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 (Coentro et al., 2018). 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, multi-stage (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 scar-wars (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).

Macromolecular crowding (MMC), a biophysical technique based on volume exclusion effect, accelerates the enzymatic conversion of water-soluble procollagen to insoluble collagen resulting in enhanced and accelerated collagen type I and associated ECM deposition (Raghunath and Zeugolis, 2021; Tsiapalis and Zeugolis, 2021; Zeugolis, 2021). In 2009, the first pathophysiologically relevant in vitro fibrosis model (termed Scar-in-the-Jar) was published that utilised the principles of MMC (to enhance and accelerate ECM deposition) and TGFβ1 (to induce myofibroblast transformation of WI-38 lung fibroblasts) (Chen et al., 2009). Since then, several fibrotic models based on MMC have been developed for screening anti-fibrotics in different fibrotic diseases (e.g., dermal (Fan et al., 2019; Fan et al., 2020), lung (Good et al., 2019; Rønnow et al., 2020), vocal fold (Graupp et al., 2015; Graupp et al., 2018) scarring). Unfortunately, these dermal scar models might be incomplete as the optimal crowding molecule was not used (Chen et al., 2009). Although MMC agents, such as Ficoll^® (Fan et al., 2019) and polyvinylpyrrolidone (Rashid et al., 2014), have been used as crowding agents, dextran sulphate has demonstrated pro-fibrotic potency by transforming corneal fibroblasts to myofibroblasts (Kumar et al., 2015), possibly due to its binding and releasing capacity of growth factors, such as TGFβ1 (Walton, 1952; Logeart-Avramoglou et al., 2002; Maire et al., 2005).

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 trans-differentiation (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).

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 organ-specific 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 TGFβ1-induced 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 anti-inflammatory 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.

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.