Targeting Phosphatidylserine with Calcium-Dependent Protein–Drug Conjugates for the Treatment of Cancer

The use of antibody-based targeted therapies for cancer has greatly expanded over the past two decades. However, almost all antibody-based agents target protein receptors that are present at higher levels on tumors compared with normal tissue. Although many tumor cells and the tumor vascular endothelium (1–3) specifically expose the negatively charged phospholipid, phosphatidylserine, in their outer membrane leaflet, few studies have taken advantage of phosphatidylserine as a therapeutic target (4).

The identification of phosphatidylserine as a tumor marker has prompted the development of molecular targeted therapies specific for this phospholipid. Among these therapeutics is bavituximab, a chimeric mAb that targets the phosphatidylserine-binding serum protein β2 glycoprotein-1 (β2GP1; ref. 5). By cross-linking β2GP1, bavituximab binds phosphatidylserine indirectly with high affinity and elicits antiangiogenic effects via antibody-dependent cell-mediated cytotoxicity (6). In addition, by masking phosphatidylserine in the tumor microenvironment, bavituximab might also play an immunomodulatory role by inducing polarization of macrophages to the inflammatory M1 phenotype and reducing the number of myeloid-derived suppressor cells (7). In combination with radiation or chemotherapy, bavituximab has been demonstrated to be effective in multiple preclinical models for glioblastoma, pancreatic cancer, prostate, breast, and hepatocellular carcinomas (6–10). However, a recent phase III clinical trial (NCT01999673; http://www.peregrineinc.com/pipeline/bavituximab-oncology.html) in non–small cell lung cancer patients was discontinued as bavituximab in combination with docetaxel did not show sufficient improvement in overall survival over treatment with docetaxel alone. This motivates the development of strategies to improve the in vivo antitumor activity of phosphatidylserine-targeting therapies.

Antibody–drug conjugates (ADCs) combine the high specificity of antibody targeting with potent cytotoxic drugs and promise to be more effective in killing tumor cells than their corresponding “naked” antibodies (11, 12). To date, three ADCs (Mylotarg, Adcetris, and Kadcyla) have received regulatory approval, although Mylotarg has been withdrawn (11). More than 50 other ADCs are in clinical development (13, 14). This suggests that a phosphatidylserine-specific ADC may have improved efficacy over naked phosphatidylserine-targeting antibodies, such as bavituximab. However, the generation of a phosphatidylserine-specific ADC faces several challenges. First, following binding to the cell surface, ADCs require effective internalization into the lysosomal pathway. The internalization behavior of phosphatidylserine and its subsequent trafficking have not been explored. Second, in contrast with protein antigens, the development of high-affinity, specific antibodies to small-molecule “haptens” such as phosphatidylserine is challenging (bavituximab targets phosphatidylserine indirectly through binding to β2GP1). Third, under certain circumstances phosphatidylserine can also be exposed on the surface of nonapoptotic cells, such as differentiating monocytes and a subpopulation of T cells (15–17).

To generate an effective phosphatidylserine-directed drug conjugate, we produced a panel of phosphatidylserine-targeting agents by fusing naturally occurring phosphatidylserine-binding domains to the Fc portion of human IgG1. To achieve efficient dissociation of the targeting agents from phosphatidylserine in sorting (early) endosomes, followed by their lysosomal delivery, we have exploited the significant decrease in Ca²⁺ levels in endosomes (18, 19) by using phosphatidylserine-binding domains that interact with phosphatidylserine in a Ca²⁺-dependent manner. We initially analyzed the pharmacokinetics and tumor targeting of the phosphatidylserine-targeting agents in mice, and observed that despite its relatively low affinity for phosphatidylserine binding, a fusion comprising the C2A domain of synaptotagmin 1 (Fc-Syt1) had superior properties compared with higher affinity protein kinase C-α (PKCα)- and Annexin A1 (AnxA1)-based fusions. Fc-Syt1 was therefore used as a platform to generate a protein–drug conjugate (PDC) with monomethyl auristatin E (MMAE). Modulation of the avidity of Fc-Syt1 for phosphatidylserine binding demonstrated that a bivalent Fc-Syt1–based PDC had improved therapeutic efficacy over a tetravalent variant due to more favorable pharmacokinetic properties. Our observations indicate the need for affinity/avidity tuning of phosphatidylserine-targeting PDCs and provide novel insight into the design of therapeutics to target this phospholipid.

2H11 mouse tumor endothelial cells (ATCC, CRL-2163) and MDA-MB-231 human breast cancer cells (ATCC, HTB-26) were cultured in DMEM supplemented with 5% and 10% FBS, respectively. T-47D breast cancer cells (ATCC, HTB-133), LNCaP and 22Rv1 human prostate cancer cells (ATCC, CRL1740 and CRL-2505) were cultured in RPMI1640 medium supplemented with 10% FBS. SK-BR-3 human breast cancer cells (ATCC, HTB-30) were cultured in McCoy's 5A medium supplemented with 10% FBS. All cell lines were authenticated with DNA fingerprinting by the University of Arizona Genetics Core on November 16, 2016, and tested for mycoplasma at monthly intervals. All cell lines were used within 15 passages following thawing of master stocks. Cells were cultured at 37°C with 5% CO₂. Expi293F cells were used for protein expression and were cultured in Expi293 expression medium (Life Technologies, catalog number A14635) at 37°C with 8% CO₂ and 80% humidity.

