Patient population
Patients with systemic sclerosis (SSc) and healthy volunteers (HV) were obtained from the European multi-center cross-sectional study of the PRECISESADS IMI consortium which involved patients from seven systemic autoimmune diseases. The PRECISESADS cohort participants did not receive any compensation. SSc patients were aged between 50 and 68 years (median 60 years) and comprised 212 women and 37 men representing 85% and 15% of patients, respectively. Healthy volunteers (HV) were aged between 47 and 59 years (median 53 years) and comprised 307 women and 58 men, representing 84% and 16% of HV, respectively. Sex rather than gender was considered for women and men, based on self-reporting. All patients and HV gave written informed consent for the study that was approved by local ethical committees of the 19 participating institutions: Fondazione IRCCS Ca’ Granda Ospedale Maggiore, Policlinico di Milano, Comitato Etico Italy; Università degli studi di Milano, Policlinico di Milano, Comitato Etico Italy; Centre Hospitalier Universitaire de Brest, Comité de Protection des Personnes Ouest VI Brest, France; Université catholique de Louvain, Comité d’Ethique Hospitalo-Facultaire, Brussels, Belgium; Centro Hospitalar do Porto, Comissao de ética para a Saude, CES do CHP Porto, Portugal; Servicio Cantabro de Salud, Hospital Universitario Marqués de Valdecilla, Comite ético de investigacion clinical de Cantabria, Santander, Spain; Hospital Clinic I Provicia, Institut d’Investigacions Biomèdiques August Pi i Sunyer, Comité Ética de Investigación Clínica del Hospital Clínic de Barcelona, Spain; Katholieke Universiteit Leuven, Commissie Medische Ethiek UZ KU Leuven/Onderzoek, Belgium; Klinikum der Universitaet zu Koeln, Geschaftsstelle Ethikkommission, Cologne, Germany; Medizinische Hochschule Hannover, Ethikkommission, Hannover, Germany; Medical University Vienna, Ethik Kommission, Borschkegasse, Vienna, Austria; Servicio Andaluz de Salud, Hospital Universitario Reina Sofía Córdoba, Comité de Ética e la Investigación de Centro de Granada (CEI–Granada), Spain; Andalusian Public Health System Biobank, Granada, Spain; Servicio Andaluz de Salud, Complejo hospitalario Universitario de Granada (Hospital Universitario San Cecilio), Comité de Ética e la Investigación de Centro de Granada (CEI–Granada), Spain; Servicio Andaluz de Salud, Complejo hospitalario Universitario de Granada (Hospital Virgen de las Nieves), Comité de Ética e la Investigación de Centro de Granada (CEI–Granada), Spain; Servicio Andaluz de Salud, Hospital Regional Universitario de Málaga, Comité de Ética e la Investigación de Centro de Granada (CEI–Granada), Spain; Hospitaux Universitaires de Genève,. DEAS, Commission Cantonale d‘éthique de la recherche Hopitaux universitaires de Genève, Switzerland; Csongrad Megyei Kormanyhivatal, University of Szeged, Hungary; Charite, Ethikkommission, Berlin, Germany.
The cross-sectional study (registered as NCT02890121 in ClinicalTrials.gov) adhered to the standards set by the International Conference on Harmonization and Good Clinical Practice (ICH-GCP) and to the ethical principles that have their origin in the Declaration of Helsinki (2013). The protection of the confidentiality of records that could identify the included subjects is ensured by the EU Directive 2001/20/EC and the applicable national and international requirements relating to data protection in each participating country. For each individual (SSc patients and HV), blood samples as well as biological and clinical information were collected, including demographic data, symptoms, comorbidities, current medications, and biological information. For each patient, disease duration and organ involvement were reported. More technical details about the sample and data collection (including inclusion/exclusion criteria and immunophenotyping) have been published previously58.
Flow cytometry analyses were performed previously in eleven centers in the context of the PRECISESADS study, requiring multi-center harmonization of flow cytometers and procedures to integrate data. The protocol has been described previously28,58,59,60, and a gating strategy is shown in Supplementary Fig. 13. After exclusion of debris, dead cells, and doublets, neutrophils (CD15hiCD16hi) were gated from CD15+ polymorphonuclear leukocytes. Lymphocytes were identified as B cells (CD19+CD3–) and T cells (CD19–CD3+). Human plasma Gal-3 was quantified using the Simple Plex assay SPCKB-PS-000490 from Biotechne (Minneapolis, USA). After quality control on transcriptomic RNAseq data28, verification of the American College of Rheumatology (ACR)/European League Against Rheumatism (EULAR) classification criteria61 and match of HV and patients based on age and gender, the final studied cohort after clustering comprised 249 SSc patients and 365 HV. Their characteristics are reported in Table 1.
