NSC-34 cells were used as previously described (Ryan et al., 2009) and were the gift of Dr. Heather Durham, (Montreal Neurological Institute, Montreal, QC). MDA-MB-468 cells were obtained from ATCC. Cells were grown in a 5% CO₂ incubator at 37°C. NSC-34 cells were maintained in DMEM with 10% fetal bovine serum (FBS) (Cashman et al., 1992). MDA-MB-468 cells were maintained in DMEM with 10% FBS. Stable transfectants that express human PGRN (NSC-34 and MDA-MB-468 cells), were generated by transfection with human PGRN (pcDNA3-hPGRN) or pcDNA3 for empty vector control transfections. Cells were transfected using Lipofectamine (Invitrogen) and selected with G418 (400 μg/ml) for 4–6 weeks according to manufacturer’s instructions. To generate cells that stably express ShRNA-mPGRN, NSC-34 cells were transfected with shRNA constructs that were designed by OriGene Technologies using murine Grn specific 29 nucleotide stretches within the coding region. NSC-34 cells were seeded onto 6 well plates and then transfected with shRNA-mPGRN (pRS/shPGRN), or control vector (pRS vector alone, Origene Technologies), using Fugene (Roche) when the cultures were 80% confluent. Stable transfectants were selected by culturing in medium containing puromycin (7 μg/ml) (Invitrogen) for 6 weeks according to manufacturer’s instructions. To prevent phenotypic drift, stocks of the original transfectants were frozen in liquid N₂ and reanimated at regular intervals.

Total RNA was isolated using Trizol reagent (Invitrogen). cDNA synthesis was performed with Revert-aid reverse transcriptase (Thermo Scientific). Human and mouse PGRN specific primer sets were designed to amplify species specific products. See Supplementary Table 1 for the sequences of the PCR primers. The polymerase chain reaction (PCR) program consisted of denaturation for 2 min at 94°C; 35 cycles at 94°C, 30 s; 55°C, 30 s; 72°C, 30 s; and a final extension of 5 min at 72°C. PCR was performed using Taq DNA Polymerase (Bio-Basic).

Radioimmunoprecipitation assay (RIPA) buffer (Sigma) mixed with Complete Protease inhibitor Cocktail (Roche Applied Science, 1 tablet for 7 ml buffer) was used to lyse and extract proteins from NSC-34 cells. Protein lysates were incubated on ice for 5 min, briefly sonicated and centrifuged at 13,000 rpm for 15 min at 4°C. Supernatants were quantified for protein content and equal amounts were mixed with 2× sample buffer, boiled for 5 min and resolved on a 10% SDS-PAGE gel. Proteins on the gel were transferred onto a nitrocellulose membrane and blocked over-night with membrane blocking agent (GE Healthcare) or non-fat milk (Thermo Scientific) at 4°C. The blots were incubated with 1:250 anti-human PGRN polyclonal antibody (rabbit anti-human PGRN generated in our laboratory) or 1:10000 anti-human PGRN (R&D system) for 1 h in blocking buffer followed by extensive washing. After incubating with an anti-rabbit or anti–mouse IgG-horseradish peroxidase (HRP)-conjugated secondary antibody (diluted 1:4,000) at room temperature for 1 h and blots were visualized using enhanced chemiluminescence (GE Healthcare) according to the manufacturer’s instructions. Other antibodies used were 1:1000 anti-Ret (Cell Signaling Technology) and anti-phospho Ret (abcam), 1:2000 anti-EGFR, anti-phosphoEGFR, anti-ErbB2, anti-phosphoErbB2 (R&D systems). Protein loading controls were performed by stripping the membrane and probing with mouse monoclonal β-actin antibody (AC-40; Sigma) at a dilution of 1:1000.

