Microchannelled chitosan sponge grafting with hydrophobic alkyl chain for enhanced noncompressible hemorrhage and tissue regeneration

Developing an anti-infective shape-memory hemostatic sponge with ability of guiding in situ tissue regeneration for noncompressible hemorrhage in civilian and battleeld settings remains a challenge. Here, hemostatic chitosan sponge with highly interconnective microchannels was engineered by combining 3D printed ber leaching and freeze-drying methods and then modied with hydrophobic alkyl chains. The microchannelled alkylated chitosan sponge (MACS) exhibited a strong capacity for water/blood absorption and rapid shape recovery. Compared to clinically used gauze, gelatin sponge, CELOX, and CELOX-gauze, the MACS demonstrated higher pro-coagulant and hemostatic capacities in lethally normal/heparinized rat and pig liver perforation models. Also, it exhibited strong anti-infective activity against S. aureus and E. coli. Additionally, it promoted liver parenchymal cell inltration, vascularization, and tissue integration in a rat liver defect model. Overall, the MACS demonstrated promising clinical translational potential in cost effectively treating lethal noncompressible hemorrhage and in facilitating wound healing.


Introduction
Hypotension and multi-organ failure caused by massive blood loss often results in high mortality in civilian and military populations 1,2 . So, rapid and e cient hemorrhage control is of paramount importance in such scenarios. The body's natural coagulation cascade process is activated in response to bleeding, but, incapable of timely stopping severe hemorrhage from a deep and noncompressible perforation wound in the absence of shape-memory hemostats 3,4 . Thus, the development of shapememory hemostats is urgently needed. In general, ideal shape-memory hemostats should possess several properties, including a highly interconnected porous structure, active coagulation, strong antiinfection activity, biocompatibility, biodegradability, ready availability, low weight, and low cost 4,5,6,7 . Notably, an interconnected porous structure permits uid to ow freely in and out of hemostats, which allows the hemostats to be xed by draining off the free water and promotes fast recovery to their initial shapes by absorbing the uid 6 . Rapid-shape recovery timely exerts pressure on the wound, leading to effective hemorrhage control 6,8 . Moreover, hemostats left in the injury site and used in directly guided in situ tissue regeneration are more practical for clinical application 9 .
Until now, several shape-memory hemostats have been developed, and some have been applied in clinical practice 10,11,12,13 . For instance, the XStat TM device composed of multiple compressed cellulose sponges was shown to rapidly expand to ll and exert pressure on a wound to control hemorrhage 10 . However, it took much more time to take out each sponge from the wound bed due to its nondegradable property, which may cause patient discomfort 8 . Moreover, such sponge lacking highly interconnected porous structure was incapable of guiding tissue repair. Many shape-memory polymer foams as hemostats have been applied to treat noncompressible hemorrhage and exhibited a certain degree of hemostatic ability 11,12,13 . However, they displayed limited absorption of blood and required decades of seconds to restore their shapes, which may cause the prologation of hemostasis time and more blood loss 5 . Injectable cryogels with high blood absorbability and rapid-shape recovery capacity have also been developed for treatment of noncompressible hemorrhage 6,14,15 . The hemostatic effect of these materials was achieved by restoring shape and applying mechanical compression on the wound. Shape-recovery property mainly originates from the reversible change of porous structure 10,11,12,13 . However, the pores inside these hemostats generated by gas foaming or ice crystal removing methods possess low interconnectivity, which might slow down the blood ow into hemostats, resulting in weakened hemostatic e ciency. The effect of pore structure, especially interconnectivity, on hemostatic performance was usually ignored in the design and construction of hemostats 5,8,10,14 . Besides, some of these hemostats lacked strong active pro-coagulant and anti-infective properties, which may result in their failure to complete the hemostasis in a timely and effective way and in their inability to protect wounds from bacterial infection. Therefore, simultaneously regulating pore structure and active modi cation is expected to improve the hemostatic and anti-infective effects of these hemostats.
Incorporating a microchannel into three-dimensional (3D) constructs is a simple and controllable architectural feature, and capable of promoting transport of nutrients, oxygen, and metabolites, host cell in ltration, vascularization, and integration with the surrounding tissue 16,17,18,19 . To create an embedded and hollow microchannel, the sacri cial brous template with a well-de ned 3D architecture was rst enclosed within a matrix material solution and later removed via external stimuli 20 . Such an approach showed better controllability and interconnectivity in pore structure than conventional pore-forming methods, including gas foaming and ice crystal removing 18 . Still, developing shape-memory hemostats with a microchannel structure has not been previously investigated.
Chitosan (CS) has been used to prepare hemostats due to its inherent properties, such as biocompatibility, biodegradability, non-toxicity, anti-infection ability, hemostasis, and so forth 21,22 . Nevertheless, as mentioned above, its hemostatic and anti-infective properties were limited, especially in cases complicated by severe hemorrhage and bacterial infections 23 . Previous studies by our group and others have demonstrated that grafting hydrophobic alkyl chains onto a CS backbone could improve its hemostatic and anti-infective abilities, attributed to the strong hydrophobic interactions between the alkyl chains and the membranes of red blood cells (RBCs), platelets, and bacteria 23,24,25,26 .
Based on these studies, we propose that the shape-memory, pro-coagulant and anti-infective properties of hemostats for noncompressible hemorrhage and in situ tissue regeneration can be improved by optimizing the materials pore structure and further active modi cation. The CS sponges with microchannels were rstly engineered by combining 3D printing polymer micro ber template leaching and freeze-drying methods. To further improve pro-coagulant and anti-infective properties, the microchannelled CS sponges were modi ed with hydrophobic alkyl chains, named MACSs. They presented a highly interconnective and controllable microchannel structure, high water/blood absorbability, a fast shape-recovery property, a strong coagulation-promoting effect, and anti-infection activity. Notably, they demonstrated better hemostatic performance compared with clinically used gauze, gelatin sponge, CELOX, and CELOX-gauze in lethally normal/heparinized rat and normal pig liver perforation wound models. Moreover, they enabled liver cell in ltration, vascularization, and tissue/sponge integration. These results suggest that the MACSs may be bene cial for treating noncompressible hemorrhage and for promoting in situ penetrating wound healing, and thus, have convincing potential for clinical and translational applications.

