Evaluation of native and non‐native biomaterials for engineering human skin tissue

Abstract A variety of human skin models have been developed for applications in regenerative medicine and efficacy studies. Typically, these employ matrix molecules that are derived from non‐human sources along with human cells. Key limitations of such models include a lack of cellular and tissue microenvironment that is representative of human physiology for efficacy studies, as well as the potential for adverse immune responses to animal products for regenerative medicine applications. The use of recombinant extracellular matrix proteins to fabricate tissues can overcome these limitations. We evaluated animal‐ and non‐animal‐derived scaffold proteins and glycosaminoglycans for the design of biomaterials for skin reconstruction in vitro. Screening of proteins from the dermal‐epidermal junction (collagen IV, laminin 5, and fibronectin) demonstrated that certain protein combinations when used as substrates increase the proliferation and migration of keratinocytes compared to the control (no protein). In the investigation of the effect of components from the dermal layer (collagen types I and III, elastin, hyaluronic acid, and dermatan sulfate), the primary influence on the viability of fibroblasts was attributed to the source of type I collagen (rat tail, human, or bovine) used as scaffold. Furthermore, incorporation of dermatan sulfate in the dermal layer led to a reduction in the contraction of tissues compared to the control where the dermal scaffold was composed primarily of collagen type I. This work highlights the influence of the composition of biomaterials on the development of complex reconstructed skin models that are suitable for clinical translation and in vitro safety assessment.

cells. 1,3 Ideally, over time, the cells start producing their own extracellular matrix (ECM), thereby remodeling the tissue. Beyond the applications for regenerative medicine, these tissues and tridimensional models of human organs have been explored as in vitro platforms for efficacy and toxicological screening of substances. 3 The engineered skin models are typically formed by fibroblasts embedded in a matrix of type I collagen to mimic the dermal layer, and layered with keratinocytes which provide a precursor for the epidermis. A combination of the optimal physiological conditions, such as media composition and exposure to air-liquid interface, promotes differentiation of the keratinocytes resulting in a stratified epidermal layer. These models have been used for a long time for efficacy and safety assessment of substances as well as start pointing for the study and development of skin grafts for regenerative medicine. Nonetheless, despite the many similarities of these structures with human skin, they still fail in recapitulating the entire complexity of the native tissue. 4 In vivo, the skin is formed by several types of cells, and comprises a rich combination of molecules such as fibrous proteins, proteoglycans, glycoproteins, soluble growth factors, and signaling molecules all of which are precisely arranged within each skin layer and skin appendages promoting proliferation, migration, and differentiation of constituent cells. [5][6][7] For example, the dermal matrix is composed mainly of collagen types I, III and V, representing, respectively, 85-90%, 8-11%, and 2-4% of the total collagen content of the dermis in adults. 6,8 Collagen I provides support and biochemicals cues to the cells in the dermal compartment and epidermal region while type III collagen is involved in the formation of the reticular skin. 9,10 Interestingly, an increase in the ratio of collagen I/III has been associated with an increase in scar formation during wound healing. 10 Other important molecules, such as the proteoglycans perlecan, hyaluronic acid and dermatan sulfate, as well as the glycoproteins fibronectin and elastin, are also present at the dermal layer. 6,11 In skin, elastin is found at a ratio of 1:9 to collagen and is responsible for the elastic properties of the tissue. 12,13 Hyaluronic acid is a natural glycosaminoglycan that has been associated with tissue homeostasis, angiogenesis, and wound healing. 14 It is a biocompatible material, with low adhesive properties and extracellular matrix hydration properties, that can be produced by recombinant technology, making it an attractive alternative to animal derived matrices that can pose a risk of immune activation in grafts intended for transplantation to address wound healing. 9,14 In wound healing, hyaluronic acid has also been associated with reduction of scar formation due to its highly hydrophilic characteristics. 15 Dermatan sulfate represents 0.3-1% of the dry weight of skin and has been associated with the processes of coagulation, cell growth, immune defense and wound repair. 16,17 In skin, dermatan sulfate is covalently bound to proteins and performs important functions such as water retention, filling up void spaces, and interacting with cytokines and cell receptors. 18 At the junction between the dermal and epidermal compartments, several biomolecules, such as proteoglycans, collagen type IV and glycoproteins (laminin 5 and fibronectin), form the basement membrane.
