3D bioprinting of an implantable xeno‐free vascularized human skin graft

Abstract Bioengineered tissues or organs produced using matrix proteins or components derived from xenogeneic sources pose risks of allergic responses, immune rejection, or even autoimmunity. Here, we report successful xeno‐free isolation, expansion, and cryopreservation of human endothelial cells (EC), fibroblasts (FBs), pericytes (PCs), and keratinocytes (KCs). We further demonstrate the bioprinting of a human skin substitute with a dermal layer containing xeno‐free cultured human EC, FBs, and PCs in a xeno‐free bioink containing human collagen type I and fibronectin layered in a biocompatible polyglycolic acid mesh and subsequently seeded with xeno‐free human KCs to form an epidermal layer. Following implantation of such bilayered skin grafts on the dorsum of immunodeficient mice, KCs form a mature stratified epidermis with rete ridge‐like structures. The ECs and PCs form human EC‐lined perfused microvessels within 2 weeks after implantation, preventing graft necrosis, and eliciting further perfusion of the graft by angiogenic host microvessels. As proof‐of‐concept, we generated 12 individual grafts using a single donor of all four cell types. In summary, we describe the fabrication of a bioprinted vascularized bilayered skin substitute under completely xeno‐free culture conditions demonstrating feasibility of a xeno‐free approach to complex tissue engineering.

a biocompatible polyglycolic acid mesh and subsequently seeded with xeno-free human KCs to form an epidermal layer. Following implantation of such bilayered skin grafts on the dorsum of immunodeficient mice, KCs form a mature stratified epidermis with rete ridge-like structures. The ECs and PCs form human EC-lined perfused microvessels within 2 weeks after implantation, preventing graft necrosis, and eliciting further perfusion of the graft by angiogenic host microvessels. As proof-ofconcept, we generated 12 individual grafts using a single donor of all four cell types.
In summary, we describe the fabrication of a bioprinted vascularized bilayered skin substitute under completely xeno-free culture conditions demonstrating feasibility of a xeno-free approach to complex tissue engineering. For example, Apligraf, a bilayered skin graft manufactured with allogeneic keratinocytes (KCs) and dermal fibroblasts (FBs) along with bovine collagen, is contraindicated in patients with known allergies to bovine collagen. 4,5 Moreover, patients grafted with cultured human KCs, for burn wounds, have been shown to make antibodies to xenogeneic antigens such as FBS, which presumably derived from culture medium. 6,7 Antigen persistence can potentially make such human KCs susceptible to immune destruction by recognition of proteins in FBS and thereby contribute to delayed loss of human KC grafts. Previous efforts have been made to reduce KC immunogenicity by growing cells in a serum-free medium with bovine pituitary extract as a serum substitute. 8 However, bovine pituitary extract contains nonhuman proteins so that human KCs cultured in medium that is supplemented with bovine pituitary extract could similarly activate an immune response. In addition, species specific variation between human and animal sources of collagen type I can substantially alter the physicochemical properties of collagen, which, in turn, can induce a cellular immune response. 9,10 Finally, xenogeneic proteins in serum or matrix can break tolerance to human homologous proteins, leading to autoimmunity. [11][12][13] Consequently, health authorities currently recommend the use of "xeno-free" protocols in the clinical translation of skin equivalents. 14 However, to our knowledge, generation of a fully xeno-free complex tissue capable of surviving after implantation into a host has yet to be demonstrated.
