Multi-modal proteomics captures holistic lysosomal biology

Lysosomes play critical roles in neurons, such as degradation, endocytosis, signal transduction, nutrient sensing, and long-distance trafficking through axons [44,45,46]. Different methods of characterizing lysosomal composition and interactions now exist, each with its own strengths [33, 47, 48]. However, a comprehensive characterization of lysosomal biology in neurons with these modern tools has not been performed. Here, we developed three complementary proteomic strategies to characterize lysosomal contents and dynamic lysosomal interactions in both living human neurons and fixed mouse brain tissues (Fig. 1A). Lysosome proximity labeling using ascorbate peroxidase (Lyso-APEX) captured lysosome membrane proteins and lysosome interactions in i3Neurons. Rapid lysosomal immunopurification (Lyso-IP) provided both lysosome lumen and membrane proteins from i3Neuron. Lysosomal biotinylation by antibody recognition (Lyso-BAR) revealed lysosome proteins and interactions in situ from fixed mouse brain tissues. The proper locations of these probes were validated by immunofluorescence and western blotting (Fig. 1B and Supplementary Figure S1). Control groups were carefully selected for each probe to reduce nonspecific labeling and ensure intracellular spatial specificity (Fig. 1C).

Fig. 1

A map of the lysosomal proteome and interactome in human neurons and mouse brains. A Schematics of lysosomal proximity labeling (Lyso-APEX) in i3Neurons, lysosomal immunopurification (Lyso-IP) in i3Neurons, and lysosomal biotinylation by antibody recognition (Lyso-BAR) in fixed mouse brain tissues. B Fluorescence imaging of Lyso-APEX, Lyso-IP, and Lyso-BAR activities in i3Neurons and fixed mouse brains. Biotinylated proteins, stained with streptavidin (SA-488), colocalize with lysosomal markers in i3Neurons and fixed mouse brain tissues. HA-tagged lysosomes colocalize with lysosomal markers in i3Neurons. Nuclei were stained by Hoechst. Scale bars are 10 μm. C Volcano plots showing significantly enriched proteins from WT Lyso-APEX compared to cytosolic-APEX, Lyso-IP compared to control neurons without HA-LAMP1 expression, and Lyso-BAR compared to control mouse brains without LAMP1 primary antibody staining (N = 4 for each group). Dotted lines denote corrected p-value of 0.05 (y-axis) and ratio of 1.5 (x-axis). The known lysosomal membrane and lumen proteins are highlighted in blue and orange colors, respectively. D GO enrichment analyses of significantly enriched proteins in Lyso-APEX, Lyso-IP, and Lyso-BAR proteomics. E Venn diagram comparison of significantly enriched proteins in three proteomics methods. F Radar plot comparison of three methods regarding the proteome coverage (lysosome lumen, membrane, and interaction), tag expression level, sensitivity, specificity, and robustness

Lyso-APEX, Lyso-IP, and Lyso-BAR proteomics provided complementary coverage of the lysosomal microenvironment in human neurons and mouse brains (Fig. 1D, E, F). Lysosomal membrane proteins such as vacuolar ATPase (v-ATPase) subunits [49], LAMP proteins, and Ragulator subunits [50] are identified and enriched by all three probes (Supplemental Figure S1). Lysosomal lumen proteins, especially hydrolases, are highly enriched in Lyso-IP proteomics, consistent with the degradative nature of the isolated organelles. Besides lysosome-resident proteins, both Lyso-APEX and Lyso-BAR proteomics captured dynamic lysosomal interaction partners related to organelle trafficking and axon transport (e.g., Kinesins, MAPs) [44]. Lyso-APEX favored surface-bound and surface-interacting proteins over luminal proteins due to the limited membrane permeability of reactive phenol-biotin generated on the cytosolic face of lysosomes during APEX-mediated labeling (Fig. 1C, D). By contrast, Lyso-BAR revealed more intraluminal lysosomal proteins since BAR activation in fixed brain tissues requires membrane permeabilization. Lyso-BAR proteomics in mouse brain also captured numerous synaptic proteins, likely due to enhanced synaptic maturation in vivo compared to cultured iPSC-derived i3Neurons (Supplemental Figure S1). Collectively, combining Lyso-APEX, Lyso-IP, and Lyso-BAR proteomic strategies allows us to obtain comprehensive lysosomal lumen and membrane proteomes as well as lysosomal interactions in both cultured human i3Neurons and fixed mouse brains.