The following antibodies were used in this study: rat anti-mouse LAMP1, mouse anti-human LAMP1, and mouse anti–β-tubulin antibodies (Developmental Studies Hybridoma Bank, clone # 1D4B, H4A3 and E7); mouse anti-human EEA1 and rat anti-mouse CD31 antibodies (BD Biosciences, catalog # 610456 and 557355); goat anti-human IgG (H + L) antibody conjugated with HRP, donkey anti-rat (H + L) antibody conjugated with Alexa Fluor 488 and donkey anti-human IgG (H + L) antibody conjugated with Cy3 (Jackson ImmunoResearch, catalog # 109-035-003, 712-545-153 and 709-165-149); goat anti-human IgG (H + L) antibody conjugated with Alexa Fluor 555, goat anti-mouse IgG (H + L) antibody conjugated with Alexa Fluor 488 and goat anti-human IgG (H + L) antibody conjugated with Alexa Fluor 647 (Life Technologies, catalog # A21433, A11029, and A21445); rabbit anti-human Ki-67 antibody (Abcam, catalog # 92742).

Proteins were labeled with Alexa 647 or IRDye800CW with the Alexa Fluor 647 NHS Ester (Life Technologies, catalog # A37573) or IRDye800CW Protein Labeling Kit (LI-COR, catalog # 928-38040), respectively, following the manufacturer's instructions. Proteins were labeled with I-125 as described previously (20). Radio-TLC analysis was used to determine radioiodination efficiency and radiochemical purity using a Rita Star Radioisotope TLC Analyzer with ITLC-SG plates and PBS as the mobile phase. The typical radiochemical yield was 68% to 76%, with radiochemical purity of >99.5% and specific activity of 10 to 12 μCi/μg. The absence of aggregation for fluorophore-labeled proteins was verified by gel filtration analyses.

For use as controls, the Fc region including the hinge region (residues 214–446 of the heavy chain) of the hen egg lysozyme-specific human IgG1, HuLys10 (21), was cloned into the pcDNA3.4 vector (Invitrogen, catalog # 14308) with the N-terminal leader peptide derived from a mouse IgG heavy chain (21, 22). Similarly, the genes encoding the heavy and light chain genes (cDNA) of the HuLys10 antibody were cloned into pcDNA3.4. For the control IgG heavy and light chain constructs, Cys214 in the light chain, which forms a sulfhydryl bridge with Cys219 in the heavy chain of HuLys10 were both mutated to serine residues using the QuikChange II Site-Directed Mutagenesis Kit (Agilent Technologies, catalog # 200523).

cDNA clones for human Annexin A1 (AnxA1), human synaptotagmin 1 (Syt1), and human PKCα were purchased from Open Biosystems (clone ID: 3459615, clone ID: 6187902 and clone ID: 40028305, respectively). Genes encoding the AnxA1 phosphatidylserine-binding core domain (amino acids 41–346), the Syt1 phosphatidylserine-binding C2A domain (amino acids 141–266) and PKCα phosphatidylserine-binding C2 domain (amino acids 157–288) were fused via a Gly₄Ser linker sequence to the CH₃ domain of the human IgG1 Fc region (residues 214–446) with a leader peptide derived from the mouse IgG heavy chain (21, 22). Cys219 in the hinge region was mutated in all Fc fusion constructs so that there are two cysteine residues per hinge. The genes encoding the Fc fusions were cloned into the pcDNA3.4 vector.

To generate Syt1-Fc-Syt1, the Syt1 phosphatidylserine-binding C2A domain (amino acids 141–266) was linked to the N-terminus of the hinge region of the Fc-Syt1 construct via a Gly₄Ser linker sequence. The leader peptide derived from a mouse IgG heavy chain (21, 22) was appended to the N-terminus of the hinge-linked Syt1 phosphatidylserine-binding C2A domain, and the resulting Fc fusion was cloned into the pcDNA3.4 vector. Mutations to reduce phosphatidylserine binding of the Syt1 C2A domain (D173N, D179N, D231N, D233N, and D239N; ref. 23) were inserted in the Fc-Syt1 construct to generate Fc-Syt1(DN) and cloned into the pcDNA3.4 vector. All constructs were generated using standard methods of molecular biology and designed oligonucleotides. Following construction of expression plasmids, sequences were verified. Oligonucleotide and construct sequences are available upon request.

All phosphatidylserine-targeting proteins used in this study were produced using the Expi293 expression system from Life Technologies following the manufacturer's instructions. Briefly, cells were transfected with expression constructs for 6 to 7 days and Fc fusion proteins were purified from culture supernatants using protein G-Sepharose. Bound proteins were eluted using 50 mmol/L diethylamine with 150 mmol/L NaCl. The eluted protein was neutralized using 2 mol/L Tris pH 7.0 followed by dialysis against PBS. All proteins were concentrated and loaded onto a Hiload 16/60 Superdex 200 gel filtration column (GE Healthcare). The homodimeric, nonaggregated form of the protein was separated, concentrated, and analyzed using a Superdex 200 15/30 gel filtration column (GE Healthcare).

Recombinant mouse FcRn was produced as described previously (24). Briefly, High Five cells grown at 27°C in EX-CELL 405 medium (Sigma, catalog # 14405) were infected at a density of 1 × 10⁶ cells/mL with recombinant baculovirus (mouse FcRn α-chain/mouse β2-microglobulin). Cells were cultured at 23°C to 24°C for 72 hours. Mouse FcRn was purified from the supernatant using Ni²⁺-NTA agarose (Qiagen) followed by the use of a Hiload 16/60 Superdex 200 gel filtration column.

For MALDI-TOF analyses, all samples (20–40 μg diluted in 20 μL PBS) were treated with 1 μL (10 U) PNGase F (Promega, catalog # V4831) at 37°C for 1 hour. Samples were extracted using C₄ LithTips (Protea, catalog # SP-410) and analyzed with sinapic acid under linear conditions with a Kratos/Shimadzu Axima CFR MALDI-TOF mass spectrometer.

