Pharmacological validation of SSc-ILD mouse model bleomycin-induced by osmotic minipump

Systemic sclerosis (SSc) is an autoimmune disease characterized by an excessive production and accumulation of collagen in the skin and internal organs often associated with interstitial lung disease (ILD). The unknown pathogenetic mechanisms of SSc-ILD and the lack of animal models mimicking the features of the human disease contribute to create a gap between the selection of antibrotic drug candidates and effective therapies. Nintedanib (NINT) was used as a tool compound to validate the pharmacological response either on lung or skin brosis in a SSc-ILD mouse model. The model is based on the continuous infusion of bleomycin (BLM) by osmotic minipumps for 1 week in the C57BL/6 female mice. Longitudinal Micro-CT analysis highlighted a signicant slowdown in lung brosis progression after NINT treatment, then conrmed by histology. However, no signicant effect was observed on lung hydroxyproline content, inammatory inltrate and skin lipoatrophy. The modest pharmacological effect reported reects the clinical outcome, lighting up the reliability of this model to serve as secondary screening to prole the best clinical drug candidates. Moreover, we have underlined the pivotal role of Micro-CT imaging, together describing the relevant readouts and the importance of their validation prior to use for drug discovery.


Introduction
Systemic sclerosis (or scleroderma) (SSc) is an autoimmune disease of unknown aetiology, characterized by vasculopathy, excessive brous tissue generation and aberrant immune activation resulting in damage to various organs including the skin, esophagus, heart, lungs, and kidneys 1,2 . Most critically, about 80% of SSc patients develop pulmonary dysfunctions such as pulmonary hypertension and associated interstitial lung disease (ILD), this latter characterized by early immune cell in ltration followed by various degree of brosis and gas exchange impairment that signi cantly reduces life expectancy [3][4][5] . The pronounced uncertainty surrounding the pathogenetic mechanisms behind SSc-ILD and the absence of effective treatments for this disorder have solicited the creation of ad hoc animal experimental models capable of mimicking the different clinical and pathological peculiarities of SSc-ILD in humans 6-8 .
In a recently published paper, an experimental model of human SSc-ILD in C57BL/6 mice has been described. This model involved a continuous and controlled release of BLM by subcutaneously implanted osmotic mini-pumps 15 . Before using a new animal model in drug discovery, its reproducibility, robustness and ability to respond to pharmacological treatments should be validated. The selection of the best drug candidate to be introduced into the clinic depends on many factors, but a head-to-head comparison with a gold standard therapy has an important role. Several crucial points, such as time, doses of administration and different readouts, are essential to provide the best view on how drugs are performing in animal models, since each model can respond to the same therapy in a different manner.
A model inducing subacute -chronic brosis through the release of BLM by osmotic mini-pumps allows evaluation about the effects of anti brotic compounds, either in the skin or in the lungs, since the target brotic lesions have been identi ed in both organs [15][16][17] .
The main purpose of this study was to examine the anti brotic activity of nintedanib (NINT), a tyrosinekinase inhibitor used to treat idiopathic pulmonary brosis and recently approved as a therapy for SSc associated with ILD 18,19 , in the aforementioned SSc-ILD mouse model.
In the present research, NINT was orally administered for two weeks following a therapeutic protocol, starting at day 14 20,21 . In order to use this brosis model for secondary screening in our pipeline several readouts were considered. The lung brosis progression and the pharmacological response to treatment were longitudinally assessed by micro-CT, and then corroborated by histological analyses in lung and skin samples at the endpoint experiment.

Animal model
Twenty-four 7-8-week-old C57BL/6 female mice (Envigo, Italy) were kept in a conventional animal facility in ventilated cages with free access to standard rodent chow and softened tap water at least 7 days prior to use. All mice were randomly subdivided into two groups: 8 were treated with saline and 16 with BLM.