Loss of progranulin results in upregulated vacuolar ATPases and elevated lysosomal pH in human neurons

Equipped with these new tools, we characterized how progranulin loss altered lysosomal biology. Using CRISPR-Cas9, we knocked out the GRN gene in wild-type (WT) iPSCs harboring the Lyso-APEX probe and differentiated them into cortical neurons (Fig. 2A). Immunofluorescence microscopy showed that the PGRN protein colocalizes with lysosomes in WT i3Neurons, and that no PGRN signal was observed in GRN KO i3Neurons (Fig. 2B). Using Lyso-APEX proteomics, we found that PGRN depletion altered the abundance of many lysosome membrane proteins and lysosome interaction partners in human neurons (Fig. 2C). Gene Ontology (GO) enrichment analysis revealed upregulation of proteins related to lysosomal acidification and autophagy, including numerous v-ATPases (ATP6Vs) and chloride channel proteins (CLCNs) (Fig. 2D, E, Supplementary Figure S2A) [51]. GO enrichment analysis of significantly downregulated proteins indicated impaired lysosomal transport and RNA processing (Supplementary Figure S2B). Several lysosome proteins (Atp6v1g2, Clcn6, Tecpr1, Pld3) were increased in Lyso-APEX but decreased in cytosolic-APEX, indicating their potential translocation from the cytosol to the lysosome in PGRN-deficient neurons (Supplementary Figure S2C, D). Given the centrality of v-ATPases in establishing the acidic lysosomal lumen pH and the strong upregulation of acidification-related proteins in PGRN deficiency, we hypothesized that lysosomal pH could be perturbed by the loss of PGRN inside the neuronal lysosome.

Fig. 2

Progranulin-null human neurons have altered lysosomal membrane and interacting proteins and elevated lysosome pH. A Schematic of Lyso-APEX in WT and isogenic GRN KO i3Neurons. B Fluorescence imaging showing the colocalization of PGRN with lysosomes in WT i3Neurons and loss of PGRN signal in GRN KO i3Neurons. Scale bar is 10 μm. C Volcano plot of Lyso-APEX proteomics in GRN KO vs. WT i3Neurons (N = 4 for each group). Cytosolic enriched proteins and nonspecific labeling were removed from the volcano plot based on WT Lyso-APEX vs. Cytosolic-APEX comparison. Red and blue colored proteins belong to lysosomal pH and protein transport GO-terms, respectively. D GO enrichment analysis of significantly upregulated biological processes in GRN KO vs. WT Lyso-APEX proteomics. E Protein network analysis of increased vacuolar-ATPase subunits and their interactors in PGRN-null neurons. F Live cell ratiometric lysosome pH assay. pH calibration curve is generated based on the ratio of pH-sensitive Oregon Green-488 dextran signal and pH-insensitive/loading control Alexa Fluor-555 red dextran in WT i3Neurons. Scale bar is 10 μm. Other linear and nonlinear curve fitting models are provided in Supplementary Figure S2E. G Lysosome pH measurements in WT vs. GRN KO i.3Neurons; three independent experiments are represented with different shapes, each with 5–10 wells of neuron culture replicates (**** denotes p-value < 0.0001)

