The SV40-immortalized human bronchial epithelial cell line BEAS-2B (ATCC) was cultured in DMEM/Ham’s F12 (EuroClone) supplemented with 5% heat-inactivated FBS (EuroClone), 2 mM L-glutamine (Lonza), penicillin (100 U/ml)-streptomycin (100 mg/ml) (Lonza) and 0.2% fungizone (EuroClone) (37). Human small airway epithelial cells (HSAECs, PCS-301-010, ATCC) were cultured as submerged monolayers in Airway Epithelial Cell Basal Medium (PCS-300-030, ATCC) supplemented with Bronchial Epithelial Growth Kit (PCS-300-040, ATCC). Both cell lines were incubated at 37°C, 5% CO₂. For challenge, when reaching 70% confluency cells were kept in medium only or stimulated using cytomix (10 nM each rHuIL-1β, TNFα, IFN-γ, GoldBio) for 48 h. For exogenous AUF-1 silencing, BEAS-2B cells were transfected when reaching 50%–60% confluency using the non-liposomal cationic vehicle FuGENE HD (Promega) according to the manufacturer’s instructions, using 100 nM AUF-1 siRNA (5′-AAGAUCCUAUCACAGGGCGATdTdT-3′) (47) or a scrambled control siRNA (5′-GAGUCAACCUUAUGAUACUdTdT-3′). After 48 h, cells were exposed to cytomix or medium for additional 48 h prior to cell harvesting. For mRNA stability assays, resting and cytomix-treated cells were either harvested at 48 h (time 0), for analysis of steady state levels or after 1, 2 and 4 h of culture with the transcriptional inhibitor actinomycin D (3 µg/ml ActD, Sigma) (51). For proteasome inhibition experiments, cells were incubated for 2h with 10 μM of MG-132 compound (Sigma), then medium was replaced and cells were treated with cytomix or medium for additional 48 h prior to cell harvesting. In all experiments, cells were harvested using trypsin/EDTA (Lonza), counted and viability was verified by Trypan Blue exclusion (EuroClone). Cell viability was ≥ 90% at harvest in all conditions.

RIP is an antibody (Ab)-based technique developed to study the interaction between a RBP and its endogenous targets (52, 53), performed according to established protocols for BEAS-2B (37, 38). Cytosolic fractions were extracted after lysing BEAS-2B cells (n=3, 10⁸ cells/condition) with polysome lysis buffer (10 mM HEPES pH 7.0, 100 mM KCl, 5 mM MgCl₂, 0.5% NP40, 1 mM DTT, 100 U/ml RNase out, 400 μM Vanadyl-Ribonucleoside Complex, 1x Protease Inhibitors) (52). An aliquot of cytosolic extract (10%) was taken as Input. For IP with, 2 mg of cytosolic extract were incubated (4°C, overnight) with 4 μg of anti-AUF-1 Ab (HPA004911, Atlas Antibodies) and for control IP, IgG isotype (02-6102, Thermo Fisher Scientific). Then, 100 μl of pre-blocked magnetic beads (Dynabeads, Thermofisher) were added and the incubation was continued (4°C, 4 h). Total RNA pools bound to AUF-1/control Ab were extracted by adding TriFast reagent (EuroClone) directly to the washed beads, following the manufacturer’s instructions. The size distribution of each RNA sample was assessed by running a 1 μl aliquot on an Agilent High Sensitivity RNA chip using an Agilent Technologies 2100 Bioanalyzer (Agilent Technologies). The RNA concentration of each sample was determined using a Quant-IT RNA Assay Kit-High Sensitivity and a Qubit Fluorometer (Life Technologies).Total RNA was used to prepare sequencing libraries (54). Briefly, 1 μg of RNA Input and 300 ng of AUF-1- and Ctrl Ab-IP RNA were used as the starting material for sequencing library preparation from three independent experiments. Immunoprecipitated RNAs were fragmented and converted to cDNA after adaptor ligations in preparation for sequencing. Principal Component Analysis (PCA) indicated that one of the biological replicates was discordant respect to the other two samples; therefore, two of the three original experiments were considered for further analysis. After normalization, a total of 12,727 transcripts were expressed in the cell line Input samples. Indexed triplicate libraries were prepared with a TruSeq Stranded Total RNA (Illumina Inc.). The quality of the libraries was evaluated with 2100 Bioanalyzer (Agilent) and Qubit dsDNA HS Assay Kits (Thermo Fisher Scientific). Libraries were sequenced at 3 pM/lane (paired-end, 2 × 75 cycles) on a NextSeq 500 (Illumina Inc.). Bioinformatic analysis was performed. as described (54). An average of 27,466,311, 36,765,708 and 25,288,124 reads were obtained from the Input, AUF-1 IP and IgG IP libraries, respectively. Despite differences in total read numbers due to the low amount of Input cDNA, the IPs consistently yielded many more mappable reads than the controls. The quality of the sequenced reads was assessed by evaluating the quality score, presence of k-mers and balance of the GC percentage, using FastQC software (55). Cutadapt software was used to remove the adapter sequences (56) using default parameters. The human transcriptome and genome (assembly hg38) was used as a reference for the alignment, which was performed using STAR version 2.7 (default parameters) (57).

