Background
The clinical development of immune checkpoint inhibitors (ICIs) has dramatically changed the landscape of cancer treatment [
1‐
3]. ICIs targeting T-cell inhibitory receptors can induce complete and durable tumor immunity in patients with metastatic and treatment-refractory cancers. Despite the clinical promise of ICIs, only a fraction of patients within the ICI responsive cancer subtypes benefit from treatment with antibodies against PD-1, PD-L1, and CTLA4. Cancer is highly heterogeneous and complex and exploits a variety of mechanisms to evade immune surveillance beyond suppression of antitumor T cell responses [
4]. Tumor-associated macrophages constitute a large fraction of the immune cell infiltrates within the tumor microenvironment of many human cancers [
5]. Dendritic cells (DCs), although low in frequency within tumors, are potent and crucial mediators of antitumor immunity [
6]. Given their prevalence and immunomodulatory activities, targeting regulators of macrophage and DC function is an attractive strategy to augment antitumor immunity and achieve additive or synergistic efficacy in combination with antitumor antibodies or ICIs.
Signal regulatory protein α (SIRPα) is an immunoinhibitory receptor expressed primarily by cells of the myeloid lineage including monocytes, macrophages, DCs, and neutrophils [
7]. Upon interaction with its principal ligand, CD47, SIRPα transmits inhibitory signals that regulate DC homeostasis, self-recognition, and macrophage-mediated programmed cell removal [
7,
8]. During malignant transformation, many human tumors exploit upregulation of CD47 thus activation of SIRPα signaling to avoid phagocytic clearance, resulting in the suppression of myeloid-mediated innate immunity and poor induction of antigen-specific immunity. SIRPα negatively regulates DC maturation, antigen presentation [
9], and proinflammatory cytokine secretion [
10]. In addition, it counteracts activating signals mediated by antibody engagement of Fcγ receptors (FcγR), which profoundly limits antibody-dependent cellular phagocytosis (ADCP) against tumors, restricts neutrophil transmigration [
11], and maintains myeloid-derived suppressor cell (MDSC) functions [
12]. Given the broad negative regulatory roles of SIRPα on innate immunity, a variety of CD47–SIRPα antagonists have been developed to promote the antitumor activity of phagocytes and myeloid cells. Blockade of the CD47–SIRPα interaction synergizes with both tumor-specific antibodies and ICIs by effectively reprogramming the myeloid compartment toward a proinflammatory phenotype improving tumor cell phagocytosis, antigen presentation, and T cell priming [
9,
13]. A number of CD47 antagonists have entered the clinic with promising anticancer activity in both hematological and solid tumors [
14‐
18].
Efforts to disrupt the CD47–SIRPα interaction have mainly focused on targeting CD47 due to its upregulation and ubiquitous expression on most human tumor types. However, CD47 is also broadly expressed on virtually all normal cells, including red blood cells and platelets, which creates a large antigen sink and CD47 blockers comprising an active Fc have shown dose-dependent cytopenia [
19,
20]. We have previously shown that safety liabilities associated with CD47 blockers with active Fc domains can be overcome by eliminating Fc effector function [
13,
16,
17,
21]. Targeting SIRPα is an orthogonal approach to inhibit the CD47–SIRPα pathway.
Herein, we explore pre-clinical pharmacology, pharmacokinetics, and exploratory safety associated with antibody-based blocking and antagonism of SIRPα. Utilizing an anti-SIRPα antibody we discovered in human antibody transgenic chickens [
22], we demonstrate that pan-allelic anti-SIRPα antibody hAB21, which blocks the CD47–SIRPα interaction, broadly recapitulates the functional properties of CD47 antagonists by promoting the anticancer activity of both tumor-specific antibodies and ICIs in a macrophage and DC dependent manner. This leads to the induction of adaptive immunity, resulting in complete and durable antitumor responses in mice. HAB21 exhibits favorable safety and pharmacokinetic (PK) profiles in monkeys with no hematological or immune-related adverse events. Thus, targeting SIRPα with hAB21 is a promising therapeutic approach to treat human cancers by reprogramming innate immunity to sensitize cancers to immunotherapy.
Methods
Expression and purification
All antibodies were expressed in Expi293 cells (Invitrogen) using standard manufacturer’s protocol. Expression cultures were typically grown for 5 days at 37 °C in 8% CO
2. Supernatants were harvested via centrifugation and sterile filtered. Proteins were affinity purified utilizing MabSelect Sure LX resin (GE Healthcare). For SPR screening, the IgV domains of human SIRPα v1 (NP_542970.1) and human SIRPα v2 (CAA71403.1) are expressed in Expi293 cells as described above as either a His-tagged or His-Avi-tagged fusions and purified using Ni-Sepharose 6 Fast Flow affinity purification and polished via gel filtration through a superdex hi-prep resin (GE Healthcare). The anti-SIRPα v1 specific antibody clone HEF-LB was generated as described in reference [
23]. For crystallography, the IgV domain of SIRPα v1 and Fab fragments was generated as described [
14].
