Engineering and Development of a Tissue Model for the Evaluation of Microneedle Penetration Ability, Drug Diffusion, Photothermal Activity, and Ultrasound Imaging: A Promising Surrogate to Ex Vivo and In Vivo Tissues

Driven by regulatory authorities and the ever‐growing demands from industry, various artificial tissue models have been developed. Nevertheless, there is no model to date that is capable of mimicking the biomechanical properties of the skin whilst exhibiting the hydrophilicity/hydrophobicity properties of the skin layers. As a proof‐of‐concept study, tissue surrogates based on gel and silicone are fabricated for the evaluation of microneedle penetration, drug diffusion, photothermal activity, and ultrasound bioimaging. The silicone layer aims to imitate the stratum corneum while the gel layer aims to mimic the water‐rich viable epidermis and dermis present in in vivo tissues. The diffusion of drugs across the tissue model is assessed, and the results reveal that the proposed tissue model shows similar behavior to a cancerous kidney. In place of typical in vitro aqueous solutions, this model can also be employed for evaluating the photoactivity of photothermal agents since the tissue model shows a similar heating profile to skin of mice when irradiated with near‐infrared laser. In addition, the designed tissue model exhibits promising results for biomedical applications in optical coherence tomography and ultrasound imaging. Such a tissue model paves the way to reduce the use of animals testing in research whilst obviating ethical concerns.


Introduction
The use of ex vivo human tissues and animals has been the main approach for evaluating the performance of emerging transdermal and intradermal drug delivery systems. Nevertheless, the use of these models is associated with several drawbacks, for example, strict ethical regulation along with sample heterogeneity and variability due to age, gender, race, and anatomical site differences. [1,2] From an ethical standpoint, small animal models should only be used when essential-for example, at an advanced stage of formulation development prior to moving to large animals or human volunteers as the immediate next step. Therefore, the use of in vivo model cannot be justified for routine quality control or for preliminary investigation. [1] Hence, depending on the targeted applications, scholars have Driven by regulatory authorities and the ever-growing demands from industry, various artificial tissue models have been developed. Nevertheless, there is no model to date that is capable of mimicking the biomechanical properties of the skin whilst exhibiting the hydrophilicity/hydrophobicity properties of the skin layers. As a proof-of-concept study, tissue surrogates based on gel and silicone are fabricated for the evaluation of microneedle penetration, drug diffusion, photothermal activity, and ultrasound bioimaging. The silicone layer aims to imitate the stratum corneum while the gel layer aims to mimic the water-rich viable epidermis and dermis present in in vivo tissues. The diffusion of drugs across the tissue model is assessed, and the results reveal that the proposed tissue model shows similar behavior to a cancerous kidney. In place of typical in vitro aqueous solutions, this model can also be employed for evaluating the photoactivity of photothermal agents since the tissue model shows a similar heating profile to skin of mice when irradiated with near-infrared laser. In addition, the designed tissue model exhibits promi sing results for biomedical applications in optical coherence tomography and ultrasound imaging. Such a tissue model paves the way to reduce the use of animals testing in research whilst obviating ethical concerns.
often employed different artificial models in place of in vivo tissues. [3][4][5][6] In light of this, there is an impetus to develop artificial models as surrogates to evaluate the insertion and drug release profile of microneedles. These models would be at a lower cost while being more accessible as there are no concerns regarding safety and ethical approval. [2] Importantly, such models would obviate the issue with sample heterogeneity thus enabling more reproducible and comparable experiments between researchers. There are many reports on the use of monolayer/ monophase artificial models either for mechanical purposes (e.g., skin insertion test) or biological goals (e.g., drug release studies or extraction of interstitial fluids for biosensing of glucose). [7][8][9] The most common skin surrogates are agarose gels (≈1.4-1.8 wt% in >98% of water) which contains a relatively higher amount of water compared to human skin (≈70 wt%). Such high water content does not mimic in vivo and ex vivo skin models may lead to erroneous estimation in drug release profiles from pharmaceutical systems. [10] The use of artificial skin models may be of heuristic value in evaluating and refining the properties of emerging pharmaceutical technologies such as microneedles (MN). [10] MN systems have been utilized for many years and have achieved considerable success within the cosmetic industry. [11] However, there has yet to be any MN-based formulation to gain FDA approval for the delivery of therapeutic compounds. Recent collaborations between academic institutions and pharmaceutical partners are thought to change this trend. Should MN-based products gain regulatory approval, this would encourage pharmaceutical companies to exploit this drug delivery system to administer a range of therapeutics into and across the skin. The development of a simple yet reproducible skin model may serve as a catalyst to help with the evaluation and refinement of MN based systems during early product development before the product is evaluated further in vivo. MN patches must be able to perforate the stratum corneum without undergoing any fracture or bending to the needle architecture during skin insertion. [10] The stratum corneum is the stiffest layer that is responsible for the stiffness and hardness during MN application. [10] Therefore, one of the shortcomings of using agarose-based skin models is the inability to mimic the mechanical characteristic of stratum corneum. [10] Furthermore, there is no accurate skin model that is capable of mimicking both the biomechanics and overall diffusion coefficient of human skin. Thus, the utilization of a such model, analogous to ex vivo skin, could be employed in order to overcome the shortcoming of various monophasic skin models. In order to achieve this aim, as a proof of concept, we fabricated a tissue model consisting of silicone rubber-based materials that mimic the human stratum corneum for microneedle penetration analysis (Scheme 1A,B). The beneath layer is composed of gelatin-based hydrogel, capable of mimicking diffusion properties of skin with similar water content.
Besides the skin, there is not an accurate in vitro model that can mimic in vivo tissues (e.g., malignant organs such as a cancerous kidney) to investigate the diffusion and release behaviour of a drug model. In light of this, we examined the release profile and diffusion of a model drug (doxorubicin) within the gel layer (Scheme 1C) and compared the release profile to agarose and ex vivo Male BALB/c mice kidney models. In addition, currently, aqueous solutions are employed to assess the photoactivity of photothermal agents. However, such method of accessing photoactivity is slightly erroneous as there is a disparity in the heating rate of photothermal agents in aqueous solution relative to in vivo and ex vivo tissues. Accordingly, the heating profile of the gel layer of the tissue model was evaluated when illuminated with a near-infrared (NIR) laser to assess its usability for assessing the thermal properties of a photothermal agent, for example, polydopamine. Finally, we also investigated the application of this tissue model in bioimaging, such as ultrasound imaging and compared to agarose and in vivo human brachial arm (Scheme 1C).

