Collaborative enhancement of antibody binding to distinct PECAM-1 epitopes modulates endothelial targeting.

Antibodies to platelet endothelial cell adhesion molecule-1 (PECAM-1) facilitate targeted drug delivery to endothelial cells by “vascular immunotargeting.” To define the targeting quantitatively, we investigated the endothelial binding of monoclonal antibodies (mAbs) to extracellular epitopes of PECAM-1. Surprisingly, we have found in human and mouse cell culture models that the endothelial binding of PECAM-directed mAbs and scFv therapeutic fusion protein is increased by co-administration of a paired mAb directed to an adjacent, yet distinct PECAM-1 epitope. This results in significant enhancement of functional activity of a PECAM-1-targeted scFv-thrombomodulin fusion protein generating therapeutic activated Protein C. The “collaborative enhancement” of mAb binding is affirmed in vivo, as manifested by enhanced pulmonary accumulation of intravenously administered radiolabeled PECAM-1 mAb when co-injected with an unlabeled paired mAb in mice. This is the first demonstration of a positive modulatory effect of endothelial binding and vascular immunotargeting provided by the simultaneous binding a paired mAb to adjacent distinct epitopes. The “collaborative enhancement” phenomenon provides a novel paradigm for optimizing the endothelial-targeted delivery of therapeutic agents.


Introduction
Drug targeting to endothelial cells (ECs) (i.e., ''vascular immunotargeting'') has the potential to improve management of diseases involving ischemia, inflammation, thrombosis, and tumor growth [1][2][3][4][5]. In particular, conjugation of therapeutics with antibodies to PECAM-1 (platelet endothelial cell adhesion molecule 1, CD31) enables their endothelial delivery, boosting specificity and efficacy of their action in animal models [3,6]. Further optimization of this promising approach is warranted to support translation into the clinical domain.
Monoclonal antibodies (mAbs) directed to different extracellular epitopes and domains of PECAM-1 have been used as probes to study the role of PECAM-1 in mediating homophilic and heterophilic binding interactions [9,10,[15][16][17][18], as well as affinity ligands for endothelial targeting of drugs, and nanocarriers [3,[19][20][21]. Antibodies directed to distinct PECAM-1 epitopes have different functional effects, either inhibiting, augmenting, or having no effect on the IgD1/IgD2-mediated homophilic binding interactions of PECAM-1 [17,22]. Further, the engagement of specific PECAM-1 epitopes controls the rate of endothelial internalization and intracellular trafficking of nanocarriers targeted by PECAM-1 mAbs [23]. These results suggest that optimization of immunotargeting and intracellular delivery is possible through the engagement of distinct PECAM-1 epitopes.
In the present study we set out to investigate the in vitro and in vivo binding parameters of mAbs directed to the IgD1 and IgD2 domains of PECAM-1 and address mutual effects of their binding. The latter aspect is a relatively uncharted one in vascular immunotargeting. Studies in this area are limited to mAbs to angiotensin-converting enzyme (ACE), a promising molecular target for drug delivery to endothelium [24,25], and show that anti-ACE mAbs directed to distinct epitopes negatively mutually interfere with binding of each other [26].
However, in contrast with this somewhat expected outcome with anti-ACE mAbs, our results indicate that endothelial immunotargeting of anti-PECAM-1 mAb can be significantly enhanced by the simultaneous binding of paired mAbs directed to adjacent, yet distinct PECAM-1 epitopes in both in vitro cell culture and in vivo mouse studies. Motivated by this hugely unusual outcome, we set out to determine whether augmentation in binding translates to an increase in therapeutic protein delivery and functional output. We used a therapeutic fusion protein targeted to PECAM-1 to demonstrate that enhanced delivery results in a significant increase in the fusion-catalyzed generation of a cell-protective species with antithrombotic and anti-inflammatory activities. This antibody-dependent ''collaborative enhancement'' phenomenon illustrates the potential of this targeting strategy for increasing the efficiency of vascular delivery in therapeutic applications.