To measure neuronal lysosomal acidification, we used a ratiometric fluorescent dextran assay. We co-generated an in-situ calibration curve using buffers of known pH, allowing accurate calculations of absolute pH within the lysosome with both nonlinear and linear curve fitting models (Fig. 2F, Supplementary Figure S2E). Lysosomal pH is significantly increased in GRN KO i3Neurons (4.81 ± 0.24) compared to WT i3Neurons (4.31 ± 0.16). While this difference in pH may seem like a subtle change, it equates to a nearly three-fold decrease in the concentration of protons in the lysosomal compartment of GRN KO i3Neurons compared to WT counterparts due to the logarithmic nature of the pH scale ([H+] in WT ≈ 52 ± 19 μM, GRN KO ≈ 18 ± 9 μM). These observations show that GRN KO i3Neurons have alkalinized lysosomes, which could trigger the upregulation of acidification machinery to compensate for this effect.

Progranulin-null lysosomes contain increased hydrolases levels but have decreased enzymatic activity

Lysosomes require acidic luminal pH to degrade proteins using hydrolases [1]. Since lysosomes from progranulin-null neurons are less acidic, we hypothesized that these lysosomes may have altered abundances or activity of pH-dependent hydrolases. Using Lyso-IP proteomics, we characterized lysosome composition in GRN KO vs. WT i3Neurons (Fig. 3A). PGRN protein was confirmed to be enriched in isolated lysosomes and absent in GRN KO i3Neurons (Supplementary Figure S3A). Proteins involved in catabolism and lysosomal acidification were significantly increased in PGRN-deficient lysosomes in human neurons (Fig. 3B, C, Supplementary Figure S3C). To investigate the impact of progranulin deficiency on lysosomes in mouse brain, we conducted Lyso-BAR proteomics in GRN−/− vs. WT fixed mouse brains (Fig. 3D). Similar protein catabolic processes were upregulated in GRN−/− mice (Fig. 3E, F, Supplementary Figure S3E). Because Lyso-IP and Lyso-BAR probes have higher variations compared to Lyso-APEX probe (Fig. 1F), we conducted western blotting and targeted parallel reaction monitoring (PRM) assay to specifically quantify v-ATPases and cathepsin proteases in Lyso-IP and Lyso-BAR samples (Fig. 3G, H). V-ATPases and cathepsins showed consistent upregulation in lysosomes from PGRN-deficient i3Neurons and mouse brains.

Fig. 3

Progranulin-null lysosomes from human neurons and mouse brains contain increased hydrolases levels but have decreased enzymatic activity. A Schematic of intact lysosomal isolation (Lyso-IP) proteomics in GRN KO vs. WT i3Neurons. B Volcano plot of Lyso-IP proteomics showing protein changes related to protein catabolic processes (red), lysosomal pH (blue), and hydrolase activities (green). Nonspecific labeling proteins were removed from the plot based on WT Lyso-IP vs. control i3Neurons without HA expression. C GO enrichment analysis of significantly changed proteins in GRN KO vs. WT Lyso-IP proteomics (left: biological processes; right: molecular functions). Color code corresponds to the volcano plot. D Schematic of mouse brain Lyso-BAR labeling in GRN−/− vs. WT mice. E Volcano plot showing Lyso-BAR protein changes in GRN−/− vs. WT mouse brains after removing nonspecific labeling proteins. F GO enrichment analysis of significantly changed proteins in GRN −/− vs. WT Lyso-BAR proteomics. (G) Western blot analysis showing elevated V-ATPases and cathepsin D levels from isolated GRN KO vs. WT lysosomes in i3Neurons. H Targeted PRM protein quantification of v-ATPases and cathepsins from Lyso-BAR mouse brains. I Fluorescence imaging of Magic Red assay to measure cathepsin B enzymatic activity in i3Neurons. CQ stands for chloroquine treatment (50 μM for 24 h). Scale bar is 10 μm. J Quantification of absolute and relative fluorescence intensities indicate decreased cathepsin B activity in GRN KO vs. WT i3Neurons