Surface plasmon resonance experiments were carried out using a BIAcore T200 (GE Healthcare). To determine the equilibrium binding affinities of mouse FcRn for the phosphatidylserine-targeting agents (Fc-AnxA1, Fc-Syt1, and Fc-PKCα), control human IgG1 and Fc, mouse FcRn was injected over immobilized proteins (coupled at ∼750–1,500 RU on flow cells of CM5 sensor chips) at a flow rate of 10 μL/minute in PBS (pH 6.0) with 0.01% v/v Tween 20. Flow cells were regenerated at the end of each run using 0.15 mol/L NaCl, 0.1 mol/L NaHCO₃, pH 8.5. The dissociation constants (K_Ds) for the interactions of mouse FcRn with different proteins were determined using custom written software (25, 26). IgG or Fc has two possible interaction sites on mouse FcRn, and equilibrium binding data were fitted as described previously (26). The equilibrium dissociation constants for the higher affinity interaction site are reported.

Lipid-coated membrane strips (Echelon, catalog # P-6002) were first hydrated with TBST (20 mmol/L Tris, 150 mmol/L NaCl, 0.1% Tween 20, pH 7.5) and then incubated with blocking solution (4% fatty acid free BSA dissolved in TBST) at room temperature for 1 hour. Proteins were diluted to a concentration of 2 μg/mL in blocking buffer and incubated with membranes at room temperature for 2 hours. The lipid strip was then washed with TBST and bound proteins detected using HRP-conjugated goat anti-human IgG (H + L) antibody.

2H11 or MDA-MB-231 cells were grown on coverslips (Zeiss, ref # 0109030091) and incubated with either control human IgG1 (HuLys10) or phosphatidylserine-specific agents diluted in growth medium for the indicated time periods in figure legends. Cells were then washed with PBS and fixed with ice-cold 4% PFA at room temperature for 20 minutes. Following fixation, cells were permeabilized with 0.1% Triton X-100 and incubated with blocking buffer (PBS, 0.1% Tween 20, 5% serum) at room temperature for 30 minutes. Primary antibodies specific for mouse LAMP-1 (clone 1D4B), human LAMP-1 (clone H4A3), or mouse early endosomal antigen-1 (EEA1) were diluted in blocking buffer and incubated with cells at room temperature for 2 hours. For staining of microtubule networks, cells were fixed with methanol at −20°C for 3 minutes. Following fixation, cells were incubated with blocking buffer (PBS, 0.1% Tween 20, 5% serum) at room temperature for 30 minutes. The primary antibody specific for β-tubulin was diluted in blocking buffer and incubated with cells at room temperature for 2 hours. After incubation with the primary antibodies, cells were washed with PBST (PBS, 0.1% Tween 20) and incubated with fluorescently labeled secondary antibodies diluted in blocking buffer for 1 hour at room temperature. To detect internalized phosphatidylserine-specific agents, fluorescently labeled goat or donkey anti-human IgG (H + L) antibody was used. Following incubation, cells were washed with PBST and mounted with ProLong Gold Antifade Mountant (Life Technologies, catalog # P36930).

Fluorescent images were acquired using a Zeiss Axiovert 200M inverted fluorescence microscope with a 63×, 1.4 NA plan apochromat objective (Carl Zeiss) and a 1.6× internal optovar. Fluorescent filter sets for GFP and Alexa 555 (Chroma Technology, catalog # 41017 and 41002b) were used for image acquisition. The acquired data were processed using the microscopy image analysis tool (MIATool) software (www4.utsouthwestern.edu/wardlab/miatool.asp).

To study the Ca²⁺-dependence of binding of the phosphatidylserine-specific Fc fusions to phosphatidylserine, proteins (100 nmol/L) were diluted in binding buffer (10 mmol/L HEPES pH 7.4, 150 mmol/L NaCl with 2 mmol/L or 2 μmol/L Ca²⁺). Phosphatidylserine-coated beads (50 μl; Echelon, catalog # P-B0PS) were added and incubated at room temperature for 2 hours. The beads were then washed with the binding buffer and bound proteins detected by immunoblotting with HRP-conjugated goat anti-human IgG (H + L) antibody. To examine the pH-dependent binding to phosphatidylserine, proteins were diluted at 100 nmol/L in PBS pH 7.4 or 6.0. Fifty microliters (bed volume) phosphatidylserine-coated beads were added, incubated at room temperature for 2 hours, and samples processed as above.

Cells (1 × 10⁶) were suspended in Annexin V–binding solution (10 mmol/L HEPES pH 7.4, 150 mmol/L NaCl, 2.5 mmol/L CaCl₂). Annexin V conjugated with Alexa 488 (Life Technologies, catalog # A13201) was added to the cell suspension at a 1:100 dilution and incubated with the cells for 10 minutes at room temperature. Cells were then washed once with Annexin V–binding solution and analyzed by flow cytometry (BD FACSCalibur). Flow cytometry data were processed using FlowJo (FlowJo, LLC).

Cells were trypsinized and resuspended in flow cytometry buffer (PBS with Ca²⁺/ Mg²⁺, 1% BSA). Phosphatidylserine-specific Fc fusions (50 nmol/L) were incubated with the cells for 30 minutes at either room temperature or on ice depending on the assay. Cells were washed with flow cytometry buffer and incubated with fluorophore-conjugated secondary antibodies on ice for 30 minutes. Cells were then washed and analyzed by flow cytometry (BD FACSCalibur). Flow cytometry data were processed using FlowJo (FlowJo LLC).

