Red blood cell aggregation within a blood clot causes platelet-independent clot shrinkage Open Access

Contraction or retraction of blood clots is the process of clot shrinkage and structural remodeling^1,2 that occurs both in vitro and in vivo.³ It affects the course and outcomes of thrombosis^3-8 and is also important in hemostasis.^3,9 Contraction of blood clots occurs by activated platelets attached to fibrin fibers.^10,11 Platelet contractility is driven by the interaction of nonmuscle myosin IIA with actin.^12,13 The intracellular force is transmitted to fibrin via the integrin αIIbβ3,¹⁴ resulting in compaction of the clot.^1,11 During contraction, a blood clot undergoes redistribution of components, such that the platelet-fibrin meshwork is accumulated at the clot periphery, whereas red blood cells (RBCs) are condensed within the clot core¹ and become tightly packed to a polyhedral shape, hence named “polyhedrocytes” or “piezocytes.”^1,15,16 

The role of RBCs in clot contraction remains poorly understood.¹⁷ The extent and rate of clot shrinkage are inversely related to the volume fraction of RBCs in blood (hematocrit).^18,19 At the same time, the deformation of RBCs activates the mechanosensitive membrane cationic channels Piezo1, followed by reduction of RBCs in size and surface exposure of phosphatidylserine, increasing the generation of endogenous thrombin,²⁰ and increased thrombin activity further augments platelet contractility. Notably, blood clot contraction depends not only on the quantity but also on the quality of RBCs, such that RBCs with increased rigidity reduce clot contraction compared with normal softer RBCs.^21,22 

Analysis of in vitro blood clots and in vivo thrombi has shown that often RBCs comprise the main constituent, sometimes reaching >80% of the clot volume fraction,^1,6,15,23-25 although after clot contraction the RBC packing is extremely tight, owing to platelet-induced compression.^15,26 It is unknown whether RBC aggregation can play a role in blood clot formation and structural rearrangements.

At low shear rates, RBCs form linear (rouleaux) or 3-dimensional (3D) aggregates,^27-29 which promotes venous thrombosis.^30,31 There are 2 mechanisms proposed for such RBC aggregation, namely the bridging and the depletion models. In the bridging model, plasma proteins (mainly fibrinogen) “bridge” the adjacent cells to form aggregates.^32-34 In the depletion model, aggregation occurs because the concentration of osmotically active macromolecules near RBC surfaces is depleted compared with their concentration in the bulk phase, resulting in an osmotic pressure gradient and a net “depletion” force.^35,36 

This study aimed to explore whether RBC aggregation can induce blood clot shrinkage per se, in the absence of platelets, and, if yes, what the mechanisms involved are. Our experiments have demonstrated that RBC aggregation plays a remarkable role in the process of blood clot shrinkage.

For detailed methods, see Section 1 of the supplemental Material.

The mixtures of washed RBCs and platelet-free plasma (PFP) from the same donor referred to as platelet-depleted reconstituted blood samples (supplemental Figures 1-3) were clotted to follow subsequent clot shrinkage.¹⁸ Briefly, clot formation and further compaction either in the blood (natural or reconstituted) or platelet-rich plasma (PRP) samples were induced by adding 2 mM CaCl₂ and 1 U/mL human thrombin. The thrombin-activated blood or plasma samples were transferred to cuvettes prelubricated with a residual layer of a detergent to prevent fibrin sticking to the walls. The cuvette was placed into a chamber (37°C) of the thrombodynamics analyzer (HemaCore), and the clot size was tracked for 60 minutes. The serial clot images were converted computationally into kinetic curves (relative extent of clot shrinkage vs time) that were analyzed for the extent of clot contraction/shrinkage at the end point (60 minutes), area under the curve, average velocity of clot contraction/shrinkage, and the lag time (supplemental Figure 4).

Clots from reconstituted platelet-depleted blood samples were formed between horizontal rheometer plates, and the clot tension was measured continuously as a negative normal (vertical) pulling force.³⁷ The RBC-related tension was determined as the difference between the maximal normal forces generated by the clots in the presence and absence of RBCs.

