Development of betabodies: The next generation of phosphatidylserine targeting agents

Externalized phosphatidylserine (PS) is a phospholipid and a selective marker of the tumor microenvironment (TME). It is exposed on the outer leaflet of the plasma membrane of tumor-associated endothelial cells, apoptotic tumor cells, and some viable tumor cells, where it functions in part to suppress immune responses by binding to PS receptors expressed on tumor-infiltrating myeloid cells. PS has been targeted with antibodies, such as bavituximab, that bind the phospholipid via a cofactor, β2-glycoprotein 1 (β2GP1); these antibodies showed excellent specificity for tumor vasculature and induce an immune stimulatory environment. We have advanced this concept by developing the next generation of PS targeting agent, a fusion protein (betabody) constructed by linking PS-binding domain V of β2GP1 to the Fc of an IgG2a. Betabodies bind to externalized PS with high affinity (∼1 nM), without the requirement of a co-factor and localize robustly to the TME. We demonstrate that betabodies are a direct PS-targeting agent that has the potential to be used as anti-tumor therapy, drug delivery vehicles, and tools for imaging the TME.

Externalized phosphatidylserine (PS) is a phospholipid and a selective marker of the tumor microenvironment (TME).It is exposed on the outer leaflet of the plasma membrane of tumorassociated endothelial cells, apoptotic tumor cells, and some viable tumor cells, where it functions in part to suppress immune responses by binding to PS receptors expressed on tumor-infiltrating myeloid cells.PS has been targeted with antibodies, such as bavituximab, that bind the phospholipid via a cofactor, b2-glycoprotein 1 (b2GP1); these antibodies showed excellent specificity for tumor vasculature and induce an immune stimulatory environment.We have advanced this concept by developing the next generation of PS targeting agent, a fusion protein (betabody) constructed by linking PSbinding domain V of b2GP1 to the Fc of an IgG2a.Betabodies bind to externalized PS with high affinity (1 nM), without the requirement of a co-factor and localize robustly to the TME.We demonstrate that betabodies are a direct PStargeting agent that has the potential to be used as antitumor therapy, drug delivery vehicles, and tools for imaging the TME.
Phosphatidylserine (PS) has been studied in pathological contexts and has emerged as a viable candidate for enhancing anti-tumor immune responses (12).PS is a negatively charged phospholipid that is part of the structure of the lipid bilayer in all cell membranes.It is sequestered to the inner leaflet of the plasma membrane in normal cells unless they become apoptotic, at which point PS is translocated to the outer leaflet to engage with PS receptors on phagocytic cells (13,14).PS engagement with PS receptors initiates efferocytosis, the clearance of apoptotic cells, and subsequent local immune suppression (15,16).Cells in the TME have been shown to exploit externalized PS signaling to inhibit the innate and adaptive immune response; environmentally stressed tumor cells, apoptotic and necrotic cells, as well as healthy, activated endothelial cells externalize PS in the TME (17,18).Macrophages, in particular, express PS receptors and, after binding PS, polarize into an anti-inflammatory phenotype and secrete cytokines that prevent dendritic cell maturation, antigen presentation, and subsequent T-cell priming and activation (12,(19)(20)(21).
PS-binding antibodies have been developed to target externalized PS in the TME (22).This suite of antibodies localizes to the TME and, in particular, tumor vasculature.Multiple studies have demonstrated that PS-targeting antibodies enhance the efficacy of standard therapy and immune therapy for cancer (20,(23)(24)(25)(26)(27).Bavituximab, a chimeric PStargeting antibody is currently in clinical testing in multiple indications [NCT03519997, NCT04150900, NCT03139916].However, these PS-targeting antibodies require the serum protein b2-Glycoprotein-1 (b2GP1) and must dimerize two units of b2GP1 to mediate binding to PS (24).Additionally, this dependency impacts the size of the therapeutic structure and its ability to permeate the tumor beyond the vasculature.
Here we report developing the next generation of PS targeting agent, a recombinant fusion protein composed of the PS-binding domain of b2GP1 fused to an IgG2a Fc.This unique fusion protein, significantly smaller in size compared to PS-binding antibodies, binds PS directly with high affinity and specifically localizes to the TME.