Fc fusion or control antibody (20 μmol/L) in PBS was incubated with 320 μmol/L TCEP at room temperature for 3 hours to reduce the hinge disulfide bonds. Maleimidocaproyl-val-cit-PAB-MMAE (MC-VC-PAB-MMAE; Levena Biopharma, catalog # SET0201) was then added to the reduced Fc fusion or control antibody at a concentration of 160 μmol/L and incubated at room temperature for 3 hours. Following the conjugation reaction, free MMAE was removed by dialysis of the protein against PBS. The conjugated Fc fusion or control antibody was stored at 4°C.

Cells were plated into 96-well plates as follows: 2H11, 10,000 cells/well; MCF-7, SK-BR-3 and LNCaP, 5,000 cells/well; MDA-MB-231, 3,000 cells/well; 22Rv1, 15,000 cells/well. Cells were grown overnight followed by the addition of phosphatidylserine-specific PDCs. Cell growth and survival were measured after a 3- to 5-day incubation with the Cell Proliferation AQ One Solution Cell Proliferation Assay Kit (Promega, catalog # G3581). Dose–response curves were plotted and IC₅₀ values determined using GraphPad Prism software.

Animal procedures used in all mouse studies were approved by the Institutional Animal Care and Use Committee of the University of Texas Southwestern Medical Center (Dallas, TX) and Texas A&M University (College Station, TX). All BALB/c SCID mice used in the study were purchased from The Jackson Laboratory (stock # 001803) and bred in-house. Pharmacokinetic studies were performed as described previously (20). Briefly, Lugol solution was added to drinking water 96 hours before the experiments. SCID BALB/c female mice (8 weeks old; 18–22 g weight) were anesthetized using 2% isoflurane in oxygen and injected (intravenously) with ¹²⁵I-labeled proteins (100–120 μCi, 10–12 μg/mouse) in 200 μL PBS with 0.1% BSA. Whole-body radioactive counts were measured using a Biodex Atomlab 100 dose calibrator. Blood radioactive counts were measured by retro-orbitally bleeding mice with 10 μL capillary tubes (Drummond) and radioactive counts (cpm) determined by gamma counting (PerkinElmer).

For whole-body, near-infrared imaging (NIR), female nude mice (6–7 weeks old; purchased from Envigo, catalog # 6903F) or BALB/c SCID mice (6–8 weeks old) were used. For implantation of MDA-MB-231 tumors, mice were anesthetized with 2% isoflurane in oxygen, and a small surgical incision was made to expose the mammary fat pad. MDA-MB-231 cells were trypsinized and dispersed into single-cell suspensions in PBS. A total of 5 × 10⁶ cells per mouse were injected in 100 μL into the mammary fat pad, using a 25 G needle, and the incision was then sealed with a wound clip. Buprenorphine was administered (subcutaneously) at 50 μg/kg immediately following the surgery and 12 hours later. Mice were monitored daily, and the wound clip was removed one week postsurgery. For imaging of nude mice, mice were divided into 3 groups (n = 3 mice/group) when tumors reached sizes of approximately 150 mm³ and administered (intraperitoneally) with 5 mg/kg docetaxel 72 and 48 hours before being injected (intravenously) with 1 nmol IRDye800CW-labeled phosphatidylserine-specific agents in PBS. Fluorescence imaging (FLI) was performed using a Caliper Xenogen IVIS Spectrum (PerkinElmer) in vivo imaging system at 0 (before injection) and 3, 24, and 48 hours postinjection. FLI was performed using 745 nm excitation, 800 nm emission, binning 8, FOV 12.9 cm, f-stop 2, and autoexposure. Data were quantitated with the Living Imaging software using absolute radiant efficiency (photons/second) in an ROI, manually drawn to outline the FLI signal of the tumor and normalized to the tumor volumes.

For analyses of tumor localization in BALB/c SCID mice, mice were divided into three groups (n = 3 mice/group) when tumors reached sizes of approximately 300 mm³ and administered (intraperitoneally) with 5 mg/kg docetaxel 72 and 48 hours before being injected (intravenously) with 1 nmol of IRDye800CW-labeled phosphatidylserine-specific agents in PBS. Forty-eight hours following injection, tumors were dissected out and imaged as above. Fluorescence in an ROI, manually drawn to outline the FLI signal of the tumor, was quantitated and normalized to the tumor weight.

Implantation of MDA-MBA-231 tumor xenografts in BALB/c SCID mice was carried out as described for the whole-body imaging experiments. For implantation of LNCaP tumors, 7- to 8-week-old male BALB/c SCID mice (18–22 g weight) were anesthetized with 2% isoflurane in oxygen and 5 × 10⁶ LNCaP cells suspended in 50% RPMI and 50% Matrigel (BD Biosciences) were injected (subcutaneously). When MDA-MBA-231 or LNCaP tumors reached a size of approximately 100 mm³, mice were injected (intraperitoneally) with 5 mg/kg docetaxel 72 and 48 hours before the treatment. Mice were then injected (intravenously) with 1 nmol unconjugated proteins, PDCs, or PBS vehicle as indicated in the figure legends twice per week. Tumors and body weights were measured twice a week. For the treatment experiment using Fc-Syt1(DN)_MMAE, mice were treated for 4 weeks and monitored for another 2.5 weeks. Experiments were terminated when the tumor size reached 2 cm in any dimension.

