Development, validation and testing of a human tissue engineered hypertrophic scar model

Adverse hypertrophic scars can form after healing of full-thickness skin wounds. Currently, reliable animal and in vitro models to identify and test novel scar reducing therapeutics are scarce. Here we describe the development and validation of a tissue-engineered human hypertrophic scar (HTscar) model based on reconstructed epidermis on a dermal matrix containing adipose derived mesenchymal stem cells (ASC). Although obtained from normal, healthy skin, ASC, in contrast to dermal mesenchymal cells, were found to facilitate HTscar formation. Quantifiable HTscar parameters were identified: contraction; thickness of dermis, collagen-1 secretion, epidermal outgrowth, epidermal thickness, and cytokine secretion (IL-6, CXCL8). The model was validated with therapeutics currently used for treating scars (5-fluorouracil, triamcinolon) and a therapeutic known to be unsuccessful in scar reduction (1,25-dihydroxyvitamin-D3). Furthermore, it was shown that atorvastatin, but not retinoic-acid, may provide a suitable alternative for scar treatment. Each therapeutic selectively affected a different combination of parameters, suggesting combined therapy may be most beneficial. This animal-free hypertrophic scar model may provide an alternative model for mechanistic studies as well as a novel in vitro means to test anti-scar therapeutics, thereby reducing the use of animals.


INTRODUCTION
Cutaneous wound healing is a natural complex response to tissue injury and normally results in a scar. The most desirable scar is thin and at and is mostly seen after super cial injury. This type of scar is called a normotrophic scar (NTscar). However extensive trauma, deep burns and sometimes even standard surgery can result in wound closure with an abnormal scar formation which is red, rm, raised, itchy and painful. This abnormal scar is known as a hypertrophic scar (HTscar) 1 . The quality of life of patients with HTscars can be severely a ected due to loss of joint mobility, contractures and dis gurements which lead to accompanying psychological problems (like depression and social avoidance) 1 .
HTscars occur more often after full-thickness wounding, where no viable dermis is left and adipose tissue is exposed. Therefore the deeper the wound, the greater the possibility of HTscar formation 2 . Super cial wounds generally heal with a NTscar. The pathogenesis of HTscar formation in humans is not well understood and although there are various treatment strategies, it is generally accepted that current strategies are still far from optimal 3,4 . A major limitation in the progress of scar management is the lack of physiologically relevant human models to explore the pathogenesis of HTscar formation and to test new therapeutics. Nowadays patients, animal models and in vitro cell culture models are used to study skin scar formation. Patient studies are essential, but are limited due to logistical and ethical problems. Common alternatives are animal studies. Despite the large number of studies describing pigs, mice, rabbits, and other animals as models to investigate hypertrophic scarring, the wound healing process in these species presents signi cant di erences when compared with human scarring 5 . Pig skin most closely represents human skin and the red durac pig model has recently been validated since these pigs have been described to develop HTscars similar to human HTscars in a number of ways 6 . However, extensive research with this model is limited due to the lack of pig speci c biomarkers such as detected by monoclonal antibodies. Rabbit skin also shows some similarities to human scar formation, however the rabbit ear scar model 7 encounters similar restrictions to the pig model. Mouse models are most extensively used even though mouse skin physiology poorly represents human skin, and mice do not form adverse scars after wounding. Therefore in order to humanize mouse models, studies have been described using CXCR3 −/− mice 8 and transplanting human skin onto the backs of nude mice 5,9 . In addition to di culties in interpreting results due to di erences in skin physiology and in particular scar formation, in icting large full thickness trauma and burn wounds to animals has substantial ethical consequences worldwide. In vitro cell culture models have been used to gain insight into di erent aspects of scar pathogenesis. For example adipose tissue-derived mesenchymal cells have been described as having a number of similar characteristics to mesenchymal cells found within HTscar tissue e.g.: both are α-SMA [10][11][12][13] . Also a scratch assay has been described in which an increase in the single parameter connective tissue growth factor (CTGF) has been proposed for testing scar therapeutics 14 . However, no attempts have been made so far to create a robust and physiologically relevant in vitro HTscar model for in vitro testing of therapeutics with multiple scar forming parameters. With increasing pressure from the EU (Directive 86/609/EEC) who strongly stimulate the replacement, reduction, and re nement of the use of animals models, there is an urgent need to develop a physiologically relevant in vitro human HTscar model, in order to investigate the pathogenesis of HTscar formation. This in turn can facilitate identifying and testing new therapeutics, and thus lead to novel treatment strategies. Therefore, we have developed and validated a tissue-engineered HTscar model consisting of a reconstructed epidermis on a dermal matrix populated with mesenchymal cells. We compared full-thickness skin equivalents (SE) constructed from mesenchymal stem cells isolated from the deep cutaneous adipose tissue (ASC) with SE constructed from more super cial mesenchymal stromal cells found within the reticular dermis (R-DSC) and papillary (P-DSC) in order to mimic HTscar formation, NTscar formation and Nskin respectively. We hypothesized that ASC in the exposed wound bed might most rapidly regenerate dermal tissue in order to close life threatening deep cutaneous wounds at the cost of HTscar formation whereas more super cial wounds are repaired from DSC within the anking and underlying dermis generally resulting in NTscar formation.
In order to develop, validate and further test the HTscar model, a number of quantiable parameters typical for HTscars were identi ed: 1) contraction since HTscars are highly contractile 15 ; 2) thickness of the dermis and 3) collagen-1 secretion since more connective tissue is formed in HTscars than in NTscars 16 ; 4) the degree of epithelialization, since it has been described that the extent of HTscar formation corresponds with delayed wound closure 2 ; 5) thickness of the regenerating epidermis, since it is known that HTscars have more epidermal cell layers than NTscars 17 . In addition to the scar forming parameters we assessed the secretion of two cytokines, IL-6 and CXCL8, known to contribute to wound healing 18 . The HTscar model was validated with therapeutics generally used in the clinic for scar treatment (5-Fluorouracil and a triamcinolone (kenacort®-A40)) 19,20 and a therapeutic known to be unsuccessful in scar reduction (1,25-dihydroxy vitamin D 3 ) 21 (Table 1). The HTscar model was further tested with two potential scar reduction therapeutics (All-trans-retinoic acid and atorvastatin calcium salt trihydrate) 20,22 .
In vitro hypertrophic scar model