Humanization of antibodies
Parental clones of hAB21 (AB21 and AB25) were isolated from SynVH, a transgenic chicken, and have fully human heavy chains but chicken light chains [
14]. In order to humanize the chicken-derived light chains, chicken hypervariable region (HVRs) of AB25 were grafted onto human lambda light chain IGLV1, IGLV2 and IGLV3 frameworks, and chicken HVRs of hAB21 were only grafted onto human lambda light chain IGLV1 framework [
24]. Combination of antibodies was generated from 4 humanized light chains with two heavy chains (derived from AB21 and AB25 respectively) in Expi293, purified and tested for SIRP binding. The top humanized antibody, with favorable expression and binding properties, was designated hAB21 and selected for further testing.
Determination of KD
The binding affinities of AB21 and hAB21 to human SIPRα v1 and v2 were determined using direct immobilization of the antibodies (via GLC chip). All antibodies and proteins were used at their nominal concentrations determined by A280 absorbance and molar extinction coefficient. Analytes (human SIRPα v1 and v2) were injected in a “one-shot” kinetic mode and flowed over AB21 and hAB21 immobilized (~ 1000 RUs) on GLC chips using ProteOn Amine Coupling Kit. For the immobilization step, GLC chip was activated with EDAC/Sulpho-NHS 1:1 (Biorad) diluted 1/100 for 300 s at 25 μL/min. AB21 and hAB21 were diluted to 80 nM concentration in 10 mM sodium acetate buffer pH 4.5 and immobilized to the chip at 30 μL/min for 50 s. Chip was inactivated with ethanolamine for 300 s at 25 μL/min. The analytes (human SIRPα v1 and v2 protein) were injected in a “one-shot” kinetic mode at nominal concentrations of 10, 3.3, 1.1, 0.37, 0.12 and 0 nM. Association times were monitored for 90 s at 100 μL/min, and dissociation times were monitored for 1200 s. The surfaces were regenerated with a 2:1 v/v blend of Pierce IgG elution buffer/4 M NaCl. Biosensor data were double-referenced by subtracting the interspot data (containing no immobilized protein) from the reaction spot data (immobilized protein), and then subtracting the response of a buffer “blank” analyte injection from that of an analyte injection. Double-referenced data were fit globally to a simple Langmuir model and the KD value was calculated from the ratio of the apparent kinetic rate constants (KD = kd/ka).
Crystallization of anti-SIRPα Fab: SIRPα complexes
A pure sample of anti-SIRPα Fab-SIRPα V1 complex at a concentration of 11.3 mg/mL in a buffer of 10 mM Tris pH 7.4, 50 mM NaCl was set for sitting drop vapor diffusion with sparse matrix crystallization screen kits available Qiagen. The condition and crystal form that gave quality diffraction leading to a complete dataset was 1. 0.1 M Sodium Acetate pH 4.0, 0.2 M Ammonium Sulfate, 18% (w/v) PEG 4000 (Cryo-protectant: 5% v/v Ethylene Glycol). Even though the crystallization condition contained a high percentage of PEG 4000, ethylene glycol was added as cryo-protectant to protect the crystal form from deteriorating and/or forming ice during freezing. These samples were screened for protein X-ray diffraction at the NSLS-II 17-ID AMX beamline; yielding a dataset with diffraction of 2.27 Angstroms resolution. Further crystallography data collected are detailed in Additional file
1: Table S1. Structure PDB ID: 7KPG.
Cell binding and blockade of CD47 binding to SIRPα cells
For detection of cell binding, AB21 was fluorescently labeled with the Alexa Fluor 647 Protein Labeling Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. 250,000 cells per well in staining buffer (PBS, 0.5% BSA or 2% FBS) were plated in 96-well plates (Falcon). Cells were first stained with fixable Live/Dead Stain (Invitrogen) and washed once in staining buffer prior to all binding assays.
To detect SIRPα binding to cells, 500 nM Alexa Fluor 647-labeled AB21 was titrated 1:4 for seven dilutions and added to cells in 100 μL volume of FACS buffer (PBS + 0.5% BSA) supplemented with a cocktail of human Fc block (Miltenyi Biotec) or mouse Fc block (Biolegend), anti-CD14 (Biolegend) for human and cynomolgus PBMCs or anti-CD11b (Biolegend) for mouse splenocytes. After a 60-min incubation on ice, cells were washed twice in staining buffer and fixed in 0.5% paraformaldehyde.
To block CD47 binding to SIRPα on PBMCs, Alexa Fluor 647 labeled CD47Fc at a concentration of 500 nM and 1:4 titration starting at 1 µM of AB21 were added to cells. After a 60-min incubation on ice, cells were washed twice in staining buffer and fixed in 0.5% formaldehyde.
Cells were analyzed on a FACS Canto II (BD Biosciences), with subsequent data analysis using Flowjo 10.7.
Cell lines
4T1, CT26, Raji, MDA-MB-231 and DLD-1 cells were obtained from the American Type Culture Collection (ATCC). MC38 cells were obtained from MuriGenics. All cell lines were cultured according to standard protocols.
