Identification of phosphatases that dephosphorylate the co-chaperone BAG3

Abstract

The co-chaperone BAG3 plays critical roles in maintaining cellular proteostasis. It associates with 14-3-3 proteins during the trafficking of aggregation-prone proteins and facilitates their degradation through chaperone-assisted selective autophagy in cooperation with small heat shock proteins. Although reversible phosphorylation regulates BAG3 function, the involved phosphatases remain unknown. Here, we used affinity purification mass spectrometry to identify phosphatases that target BAG3. Of the hits, we evaluated the involvement of protein phosphatase-1 (PP1) using chemical inhibitors and activators in in vitro and cellular approaches. Our results demonstrate that PP1 can dephosphorylate BAG3-pS136 in cells and counteract 14-3-3γ association with BAG3 at this motif. Furthermore, protein phosphatase-5 (PP5) co-enriched with proteostasis-related proteins, and it has the capacity to dephosphorylate a BAG3 phosphorylation-site cluster regulating the interaction of BAG3 with small heat shock proteins and BAG3-mediated protein degradation. Our findings provide new insights into the regulation of BAG3 by phosphatases. This paves the way for future research focused on the precise control of BAG3 function through its regulatory proteins, potentially holding therapeutic promise for diseases characterized by disrupted proteostasis.

Graphical abstract

Figure 1.. BAG3 is a part of the protein maintenance network involving phosphatases.

(A) Linear illustration of human BAG3 protein sequence with reported p-sites annotated. (B) Co-IP of FLAG-BAG3 from HEK293 cells revealed significantly enriched proteins, assessed using the rankprod method (n = 4), with a minimum fivefold increase and a P-value ≤ 0.05. Enriched proteins in the sample are marked in respective colors for subcategories. Selected hits are annotated with their gene name. A false discovery rate of 5% is indicated by a dashed line. (C) Enriched phosphatases and respective phosphatase regulators from the BAG3 co-IP sample were assigned to their respective (super-)families of human phosphatases. The phosphoprotein phosphatase family is colored; other Ser/Thr-specific phosphatases are in shades of gray. Gene names are listed next to the corresponding assigned phosphatase identifications. PPMs, metal-dependent protein phosphatases; PSTPs, protein serine/threonine-specific phosphatases. (A, D) Gene ontology enrichment analysis of significantly enriched genes from the BAG3 co-IP sample (n = 277) in (A). The 10 most abundant biological process GO terms are displayed as bars, ranked based on protein counts. Biological processes related to BAG3’s role in protein homeostasis are presented in blue. (E) Gene ontology enrichment analysis of the co-IP sample compared with overall human gene expression, displayed as a heatmap. Fold enrichment was calculated for the BAG3 enriched, not enriched, and total co-IP sets, with overall GO categories annotated at the respective cluster. Statistical significance was determined using Fisher’s test (P-value < 0.05) with the Bonferroni correction implemented in the PANTHER database. Non-significant changes are colored gray within the heatmap.

Figure S1.. (Related to Fig 1). Sequence conservation of BAG3 orthologs in vertebrates.

(A) Illustration of the used FLAG-control construct to determine unspecific binding and the N-terminally triple-FLAG–tagged human BAG3 used as the bait for the FLAG co-IP. (B) A schematic overview displays annotated BAG3 domains for orientation. BAG3 sequences available from public databases (377 species, all vertebrates) were aligned by geneious using the MUSCLE algorithm. A heat map representation of the alignment identity score was generated to visualize the sequence conservation of BAG3. Regions in the alignment corresponding to gaps in the reference (BAG3 of homo sapiens) were excluded. (C) The multiple sequence alignment for the p-site stretches of interest in this study are underlined and represented in more detail using sequence logos (see (C)). (B, C) Phylogenetic analysis of the multiple sequence alignment displayed in (B). A consensus tree was built with Jukes-Contor model and Neighbor-Joining method after 200 bootstrap replications. The consensus tree for the 377 BAG3 orthologs reveals three primary clusters, corresponding to Teleostei, Sarcopterygii, and mammals. Sequence logos display the context of BAG3-pS136 and the p-site cluster (pS284-pS291) phosphorylation sites for these three principal orthologous groups. Phosphorylation sites are not conserved across all vertebrate classes, but they are conserved in mammals. In the tree, the BAG3 positions of mouse, rat, and human are indicated by black triangles, arranged from left to right.

Figure S2.. (Related to Figs 2 and 5). Immunoblot and quantitative analysis of BAG3 expression in HEK293 and A7r5, and phosphopeptides used for kinetic studies.