Cells were plated in 6 well plates at 7000 cells per well containing DMEM with 10% FBS. After 24 h, the media was replaced with DMEM containing 0% FBS, 1% FBS, or 10% FBS. The cultures were monitored for 25 days, and photographs were acquired using an inverted phase contrast microscope. The photographs were analyzed using ImageJ (Schindelin et al., 2012) by enabling the “Area” and “Shape” descriptors, and “Perimeter” and “Feret’s diameter” measurements (Rasband, 1977). Feret’s diameteris the longest distance between any two points along the selection boundary (Ferreira and Rasband, 2012), sometimes also called the “caliper length,” and is a measure of asymmetry. The neurite-like projections were measured using the NeuronJ plugin.¹ This facilitates the tracing and quantification of neurite extension using the software ImageJ (Meijering et al., 2004; Meijering, 2010).

Cells were cultured in 12 well plates coated with nitrocellulose which was prepared by dissolving 5 cm² of nitrocellulose (membrane 0.22 um, Bio basic) in 6 ml of methanol. Briefly 0.1 ml aliquots of this solution were rapidly spread over the surface of each well in 12 well plate and allowed to dry under a laminar flow hood as described in Lagenaur and Lemmon (1987). PGRN was applied in 3 μl droplets containing 100 μg/ml. After one minute, the droplets were removed gently using a pipette. The plates were then blocked by washing twice with 2 ml of culture media (DMEM with 10% FBS) for 10 min each. NSC- 34 cells were cultured on the nitrocellulose coated plates at the concentration of 6000 cells per ml in 2 ml medium supplemented with 10% FBS for 2 days. The neurite length was measured using ImageJ by tracing the neurite-like projections.

Statistical significance among experimental groups was determined by one-way ANOVA followed by Tukey’s Multiple Comparisons Test (p < 0.001-^***, p < 0.01-^**, p < 0.05-*) using GraphPad software (GraphPad Prism Software Inc., San Diego, CA). Error bars represent s.e.m.

NSC-34/shPGRN, NSC-34/hPGRN, or control constructs were cultured in T75 flasks. They were sub-cultured in four replicates. NSC-34 cells (200,000) from each of 12 flasks were plated in T25 flasks with serum. After 24 h the media containing 10% serum was removed and replaced with media containing 5% serum. After a further 24 h the media was changed to one containing 1% serum. 24 h later cells were lysed in Trizol and total RNA was extracted. The RNA concentration was determined using a NanoDrop chip. cDNA synthesis was carried out using RevertAid First Strand cDNA Synthesis Kit according to the manufacturer’s protocol (Thermo Scientific). Following cDNA synthesis reverse transcription polymerase chain reaction (RT-PCR) was performed to confirm over/under expression of PGRN as indicated earlier. Reduced levels of mPGRN were noticed in all shPGRN constructs and expression of hPGRN was detected only in the over-expressors as expected. The over/under expression of PGRN levels were further confirmed by Western blot as mentioned previously.

The RNA quality/integrity was analyzed using a Nano ChiP Bioanalyzer. The gene expression profile was determined using Illumina Mouse Whole-Genome Expression BeadChips (MouseWG-6 v2.0, multiple 12 samples) at the McGill University and Genome Quebec Innovation Centre. The data replicates were validated by principal component analysis which confirmed that all 4 replicates from the same construct segregate together and separately from the other constructs. The lumi normalization followed by an empirical Bayesian, EB (Wright and Simon, 2003) algorithm was applied to the data using FlexArray 1.6.1 software (available at http://gqinnovationcenter.com/services/bioinformatics/flexarray/tutorials.aspx?l=e) to identify genes that were either significantly reduced or induced in the PGRN over/under expressing NSC-34 cells compared to controls. The Database for Annotation, Visualization, and Integrated Discovery (DAVID) v6.8 (Huang da et al., 2009b,a) was employed to provide tables of enriched Gene Ontology (GO) terms and predicted functional pathways based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. Gene networks were analyzed using the Ingenuity Pathway analysis software (IPA). IPA is a web-based bioinformatics application that allows data analysis from microarray experiments by identifying related proteins within a pathway. 1.5-fold enrichment or depletion of transcript signal was used as the cut-off (Kramer et al., 2014). Additional functional assignments for individual genes were obtained manually by searching for their entries in Uniprot,² Genecards³ and On-line Mendelian Inheritance in Man (OMIM⁴). Gene expression data was submitted to Gene Expression Omnibus (GEO) public repository, accession number: GSE185877.

