Human SH-SY5Y neuroblastoma cells (94030304, Sigma-Aldrich) were cultured in complete Dulbecco’s Modified Eagle Medium (61965–059, Gibco) supplemented with 1% non-essential amino acids (11140035, Gibco), 1% penicillin-streptomycin (15140122, Gibco) and 10% fetal bovine serum (FBS; 10270106, Gibco). Cells were grown at 37°C with saturated humidity and 5% CO₂ and maintained in culture for less than 1 month. MAPT knock-out cells were described previously (Sola et al., 2020).

Total RNA extraction with the TRIzol™ Reagent (15596026, Invitrogen) was done according to the instructions of the manufacturer. Extracted RNA was processed with the NEBNext Ultra Directional II RNA library preparation kit for Illumina and sequenced on the Illumina NextSeq500 with single-end, 75 base pair long reads, and 10.6–12.9 × 10⁶ number of reads (Supplementary Table S4). The overall quality of sequencing reads was evaluated using a variety of tools, namely FastQC (Wingett and Andrews, 2018), RSeQC (Wang et al., 2012), AfterQC (Chen et al., 2017) and Qualimap (García-Alcalde et al., 2012). Sequence alignments to the reference human genome (GRCh38) was performed using STAR (v.2.5.2a) (Dobin et al., 2013). Transcript expression was quantified at gene level with the comprehensive annotations v27 release of the Gene Transfer File (GTF) made available by Gencode (Harrow et al., 2012). Raw-counts were further processed in the R Statistical environment and downstream differential expression analysis was performed using the DESeq2 pipeline (Love et al., 2014). Transcripts characterized by low mean normalized counts were filtered out by the Independent Filtering feature embedded in DESeq2 (alpha = 0.05). The RNA-Seq data have the accession no. E-MTAB-8166 and were uploaded on ArrayExpress https://www.ebi.ac.uk/biostudies/arrayexpress/studies/E-MTAB-8166?key=64a67428-adb9-4681-99c9-98910b78ed4c. For the study presented herein, the analysis of the RNA-Seq data was performed with control, non-treated cells utilizing three Tau expressing samples (two parental SH-SY5Y lines and the unsuccessful CRISPR-Cas9 SH-SY5Y 231A line) and the three Tau-depleted SH-SY5Y cell lines 231P, 231K and 232P (Sola et al., 2020); Supplementary Table S4). The genes differentially upregulated in Tau-KO cells (723) were used to interrogate a possible gene-set enrichment utilizing the transcription tool of the EnrichR portal (Chen et al., 2013; Kuleshov et al., 2016; Xie et al., 2021).

Pseudolentiviral particles were produced by transient transfection of HEK293FT cells with 2 μg of the pSIF-H1-puro-IGFBP3 shRNA-2 or of the control plasmid pSIF-H1-puro-luciferase shRNA and 8 μg of the feline immunodeficiency virus (FIV) packaging plasmid mix (pFIV-34N and pVSV-G); all plasmids were kindly provided by Prof. Yuzuru Shiio (Greehey Children’s Cancer Research Institute, University of Texas). Cell conditioned medium was collected 2 days after transfection and cleared by centrifugation at 300 g for 5 min, 4°C. Pseudo-lentiviruses were 20-fold concentrated with centrifugal filters (MWCO 30 kDa, UFC903024, Amicon) at 3’000 g for 30–45 min, 4°C, aliquoted and stored at −80°C.

Human SH-SY5Y neuroblastoma cells (1 × 10⁵) were seeded into a 24-well plate coated with poly-D-lysine (P6407, Sigma-Aldrich) 1 day before pseudo-lentiviral particle transduction. One day after transduction, cells were supplemented with fresh complete medium and selected in the presence of 2.5 μg/mL puromycin (P8833, Sigma-Aldrich) for 2 weeks.

Treatment of SH-SY5Y cells with Tazemetostat (CAS No. 1403254–99-8, S7128, Selleckchem) was performed at 10 µM for 4 days starting from a 10 mM stock solution in DMSO; vehicle 0.1% DMSO was added to the controls.

