Prolonged Scar-in-a-Jar: an in vitro screening tool for anti-fibrotic therapies using biomarkers of extracellular matrix synthesis

Most drug candidates for pulmonary fibrosis fail in human clinical trials [1, 2]. To reduce the attrition rates in the clinic it is essential that novel anti-fibrotic compounds are screened in reliable and disease relevant pre-clinical models of fibroproliferative diseases. It is important that these preclinical models replicate key events in human pulmonary fibrosis such as dysregulated fibroblast activity and aberrant remodeling of the extracellular matrix (ECM) [3].

Pulmonary fibrosis includes several lung disorders characterized by the formation of excessive scar tissue in the lungs. Idiopathic pulmonary fibrosis (IPF) is a particularly severe and progressive form [4], with a mean survival of 3–5 years after the time of diagnosis [5]. The incidence of IPF in Europe and North America has risen in recent years and is estimated to range between 2.8 and 18 cases per 100.000 people per year [6, 7]. During the development of IPF, healthy tissue is replaced by rigid ECM, destroying the lung architecture and leading to disrupted gas exchange and ultimately respiratory failure and death [3]. Transforming growth factor (TGF)-β1 plays a critical role in the differentiation of fibroblasts into myofibroblasts, which in turn produce ECM proteins driving the abnormal repair response and scar formation in IPF [8, 9]. During the progression of fibrosis, ECM alignment and composition is altered [10]. Imbalanced ECM remodelling leads to increased release of tissue- and pathology-specific protein fragments into the circulation [11, 12]. Such protease-generated fragments represent neo-epitopes which can be recognized by specific antibodies employed in enzyme-linked immunosorbent assays (ELISAs) and utilised as biomarkers. Some of these biomarkers have previously been shown to correlate with the progression of IPF [13]. Currently, no circulating biomarkers are routinely used for IPF in the clinic, neither for diagnosis, prognosis, prediction or monitoring. Some of the most commonly studied biomarkers include SP-D and KL-6, reflecting epithelial injury; MMP-7, periostin and ECM neo-epitopes such as C1M, C3M, C6M and CRPM reflecting ECM remodelling [14‐16].

The first effective disease-modifying drugs to be approved by the U.S. Food and Drug Administration (FDA) and European Medicines Agency (EMA) were pirfenidone and nintedanib, which have succeeded in attenuating lung function decline in patients with IPF [17, 18]. There is still no cure for IPF, thus new therapeutic options are being explored [19, 20]. One group of therapies that is being tested in clinical trials is inhibitors of the mammalian target of rapamycin (mTOR). These were initially introduced into clinical practice to prevent transplant rejection and later to treat mTOR diseases such as lymphangioleiomyomatosis [21, 22]. The anti-fibrotic effect of the mTOR inhibitors is mediated by a decrease of type I and III collagen synthesis [23]. Omipalisib (GSK2126458), a potent inhibitor of mTOR and phosphatidylinositol 3-kinase (PI3K) [24], has been tested in a randomised, placebo-controlled study of IPF, which found that omipalisib has an acceptable tolerability and that target engagement was confirmed (NCT01725139) [25]. Additionally, recent data from a proof of mechanism study indicate that omipalisib reduces collagen synthesis as shown by a rapid decrease of the biomarkers PRO-C3 and PRO-C6 in serum of IPF patients [26]. By using human lung fibroblasts, Woodcock et al. showed that TGF-β1 signalling through mTOR is critical for fibrogenesis, by demonstrating that mTOR signalling, was required for type I collagen deposition in vitro [27]. Additional results from precision-cut IPF lung slices confirmed that synthesis of type I collagen, as measured by PRO-C1 ELISA in the supernatants, was mediated through mTOR [27].