Another challenge in achieving successful clinic translation of bioengineered tissues and organ is vascularization. 15 As more becomes known about skin function, it also becomes clear that current skin grafts comprising of FBs and KCs alone lack the complexity of native human skin. Notably, Apligraf and other approved bilayered skin replacements fail to permanently engraft because they lack dermal vascular networks important for perfusion. 16,17 The complexity of skin microvasculature and other three-dimensional (3D) structures that are present in human skin are difficult to replicate in vitro through simple manual fabrication methods. Advanced manufacturing tools, such as 3D bioprinting, are being developed to generate bioengineered organs with high spatial resolution and in a high-throughput manner. 18,19 Several bioprinting approaches such as inkjet, microextrusion, and laser-assisted printing have been explored for fabricating synthetic skin models. 20,21 The specific mode of material deposition and subsequent phase transition is one of the first factors that needs to be considered to accommodate the material properties. [22][23][24] Other important aspects to consider are the material viscosity and gel-tosolid phase transitioning process required for hydrogels. 25 To date, multiple approaches using 3D bioprinting technology have successfully shown the incorporation of different cell populations beyond KCs and FBs, such as endothelial cells (EC) for the generation of perfusable capillary-like structures, and melanocytes, another important component of the epidermis. [26][27][28][29][30] Furthermore, recent attempts to include appendages such as sweat glands and hair follicles in a human reconstructed skin model have also been made. [31][32][33] Focusing on the 3D structure of the hair follicle Pan et al., using soft lithography, created an array of microwells that resembled the architecture of this skin appendage. 33 The present study addressed two specific challenges: creating synthetic human skin with a functional microvascular network and doing so under conditions that avoid any introduction of xenogeneic cells or antigens.
In previous experiments using skin constructs generated by repopulation of decellularized dermis we showed that allowing human ECs to infiltrate the dermis led to microvessel formation and early perfusion following engraftment on the dorsum of immunodeficient mice, thereby improving graft survival. 34 More recently we demonstrated that inclusion of human vascular cells into the dermal bioink used for 3D printing of skin also produced a bilayered, vascularized skin construct that is perfused, through both graft and host microvessels, early after implantation on immunodeficient mice. 26 Building upon this work, we now demonstrate the fabrication of a bioprinted vascularized skin graft generated under strict xenogeneic-free conditions starting from isolation, culture and cryopreservation of primary cells, bioink/scaffold development, construct maturation in vitro, and successful engraftment on an immunodeficient mouse host. We also improve the mechanical properties of our previously described 3Dprinted graft by incorporation of a biodegradable polyglycolic acid (PGA) mesh into the dermal layer.

| MATERIALS AND METHODS
2.1 | Cell isolation, expansion, and cryopreservation of primary human cells under xeno-free conditions EC cultures were established from human endothelial colony forming cells (HECFCs) present in human umbilical cord blood by culturing whole cord blood mononuclear cells from deidentified donors in EC-Cult-XF medium (STEMCELL Technologies) supplemented with 2 mg/ml heparin (#07980; STEMCELL Technologies) on 10 μg/ml human plasma fibronectin (#FC010; EMD Millipore)-coated plates, as previously described. EC colonies spontaneously arose as "late outgrowth cells" by Day 14, were allowed to reach confluence, and were then serially passaged in the same medium. Human microvascular placental pericytes (PCs) were isolated from discarded and deidentified placentas as explant-outgrowth cells in EC-Cult-XF supplemented with 2 mg/ml heparin on 10 μg/ml human plasma fibronectin-coated plates, also as previously described. 35 Human FBs and KCs were isolated from freshly discarded human foreskin samples according to previously published protocols. 26

| Preparation of xeno-free dermal and epidermal bioinks
Xeno-free dermal and epidermal bioinks were modified from a previously described protocol, substituting human for xenogeneic matrix proteins. 26 Specifically, the xeno-free dermal bioink was formulated Hydrogel Matrix (The Well Bioscience). VitroGel, a xeno-free hydrogel, was added to the bioink to minimize in vitro gel contraction and accelerate gelation of the dermal bioink. The ratio of VitroGel to human collagen type I in the dermal bioink was optimized in vitro. The condition that allowed for faster gelation while promoting selfassembly of EC-networks was selected (data not shown). The xenofree epidermal bioink was formulated with 2 Â 10 6 /ml human KCs in 500 μl of EpiLife medium (#MEPI500CA; Thermo Fisher Scientific) supplemented with HKSdaFREE and HKGE (AvantBio) and 5 ng/ml human recombinant animal-free KGF (Sigma-Aldrich), 1.64 mM CaCl 2 and 50 μg/ml ascorbic acid, a xeno-free medium formulation adapted from Poumay et al. 37 All cells were maintained at 37 C in a 5% CO 2 atmosphere until printing. Where indicated, cultured cells were compared to their counterparts isolated and cultured using previously conditions involving use of nonhuman-derived materials. 26 2.5 | 3D bioprinting of xeno-free skin equivalents Constructs (~2 cm 2 in size) were transferred to six-well plates and submerged in a larger volume of media (4 ml) to prevent rapid exhaustion of nutrients. Medium was changed daily. After 4 days under EC-Cult-XF medium submersion to allow self-assembly of endothelial networks, the dermal construct was carefully transferred to a 24-well plate and the epidermal bioink containing KCs (1 Â 10 6 in 500 μl) was printed from a 3 ml syringe, 30-gauge stainless steel blunt (CELLINK), at a pneumatic pressure of 20 kPa, on top of the dermal construct.