In addition to ensuring the adhesion between the two compartments, this thin bilayer membrane regulates the differentiation and proliferation of epidermal cells and is an important reservoir of cytokines and growth factors. 19,20 Furthermore, the basement membrane is a key component for adhesion of melanocytes and melanin production which directly influences skin pigmentation. 4 Collagen IV is an adhesive protein of the basement membrane that regulates keratinocyte attachment, proliferation, and differentiation in vivo and in vitro. 21,22 Laminin 5 is an adhesive protein formed by three chains: α, β, and γ. 23 It promotes adhesion of keratinocytes to the dermal layer and is a known key component of keratinocyte migration in epidermal wound healing (reviewed by Varkey et al.). 20 Fibronectin is also connected to enhanced cell attachment and migration during re-epithelization process in wound healing (reviewed by Jahoda et al.). 24 In cutaneous wounds, the provisional matrix in the damaged region is enriched with fibronectin and fibril-like proteins produced by fibroblasts that migrate from the subdermal tissue. 25,26 Biomaterials currently employed for engineering skin models are typically collagen-based with chemical/temperature-controlled crosslinking mechanisms. The lack of complexity as well as the diversity of biomolecules and cells found in the native tissue contribute, for example, to the poor mechanical properties characteristics of most in vitro models. 27 Different approaches have been explored to successfully produce stable dermal layer such as chemical crosslinking associated with lyophilization of scaffold, use of non-woven hyaluronic acid-based fibrous material, highly porous polystyrene scaffold among others, but these still rely on a single or a minimum number of components. [28][29][30][31][32] Ralston et al. showed that human acellular dermis separated from the epidermis but retaining the dermal epidermal junction (DEJ) basement membrane induced a significantly higher production of soluble fibronectin by the epidermal cells in in vitro models compared to the control (lack of the basement membrane antigens). 33 They further demonstrated that the presence of this layer in the DEJ model improved attachment and morphological aspects of the epithelial cells and exhibited the presence of consistent amounts of collagen IV and laminin 5 at the basement membrane as well as increased expression of soluble collagen IV and fibronectin compared to control. 34 These results suggest that the incorporation of basement membrane proteins within the DEJ can promote not only cell attachment, migration and differentiation, but also the expression of extra cellular matrix proteins, supporting tissue remodeling. Work by El Ghalbzouri et al. demonstrated that the source, that is, humanversus animal-derived, and composition of biomaterials used to generate the skin model have significant influence also on the time over which these tissues can be cultured in vitro. 35 This is particularly important because the short life span of most current in vitro skin models limits their use in long-term efficacy or toxicity studies. 35 They further demonstrated that matrices derived from human primary fibroblasts can support the culture of in vitro models, with a live layer of human primary epithelial cells in the epidermis, for a longer period of time compared to models employing animal-derived matrices such as rat tail collagen. 35 In these studies, the model generated with animal-derived matrix presented a higher degree of tissue contraction compared to a human-derived matrix, which resulted in poor epidermal homeostasis. The investigators speculate that the lack of a human-derived matrix could be attenuating the self-renewal process of keratinocytes. 35 These characteristics present in in vitro models developed with animal derived matrices can be associated with their short life span. Recently, Kim et al. showed that a complex dermal hydrogel composed of a porcine skin-derived extracellular matrix, which retained the major biomolecular composition of skin, was able to support its regeneration in vitro with reduced contraction, improved barrier properties and epidermal organization compared to the control (a collagen I-based hydrogel). 27 Furthermore, they demonstrated that the porcine skin-derived extracellular matrix generated a skin construct with a complex moduli 10 times greater than the control sample generated with only collagen I, indicating the importance of the synergistic physical and chemical interaction between the different extracellular matrix components of the skin. 27 Researchers have also used porous scaffolds to support production of de novo ECM by skin cells, contributing to improved biomolecular complexity, mechanical properties and reproducibility of the skin model. 32,36,37 A comprehensive effort to develop defined compositions that could sufficiently mimic the inter-and intra-complexity of the various compartments of human skin is currently lacking. This creates an opportunity for the formulation of materials for the development of skin models with increased complexity and defined composition aiming at the generation of reconstructed human skin models that better mimic the native tissue. These efforts should focus not only on increasing the complexity of current skin models in terms of structures, cells, and composition but also in terms of their mechanical properties, host resorbability and integration (relevant for regenerative medicine) and in vitro life span (relevant for efficacy testing). 38 In this work, we aim to address some of these challenges by investigating the effect of incorporating a broader and expanded repertoire of native biomolecules within the dermal layer and at the dermalepidermal junction. We hope that this work will encourage the development of complex reconstructed skin models by including biomaterial combinations that could support the incorporation of more cell types.