For direct analysis by western blot, total lysates from cells cultured in 6-well plates were prepared in 50 μL of SDS-PAGE sample buffer (1.5% SDS, 8.3% glycerol, 0.005% bromophenol blue, 1.6% β-mercaptoethanol and 62.5 mM Tris pH 6.8) and incubated for 10 min at 100°C. 15 μL of the sample per lane was loaded on SDS-polyacrylamide gels (SDS-PAGE).

For immune isolation, the cells were rinsed with PBS and lysed on ice in 100 μL of AlphaLisa Lysis Buffer (AL003, PerkinElmer) supplemented with protease and phosphatase inhibitor cocktails (S8820 & 04906845001, Sigma-Aldrich). Cell lysates were treated with benzonase (707463, Novagen) for 15 min at 37°C, centrifuged at 20’000 g for 10 min at 4°C and supernatants were collected as cell extracts. These latter were diluted in HiBlock buffer (10205589, PerkinElmer) and incubated overnight at 4°C with 0.5 μg of primary antibodies against SUZ12 (3,737, Cell Signaling Technology), or EZH2 (5,246, Cell Signaling Technology). Protein G-Sepharose beads (101241, Invitrogen) were added for 1 h at room temperature (RT) and the beads were washed three times in PBS with 0.1% Tween-20. Bead-bound proteins were eluted in SDS-PAGE sample buffer by boiling for 10 min at 100°C.

After SDS-PAGE, PVDF membranes with transferred proteins were incubated with primary antibodies: 0.084 μg/mL SUZ12, 0.421 μg/mL EZH2, 0.18 μg/mL GAPDH (ab181602, Abcam), 0.1 μg/mL H3K27me3 (C15410069, Diagenode), 0.02 μg/mL H3 (ab176842, Abcam), or 0.4 μg/mL IGFBP3 (sc365936, Santa Cruz Biotechnology). Primary antibodies were revealed with anti-mouse IgG coupled to IRDye RD 680 or anti-rabbit IgG coupled to IRDye 800CW (Licor Biosciences, 926–68070 & 926–32211) on a dual infrared imaging scanner (Licor Biosciences, Odyssey CLx 9,140) and quantified with the software provided (Licor Biosciences, Image Studio V5.0.21, 9,140–500).

For immune staining, cells were grown on poly-D-lysine coated 8-well microscope slides (80826-IBI, Ibidi). Cells were fixed in 4% paraformaldehyde and stained (Ulrich et al., 2018) with primary antibodies: 0.168 μg/mL SUZ12, 0.842 μg/mL EZH2, 1.6 μg/mL H3K27me3 or 1.5 μg/mL P16 (ab108349, Abcam). Detection by fluorescent laser confocal microscopy (Nikon C2 microscope) was done with 2 μg/mL secondary antibodies anti-mouse IgG Alexa594, anti-rabbit IgG -Alexa 488 or anti-rabbit IgG-Alexa 647 (A-11032, A-11034, A21245, Thermo Fisher Scientific). Nuclei were counterstained with 0.5 μg/mL DAPI (D9542, Sigma-Aldrich). Images were acquired by sequential excitations (line-by-line scan) with the 405 nm laser (464/40 emission filter), the 488 nm laser (525/50 nm filter), the 561 nm laser (561/LP nm filter) and the 650 nm laser (594/633 emission filter). ImageJ was used for all image quantifications.

Total RNA extraction using the TRIzol™ Reagent (15596026, Invitrogen) and cDNA synthesis using the GoScript Reverse Transcription Mix Random Primers (A2800, Promega) were done according to the instructions of the manufacturer. Amplification was performed with SsoAdvanced Universal SYBR Green Supermix (1725271, Bio-Rad) with 43 cycles at 95°C for 5 s, 60°C for 30 s and 60°C for 1 min using specific primers for EZH2, SUZ12, IGFBP3, GPR37, ITGA3, MRC2 and IRF6 gene transcripts (Supplementary Table S5). Relative RNA expression was calculated using the comparative Ct method and normalized to the geometric mean of the GAPDH and HPRT1 mRNAs.