Animals
All mouse experiments except for MDA-MB-231 and batf3 KO mice were conducted according to protocols approved by Institutional Animal Care and Use Committee (IACUC) of ALX Oncology. Tumor model with MDA-MB-231 and batf3 KO mice were conducted in compliance with UTSW Human investigation and UTSW Institutional Animal Care and Use Committee protocols. Female Batf3−/− mice in the C57BL6/J background and NSG-SMG3 mice were purchased from The Jackson Laboratory. Female BALB/c, C57BL/6 and NOD-SCID animals, age 6–8 weeks old, were purchased from Charles River Laboratories International (Hollister, CA). All animals were housed according to institutional IACUC guidelines.
Female naïve cynomolgus monkeys (Macaca fascicularis) were provided by Charles River Laboratories. The in-life portion of the study was conducted by CRL’s testing facility (Reno NV), and the animals were released to the CRL testing facility’s colony at the end of the study.
Generation of monocyte-derived macrophages and phagocytosis assay
Human CD14+ cells were purified from Trima residuals (Vitalant) with Ficoll-Paque Plus and negative selection (Monocyte Isolation Kit II, Miltenyi Biotec) according to the manufacturers’ protocols. Monocyte-derived macrophages (MDM) were made by seeding 10 million CD14+ cells into 150 mm tissue culture dishes (Corning) in growth medium supplemented with 10% human AB serum (Corning) or 10% FBS and 50 ng/mL MCSF. Cells were cultured for 7–11 days. Adherent cells were detached from culture plates with TrypLE Select (Thermo Fisher Scientific). Target cells (DLD-1) were labeled with the Celltrace CFSE Cell Proliferation kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. 100,000 target cells and 50,000 MDMs were incubated in ultra-low attachment U-bottom 96-well plates (Corning) with anti-SIRPα antibodies and the corresponding tumor-specific antibody for 2 h at 37 °C. Cetuximab was added at a concentration of 0.01–0.1 µg/mL.
For flow cytometry, cells were incubated in human FcR blocking reagent (Miltenyi Biotec) and stained with fluorochrome-labeled antibodies against CD33 (clone WM53, Biolegend) and CD206 (clone 15–2, Biolegend). To eliminate macrophage/target cell adhesion from analyses, antibody against CD326 (clone 9C4, Biolegend) was included. Furthermore, a pulse geometry gate of forward scatter signal area vs height was used to select for single cells. Fixable viability dye (Thermo Fisher Scientific) was used to identify live cells. Cells were acquired on a FACS Canto II flow cytometer (BD Biosciences) with subsequent analysis using FlowJo software. Percent phagocytosis indicates the percentage of viable CD33+ CD206+ macrophages that stain negative for CD326 and positive for CFSE. Where applicable, 4 parameter fit curves were generated with Prism 7 software (GraphPad).
In vitro PBMC culture
Peripheral blood mononuclear cells (PBMC) were isolated from Trima residuals of healthy individuals with Ficoll-Paque Plus. 500,000 PBMCs were incubated in U-bottom 96-well plates (Falcon) with anti-SIRP at a concentration of 10 ug/mL for 48 h at 37C.
For quantification of PBMC subsets by flow cytometry, cells were incubated in human FcR blocking reagent and stained with a cocktail of fluorochrome-labeled antibodies against lin—(CD3, CD14, CD16, CD19, CD56) and HLADR. Fixable viability dye was used to identify live cells. After staining, cells were washed and fixed with 0.5% paraformaldehyde in PBS. Prior to acquisition, absolute counting beads (Thermo Fisher) were added and samples were acquired with Canto II flow cytometer and analyzed using FlowJo software.
Tumor studies
Isoflurane anesthesia was used on mice to eliminate or minimize pain and distress during tumor implantation. 2 × 106 MC38, 2 × 106 CT26, 5 × 105 B16F10 and 5 × 105 4T1 cells were resuspended in 100ul of PBS or RPMI and implanted subcutaneously into the flank of female C57BL/6 mice for MC38 and B16F10 and BALB/c for CT26 and 4T1. 5 × 106 Raji cells were resuspended in 100 ul 1:1 PBS:Matrigel (Corning) and implanted subcutaneously in the flank of NOD-SCID mice.
When tumors reached an average of 50-180mm3, as calculated with the formula volume = (width2 × length)/2, mice were randomized into treatment groups. All treatments were dosed intraperitoneally (i.p.) or intratumorally. For xenograft models in NOD-SCID mice, rituximab was dosed at 3 mg/kg and AB21 at 10 mg/kg, five times every 3 days. In syngeneic models, mice were treated with 10 mg/kg for CT26, MC38, B16F10 or 30 mg/kg for 4T1 of clone AB21, 10 mg/kg anti-PD-1 (BioXcell, clone RMP1-14) and 2 mg/kg (MC38 models) anti-PD-L1 (murine IgG1, ALX Oncology). For intratumoral injections, 50 µg of AB21 was given four times every 3 days.
