Engineering tough blood clots for rapid haemostasis and enhanced regeneration

2026-05-21Jiang,Shuaibing,Bao,Guangyu,Yang,Zhen,Wu,Jing,Yang,Xingwei,Kim,JooEunJune,Jiang,Roselyn,Zhan,Ying,Nottegar,Alexander,Liu,Yin,Gao,Zu-hua,Beckett,Andrew,Nijnik,Anastasia,Long,Rong,Kastrup,Christian,Li,Jianyu

Nature 2026年8115期

关键词：公式

Abstract

Blood clots are pivotal for haemostasis and regeneration 1 , but they are mechanically weak and form slowly 2 , posing risks for life-threatening haemorrhage and limiting broader applications 3 , 4 , 5 . These limitations are attributed to complex coagulation cascades, abundant mechanically ineffective cells and little structural polymers. Strategies that strengthen polymer networks are inapplicable to these highly cellularized materials. Here we report a strategy that rapidly crosslinks red blood cells into tough cytogels and integrates them within blood clots. The resulting engineered blood clots (EBCs) form within seconds and exhibit a 13-fold increase in fracture toughness, and a 4-fold improvement in adhesion energy compared with native clots. Experiments and modelling identify the rupture of mechanically integrated cells as a key toughening mechanism. In vivo studies demonstrate that EBCs can rapidly halt haemorrhage, promote tissue regeneration, mitigate inflammation and foreign body reactions, and prevent postoperative adhesion. The safety and efficacy of both autologous and allogeneic EBCs were also validated. Our strategy is applicable to a range of cells and polymers. This work may motivate the development and translation of highly cellularized materials for bleeding control, wound management, tissue repair and regenerative medicine.

Main

Highly cellularized materials, in which biological cells constitute the dominant volume fraction, include blood clots, cell aggregates, organoids and many biological tissues 6 , 7 , 8 , 9 . Improving their mechanical properties is challenging because cells are often treated as mechanically ineffective inclusions and structural polymers are sparse 6 , 10 . Blood clots exemplify this limitation: red blood cells (RBCs) occupy about 45% of the volume 11 , whereas fibrin comprises less than 1% (ref. 12 ). As a result, native clots are mechanically weak and prone to fracture 13 , 14 , and their formation requires minutes because of slow native coagulation. These constraints limit their haemostatic performance under severe haemorrhage 15 , 16 . Moreover, failed clot formation, rupture and detachment at wound sites can impair regeneration in organs, including bone, cartilage, muscle, liver and periodontal tissues 17 , 18 , 19 .

Strategies to strengthen biomaterials typically focus on engineering polymer network architecture 20 , 21 , 22 . However, these approaches are poorly suited to highly cellularized materials, due to their deficiency of structural polymers and sensitivity of cells and associated biological processes. In blood clots, coagulation factors, fibrin crosslinking or platelet contraction could stiffen clots, but does not fundamentally address fracture resistance or rapid formation 23 , 24 . Alternatively, leveraging the abundant cellular components could potentially overcome these limitations. Yet attempts to accelerate clotting by inducing RBC aggregation through electrostatic or hydrophobic interactions often yield mechanically weak gels, haemolysis or disrupted fibrin organization 25 , 26 , 27 , 28 . An ideal strategy that directly converts densely packed cells into a continuous load-bearing network while remaining compatible with coagulation could potentiate the performance of blood clots and other highly cellularized materials 29 (Extended Data Fig. 1).

Here we report a strategy termed ‘click clotting’ to mechanically integrate cells into tough cytogels. Using RBCs as a model system, we covalently crosslink high-density cells with a trace of polymer linkers through ultrafast bioorthogonal click chemistry (Fig. 1a), forming cytogels in which cells serve as structural building units rather than passive fillers. These cytogels differ from conventional cell-laden hydrogels, whose mechanical behaviour is primarily governed by polymer networks, and from cell aggregates that possess limited mechanical properties and require lengthy preparation (Supplementary Scheme 1). We further implement click clotting with native coagulation to generate engineered blood clots (EBCs) composed of an RBC cytogel interpenetrated with fibrin (Fig. 1b). Our strategy enables instantaneous clotting and markedly enhanced fracture resistance despite low structural polymer content (Fig. 1c), while preserving the intrinsic bioactivity of blood clots to enhance haemostasis and regeneration (Fig. 1d).

**Fig. 1: Schematics and characteristics of click clotting and EBC.**

Design of tough cytogels

To form tough cytogels in which living cells are covalently crosslinked into a continuous network, we propose two design criteria: (1) covalent linkages should be formed bioorthogonally between cells and polymer linkers; and (2) a long-range connected cell network should be constructed by sufficient linkages using proper cell density, polymer size and concentration. We embody our design strategy with the RBC cytogel as a model system. To satisfy the first criterion, we choose the click reaction of trans-cyclooctene (TCO)/tetrazine (TZ), which is bioorthogonal and ultrafast to form covalent linkages 30 , 31 . This chemistry has been widely applied in in vivo and cellular applications, including for hydrogel crosslinking in cell encapsulation and targeted bioconjugation 32 , 33 , 34 , 35 . Instead of inserting click motifs into the fluid-like lipid bilayer of cell membranes, we conjugate TCO groups covalently to membrane proteins, some of which (for example, BAND 3) are tethered to the underlying cytoskeleton 36 (spectrin network) and can provide robust anchorage (Fig. 1a). The TZ groups are grafted to long-chain polymers such as hyaluronic acid (HA) by carbodiimide chemistry (Supplementary Fig. 1). Confocal imaging confirms the presence and even distribution of TCO on the RBC surface (Fig. 1e) and the lack of fluorescence for unmodified RBCs (Supplementary Fig. 2). Quantitively, we measured a high density of TCO of 5.4 × 10 7 per RBC, and the TCO is stable on the RBC surface over 21 days (Supplementary Fig. 3). The degree of substitution of HA (TZ to carboxyl) is 2–7% dependent on the feed ratio of TZ (Supplementary Table 1).

Mixing modified RBCs (RBC-TCO) and polymer linkers (HA-TZ) results in cytogels by click clotting in less than 5 s (Supplementary Video 1). The plateaued storage modulus G′ of 200 Pa is much larger than the loss modulus G″ of 25 Pa (Fig. 1f), whereas the mixture of unmodified RBCs and polymer linkers remains as a viscous liquid. Our RBC cytogels are stiffer as shown with much higher G′, and more elastic over a wide frequency range, compared with previous viscoelastic blood gels formed by electrostatic (chitosan) 26 , enhanced electrostatic (polymers consisting of cationic and adjacent aromatic residues) 28 and hydrophobic interactions (hydrophobically modified chitosan) 26 (Fig. 1g). Previously reported blood gels exhibit similar tan δ values of about 0.5, in contrast to our RBC cytogel with tan δ of approximately 0.1 (Supplementary Fig. 4), indicating substantially reduced viscous dissipation and a more elastic, solid-like response enabled by covalent intercellular crosslinks. Consistently, EBCs containing crosslinked RBCs sustain large extensions while remaining cohesive (Fig. 1h). We ascribe these mechanical properties to the effective integration of the cytoskeleton by protein-mediated covalent linkage: although the cell membrane is lysed, the proteins tethered to the cytoskeleton still serve as crosslinkers to maintain gel integrity 37 (Supplementary Fig. 5).

To meet the second criterion of long-range connectivity (Fig. 2a), we systematically vary the volume fraction of RBCs and the size and concentration of polymer linkers. When less than 2% by volume (1.3 × 10 8 cell ml −1 ), the cells are too sparse to form a long-range network. With increasing cell content, the shear modulus increases to a peak value at a cell volume ratio of 33% (2.1 × 10 9 cell ml −1 ) (Fig. 2b). Even higher cell content (50%) leads to a decrease of modulus possibly because of poor mixing of the overcrowded cells and the polymer linkers. Using the optimal cell density, we further study the effect of polymer linkers. By testing HA linkers with varying molecular weights (Supplementary Table 2), we found a crucial hydrodynamic size (33–53 nm), which decides whether the polymer linker can crosslink cells into gels (Fig. 2c and Extended Data Fig. 2). Considering that the polymer linker can react only with modified cells and that a finite gap (more than 20 nm) exists between RBCs because of the electrostatic repulsion of negatively charged cell membranes 38 (Fig. 2a), we rationalize this finding that large polymer linkers can mend the gap between cells for effective crosslinking. Apart from the polymer size, the concentration and TZ ratio of polymer linkers determine the linkage density among cells and thus the shear moduli of RBC cytogels (Fig. 2d and Supplementary Fig. 6). Altogether, we demonstrate that proper RBC volume ratio (33%), polymer size (137 nm) and polymer concentration (1–2%) jointly result in sufficient linkages, which are important for the long-range continuous network formation of tough cytogels.