Feature-count was used with default parameter to compute gene-level read counts (58). Only the genes whose read count was ≥ 10 in at least one sample were considered for the further analysis. The R bioconductor package DESeq2 was used to test the differential expression of genes from RIP-Seq data compared to controls (59). RNAs showing Enrichment Factors (EFs) ≥1.5 and False Discovery Rates (FDRs) ≤0.05 computed according to Benjamini–Hochberg were considered for further analysis. Transcript per million (TPM) was computed using RSEM (60). Scatter plot and box plot were elaborated with R (v3.6.2) (61). Raw Rip-sequencing data are deposited in the EBIArrayExpress database with accession number E-MTAB-12583.

The p45^AUF-1 sequence from NCBI was used for analysis with the CatRAPID algorithm (62). The list of coding and non-coding targets of AUF-1 protein was filtered considering a Discriminative Power (DP) ≥0.75. Enrichment ratios for every transcript in each RIP-Seq experiment were log transformed. Graphic visualization was elaborated with R version 3.6.2. Prediction of binding motifs for AUF-1 were identified on 3’UTR sequences of experimental targets with Sequence & Structure Motif enrichment Analysis for Ranked RNA data generated from in vivo binding experiments (SMARTIV) (63, 64), with standard setting and Multiple Expectation maximizations for Motif Elicitation (MEME version 5.4.1) (65), a position weight matrix-based tool for motif identification, using the following parameters: number of repetitions, any; minimum width for each motif: 5; maximum width for each motif: 35; and maximum number of motifs to be found: 20. Motifs with E-value ≤ 0.05 according to minimum hypergeometric statistical approach (mmHG) were considered significant.

Biotinylated 3’UTRs were generated by PCR of BEAS-2B RNA ( Table S2 ). Long 3’UTRs were fragmented in adjacent sequences to allow correct in vitro transcription and biotinylation. PCR products were purified from agarose gels and used as templates for biotinylated RNAs synthesis using MAXIscript™ T7 Kits (AM1312, Invitrogen) and Biotin-11-cytidine-5’-triphosphate (ENZ-42818, Enzolife). Unstimulated BEAS-2B cells were lysed with polysomal extraction buffer (100 mM KCl, 5 mM MgCl₂, 10 mM Hepes pH 7.0, 0.5% NP-40, 1X protease inhibitor) to obtain cytoplasmic fractions. Cytosolic lysates (500 µg) were incubated with 1 µg of biotinylated transcripts (30 min) and, then, ribonucleoprotein complexes were isolated with streptavidin-conjugated Dynabeads (11205D, Invitrogen). The presence of AUF-1 in the pulldown material was verified by immunoblot analysis (38, 66).

Proteins were separated, quantified, and subjected to Western blot analysis as described (48). For total protein extraction, cells were directly lysed in buffer containing 50 mM Tris/HCl at pH 7.5, 150 mM NaCl, 2 mM EDTA, 2 mM EGTA, 25 mM NaF, 25 mM β-glycerolphosphate, 0.1 mM Na3VO4, 0.1 mM PMSF, 0.2% Triton X-100, 0.3% NP40, and a cocktail of protease inhibitors (100 X, EuroClone). After incubation (4°C, 30 min) the lysates were centrifuged (15,700 xg, 4°C, 15 min). A total of 15 µg of protein per well was separated using 4–15% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes. Membranes were then blocked with 5% milk-TBS-Tween buffer (TBS plus 0.1% Tween-20, room temperature, 1h) and incubated with primary antibodies at 4°C overnight, washed with TBS-Tween buffer three times and incubated with corresponding horseradish peroxidase-conjugated secondary antibodies (room temperature, 45 min). Signals were detected using the “Pierce ECL Western Blotting Substrate” method (Thermo Fisher Scientific) and analysed using ImageLab software with the Chemidoc image acquisition and analysis tool (BioRad).

The following primary antibodies were used: anti-AUF-1 (HPA004911, Atlas), anti- Human antigen R (HuR) (sc-5261, Santa Cruz), anti- Tristetraprolin (TTP) (ab83579, Abcam), anti-p53 (sc-126, Santa Cruz), from Cell Signaling Technology, anti-phospho-Rb (9308), anti-p21 (2947), anti-phospho-p53 (9286), anti- β-actin (3700), anti-β-tubulin (9F3), anti-CD9 (10626D); anti-CD63 (10628D, Invitrogen, used in non-reducing conditions).

Total RNA was extracted using TriFast reagent (EuroClone) and reverse transcription was achieved using LunaScript^® RT SuperMix Kits (New England Biolabs) following the manufacturer’s protocol. Template complementary DNA (cDNA) was subjected to qRT-PCR using FluoCycle II SYBR Master Mix (Euroclone) according to the manufacturer’s protocol (48). Primers were published or designed with Primer-BLAST software https://www.ncbi.nlm.nih.gov/tools/primer-blast/ ( Table S2 ). Reactions were run in duplicate on a LightCycler 480 II (Roche), using the following setup: denaturation at 95°C for 5 minutes; amplification at 95°C for 15 seconds, and 60°C for 60 second (45 cycles). Target expression was normalized to GAPDH by the cycle threshold (Ct) method and expressed using the 2^−ΔΔCt calculation as fold over control.