Fabrication and Characterization
Scheme 1A shows the fabrication of the tissue model. FTIR was used to study the interaction between acacia gum, gelatin, agarose and PVA in the tissue skin model. In the FTIR Scheme 1. Fabrication and biomedical applications of the tissue model. A) Schematic illustration of the fabrication of tissue model. B) Interaction between the components. C) Schematic representation of potential application of the tissue model: i) fluid extraction and biosensing; ii) investigation of insertion ability of microneedle; iii) evaluation of the diffusion profile of a model drug injected with a needle; iv) bioimaging using ultrasound imaging and optical coherence tomography (OCT) microscopy; v) evaluation of the photothermal activity of a photothermal agent in place of water solution. spectra of tissue model ( Figure S1, Supporting Information) the characteristic peak related to carbonyl of amide (1700 cm −1 ) is observed which verify that the amine groups of gelatin are reacted with carboxylic groups of acacia gum to produce amide groups (Scheme 1B). [12] In addition, hydrogen bonding between hydroxyl groups of PVA and amine (in gelatin) and carboxyl (in acacia gum) groups led to blue shift and decrease in the intensity in IR spectrum. On the other hand, the presence of characteristic peaks of agarose observed in the FTIR of the tissue skin mode which confirmed the physical interaction between agarose and other components. [13] These results showed that the mixture of Acacia gum, gelatin, and poly(vinyl alcohol) (PVA) in presence of agarose led to a 3D hydrogel. Scheme 1C exhibits the potential applications of the tissue model. Photograph, optical microscopy, and scanning electron microscopy (SEM) images of the tissue model are shown in Figure 1A,B.

Swelling Behaviour and Wettability
The swelling ratio of the gel layer (with and without freeze dying) is shown in Figure 1C. It was apparent that the swelling ratio of the gels increased over time till reaching equilibrium. As anticipated, the lyophilized samples (≈5.1 g g −1 ) displayed higher fluid absorbance, compared to control (≈0.6 g g −1 ). Water contact angle test was employed to evaluate the surface wettability of the samples ( Figure 1D). Smooth-Sil 960 (124.5°) and Dragon Skin 30 (113.3°) showed the highest contact angle than other samples, indicating that these surfaces were considerably hydrophobic with reduced surface energy. The contact angle of ex vivo stratum corneum of porcine skin (37.8°) and gel layer (24.1°) were lower relative Smooth-Sil 960 (124.5°) and Dragon Skin 30 (113.3°). The contact angle of agarose was not detectable. Among the samples, agarose showed the highest hydrophilicity, indicating that the gel exhibited high surface hydrophilicity with enhanced surface energy. Figure 1E shows the shear viscosity versus shear rate of the tissue model and agarose gel. As shown in the flow curves, both samples display a similar rheological trend, which is a reduction in viscosity with an enhancement in shear rate. Besides, the tissue model possesses higher viscosity relative to the agarose gel over the whole range of tested shear rates. Upon increasing the shear rate, the hydrodynamic forces exerted cause the aggregates to become deformed. This eventually culminates in the disruption in the overall macromolecular alignment and leads to a decrease in the viscosity. [14] In addition, small amplitude oscillatory shear tests was also performed to evaluate the linear viscoelastic properties, that is, elastic (G′) and viscous (G″) moduli of the samples. The rheometric curve mechanical spectra in terms of G′ and G″ as a function of frequency, for the tissue model and the agarose gel are shown in Figure 1F,G. By comparing the elastic and viscous moduli of the samples, it is apparent that porcine skin displayed a higher G′ and G″ relative to the agarose gel and tissue model. In addition, all samples exhibited an elastic like behavior as evidenced by G′ > G″. Furthermore, the oscillatory shear test also showed that the tissue model is far more elastic than the agarose gel.

Impedance Spectroscopy
The electric properties of the tissue models were also evaluated in comparison to porcine skin and agarose via impedance spectroscopy. The impedance phase and modulus of the gel layer of the tissue model showed similar trends relative to the porcine skin ( Figure 1H,I). The gel layer with 1.8% NaCl concentration is also compared to porcine skin and agarose, where 0.9% NaCl concentration was proven to be more similar to porcine skin than 1.8% ( Figure S2, Supporting Information). Those properties were largely governed by the concentration of the salt in the gel layer. The effect of the top silicone layer was not largely observed here, due to its small thickness and the use of a concentric needle, where the electric field propagates only in the gel layer. The dissipation of the saline in the gel was tested by the transient impedance measurement and the result is shown in Figure 1J. The impedance readings responded to the injected saline and dropped largely when the saline affected the electric field path of the measurement. As seen in the graph, there are two drop points attributed to two injections. Although the concentric needle was not in the direct contact with the saline solution, the impedance reading largely reacted to the presence of the saline injected. This is attributed to optimal permittivity of the gel material used which is similar to skin tissue. This gives an opportunity for the gel to be a model for indirect yet closecontact and sensitive measurement applications.

Microindentation and Micropenetration Analysis
A wide variety of mechanical analysis (e.g., compression, tensile, penetration, etc.) were designed to elucidate the behaviour of skin and its relevant models. [15,16] Among them, indentation (without perforation of tissue model, Figure 2A) and penetration (with perforation of tissue model, Figure 2B) can better simulate the insertion of microneedle. The microindentation test entails applying a compressive stress via an indentor onto the targeted tissue or skin model. The mechanical behaviour of the designed tissue model is lower than ex vivo porcine but higher than in vivo human skin. The results exhibit a raise in the normal contact force as a function of the indentation depth.
Skin micropenetration analysis performed on the respective tissue models are shown in Figure 2B. Porcine skin demonstrated the highest mechanical properties in terms of penetration test than other models. In contrast, agarose and chicken tail showed the lowest mechanical strength in the penetration test. The effect of the underneath underlying layer on penetration efficiency of a 24G needle through comparing the penetration results of a chicken wing and a chicken tail. Skin deformation is limited by a stiffer beneath rigid underlying layer (similar to the bone underlying the skin in a chicken wing) whereas a softer beneath underlying layers (similar to the chicken tail) allow more deformation. As a result, in the first condition, a much larger force is required, whereas in the second case, less force is needed relative to the latter. The result from micropenetration test ( Figure 2B) echoes the finding observed in Figure 3A-C. The designed tissue models covered with different silicone-based materials mimicked chicken wing and was similar to porcine skin. In most instances, the tissue model covered with silicone-based material was more analogous to ex vivo tissue when compared to the monophasic agarose gel. With respect to designing these tissue models, film formation is one of the limitations of some of these silicone-based materials. Poly(dimethylsiloxane) (PDMS) may not be peeled off easily without being damaged. Consequently, although the mechanical properties of PDMS in terms of the penetration test is similar to human skin, it may not be considered as stratum corneum. We propose Dragon Skin 30A and Smooth-Sil 960 as the topmost layer because of the stiffness and capability of filmforming. Figure 2C,D show the optical images of micropenetration on ex vivo porcine and in vivo human skin, respectively.

Simulation
Experimental uniaxial compression test data are employed to find a proper hyperelastic model of the respective tissues regarding that hyperelastic models capture the behaviour of soft materials among all available hyperelastic models, the Ogden 1st order model with the coefficients reported in Table S4, Supporting Information, offered the best fit to the experimental data as shown in Figure 2E. This model is utilized in a simple indentation numerical simulation with a maximum strain of 20%. The same test is performed experimentally as mentioned in Section 2.5. Finally, stressstrain diagrams of both cases are plotted and compared in Figure 2F. Also, Figure 2G demonstrates the simulated deformation caused by the indentor. Figure 2F demonstrates a perfect coincidence between simulation results and experimental results in the region of 5-15% compressional strain and for the other regions, the coincidence is   . D) Microscopy and optical coherence tomography images of microneedle penetration into full-thickness neonatal porcine skin, tissue model and agarose gel. The red arrow indicates the surface of the skin or tissue models. E) Insertion depth of microneedles into different tissue models. F) Pore size generated by microneedles in different tissue models. The results are presented as mean ± S.D. (n = 10). Differences were calculated using one-way ANOVA, followed by Tukey's post hoc test, and deemed significant at p < 0.05; n.s. indicate no significant difference.