Results And Discussion
Fabrication and characterization of the MACSs According to our design criteria, the MACSs were fabricated by the procedure illustrated in Fig. 1A. First, the sacri cial PLA micro ber templates were printed by a 3D printer ( Fig. 1B and Supplementary Fig. 1). Then, the templates were lyophilized after lling with a 4% (w/v) CS solution. A CS sponge with a uniform microchannel structure was obtained following complete removal of the PLA templates, which was con rmed by FTIR measurement (Supplementary Fig. 2). The resultant CS sponge was further grafted with hydrophobic alkyl chains to improve its pro-coagulant and anti-infective properties. The grafting was carried out via a highly e cient Schiff-base reaction between the amine group of CS and aldehyde group of DA ( Fig. 2A). The unstable imine bonds (C=N) were converted into stable alkylamine (C-N) linkages using a reductant (NaCNBH 3 ). Compared to the N1s spectrum of the CS sponge, the appearance of C-N * H-C with a peak area of 39.84% and reduction of the peak area of C-N * H 2 in the N1s spectrum of the alkylated CS sponge indicated the successful reaction of the amine and aldehyde groups ( Fig. 2B-D). Moreover, the modi ed CS sponge showed increased Atom Conc % and Mass Conc % of C1s, further demonstrating the successful grafting of hydrophobic alkyl chains (Fig. 2E).
Interconnected pores of the hemostatic sponge could endow itself with the ability to concentrate blood clotting factors and rapidly recover initial shape 5, 10, 29 . Moreover, they were able to provide a comfortable niche to support host cell in ltration, vascularization, and tissue ingrowth 30 . Micro-CT images showed that the alkylated CS sponges with different porosity (MACS-1/2/3) fabricated by a combination of the template leaching method and freeze-drying possessed a uniform microchannel structure with an increased microchannel density (Fig. 1C). The alkylated CS sponge (ACS) prepared by direct freeze-drying presented dense structure. Furthermore, SEM images displayed a hierarchical porous structure including microchannel (138 ± 4.3μm) and micropores (8.7 ± 1.5μm) in the MACS-1/2/3 ( Fig. 1D-F), while only micropores (8.4 ± 0.9μm) randomly distributed throughout the ACS. The microchannel structure was highly interconnected and tunable, and distributed uniformly across the MACS-1/2/3 (Supplementary Movies 1-3). However, the micropores distributed in the ACS showed a dense structure and low interconnectivity (Supplementary Movie 4). The interconnectivity of the porous structure played a key role in accelerating hemostasis and guiding tissue regeneration, which usually was ignored in most previous studies 5, 6,10,13 . The MACSs were expected to exhibit an obvious advantage in the treatment of noncompressible hemorrhage and in situ tissue regeneration in comparison with reported porous hemostats 5,6,10,14 . Accordingly, the porosity of the MACSs gradually increased from 70 ± 2.0 to 90 ± 0.6% with an increase in lling ratio of PLA micro ber, which were signi cantly higher than the 31 ± 0.7% of the ACS (Fig. 1G). Hemostats lled into the wound cavity should possess desirable mechanical strength to prevent their shape deformation caused by external stress from surrounding tissues, thereby providing durable compression on the bleeding site. We rst examined the effect of CS concentration on the compressive stress of the MACSs. As the CS concentration increased from 1 to 4% (w/v), the