2 | RESULTS 2.1 | Evaluation of basement membrane proteins for the reconstruction of the dermal-epidermal junction of skin 2.1.1 | Influence of the source of primary cells and collagen coating on proliferation rate of keratinocytes We tested primary human keratinocytes isolated from discarded skin tissue obtained from three independent and de-identified donors: donor A (adult female, breast, phototype I-II), B (neonatal male, foreskin, phototype I-II), and C (neonatal male, foreskin, phototype V-VI). Figure 1 shows the proliferation rates of cells from each donor in culture on collagen IV (2 μg/cm 2 ) coated and uncoated substrates. The proliferation rate was dependent on the age of the donor (adult vs. neonatal), and, as expected, cells isolated from neonatal foreskin (donors B and C) exhibited a higher proliferation rate compared to those isolated from adult breast skin. It can also be observed that cells seeded on substrate coated with collagen (specifically, Type IV) proliferated faster compared to those seeded on an uncoated substrate.  The presence of skin markers throughout the epidermal layer is indicative of the expected stratification of the epidermis. The differentiated layers (stratum granulosum and corneum) are characterized by the presence of proteins associated with the cornified envelope such as filaggrin. 40,[44][45][46][47] In all conditions tested, filaggrin expression was similarly detected in the upper layers of the epidermis (Figure 4-ii (ae)). Cytokeratin 14 and 10 are two proteins expressed, respectively, F I G U R E 2 Heat map representing the effect of basement membrane proteins at the dermal-epidermal junction on proliferation of keratinocytes obtained from donors A, B, and C at Day 4. The color scale represents scoring of the combinations. Blue (corresponding to rate of proliferation > control [uncoated substrate]) represents the highest proliferation rate and Red (corresponding to rate of proliferation < control) represents the lowest proliferation rate of keratinocytes for each donor. Protein concentrations tested: Collagen IV (C)-2 μg/cm 2 (C1) and 8 μg/cm 2 (C2), fibronectin (F)-1 μg/cm 2 (F1), and 4 μg/cm 2 (F2), and laminin 5 (L)-0.5 μg/cm 2 (L1) and 2 μg/cm 2 (L2). Proteins and their combinations were tested in two independent sets of experiments each performed in triplicate (n = 6) by highly proliferating undifferentiated cells 23,40,41,45,[48][49][50] and by cells in the suprabasal layers. 40,41,45,50,51 As can be seen in Figure 4-iii (ae), cytokeratin 14 was strongly detected in the basal and suprabasal layers of the epithelium, and cytokeratin 10 was detected in the suprabasal layers ( Figure 4-v (a-e)). The proliferative capacity of the cells in the epidermis can be confirmed by the detection of Ki67 40,52 in the keratinocytes of the stratum basale. As can be seen in

| Screening of biomaterials for reconstruction of dermal compartment
We first evaluated the influence of the source of Type I collagen on the rate of proliferation of fibroblasts. Specifically, we tested Type I collagen from rat, human (VitroCol ® ), and bovine (PureCol ® ) sources.