Live-cell radioimmunoassay (RIA) of 125 I-labeled mAbs ([ 125 I]-mAb) was used for quantitative assessment of equilibrium binding parameters (K d ), including the number of maximum available binding sites (B max

Modulation of in vitro PECAM-1 targeting
We next investigated the mutual binding effects of mAb 37 and 62 to their epitopes in IgD1 of huPECAM-1. Expectedly, endothelial binding of [ 125 I]-mAb 62 and [ 125 I]-mAb 37 was competitively inhibited by their respective unlabeled mAb counterparts directed to the same epitope (''self-paired'') ( Figure 4A; Figure S3). However, binding of [ 125 I]-mAb 62 was enhanced 1.52fold by unlabeled mAb 37 (''paired'') ( Figure 4A,B). This enhancement effect was not mutual, as unlabeled mAb 62 did not alter the binding of [ 125 I]-mAb 37 ( Figure S3). [ 125 I]-mAbs 62 and 37 bind to immobilized huPECAM-1, but not to mAb pairs or control IgG ( Figure S4). This result confirms that modulation of anti-PECAM mAb binding to endothelial cells is due to binding through cellular PECAM-1 and not due to binding to cell-associated antibodies.
RIA  Table S2). Taken together, these data suggest that the To test whether the collaborative binding phenomenon is unique to human PECAM-1, we investigated mAb modulatory effects on muPECAM-1-expressing cells. Binding of [ 125 I]-mAb 390 and [ 125 I]-mAb MEC13.3 to REN-muP cells expressing recombinant muPECAM-1 was inhibited by its unlabeled selfpaired mAb, yet enhanced by paired mAb directed to a distinct muPECAM-1 epitope ( Figure 4C). These results were recapitulated in murine MS1 endothelial cells expressing native muPE-CAM-1 ( Figure S5). Interestingly, the most dramatic collabora-

Collaborative enhancement increases targeting and effect of a therapeutic fusion protein
Collaborative enhancement of anti-PECAM mAbs binding was validated using a novel protein therapeutic prodrug, i.e., the extracellular domain of mouse thrombomodulin (TM) fused to a single-chain variable fragment (scFv) targeted to the 390 epitope of muPECAM-1 (390 scFv-TM [31]). Live-cell ELISA demonstrated that paired mAb MEC13.3 increased the apparent binding affinity of 390 scFv-TM ,42fold relative to fusion alone (IC 50 0.91 nM vs. 3.49 nM) ( Figure 5A). Self-pairing the epitope with maternal mAb 390 inhibited 390 scFv-TM binding close to control levels with REN cells. This increase in binding affinity is accompanied by an increase in 390 scFv-TM bound to muPECAM-1, as made apparent by a higher maximum OD 490 value compared to 390 scFv-TM alone.
We further examined whether enhanced delivery of 390 scFv-TM may have therapeutic consequences. TM captures the serineprotease thrombin and modulates its pro-thrombotic activity to convert protein C to activated protein C (APC), which itself has cell-protective anti-thrombotic and anti-inflammatory effects [32]. Targeting of the TM fusion protein to the luminal endothelial surface helps to control coagulation and inflammation in animal models of acute lung injury and ischemia/reperfusion via APCmediated pathways [3,31]. 390 scFv-TM bound to REN-muP cells, which have no endogenous TM, generates APC from protein C zymogen in the presence of thrombin. We found that REN-muP cells co-incubated with 390 scFv-TM and MEC13.3 demonstrated a ,62fold increase in APC generation relative to 390 scFv-TM alone ( Figure 5B). Moreover, pairing of mAb MEC13.3 with 390 scFv-TM seemed to shift the potency of the prodrug (based on APC generation levels) to lower concentrations of 390 scFv-TM. These observations closely parallel the ELISA results and indicates an increase in both binding affinity and absolute fusion protein bound.