b2GP1 null C57Bl/6 mice demonstrate the function of b2GP1 in targeting PS
A drawback of the PS-targeting antibodies is that they do not bind PS directly.PS binding requires two units of a cofactor, b2GP1 (23) (Fig. 1A).We developed b2GP1 (Apoh) deficient mice on a C57Bl/6 background to demonstrate the

Development of novel PS-targeting agents, betabodies
essential relationship between PS-targeting antibodies and b2GP1 in vivo.Mice were created as described in the Experimental Procedures resulting in the complete deletion of the APOH gene via insertion of the LacZ cassette (Fig. 1B).Homozygous Apoh knockout mice were verified by genotyping for the WT gene and LacZ cassette; they were further characterized via qPCR of RNA extracted from mouse livers and Western blot for b2GP1 protein in liver lysates (Fig. 1C).
Dimerized b2GP1 is an essential cofactor for PS-targeting antibodies to bind to externalized PS and localization to the tumor microenvironment To demonstrate the requirement of b2GP1 for binding to PS, we utilized Apoh −/− mice.Apoh −/− mice and WT C57Bl/6 mice bearing established subcutaneous mouse pancreatic tumors (KPfC) were injected i.v. with 1N11 alone or 1N11 preincubated with purified b2GP1.After 2 h, mice were euthanized and perfused, and tumors were resected and snapfrozen.Frozen tumor sections were stained for 1N11 localization to the TME; sections were also stained for CD31, a marker of endothelial cells to show localization of the antibody to tumor vasculature as demonstrated prior (28).Immunofluorescence (IF) staining of each group illustrated that 1N11 localizes to tumor vasculature in a b2GP1-dependent manner, showing no localization in tumor-bearing Apoh −/− mice without prior incubation of 1N11 with a 2 to 1 M ratio of b2GP1 (Fig. 1D).

Evolution of betabodies from PS-targeting antibodies to a single domain Fc fusion protein
To develop direct PS binders, full-length b2GP1 was initially fused to an IgG Fc domain (FLB).A concern with this design was the size (150.6 kDa) and potential toxicity.Antiphospholipid antibodies, the causative agent in antiphospholipid syndrome (APS) typically bind to domain I of b2GP1 (29)(30)(31)(32)(33)(34)(35).Although a small number of studies have claimed that there are some endogenous anti-b2GP1 antibodies that target domains II-IV, pathological antibodies to b2GP1 isolated from human APS patients only interact with domain I (31).PS-targeting antibody 1N11 is able to diminish APS pathology by interfering with APS antibody-b2GP1 interaction (36).
We tested multiple variations of b2GP1 domains fused to Fc (Table 1) in vitro using a PS-binding ELISA assay and flow cytometry to show binding to cells.For flow cytometry, cells were irradiated or treated with H 2 O 2 , as indicated, to induce PS externalization.To determine which domains of b2GP1 were optimal for binding to PS, we produced constructs lacking domain I.These "domain betabodies" contained domains II-V or III-V fused to Fc.Two constructs, Fc-II-V and Fc-III-V, with and without a linker were transiently expressed in CHO cells and tested.Each construct bound PS with activity similar to 2aG4, a mouse PS-targeting antibody (18) as shown by ELISA (Fig. 2A, left panel); however, they displayed significantly worse activity than 2aG4 on irradiated NS0 cells quantified by flow cytometry (Fig. 2A, right panel).Next, we tested constructs consisting of the full-length wild-type mouse b2GP1 fused to the N-terminus (BLF) or the C-terminus (FLB) of IgG2a Fc, as well as versions with multiple linker lengths: GGGGS -BLF, FLB; (GGGGS) 3 -B3LF, F3LB; (GGGGS) 5 -B5LF, F3LB (Table 1).The C-terminal constructs bound to PS via ELISA similarly to 2aG4 and with higher affinity than the N-terminal versions (Fig. 2B, left panel).Flow cytometry analysis of irradiated cells revealed that only the Cterminal construct containing a single linker (FLB) binds to PS exposed on the cell surface similarly to 2aG4 (Fig. 2B, right panel).Subsequent studies investigated the pharmacokinetics of the FLB constructs in vivo.These constructs were cleared in approximately 30 min.Hypothesizing that the heavy glycosylation pattern of domains II and III of b2GP1 may be causing clearance in the liver (37), https://www.uniprot.org/uniprotkb/Q01339/entry#function, we performed a 10-min PK study and found that the majority of FLB had already accumulated in the liver (Fig. S1A).Subsequent deglycosylation of domains II and III of b2GP1 chemically with PNGase F or genetically (dgFLB, dgF3LB, dgF5LB) resulted in significantly diminished PS binding (Fig. S1B).In summary, we found that the C-terminal Fc-b2GP1 constructs bound PS-positive cells as efficiently as 2aG4 but were cleared quickly by the liver in vivo.
As a result of the abrupt clearance of the FLB constructs in vivo due to the glycosylation of domains II and III, we returned to the truncated domain betabodies.We produced betabodies that fused domains I and V (FL15, F3L15, F5L15) or domains I, II, and V (FL125, F3L125, F5L125) with 1, 3, or 5 linkers, respectively (Table 1).The glycosylation sites in domain II were mutated (N86Q and N98Q) to eliminate the clearance issue.All six constructs were tested for PS binding in vitro; FL15, F3L15, and F5L15 bound PS by ELISA only marginally less effectively than FLB, whereas the I, II, V constructs bound to PS significantly less in comparison (Fig. 2C, left panel).Flow cytometry analysis showed that I, V constructs bound externalized PS at the same rate as FLB, but, again, the I, II, and V constructs showed much less effective binding (Fig. 2C, right panel).Interestingly, the number of linkers did not make a significant difference functionally, though as more linkers were included, there was a slight reduction in binding.In conclusion, FL15, a C-terminal fusion containing b2GP1 domains I and V and a single linker of the Fc and b2GP1 domains proved to be just as effective in binding PS by ELISA and flow cytometry as the Fc-b2GP1 construct.

