells were studied to generalize species effects and because they are widely used in preclinical immunotherapy, including CD47–SIRPα blockade (Ingram et al., 2017; Chowdhury et al., 2019; Mandal et al., 2019).

CRISPR/Cas9-mediated knockout (KO) of CD47 or SIRPα was performed in B16 cells, alongside a line with a non-targeting guide RNA for control, and all B16 cells were equally opsonized with anti-Tyrp1 (Figs S1B and S4A) before being added to mouse bone-marrow derived macrophages (BMDMφs). Flow cytometry and imaging both show that engulfment of SIRPα-KO cells was reduced by 30–40% relative to control B16s, compared to the 2–3-fold increase for CD47-KO cells (Fig. 3B–D; Fig. S4B); also 1–2 cells were again engulfed per BMDMφ (Fig. 3C,D). These results support the general hypothesis that CD47–SIRPα cis interactions modulate trans interactions.

Heterologous display of human CD47–GFP on Chinese hamster ovary (CHO) cells varies broadly, but the variation includes physiological levels of CD47 (Subramanian et al., 2007). Co-expression (or not) of human SIRPα (unlabeled but expressed similarly) with CD47–GFP, allowed us to assess trans binding of anti-CD47 or soluble SIRPα. Co-expressing cells show a ∼90% reduction of the red fluorescence from anti-CD47 binding compared to cells expressing CD47–GFP alone (Fig. 4A,B). Such a reduction is consistent with cis-trans competition. For red-fluorescent soluble SIRPα, co-expressing cells showed almost no trans-SIRPα binding relative to cells expressing CD47–GFP alone (Fig. 4B). The larger effect on trans-SIRPα binding relative to anti-CD47 is consistent with weaker association of SIRPα with CD47.

Among its three Ig domains, the N-terminal domain of SIRPα binds the single Ig domain of CD47 in a crystallizable interaction (Hatherley et al., 2008). The Ig domains of SIRPα move as near-rigid bodies linked by hinges in computations of normal mode motions (Fig. 4C). Insertion of SIRPα with its transmembrane domain into a simulated lipid bilayer while in crystallographic association with CD47 in the same membrane (Fig. 4D) provides a molecularly detailed view of SIRPα bending over and binding to CD47 in cis. Bending energy could destabilize binding energy, and a first step toward determining this might be to relate SIRPα-docking probabilities for CD47-derived ‘self’ peptides to measured binding affinities (Rodriguez et al., 2013).

Safety of CD47 blockade is a concern across multiple clinical trials, with loss of various blood cells (RBCs, platelets, etc.; see Andrechak et al., 2019). Clearance by macrophages has been claimed to result from opsonizing signals on aged cells (Advani et al., 2018), such as opsonizing IgG (against oxidation-generated epitopes) or membrane rigidity (Sosale et al., 2015). Paradoxically, fresh hRBCs are minimally eaten by macrophages and are not engulfed more when treated with anti-CD47 (Fig. 1C,D). This paradox is resolved by disruption of the inhibitory CD47–SIRPα cis interaction on the macrophage. Engulfment clearly increases, even though blocking CD47 on only the macrophage should release SIRPα molecules to conceivably interact in trans and increase inhibition. More study is clearly needed, but this is the first study to illustrate the importance of CD47 on the macrophage, with clear implications for clinical blockade of CD47.

The presence of cis interacting CD47–SIRPα is surmised from our four experimental approaches, including anti-CD47 on human-derived THP-1 macrophages (Fig. 1C,D; Fig. S2). These approaches are shRNA-mediated knockdown of CD47 on THP-1 macrophages (Fig. 2; Fig. S3), phagocytosis by primary mouse macrophages of mouse melanoma B16 cells with SIRPα or CD47 deleted (Fig. 3; Fig. S4), and heterologous co-display of SIRPα and CD47 (Fig. 4A,B). Simulations further indicate a cis interaction is physically reasonable. The potency of the phospho-signaling of SIRPα in cis versus that seen in trans remains unclear, but phospho-signaling in trans saturates with increased CD47 on a target (Tsai and Discher, 2008), which differs from the linearity seen upon knockdown (Fig. 2B) and perhaps reflects a limit to SIRPα accumulation at the phagocytic synapse.