Microneedle Penetration and Imaging
Following extensive characterization of the tissue model, a microneedle insertion study was conducted in order to evaluate the capability of the model being used as a surrogate to ex vivo porcine skin, a common model which is frequently applied in microneedle research. [17] The microneedle insertion test was conducted by applying the patches onto the model under a pre-set force of 32 N which is analogous to thumb pressure.
It was shown that the mechanical properties of underlying strata found beneath the skin affected the microneedle insertion profile. [18] In our experiment, we also found that the maximum force needed for a hypodermic needle to puncture a chicken tail was significantly lower in comparison to the chicken wing, highlighting the effect of elasticity from the underlying subcutaneous layers beneath the skin ( Figure 2B). Relative to the chicken wing, the chicken tail possesses much more fat and lacks the structural support from connective tissue such as bones which culminated in enhanced elasticity. Accordingly, in our model, we evaluated the effect of the underlying layer (i.e., that mimic subcutaneous tissues) on penetration profile of microneedles ( Figure 3A-C). It can be seen that the insertion depth of microneedles into the tissue model that is void of any subcutaneous layer is much deeper relative to the model that incorporated the underlying subcutaneous layer. Based on our finding, this suggests that it would be more physiologically accurate to fabricate a tissue model without an underlying subcutaneous layer in order to avoid overestimating the insertion profile of microneedles. Figure 3D shows an OCT image of the hydrogel forming microneedle arrays being inserted into respective ex vivo and in vitro tissue models. It was apparent from the OCT images that microneedle showed the deepest insertion into neonatal porcine skin and tissue model. In contrast, we observed shallower penetration depth when the microneedles were inserted into the agarose gel. In addition, using OCT, we also evaluated the in situ penetration depth of hydrogel-forming microneedles when the microneedles were inserted into the respective skin simulants ( Figure 3E,F). Accordingly, we were able to measure the penetration depth along with the size of the microneedle channels created in the respective tissue models. It is apparent that microneedle insertion was deepest in ex vivo porcine skin as the needles managed to be inserted to a depth of ≈600 µm. As can be seen, the agarose showed shallower penetration than tissue model, relative to ex vivo neonatal porcine skin. In addition, when the size of the microneedle channels was evaluated, we can see that the size of the microneedle channels created in ex vivo porcine skin (≈300 µm) and tissue model (≈280 µm) were much significantly (p < 0.05) wider than agarose gel (>10 µm). In summary, the results show that the tissue model can mimic the insertion test better than agarose gel. It should be noted that the silicone layer was used to mimic the stratum corneum in order to evaluate the insertion capability of microneedles across the outermost layer of the skin. Accordingly, in the following tests, only the gel layer was used (tissue model without silicone layer).

Drug Diffusion Study
To evaluate the diffusion small molecules across the respective skin model, fluorescein, a model drug, was injected wthin the tissue model, agarose, and rat skin. During the imaging process, the changes of the colored areas were imaged. As shown, it was obvious that the signal intensity decreased over time, exhibiting a significant decline in fluorescent intensity at the injection site ( Figure 4A). The relative fluorescence curves are shown in Figure 4B (left panel) versus time. The behavioral similarities (R 2 ratio) between the tissue model and the rat skin were shown to be analogous ( Figure 4B, right panel). Figure 4C displays fluorescence intensity variations among the three model of skins after percutaneous injection of the model gels within the skins. The signal intensities were obtained upon illumination of the samples using a black light lamp. As displayed in Figure 4C, the variation in signal intensity in the designed tissue model was much less noticeable. In contrast, such variations were much more significant in the case of rat skin. Therefore, one can deduce that similarity in the optical performance between the rat skin and skin tissue model was close together indeed. In summary, such similarities and performances in the optical activities of the developed tissue model endows it with promising alternative candidate for future investigation focusing on the biomedical application including many nanoparticles, biomaterials and microneedle patches.
To evaluate the drug diffusion behaviour in tumor microenvironment, doxorubicin was injected to cancerous kidney, as well as to the agarose and tissue models having acidic pH: 4.5-5.5 ( Figure 4D). The results showed that doxorubicin penetrated in agarose models (≈60 µm) more than tissue model (≈30 µm) and kidney (≈40 µm) ( Figure 4E), showing that cancerous kidney has similar behaviour to the tissue model.

Photothermal Behaviour
Photothermal behaviour of the tissue models (gel layer), ex vivo, and in vivo models was evaluated after irradiation with NIR laser. The gradual changes in signal intensity (2 to 10 min) are shown in Figure 5A. The corresponding heating profile of the samples are shown in Figure 5B. It can be seen that the raised temperature profile of the skin tissue model is higher than agarose and is similar to that of rat skin. Similar behaviour of ex vivo and in vivo models were calculated against agarose and tissue model ( Figure 5C). The results showed that the R 2 of the tissue model was much higher than agarose. However, the photothermal behaviour of the tissue was much more similar to that ex vivo and in vivo tissues. Photothermal heating and natural cooling cycles were evaluated in 4 cycles during NIR laser illumination up to 80 min ( Figure 5D). Alteration in minimum temperatures for agarose gel was observed during four cycles of laser irradiation. As can be seen, agarose did not show reproducible behaviour whereas the tissue model and rat skin exhibited similar behaviour with an approximately repeatable constant cycle, signifying their excellent photostability.
To further investigate the thermal behaviour of the tissue models, PDA as a model photothermal agent was injected and the temperature rise (∆T) was recorded. As evidenced in Figure 5E,F, our proposed tissue model exhibited similar thermal behaviour to mice skin, compared to agarose. In detail, PDA-loaded agarose showed a higher rise temperature relative to   the tissue model and mice skin tissue. Since water is the most used media for evaluation of photothermal activity, we measured the temperature changes of PDA in water as well. Although results showed that water is a better system than agarose, still our designed tissue model was more similar to mice tissue.

Laser Ablation
The use of lasers in surgery is becoming increasingly common in many medical fields. [19] Most of these laser procedures involve either excision or tissue vaporization. The specific effects on tissue depend on complex interaction between the dose of laser administered and the target tissue. In many experimental setups, such as platforms for training or research, [20] there is a need to simulate the laser ablation effect on the target tissue. These experiments typically require the use of animal tissues, which are restrictive in terms of expiration period, cleaning and disinfection, and may also impose significant costs to the overall study. A realistic phantom model for laser ablation experimentation would be a good alternative to eliminate the need for animal tissues.
Accordingly, we verified that the proposed tissue model is appropriate for laser ablation experiments, as well as being a better alternative to agarose gel, to simulate the interactions between the laser with real tissues. The results of CO 2 laser ablation tests under a common set of laser parameters was captured via OCT imaging as illustrated in Figure 5G. It is evident from the figure that the tissue model presents a laser cut profile similar to the one obtained in porcine tissue. The laser cut on the agarose model, on the other hand, showed a different ablation profile relative to porcine tissue. The relevant depths and widths of the ablation craters created during the experiment are plotted and shown in Figure 5H. As can be seen, for both dimensions, the values of the tissue model were close to the porcine ones, compared to agarose model.

