Artificial cell membrane binding thrombin constructs drive in situ fibrin hydrogel formation

Cell membrane re-engineering is emerging as a powerful tool for the development of next generation cell therapies, as it allows the user to augment therapeutic cells to provide additional functionalities, such as homing, adhesion or hypoxia resistance. To date, however, there are few examples where the plasma membrane is re-engineered to display active enzymes that promote extracellular matrix protein assembly. Here, we report on a self-contained matrix-forming system where the membrane of human mesenchymal stem cells is modified to display a novel thrombin construct, giving rise to spontaneous fibrin hydrogel nucleation and growth at near human plasma concentrations of fibrinogen. The cell membrane modification process is realised through the synthesis of a membrane-binding supercationic thrombin-polymer surfactant complex. Significantly, the resulting robust cellular fibrin hydrogel constructs can be differentiated down osteogenic and adipogenic lineages, giving rise to self-supporting monoliths that exhibit Young’s moduli that reflect their respective extracellular matrix compositions.

A s advanced cellular-based therapies approach clinical translation, there is an increasing demand for new cell specific matrices that provide improved therapeutic performance 1 . However, rational matrix design is extremely challenging, as cell phenotype and fate are reliant on a wide range of scaffold-dependent environmental factors, which include cell adhesion, chemical composition, cell receptor stimulation, microand nanomorphology, and mechanical stiffness 2,3 . These factors come into play almost immediately, as in vitro tissue engineering typically begins with the seeding and adhesion of cells onto a biocompatible and biodegradable scaffold, which acts as a surrogate for the extracellular matrix (ECM) 4 . As the cells proliferate and/or differentiate, they produce ECM that gradually replaces the scaffold material, giving rise to a structurally self-supporting entity 5 .
A range of biocompatible natural polymers have been used to produce transient hydrogel scaffolds for tissue engineering, including chitosan, gelatin, collagen, hyaluronic acid and fibrin 6 . Of these biopolymer-based scaffolds, fibrin hydrogels are amongst the most popular, as they can be readily produced in situ at room temperature via proteolytic cleavage 7 . Fibrin formation in vivo occurs as a response to injury, culminating with the proteolytic cleavage of prothrombin to thrombin, which is a serine protease that catalyses the polymerization of fibrinogen to fibrin, resulting in the formation of a fibrin hydrogel clot 8 . Importantly, fibrin hydrogels display specific peptide motifs, such as RGD, which promote adhesion via integrin signalling, and mediate the recruitment, activation and presentation of growth factors (e.g. BMP-2 and TGF-β1), which help regulate osteogenic and chondrogenic differentiation in human mesenchymal stem cells (hMSCs) [9][10][11] . Moreover, a key advantage of (fibrin) hydrogels over solid porous scaffolds such as poly(lactide-co-glycolide) is their ability to be delivered via injection 12 , although extrusionbased shear stress can result in a reduction in cell viability 13 .
Cell membrane engineering is emerging as a premier approach for cell augmentation, as it can be used to rapidly provide additional cell functionality, including immunoevasion, adhesion and homing 14 . However, there are few examples where cell membrane modifications have been used to drive in situ scaffold formation. Wu et al. 15 and Cruise et al. 16 irradiated cells coated in the photosensitive dyes eosin and polyethylene (glycol) diacrylate to generate hydrogel structures on the surface of lung adenocarcinoma cells and porcine islets, respectively, although both systems required synthetic monomers and an external source of radiation.
Here, we describe a methodology involving the synthesis of supercationic thrombin-polymer surfactant complexes that spontaneously bind the plasma membrane of hMSCs and drive in situ fibrin hydrogel nucleation and growth (Fig. 1). The resulting self-supporting hydrogel constructs support high levels of metabolic activity, and provide a matrix for adipogenic and osteogenic differentiation. Moreover, we demonstrate that this cell functionalisation methodology is facile and can be used to inject thrombin labelled GFP-expressing fibroblasts into an in vivo zebrafish skin wound model.