Lipidomics and Proteomics mass spectrometric analyses were performed by the Proteomics Technology Platform, Research Institute of the McGill University Health Centre, Montreal (manager Lorne Taylor). NSC-34/shPGRN or NSC-34/hPGRN cells were cultured in T75 flasks in DMEM with 1% FBS. Cells were collected by trypsinization at about 70% confluency. Three independent extractions and mass spectrometric analyses were performed. Cells were centrifuged, resuspended in 1 ml and cell numbers were determined. Twenty million cells from each sample were collected by centrifugation at 3000 rpm for 5 min and incubated with 150 μl methanol and 500 μl Methyl tert-butyl ether (MTBE, Fisher Scientific) for 2 h on a rocker at room temperature. Phase-separation was performed by adding 125 μl of MS grade water (Fisher Scientific). After 10 min incubation at room temperature each sample was centrifuged at 3000 rpm for 10 min. The upper organic phase was collected, and the lower phase was re-extracted by adding methanol, MTBE, and water. Both fractions were combined, dried using a SpeedVac lyophilizer and were introduced into the TripleTOF™ 5600 System by direct infusion using the Nanospray III source and an Eksigent NanoLC Ultra and Autosampler at a flow rate of 1 μL/min. The data were analyzed using peak view Software and lipid identification and quantification was done in lipid view Software.

NSC-34/ShPGRN, NSC-34-/hPGRN, or vector controls were cultured in T75 flasks in DMEM with 1% FBS until about 80% confluency. Cells were washed with PBS and lysed by homogenizing in solution containing 20 mM HEPES, 150 nM NaCl, and 0.320M sucrose. Supernatants were collected by centrifugation at 3000 rpm at 4°C for 10 min. Samples were normalized for input by cell numbers and subsequently by total peptides using the Scaffold program (see below) at the data analysis stage. For each sample, proteins were loaded onto a single stacking gel band to remove lipids, detergents and salts. The gel band was reduced with dithiothreitol, alkylated with iodoacetic acid and digested with trypsin. Extracted peptides were re-solubilized in 0.1% aqueous formic acid and loaded onto a Thermo Acclaim Pepmap (Thermo, 75 uM ID × 2 cm C18 3 uM beads) precolumn and then onto an Acclaim Pepmap Easyspray (Thermo, 75 uM × 15 cm with 2 uM C18 beads) analytical column separation using a Dionex Ultimate 3000 uHPLC at 230 nl/min with a gradient of 2–35% organic (0.1% formic acid in acetonitrile) over 3 h. Peptides were analyzed using a Thermo Orbitrap Fusion mass spectrometer operating at 120,000 resolution (FWHM in MS1) with HCD sequencing (15,000 resolution) at top speed for all peptides with a charge of 2+or greater. The raw data were converted into *.mgf format (Mascot generic format) for searching using the Mascot 2.6.2 search engine (Matrix Science) against mouse protein sequences (Uniprot 2018). The database search results were loaded onto Scaffold Q+ Scaffold_4.8.9 (Proteome Sciences) for statistical treatment and data visualization. Proteomics data was submitted to MassIVE public repository by participating in ProteomeXchange, accession number: MSV000088451. The Database for Annotation, Visualization, and Integrated Discovery (DAVID) v6.8 (Huang da et al., 2009b,a) was employed to analyze Gene Ontology (GO) as above using a 1.2-fold change cut-off and p < 0.05.

