Experimental animals and histology
A mouse model of bleomycin-induced lung fibrosis, currently in use in our institution, has been previously published 30 and has been approved by the internal AWB (Animal Welfare Body) of Chiesi Farmaceutici under protocol number: 841/2019-PR and comply with the European Directive 2010/63 UE, Italian D.Lgs 26/2014 and the revised “Guide for the Care and Use of Laboratory Animals” and the Animal Research: Reporting of In Vivo Experiments (AARIVE) guidelines (www.arriveguidelines.org).
In brief it consists of the oropharyngeal BLM administration to 7-8 weeks old C57Bl/6 female mice of 15 µg/mouse at each day of treatment respectively. Animals were sacrificed at 7, 14, 21 and 28 days after the administration.
After the sacrifice, lungs were removed, formalin fixed and paraffin embedded (FFPE). Three serial sections, 5 μm thick were obtained for the stainings: Hematoxylin and Eosin, Masson’s trichrome (TM) and a multiplex immunostaining.
Whole-slide images were acquired by a NanoZoomer S-60 Digital slide scanner (NanoZoomer S60, Hamamatsu, Japan) at 20x magnification. Fibrotic lung injury was assessed histologically through the Ashcroft scoring system 31,32 on whole parenchyma by two independent researchers (blinded to the experimental design). Detailed description and histomorphometric characterization are included in the Supplementary methods and in Supplementary Figure 1. The most relevant areas on the slide were marked for subsequent tissue microarray construction (see below).
Tissue Microarray design and construction
For the TMA preparation, we used FFPE tissue blocks of lungs from BLM-treated female mice. 13 cases of treated and control lungs were selected; cores were extracted, selecting specific regions representing normal lung tissue with alveolar parenchyma, bronchioles and vessel for saline samples and fibroproliferative foci with its alterations for bleomycin ones from the C57BL/6 female mice FFPE lungs of BLM and saline treated mice across four time points (7, 14, 21 and 28 days, see Supplementary Table 1). The TMA was constructed with a Tissue Microarrayer Galileo CK4500 (Tissue Microarrayer Model TMA Galileo CK4500; Integrated Systems Engineering srl, Milano, Italy) using Galileo Software to match the annotated tissue on histological slide with their corresponding areas on the surface of the paraffin donor blocks. After the core transfer in the recipient block, to allow the samples to be properly embedded into the block, the TMA was incubated for 24h at +38°C. Finally, 5 μm thick serial sections were cut from the TMA with a rotary microtome (Slee Cut 6062, Slee Medical, Mainz, Germany) and placed on Polysine adhesion glass slides (Thermo Fisher Scientific). Routine histology were performed as described previously 30.
In order to select the appropriate Tissue Microarray (TMA) core dimension, after applying a high-plex MILAN staining method 13 on the whole slide of a mouse lung, we selected 4 virtual cores (with ImageJ) of 1 mm and 4 of 2 mm in diameter. Thus we run single cell analysis as previously described on 8 cores and on a whole slide and we compare the results. The level of adequacy of tissue portion was set when all main cell populations found in whole tissue by clustering analysis (Rphenograph package ) were rediscovered in cores.
Lung samples were processed as per the Milan (Multiple Iterative Labeling by Antibody Neodeposition) protocol 13 modified for mouse FFPE tissue staining. The protocol consists in the cyclic application of primary antibodies (Table 1) raised in multiple species, including mouse 12, applied on the same section, nuclear staining with DAPI, autofluorescence subtraction 9. For each round, the stained slides were scanned and saved as digital images using a multichannel fluorescence acquisition instrument (NanoZoomer S60, Hamamatsu, Japan). The antibody stripping preceded a subsequent staining round. A multichannel fluorescence acquisition was performed after each stripping, in order to check the complete antibody removal.
In order to generate an antibody panel which would produce single cell lung classification, we mined existing single cell RNA sequencing (scRNAseq) studies 24,25 for highly expressed, lineage-restricted messages which would correspond to proteins against which antibodies would be available. The staining sequence design and antibodies specifications were provided in Table 1. The methods used for the antibody selection and validation are reported in Supplementary methods.
Sections were incubated overnight with primary antibodies, which were then revealed by secondary fluorochrome-tagged antibodies; both isotype control and omission of primary have been performed as negative control. Sequential stripping was obtained with a beta-mercaptoethanol and sodium dodecyl sulphate mix 9.
Mouse tissue preservation after each of multiple staining and stripping rounds showed no or negligible cell loss (Supplementary Fig 2 and 3), as measured by DAPI nuclear stain and quantification on four other tissues (kidney, liver, lung and heart).
Single cell analysis
After the stainings were acquired, digital slide images (.ndpi) were imported as .tiff and registered with the AMICO software 33, based on Fiji. All DAPI images from different staining rounds were registered together and their coordinates of rotation and translation were used to align the individual marker images of each round. Once all images were aligned, autofluorescence was subtracted from FITC and TRITC channels.
DAPI stainings were used for cell segmentation with Fiji (Threshold Default -Watershed_Analize particles and Count mask creation)
Mean intensity value of all markers within each segmented cell and spatial coordinates of centroids of nuclei were recorded together in a .csv file.
Then .csv files were uploaded to the R Studio software (version 1.4) for a more detailed analysis. Rtsne 34, umap 35 and Rphenograph36 (n=30) algorithms (R packages) were used respectively for dimensionality reduction and clustering of data. UMAP plots were decorated with single individual relevant marker intensity to evaluate the distribution among phenogroups.
A single tSNE plot from homogeneous groups of animals was used to compare treatments vs control, then the samples were analyzed individually.
Clusters obtained were further explored via: i) individual cell immunoprofiling through dedicated hierarchical clustering heatmaps as published 37, ii) visualization of the spatial distribution on tissue. Clusters which satisfy both criteria were manually classified according to the expressions of key markers (Table 1).
Major cell types were created by merging clusters with a similar phenotype, after removing artifacts. Main types were organized as follows: 1) Alveolar Epithelial cells subdivided in AT1, AT2 or transitional AT1-AT2 subpopulations, 2) Bronchial and Goblet Epithelial cells, 3) Macrophage cells, 4) B cell, 5) T cells, 6) Vasculature and 7) Stromal cells, Neutrophils (8). For graphic spatial representation, all epithelial cells were grouped in a single entity (Alveolar Epithelial Cells).
Phenoclusters for which no coherent phenotype was identifiable were excluded from the analysis (see supplementary material for further information).
The transporter expression was analyzed by measuring the mean of the intensity of the signal (on 8 bit grayscale images, values 0-255) within each main type.