For re-challenge experiments, treated mice with complete tumor eradications were re-challenged with either 2 × 106 MC38 cells on one flank or 0.5 × 106 B16F10 tumor cells on the opposite flank. Age-matched naïve mice were used as control.
For cellular depletion experiments, mice were dosed intraperitoneally with 250 ug of either anti-CSF1R (BioXcell, clone AFS98), anti-CD8b (Bioxcell, clone 53-5.8) or anti-GR1 (BioXcell, clone RB6-8C5) on days 2, 5, 10 and 15 post-tumor implantation. Confirmation of cellular depletion was performed on tumor and spleen of spare cellular depleted mice on days 1, 3, 4 and 8 post-injection by flow cytometry using anti-CD45 (Biolegend, clone 30-F11), anti-CD3 (Biolegend, clone 145-2C11), anti-CD8 (Biolegend, clone 53.6.7), anti-CD11b (ebioscience, clone M1/70), anti-Ly6C (Biolegend, clone 145-2C11), MHCII (ebioscience, clone M5/114.15.2) and F4/80 (Biolegend, clone BM8).
Humanized mice were generated with 4-week NSG-SGM3 female mice and irradiated with 100 cGy (X-ray irradiation with X-RAD 320 irradiator) 1 day prior to CD34+ cells transfer. Irradiated mice were treated with Bactrim (Aurora Pharmaceutical LLC) water for 2 weeks. Cord blood was obtained from UT Southwestern (UTSW) Parkland Hospital. Human CD34+ cells were purified from cord blood by density gradient centrifugation (Ficoll® Paque Plus, GE healthcare) followed by positive immunomagnetic selection with anti-human CD34 microbeads (Stem Cell). 105 CD34+ cells were intravenously injected into recipient mice.
12 weeks after engraftment, humanized mice with over 40% human CD45+ cells reconstitution and age and sex matched non-humanized mice were inoculated with 2 × 106 MDA-MB-231 tumor cells subcutaneously on the right flank. Tumor volumes were measured by length (a), width (b) and height (c) and calculated as tumor volume = abc/2. 8 days later when tumors were around 50 mm3, mice were intratumorally treated with 50 µg AB21, HEF-LB or PBS, four time every 3 days.
BALB/c mice were implanted with 4T1 cells and lungs were harvested 8–9 days post-last injection for metastatic nodule quantification. In brief, lungs were harvested in ice-cold 1xPBS, minced into small pieces then transferred into digestion solution consisting of 2 mg/mL collagenase type V (Worthington) supplemented with 50 ug/mL DNAse (Sigma) and incubated for 2 h in a 37C incubator with end-over-end rotation. Suspension was transferred into 70-um strainer, washed once in 1 × PBS then transferred into 10 mL selection media consisting of RMPI 1640 supplemented with 10% FBS, penicillin–streptomycin and 10 ug/mL 6-thioguanine (Sigma). Three to four 1:10 serial dilutions were plated either in 6-well plates or 10-cm dishes and cultured for 10–14 days at 37C, 5% CO2. Metastatic plaques were then fixed in methanol for 5 min at room temperature, re-hydrated in distilled water then stained with 0.03% methylene blue (Sigma) for 5 min at room temperature. Dye was then discarded, plate was rinsed gently with distilled water and allowed to air-dry prior to counting plaques.
Immunophenotyping
For immune response monitoring in tumor-bearing mice, anti-SIRPα was dosed either three times in combination with anti-PD-1, or twice with anti-SIRPα and once with anti-PD-1, all injections were administered i.p 3 days apart at 10 mg/kg.
Spleens and tumors were harvested either 2 or 3 days post-last injection for immunophenotyping. Spleens were processed into single-cell suspension in ice-cold PBS, lysed with ACK lysis buffer (Gibco), washed twice and re-suspended in PBS supplemented with 2% FBS. Tumor-derived single-cell suspensions were prepared using a cocktail of Collagenase I (Worthington), Collagenase IV (Sigma) and DNAse (Sigma) for 45 min at 37° C. Cell counts were performed using ViCell counter (Beckman Coulter) for spleen and lymph node and trypan blue exclusion with hemocytometer for tumor. Aliquots of 1–2 106 cells were either used for cell-surface antigen staining or stimulation for cytokine assessment. For surface staining, cells were stained with LIVE/DEAD fixable dye (Thermo Fisher), followed by mouse Fc-block (Biolegend) and subsequently stained with antibodies according to cell-type specific antibody panels for at least 30 min at 4 °C. CD4 clone GK1.5, CD8 clone 53-6.7, CD25 clone PC61, CD3 clone 145-2C11, CD45 clone 30-F11, CD47 clone MIAP301, NKp46 clone 29A1.4, PD1 clone J43, FoxP3 clone FJK-16s, Ki67 clone SolA15, Granzyme B clone QA16A02, CD44 clone IM7, CD62L clone MEL-14, TNFa clone MP6-XT22, IFNg clone XMG1.2, SIGLEC H clone 551, CCR7 clone 4B12, CD172 clone P84, CD86 clone GL-1, MHCII clone M5/114.15.2, GR-1 clone RB6-8C5, 33D1 clone 33D1, CD11b clone M1/70.15, CD11c clone N418, CD103 clone 2E7, Ly6C clone NH1.4, F4/80 clone BM8, CD24 clone M1/69. All flow antibodies were purchased from either Biolegend or Thermo Fisher.
