Adapting the Scar-in-a-Jar to Skin Fibrosis and Screening Traditional and Contemporary Anti-Fibrotic Therapies

Skin fibrosis still constitutes an unmet clinical need. Although pharmacological strategies are at the forefront of scientific and technological research and innovation, their clinical translation is hindered by the poor predictive capacity of the currently available in vitro fibrosis models. Indeed, customarily utilised in vitro scarring models are conducted in a low extracellular matrix milieu, which constitutes an oxymoron for the in-hand pathophysiology. Herein, we coupled macromolecular crowding (enhances and accelerates extracellular matrix deposition) with transforming growth factor β1 (TGFβ1; induces trans-differentiation of fibroblasts to myofibroblasts) in human dermal fibroblast cultures to develop a skin fibrosis in vitro model and to screen a range of anti-fibrotic families (corticosteroids, inhibitors of histone deacetylases, inhibitors of collagen crosslinking, inhibitors of TGFβ1 and pleiotropic inhibitors of fibrotic activation). Data obtained demonstrated that macromolecular crowding combined with TGFβ1 significantly enhanced collagen deposition and myofibroblast transformation. Among the anti-fibrotic compounds assessed, trichostatin A (inhibitors of histone deacetylases); serelaxin and pirfenidone (pleiotropic inhibitors of fibrotic activation); and soluble TGFβ receptor trap (inhibitor of TGFβ signalling) resulted in the highest decrease of collagen type I deposition (even higher than triamcinolone acetonide, the gold standard in clinical practice). This study further advocates the potential of macromolecular crowding in the development of in vitro pathophysiology models.


INTRODUCTION
Skin fibrosis is characterised by the formation of excessive fibrous connective tissue, which leads to alteration of the architecture of the dermis and compromises skin's function and mechanical properties . Skin fibrosis manifests either locally (after skin wounding) or systemically (as a result of autoimmune skin disease), with clinical outcomes ranging from small cosmetic imperfections to functional impairment. Skin fibrosis affects over 100 million patients every year (Sund and Arrow, 2000) and is associated with annual healthcare expenditure in excess of US$ 12 billion in the US alone (Griffin et al., 2020).
Fibrosis and skin wound-related scarring are complex, multistage (inflammatory, proliferative and remodelling) processes, involving numerous cells, molecules and signalling pathways (Zeng et al., 2011). The key feature in fibrosis formation is the transformation of normal fibroblasts to myofibroblasts, which are contraction capable cells and responsible for scar and fibrosis formation in different diseases (Schulz et al., 2018;Pakshir et al., 2020). De novo expression of α smooth muscle actin (αSMA), a marker of late stage myofibroblast transformation (Pakshir et al., 2020), is ultimately associated with fibrosis. Biological (e.g., transforming growth factor β1, TGFβ1) and biophysical (e.g., mechanical stress (Seo et al., 2020)) stimuli trigger fibroblast transition into myofibroblast lineage (Hinz et al., 2012), which is associated with the establishment of several characteristic hallmarks of fibrosis, such as atypical collagen synthesis and deposition, alterations in collagen type I/III ratio and distorted extracellular matrix (ECM) architecture (Jarvinen and Ruoslahti, 2010;Henderson et al., 2020;Pakshir et al., 2020).
Anti-fibrotic therapeutics are the first line of defence in scarwars (Jarvinen and Ruoslahti, 2010;Desallais et al., 2014). Different classes of molecules have been assessed over the years, largely classified as: corticosteroids; inhibitors of histone deacetylases, collagen crosslinking and deposition, TGFβ signalling or pleiotropic fibrotic activation ( Supplementary  Table S1). Unfortunately, the development of anti-fibrotic approaches has been hindered by side effects encountered. For example, TGFβ inhibitors may compromise immunity and induce autoimmune diseases (Henderson et al., 2020). Other potential factors that have further limited the development of anti-fibrotic therapies include the use of time consuming and low throughput and specificity (due to genetic, epigenetic, immune status and physiological differences between humans and animals) in vivo models that fail to recapitulate human disease states and effectively screen potential drugs (Padmanabhan et al., 2019). In vitro models have their share of shortcomings (Supplementary Table S2). For example, the low ECM levels present in many traditional in vitro models (an oxymoron for a fibroplasia model) is liable for cell genetic and epigenetic drift and restrains/inhibits cell-ECM interactions and paracrine signalling cascades, resulting in failure of the models to predict in vitro relevant in vivo toxicity of the under investigation molecules (Chen et al., 2009).
Considering the above, herein, we first modified and adopted the Scar-in-the-Jar model (Chen et al., 2009;Stebler and Raghunath, 2021) for skin fibrosis by using dextran sulphate as MMC agent, primary dermal fibroblasts as tissue-specific cell population and TGFβ1 to induce their myofibroblast transdifferentiation (Supplementary Figure S1). We then assessed the model's anti-fibrotic screening potential (through collagen deposition and cell metabolic activity, DNA concentration and viability) by using different anti-fibrotic compounds (corticosteroids: Triamcinolone acetonide, TAC; inhibitors of histone deacetylases: Trichostatin A, TSA; inhibitors of collagen crosslinking: β-aminopropionitrile, BAPN; inhibitors of TGFβ signalling: soluble TGFβ type II receptor-based 2 traps, recombinant proteins T22d35 and T122bt and an activin IIB receptor inhibitor, ACVR2B; and pleiotropic inhibitors of fibrotic activation: Serelaxin, RLX-2 and Pirfenidone, Pirf).