Cells were grown in submerged cultures in presence of: media only; cytomix, for 48h: etoposide (6 µM) for 24 h, used as positive control for senescence. The culture media was replaced with no further stimulation (as post-challenge time 0) and cells were cultured at 37°C for an additional 5 days (as post-challenge time 5). Cellular senescence was then detected using flow cytometric analysis of β-D-galactopyranoside, a fluorogenic β-galactosidase substrate, using Cell Meter™ Cellular Senescence Activity Assay Kits (23005, AAT Bioquest) according to the manufacturer’s protocol (67). Fluorescence was detected using a FACSVerse™ flow cytometer (BD Biosciences), in the FITC channel, and the data were analyzed using BD FACSuite™ software.

BEAS-2B and HSAEC supernatants were screened for inflammatory cytokines associated with SASP (IL-1β, MCP-1, IL-6, IL-8) using bead-based immunoassays LEGENDplex™ (740809, Biolegend), following the manufacturer’s protocol. Briefly, cell supernatants were incubated for 2 h with two sets of specific anti-cytokine antibodies-conjugated beads, differentiated by size and internal fluorescence intensities, followed by a 1 h incubation with a detection antibody. After 30 min incubation of the mix with streptavidin-conjugated phycoerythrin, the fluorescent signal intensity, which is proportional to the amount of bound analytes, was detected using a FACSVerse™ flow cytometer (BD Biosciences) and data were analyzed with the LEGENDplex Data Analysis Software.

BEAS-2B and HSAEC cells were seeded with equivalent cell numbers between conditions (40 x10⁶ cells/condition) and cultured with corresponding EV-depleted culture mediums, obtained by overnight centrifugation (100,000 xg). Cell-derived EVs were isolated from culture media of unstimulated and cytomix-stimulated cells by differential centrifugation following an established protocol (68, 69). Briefly, cell supernatants were centrifuged at 300 xg for 10 min to pellet cells, then at 2,000 xg for 10 min to pellet dead cells and 10,000 xg for 30 min to remove cell debris. Finally, supernatants were ultra-centrifuged (100,000 xg, 70 min) using a Beckman Coulter Optima XE-100 Ultracentrifuge with SW 32.1 swinging-rotor. EV pellets were resuspended in 50 µl of PBS.

DLS analyses the velocity distribution of particle movement caused by Brownian motion by measuring fluctuations of scattered light intensity. Then, the particle size is calculated size via the Stokes-Einstein equation (70). 10 µL of EVs were diluted in 990 µL of water and size was measured using a Nano ZS Malvern Zeta Sizer (model 1000HSa, UK, 25°C), equipped with a He-Ne laser of 633 nm and detector angle of 173°C. Analyses were performed in three independent technical replicates for each sample. EV size was expressed as mean ± Standard deviation (SD).

The NTA determines the concentration and size of particles in EV samples through identification and tracking of individual nanoparticle movements under Brownian motion. In detail, 10 µL of EVs suspension was diluted in 1000 µL of water and then injected into the Nanosight NS300 (Malven) for measurement. EVs size and concentration were expressed as mean ± SEM of nm and number of particles/ml, respectively.

One drop of sample solution (~25µl) was placed on 400 mesh holey film grids. After staining with 2% uranyl acetate (2 min) samples were observed with a Tecnai G2 (FEI) transmission electron microscope operating at 100 kV. Images were captured with a Veleta (Olympus Soft Imaging System) digital camera. For cell monolayer samples, seeded cells were washed in 1x HBSS and fixed in 2.5% glutaraldehyde (Sigma-Aldrich) in 0.1M Hepes buffer (4 °C, 1 h, pH 7.4). After three water washes, samples were dehydrated in a graded ethanol series and embedded in epoxy resin (Sigma-Aldrich). Ultrathin sections (60-70 nm) were obtained with an Ultrotome V (LKB) ultra-microtome, counterstained with uranyl acetate and lead citrate and viewed with a Tecnai G2 (FEI) transmission electron microscope. Images were captured with a Veleta (Olympus Soft Imaging System) digital camera. The mean ± SD EVs size was calculated using ImageJ software https://imagej.nih.gov/ on 100 particles chosen randomly in pictures from control and cytomix-treated samples.

A drop of sample solution (~25 µl) was placed on a 400 mesh holey film grid for 2-3 min. Subsequently they were incubated with blocking solution (0.5% bovine serum albumin (BSA), in PBS, room temperature, 30 min). Immediately the grids were then incubated (30 min, RT) with a primary antibody anti-AUF-1 (HPA004911, Atlas) diluted 1:40 in blocking solution and then washed three times with PBS (5 min each, RT). The grids were then incubated with an IgG Gold II secondary anti-rabbit antibody coupled to gold particles (5nm, Sigma Aldrich G3779, RT, 30 min). After washing, in PBS (3X) and water (2X), grids were counterstained with uranyl acetate and lead citrate and viewed with a Tecnai G² (FEI) transmission electron microscope. Images were captured with a Veleta (Olympus Soft Imaging System) digital camera.