For PMA/ionomycin ex-vivo stimulation to quantitate IFNγ+ cells, total splenic and tumor cells were plated at 1 × 106 cells/well in complete RPMI 1640 comprised of 10% heat-inactivated FBS, 2% Pen/Strep, 1% Glutamax, 1% MEAA, 1% sodium pyruvate, 25 mM HEPES and 5 μM ß-mercaptoethanol supplemented with 50 ng/ mL PMA (Fisher Scientific) and 1 μM ionomycin (Sigma) in the presence of Golgi-Stop for at least 4 h at 37 °C, 5% CO2, and subsequently stained with antibodies to surface and intracellular markers. Samples were then acquired using Attune NxT (Thermo Fisher) or Canto II (BD). Analysis was performed using FlowJo 10.0 (BD) and tabulated using GraphPad Prism 7.3.
HAB21 and soluble SIRPα serum ELISA
Immulon 96-well ELISA plates (Thermo Fisher Scientific, 3855) were coated overnight with human wild type SIRPα, variant 1 (ALX Oncology) in PBS. Plates were washed with Tris-Buffered Saline Tween-20 (TBST, 25 mM Tris, 0.15 M NaCl, 0.05% Tween-20, pH 7.5) and blocked for 1 h with assay buffer (PBS, 1% BSA, 0.05% Tween-20, 0.25% CHAPS, 5 mM EDTA, 0.35 M NaCl). Serum samples diluted a minimum of 1:50 in assay buffer or hAB21 standard curve protein (two-fold serial dilutions from 160 to 1.25 ng/mL, in 1:50 normal cyno serum diluted in assay buffer) were added to blocked plates for 1 h. Plates were washed with TBST. Standard curves and samples were incubated for 1 h with biotinylated goat anti-human IgG (H + L) antibody (Bethyl, A80-319B), washed with TBST, incubated for 30 min with HRP-conjugated Avidin D (Vector, A2004), and washed with TBST. All plates were incubated with 1-Step Ultra TMB ELISA solution (Thermo Fisher Scientific, 34028) and the reaction was stopped with 0.16 M sulfuric acid solution (Thermo Fisher Scientific, N600). Plates were read at an O.D. of 450 nm with a background reference reading at 570 nm on a SpectraMax i3 plate reader (Molecular Devices). Protein concentrations of serum samples were interpolated from the hAB21 standard curve with a 4-parameter fit curve using Prism software (GraphPad).
For the detection of soluble SIRPα, similar assay to hAB21 serum ELISA was performed except for the following. 2 ug/mL anti-SIRPα mouse IgG1 antibody (AB136b, ALX Oncology, a non-CD47 blocking antibody that binds both cyno and human SIRPα with high affinity) in PBS was used to coat the plates. Cyno PK serum samples diluted 1:25 in assay buffer or SIRPα standard curve protein (ALX135, ALX Oncology, 25 to 0.024 ng/mL, diluted in assay buffer) were added to blocked plates for 1 h. Pre-dose cynomolgus serum samples spiked with or without 200 ug/mL of anti-SIRPα human IgG1 antibody were run as controls since most of the PK samples contained at least 200 ug/mL anti-SIRPα human IgG1 antibody. After incubating samples for 1 h in the ELISA, plates were washed with TBST. 5 ug/mL anti-SIRPα human IgG1 kick-off Fab + 6 × His tag (AB115f, ALX Oncology, binds both cynomolgus and human SIRPα with high affinity) was added to samples for 1 h. Plates were washed with TBST. Samples were incubated for 1 h with 0.15 ug/mL rabbit anti-6 × His Tag HRP conjugated antibody (abcam cat. # ab1187).
HAB21 receptor occupancy
For the analysis of receptor occupancy, 75 μL aliquots of whole blood sample were washed in 1.5 mL FACS buffer (PBS, Thermo Fisher Scientific + 0.5% bovine serum albumin, Sigma) and stained with fluorochrome-conjugated antibodies against CD14 (M5E2, Biolegend) and HLA-DR (L243, Biolegend), fixable viability dye eFluor 506 (Thermo Fisher Scientific), and FcR blocking reagent, human (Miltenyi Biotec). SIRPα occupancy was detected using labeled SIRPα antibodies which would compete with test article for binding to SIRPα. For detection of SIRPα occupancy, Alexa-Fluor 647 conjugated (Thermo Fisher Scientific) hAB21 was added to the lineage stains and cells were incubated at 4 °C for 1 h. All stains were then washed in 1.5 mL FACS buffer and erythrocytes were lysed in 1.5 mL FACS lysing solution (BD Biosciences) for 10 min at room temperature. Cells were washed in 1.5 mL FACS buffer and resuspended in FACS buffer. Cells were analyzed by flow cytometry on a FACS Canto II (BD Biosciences).