Molecular subgroup discovery and identification of a Gal-3 fingerprint associated with SSc clusters
Patient subgroup discovery was based on the pre-processed RNAseq data and clustering methodologies previously used for primary Sjögren syndrome patients from the PRECISESADS cohort28. Except when indicated, data analyses were carried out using either an assortment of R system software (http://www.R-project.org, V4.0.1) packages including those of Bioconductor v3.17, or original R code. R packages are indicated when appropriate. In order to analyze RNAseq data, first, bcl2fastq2 Conversion Software v2.20 was used to demultiplex sequencing data and convert BCL files. Quality control was obtained with FastQC tools v0.11.18 and adapters were removed with Cutadapt v1.18. Transcriptome alignment was done with STAR v2.5.2b on GENCODE v19 annotation (hg19) and read counts were obtained with RSEM v1.2.31. For normalizations and batch correction, read counts were normalized by the variance stabilizing transformation vst function from the DESeq2 v1.30.0R package. To reduce the effect of the RIN, a correction was applied using the ComBat function from the sva v3.38.0R package, after categorization of RIN values into 7 classes: [6.5–7.0], [7.0–7.5], [7.5–8.0], [8.0–8.5], [8.5–9.0], [9.0–9.5], [9.5–10].
Our objective was to obtain a robust classification scheme ensuring the identification of highly homogeneous SSc subgroups. The strategy used iterates unsupervised and supervised steps and is therefore designated as a “semi-supervised” approach. The training set comprised 263 SSc samples and the test set comprised 65 samples. To determine the number of clusters of patients (unsupervised step), a consensus clustering between three methods was performed: (i) Agglomerative Hierarchical Clustering (hclust function from stats v4.0.2R package) with Pearson correlation as a similarity measure and Ward’s linkage method, (ii) k-means clustering (k-means function from stats R package) with four groups and (iii) Gaussian mixture clustering (mclust function from mclust v5.4.6R package). The number of clusters was determined as the best consensus between three unsupervised clusterings (https://github.com/psBiostat/GAL3_PAPER.git)62. From 328 patients, 249 (219 in the training and 30 in the discovery test) were considered as consensus in the clustering (Supplementary Fig. 2). The top discriminating genes were defined with randomForest function from randomForest v4.6-14R package. Heatmaps were obtained with the ComplexHeatmap v2.6.2R package. Enrichment analysis was performed using transcriptional module repertoire and generated fingerprint representations with BloodGen3Module v0.99.36.
A non-redundant list of Gal-3 interacting proteins was compiled by bioinformatics by querying the GPS-prot database63, allowing the referencing of 210 Gal-3 interactants, and the Ingenuity Pathway Analysis database64, allowing the referencing of 276 Gal-3 interactants. Merging the 2 databases finally led to the identification of 307 proteins, listed in Supplementary Data 1 (column A), among which 59 were found absent in datasets of the PRECISESADS SSc cohort (Supplementary Data 1, column B). Among the 248 remaining genes, we searched for those that were significantly associated with PRECISESADS SSc clusters by applying the randomForest method (randomForest function from randomForest R package65). To obtain a robust selection, we generated 200 bootstrap replicates. Only genes frequently chosen by randomForest over bootstrap samples were selected, which improved the stability of the results. The frequency threshold was specified to be conservative (100%) so as not to exclude any cluster-associated gene. The most robust Gal-3 fingerprint associated with SSc patient clusters was defined by 69 Gal-3 interactants, 48 of which were upregulated and 21 downregulated as compared to C3. ‘Gal-3up’ and ‘Gal-3down’ scores were calculated as follows:
$$ {{{{{\rm{Score}}}}}}\; {{{{{\rm{Gal}}}}}}{\mbox{-}}{3}^{{{{{{\rm{up}}}}}}}{\mbox{}}={{{{{{\rm{Median}}}}}}}_{j=1},\ldots,\,{G}^{{{{{{\rm{up}}}}}}}\left({Y}_{{{{{{\rm{ij}}}}}}}{{{{{\rm{{-}}}}}}}{{{{{{\rm{median}}}}}}}_{k=1,\ldots,N_{{{{{\rm{ctrl}}}}}}}\left({{{{{{\rm{Ctrl}}}}}}}_{{{{{{\rm{kj}}}}}}}\right)\right)$$
(1)
$${{{{{\rm{Score}}}}}}\; {{{{{\rm{Gal}}}}}}{\mbox -}{3}^{{{{{{\rm{down}}}}}}}{\mbox{}}={{{{{{\rm{Median}}}}}}}_{j=1},\ldots,\,{G}^{{{{{{\rm{down}}}}}}}\left({Y}_{{{{{{\rm{ij}}}}}}}{{{{{\rm{{-}}}}}}}{{{{{{\rm{median}}}}}}}_{k=1,\ldots,N_{{{{{{\rm{ctrl}}}}}}}}\left({{{{{{\rm{Ctrl}}}}}}}_{{{{{{\rm{kj}}}}}}}\right)\right)$$
(2)
Where:
Yij represents the jth expression gene of ith patient,
Ctrlkj represents the jth expression gene of the kth HV,
Gup = 48 and Gdown = 21 represent the total number of the upregulated and downregulated genes, respectively,
NCtrl represents the total number of HV.