GO analysis was performed with Ingenuity Pathway Analysis (IPA) software (71). Heatmaps and Pearson correlation matrices for correlated expression changes were generated using tMEV. GOPlot was used to visualize the Circos plot (72, 73).

Bronchial rings and peripheral lung samples were obtained from subjects recruited from the Respiratory Unit of the University Hospital of Messina, Italy, among patients undergoing lung resection for peripheral lung carcinoma ( Table S3 ). Smokers with mild-to-moderate stable COPD (n=7) were compared with age- and smoke history-matched smokers with normal lung function (NLF, n=5). Diagnosis of COPD was defined according to international guidelines as the presence of post-bronchodilator forced expiratory volume in 1 s (FEV₁)/forced vital capacity (FVC) ratio <70% or the presence of cough and sputum production for at least 3 months in each of two consecutive years (24, 74). All patients were in a stable condition at the time of the surgery and had not suffered acute exacerbations or upper respiratory tract infections in the preceding two months. None had received glucocorticoids or antibiotics within the month preceding surgery, or inhaled bronchodilators within the previous 48 h. Patients had no history of asthma or other allergic diseases. All former smokers had stopped smoking for >1 year. Each patient underwent medical history collection, physical examination, chest radiography, electrocardiogram, routine blood tests, and pulmonary function tests during the week prior to surgery. Pulmonary function tests (Biomedin Spirometer, Padova, Italy) were performed as described (75) according to published guidelines. The study was approved by the local Ethics Committees of the University Hospitals of Messina and participating patients and control subjects signed the approved informed consent forms.

For gene expression analysis libraries were prepared with the Lexogen QuantSeq 3′ mRNA-Seq Library Prep Kit (FWD) for Illumina (cat. no. 015.96), as per the manufacturer’s instructions. The modified protocol for FFPE samples was used and RNA input was 250ng. qPCR was performed to find the optimal cycle number for endpoint PCRs, using the PCR Add-on Kit for Illumina (cat. no. 020.96) to quantify cDNA before final library amplification. Library pooling was performed by BGI and NGS was run on their DNBSEQ platform (BGI Genomics, HK) with PE100 reads. Read count was performed on the BlueBee platform using standard settings for the Quantseq 3’ kit (www.lexogen.bluebee.com). Differential expression analysis was performed using DESeq2 package from the Bioconductor (https://bioconductor.org/packages/release/bioc/html/DESeq2.html). Heatmaps were generated using tMEV tools v4_9_0.45. (76, 77).

This statistical method evaluates variations in underlying mechanisms between groups (78) and was used to compute the Enrichment Score (ES) of the AUF-1 RIP-seq gene set in all subjects included in the database GSE5058 (79, 80). The analysis was performed according to described parameters (81). Gene signatures were considered significantly differentially expressed with differences in Ess (dES) ≥ 0.2 between the groups and p-value <0.05.

AUF1 target expression was evaluated in scRNAseq datasets generated from primary broncho-epithelial cells from stable COPD patients and healthy control subjects differentiated at the air-liquid interface (ALI) (82).

For RIP-seq data statistical analysis, FDR ≤ 0.05 computed according to Benjamini–Hochberg were considered for further analysis. Data from immunoblot densitometry and qRT-PCR were analyzed using Student’s paired t-test. For cellular senescence activity assays, ANOVA test with FDR post hoc multiple comparison analysis was performed. Statistical analysis was performed using GraphPad Prism 5 (GraphPad Software Inc.). A probability p ≤ 0.05 was considered significant.

Cytosolic extracts of unstimulated BEAS-2B cells (n=3) were isolated and subjected to RIP-Seq analysis (see Methods). Immunoblot analysis of the protein fraction ( Figure 1A ) revealed a high level of enrichment in IP AUF-1 compared to IP IgG, No-Ab IP and unbound fraction controls. To visualize the enrichment data, each sequenced transcript in the IP sample vs the Input samples were plotted. A scatter plot was constructed using the log-transformed and normalized read numbers ( Figure 1B ). Enrichment analysis was set with EF ratio in AUF-1 IP vs Input and IgG IP vs Input at ≥ 1.5 and FDR ≤ 0.05. With these cutoff values, 1,078 transcripts were significantly immunoprecipitated in AUF-1 IP vs Input and 1,149 transcripts in IgG IP vs Input samples. Subsequently, the two datasets were crossed and overlapping targets with IgG IP vs Input transcripts were excluded. As a result, 494 RIP-Seq-identified AUF-1 targets were considered for further analysis ( Figure 1C ), a number in line with previous RIP-based transcriptomic studies in these cells (38). Table 1 lists the top 20 RIP-Seq-identified AUF-1 targeted transcripts ranked by enrichment value (full target list in Table S4 ). For these selected targets, the average of reads enrichment was significantly higher in AUF-1 IP compared to the Input and to IgG IP (median normalized reads: 1,024, 583 and 762 in AUF-1 IP, Input and IgG IP samples, respectively, Figure 1D ). Of relevance, the AUF-1 IP-target pool includes transcripts previously identified as associated to AUF-1, such as DICER (109) and ZFP36L1 (110) and almost all targets (n= 490) were listed among the over 2,000 mRNAs identified as bearing AUF-1 binding sites in a model of AUF-1-overexpression in human embryonic kidney cells by the more stringent PAR-CLIP analysis (47).