Data were analyzed with FlowJo 10.4 software (Becton Dickinson). To measure SIRPα occupancy on monocytes, geometric mean fluorescence intensity (MFI) in the Alexa-Fluor 647 channel was determined for CD14+HLADR+ cells. To calculate percent SIRPα occupancy, this value was normalized to the MFI for the predose timepoint as follows: MFI was calculated as a percentage of MFI at the predose timepoint. This number was subtracted from 100%, with the result being the percent occupancy. Thus, MFI greater than or equivalent to the predose level is reported as zero percent occupancy and MFI equivalent to background is reported as 100% occupancy. All samples for which SIRPα staining was equivalent to or greater than staining of predose samples were considered to have no occupancy.
Cynomolgus monkey study
The in-life of monkey study was conducted by Charles River Laboratories (Reno, NV). The study was performed in accordance with the standard operating procedures and Good laboratory Practices (GLP). Male cynomolgus monkeys (n = 2/group) of Chinese origin were administered hAB21 on Days 1 and 8 for a total of 2 intravenous doses at doses 10 and 30 mg/kg by slow bolus IV injection. Hematology was assessed by Charles River Laboratories and were collected from all animals on Days—4, 4, 8, 15, and 22. For PK bioanalysis, serum was collected from animals at the following time points: 0 (predose), 1, 3, 8, 24, 72, and 168 h postdose on Day 1. On Day 8, serum was collected from animals at the following time points: 1, 3, 8, 24, 72, 168 h, Day 18 and D22 postdose. For receptor occupancy assay, blood samples were drawn and collected for all animals at the following time points: predose (Day-4), 4, 24, and 72 h after the Day 1 dose; predose, 4, 24, 72, and 168 h after the Day 8 dose; Day 18 and Day 22.
Discussion
Blockade of the CD47–SIRPα interaction represents a promising approach to boost the antitumor activity of cancer immunotherapies when used as an adjuvant in antitumor antibody or ICI therapy. A variety of CD47 blocking antibodies or Fc-fusion proteins have shown objective responses in patients with advanced hematologic or solid tumor malignancies [
14‐
17,
39,
40]. While a multitude of agents that target CD47 have been investigated in pre-clinical models and clinical trials, only a limited number of anti-SIRPα antibodies have been reported [
32,
41]. However, due to the lack of species cross-reactivity, no anti-SIRPα antibody has, to our knowledge, been investigated in both mouse and monkey to evaluate safety and efficacy. The development of antibody-based SIRPα antagonists suitable for clinical translation has been hindered in part by polymorphisms within the CD47-binding domain of SIRPα, which necessitates pan-allele reactive anti-SIRPα antibodies for therapeutic intervention in diverse patient populations. Therefore, the impact of targeting SIRPα compared to CD47 was investigated as differences in target expression profile and function may result in divergent activities of anti-CD47 versus anti-SIRPα targeted therapies that could yield potentially distinct pharmacokinetic and safety profiles.
We previously identified anti-SIRPα antibodies from immunized chickens, which are phylogenetically distant from human, monkey, and mouse, which enabled the discovery of a diverse array of pan-allelic and species cross-reactive anti-SIRPα antibodies [
22]. Herein, we selected the anti-SIRPα antibody clone AB21 for humanization and further functional characterization due to its high affinity binding to both human SIRPα v1 and v2 alleles, comparable binding to cynomolgus monkey SIRPα, cross-reactivity with various mouse SIRPα alleles, and ability to potently block the interaction between CD47 and SIRPα. Allele cross-reactivity is an important design criterion for anti-SIRPα antibodies intended for clinical use. Several groups have reported the presence of multiple SIRPα alleles in the human population [
32,
42]. However, in our analyses of 2535 individuals from 1000 Genome Project, we determined that only two dominant SIRPα alleles, v1 and v2, are present at various frequencies dependent on ethnicity [
22]. Blockade of both SIRPα alleles is necessary to maximize efficacy; a single functional SIRPα variant can compensate for the antagonized variant. As demonstrated in Fig.
3, the v1 allele-specific anti-SIRPα antibody HEF-LB enhanced macrophage-mediated ADCP only when v1/v1 homozygous macrophages were used as effectors. HEF-LB has little to no impact on ADCP mediated by v1/v2 heterozygous or v2/v2 homozygous macrophages and on antitumor immunity by v1/v2 heterozygous donor (Fig.
3). In contrast, hAB21 is efficacious independent of SIRPα genotype (Fig.
3). SIRPα v1 specific antibodies, such as HEF-LB, would be limited to use in v1/v1 homozygote patient populations, which ranges from only 13.3–49.1% frequency among the global sub-populations [
22]. Pan-allelic antibodies such as hAB21 can potentially impact the entire population, negating the need to stratify patients prior to therapy, a desirable property for clinical translation.
