Idiopathic pulmonary fibrosis (IPF) is a progressive lung disease with a median survival of only 3–5 years, worse than many common cancers (Navaratnam et al., 2011; Raghu et al., 2011). It is characterised by parenchymal lung fibrosis leading to impairment of gas exchange and death (King et al., 2011). Although there are recognised risk factors such as smoking, the aetiology of IPF remains unknown. Repetitive damage to the alveolar epithelium and capillary endothelium, activation and proliferation of myofibroblasts and the disordered deposition of extracellular matrix (ECM) result in destruction of the alveolar architecture. Transforming Growth Factor beta 1 (TGFβ1) released following epithelial cell damage is a key upstream pro-fibrotic growth factor driving IPF pathophysiology (Raghu et al., 1989; Border and Noble, 1994; Branton and Kopp, 1999; Leask and Abraham, 2004).

There are only two drugs approved for the treatment of IPF, nintedanib and pirfenidone. Both drugs slow disease progression, but only modestly, and are often poorly tolerated (King et al., 2014). There therefore remains a major unmet clinical need. Pirfenidone exerts anti-fibrotic, anti-oxidant and anti-inflammatory effects to reduce lung collagen synthesis and deposition in bleomycin animal models, and in animal and human lung fibroblasts (Hilberg et al., 2012; Lehtonen et al., 2016; Knüppel et al., 2017). However, the molecular mechanisms of pirfenidone have not been fully elucidated as most mechanistic studies have concentrated on its effects in isolated lung cells, such as fibroblasts (Conte et al., 2014; Epstein Shochet et al., 2018; Jin et al., 2019). Nintedanib, in comparison, is an inhibitor of the platelet-derived growth factor (PDGF), vascular endothelial growth factor (VEGF) and fibroblast growth factor (FGF) receptor tyrosine kinases, collectively inhibiting signalling pathways involved in proliferation, migration and maturation of lung fibroblasts (Hostettler et al., 2014). Its anti-fibrotic activity has been demonstrated in animal models, in vitro assays and in clinical trials (Richeldi et al., 2014; Wollin et al., 2014; Wollin et al., 2015; Knüppel et al., 2017). To date, the molecular effects of these drugs have not been tested extensively in human lung tissue where cell function can be studied in the natural 3D tissue environment. Understanding the mechanisms of action these drugs have in human lung tissue and their relative efficacy may uncover anti-fibrotic pathways that can be targeted more effectively, clarify where these drugs fail to influence the fibrotic process, as well as providing a benchmark for the assessment of novel putative anti-fibrotic compounds.

Several studies have confirmed that human lung tissue is a valuable tool for the study of early pro-fibrotic signalling pathways (Alsafadi et al., 2017; Lehmann et al., 2018; Roach et al., 2018). Perhaps more importantly, these models are capable of screening and validating novel anti-fibrotic therapies rapidly and effectively, and therefore have great potential as more accurate predictors of clinical drug efficacy than currently used animal models. Our recently described model uses TGFβ1 to stimulate early fibrotic events in ex vivo human lung tissue, and recapitulates many of the gene expression changes present in IPF tissue (DePianto et al., 2015; Roach et al., 2018). This model does not require lung inflation with agarose which is major technical advantage, and is sensitive to drug interventions (Roach et al., 2018).

To-date the only study of pirfenidone and nintedanib in human lung tissue investigated the effects of these drugs on alveolar epithelial cell gene expression in precision cut lung slices stimulated with a pro-fibrotic cocktail from three donors (Lehmann et al., 2018). We have therefore used our model to investigate the molecular and histological pathways impacted by pirfenidone and nintedanib, and have compared these with the putative anti-fibrotic compounds senicapoc, an inhibitor of K_Ca3.1 ion channels (Roach and Bradding, 2020), and the resolving mediator lipoxin A4 (LXA₄) (Roach et al., 2015a).

Assessment of tissue viability, cellular senescence, gene expression and protein expression in human lung tissue exposed to TGFβ1.