Similar to what was observed with the keratinocytes, there is also a difference in the rate of proliferation of fibroblasts depending on the cell source. However, as it can be inferred from Figure 5, here, the main difference seems to be related to the donor in general rather than specific age or anatomical location of the skin samples. Furthermore, these results show an influence of the source of the Type I collagen used in the hydrogels on the metabolic activity of the F I G U R E 3 Growth, spread, and migration of keratinocytes on surfaces coated with dermal-epidermal junction proteins: collagen IV (2 μg/cm 2 ) alone, fibronectin (1 μg/cm 2 ) alone, laminin 5 (2 μg/cm 2 ) alone, and a combination comprising collagen IV (8 μg/cm 2 ), fibronectin (4 μg/cm 2 ), and laminin 5 (0.5 μg/cm 2 ). An uncoated surface was used as control. Twenty-four hours after cell seeding, the cultures were transferred to the Olympus Vivaview ® microscope and live images were acquired every 30 min and are presented here at 12-h intervals. Each experiment was performed in duplicate and representative images are shown fibroblasts over time. Throughout the period evaluated, and among the three cell donors, the hydrogels constituting rat tail Type I collagen exhibited the highest proliferation rate, followed by the hydrogels constituting bovine and human collagens Type I, respectively.
To visualize different collagens, second harmonic generation images of the constructs with and without fibroblast were acquired at Day 0 and 6 days after seeding ( Figure 6). Second harmonic generation microscopy enables label free imaging of the collagen fibers. 53 Bovine collagen showed finer fibers of collagen compared to collagen derived from human or rat, while the rat collagen was seen to have the largest fibers. Evaluation of collagen from different origins showed no visible contraction of the fibers when fibroblasts were absent.
Addition of fibroblasts to the collagen showed distinct signs of contraction on Day 6. Early signs of contraction could be detected in PureCol ® samples (bovine collagen) as early as day 0 with Fibroblasts, in contrast to collagen from human and rat.
The shrinkage of the gels was measured using a Vernier caliper.
While no significant differences can be found between Day 0 and Day 6 for samples without cells, the addition of fibroblasts led to a contraction of the gels to less than half their original size, from 2.0 to 0.85 and 0.94 cm ( Figure S1).
To quantify the contraction of the collagen taking place on a micrometer scale, the area filled by the collagen in the images acquired using 2-photon microscopy. The collagen area was found in each image by segmenting the images into background and collagen.
Using the determined area of collagen, the average area percentage F I G U R E 4 Images of reconstructed skin model including a dermal-epidermal junction layer. Column (a): Control (only dermal and epidermal compartment); Column (b): collagen IV (2 μg/cm 2 ); Column (c): fibronectin (1 μg/cm 2 ); Column (d): laminin 5 (2 μg/cm 2 ); Column (e): combination of collagen IV (8 μg/cm 2 ), fibronectin (4 μg/cm 2 ), and laminin 5 (0.5 μg/cm 2 ). Row (i): Histological analysis-hematoxylin and eosin stain (10 μm sections). Rows (ii)-(vi): Analysis of skin markers by immunofluorescence (10 μm sections); Row (ii): filaggrin (orange), fibronectin (green), DAPI (blue); Row (iii): cytokeratin 14 (orange), DAPI (blue); Row (iv): Laminin 5 (orange), DAPI (blue); Row (v): cytokeratin 10 (green), collagen IV (orange), DAPI (blue); Row (vi): Ki 67 (green). The in vitro skin models were generated using a mixed population of human primary cells (fibroblasts, keratinocytes, and melanocytes) isolated from neonatal foreskin samples. Scale bar = 50 μm taken up by collagen was calculated. Overall, the average total area percentage taken up by the collagen network was found to be between 20% and 30% for samples on Day 0, and between 30% and 40% on Day 6 (see Figure S2). To quantify the ratio of change due to the fibroblasts, the data from Day 6 were normalized by calculating the ratio of the data from Day 6 against the data obtained from the corresponding fibroblast-free gels from Day 0 ( Figure 7). In general, the value for gels without fibroblasts was close to 1, suggesting that the density of the collagen does not change considerably. For the gels with fibroblasts, a ratio of 1.19-1.48 is seen, showing that the collagen network contracts and becomes 19%-48% denser. No significant difference was found between different collagen density ratios. The lowest collagen density ratio for hydrogels with fibroblasts was observed for VitroCol ® , followed by Rat Tail Collagen and PureCol ® .