Female BALB/c SCID mice bearing MDA-MB-231 tumors were treated (intraperitoneally) with 5 mg/kg docetaxel 72 and 48 hours before intravenous delivery of either PBS vehicle or 1 nmol Fc-Syt1 conjugated with MMAE. At different time points, mice were perfused (intracardiac) with PBS followed by 4% PFA. Tumors were then dissected out, embedded in OCT (Thermo Fisher Scientific, catalog # 23-730-571) and stored at −80°C. Ten-micron tissue sections were cut and hydrated with PBS at room temperature before fixation with 4% PFA for 30 minutes. Tumor sections were then washed with PBS and incubated with the permeabilization/blocking solution (PBS, 0.5% Triton X-100, 3% BSA) at room temperature for 1 hour. Primary antibodies for human Ki-67, mouse CD31, and F4/80 were diluted in the blocking buffer (PBS, 0.1% Tween 20, 5% serum) and incubated with the tissue sections at 4°C overnight. The next day, tissue sections were washed with PBST (PBS, 0.1% Tween 20) and incubated with the fluorophore-conjugated secondary antibodies diluted in the blocking buffer at room temperature for 2 hours. After washing with PBST, tissue sections were mounted with ProLong Gold antifade mounting medium. Confocal images were acquired using a Nikon A1R confocal microscope equipped with a 40×, 1.3 NA plan fluor objective and processed with the NIS-Elements software (Nikon).

A panel of phosphatidylserine-targeting agents was generated by fusing the Fc region of human IgG1 to the following phosphatidylserine-binding domains: core domain of Annexin A1 (AnxA1), C2A domain of Synaptotagmin 1 (Syt1), and C2 domain of PKCα (Fig. 1A). The three domains bind to phosphatidylserine in a Ca²⁺-dependent manner (4, 27, 28), and the resulting fusion proteins were designated Fc-AnxA1, Fc-Syt1, and Fc-PKCα, respectively. The phosphatidylserine-targeting agents were expressed and purified as homodimers (Fig. 1A; Supplementary Fig. S1) and bound to phosphatidylserine in a lipid-binding assay (Fig. 1B). They also bound to cardiolipin, which is located on the inner mitochondrial membrane of eukaryotic cells and therefore not relevant to targeting. In contrast with Fc-Syt1 and Fc-PKCα, Fc-AnxA1 exhibited a broader lipid-binding profile and bound both neutral and negatively charged lipids (Fig. 1B). Importantly, none of the phosphatidylserine-binding agents bound phosphatidylcholine and sphingomyelin, lipids that are present in the plasma membrane outer leaflet.

The tumor endothelial cell line 2H11 (29) and triple-negative breast tumor cell line MDA-MB-231 were used to investigate the ability of the phosphatidylserine-binding agents to interact with lipids on the cell surface. Binding of fluorescent Annexin V showed that both of these cell lines expose phosphatidylserine and that phosphatidylserine exposure increases following docetaxel treatment (Supplementary Fig. S2). All phosphatidylserine-binding agents interacted with phosphatidylserine-positive cells, with Fc-Syt1 showing lower levels of binding than Fc-AnxA1 and Fc-PKCα (Fig. 1C).

To determine which recombinant proteins were suitable for further development as PDCs, we evaluated the pharmacokinetic behavior and tumor localization of the three phosphatidylserine-binding agents in mice. Pharmacokinetic studies of the phosphatidylserine-binding agents revealed that Fc-Syt1 persisted at both the whole-body and blood levels for significantly longer than the other two proteins (Fig. 1D and E). Binding analyses demonstrated that these three fusion proteins had similar affinities for binding to recombinant mouse FcRn at pH 6.0, with no detectable binding or affinity too low to accurately determine at pH 7.4 (Table 1), indicating that affinity differences do not contribute to the longer persistence of Fc-Syt1. The proteins were labeled with the residualizing dye, IRDye800CW, and injected (intravenously) into female nude mice bearing MDA-MB-231 xenografts and imaged at different time points (Fig. 1F) with tumor fluorescence quantitated 48 hours postinjection (Fig. 1G). Similar experiments were carried out in BALB/c SCID mice, and 48 hours later, tumors were excised and dye levels determined (Fig. 1H and I). Among the three phosphatidylserine-specific agents, Fc-Syt1 exhibited the highest level of tumor localization. Collectively, these results indicated that of the three phosphatidylserine-targeting agents, Fc-Syt1 is the most promising for conversion to a PDC for use in therapy.

For therapeutic efficacy of an antibody–drug conjugate (ADC), target cell internalization of the ADC and release of the toxic payload in late endosomes or lysosomes is essential (11). In several receptor systems, it has been shown that multivalent ligands, or mixtures of cross-linking ligands, such as antibodies, promote receptor internalization and degradation (30–32). To study the role of avidity in the behavior of phosphatidylserine-targeting PDCs, we also generated tetravalent Syt1-Fc-Syt1 that contains four Syt1 C2A domains (Fig. 2A). The tetravalent protein was purified as a homodimer (Fig. 2A; Supplementary Fig. S3A). As designed, Syt1-Fc-Syt1 has higher affinity/avidity for phosphatidylserine than Fc-Syt1 in lipid-binding assays (Fig. 2B) and the same lipid selectivity as its bivalent parent, Fc-Syt1 (Fig. 2C). Consistent with the binding data shown in Fig. 2B, the tetravalent Syt1-Fc-Syt1 bound to 2H11 cells at significantly higher levels than bivalent Fc-Syt1 protein (Fig. 2D).