For the global differential gene expression analysis between HV and each patient subgroup, we applied a linear model (lmFit function from limma v3.46.0R package) on vst transformation gene expression dataset. The resulting p-values were adjusted for multiple hypothesis testing and filtered to retain differentially expressed genes with false discovery rate (FDR) adjusted p-values ≤ 0.05 and an absolute fold-change (|FC|) ≥ 1.3. Ingenuity Pathway Analysis (IPA) was applied to determine the most significantly dysregulated canonical pathways with Benjamini-Hochberg FDR adjusted p-values ≤ 0.05 and |FC| ≥ 1.3.
Criteria for selection and characterization of candidate therapeutic monoclonal antibodies against Gal-3
The criteria for generating Gal-3 blocking mAb were defined as follows: ability to interact with the carbohydrate recognition domain (CRD) of Gal-3, and full-length (FL) Gal-3; ability to bind primarily to human Gal-3 (hGal-3) and also to mouse Gal-3 (mGal-3) for in vivo testing; selectivity versus close galectin family members represented by Gal-1 and Gal-7; ability to displace natural Gal-3 ligand from the CRD.
ScFv selection, production, and binding characteristics
The naive scFv library (HuscI™, Mabqi proprietary library) was used for phage display panning. HuscI™ is a synthetic library based on a single cytosoluble hyper-stable human framework scFv termed 13R466,67,68 with side chain diversity incorporated at positions that contribute most to the antigen binding energy and least to intra-scFv contacts69. For the selection of scFv, recombinant His-tagged FL mGal-3, and hGal-3 were produced by Novalix (Illkirch, France) and used sequentially as baits to select human and mouse cross-reactive Gal-3 binders through successive rounds of enrichment by phage display with a human-mouse-human panning strategy. Recombinant proteins were immobilized through their His-tag on NiSO4/Tris-NTA-Biotin-coated microplates following the method described previously70. The naive HusclTM phage library was depleted on wells coated with streptavidin and NiSO4/Tris-NTA-biotin before panning. Following the last round of panning, bacteria colonies containing the selected scFvs were picked, and grown, and the scFvs were expressed as monoclonal fragments in deep-well plates. Finally, cultures were centrifuged and polymyxin B sulfate was used for periplasmic extraction of scFv from pellets.
ScFv binding and selectivity ELISA on human and mouse Gal-3, hGal-1, and hGal-7
Recombinant proteins hGal-3-FL-His Nter, hGal-3-CRD-His Nter, and mGal-3-FL-His Nter (Novalix, Illkirch, France) were immobilized on Nunc maxisorp plates via their His tag as described above. Recombinant proteins hGal-1 and hGal-7 (Novalix) were immobilized passively on Nunc maxisorp plates at 3 µg/mL in PBS, 100 µL/well, overnight at 4 °C, then blocked with 4% BSA. Plates were washed three times with PBS containing 0.1% Tween 20 (PBST), and periplasmic extracts were added at a 1:2 dilution in blocking buffer (PBS containing 4% of skimmed milk powder) for 1 h at room temperature. For detection of scFv via their myc-tag, a secondary anti-myc-HRP antibody (Santa Cruz sc-40) was added in blocking buffer at a 1:2000 dilution for 1 h at room temperature. Commercial antibodies directed against hGal-3 (Abcam 53082), hGal-1 (Abcam 25138), and hGal-7 (Abcam 108623) were used at 5 μg/mL as positive controls and detected with an anti-rabbit HRP secondary antibody (Cell Signaling 7074S) diluted 1:2000. Plates were washed 3 times using PBST. TMB chromogenic substrate for HRP detection was added to the wells and the reaction was stopped with H2SO4 1 M before optical density measurements.