compressive stress was enhanced from 0.6 ± 0.2 to 23 ± 1.5kPa (Fig. 1H, I). When the CS concentration was lower than 4%, the sponges could not maintain their shapes ( Supplementary Fig. 3A). The CS solution with concentration higher than 4% possessed higher viscosity ( Supplementary Fig. 3B, C), and was di cult to be sucked into the gap of the PLA micro ber template under negative pressure. So, the 4% CS solution was selected to fabricate the MACSs. Next, we investigated the effect of the lling ratio of the PLA micro ber template on the compressive stress. The compressive stress decreased from 46.2 ± 8.0 to 8.1 ± 0.9kPa by increasing the lling ratio of the PLA micro ber template from 20 to 60% (Fig. 1J, K). Indeed, the compressive stress of the MACSs was signi cantly lower than the 138.0 ± 16.3kPa of the ACS due to the incorporation of the microchannel structure. To better approach practical application, we further detected the compression stress of the sponges after absorbing blood. All the sponges exhibited reinforced mechanical strength (Fig. 1L, M), attributing to the formation of blood clots within the sponges. Both the CS and hydrophobic alkyl chain have been proven to facilitate blood clotting by promoting the adhesion and activation of platelets and the aggregation of RBCs. The MACSs had a higher mechanically reinforced fold than the ACS (Fig. 1N) 1,9 . Also, the mechanically reinforced fold of the MACSs gradually enhanced with the increase in porosity (Fig. 1N). The MACSs with high porosity and large surface area could absorb more blood and facilitate the blood to fully contact with the matrix to form more blood clots. Also, the alkylated CS sponge (MACS-2) displayed an improved mechanically reinforced fold compared to the unmodi ed CS sponge (MCS-2) due to the introduction of hydrophobic alkyl chains (Fig. 1N).  Water/blood absorbability of the MACSs The main hemostatic mechanism of expandable hemostats was mechanical compression on the bleeding site, which mainly resulted from water/blood-triggered shape recovery and volume expansion 1,5,14,27,29 . Thus, strong water/blood absorbability was indispensable for expandable hemostats. After absorbing water and blood, the MACSs rapidly sank to the bottom of the container, while the ACS suspended in water and blood (Fig. 3A, B), revealing that the MACSs could absorb a higher volume of water and blood compared with the ACS. The maximum water and blood absorption capacity of the MACSs was signi cantly higher than that of the ACS and gradually improved with an increase in the porosity ( Fig. 3C-F). Notably, the MACSs took much less time to achieve saturated water/blood absorption than that of the ACS (Fig. 3C, D). The water and blood absorption rate of the MACSs was higher than that of the ACS (Fig. 3G, H), which resulted from the increased number of microchannels. The more microchannels present, the higher the water/blood absorption rate. We further stimulated the uid absorption behavior of the sponges, whose pore size originated from the statistical analysis of SEM images, as shown in Fig. 3I. We found that the uid speed in the microchannels of the alkylated sponges (MACS-1/2/3) was higher than that in micropores of the ACS. The higher number of microchannels resulted in a larger area of distribution of the high uid speed. The total uid speed of the MACSs was notably higher than that of the ACS and gradually improved as the number of microchannels increased.