We next compared the internalization of Fc-Syt1 and Syt1-Fc-Syt1 using Alexa 647–labeled proteins. 2H11 cells were incubated with labeled Fc-Syt1 and Syt1-Fc-Syt1 on ice at different concentrations to achieve similar surface binding followed by incubation at 37°C to allow internalization for different times. Surface-bound proteins were stripped off with EDTA (due to the Ca²⁺-dependent of binding), and internalized levels (resistant to stripping) were determined by flow cytometry. These studies indicated that although both proteins efficiently accumulated within cells, the tetravalent Syt1-Fc-Syt1 was internalized more rapidly compared with bivalent Fc-Syt1 (Fig. 2E). Fluorescent microscopy was also used to study the subcellular trafficking behavior of the Syt1-Fc fusion proteins. In both 2H11 and MDA-MB-231 cells, Fc-Syt1 and Syt1-Fc-Syt1 were internalized and delivered into LAMP-1 positive lysosomes (Fig. 2F and G; Supplementary Fig. S3B).

The lysosomal trafficking and internalization behavior of Fc-Syt1 and Syt1-Fc-Syt1 indicated that they could be effective as delivery vehicles for conjugated drugs. We therefore conjugated maleimidocaproyl-val-cit-PAB-MMAE (33) to hinge cysteines (Fig. 3A, left). For use as a negative control, MMAE was conjugated to a hen egg lysozyme-specific human IgG1 (21) in which the light/heavy chain-interacting cysteine residues were mutated to serine to result in four possible conjugation sites. SDS-PAGE analyses indicated that the conjugations had gone to completion and resulted in drug-to-antibody ratios (DAR) of four (Fig. 3A, right). The DARs of four for both MMAE-conjugated Fc-Syt1 and Syt1-Fc-Syt1 were confirmed by MALDI-TOF mass spectrometry. As complete conjugation disrupted the two disulfide bonds in the hinge region, apparent molecular weights of 43.6 kDa for Fc-Syt1_MMAE and 58.6 kDa for Syt1-Fc-Syt1_MMAE, respectively, were observed in these analyses (Supplementary Fig. S4A and S4B, top). This contrasted with the unconjugated or partially conjugated protein that retained two or one disulfide bonds, respectively, resulting in an apparent molecular weight of approximately 82 kDa for Fc-Syt1 and 112 kDa for Syt1-Fc-Syt1 (Supplementary Fig. S4A and S4B, bottom). Importantly, the conjugation process did not lead to protein aggregation (Supplementary Fig. S4C).

The Syt1 C2A domain requires Ca²⁺ for phosphatidylserine binding (23). The lower Ca²⁺ concentration in early/sorting endosomes (∼2 μmol/L) compared with the extracellular Ca²⁺ levels (1–2 mmol/L) suggests that following internalization, phosphatidylserine-specific PDCs will dissociate from the limiting membrane of these endosomes (18, 19). This dissociation is expected to lead to improved lysosomal delivery. Both phosphatidylserine-specific PDCs containing Syt1 domains bound to phosphatidylserine-beads in buffer containing 2 mmol/L Ca²⁺, but no detectable interaction was observed when the Ca²⁺ concentration was decreased to 2 μmol/L (Fig. 3B, left). In addition, because the pH within sorting (early) endosomes is acidic (pH 6.0–6.5; ref. 34), we analyzed the effect of pH on PDC:PS interactions. Both PDCs bound to phosphatidylserine at similar levels in the pH range 6.0 to 7.4 (Fig. 3B, right). Consistent with the in vitro binding analyses demonstrating Ca²⁺-dependent binding, fluorescence microscopy analyses showed the presence of the phosphatidylserine-specific PDCs in the lumen, rather than limiting membrane of sorting endosomes in MDA-MB-231 cells following 15- and 30-minute coincubation with PDC (Fig. 3C; Supplementary Fig. S5A). The relatively large size (∼1.5 μm diameter) of the sorting endosomes in MDA-MB-231 cells, which was also observed with untreated cells (Supplementary Fig. S5B), enabled PDC dissociation from the limiting endosomal membrane to be quantitated (Fig. 3C, right). Fc-Syt1_MMAE and Syt1-Fc-Syt1_MMAE could be detected in lysosomes within 30 minutes, and the level of Fc fusion increased following 4 hours of coincubation (Fig. 3D; Supplementary Fig. S5C). In addition, both PDCs disrupted the microtubular network in 2H11 and MDA-MB-231 cells (Fig. 3E and F).

We examined the effects of phosphatidylserine-specific PDCs on the growth of multiple cell lines, including tumor endothelium (2H11), ER-positive breast cancer (T-47D), HER2-positive breast cancer (SK-BR-3), triple-negative breast cancer (MDA-MB-231), androgen-sensitive prostate cancer (LNCaP), and castration-resistant prostate cancer (22Rv1). Staining of the cells with fluorescently labeled Annexin V showed that all of these cell lines were phosphatidylserine positive (Supplementary Fig. S6A). Incubation with phosphatidylserine-specific PDCs effectively inhibited the growth and survival of the cells in a dose-dependent manner (Fig. 4; Table 2). Although the tetravalent Syt1-Fc-Syt1_MMAE was more potent in inhibiting T-47D cell growth than the bivalent Fc-Syt1_MMAE, the two PDCs exhibited similar effects on the other cell lines. Consistent with the growth-inhibitory effects, the internalization of the two PDCs at concentrations close to the corresponding IC₅₀s were similar for all cell lines, except that Syt1-Fc-Syt1_MMAE was internalized at higher levels than Fc-Syt1_MMAE for T-47D cells (Supplementary Fig. S6B). Relatively high concentrations of the control IgG (hen egg lysozyme-specific human IgG1) conjugated with MMAE resulted in inhibition of cell growth, possibly due to nonspecific fluid phase uptake of the drug, whereas unconjugated phosphatidylserine-targeting proteins showed no effect on cell growth (Supplementary Fig. S6C). Thus, our data indicate that Syt1-based PDCs are potent inhibitors of tumor endothelial and cancer cell growth in vitro.