To better understand the mechanisms underlying the RBC-induced clot compaction, we minimize the free energy in a continuum model for fibrin clots containing RBC (for details, see Section 3 of the supplemental Material). The model builds on the continuum analysis of fibrin gels³⁸ by adding the contributions of aggregated RBC to the free energy. Briefly, the elastic energy density of the fibrin network is based on an 8-chain model of polymer elasticity as adapted to fibrin and described previously.^39,40 The energy density of the bridging interactions of RBCs is taken to be neo-Hookean given that it represents the elastic energy stored in stretched (or compressed) biopolymer chains anchored in the membranes of adjacent RBC.⁴¹ The energy density of interactions between RBC and fibrin is based on measured adhesive interactions between RBC and fibrinogen.⁴² The model also includes the energy density associated with electrostatic and depletion interactions of RBC based on the theory of Neu and Meiselman⁴³ and the free energy density of the mixing of liquid and other components of the clot based on a Flory-Rehner type model that is used in elastomeric gels.³⁸ 

The study involving human subjects was approved by the University of Pennsylvania Institutional Review Board.

While studying the effects of various platelet inhibitors on blood clot contraction, we discovered that their efficacy was different in whole blood and PRP. To suppress platelet contractility, we applied a number of inhibitors at highly effective concentrations (see the supplemental Methods): PGE₁ (increases cyclic adenosine monophosphate level), latrunculin A (an inhibitor of actin polymerization), blebbistatin (an inhibitor of nonmuscle myosin IIa), abciximab (a Fab-antibody fragment, inhibitor of fibrin interaction with the integrin αIIbβ3), atopaxar (PAR1 antagonist), and tirofiban (synthetic integrin αIIbβ3 inhibitor). Expectedly, the inhibitors suppressed all the parameters of clot contraction (Figure 1; supplemental Table 1). The only exception was the effect of PGE₁, which prolonged the lag time and reduced the area under the kinetic curve, but did not significantly reduce the average velocity and final extent of clot contraction. However, in whole blood, none of the inhibitors suppressed clot contraction completely, allowing for the residual 70% to 85% of the initial clot size (except for PGE₁). When the same platelet inhibitors at the same concentrations were added to PRP, the contraction of PRP clots was almost completely prevented compared with whole blood, that is, retaining almost 100% of the initial clot size after 1 hour (Figure 1; supplemental Table 1). In particular, the inhibitors completely abolished clot contraction in PRP from normal ∼90% down to 0, whereas in whole blood the inhibitors reduced the extent of contraction from normal 45% to 51% down to 15% to 23%, that is, twofold to threefold (with the exception of PGE₁) (supplemental Table 1). The inhibitory effect of ticagrelor on clot contraction was moderate, although ticagrelor was still more efficient in PRP than whole blood (supplemental Figure 5; supplemental Table 1).

These results suggest that blood clot compaction may be caused by RBCs and that this effect, unlike the platelet-driven clot compression, is unrelated to the intracellular contractile actomyosin machinery associated with αIIbβ3-mediated transmission of the traction force to the extracellular fibrin network.

To rule out the possibility that the difference in the inhibitors’ efficacy between whole blood and PRP was because of absorbance by RBCs, requiring higher concentrations, we performed experiments on whole blood clot contraction at various concentrations of the inhibitors and plotted the dose-response data (supplemental Figure 6). The results show that at the final concentrations >100 μM for blebbistatin, >2 μM for latrunculin A, >50 μg/mL for abciximab, >5 μM for atopaxar, and >10 μM tirofiban, the final extent of contraction remained constant at a level of ∼20% to 25%. Therefore, the results obtained confirm that the distinct difference in sensitivity to the inhibitors between whole blood and PRP is real and not because of a concentration shift.

The results obtained confirm the key role of platelet contractility, but they also suggest that there are noncontractile cellular reactions, likely related to the presence of RBCs, that can also contribute to the fibrin clot shrinkage.