Next, we evaluated the effect of other components (collagen Type III, hyaluronic acid, elastin, and dermatan sulfate) along with collagen Type I on the rate of proliferation of fibroblasts as shown in Based on these findings, we further investigated the distribution of rat tail collagen and the influence of dermatan sulfate on the contraction of the hydrogels. Imaging of the hydrogels showed sparser collagen density for samples supplemented with dermatan sulfate ( Figure 9). Addition of fibroblasts showed the expected contraction of collagen fibers on Day 6 (see Figure S3); however, the addition of dermatan sulfate resulted in less densely packed hydrogels.
To quantify the density of the collagen network, the area filled by collagen in the images was calculated. An overall lower collagen density for samples including dermatan sulfate was observed ( Figure 10

| DISCUSSION
Reconstructed skin models have been developed and studied for more than two decades. These models have been extensively characterized for their morphology, functionality, and similarity to the human skin. 12,23,28,29,40,42,47,51 They have been validated as alternative methods to animal models for the safety assessment of substances and explored as relevant tools for efficacy evaluation of topically applied products. [54][55][56][57] Furthermore, the field of regenerative medicine has evolved exponentially toward the development of skin grafts that could be used for clinical applications. 27,58,59 Nonetheless, engineered skin models still fail to recapitulate the complexity of the native tissue in terms of cellular diversity (e.g., lacking melanocytes, neural and immune cells), presence of adnexal structures (e.g., lacking hair follicle, sebaceous, and sweat glands), and vascularization. Furthermore, the currently available reconstructed skin models employ a very limited array of matrix proteins and scaffold biomaterials compared to the diversity of biomolecules present in the human skin, which are important for tissue homeostasis. 40,51,54,60,61 In this study, we investigated the effect of employing a more complex and defined matrix on the growth and support of cells in 2D as well as on promoting proper development of a full thickness 3D skin model. Using this approach, we first analyzed the impact of collagen IV, fibronectin, and laminin 5, found at the dermal-epidermal function of the native skin, on the growth of keratinocytes in a monolayer.
We observed that some of these proteins or their combinations pro- , PureCol ® , and VitroCol ® ) over 6 days. Each image was segmented in FIJI using Li's Minimum Cross Entropy thresholding method to calculate the area covered by collagen. To calculate the collagen density ratio, the average area percentage from Day 6 was divided by the data from Day 0. Every experiment was performed in duplicate and three images were acquired in different xyz positions for each duplicate (n = 6). The results present the mean ± SD of the data One possible explanation is that in keratinocytes, fibronectin receptors are trypsin-sensitive, which suggested that, in vitro, their adherence to fibronectin depends on de novo synthesis of this receptor. 25 The need for the synthesis of new receptors prior to the adherence of the cells to the surface coated with fibronectin could explain the decreased proliferation rate. A few combinations showed variability in proliferation response, which may be attributed to factors of genotypic and/or epigenetic origin that while interesting were beyond the scope of this study and will be probed in future studies.