Ethical statement
All experiments were carried out in the animal unit of Chiesi Farmaceutici, an AAALAC Internationalaccredited facility, in accordance with the intramural animal-welfare practices for animal experimentation approved by the Animal Welfare Body of Chiesi Farmaceutici and authorized by Italian Ministry of Health (protocol number: 449/2016-PR). The study was also in compliance with the European Directive 2010/ 63 UE, Italian D.Lgs 26/2014 as well as the revised "Guide for the Care and Use of Laboratory Animals" released by the National Research Council Committee (US, 2011) 22 and reported in accordance with ARRIVE guidelines 23 .

Experimental design
Each mouse was anesthetized with 2.5% iso urane mixed with oxygen, the implantation site was shaved and, through a small incision, a subcutaneous pocket was created on the left-hand side of its back using the jaws of a hemostat clamp. The osmotic minipumps [ALZET 1007D; DURECT, (release rate 0.5 µl/h for 7 days), Cupertino, CA] containing either 100 µl saline or BLM (Baxter Oncology GmbH, 60mg/kg dissolved in saline) were implanted and removed after 8 days. At day 14, BLM mice were randomly divided in two groups, receiving either NINT (Carbosynth, 60mg/kg/die) dissolved in Tween80 0.05% in saline or vehicle (Tween80 0.05% in saline), by gavage, daily for two weeks (Fig. 1).
The animals were monitored and weighed daily throughout the experimental procedure.
A Visual Analogue Scale (0-10) for pain assessment was assessed daily by a designed veterinarian or trained technicians. VAS ≥7 and/or body weight loss ≥ 20% were considered as humane endpoints, as well as signs of dyspnoea or apathy evaluated by a designed veterinarian.

Micro-computed tomography acquisition protocol
Micro-computed tomography (micro-CT) lung imaging was performed longitudinally at day 14 and 28 by Quantum GX Micro-CT (PerkinElmer, Inc. Waltham, MA). Each mouse was anesthetized using 2% iso urane and then positioned inside the CT scan. Images were acquired with the following parameters: X-ray tube voltage 90 KV, X-ray tube current 88 µA and total scan time of 4 minutes. A ring reduction correction was applied to the sinograms and the entire set of projection radiographs was entered into a GPU-based ltered back-projection algorithm with a Ram-Lak lter 24 . The acquisition protocol in 'high resolution' mode resulted in one 3D dataset with 50 μm isotropic reconstructed voxel size.
Image post-processing: lung segmentation protocols and analysis.
For each acquisition, a stack of 512 cross-sectional images was produced. The reconstructed datasets were analyzed using the Perkin Elmer Analyze software (Analyze 12.0; Copyright 1986-2017, Biomedical Imaging Resource, Mayo Clinic, Rochester, MN). The images stacks were ltered and converted from grey levels to CT numbers (Houns eld Units -HU). The conversion is a linear transformation setting -1000 HU as the density of air and 0 HU as the density of water. A semi-automatic segmentation was used to extract airways and lungs. For the quantitative assessment of the lung parenchyma, HU clinical ranges were applied on rescaled HU images to the segmented lung volume to de ne normo-aerated [(-900, -500) HU] and poorly-aerated [(-500, -100) HU] tissues 25 . The two compartments with a different aeration degree were expressed as percentage of the total lung volume. The poorly-aerated tissue refers to a low gas/tissue ratio and it was used to quantify lung brosis progression and evaluate the e cacy of NINT 26 .
After micro-CT imaging at day 28, all mice were euthanized with an overdose of anesthetic follow by bleeding from the abdominal aorta. Bronchoalveolar lavage uid (BALF) was collected by gently washing the bronchial tree using 0.6 mL sterile solution three times [Hank's balanced salt solution (HBSS) ×10; ethylenediaminetetraacetic acid (EDTA) 100 mM; 4-(2-hydroxy-ethyl)-1-piperazineethansulphonic acid (HEPES) 1 mM; distilled water]. The samples were centrifuged at 300 x g for 10 minutes at 4°C and the supernatant collected and frozen for further investigation. The cellular pellet was resuspended in 0.2 mL of BALF solution and total white blood cells (WBC) were measured using an automated cell counter (Dasit XT 1800J). Afterwards, about 1.0x10 6 cells were also used to quantify the M2 macrophage population by ow cytometry. The cells were suspended in FACS Buffer (PBS; 0.5% BSA) and in the lysis buffer to remove red blood cells (BD Bioscience). Then the cells were stained with anti-CD206 (Bio-Rad), anti-F4/80 (BioLegend) and anti-CD11b (BioLegend) antibodies, washed, and nally acquired using a FACS Canto II Cytometer (BD Bioscience) and analyzed with FACS Diva software. The total macrophage population was selected based on forward (FSC) versus side scatter (SSC) plots, and, subsequently, M2 macrophages were identi ed in terms of total number of anti-CD206 + events within a F4/80-CD11b positive selected population.
The matrix metalloproteinases 2 and 9 (MMP-2 and MMP-9, respectively) and metallopeptidase inhibitor 1 (TIMP-1) concentrations in BALF were assessed by enzyme-linked immunosorbent assay (ELISA) kit (R&D Systems, Minneapolis, USA). The protein concentrations were measured by interpolation from the standard curve and were expressed in fold of increase (FOI).