Point mutations in domain V are required to maintain stability in vivo
Because one of the residues (K308) in b2GP1 domain V is involved in binding to LDL receptor ApoER2 and these interactions have been implicated in the pathogenicity of APS (38), we introduced a point mutation K308A, K308D, and K308S and generated three variants of FL15.From this point, the non-mutated version will be referred to as KL15, where K denotes K308.FL15 mutants K308A, K308D, and Development of novel PS-targeting agents, betabodies K308S will be referred to as AL15, DL15, or SL15, respectively; all variants, including KL15, bound PS in the ELISA equally well, except for AL15 (K308A) which completely lost binding activity (Fig. S1C).By flow cytometry, KL15 bound irradiated cells better than the other variants, though SL15 (K308S) had the closest binding pattern to KL15 (Fig. S1D).We then proceeded to test the pharmacokinetics of KL15 and SL15 in vivo by injecting 125 I-labled KL15 or SL15 via the tail vein.Blood was sampled at multiple time points and 125 I counts were determined with a g-counter.KL15, at approximately 125 h, had a longer half-life than SL15 (Fig. S2A), and both lasted longer in vivo than bavituximab (48 h) (39), a chimeric PS-targeting antibody that is currently in clinical testing.KL15 was the most promising construct in vitro and in vivo to this point.Given the potential APS-like toxicity due to domain I, we also produced and characterized the smaller KL5, a construct with the same structure as KL15 but only including domain V (Table 1).
Subsequent PK studies with KL5 and KL15 identified that each betabody was losing PS binding efficacy in vivo.For example, after overnight circulation in SCID mice, the PS binding of KL15 was reduced to 65%.KL5 fared better with a PS-binding efficacy of 60 to 70% after 3 days in SCID mice (Fig. S2B).Next, purified KL15c was added to either freshly collected plasma or serum from naïve mice, incubated for indicated time periods, and analyzed for PS-binding activity by PS ELISA.Results indicate that KL15c is stable in plasma but relatively unstable in serum over time (Fig. S2C).KL5c and KL15c fragmentation/degradation was confirmed by SDS-PAGE 24 h after injecting these betabodies into SCID mice (Fig. S2D).To identify the sites of possible degradation affecting PS-binding, we repeated in vivo PK with an n-terminal betabody (KL15n); it was purified from the plasma by protein A, run on SDS-PAGE, and submitted for N-terminal sequencing to identify the site(s) of fragment loss.Sequencing revealed that betabodies were cleaved at Lys 317 -Thr 318 site in domain V (Fig. 2D).Factor XI and plasmin are known to cleave b2GP1 at this site as part of thrombus generation (40,41).This site is significant in that it is directly adjacent to the hydrophobic loop necessary for PS-binding (42).To preserve betabody integrity, we inserted a point mutation K317E which eliminated the cleavage site and preserved PS-binding.The final candidate betabodies that we moved forward with for in vivo localization assays were the C and N-terminal versions of KL5 and KL15, all containing the K317E mutation.