**Fig. 2: Click clotting and fracture of RBC cytogels.**

Our strategy is applicable across polymers and cell types when these design criteria are met. By screening polymer linkers, we formed cytogels with alginate, chitosan and polyacrylic acid of large hydrodynamic sizes (Supplementary Fig. 7). Also, we demonstrated the versatility of our design by forming living cytogels from fibroblasts, adipose-derived stem cells and microalgae (Supplementary Fig. 8). Our strategy can rapidly assemble cells into mechanically robust living materials with high cell content and high cell viability (>80%), which are technically challenging and time-consuming to achieve using conventional methods 39 . This would pave new ways for developing engineered tissues and engineered living constructs with tailored functionality 40 , 41 .

Mechanical properties of tough cytogels

The design of RBC cytogels endows them with unique mechanical responses. Despite previous reports on cell crosslinking, the mechanism and mechanics of cell-crosslinked gels are poorly understood. As the cells are mechanically integrated and highly dissipative, we propose that the cells could contribute substantially to the fracture properties of RBC cytogels. To test this point, we characterized the fracture energy of the RBC cytogels using modified lap-shear tests (Fig. 2e,f). The measured fracture energy shows a non-monotonic trend with respect to the cell volume ratio (Fig. 2g), consistent with that of the shear modulus (Fig. 2b). Specifically, the fracture energy peaks at the RBC volume ratio of 33%, comparable to the RBC content in native whole blood. However, the fracture energy is not sensitive to the concentration of polymer linkers (Fig. 2h). The result implies that the fracture of cell-crosslinked gels is dominated by cell rupture, which is different from conventional polymeric hydrogels, whose fracture toughness increases with increasing polymer content. In support of this point, we also found a weak dependence on loading rate (Supplementary Fig. 9), unlike the strong rate dependence commonly seen in viscoelastic hydrogels. As the energy required to rupture an individual cell becomes the limiting factor, the macroscopic fracture energy is closely related to cell volume ratio rather than the polymer concentration. In line with this mechanism, two control experiments: (1) direct EDC/N-hydroxysuccinimide (NHS)-mediated crosslinking of RBCs with HA and (2) bulk click-crosslinked HA hydrogels encapsulating unmodified RBCs (Supplementary Fig. 10), both yielded much weaker and brittle clots, underscoring the necessity of the cytogel-forming click clotting mechanism for mechanical reinforcement.

To further investigate the mechanism underlying the fracture of RBC cytogels, we evaluated the dissipative property of the gels with cyclic tensile tests (Fig. 2i). The RBC cytogels were subject to two loading–unloading cycles at three different maximum stretches. The hysteresis (the area between loading and unloading paths) is significant during the first cycle but disappears in the following cycles (Fig. 2j), reminiscent of the Mullins effect in elastomers and tough hydrogels 21 , 42 . This characteristic is presumably attributed to the rupture of RBCs, dissipating energy under large deformation and causing permanent damage. As the rupture of RBCs would release intracellular haemoglobin, we further measured the amount of haemoglobin released from the RBC cytogel post-test. We observed a linear correlation between released haemoglobin and stretch ratio (Fig. 2k), supporting our hypothesis.

Based on the mechanical testing results, we further developed a finite-element model to simulate the fracture of RBC cytogels. We use a cohesive zone model to simulate the crack tip opening and the Mullins model to capture the energy dissipation in the bulk (Methods and Extended Data Fig. 3). Parameterized with the cyclic test results, our finite-element model with the Mullins effect captures the force–displacement curve and the fracture energy calculated from the lap-shear tests (Fig. 2l and Extended Data Fig. 4). By contrast, the model without the Mullins effect predicts a much lower fracture energy, signifying the role of energy dissipation in toughening RBC cytogels, which is deduced to be because of cell rupture.

Synthesis and characterization of EBC

Because cytogel formation proceeds through a bioorthogonal mechanism, click clotting can act synergistically with native coagulation cascades in whole blood. The resulting EBCs thus integrate the cytogel within a fibrin network to achieve instantaneous gelation and enhanced mechanical performance. To test so, we formed EBCs by mixing RBC-TCO in anticoagulated whole blood and HA-TZ in PBS or platelet-rich plasma (PRP) solutions supplemented with calcium. Calcium was added to recalcify the anticoagulated blood and restore native coagulation activity, enabling fibrin formation alongside the click clotting process. Given the distinct reaction kinetics, we expect rapid solidification due to RBC crosslinking and subsequent stiffening by the formation of a fibrin network. The former can be readily tuned by varying the bioorthogonal reaction (Fig. 3a). With the ultrafast TCO-TZ reaction, the storage modulus (G′) well exceeds the loss modulus (G″) instantaneously at the beginning, and gradually increases to a plateau of 250 Pa (Fig. 3b), indicating both clotting mechanisms take place. The measured gelation time (that is, the time when G′ > G″) of EBCs is significantly smaller than that of native blood clots (NBCs), which is beneficial for rapid haemorrhage control (Fig. 3c). The clotting time can be tuned with other click reactions, for instance, dibenzocyclooctyne (DBCO)-azide reaction 43 , which leads to a clotting time around 2–3 min (Supplementary Fig. 11). To this end, our strategy of forming EBCs is amendable to various click reactions tailorable for specific applications requiring varied time windows.

**Fig. 3: Structural and mechanical characterization of EBC.**

To verify the formation of the RBC network and fibrin network, we characterized the microstructure of EBC with confocal imaging and scanning electron microscopy (SEM). We labelled the individual components, including RBCs, HA-TZ and fibrin, with fluorescence, and found that RBCs with red fluorescence clustered continuously by the blue HA-TZ, proving the click clotting of the RBC network (Fig. 3d). A green fibrous network interpenetrated with the RBC network reflects the formation of fibrin. The interpenetrating network of EBC was further revealed in the reconstructed 3D confocal images (Extended Data Fig. 5). By contrast, unmodified RBCs tend to be trapped in the fibrin network as separate aggregates (Supplementary Fig. 12). SEM images show the cell–cell adhesion by polymer crosslink in EBC, whereas unmodified RBCs did not adhere to each other within the blood clots (Fig. 3e and Supplementary Fig. 13). These results substantiate excellent compatibility between click clotting and native coagulation mechanism and full structural integration of RBCs into the EBC.

The structure of EBC is expected to result in high fracture toughness. To test this point and compare EBC with NBC and a commercial haemostat, Floseal (Baxter), we performed modified lap-shear tests to measure their fracture energy (Fig. 3f). The fracture energy of EBC (25.6 J m −2 , around 0.15 mg ml −1 fibrin) is 10 times and 9.5 times higher than NBC and Floseal, respectively (Fig. 3g). By raising the fibrin content to its physiological level (around 0.3 mg ml −1 ), we made EBC h with even higher fracture energy of 32.4 J m −2 . Moreover, we measured the burst pressure of different materials using a burst chamber setup (Supplementary Fig. 14). Both RBC cytogels and EBCs achieve much higher burst pressure, compared with NBC and Floseal (Fig. 3h and Supplementary Video 2). Specifically, the burst pressure obtained with EBC h (59 mmHg) is around thrice that of NBC. Given the coupling between bulk and adhesion properties, we propose that our design could enhance the adhesion energy of EBC on biological tissues. We performed peeling tests (Supplementary Fig. 15) and measured the adhesion energy of EBC to be 6.3 J m −2 , around four times higher than that of native clots (Fig. 3i). These results support the use of EBCs for haemorrhage control in solid organs and low- to moderate-pressure vessels. For major arterial bleeding, further mechanical reinforcement, such as rigid backings or fabric meshes, would be required to enable effective sealing.

Our strategy is advantageous over previously reported methods using chitosan to crosslink RBCs, which lead to brittle clots, haemolysis or inconsistent clotting (Supplementary Fig. 16). EBCs outperform NBCs and many polymeric hydrogels (for example, gelatin and alginate) in terms of fracture resistance at the comparable content of structural polymers (Supplementary Fig. 17). Compared with NBCs, EBCs show markedly improved elastic recovery on loading (Extended Data Fig. 6), higher tensile and compressive strengths (Supplementary Fig. 18). These enhancements reflect the reinforcing and elastic contributions of covalent polymer linkages between RBCs introduced by click clotting. Its excellent mechanical performance at the low-concentration region would enlighten the design for robust, highly cellularized materials. Crosslinked RBCs dissipate considerable energy through membrane rupture when EBC is stressed, while the fibrin network maintains physical integrity. Unlike conventional polymeric hydrogels that rely on dense polymer networks, EBCs achieve exceptional mechanical performance through direct cell–cell crosslinking and rupture-driven energy dissipation. Furthermore, as the matrix of EBC is more dissipative and fibrin and platelets enable interfacial bonding with tissues, the EBC affords tougher bioadhesion. The mechanical properties, together with ultrafast clotting kinetics, make EBCs appealing for various applications such as surgical sealants, drug/cell delivery carriers and adhesive scaffolds.