To see whether clot compaction can be driven by forces other than platelet contractility, we induced clot formation in the platelet-depleted reconstituted blood samples prepared by mixing washed RBCs and PFP from the same donor at a physiological RBC count of 4 × 10⁶/μL. Continuous registration of the clot size revealed a progressive volumetric shrinkage that reached >20% of the initial clot size within 1 hour after thrombin-induced clot formation (Figure 2A). Importantly, control clots formed in the same PFP without RBCs remained visually unchanged within the time of observation. The results indicate that fibrin clots in plasma can be partially compacted in the absence of platelets owing to the presence of RBCs. This process was insensitive to the inhibitors of nonmuscle cell contractility, such as blebbistatin, latrunculin A, and abciximab, given that none of the kinetic parameters of the RBC-induced clot shrinkage were affected (Figure 3; supplemental Table 2). In parallel control experiments, the inhibitors applied at the same concentrations effectively suppressed the platelet-driven clot contraction (Figure 1; supplemental Table 1). The results obtained clearly show that blood clot compaction may be caused by RBCs and that this effect, unlike the platelet-driven clot compression, is unrelated to the intracellular contractile actomyosin machinery associated with αIIbβ3-mediated transmission of the traction force to the extracellular fibrin network. Notably, the RBC-induced clot shrinkage increased with an increase of the RBC count from 4 × 10⁶/μL to 6 × 10⁶/μL, which caused a significant increment in the extent and average velocity of clot shrinkage without affecting the lag time and area under the kinetic curve (supplemental Figure 7; supplemental Table 3), confirming that the process is RBC dependent.

Although the complete absence of platelets in blood is hardly possible, there are a number of pathological conditions where the platelet-driven clot contraction is compromised and the RBC-induced clot compaction could be substantial. To test this presumption, we investigated the effects of RBCs on the extent of clot shrinkage in the presence of low platelet counts (20×10³/μL, 50×10³/μL, and 70×10³/μL), mimicking severe to moderate thrombocytopenia. Figure 2B shows a strong promoting effect of RBCs on the impaired clot contraction when the platelet number is low. The stimulating effect of RBCs was proportional to the background low platelet counts, suggesting that the RBCs enhance the reduced bulk platelet contractility in conditions resembling thrombocytopenia.

To get insights into the structural mechanisms of the RBC-induced clot shrinkage, we compared the morphology of the platelet-depleted reconstituted blood clots right after formation and after 1 hour of unconstrained compaction. As revealed by light microscopy and scanning electron microscopy, the initial clot contained relatively sparse mostly single RBCs (Figure 4A) that had a biconcave shape and were embedded randomly into a network of fibrin fibers (Figure 4C; supplemental Figure 8A,C). After the clot shrinkage, RBCs were arranged into aggregates (Figure 4B), either linear stacks (rouleaux) or unstructured 3D accumulations with the intercellular space filled with amorphous protein mass (Figure 4D; supplemental Figure 8B,D). Thus, the main structural difference between the platelet-depleted blood clots before and after shrinkage was the intercellular adhesion of RBCs, which suggests that the RBC aggregation is the driving force of clot compaction in the absence of platelets.

To see whether the observed RBC aggregation within a clot is because of fibrin formation, not other triggers, control experiments were performed in which RBCs were suspended and allowed to settle for 1 hour in PFP, thrombin-induced PRP-derived serum, and thrombin in the absence and presence of bovine serum albumin (supplemental Figure 9). When RBCs were suspended in PFP, minor macroscopic signs of RBC aggregation were observed (supplemental Figure 9Ai) confirmed by microscopically revealed small (<10 RBCs per aggregate) sporadic stacks of RBCs (supplemental Figure 9Aii). In the other environmental conditions studied, no signs of RBC aggregation were observed, either macro- or microscopically (supplemental Figure 9B-E). Therefore, the RBC aggregation inside a clot is triggered owing to the presence of fibrin, which entraps RBCs and brings them close together inside the network pores. No other components of plasma or serum, including active thrombin, were able to induce significant RBC aggregation by themselves, but serum proteins could augment the RBC aggregations via the osmotic depletion mechanism analyzed later.