In addition to species and allele cross-reactivity, the choice of IgG subclass is an important consideration when designing therapeutic antibodies. The efficacy of both antitumor and immunomodulatory antibodies is highly dependent on Fc-domain engagement of FcγRs [
43‐
45]. FcγR interactions with tumor antigen specific antibodies are crucial for the induction of antitumor immunity [
45] but can be detrimental to the activity of immunomodulatory antibodies due to unwanted depletion of antigen-positive effector cells [
43]. Toxicity may also result from FcγR interactions that promote antibody-mediated destruction of undesirable cell types and tissues, as is the case for CD47-targeted therapies capable of engaging activating FcγRs [
20]. CD47 is expressed broadly, including on RBCs and platelets. Hematological toxicities including thrombocytopenia and anemia are observed in patients treated with anti-CD47 antibodies capable of triggering FcγR-dependent effector functions [
19,
20,
39]. We have previously demonstrated that safety liabilities associated with targeting CD47 can be mitigated by eliminating binding to FcγRs, leading to the clinical development of ALX148 [
13], a fusion protein consisting of a high-affinity SIRPα variant fused to an inactive Fc-domain [
46].
To investigate the role of FcγR interactions on anti-SIRPα antibody activity, we substituted the Fc-domain of hAB21 with various natural or engineered IgG Fc-domains that have differential binding to FcγRs. Inhibiting the CD47–SIRPα interaction generally enhances macrophage-mediated ADCP of tumor cells. However, we found that potentiation of ADCP was highly dependent on the hAB21 Fc-domain; hAB21 with active Fc-domains (IgG1 and IgG4) failed to enhance ADCP, whereas hAB21 with inactive Fc-domains potentiated ADCP (Fig.
2a). This differential activity was demonstrated to be due to competition between the antitumor antibody cetuximab and hAB21 for binding to effector cell FcγR. Our data corroborate Voets et al. [
32] showing abrogation of rituximab induced phagocytosis when anti-SIRPα is expressed with an active Fc. Interestingly, DCs numbers were reduced significantly in PBMCs cultured in vitro with hAB21 containing active Fc-domains, whereas hAB21 with an inactive Fc-domain had no effect on DC numbers in PBMC cultures (Fig.
2b). Monocytes, despite expressing high levels of SIRPα, were unaffected by hAB21 irrespective of IgG subclass (Additional file
1: Fig. S1). SIRPα expression on macrophages and DCs is downregulated by hAB21 (Additional file
1: Fig. S4C) and SIRPα downregulation can reduce SIRPα tonic signaling resulting in DC activation, maturation, and turnover, a mechanism previously reported in mice treated with an anti-SIRPα antibody [
47,
48] and similar to what we observed in mice treated with hAB21 (Fig.
6). SIRPα downregulation by hAB21 may contribute to the DC depleting effect of hAB21; however, all anti-SIRPα antibodies regardless of subclass downregulate SIRPα but only hAB21 with active Fc-domains deplete DCs. We speculate that the DC depleting effect of hAB21 with an active Fc-domain is due to monocyte- and/or NK cell-mediated killing of DCs opsonized with hAB21; however, we cannot rule out that the combination of SIRPα downregulation and FcγR signaling induced by hAB21 with an active Fc-domain contributes to DC maturation and subsequent apoptosis.
A similar relationship between FcγR-binding and hAB21 activity was observed in vivo. hAB21 with a mouse IgG2a Fc-domain, which preferentially binds activating FcγRs, reduced the antitumor activity of rituximab whereas hAB21 with a mouse IgG1-N297A Fc-domain (with limited FcγR binding) enhanced rituximab antitumor activity. Although we did not investigate the in vivo mechanism responsible for these divergent activities, based on our in vitro results it is likely due to “scorpion effect”, competition between rituximab and AB21 for binding to effector cell FcγR, unwanted depletion of DCs, or all. Thus, similar to CD47-targeted therapies, ablation of FcγR-dependent effector function is desirable when targeting SIRPα, albeit for distinct reasons. Eliminating Fc-effector function from CD47 targeted therapies mitigates cytopenia [
13], while eliminating effector function from anti-SIRPα antibodies is necessary to enhance macrophage-mediated ADCP and to prevent DC depletion in vitro.
We and others have demonstrated that blocking the CD47–SIRPα interaction has a profound impact on both antitumor antibody therapy and T cell checkpoint blockade, resulting in a coordinated innate and adaptive antitumor immune response mediated by multiple cell types including macrophages, dendritic cells, neutrophils, and T cells [
9,
13,
41,
49,
50]. In most reported studies the target is CD47, and little is known about the in vivo pharmacology of targeting SIRPα. Therefore, we investigated the impact of SIRPα antagonism on antitumor immunity in mouse tumor models using hAB21 as a monotherapy or in combination with an antitumor antibody or T cell checkpoint inhibitors. We selected hAB21 containing an inactive Fc-domain for all in vivo studies due to the impact of Fc-effector function on hAB21 activity in vitro. Similar to targeting CD47 [
13,
31], SIRPα blockade with hAB21 improved responses to rituximab (anti-CD20) in the Raji human tumor xenograft model, indicating that a major mechanism of action of both anti-CD47 and anti-SIRPα therapies in xenograft models, which lack NK, B and T cells, is to enhance the antitumor activity of myeloid cells. Although we did not investigate the effector cells responsible for hAB21-mediated tumor eradication in the Raji xenograft model, Ring et al. demonstrated that neutrophils and macrophages are the main effector cells responsible for the anti-tumor activity of a SIRPα blocking antibody in human tumor xenograft models [
41]. These data are consistent with a large body of evidence implicating phagocytes as the dominant effector cell responsible for the efficacy of agents that block the CD47–SIRPα interaction in xenograft models [
31,
46,
51].