Biocompatibility and biodegradability

To evaluate the translation potential of EBCs, we conducted a series of tests, including in vitro experiments to evaluate cytocompatibility and haemocompatibility, as well as in vivo rodent experiments to assess biocompatibility and biodegradability. We first cultured immortalized human vocal fold fibroblasts with EBC extract for 24 h. The LIVE/DEAD assay indicated excellent cytocompatibility of EBC, with approximately 100% cell viability (Supplementary Fig. 19). Moreover, EBC demonstrated the ability to promote cell proliferation when serum supplement is absent in the media, because of the regenerative properties of blood clots, containing growth factors and cytokines. These results indicate that EBC could serve as a bioactive matrix to facilitate cell growth.

Regarding haemocompatibility, we investigated the impact of EBC on haemolysis, platelet function and immune cell responses within the blood. First, our cytogel formulation induced no haemolysis, in contrast to chitosan, a commonly used haemostatic agent, which induced significant haemolysis (Supplementary Fig. 20). Furthermore, EBC did not adversely affect platelets and leukocytes in circulation as evaluated by flow cytometry (Supplementary Figs. 21 and 22). EBC did not activate or destroy these blood components, including platelets, monocytes, granulocytes, B lymphocytes and natural killer cells. Note that there is a minimal induction of the MHC-II activation marker on classical monocytes and of the CD44 activation marker on T cells (Supplementary Fig. 23). Furthermore, we examined immune responses under exposure to allogeneic EBCs prepared using blood-type-matched donor blood, and found that T cells, B cells and natural killer cells remained quiescent, indicating minimal immune activation even in the allogeneic setting (Supplementary Fig. 24). The excellent blood compatibility of EBC can be attributed to its native components and bioorthogonal crosslinking, making it suitable for blood-contacting applications, such as haemostatic sealants and embolization agents.

To further confirm the in vivo biocompatibility of EBC, we performed subcutaneous implantation in rodent models and compared it with NBC, RBC cytogel and Floseal (Fig. 4a). Histological analysis was conducted on explanted samples on days 3, 7, 14 and 28 (Extended Data Fig. 7). On day 3 post-implantation, all materials led to minimal presence and invasion of lymphocytes and plasmocytes, indicating mild levels of inflammation (degree 1 or 2). Importantly, the inflammatory response to EBC and RBC cytogels was reduced to no or very mild levels of inflammation (degree 0 or 1) on day 7 and thereafter (Fig. 4b). By contrast, Floseal exhibited more inflammatory reactions, with increased fibrosis and numbers of lymphocytes, plasmocytes, fibroblasts and macrophages, resulting in moderate inflammation that persisted on day 14. Overall, EBC demonstrated excellent biocompatibility comparable to autologous materials.

**Fig. 4: Biocompatibility, haemostasis and regeneration with EBC.**

We also evaluated the biodegradation of EBC through subcutaneous implantation in rats for 8 weeks (Fig. 4c and Supplementary Fig. 25). On day 3 post-implantation, significant swelling was observed in Floseal and RBC cytogel, while EBC exhibited minimal swelling. This can be attributed to the fibrin network of EBC, which can resist the swelling of hydrophilic polymers. These implants, while maintaining overall integrity, gradually decreased in size over time. EBC and RBC cytogel degraded significantly more slowly and were completely absorbed within 8 weeks, in contrast to NBC and Floseal. The different degradation profiles can be explained by the availability and activity of enzymes that can degrade gelatin and HA 44 . Prolonged degradation of EBC can extend the duration of implants in vivo. Furthermore, we investigated the biodistribution of EBC degradation products using fluorescent labelling and In Vivo Imaging System (IVIS). As expected, the biodistribution mapping over 28 days showed the liver and kidney as the main clearance pathways (Supplementary Fig. 28). Histological analysis showed no evidence of systemic toxicity in important organs on day 28, suggesting the long-term safety of EBC for implantation.

EBC promotes haemostasis and regeneration

The combination of excellent mechanical and biological properties of EBC makes it a promising solution for haemostasis, tissue repair and regeneration. EBCs can be easily delivered using common syringes and catheters and offer high usability. Moreover, their precursors can be stored in a fridge before usage, as modified RBCs stored in a refrigerator for 7 days showed no significant changes in shear moduli (Supplementary Fig. 26). EBCs can be prepared from either autologous blood or ABO-matched donor blood from a blood bank (Supplementary Scheme 2). Our preliminary study demonstrates that EBCs can be prepared within 30 min from blood collection. Together with the clinical use of human blood and the TCO-TZ chemistry 45 , our EBCs offer usability, stability and suitability for clinical use.

We initially evaluated the efficacy of EBC as an injectable haemostatic sealant for managing noncompressible haemorrhage, which is an unmet clinical challenge associated with major mortality. Using a rat liver deep incision haemorrhage model, we compared the haemostatic efficacy of EBC to Floseal. We created incisions of 7 mm in length and 3 mm in depth and then applied EBC or Floseal to the bleeding liver (Fig. 4d). In the injury-only group, native clot formation occurred in situ but was slow and led to substantial blood loss. The instantaneous clotting ability of EBC allowed us to quickly seal the bleeding site (Supplementary Video 3). Moreover, the EBCs formed in situ exhibited toughness, stretchability and adhesiveness, preventing clot rupture or detachment (Fig. 1g). Our results demonstrated that EBC significantly reduced blood loss from 1,813 mg to 24 mg (Fig. 4e) and decreased bleeding time from 265 s to 5 s (Fig. 4f). By contrast, Floseal absorbed the blood and slowly formed mechanically weak clots, resulting in 148 mg of blood loss and 91 s of bleeding time (Supplementary Video 4). Apart from the liver injury model, we further evaluated EBCs in a rat tail amputation model, in which EBCs prepared from the blood of one rat were applied to another rat, achieving fast haemostasis and markedly less blood loss compared with untreated controls (Supplementary Fig. 27 and Supplementary Video 5). The haemostatic performance of EBC can be attributed to its ultrafast gelation, high fracture toughness and strong adhesion.

To further examine the haemostatic and regenerative capacity of EBC, we used it as a scaffolding material for treating volume defects in a rat liver puncture haemorrhage model. We created a tissue defect measuring 4 mm in diameter and 3 mm in depth, and then applied EBC or Floseal to the wound for haemostasis and subsequent regeneration (Fig. 4g). Similar to the previous test, EBC significantly reduced both blood loss and bleeding time (Fig. 4h,i). Although there was no significant difference in haemostatic efficacy compared with Floseal, we proposed that EBC could better promote regeneration and minimize foreign-body reactions. This assumption is based on the use of autologous blood and a minimal amount of biodegradable HA in EBC.

To test this hypothesis, we monitored systemic responses and local regeneration profiles through blood tests and histological analysis over time. The blood test results indicated no effect on RBCs, but an increase in platelets and white blood cells was observed in the Floseal group on day 5 post-surgery (Supplementary Fig. 29). The increase in inflammation-related cells was primarily linked to lymphocytes (Fig. 4j ). Furthermore, we examined physiological parameters related to organ function, including glucose, urea, cholesterol, lipase, total protein, globulins, bilirubin, albumin, alkaline phosphatase (ALP), γ-glutamyltransferase (GGT), alanine transaminase (ALT) and aspartate transaminase (AST) (Extended Data Fig. 8). Most parameters remained within normal levels, indicating normal organ functions despite the liver defect and the presence of biomaterials. Notably, the albumin level, an indicator of liver function, showed a slight decrease on day 5 and recovery on day 10, with the EBC group consistently exhibiting higher levels compared with the Floseal group. This difference may be attributed to variations in liver regeneration rates. These results indicate that the foreign materials applied locally to the injury sites caused systemic inflammation, whereas EBC did not significantly affect blood cell parameters because of its excellent biocompatibility.