Results
Artificial membrane binding thrombin synthesis. The synthesis of the cell membrane binding thrombin (bovine) complex was performed using a two-step process (Fig. 2a). The initial step involved the kinetically controlled cationization of thrombin via N-(3-dimethylaminopropyl)-N′ethylcarbodiimide hydrochloride (EDC) mediated covalent attachment of cationic N,N′-dimethyl-1,3-propanediamine (DMPA) to surface exposed carboxylic acid functional groups to generate an active supercationic thrombin (sc_thrombin) construct ( Supplementary Fig. 1) 17 . Thrombin not only contains glutamic acid (21) and aspartic acid (18) residues, but also acid functional groups arising from posttranslational modifications, including up to 10 ɣ-carboxyglutamyl residues that each contain two carboxylic acid groups 18 and carboxylic acid groups resulting from the N-linked glycosylation of three asparagine residues 19 . Accordingly, the reaction conditions were carefully controlled (pH, temperature, and the number of EDC and DMPA equivalents) and the enzyme cationization reaction progress was monitored using zeta potentiometry, which shifted from −13.9 ± 0.9 to +5.8 ± 0.3 mV over a 2 h period (Fig. 2b). As the active site of thrombin contains the serine protease catalytic triad including a reactive aspartic acid (Asp189) 20 , it was necessary to concomitantly evaluate the activity of the thrombin during the reaction. This was achieved by monitoring the increase in fibrinogen solution turbidity resulting from fibrin gel formation, which showed a steady decrease in reaction rate as the cationization reaction progressed over the 2 h period (Fig. 2c). To balance the effective cationic surface charge density required for electrostatic surfactant conjugation with high enough levels of enzymatic activity, the reaction was quenched at 60 min and these conditions used for all subsequent experiments. Matrix assisted laser desorption ionisation time of flight mass spectroscopy performed on the resulting sc_thrombin ( Fig. 2d), showed an average increase in mass of 3.25 kDa, which signified charge reversal (negative to positive) at 39 sites.
While it has been shown that supercationic proteins such as supercharged green fluorescent protein have a high membrane binding affinity, strong electrostatic interactions with the anionic sulfonated proteoglycans of the glycocalyx results in rapid endocytosis 21,22 . This transient behaviour has been used to good effect for the delivery of siRNA 23 , as has the cationic polymer poly(ethylenimine) (PEI) for nucleic acid 24 and protein 25 delivery. However, our principle objective was to display the active thrombin at the cell surface in order to nucleate a fibrin gel from the plasma membrane. Therefore, a second step involved electrostatically conjugating the anionic polymer surfactant glycolic acid ethoxylate 4-nonylphenyl ether (ox890), which comprises an anionic carboxylic acid headgroup, a hydrophilic poly(ethylene)glycol central moiety (E40) and a hydrophobic nonylphenyl tail (Supplementary Fig. 2 & Supplementary Methods). Significantly, we have recently shown that this approach can be used to generate a polymer surfactant corona on the surface of myoglobin that reconfigures to insert into the plasma membrane of hMSCs 17 . Here, electrostatic conjugation of the polymer surfactant (≈120 molecules per protein) to the Fig. 1 Schematic showing in situ fibrin hydrogel formation from the membranes of bone-marrow derived human mesenchymal stem cells (hMSCs). Artificial membrane binding thrombin constructs comprising supercationic thrombin molecules (white) surrounded by a polymer surfactant corona (yellow) that associates with surface exposed cationic (red) residues spontaneously insert into bilayer regions of hMSC plasma membranes. In the presence of fibrinogen, the membrane-immobilised thrombin catalyses fibrin formation (blue fibres) within the interstitial spaces between the cells giving rise to a self-supporting hydrogel monolith sc_thrombin to yield [sc_thrombin][ox890] resulted in an negative shift in the zeta potential distribution ( Supplementary  Fig. 3). Dynamic light scattering experiments ( Supplementary  Fig. 4) performed on solutions of thrombin, sc_thrombin and [sc_thrombin][ox890], gave particle size number distributions centred at approximately 5 nm, with a small increase in hydrodynamic diameter distribution after surfactant conjugation, which is consistent with previous reports 17, [26][27][28] .