Co-immunoprecipitation (co-IP) studies revealed formation of a tri-molecular complex between 390 scFv-TM, PECAM-1 and MEC13.3 mAb ( Figure 5C, lane 8). The simultaneous binding of the antibody ligands to adjacent non-overlapping epitopes of PECAM-1 suggests that the increased binding and functional effect of the fusion protein are mediated through modulation of its interaction with PECAM-1 by the enhancing antibody.

In vivo PECAM-1 targeting
In vitro studies suggest that mAb-mediated modulation of endothelial binding may have important implications for the vascular immunotargeting using PECAM-1 antibodies. To evaluate collaborative enhancement of immunotargeting in vivo and recapitulate cell culture findings, we studied effects of nonlabeled mAbs on the pulmonary uptake of [ 125 I]-mAb 390 and [ 125 I]-mAb MEC13.3 injected in mice ( Figure 6). The pulmonary vasculature, due to the privileged perfusion and extended endothelial surface area [33], is the preferential target of mAbs directed to PECAM-1 [3,6]. Pulmonary targeting of [ 125 I]-mAb 390 and [ 125 I]-mAb MEC13.3 alone was reconfirmed and determined to be 67% ID/g and 41% ID/g, respectively ( Figure 6A,B). Subsequently, [ 125 I]-mAbs were co-administered with self-paired or paired mAb, and the in vivo results recapitulated cell culture findings. The pulmonary uptake of [ 125 I]-mAbs was inhibited by co-injection of non-labeled self-paired mAb down to levels observed with control [ 125 I]-IgG. Co-administration of paired mAb led to 2.12fold and 1.92fold enhancement in the pulmonary uptake of [ 125 I]-mAb 390 and [ 125 I]-mAb MEC13.3, respectively ( Figure 6C). Correcting pulmonary uptake levels for residual blood activity yields a more accurate reflection of collaborative enhancement due to active vascular immunotargeting of anti-PECAM-1 mAb. As compared to [ 125 I]-mAbs alone ( Figure 6B), the lung:blood localization ratio for both muPE-CAM-1 mAb pairs is enhanced 3.42fold over mAb alone ( Figure 6D).

Discussion
The binding of ligands, including antibodies to epitopes of target molecules can block the delivery of ligands directed to the same epitope, or potentially modulate (i.e., block or enhance) the binding of ligands directed to secondary epitopes. Herein, we examined the interaction of a panel of four monoclonal antibodies (mAbs) directed to distinct extracellular epitopes of PECAM-1 domains IgD1 (human) and IgD2 (murine) (Figure 1) for understanding and optimizing endothelial immunotargeting. PECAM-1 mAb binding exhibits properties characteristic of mAb-antigen interactions: high affinity and specificity contributed by the steric complementarity between the antibody and antigen surface (Figures 2, 3). Interestingly, for the mAbs evaluated it was clear that not all epitopes are displayed on PECAM-1 equally. In this study, we found that the mAb with higher affinity was accompanied by lower epitope accessibility, as reflected by B max ( Figure 3B). Variable accessibility to different antibodies could result from differences in: (1) masking of an epitope (e.g., due to tertiary structure of Ig-like domain, or masking by protein glycosylation and/or other components of the plasmalemma), (2) protein associations (e.g., different cell surface distribution and/or cytoskeletal associations), (3) membrane turnover of PECAM-1 sub-populations, or (4) Ab-induced shedding of PECAM-1 resulting in diminished epitope expression. However, K d and B max binding parameters can serve as valuable empiric criteria in judiciously selecting the most effective ligand (i.e. high affinity and accessibility) for therapeutic vascular immunotargeting to PE-CAM-1.