We further examined the local liver regeneration profiles through visual examination and histological analysis, revealing significant differences between EBC and Floseal in promoting tissue regeneration (Fig. 4k). Digital photos showed injured livers gradually regenerated with the degradation of EBC or Floseal. On day 14, a visible capsule with newly formed blood vessels was observed in the Floseal group, indicating a typical foreign body response (Extended Data Fig. 9). EBC fully regenerated the injured liver to its pristine state on day 28, whereas scars were found on the Floseal-administrated livers (Fig. 4k). When examining microscope images of the explants, we found that EBC greatly promoted liver regeneration with a fast regeneration rate (78% on day 5, 82% on day 10 and around 100% on day 28), whereas Floseal led to much slower regeneration (20% on day 5, 51% on day 10 and 84% on day 28) (Fig. 4l). Histological analysis of inflammation levels showed that EBC led to very mild to mild inflammation within the 2 weeks post-injury (Fig. 4m), resembling the native wound healing process and facilitating tissue regeneration. By contrast, Floseal caused severe inflammation, as evidenced by a significant accumulation of lymphocytes and some plasmocytes around the material (Extended Data Fig. 9), which is consistent with the haematology test results and likely to impair regeneration. Moreover, a thick fibrosis capsule was observed surrounding Floseal, whereas there was no fibrosis observed with EBC, further highlighting the advantages of the cytogels formed with autologous cells over foreign body responses in engineered materials. Moreover, EBC significantly reduced the occurrence of postoperative adhesion complications compared with Floseal (17% compared with 83%; Fig. 4k,n). Apart from the autologous EBC, we evaluated the efficacy and biosafety of allogeneic EBCs prepared from different rat strains with matched RBC antigens. Histological and morphological analyses of the liver demonstrated that allogeneic EBCs achieved comparable healing, tissue regeneration and minimal inflammation to the autologous condition, confirming their cross-compatibility and safety (Extended Data Fig. 10). Compared with previously reported biomaterials for liver regeneration, EBC demonstrated milder inflammation and more efficient tissue regeneration (Supplementary Table 3). Overall, EBC, as a native scaffolding material, can promote tissue regeneration with minimal inflammation and foreign body responses, and prevent postoperative adhesions, outperforming the clinically used products.

In this study, we established click clotting as a strategy to mechanically integrate cells into tough cytogels and demonstrated the potential of EBC to enhance haemostasis and regeneration. Future studies should validate performance in specific clinical scenarios, particularly involving coagulopathies and other blood disorders, and develop predictive mechanical models that link cytogel architecture and cell rupture to macroscopic mechanical responses. This work opens avenues for designing and translating highly cellularized materials.

Methods

Materials

Reagents and solvents were purchased from commercial sources and were used without any further purification. Methyltetrazine amine (1011), methyltetrazine acid (1125), sulfo-DBCO amine (1227), TCO-PEG 4 -NHS ester (A137), Azido-PEG 4 -NHS ester (AZ103), AZDye 405 cadaverine (1418), AZDye 568 TZ (1363) and Cy5 DBCO (A130) were purchased from Click Chemistry Tools. HA (MW ≈ 2,000 kDa) was from Lyphar Biotech. Alginate (high molecular weight, I-1G) was from KIMICA. Chitosan (high molecular weight, 419419), polyacrylic acid (PAAc, MW ≈ 1,250 kDa, 306215), N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDC, 03450), N-hydroxysuccinimide (NHS, 130672), fluoresceinamine (201626), MES hydrate (M8250), sodium hydroxide solution (S2770), sodium chloride (NaCl, S9888), dimethyl sulfoxide (DMSO, 276855), calcium chloride (CaCl 2 ), dialysis tubing (MWCO 14 kDa, D9527) and citrate-phosphate-dextrose solution with adenine (CPDA, C4431) were purchased from Sigma-Aldrich. Normal human RBCs (pooled sex, CPDA anticoagulant) and normal human whole blood (pooled gender, CPDA anticoagulant) were from BioIVT.

Surface modification of RBCs

The RBC surface was decorated with TCO or azide by NHS ester chemistry. Human RBCs (1.0 ml packed cells) were washed thrice with cold PBS (9 ml, pH = 7.4) and centrifuged (1,000 × g, 5 min). The washed RBCs were then resuspended in 5.0 ml PBS. A stock solution of TCO-PEG 4 -NHS ester (400 μl, 25 mg ml −1 dissolved in DMSO) was diluted into 3.6 ml PBS, and then immediately added to the RBC PBS suspension. The mixture was incubated at 37 °C for 40 min. To remove unreacted reagents, cells were purified by sequential centrifugation (1,000 × g, 5 min) and washing with cold PBS (2 ml, 4 ml, 6 ml, 8 ml), followed by three additional washes with PBS.

Using the same protocol, the RBC surface was also decorated with azide by incubating cells with Azido-PEG 4 -NHS ester. The presence and distribution of TCO on RBCs were detected by a fluorescent probe, AZDye 568 TZ. The modified RBCs were incubated with AZDye 568 TZ (2 μg ml −1 in PBS) for 5 min at room temperature, and washed with PBS thrice. The stained cells were imaged using a confocal laser scanning microscope (Zeiss, LSM710).

Flow-cytometric quantification of TCO groups on RBCs

The amount of TCO on RBC surfaces was quantified using flow cytometry following a previously reported titration method 46 . In brief, a fixed amount of TCO-modified RBCs was added into PBS containing varying concentrations of AZDye 568 TZ, resulting in final TZ-to-RBC ratios ranging from 10 6 to 10 8 . After 5 min of incubation at room temperature, the cells were washed thrice with PBS to remove unbound dye. The fluorescence intensity of the labelled RBCs was then measured using a flow cytometer (BD LSR Fortessa). The measured fluorescence intensity increased with the TZ-to-RBC feeding ratio and plateaued at high ratios, indicating saturation of reactive TCO sites. The number of TZ molecules bound per RBC was estimated from the inflection point of the fluorescence compared with TZ/RBC ratio curve. Given the high efficiency of the TCO-TZ click reaction 31 , the number of bound TZ groups approximates the number of TCO moieties on the RBC surface.

To assess the in vitro stability of TCO on RBCs, the number of TCO groups presenting on RBC was monitored over time. TCO-modified RBCs were stored in CPDA solution at 4 °C for defined time periods. At each time point, the cells were incubated with an excess amount of AZDye 568 TZ (10 8 TZ per RBC) for 5 min at room temperature, followed by three PBS washes. The residual TCO levels were then quantified by flow cytometry as described above.

Synthesis of HA polymer linkers

Sodium hyaluronate (HA; molecular weight about 2,000 kDa; Lyphar Biotech) was used as-received and functionalized with methyltetrazine or dibenzocyclooctyne (DBCO) moieties through carbodiimide-mediated coupling. HA (500 mg, with about 1.25 mmol COOH) was first dissolved in 100 ml of 0.1 M MES buffer containing 0.3 M NaCl (pH 6.0) to afford a 0.5% w/v solution. The solution was stirred at room temperature until fully dissolved. EDC (598 mg, 3.12 mmol, 2.5 equiv. relative to HA carboxyl groups) and NHS (360 mg, 3.12 mmol, 2.5 equiv.) were added sequentially under stirring. After 15 min of activation, methyltetrazine amine (corresponding to 0.25 mmol g −1 , 0.50 mmol g −1 or 0.75 mmol g −1 HA; for example, 125 μmol, 250 μmol or 375 μmol for 500 mg HA) was added dropwise to the activated HA solution to achieve nominal substitution levels of 10%, 20% and 30% (Supplementary Table 1). The reaction mixture was stirred at room temperature for 24 h under ambient atmosphere. The solution was then transferred into dialysis tubing (MWCO 14 kDa) and dialysed against 4 l of 150 mM NaCl solution for 24 h (three changes), followed by dialysis against 4 l deionized water for 4 days (water changed twice daily). The purified solution was frozen and lyophilized using a benchtop freeze dryer (Labconco) to yield HA-TZ as a pink fluffy solid. HA-DBCO conjugates were synthesized using sulfo-DBCO amine under identical activation and purification conditions, using the same molar equivalents relative to HA carboxyl groups.

Preparation of lower-molecular-weight HA-TZ

HA-TZ derivatives were prepared by enzymatic depolymerization using bovine testicular hyaluronidase (400–1,000 U mg −1 , Sigma-Aldrich, H3506). HA-TZ was first synthesized from as-received HA as described above. HA-TZ (200 mg) was dissolved in 40 ml of 0.1 M MES buffer (pH 6.0) to obtain a 0.5 wt% solution. Hyaluronidase (20 mg; final concentration 0.5 mg ml −1 ) was added, and the mixture was stirred at 37 °C for 20 min, 40 min or 60 min. The reaction was quenched by immersion in boiling water for 5 min, followed by rapid cooling in an ice bath. The solution was filtered (0.45 μm), dialysed and lyophilized to afford HA-TZ derivatives with reduced molecular weights, designated HA20-TZ, HA40-TZ and HA60-TZ. Their structure was confirmed with NMR as described above (Supplementary Fig. 30). Hydrodynamic sizes were determined at 0.5 mg ml −1 in PBS (pH 7.4) using dynamic light scattering (Malvern Zetasizer Nano ZS), and number- and weight-average molecular weights were determined by gel permeation chromatography (Extended Data Fig. 2).