Surfactant conjugation did not result in a reduction in catalytic activity (cf. sc_thrombin) ( Supplementary Fig. 5 Fig. 7) that were consistent with soft fibrin hydrogels 29 .
Thrombin construct membrane affinity and hydrogel formation. A clinical application for thrombin functionalised cells could be readily derived from in vitro tissue engineered constructs that utilise transient fibrin scaffolds. Accordingly, bone marrow derived hMSCs with well-characterised differentiation pathways (e.g. adipogenic, chondrogenic and osteogenic) were selected to assess membrane adhesion and cell metabolic activity 30 . Initially, a monolayer of hMSCs was incubated with 1 μM    Supplementary Fig. 9). MTS assays 31 were performed to measure the relative metabolic activity of hMSCs labelled with the [sc_thrombin] [ox890], which showed no significant reduction up to concentrations as high as 2.6 μM (p = 0.087) (Fig. 3d). Accordingly, a 1 μM [sc_thrombin][ox890] incubation concentration was selected for all subsequent experiments, as this afforded suitable catalytic activity, membrane loading potential and biocompatibility, and was within the physiological concentration range of (pro) thrombin in plasma 32 . Moreover, control cell metabolic activity assays were also performed using the neat oxidised surfactant  Fig. 11). The membrane labelling protocol was then modified to generate 3D fibrin hydrogel constructs containing 1 × 10 7 cell mL −1 (Dia. = 6.5 mm; Vol. = 100 μL; ≈1 × 10 6 cells) (Fig. 4b). Here, confluent hMSCs were labelled with 1 μM [sc_thrombin][ox890], re-suspended in FBS-free culture media supplemented with 6 mg mL −1 fibrinogen, and pipetted into wells coated in a 1 wt.% agarose gel. Fibrin gelation was allowed to proceed for a period of 60 min prior to the addition of an equal volume of high glucose media supplemented with 20% (v/v) FBS and 0.1 TIU mL −1 aprotinin to limit fibrin degradation 33 . After 24 h, the constructs were fixed and imaged using scanning electron microscopy ( Fig. 4c), which showed dense cellular aggregates with individual cells surrounded by a 3D fibrin matrix. Z-stacks produced using live cell confocal fluorescence microscopy showed a welldispersed distribution of hMSCs, uniformly suspended within the fibrin hydrogel matrix (Supplementary Movie 4 & Supplementary Fig. 12).
hMSC differentiation and mechanical properties. In order to investigate the ability of the fibrin hydrogel system to sustain 3D cultures for the extended periods of time required for tissue engineering, the metabolic activity of hMSCs within the fibrin constructs was monitored over 22 days using a MTS cell metabolic assay. Cells were cultured in a medium that maintained hMSC multipotency to avoid changes in innate cell metabolic activity due to differentiation, which allowed the relative increase in cell number to be estimated. Significantly, a constant increase in the ratio of the formazan product (490 nm) to the MTS reactant (387 nm) was observed over the entire period, which indicated sustained levels of hMSC proliferation (Fig. 5a).
Reverse transcription quantitative polymerase chain reaction (RT-qPCR) experiments were performed to monitor the earlystages of differentiation of the [sc_thrombin][ox890] modified hMSCs in the fibrin hydrogel system. The PPAR-γ gene is a pivotal ligand-activated transcription factor that upon activation is upregulated and drives hMSCs towards an adipogenic fate 34 . Accordingly, upregulated PPAR-γ expression was used as an early indicator of adipogenic differentiation (14 days), which showed a 7-fold increase when the hMSCs were cultured in adipogenic media (cf. standard media) (Fig. 5b). To probe the capability of cells to undergo chondrogenesis, the relative expression of the chondrogenic gene SOX9 was explored 35 . SOX9 is upregulated in response to the addition of chondrogenic factors (e.g. TGF-β3) and downregulated in the presence of osteogenic factors (e.g. BMP-2), with its expression linked to the activity of the osteoresponsive gene RUNX2 36 . Here, SOX9 expression in the fibrin constructs supplemented with chondrogenic media resulted in a 4-fold increase in SOX9 expression (cf. osteogenic media) after 7 days (Fig. 5c). However, no significant increase in RUNX2 expression was apparent in the fibrin constructs supplemented with osteogenic media (cf. standard media) after 7 days ( Supplementary Fig. 13).