Materials
All labware were obtained from Sarstedt (Ireland) and Thermo Fisher Scientific (Ireland) and all chemicals and reagents were purchased from Sigma-Aldrich (Ireland), unless stated otherwise. amino acid long native linker (O'Connor-Mccourt et al., 2013) (Supplementary Figure S2). Both TGFβ traps were expressed in a mammalian expression system, purified by chromatography and characterized in detail. TGFβ neutralisation curves were plotted, and the determined IC 50 -values were tabulated (Supplementary Table S3).

SDS-PAGE Analysis
Cell layers were analysed by SDS-PAGE as described elsewhere (Capella-Monsonís et al., 2018). Briefly, culture media were aspirated, cell layers were washed with PBS and digested with 0.1 mg/ml pepsin solution (porcine gastric mucosa, 3,500-4,200 U/mg) in 0.5 M acetic acid. The cell layers were then scraped, neutralised with 1 M NaOH, denatured at 95°C and resolved under non-reducing conditions using in-house resolving and stacking polyacrylamide gels (5 and 3% respectively) on a Mini-Protean 3 (Bio-Rad Laboratories, United Kingdom) system. Purified collagen type I (Symatese, France) was used as standard. Samples were stained using a SilverQuest ™ Silver Staining Kit (Invitrogen, Ireland) according to the manufacturer's instructions. Densitometric analysis was performed on α1(I), α2(I), β11(I), β12(I) or c(I) bands, as appropriate, using ImageJ software (NIH, United States).

DNA Concentration Analysis
Cell proliferation was assessed using the Quant-iT ™ PicoGreen ™ dsDNA Assay Kit for quantifying DNA concentration (Invitrogen, United States) as per the manufacturer's instructions. Briefly, cells were washed with PBS, ultrapure water was added and three cycles of freezing and thawing to promote cell lysis were followed. DNA standards of known concentrations were prepared, both samples and standards were mixed with Tris-HCl-Ethylenediaminetetraacetic acid (EDTA) buffer and PicoGreen ™ reagent and incubated in dark. Fluorescence was measured at 480 nm excitation and 520 nm emission with a Varioskan Flash Spectral scanning multimode reader (Thermo Fisher Scientific, United States).

Metabolic Activity Analysis
Cell metabolic activity was assessed using the alamarBlue ® assay (ThermoFisher Scientific, United States) as per the manufacturer's instructions. Briefly, cells were washed with PBS and incubated with a 10% alamarBlue ® solution in PBS for 3 h at 37°C in a humidified atmosphere of 5% CO 2 . Absorbance was then measured at 550 and 595 nm with a Varioskan Flash Spectral scanning multimode reader (Thermo Fisher Scientific, United States). Cell metabolic activity was expressed as percentage reduction of the alamarBlue ® dye and normalised by the respective quantity of DNA and to the nontreated control.

Viability Analysis
Calcein AM (live cell marker) and ethidium homodimer I (dead cell marker) stainings were used to assess the influence of MMC, TGFβ1 and anti-fibrotic molecule supplementation on cell viability. Briefly, at each time point, cells were carefully washed with PBS and incubated with a solution of calcein AM (4 μM) and ethidium homodimer I (2 μM) in PBS for 30 min at 37°C in a humidified atmosphere of 5% CO 2 . Afterwards, cells Frontiers in Bioengineering and Biotechnology | www.frontiersin.org October 2021 | Volume 9 | Article 756399 were imaged with an inverted fluorescence microscope Olympus IX-81 (Olympus Corporation, Japan), using the FITC filter for calcein AM and the Texas Red filter for ethidium homodimer.