Ligand receptor capture (LRC) identifies the interactions of extracellular soluble proteins with membrane proteins in live cells by making use of a biotinylated trifunctional reagent TRICEPS (Frei et al., 2012). All reactions were performed according to the manufacturer’s protocols. 300 μg PGRN (R&D Systems), or insulin (Sigma-Aldrich) as the positive control, were coupled to 1.5 μl TRICEPS by an amide linkage. The second functional group of the TRICEPS reagent, a protected hydrazine, reacts with aldehydes on the glycan side chains of the target receptor proteins. As aldehydes are not typically present on the cell surface, they are first generated by mild sodium metaperiodate oxidation using 6 × 10⁸ NSC-34 cells. The cells were cultured in DMEM with 10% FBS and washed once with 1× PBS then in DMEM without serum for 30 min prior to cells collection. The LRC reaction was initiated by incubating mildly oxidized and washed living cells with 50 μl of the PGRN-TRICEPS conjugate, or 50 μl of Insulin-TRICEPS for 90 min at 4°C during which time PGRN-TRICEPS or insulin-TRICEPS reacts covalently through hydrazine coupling to the glycan chains of any cell surface membrane protein to which it binds. The cells were lysed, digested with trypsin and the TRICEPS labeled peptides recovered by biotin affinity using streptavidin beads. For mass spectrometry the streptavidin-purified peptides were de-glycosylated using PNGaseF, an enzyme which cuts between the proximal N-acetylglucosamine of the glycan side chain and the glycopeptide asparagine in the N-X-S/T glycosylation site motif (N stands for asparagine, X is any amino acid except proline and S/T is serine or threonine). The PNGase reaction deamidates the asparagine and modifies the N-X-S/T glycosylation motif to a D-X-S/T motif (where D is aspartic acid), the corresponding mass shift being used to identify peptides that were crosslinked by the PGRN or Insulin-TRICEPS conjugates. The TRICEPS labeled peptide recovery and the mass spectrometric analyses were performed by DualSystems Biotech. The samples were analyzed on a Thermo LTQ Orbitrap XL spectrometer. Peptide identifications were filtered to a false-discovery rate of ≤1% and quantified using an MS1-based label-free approach. For the MS1 quantification, the non-linear DYNAMICS Progenesis QI for proteomics was used. The experiment was done in triplicate. All six samples, PGRN and insulin control, showed an alignment score of >90%. Therefore, none of the six samples had to be excluded from statistical analysis. The Data is presented as a volcano plot where all the membrane proteins that crosslinked to either PGRN or the insulin control fall as statistical outliers on either arm of the plot. Positive detection of the insulin receptor control signal confirms that the TRICEPs reaction worked under the conditions used, and that there was no protein denaturation. Non-specific binding is accounted for as signals that are statistically indistinguishable between the insulin and progranulin arms of the experiment. Possible interactions among proteins identified in these assays was performed using the STRING program (Search Tool for the Retrieval of Interacting Genes/Proteins) (Szklarczyk et al., 2019).

The proteome profiler mouse or human Phospho-Receptor Tyrosine Kinases (RTK) antibody array membranes (R&D Systems, Minneapolis, MN) were incubated with NSC-34 or MDA-MB-468 cell lysates and processed as per the manufacturer’s protocol using a Phosphotyrosine specific antibody. In brief, approximately 3 × 10⁷ cells of NSC-34/shPGRN cells expressing reduced PGRN, NSC-34/hPGRN cells overexpressing PGRN and WT NSC-34 control, MDA-MB-468 WT and PGRN expressing cells were cultured in serum free medium overnight. After incubation, cells were washed with ice-cold Dulbecco’s phosphate-buffered saline (DPBS) and lysed with lysis buffer containing aprotonin/leupeptin (10 μg/mL each) for 30 min, centrifuged at 13000 rpm for 5 min. Supernatants were collected and stored at –80 degrees A protein assay (BioRad Biotechnologies, Hercules, CA) was performed before incubating with the RTK membranes to normalize loadings.