Betabody constructs exclusively bind phosphatidylserine in vitro
The final betabody candidates (Fig. S3) have a similar structure and include a single GGGGS linker between the Fc and b2GP1 domain(s).KL15c and KL5c include b2GP1 domain V (KL5c) or b2GP1 domain I and V (KL15c) connected to the C-terminus of mouse IgG2a Fc.KL5n and KL15n incorporate the same b2GP1 domains fused to the N-terminus of mouse IgG2a Fc (Fig. 3A).We also constructed KL4 incorporating b2GP1 domain IV with Fc as a negative control because domain IV does not bind to PS or other lipids.The five betabodies were tested for binding to PS via ELISA and compared to the PS-targeting antibody 2aG4 (a mouse IgG2a ( 26)).As predicted, KL5c and KL15c bound PS comparably to 2aG4, whereas KL5n and KL15n demonstrated diminished binding (Fig. 3B).All five betabodies were also tested for binding to phosphatidylethanolamine (PE) and phosphatidylcholine (PC) via ELISA.
a Constructs are the same.
Development of novel PS-targeting agents, betabodies highest percentage of cells, with KL5c ultimately having the best cell binding efficacy (Fig. 3D).All betabodies also showed limited or no binding to untreated 4T1 cells, demonstrating exclusive PS binding (Fig. S5).With KL5c as the leading candidate, we visually verified the binding of that betabody to H 2 O 2 -treated E0771 cells grown in 3D culture.It is important to note that, because we were staining for binding to lipids, we did not use detergents during the staining process.This resulted in some punctate background staining with the Alexa-488 conjugated secondary antibody throughout the 3D matrix.While the negative control KL4 showed no cell membrane interaction, even with H 2 O 2 treatment of the cells, KL5c robustly localized to the cell membrane after H 2 O 2 treatment, in a primarily nuclear-adjacent pattern (Fig. 3E).All betabodies were tested for PS-binding on E0771 cells (Fig. S6).

Development of novel PS-targeting agents, betabodies
KL5c selected as the final betabody with robust in vivo localization to the TME of 4T1 orthotopic breast tumors To demonstrate that KL5c localizes to externalized PS in a b2GP1-independent manner, localization in KPfC tumors grown in Apoh −/− mice was performed.Intravenous injection of an equal amount of betabody into three Apoh −/− mice showed that betabodies containing dimerized domain V localized throughout the tumors in the absence of endogenous b2GP1 (Fig. 4A) indicating that binding of the betabody to externalized PS is b2GP1-independent.
We next investigated in vivo localization of the betabodies in tumor-bearing mice.We injected 4T1 tumors orthotopically into a cohort of NSG mice and let tumors grow for 26 days to allow for metastasis to the lungs to manifest.KL5c (50 mg) or other functional betabodies were mixed with biotinylated KL4 (50 mg), a non-functional betabody, and injected intravenously Development of novel PS-targeting agents, betabodies into 3 mice per each betabody combination.KL4 served as an internal control.Betabodies were allowed to circulate for 18 h before tumors and organs were removed and snap-frozen for IHC.All tumors and normal organs were stained for functional betabody, KL4, and vasculature (CD31).KL4 failed to localize to the tumor microenvironment (Fig. 4B) or normal organs.Nterminal and C-terminal betabodies localized to the tumor microenvironment; KL5c demonstrated the most robust localization, with clear binding to the vasculature (inset) and distribution throughout the tumor (Fig. 4C).Healthy tissues stained by the same methods as the tumor sections showed no betabody localization (Fig. 4, D-F) apart from lung tissue (Figs. 4G and S10).Strong staining of betabodies in compact, highly nucleated regions of the lungs indicated that betabodies localize to lung metastases (Fig. 4G) and the primary tumor (Fig. 4C).Identification of the highly nucleated areas as 4T1 metastases was confirmed with IHC staining of a known 4T1 marker, Gpa33 (43) (Fig. S7).All four betabodies demonstrated efficacy in vivo, but KL5c illustrated its superiority based on more widespread dispersion throughout the tumor, quantified by ImageJ and analyzed for significance using a 2Way ANOVA (Figs. 4H and S8).Additionally, KL5c was tested in competition with the well-characterized PS-binder Annexin V and was able to effectively compete for PS binding (Fig. S11).Therefore, KL5c is the final betabody construct that will be used in future experiments.