Prior studies of GRN−/− mouse models have suggested that cathepsins may be less active in progranulin-null cells despite increased abundance [30, 52, 53]. To directly evaluate the impact of progranulin depletion on lysosomal activity in human neurons, we quantified cathepsin B activity using a Magic Red assay in living WT and GRN KO i3Neurons. We observed a significant decrease in cathepsin B activity in PGRN-null i3Neurons compared to WT, indicating impaired proteolytic function (Fig. 3I, J, Supplementary Figure S3B). To mimic alkalinization-related phenotypes observed in GRN KO i3Neurons, we treated neurons with chloroquine, an agent that neutralizes lysosomal pH. As predicted, direct alkalinization of lysosomes with chloroquine treatment reduced Magic Red fluorescence (Fig. 3I). These findings confirm that although lysosomal hydrolases were upregulated in the absence of progranulin, their activity was decreased, likely due to alkalinized lysosomal lumens.

Characterizing global protein turnover in human iPSC-derived neurons

Since lysosomes are major proteostatic organelles and their degradative function is impaired in progranulin-depleted neurons, we hypothesized that progranulin deficiency could influence global proteostasis. To measure the global protein turnover in neurons, we designed a dynamic SILAC proteomic method in cultured i3Neurons to calculate neuronal protein half-lives (Fig. 4A). By modeling the peptide degradation curves in WT i3Neurons, we found that most peptide degradations follow first-order exponential decay, consistent with other cell types in prior studies (Fig. 4B, Supplementary Figure S4A) [54]. Peptide level and protein level half-lives correlate well with each other, with a median half-life of around 4 days (Fig. 4C, D and Supplementary Figure S4B, C). Therefore, peptide and protein half-lives can be calculated using a single time point at 4 days (96 h) after heavy medium switch (Supplementary Figure S4D). As we examined the distribution of protein half-lives, we found that numerous histones, nucleoporin proteins (Nups), and inner mitochondrial membrane proteins have extremely long half-lives (> 20 days) in i3Neurons, in agreement with recent studies in primary rodent neurons and brain tissues [55,56,57]. In contrast, proteins related to neurosecretion (GPM6B, VGF), axonal transport (kinesins), and ubiquitination (UBL4, USP11) have very short half-lives (0.3–2 days) (Fig. 4E, Supplemental Figure S4B). Notably, one of the shortest half-life proteins in the entire neuronal proteome was STMN2, a microtubule-binding, Golgi-localized protein implicated in ALS pathogenesis [58, 59]. Lysosomal proteins have a median half-life of 3.6 days, slightly shorter than the median half-life of global neuronal proteins. Further investigation into the lysosomal compartment revealed a median half-life of 7.5 days for degradative enzymes, 3.5 days for v-ATPases, 6.2 days for lysosome-associated membrane glycoproteins (Lamps), 3.5 days for LAMTOR and HOPS complex subunits, and 3.1 days for BLOC1 complex subunits (Fig. 4F). Together, this method enabled us to calculate global protein half-lives in living human i3Neurons for the first time.

Fig. 4

Measuring global protein half-lives in cultured human i3Neurons. A Schematic of dSILAC proteomics to measure global protein half-lives in cultured human i3Neurons. Cortical neurons were grown in normal medium until day 10 and then switched to heavy lysine-containing medium. Neurons are harvested at 1, 2, 4, and 6 days after medium switch followed by bottom-up proteomics. B Time-dependent changes of relative light and heavy protein abundances indicating protein degradation and synthesis processes, respectively, in WT i3Neurons. C Scatter plot of ranked protein half-lives measured in WT i3Neurons. D Histogram distribution of protein half-lives in WT i3Neurons. An example cathepsin B (CTSB) protein with five unique peptide sequences is illustrated in the inset. E GO enrichment analysis of the fast (left) and slow (right) turnover proteins in WT i3Neurons. F Violin plots of half-life distributions from lysosomal proteins in WT i3Neurons

Loss of progranulin alters neuronal protein turnover and decreases lysosomal degradative function