The ChIP experiment followed standard protocols with an H3K27me3 antibody (C36B11, Cell Signalling) or control IgG (12–370, Sigma-Aldrich) as negative control. In brief, 40 × 10⁶ cells were resuspended and fixed with 1% methanol-free formaldehyde in complete medium for 5 min at room temperature, followed by 5 min with 125 mM glycine. Cells were washed twice with cold PBS and lysed (Buffer B; PN 520237, Covaris). Nuclei were washed (Buffer C; PN 520237, Covaris) and resuspend in sonication buffer (10 mM Tris pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.5% N-lauroylsarcosine sodium salt and fresh 0.1% Na-deoxycholate and proteinase inhibitors). Chromatin was fragmented by sonication with a Bioruptor to an average size of 100–400 bp and brought to 1% TritonX-100. A 50 μL aliquot was saved as input, whereas 500 μL each was incubated with 5 μg antibody, overnight at 4°C. Protein A magnetic beads (16–661, Merk Millipore) were washed three times with 0.5% BSA in PBS and added for 5 h at 4°C. DNA-protein complex was recovered, washed and eluted overnight at 65°C (GeneJET PCR purification; K0701, Thermo Fisher Scientific).

Library preparation from 5 ng purified chromatin was performed using NebNext Ultra II DNA Library Prep with magnetic purification beads (E7103S, NewEngland Biolabs) and multiplex oligonucleotides (Dual Index Primers; E7600S, NewEngland Biolabs), followed by adaptor ligation according to the manufacturer’s protocol. Quality controls of amplified chromatin fragments were performed on Bioanalyzer 2,100 (Agilent Technologies) and Qubit V4 (Thermo Fisher Scientific). Next-generation sequencing was performed on NextSeq 2000 (Illumina) with the P2 reagents kit V3 (100 cycles; Illumina). Samples were processed starting from stranded, single-ended 120bp-long sequencing reads.

Raw data in FastQ format were first preprocessed to ensure suitability for downstream analysis. Adapter sequences were trimmed from the reads using trimmomatic algorithm. Quality control was performed to filter out reads with low-quality bases, short lengths, or those suspected to be sequencing artifacts. Reads were mapped to the HG38 human genome using BWA alignment. Peak calling was performed using MACS (version 2) with “broadPeak” setting to identify broad regions of H3K27me3, then annotated to the nearest genes with HOMER considering both transcription start and end sites. Peak counts were between 82’000 and 130’000 for the six samples. To detect differential binding patterns, the analysis was performed with the Limma package. The raw count data were normalized with TMM, and statistical testing assessed the significance of differential binding between conditions, considering factors such as biological variability and technical noise. To control the false discovery rate, the p-values resulting from the differential analysis were adjusted using the Benjamini-Hochberg procedure. This correction minimized the risk of reporting false positives while maintaining statistical power. Raw and processed data have the accession number (GSE242694) on Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE242694).

Senescence-associated β-galactosidase (SA-βGal) staining was determined on cells grown in 6-well plates, fixed with 2% paraformaldehyde for 10 min at RT and washed twice with gentle shaking for 5 min at RT. Cells were then incubated with 1 mg/mL X-gal (20 mg/mL stock in DMF; B4252, Sigma-Aldrich) diluted in pre-warmed 5 mM K₃[Fe(CN)₆] (P8131, Sigma-Aldrich), 5 mM K₄[Fe(CN)₆]·3H₂O (P-3289, Sigma-Aldrich), and 2 mM MgCl₂ (M8266, Sigma-Aldrich) in PBS at pH 6.0. Acquisition and quantification of the images for SA-βGal activity and cell area were done with an automated live cell imager (Lionheart FX, BioTek).