It was additionally observed (data not shown) that in the control wells (uncoated substrates), keratinocytes preferred to proliferate as isolated colonies while on the coated substrates the cells spread more uniformly as a monolayer. Lamb and colleagues have previously observed that keratinocytes grown in media supplemented with serum and high concentration of calcium were highly motile compared to serum-free and low calcium condition, demonstrating the influence of growth conditions on cell behavior. 49 Similarly, Huang and colleagues showed that media composition not only affects cell morphology and proliferation but also influences colony formation. 62 To further investigate the colony formation behavior, we performed a live analysis of the movement and distribution of keratinocytes over a period of time. We demonstrated that keratinocytes cultured on uncoated substrates grow as colonies while cells grown on substrates coated with the proteins from the dermal-epidermal junctions present a dynamic behavior and uniform spreading. This confirmed our observation that, beyond regulating cell proliferation, these proteins promote cell migration and result in uniform cell distribution. In vivo, progression of keratinocytes through the cell cycles is affected by the degree of differentiation, matrix adhesion, and growth factors. 63 The proper balance between the proliferation and differentiation of keratinocytes, which presents a turnover of around 30 days, results in tissue homeostasis. 64,65 In vitro, the loss of a physiological state and F I G U R E 8 Viability of fibroblasts in hydrogels formed by collagen type I from Rat tail and (a) Type III collagen, (b) elastin, (c) hyaluronic acid, and (d) dermatan sulfate. Cell viability was evaluated using PrestoBlue ® solution. Relative cell viability was calculated by normalizing fluorescence readings of hydrogels containing cells to controls of hydrogels without cells. The results present the average ± SD of the data from two independent experiments each performed in triplicate (n = 6) inclusion of artificial growth conditions favors a proliferative profile that contributes to cellular expansion in monolayer. 63,64,66 Furthermore, media composition, such as presence of serum and antibiotics, concentration of calcium and growth factors, or culture conditions, can modulate cell behavior in monolayer as well as influence outcomes in 3D in vitro models. 49,62,67 We hypothesized that the inclusion of a DEJ layer in the skin reconstruction protocol could accelerate early keratinocyte proliferation and distribution contributing to the formation of a uniform confluent monolayer prior to the beginning of the cell differentiation process when exposed to the airliquid interface. Furthermore, we believe that this characteristic could contribute toward accelerating the host integration of skin graft placed in wounds as it is known that proliferation and migration of keratinocytes are fundamental for tissue reepithelization. 58,66 Considering this hypothesis and understanding that beyond cell attachment and cell proliferation rate, these DEJ proteins must promote proper keratinocyte differentiation, we evaluated the effect of four distinct protein compositions in the skin reconstruction context. Following 14 days of tissue maturation, we did not observe any significant morphological differences between the control and test conditions. Nevertheless, the staining with Ki67 suggested that these compositions could indeed be supporting prolonged keratinocyte proliferation, which remains to be further investigated. Based on these results, we speculate that the in vitro period required for the maturation of the tissue could be long enough that any early positive or negative effect of the inclusion of these proteins at the dermal-epidermal junction would be compensated over time by the innate capacity of cells to produce extracellular matrix molecules and reach homeostasis. On the micrometer scale, the second harmonic generation images of collagen showed first signs of contraction within a few hours after seeding, thus agreeing with previously published macroscopic measurements of collagen contraction. 69,70 Quantification of the images of collagen showed similar overall collagen densities; however, there were differences in the collagen fiber size, with bovine and human collagen having finer fibers than rat tail collagen. The collagen density ratios showed differences between the three gels after 6 days of Modification of the collagen organization has previously also been shown by Stuart and Panitch, showing that addition of dermatan sulfate led to an increased amount of void space. 72 The study was, however, performed on cell-free gels. Taken together with the results presented in this study, we suggest that the effect of dermatan sulfate may not be cumulative, meaning the relative amount of void space present in the gel is determined from the moment of polymerization, and the relative contraction of the gel is only minimally influenced by the addition of dermatan sulfate. The results do however suggest a more regulated organization and polymerization of the fibers, showing more coherent results in gels with dermatan sulfate compared to gels without it.