Statistical Analysis
Statistical evaluation of the data was performed using the statistical program MiniTab ® version 17 (Minitab Inc., United States). All data are expressed as mean values ± standard deviations. Datasets were assessed for normal distribution (Anderson-Darling) and equal variance (Levene's test for homogeneity of variances). When the assumptions of parametric analysis were confirmed, one-way analysis of variance (ANOVA) was used for multiple comparisons and Tukey's post hoc test was used for pairwise comparisons. When either or both assumptions were violated, non-parametric analysis was conducted using Kruskall-Wallis test for multiple comparisons and Mann-Whitney U test for pairwise comparisons. Statistical significance was accepted at p < 0.05.

Fibrotic Model Establishment
SDS-PAGE ( Figure 1A) and complementary densitometry analysis of collagen type I α(I)1 and α(I)2 bands ( Figure 1B) made apparent that at day 4 and day 7 almost no collagen was deposited in the control and the TGFβ1 groups, whilst MMC groups significantly (p < 0.05) increased collagen deposition at all time points, which was further increased (p < 0.05) with +MMC+TGFβ1 at day 7. Densitometry analysis also showed a significant increase (p < 0.05) of β11(I), β12(I) dimers for the +MMC+TGFβ1 group at day 4 and day 7 (Supplementary Frontiers in Bioengineering and Biotechnology | www.frontiersin.org October 2021 | Volume 9 | Article 756399 Figure S3A) and of γ(I) trimers for the +MMC+TGFβ1 group at day 4 and day 10 (Supplementary Figure S3B). Immunocytochemistry (Figure 2A for αSMA and Figure 3A for collagen type I) analysis made apparent that when the cells were cultured with +MMC+TGFβ1, clear stress fibres were observed and the collagen fibres were aligned parallel to the stress fibres, albeit collagen type I deposition showed a granular pattern when MMC was used, whilst in its absence, a meshwork architecture was evidenced. Complementary image intensity ( Figure 2B for αSMA and Figure 3B for collagen type I) analyses revealed that, in comparison to the control, the addition of TGFβ1 resulted in a significant (p < 0.05) increase in αSMA expression at day 7 and day 10 and collagen type I deposition at day 4. MMC resulted in a significant (p < 0.05) increase compared to the control for both molecules in almost all time points. This was also observed for +MMC+TGFβ1, which led to an even greater (p < 0.05) increase in αSMA at day 7 and collagen I at day 4.  Figure S4B) were observed. At day 7 the +MMC+TGFβ1+TAC groups exhibited significantly (p < 0.05) higher and at day 10 significantly (p < 0.05) lower DNA concentration than the +MMC+TGFβ1 group (Supplementary Figure S5A). At day 4 and 7 almost all the +MMC+TGFβ1+TAC groups exhibited significantly (p < 0.05) lower and at day 10 significantly (p < 0.05) higher metabolic activity than the +MMC+TGFβ1 group (Supplementary Figure S5B). SDS-PAGE ( Figure 5A) and densitometry of α(I)1 and α(I)2 bands ( Figure 5B) analyses revealed that all TSA concentrations in +MMC+TGFβ1 at day 7 and the 1, 2.5 and 5 µM TSA  Figure S6B). The 1, 2.5 and 5 µM TSA concentrations in +MMC+TGFβ1 resulted in significant (p < 0.05) reduction of DNA concentration at day 4 and the 2.5 and 5 µM TSA concentrations in +MMC+TGFβ1 resulted in significant (p < 0.05) reduction of DNA concentration at day 7 and day 10, all in comparison to the +MMC+TGFβ1 group (Supplementary Figure S7A). The 0.5 and 2.5 µM TSA concentrations in +MMC+TGFβ1 resulted in significant (p < 0.05) reduction of metabolic activity at day 7 in comparison to the +MMC+TGFβ1 group (Supplementary Figure S7B). SDS-PAGE ( Figure 6A) and densitometry analysis of α(I)1 and α(I)2 ( Figure 6B), β11(I), β12(I) (Supplementary Figure  S8A) and γ(I) bands (Supplementary Figure S8B) revealed that FIGURE 4 | TAC moderately reduces collagen deposition. All TAC concentrations resulted in reduced collagen deposition at day 7 and day 10, in comparison to +MMC+TGFβ1 group, as judged by SDS-PAGE (A) and densitometry of α(I)1 and α(I)2 bands analyses (B). Col I STD: 0.1 mg/ml. One-way ANOVA and Tukey's posthoc comparison tests were conducted. *: p < 0.05 denotes a significant difference when compared to the +MMC+TGFβ1 group of the respective time point. n 3.
Frontiers in Bioengineering and Biotechnology | www.frontiersin.org October 2021 | Volume 9 | Article 756399 7 all RLX-2 concentrations in +MMC+TGFβ1 at all time points resulted in a significant (p < 0.05) decrease of the deposition of α(I)1 and α(I)2, β11(I), β12(I) and γ(I) components, when compared to the +MMC+TGFβ1 group at the respective time point. At day 7, all concentrations of RLX-2 in +MMC+TGFβ1 resulted in significant (p < 0.05) increase of DNA concentration in comparison to the +MMC+TGFβ1 group (Supplementary Figure S9A). At day 7, all concentrations of RLX-2 in +MMC+TGFβ1 resulted in significant (p < 0.05) decrease of metabolic activity in comparison to the +MMC+TGFβ1 group (Supplementary Figure S9B).
SDS-PAGE ( Figure 7A) and densitometry analysis of α(I)1 and α(I)2 bands ( Figure 7B) revealed that all Pirf concentrations at day 4, the 1 and 1.5 mM Pirf concentrations at day 7 and all Pirf concentrations at day 10, all in +MMC+TGFβ1, resulted in significant (p < 0.05) FIGURE 5 | High concentrations of TSA reduce collagen deposition. The 2.5 and 5 µM TSA concentrations in +MMC+TGFβ1 at day 7 and day 10 induced the lowest collagen deposition in comparison to +MMC+TGFβ1 group, as revealed by SDS-PAGE (A) and densitometry of α(I)1 and α(I)2 bands (B). Col I STD: 0.1 mg/ml. One-way ANOVA and Tukey's post-hoc comparison tests were conducted. *: p < 0.05 indicates a statistically significant difference when compared to the +MMC+TGFβ1 group of the respective time point. +: p < 0.05 indicates a statistically significant difference when compared to the +MMC+TGFβ1+TSA 0.5 μM group of the respective time point. #: p < 0.05 indicates a statistically significant difference when compared to the +MMC+TGFβ1+TSA 1 μM group of the respective time point. n 3.