NSC-34 vector control cells, and cells that either stably over-express human PGRN (NSC-34/hPGRN) or under-express murine PGRN (NSC-34/shPGRN) were generated (Supplementary Figure 1). PGRN levels were depleted by approximately 50% in NSC-34/shPGRN cells, a reduction equivalent to that associated with FTD-GRN (Supplementary Figures 1A,B). To confirm their neuron-like identity NSC-34 cells were stained for the motor neuron marker SMI32, which recognizes dephosphorylated epitopes of neurofilaments (Supplementary Figure 1C). Neither depletion nor over-expression of PGRN altered TDP-43 mRNA or protein expression (Supplementary Figure 1D).

Progranulin expression had a strong impact on the shape and size of the cell body. At 48 h of culture, cell bodies of NSC-34/hPGRN have greater area and are flattened compared to the controls and NSC-34/shPGRN cells, which are typically small and circular (Figures 1A,B). Neurite length was greater in NSC-34/hPGRN cells than control or NSC-34/shPGRN cells when serum was present (10 or 1%). Extension in NSC-34/shPGRN often appeared to be poorly condensed and disorganized, which we refer to as “tortuous” (Figures 1A,B). Serum-deprivation stimulates differentiation in NSC-34 cells (Eggett et al., 2000) and, correspondingly, over an 8 day period in the absence of serum, all cells exhibited neurite extension with no significant difference detected in neurite length between NSC-34/hPGRN and NSC-34/shPGRN cells (Figure 1C). Extensive cell death occurs in the NSC-34/shPGRN cells after 8 days in serum free medium, at which point measurements were discontinued. PGRN is secreted but also localizes cytoplasmically, due, at least in part, to endocytosis into lysosomes (Hu et al., 2010; Petoukhov et al., 2013; Zhou et al., 2015, 2017). Neurons secrete PGRN in an activity dependent manner (Petoukhov et al., 2013). There are in addition multiple paracrine extracellular sources of PGRN including microglia (Ryan et al., 2009; Hu et al., 2010) and de-differentiated Schwann cells (Hyung et al., 2019). We therefore tested whether extracellular PGRN, mimicking the secreted protein, reproduces the morphological and survival phenotypes associated with endogenous PGRN expression. Exogenous PGRN added to the medium reproduced the effects of stable PGRN expression as WT NSC-34 cells displayed longer neurite outgrowth compared to cells cultured without PGRN on nitrocellulose coated plates pre-treated with PGRN in the presence of 10% serum (Figure 1D). Cell survival of NSC-34/shPGRN cells in 8-day serum-reduced culture decreased relative to WT NSC-34 cells as expected (Ryan et al., 2009), but adding PGRN to the medium restores NSC-34/shPGRN cell numbers to the levels of control NSC-34 cells (Figure 1E).

Given that the cells undergo active morphological differentiation within 2 days incubation in 1% serum with no loss of viability (Figure 1), all subsequent molecular analyzes were performed under these conditions. Gene expression microarrays were used to identify how the mRNA phenotype of NSC-34 cells responds to PGRN expression levels. The anticipated expression level of PGRN were confirmed in each of the NSC-34/shPGRN, control and NSC-34/hPGRN cell lines prior to the analysis (Supplementary Figure 2A). Genes showing statistically significant changes in expression were identified by FlexAarray analysis software. Principal component analysis confirmed that the data from each of NSC-34/shPGRN, NSC-34/hPGRN and vector controls cells segregates independently (Supplementary Figure 2B). Among genes that showed the greatest fold down-regulation in NSC-34/shPGRN cells compared to NSC-34/hPGRN cells (Table 1) were several associated with neurotransmission (dopamine decarboxylase, vesicular GABA transporter, alpha-2 adrenergic receptor), cytoskeletal organization (Stathmin 2, FGD2, SRGAP3), and RNA-binding proteins with roles in neuron development, survival, or congenital motor disorders (IGF2BP1, TSHZ1, GLE1).