The efficacy of anti-CD47 in human cancer trials depends on macrophage activation, such as with tumor-opsonizing antibodies (Fig. 1Ai) (Advani et al., 2018; Andrechak et al., 2019). Expression of anti-CD47 from bacteria within a mouse tumor might activate macrophages via bacterial pathways (Chowdhury et al., 2019), but the same anti-CD47 is also likely to disrupt CD47–SIRPα cis interactions on tumor-associated macrophages and thereby drive hyper-phagocytosis. SIRPα blockade should lead to similarly high activation and hyper-phagocytosis, which is interesting to consider in light of anti-tumor efficacy with systemically injected anti-SIRPα blocked macrophages (Alvey et al., 2017).

Dulbecco's phosphate-buffered saline (DPBS) without Ca²⁺ or Mg²⁺ (Invitrogen, Carlsbad, CA) was supplemented with either 1% bovine serum albumin (BSA) or 1% BSA and 0.05% Tween 20 (Millipore Sigma, Darmstadt, Germany). Tris-buffered saline (TBS) and TBS with 0.5% Tween 20 (TTBS) were used in western blotting. Hoechst 33342 (Invitrogen) was used for DNA stains.

The anti-human CD47 antibody clones B6H12 (catalog no. 556044, BD Biosciences, San Jose, CA), 2D3 (catalog no. 14-0478-82, Thermo Fisher Scientific, Waltham, MA), and 6H9 (a gift from Marilyn Telen at Duke University, Durham, NC) were used for detection of human CD47 on macrophages or CHO cells expressing full-length human CD47. Quantification of CD47 and SIRPα was performed using B6H12–FITC (catalog no. 556045, BD Biosciences, San Jose, CA) and anti-SIRPα clone SE7C2 (catalog no. sc-23863, Santa Cruz Biotechnology, Dallas, TX), respectively. Human SIRPαex [produced as a secreted and purified SIRPα1-GST as per Subramanian et al. (2007), see below] was used for experiments comparing binding affinities between species. The following antibodies were used as IgG opsonin in our phagocytosis assays: polyclonal rabbit anti-human RBC (referred to as anti-hRBC in the text) (catalog no. 109-4139, Rockland Immunochemicals), polyclonal rabbit anti-sheep RBC (catalog no. 113-4139, Rockland Immunochemicals, Limerick, PA) to target sheep red blood cells (anti-shRBC), rabbit anti-streptavidin (catalog no. S6390-1ML, Sigma-Aldrich), and rabbit anti-streptavidin conjugated to FITC (catalog no. 200-402-0955, Rockland Immunochemicals) to target streptavidin coated polystyrene beads (catalog no. SVP-50-5, Spherotech, Lake Forest, IL). Anti-Cas9 (7A9-3A3 clone) antibody (catalog no. ab191468, Abcam, Cambridge, MA) was used to detect Cas9 expression in transduced cell lines in western blot analysis. Secondary antibodies used for detecting opsonin levels and uningested beads included goat anti-rabbit FITC or goat anti-rabbit F(ab′)2 R-PE (Millipore Sigma). Secondary antibodies used for detecting SIRPαex binding included anti-GST Alexa Fluor 488 (Invitrogen). Secondaries antibodies used for confirming complete RBC engulfment included donkey anti-rabbit Alexa Fluor 647 (Invitrogen). Cytokine granulocyte colony stimulating factor (G-CSF) was purchased from R&D Systems (Minneapolis, MN).

COS-1, CHO-K1, A549, THP-1 and B16F10 cells (American Type Culture Collection, Manassas, VA) were maintained in culture with DMEM, MEMα, F-12, RPMI 1640 and RPMI 1640 (Invitrogen), respectively. All cell culture media, unless otherwise specified, was supplemented with 10% (v/v) fetal bovine serum (FBS; F2442, Sigma), 100 U/ml penicillin and 100 μg/ml streptomycin (Gibco, Gaithersburg, MD). Human mesenchymal stem cells (MSCs; Osiris Therapeutics, Columbia, MD) were maintained in low-glucose DMEM, supplemented with 20% (v/v) FBS with the previously mentioned penicillin-streptomycin concentrations. All cell lines were regularly tested for mycoplasma contamination. Cells were detached using 0.05% trypsin and 0.5 mM EDTA (Invitrogen) for passaging. Human blood was obtained from finger pricks of healthy donors. Blood from other species was obtained from Covance (Princeton, NJ) and washed three times in 0.4% BSA. Human hematopoietic stem cells (HSCs) were obtained from fresh purified bone marrow (BM)-derived human CD34⁺ cells from Lonza (Basel, Switzerland). All HSC experiments were performed in hematopoietic stem cell expansion medium (StemLine-II; Millipore Sigma) and supplemented with 1× antibiotics and G-CSF (100 ng/ml) for 7 days for cultured cells. Peripheral blood monocytes from human donors were obtained through the Human Immunology Core (University of Pennsylvania, Philadelphia, PA).