ScFv reformatting to full-length IgG
Anti-Gal-3 scFvs were reformatted to full-length antibodies (IgG1 subclass) and further engineered with the double LALA mutation in the Fc region known to strongly reduce immune effector functions both in human and mouse71,72, and produced in CHO-K1 cells at Evitria (Zürich, Switzerland). Briefly, CHO-K1 cells were grown in eviGrow medium then transfected with eviFect and kept in eviMake2 production medium. Before protein A affinity purification using MabSelect™ SuRe™ (Cytiva), transfection supernatants were harvested by centrifugation followed by sterile filtration (0.2 μm). A dialysis was performed to formulate the IgG into DPBS leading to high-quality mAbs with endotoxin levels <1 EU/mg. For quality control, antibodies were analyzed by SDS-PAGE, size exclusion chromatography-multi angle light scattering (SEC-MALS), and mass spectrometry (MS). SDS-PAGE was performed using pre-cast NuPAGE 4–12% in MOPS buffer (ThermoFisher Scientific) at 150 V. Three micrograms of proteins were mixed with loading buffer with or without reducing agent and separated using standard conditions. The gels were then stained with Instant Blue (Expedeon, Cambridge, UK) according to the manufacturer’s recommendations. Homogeneity was determined by SEC-MALS using an ACQUITY UPLC Protein BEH SEC column, 200 Å (Waters, Saint-Quentin-en-Yvelines, France) at a flow rate of 0.4 mL/min in ammonium acetate, pH 5.4 running buffer. After centrifugation at 13,000 x g for 20 min, 5 µL of each purified mAb (25 µg) was injected into the column. Protein elution was monitored at 280 nm using a diode array detector (Agilent) and homogeneity was measured using a Heleos 8 + MALS detector (Wyatt, Toulouse, France) and Astra software v3.1.4 (Wyatt). For mass measurements, mAb was enzymatically deglycosylated with PNGase F (New England Biolabs, Evry, France) for 1 h at 37 °C and 1 µg of mAb was injected on a BioResolve RP 450 Å, 2.1*150 mm, 2.7 µm column (Waters) at 0.1 mL/min using a 20 to 90% acetonitrile gradient in 0.1% formic acid. Intact experimental masses were determined using a Xevo G2-XS Q-ToF (Waters) mass spectrometer coupled to a H-Class Bio UPLC system (Waters) followed by MaxEnt deconvolution. The thermal stability of mAbs was studied in comparison with a reference IgG1 approved in the EU, namely trastuzumab (EU/1/00/145/001). mAb samples stored at 4 °C or 40 °C for 2 weeks were analyzed by SEC using a BEH SEC column, 200 Å (Waters), with a of 25 mM ammonium acetate, pH 5.4 running buffer, and UV absorbance was recorded at 280 nm. Thermal stability was also assessed using nano differential scanning fluorimetry (nanoDSF) with mAb at 2 mg/mL in 50 mM histidine, 0.2 M glycine buffer, pH = 5.5. NanoDSF was averaged from duplicates using a Prometheus instrument (NanoTemper Technologies, München, Germany) from 20 to 95 °C with a 1 °C/min temperature increase and readouts at both 350 and 330 nm. Nucleotide sequences of D11 and E07 anti-Gal-3 mAbs have been filed in the patent application EP22305372 in March 2022.
IgG ELISA on Gal-3 and selectivity toward other galectin members
Single-dose IgG ELISA on hGal-3-FL, hGal-3-CRD, mGal-3-FL, hGal-1, and hGal-7 were performed using the procedure described for scFv with IgG at 20 µg/mL. Selectivities towards human Gal-2 (R&Dsystems 9874-GA), -4 (R&Dsystems 1227-GA), -8 (R&Dsystems 1305-GA), -10 (abcam ab107951) and -14 (Novus Biologicals NBP1-50082) were determined following a similar procedure using their respective commercial antibodies as positive controls. Gal-9 was excluded from the analysis due to low-quality control. For dose–response binding ELISA on hGal-3-FL and mGal-3-FL, IgGs were tested at concentrations ranging from 0.0009 to 500 nM. A secondary anti-hFab-HRP antibody (Sigma, A0293) was used for IgG detection using a classical TMB chromogenic procedure.
Asialofetuin dose–response inhibition of IgG1 mAbs on human Gal-3-CRD by ELISA
Recombinant hGal-3-CRD was coated as described above at 1 µg/mL and plates were blocked with 200 µL of TBS-BSA 4 % for 2 h, then washed three times with PBST, and IgGs were added to the wells in blocking buffer at concentrations ranging from 0.0006 to 780 nM for 1 h at room temperature. Plates were washed three times using PBST and asialofetuin (Sigma A1908) was added to the wells at 8 µg/mL in PBS containing BSA 4%, for 30 min at room temperature. The binding of asialofetuin to immobilized Gal-3 was detected using an anti-asialofetuin antibody (Abcam, ab35184) labeled using an HRP conjugation kit (Abcam ab102890) diluted 1:1500.
Binding kinetics between anti-Gal-3 mAbs and multiple Gal-3 species by surface plasmon resonance (SPR)
The interaction of anti-Gal-3 mAbs with recombinant Gal-3 from multiple species (Novalix) was monitored by SPR detection using a Biacore T-200 instrument (Biacore AB, Uppsala, Sweden). Binding studies were performed in freshly prepared, filtered, and degassed running buffer containing 10 mM HEPES, 150 mM NaCl, and 0.05% P20 (HBS-P, Cytiva Life Sciences) at 25 °C. Before the experiment, the CM5 sensor surface immobilized with anti-human Fc mAb was equilibrated with HBS-P by priming the instrument at least three times. The CM5 sensor surface was immobilized with an anti-human Fc mAb using a standard protocol. The entire sensor surface was then activated by injecting a 1:1 (v/v) mixture of 400 mM 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC) and 100 mM NHS at a flow rate of 10 μL/min for 7 min. Monoclonal anti-human Fc antibody (Millipore AP113), 25 µg/mL in 10 mM sodium acetate pH 5, was flowed for 6 min at 10 µL/min to reach 10000 resonance units (RU) on both flow cells followed by the injection of 1 M ethanolamine, pH 8.5 for 7 min at a flow rate of 10 μL/min. HBS-P was used as the running buffer during the entire immobilization procedure. Antibody samples at 5 µg/mL were injected for 30 s at 10 µL/min. In general, 400 to 680 resonance units of antibodies were captured on the chip. During the single-cycle kinetics (SCK) analysis assays, a range from 0.5 to 50 nM of human, mouse, rat, dog, or monkey Gal-3 were injected over the mAb capture surfaces for 210 s followed by a dissociation phase of 600 s at 30 µL/min, in duplicate conditions. At the end of each cycle, capture surfaces were regenerated by 2 injections of 10 mM glycine, pH 1.5, for 30 s at 10 µL/min, followed by an extra wash with 10 mM glycine, pH 1.5. Fresh mAb was captured at the beginning of each cycle. Each analyte injection cycle was preceded by one buffer injection cycle. The double reference subtracted SCK data was fitted with a 1:1 Langmuir using the SCK binding model with the Biacore T-200 evaluation software (version 3.2). Each interaction was investigated in duplicate (using two independent analyte dilution series).