Twenty-four hours after incubation at 37 C, the skin equivalents were transferred to a six-well plate with EpiLife medium supplemented with HKSdaFREE, HKGE, 5 ng/ml KGF, 1.64 mM CaCl 2 , and 50 μg/ml ascorbic acid. The medium was changed every other day for three additional days.          (Figures 6a and S1A). Similarly to nonperfused human vessels, the presence of murine vessels in the dermis of vascularized bioprinted grafts was significantly reduced by 4 weeks (5.5%), but examination of grafts harvested at 6 weeks (4.3%) suggested that mouse vessel number had stabilized (Figure 6e).

Previous work both in vitro and in vivo suggests that placental
PCs associate with and stabilize microvessels formed by self-assembly of human ECs. 26 Quantification of SMA staining in xeno-grafts 4-week postengraftment showed that about 20% of the total SMA+ cells were associated with vessels in the dermis ( Figure S2). This population would include human PCs, but based upon our in vitro phenotyping, could also include human FBs, present in 10-fold greater numbers in the bioink, and, since this stain is not human-specific, The use of collagen type I as a biomaterial has several advantages that include biocompatibility, biodegradability, and the presence of bioactive sites that allow for cell attachment, growth, and differentiation. 41 Currently, most commercially available sources of collagen type I for both research and clinical applications are derived from animal tissues, such as bovine, porcine, rat, and fish. 42,43 Despite some promising clinical outcomes, immunogenic responses (i.e., allergies in 2%-4% of the population) and potential for transmission of prion diseases are two major clinical concerns associated with the use of xeno-derived collagen grafts. In this study, we demonstrate that human collagen type I and human fibronectin can be used in the development of a bioink that supports the generation of an implantable and perfusable xeno-free skin graft with a mature stratified epidermis. We also improved the mechanical properties of the 3D-printed graft by incorporation of a PGA mesh into the dermal layer. PGA has been available for implantation into humans for over 30 years in the form of surgical sutures. 44 Because it is nontoxic in vivo, this polymer has been widely used as a scaffold for tissue engineering: for example, recent studies utilized PGA as a scaffold for vascular tissue engineering using porcine and human iPSC-derived smooth muscle cells (SMCs). 45,46 Here, we used a PGA nonwoven mesh to provide initial mechanical stability to the graft. The presence of the mesh also allowed for easier handling of the construct during formation and easier suturing of skin grafts at the time of engraftment. Within 4 weeks, the mesh was completely hydrolyzed-in a response mediated by host macrophages-and replaced by new native ECM. The inflammation associated with the presence of the mesh appeared to be largely resolved by 4 and 6 weeks.
While there seems to be a high degree of matrix remodeling occurring in the graft at earlier times (e.g., 2 weeks), no significant changes were observed regarding the number of perfused human vessels that persist in the dermis between 4 and 6 weeks. There are some potential disadvantages to using PGA scaffolds in this setting: for example, it has been shown that SMCs grown in high concentrations of PGA breakdown products showed decreased expression of calponin, a marker for SMC differentiation, and decreased SMC proliferation. 47 Although PGA may not be the optimal scaffold for the broad range of vascular tissue engineering applications where complete repopulation of the scaffold with host SMC is desired, it may offer significant advantages for skin tissue engineering applications where graft contraction and scar formation are still current limitations. [48][49][50] Although contraction of a wound bed is a normal part of healing potential clinical issues are: (1) whether the wound heals to restore a barrier; and (2) whether the contraction is excessive, limiting functional mobility.