Synthesis of fluorescently labelled polymers

Two fluorescent HA-TZ linkers were synthesized for imaging and biodistribution studies: HA-TZ-FITC, labelled with fluoresceinamine for subcutaneous implantation and biodistribution analysis; and HA-TZ-405, labelled with AZDye 405 cadaverine for confocal imaging of EBC.

For HA-TZ-FITC, HA-TZ (200 mg) was dissolved in 40 ml of 0.1 M MES buffer containing 0.3 M NaCl (pH 6.0) to afford a 0.5 wt% solution. EDC (1.0 equiv. relative to –COOH) and NHS (1.0 equiv.) were added to the solution and allowed to activate for 30 min under stirring. Separately, fluoresceinamine was dissolved in DMSO and added dropwise to the activated HA-TZ solution at a molar ratio corresponding to 2% of the available –COOH groups of HA. The reaction was maintained for 24 h at room temperature in the dark. The resulting solution was dialysed sequentially against 1:1 (v/v) ethanol/water for 2 days and then against deionized water for 3 days, followed by lyophilization to obtain the green-fluorescent polymer HA-TZ-FITC. For the preparation of HA-TZ-405, AZDye 405 cadaverine was added to a 0.5 wt% HA-TZ solution at a molar ratio of 1% (amine to carboxyl), together with EDC and NHS at 2.5-fold molar excess. The reaction was carried out for 24 h at room temperature under light protection, followed by dialysis and lyophilization to yield the blue-fluorescent conjugate HA-TZ-405.

Synthesis of other polymer linkers

Other polymer linkers, including alginate (Alg), chitosan (Chi), poly(acrylic acid) (PAAc) and four-arm poly(ethylene glycol)-amine (PEG), were functionalized with methyltetrazine groups through the same EDC/NHS coupling reactions as described above. Alginate (0.5 wt%) and PAAc (0.3 wt%) were reacted with methyltetrazine amine at 0.75 mmol g −1 polymer in 0.1 M MES buffer (pH 6.0). Chitosan (0.3 wt%) was functionalized with methyltetrazine acetic acid at 0.75 mmol g −1 polymer in MES buffer adjusted to pH 5.5. Four-arm PEG-amine (10 kDa, 2 wt%) was reacted with methyltetrazine acetic acid at 1.5 mol equiv. per amine group under identical EDC/NHS coupling conditions. All reactions were stirred for 24 h at room temperature, followed by dialysis against deionized water with decreasing NaCl gradients (from 150 mM to 0 mM) and lyophilization.

NMR measurements

The degree of substitution was determined by 1 H NMR spectroscopy (400 MHz, D 2 O). For HA-TZ, the aromatic protons of the phenyl ring of methyltetrazine (about 8.2 ppm, 2H) were integrated relative to the acetyl methyl resonances (about 2.0 ppm, 3H) of HA backbone. For HA-DBCO, the DBCO aromatic envelope (7.80–6.80 ppm, 8H) was integrated relative to the acetyl methyl resonances (about 2.0 ppm, 3H) of HA backbone. For Chi-TZ, the TZ aromatic resonances (about 8.20 ppm, 2H) were integrated relative to sugar ring protons H2 to H6 (3.0–4.0 ppm, 6H) of the chitosan. For Alg-TZ, the TZ aromatic resonances (about 8.20 ppm, 2H) were integrated relative to the alginate anomeric H1 (4.80–5.00 ppm, 1H) region as the internal reference. For PAAc-TZ, the TZ aromatic resonances (about 8.20 ppm, 2H) were integrated relative to the methine resonances (2.2–2.4 ppm, 1H) of the polyacrylic acid.

Preparation of RBC cytogels and control hydrogels

To prepare RBC cytogels by click clotting, polymer linkers were dissolved in PBS at a concentration of 1–3% (w/v). Modified RBCs were collected by centrifugation and transferred into a Luer-lock syringe. The polymer solution was loaded into a second syringe and connected using a female–female Luer adaptor. The two solutions were syringe-mixed until homogeneous. For example, mixing 1.5% (w/v) HA-TZ with modified RBC suspension at a 2:1 (v/v) ratio yielded a final polymer concentration of 1% (w/v) and an RBC volume fraction of 33%. The resulting cytogel formed immediately and was extruded for subsequent testing.

For comparison, two RBC-containing hydrogel controls were prepared. In one control, unmodified RBCs were mixed with HA solution (1% w/v) containing freshly added EDC (12 mg ml −1 ) and NHS (12 mg ml −1 ) to allow carbodiimide-mediated crosslinking. In another control, two HA polymers pre-functionalized with TZ or TCO were synthesized as described above. They are syringe-mixed, crosslinked into an HA hydrogel with unmodified RBCs physically encapsulated inside at volume fractions of 5% or 33%. All control samples were incubated at room temperature for 1 h before characterization.

Cell culture and preparation of other cytogels

Immortalized human vocal fold fibroblasts (hVFFs) were cultured in Dulbecco’s modified Eagle medium (DMEM, Corning) supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin and 1% MEM nonessential amino acids. Human adipose-derived stem cells (ASCs, Cyagen, Cat. HUXMD-01001) were cultured in DMEM/F12 basal media supplemented with 10% (v/v) FBS, 1% (v/v) penicillin/streptomycin, 2.5 μg ml −1 amphotericin B and 5 ng ml −1 recombinant human basic fibroblast growth factor 2 (rhFGF2). Cells were incubated at 37 °C, in a 5% CO 2 humidified atmosphere. The media were changed every 3 days. Cells were disassociated using 0.25% trypsin-ethylenediaminetetraacetic acid (EDTA) when the cell confluency reached 80%. Microalga Chlamydomonas reinhardtii (+) was from Carolina Biological Supply and cultured in a standard TAP medium at room temperature under continuous light illumination 47 . Microalga cells were harvested by centrifuge (1,000 × g, 5 min). Various cells were surface-modified with TCO and then mixed with HA-TZ solution to achieve cytogels using the same method for preparing RBC cytogels.

Preparation of EBC

EBC was prepared by forming a cellular network of RBCs integrated with a fibrous polymer network of fibrin and other blood components. Human whole blood was first processed by substituting native RBCs with the modified RBCs. To initiate the native and click clotting simultaneously, whole blood was recalcified by using calcium chloride (CaCl 2 , final concentration 30 mM) and then mixed with an equal volume of polymer solution with varied concentration (for example, 2% w/v HA-TZ).

Rheological measurement

The mixed gels were immediately ejected from the syringe onto the bottom plate of a torsional rheometer equipped with a 20 mm flat upper plate geometry (DHR-2, TA Instruments). Mineral oil was applied to the gel periphery to prevent the gel from dehydration during the test. The storage and loss moduli (G′ and G″) of gels made of various polymer species, degree of TZ substitution, polymer concentration and RBC ratio were measured using 0.1% strain at 1 Hz.

Confocal imaging

To visualize the structure of EBC, the RBCs were labelled with Cy5 by click chemistry, and the synthesized fluorescent polymer HA-TZ-405 and Fibrinogen-Alexa Fluor 488 Conjugate were used. First, human RBCs were grafted with TCO and azide by sequence. The modified RBCs were then incubated with DBCO-Cy5 to be labelled with fluorescence by DBCO-azide ligation. Labelled RBCs with TCO moiety were resuspended with platelet-rich plasma with the addition of Fibrinogen-Alexa Fluo 488 (5%) and calcium chloride, then mixed with HA-TZ-405 solution to form EBC. Unmodified RBCs were used as a control. The structure was imaged by using confocal laser scanning microscopy (Zeiss, LSM710) under the DAPI, AF488 and Cy5 channels. Stack images were also obtained to construct the three-dimensional structure of EBC by using the software ZEISS Zen.

SEM imaging

The morphology of materials was observed using a field emission scanning electron microscope (FE450, FEI) with an accelerating voltage of 10 kV. The prepared EBC and other materials, including un-crosslinked EBC, NBC and RBC cytogel, were fixed with 4% paraformaldehyde and then dehydrated with gradient ethanol (30–100%). The dry samples were coated with 4 nm platinum using a high-resolution sputter coater (ACE600, Leica) to increase surface conductivity.

Fracture toughness measurement

Fracture toughness of RBC cytogels and EBC was measured by modified lap-shear tests. To prepare RBC cytogels of defined shape yet with ultrafast gelation kinetics, the modified RBCs and polymer precursors were pre-cooled at 4 °C in the fridge for 30 min before mixing to slow down the reaction rate. The mixture was immediately poured into a PTFE mould of 20 × 15 × 1.5 mm 3 and covered with a PTFE sheet. After 10 min at room temperature, the RBC cytogel was taken out of the moulds and glued with two thin PET films on each side using Krazy glue. The length of overlapping joint is 20 mm. An initial edge crack of 5 mm was made with a blade in the middle of each sample. The samples were tested using an Instron machine (model 5965) equipped with a 10 N load cell. The displacement rate was set at 2 mm min −1 , unless otherwise specified. EBC samples were prepared in the same mould, incubated at 37 °C for 1 h to complete native clotting and tested using the same method. For comparison, NBCs, Floseal haemostatic matrix (Baxter) infused with NBCs and blood/chitosan clots were prepared and measured using the same method. The fracture toughness was calculated by the area under the shear force–displacement curve up to the failure point normalized by the overlapping joint area.