Discussion
b2GP1 has been studied in the context of APS, thrombosis, and pregnancy loss (44)(45)(46).However, outside of its contribution to PS-targeting antibodies, b2GP1 has not been utilized as a tool to target the TME.Though we have shown previously in vitro that b2GP1 is a required cofactor for the interaction of PS-targeting antibodies with PS ( 24), here we demonstrate that it is unequivocally required for the localization of PS-targeting antibodies (e.g., bavituximab) to externalized PS in vivo.Utilizing this relationship, we designed a construct that binds PS directly and contains an Fc domain.Multiple permutations of b2GP1 domains and Fc orientations were tested for efficacy, however the simplest design was the most effective.The combination of Fc fused to domain V, the sole PS-binding domain of b2GP1, with a single linker is the superior iteration of betabody for PS localization to the TME.
We developed the Apoh −/− mouse as a negative control to demonstrate the indispensable relationship between PS and b2GP1 for PS-targeting.Because domain V is the only domain of b2GP1 that binds to PS, an alternative strategy could have been to produce a mouse that only expressed b2GP1 domains I-IV.While, this strategy would definitively isolate the PS-b2GP1 relationship to domain V in vivo, the requirement of domain V for PS interaction has already been demonstrated (47).Eliminating b2GP1 in its entirety was an efficient strategy given the mice were available from UC Davis KOMP Repository Knockout Mouse Project.In addition, b2GP1 domains are associated with interaction with other proteins.For example, sequences of residues within the primary structure of domain III of b2GP1 mimic peptides from infectious bacteria and viruses; these peptides can give rise to antibodies associated with an APS phenotype (48).In addition, domain V serves to hold b2GP1 in an inactive "Q" confirmation in circulation through interaction with domain I, masking cryptic epitopes on the domain I that are recognized by APS antibodies (32).
Eliminating domain V alone might make endogenous b2GP1 more immunogenic.Our development of KL4, utilizing non-PS-binding domain IV, demonstrates specific binding of domain V of b2GP1 to externalized PS.
Though encouraging that in vivo experiments have demonstrated that betabodies function as anticipated in tumor-bearing animals, the question about the specificity of these constructs for the TME remains to be answered.PS is exposed regularly in multiple healthy tissues.For example, the epithelium of the gastrointestinal tract, cells of the thymus and bone marrow, and germ cells in the testes undergo apoptosis regularly ( 49), yet we have not detected betabody localization to these tissues in vivo.Additionally, there are other cells that are non-apoptotic and not associated with tumors that externalize PS as a signaling mechanism, including differentiating monocytes and certain populations of T cells (50,51).These cells might be bound and affected by betabodies (52).However, in our experiments, betabodies do not appear to be accumulating in these tissues or being cleared by immune cells.Instead, betabodies localize strongly to the TME.Similarly, duramycin, a highly specific phosphatidylethanolamine (PE)-binding peptide, has been used for imaging of the tumor vasculature; PE is also externalized on apoptotic cells, and duramycin localizes specifically to the TME (53).The reason for this is not yet clear.
An additional concern is that the circulation of perpetually dimerized domains of b2GP1 may interfere with the function of endogenous b2GP1.While b2GP1 must have a biological function as it is known to be maintained at a high concentration, approximately 200 mg/ml, in serum and is highly conserved throughout the mammalian class (30), the specifics as to what the function of b2GP1 is have not been definitively identified.It is well known that b2GP1 is the primary antigen for pathogenic antibodies in APS (45,46); however, the mechanisms by which b2GP1 is involved in the pathology are unclear.Indeed, humans and mice that lack b2GP1 are viable and healthy (54).
One of the advantages of betabodies over PS-targeting antibodies is their size, but the inclusion of the Fc domain increases the size.The Fc domain is highly useful as a tag for purification and for attachment of labels for imaging purposes; however, generation of a cleavable Fc domain and its removal would permit more rapid and efficient penetration into solid tumors and allow for more rapid clearance of excess betabody from circulation.This could be a favorable characteristic if betabodies had cytotoxic drugs conjugated to them in the manner of an antibody drug conjugate (ADC) or if betabodies were used to image the TME.Indeed, duramycin-conjugates, which are cleared rapidly (within 4 min rats), have been used to non-invasively image the TME (53,55).
Several labs, including our own, have developed PStargeting agents that have shown efficacy in preclinical models and in human patients (25-27, 52, 56-58).For example, Bavituximab has advanced into clinical trials (NCT03519997, NCT04150900, NCT03139916).Some of these constructs exhibit internalization which could contribute to off-target effects, especially in immune cells that externalize PS, though neither PS-targeting antibodies nor betabodies have been shown to internalize when bound to PS.The mechanism of action of PS-targeting antibodies appears to be driven on the outside of tumor cells (20).Beyond just PStargeting antibodies, other PS-specific agents have shown efficacy against multiple tumor types.SapC-DOPS nanovesicles have been developed that directly target tumor cells by binding to externalized phosphatidylserine and inducing caspasemediated apoptosis (59)(60)(61).This approach bypasses the immune response, targeting tumor cells directly.Another group has also developed a fusion protein targeting PS, Fc-Syt1, by which they deliver a cytotoxic drug through internalization of the construct (52).We have not seen evidence of internalization of PS-targeting antibodies or betabodies by tumor cells, and, in fact, the bivalent Fc-Syt1 protein, analogous to betabodies, exhibited limited internalization compared to their tetravalent agent (52).
We envision the use of betabodies such as KL5c as a delivery agent for therapeutic payloads to the TME, where payloads are released from the betabody through extracellular proteases (62).This would allow for an additional layer of specificity for drug targeting.This is the next step for betabodies, the next generation of PS-targeting agents.