Using our dynamic SILAC proteomics approach, we evaluated protein turnover in WT vs. GRN KO i3Neurons (Fig. 5A). The median of global protein half-lives remained unchanged, but a remarkable 25% of all measured proteins presented significantly altered half-lives in GRN KO vs. WT i3Neurons (corrected p-value < 0.05) (Fig. 5B and C). Because protein half-life measurements have much smaller fold changes compared to protein abundance levels, we did not use any fold change ratio cutoff. Proteins related to polymerization and fiber organization showed significantly slower turnover, which may indicate a propensity for protein misfolding and aggregation in GRN KO neurons related to FTD pathogenesis (Fig. 5D) [60]. Despite the significantly slower turnover of both cathepsin B and cathepsin D, proteins related to RNA catabolic processes showed faster turnover, which further implicates the disturbance of molecular degradation pathways (Fig. 5E). Go enrichment analysis from altered proteins showed enrichment in ALS/FTD and other neurodegenerative diseases, suggesting potential converging pathways among different neurodegenerative diseases and dysfunction of key regulators of proteostasis (Fig. 5F).

Fig. 5

Global protein turnover and lysosomal degradative function are impaired in progranulin-null human neurons. A Schematic of protein half-life measurements in GRN KO vs. WT i3Neurons using dynamic SILAC proteomics. B Histogram distribution of global protein half-lives in GRN KO (blue) vs. WT (green) i3Neurons. C Volcano plot of protein half-life changes in GRN KO vs. WT i3Neurons. D GO enrichment analysis of biological processes from proteins with significantly slower turnover. E GO enrichment analysis of biological processes from proteins with significantly faster turnover. F KEGG pathways enriched from significantly altered protein half-lives in GRN KO vs. WT i3Neurons. G Schematic of the DQ-BSA Red assay to measure lysosomal degradative function. Extracellular DQ-BSA with self-quenched dye is endocytosed into i3Neurons and trafficked to the lysosome, where it is degraded into smaller protein fragments with isolated fluorophores with fluorescence signals. H Representative fluorescence imaging of DQ-BSA Red assay showing DQ-positive lysosomes in i3Neurons. Scale bar is 10 μm. I Quantification of the fluorescence intensities of the DQ-BSA Red assay in WT vs. GRN KO i3Neurons, normalized to the total number of puncta in two groups

Given our observations that lysosomes within GRN KO i3Neurons are alkalinized, have reduced cathepsin activity, and exhibit substantial changes in global protein homeostasis, we predicted that GRN KO lysosomes would exhibit impaired lysosome-mediated protein degradation. We directly assayed lysosomal degradative capacity using a fluorescent DQ-BSA Red assay (Fig. 5G, H) [61, 62]. The DQ-BSA substrate is initially self-quenching due to the close spatial proximity of the fluorophores. Once cleaved in acidic lysosomes, the DQ-BSA substrate exhibits bright fluorescence signals. The mean DQ-BSA intensity in GRN KO i3Neurons was significantly decreased compared to WT neurons (Fig. 5I), similar to pharmacological inhibition of lysosomal degradation using chloroquine (Supplemental Figure S5A). The change in active proteolysis was independent of lysosomal biogenesis, as there was no change in the number of puncta per cell in GRN KO vs. WT (Supplemental Figure S5B). Taken together, these results show that GRN KO lysosomes have significantly hindered proteolytic capacity, consistent with our observations of pathological impairment in lysosomal acidification and impaired lysosomal hydrolase activity.