Normal skin and scar tissue
Human adult skin samples were obtained from healthy individuals undergoing abdominal dermolipectomy or breast reduction surgery (n = 9; age: 25-50 years; sex: 8 x female, 1 x male). Scar tissue samples were obtained from patients who underwent plastic surgery for scar excision (HTscar n = 8; age: 25-55 years; sex: 7 x female, 1 x male; location: abdomen, breast and ank; age of scar: >1 year and NTscar n = 7; age: 15-60 years; sex: 6 x female, 1 x male; location: abdomen and breast; age of scar: >1 year). HTscars were de ned as raised above skin level (>1 mm) for at least 1 year and NTscar were de ned as never raised above skin level. VU University medical center approved all the experiments described in this manuscript. The study was conducted according to Declaration of Helsinki 1975.
Cell isolation and culture of normal healthy skin Epidermal keratinocytes were isolated from healthy (non scarred) human adult skin and cultured as described earlier 23 . Keratinocytes were cultured until 80% con uency and then stored in the vapor phase of liquid nitrogen for later use.
Papillary dermal, Reticular dermal and Adipose tissue-derived mesenchymal cells were isolated by collagenase type II / dispase II treatment from healthy (non scarred) human adult skin as previously described by Kroeze et al 24 . In short, split thickness skin (0.4 mm) was removed using a dermatome (Acculan II, Braun, Tuttligen, Germany) to separate the papillary dermis from reticular dermis and adipose tissue 25 . The cells in the papillary dermis (upper layer) are further referred to as papillary dermal derived mesenchymal stromal cells (P-DSC). From the remaining reticular dermis all adipose tissue was removed. Cells in the reticular dermis are further referred to as reticular dermal derived mesenchymal stromal cells (R-DSC). Adipose derived mesenchymal stem cells (ASC) were isolated in the same way as P-DSC and R-DSC. All mesenchymal cells were cultured under identical conditions and upon reaching 80% con uency were stored in the vapor phase of liquid nitrogen until required. Notably within a single experiment KC, P-DSC, R-DSC and ASC were all from the same donor. Cells at passage 3 were used to construct SE and DE. Of note P-DSC and R-DSC are the same cell population often referred to as dermal broblasts 24 .