Disrupting the CD47–SIRPα interaction bridges innate and adaptive antitumor immune mechanisms [
9,
13]; therefore, we determined the activity of hAB21 in multiple syngeneic murine tumor models as a monotherapy or in combination with immune checkpoint inhibitors. These models encompass various levels of immunogenicity and sensitivity towards T cell checkpoint inhibition. As monotherapy, interfering with the CD47–SIRPα interaction generally has limited efficacy but is a powerful adjuvant to a variety of cancer immunotherapies [
31,
46]. Treatment with hAB21 alone had no effect on tumor growth in syngeneic 4T1, CT26, or B16F10 tumors (Fig.
5e and Additional file
1: Fig. S3). Modest hAB21 single agent activity was observed against MC38 tumors (Figs.
4a,
5a, b), which was dependent on the presence of CD8α
+ DCs, CD8
+ T cells and neutrophils (Fig.
4). Depletion of neutrophils by anti-GR1 but not macrophages by anti-CSF1R prior to hAB21 treatment in immunocompetent mice showed no antitumor activity, suggesting cell depletion by anti-GR1 impacted hAB21-induced antitumor activity. Although, anti-GR1 has been used extensively to deplete neutrophils in C57BL/6 mice, it does also bind GR1+
+ non-neutrophils, specifically CD8
+GR1
+ T cells [
52,
53]. The lack of antitumor activity with anti-GR1 depletion may be due to the elimination of CD8
+GR1
+ T cells, known to secrete IFNγ, neutrophils or both cell types. CD8α
+ DCs are indispensable for the production of a CD8
+ antitumor T cell response [
3] and DCs play a critical role in the responses to anti-CD47 targeted therapies [
9,
13,
33]. Administration of hAB21 to MC38 tumor bearing mice resulted in DC activation in the spleen and tumor that when further combined with anti-PD-1/PD-L1 therapy promoted both CD4+
+ and CD8
+ T cell effector functions and reduced immunosuppressive CD4
+ T
regs and TAMs leading to eradication of MC38 tumors and long-term, durable immunity. Similar mechanisms of MC38 tumor control have been reported for anti-CD47 antibodies [
33,
54], and for ALX148, a CD47 blocking Fc-fusion protein [
13], suggesting that the antitumor effects of targeting SIRPα broadly recapitulate that of CD47-targeted therapies despite differences in the ligand expression patterns and molecular and cellular functions of SIRPα and CD47.
Responses to the combination of hAB21 with anti-PD-1 or anti-4-1BB were less pronounced in more aggressive and less immunogenic tumors including CT26, 4T1, and B16F10 (Additional file
1: Fig S3). These results are consistent with additional studies that have investigated the combination of CD47 antagonists with checkpoint regulators in syngeneic mouse tumors [
49]. In more aggressive and poorly immunogenic tumors, additional treatment modalities may be necessary to improve tumor control and eradication, such as triple combinations with tumor antigen specific antibodies, CD47/SIRPα antagonists, ICIs, vaccines, or chemotherapy [
49,
55].
Due to the widespread expression of CD47, targeting SIRPα may overcome some obstacles associated with targeting CD47 [
32,
56] such as the large antigen sink that promotes target mediated clearance of anti-CD47 therapies and off-tumor toxicities due to CD47 expression on RBCs and platelets by certain anti-CD47 therapeutics. We investigated safety and PK/PD of hAB21 in an exploratory toxicology study in cynomolgus monkey. No dose-dependent adverse clinical observations or changes in hematology and serum chemistry assessment were observed with hAB21 treatment (Fig.
7), whereas anti-CD47 therapeutics capable of activating FcγR-dependent effector functions causes dose-dependent cytopenia in NHPs and humans [
19,
20]. Administration of hAB21 with an inactive Fc at both 10 mg/kg or 30 mg/kg resulted in complete target occupancy on circulating monocytes and linear PK with a half-life of 5.3 days and 8.1 days, respectively. This contrasts with anti-CD47 antibodies or anti-CD47 Fc-fusion proteins which display non-linear PK and accelerated clearance at similar dose levels due to the large CD47 antigen sink [
20]. However, soluble SIRPα was detected in the monkeys and accumulation observed during the course of the study (Fig.
7d). This is indicative of serum stabilization of the naturally produced soluble SIRPα by binding to the anti-SIRPα antibody. Soluble SIRPα-bound antibody is not available for blocking cellular CD47–SIRPα interaction, potentially limiting the impact of the longer half-life of anti-SIRPα antibodies in comparison to anti-CD47 antibodies.