Shape-memory property of the MACSs
We further evaluated the water-and blood-triggered shape-memory property of the MACSs and ACS. All sponges could be compressed and shape-xed after squeezing out the free water (Fig. 4A, B). Upon absorbing the water, they could recover to their original shapes (Fig. 4A, C), giving a 100% recovery ratio. The recovery time (3.3 ± 0.6s, 2.0 ± 0.1s, 1.7 ± 0.6s) of the MACSs was signi cantly shorter than the 41 ± 3.6s of the ACS (Fig. 4D and Supplementary Movies 5, 6). After absorbing blood, the shape-xed MACSs could achieve full shape recovery (4.0 ± 1.0s, 2.5 ± 0.5s, 2.0 ± 0.1s) (Supplementary Movie 7); however, the ACS kept a compressed shape and could not recover any further (Fig. 4B, D, F and Supplementary Movie 8).
The microstructure of the compressed sponges after absorbing water and blood was further observed by SEM (Fig. 4G). In their original state, homogeneous and circular microchannels with gradient numbers distributed throughout the MACSs. The circle microchannels changed to at channels under compression stress. After absorbing water/blood, the deformed microchannels recovered to their original shapes, and the size of the microchannels had no obvious change before and after absorbing water and blood (Supplementary Fig. 4A-C). Furthermore, a large number of RBCs aggregated on the surface of the microchannels. The deformed micropores of the ACS recovered to their original state after absorbing water; however, they did not recover to their original shape after absorbing blood, and almost no RBCs were observed within the ACS (Fig. 4G). In addition, the shape-recovery time of the MACSs was signi cantly shorter (especially absorbing blood) than that of reported shape-memory hemostats (Fig.   4H). Indeed, a large number of studies have demonstrated that, compared to water, blood is more likely to prolong the shape recovery time of hemostats due to its higher viscosity 14,15 . In contrast, there was no signi cant difference in shape recovery time for the MACSs after the absorption of water and blood. This was attributed to the highly interconnected microchannel structure, which allowed the blood to freely penetrate into the sponges. The pore structure inside the ACS and reported shape-memory hemostats generated by the removal of ice crystals and by gas foaming methods exhibited low interconnectivity, which slowed down the ow speed of the blood. 6,9,10,11,15,16 . We also assessed the pro-coagulant ability of the gauze, GS, CELOX, CELOX-G, ACS, MCS-2, and MACSs by the BCI test, in which the lower the BCI value, the stronger the pro-coagulant ability. The BCI values of the MACSs decreased as the porosity increased at 5 and 10min (Fig. 5A), indicating a positive correlation between the promotion coagulation ability and porosity. The BCI values of the MACSs were signi cantly lower than that of the ACS (Fig. 5A). Also, the alkylated CS sponge (MACS-2) exhibited stronger procoagulant ability than the unmodi ed CS sponge (MCS-2) due to the introduction of alkyl chains 24,25,26 . Notably, the MACSs demonstrated better pro-coagulant performance compared with clinically used gauze, GS, CELOX, and CELOX-G due to the synergistic effects of the microchannel structure, CS itself, and hydrophobic modi cation.
The active coagulation cascade mainly relied on the aggregation of RBCs and adhesion and activation of platelets 5 . Thus, we further evaluated the blood coagulation effect of various samples using RBCs and platelets adhesion assays. The number of adhered RBCs and platelets to the MACSs was remarkably higher than that on the gauze, GS, CELOX, CELOX-G, ACS, and MCS-2 (Fig. 5B, C). Additionally, the higher porosity resulted in a higher number of adhered RBCs and platelets. Consistently, as observed in SEM images, more RBCs and platelets adhered to the MACSs than on other samples (Fig. 5D, E). A higher number of aggregated RBCs and activated platelets were detected in the MACSs than that in other samples (Fig. 5D, E), which accelerated blood coagulation 31 . CS has been proven to accelerate platelet adhesion and activation, and the aggregation of RBCs through electrostatic interactions 32, 33 . The microchannel structure was able to promote penetration of the blood and aggregation of RBCs and platelets. The hydrophobic alkyl chains could insert into membranes of the RBCs and platelets, further promoting active capture and aggregation 24,25,34 . We concluded that the CS, microchannel structure, and hydrophobic alkyl chains synergistically contributed to the strong pro-coagulant ability of the MACSs (Fig. 5F).