Following on from the RT-qPCR experiments, the hMSC fibrin constructs were differentiated down adipogenic or osteogenic lineages over a 21 day period to enable the potential for development of typical phenotypic traits 30,37 . Aside from visual changes in cell morphology, analysis of the resulting constructs were probed by the addition of specific fluorescent stains relevant for each lineage. This included Oil Red O for lipid vacuole formation during adipogenesis 38 and Alizarin Red for calcium deposition resulting from osteogenesis 39 . For the modified cells exposed to the adipogenic media, confocal fluorescence microscopy images showed clusters of lipid vacuoles, emanating from cells with a globular morphology, which was consistent with the formation of mature adipocytes (Fig. 5d). Conversely, modified cells exposed to the osteogenic media exhibited extensive calcium deposition, signifying osteogenesis, which was accompanied by subtle changes away from a spindle-like morphology (but not cuboidal), highlight the ongoing transition toward the formation of fully mature osteoblasts (Fig. 5d) 40 . Both phenotypes were also observed across a wider population of cells, liberated, re-plated (overnight) and imaged in 2D (Supplementary Fig. 14).
The differentiation pathways of hMSCs are dependent on the mechanical properties of their environment, with stiffer interfaces generally favouring an osteogenic fate and softer surfaces adipogenesis [41][42][43] . In practice, this means that ECM formation during tissue engineering can provide positive feedback to differentiation, as it can directly impact on the stiffness of the surrounding environment. For example, the deposition of calcium phosphate during osteogenic biomineralization can result in an increase in stiffness in the surrounding environment, which in turn can stabilise the osteoblast phenotype 44 . To explore the impact of ECM formation on the mechanical properties, unconfined compression testing (Fig. 6a, b) was performed on the constructs (Dia. = 8.0 mm; Vol. = 400 µL; ≈ 4 × 10 6 cells) formed from 6 mg mL −1 fibrinogen solutions cultured in either osteogenic, adipogenic or standard expansion medium for a period of 23 days. These were compared to constructs cultured for 1 day (containing hMSCs) in standard expansion media, and with constructs devoid of hMSCs and catalysed using 200 nM native thrombin. Significant differences between each system were observed, with day 1 Young's compressive moduli of 8.1 kPa in fibrin only constructs, which increased to 12.2 kPa in the presence of hMSCs. After 23 days, hMSCs undergoing osteogenesis exhibited a much higher stiffness (34.2 kPa), when compared with those undergoing adipogenesis (24.6 kPa) or multipotency culture conditions (15.4 kPa) (Fig. 6c). These results indicate that the cell membrane nucleated fibrin structures are amenable to scaffold-matrix remodelling by the differentiated cells, and that the resulting ECM influences the bulk compressive mechanical properties 45 .