SDS-PAGE (Supplementary
SDS-PAGE ( Figure 9A) and densitometry analysis of α(I)1 and α(I)2 ( Figure 9B), β11(I), β12(I) (Supplementary Figure  S17A) and γ(I) (Supplementary Figure S17B) bands revealed  Figure  S18A). The 25, 50 and 100 nM ACVR2B concentrations in +MMC+TGFβ1 significantly (p < 0.05) decreased metabolic activity, in comparison to +MMC+TGFβ1 group, at day 7 and the 250 nM concentration at day 10 (Supplementary Figure  S18B). Figure S19A) and densitometry of α(I)1 and α(I)2 bands (Supplementary Figure S19B) revealed that all concentrations of BAPN in +MMC+TGFβ1 at day 4 and the 0.1, 0.25 and the 0.5 mM concentrations of BAPN in +MMC+TGFβ1 at day 10 significantly (p < 0.05) increased α(I)1 and α(I)2 deposition in comparison to the FIGURE 9 | ACVR2B reduces collagen deposition. All ACVR2B concentrations in +MMC+TGFβ1 at all time points significantly reduced collagen deposition, in comparison to +MMC+TGFβ1 group, as judged by SDS-PAGE (A) and densitometry analysis of α(I)1 and α(I)2 bands (B). Col I STD: 0.1 mg/ml. One-way ANOVA and Tukey's post-hoc comparison tests or Kruskal Wallis and Mann Whitney post-hoc analyses were conducted, as appropriate. *: p < 0.05 indicates a statistically significant difference when compared to the +MMC+TGFβ1 group of the respective time point. n 3.