For LysoTracker staining, cells were seeded in poly-D-lysine coated 8-well microscope slides, incubated for 10 min at 37°C with 0.25 μM Lysotracker Red (L7528, Thermo Fisher Scientific). Nuclei were counterstained with 2.5 μg/mL Hoechst (H3570, Invitrogen) for 10 min at 37°C, afterwards cells were washed with complete medium. Images of living cells were acquired on a fluorescent laser confocal microscope (C2, Nikon) by sequential excitations (line-by-line scan) with the 405 nm laser (464/40 nm emission filter), and the 561 nm laser (561/LP nm filter). ImageJ was used for all image quantifications. Both the lysosomes area and the mean number of lysosomes per cells and per images were determined.

We performed a next-generation transcription (RNAseq) analysis of human SH-SY5Y neuroblastoma cells knocked-out for Tau when compared to normal Tau expressing cells (Figure 1A). The sequences obtained from the six samples analyzed (three Tau expressing cell lines, three Tau-KO cell lines) mapped reliably to ∼16’000 genes. Additional 14’000 transcripts were expressed at low levels and were not included in the differential expression analysis. The primary data were stored at https://www.ebi.ac.uk/biostudies/ with open access (E-MTAB-8166). When filtering for differentially expressed transcripts in Tau-KO cells, 1,388 RNAs displayed a significant change (Adj p < .05), of which 723 RNAs were upregulated in the log2(FC) range between 0.31 and 11.05 (between 1.24 and ∼2000 fold higher than in control Tau expressing cells) (Supplementary Table S1).

We selected these 723 differentially expressed genes to interrogate a possible gene-set enrichment utilizing the transcription tool of the EnrichR portal (Chen et al., 2013; Kuleshov et al., 2016; Xie et al., 2021). The CHIP-sequencing datasets (ChEA 2016) identified an overrepresentation of Polycomb Repressive Complex-associated proteins among the 68 datasets showing a significant (Adj p < .01) difference. Indeed, almost half of the enriched CHIP-sequencing datasets were obtained from core components or known regulators of PRC2 (32%) or of PRC1 (16%) (Figure 1B; Supplementary Table S2). PRC2 actively catalyzes the trimethylation of histone 3 (H3) at lysine 27 (H3K27me3) (Laugesen et al., 2016; Moritz and Trievel, 2018). In agreement with the identification of PRC2 in the CHIP-sequencing datasets, mining of the epigenomics roadmap (HM ChIP-seq) resulted in a 73% enrichment of the H3K27me3 signature (Adj p < .01) in 45 datasets (Figure 1C; Supplementary Table S3). Altogether, analysis of the RNAseq data suggested that upregulation of transcription of a specific set of genes in Tau-depleted SH-SY5Y neuroblastoma cells might ensue from a relief of PRC2-mediated repression of gene transcription.

We subsequently analyzed the amount of two PRC2 core components in Tau-KO cells by western blot. As suggested by the transcriptomics data, we observed reduced amounts of the catalytic subunit EZH2 and the scaffold subunit SUZ12 of PRC2 in Tau-KO cells when compared to Tau expressing cells (Figure 2A). Reduced proteins were found also by quantitative immune staining of the cells utilizing specific antibodies (Figure 2B; Supplementary Figure S1). RT-qPCR analysis excluded that the effect on PRC2 protein resulted from reduced transcription since no difference was found for the EZH2 and SUZ12 mRNAs (Figure 2C), data that suggested a Tau-dependent effect on PRC2 protein stability. Nonetheless, co-immune isolation revealed the presence of the EZH2-SUZ12 core complex of PRC2 in Tau-KO cells, albeit at reduced levels when compared to controls (Figure 2D).

We determined by western blot and quantitative immune staining the extent of modified H3K27me3, an epigenetic mark produced by the histone methyl transferase activity of PRC2 (Guo et al., 2021), and reduced by KDM6A/B demethylase activity (Lee et al., 2007). Confirming the lower amounts of the PRC2 complex, we found that Tau-KO cells displayed decreased H3K27me3 when normalized for total H3 (Figures 3A, B). Among the upregulated transcripts found by RNA-seq in Tau-KO cells, we selected five known PRC2 targets displaying close to average signals (Supplementary Table S1): IGFBP3 (19.0x of WT, Adj P .004), GPR37 (9.3x, .015), ITGA3 (6.3x, .017.3x), MRC2 (5.1x, .016), and IRF6 (3.2x, .0498). Determination of mRNA expression by RT-qPCR validated their upregulation (Figure 3C); suggesting that repression of transcription may be relieved by reduced PRC2 and H3K27me3 as plausible consequence of Tau-depletion.