In vivo localisation of thrombin-modified fibroblasts. Although the development of the membrane re-engineering approach described above could prove advantageous for in vitro tissue engineering, the ability to produce thrombin coated cells ad hoc that could be injected into a site of injury provides new opportunities for in vivo tissue engineering. Here, an injury site would be preferable as endogenous fibrinogen is freely available within intravascular fluids such as blood plasma at between 1.5 and 4 mg mL −1 46 . However the clinical use of exogenous fibrinogen within fibrin sealants may also permit administration to other sites 47 . Accordingly, we performed preliminary cell transplant studies using an in vivo zebrafish model system to examine the isolation, labelling, delivery and survival of a different cell type, namely, primary zebrafish fibroblasts. Zebrafish (Danio rerio) are an established model organism and provide an excellent platform for live cell imaging with numerous cell type specific fluorophoreexpressing strains available (Supplementary Fig. 15) 48 . Zebrafish have also been considered as a model organism for studying thrombolytic and haemostatic processes 49 and furthermore, haemostatic activity has been observed by the infusion of activated bovine thrombin to zebrafish in vivo as a method to assess the availability of clottable fibrinogen in antithrombin III mutants 50 . Consequently, fibroblasts from GFP-expressing ET37 fish 51 (Supplementary Fig. 16   re-suspended, with donor labelled (GFP+) fibroblasts (micro) injected around the site of a fresh ventral incisional injury in adult, non-transgenic recipient zebrafish (Fig. 7a, b). Significantly, similar numbers of viable unlabelled and thrombin-functionalised labelled cells were observed at 3 days post (micro)injection (dpi), suggesting normal cell survival (Fig. 7c, d). Macroscopic observations and histological analysis of sections through the incisional injury revealed no adverse effects between fish injected with labelled and unlabelled fibroblasts (Fig. 7e, f), however further investigations will be required to determine precise effects on specific wound healing processes. Herein, we described the synthesis and characterisation of a new membrane active thrombin construct that can be used to drive in situ fibrin formation from the plasma membranes of stem cells. The resulting hydrogel constructs support high levels of viability and proliferation in hMSCs cultures over 22 days, which can be readily differentiated to produce self-supporting tissue engineered constructs. Unconfined compression testing showed that differentiation of the hMSCs down well-defined adipogenic and osteogenic pathways directly impacted the Young's moduli of the resulting structures, which indicates high levels of integration between the nucleated fibrin hydrogel and the resulting ECM.
The facile membrane labelling strategy and tissue engineering protocols described above provide a new approach to both in vitro and in vivo tissue engineering, although a detailed understanding of the temporal behaviour of the membranebound thrombin constructs (e.g., internalisation and deactivation of the enzyme) would need to be ascertained before clinical translation. This is key, as controlling the concentration of bioavailable thrombin, is a fundamental requirement in the development of haemostatic products.

Methods
Materials. All materials unless otherwise specified were purchased from Sigma-Aldrich Company Ltd (Poole, UK) or ThermoFisher Scientific Ltd (Loughborough, UK) and were of the highest available purity.   was repeated a minimum of five times to limit further cationization and protein concentration verified using a Pierce™ BCA protein assay kit (ThermoFisher Scientific Ltd, Cat. No. 23225). Samples were then stored at −20°C prior to further use for up to 72 h. Surfactant oxidation was achieved by the complete oxidation of the terminal hydroxyl group of IGEPAL Co890 to yield a terminal carboxylic acid functional group. Two grams of IGEPAL Co890 dissolved in 50 mL of deionized water was mixed with 30 mg 2,2,6,6,-tetramethyl-1-piperidinyloxyl (TEMPO), 50 mg sodium bromide and 5 mL of a sodium hypochlorite solution containing 10-15 % available chlorine. The solution was adjusted to pH 11 and then stirred overnight at room temperature before the reaction was quenched with the addition of 10 mL of ethanol and adjusted to pH 1 prior to a solvent extraction with three 80 mL aliquots of chloroform. The combined chloroform fractions were washed with three 80 mL aliquots of ddH 2 O at pH 1 and chloroform removed by rotary evaporation (40°C/150 mbar). The resulting product was dissolved in 40 mL of ethanol at 65°C and then recrystallized overnight at −20°C. The following day the ethanol was decanted with this process repeated once more prior to removing any residual ethanol by rotary evaporation (50°C/120 mbar) to yield a fully oxidized product. Surfactant conjugation was achieved by dissolving the oxidized product directly into the desired protein solution at room temperature at a concentration equivalent to 1.4 the total number of glutamic acid, aspartic acid, lysine and arginine residues and stirred for a period 1 h and protein concentration verified using a Pierce™ BCA protein assay kit. Samples were then used immediately or stored at −20°C prior to further use for up to 72 h. buffer. Conjugation efficiency was then determined using UV-visible spectroscopy by measuring relative intensities at 280 nm (protein) and 555 nm (rhodamine) 52 . Cells were labelled with 1 μM of the desired thrombin solutions in DMEM media (no FBS) with 60 mM HEPES (pH7) for 20 min prior to washing with PBS. Thereafter, cells were labelled with CellMask™ DeepRed following the as supplied instructions prior to imaging in complete medium ((DMEM low (1000 mg L −1 ) glucose supplemented with 100 μg mL −1 streptomycin, 2 mM Glutamax, 100 U mL −1 penicillin, 10% (v/v) FBS, 5 ng mL −1 fibroblast growth factor).