SDS-PAGE (Supplementary
Frontiers in Bioengineering and Biotechnology | www.frontiersin.org October 2021 | Volume 9 | Article 756399 +MMC+TGFβ1 group at the respective time points. No significant (p < 0.05) differences in the deposition of β11(I), β12(I) (Supplementary Figure S20A) and γ(I) components (Supplementary Figure S20B) were observed. The 1 mM concentration of the BAPN in +MMC+TGFβ1 significantly reduced DNA concentration at day 7 and day 10 in comparison to the +MMC+TGFβ1 group (Supplementary Figure S21A). The 0.5 mM and the 1 mM BAPN concentration in +MMC+TGFβ1 resulted in significantly increased metabolic activity at day 10 in comparison to the +MMC+TGFβ1 group (Supplementary Figure S21B). Qualitative cell viability assessment revealed that TAC did not have a negative affect; TSA, BAPN and T122bt had a negative effect at high concentrations (the latter, only at the latest time point); and RLX-2, Pirf, T22d35 and ACV2R had a negative effect mostly at later time points, regardless of concentration ( Supplementary Figures S22-S29).

DISCUSSION
Fibrosis is not just the outcome of devastating skin diseases, but it can also be the result of abnormal skin wound healing and scarring. One of the fundamental roadblocks in the use of traditional in vitro models for screening anti-fibrotic molecules is that they do not recapitulate the excessive and altered ECM characteristic of human fibrotic diseases (Chen et al., 2009). To address this deficiency, we ventured to assess whether MMC (that dramatically enhances and accelerates ECM deposition) coupled with TGFβ1 (that induces myofibroblast trans-differentiation) can generate an efficient skin fibrosis model.
Starting with the establishment of the fibrotic model, TGFβ1 is a key player in many fibrotic conditions (Lodyga and Hinz, 2020), with its signalling pathway involved in transformation of fibroblasts into myofibroblasts (Sarrazy et al., 2011;Lichtman et al., 2016). MMC significantly increased collagen deposition and aSMA expression in TGFβ1 supplemented cultures and resulted in the formation of an ECM-rich substrate, magnifying the profibrotic effect of TGFβ1, a prerequisite for the development of a scarring model.
Although a quantitative analysis of cell and ECM orientation was not performed in this study, it was still possible to observe that treatment with macromolecular crowding and TGFβ1 resulted in an alteration of the deposited ECM. Deposition of densely packed granular collagen type I (resembling the formation of scar tissue (Pakshir and Hinz, 2018)) was observable. On the other hand, in the absence of macromolecular crowding treatment, it was observed the normal meshwork architecture of healthy skin ECM (Goto et al., 2020). The formation of αSMA stress fibres alongside the deposition axis of collagen type I also seems to confirm this.
In accordance with previous publications, several other organspecific fibrotic models induced by MMC supplementation have been developed (Chen et al., 2009;Graupp et al., 2015;Graupp et al., 2018;Fan et al., 2019;Good et al., 2019;Fan et al., 2020;Rønnow et al., 2020;De Pieri et al., 2021). It is worth noting that cocktails of pro-inflammatory cytokines have been used for more accurate recapitulation of fibrosis in vitro (Chawla and Ghosh, 2018) and such approach should be studied further in the future in combination with MMC.
After establishment of the in vitro model of skin fibrosis, several anti-fibrotic molecules with different mechanisms of action were tested to assess their capacity to decrease collagen synthesis and/or deposition. We also assessed their effect on basic cellular functions (cell metabolic activity, viability, DNA concentration) to assess how selective their mode of action is. Tested drug concentrations were based on previously published data that had been shown to have a therapeutic effect in vitro. This explains the different orders of magnitude in concentrations used for different anti-fibrotic molecules, which allowed us to validate the model and select the best, as judged by maximum collagen reduction and adequate profile regarding basic cellular functions, anti-fibrotic molecules.
The first molecule tested was TAC, and in this study, only a very moderate effect in the reduction of the deposition of collagen type I was observed. This synthetic corticosteroid, used to reduce inflammation, is currently one of the most used molecules for the treatment of hypertrophic scars and keloids in humans and arguably the "gold standard" treatment (Ogawa, 2010;Berman et al., 2017;Hietanen et al., 2019). Despite its wide usage, TAC presents only moderate response rates, while presenting relatively high recurrence rates (Ogawa, 2010;Hietanen et al., 2019) and considerable side effects on skin (Sadeghinia and Sadeghinia, 2012). We have recently demonstrated in a randomised controlled trial that keloids that respond to either TAC or 5-FU show a reduction in myofibroblasts (Hietanen et al., 2020). This is a secondary effect as neither TAC nor 5-FU have a direct influence on myofibroblast transformation. Our results support clinical data and illustrate that more effective and safer therapies than TAC should be developed for keloids and hypertrophic scars.