Betabody production and purification
Betabody were produced and purified as previously described (63).Briefly, plasmids for each construct were transiently transfected into Expi293F suspension cells (Gibco A14528).After 5 days, the supernatant was harvested by ultracentrifugation and filtration.Supernatants were diluted by 50% with binding buffer (50 mM Tris, 50 mM NaCl, pH 8.2) and run over a protein A column (Prosep-vA High-Capacity Affinity Chromatography Media, Millipore C175805).Betabodies were eluted using acetate buffer (100 mM sodium acetate, pH 2.8) and neutralized with 10% volume of 1 M Tris pH 9.0.Betabodies were purified using Cytiva AKTA pure chromatography system.
Generation of b2GP1 null C57Bl/6 mice Two heterozygous/hemizygous embryonic stem cell clones derived from a C57Bl/6NTac background were procured from the UC Davis KOMP Repository Knockout Mouse Project and microinjected into blastocyst embryos.Both clones contained a ZEN-UB1 lacZ reporter/neo selection cassette (Apoh tm1(- KOMP)Vlcg ) inserted into the Apoh locus on chromosome 11 via homologous recombination between location 108,395,347 and location 108,414,365, resulting in a complete deletion of the gene.Heterozygous knockouts of the Apoh gene were backcrossed with C57Bl/6 mice for six generations, before being crossed with Cre + C57Bl/6 mice (B6.C-Tg(CMV-cre)1Cgn/J; IMSR_JAX:006054) to remove the Neo cassette (Fig. 1B).Apoh +/− Cre + offspring were backcrossed with WT C57Bl/6 mice for a final heterozygous N7 generation lacking Cre.The genetic background of the N7 offspring was tested by the DartMouse Lab at the Dartmouth Geisel School of Medicine and confirmed to be 99% C57Bl/6 background.N7 Apoh +/− Cre -mice were crossed to produce the first generation of Apoh −/− mice.Homozygous Apoh knockout mice were verified by genotyping for the WT gene and LacZ cassette using primers 26,27 and 24,25 respectively (Table S1), RNA and protein fractions extracted from mouse livers were tested by qPCR using primers 21,22 (Table S1), and Western blot (Proteintech 11892-1-AP) for b2GP1 RNA and protein expression, respectively.Apoh−/− mice were generated as described in the Results (Fig. 1B).All animals were housed in a pathogen-free facility with access to food and water ad libitum.Experiments were performed under protocols approved by the institutional animal care and use committee at the University of Texas Southwestern Medical Center in Dallas.