Application of therapeutics
SE containing ASC were generated as described above with the addition of therapeutics supplemented in culture medium from the rst medium renewal after starting the culture onwards. The constructs were cultured with 10 −7 M all-trans-retinoic acid . Corresponding vehicles were used as controls. The concentrations were determined from dose response studies on ASC monolayers and were the concentrations where no inhibition of ASC metabolic activity, which corresponds to proliferation (2 days exposure), was observed by MTT assay (see below).

Histological and immunohistochemical analysis
Para n embedded sections of normal tissue, scar tissue and SE were used for morphological (haematoxylin and eosin staining) and immunohistochemical analysis (alphasmooth muscle actin (α-SMA) (clone 1A4; 1:200, Dako, Glostrup, Denmark)) 23 .The dermal thickness of SE was quanti ed from photos of H&E stainings (Nikon Eclipse 80i Düs-seldorf, Germany) taken at 200 fold magni cation using NIS-Elements AR 2.10 software. The epidermal thickness was quanti ed by taking the mean of the number of living cell layers at 5 di erent regions within a single tissue section .

Measurement of matrix contraction and outgrowth of epidermis
Matrix contraction and outgrowth of the epidermis were determined by taking photographs of the constructs at the rst medium change and then again at the time of harvesting of the cultures. Photographs were taken with a Nikon coolpix 5400 digital camera (Japan). The surface area of the constructs and the outgrowth of the epidermis outside of the original 1 cm diameter seeding area were determined using NIS-Elements AR 2.10 imaging software (Nikon).

Keratinocyte migration
Chemotactic migration of keratinocytes towards DE conditioned medium (dose response of 0,3%, 3% and 30%) with the aid of a modi ed Boyden well chamber technique using a 24-transwell system with 8μm was assessed and quanti ed as previously described 26 .
Cell proliferation ASC: A MTT assay was used to measure mitochondrial activity, which is representative of viable number of ASC 27 . The assay was performed as described by the supplier (Sigma-Aldrich, St. Louis, MO, USA).
Keratinocytes: Relative proliferation of keratinocytes was determined by quantifying the amount of house keeping enzyme lactate dehydrogenase (LDH) released into the supernatant after 100% cell lysis with 0,1% Triton X-100 as earlier described by Kroeze et al 26 .
Enzyme-linked immunosorbent assay for cytokine production All reagents were used in accordance to the manufacturer's speci cations. For collagen I quanti cation commercially available ELISA antibodies and recombinant proteins obtained from Rockland, Gilbertsville, PA, USA were used. For IL-6 commercially available paired ELISA antibodies and recombinant proteins obtained from R&D System Inc. (Minneapolis, MN, USA) were used. For CXCL8 quanti cation, a Pelipair reagent set (CLB, Amsterdam, The Netherlands) was used.

Statistical analysis
At least three independent experiments were performed with each experiment being from a di erent donor and having an intra experimental duplicate. Importantly all experiments using KC, P-DSC, R-DSC and ASC were donor matched and performed in parallel. Di erence in thickness and contraction of the matrix, outgrowth of the epider-mis, number of epidermal cell layers and collagen 1 secretion were compared between the di erent constructs using a repeated measures ANOVA test followed by Bonferroni's multiple comparison test. Di erence in number of epidermal cell layers in native skin and scar tissues were compared using a one-way analysis of variance test, followed by Bonferroni's multiple comparison test. Di erences in biomarkers of ASC model treated with therapeutic (compared to vehicle control) were compared using paired t-test. Differences were considered signi cant when p < 0.05.