Ultrasound Imaging
The proposed tissue model was tested in terms of its visibility in ultrasound imaging. Moreover, the quality of the acquired images was compared to those images taken from the agarose model and from real femoral vessels in a human brachial arm (in vivo). For this, two different views of the tissues were considered: transverse and sagittal ( Figure 6A). The obtained images show that the proposed tissue model enables good visualization of the embedded blood vessels from a transverse and sagittal perspective. This indicates that the ultrasound waves were transmitted well through the proposed tissue model. In contrast, when ultrasound imaging was conducted on the agarose tissue model, we observed poor visualization of the embedded vessels as well as resulting inconsistent ultrasound images. In addition, agarose-based tissue models were shown to be limited in size, as it needs a specific surface-to-volume ratio to solidify properly. This limits their maximum thickness to around 5 cm, which may be restrictive as a phantom for in many medical applications. [21] Although direct comparison with the real human anatomy may reveal significant differences, the acquired images demonstrate the proposed tissue model is a suitable surrogate for vascular imaging, presenting realistic features during ultrasound scanning. Finally, it is worth noting that the proposed tissue model was able to maintain its properties over time. The experiments described here were repeated over 10 days with the same prototype, with minimal to no observable changes on the ultrasound imaging results. In contrast, the agarose-based tissue was able to maintain its imaging response only up to 5 days. Besides ultrasound imaging, we also demonstrated that the tissue model is useful for OCTrelated research, allowing the visualization of different layers and subsurface structures. Figure 6B provides OCT images of an embedded thin wire within the tissue model.

Biocompatibility and Hemocompatibility
The CLSM images in Figure 7A demonstrate the healthy aspect of the human primary skin fibroblasts grown on control substrates and on the tissue model parenchyma substrates, confirming the data provided by the cell proliferation test. The cell proliferation experiments showed that the tissue model parenchyma supports the growth of human primary skin fibroblasts both at 24 and 72 h ( Figure 7B). Compared to the control growth, tissue model parenchyma reports a cell proliferation extent of 89.8% at 24 h and 109.2% at 72 h, whereas the agarose substrates exhibited a proliferation extent of 42.9% at 24 h and 73.4% at 72 h. Such data highlights that agarose exhibit low biocompatibility with the human primary skin fibroblast as the model resulted in poor fibroblast proliferation. Through "live and dead" assay we tested the amount of viable human primary skin fibroblasts on the substrates ( Figure 7C). With this assay, it is possible to discriminate the living cells from apoptotic cells.
The live and dead assay showed that there is no significant difference between the number of living cells in the control group relative to cells exposed to agarose gel and the tissue model over the course of 24 and 72 h. The graphs reported in Figure 7D,E show that the green positive cells (live cells) at 24 h have an average cell count of 97.6% ± 1.4% for the control group, 97.6% ± 0.75% for the agarose gel control and 97.1% ± 1.5% for the cells seeded on tissue model. On the other hand, the cells with redlabeled nuclei (dead cells) appear in very small amounts, with an average of 2.4% ± 1.4% for the control group, 2.3% ± 0.7% for agarose gel and 2.8% ± 1.5% for tissue model. A similar situation was also observed at 72 h. These results indicate that the tissue model exhibits good biocompatibility as the model resulted in low cell mortality under all conditions tested relative to the control arm which was seeded on pristine glass coverslip.
In order to evaluate the growth properties offered by the tissue model, a tumor cell line was also seeded on these substrates. The images showed a sustained cell growth even in the case of the use of the U87MG GFP + tumor line on the parenchyma model ( Figure 7F). In contrast, tumor cells grown on agarose substrates appeared to agglomerate into small clusters, suggesting poor adhesion to the agarose. These results demonstrate that the tissue model support the growth of both cell models employed in this study, namely human primary skin fibroblasts and tumor cells, suggesting its suitability for in vitro biomimetic studies.
The biocompatibility of the tissue model was evaluated using Cell Counting Kit-8 (CCK-8) assay [22] on the human umbilical vein endothelial cell (HUVEC) cell line. The graphs reported in Figure 7G show that the viability of HUVECs at 24 and 72 h were ≈93.5% ± 4.6% and 105.38% ± 2.2% for the tissue model relative to the control. It can be seen that after 24 or at 72 h there is no statistically significant difference between the percentages of viable cells seeded on the tissue model relative to the control. Based on these findings, this suggests that the tissue model is not cytotoxic to the HUVECs over the 72 h period. Besides that, the hemolytic property of a material is an important criterion that ought to be assessed prior to any biomedical applica-tions. [23] Therefore, we have conducted an ex vivo erythrocyte hemolysis assay in order to understand the effect of the tissue model on red blood cells. The graphs reported in Figure 7H showed that the tissue model was non-hemolytic (hemolysis below 2%) upon contact with erythrocytes over a period of 60 min. [22] C-reactive protein (CRP) is a proinflammatory induction and stimulates the production of IL-1 by monocyte. [24] Figure 7I shows the effects of the subcutaneously implanted tissue model in mice on the level of serum CRP. [25] The results indicate the prepared tissue model displayed similar CRP level in treated group compared to the control group after 2 and 7 days with CRP levels decreasing after day 7 post-surgery. [25] Adv. Mater. 2023, 35, 2210034

Immunomodulating Effects of the Tissue Model on Dendritic Cells
In order to further characterize the biological activity of the tissue model, we analyzed the activation of dendritic cells, typically involved on inflammation and immune reactions, to evaluate the possible influence of the material composition along with sample degradation. In particular, we evaluated the phenotype and the cytokine secretion in bone marrowderived dendritic cells (BM-DCs) upon exposure to the tissue model. BM-DCs were derived from bone marrow precursors and immature BM-DCs were cultivated on either the tissue sample or agarose gel for 20 h. The modulatory effect of the tissue model on the expression levels of CD80 and CD86 activation marker, indicators of DC maturation, were assayed by cyto-fluorimetric analysis. As shown in Figure 8A, the exposure of DCs to agarose and tissue sample induces the upregulation of CD80 and CD86 expression, compared to the untreated cells. Moreover, different pro-inflammatory cytokines were measured in the supernatants of BM-DCs upon 24 h of exposure to the tissue model. The secretion of IL-6 and TNF-α were also upregulated upon exposure to the tissue model ( Figure 8B,C), whereas IL-12 production is not significantly affected by tissue model or agarose gel exposure ( Figure 8D). Overall, these results demonstrate that the tissue model does have an effect in modulating the biologic function and then activation status of DCs.