Treatments and sampling in the mouse model of HOCl-induced SSc
Sixty 6-week-old Balb/c AnNRj mice, initially selected by Mac Dowell starting from a stock of inbred Albino mice (https://janvier-labs.com/en/fiche_produit/balb-cjrj_mouse/) purchased from Janvier(Le Genest-St-Isle, France) were included in the study. Only female mice were used, as described previously73. All mice were maintained in a specific pathogen-free (SPF) facility at the Pasteur Institute of Lille and the experimental/control groups were bred in separate cages. Mice were fed with a standard diet, given free access to water, and housed at a temperature of 22 ± 2 °C and a 35–70% humidity atmosphere, with 12 h light /12 h dark cycles. Animal experiments were performed in compliance with the European guidelines No. 68/609/EEC (approval number: #19603-2020061914271271 v6). The mouse study was approved by the local ethics committee: Comité d’éthique en expérimentation animale (CEEA 75), Nord Pas-de-Calais, France. The model was induced by daily intradermal injection of 300 µL of HOCl into the shaved backs of mice for 5 consecutive days per week, for a duration of 6 weeks. The control group (n = 12) comprised PBS-receiving mice. The HOCl groups (n = 12 each) comprised mice receiving 100 mM KH2PO4 + 0.08% of bleach to achieve a pH of 6.2. D11 and E07 mAbs were administered by subcutaneous injection at 20 mg/kg at day minus 1 and every 5 days starting from day 5 until euthanasia. Each mAb was solubilized in sterile, endotoxin-free PBS. TD139 (Syngene, India) was administered by intratracheal instillation twice a week starting from day 21 at a concentration of 0.5 mg/kg. TD139 was dissolved in 100% DMSO and diluted in endotoxin-free PBS to obtain a final concentration of 2% DMSO and 0.25 mg/mL of TD139. Body weights were measured at day minus 1 and every two or three days until the termination day. The evolution of skin thickening was monitored by using an external caliper, every 7 days until euthanasia. Longitudinal data were analyzed using a repeated measures ANOVA for comparison between HOCl and control group (model induction) and with a one-way ANOVA for comparison between all tested items and their HOCl control group, using Dunnett’s adjustment. The baseline (day minus one) was added as covariable. Statistical analyses were performed using SAS software, v9.4. Blood samples were collected from the venous sinus by retro-orbital puncture at day minus one, day 8, and 14 and processed to plasma by using lithium-heparin tubes for PK analyses and cytokine measurements. Mice were euthanized by cervical dislocation under deep CO2 anesthesia. Following confirmation of death, an incision was made in the neck and the muscle layers were separated by blunt dissection and the trachea was isolated. A small incision was made in the trachea and a cannula was inserted. The airway was lavaged with 0.6 mL of PBS. PBS was left in the airway for approximately 10 s while the chest was gently massaged before the removal of bronchoalveolar lavage (BAL) fluid. This procedure was repeated twice more. Terminal blood was collected, and 100 µL was transferred in RNA-protected Animal Tubes (Qiagen) for transcriptomic analyses. The remaining volume was processed into plasma by using lithium-heparin tubes for PK analyses and cytokine measurements.
Immunophenotyping in the mouse model of HOCl-induced SSc
For immunophenotyping, the BAL fluid from the three lavages was pooled and processed for flow cytometry. The BAL fluid was centrifuged at 400 x g for 7 min at 4 °C. The supernatant was harvested and stored at –80 °C. Cell pellets were incubated for 2 min in Red Blood Cell Lysing Buffer (Hybri-Max, Sigma-Aldrich) and washed in PBS-SVF 2%-EDTA 1 mM. Cells were resuspended in 130 µL of PBS-SVF 2%-EDTA 1 mM containing 10 µg/mL Fc Block (BD Biosciences 553142, clone 2.4G2) and incubated at 4 ˚C for 5 min. For the Numeration of viable CD45+ cells in BAL fluid, cell suspensions (50 µL) were incubated with anti-CD45-APC antibody diluted 1:120 (Biolegend 103111, clone 30-F11) or the corresponding isotype control at the same dilution (Biolegend 400611, clone RTK4530) for 20 min at 4 °C, protected from light, and with propidium iodide (50 ng, Sigma-Aldrich) for 5 min. Flow-Count Fluorospheres (Beckman Coulter) were added and data were acquired on a 4-laser cytometer (CytoFLEX, Beckman Coulter) and analyzed with dedicated Kaluza software v2.1 (Beckman Coulter). For the phenotypic analysis of the cellular content in BAL