Qualitative macroscopic and microscopic comparison of native scars with the in vitro HTscar model
In order to determine which characteristics are typical for a HTscar we rst compared HTscar with native human NTscar and Nskin. Macroscopically, HTscar is more raised and red than NTscar and Nskin ( Figure 1a). Microscopically, HTscar has a thicker epidermis than NTscar and Nskin. Rete ridges are almost absent in HTscar and occur to a lesser extent in NTscar compared to Nskin 15 (Figure 1a). In order to identify the presence of myo broblasts, which are thought to be mainly responsible for skin contraction after wounding, an α-SMA staining was performed. In HTscars α-SMA positive staining was not only observed around blood vessels but also in single cells in lower regions of the dermis. In contrast, both in NTscar and Nskin α-SMA staining was mainly restricted to blood vessels (Figure 1a).
Next we determined whether the SE constructed with either ASC, R-DSC or P-DSC showed typical macroscopic and microscopic characteristics of HTscar, NTscar and Nskin, respectively. Macroscopically, the SE ASC are more contractile than SE R-DSC and SE P-DSC (Figure 1b). Similar to HTscar, microscopical examination of tissue sections showed that SE ASC had increased thickness of the epidermis. This was not observed with SE R-DSC and SE P-DSC . There was increased α-SMA staining in SE particularly in SE ASC , where it is mainly located directly underneath the epidermis. The α-SMA staining was much less and spread throughout the dermis in SE R-DSC and SE P-DSC (Figure 1b).
Clearly SE constructed with ASC populated matrixes represent HTscars both macroscopically and microscopically, and have the potential for use in an in vitro HTscar model. In contrast R-DSC and P-DSC visually represent NTscar and Nskin, respectively. Before the HTscar model can be implemented, quanti able and relevant parameters typical for HTscar need to be identi ed. Therefore we next determined whether: thickness of dermis; contraction; collagen 1 secretion; number of epidermal cell layers and outgrowth of epidermis were suitable parameters. In addition we determined whether the secretion of two cytokines, IL-6 and CXCL8, related to wound healing di ered in the 3 di erent models.

Identi cation of dermal parameters in HTscar model
In skin wound healing, the development of HTscar is characterized by an overproduction of extracellular matrix, increased contraction and augmented α-SMA expression compared to NTscar 15 . For this reason we rst compared SE ASC with SE R-DSC and SE P-DSC with regards to thickness of the dermis, contraction and collagen 1 secretion (Figure 2).
The dermal thickness was not signi cantly di erent between the three SE ( Figure  2a). An increase in contraction, is represented by a decrease in surface area of the SE. The contraction in SE ASC was increased compared to SE R-DSC and SE P-DSC (Figure 2b). SE ASC secreted signi cantly more collagen 1 compared to SE P-DSC (Figure 2c).
From these results, contraction and collagen 1 secretion were identi ed as suitable dermal parameters for assessing HTscar formation in vitro using SE. It was observed that native HTscar had a thicker epidermis than NTscar and Nskin (Figure 1a). This observation was con rmed by quanti cation of the number of epidermal cell layers: HTscar showed more epidermal cell layers (7.9 ± 1.6) than NTscars (6.9 ± 1.0) and Nskin (5.8 ± 0.6) (Figure 3a). Next we determined whether this increased epidermal thickness in native epidermis also occurred in the HTscar model. Indeed, SE ASC had increased number of epidermal cell layers (8.00 ± 1.3) compared to SE R-DSC (6.5 ± 0.6) and SE P-DSC (5.3 ± 1,1) (Figure 3b). Notably, all of these ndings correlated very closely to native tissue and in particular HTscars had the same number of epidermal cell layers as SE ASC .
Since the probability of HTscar formation is increased in wounds with delayed wound closure 2 , we next determined whether ASC were responsible for the delayed epidermal outgrowth compared to DSC. Indeed, SE ASC had signi cant slower outgrowing epidermis compared with SE R-DSC and SE P-DSC (Figure 3c). However since the contraction is also greater in SE ASC compared with SE R-DSC and SE P-DSC it could not be entirely excluded from these ndings that contraction confounded this result. To exclude the confounder a chemotactic transwell migration experiment was performed with keratinocytes using conditioned supernatant derived from the three types of DE. The keratinocyte migration was reduced with supernatant derived from DE ASC compared with supernatants derived from DE R-DSC and DE P-DSC (Figure 3d). The parallel proliferation experiment showed that this decrease in migration was not due to changes in keratinocyte proliferation (Figure 3e) indicating that ASC do indeed stimulate less epidermal migration than R-DSC and P-DSC.
From these results, the increase in number of epidermal cell layers and delayed outgrowth of epidermis were identi ed as suitable epidermal parameters for assessing HTscar formation in vitro using SE.