Tensile test of RBC cytogels

To measure the tensile behaviour under cyclic loading, the RBC cytogels were prepared into strips in a PTFE mould with dimensions of 30 × 8 × 1.5 mm 3 . The accurate size of samples was measured after demoulding. Samples were tested by an Instron machine (10 N load cell) with a displacement rate of 3 mm min −1 . Four tensile cycles were applied to samples at each stretch ratio. The nominal stress was obtained by dividing the force by the cross-sectional area. The nominal strain was obtained by dividing the change in length by the original length.

Measurement of haemoglobin released from ruptured RBCs

Haemoglobin was released from stretched RBC cytogels owing to the cell rupture. The release amount was quantified for each RBC cytogel that underwent tensile tests. In brief, after the tensile test, RBC cytogels were cut into cubes and weighed for the mass (around 50 mg). The gel pieces were immersed in 2 ml PBS and incubated overnight at 4 °C in the fridge. After centrifuging, the absorbance at 540 nm of the supernatant was measured by using a microplate reader (Synergy HTX, Agilent). The result was reported by normalizing the absorbance to the mass of RBC cytogels.

Finite-element modelling of lap-shear test

To study the toughening mechanism based on the Mullins effect, we develop a finite-element model to simulate the modified lap-shear test of the RBC cytogels. Extended Data Fig. 3 shows the geometry of the analysed problem. An initial crack is introduced to an RBC cytogel, which is deformable and dissipative. The RBC cytogel is damaged on loading, for example, by ruptures of RBCs. These damages lead to the loss of modulus of the RBC cytogel, giving rise to the softening behaviour along the unloading path and forming a hysteresis loop. This behaviour is referred to as the Mullins effect. Here, we use the combined neo-Hookean and modified Ogden–Roxburgh model to capture the Mullins effect 48 , 49 . In the model, the softening behaviour is empirically described by a pseudo-elastic energy function W(F, η):

(1)

where hatW is the free energy without the Mullins effect; ϕ(η) is the damage function in the form of phi (eta )=int 1 eta (m+beta W max )erf -1 [r(1-eta )] -W max deta, with W max being the maximum strain energy density stored in the material before unloading and erf being the error function; r, m and β are the Mullins coefficients that characterize the energy dissipation as detailed below. In equation (1), F is the deformation gradient tensor and η is a damage variable, which can be expressed as

(2)

where W max is the maximum strain energy stored in the material upon loading, and hatW is the strain energy without considering the Mullins effect. We adopt the neo-Hookean free energy function for hatW:

(3)

where I 1 =lambda 1 2 +lambda 2 2 +lambda 3 2, λ i is the ith principal stretch, and mu is the shear modulus. The shear modulus is taken to be μ = 0.88 kPa, estimated by the slope of the shear stress–shear strain curve in Fig. 2f. The gel is assumed to be incompressible, so the Young’s modulus is thrice the shear modulus, E = 3μ.

Owing to the Mullins effect, only part of the strain energy becomes recoverable on unloading, whereas the rest is dissipated because of the Mullins effect. As shown in Extended Data Fig. 3, the area of the hysteresis loop is the dissipated energy density defined by U D and the shaded area under the loading path is the total mechanical work done on a unit volume before unloading defined by U. Based on the modified Ogden–Roxburgh model 45 , 46 , the hysteresis ratio is defined as

(4)

On evaluating U D and U from our data in Fig. 2j, we establish h as a function of U for the RBC cytogels (Extended Data Fig. 3e). By fitting equation (4) to the data, we can obtain the Mullins coefficients for the RBC cytogels, which will be input into finite-element modelling to capture the energy dissipation during fracture process. Extended Data Fig. 3e shows the h–U relation for tough gels and porcine skin, extracted from literature 49 . Compared with other dissipative materials, the RBC cytogels show higher rate-independent energy dissipation capacity at lower mechanical energy input, indicated by the high h values at small U values 50 .

The fracture of the RBC cytogel is modelled using a bilinear triangular cohesive zone model (Extended Data Fig. 3b), which has been widely implemented to study the coupling between the interfacial properties and the bulk deformation 51 , 52 . We define the intrinsic fracture toughness Γ 0 using the interfacial strength σ max and the maximum separation distance δ f using

(5)

The interfacial strength σ max is difficult to measure precisely in experiments. We treat σ max as a fitting parameter and vary it between approximately 1 kPa and 10 kPa. This range is chosen according to the representative range of adhesive strength reported for tissue adhesives using the lab-shear test 53 , 54 . We use the commercial FE package Abaqus (6.14, SIMULIA) to carry out the finite-element simulations. The problem is simplified to be two-dimensional by adopting a plane-strain condition. We add rigid backings to the bottom and the top surfaces in the simulation (Extended Data Fig. 3a). The top backing is subjected to a constant velocity boundary condition in the direction parallel to the interface. The bottom surface is constrained in a frictionless manner, whereas one node near the left boundary is pinned to prevent the rigid-body motion. The fracture energy can thus be determined using the force–displacement curve F(δ):

(6)

where A is the bonded area. The bulk of the RBC cytogel is modelled with CPE4R element, and the interface is modelled with COH2D4 element. The explicit solver in Abaqus is used to simulate the lap-shear process.

We found that the combination of Γ 0 = 3 J m −2 and σ max = 15.4 kPa yields Γ ≈ 15 J m −2 and shows a good agreement between the experimental and computational force–displacement curves of lap-shear test in Extended Data Fig. 4a. We then remove the Mullins effect in the model and find that the crack propagates at a much lower critical displacement and force (Extended Data Fig. 4b), yielding a lower fracture toughness Γ ≈ Γ 0 = 3 J m −2 . Comparing the models with and without the Mullins effect, the one with Mullins effect dissipates considerable amount of energy during the clot deformation (Extended Data Fig. 4, insets), causing a higher fracture toughness. The use of finite-element models shows the Mullins effect caused by the RBC rupture as the key toughening mechanism in RBC cytogels.

Adhesion energy measurement

T-peel test was performed with an Instron machine (Model 5965, 10-N load cell) to measure the adhesion energy between blood clots and biological tissues. Fresh porcine skin from a local grocery store was cut into strips, and the skin surface was cleaned with deionized water. An elastomer mould with an inner dimension of 30 mm in length, 10 mm in width and 1.5 mm in thickness was attached onto the skin with Vaseline applied to the interface for complete sealing. Precursor solutions of EBC or recalcified human whole blood were poured into the moulds. A porous fabric saturated with recalcified whole blood was put on whole blood specimens as a rigid backing. The samples were left for coagulation in a sealed and humidified chamber at 37 °C for 1 h. The elastomer moulds were carefully removed without disturbing the clot–tissue specimens. Thin polyethylene terephthalate (PET) films were adhered to EBC and skin by using Krazy glue as a rigid backing. An initial crack of 5 mm was created between the clot and skin. The specimens were then gripped to the Instron machine and peeled at a displacement of 100 mm min −1 . The adhesion energy of the specimen was calculated by dividing twice the average force at the plateau (F avg ) by the width of the specimen: Γ = 2F avg /W.

Burst pressure measurement

A burst chamber was manufactured according to the ASTM F2392: standard test method for burst strength of surgical sealants. Collagen casing was used as a model tissue. A biopsy punch was used to create a 3-mm diameter hole in the centre of the collagen casing. Tested materials (EBCs, NBCs, RBC cytogels and Floseal blood clot gels) were prepared into a disc shape with a diameter of 15 mm and a thickness of 3 mm. The clot samples were glued to the collagen casing using Krazy glue to test the cohesive burst pressure of the materials. A custom-made elastomer ring was placed onto the samples to avoid leaking of the burst chamber. During the test, a syringe filled with PBS was connected to both the burst chamber and a pressure gauge. A syringe pump was used to feed the PBS to the chamber. The peak pressure before the cohesive failure was recorded and used as the burst pressure.

Tensile and compressive tests

The mechanical tests were performed using an Instron universal testing machine. For compression tests, disc-shaped samples (10 mm in height and 7 mm in diameter) were subjected to unconfined monotonic compression until failure. For tensile tests, rectangular samples (30 mm in length, 8 mm in width and 1.5 mm in thickness) were tested under uniaxial tension. For cyclic loading tests, the same rectangular samples were repeatedly stretched between 0% and a maximum strain (10%, 20% and 40%) for up to five cycles. The recovery ratio was calculated as the ratio of the mechanical work during unloading to that during loading in the first cycle (W unload /W load ).