Given that the RBC-induced compaction of a clot is a biomechanical process, it must have a measurable force that pulls on a fibrin network, making it more condensed. The average isometric tension generated by RBCs within a constrained, platelet-depleted clot in a rheometer was measured as 143 ± 121 mN. When normalized by the total number of RBCs in the sample, the estimated force per 1 RBC was ∼120 ± 100 pN (Figure 5). This measurement reflects the aggregation forces produced by aggregating RBCs, which play a role in clot compaction in the absence of platelets.

To assess the necessity of plasma proteins in the RBC-induced clot shrinkage, we made clots from purified fibrinogen mixed with RBCs at a physiological count of ∼4 × 10⁶/μL. The results show that, in the absence of plasma proteins, RBCs did not induce shrinkage of a fibrin clot, at least within 1 hour of observation (Figure 6A), indicating that some plasma proteins, other than fibrinogen, are indispensable to the RBC aggregation and shrinkage of fibrin clots. This result also suggests that thrombin by itself is not a trigger of RBC aggregation because of the lack of RBC aggregation/agglutination in the presence of thrombin without fibrinogen (supplemental Figure 9). It is noteworthy that clot formation, either in plasma or in a purified system, results in fibrinogen to fibrin conversion without free fibrinogen left in the serum; therefore, the fibrinogen-mediated bridging mechanism for RBC aggregation within a clot seems highly unlikely.

To test whether the RBC-induced shrinkage of fibrin clots is osmotically driven, we added an osmotically active macromolecule, 100 kDa dextran, to the mixture of purified fibrinogen and platelet-depleted RBCs followed by clotting with thrombin. In the presence of dextran, the purified fibrin clots containing RBCs underwent apparent shrinkage (Figure 6A) and the effect was concentration dependent (Figure 6B-E). There was a clear dose-dependent increase in the extent of clot shrinkage (Figure 6C), shortening of the lag time (Figure 6D), and an increase in average velocity (Figure 6E) after 1 hour (supplemental Table 4). These results strongly suggest the importance of the osmotic depletion interactions between aggregating RBCs as a driving force of the RBC-induced shrinkage of fibrin clots.

As a negative control for the role of RBC aggregation in clot shrinkage, we performed experiments in platelet-depleted plasma mixed with RBCs rigidified by pretreating them with low concentrations of glutaraldehyde to reduce deformability and aggregability. Glutaraldehyde-treated RBCs had a significantly (on average of more than twofold) reduced ability to induce clot shrinkage (supplemental Figure 10A-D). This effect was because of the increased rigidity of glutaraldehyde-treated RBCs confirmed with ektacytometry (supplemental Figure 10E-H).

A plot of the extent of contraction, (1-V/V₀), as a function of dextran concentration $c2$ at equilibrium is shown together with experimental data in Figure 7. The modeling parameters were consistent with measured characteristics of RBC interactions (supplemental Material, Section 3). The increased extent of contraction with an increase in dextran concentration is captured very well. We used the model to show how various parameters dictating the energetics of fibrin, RBC, and liquid govern the trends in clot shrinkage as a function of the osmolyte concentration. Supplemental Figure 11A shows that increasing the elastic modulus of the fibrin clot reduces clot shrinkage given that osmotic depletion forces encounter greater resistance to deforming the clots, whereas supplemental Figure 11B shows that changing the RBC-fibrin interactions by a similar multiple has a smaller effect. The latter suggests that the RBC-fibrin interactions play a weaker role than clot elasticity and RBC electrostatic interactions in clot shrinkage. The effects of various model parameters on clot shrinkage are discussed in detail in the supplemental Material, Section 3.

Incorporation of RBCs into fibrin clots has important biological consequences, including alterations in fibrin network structure and clot mechanical properties,²⁵ predisposition to fracture,⁴⁴ lytic resistance,⁴⁵ and more. Here, we establish a hitherto unknown ability of RBCs embedded into a fibrin network to induce clot shrinkage or contraction.