We have previously demonstrated that control and TGFβ1-stimulated human lung tissue is viable over 7 days with no evidence of nuclear fragmentation, tissue necrosis or reduction in metabolic activity (MTS turnover), and that RNA integrity is maintained (Roach et al., 2018). This was mirrored in the current experiments, and RNA integrity was again preserved with all RIN values > 8. To examine whether culture of the tissue in TGFβ1 promoted cellular senescence we examined senescence markers via RT-PCR. We found no difference between control and TGFβ1-treated tissue after 7 days, using a panel of diverse senescence markers, including p16, p21, p53 (Herranz and Gil, 2018) together with markers of the senescence secretome IL-1β, IL-6 (McHugh and Gil, 2018) and recently identified membrane-specific proteins preferentially expressed in senescence cells (Althubiti et al., 2014) (n = 5 donors) (Figure 1A).

Early fibrotic changes were present in the TGFβ1-stimulated tissue at both day 4 and day 7 identified using RT-PCR on a select panel of key fibrosis-associated genes, including the myofibroblast activation marker αSMA (ACT2A), and extracellular matrix deposition markers fibronectin (FN1), collagen type 1 (COL1A2) and collagen type 3 (COL3A1) (n = 3) (Figure 1B).

Using the 84-well RT² fibrosis array, at day 7 there was upregulation of fibrosis-associated gene expression as seen previously (Roach et al., 2018). In keeping with our previous work, 45 out of 84 pro-fibrotic and pro-inflammatory gene transcript levels associated with IPF were increased significantly in the TGFβ1-treated tissue and 36/45 of these demonstrated a biologically relevant ≥0.5 log₂FC increase when compared to control (n = 32 donors). Genes that were upregulated encoded tissue remodelling growth factors (CTGF, PDGFA and B, TGFβ1), cell adhesion molecules (integrins αv, β1, β6), ECM remodelling enzymes (matrix metalloproteinases [MMPs] 2,-3,-13,-14), extracellular matrix constituents (collagens type I and 3), transcription factors (SMAD6,-7) and αSMA (Figures 2A,B ). Immunohistochemical analysis showed that following TGFβ1 treatment there were quantifiable increases in lung parenchymal collagen type 1 (n = 17) and 3 (n = 15), and αSMA+ (n = 17) and FSP + cells (n = 14) (Figures 3A,B). The PCR and immunohistochemistry results are consistent with our previously published work (Roach et al., 2018) and observations in IPF tissue (DePianto et al., 2015; Zhang et al., 2019).

In addition to upregulating mRNA for pro-fibrotic molecules, TGFβ1 also downregulated mRNA for three molecules with a ≥0.5 log₂FC (IL13RA2, MMP9, caveolin 1). Downregulation of these molecules potentially promotes fibrosis (Cabrera et al., 2007; Lumsden et al., 2015; Marudamuthu et al., 2019). Diagrams summarising the fibrotic pathways regulated by TGFβ1 exposure in this model are summarised in Supplementary Figure S2.

Resident tissue macrophages are key regulators of tissue repair and fibrosis, and disturbances in their function have been associated with IPF (Wynn and Vannella, 2016; Misharin et al., 2017). We found no change in total, airspace or tissue macrophages as a percentage of total cells between control and TGFβ1-exposed lung tissue (Figure 3C).

We have used our human tissue model of early lung fibrosis to perform the first detailed analysis of the anti-fibrotic effects of pirfenidone and nintedanib in human lung tissue stimulated with the key pro-fibrotic growth factor TGFβ1. Our data show that this model is highly reproducible in terms of its response to TGFβ1, and replicates many of the molecular changes characteristic of IPF such as increases in pro-fibrotic gene expression, tissue and soluble collagens, and increased myofibroblast numbers (DePianto et al., 2015). In total, the mRNA for 36 out of 84 fibrosis-associated genes was upregulated with a significant ≥0.5 log₂FC in response to TGFβ1, representing ECM remodelling enzymes, ECM components, integrins, growth factors and TGFβ1 signalling components. A further three potentially anti-fibrotic genes were also downregulated by TGFβ1 with a significant ≥0.5 log₂FC. Similar to IPF, we did not see a strong or persistent response of inflammatory cytokines and chemokines. In parallel, there were increases in fibroblasts expressing fibroblast-specific protein, cells expressing αSMA, deposition of type 1 and 3 collagens, and secretion of soluble collagen. This model therefore replicates several fibrotic processes and gene expression changes present in human IPF tissue (DePianto et al., 2015), and is reproducible.