Regarding the more complex scaffold generated by the mixture of Type I collagen and other matrix molecules, when we compared each condition that contains hyaluronic acid to its respective control in terms of collagen Type I dilution, we can infer that the decrease on the metabolic activity of fibroblasts observed can be associated with the lower concentration of type I collagen and not the presence of the hyaluronic acid. Previously, and consistent with our findings, Kreger and Voytik-Harbin also observed that addition of hyaluronic acid to collagen I hydrogels did not significantly change the proliferation of human dermal fibroblasts embedded in the gels. 73 Independent of the hydrogel compositions, we observed only a small increase in the metabolic activity or total number of cells over the period evaluated. In the native tissue and under steady state conditions, the mitotic activity of adult fibroblasts in most locations is very low compared to fibroblasts grown in vitro or even epidermal keratinocytes in the stratum basale. 74 Considering that our goal is to generate a skin model that mimics healthy human tissues, the lack of significant increase in the rate of proliferation of the fibroblasts within the 3D hydrogel is normal and expected. We concluded that the formulated hydrogels tested were non-cytotoxic to the fibroblasts and did not induce abnormal cell growth.
One of the main challenges in the fabrication of the reconstructed skin models in vitro is the contraction (x, y, and z directions) that most of the constructs developed undergo during the submerged condition and more prominently when exposed to the air liquid interface. 29,75,76 It has been shown that the contraction of skin models in vitro as well as hydrogels without epidermal cells is influenced not only by the fibroblast density and culture period but also by the concentration of collagen type I. 77 This is one of the aspects that limits the life span of most in vitro models to a maximum of 8 weeks in culture. 35 Over the years, several groups have been studying the mechanisms underlying this phenomenon and investigating strategies to modulate and control, but with limited success. Some of these successful strategies have used stabilized scaffolds to support the development of welldifferentiated and mechanically stable skin models that could be maintained for up to 12 weeks at the air liquid interface. 28,29 Boehnke and colleagues demonstrated that a hyaluronic acid esterified with benzylic alcohol scaffold provided counteracting forces to prevent skin contraction but also supported the synthesis of de novo extracellular matrix. 29 Similarly, in the work done by Mewes and colleagues, fibroblasts embedded on the stabilized collagen scaffold secreted extracellular matrix proteins and created a network of elastic-fibers similar to that observed in the human skin, which supported tissues homeostasis for 51 days. 28 We speculated that the inclusion of the glycosaminoglycans such as hyaluronic acid and dermatan sulfate could reduce the degree of tissue contraction by controlling the water-binding capacity or ECM modification 9,78-81 In our experiments, the poor gelation of the gels containing hyaluronic acid could be a result of the absence of crosslinking of the hyaluronic acid. Furthermore, the hydrogels containing hyaluronic acid were not able to reduce the contraction of the tissue samples. Accordingly, Kreger and Voytik-Harbin also demonstrated that the inclusion of hyaluronic acid in collagen Type I hydrogels did not affect skin contraction by fibroblasts. 73 However, the hydrogels containing dermatan sulfate and Type I collagen reduced the contraction of the samples compared to the control, which could be associated with higher water retention (substrate-bound and free water). 80 Also, the combination of the in vitro condition and the presence of dermatan sulfate could be inducing fibroblasts differentiation into myofibroblasts. This phenotype displays exacerbated ECM production and presents a contractile apparatus that helps them to remodulate the ECM, which is fundamental for physiological tissue repair. 81,82 In vitro, the control of ECM composition and structural modification by differentiated fibroblasts could also influence tissue integrity and contraction profile. Considering both hypothesis, the inclusion of dermatan sulfate could provide the right condition to promote de novo extracellular matrix production and improve the mechanical properties. 32 Based on these results, we infer that dermatan sulfate can modulate skin contraction while promoting proper tissue homeostasis.
Finally, analyzing the skin contraction results from several independent experiments (data not shown) we also observed considerable variability in the extent of skin contraction in the constructs with the same dermal composition. This could be associated with the specific combination of cells used in each of the independent experiments.