We next investigated the therapeutic effects of the phosphatidylserine-specific PDCs against tumor xenografts in SCID mice. Prior to therapy, pharmacokinetic studies of the phosphatidylserine-specific PDCs demonstrated that tetravalent Syt1-Fc-Syt1_MMAE had a shorter half-life than bivalent Fc-Syt1_MMAE (Fig. 5A and B), possibly due to increased target-mediated uptake. Because of the relatively short half-lives, we delivered the PDCs (1 nmol/mouse) into tumor-bearing mice twice per week. In female SCID mice bearing orthotopic MDA-MB-231 breast tumors that had been pretreated with docetaxel, bivalent Fc-Syt1_MMAE potently blocked breast tumor growth (Fig. 5C). Tetravalent Syt1-Fc-Syt1_MMAE also inhibited tumor growth, but was less effective than bivalent Fc-Syt1_MMAE. Treatment with unconjugated Fc-Syt1 and Syt1-Fc-Syt1 had no effect on tumor growth (Fig. 5C).

In addition to efficacy in the breast tumor model, bivalent Fc-Syt1_MMAE completely blocked tumor growth in male SCID mice bearing prostate cancer LNCaP xenografts that had been pretreated with docetaxel. Consistent with the in vitro data (Supplementary Fig. S6C), the unconjugated protein had no effect (Fig. 5D). However, tetravalent Syt1-Fc-Syt1_MMAE did not significantly inhibit tumor growth. The most likely explanation for this difference is the shorter in vivo persistence of the tetravalent PDC. Taken together with the breast tumor data, these results indicate that the valency of the phosphatidylserine-specific PDCs has a profound influence on their therapeutic outcomes. Importantly, we did not observe weakness or loss of body weight in any of the treatment groups (Fig. 5E; Supplementary Fig. S7), indicating that phosphatidylserine-specific PDCs are well tolerated in vivo.

To further validate that Fc-Syt1_MMAE binds phosphatidylserine-positive cells in tumor tissue following docetaxel treatment, we performed IHC following delivery of this PDC into tumor-bearing mice. We found that Fc-Syt1_MMAE was localized to CD31-positive blood vessels (Fig. 5F), tumor cells, and tumor-infiltrating F4/80-positive macrophages (Fig. 5G), which can expose phosphatidylserine (5). The data indicate that cancer cells not only expose phosphatidylserine in vitro (Supplementary Fig. S6A; refs. 3, 35), but retain this loss of phosphatidylserine asymmetry in vivo.

To exclude the possibility that the drug accumulated in the tumor through nonspecific mechanisms, such as the enhanced permeability and retention effect (36), and to show that the in vivo efficacy was dependent on phosphatidylserine binding, we generated a mutated variant of the synaptotagmin 1 C2A domain with decreased affinity for phosphatidylserine. The C2A domain of synaptotagmin 1 interacts with phosphatidylserine through three Ca²⁺ ions chelated by five aspartic acids (D) in domain loops I and III (23). To ablate Ca²⁺ binding, we mutated all five aspartic acid residues to asparagines (N) to produce Fc-Syt1(DN). Fc-Syt1(DN) was purified as a homodimer and interacted with phosphatidylserine at essentially background levels in the protein–lipid overlay assay (Fig. 6A and B). Moreover, Fc-Syt1(DN) had significantly reduced binding to phosphatidylserine-positive cells compared with its wild-type counterpart (Fig. 6C). This mutated protein and as a control, human IgG1-derived Fc, were conjugated to MMAE and analyzed using mass spectrometry. These analyses indicated conjugation ratios of four drugs per protein molecule (Supplementary Fig. S8A and S8B). In addition, size exclusion analyses demonstrated that the resulting MMAE-conjugated proteins were not aggregated (Supplementary Fig. S8C). Mice bearing orthotopic MDA-MB-231 tumors were pretreated with docetaxel prior to treatment with PDCs (1 nmol/mouse) or vehicle control for 4 weeks, until mice in the control (PBS) group were euthanized due to their large tumor sizes. Treatment of MDA-MB-231 tumors with Fc-Syt1_MMAE led to potent growth inhibition (Fig. 6D). More importantly, tumor growth remained inhibited after the treatment with Fc-Syt1_MMAE was stopped at 4 weeks. No tumors could be isolated from 3 of 6 mice at the end of the experiment (Fig. 6E). Although the delivery of Fc_MMAE or Fc-Syt1(DN)_MMAE slowed tumor growth initially, rapid proliferation was observed following the end of treatment (Fig. 6D). Collectively, the data indicate that phosphatidylserine-binding is essential for the activity of Fc-Syt1_MMAE.

ADCs represent a rapidly expanding class of therapeutics to target cancer. For the majority of ADC or PDC platforms, therapeutic efficacy is dependent on internalization and trafficking of the cytotoxic drug to lysosomes (11, 12). The current study describes a novel approach for the generation of a PDC to its target, namely exposed phosphatidylserine on tumor cells and tumor vascular endothelium. First, we have identified a phosphatidylserine-targeting protein that is efficiently internalized into cells that expose phosphatidylserine. Second, we have exploited the Ca²⁺ dependence of the phosphatidylserine–Syt1 interaction to enable endosomal dissociation, and consequent lysosomal delivery, of the Syt1-based PDC. Third, we have tuned the binding strength of the phosphatidylserine-targeting agent to optimize pharmacokinetic behavior and delivery to the tumor site. We demonstrate that the resulting PDC, Fc-Syt1_MMAE, has potent antitumor effects in mouse models of breast and prostate cancer.