The effects of both pirfenidone and nintedanib have been tested extensively in animal models of fibrosis and primary IPF-derived cells. Pirfenidone inhibits fibrosis in bleomycin-treated animals and inhibits human lung fibroblast proliferation, differentiation and collagen secretion through the inhibition of TGFβ1-dependent signalling (Hilberg et al., 2012; Lehtonen et al., 2016; Knüppel et al., 2017; Epstein Shochet et al., 2018; Jin et al., 2019). However, it’s precise molecular mode of action remains elusive. Nintedanib is an inhibitor of PDGF, VEGF and FGF receptor tyrosine kinases, and similar to pirfenidone, inhibits lung fibrosis in animal models and inhibits pro-fibrotic human lung fibroblast activity (Hostettler et al., 2014; Wollin et al., 2015; Lehtonen et al., 2016; Knüppel et al., 2017; Epstein Shochet et al., 2018). However, to-date only one study has reported the effects of pirfenidone and nintedanib in n = 3 human lung samples using a precision cut lung slice (PCLS) model that uses agarose to inflate whole lobes and uses a cocktail of TGFβ1, PDGF-AB, TNFα and lysophophatidic acid (Lehmann et al., 2018). This study focused on alveolar epithelial cell function, and showed that pirfenidone and nintedanib did not alter tissue viability, and that nintedanib treatment restored epithelial gene expression, prosurfactant protein C protein expression, and surfactant protein secretion that was suppressed by the pro-fibrotic cocktail (Lehmann et al., 2018). Pirfenidone did not have an impact on any of these markers of alveolar epithelial cell function. These data were supported by experiments in mouse bleomycin model lung slices and mouse type II alveolar epithelial cells.

Using our model of TGFβ1-dependent fibrogenesis, neither pirfenidone nor nintedanib inhibited type 1 or 3 collagen gene expression, although both drugs had significant inhibitory effects on interstitial type 1 and 3 collagen deposition as assessed by immunohistochemistry, and soluble collagen secretion, raising the possibility that they inhibit collagen production/secretion in human lung tissue through post translational effects. Nintedanib inhibited the gene expression for the chemokine receptor CXCR4 (implicated in cell recruitment and migration), the ECM remodelling enzymes MMP1, −9, −13 and −14, PDGFB, and the integrin α2. In addition, nintedanib increased the expression of decorin, a glycosaminoglycan that may inhibit TGFβ1 activity (Giri et al., 1997). Thus nintedanib clearly displays anti-fibrotic activity in this model. However, nintedanib also enhanced the TGFβ1-dependent expression of several profibrotic molecules (CTGF, thrombospondin 1 and 2, angiotensinogen, and αSMA) which may counteract some of the inhibitory effects on fibrosis. Furthermore, αSMA-positive cells were not reduced in the tissue. This might explain why nintedanib is less effective clinically than might be predicted from in vivo animal experiments and studies on lung cells.

Pirfenidone exhibited only weak anti-fibrotic activity in this model at the level of gene expression, and did not inhibit αSMA expression at the gene level or the protein level in the tissue. However, pirfenidone did not increase the gene expression of any pro-fibrotic molecules, and thus while exhibiting potentially less anti-fibrotic activity than nintedanib in human lung tissue, it did not counteract these inhibitory effects, which may explain why it is as effective as nintedanib clinically (King et al., 2014; Richeldi et al., 2014).

Fibroblast numbers and αSMA-positive cells were increased significantly in human lung tissue following 7 days of TGFβ1 treatment. There were more αSMA-positive cells than fibroblast-specific protein-positive cells indicating that not all αSMA-positive cells are (myo) fibroblasts, in keeping with the known expression of αSMA by other cell types. The reduction in fibroblast-specific protein-positive cells with both pirfenidone and nintedanib would be in keeping with the inhibition of acutely proliferating fibroblasts within the tissue, which is consistent with the ability of these drugs to inhibit the proliferation of human lung fibroblasts in culture (Conte et al., 2014; Hostettler et al., 2014). However, the failure to inhibit αSMA mRNA or protein expression in the tissue suggests that cells transitioning to myofibroblasts are not inhibited. Some of these cells could potentially be epithelial cells undergoing epithelial-mesenchymal-transition which is well recognised in IPF (Willis et al., 2005). These data on αSMA expression conflict with animal models of fibrosis, which suggest that pirfenidone and nintedanib decrease myofibroblast accumulation and differentiation, but are in keeping with the most recent histopathological results from IPF patients treated with either nintedanib and pirfenidone, where both drugs failed to reduce the density of the pathognomonic fibroblast foci (Zhang et al., 2019).