Histological analysis and brosis quanti cation
The whole left lung and skin from the left gluteal region (i.e., distant from the implant site to avoid dermal brosis artifact) were excised. The skin and the lung were xed with 10% paraformaldehyde and embedded in para n. Serial 5 μm thick sagittal sections were stained with hematoxylin-eosin (H&E) to demonstrate the general tissue morphology, and with Masson's trichrome (MT) to evaluate the collagenbased matrix. Whole slide images (WSI) were acquired using NanoZoomer S60 scanner (Hamamatsu Photonics, K.K., Japan). Two independent researchers with experience in animal models of lung brosis performed blind histological analyses.
The extent of brosis was morphologically and qualitatively assessed in a subpleural frame (Region of Interest) of lung parenchyma (250 µm thickness, Fig. 2(a)). The broproliferative modi cations in the frame area were evaluated through the semiquantitative 0-to-8 Ashcroft score. The Ashcroft frame scores were subsequently categorized into mild (mean score from 0 to 3), moderate (4) and severe (5)(6)(7)(8). Moreover, the brotic foci within the frame were quanti ed based on morphological and colorimetric thresholds and labelled as "areas of interest" (AOI). The extent of brosis was evaluated through the following histomorphometric parameters: 1) mean number of brotic foci per mm 2 of parenchyma; 2) focus size (small, if its area was <7,500 mm 2 ; large, if >7,500 mm 2 ); 3) fraction of Frame occupied by total AOI area (SAOI area/Frame surface), as a percentage value.
For the skin samples, the histomorphometric parameters considered in the MT stained sections were: 1) dermis thickness, de ned as the mean distance between the epidermal-dermal junction and the dermal-subcutaneous junction 16 ; 2) hypodermis thickness, de ned as the mean distance between the dermalsubcutaneous junction and the muscle layer. The in ammatory in ltrate was evaluated on sections stained with H&E by a semiquantitative method using a 0 to 4 grade score, re ecting increasing in ammation, as described by Gallet et al. (2011) 27 . Measurements were carried out in ve randomly selected elds from one sample from each animal.

Collagen content
The collagen-based extracellular matrix was measured using the image analysis software NIS-Elements AR 3.1 (Nikon Tokyo) in the TM-stained lung sections after selection of a correct green threshold detected on the Light Green stained collagen bers to eliminate air spaces and bronchial epithelium 28 .

Hydroxyproline quanti cation
The right lung lobes were used to quantify the collagen indirectly through hydroxyproline (Hyp) concentration using a commercial kit (Sigma Aldrich) in accordance with the manufacturer's protocol. In brief, the lobes were homogenized in PBS, hydrolyzed in 6 N HCl for 24 hours at 100°C and nally neutralized in 6N NaOH. The nal Hyp concentration was determined by the reaction of oxidized Hyp with 4-Dimethylamino benzaldehyde (DMAB), which resulted in a colorimetric product, proportional to the Hyp content. This reaction was measured at a wavelength of 560 nm and, nally, each total amount of Hyp was normalized for the relevant right lobe weight.