In vivo hemostatic effect of the MACSs
The MACS-2 was selected and used for in vivo hemostasis based on its mechanical strength, water/blood absorbability, blood-triggered shape-memory property, and pro-coagulant capacity ( Supplementary Fig. 5). The hemostatic effect was explored in the normal rat liver perforation wound model, as illustrated in Fig. 6A. After treating the wound with the MACS-2, a small area of bloodstain was observed on the surface of the lter paper beneath the liver, while a large area of bloodstain was sighted in the gauze, GS, CELOX-G, CELOX, ACS, and MCS-2 groups (Fig. 6B and Supplementary Movie 9). Quantitatively, the total blood loss of the MACS-2 group was signi cantly lower than that of other groups (Fig. 6C). Also, the hemostatic time was signi cantly shorter than that of other groups (Fig. 6D). Hemorrhage control of anti-coagulated patients remains a challenge in the clinical setting 35 . To simulate clinical application, a heparinized-rat liver perforation wound model was used to evaluate the hemostatic capacity of various samples (Supplementary Fig. 6A). After applying the MACS-2, only a small area of bloodstain distributed on the surface of the lter paper under the liver (Supplementary Fig. 6B and Supplementary Movie 10). In contrast, a large area of bloodstain was observed after applying other hemostats. Statistical analysis showed that the hemostatic time of the MACS-2 group was much shorter than that of other groups (Supplementary Fig. 6C). Also, the MACS-2 was superior in reducing the total blood loss when compared with the other hemostats ( Supplementary Fig. 6D).
To further explore the clinical translation potential of the MACSs, a lethal pig liver perforation wound model was used to evaluate its hemostatic capacity (Fig. 7A). Commercial CELOX as a control is a commonly used hemostat in prehospital and hospital scenarios in military and civilian settings 1, 36 . As the shape-xed MACS-2 was lled into the wound cavity (diameter of 1.5cm), it rapidly recovered its initial cyclical shape by absorbing blood, and then lled the cavity and exerted pressure on the wound wall, achieving hemostasis within 2.0 ± 0.5min (Fig. 7B, C and Supplementary Movie 11). However, the untreated wound continued to bleed at least 10min (Supplementary Movie 12), and the CELOX-treated wound stopped bleeding at 9.0 ± 0.3min (Supplementary Movie 13). The MACS-2 was xed on the bleeding cavity by its shape recovery. In contrast, the CELOX was prone to being washed away by the blood without external compression. In fact, manual pressing is very inconvenient in emergencies and it is di cult for the wounded to complete self-rescue on the battle eld 31 . We further quanti ed the total blood loss by determining the sum of the weight of the blood absorbed by the lter paper and hemostat.
The total blood loss (17.6 ± 4.5g) in the MACS-2 group was much lower than that in untreated (153.0 ± 15.5g) and CELOX (143.0 ± 6.6g) groups (Fig. 7D). The MACS-2 demonstrated superior in vivo hemostatic ability for lethal noncompressible hemorrhage compared to clinically used gauze, GS, CELOX, and CELOX-G, which was due to the synergistic effect of CS itself, microchannel structure, and hydrophobic modi cation (Fig. 7E). The highly interconnected and controllable microchannel structure enhanced the blood adsorption capacity of the sponge, allowed the blood to perfuse into the interior of the sponge quickly, and then facilitated the recovery of its original shape, which pressed the wound and achieved rapid hemostasis. CS and alkyl chains actively captured RBCs and platelets via electrostatic and hydrophobic interactions, and also promoted aggregation of the RBCs and platelet activation. This action triggered the coagulation cascade reaction by brinogen-mediated interaction with the activated platelet integrin glycoprotein IIb/IIIa, which further improved hemostasis e ciency 1,9,23 . For clinic application, the MACSs could be customized into different shapes to meet special requirements in practical applications ( Supplementary Fig. 7).