The interaction of AUF-1 with its target mRNAs is mediated predominantly by motifs located in the 3′UTR of the transcripts (39, 47, 111, 112). Thus, we focused our analysis of enriched elements to the 3’UTR of the AUF-1-bound targets. We screened the 494 epithelial AUF-1 targets for the occurrence of 3’UTR motifs using the SMARTIV tool (63, 64). Figure 2A shows 4 core motifs with k-mer length of 5 and 6 nucleotides. Within the experimental dataset 12 enriched gapped k-mer motifs were further identified, mostly comprising Guanine-Cytosine (GC) nucleotides, which had the highest frequency of hits over the entire SMARTIV database ( Figure S1A ). We also searched for an extended 30-mer core motif using a MEME Suite tool. This analysis confirmed that the majority of experimental AUF-1 epithelial targets shared a GC-rich motif. We selected for validation the motif displaying the most significant E-value (see Methods) ( Figure 2B for validated motif; remaining are shown in Figure S1B ).

To validate AUF-1 association with targets identified by RIP-Seq analysis, biotin pull-down experiments were set for selected transcripts, chosen to represent the spectrum of EF ( Figure 2C ). Cytoplasmic lysates from unstimulated BEAS-2B cells (n=2) were incubated with biotin-labelled synthetic RNAs corresponding to full-length or segments of targets’ 3’UTRs, in case the 3’UTR was too long for synthesis of a single biotin-labelled molecule ( Table 2 ). IL-6 mRNA was included since it is an important mediator of SASP also known to be an ARE-bearing gene whose mRNA decay is accelerated by AUF-1 binding (44, 113, 114).

For the newly identified GC-rich motif selected for validation, a synthesized biotinylated sequence containing the nucleotides with highest frequency was utilized. The 3’UTR of Cyclin D1, a known target of AUF-1 (115) and a beads-only sample and a non-AUF-1 target sequence (PD-L1 coding sequence) were included as positive and negative controls, respectively. Immunoblot analysis ( Figure 2C ) revealed the presence of AUF-1 in the starting lysate (Input) and in the pulldown fractions obtained with biotinylated 3’UTR sequences of all experimental transcripts. In particular, AUF-1 was detected in the pulldown fraction obtained with full-length 3’UTRs (EGR1, TGF-β1, isoform 1 of MUC1 and IL-6) and specific segments for other transcripts: segment A and, more abundantly, with segment D of HDAC2 3’UTR, segment A of FoxP4-3’UTR, all segment of DDX17-3’UTR, segment B of GLIS2-3’UTR. Importantly, AUF-1 was detected in association with the sequence modelled on the motif shown in Figure 2B . These results support the data obtained from RIP-Seq analysis, confirming the association of AUF-1 to the 3’UTR of the selected transcripts, pointing in some cases to specific regions of 3’UTRs.

We previously documented that cytomix- and CSE-induced loss of AUF-1 in BEAS-2B cells occurred along with changes in expression levels of many established AUF-1-regulated cytokines and chemokines. This modulation was replicated with greater AUF-1 loss induced by siRNA (48). We hypothesized that loss of AUF-1 occurring upon cytomix stimulation might be reflected in changes in mRNA stability of its newly identified epithelial targets. We therefore evaluated steady-state levels and mRNA decay rates of selected targets displaying different EF from the RIP-Seq analysis (GLIS2, MUC1, FOXP4, CRTC1, TGFβ1, EGR1, DDX17, DHX36, HDAC2, IGF1R). Since diseased phenotypes were already observable in heterozygous Auf1 ^-/- mice (39, 43, 44), these parameters were evaluated in conditions associated with three different levels of AUF-1 expression (48): in resting cells (basal AUF-1 levels), cytomix-stimulated (lower AUF-1 levels compared to basal) and cytomix-stimulated, AUF-1 siRNA-transfected cells (near-complete AUF-1 loss), which we show to be present in the experimental system ( Figure 3A ): in BEAS-2B cells transfected with scrambled siRNA, cytomix stimulation significantly decreased basal AUF-1 protein level (by 53.1%). Upon AUF-1 silencing, AUF-1 levels in resting and cytomix-treated cells were reduced by 67% and by 96.6%, respectively, compared to scrambled siRNA-transfected, unstimulated cells.

We first examined steady-state mRNA levels of the chosen AUF-1 targets. In unstimulated cells, silencing of AUF-1 did not change significantly their basal expression ( Figure S2 ). In cells stimulated with cytomix ( Figure 3B ), near-complete loss of AUF-1 by siRNA-mediated silencing reproduced the changes in expression induced by cytomix in scrambled siRNA-transfected cells, where AUF-1 levels were lowered by this treatment, with no further enhancement.

Examining mRNA decay by ActD assay, in cells transfected with scrambled siRNA ( Figure 3C , left panels) cytomix induced mRNA stabilization for GLIS2 and EGR1 along with IL-6; however, AUF-1 silencing changed cytomix-induced decay rates differently ( Figure 3C , right panels).