In vitro haemolysis test

The haemolysis test was performed as the first indicator of blood compatibility. To evaluate whether the polymer or cell crosslinking caused haemolysis, 100 μl of 1% HA-TZ solution was mixed with 100 μl native RBCs or modified RBCs (RBC-TCO) for 15 min. PBS (10 ml) was then added to the mixture and incubated at 37 °C for 30 min. For comparison, a chitosan solution was also mixed with RBCs in the same way. Native RBCs (100 μl) were added to deionized water to achieve complete haemolysis as the positive control. Native RBCs (100 μl) were added to PBS as the negative control. The mixtures were centrifuged (1,000 × g, 5 min), and the absorbance at 540 nm of the supernatant was measured by using a microplate reader. The percentage of haemolysis was calculated according to the following equation: % haemolysis = (Abs sample − Abs negative )/Abs positive × 100%.

In vitro human blood immunity test

The immune compatibility of EBCs was evaluated using human whole blood and peripheral blood immune cells. For the autologous EBC test, both the EBCs and the corresponding PBMCs were prepared from the same pooled donor blood purchased commercially. Two experimental conditions were included: (1) EBC-treated and (2) blank (non-treatment) control. In brief, EBCs (approximately 10 × 10 × 1.5 mm 3 ) were prepared from human whole blood as described above. Each sample was first incubated with 10 ml human whole blood under agitation at 37 °C for 10 min, and then removed to obtain the EBC-conditioned whole blood. The conditioned blood was diluted with an equal volume of PBS (pH 7.4) and separated by density-gradient centrifugation over Lymphoprep (Stemcell Technologies). The mononuclear cell layer was collected, washed twice in PBS with 2% FBS and counted using a trypan blue exclusion method.

The allogeneic EBC test was performed with freshly isolated human blood. Whole blood from a type O donor (donor A) was used to prepare both EBCs and unmodified RBC controls, whereas ABO/Rh-matched whole blood from another healthy type O donor (donor B) was used to isolate PBMCs. PBMCs were obtained by Ficoll density-gradient centrifugation, washed twice with PBS and resuspended in complete RPMI 1640 medium supplemented with 10% fetal bovine serum, 2 mmol l-glutamine and penicillin–streptomycin. Cell viability exceeded 95% by trypan blue exclusion. Three experimental conditions were included: (1) PBMCs alone (non-treatment control); (2) PBMCs co-incubated with EBC (EBC group); and (3) PBMCs co-incubated with unmodified RBCs (RBC group). For each condition, one equilibrated EBC (0.3 ml; 2% hyaluronic acid with modified RBC and PRP in the ratio of 2:1:2) or an equivalent volume of unmodified RBCs was placed in a single well containing 1 ml of complete medium. PBMCs were added at 1.0–1.5 × 10 6 cells per well and incubated at 37 °C with 5% CO 2 for 4 h or 16 h without disturbance. Positive activation controls were prepared by stimulating PBMCs with lipopolysaccharide (LPS, 100 ng ml −1 , Escherichia coli) to activate monocytes, or with phorbol 12-myristate 13-acetate (PMA, 50 ng ml −1 ) plus ionomycin (1 µg ml −1 ) for broad lymphocyte activation. All cultures were carried out in low-adhesion plates to minimize artefactual monocyte adherence. After incubation, the cells were gently collected and washed in PBS.

For flow cytometry staining, the cells were divided into triplicates of 2 × 10 6 cells. The cells were pre-incubated with eBioscience Fixable Viability Dye eFluor 506 (Invitrogen, 1:1000) and Human TruStain FcX (BioLegend, 1:50) in PBS for Fc receptor blocking and viability staining for 20 min on ice. The cells were then stained for surface markers in PBS with 2% FBS for 30 min on ice, using the following antibodies: PE-conjugated anti-human CD3 (UCHT1, BioLegend, 1:50), anti-human CD16 (B73.1, BioLegend, 1:20) and anti-human CD62P P-selectin (AK4, BioLegend, 1:20); BV785-conjugated anti-human CD4 (RPA-T4, BioLegend, 1:40); APC-Cy7-conjugated anti-human CD8a (HIT8a, BioLegend, 1:20); FITC-conjugated anti-human CD20 (2H7, BioLegend, 1:20) and anti-human CD41 (HIP8, BioLegend, 1:20); BUV395-conjugated anti-human CD56 (NCAM16.2, BD Biosciences, 1:20); BV650-conjugated anti-human CD14 (M5E2, BioLegend, 1:20); BV421-conjugated anti-human CD45R (HI100, BioLegend, 1:20); PE-Cy7-conjugated anti-human HLA-DR (L243, BioLegend, 1:60); APC-conjugated anti-human CD86 (BU63, BioLegend, 1:20) and anti-human CD41/CD61 (A2A9/6, BioLegend, 1:20); BUV737-conjugated anti-human CD3 (UCHT1, BD Biosciences, 1:20) and anti-human CD16 (3G8, BD Bioscience, 1:60); PerCPCy5.5-conjugated anti-mouse/anti-human CD44 (IM7, BioLegend, 1: 80); and Pacific Blue-conjugated anti-human CD11b (ICRF44, BioLegend, 1:30), using the eBioscience Super Bright Complete Staining Buffer (Invitrogen, 1:50).

The samples were analysed by flow cytometry, gating on monocytes (CD14+CD20−CD3−CD56−), non-classical monocytes and granulocytes (CD16+CD14−CD20−CD3−CD56−), B lymphocytes (CD20+CD3−), CD4 T lymphocytes (CD3+CD4+CD20−), CD8 T lymphocytes (CD3 + CD8 + CD20-), NK cells (CD56 + CD20-CD3−), and platelets (CD41 + ). Expression of activation markers on each cell type was measured as the median fluorescence intensity. Viability Dye eFluor 506 (eBioscience) was used to discriminate dead cells. Compensation was performed with BD CompBeads (BD Biosciences) or with UltraComp eBeads Plus (01−3333-42, Thermo Fisher). The data were acquired on BD Fortessa (BD Biosciences) or Cytek Aurora; the data were analysed with SpectroFlo (Cytek Biosciences) or FlowJo (BD Biosciences) software.

In vitro cytocompatibility test

The cytotoxicity of EBC was evaluated using immortalized human vocal fold fibroblasts, following the International Organization for Standardization (ISO) 10993-5: Tests for in vitro cytotoxicity. For every 200 mg of EBC, 1 ml of DMEM was added for extraction. Meanwhile, 20,000 cells per well were seeded in a 96-well plate. The extracts, after a 24-h incubation period, were collected and supplemented with 1% penicillin–streptomycin, and with or without 10% fetal bovine serum. The culture medium within the 96-well plate was then replaced with the supplemented EBC extracts. Completed DMEM was used as the control. The cells were then cultured for 24 h inside an incubator, in an environment at 37 °C, 95% relative humidity and 5% CO 2 atmosphere. Cell viability was assessed using a Live/Dead viability kit (Invitrogen, L3224), following the protocol of the manufacturer. Confocal laser scanning microscopy (Zeiss, LSM710) was used for investigation. Live cells were visualized in green and dead cells in red.

Ethics statement

Human ASCs were commercially obtained from Cyagen (Cat. HUXMD-01001). The use of human ASCs and human blood was approved by the Research Ethics Office of McGill University (protocol nos. A03-M11-21A and A05-M46-20B).

In vivo biocompatibility, biodegradation and biodistribution

All in vivo experiments were approved by the McGill University Animal Care Committee (protocol no. 2019-8098) and performed according to the guidelines of the Canadian Council on Animal Care. Female Sprague Dawley rats (250–300 g, 8−10 weeks) were purchased from Charles River Laboratories and used for all the in vivo studies.