Discussion
Over the years, considerable strides have been made in an attempt to fabricate stable and reproducible skin models that could function as a surrogate biological tissue. Such models aims to obviate ethical concerns, regulatory issues, cost and sample heterogeneity which are disadvantages associated with ex vivo skin tissues. [26][27][28] In line with this, several tissue models, aiming to mimic the diverse stratum of the human skin, have been reported in the literature, such as lipid bilayer, [29] lipid monolayers, [30] and polymeric-based membranes, [31,32] to name a few. For instance, polymeric films such as Parafilm M which is a blend of hydrocarbon wax and polyolefin has been utilized as a model membrane to evaluate MN insertion depth. [1] Paraffin wax has also been exploited in some studies. However, the aforementioned models cannot be utilized for investigating drug delivery as they only simulate MN insertion into the skin.
The outermost layer of the skin, the stratum corneum governs the general mechanical behaviour of skin for insertion test. [15] In our proposed tissue model, a translucent material based on silicone rubber with similar mechanical properties and hydrophobicity to stratum corneum was employed to mimic the outmost upper layer of skin. The silicone-based material was utilized as it was harder than agarose and gelatinbased tissue model. In addition, as the silicone layer is intrinsically hydrophobic this enabled this layer to mimic the highly lipophilic stratum corneum. Table 1 represents the available tissue models for potential applications in microneedle skin penetration and drug delivery applications.
Agarose gel has also been investigated as a tissue model due to its high homogeneity and semi-clarity. [33] It has a 95-99% water content higher than human skin, with only between 65% to 75% water content. [34] Such a high amount of water affects the drug release profile, for example, for microneedles and or injected drug through subcutaneous syringes. This also brings its limitation to being employed in potential applications in cancer field such as photothermal therapy. Our results showed that agarose cannot mimic in vivo and ex vivo tissues may be due to its high-water content. On the contrary, our tissue model showed similar behaviour in photothermal test, close to that of in vivo model.
Differences in measurement procedures, dimension of the probe (e.g., needle size for micropenetration test) and testing conditions are translated into a broad range of biomechanical features found in the literature. [10] The experimental evaluation of stress/strains test of skin is a fabulous challenge because of its sensitivity to environmental conditions, dimension,  [15] Since the properties of living and dead tissues are different, the available experimental and theoretical data on the viscoelastic properties of human ex vivo skin are not reliable. In particular, it is almost impossible to collect just one separated skin strata. Besides that, as these respective layers are interlocked with each other, any changes to one of these layers would affect the overall property of the model when evaluated. The scenario is far more comlex in an in vivo settings due to the presence of other tissues underlying and peripheral tissues (like bones, muscle, and fat) that can collectively affect the overall biomechanics of the skin. For instance, Moronkeji et al. [18] showed that the stiffness of subcutaneous tissue, including fat and muscle affects the performance of MN patches. The fat thickness of human hands varies in the range of 3.0-8.2 mm depending on the weight, age, activity, and ailment. For instance, the thickness of the subcutaneous layer in human is strongly affected by one's physique and age which would ultimately impact the overall properties of the skin. [35] Unfortunately, most in vitro and ex vivo models do not reflect the biophysical properties of skin or that of the subcutaneous layers. In order to overcome these limitations, we utilized silicone-based materials with adjustable thicknesses to mimic subcutaneous fat and muscle tissues. The proposed tissue model showed high similarity upon penetration of a microneedle (needle 30G) till 800 µm insertion depth. In vitro drug release and diffusion studies are valuable tools within the academic and pharmaceutical industry, particularly during the development and quality control. [36,37] To evaluate the potential application of the designed tissue model to investigate drug diffusion, we injected fluorescein and doxorubicin into the tested models. The results revealed that the profile of fluorescein distribution in the designed tissue model (gel layer) was similar to the ex vivo rat tissue model, which were different to that of agarose model. In another experiment, to mimic the tumor microenvironment for drug delivery purposes, we injected doxorubicin into cancerous kidney tissue as well as agarose and the designed tissue model with acidic pH. The results exhibited that doxorubicin diffuse to a higher depth in the agarose model (≈60 µm), compared to the kidney (≈40 µm), and tissue model (≈30 µm), showing that the designed tissue can mimic tumor tissue better than agarose (Figure 4). Such observation may be attributed to the lower water content of the designed tissue model, as well as its compactness, and porosity. In this context, considering the possibility of pharmaceutical industry acceptability, feasibility and scale-up, agarose gels evidenced a strong disadvantage compared to the tissue model, limiting their application in these experimental settings. In contrast, the developed tissue model in this work emerges as a promising alternative to biological membranes, due to the capacity to closely mimic the in vivo tissues-which is marked by ethical issues and interindividual variability. In addition, our proposed tissue model showed similar photothermal behaviour and stability when irradiated with NIR laser. Thus, such a model can be employed for evaluating photothermal properties of different compounds as the rate of temperature rise in the tissue model is similar to in vivo models.
Ultrasound imaging is extensively employed in medical applications such as for abdominal, urological, cardiac or vas-cular, gynecological and breast examination. [38,39] Tissue-mimicking phantoms with known acoustic properties, dimensions, and internal characteristics are required to operate and calibrate ultrasound imaging systems. [40,41] Tissue phantoms are commercially available for a range of medical applications, such as sonography, echocardiography, ultrasound contrast agent, or Doppler imaging. However, commercial phantoms are typically designed with specific requirements and may not cover the wide variety of medical applications in terms of system design, clinical training, or performance evaluation for different imaging modalities. [21] Some experiments, such as those for needle insertion procedures, are quite invasive and destructive which may reduce the lifespan of commercial models due to wear and tear. At the same time, silicone-based commercial phantoms fail to facilitate custom applications such as the evaluation of a contrast agent while often presupposes excessive costs and availability. [42] For all these reasons, the development of skin tissue phantoms that could be customized for ultrasound imaging is of high importance in order to meet custom specifications, targeted tissue properties, dimensions, or specialized embedded features.
The applicability of the tissue model was also demonstrated by the laser ablation experiments. By these means we can overcome the issues associated with extensive use of animal tissue for research experiments. OCT visualization showed that the ablation on the proposed tissue model was successful and the measured dimensions were very similar to those measured on real tissues. The proposed tissue model reveals important features that make it suitable for many research setups as demonstrated in this work compared to the other two models.
The interactions of the skin tissue models with immune cells were analyzed. Dendritic cells represent the most efficient antigen presenting cells, able to effectively activate way CD4+ and CD8+ T cells. [43] We found that dendritic cells in contact with the tissue model differentiate in fully competent antigen presenting cells with upregulation of CD80 and CD86 costimolatory markers. In addition, these cells produce IL-6 and TNF-α, modulating the biological function and activation status of DCs, [44,45] but not the IL-12, a cytokine involved in IFN-gamma production. [46]

Conclusion
Despite considerable developments in the field of microneedles, the complexity of the skin biophysics coupled with the intrinsic elasticity of the stratum corneum still poses some challenges for the prediction of microneedle penetration prior to clinical trial. In light of this, in vitro tissue models can pave the way for the analysis of transdermal patches before in vivo experiments. Herein, we developed a tissue model that can mimic the stiffness of the stratum corneum for MN insertion test. The results showed similar behaviour to ex vivo models. In addition, the designed tissue model is able to mimic drug diffusion profiles and photothermal behaviour similar to that within in vivo organs (e.g., cancer-infected kidney). Consequently, this designed tissue can be used for the evaluation of the photothermal activity of a model drug. Finally, the designed and customizable tissue model showed good results when analyzed via ultrasound and OCT imaging enabling it to mimic vascular visualization in a manner which is similar to in situ visualization of human blood vessels in vivo. Overall, the engineered skin model may serve as a promising model for both drug delivery analysis and biodiagnostics. Although, our tissue model is not proposed for any therapeutic applications or to be implanted in human/animal body (e.g., regenerative medicine or cancer therapy), they may be loaded with cells due to its biocompatibility.