Our constructs do contract in the first few weeks ( Figure S4), but we believe that our data suggest, although do not rigorously prove, that wounds are healing with stable engraftment of our bioengineered xeno-free vascularized skin construct. We do not yet know whether the degree of contraction is appropriate or excessive. Future experiments will be required to test this question.
The major argument for incorporating vascular cells into the dermal layers is to accelerate revascularization of the graft. Grafts produced that lack ECs and PCs underwent extensive early necrosis and KC loss. Although an epidermis did eventually form in grafts lacking ECs and PCs-likely from surviving human KCs localized near the graft edges-preassembled vessels prevented the early necrotic phase. Furthermore, the presence of ECs and PCs in the bioprinted grafts enhanced host angiogenesis compared to grafts containing only FBs, xeno-free matrix components, and PGA mesh. At 4-week postengraftment about 20% of total α-SMA+ cells associated with vessels.
We have previously shown in a coculture system of RFP-transduced ECs, AmCyan-transduced FBs, and GFP-transduced PCs, that PCs directly associate with ECs while FBs remain randomly distributed. 26 Given the 1:10 ratio of human PCs to FBs in the bioink, 20% of the SMA+ cells number probably reflects the majority of PCs. In addition, we found no detrimental effects on cell function by exposure to the xeno-free coculture medium in our 3D-printed cocultures. Specifically, we show preservation of KC, FB, and EC function by the presence of a mature epidermis of human origin and perfused human EC-lined vessels 2-, 4-, and 6-week postengraftment. We and others have previously reported that the presence of human PCs in the dermal layer promote a mature stratified epidermis with rete ridge-like structures. 26,51,52 This beneficial paracrine effect on KC maturation introduced by the addition of PCs was confirmed in this xeno-free model, further demonstrating preservation of human PC function.
Other cases of crosstalk between vascular ECs and KCs have been reported. [53][54][55] It was recently shown that soluble heparinbinding epidermal growth factor (s-HBEGF) released from vascular ECs activates the epidermal growth factor receptor on KCs and promotes stabilization of sirtuin 1, an essential keratinization inducer. 53 Although the proliferation of human primary KCs with s-HBEGF recombinant protein was almost unchanged in a cell survival assay, a significant increase was detected in the mRNA levels of differentiation markers, corroborating the role of s-HBEGF in facilitating keratinization. These results suggest that that the presence of host and donor ECs in xeno-free skin grafts may not act on KC survival by inducing KC proliferation, but rather through a paracrine effect on keratinization induction.
Cell and ECM sourcing is a critical issue in design of bioinks.