As expected for IL-6, in scrambled siRNA-transfected cells (left panel) cytomix treatment increased the mRNA stability (55%, 70% and 63% of mRNA remaining vs time 0) compared to resting cells (24%, 25% and 14% vs time 0, half-life: > 4 h in cytomix-treated vs 0.4 h in resting cells). In AUF-1-silenced cells (right panel), IL-6 mRNA stabilization by cytomix was further accelerated, with 96% remaining mRNA at 1h vs 55% left at 1h vs time 0 in scrambled-transfected cells.

Also for GLIS2 mRNA, in scrambled siRNA-transfected cells (left panel) cytomix treatment increased mRNA stability over unstimulated cells, with 74% and 60% mRNA remaining at 2 h and 4 h, respectively vs 56% and 39% left in resting cells vs time 0 (half-life: > 4 h in cytomix-treated vs 3 h in resting cells, a basal decay rate slower than IL-6); in AUF-1-silenced cells (right panel) no further stabilization occurred.

EGR1 mRNA had a yet different pattern according to partial or total loss of AUF-1. In scrambled siRNA-transfected cells (left panel), similar to IL-6 mRNA, cytomix treatment triggered a marked mRNA stabilization (64%, 63% and 30% vs time 0) over the rapid mRNA decay rate in unstimulated cells (16%, 15% and 5% vs time 0, half-life: 2 h in cytomix-treated vs 0.3 h in resting cells). Surprisingly, in AUF-1 silenced cells (right panel) EGR1 mRNA decay rate became faster in all timepoints in both cytomix-treated cells (39%, 30% and 14% vs time 0) and in resting condition (8%, 8% and 3% vs time 0, half-life: 0.5 h in cytomix-treated vs < 0.5 h in resting cells) though differences with the corresponding values in scrambled-transfected cells was statistically non-significant due to data variability.

For the remaining validated transcripts (MUC-1, FOXP4, CRTC1, TGFβ1, DDX17, DHX36, HDAC2 and IGF1R), we detected a slow rate of mRNA decay in resting cells with half-lives > 4 h, with small measurable changes induced by cytomix regardless of AUF-1 levels ( Figure S3 ).

Findings in Auf1^-/- KO mice indicate that its loss favours exaggerated cytokine responses and promotes cellular senescence through multiple mechanisms, providing a strong rationale for pathogenic relevance of the described decrease in AUF-1 bronchiolar expression in COPD (48). We therefore investigated whether epithelial cell senescence was present in conditions of AUF-1 loss, modelled by cytomix stimulation. We assessed this in BEAS-2B and also in HSAEC, in which we confirmed the decrease in AUF-1 protein by cytomix without concurrent changes in mRNA levels, and lack of changes in expression of other two relevant RBPs, TTP and HuR ( Figure S4 ), as previously reported in BEAS-2B (48). Both cell types were stimulated with cytomix for 48 h or with low-dose etoposide (6 μM) as trigger control for senescence (67, 116) for 24 h ( Figure 4A, B ). At the end of the treatment period, set as time 0, culture media was replaced without adding any further stimulus and cells were incubated for 5 days, set as time 5.

Immunoblot analysis showed that cytomix induced a significant decrease of AUF-1 that persisted at time 5 (28.7% and 20.1% of corresponding controls at time 0 and 5, respectively). Importantly, levels of AUF-1 were significantly reduced also by etoposide treatment, more markedly at time 0 than at time 5 (9.6% and 45.0% of expression in controls, respectively). In the same experiments, markers of senescence displayed corresponding time-dependent changes: cytomix pre-stimulation significantly decreased levels of phospho-Retinoblastoma (56.1 and 40.4% of controls at time 0 and 5, respectively), it increased levels of the cyclin-dependent kinase (CDK) inhibitor p21 (563.4 and 280.1% of controls at time 0 and 5, respectively) and of phospho-p53/p53 ratio (1596,6% and 1407,7% of controls at time 0 and 5, respectively). A similar effect was induced by etoposide pre-treatment, which induced a decrease in phospho-Retinoblastoma levels (45.0% and 84.1% of controls at time 0 and 5, respectively), an increase in p21 levels (3998.4% and 404.4% of controls at time 0 and 5, respectively) and of phospho-p53/p53 ratio (37364,5 and 1887,9% of controls at time 0 and 5, respectively). Overall, cytomix effects were less marked, or comparable in amplitude, to those exerted by etoposide yet they were more persistent, being in all cases present at time 5, although experimental variability hampered in some case the finding’s statistical significance. Concurrently, real time-PCR showed significant decrease of mRNA levels of AUF-1 (42% inhibition over unstimulated control) after 5 days of cytomix stimulation, in contrast to unchanged mRNA levels determined after 48 h, at time 0 ( Figure 4C ).

In the same model, flow-cytometric senescence-associated β-galactosidase (SA-β-gal) assay detected a consistent and significant increase in β-gal activity upon cytomix treatment at time 5 in both cell types (up to 26.4% for BEAS-2B and 23.6% for HSAEC vs controls, p ≤ 0.05) which was in this case comparable in amplitude and duration to the effect seen at time 5 with etoposide treatment (39.1% for BEAS-2B and 28.7% for HSAEC vs controls, p ≤ 0.05) ( Figures 4D, E ).