Cytokine IL-6 and CXCL8 secretion
Most probably, already at the onset of wound healing, scar formation is initiated. Cytokines such as IL-6 and CXCL8 are reported to play a role in in ammation and granulation tissue formation during the wound healing process 18 . Therefore the secretion of IL-6 and CXCL8 was assessed in culture supernatants derived from SE for their use as potential future novel scar parameters (Figure 4).
The secretion of IL-6 was slightly lower (trend) when ASC were incorporated into SE than when P-DSC were used. The secretion of CXCL8 by the SE was signi cantly lower when ASC were incorporated into SE than when P-DSC were used.
From these results, decreased IL-6 and CXCL8 secretion were identi ed as a characteristic of SE ASC .   Clearly SE constructed from ASC populated matrixes not only visually represent HTscars, but also enabled quanti able parameters to be identi ed, which are representative for HTscars. These were increases in 1) thickness of dermis; 2) contraction; 3) collagen 1 secretion; 4) number of epidermal cell layers and decreases in 5) degree of epithelialization. In addition SE ASC showed reduced IL-6 secretion and reduced CXCL8 secretion. The HTscar model was next validated by culturing with two standard therapeutics (5-uorouracil (5FU) and triamcinolon (TC)) which result in partial scar correction in patients, (positive controls) and a therapeutic that is known to be not e ective in scar reduction (1,25-dihydroxy vitamin D 3 (VitD 3 ), which functioned as negative control therapeutic (Table 1). Additionally potential novel scar therapeutics (Atorvastatin and All-trans-retinoic acid (RA)) were tested (Table 1). For all therapeutics, vehicle controls were tested in parallel. No signi cance was found between control condition (nothing added) and vehicle control conditions for the selected parameters. The results of this validation study are described below and summarized in Table 2, Figure 5 and Figure 6.

5-uorouracil (5FU): standard care (partially e ective therapeutic)
Supplementing SE ASC with 5FU led to reduced contraction (Figure 5a and 6a) and reduced number of epidermal cell layers of SE compared to control (6.3 ± 0.8 versus 7.8 ± 0.9) (Figure 5b and 6b). Notably, SE ASC treated with 5FU had approximately the same number of epidermal cell layers as NTscars (6.9 ± 1.0) and SE R-DSC (6.5 ± 0.6) (Figure 3a and  b). No di erences were found with regards to the other parameters ( Figure 6).  (Figure 6a). Also the number of epidermal cell layers of SE ASC decreased after treating with TC compared to control (7.0 ± 1.2 versus 7.8 ± 0.9) (Figure 5b and 6b). No di erences were found with regards to the other parameters.
All-trans-retinoic acid (RA): potential novel scar therapeutic Supplementing potential novel scar therapeutic RA only partially normalized collagen 1 secretion of the HTscar model (Figure 6a). No di erences were found after supplementing SE ASC with RA with regards to the other parameters. These results indicate that RA was not an e ective anti scar therapeutic in the HTscar model.
Atorvastatin: potential novel scar therapeutic Supplementing SE ASC with atorvastatin reduced the thickness of the dermis (Figure 5b and 6a). The number of epidermal cell layers of SE ASC decreased after treating with atorvastatin compared to control (6.4 ± 1.0 versus 7.8 ± 0.9) (Figure 5b and 6b). Notably, SE ASC treated with atorvastatin had approximately the same number of epidermal cell layers as NTscars (6.9 ± 1.0) and SE R-DSC (6.5 ± 0.6) (Figure 3a and b). The secretion of CXCL8 by SE ASC was increased by adding atorvastatin (Figure 6c). No di erences were found with regards to the other parameters. Atrovastatin was the only therapeutic tested which resulted in partial normalization of three parameters.  In vitro hypertrophic scar model