[ [ox890] in 60 mM HEPES (pH7) subsequently added prior to incubation at 37°C/5% CO 2 for 20 min. Afterwards cells were immediately washed with 25 mL PBS twice and trypsinised (8 mL) for 5 min in order to liberate adherent hMSCs into a suspension, to which 16 mL of complete medium is added (in order to inhibit further trypsin activity). The resulting suspension was then centrifuged (400 × g, 25°C, 5 min) with the supernatant discarded and replaced with 25 mL of complete medium (without FBS) and centrifuged (400 × g, 25°C, 5 min) once more with the supernatant discarded, this washing (i.e. FBS removal) step was then repeated once more. The visible cell pellet was then resuspended in complete medium (without FBS) to a total volume between 75-150 μL (dependant on the number of constructs to be created). A 20 μL aliquot of resuspended hMSCs (5 × 10 7 cells mL −1 was then added to 80 μL 7.5 mg mL −1 human fibrinogen preplaced into a precoated (agarose) well and gently mixed by pipetting. The resulting mixture was then incubated at 37°C/5% CO 2 on an orbital rotator (50 rpm) for 60 min to promote and consolidate fibrin gel formation (6 mg mL −1 ). Afterwards cells were supplemented with 100 μL of the relevant medium (e.g. adipogenic) containing 0.1 TIU mL −1 aprotinin with 3-5 media changes per week for up to 23 days. No. ant-nr-1) with FBS at 10% (v/v). Cells were then centrifuged (250 × g, 4°C, 10 min) with the resulting supernatant discarded. Cells were then resuspended (1 × 10 6 cells mL −1 ) and washed once more by centrifugation before resuspension at a final concentration of 4 × 10 7 cells mL −1 prior to microinjection.
Zebrafish adoptive cell transfer. Wildtype, non-fluorescent recipient fish were anaesthetised in 0.13% MS-222 and placed ventral side up in a pre-cut sponge soaked in aquarium water containing anaesthetic. A 4-mm incision was made through the skin and the pericardial sac directly above the heart. This area was chosen as the underlying dermis is thicker than other sites and the procedure is routinely preformed. Approximately 1000 GFP+ labelled or unlabelled cells in a 50 nL volume were microinjected at six sites around the edge of the incisional injury and the fish allowed to recover in fresh system water without anaesthetic.
Zebrafish imaging and staining procedures. At the desired time-points, recipient fish were terminated and the area around the incision dissected and fixed in 4% (v/ v) paraformaldehyde (PFA) for 1 h at room temperature, washed twice in PBS and mounted in 1.5 wt.% low melting point agarose. Images of endogenous GFP expression were acquired on a Leica DM6000 SP8 AOBS upright confocal laser scanning microscope. For histological analysis, skin samples were decalcified in 0.5 M EDTA (pH 7.4) at room temperature for 5 days, embedded in paraffin wax, sectioned and stained as described previously 53 . Sections of 8 μm thickness were stained with AFOG using standard protocols. Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All data supporting the research findings in this study are available from the corresponding author upon reasonable request. The source data underlying Figs. 2b, 3d, 5a, 5b, 5c, 6c and Supplementary Figs. 7a, 10 and 13 are provided as a Source Data file.