Harnessing the translational power of bleomycin model: new insights to guide drug discovery for idiopathic pulmonary fibrosis

Abstract

Background: Idiopathic pulmonary fibrosis (IPF) is a chronic, progressive, age-related interstitial lung disease (ILD) with limited therapeutic options. Despite the wide variety of different in vivo models for IPF, these preclinical models have shown limitations that may significantly impair their translational potential. Among the most relevant limitations are the methodologies used to assess the efficacy of anti-fibrotic treatments, that are not the ones used in humans. In this scenario, the goal of the work presented in this paper is to provide translational relevance to the bleomycin (BLM)-induced pulmonary fibrosis mouse model, introducing and validating novel readouts to evaluate the efficacy of treatments for IPF. Methods: The BLM model was optimized by introducing the use of functional assessments such as the Forced Vital Capacity (FVC) and the Diffusion Factor for Carbon Monoxide (DFCO), that are respectively the primary and secondary endpoints in clinical trials for IPF, comparing them to more common readouts such as lung histology, improved by the application of Artificial Intelligence (AI) to detect and quantify fibrotic tissue deposition, and metalloproitenase-7 (MMP-7), a clinical prognostic biomarker. Results: Lung function measurement and DFCO changes well correlated with Ashcroft score, the current gold-standard for the assessment of pulmonary fibrosis in mice. The relevance and robustness of these novel readouts in the BLM model was confirmed by the results obtained testing Nintedanib and Pirfenidone, the only drugs approved for the treatment of IPF patients: in fact, both drugs administered therapeutically, significantly affected the changes in these parameters induced by BLM treatment, with results that closely reflected the efficacy observed in the clinic. Changes in biomarkers such as MMP-7 were also evaluated, and well correlated with the modifications of FVC and DFCO. Conclusion: Novel functional readouts such as FVC and DFCO can be efficiently used to assess pathology progression in the BLM-induced pulmonary fibrosis mouse model as well as compound efficacy, substantially improving its translational and predictivity potential.

Keywords: Ashcroft score; Broncho alveolar lavage fluid; Nintedanib; Pirfenidone; automated histology imaging analysis; diffusion factor for carbon monoxide; forced vital capacity; metalloproteinase-7.

FIGURE 1

Experimental protocol of a therapeutic BLM OA Study.

FIGURE 2

Schematic classification of lung fibrosis severity distribution by AI-based Visiopharm^® APPs.

FIGURE 3

Effect of Nintedanib (A) and Pirfenidone (B) on DFCO decline at day 21 after the first BLM OA administration in mice. Data represent mean ± SEM values for BLM/Veh (n = 25 for Nintedanib study and n = 23 for Pirfenidone study; black bar) and BLM/Nintedanib (n = 21; blue bar) or BLM/Pirfenidone (n = 24; green bar) groups of animals expressed as difference vs. mean of Saline group (n = 12 for each study). Statistical analysis was assessed using unpaired Student’s t-test vs. the BLM group; *p ≤ 0.05, **p ≤ 0.01. BLM, bleomycin; Veh, vehicle.

FIGURE 4

Effect of Nintedanib (A) and Pirfenidone (B) on the decline in FVC at day 21 after the first BLM OA administration in mice. Data represent mean ± SEM values for BLM/Veh (n = 25 for Nintedanib study and n = 23 for Pirfenidone study; black bar) and BLM/Nintedanib (n = 21; blue bar) or BLM/Pirfenidone (n = 24; green bar) groups of animals. Data are expressed as difference of mL vs. the observed mean of the Saline group (n = 12 for each study). Statistical analysis was carried out using unpaired Student’s t-test in comparison with the BLM group; **p ≤ 0.01, ***p ≤ 0.001. BLM, bleomycin; Veh, vehicle.

FIGURE 5

Pearson’s correlation coefficient and linear regression analysis of the relationship between FVC (mL) and DFCO among all the experimental groups (Saline, Bleomycin, Nintedanib or Pirfenidone) in Nintedanib (A) and Pirfenidone (B) studies. ***p ≤ 0.001.

FIGURE 6

Effect of Nintedanib (A) and Pirfenidone (B) on Ashcroft score. Data represent mean ± SEM values for SAL/Veh (n = 12 for each study; white bar), BLM/Veh (n = 25 for Nintedanib study and n = 23 for Pirfenidone study; black bar) and BLM/Nintedanib (n = 21; blue bar) or BLM/Pirfenidone (n = 24; green bar) groups of animals. Statistical analysis was assessed using one-way ANOVA followed by Dunnett’s test in comparison with the BLM group; ***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05. SAL, Saline; BLM, bleomycin; Veh, vehicle.

FIGURE 7

Effect of Nintedanib (A) and Pirfenidone (B) on tissue classification of lung fibrosis severity distribution using AI-based Visiopharm^® APPs. Data represent mean ± SEM values for BLM/Veh (n = 25 for Nintedanib study and n = 23 for Pirfenidone study) and BLM/Nintedanib (n = 21) or BLM/Pirfenidone (n = 24) groups of animals. Statistical analysis was assessed using one-way ANOVA followed by Dunnett’s test in comparison with the BLM group; ***p ≤ 0.001, **p ≤ 0.01. BLM, bleomycin; Veh, vehicle.

FIGURE 8

Effect of Nintedanib (A) and Pirfenidone (B) on the levels of MMP-7 in BALF. Data represent mean ± SEM values for SAL/Veh (n = 12 for each study; white bar), BLM/Veh (n = 25 for Nintedanib study and n = 23 for Pirfenidone study; black bar) and BLM/Nintedanib (n = 21; blue bar) or BLM/Pirfenidone (n = 24; green bar) groups of animals. Statistical analysis was assessed using one-way ANOVA followed by Dunnett’s test in comparison with the BLM group; ***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05. BLM, bleomycin; Veh, vehicle.

FIGURE 9

Correlation between MMP-7 levels in BALF and the decline of DFCO assessed in both Nintedanib (A) and Pirfenidone (B) studies. The correlation between MMP-7 (pg/mL) and decline of DLCO was carried out by Pearson’s correlation coefficient and linear regression analysis ***p ≤ 0.001.

FIGURE 10

Correlations between Ashcroft score analysis, moderate and severe fibrotic tissue by AI, FVC, DFCO and MMP-7 at day 21 after BLM OA administration in mice treated with Nintedanib (A) or Pirfenidone (B) or vehicle. Correlations were determined using Pearson’s correlation coefficient matrix tests.