Immuno uorescence staining
Immuno uorescent (IF) reactions to detect M2-like polarized macrophages were performed on para n embedded sections (16). Brie y, sections were incubated using an anti-CD206 primary antibody (1 µg/mL, AF2535; R&D Systems). This reaction was revealed by a secondary donkey anti-goat antibody Alexa Fluor 488 conjugate (3 µg/mL, AB2336933; Jackson Laboratories). Lastly, the nuclei were counterstained with DAPI (Invitrogen). For negative control the primary antibody was omitted and tissues were incubated in 10 mM phosphate buffer or, alternatively, with unlabelled rabbit IgG nonimmune isotype control (2009-1; Alpha Diagnostic International) used at the same concentration as the selective antibody. Fluorescent WSI were acquired using NanoZoomer S60 (Hamamatsu Photonics, K.K.).

Statistical analysis
Statistical analysis was performed using one-way ANOVA followed by Dunnett's t test, to compare each group with the BLM group as control. For micro-CT data a two-way ANOVA test was performed to compare each group with the BLM group and to compare different time points of observation, using Dunnett's and Sidak's tests for multiple comparisons, respectively. The comparison of the frequency distribution was performed with a Chi-squared test. Statistics were carried out using GraphPad Prism 7.0 software (GraphPad; La Jolla, CA, USA). Sample size was calculated with A-priori Power Analysis (G*Power Version 3.1.2) considering Ashcroft Score as endpoint. A value of P<0.05 was considered statistically signi cant. Data are expressed as means and S.E.M.

Results
Experimental Animals BLM induced a weight loss up to a maximum reduction of 18% at day 14 but, as expected, mice recovered at later time points.
NINT showed a well-tolerated pro le, since no difference in body weight was identi ed compared to the BLM group. The Saline group did not exhibit any distress and no weight loss was observed ( Supplementary Fig. 1).

Morphology and brosis quanti cation in subpleural tissue
MT staining highlighted different morphological features in the lungs of BLM and BLM+NINT groups, as compared to the normal parenchyma of the Saline group. The brotic lesions were mainly located in the subpleural lung parenchyma (frame, Fig. 2(a)) in both BLM and BLM+NINT groups and were characterized by collagen deposition with thickening alveolar septa and moderate in ammatory in ltrate.
Representative foci histological features and conglomerations are presented in Fig. 2(b-d).
In accordance with the Ashcroft score, collagen content percentage was signi cantly reduced in NINTtreated mice compared to the BLM group (18%) (Fig. 2(g)).
The effect of NINT was also investigated for number of foci/mm 2 and SAOI area/Frame (%). This analysis highlighted signi cant differences between the Saline and BLM groups, but was not able to discriminate the e cacy of the anti brotic therapy, since both these parameters were not signi cantly reduced by the NINT treatment ( Fig. 2(h-j)). Similarly, the quanti cation of Hyp in the lungs of the BLM group was signi cantly higher compared to Saline but not signi cantly reduced by NINT treatment (Fig.  2(i)). Fig. 2(a)). NINT treatment signi cantly reduced (40%) the total number of WBC measured in BALF, but only a modest inhibition was observed on the macrophage (26%), lymphocyte (24%) and neutrophil (33%) populations ( Supplementary Fig. 2(b-d)). We have investigated the effect of NINT on the M2-like macrophages cell population in BALF, using a ow cytometry technique with CD206 as a surface marker. FACS analysis revealed a marked increase in M2-like cells in BLM-treated mice, compared to Saline, however no signi cant level of inhibition was observed in the BLM+NINT group ( Fig. 3(g)).