Ingenuity Pathway analysis identified potential PGRN-sensitive functional networks (Table 2). Among “biological function” networks PGRN knockdown was associated with neurological disease, behavior and nervous system development and function. The strong neurological disease-related signature in the downregulated data set of NSC-34/shPGRN cells implies that they exhibit disease-relevant molecular signals of neuronal disturbance, and this data set was, therefore, prioritized for further analysis. The top five IPA Associated Network Functions for PGRN knockdown cells versus PGRN over-expressors, based on a 1.5-fold cutoff difference for transcript expression, are given in Table 3. Details of each of the IPA Associated Network Functions are given in Supplementary Figure 3. Manual inspection of IPA network 1 predicts a potential sub-network associated with histone remodeling. Network 2 predicts two sub-networks associated with sterol metabolism and triglyceride metabolism, respectively. Network 4 predicts a sub-network associated with activation of RNA polymerase or RNA processing and Network 5 predicts two possible sub-networks associated with neuronal voltage-activated calcium channel ion-pore structure or regulation, and synaptic neurotransmitter vesicle formation/exocytosis.

Database for Annotation, Visualization, and Integrated Discovery Bioinformatics Resources were used to determine enrichment of GO terms and provide functional annotation clustering with groups, or clusters, of functionally related GO and KEGG terms ranked by an enrichment score (Huang da et al., 2009b,a). The DAVID analyses were stratified into: (i) signals that are downregulated in PGRN knockdown cells versus PGRN overexpressing cells; (ii) signals that are down regulated in PGRN knockdown cells with respect both to PGRN overexpressing cells and control cells (“commons”); (iii) the “neuronal commons,” that is genes extracted from the “common” subset that have a well-defined CNS function assessed by manually surveying their entries in the Uniprot, Genecards and OMIM databases. 657 genes in total go down in the [shPGRN vs hPGRN] gene set. The common gene set [shPGRN vs CTL/hPGRN] contains 264 molecules (40% of all down-regulated genes in the [shPGRN vs hPGRN] gene set). There are 55 genes in the “neuronal commons” gene subset, representing one fifth (20.8%) of the common gene set. Predicted functional annotation clustering is given in Figure 2A for clusters with an enrichment score of one or above. The top 10 most enriched individual GO-terms are given in Figure 2B. Both the cluster analysis and individual GO-terms indicate effects on synaptic function in the NSC-34/shPGRN cells (compare with IPA network 5, Supplementary Figure 3), associated, in particular, with synaptic vesicle cycling and SNARE interactions in vesicular transport and neuron projection. Functional annotation clustering (Figure 2A) and the list of top individual GO terms (Figure 2B) also show a strong signal for sterol biosynthetic pathways (compare with IPA network 2, Supplementary Figure 3).

The top GO-terms and annotation clusters that are upregulated in NSC-34/shPGRN cells relative to both control cells and PGRN over-expressing cells (commons-up) are summarized in Supplementary Table 2. This revealed several signals for neuron-related terms, but, rather than being centered on synaptic structural functions, as was observed among the down-regulated data set (Figure 2), the up-regulated terms are related to semaphorin signaling and extracellular axon guidance pathways. The GO term “membrane” was enriched in both the down-regulated (Figure 2B) and up-regulated transcript data sets (Supplementary Table 2). Among the molecules included in the two “membrane” GO terms (up and down regulated), are many with receptor-like properties (Table 4). Substantially more signals for receptor tyrosine kinase transcripts (AXL, EPHB4, NGFR, and PDGFRB) are up-regulated in NSC-34/shPGRN cells compared to those that are down-regulated (RET). Similarly signals for molecules in the semaphorin-signaling pathway (NRP1, NRP2, PLXNA1, SEMA7A, SEMA6B), Notch-related receptors (NOTCH1 and NOTCH4), and receptors- in the Wnt-signaling pathway (FZD3, FZD7, LGR6) were up-regulated when PGRN was depleted. In comparison, the number of differentially regulated G-protein linked receptor mRNAs in either the up- or the down-regulated groups is similar (5 vs 6). Receptor-like proteins associated with the immune response were down-regulated in NSC-34/shPGRN cells (CD109, CD180, CD40, Fc receptor IgA IgM, CSF2RA, TREML1, and TNFSF8). Although PGRN is a powerful survival agent in NSC-34 cells we did not detect any signals associated with apoptosis or its inhibition in NSC-34/shPGRN cells. This is, however, unsurprising as the conditions chosen for the gene expression analyses were not associated with extensive cell death.