Among the three phosphatidylserine-targeting agents developed here, both Fc-AnxA1 and Fc-PKCα exhibited higher levels of binding to target cells than Fc-Syt1. However, of these three agents, Fc-Syt1 exhibited the highest tumor accumulation in mice. The data indicate that high-affinity interactions with phosphatidylserine negatively impact the in vivo behavior with respect to tumor localization. There are two possible reasons for this behavior: First, phosphatidylserine exposure has also been identified in nonapoptotic normal cells, such as differentiating monocytes and a subpopulation of T cells (16, 17). Although the SCID mice used in the current study lack T cells, monocytes may therefore act as a sink for target-mediated clearance outside the tumor tissue. This sink effect is expected to be greater for higher affinity phosphatidylserine-specific agents, resulting in shorter in vivo persistence. Consistent with this, pharmacokinetic studies demonstrate that Fc-Syt1 has a significantly longer half-life compared with Fc-AnxA1 and Fc-PKCα. Furthermore, bivalent Fc-Syt1_MMAE also displayed a longer half-life and better in vivo efficacy than higher avidity, tetravalent Syt1-Fc-Syt1_MMAE. Second, besides the influence on in vivo persistence, higher affinity HER2-specific antibodies have been shown to penetrate tumors less efficiently than their lower affinity counterparts [the “binding barrier theory” (37–39)]. Similar pathways may operate during phosphatidylserine targeting.

Interestingly, our study showed Fc-Syt1_MMAE could potently inhibit tumor growth in mice, whereas unconjugated Fc-Syt1 had no effect. This contrasts with bavituximab, which has been reported to be therapeutically effective as an unconjugated antibody in preclinical models when combined with radiation, chemotherapy, or other procedures that increase phosphatidylserine exposure in tumor tissues (6–10). One possible reason for this difference is that we delivered the chemotherapeutic drug, docetaxel, as two doses at the beginning of the treatment to elevate tumor phosphatidylserine exposure, whereas in the bavituximab study, mice were treated in combination with antibody and docetaxel throughout therapy. The importance of increasing phosphatidylserine exposure using either chemotherapy or radiation for tumor localization of phosphatidylserine-targeting agents was indicated by a recent study in which two antibodies with β2GP1-independent specificity for phosphatidylserine, phosphatidic acid, and cardiolipin failed to accumulate in tumor-bearing mice that were not treated with docetaxel or other “phosphatidylserine-inducing” agents (40).

The induction of endosomal dissociation of antibodies from their target antigen, by exploiting the lower pH or Ca²⁺ concentration within early/sorting endosomes relative to the extracellular environment, has been described by us and others as a strategy to clear antigens such as inflammatory cytokines or their soluble, signaling competent receptors (41–43). This release of target is followed by recycling and exocytosis of the antibody through binding to the Fc receptor, FcRn, which is expressed in most normal parenchymal and hematopoietic cells (20, 44, 45). To date, this approach has been implemented by using antibody engineering to generate antibodies that bind with pH-dependence or Ca²⁺-dependence (41–43). Here, we use naturally occurring proteins that have intrinsic Ca²⁺-dependence for binding to target, thereby avoiding the need for engineering. Recent data including our own indicate that, in general, tumor cells express very low or undetectable levels of the recycling receptor FcRn (46, 47). This loss of expression is critical for efficacy of Ca²⁺-switched ADCs/PDCs, as it avoids the possibility of FcRn-mediated salvage of the drug conjugate away from lysosomal delivery following endosomal dissociation from target.

To summarize, we have developed a novel PDC that dissociates from target in early endosomes through Ca²⁺ switching, enabling lysosomal delivery of the drug regardless of ligand (phosphatidylserine) trafficking. Our study also reveals a complex relationship between in vivo efficacy of the phosphatidylserine-targeting agent and in vitro binding properties, with the lowest affinity PDC, Fc-Syt1_MMAE, showing superior pharmacokinetic properties and potent antitumor effects. Given that the subcellular trafficking behavior of many tumor surface markers following endosomal delivery remains poorly characterized, and most solid tumor antigens are not exclusively tumor specific, our study indicates that approaches involving Ca²⁺ switching and affinity/avidity tuning may have value for the design of effective ADC or PDC platforms.

No potential conflicts of interest were disclosed.

Conception and design: R. Li, L. Li, R.P. Mason, R.J. Ober, E.S. Ward

Development of methodology: R. Li, S. Chiguru, L. Li, R.P. Mason, E.S. Ward

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Li, S. Chiguru, L. Li, S.C. Devanaboyina, H. Tian, A. Schroit, R.P. Mason, E.S. Ward

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Li, S. Chiguru, D. Kim, R. Velmurugan, D. Kim, S.C. Devanaboyina, A. Schroit, R.P. Mason, R.J. Ober, E.S. Ward

Writing, review, and/or revision of the manuscript: R. Li, A. Schroit, R.P. Mason, R.J. Ober, E.S. Ward

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Li, S. Chiguru, E.S. Ward

Study supervision: A. Schroit, R.J. Ober, E.S. Ward

This study was supported in part by grants from the Cancer Prevention and Research Institute of Texas (CPRIT; RP 110441, awarded to E.S. Ward, R.J. Ober, A. Schroit, and R. Mason) and the NIH (Simmons Comprehensive Cancer Center support grant; 5P30CA142543, awarded to M. Cobb). The IVIS Spectrum was purchased with support (1S10RR024757, awarded to R.P. Mason) from the NIH.

This study is dedicated to the memory of our colleague, Philip Thorpe. We thank Alex Winters and Jeni Gerberich for assistance with the whole-body imaging experiments, Jacob Newman for assistance with the construction of expression plasmids for the phosphatidylserine-specific Fc fusions, and Drs. Dilip Challa and Priyanka Khare for providing recombinant mouse FcRn.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.