Differentiation of THP-1 cells followed a protocol based on several recent publications that all refer to the differentiated cells as THP-1 macrophages (Genin et al., 2015; Starr et al., 2018; Sedlyarov et al., 2018). Differentiation was achieved in 100 ng/ml phorbol myristate acetate (PMA) (Millipore Sigma) for 2–3 days, which leads to these cells switching from growth in suspension (as monocytic cells) to attaching to tissue-culture plastic. Gene expression data shown in Fig. S1B confirms that PMA-differentiated THP-1 macrophages and primary macrophages have similar expression profiles for key macrophage markers, while differing greatly from two epithelial cell types.

Human CD47 (hCD47; isoform 2) was PCR amplified, digested with XhoI and BamHI (New England Biolabs, Ipswich, MA) and ligated to a similarly digested vector pEGFP-N3 (Takara Bio USA, Inc., Mountain View, CA). CHO cells were plated at 1×10⁵ cells/cm² at 1 day prior to transfection. On the day of transfection, medium was replaced with 2 ml Opti-MEM I (Invitrogen) per 25 cm² surface area and 10–15 μl Lipofectamine 2000 and 5 μg plasmid DNA were diluted in 0.25 ml Opti-MEM I separately and, 5 min later, mixed and incubated for a minimum of 20 min at 25°C. Lipid–DNA complexes in a total volume of 0.5 ml Opti-MEM I were added to the flasks and incubated for 4–6 h. The flasks were then replenished with fresh growth medium. Transfected cells were harvested using DPBS supplemented with 2 mM EDTA (Invitrogen) at 1–2 days post transfection for analysis. Full-length human SIRPα was expressed in CHO cells in a similar manner to human CD47–GFP.

COS-1 cells were transfected with pcDNA3-based vector encoding a human SIRPα extracellular domain fused to GST using Lipofectamine 2000 (Invitrogen). Secreted SIRPα1–GST (referred as hSIRPαex) was affinity-purified using glutathione–Sepharose 4B (Amersham Biosciences, Little Chalfont, UK) and dialyzed against DPBS (Invitrogen). The protein was stored at −20°C with or without addition of 10% (v/v) glycerol (Thermo Fisher Scientific).

Plasmid encoding the extracellular domain of human CD47 or mouse CD47 were PCR amplified, digested with XbaI and SalI (New England Biolabs) and ligated to a similarly digested vector, pEF-BOS-XB, which results in an in-frame fusion of CD4d3+4-biotin at the C-terminus of the extracellular domain of CD47. The above vector containing the extracellular domain of CD47 was transfected into CHO-K1 cells using Lipofectamine 2000 (Invitrogen). Secreted CD47-CD4d3+4 was concentrated using a 10 kDa molecular mass cut-off (MWCO) Amicon filter (Millipore Sigma) and biotinylated at the C-terminus using a biotin-protein ligase (Avidity, LLC) and dialyzed against DPBS (Invitrogen). The protein was affinity-purified using a monomeric Avidin (Promega) and dialyzed against DPBS (Invitrogen).

Human RBCs, THP-1 cells, A549 cells, peripheral blood monocytes and human MSCs were labeled with 15 µl of B6H12-FITC at saturating levels (BD Biosciences) against human CD47 for 30 min at room temperature. Cells were washed and resuspended in DPBS supplemented with 5% FBS for flow cytometry analysis. For measurement of human SIRPα expression on CHO-K1 cells anti-SIRPα (clone P3C4) (catalog no. LS-C179629, MBL International, Woburn, MA) was used. Anti-human CD47 measurements and detections were performed using different antibodies above recognizing different epitopes of CD47. Saturating concentration of anti-SIRPα or anti-human CD47 antibodies were used as similarly described for the above cells.

shRNA lentiviral supernatants to CD47 were purchased from Millipore Sigma (TRC#: TRCN0000007836, TRCN0000007837) to target CD47 and resulted in 52 and 87% knockdown, respectively. Further details of these clones are available from the Millipore Sigma website. Target THP-1 cells were infected with lentiviral supernatants at a multiplicity of infection (MOI) of 10 in the presence of 80 μg/ml polybrene (hexadimethrine bromide) (Millipore Sigma) and remained in culture for 24 h at 37°C and replaced with fresh RPMI 1640 supplemented with 10% FBS. Cells with integrated viral sequence were selected using puromycin (Millipore Sigma) at 2 μg/ml and then passaged with continuous puromycin selection. The degree of CD47 silencing was regularly monitored by flow cytometry and western blotting with 1:200 anti-SIRPα (clone SE7C2, Santa Cruz Biotechnology). Control cell cultures were generated with control lentivirus in parallel.