The reduced expression of genes involved in the neuronal cytoskeleton and neurotransmission in cells with depleted PGRN expression was confirmed by RT-PCR (Figure 3A). This is consistent with a role for PGRN expression in the maintenance of a differentiated state in NSC-34 cells (Figure 3A). Interestingly, there was often little difference in mRNA expression between control and NSC-34/hPGRN cells. We tested specifically for active growth cone formation using growth associated protein 43 (GAP-43, also called neuromodulin) (Maier et al., 2013), and, for evidence of cholinergic differentiation using vesicular acetylcholine transporter (SLC18A3) and choline acetyl transferase (ChAT) as markers (Figure 3B). Expression of GAP-43, and SLC18A3 cells were unchanged between control and NSC-34/hPGRN cells but were decreased in NSC-34/shPGRN, further supporting a role for PGRN in supporting neuron-like structural differentiation in NSC-34 cells. In contrast ChAT mRNA levels, were unaffected by PGRN depletion. Increased expression of neurotrophic receptor transcripts in NSC-34/shPGRN cells was confirmed, with strong signals observed for NGFR (nerve growth factor receptor) and FZD3 (Frizzled 3, a wnt receptor) (Figure 3C). A concerted change in genes associated with lipid metabolism was confirmed (Figure 3D) with most genes downregulated in NSC-34/shPGRN cells compared to control or NSC-34/hPGRN cells as predicted. Not all lipid metabolism genes were downregulated, however, as mRNA for ceramide synthase 2 (CER2) was strongly upregulated.

Neither IPA nor DAVID predicted a strong lysosomal response to PGRN depletion even though deletion of PGRN results in altered expression of lysosomal genes and proteins in other cell lines and Grn knockout mice (Tanaka et al., 2017; Huang et al., 2020). To evaluate whether PGRN depletion alters lysosome gene expression in NSC-34 cells we measured the relative mRNA levels of the lysosome proteins lysosomal associated membrane protein 1 (LAMP1) and cathepsin D (CTSD). There was no indication of increased expression of these genes in NSC-34/shPGRN cells (Figure 3E), and an apparent decrease in expression of cathepsin D, which implies a lack of extensive lysosome biogenesis upon partial PGRN depletion in this model.

Since the transcriptomic data suggested that lipid metabolic pathways are sensitive to PGRN depletion, we undertook mass spectrometric lipidomic analyses on NSC-34/shPGRN and NSC-34/hPGRN cells. Principal component analysis confirmed that all 3 replicates from the same cell construct segregate separately from the other constructs (Supplementary Figure 4). High or low PGRN expression had no effect on the levels of the quantitatively dominant lipid groups (Figure 4A). In contrast, three lipid classes that are quantitatively less abundant, phosphatidic acid, ceramides and cholesteryl ester show upregulation of specific lipid members between high and low PGRN expressors (Figures 4B–D).

The influence of PGRN expression on cellular protein composition was determined comparing NSC-34/hPGRN, NSC-34/shPGRN cells, and control cells, respectively (Supplementary Figure 5). The largest variability was detected between PGRN enriched NSC-34/hPGRN cells and NSC-34/shPGRN knockdown cells. Enrichment of GO and Kegg terms identified functional annotation clusters associated with cell adhesion, and metabolism (Figures 5A,B). Functional annotation clusters (Figure 5C) and enriched GO terms (Figure 5D) for proteins that were overexpressed in NSC-34/shPGRN cells were linked to cell adhesion, RNA binding, RNA processing and metabolism.