fluid, cell suspensions (50 µL) were incubated with the following antibodies: anti-CD45-FITC diluted 1:200 (Biolegend 103107, clone 30-F11), anti-CD11b-APC-Cy7 diluted 1:200 (BD Biosciences 561039, clone M1/70), anti-CD3-PC7 diluted 1:200 (Biolegend 100219, clone 17A2), anti-CD4-PE diluted 1:100 (BD Biosciences 553653, clone H129.19), anti-CD8-Pacific Blue diluted 1:100 (BD Biosciences 558106, clone 53-6.7), anti-CD19-BV510 diluted 1:100 (BD Biosciences 562956, clone 1D3), anti-CD335-APC diluted 1:50 (Biolegend 137607, clone 29A1.4), or their corresponding isotype controls at similar dilutions (FITC, PC7 and APC from Biolegend, #400634 clone RTK4530, #400617 clone RTK4530, #400511 clone RTK2758, respectively; APC-Cy7, PE, Pacific Blue and BV510 from BD Biosciences, #552773 clone A95-1, #553930 clone R35-95, #558109 clone R35-95, #562952 clone R35-95, respectively), for 20 min at 4 °C protected from light. 7-AAD (Biolegend) was added and cells were incubated for 5 min at 4 °C protected from light. Cells were washed in PBS-SVF 2%-EDTA 1 mM and fixed in 1% paraformaldehyde (PFA)/PBS solution. Data were then acquired on a 4-laser cytometer (Cytoflex, Beckman Coulter) and analyzed with Kaluza software v2.1. The gating strategy was as follows (Supplementary Fig. 14): Immune cells = CD45+ cells; B cells = CD19+ cells; NK cells = CD335+ cells; CD4+T cells = CD3+CD4+ cells; CD8+T cells = CD3+CD8+ cells. Among CD19- CD3- CD335- cells: alveolar macrophages = autofluorescencehigh CD11b-/low and other myeloid cells (monocytes, dendritic cells, neutrophils, and eosinophils) = autofluorescencelow/int CD11b+/high. Statistical analyses for immune cells in BAL were determined with SAS software v9.4 using a one-way ANOVA for comparison between the HOCl-treated and control group (model induction) and with a one-way ANOVA for comparison between all tested items and their HOCl control group, using Dunnett’s adjustment on log-transformed data.
Lung and skin histology
Following completion of the BAL procedure, the thorax of mice was opened, lungs were removed, 0.5 cm skin samples were collected and all tissue samples were fixed in 4% paraformaldehyde (PFA)/PBS solution according to the standard method for paraffin embedding for histological assessment. Three 4-µm sections of skin and lung were cut from the blocks and, after rehydration, stained with Picrosirius Red to determine the level of collagen deposition. Briefly, tissue sections were stained with 0.1% Direct Red stain (Sigma-Aldrich)/0.5% Picric Acid (Sigma-Aldrich) for 60 min after deparaffinization and rehydration. Subsequently, each slide was examined and analyzed using the web-based ImageJ software v1.53t. Quantitative analysis of connective tissue deposition was performed using a threshold detection method for the grayscale image after a color deconvolution to separate the stains and the area occupied by red-stained collagen74. The levels of skin collagen content were analyzed using a one-way ANOVA for comparison between the HOCl-treated and control group (model induction) and with a one-way ANOVA for comparison between all tested items and their HOCl control group, using Dunnett’s adjustment. For the levels of lung collagen content, the same analyses were performed on log-transformed data, adding the animal as a random factor (for replicate measurements of transversal and longitudinal slices).
The macrophage surface marker F4/80 was evaluated on 4-µm fixed lung sections by immunofluorescence assay. After permeabilization in 0.1% Triton-X 100, lung sections were incubated overnight at 4 °C with a monoclonal anti-F4/80 primary antibody (ab111101, Abcam) diluted 1:100. Then, the samples were incubated with a specific conjugated anti-mouse secondary antibody conjugated to AlexaFluor 488 (A32723, ThermoFisher Scientific) diluted 1:2000 for 1 h at room temperature. Nuclei were stained with Hoechst 33342 dye (62249, ThermoFisher Scientific). Digital images were processed with the Zeiss LSM Browser. The number of F4/80-expressing cells was quantified using QuPath software (v.0.4.1) as a percentage of the number of cells. Data were expressed as the fold-change compared to the control group. Statistical analyses were performed using a one-way ANOVA for comparison between the HOCl-treated and control group (model induction) and with a one-way ANOVA for comparison between all tested items and their HOCl control group, using Dunnett’s adjustment. All statistical analyses were performed using SAS software v9.4.