Comparison of in vitro anti-infective property of the MACS-2 with other hemostats
Severe bacterial infection, similar to massive blood loss, is also responsible for trauma-associated deaths 37 . Thus, ideal hemostats should possess robust anti-infection property. The anti-infective capacity of the MACS-2 against S. aureus and E. coli was evaluated by a contact-killing assay and compared with the gauze, GS, CELOX-G, CELOX, ACS, and MCS-2 (Fig. 8). Qualitative and quantitative analysis showed that, after contacting the MACS-2, the CFUs number of S. aureus was signi cantly lower than that of the gauze, GS, and ACS groups. There was no obvious difference in the CFUs number between the MACS-2 and CELOX-G, CELOX, as well as MCS-2 (Fig. 8A, C), because the hydrophobic alkyl chains could not interact with the membranes of S. aureus 38 . After contacting the MACS-2, the CFUs number of E. coli was remarkably lower than that of the gauze, GS, CELOX-G, CELOX, ACS, and MCS-2 (Fig. 8B, D). This enhanced anti-infective activity was ascribed to the synergistic effects of the microchannel structure, grafted hydrophobic alkyl chains, and CS itself 24,26 .

MACS-2 guided in situ liver regeneration
The removal of hemostats may result in secondary bleeding and cause great distress to patients. If hemostats could be left in the injury site and directly guide in situ tissue regeneration, this would be favorable to patients and surgeons 9 . In situ liver regeneration as a representative model was used to evaluate the pro-regenerative ability of the MACS-2 and ACS. Rapid host cell in ltration was the rst and crucial step for endogenous tissue regeneration 39,40 . DAPI and H&E staining showed that the host cells migrated into the interior of the MACS-2, but were mainly distributed around the edge of the ACS due to its dense structure (Fig. 9A). Accordingly, the cell number inside the MACS-2 was signi cantly higher than that of the ACS (Fig. 9B). In ltrated cells secreted a large amount of extracellular matrix and formed neotissue. The tissue ingrowth area within the MACS-2 was much larger than that of the ACS. However, almost no neotissue grew inside the ACS (Fig. 9A, C). A rich capillary network capable of delivering adequate oxygen and nutrients is indispensable for newly formed tissue survival. Thus, vascularization was assessed by immunostaining for von Willebrand Factor (vWF). A high density of capillaries distributed inside the MACS-2 (Fig. 9D); in contrast, almost no capillary was observed within the ACS. A large number of ALB positive cells were observed in the interior of the MACS-2, indicating ingrowth of liver parenchymal cells and liver tissue regeneration. In comparison, almost no liver parenchymal cells in ltrated into the ACS (Fig. 9A, E) 41 . The improved ability of cellularization, vascularization, and tissue ingrowth of the MACS-2 attributed to the highly interconnected microchannels, high porosity, and good biocompatibility (Fig. 9F) 39 . To our knowledge, there has not been any report to date regarding the use of a shape-memory hemostatic sponge for internal penetrating wound repair. Our MACS-2 simultaneously achieved hemostasis and in situ tissue regeneration, which broadens the application of hemostats and opens up an opportunity for the design and construction of clinically bene cial hemostats. Speci cally, the application of our hemostatic sponge will reduce patient discomfort, simplify treatment procedures, and potentially decrease healthcare costs.

Conclusion
In this study, we incorporated a microchannel structure into a CS sponge and further modi ed it with hydrophobic alkyl chains. The MACSs achieved rapid shape recovery by absorption of water and blood. Compared with clinically used gauze, GS, CELOX, and CELOX-G, the MACSs demonstrated stronger procoagulant ability in vitro and hemostatic capacity in lethally normal/heparinized rat and normal pig liver perforation wound models. Moreover, they exhibited enhanced anti-infective activity against S. aureus and E. coli. Notably, the MACSs could be left in the wound bed and guided in situ liver regeneration. All results in this study indicate that MACSs have the clinical translational capacity to provide effective treatment of potentially lethal noncompressible hemorrhage and to facilitate tissue repair.