Complex in vitro models of fibrogenesis using primary human lung fibroblasts offer a physiologically relevant alternative to in vivo models which may not fully recapitulate the complex pathogenesis exhibited by IPF patients [28, 29]. Cells grown in a crowded, pseudo-3D environment more closely resemble the physiological environment and the pathological ECM compared to standard 2D cell cultures [30]. The Scar-in-a-Jar model developed by Chen et al. [31] uses healthy human lung fibroblasts grown in medium with macromolecules (ficoll or dextran sulphate) to provide crowding conditions. Ficoll contains neutral macromolecules of 70 and 400 kDa that occupy the space in the well, making it unavailable for other molecules and thereby causing the excluded volume effect [32]. These crowded conditions cause increased interactions between the substrate and their respective enzymes, increasing enzymatic activity and accelerating collagen deposition [33]. The use of ficoll allows us to mimic the dense cellular environment within the ECM and improve the release of collagen pro-peptides as compared to systems without crowded conditions. Furthermore, Chen et al. demonstrated that the presence of ficoll leads to a higher degree of cross-linking in the deposited collagen [31], thus better representing in vivo conditions. Ascorbic acid is another crucial component in this model, as it is essential for the post-translational modification of proline to hydroxyproline that allows the correct folding of collagens during synthesis. The addition of human TGF-β1 induces fibroblast activation and differentiation to myofibroblasts, increasing ECM production and thereby imitating fibrogenesis [31]. Chen et al. showed that within 6 days the collagen deposition doubled in the Scar-in-a-Jar assay as compared with non-crowded cultures and that the addition of TGF-β1 increased the deposition 2-fold as compared with a crowded culture without stimulation. Furthermore, they showed a more aligned pattern of the fibrillar type I collagen and an effect of potential anti-fibrotic compounds on type I collagen deposition. Interestingly, the addition of a crowding agent induced complete cleavage of the type I collagen C-propeptide, whereas the fibroblasts in the non-crowded culture conditions continued to secrete intact procollagen. Recently, Good et al. described a 72 h high content screening Scar-in-a-Jar assay utilizing fibroblasts from IPF patients, that allowed the quantification of disease-relevant ECM deposition in vitro [34].

Here, a prolonged Scar-in-a-Jar model was used to allow for correct collagen processing, and biochemical markers quantified ECM synthesis for up to 12 days. These markers include PRO-C3 and PRO-C6, biomarkers of type III and VI collagen synthesis, respectively, which have been found to be increased in blood from progressing IPF patients as compared to stable IPF patients in the PROFILE (Prospective Observation of Fibrosis in the Lung Clinical Endpoints) study [35]. In addition, serum levels of a marker of myofibroblasts, alpha-smooth muscle actin (α-SMA), has also been shown to be increased in IPF patients compared to healthy controls [36]. The effect of TGF-β1 stimulation on ECM synthesis was investigated after 4, 8, and 12 days to determine the kinetics of the different ECM proteins produced. Further, the use of the prolonged Scar-in-a-Jar model as a drug screening tool for novel anti-fibrotic compounds was evaluated by testing nintedanib, pirfenidone and the mTOR/PI3K inhibitor omipalisib. The model is depicted in Fig. 1.

Many anti-fibrotic drug candidates fail when entering clinical trials, despite promising results obtained in a pre-clinical setting. This suggests that improved pre-clinical assays are needed to successfully translate the preclinical results to the clinical setting. Here, we propose a novel tool for drug screening that has the potential to aid the anti-fibrotic drug development process. We developed a prolonged Scar-in-a-Jar in vitro model and combined the assay with the measurement of clinically validated biomarkers of ECM synthesis in the cell supernatant. The model was validated by testing first a TGF-β1 receptor kinase inhibitor, and subsequently the two FDA approved anti-fibrotic compounds nintedanib and pirfenidone together with a mTOR/PI3K inhibitor in development, omipalisib. We found that stimulating primary lung fibroblasts with TGF-β1 increased the ECM synthesis and α-SMA expression over time, and that this was modulated by anti-fibrotic treatments in a concentration dependent manner.