PK and ADA measurements
Plasma samples were assayed for mAb concentrations using a generic method implemented on a Gyrolab xPlore automated immunoassay system75. Briefly, in-house biotinylated mouse anti-human IgG CH2 capture antibody (ThermoFisher, MA5-16929) was immobilized at 100 µg/mL on the surface of streptavidin-coated beads in Gyrolab Bioaffy CD200 (P0004180, Gyros Protein Technologies), and diluted plasma samples were loaded. Then, in-house 647-Alexafluor conjugated mouse anti-human IgG antibody (BD bioscience 555784) was used as a detection reagent at 10 nM. Sample concentrations were calculated using the Gyrolab software (Gyros Protein Technologies). Anti-drug antibodies (ADA) were measured with a generic method implemented on a Gyrolab xPlore automated immunoassay system. Mouse plasmas were coincubated with excess concentration (10 µg/mL) of D11 or E07 mAbs directly in the mixing chamber of the Gyrolab mixing CD96 (P0020455, Gyros Protein Technologies). Then, the ADA-mAb complexes were captured by biotinylated Goat anti-human IgG (Southern Biotech 2014-08) coated at 100 µg/mL on the streptavidin beads within the compact disc. Finally, 10 nM of AlexaFluor 647 goat anti-mouse IgG (H + L) cross-adsorbed secondary antibody (ThermoFisher A-21235) was used as a detection agent. Values were expressed as fluorescence units with a significant cut-off threshold in each experimental run determined as the mean of 3 standard deviations calculated from the mean of signals obtained from 15 control untreated mice. Statistical analyses were performed using SAS software v9.4.
Mouse Gal-3 and cytokine assays
Mouse plasma Gal-3 was quantified using the ELISA kit DY1197 from BioTechne. Ten soluble cytokines (namely IFN-γ, IL-10, IL-1β, IL-2, IL-4, IL-5, IL-6, KC/GRO, IL-12p70, and TNF-α) were also assayed in mice plasma using the V-plex® Plus mouse proinflammatory panel 1 K15048G from Meso Scale (Discovery, Rockville, USA). Statistical analyses were performed using SAS software v9.4 on log-transformed data using a mixed ANCOVA model (baseline as covariate and animal as random) for comparison between the HOCl-treated and control groups (model induction) and using a mixed ANCOVA model for comparison between all tested items and their HOCl control group, using Dunnett’s adjustment.
Mouse RNAseq studies and canonical pathways
Total RNA was extracted from whole blood collected in RNA protect tubes (Qiagen) and proceeded for extraction using the RNeasy Protect Animal blood kit with the optional on-column DNAse I digestion. After extraction, total RNA samples were quantified on Nanodrop 2000 (ThermoFisher Scientific) and analyzed on Bioanalyzer (Agilent) using RNA 6000 Nano chips to determine RNA quality (RIN). Total RNA matching the QC criteria (RIN > 7) was included in the study. At least 250 ng of purified RNA samples were used as input for the TruSeq Strand-specific RNA-seq (poly A selection and globin depletion) according to the manufacturer’s instructions (Illumina). Following the preparation of libraries, samples matching the requested criteria were sequenced on a NovaSeq 6000 sequencer (Illumina). Sequencing parameters were adapted for 2 × 150 bp configuration with at least 20 M pair-end reads per sample. Raw reads were processed using the in-house RNAExp Expression pipeline 2.1 and the quality was assessed using FastQC tools v.11.18. Then, reads were trimmed for adaptors using Cutadapt and trimmed reads were aligned to the mouse mm39 reference genome (https://www.ncbi.nlm.nih.gov/assembly/GCF_000001635.27/) using the STAR aligner. Aligned data were evaluated for quality using several quality metrics (e.g., mapping rate, %rRNA, %mRNA) using Samtools and Picard tools. Then, aligned reads per gene were quantified using FeatureCount. Read counts were normalized by the variance stabilizing transformation vst function from the DESeq2 R package v1.30.0. 55,416 genes were detected in the data. Genes not coding for a protein were filtered (29,947 genes removed). Those with 0 count over all the samples or having an expression level below 1 in more than 95% were filtered. The final RNAseq dataset comprised 10,508 genes.
Partial Least-Squares-discriminant analysis (PLS-DA) performs a dimension reduction by preserving the covariance between the expression matrix and the outcome variable. It was applied to represent the samples while maximizing their separation with regard to the outcome. In order to identify genes differentially expressed between groups, a linear model (lmFit function from limma v3.46.0R package) based on the vst transformation gene expression dataset was applied. The resulting p-values were adjusted for multiple hypothesis testing and filtered to retain differentially expressed genes with a FDR adjusted p-value ≤ 0.05 and a |FC| value ≥ 1.3. Ingenuity pathway analysis was applied to determine the most significantly dysregulated canonical pathways with Benjamini-Hochberg FDR and FC values as above.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.