Fabrication of the MACSs
The fabrication of the MACSs was as follows: First, the PLA micro ber templates with lling ratios of 20, 40, and 60% were printed using a 3D printer (Shenzhen Creality 3D Tech Co., LTD., China). Second, the templates were lled with CS solution (1, 2, and 4%, w/v) dissolved in acetic acid aqueous solution (2%, v/v), followed by freezing in liquid nitrogen and lyophilization. Third, the CS sponges with microchannel structure were obtained by leaching out the templates with dichloromethane. Residual acetic acid was neutralized with a mixed solution of ethyl alcohol/NaOH (9/1, v/v). The resultant CS sponges were further modi ed with DA in the presence of NaCNBH3. Unreacted DA and NaCNBH3 were removed by rinsing with ethyl alcohol and deionized water (DIW) in turn. The MACSs generated from PLA micro ber templates with lling ratios of 20, 40, and 60% were named as the MACS-1, MACS-2, and MACS-3, respectively. An unmodi ed microchannelled CS sponge generated from a PLA micro ber template with a ratio of 40% was abbreviated as the MCS-2. The alkylated CS sponge prepared by direct freeze-drying was named ACS.

FTIR spectrum test
The spectra of CS powder, PLA micro ber template, PLA/CS composite, and microchannelled CS sponge were recorded in the range of 4000-500cm-1 by using a fourier transform infrared spectrometer (FTIR, TENSOR II, Germany).

XPS analysis
The super cial chemical structure and element content of the CS sponges with or without modi cation were detected using an X-ray photoelectron spectrometer (XPS, Axis Ultra DLD, England).

Characterization of macro/microstructure and porosity
The macro and microstructure of the MACSs and ACS were characterized by the micro-computed tomography (Micro-CT, Germany) and scanning electron microscopy (SEM, MERLIN Compact, Germany) 19 . The average pore size was measured using Image-J software (ImageJ 1.44p). The porosity was calculated using Micro-CT.

Mechanical test
The mechanical strength of the MACSs generated with different CS concentrations (1, 2, and 4%, w/v) and PLA micro ber lling ratios (20,40, and 60%) were prepared into cylindrical shapes (5mm in height and 8mm in diameter) and tested in a universal mechanical tester (Instron, 3345). The compression strain and speed were xed at 70% and 1mm/min, respectively. The maximum compressive stress was obtained from a stress-strain curve. The compressive stress of the MACSs (5mm in height and 8mm in diameter) after absorbing the blood was also measured.

Water/blood absorption behavior
After squeezing out water, the compressed MACSs and ACS contacted the water and blood. Their positions in water and blood were recorded by a digital camera. To quantitatively evaluate absorption behavior, the compressed MACSs and ACS were weighed (W d ) and soaked into water and blood from rats. At different time intervals, they were taken out and weighted (W w ). The water/blood absorption capacity was calculated according to the following equation: Water/blood absorption capacity (g/g) = (W w W d ) / W d The water/blood absorption rate was calculated by measuring the slope of the water/blood absorption capacity-time curve within 2s. Moreover, the absorption behavior was further measured by digital uid simulation. The MACSs and ACS were modeled by using the software Solidworks Flow Simulation. The ow orientation of water with a dynamic viscosity of 1.7912×10-3Pa. s was parallel to the axial direction of the sponges. The working temperature and pressure were set as 273.2K and 101325Pa, respectively.
The mass ow at the inlet was 0.001m/s. The stimulated pore size was consistent with statistical results from SEM images. To simplify the simulation, the micropore with low interconnectivity in the sponge was replaced by microchannel with comparable size to the micropore.

Shape-memory property
We assessed the shape-memory property of the MACSs and ACS using the reported method 6 . The MACSs and ACS were rst compressed to squeeze out the free water. Then, both water and blood were dropped onto their top surfaces. The shape-recovery process was recorded using a digital camera. The shaperecovery ratio and time were measured. Also, the microstructure recovery of the MACSs and ACS before and after absorbing water and blood was further observed by SEM.
The MACSs were compressed to squeeze out water and placed in EP tubes. After warming for 10min at 37℃, 50μL of the citrated whole blood (CWB) from rats was dropped onto their top surfaces. After incubation for 5 and 10min at 37℃, 3mL of DIW was added into each EP tube, and optical density (OD) value at 540nm of the supernatant was determined using a microplate reader (BIO-RAD, iMARKTM). The mixed DIW/CWB (3mL/50μL) solution was used as a negative control and its OD 540nm value represented as 100%. The BCI was calculated based on the following equation: BCI (%) = OD material / OD reference value × 100%