DISCUSSION
In this study we show that ASC and keratinocytes both isolated from healthy full thickness human skin which is readily obtained as rest material after standard surgical procedures may be used to make an in vitro HTscar model to test anti-scarring therapeutics. The HTscar model had similar characteristics as HTscars and enabled relevant and quanti able HTscar parameters to be identi ed and tested. Our rst results shown Experiments were performed with SE from three di erent donors, each in duplicate. Keratinocytes, P-DSC, R-DSC, and ASC were all from the same donor within a single experiment. Data are presented as the mean ±SEM (n=3). Statistical signi cance between the HTscar SE exposed to therapeutic and its corresponding vehicle was calculated using a paired t-test. For all therapeutics, vehicle controls were tested in parallel. No signi cance was found between control condition and vehicle control conditions for the selected parameters (data not shown). Therefore, all control conditions are grouped together in the white bar. The experiments were performed with three donors each in duplicate. *, p<0.05.
in this study indicate that the in vitro HTscar model may be used to test new anti-scar therapeutics. Testing with combinations of known therapeutics and novel therapeutics is now required to further indicate the true value of the HTscar model with regards to replacement, reduction, and re nement of the use of animal models. The rst part of this study involved developing the HTscar model and selecting relevant and quanti able HTscar parameters. We found that SE constructed with ASC visually represents HTscars. In contrast, incorporation of R-DSC and P-DSC, which are cells isolated from the more super cial layers of the skin, led to SE visually representing NTscar and Nskin respectively. This observation is in line with the clinical observation that HTscars occur more often after the closure of full-thickness wounds 2 . Relevant and quanti able parameters typical for HTscars that were identi ed in the HTscar model were contraction; collagen 1 secretion; outgrowth of epidermis and epidermal thickness. Additionally 2 cytokines typically involved in wound healing were assessed. The decrease in both IL-6 and CXCL8 secretion was characteristic for the HTscar model only and therefore it would now be interesting to determine whether HTscar in vivo also show decreased expression of these cytokines. In literature no consensus was reached whether IL-6 and CXCL8 are up or down regulated during HTscar formation [28][29][30] . The confusion may be due to size, location and age of the studied scars samples. Although we did observe an increase in collagen 1 secretion no increased thickness of the dermis was observed in the presented HTscar model compared to SE composed with R-DSC and P-DSC. However, the thickness of the dermis was greater in DE when only ASC were incorporated into the matrix (without keratinocytes on top) than when R-DSC or P-DSC were used (data not shown). At present the reason for this is unknown, however this discrepancy between SE and DE may be related to cultured keratinocytes being very active in secreting proteins which degrade the collagen matrix as it forms 31,32 .
Our results showed that dermal broblasts exhibited less hypertrophic scar characteristics than ASC even though they have been reported to produce TGF beta 1 and many cytokines involved in wound healing and scar formation 33 . This indicates, in line with others, that dermal broblasts are involved in normal wound healing whereas ASC may be involved in adverse scar formation 11,12 . Our nding that the SE ASC model secreted less cytokine IL-6 and CXCL8 may be of signi cance for the pathophysiology of scar formation and our model now provides an excellent means to investigate this further in parallel with in vivo patient derived data in the future. Of note, previously we have shown that ASC and dermal broblast both display a mesenchymal stem cell phenotype (CD31−, CD34+, CD45−, CD54+, CD90+, CD105+, CD166+) and show similar multi-lineage di erentiation potential 24 . These characteristics were more pronounced for ASC. This suggests that, possibly, potent mesenchymal stem cell capacity may correlate to poor scar quality and requires further investigation. Although our results are in line with the clinical observation that HTscars show increased α-SMA compared to NTscar and Nskin, it was noticed though that α-SMA was strongly expressed directly below the basement membrane in the SE ASC HTscar model. This indicates that cultured keratinocytes may secrete a factor which stimulates di erentiation into α-SMA positive cells. Interestingly DE ASC showed very little α-SMA expression supporting this hypothesis. Since the immunohistochemical staining of α-SMA positive cells is di cult to quantify, this biomarker was not selected as a scar forming parameter.
The second part of this study was to validate the HTscar model with two therapeutics regularly used in the clinic for scar treatment (5FU and TC) 20 and one therapeutic known to be unsuccessful in scar reduction (VitD 3 ) 21 . Supplementing the HTscar model with 5FU resulted in partial normalization of the contraction and the epidermal thickness. Interestingly, the other therapeutic, TC, resulted in partial normalization of two di erent parameters: collagen 1 secretion and epidermal thickness. This nding indicates that combined therapy with 5FU and TC may have a better therapeutic a ect than either single therapy. Indeed it has been shown in a clinical study (60 patients) that the combination of 5FU and TC does give a better response rate than either therapeutic alone 34 . Not all parameters (thickness of dermis, outgrowth of epidermis; IL-6 secretion; and CXCL8 secretion) were favorably in uenced by these two therapeutics. This result is in line with clinical results for 5FU and TC since it is known that neither of these therapies can completely restore scar tissue to a normal skin phenotype in all patients (Table  1) 4,20,35 . Both standard therapeutics normalized only