Experimental Section
Fabrication of Tissue Model: The tissue model was fabricated by a two-step process, that is, preparation of the gel and silicone layers and subsequent integration of the layers.
Topmost Layer (Stratum Corneum): Dragon Skin 20A, 30A, Mold Max 30, Smooth-Sil 960, and PDMS have an acceptable pot life (>30 min). However, the pot life of Mold Star 20T was about 6 min, which was quite short. Therefore, before mixing part A and B of Mold Star 20T resin, they were put in a refrigerator to cool down to 0 °C. Then, part A and B of the resin were mixed using a mixer (Thinky, ARE-250, Japan) and put in a vacuum oven to remove the air bubbles (vacuum degassing). Subsequently, the resin was coated using automatic film applicators (Zhentner, Proceq ZAA2300, Switzerland) on polymeric sheets, for example, teflon or poly(ethylene terephthalate) (PET) sheet. The scan speed (the blade speed) was 20 mm s −1 for coating the silicone resin with a desired thickness (from 20 to 50 µm). Post-curing (80 °C for 2 h) was recommended as it can aid in attaining higher mechanical properties of the silicone layer. Subsequently, the film was delicately peeled off, and then it was covered and fixed on the gel layer using a very thin layer of adhesive (Loctite 406, Henkel, Germany) or PVA solution (25 wt%). The final tissue model was stored in a refrigerator in a closed container to avoid vaporizing of water as well as to be preserved for about 30 days.
Hydrogel Layer (Viable Epidermis and Dermis): The hydrogel layer consisted of sodium chloride (NaCl), gelatin as protein, and acacia gum as polysaccharides, whist exhibiting an overall pH of 7.4 similar to physiological conditions. These materials were selected to mimic the respected biomaterials in human skin. Sodium chloride was the highest available salt in interstitial fluid and was used in place of all salts in our model. [47,48] Acacia gum (Arabic gum) was an anionic carbohydrate polymer that was employed in place of hyaluronic acid to fabricate a more affordable skin model. [49] To fabricate this tissue model, NaCl (0.9 wt%) and acacia gum (1 wt%) were mixed in PBS buffer containing PVA (5 wt%, 13-23 kDa). It was worth noting that in order to speed up the solubilization process, it was recommended to have a stock solution of concentrated PVA because PVA granules/powder may need overnight to be solubilized. Then, gelatin (27.6 wt%) was added and the mixture was heated at 40-50 °C till dissolution of the mixture. Then, glycerol (5 wt%) and agarose (1 wt%) were added to the previous solution. A preservative agent preferably should be employed to prevent gel contamination by microorganisms such as fungi, yeasts, molds, and bacteria. Accordingly, 1 wt% of amphotericin B solution (250 µg mL −1 in deionized water, Sigma Aldrich) or 0.05 wt% of sodium azide (NaN 3 ) should be added to the mixture as an anti-mold agent. Subsequently, the mixture was put in a microwave for 30 s to reach boiling temperature. The solution was immediately poured into a mold placed on the silicone film (the topmost layer) covered with a very thin layer of adhesive to enhance the integration of the layers. The gel (2 mm thickness) was left to cool down to room temperature to solidify which may take about 30 min. After demolding, the gel can be integrated and fixed on the subcutaneous silicone tissue using a very thin layer of adhesive (Loctite 406, Henkel, Germany) or PVA solution. The fabrication process of the tissue model is shown in Scheme 1.
Characterization: SEM images of the samples were obtained with a Dual Beam FIB/SEM Helios Nano-Lab 600i (ThermoFisher Scientific, Hillsboro, OR, USA). SEM images of each layer were obtained with accelerating voltage 5 kV. Gel samples were prepared via freeze-drying to minimize morphological change.
The rheological parameters, that is, elastic (G′) and viscous (G″) moduli were examined in a frequency range of 0.01-10 Hz. The measurements were carried out through a rotational rheometer (MCR302 Rheometer, Anton Paar, Waltham, MA, USA), using the smooth parallel plate geometry at 25 °C. For porcine skin, sandblasted plates were used to prevent slippage. In order to identify the linear viscoelastic response range of the materials, preliminary strain sweep tests were performed on the samples, at the oscillation frequency of 1 Hz. During the tests, the samples were placed into a chamber to mitigate solvent evaporation. Steady-state shear test in term of low curves was performed to evaluate the dependence of viscosity upon the shear rate. The tests were repeated at least three times on each sample.
The behaviour of the tissue model was investigated through the tensile test. A molded dog bone sample with 3 mm thickness was prepared with dimensions introduced by the ASTMD-412-C standard. [50] The tensile force was applied by a Universal Testing Machine (Z005, ZwickRoell, Ulm, Germany) and measured by the attached load cell (Xforce-P, 1 kN, ZwickRoell, Ulm, Germany). The grippers separation/ approaching speed was 20 mm min −1 . The force varied from 0 to 0.6 N during 20 cycles.
Impedance Spectroscopy: Impedance measurement was carried out for porcine skin, agarose, silicone, gel/silicone models. A concentric needle (length 25 mm, diameter 0.45 mm, GA24A31, SEI EMG s.r.l., Cittadella, Italy) was inserted into the sample and impedance spectroscopy was measured using an LCR meter (Bench LCR Meters Model 895, B&K Precision Cooperation, Yorba Linda, CA, USA).
Transient impedance measurement was carried out on the tissue model under saline injections to evaluate the potential utility of the tissue model for biosensing applications. The concentric needle was inserted into the sample and impedance spectroscopy was measured using an LCR meter. For saline injection, two syringes were filled with saline solution while a 24G plastic catheter (Introcan, Braun, Melsungen, Germany) was attached for each. In this process, the use of normal metal needles for injection was avoided. This was because the presence of any metallic component in the gel may result in erroneous impedance reading. The catheters were inserted into the gel followed by the injection saline solution through the catheter. The tissue model had physiological saline concentration (0.9%), and thus to change the impedance of the tissue model concentrated saline solutions (≈1.8-3.6%) were injected.
Microindentation and Micropenetration Tests: Microindentation and micropenetration tests were performed on different tissue models. The displacement of the indenter or penetrator was controlled by a displacement (precision linear stage L-509.20AD10, PI, Germany). Micropenetration test was conducted using stainless steel needle 30G (inner diameter: 150 µm, outer diameter: 300 µm) at ambient temperature (22-24 °C). A load cell (LRF400, Futek Advanced Sensor Technology Inc., Irvine, CA, USA) with a maximum force 1000 mN was employed. The experiments were conducted at penetration speed V = 5 µm s −1 . During the perforation, the variation of the penetration force as a function of the penetration depth was measured.
The indentation test was performed with the same instrument but containing a steel cylinder indenter (diameter 1.5 mm). The difference between penetration and indentation analysis was that in the indentation test the indenter did not perforate the samples. [16] The microindentation experiments were conducted at indentation speed V = 50 µm s −1 . During the indentation, the variation of the indentation force as a function of the indentation depth was measured. Details about the experimental setup, associated testing procedures, and mechanics analysis for determining the indentation test can be found in the Supporting information. The indentation test was also performed on the index fingers of 32-and 34-year-old men as well. After each test, the indentors were replaced and sterilized with ethanol. The skin of volunteers before and after the tests was sanitized as well.