An isogenic series of FTD patient-derived i3Neurons with deficient PGRN exhibit altered protein homeostasis

To further explore the relationship between GRN insufficiency and protein homeostasis abnormalities, we created i3Neurons from an FTD patient-derived iPSC line with a heterozygous GRN mutation [35] (c.26 C > A, p.A9D; referred subsequently as ptMut), as well as the isogenic iPSC control line with corrected GRN mutation (ptWT). We further knocked out GRN in this control line to create an additional isogenic GRN KO iPSC line (ptKO) (Fig. 6A, Table 1). After differentiating each iPSC line to i3Neurons, performing dSILAC, and measuring protein half-lives, we found that over 25% of proteins showed significantly altered half-lives (corrected p-value < 0.05) in ptKO compared to ptWT i3Neurons (Fig. 6B), consistent with GRN-KO vs. WT comparison in Fig. 5C. About 15% of protein half-lives were significantly altered in ptMut compared to ptWT group (Fig. 6B and C). Principal component analysis and hierarchical clustering showed complete separations of both genetic background and GRN genotypes from five i3Neurons lines (GRN-KO, WT, ptKO, ptMut, ptWT) based on protein half-lives (Fig. 6D, E). The overall protein half-life changes also suggested a potential gene dosage effect, in which many proteins have greater fold changes in GRN-KO neurons compared to GRN-mutant neurons (Fig. 6F, Supplementary Figure S6A). Half-life changes of key overlapping proteins in the three comparisons (GRN-KO vs. WT, ptKO vs. ptWT, ptMut vs. ptWT) are highlighted in Fig. 6G and Supplementary Figure S6B.

Fig. 6

FTD patient-derived i3Neurons with mutant GRN reveal altered protein turnover of lysosomal enzymes and FTD-associated proteins. A Generation of a set of FTD patient fibroblast-derived i3Neurons. First, CRISPR-Cas9 was used to insert an inducible NGN2 cassette into the AAVS1 locus of a patient fibroblast-derived iPSC line (ptMut). Next, CRISPR-Cas9 was used to correct the GRN mutation in ptMut to create an isogenic control iPSC line (ptWT) and then to knockout GRN in pWT to create the ptKO iPSC line. These iPSC lines were then differentiated into i3Neurons and dSILAC proteomics was performed. B Volcano plot of protein half-life changes in ptKO vs. ptWT i3Neuron. C Volcano plot of protein half-life changes in ptMut vs. ptWT i3Neuron. D Principal component analysis using protein half-lives in GRN-KO, WT, ptKO, ptMut, and ptWT i3Neurons groups. E Hierarchical clustering of five i3Neurons groups with five biological replicates in each group. F Scatter plot of protein half-life changes in ptKO vs. ptWT and ptMut vs. ptWT comparisons showing the consistency and potential gene dosage effect of ptKO and ptMut i3Neurons. G Heatmap showing key overlapping protein turnover changes in GRN KO vs. WT, ptKO vs. ptWT, and ptMut vs. ptWT i3Neurons. Heatmap colors represent the absolute half-life differences in days between comparison groups. Key proteins from lysosomes and relevant to FTD/ALS are highlighted in red and blue, respectively. H Schematic of proposed lysosomal impairment in progranulin-deficient neurons caused by GRN mutations in FTD patients

The findings in patient-derived GRN mutant and KO neurons validate our observations of dysregulated protein homeostasis in settings of GRN depletion and insufficiency, including alterations in the half-lives of numerous neurodegeneration-associated proteins. Many lysosomal enzymes showed prolonged protein half-lives, such as cathepsins (CTSD, CTSB), which was especially notable given our direct measurements of increased cathepsin levels and reduced CTSB activity in GRN KO neurons. Our findings additionally show that substantial upregulation of numerous lysosomal-associated proteins and enzymes occurs in GRN-deficient neurons – many via prolongation in protein half-lives – but that these homeostatic changes are insufficient to normalize lysosomal degradative capacity. A most recent study also demonstrated the elevated lysosome pH in FTD patient-derived neurons which agrees well with our findings [63]. As summarized in Fig. 6H, we propose that GRN mutations that cause PGRN deficiency inside neuronal lysosomes result in alkalinized lysosomal pH, decreased proteolytic activities, and impaired global protein homeostasis that eventually lead to frontotemporal dementia.