RBCs and platelets adhesion assays
The interactions between the MACSs and RBCs were investigated with the previously reported method with some modi cation 27 . The gauze, GS, CELOX, CELOX-G, ACS, and MCS-2 were used as controls.
Before testing, RBCs suspension was obtained by centrifuging the CWB for 10min under 400G. The MACSs were compressed to drain off water and placed in a 24-well microplate. Next, 100μL of RBCs suspension was dropped onto their top surfaces. After incubation for 30min at 37℃, they were rinsed with a phosphate buffer solution (PBS, pH=7.4) to remove nonadherent RBCs, and then transferred into DIW (4mL) to lyse adhered RBCs to release hemoglobin. After 1h, 100μL of the supernatant was taken out and placed into a 96-well microplate followed by measuring its OD 540nm value. The OD 540nm value of a solution composed of 100μL of RBCs suspension and 4 mL of DIW was used as a reference value. The percentage of adhered RBCs was calculated by the following equation: RBCs adhesion (%) = OD material / OD reference value × 100% The interactions between various hemostats and platelets were further evaluated by a platelet adhesion assay 27 . Before measurement, the platelet-rich plasma (PRP) was obtained by centrifuging the CWB for 10min under 400G. The MACSs were compressed to squeeze out water and placed into a 24-well microplate. Then, 100μL of PRP was dropped on their top surfaces followed by incubation for 30min at 37℃. Next, they were washed with PBS to remove nonadherent platelets and soaked into a 1% Triton X-100 solution to lyse platelets to release the lactate dehydrogenase (LDH) enzyme. The LDH was determined with an LDH kit (Biyuntian, China) according to its instruction. Finally, the OD 490nm value of the supernatant was measured. The OD 490nm value of a solution composed of 100μL of PRP unexposed with hemostats was measured and used as a reference value. The percentage of adhered platelets was calculated by the following equation: Adhered platelets (%) = OD material / OD reference value × 100% The adherence of RBCs and platelets on the various hemostats was observed by SEM. Brie y, hemostats were placed into each well in a 24-well microplate and contacted with 100μL of RBCs and PRP suspensions. After 30min at 37℃, they were rinsed with PBS, and then xed with 2.5% glutaraldehyde and dehydrated using a series of graded alcohol solutions. After drying, they were cut, and the longitudinal sections were sputtered with gold and observed by SEM.

Hemostasis in vivo
The hemostatic ability of the MACS-2 was evaluated by lethally normal/heparinized rat and normal pig liver perforation wound models, and compared with gauze, GS, CELOX, CELOX-G, ACS, and MCS-2. All animal experiments were performed with the approval of the Animal Experimental Ethics Committee of Nankai University.
Normal and heparinized rat liver perforation wound models: A rat (male, weight of 250~300g) was anesthetized by injecting 10wt% chloral hydrate in a dose of 1mL/300g. Then, the rat's abdomen was incised, and the liver was lifted and placed onto the surface of the pre-weighted lter paper. Next, a circular perforation wound (diameter of 6mm) was created on the liver to induce hemorrhaging. Finally, the cylindrical MACS-2 (diameter of 8mm) was compressed to squeeze out water and lled into the wound cavity. The hemostatic process was recorded with a digital camera. The blood loss was measured by determining the total weight of the blood absorbed by the lter paper and hemostats. The hemostatic time was recorded with a timer. The heparin solution (50UI) was injected into the rat (male, weight of 250~300g) through a vein at a dose of 2mL/kg and used for the construction of the heparinized rat liver perforation wound model. Other procedures were similar to the method mentioned above.
Lethal pig liver perforation wound model: Bama miniature pig (3 months, weight of 15kg) was anesthetized by injecting a mixed solution of midazolam and xylazine hydrochloride (2/1, v/v) into its muscle at a dose of 0.14mL/1kg. Then, the abdomen of the pig was incised, and its liver was taken out and placed onto the surface of the lter paper. Next, a 15mm-diameter circular perforation wound was made on the liver. After bleeding, the cylindrical MACS-2 (diameter of 18mm) was compressed to squeeze out the free water and lled into the wound cavity. The hemostatic process was recorded with a digital camera. The total blood loss from each liver was weighed and the hemostatic time were recorded.
In situ liver regeneration