Efforts to create platelet-like structures for the augmentation of hemostasis have focused solely on recapitulating aspects of platelet adhesion ¹; more complex platelet behaviors such as clot contraction ² are assumed to be inaccessible to synthetic systems. Here, we report the creation of fully synthetic platelet-like particles (PLPs) that augment clotting in vitro under physiological flow conditions and achieve wound-triggered hemostasis and decreased bleeding times in vivo in a traumatic injury model. PLPs were synthesized by combining highly deformable microgel particles with molecular-recognition motifs identified through directed evolution. In vitro and in silico analyses demonstrate that PLPs actively collapse fibrin networks, an emergent behavior that mimics in vivo clot contraction. Mechanistically, clot collapse is intimately linked to the PLP's unique deformability and affinity for fibrin fibers, as evidenced by dissipative particle dynamics simulations. Our findings should inform the future design of a broader class of dynamic, biosynthetic composite materials.

Uncontrolled bleeding is the major cause of death in civilian and battlefield traumas ^3,4, highlighting the critical need for better technologies for wound management. Current hemostasis technologies including topical sealants, exothermic zeolites, advanced dressings and recombinant clotting factors ^5,6 have demonstrated modest successes, yet all have significant drawbacks and none are as “evolved” as the natural hemostasis system. More recent efforts have focused on creation of synthetic analogs of clotting constituents, most notably platelets. The vital platelet functions ^2,7 that one would like to recapitulate include 1) binding, stabilization, and enhancement of fibrin clot formation in dynamic flow conditions, 2) clot contraction and 3) cytokine and growth factor release to stimulate wound healing. To date, all artificial platelet approaches ranging from purely synthetic to reconstituted freeze-dried harvested native platelets fail to fully recapitulate these key functions. Most approaches claiming success achieve only the binding and augmentation of clot formation through multivalent display of platelet-binding motifs or platelet-cell surface adhesion motifs on a micro/nano-sized vehicle ¹.

Such approaches are sufficient to recruit clotting components and thereby decrease clotting time, however, these studies rely upon vehicles that lack the natural platelet's ability to deform within and in response to the fibrin mesh. To more accurately mimic platelet function, we created a highly deformable platelet “body” that enables multivalent display with much greater conformational flexibility. To that end, ultra-low cross-linked (ULC) poly(N-isopropylacrylamide-co-acrylic acid) (pNIPAm-AAc) microgels (μgels) were synthesized via a nonstandard, `cross-linker free', precipitation polymerization method ⁸. This unique method leverages rare chain transfer associated chain branching events to create μgels with exceedingly low (<0.5%), core-localized, cross-linking densities resulting in particles with an unmatched deformability ⁹. Like natural platelets, ULC μgels are ~1 μm in diameter in solution. Upon initiation of clotting, platelets become activated, bind to nascent fibrin fibers, actively spread within the fibrin network, and over time engage their actin-myosin machinery to contract the clot (Figure 1A). Through the course of these events, platelets undergo significant shape changes. Our ULC μgel platelet “bodies” are capable of undergoing large degrees of deformation as suggested by AFM images of μgels spread on a glass surface with a diameter of approximately 2 μm but an approximate height of only 4 nm.

The μgels were enabled with clot specificity via evolved variable domain-like recognition motifs (sdFv) displaying high selectivity for nascent fibrin protofibrils and minimal binding to the soluble circulating precursor, fibrinogen. Fibrin specificity is a critical design feature of our platelet-like particles (PLPs), as it allows for unrivaled specific targeting to a wound site without nonspecific interactions with the soluble precursor fibrinogen or circulating platelets. This seemingly simplistic approach is critically complicated by the molecular homology between fibrin and fibrinogen molecules; previous fibrin targeting attempts are also known to interact to some degree with fibrinogen domains ^10–12. We overcame this challenge and successfully identified fibrin-specific sdFvs through phage display biopanning. We performed screens with three different phagemid libraries ¹³ in vitro against fibrin clots. Following three rounds of screening (figure S1), 96 clones from each library were tested for binding to fibrin and fibrinogen (figure S2). The four most promising clones, based on their selectivity for fibrin over fibrinogen, and a random clone were then evaluated through SPR (figure S3). The clone found to have the highest affinity for fibrin (H6), and the random, nonbinding clone (S11) were utilized for creation of PLPs and control particles, respectively. Interferometery analysis verified that H6-μgels, the so-called PLPs, maintained their fibrin-binding capabilities (figure 1C), while no binding of S11-μgels (control PLPs) to fibrin was observed.