BLM induced pulmonary in ammation by recruiting white blood cells (WBC), (Supplementary
IF staining of lung tissue after bronchoalveolar lavage con rmed the presence of residual M2-like macrophages in both BLM and BLM+NINT-treated mice (Fig. 3(a-f)).
Another goal of this study was to evaluate the activity of NINT treatment on skin brosis. The most evident changes in skin morphology of BLM and BLM+NINT, compared to Saline, are due to lipoatrophy (Fig. 5(a-c)). No signi cant differences were found in dermis thickness (Fig. 5(d)), while a signi cant decrease in hypodermis thickness was observed in both BLM and BLM+NINT compared to Saline-treated mice (Fig. 5(e)). Finally, a moderate pro-in ammatory effect of BLM was revealed in skin and only a slight modulation was induced by NINT treatment (Fig. 5(f)).

Micro-CT
Longitudinal micro-CT imaging was performed at 14 (baseline) and 28 days, representing the beginning and the end of NINT treatment.
Representative micro-CT scans of BLM and BLM+NINT-treated mice showed an increase in the poorlyaerated tissue along with a decrease in the normo-aerated tissue (pink and blue, respectively) at 28 days compared to baseline (Fig. 6(a)). On the contrary, Saline lungs were largely composed of normallyaerated tissue and remained unchanged over time.
Poorly-aerated lung compartment was used as a marker of brosis, thus longitudinally quanti ed for each group (Fig. 6(b)). These data revealed that brosis was uniformly distributed at the baseline for BLM and BLM+NINT groups since the percentage of poorly-aerated tissue was not signi cantly different.
The brosis progression was detected in both BLM and BLM+NINT groups at day 28; however statistical signi cance was achieved only for BLM group compared to baseline (Fig. 6(b)).
As previously reported, no alteration in Saline lung parenchyma was reported, in fact, the amount of poorly-aerated tissue was constant throughout the experiment (Fig. 6(b)).
The quanti cation of normally and poorly-aerated lung compartments is usually performed at the end of the experiment to evaluate the e cacy of the pharmacological treatment with respect to BLM control group (Fig. 6(c)). Although poorly-aerated tissue was higher in both BLM or BLM+NINT-treated animals (70 and 54%) compared to Saline (22%), NINT treatment signi cantly limited this worsening (-23%) compared to vehicle-treated mice (Fig. 6(c)).