Often, the need for a graft is urgent and it must be available "off-theshelf" to be most useful. For example, a skin graft could be fabricated using differentiation of patient-derived induced pluripotent stem cells (iPS cells), 56 but that would require many weeks to accomplish. Consequently, cells used for clinically relevant graft construction will most likely be derived from another (i.e., allogeneic) individual. Moreover, recent studies have clearly established that human pluripotent stem cell lines display variable capacities to differentiate into specific lineages, even when using iPS cells derived from the same donor, and that the number of cell divisions required creates a mutational burden producing a significant risk for cancer formation. 57 in the formulation of bioinks should also be considered for the treatment of larger wounds. In our protocol, we use approximately 3.6 mg of human collagen type I and 60 μg of human plasma fibronectin to generate 1 cm 2 of bioprinted skin tissue. According to a report by the American Association of Tissue Banks, in 2015 the total number of skin donors as well as total number of square feet of skin distributed by skin banks throughout the United States and Canada was approximately 10,000 donors and 30,000 ft 2 , respectively. 59 Bakerman has studied the variation in yield of acid-extractable human skin collagen with age and his data indicate that a maximum of 0.5 mg of citric acidsoluble collagen per gram of skin can be extracted from individuals older than 15 years. 60 We can extrapolate this data to assume that sourcing human collagen for the treatment of large wounds should be feasible. In addition, sourcing of human fibronectin from human plasma should also be feasible as it is relatively abundant in the plasma (about 30 mg/100 ml) in comparison to the small amount required in the generation of a skin graft. 61 Our current results provide proof of principle that the human sources of cell types and extracellular matrices are practical for tissue engineering of skin.Further improvements and modifications to our current methodology are likely to be required for optimization of clinical skin substitutes. Natural human skin is highly immunogenic in patients and in human immune system reconstituted mice models, but avascular synthetic skin grafts, containing only FBs, KCs, and matrix, such as Apligraf are not. 16 Our past work has provided a potential mechanism to explain this observation: graft ECs, which largely remain of donor origin in clinical transplantation, become "activated" through deposition of complement and evoke strong allogeneic host immune responses. [62][63][64][65] We have proposed that modification of graft ECs to reduce their immunogenicity could reduce the frequency and intensity of rejection without increasing (or possibly even reducing) immunosuppression. 66,67 For example, we found that elimination of MHC class I and class II molecule expression on human ECs, achieved using CRISPR/Cas 9 technology, can diminish recognition of human ECs by both T cells and DSA, and did not cause "missing-self" recognition by NK cells. 68,69 Furthermore, implantation of self-assembled perfusable networks generated with double knockout class I and II MHC-EC into a human immune system mouse model, proved to be sufficient to eliminate rejection. We anticipate that future incorporation of genetically altered cells in bioprinted skin grafts-using cells that are designed to evade alloimmune responses under xeno-free conditions-will allow the generation of an "off-the shelf" vascularized, allogenic, bioprinted skin graft that can evade rejection or minimize the degree of immunosuppression.
Although there is strong evidence that ECs play an important role in triggering human allograft rejection, it is unknown if or which other cell types within a human graft are significant contributors to rejection and thus must also be treated to protect the graft. The experiments with Apligraf cited above appear to argue against an important role for KCs and FBs, but it is equally possible that the lack of vascularization simply kept circulating lymphocytes from contacting these other cells. Interpreting the role of different cell types in bioengineered grafts requires that they come from the same donor and share MHC alleles if the results are to be applied to natural organs. Here, we have developed an implantable bioprinted skin equivalent in which all of the cell populations used were derived from the same neonatal male donor, using cord blood as a source of ECs, placenta as a source of PCs, and foreskin (from neonatal circumcision) as a source of FBs and KCs. The development of this model provides an opportunity to study rejection of bioengineered grafts and to better understand mechanisms of alloimmunity in the context of bioengineered tissues and organs.
While organ transplantation is the single most effective treatment for end stage organ failure, it still requires the use of immunosuppressive regimens with significant side effects, including direct toxicities and increased susceptibility to infections and cancer due to immunological responses of the host to allogeneic tissues. 70 One promising solution is manufacturing of replacement organs through tissue engineering, where significant advances have been made through 3D printing. 24,71,72 3D printing offers the opportunity to produce complex macro-and microstructures with different cell types that have been gene-edited or drug-loaded to alter immunogenicity or antigenicity.
Reducing rejection will not only enable bioengineered organs to function for longer, but also minimize a need for second transplants.
Reduction of the incidence and severity of allograft rejection could both significantly improve current transplant outcomes and increase the likelihood of developing successful therapies to replace multiple tissues and organs.

ACKNOWLEDGMENTS
This work was supported by NIH grants R01-HL085416 and R21-AI159580. Tania Baltazar was supported by the PhRMA Foundation's Postdoctoral Fellowship in Translational Medicine.

CONFLICT OF INTERESTS
Mohammad Z. Albanna is a founding member of Humabiologics, a start-up company developing human-derived biomaterials for tissue engineering. Other authors declare no conflicts of interests.

DATA AVAILABILITY STATEMENT
Any data associated with this study are available from the authors upon reasonable request.