To first investigate the ability of our PLPs to recapitulate platelet function, we tested clotting of platelet-poor plasma in relation to platelet-rich plasma in vitro. These dynamic clotting experiments were performed in an endothelialized microfluidic device that recapitulates the cellular, physical, and hemodynamic environment of microcirculation (described in detail in supplemental methods) ¹⁴. As is widely accepted, we found that normal human platelet-rich plasma under flow conditions displays robust formation of a fibrin-based clot in the presence of thrombin whereas the depletion of platelets diminishes this response. Supplementation of platelet-poor plasma with PLPs successfully rescued fibrin clot formation in response to thrombin, even in 1:1 saline-diluted platelet-poor plasma (figure 2 and videos 1–5). This triggered response was presumed to be an effect of multivalent presentation of the fibrin-specific sdFv only. Indeed, particle deformability does not appear critical to this effect since H6-particles with higher degrees of cross-linking (decreased deformability) support thrombin-induced clot formation in platelet-poor plasma (figure S4). PLPs are additionally able to augment clotting in post-surgical platelet-poor plasma from neonates undergoing cardio-pulmonary bypass to correct congenital defects (figure S5). This challenging population has an underdeveloped coagulation system and experiences additional coagulopathy induced by plasma dilution effects of the cardiopulmonary bypass machine ¹⁵. In contrast, PLPs cannot augment clotting in severe hemophilia patients in which a FVIII/FIX deficiency prevents formation of fibrin, the PLP `trigger'.

In addition to augmenting fibrin formation, PLPs spread significantly within fibrin matrices, similarly to levels seen with natural platelets (figure 3A), while non-binding S11-ULC or hard H6-polystyrene (PS) particles spread significantly less than PLPs. Surprisingly, clots formed in the presence of PLPs were found to have a significantly altered structure. These clots were highly heterogeneous, containing regions of dense fibrin observed in conjunction with PLPs (figure 3B), with the dense features becoming more pronounced over time, similar to features observed from clots formed from platelet-rich plasma, although occurring on a different time scale. Clots formed in the presence of non-binding S11-ULC μgels were found to result in a more porous network compared to control (i.e. fibrin-only) clots, but the structure remained largely homogenous, and was similar to fibrin-only clots. We next characterized clot collapse at the macroscopic level by performing standard gross clot contraction experiments routinely utilized by platelet biologists ^16,17. Clot collapse was observed within 24 – 48 hours in the presence of PLPs but not under any other conditions tested (figure 3C). No significant differences in the overall degree of clot collapse were observed when comparing platelet-poor plasma containing PLPs and native platelet-rich plasma (figure 3D). However, not surprisingly, the rate of PLP-induced collapse was slower than that observed in native platelet-containing clots. We analyzed the role of particle deformability in the observed alterations in clot structure by comparing clots formed in the presence of PLPs, 2% BIS, 4% BIS, or 7% BIS μgels, or PS particles conjugated to the fibrin-specific sdFv and found that only PLPs significantly altered clot structure at both the microscopic (Figure S6) and macroscopic level (Figure S7). These results suggest that fibrin binding coupled with high levels of particle deformability likely contribute to the significant alterations in clot structure observed. Clot contraction is a significant function of natural platelets and has been correlated to enhanced clot stability, decreased fibrinolysis and is a strong contributing factor to subsequent wound healing ¹⁸. Similarly, clots formed in the presence of PLPs were more resistant to degradation than control clots or clots formed in the presence of S11-ULC μgels (figure S7). It should be noted that the role of gross clot collapse mediated by natural platelets is not a primary contributor to initial clot formation but rather occurs after establishment of the platelet-fibrin hemostatic plug.

To gain insight into the mechanisms by which PLPs contribute to macroscopic clot collapse, we developed a computational model of a compliant polymer network interacting with soft, deformable particles (figure 3E and video 6). Non-adhesive particles have no effect on network size. However, even weak attraction between particles and network filaments led to a significant reduction of the network size. When particles attach to the network, they join neighboring filaments (figure 3G), leading to local alterations in microstructure that not only disrupt network homogeneity, but also reduce its stability, which in turn ultimately causes compete network collapse (figure 3F). Local collapse is governed by the interplay between network elasticity and particle adhesiveness; more adhesive particles cause faster network collapse (figure 3H). Furthermore, our simulations confirm that particle compliance plays a key role in network collapse. Networks seeded with softer particles show a larger degree of collapse than those with stiff particles (figure S7 and video 7). This can be related to a higher ability of soft particles to form intimate contacts with network fibers (figure 3G), thereby reducing the energy barrier for clot collapse.