As the observed PGRN-dependent phenotypes in NSC-34 cells are reproduced by extracellular PGRN (Figures 1D,E), implying an action on extracellular membrane proteins (Chitramuthu et al., 2017a; Paushter et al., 2018), we therefore extended the proteomic analysis to probe specifically for potential PGRN membrane binding proteins using a LRC proteomic strategy (Frei et al., 2012, 2013; Figure 6A). In total 1746 glycopeptides were identified and quantified. Of these, the cut-off criterion employed for inclusion as putative PGRN-binders was a log2 fold binding enrichment > 0.6 (which is equivalent to an approximately 1.5-fold change in absolute values), and a p value < 0.02 (Supplementary Table 3). This yields 33 proteins. Most of the putative PGRN-binding proteins are associated with the plasma membrane, ER, or lysosome membranes (Figure 6B). Applying STRING analysis to this data set generated a potential interaction network of candidate lysosomal PGRN binding proteins that includes sortilin (SORT1, log2 fold enrichment 1.00 or 2 in natural numbers) and prosaposin (PSAP, log2 enrichment 0.716 or 1.64 in natural numbers) (Figure 6C). GO analysis of biological processes gave the highest score for lysosomal transport (Figure 6D). The presence of ER/Golgi/lysosome proteins is likely to be due to their exposure on the cell surface rather than uptake of labeled PGRN into the cells as the binding experiments were all conducted at 4°C. The highest scoring plasma membrane protein (Supplementary Table 3) was the tetraspanin TSPAN4 (log2 fold enhancement 3.96 or 15.5-fold in natural numbers). A second tetraspanin, TSPAN31 (log2 fold enhancement 1.86 or 3.6-fold in natural numbers), was also captured. Two members of the RTK-family, RET (log2 fold enrichment 1.25 or 2.4-fold in natural numbers), which is the kinase subunit of the glial cell-derived nerve growth factor (GDNF) receptor, and PTK7 (protein tyrosine kinase 7, log2 fold enrichment 0.942875 or 1.92-fold in natural numbers), which is an inactive RTK homolog. In the insulin-binding control arm, multiple peptides for insulin (INSR) were identified (log2 fold enrichment 2.23 or 4.69 in natural values).

Receptor phosphotyrosine antibody arrays were employed to screen for RTKs that are sensitive to PGRN expression levels (Figure 7A). RET, PDGFR-alpha (Platelet derived growth factor -alpha), Mer, and weakly, TrkA (high affinity nerve growth factor receptor), MusK (Muscle-Specific Kinase), and EphA3 (EPH Receptor A3) were more highly tyrosine phosphorylated in NSC-34/hPGRN cells than in vector controls or NSC-34/shPGRN knockdown cells (Figure 7B). Other reports, using the same phosphotyrosine receptor array screening technique, but analyzing non-neuronal cells, revealed that EphA2 was the likely RTK target for PGRN (Neill et al., 2016). Given that EphA3 phosphorylation is only weakly enhanced by PGRN expression in NSC-343 cells, we tested PGRN in MDA-MD-468 mammary epithelial cells and detected phosphorylation of the insulin receptor, the epidermal growth factor (EGF) receptor, and ErbB2 (an EGF-receptor family member), as well as members of the EphA family of receptors (Figure 7A). PGRN expression, therefore, activates different sets of tyrosine kinase receptors in different cell types implying cell context-dependent signaling. The antibody array results were validated by Western blot analyses using phospho-receptor specific antibodies (Figures 7C,D). RET expression is greatly reduced in NSC-34/shPGRN cells, suggesting that at least some of the differences observed in receptor RET tyrosine phosphorylation reflect differential expression of the RET protein, as was also predicted from the mRNA microarray analyses.