Currently, many in vitro models only allow investigations of fibrogenesis in acute studies of 1–3 days [28, 34] although some models extend to a few weeks [44]. To our knowledge, this is the first time that a 12 days fibroblast culture has been combined with the investigation of ECM synthesis in the supernatant. The prolonged culture allows investigation of a drug in a system which is not limited to the acute response. In this study we investigated the ECM production after 4, 8 and 12 days of fibroblast culture, and observed that synthesis of different proteins occur at different times. Interestingly, type III collagen and α-SMA synthesis was not evident at day 4 but increased with TGF-β1 stimulation at day 8 and even more so at day 12. Levels of type I collagen and fibronectin peaked at day 8 and stayed elevated at day 12, while remaining biomarkers continued to increase over the whole course of the experiment. This could indicate that there is a continuous activation of fibroblasts and additive contribution to the ECM synthesis throughout the 12-day period. In the prolonged Scar-in-a-Jar model, TGF-β1 significantly induced accumulation of type I (PRO-C1), III (PRO-C3) and VI (PRO-C6) collagen, fibronectin (FBN-C) and α-SMA in comparison to unstimulated cells but did not stimulate type IV (PRO-C4) or V (PRO-C5) collagen production. Type I and III collagen are the main ECM proteins associated with fibrogenesis, and TGF-β1 has previously been identified as a stimulator of type I and III collagen production by lung fibroblasts in vitro [45]. In line with our data, Walker et al. observed a small increase in fibronectin expression in airway fibroblasts in response to TGF-β stimulation after one day of culture while the increase was more pronounced after 20 days [44]. This underlines the importance of the temporal aspect when investigating fibrotic processes. Normally, fibroblasts do not produce type IV collagen, but Roach et al. showed that myofibroblasts had increased type IV collagen mRNA expression upon stimulation with TGF-β1 [46]. Tuan et al. demonstrated that fibroblasts in a fibroplasia model were able to produce type V collagen, albeit to a much smaller extent than type I collagen [47]. Additionally, Knüppel et al. observed that TGF-β stimulation of primary fibroblasts resulted in increased type V collagen mRNA expression and protein concentration [48]. This indicates that the conditions in the prolonged Scar-in-a-Jar model are not favorable for the production of type IV and V collagen, or that amounts are too small to observe. Using a different fibrogenic stimuli, e.g. platelet-derived growth factor (PDGF), may induce fibrogenesis with a different ECM protein profile. Furthermore, the degree of collagen cross-linking in the prolonged Scar-in-a-Jar model was significantly elevated by TGF-β1 stimulation and could be abolished by a pan-LOX inhibitor. Hence, providing a model which could be used to investigate the effects of current anti-fibrotic compounds on collagen cross-linking, an effect which could promote fibrolysis and be beneficial for patients with pulmonary fibrosis.

ALK5i, an inhibitor of the type I TGF-β receptor kinase, was able to significantly decrease the TGF-β1-induced levels of PRO-C1, PRO-C3, PRO-C6, FBN-C, and α-SMA. Thus, the pro-fibrotic effects that TGF-β1 exerts on lung fibroblasts can be modulated in this system, and the combination of the prolonged Scar-in-a-Jar model and biochemical markers of ECM synthesis may be a useful tool for evaluating anti-fibrotic compounds. In line with this, Epstein et al. showed that an ALK5i was able to reduce α-SMA and type I collagen mRNA levels after 24 h in normal fibroblasts cultivated in an IPF conditioned matrix [49].

Nintedanib and omipalisib were able to modulate fibrogenesis in the prolonged Scar-in-a-Jar model as evident by a significant decrease of PRO-C1, PRO-C3, PRO-C6, FBN-C and α-SMA levels in supernatant. In line with the current results, Gao et al. showed that an mTOR inhibitor was able to significantly reduce TGF-β1 induced type III collagen and fibronectin expression within 24 h in primary human lung fibroblasts [23]. Studies using precision-cut IPF lung slices have shown that omipalisib is also capable of modulating the accumulation of PRO-C1 from fibrotic lung tissue in an ex vivo setting [24, 27]. Additionally, a proof of mechanism study showed that treatment of IPF patients with omipalisib for only 7–10 days significantly reduced serum levels of PRO-C3 and PRO-C6, and that these levels correlated with levels of omipalisib in plasma [26]. Nintedanib has been shown to decrease TGF-β1 induced collagen secretion from IPF fibroblasts and non-fibrotic control fibroblasts within 48 h [50]. Nintedanib was also able to prevent an increase in type I collagen mRNA when added to normal fibroblasts cultured for 24 h in an IPF conditioned matrix [49]. These data indicate that mTOR inhibitors and nintedanib have direct effects on the fibroblasts producing collagens and other ECM proteins. Our results indicated that both nintedanib and omipalisib may have cytotoxic effects on the fibroblasts at high concentrations. There was a clear effect on Alamar Blue and partially on LDH release in response to these compounds, indicating a reduction in cell metabolic activity, and a concurrent increase in cytotoxicity, both of which could be beneficial at the correct dose in a fibrotic setting.