Microneedle Penetration and Imaging: A blank hydrogel-forming microneedle array patches (MAPs) were used as a model to evaluate the mechanical properties of the skin model, compared to agarose and neonatal porcine skin. Hydrogel-forming MN arrays were prepared using silicone mold with pyramidal needle density of 16 × 16, 850 µm height, 300 µm width at the base and 300 µm interspacing. In this study, hydrogel-forming MAPs were prepared from an aqueous blend containing 20 wt% Gantrez S-97 and 7.5% PEG 10000. The insertion and penetration depth of hydrogel-forming MAPs were then determined using an optical coherence tomography (OCT) microscope (Michelson Diagnostics Ltd., Kent, UK), as reported previously, [17] following insertion into full-thickness neonatal porcine skin, Parafilm M, skin model, and agarose (1.5 wt%). ImageJ (National Institutes of Health, Bethesda, MD, USA) was used to measure the height of needles inserted. [51] Simulation: Finite element simulation, as a powerful engineering tool, helped to improve the sight to the behaviour of the skin. Here, the static structural simulation of the skin model is performed by the software ANSYS Workbench 19.2. A simple indentation test was simulated numerically and results were compared with a real test. The suitable coincidence between the simulation results and experimental data validated the utilized material model of the skin, simulation procedure, and further predictions of the skin model behaviour based on simulations.
The material model, employed to describe the hyperelastic behaviour of the skin, strongly affects the simulation results. Uniaxial compression test data of a 5 mm × 5 mm × 5 mm cubic specimen of the tissue model was used to extract the coefficients of the Ogden 1st order hyperelastic model by curve fitting. On the other hand, the indentor was modeled with linear elastic structural steel. In the simulation, a 20 mm cylindrical indentor with a 3 mm diameter was pushed into a 50 mm × 50 mm face of the skin model with 10 mm thickness while the opposite face was fixed. The maximum indentation value was 2 mm. The indentor and skin specimen both were meshed by 20 node hexahedral elements (SOLID186). This mesh, among other discretization adopted, had turned out to be the best, in terms of convergence and consistency of the results. The indentor's displacement caused strain and stress in both skin and indentor. Since the indentor was strongly rigid, compared to the skin, its deformation was neglected. The simulation was compared with an experimental indentation test where the initial thickness of the skin model specimen was 5 mm, the indentor's diameter was 1.5 mm, the indentation velocity was 50 µm s −1 , and the maximum indentation was 1000 µm.
Drug Deposition and Diffusion Study: Fluorescein as fluorescent agent was injected to the test models (e.g., agarose, tissue model, and the hypoderm of a rat via a syringe). Next, each sample was illuminated by black light and then fluorescent images were taken by a digital camera after post injection, 2, 4, 6, and 8 min. The corresponding curves were plotted using MATLAB software.
In another test (doxorubicin diffusion analysis), a total of 5 × 10 6 Caki-1 renal cancer cells were injected subcutaneously into the flank of Male BALB/c mice (4-5 weeks old). 1 week after cell inoculation, all mice developed single subcutaneous palpable tumorous of ≈50-100 mm 3 . Mice were arbitrarily placed in the groups of treatment. The mice were sacrificed on the 8th day after the injection. The targeted main organ (kidney) was separated, then, doxorubicin (10.5 µg mL −1 ; 2 mL; deionized water) was injected on the surface of the separated kidney (with a depth of 2-3 µm using an adjustable needle). In the next step, the kidney was fixed in 4% (W/V) PBS-buffered paraformaldehyde (PFA) overnight, and then embedded in paraffin. The paraffin-embedded tissues were cut into slices (each step with 10 µm diameters sliced with microtome-serial sections (10 µm) were prepared with a rotary microtome, Leica RM 2135, Germany) for hematoxylin and eosin (H&E) staining.
Photothermal Behaviour: Delivery of bioactive compounds has been widley used in nanomedicine. Understanding drugs diffuse through tissues is essential for optimizing their therapeutic efficacy, drug formulation, dosage, and delivery route. We evaluated diffusion of fluorescein and doxorubicin in the tissue models. Each sample was exposed by NIR laser (808 nm wavelength) at 1 W cm −2 laser power for 10 min. The temperature of the samples was determined by a thermal camera (Teledyne FLIR, Wilsonville, OR, USA) at an interval of 1 min. Similarly, polydopamine (PDA, 100 µL, 0.03 wt%) as photothermal agents were injected into agarose, tissue model, mouse, and at models and irradiated by NIR.
The animals were provided from the animal facility of Zanjan University of Medical Sciences (ZUMS) (200-220 g). They were kept under standard conditions with free access to water and food. The ethics guidelines were strictly followed for animal care and use approved by ZUMS, Iran (Protocol approval number: IR.ZUMS.REC.1400.489).
Laser Ablation: A set of experiments were implemented under controlled conditions, using a surgical laser source to study the effects of the laser parameters on the resulting incision depth of the three studied models, namely the proposed tissue model, the agarose phantom model, and a porcine skin model.
For the laser ablation of the different tissue models, a commercial surgical laser source, a SmartXide 2 Carbon dioxide (CO2) laser by DEKA (DEKA M.E.L.A. srl, Florence Italy) was used, that possessed a configurable power setting ranging from 2 to 25 W. In the system used in this research, the laser power was set to 5.0 W in all the experiments and the laser spot consumed a total of 4.5J energy within a timeframe of 0.9 s.
Ultrasound and Optical Coherence Tomography Imaging: Ultrasound was widely used in medical imaging due to its properties, including being safe, relatively cheap, and capable of providing real-time images of deep anatomy. It was also particularly useful as an assistive technology for challenging percutaneous procedures, offering critical guidance information during operations such as needle biopsy, precise delivery of treatments, or catheterizations. [52] In fact, the evolution of ultrasound-guided methods and associated assistive technologies were active research areas that could largely benefit from realistic tissue analogues. Therefore, it was useful to demonstrate the performance of our tissue models with this important medical imaging technique. In this respect, standard parameters of tissue models for ultrasound imaging are described in the literature. These included sound speed, attenuation coefficient, acoustic impedance, density, and elasticity, which can be tuned to simulate specific medical applications. [53,54] For a comparative assessment of ultrasound imaging compatibility, both the proposed tissue model and the agarose model were produced to be of the same size (5 × 10 × 3 cm). Vein analogues based on silicone tubes (10 cm length with inner and outer diameter of 0.2 cm and 0.4 cm, respectively) were embedded in both models during the fabrication process. The veins were placed in the same depth (1.7 cm from the top) on both tissue models. Synovial fluid had been used as blood analogue, as this was the recommended fluid for use with commercial phantom models for ultrasound imaging. [55] In this study, the ultrasound images were obtained using a high-frequency ultrasound machine, (GE Healthcare's Venue Go, USA). Images were acquired using a linear array ultrasound transducer (probe model: 12L-RS) with 5-13 MHz bandwidth.