two parameters out of seven, indicating that for a therapeutic to be potentially e ective it should also partially normalize at least two parameters. VitD 3 was used as a negative control therapeutic in our study based on clinical evidence 21 . In line with the negative clinical results, we found an increased number of epidermal cell layers after adding VitD 3 . After adding VitD 3 both IL-6 and CXCL8 were even further reduced. However, we also observed a decrease in contraction in SE which may be due to VitD 3 inhibiting ASC proliferation resulting in fewer cells in the matrix at time of harvesting. Indeed, FACScan ow cytometry analysis of 3 mm punch biopsies isolated from SE showed 48% less CD90+ cells within the dermis of VitD3 exposed SE compared to control vehicle exposed SE (data not shown). Despite the thusfar reported clinical results properly dosed VitD3 may possibly prove to be bene cial to the patient since a decrease in the number of broblasts would result in fewer α-SMA positive cells and less contraction. Therefore further clinical studies are justi ed.
After testing the positive and negative controls, two therapeutics of unknown capacity to reduce HTscar characteristics (RA and atorvastatin) were tested. RA is an active metabolite of vitamin A and was included into this study since it decreases broblast proliferation and collagen production 20 . However our results indicate that RA may only have limited value for scar treatment since it only partially normalized one HTscar parameter (reduction of collagen 1 secretion). Furthermore this favourable e ect may be counteracted by simultaneously decreasing collagen degradation 20 . On the other hand, atorvastatin shows distinct therapeutic potential since it was the only therapeutic to partially normalize three parameters (thickness of the dermis, epidermal cell layers and CXCL8 secretion. These results are in line with literature describing atorvastatin to prevent cardiac hypertrophy in rabbits and brotic adhesions in rats 22,36 . Of note, this was the only therapeutic to reduce the thickness of the dermis, a major parameter for a HTscar model. Since we showed in vitro that both 5FU and TC have partly complementary properties compared to atorvastatin, they may potentially be used as combined therapies with atorvastatin. Our in vitro HTscar model will easily permit such pre-clinical investigations in the future.
The HTscar model constructed with ASC not only assesses HTscar reduction but also HTscar prevention since therapeutics were already applied to the culture medium from day 4 before SE were fully developed. This mimics early treatment after surgery. All selected parameters typical for HTscar were a ected in SE by at least one of the tested therapeutics with the exception of the outgrowth of the epidermis. This indicates that the model may be able to identify combinations of therapeutics which compliment each other in correcting adverse scar formation.
As with all in vitro models, the HTscar model has a number of limitations which should be addressed. The main limitation is that it lacks an immune component since it is well known that in ltrating cells e.g. macrophages, monocytes etc in uence wound healing 16 . Currently the model is being further developed to include these immune cells in co-culture with the HTscar model. Also neuro-endocrine signals 37 and an angiogenic component 38 are not incorporated in this HTscar model yet. Furthermore the current number of scar forming parameters might be further expanded. Extensive screening for more parameters such as increased TGFβ1 39 or CTGF 14 might further improve the model and provide more insight into human HT scar formation. Another limitation is that only therapeutics which can be dissolved in the culture medium have been studied so far.
It has yet to be determined whether similar results will be obtained if therapeutics are added topically to the stratum corneum of the SE. If this is the case, the model will also be suitable for testing water insoluble therapeutics in the form of creams and ointments. Also the model will need further adapting if it is to test pressure and silicone dressings, both widely used in HTscar treatment 40 . The negative control therapeutic VitD 3 gave one false positive result (contraction) and one correctly assessed result (increase in epidermal thickness) in addition to a decreased IL-6 and IL-8 secretion. However, it may be possible that the false positive result is a valid result and that the single clinical study described was performed under sub optimal conditions with regards to VitD 3 concentration. In general though, a single false positive result can be minimized due to the assessment of multiple scar parameters.
In most academic research and during drug discovery studies, many animal experiments are used in the early phases to de ne and re ne research questions and potential future applications. It is possible that these early stages in drug development can be replaced by our human in vitro HTscar model system, limiting animal experiments to the nal in vivo con rmation and risk assessment phases. Generally, these nal phases require maximally one-tenth of the total number of animals used (http://www.buzzle. com/articles/animal-testing-statistics.html).
In summary we developed and validated a HTscar model using ASC and keratinocytes isolated from healthy skin and identi ed relevant and quanti able parameters typical for HTscars. In line with the clinical experience, 5FU and TC only partially restored HTscar to normal skin phenotype. Each therapeutic selectively a ected a di erent combination of parameters. These ndings indicate that the in vitro model may be useful for selecting combinations of therapeutics with complementary properties. This will be a future area for investigation. Although the number of therapeutics tested in this initial study is small, our results indicate that this animal free HTscar model may be used to test novel anti-scar therapeutics and thereby may lead to the reduction of the use of animals in HTscar research.