In spite of a pronounced effect of the solvent at high concentrations, a reduction of PRO-C1, PRO-C3, PRO-C6 and FBN-C, but not α-SMA, was observed in response to pirfenidone treatment after 8 and 12 days of culture. Interestingly, after 4 days of culture, only PRO-C6 levels were significantly reduced by pirfenidone as compared with vehicle. Pirfenidone at concentrations of 300 and 1000 μM had small but significant cytotoxic effects, as determined by LDH release, as compared to the solvent alone, while metabolic activity assessed by Alamar Blue was not altered. The biomarker data showed no effect on α-SMA assessed in cell supernatant which could indicate that pirfenidone does not alter fibroblast activation but may have a beneficial effect on ECM synthesis. The anti-fibrotic and anti-inflammatory effects of pirfenidone have not been fully established [51]. Some studies have shown that pirfenidone had a direct effect on fibroblasts, whereas others indicated that the effect was very weak [49, 52‐55]. However, these studies were performed for a shorter period of 1–2 days which may influence the effect on protein synthesis, as indicated by our results. In a study of IPF fibroblasts, Knüppel et al. found nintedanib to reduce gene expression of type I collagen and fibronectin and attenuate the secretion of type I and III collagen, whereas pirfenidone showed less pronounced effects even in concentrations of up to 1000 μM [48]. Interestingly, both compounds affected type I collagen fibrils by delaying their formation and resulting in fewer and thinner collagen fibrils as compared with untreated controls. Using the same concentration of pirfenidone in the current assay, anti-fibrotic effects of pirfenidone were evident on fibroblasts although effects were more pronounced for nintedanib and omipalisib. When comparing our data with those published by others, it should be taken into consideration that the biomarkers used to assess ECM protein synthesis in the prolonged Scar-in-a-Jar assay are dependent on the release of collagen pro-peptides to the supernatant during protein maturation. Thus, if changes in the amount of collagen that is laid down in the matrix without proper processing occurred, it would not be registered by the biomarkers. Additionally, any effect on fibrosis resolution, which may be enhanced by anti-fibrotic compounds, may not be assessed in this system.

In our experiments we aimed to test anti-fibrotic drugs at concentrations similar to those achieved in patients. After standard dosing of nintedanib, plasma concentrations of approximately 70 nM may be reached [56] and concentrations of 1–1000 nM were used with only minor cytotoxic effects in this model. In the literature, pirfenidone is applied in concentrations up to 10 mM in vitro [54, 57‐59] but in an effort to keep within the physiologically relevant range, pirfenidone was applied in concentrations of 30–3000 μM as plasma levels may reach concentrations of 80–100 μM after standard dosing of pirfenidone [56]. Notably, no marked effects were observed for pirfenidone in concentrations below 300 μM. These relatively high concentrations of pirfenidone may force an effect in vitro which is not physiologically relevant. A limitation in our assay is that the higher concentrations of pirfenidone required relatively high concentrations of DMSO which in itself had significant cytotoxic effects at the highest used concentration. However, at concentrations of less than 1000 μM pirfenidone, effects of the solvent were not evident based on the Alamar blue or LDH release.

Even though the investigated in vitro model does not reproduce the complexity of the lung tissue during IPF, it allows for an efficient investigation of the ECM production by lung fibroblasts that resembles the process in vivo, since growing the fibroblast in a ficoll-rich moiety ensures that collagens are deposited in the matrix in the natural conformation [33]. Collagen pro-peptides are cleaved off to allow for correct processing and maturation, including cross-linking. However, a simple single-cell culture which does not include other cell types or production of all relevant proteinases is limited in the investigation of the fibrolytic aspect of collagen turnover. Another limitation of the model is the use of a single pro-fibrotic stimulus, meaning that anti-fibrotic compounds tested in the current model will need to work through the TGF-β1 signaling pathway. Future experiments should investigate the use of other pro-fibrotic mediators, including PDGF, to extend the current model to better represent fibrogenesis. The presented experiments consists of only a few healthy donors of lung fibroblasts, even though the experiment was performed three times to account for technical variability and showed excellent reproducibility. Fibroblasts originating from a range of healthy donors, as well as IPF patients, should be evaluated for their fibrotic response in the prolonged Scar-in-a-Jar model to ensure a continuously working and consistent model.

In conclusion, the combination of a prolonged Scar-in-a-Jar model with clinically validated biomarkers of ECM synthesis and the evaluation of anti-fibrotic treatments already approved by the FDA, allowed us to develop a robust fibroblast model to use as a screening tool in early drug development for IPF. Additionally, we have shown the importance of the temporal aspect when evaluating ECM accumulation and that more long-term effects should be assessed. This tool may enable the selection of drug candidates and aid translational science, potentially favoring a more successful path in drug development.