In the presence of platelet inhibitors applied at highly effective concentrations, there is incomplete inhibition of the contraction process when clots are formed in whole blood, whereas the same inhibitors applied in PRP clots almost fully suppress contraction (Figure 1; supplemental Table 1). These observations have led us to the assumption that RBCs can play a mechanically active role in contraction of blood clots. To test this hypothesis, we have studied macroscopic thrombin-induced clots from reconstituted human blood samples made of PFP and platelet-depleted RBCs. These clots have no detectable platelets, yet they are seen to experience substantial clot shrinkage, exceeding 20% of the initial clot volume within an hour (Figure 2A). However, any active contractility of RBCs or other cells has been excluded because the clot shrinkage is insensitive to the inhibitors of nonmuscle myosin IIa (blebbistatin) and actin polymerization (latrunculin A) (Figure 3; supplemental Table 2), known to generate traction forces.^46,47 An alternative explanation for the mechanical activity of RBCs is their aggregation, which was confirmed morphologically in the shrunken clots containing RBCs (Figure 4). Moreover, we have shown directly the mechanical potential of aggregating RBCs and quantified the tensile force generated by the aggregated RBCs within a clot, which was found to be ∼120 pN per 1 RBC, when the steady state is reached (Figure 5). Notably, this number is based on a unidirectional bulk force measurement and does not account for potential redistribution of tensile forces across the 3D structure of the clot. Nevertheless, this rough evaluation is consistent with an earlier study where the adhesion force between 2 RBCs was found to be in the range of 43 to 169 pN.⁴⁸ In addition, the mean adhesion force between 2 RBCs in the presence of lipopolysaccharide was determined to be 100 ± 84 pN.⁴⁹ Therefore, we have uncovered and quantified the mechanical interactions between aggregating RBCs that generate the forces driving the process of blood clot shrinkage. Theoretically, the fibrin gel may undergo spontaneous shrinkage known as gel syneresis,⁵⁰ but this pure physicochemical reaction can be excluded here, because there has been no fibrin clot shrinkage in the absence of RBCs (Figure 6A).

To create a more detailed mechanistic picture of the RBC-induced fibrin clot shrinkage, the experimental observations described earlier have been captured and analyzed in a new model for the mechanical behavior of clots containing fibrin and RBC (no platelets). The proposed model is a novel combination of a continuum gel model and a separate model for depletion interactions of RBCs, and it provides predictions about the extent of clot contraction as a function of osmolyte concentration in response to changes in RBC electrostatic interactions, fibrin network elasticity, RBC volume fraction, and strength of adhesive interactions between RBCs and fibrin (see the supplemental Material, Section 3; supplemental Figures 12-17). This continuum model takes account of fibrin gel poroelasticity, electrostatic, depletion, and bridging interactions of aggregated RBC and RBC-fibrin adhesive interactions. We computed the moduli corresponding to each of these interactions based on experimental data and found that bridging interactions of RBCs are relatively weak, RBC-fibrin adhesive interactions are stronger, and electrostatic and depletion interactions are strongest, but only when RBCs are in close proximity to each other. It is the formation of a fibrin network that provides a trigger for the aggregation of RBC by bringing them close together through embedment and constrained packing within network pores, so that electrostatic, depletion, and bridging interactions can come into play. The aggregation of RBCs can cause shrinkage of a clot when the concentration of osmolyte (eg, dextran or plasma proteins) is high enough.⁵¹ The osmotic depletion forces push the RBCs closer, whereas electrostatic repulsion drives them apart. Osmotic forces also work against the elasticity of the composite material formed by the fibrin network and aggregated RBCs interacting among themselves and fibrin. The forces owing to adhesion between RBC and fibrin help osmotic forces to bring RBCs closer together, but they are of much smaller magnitude. As such, the parameters controlling the electrostatic interactions of RBCs, the elasticity of the fibrin gel, and the adhesive interactions between RBCs and fibrin contribute to the physics of shrinkage of these clots, in addition to the concentration of the osmotic (oncotic) agents that control the depletion forces.

Based on our results and the literature on RBC aggregation, we postulate that there are 3 interconnected processes that result in the RBC-induced fibrin clot shrinkage. (1) After clot formation, RBCs get entrapped and brought close together within the pores of the fibrin network. (2) As a result of compaction, RBC aggregation occurs, driven by the osmotic gradient and depletion interactions, that is, a lower localized plasma protein concentration near the cell surface than the surrounding medium. (3) Owing to the RBC-fibrin binding interactions, the traction force generated during RBC aggregation is transmitted to fibrin fibers, causing shrinkage or contraction of the entire fibrin network.