Reconstructed skin models were generated using human primary cells isolated from skin samples discarded during surgery, and consequently the number of cells obtained from these samples is finite and their life span in culture is also reduced compared to immortalized cell lines. This implies that there is a limited supply of each cell type and, as consequence, it is frequently necessary to use different combinations of cells in the experiments. For example, it has been shown that keratinocytes isolated from African and Caucasian skin exhibit significant differences in their stratification and differentiation when included in reconstructed skin models. 83 The differences observed here highlight, once more, the importance of using a representative population of cells in tissue reconstruction. This would not only substantiate the study in terms of observations that can be made based on specific cell populations but can also reduce the variability in these observations. The data collected in this work could support the optimization of skin reconstruction protocols by better modulating the tissue maturation process as well as rheological properties of gel-based models. and the plates were incubated for 30 min at 37 C and 5% CO 2 . After gelation, 1.5 ml of culturing media was added on top of the gels and changed every other day. For image acquisition, the nuclei were stained using Hoechst 33342 (Invitrogen, H3570) for 10 min. The media was replaced using culturing media prepared with phenol-free DMEM (Sigma, D1145). The images were acquired using a custom-build 2-photon microscope based on the Nikon Eclipse Ti-E Inverted microscope system described here. 86 The excitation laser was Spectra-Physics Mai-Tai DeepSee Tuneable Ultrafast Laser tuned to 800 nm. The microscope was controlled using Micromanager, and VistaVision (ISS). The emission signal was split by a beamsplitter (460 LP filter), the DAPI signal was detected using a (494/20 nm bandpass filter) while the SHG from the collagen was detected using a 405/14 nm bandpass filter. The area covered in collagen was calculated in FIJI 87 by segmentation using Li's Minimum Cross Entropy thresholding method and subsequent measurement of the average area percentage.

| Reconstruction of human skin tissue
To generate the reconstructed skin models, we started by depositing the dermal solution (800 μl) with specific hydrogel compositions and Vectashlied ® with DAPI (nuclear blue stain). All images were obtained and analyzed using an Olympus IX51 Fluorescence Microscope.

| CONCLUSION
Despite being considered equivalent to human skin, reconstructed skin models still fail to reproduce the same level of complexity and physiological characteristics of the native tissue. These gaps limit not only their application as platforms for toxicological screening of substances in vitro but also their use as grafts for the treatment of wounds due to impaired host integration. In the present study, we demonstrate that proteins from the DEJ (collagen IV, laminin 5, and fibronectin) can modulate not only the proliferation of keratinocytes grown as a monolayer but also their migration and distribution in 2D.
More specifically, we show that collagen IV can singularly induce a higher proliferation rate compared to the other proteins, individually or in combination. These results suggest that the proper combination of dermal-epidermal junction proteins could support skin graft integration as well as modulate cellular turnover in in vitro models. However, in the 3D context of the fabricated skin, we observed that the cells can produce these proteins and form a DEJ independent of the deposition of any of the proteins initially, suggesting that a positive effect could be limited to early stages of tissue maturation In the investigation of the effect of components from the dermal layer (collagen I and III, elastin, hyaluronic acid, and dermatan sulfate), the primary influence on the viability of fibroblast was attributed to the source of the collagen type I (rat tail, human and bovine) used as a scaffold material. Furthermore, the incorporation of dermatan sulfate in the dermal matrices lead to a reduction of the contraction of skin samples compared to the control samples whose dermal scaffold is composed primary of collagen type I. Finally, we observed differences in response of cells from different donors (proliferation and metabolic activity).
Our results highlight the relevance of the biomaterial composition and cell source to the development of complex reconstructed skin models. Biomaterial enrichment, as well as culture optimization, could be a key step for supporting cellular diversity in tissue models and hence, truly achieving tissue complexity. Furthermore, we believe that these findings along with work done by other colleagues in the field of tissue engineering can contribute, for example, to the design of bioinks to be used to 3D bioprint a new generation of physiologically more relevant reconstructed skin models.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.