Lastly, we tested PLPs' ability to recapitulate the important platelet function of augmenting clotting in vivo and homing to sites of injury by utilizing a well-established rat femoral vessel traumatic injury model ^19–21. Experimental groups or vehicle were injected intravenously and allowed to circulate for five minutes prior to induction of injury to the femoral vein. Bleeding time following injury was found to significantly decrease in the presence of PLPs (p<0.01) compared to vehicle only and were similar to those in the presence of the current clinical standard, Factor VIIa. PLPs resulted in a more significant reduction in bleeding time than transfusion of 100-fold greater numbers of infused fresh platelets (figure S8). S11-ULC μgels did not significantly affect bleeding times compared to vehicle only control (figure 4), and total blood loss was significantly less in the presence PLPs compared to S11-ULC μgels (p<0.05). Analysis of bleeding dynamics also demonstrated that PLPs resulted in the slowest blood loss over time, while S11-ULC μgels resulted in the most rapid blood loss (figure 4C–D). Wound tissue was analyzed postmortem for fibrin and PLP deposition through MSB staining for fibrin and immunohistochemical staining for the MYC-tag encoded on the sdFvs. Co-localization of PLPs within fibrin clots (figure 4D, arrows) was observed, while minimal MYC staining was observed in S11-ULC μgels tissue samples. Furthermore, higher levels of fibrin staining were observed in vessels collected from animals receiving PLPs. These data indicate that PLPs are capable of localizing to the site of injury, enhancing fibrin clot formation, and decreasing bleeding time and blood loss.

We demonstrate that our PLPs can recapitulate some key functions of platelets including binding, stabilizing and enhancing fibrin clot formation, responsiveness to injury cues, and induction of clot collapse. To our knowledge, no platelet-mimetic materials recapitulate all of these important functions of platelets involved in hemostasis. It should be noted that several differences exist between our PLPs and natural platelets. While PLPs bind only fibrin, resting platelets bind neither fibrinogen nor fibrin, but activated platelets bind both and serve as initiators of wound healing. Furthermore, natural platelets serve as a platform for recruitment of procoagulant factors that enhance the hemostatic response and actively generate thrombin on their surfaces, resulting in fibrin fibers originating from platelet aggregates. Platelet phosphatidyl serine also contributes to this response by binding plasma coagulation proteins. PLPs bind nascent fibrin fibers, serving predominantly as nucleating agents and cross-linkers during clot formation, and are trapped within the network. In other words, platelets initiate clotting, while PLPs enhance fibrin formation that has already begun. This is a nice advantage of our system since PLPs should not initiate adverse clot formation; our data with hemophilia patients further supports this premise.

Perhaps the most exciting, and surprising, result generated by our approach is the ability to induce clot collapse without the active contractile machinery required for natural platelet-mediated clot contraction. In the native platelet, ATP-driven actinomyosin mediated contraction enables rapid collapse over a few hours while ATP-independent collapse resulting from our PLPs occurs on a much longer time scale (tens of hours). The multivalent interactions between μgel polymer chains and fibrin fibers likely cause local deformation of the fibrin network, acting as cross-linkers or bridging sites between adjacent fibers. Such bridging sites likely make subsequent μgel binding more favorable via a cooperative “zipper effect” ^22–24. Over time these multiple small collapses and concomitant μgel binding will lead to an overall (macroscopically observable) network collapse. Recent studies characterizing collapse of actin polymers by myosin demonstrate that actin network connectivity (determined by the degree of cross-linking) controls length scales of collapse, with weakly connected networks contracting locally, medially cross-linked networks contracting into multiple disjointed clusters, and strongly cross-linked networks contracting into a single dense cluster ²⁵. We observe a similar phenomenon in our fibrin-PLP systems, however in our system, the collapse changes over time as the PLPs, presumably, increase the degree of network connectivity. Over time, we observe that PLPs induce local network collapse, followed by collapse into multiple dense fibrin clusters, ultimately inducing collapse into a single dense fibrin cluster. A few features of the PLPs appear to be critical for this process. First, the high fibrin affinity imparted by H6 permits the particles to remain bound to the fibers as the clot collapses. Second, the extremely low density of the ULC μgels likely permits the pendant sdFvs to engage fibrin with near-native affinity. In contrast, more densely cross-linked (and therefore not as deformable) particles would restrict the ability of pendant sdFvs to bind to multiple fibers, therefore decreasing the degree of fiber deformation and diminishing the cooperativity of subsequent μgel binding events. This is supported by our clot collapse simulations, which demonstrate that loosely cross-linked particles take on multiple conformations to interact with the fibrin network and induce clot collapse to a greater degree than non-deformable hard particles. It is also possible that PLP binding to preformed fibrin oligomers and subsequent assembly of the network introduces an uneven distribution of tension in the network, resulting in instabilities that further contribute to network collapse.

Our ability to recapitulate many functions of natural platelets demonstrates the power of evolution-based approaches in designing biomimetic biomaterials. Although this report specifically focuses on PLPs and their emergent dynamic behaviors, we believe that the use of ultra-soft/deformable colloidal systems represents a unique strategy toward the development of a new class of biosynthetic composite materials whose component parts more intimately engage and augment one another.