Lastly, we evaluated in the culture supernatants the expression of SASP-related inflammatory mediators IL-1β, CCL2 (MCP-1), IL-6 and IL-8 (117) using LEGENDplex immunoassays ( Figure 4F ). In this setting, cytomix and etoposide displayed markedly different effects. In BEAS-2B cells (left panel), cytomix stimulation induced a robust and prolonged upregulation of IL-1β, IL-6 and IL-8 with little CCL2 modulation, in contrast with little or no cytokine increase by etoposide treatment. Differently from BEAS-2B, in HSAEC (right panel) cytomix induced a cytokine profile with smaller and more transient IL-6 and IL-8 upregulation, while IL-1β and CCL2 levels remained elevated at time 5. Despite increased β-gal activity ( Figure 4D, E ), in HSAEC etoposide did not elicit cytokine release.

These results suggest that cytomix pre-stimulated cells, while expressing low levels of AUF-1 also underwent cell cycle arrest and displayed features of senescent phenotype, likely with mechanisms only partially common to those occurring in etoposide-treated cells.

Given that AUF-1 mRNA expression was unchanged upon 48h cytomix stimulation in HSAEC ( Figure S4 ) and in BEAS-2B (48) and decreasing significantly only at time 5 ( Figure 4C ), we initially evaluated whether cytomix decreased AUF-1 protein via proteasome-regulated degradation (118, 119). Cell preincubation with proteasome inhibitor MG 132 (10 μM) did not increase AUF-1 protein in resting and cytomix-treated BEAS-2B ( Figure S5 ). In parallel, we assessed whether cytomix triggered an increase in EVs with relative transfer of AUF-1 in this compartment. Supernatants from resting and cytomix-treated BEAS-2B and HSAEC cells, both seeded at equal cell density/condition, were collected and EV isolated by differential centrifugation were characterized by immunoblot, DLS and TEM. Immunoblot analysis ( Figure 5A ) revealed that in parallel with the decrease in intracellular AUF-1 levels, in both cell types cytomix induced an increase in AUF-1 detection in the extracellular EVs fraction. Together with AUF-1, CD63 and CD9 were included as markers for EVs, and supernatants of EVs pellet were also loaded to confirm isolation of EVs. DLS analysis ( Figure 5B ) confirmed the release of particles with similar size range in BEAS-2B (average size of 226.6 nm and 307.1 nm) and in HSAEC (average size of 293.3 nm and 366.8 nm) in resting and cytomix-treated cells. Quantification of EVs by NTA analysis showed a concentration of 3.94e+10 ± 1.71e+09 (mean ± SEM) particles/ml and 8.88e+10 ± 2.34e+09 particles/ml released by resting and cytomix-stimulated cells, respectively; the mean ± SEM size in these conditions were 193.1 ± 1.5 nm and 209.7 ± 0.7 nm, respectively. Further analysis by TEM of EVs from BEAS-2B cells ( Figure 5C ) revealed EVs of spherical shape surrounded by a bilayer. Assessment of particle size by Image software by randomly selected vesicles (100 in each condition) indicated a mean average diameter of 116.73 nm in unstimulated and 117 nm in cytomix-stimulated samples. Differences in size measurement in TEM compared to DLS for EVs have been previously documented (120, 121). TEM analysis of the corresponding BEAS-2B cell monolayers ( Figure 6A ) showed an enrichment of membrane protrusions in cytomix-stimulated cells (lower panels) suggestive of budding vesicles. Immunogold labelling with anti-AUF-1 antibody ( Figure 6B ) revealed detectable staining in EVs, further supporting localization of AUF-1 in this extracellular compartment.

Given decreased expression of AUF-1 in COPD epithelium (48), changes in the 494 RIP-Seq-derived AUF-1 target mRNAs levels were investigated in multiple transcriptomic studies of airway bronchiolar epithelium, to further study the potential impact of AUF-1 regulation in airway epithelial responses. A first analysis was conducted in a public microarray database from small airway epithelium obtained by bronchial brushings of stable COPD patients, smokers and non-smokers both with NLF (GEO ID: GSE5058) (79) ( Figure 7A ). Genes whose relative probes showed discordant FC values (up- and down-regulated) were not included in the total gene count. Out of the 494 genes, 150 (30%) were differentially expressed genes (DEG) in COPD patients vs smokers with NLF (FC ≥ |1.5|, FDR ≤ 0.05), with the large majority (102 of them, 66% of DEG, with FC ≤ -2) down-regulated, with changes of decreasing amplitude in non-smokers and healthy controls ( Figure 7A , Table S5 ). The same search was implemented in a newly generated RNA sequencing database from whole lung biopsies of stable moderate-to severe COPD patients and age- and smoking history-matched smoker subjects with NLF ( Table S3 ). Fifty-two (10%) of the AUF-1- bound target mRNA were expressed as DEG, and in this case as well the majority (41 of them, 79%) were significantly downregulated (log₂FC ≤ -0.40) ( Figure 7A , Table S6 ). Cross comparisons of the two AUF-1 target DEG lists identified 24 common transcripts, also in this case largely down-regulated, with only 4 genes (LPP, SF3A2, NEU3, DOCK1) upregulated. These 24 transcripts encode for a core of DNA/RNA binding proteins involved in regulation of transcription, DNA repair and genomic stability, telomere maintenance, RNA metabolism and translation, which are all within described AUF-1 regulatory functions; a part of them is involved in cell adhesion, cytoskeletal rearrangements for motility, migration, endocytosis, phagocytosis and nucleocytoplasmic shuttling, and protein ubiquitination ( Table 3 ).