(A) HEK293 cells and A7r5 smooth muscle cells were lysed, and protein levels of BAG3 were assessed in relation to the housekeeping protein GAPDH using immunoblots. Quantitative analysis of BAG3 expression validated elevated expression in A7r5 cells. Statistical significance was evaluated through t tests, incorporating data from eight independent replicates. (B) Alignment of human, mouse, and rat BAG3 orthologs to visualize the conservation of the p-sites of interest. Functional p-site studies were carried out in these three organisms in preliminary studies and in rat and human in this work. (C) Summary of the phosphopeptides used in enzyme kinetic studies for the assessment of PP1c and PP5c-mediated dephosphorylation of BAG3 mimetics. (D) Monitoring of CDC37-pSer13 dephosphorylation in A7r5 lysate upon incubation with PP5, for the indicated incubation time to validate enzymatic activity. Endogenous phosphatases were inhibited by the addition of 20 nM CalA. Quantification depicts the results of five independent experiments with adjusted P-values obtained from one-way ANOVA with Sidak correction (***P = 0.0004), ****P < 0.0001). Mean is shown and error bars represent the SD. Source data are available for this figure.

Figure 2.. PP1c dephosphorylates BAG3-pS136 in vitro.

(A) A7r5 smooth muscle cells were treated with increasing concentrations of phosphoprotein phosphatase inhibitors CalA or OA as indicated, or DMSO as control. After 20 min of incubation, cells were lysed, and modulator effects on BAG3-pS136 were determined with immunoblots. The quantification is presented as a bar plot, with the mean depicted with error bars that represent the SEM based on five independent experiments. Statistical significance between inhibitors was determined with t test and P-values are displayed as stars (*P = 0.020, ***P < 0.001). (B, C) Michaelis–Menten kinetic parameters of PP1c were determined against the indicated mono- or bisphosphorylated peptides. Error bars represent the SEM of three independent replicates with technical duplicates. k_cat/K_m was calculated by comparison with a phosphate standard curve. (D) Overexpressed FLAG-BAG3 from transiently transfected HEK293 cells was single-step affinity immobilized with anti-FLAG beads and treated with PP1c for 30 min. Immunoblot quantification depicts the mean of four independent experiments, the error bar represents the SD, with P-values obtained from a t tests (two-tailed, paired, ****P < 0.001). (E) Monitoring of BAG3-pS136 dephosphorylation in A7r5 lysate upon incubation with PP1c, PP5, or without phosphatase as a control treatment for the indicated incubation time. Endogenous phosphatases were inhibited by the addition of 20 nM CalA. Quantification depicts the results of four independent experiments with P-values obtained from two-way ANOVA with Sidak correction (**P = 0.006 [PP1c], **P = 0.005 [PP5], ****P < 0.001). Mean is shown and error bars represent the SEM. Source data are available for this figure.

Figure 3.. PP1 dephosphorylates BAG3-pS136 in cells.

(A) Illustrative overview of the small molecule and peptide modulator screen to tune PP1 activity in cells. HEK293 cells transiently expressing FLAG-BAG3 were treated as illustrated, lysed, and captured by a single-step affinity enrichment using anti-FLAG beads. Samples were analyzed by immunoblots. Quantification depicts results of three independent experiments with P-values obtained from t tests (two-tailed). Mean is shown with replicates as scatter plot and error bars represent the SD (*P = 0.015, ***P < 0.001). (B) A7r5 smooth muscle cells were incubated with increasing concentrations of Tautomycetin as indicated. Immunoblots of lysate were used to determine the sensitivity of the titrated inhibitor towards BAG3-pS136. Mean is shown as a barplot with replicates as scatter plots, and error bars represent the SD of four independent experiments. Statistical significance between concentrations is determined with t test with Welch’s correction (*P < 0.05). (C) Analysis of phosphorylation-dependent binding of 14-3-3γ to FLAG-BAG3 in A7r5 muscle cells after 20-min modulation of endogenous PP1 before lysis and FLAG-IP enrichment. Results are presented as a barplot, P-values obtained from a t test (***P = 0.0002, ****P < 0.0001). (D) Immunoblots depicting siRNA-mediated knockdown of all PP1 isoforms combined (PP1α/β/γ) or individual PP1 isoforms (PP1α, PP1β, PP1γ) separately. The displayed blots represent the analyzed protein levels of 24 or 48 h after transfection. (E) Quantification of immunoblotted lysates show level changes of PP1α/β/γ and BAG3 (n = 3 or 4). Quantification depicts results of three or four independent experiments with P-values obtained from t tests (two-tailed, paired). Error bars represent the SD. Source data are available for this figure.

Figure S3.. Knockdown of PP1c isoforms in HEK293 cells. (Related to Fig 3).

(A) Schematic illustration of the protein level assessment after treatment of HEK293 cells with pre-designed siRNA. HEK293 cells were transfected with siRNA of PP1 isoforms (PPP1CA/B/C) for 24 and 48 h. (B) Quantification of immunoblotted lysates show phosphorylation level changes of BAG3-pS136 relative to the overall BAG3 levels (n = 3 or 4). Quantification depicts results of three or four independent experiments with P-values obtained from t tests (two-tailed, paired). Error bars represent the SD. Source data are available for this figure.

Figure S4.. (Related to Fig 4). FLAG-PP5 co-IP and co-enriched network.