It has been reported that specific mAbs to huPECAM-1 IgD1 augments IgD1-mediated trans-homophilic interactions between adjacent PECAM-1 molecules [22]. Based on these observations, it stands to reason that if the binding of one mAb to PECAM-1 can increase the binding to an adjacent PECAM-1 molecule, then it may also increase binding of a second mAb directed to a different epitope, particularly in those domains that are implicated in homophilic PECAM-1 binding. Similar types of ''enhanced binding'' phenomena, attributed to conformational changes induced in the target molecule due to protein allostery [34][35][36], have been reported with binding of multiple ligands to isolated proteins [37], cells [38] and tissue homogenates [39]. We are observing this unusual behavior for the first time with antibodies directed to an endothelial determinant, specifically PECAM-1 which has demonstrated potential for vascular targeting of therapeutics, including immunoconjugates [19,21,40], fusion proteins [3,20,31], and nanocarriers [23].
The results presented in this report show that the binding of certain mAbs to epitopes in PECAM-1 domains 1 and 2 enhances the binding of a second paired mAb to a distinct epitope in the same domain, both in vitro (Figures 4, 5, S3, S4) and in vivo ( Figure 6). However, not all mAb pairs exhibit ''collaborative enhancement'' nor to the same degree. Augmentation of [ 125 I]-mAb binding is most pronounced using the paired ''enhancer mAb'' with a higher affinity for PECAM-1 (as is the case with mAb 37 and mAb 390). This observation is likely due to the fact that lower affinity mAb have greater potential for affinity elevation, hence the more robust differences in their binding with an enhancer mAb. The innocuous effect of lower affinity mAb 62 on [ 125 I]-mAb 37 binding ( Figure S3) further suggests that a higher affinity mAb ligand drives the increase in total binding of a paired mAb to PECAM-1.
Additional studies reveal that [ 125 I]-mAb affinity to PECAM-1 also increases in the presence of an enhancer mAb. This is evidenced by the 1.52to242fold decrease in the apparent K d when [ 125 I]-mAb 62 is co-incubated with enhancer mAb 37 both in live cells and with immobilized PECAM-1 (Table S2). An increase in binding affinity is also implied in the left shift of the ELISA binding curve of the therapeutic 390 scFv-TM fusion construct targeted to the mAb 390 epitope of muPECAM-1 when modulated with mAb enhancer MEC13.3 (IC 50 = 3.49 nMR0.91 nM, P,0.001) ( Figure 5A). We hypothesized that the improved affinity combined with an enhancement in absolute 390 scFv-TM anchored to the endothelium would result in more efficient production of APC at sites of injury. Indeed, in vitro studies reveal a significant increase in APC generation of 390 scFv-TM paired with mAb MEC13.3 (,62fold, P,0.001) at much lower fusion concentrations than 390 scFv-TM alone ( Figure 5B). The clinical and translational impact of these findings in an in vivo model of lung injury is of great significance and we are currently resolving this question.
Collaborative enhancement is only realized if there exists a ternary complex comprised of the mAb-ligand, the enhancer mAb-ligand, and PECAM-1; Co-IP experiments with 390 scFv-TM demonstrate that there is a complex between 390 scFv-TM/ muPECAM-1/MEC13.3 mAb ( Figure 5C). This lends further support that enhanced mAb binding and increased production of APC is mediated directly through modulation of PECAM-1 epitope engagement.
Importantly, the collaborative enhancement of muPECAM-1 immunotargeting in vivo was confirmed when measuring the  Figure 6). The results of in vivo studies in mice highlight the difficulty in predicting unambiguously the best mAb for in vivo immunotargeting based on in vitro mAb affinity and epitope accessibility from ELISA and RIA. For instance, following normalization of pulmonary uptake for residual blood levels (localization ration, LR) there is only 1.42fold higher endothelial selectivity of [ 125 I]-mAb 390 versus [ 125 I]-mAb MEC13.3 ( Figure 6B, P,0.001). This is despite mAb 390 having ,6.42fold higher binding affinity, albeit a 22fold lower epitope accessibility relative to mAb MEC13. Our findings are consistent with a model in which an enhancer mAb binds to PECAM-1 to mediate collaborative enhancement of paired mAb binding via a single PECAM-1 molecule or through a PECAM-1-PECAM-1 homodimer. An enhancer mAb may influence intermolecular interactions between PECAM-1 molecules in the endothelial plasmalemma in many ways, including ligand-mediated disruption of homologous dimerization and oligomerization, as has been described, for example, with VEGFR [41], EGFR/HER2 receptors [42], and ACE [43,44]. It is known that mAbs 62 and 390 can inhibit formation of homophilic PECAM-1/PECAM-1 interactions [10,14,22], although it is not clear if these mAbs can actually disrupt existing PECAM-1 homodimers. In theory, the binding of anti-PECAM mAbs might illicit surface exposure of additional PECAM-1 copies via more generalized EC activation involving cytoskeletal rearrangements  [45]. The fact that EC activation by antibody-engagement of the cell adhesion molecule ICAM-1 does not enhance anti-PECAM-1 mAb binding would argue against this scenario.