We further probed the transcriptomic profiles of GSE5058 database by applying GSVA analysis (78) to calculate the enrichment score of those AUF-1-RNA targets characterized by an EF≥2 and FDR ≤ 0.05 (n=73). The global gene signature was significantly enriched in COPD patients compared to smokers with NLF (difference of ES (dES) = 0.20; p-value<0.05) ( Figure 7B ). Lastly, we examined the expression of AUF1-bound epithelial targets in scRNA-Seq datasets generated from primary airway epithelial cells from stable COPD patients and healthy control subjects differentiated at ALI (82). The dataset contains transcriptomic profiles identifying nine different cell clusters included in three main cell types: club/goblet, basal and ciliated cells identified according to specific gene expression markers. Overall, AUF-1 targets were significantly overexpressed in COPD epithelial cell transcriptome over that of healthy subjects ( Figure 7C ), but expression profiles vary among the cell type clusters, with higher enrichment in basal.1 cells and lower in club/goblet.1. from COPD donors compared to healthy controls. The AUF-1 expression signature was also not detected in cycling basal and lower, although not statistically different, in ciliated.1 and ciliated.2 cells from COPD donors compared to healthy controls ( Figure 7D ). Differential expression analysis between COPD and healthy subjects showed that AUF-1 targets are specifically divergent in club/goblet.4 subcluster, as well as in cycling basal cells and in ciliated.2 cells ( Figure 7E ). Interestingly, club/goblet.4 subcluster was characterized by high expression of interferon response genes (82).

The list of 494 RIP-Seq-derived AUF-1 targets ( Table S4 ) was subjected to Ingenuity Pathway Analysis (IPA) to map the main biological functions putatively affected by AUF-1. GO analysis demonstrated significant enrichment of epithelium-derived Canonical Pathways (CP) previously associated with COPD, or cigarette smoking and smoking-related lung disease (171, 172) such as intracellular and signalling molecules relevant in inflammation (SAPK, ERK/MAPK, CXCR4, EGF, TGFβ, FGF, GM-CSF, IL-9, IL17R, CCR3), cell cycle and DNA repair (p53, mitotic roles of polo-like kinase, DNA Double-Strand Break Repair by Non-Homologous End Joining), metabolism signalling (Aryl Hydrocarbon Receptor Signaling, PPARα/RXRα Activation, TR/RXR Activation) ( Figure 8A , top circus plot; Table 4 for selected pathways, full CP list in Table S7 ). Importantly, IPA performed over the 150 RIP-Seq-derived AUF-1 targets identified as DEG in COPD in the bronchiolar epithelium-derived transcriptomic GSE5058 database shared about a third of this profile but also acquired enrichment in important CP associated with cellular senescence and inflammation (senescence pathway, AMPK signaling, autophagy, macrophage alternative activation pathway) and with relevant and diverse posttranscriptional regulation (Inhibition of ARE-mediated mRNA degradation pathway, MicroRNA biogenesis signaling pathway, lnc HOTAIR regulatory pathway) ( Figure 8A , bottom circus plot; Table S8 for full CP list). Furthermore, AUF-1 targets were searched in the human SASP database of secreted proteins and exosomal cargo-associated SASP factors (www.SASPAtlas.com) originating from primary renal cortical epithelial cells and lung fibroblasts, in which senescence was induced by X ray-irradiation (IR) and inducible RAS overexpression (30). In this database, 26 AUF-1 targets were regulated in IR-induced senescent epithelial cells vs control, 10 of which (ACADM, AHNAK, CLTC, DYNC1H1, LRB4, MACF1, MIOF, PKM, PSMD12, USP9X) reached statistically significant increases (log₂(SEN/CT) ≥ 0.58) ( Figure 8B ) and 7 (AKAP12, ATP6V1A, CLTC, COPB2, DYNC1H1, MYOF, PSMD12) were also found in exosome SASP of IR-induced senescent fibroblasts (log₂(SEN/CT) ≥| 0.58|).

Full length p45^AUF-1 protein sequence (containing all exons) was submitted to target search through CatRAPID algorithm, which estimates the binding propensity of protein-RNA pairs (62). This bioinformatic analysis listed 3,367 coding genes as putative AUF-1-binding targets. Of these, 123 genes were expressed in our Input dataset, with TPM cutoff ≥ 0.5 in a least one biological replicate and showed a DP≥0.75 computed by catRAPID tool. Of these 123 computationally derived AUF-1 targets, 70 (56.9%) were shared with the RIP-Seq experimental dataset ( Figure S6 , Table S9 ). Some transcripts of particular interest for COPD emerged, such as HDAC2, a deacetylase critically involved in epigenetic control of inflammatory responses whose expression and activity are repressed in COPD in several cell types (173, 174).