LentiV-cas9_puro and Lenti_sgRNA_EFS_GFP plasmids (Addgene #108100 and 65656, respectively) were gifts from Christopher Vakoc. The single guide RNA (sgRNA) oligonucleotides (CD47, 5′-TCCCCGTAGAGATTACAATG-3′; SIRPα, 5′-TAATTCTAAGGTCATCTGCG-3′) were designed using the Broad CRISPR algorithm. sgRNAs were cloned into the sgRNA vector using a BsmBI restriction digest.

HEK293T cells were used for lentiviral delivery of Cas9 and sgRNA and cultured in high-glucose DMEM supplemented with 10% (v/v) FBS (F2442, Sigma), 100 U/ml penicillin and 100 μg/ml streptomycin (Gibco). To generate lentivirus, transfer plasmid, pVSVg and psPAX2 were co-transfected into HEK293T cells at a 2:2:1 ratio using Mirus TransIT (MIR6603) transfection reagent (Mirus, Madison, WI) following the manufacturer's protocol. Viral supernatants were collected 48 h after transfection and added to target cells at a 1:1 ratio of culture medium to viral supernatural. Successfully transduced cells were selected using puromycin at 2 μg/ml. Cas9 expression was confirmed using western blot analysis. Clonal knockout cells lines were developed and confirmed using both flow cytometry and western blot analysis using the previously described antibodies.

The binding isotherm of soluble human CD47 was performed for THP-1 wild-type and CD47 knockdowns, as noted, over a range of concentration using flow cytometry. Forward scatter, side scatter and fluorescence (FL1, FL2, FL3, FL4 channels in logarithmic mode) were acquired for at least 10⁴ events using a FACScan, FACSCalibur or LSR II (BD Biosciences, San Jose, CA). Data points from flow cytometry were plotted and analyzed using a hyperbolic fit of the form y=A x/(K+x) to obtain the K_d values shown.

The rates of association (k_on) of SIRPα with soluble CD47 were measured at room temperature. Mixtures of 8, 4 and 2 µM soluble CD47 with 15 µl of B6H12–FITC against human CD47 at saturating levels (BD Biosciences) were pre-equilibrated for 3 h to produce the primary complex. After mixing, aliquots were periodically withdrawn and added to the human THP-1 cells, then washed and diluted in PBS (1:30) to measure the CD47 and SIRPα binding using flow cytometry. Forward scatter, side scatter and fluorescence (FL1, FL2, FL3, FL4 channels in logarithmic mode) were acquired for at least 10⁴ events using a FACScan, FACSCalibur or LSR II (BD Biosciences). Data points from flow cytometry were plotted and fitted to obtain the k_on values shown.

For the phagocytosis assays, macrophages were plated in 4 cm² Lab-Tek II Chambered Coverglass (Nalge Nunc International, Rochester, NY) or 12-well tissue culture plates (Corning, Corning, NY) at 1×10⁵ cells per 4 cm². Streptavidin polystyrene beads or RBCs were added to macrophages at a ratio of 20:1 and allowed to incubate at 37°C for 60 min. Non-phagocytosed beads or RBCs were washed with DPBS. For assays using RBCs as targets, lysis of non-ingested RBCs was performed by adding deionized H₂O for 30 s, followed by immediate replacement with 0.4% BSA and fixing with 5% formaldehyde for 5 min.