OCT had garner considerable attention over the year as a potential tool for diagnosis, management, and monitoring of many retinal diseases or neovascular disorders. [56,57] In contrast to ultrasound, OCT exhibited a narrower field of view of 0-2 mm. OCT trials were performed in the lab, using a commercial OCT device by Thorlabs, Inc. (US). The tissue model investigated using OCT had a dimension of 4 × 5 mm and contained two embedded wires of 100 µm diameter. For optimal visualization, before each scan, the focal length and the scanning parameters were specified.
In Vitro Viability: A drop of 200 µL of hot-liquid gel (agar 2% and skin model parenchyma) was disposed on glasses of 1.9 cm 2 diameter, and when the agarose gel and skin model parenchyma were solidified, they were freeze-dried for 24 h. After, the substrates were exposed to UV light for 20 min. Then, 15 × 10 3 of human primary skin fibroblasts were seeded on the different substrates and on pristine glass, used as control, in 24-well plates with 500 µL of Dulbecco's modified Eagle's medium (DMEM) high glucose (Gibco, Waltham, MA, USA) supplemented with 10% heat-inactivated fetal bovine serum (FBS) (Sigma-Aldrich), 1:100 penicillin/streptomycin (pen/strep), and 1:100 sodium pyruvate (stock 100 mm; Gibco).
The proliferation assay was performed 24 and 72 h after seeding. WST-1 (abcam) reagent was added in DMEM high glucose complete medium without phenol red to a final dilution factor of 1:20, and 300 µL of this solution was replaced in each well. After 40 min of incubation at 37 °C and 5% CO 2 , the solutions were collected in transparent 96-well plate and their absorbance read at 440 nm with a Victor X3 plate reader. Each experimental class was then normalized to the absorbance value of their corresponding control.
15 × 10 3 of human primary skin fibroblasts were seeded in 24-well plate (Ibidi, Gräfelfing, Germany) on glasses of 1.9 cm 2 diameter previously coated with two different substrates (agar 2% and skin model parenchyma), and on pristine glass coverslips in 500 µL of DMEM high glucose complete medium as previously described. After 24 and 72 h from the seeding, the samples were fixed with 300 µL 4% PFA (Sigma-Aldrich) for 20 min at 4 °C. Samples were then washed three times with 500 µL DPBS (Sigma-Aldrich); thereafter, a permeabilization step with PBS and 1:1000 Triton ×100 for 10 min was performed. 300 µL of DPBS solution with 1:100 TRITC-phalloidin (Sigma-Aldrich) for f-actin staining and 1:1000 Hoechst (Invitrogen) for nuclei staining were added in each well, and incubated for 2 h at 37 °C. Three washing steps with DPBS concluded the staining procedure. Images were acquired with a confocal laser scanning microscope (C2s system, Nikon) equipped with a 10× objective.
The ability to support the cell growth of the skin model parenchyma was also verified by seeding a human tumor line on the substrates. To this aim, 20 × 10 3 of U87-MG expressing green fluorescent protein (GFP) (Cellomix) were seeded in 24-well plate (Ibidi) on the different substrates as previously described, in 500 µL of DMEM high glucose with HEPES (Gibco) complete medium. After 72 h, cells were treated with 1:1000 Hoechst (Invitogen) for 20 min in DMEM complete medium and three washes with DMEM complete medium were performed after the staining step. The samples acquisition was obtained with confocal laser scanning microscope (C2s system, Nikon) with a 10× objective.
In order to biocompatibility of the tissue model, a live and dead assay was carried out. As reported before, the substrates were obtained by depositing 200 µL of hot-liquid gel (agar 2% and skin model parenchyma) on glass coverslips of 1.9 cm 2 diameter (Thermo Fisher) prior to lyophilization. Before the seeding the substrates were exposed under UV light for 20 min. Pristine glass coverslips were employed as control. 15 × 10 3 of human primary skin fibroblasts were seeded on the different substrates. Cells were cultured for 24 and 72 h in 24-wells plates (Ibidi, Gräfelfing, Germany) with 500 µL of DMEM high glucose without phenol red (Gibco) supplemented with 10% FBS (Sigma-Aldrich), 1:100 pen/strep (Euroclone), and 1:100 sodium pyruvate (stock 100 mm; Gibco).
In order to perform the assay, the cell culture medium in each well was replaced with 500 µL of staining solution according to the live/ dead staining kit (Invitrogen) instructions. Briefly, a working solution was obtained adding calcein-AM (green fluorescent dye, 1:1000), ethidium homodimer-1 (red fluorescent dye, 1:1000) plus Hoechst 33342 (nuclei blue dye, 1:1000, Invitrogen) to DPBS with Ca 2+ and Mg 2+ and this solution was incubated with the cells for 40 min at 37 °C and 5% CO 2 .
The samples' acquisitions were obtained with confocal laser scanning microscope (C2s system, Nikon) with a 10× magnifying objective. The cell count was carried out with ImageJ software and the data were normalized considering the total number of the cells in the sample. Both live and dead cells were expressed as % with respect to the total number of cells.
For better assessment and evaluation of biocompatibility, the viability, as well as, cell proliferation using CCK-8 using HUVEC were evaluated. Cells were seeded on skin tissue model and well as control in 96-well plates at 5 × 10 3 cells per well. Cell proliferation was investigated after 1 and 3 days using CCK-8 assay. After preparing the wells with known numbers of viable cells, 10 µL of the CCK-8 solution was added to each well of the plate to create a calibration curve. The plate was incubated for 1 to 4 h in the incubator, and the absorbance was measured at 450 nm using a microplate reader. The calibration curve was prepared using the data obtained from the wells that contained known numbers of viable cells.
C-Reactive Protein Assay: The skin tissue model sample was punched using 5 mm circular biopsy punch and implanted into a subcutaneous pocket in the left dorsum of the animals. The procedure was performed under anesthesia by intraperitoneal injection of xylazine/ ketamine. In the control arm only the created wounds were sutured. The animals were euthanized on day two and on day-7 post-surgery by anesthetic overdose followed by exsanguination via cardiac puncture in order to collect blood for analysis. The collected blood samples were centrifuged at 3000 rpm for 10 min. Then, after the separation of serum at room temperature, serum was analyzed for CRP levels, using the CRP kit by an automatic analyzer. The CRP level was expressed in mg L −1 of serum.
Statistical Analyses: Statistical analyses were performed using OriginLab Software (SR1, 2018) or GraphPad Prism 7.0. Statistical evaluations among groups were analyzed using the one-way analysis of variance, while statistical evaluations between two groups were analyzed using unpaired two-tailed t-test. p < 0.05 was taken to indicate significance.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.