Discussion
Despite the great advances in knowledge of the etiopathogenesis of SSc-ILD in recent years, medical need remains very high. In particular, it is imperative to nd reproducible and relevant animal models capable of reproducing the chronic and progressive aspects of the disease, and to provide robust readouts in order to test putative new drugs 20,21 . Furthermore, new technologies need to be fully integrated into the anti brotic drug development process for the screening of the best compounds to advance in clinical therapy.
In this study, we explored the anti brotic effects of NINT on lung and skin brosis in a SSc-ILD mouse model, by using different readouts including micro-CT imaging.
NINT showed a signi cant inhibition of the total Ashcroft score and collagen deposition.
Although the anti brotic treatment effect was less pronounced on the Ashcroft frequency distribution, the number of brotic foci/mm 2 and Σ AOI/frame (%), these parameters may still provide important information either to monitor the inter-experiment reproducibility of brosis or to evaluate drug e cacy.
Hyp was not signi cantly modulated by the NINT treatment. Even though Hyp is commonly considered an important readout, in our drug screening experiments concerning BLM-induced lung brosis models, we always observed high variability and poor inter-experiments reproducibility. As previously reported, the brotic lesions, mainly localized in the subpleural area, could be underestimated if evaluated over the whole parenchyma 29 . Similarly, the quanti cation of Hyp could be affected by the size of the sampling site. For this reason, a FOI of 2 between Saline and BLM raises serious questions about whether this range will be su cient to evaluate any anti brotic effect in this model.
Hyp determination, being a destructive assay, precludes other histological analyses or alternative readouts that might be considered 30 .
A signi cant anti-in ammatory activity of NINT was noted only on WBC in BALF; on the contrary, only a mild and non-signi cant inhibition was observed in the M2-like population by FACS. Although the pivotal role of macrophages in IPF has been recently well reported 31,32 , the modulatory effect of NINT on macrophage polarization in vivo has been only demonstrated in Fra2 transgenic mice, ameliorating histological features of pulmonary arterial hypertension 33 . In human macrophages, NINT treatment was able to downregulate M2 markers of expression in vitro 34 . Although NINT showed a modest inhibitory effect in our mouse model, we found that M2-like cells may represent a useful readout for evaluating the anti brotic drug effect, since it might be directly linked to brosis.
NINT modulated levels of matrix metalloproteinases in BALF samples compared to vehicles, however achieving statistical signi cance only for MMP-9. This could be explained by the fact that MMP-9 is mainly expressed by in ammatory cells, whereas MMP-2 is especially released by epithelial cells 35 .
Their tissue inhibitor, TIMP-1, is expressed by the interstitial cells in the brotic areas during wound healing 35 . TIMP-1 levels remain high in the BALF up to day 21 29 , corresponding to the peak of activated macrophages in the alveolar spaces, but decrease after 28 days, probably caused by the progressive depletion of cellular reserves of the inhibitor. For this reason, it is acceptable that NINT may not have any effect upon it.
NINT treatment did not signi cantly reduce either the in ammatory in ltrate in the skin or lipoatrophy, which are the most evident changes induced by BLM administration through osmotic minipumps. This is in accordance with the evidence that, even though NINT has been approved for scleroderma, it has been shown not ameliorate the status of the skin 18, 19 .
Unfortunately, many pre-clinical readouts focused on evaluating anti brotic treatments are terminal procedures which don't re ect the clinical situation. In this scenario, the inhibition of the drug group compared to vehicles remains the unique informative result, precluding any other intra-subject evaluation about the disease development.
Imaging technologies, such as micro-CT, allowed longitudinal studies in the same mice before (baseline) and after drug treatment as its own control 36 .
This pre-clinical setting re ects a more relevant clinical situation along with a drastic reduction in both the variability and number of mice used.
The therapeutic protocol used for NINT, starting the treatment at day 14, could decrease the anti brotic effect, as reported in the study, on different readouts; however, we are committed to bring out the real potentiality of the drug tested and no amplify the pharmacological activity.
NINT only partially reduced lung brosis progression in ILD patients 19 , since it has been reported a decline in the forced vital capacity (FVC) despite the treatment of 52 weeks. In agreement with the clinical outcome, the progressive increase in poorly-aerated tissue observed in BLM and NINT groups, compare to baseline, revealed a worsening of lung brosis which was only partially reduced by the anti brotic treatment.
Taking together, these ndings light up the reliability of this model as a translational tool to pro le new anti brotic drugs.
In this study we pharmacologically validated a new murine model of ILD and SSc, using an FDAapproved anti brotic drug.
We strongly believe that an integrative approach is needed in drug discovery to establish reliable, reproducible and robust readouts in order to better pro le putative drug candidates. We sought to address the strengths and weaknesses of the relevant readouts, as well as the importance of their validation prior to use for drug discovery. Figure 1 Experimental set up. Experimental timeline of bleomycin-induced lung mouse brosis by osmotic minipumps. 24 females C57BL/6 were randomly assigned to receive either saline (8 mice) or BLM (60U/kg) by osmotic minipumps at day 0. From day 14, 8 animals from BLM group received nintedanib and the others 8 received vehicle until day 28. At the endpoint, the effect of nintedanib was assessed by micro-CT and ex-vivo analyses. the BLM group using one-way ANOVA followed by Dunnett's test (**P<0.01; ***P<0.001). In E statistical comparisons with the BLM group were performed using Chi-squared test (**P<0.01; ***P<0.001). The percentages on the BLM+NINT bars represent the inhibition effect of the treatment compared to BLM.