Understanding the potential anti-fibrotic effects of nintedanib and pirfenidone in TGFβ1-stimulated human lung tissue provide a potential benchmark with which to compare novel putative anti-fibrotic compounds. K_Ca3.1 ion channels are implicated in fibrosis in several organs, and demonstrate a major role in pro-fibrotic human lung (myo) fibroblast activities (Roach and Bradding, 2020), and in a sheep model of bleomycin-dependent lung fibrosis (Organ et al., 2017). We previously described the anti-fibrotic effects of the K_Ca3.1 ion channel in our ex vivo human lung model of fibrogenesis and showed significant inhibition of 28 TGFβ1-dependent genes including many molecules implicated in IPF pathophysiology, as well as reduced tissue fibroblast numbers, αSMA protein expression and tissue type I and III collagen expression. There was no upregulation of pro-fibrotic gene expression. We have further explored the molecular pathways that are likely to be impacted by nintedanib, pirfenidone and senicapoc, such as ECM-related molecules, cytokines and chemokines, growth factors, TGFβ superfamily members, and transcription factors (Supplementary Figures 3, 4, 5). Senicapoc inhibited molecules within all of these pathways, whereas nintedanib was more restricted and pirfenidone only inhibited MMPs within the ECM remodelling pathway. Coupled with the factor analysis of co-regulated genes, where senicapoc shows greater activity than either nintedanib or pirfenidone, and the potentially pro-fibrotic activity of nintedanib, we would predict that senicapoc will be more effective at treating human lung fibrosis than the two currently licensed drugs. Senicapoc was well tolerated in phase 3 trials of sickle cell disease over 12 months (Ataga et al., 2011), and there is clearly a strong case for progressing to clinical trials in human IPF. We expect that benchmarking against nintedanib and pirfenidone will be useful in predicting the clinical efficacy of other novel anti-fibrotic compounds.

Lipoxin A4 and a stable analogue have demonstrated anti-fibrotic activity in HLMFs and the mouse bleomycin lung fibrosis model (Martins et al., 2009; Roach et al., 2015a). There was no effect of lipoxin A4 in this ex vivo human lung model, which suggests that lipoxin A4 is unlikely to demonstrate efficacy in human IPF. It may simply not be stable enough in this model to exert anti-fibrotic activity, even with daily dosing as used here, and it remains possible that stable analogs might be more effective.

No model of disease, animal or human will be perfect, but if a model can be shown to reliably predict drug efficacy in IPF this would be a major advance. Our model does not develop over years unlike clinical IPF, and does not allow for the recruitment of cells from the blood such as fibrocytes. While the gene expression changes in our model are very similar to those reported in IPF tissue, CXCR4 for example is down regulated and so this model would not be suitable for studying CXCR4-related pathways. Further approaches can also shed light on drug mechanism of action in IPF, such as the work of Kwapiszewska et al. (Kwapiszewska et al., 2018). They used transcriptomic analysis to examine the molecular changes in the fibroblasts and lungs of people with IPF who had progressed to lung transplant, and demonstrated alteration of several pathways linked to extracellular matrix architecture and inflammation in patients receiving pirfenidone compared to those who had not received it. Explanted IPF tissue like this is in short supply and therefore difficult to obtain and work with, and the fact that these patients had progressed to lung transplantation shows that pirfenidone was of limited benefit, and this approach is not practical for assessing drugs at the pre-clinical development stage. Studying the effects of drugs in IPF tissue ex vivo is also potentially useful (Mercer et al., 2016) but again limited by tissue availability, and it is unclear whether explanted severely fibrotic tissue is too far down the fibrotic pathways to provide useful information on drugs that might halt disease progression. Clinical trials will ultimately prove whether our model described here is better than other approaches used to-date.

In summary, we have provided novel data regarding the molecular effects of nintedanib and pirfenidone in human lung tissue displaying early features of fibrosis after exposure to TGFβ1. Both drugs demonstrated some anti-fibrotic activity, but the effects of both drugs on gene expression were modest, and nintedanib also upregulated some pro-fibrotic molecules. This may explain why these molecules are only partially effective at treating human IPF, and suggests that there is substantial room for improvement with the development of new compounds. Towards that goal, the K_Ca3.1 channel blocker senicapoc holds great promise when benchmarked against both nintedanib and pirfenidone.