The (patho)physiological importance of the revealed RBC-induced blood clot shrinkage is likely to be substantial in pathological conditions where the platelet-driven clot contraction is compromised. These conditions are related to many hereditary and acquired diseases associated with thrombocytopenia and/or thrombocytopathies.^52-62 This presumption has been confirmed in the experiments where the effects of RBCs on clot shrinkage were studied in the presence of low platelet counts, mimicking severe to moderate thrombocytopenia. The promoting effect of RBCs on the impaired clot contraction is quite substantial and proportional to the platelet counts, suggesting that the RBCs work in synergy with the platelet contractility, when it is not strong enough to effectively shrink a clot by itself (Figure 2B). These results strongly support the concept that, in case of weak bulk platelet contractility, RBCs at a physiological level can substantially improve clot contraction via a complementary or spare mechanism for clot compaction.

There are several other conceivable options that remain to be studied. RBC aggregation can normally contribute to the platelet-driven clot contraction of hemostatic clots and pathological thrombi. The extent of such contribution may be relatively small because the weaker RBC aggregation forces can be swamped out by the stronger contractile forces of activated platelets. In addition, RBC-induced shrinkage is likely to compete with platelet-driven contraction in venous thrombi with especially small amounts of platelets, such as chronic thrombi, thrombi in larger veins, or those associated with slower blood flow that may have a higher proportion of fibrin and RBCs. Notably, the volume fraction of platelets in venous thrombi is <1% compared with >60% for RBCs.²⁴ Finally, the contribution of RBCs to blood clot shrinkage is likely to be substantial when there are few and/or dysfunctional platelets in the blood in combination with high RBC counts, as in chronic myelogenous leukemia, myelofibrosis, polycythemia vera, and other myeloproliferative neoplasms,⁶³ as well as RBC transfusion in severe pancytopenia.⁶⁴ 

In summary, this study provides compelling evidence that RBCs play a role in blood clot shrinkage unrelated to platelets. The main mechanism underlying the RBC aggregation within a clot is depletion interactions occurring when RBCs are embedded in close proximity to each other within the pores of the fibrin network. The potential pathophysiological significance of the RBC-induced clot shrinkage may be attributed to the ability of RBC aggregation to reinforce the platelet-driven blood clot contraction and/or promote clot compaction in thrombocytopenia and/or thrombocytopathies, especially associated with polycythemia, and in other pathological conditions with extremely low platelet content in thrombi. Certainly, the newly described phenomenon of RBC-mediated clot shrinkage needs further research to fully elucidate its (patho)physiological significance. As a part of these investigations, future studies must focus on biological and physical properties of RBC-shrunken clots, including permeability, sensitivity to fibrinolysis, mechanical properties, and much more.

The authors thank HemaCore Ltd for providing the thrombodynamics analyzer used in this study. The authors thank Chandrasekaran Nagaswami from the University of Pennsylvania for training on scanning electron microscopy.

This study was supported by the National Institutes of Health grants R01 HL148227, P01 HL146373, R01 HL148014, and R01 HL159256, and American Heart Association grant 25POST1357254/2025 (A.D.P.).

Contribution: A.D.P. contributed to project administration, methodology, and investigation, and wrote the original draft; E.K.R., R.R.K., and O.V.K. contributed to methodology and investigation; V.R.M. reviewed and edited the manuscript, supervised the study, and contributed to funding acquisition; P.K.P. wrote the original draft, reviewed and edited the manuscript, conceptualized the study, and contributed to funding acquisition; R.I.L. wrote the original draft, reviewed and edited the manuscript, and conceptualized and supervised the study; and J.W.W. reviewed and edited the manuscript, conceptualized and supervised the study, and contributed to funding acquisition.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: John W. Weisel, Cell and Developmental Biology, School of Medicine, University of Pennsylvania, 421 Curie Blvd, BRB II/III, Room 1154, Philadelphia, PA 19104-6058; email: weisel@pennmedicine.upenn.edu.