For stimulated phagocytosis assays, beads (with or without CD47) were incubated with rabbit anti-streptavidin serum, sheep RBCs with rabbit anti-sheep RBC antibody, and human RBCs with rabbit anti-human RBC antibody, respectively, as the opsonin. Beads or RBCs were opsonized at their respectively titrated concentrations for 45 min at room temperature. Opsonized beads and RBCs were washed twice and resuspended in 50 μl of PBS containing 0.4% BSA. If the opsonized target required an additional CD47 block, they were treated with anti-CD47 (B6H12 clone) diluted in DPBS to a final concentration of 0.5 μg/ml for 45 min at room temperature. For phagocytosis assays in which the interactions in cis on THP-1 cells was disrupted, we also added 1.5 µM anti-CD47 clone B6H12 antibody to the THP-1 cells prior to the addition of RBCs. The antibody was allowed to bind for 1 h prior to the assay and then washed out with DPBS. Finally, if RBCs were used, they were stained with PKH26 membrane dye (Sigma-Aldrich) for 45 min at room temperature prior to their addition to THP-1 cultures.

To identify non-internalized beads, beads were labeled with a primary antibody, rabbit anti-streptavidin (catalog no. S6390-1ML, Millipore Sigma) at 1:1000 in DPBS for 20 min at 25°C. A second antibody, anti-rabbit IgG conjugated to R-phycoerythrin (R-PE) (catalog no. 52412-1MG-F, Millipore Sigma), was added at 1:1000 in PBS to the cells and incubated for an additional 20 min at 25°C. Cells were then washed with 0.4% BSA and then quantified by light and fluorescence microscopy. For differentiation of non-internalized RBCs, fixed THP-1 cells were blocked with 5% BSA for 30 min followed by addition of donkey anti-rabbit IgG (to target rabbit anti-human RBC) conjugated to Alexa Fluor 647 at 1:500 dilution in DPBS for 45 min at 25°C. Secondary antibody would not be able to bind to completely engulfed RBCs but would stain RBCs that simply adhered to the THP-1 surface. For quantification purposes, we define true phagocytic events as those in which RBCs are completely internalized and do not stain for secondary antibody. At least 200 macrophages were scored per well and experiments were repeated at least three times.

Bone marrow cells were isolated from femurs and tibias of healthy male C57BL6/J mice (The Jackson Laboratory 000664) and cultured in 10-cm petri dishes containing Iscove's modified Dulbecco's medium (IMDM, Gibco 12440-053) supplemented with 10% (v/v) FBS, 100 U/ml penicillin, 100 μg/ml streptomycin (Gibco), and 20 ng/ml M-CSF (Biolegend 576404) for 7 days at 37°C, 5% CO₂. All animal experiments were performed according to protocols approved by the University of Pennsylvania's IACUC (protocol #805977 and #804455). The resulting BMDMφs were detached with 0.05% trypsin-EDTA (TRED, Gibco) and re-plated in 6- or 24-well tissue culture plates at a density of 2.1×10⁴ cells/cm².

The following day, macrophages were stained with 0.5 μM CellTracker Deep Red (C34565, Thermo Fisher Scientific) for 10 min at 37°C. The labeled BMDMφs were washed twice with serum-free IMDM. B16 cell lines were labeled with 1 μM CFDA-SE (V12883, Thermo Fisher Scientific) and detached with 0.05% TRED, counted and resuspended at 2×10⁵ per ml in serum-free IMDM. Labeled B16 cells were opsonized in suspension with 5 μg/ml anti-Tyrp1 (clone TA99, Bio X Cell BE0151, Lot #715419A1) or mIgG2a isotype control (clone C1.18.4, Bio X Cell BE0085, Lot#692418S1) for 30 min on ice.

For analysis of phagocytosis by flow cytometry, 2 ml (∼4×10⁵) of the CFDA-labeled, opsonized B16 suspension was added to each well of a six-well plate containing Deep Red-labeled BMDMφs and incubated at 37°C, 5% CO₂ for 2 h. All cells were detached with 0.25% TRED and resuspended in FACS buffer [PBS plus 1% (w/v) BSA and 0.1% sodium azide] and analyzed on an LSRII flow cytometer (BD Biosciences). For analysis of phagocytosis by fluorescence microscopy, 0.5 ml containing opsonized B16 cells was added to each well of a 24-well plate containing Deep Red-labeled BMDMφs and incubated at 37°C, 5% CO₂ for 2 h. For the final 15 min of the incubation period, cells were stained with 1 μg/ml Hoechst 33342 (Thermo Fisher Scientific) before fixation with 4% formaldehyde. Epifluorescence images were acquired on an EVOS FL auto imaging system with either a 20× objective or on an Olympus IX inverted microscope with a 40× objective. For imaging, we quantified the phagocytic index, which is defined as the product of the percentage of the total BMDMφs population phagocytosing and the number of target cells engulfed per BMDMφ.