The exact mechanism of antibody-mediated collaborative enhancement of PECAM-1 is worth further investigation. The fact that collaborative enhancement of mAb binding occurs in vivo implies that this phenomenon may be employed to further optimize vascular PECAM-1 immunotargeting of diverse therapeutic cargoes, from anti-thrombotic agents to nanocarriers carrying antioxidants.
The modulation of muPECAM-1-targeted 390 scFv-TM binding in the presence of self-paired parental mAb 390 or paired mAb MEC13.3 was performed by incubating REN-muP cells with a series dilution of 390 scFv-TM co-mixed with 22fold excess of muPECAM-1 IgG mAbs. Data were collected as described above, and the observed specific binding was plotted as a function of 390 scFv-TM added.
All ELISA binding data were analyzed using Prism 5.0 (GraphPad, San Diego, CA) software to determine relative binding affinity constants, as defined by IC 50 . Data were fit using equation (1) for the ''four-parameter logisitic (4PL) non-linear regression model'' most commonly used for sigmoidal curves such as ELISAs: OD(x) is the OD 490 value as a function of X, the mAb concentration [mAb].
[X] 50 is [mAb] at the inflection point of the curve when binding is half-maximal (IC 50 ). B is the Hill Slope coefficient. IC 50 values are reported as the mean 6 standard deviation (SD) of three independent experiments, with each experiment performed in triplicate. RIA. Cells were grown to confluence in 1% gelatin-coated 96strip-well microplates (Corning Life Sciences, Lowell, MA). For binding assay, monolayers of cells were incubated with increasing concentration of [ 125 I]-mAb (1.8 pM-5 nM in assay buffer) in quadruplicate at 4uC for 2 h. At the end of incubation, cells were washed five times with ice-cold assay buffer. The cell-associated radioactivity was measured by a gamma counter and was normalized to the total number of cells, as counted by a hemocytometer. Non-specific binding (NSB) was calculated by subtracting the total binding calculated from performing the binding assays in the presence of 1002fold excess of unlabeled protein or by subtracting radiolabeled ligand binding to wild-type cells. The data from the live-cell RIA experiments were analyzed by Scatchard analysis using Prism 5.0 (GraphPad) software to determine equilibrium binding constant and the number of functional binding sites.
The apparent binding affinity, K d , for specific binding was calculated using non-linear regression analysis of a one-site binding hyperbola: where B max is the maximum number of binding sites per cell at the asymptotic maximum; X is [mAb], and K d is the apparent equilibrium dissociation constant. K d and B max values represent the mean 6 SD of three or more independent experiments, and each independent experiment was performed in quadruplicate. The modulation of [ 125 I]-mAb binding in the presence of unlabeled self-paired or paired mAb was performed by incubating cells with a series dilution of nonlabeled mAb co-mixed with a fixed concentration of [ 125 I]-mAb (0.3-0.6 nM) for 2 h at 4uC to allow binding. NSB was determined in the presence of 1002fold excess of unlabeled self-paired mAb. Data were collected as described above, and the observed specific binding was plotted as a function of unlabeled mAb concentration. The IC 50 for mAb self-pairs and pairs was determined by fitting this data to a fourparameter fit (see Equation 1).