Another molecule tested was TSA and we observed an overall reduction in collagen type I deposition and cell proliferation. This is not surprising considering that it is an inhibitor of class I and II mammalian histone deacetylases (HDACs), which alter gene expression by altering the access of transcription factors to DNA. TSA has been shown to reduce αSMA expression and collagen type I deposition; inhibit cellular proliferation in fibroblasts; and promote apoptosis in certain experimental skin fibrosis models (Rombouts et al., 2002;Huber et al., 2007;Diao et al., 2011). HDACs, in general, have been implicated in TGFβ1induced epithelial-mesenchymal transition (EMT) (Jones et al., 2019;Qiao et al., 2020). HDAC suppress epithelial-specific genes and mediate TGFβ1-induced mesenchymal enhancer reprogramming, which results in EMT of the cells (Qiao et al., 2020). As myofibroblast transformation is a central feature in fibrosis and our results demonstrate that TSA can inhibit it, this makes it a viable therapeutic candidate.
After treatment with RLX-2, a decrease in collagen type I deposition was observed, with no significant negative effects on basic cellular functions. To substantiate this, one should consider that recombinant human relaxin-2 (RLX-2), a naturally occurring peptide, binds to its receptor, RLX family peptide receptor 1. This has been shown to suppress not only Smad 2/3 and consequently, TGFβ1 signalling pathways in cardiac fibroblasts, but also angiotensin II type 2 receptors and interleukin-1β and the potential interaction of these signalling pathways with the TGFβ axis (Yuan et al., 2017;Wu et al., 2018), as well as with the inflammasome (Pinar et al., 2020). In addition to those, RLX-2 can increase the expression and activity of matrix metalloproteinases that facilitate ECM degradation (Li et al., 2021). Its application in other diseases, mainly fibrotic, has been postulated, although the clinical trials have been inconclusive on its therapeutic efficacy (Samuel et al., 2017;Yuan et al., 2017;Blessing et al., 2019;Hinz and Lagares, 2020).
When testing Pirf, our study showed decreased collagen deposition and an absence of significant adverse effects on basic cellular functions at the lowest drug concentrations tested. However, when compared to the other molecules tested, very high relative concentrations of the molecule were required to obtain anti-fibrotic activity. This follows the reported literature, as Pirf is a small molecule inhibitor and a Food and Drug Administration approved drug for idiopathic lung fibrosis, but large doses are required (Noble et al., 2011;Lancaster et al., 2017). It has been investigated as a treatment for several other fibrotic diseases, such as liver (Seniutkin et al., 2018), kidney (Salah et al., 2019), intestinal (Sun et al., 2018) and skin (Hall et al., 2018). It functions by inhibiting TGFβ1 (and its production) and platelet-derived growth factor-activated signalling pathways (Lv et al., 2020) in conjunction with antiinflammatory activity (Mora et al., 2015;Hall et al., 2018). Due to the high doses required, Pirf is associated with side-effects and organ toxicity (Anderson et al., 2016;Lancaster et al., 2017), which require rigorous supervision and limit its therapeutic potential. One option could be to combine a low dose of pirfenidone with another anti-fibrotic molecule. By that approach, side-effects could hopefully be avoided, and the therapeutic effect enhanced.
Following treatment with both TGFβ traps, no significant effects on collagen deposition were observed for T22d35, while T122bt resulted in a significant decrease of collagen deposition, at a concentration several orders of magnitude lower than for some of the other molecules. Neither of the TGFβ trap molecules elicited a negative response regarding basic cellular functions. This is expected, as given the importance of the TGFβ signalling pathway in the transition to a fibrotic phenotype, recombinant TGFβ traps, i.e., soluble ligand binding parts of the TGFβ receptors can act as inhibitors of this pathway. They can bind different isoforms of TGFβ, namely -β1 and -β3 for T22d35 and all three isoforms for T122bt. Both molecules can inhibit different isoforms of TGFβ at almost picomolar concentration, which is to our knowledge, the lowest concentration that has been attained against TGFβ (Zwaagstra et al., 2012;O'Connor-Mccourt et al., 2013). Concerning the difference between the two traps and TGFβ isoforms, TGFβ2 augments the profibrotic functions of TGFβ1 (Jagadeesan and Bayat, 2007), whereas TGFβ3 has been hypothesised to possess anti-fibrotic functions (Occleston et al., 2008;Lichtman et al., 2016). Thus, the T122b trap has a preferential TGFβ inhibitory profile than T22d35. Furthermore, both tested molecules have demonstrated substantial anti-tumour effect in TGFβ-driven cancer models in vivo (Zwaagstra et al., 2012;O'Connor-Mccourt et al., 2013). As TGFβ1 was the only growth factor supplied in the media, our results indicate that these traps are very potent and specific in their inhibitory activity.
We also tested an activin IIB receptor antagonist, a soluble ligand binding domain part of the receptor. Despite a substantial decrease in collagen type I deposition, a decrease in basic cellular functions was observed for ACVR2B. This can be explained as this antagonist blocks signalling of activins A and B, myostatin (MSTN/GDF-8), its close homolog, growth differentiation factor-11 (GDF-11) and bone morphogenetic protein-10 (Lautaoja et al., 2019;Magga et al., 2019;Szabó et al., 2020). Both activins are involved in Smad 2/3 signalling and consequently play a role in fibrosis (Werner and Alzheimer, 2006;Canady et al., 2013;Walton et al., 2017;Itoh et al., 2018). The expression of activins is induced during wound repair and activin A leads to accelerated wound healing (Cangkrama et al., 2020). However, its most striking effect is on the granulation tissue formation as it induces excessive scar formation (Cangkrama et al., 2020;Wietecha et al., 2020). This is attributed to the expression of its target genes in fibroblasts, which include both ACTA2 and COL1A1 (αSMA and collagen type I) (Cangkrama et al., 2020). Both MMC and TGFβ1, in turn, induce the expression of activins and myostatin in fibrotic disorders (Erämaa and Ritvos, 1996;Cianciolo et al., 2020;Winter et al., 2020). As activins are produced in mature, active form, they can induce cell signalling immediately (Cangkrama et al., 2020). Furthermore, whereas TGFβ1 signalling becomes refractory for extended periods of time due to receptor internalisation and degradation, the activin signalling remains active all the time due to constant receptor renewal at cell surface (Miller et al., 2019). Thus, our model could involve activin activity although activins were not supplemented to the culture media.
Treatment with BAPN resulted in decreased cell proliferation, although no significant decrease in collagen type I deposition was observed, which is in accordance with what was reported in previous studies. Given its action as a lysyl oxidase (LOX) inhibitor, which inhibits collagen crosslinking (Redden and Doolin, 2003), its use has been previously suggested as an anti-fibrotic (Trackman, 2016). However, several safety concerns related to osteolathyrism, a collagen cross-linking deficiency (Wilmarth and Froines, 1992;Rosenthal, 2003) have been reported following its use. The target enzyme, LOX, has a wide variety of biological effects beyond collagen cross-linking, which also influences matrix stiffness and cell proliferation and migration (Saatci et al., 2020;Freeberg et al., 2021;Kozma et al., 2021;Sflomos et al., 2021). This is an indicator of its unsuitability as an anti-fibrotic molecule and potentially the reason for the molecule's failures in clinical trials.

CONCLUSION
The low extracellular matrix content in the traditional in vitro fibrosis models results in poor imitation of the tissue pathology and to scattered predictive capacity. This study advocates the use of macromolecular crowding (to enhance and accelerate extracellular matrix deposition) and TGFβ 1 (to induce dermal fibroblast trans-differentiation to myofibroblast) in the development of skin fibrosis specific in vitro models. We further identified trichostatin A, serelaxin, pirfenidone and soluble TGFβ trap as potent anti-fibrotic therapies.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

FUNDING
This work has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme, grant agreement No. 866126. This publication has emanated from research supported in part by grants from Science Foundation Ireland (SFI) under grant numbers 15/CDA/3629 and 19/FFP/6982 and Science Foundation Ireland (SFI) and European Regional Development Fund (ERDF) under grant number 13/RC/ 2073_2. This research was also funded by the Finnish National Agency for Education (EDUFI), Academy of Finland, Päivikki and Sakari Sohlberg Foundation, Tampere Tuberculosis Foundation, Tampere University Hospital Research Fund and Pirkanmaa Hospital District Research Foundation. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.