Betabody in vivo localization
Four groups of Nod.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice (n = 3) were injected with 100,000 4T1 cells (100 ml PBS) into the fourth mammary fat pad.Primary tumors were allowed to grow for 3 weeks to allow for metastasis to the lungs.Each group was treated with a combination of 50 mg of one of the functional betabodies (KL5c, KL5n, KL15c, or KL15n) and 50 mg of a biotinylated negative control betabody (KL4) via tail vein injection.Betabodies were allowed to circulate for 18 h; mice were then humanely euthanized and perfused with PBS before resection of tumors and organs.
Resected tissues were bisected, with half of each preserved in formalin and half snap frozen in liquid nitrogen for betabody localization staining.IHC staining was performed using the same methods as for the 1N11 localization except that: functional betabodies were identified with anti-mouse IgG HRP followed by Opal 570 (red) and KL4-biotin was identified by HRP-streptavidin followed by Opal 570 on a separate tissue section.Vasculature was stained with anti-CD31 followed by anti-goat HRP and Opal 520 (green).Nuclei were counterstained with Hoescht (blue).

Anti-PS ELISA
Phosphatidylserine (Avanti Polar Lipids, Inc. 840032C-25 mg) was dissolved in n-hexane to a concentration of 10 mg/ ml.50 ml of the solution was added to each well of a 96-well Immulon 1B microtiter plate (#4466966), and the solvent was allowed to evaporate.Plates were blocked for 1 h at room temperature with 5% nonfat milk in PBS (w/v).Betabodies (100 nM) were subjected to serial 2-fold dilution in the blocking buffer (100 ml per well), and plates were then incubated for 1 h at 37 C. Betabodies were detected with HRPconjugated anti-mouse (Fcg specific) (1:3000) for 1 h at 37 C. Plates were washed four times with 200 ml of PBS per well after betabody and secondary antibody incubation.2aG4, a mouse PS-targeting antibody was used as a positive control (18).Secondary reagents were detected by chromogenic substrate TMB and stopped with equal volume of 10% HCL.
Plates were read at 450 nm using a microplate reader.Negative controls included incubation of non-PS binding betabody KL4 and incubation of betabodies in uncoated wells.

Flow cytometry
4T1 cells were treated with 4 mM H 2 O 2 in serum-free DMEM for 30 min.Cells were washed with cold PBS, trypsinized, and washed three times with flow buffer (Dulbecco's Phosphate Buffered Saline + 0.5% BSA).Cells were fixed with 2% PFA and then betabody was added directly to the cell suspension, final concentration of 10 mg/ml, for 1 h on ice.
After 1 h, cells were spun down and washed three times with buffer.Cells were suspended in 500 ml of anti-mouse Alexa Fluor-488 (1:200) and incubated on ice in the dark for 30 minutes.Controls included unstained cells and AnnexinV-FITC (positive control, 1:100).Cells were spun down and washed three times with buffer and analyzed on a FACSCaliber machine in the UTSW Flow Cytometry Facility.

Figure 1 .
Figure1.b2-Glycoprotein 1 is an essential cofactor for binding of PS-targeting antibodies to externalized PS.A, schematic of PS-targeting antibody binding to PS via b2GP1.PS-targeting antibodies (e.g., 1N11, 2aG4) require b2-Glycoprotein-1 (b2GP1) to bind to externalized phosphatidylserine (PS)(24).PS-targeting antibodies bind b2GP1 resulting in dimerizing of b2GP1 which binds to PS via domain V. Dimerization of b2GP1 increases its affinity for PS from 1 mM to 1 nM.B, schematic for deletion of the Apoh gene.C, Apoh −/− mice were verified by RT-PCR and Western blotting.Litters from an N7 Apoh +/− x N7 Apoh +/− cross were genotyped using primer sequences provided by the KOMP repository for the Apoh WT gene and the lacZ cassette, indicating gene loss.Pups positive for lacZ and negative for Apoh WT were retained as Apoh −/− and bred.The genotyping was further verified by RT-PCR and Western blot of liver lysates from a C57Bl/6 WT mouse, an Apoh +/− heterozygous mouse, and an Apoh −/− mouse.D, in vivo localization of PS targeting antibodies.Human PS-targeting antibody, 1N11 (50 mg), was injected i.v.into KPfC tumor bearing mice, with three groups: 1N11 into Apoh −/− mice, 1N11 preincubated with b2GP1into Apoh −/− mice, and 1N11 into C57Bl/6 WT mice (n = 3).After 2 h of circulation, tumors were resected and snap frozen for IF staining of 1N11 localization.
KL5 and KL15 constructs bound PS exclusively (Figs.3C and S4).Betabody binding to externalized PS on cells was tested by flow cytometry of H 2 O 2 -treated 4T1 cells.The results mirrored the PS ELISA.While no betabody matched 2aG4, again the c-terminal constructs bound the