Co-IP. REN-muP cells were grown to confluence in a 6 well plate. Cells were incubated with MEC13.

Activated protein C (APC) activity assay in live-cells
Previously reported assays of APC generation on the endothelial cell surface [48,49] were modified to allow measurement of APC generation by 390 scFv-TM bound to muPECAM-1 expressing cells. REN-muP and control REN cells were grown to confluence in 1% gelatin-coated 24-well plates (BD Biosciences). Cells were washed with serum free media then incubated with specified concentrations of 390 scFv-TM (622fold excess mAb MEC13.3) for 30 min at 37uC. Cells were washed three times with assay buffer (20 mM Tris, 100 mM NaCl, 1 mM CaCl 2 , 0.1% (w/v) BSA, pH 7.5) then incubated with 1 nM bovine thrombin (Sigma-Aldrich) and 100 nM protein C (Haematologic Technologies, Essex Junction, VT) in assay buffer for 1 hour at 37uC. Aliquots were removed and APC activity was measured by adding 100 nM hirudin (to inhibit thrombin; Sigma-Aldrich) and 0.5 mM of the APC substrate S-2366 (Diapharma, West Chester, OH). All samples were run in duplicate. The rate of substrate hydrolysis was measured by monitoring the change in absorbance at 405 nm over time (mOD 405 /min) at room temperature using a Multiskan FC Microplate reader. These mOD 405 /min values were subsequently converted to nmol APC using a standard curve generated using purified APC.

Ethics Statement
Animals were cared for and handled in accordance with the Guide for the Care and Use of Laboratory Animals as adopted by the NIH, under a protocol approved by the University of Pennsylvania Institutional Animal Care and Use Committee (IACUC). The approved protocol number was 802060.
In vivo targeting to the pulmonary endothelium The pulmonary vasculature represents approximately 30% of total endothelial surface in the body and gets preferential perfusion by 50% of the total cardiac blood output [33], thus pulmonary uptake of the PECAM-1 targeted [ 125 I]-mAb, once corrected for blood activity, is reflective of specific mAb binding to endothelial cells.

Data analysis and statistics
All experiments were performed at least in triplicate with a minimum of three independent experiments. Results are expressed as mean 6 SD unless otherwise noted. Significant differences between means were determined using one-way ANOVA followed by post-hoc Bonferroni multiple comparison test, or unpaired student t-test, as appropriate. P,0.05 was considered statistically significant. All curve fitting and statistical analyses was conducted using Prism 5.0 software. Figure S1 Schematic diagram of PECAM-1 (CD31) protein domain structure and sites of molecular binding interactions. PECAM-1 is a 130 kDa type 1 transmembrane glycoprotein belonging to the Ig-like superfamily of cell adhesion molecules (CAM). It consists of six extracellular Ig C2-type domains defined by disulfide bonds (S-S), a short transmembrane spanning domain, and a long cytoplasmic tail containing two ITIM [51]. Ig-domains 1 and 2 are implicated in homophilic trans-binding interactions with endothelial PECAM-1 molecules on adjacent cells and with PECAM-1 on circulating leukocytes. Igdomains 2, 3, 5, and 6 mediate heterophilic binding interactions with other cells surface antigens (e.g. CD177 on leukocytes) [16,[52][53][54].   I]-mAb 62 to huPECAM-1 on live HUVECs or to immobilized rhuPECAM-1 is studied alone or in the presence of 50 nM mAb 37. Note that total binding was corrected for NSB using 1002fold excess unlabeled mAb 62. Co-treatment of HUVECs with [ 125 I]-mAb 62 and mAb 37 led to a 1.42fold increase in binding affinity over solo binding, whereas the binding affinity increases nearly four2fold following collaborative enhancement with immobilized rhuPECAM-1. Results were determined by three independent RIA experiments performed in quadruplicate, with data expressed as mean 6 S.D. (TIF)