Wild-type THP-1 phagocytes and CD47 knockdowns (2×10⁶) were cultured and differentiated in six-well plates for 48 h after PMA differentiation. Human CD47 was attached to 2.1 μm diameter beads at specific densities as described above and added at a bead-to-cell ratio of 20:1 for 10 min. Following the incubation time, the cells were washed with ice-cold PBS and then lysed on ice in 300 μl of lysis buffer [50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA, 1% NP-40, 1% protease inhibitor cocktail (Millipore Sigma) and 2 mM activated sodium orthovanadate]. For immunoprecipitation, whole-cell lysate was mixed with 1:200 anti-SIRPα (SE7C2 clone) antibody (Santa Cruz Biotechnology, Inc.) with Pierce Protein G agarose (Thermo Fisher Scientific) at 4°C overnight. Precipitated protein was placed in 4–12% SDS-PAGE gels in MOPS buffer (Invitrogen), transferred onto polyvinylidene difluoride (PVDF) membrane, blocked and labeled via the anti-phospho-tyrosine IgG conjugated to horseradish peroxidase (HRP)- (1:1000, catalog no. ab16389, Abcam) and anti-SIRPα (C-20 clone) (1:500, catalog no. sc-6922, Santa Cruz Biotechnology) primary antibodies and anti-goat-HRP (Amersham Biosciences). All western blots were run in duplicate, along with an additional blot for actin to ensure constant protein load among samples.

A mix of 5 µl of soluble SIRPα (final concentration ∼1 µM), 5 µl of 2 mg/ml Alexa Fluor 647-conjugated rabbit anti-GST, 40 µl DPBS, 1% BSA and 2.5×10⁶ CHO cells was prepared, and incubated at room temperature for at least 30 min. Cells were pelleted (500 g for 5 min) and resuspended in 1 ml cold DPBS and analyzed immediately. For antibody labeling, saturating levels of anti-CD47 antibody in 50 µl of DPBS, 1% BSA and 2.5×10⁶ CHO cells were mixed together and incubated as above. Cells were washed in 0.5 ml DPBS with 1% BSA and then resuspended in 50 µl of DPBS with 1% BSA containing 5 µl of secondary antibody (2 mg/ml). After incubation for 30 min at room temperature, cells were washed once in 0.5 ml DPBS and resuspended in 1 ml of DPBS with 1% BSA and imaged immediately.

Images were acquired with an inverted microscope (Olympus; IX71) with a 60× (oil, 1.4 NA) objective using a Cascade CCD camera (Photometrics, Tuscon, AZ). Image acquisition was performed with Image-Pro Plus software (Media Cybernetics, Silver Spring, MD). All subsequent image analysis was done using ImageJ.

Data from the GEO database was used to obtain gene expression data for key genes associated with macrophage identity (Fig. S1B). The cell types included in this analysis were human HEK 293T (GEO accession GSE28715), human PMA-differentiated THP-1 macrophages (GEO accession GDS4258), primary mouse macrophages (GEO accession GDS2454) and mouse B16F10 melanoma (GEO accession GSE33607).

The cylindrical shape of the protein in the crystal structure yields normal modes that are very close in energy, but three different regions emerge in the normal mode analysis matrix and unsurprisingly correspond to the three Ig domains in the crystal structure. The regions along the diagonal of the matrix with values close to zero (i.e. regions with reduced mobilities) are expected to behave as hinges (Flores et al., 2008). In this way, we can determine that the regions delimited by residues 113–117 and 220–222 act as hinges. Since we were also interested in having a spatial description of the normal modes, we needed to make use of the Anisotropic Network Model (ANM) (Hinsen et al., 1999). Despite being less realistic than GNM, we used ANM since it is better suited for assessing the directions of motions (Cui and Bahar, 2007).

All statistical analyses were performed using GraphPad Prism 7. Unless otherwise noted, all statistical comparisons were made by unpaired two-tailed Student's t-test and were considered significant if P<0.05. Unless otherwise stated, all plots show mean±s.e.m. or mean±s.d. Figure legends specify the exact meaning of ‘n’ for each figure.

We thank Jason Andrechak (Penn) for helpful discussions in manuscript preparation. We thank the University of Pennsylvania Flow Cytometry & Cell Sorting Facility and Cell & Developmental Microscopy Core for providing instrumentation and technical assistance in the experiments performed in this study. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Institutes of Health, National Science Foundation, and other funding agencies.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.237800.reviewer-comments.pdf