Figure 2 .
Figure 2. Betabodies evolved from full-length b2GP1 fusion proteins to small single-domain constructs.A-C, betabodies containing only partial b2GP1 sequences, Fc-II-V and Fc-III-V (A), full-length b2GP1 fused to either N or C terminus of the Fc (B), domains I and V or I, II, and V contain 1, 3, or 5 linkers (C) and C44-B2V (IgG2a isotype control antibody linked to domain V) and 2aG4 Ab were tested for binding to PS by ELISA and flow cytometry on irradiated NS0 cells (A) or Daudi cells (B and C).Constructs were evaluated for PS binding on cells by comparing mean fluorescence intensity (MFI).D, identification of the betabody cleavage site by N-terminal sequencing was performed on Domain I/V betabodies recovered from in vivo circulation in SCID mice.

Figure 3 .
Figure 3. Betabodies exclusively bind phosphatidylserine in vitro.A, schematic of final betabody candidates.The betabodies were designed by fusing functional PS binding domain of b2GP1 (domain V) to the GGGGS linker and mouse IgG2a Fc domain.KL5c and KL5n contain b2GP1 domain V fused to the C terminus or N terminus of Fc, respectively.KL15c and KL15n contain b2GP1 domain I and V fused to the C terminus or N terminus of Fc, respectively.KL4 mimics KL5c in structure, except it fuses b2GP1 domain IV to Fc. B, binding of betabodies to PS via ELISA.2aG4, a PS-targeting antibody, was used as a positive control; KL4 is used as a negative control.C, binding to PS, phosphatidylethanolamine (PE), or phosphatidylcholine (PC) via ELISA.KL15c and KL4 results presented.D, 4T1 cells were treated with 4 mM H 2 O 2 for 30 min, fixed with 2% paraformaldehyde and then stained with 10 mg/ml of each of the betabodies.Cells were then stained with goat anti-mouse IgG -Alexa Fluor 488 secondary antibody (Invitrogen A-11001; 1:200) and subjected to flow cytometry.E, E0771 cells were grown in 3D, treated with 4 mM H 2 O 2 for 30 min, harvested with trypsin and embedded in 70% collagen with continued treatment with 200 mM H 2 O 2 in low serum overnight.Cells were fixed with 2% PFA for 1 h and then incubated with 10 mg/ml betabody in PBS.Cells were then stained with secondary antibodies: [Betabody (Alexa Fluor-488); Phalloidin (Alexa Fluor-564); Nuclei (Hoescht)].

Figure 4 .
Figure 4. Betabodies localize to the tumor microenvironment in vivo.A, in vivo localization in b2GP1-null mice.Tumor-bearing Apoh −/− mice (n = 3) were injected i.v. with KL5c (50 mg) that were allowed to circulate for 2 h.Tumors were resected, snap-frozen, and used for IHC to detect KL5c and CD31.Hoescht is used as a counterstain.B-G, In vivo localization studies.NSG mice bearing 4T1 tumors were injected i.v. with biotinylated KL4 (50 mg) plus the indicated unlabeled betabody (50 mg).Mice were harvested 18 h post-injection, perfused, and organs including tumors collected, snap-frozen, and sectioned.Tumor (b, c), brain (d), kidney (e), liver (f) and lung (g) were stained for KL4 (streptavidin-HRP and Opal 570, red) and the betabodies (anti-mouse-HRP and Opal 570, red), blood vessels (CD31, anti-goat-HRP, Opal 520, green) and nuclei (Hoescht, blue).Representative images taken at 20× are shown.H, 4× images of all tumors and organ tissues sectioned and stained were analyzed for individual betabody localization using ImageJ.Dots represent biological replicates.Individual betabody signal was compared to the other three using a 2way ANOVA analysis.