Dorsal subcutaneous implantation was used to evaluate the in vivo biocompatibility, biodegradation and biodistribution of RBC cytogel and EBC. Autologous whole blood from the rat saphenous vein was collected to prepare the RBC cytogel and EBC for implantation. The rat was restrained, and the hind leg was immobilized in the extended position. The collection site was shaved with an electric clipper and cleaned with 70% alcohol. Blood was collected in a BD EDTA blood collection tube by using a 23-gauge needle from the lateral saphenous vein. Blood flow is stopped by applying pressure with sterile gauze to achieve haemostasis. The collected whole blood was processed, and then combined with the fluorescently labelled polymer (HA-TZ-FITC) to fabricate RBC cytogel and EBC. Samples were prepared into a disk shape of 6 mm in diameter and 1.5 mm in thickness using aseptic techniques. For implantation, the rats were anaesthetized using isoflurane (4% isoflurane in oxygen) in an induction chamber. Anaesthesia was maintained at 2% isoflurane using a nose cone during the surgery. A volume of 1 ml of saline was injected subcutaneously. Dorsal hair was removed, and the rats were placed over a heating pad for the duration of the surgery. The subcutaneous space was accessed by a 1 cm skin incision per implant at the back of the rat. Blunt dissection was performed from the incision point towards the shoulder blades of the rat to create a subcutaneous space for implantation. RBC cytogel and EBC were implanted into the subcutaneous spaces, and NBCs and Floseal were used as a control. Four implants were placed per rat, and the skin incisions were closed with sutures. On days 3, 7, 14, 28 and 56, the rats were euthanized by 5% isoflurane induction followed by CO 2 inhalation. Subcutaneous regions of interest were excised. The size of the implants was measured and normalized to the initial size to represent the in vivo degradation. Excised samples were then fixed in 4% paraformaldehyde solution and embedded in paraffin for histological analysis to evaluate the biocompatibility.

On days 3, 7, 14, 21 and 28, vital organs, including heart, lungs, liver, kidneys and spleen, were harvested. The biodistribution of the degradation product (that is, HA-TZ-FITC) was evaluated by visualizing the fluorescence in those organs using the IVIS spectrofluorophotometer in vivo imaging. To determine systemic toxicity, on day 28, vital organs were collected, formalin-fixed, and paraffin-embedded for further histological analysis.

In vivo rat haemorrhage model of liver incision

For haemostatic sealing of the deep incisional hepatic injury, the rats were anaesthetized using isoflurane (4% isoflurane in oxygen) in an induction chamber. Anaesthesia was maintained at 2% isoflurane using a nose cone during the surgery. A volume of 1 ml of saline was injected subcutaneously. Abdominal hair was removed, and the rats were placed over a heating pad for the duration of the surgery. The liver was exposed by a laparotomy. A laceration wound of 7 mm length and 3 mm depth was made to the liver using a #11 scalpel. Immediately after wiping off the blood using gauze, around 500 μl of EBC or commercial haemostatic matrix Floseal was applied to the injury site. The time to haemostasis was recorded by a stopwatch. The amount of blood loss until bleeding stops was determined by measuring the weight of blood lost and absorbed to a pre-weighed gauze. After the surgery, the rats were euthanized by 5% isoflurane induction followed by CO 2 inhalation.

In vivo rat haemorrhage model of tail amputation

For evaluation of haemostatic performance after tail transection, the rats were anaesthetized with isoflurane (4% in oxygen) in an induction chamber and maintained at 2% using a nose cone. The animals were placed on a heating pad, and the distal tail was disinfected with 70% ethanol followed by povidone–iodine. Using a sterile scissor, a transverse amputation was performed at 3 cm proximal to the tail tip (about 4 mm in diameter) to create a standardized circular bleeding wound. Immediately after gently blotting excess blood with gauze, about 500 μl of EBC was applied to the cut surface, or the wound was left untreated as a control. Time to haemostasis was recorded with a stopwatch. Blood loss until bleeding ceased was determined by weighing pre-weighed gauze and absorbent pads used to collect shed blood beneath the tail. After data collection, the rats were euthanized by 5% isoflurane induction followed by CO 2 inhalation.

In vivo rat haemorrhage model of liver puncture

A standardized rat liver puncture model was used to evaluate the haemostatic performance and regenerative potential of EBCs. Following anaesthesia and laparotomy, a volumetric injury measuring 4 mm in diameter and 3 mm in depth was created on the liver surface using a sterile biopsy punch. After gentle blotting of excess blood with gauze, approximately 500 μl of EBC or a commercial haemostatic matrix (Floseal, Baxter) was applied directly to the bleeding site. Both autologous and allogeneic EBCs were examined using two rodent models as described below.

Autologous model with Sprague Dawley rats

Adult, male Sprague Dawley rats (250–300 g) were anaesthetized with 4% isoflurane and maintained under 2% isoflurane during surgery. After a midline laparotomy, a volumetric injury was created on the liver lobe using a sterile biopsy punch. Immediately after gentle blotting of excess blood with gauze, approximately 500 μl of autologous EBC or a commercial haemostatic matrix (Floseal, Baxter) was applied to the wound bed. The time to haemostasis was measured using a stopwatch from the moment of injury to complete cessation of bleeding. Blood loss was quantified by weighing pre-weighed gauze before and after the procedure. After achieving haemostasis, the treated liver was repositioned, and the abdominal incision was closed with 4-0 absorbable sutures. The animals were allowed to recover and monitored postoperatively. Blood samples were collected from the saphenous vein on postoperative days 5, 10, 14 and 28. Whole blood was used for haematology, and serum was used for biochemical analyses (such as ALT and AST). At each designated time point (days 5, 14 and 28), the animals were euthanized with 5% isoflurane induction followed by CO 2 inhalation. The liver containing the haemostatic site was excised and fixed in 4% paraformaldehyde for 48 h before histological processing.

Allogeneic model with mixed-strain rats

Fischer 344 rats (donors) and Brown Norway rats (recipients) with matched erythrocyte antigens were used as a mixed-strain allogeneic model. The whole blood of the donor was collected by cardiac puncture, and RBCs were isolated and modified with TCO click reagents following the same protocol as for autologous samples. The recipient rats underwent the same liver puncture injury procedure described above. Allogeneic EBCs prepared from Fischer 344 blood of the donor were immediately applied to the wound site of Brown Norway recipients (about 500 μl per site). The surgical closure and postoperative monitoring followed the same protocol as for the autologous EBC model. The liver samples were harvested on postoperative days 5, 14 and 28, fixed in 4% paraformaldehyde, and processed for histological evaluation. Gross morphology and H&E (haematoxylin and eosin) staining were examined to assess wound healing, inflammation and potential immune response. Consistent tissue regeneration and minimal inflammatory infiltration indicated the biocompatibility of ABO/Rh-matched allogeneic EBCs.

Additional discussion on preparation and immune responses of EBCs

Both autologous and allogeneic EBCs were evaluated for their safety and feasibility for clinical applications in humans. Autologous EBCs, prepared from the patient’s own blood, offer superior biocompatibility, avoid the need for cross-matching, and potentially simplify regulatory approval (similar to autologous PRP therapies). RBCs, which are typically discarded during PRP preparation, can be repurposed for EBC formation, and the entire preparation can be completed within 30 min (Supplementary Scheme 2), fitting within procedural windows for surgical and emergency interventions. For allogeneic EBCs, prepared using ABO/Rh-matched packed red blood cells (PRBCs) and platelet-poor plasma (PPP) from the blood bank, can be completed in approximately 10 min and align with established transfusion practices, making them suitable for trauma and emergency care. Importantly, EBC formation relies on a click clotting mechanism independent of the native coagulation cascade and does not require platelets. Unlike PRP, both PPP and PRBCs are standard, stable blood bank components, facilitating seamless integration into existing clinical protocols.

The immune compatibility is an important consideration for allogeneic application. Our in vitro experiments using human whole blood peripheral blood mononuclear cells (PBMCs) show that neither autologous nor allogeneic EBCs induced activation of T cells (CD4+ and CD8+), B cells (HLA-DR+ and CD86+) or natural killer (NK) cells (CD44+ and CD16+) after co-incubation with human immune cells. By contrast, these lymphocyte subsets responded robustly to LPS and to PMA plus ionomycin, confirming preserved immune responsiveness. The lack of upregulation of costimulatory, activation or cytotoxic markers in the EBC-treated groups indicates that EBCs do not trigger adaptive or innate lymphocyte activation, even with prolonged exposure. Monocytes exhibited only a transient and modest increase in activation markers at 4 h, which resolved by 16 h, consistent with a mild, self-limiting response expected for short-term blood–material contact. Together, these results demonstrate that both autologous and allogeneic EBCs maintain excellent haemocompatibility, supporting their use as haemostatic biomaterials without provoking inflammatory or immune sensitization. Although further preclinical and regulatory validation is warranted, these findings underscore the translational potential of both preparation pathways.

Histology and analysis

Fixed tissue samples were placed in 70% ethanol and submitted for histological processing and H&E staining at the Histology Core at McGill University. Z.-h.G., the pathologist-in-chief at the University of British Columbia, examined all histological sections. Representative images of each group are shown in the corresponding figures.

Statistical analysis

A sample size of n ≥ 3 was used for all experiments. Data are shown as mean ± s.d. Statistical analysis was performed using two-sided one-way analysis of variance (ANOVA) and post hoc Tukey tests for multiple comparisons or Student’s t-tests for comparison between two groups (Prism 9). P < 0.05 was considered statistically significant.

登录APP查看全文