IGFBP3 protein is a component of the senescence-associated secreted phenotype (SASP) (Basisty et al., 2020). Having previously reported that Tau-depletion favored cellular senescence (Sola et al., 2020), we interrogated the role of IGFBP3 in this process. As anticipated from the mRNA data, a strong overproduction of endogenous IGFBP3 was present in Tau-KO cells (Figure 4A). Reinforcing the link between PRC2 and IGFBP3, treatment of SH-SY5Y cells with Tazemetostat, a specific blocker of the histone methyl transferase activity of EZH2 (Straining and Eighmy, 2022), reduced H3K27me3 and increased IGFBP3 (Figures 4B, C). Lower EZH2 and SUZ12 and higher IGFBP3 was confirmed in an independent Tau-KO cell line (Supplementary Figure S2).

Increased cellular expression of IGFBP3 is associated with autocrine and paracrine senescence induction (Elzi et al., 2012), and reduced PRC2 is also linked to increased cellular senescence (Ito et al., 2018). Furthermore, increased senescence is observed in Tau-KO cells (Sola et al., 2020). Thus, we postulated that PRC2-dependent de-repression of IGFBP3 in Tau-depleted cells may explain the induction of cellular senescence. We observed first that Tau depletion as well as PRC2 inhibition with Tazemetostat both increased the percentage of SH-SY5Y cells entering in senescence, as assessed by three independent markers: senescence-associated β-galactosidase (SA-βGal), P16, and the size and number of lysosomes labelled with the acidotrophic LysoTracker dye (Figure 5A; Supplementary Figure S3). Next, we reduced IGFBP3 expression in Tau-KO cells by a shRNA-based approach (Figure 5B) and found that this inhibited senescence induction in Tau-KO cells (Figure 5C; Supplementary Figure S3). Thus, we validated our hypothesis that increased senescence in Tau-KO SH-SY5Y cells was likely the consequence of decreased PRC2-dependent repression of IGFBP3 expression.

Finally, we performed a chromatin immune precipitation (ChIP) experiment with an antibody against H3K27me3 validated for this purpose to analyze any difference between two Tau-KO SH-SY5Y cell lines when compared to normal Tau expressing cells. Raw values and processed data were stored at (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE242694) with accession number (GSE242694). Principal component analysis grouped the four Tau-KO samples above the two Tau expressing samples, when considering the difference in H3K27 trimethylation expressed by PC2 (Figure 6A). 73’041 H3K27me3 marks were found for the six samples analyzed, 21’913 of which were present in all six samples, 1’237 were only present in the two Tau samples, and 90 were shared only in four two Tau-KO samples originating from the two independent lines (Figure 6B). The volcano plots show that when comparing the Tau-KO samples to control samples, there was a significant difference (Adj p < .05) for only 199 marks (153 reduced in Tau-KO cells). For the second Tau-KO cell line, a significant difference was found for (Adj p < .05) for 210 marks (191 reduced in Tau-KO cells) but, notably, the two Tau-KO lines had an overlap for 90 significantly different H3K27me3 marks (Figure 6C). However, no significant differences were found at this level of analysis for the two H3K27me3 marks found for the IGFBP3 gene (Figure 6D) and no gene-set enrichment for senescence-related mechanisms were found (not shown). Altogether, analysis of the ChIP data confirmed a reduced number of H3K27me3 marks in two Tau-KO SH-SY5Y cell lines, which correlated with reduced PRC2 and H3K27me3 observed in SH-SY5Y neuroblastoma cells upon Tau-depletion.