(A) Schematic illustration depicting the N-terminally triple-FLAG–tagged human PP5 construct and the control construct used to identify co-enriched proteins. (B) Depiction of the WPIN (Whole Protein Interaction Network) of 235 co-enriched proteins in the FLAG-PP5 co-IP clustered with the MCODE plugin resulting in 14 clusters (C1–C14). Proteins illustrated as blue nodes and annotated with their gene name. Connecting lines (edges) between nodes display interactions; the color depicted the confidence score, with darker lines indicating higher confidence scores. (Source data in Table S2).

Figure 4.. PP5 is part of the protein maintenance network.

(A) Co-IP of FLAG-PP5 from HEK293 cells with significantly enriched proteins highlighted. Enriched proteins in the sample are marked in the respective colors for the subgroups. Enrichment was determined through rankprod (n = 4). Selected hits are annotated with their gene name. Gene names of proteins of interest in this work are highlighted in bold. Dashed line indicates a false discovery rate of 5%. (A, B) Gene ontology analysis of the significantly enriched genes from the PP5 co-IP sample (n = 235) from (A). The 10 most abundant biological process GO terms are shown as bar charts, ranked based on the protein counts. Biological processes related to protein folding and stabilization are illustrated in red. (A, C) Enrichment analysis of enriched genes from the co-IP experiment (A) compared with overall human gene expression displayed as a heatmap. Fold enrichment was calculated for the enriched, not enriched, and total sample, with overall GO categories annotated at the respective cluster. Statistical significance was determined using Fisher’s test (P-value < 0.05) with the Bonferroni correction implemented in the PANTHER database. Non-significant changes are colored as gray within the heatmap. (D) Cluster 1 of PP5 co-enriched proteins, with key characteristics visually represented through color-coding and labeled with their respective gene name. Proteins are presented as nodes: darker blue hues signify more interactions within the cluster. Node size reflects the fold change in the co-IP, with larger nodes representing higher fold change values. Black encirclement of nodes indicates the GO biological process “cellular response to stress.” Connecting lines (edges) between nodes display interactions; the color depicted the confidence score, with darker lines indicating higher confidence scores.

Figure 5.. PP5 dephosphorylates a p-site cluster in BAG3 and enables HspB8 binding.

(A) Phosphorylation levels of BAG3-pS136 were compared in HEK293 cells overexpressing FLAG-BAG3 or both FLAG-BAG3 and FLAG-PP5 for 24 h. Immunoblot quantification demonstrates that PP5 overexpression does not affect BAG3-pS136 phosphorylation. Statistical significance was determined through t test (n = 4). (B) Michaelis–Menten kinetic parameters of PP5c were determined against the indicated mono- or bisphosphorylated peptides. Error bars represent the SEM of three independent repeats with technical duplicates. k_cat/K_m was calculated by comparison with a phosphate standard curve. (C) HEK293 cells overexpressed FLAG-BAG3 and were treated with 1 μM recombinant PP5 after lysis for 1 h (in vitro), or overexpressed FLAG-BAG3 and FLAG-PP5 simultaneously for 24 h (in-cell), or were not treated with phosphatase (Ctrl). FLAG-BAG3 was subjected to affinity enrichment and subsequently evaluated using LC-MS/MS. The theoretical phosphopeptide generated through trypsinization is shown at the top. Targeted phosphoproteomics quantification reveals the complete dephosphorylation of the p-site cluster for both PP5 treatments. Statistical significance was determined through two-way ANOVA (n = 4), error bars represent the SD (****P < 0.0001, ***P = 0.0002, *P = 0.0021). Indexed retention time (iRT). (D) FLAG-BAG3 mutants mimicking the phosphorylated (4D) and dephosphorylated (4A) state of the BAG3 p-site cluster were overexpressed in HEK293 cells followed by co-IP and MS analysis to assess PP5 binding to each variant compared with WT. Bars represent the mean (n = 3), with significance and error bars calculated using t test (*P = 0.03). (C, E) Samples were subjected to the same treatment as described for (C) to induce dephosphorylation of p-site cluster. This was followed by the analysis of protein quantities bound to BAG3 via immunoblotting, normalized to the amounts bound to the control sample. Statistical significance of changes in protein levels was assessed using data from seven or nine independent replicates, with means presented as bar plots. P-values were calculated using a one-way ANOVA with a 95% confidence interval and analyzed using PRISM/GraphPad version 6 (HspB8: *P = 0.0188; **P = 0.0002; 14-3-3: P [in vitro] = 0.1668, P [in-cell] = 0.6241). (F) A7r5 cells were transfected with PPP5C siRNA for the specified incubation time and protein levels were determined using immunoblots. Statistical significance in the changes of protein levels was determined based on data from five independent replicates, and the means are presented as bar plots. The P-value was calculated using a two-tailed test with a 95% confidence interval and determined using PRISM/GraphPad version 6 (PP5: ***P = 0.001, **P = 0.0045, ***P = 0.0006; BAG3: **P = 0.0097, *P = 0.046, *P = 0.0328). (G) Smooth muscle cells were subjected to treatment with control siRNA or PP5-targeting siRNA (siPP5) followed by treatment with BafA1 to inhibit CASA mediated BAG3 degradation. Statistical significance of changes in protein levels was assessed using data from seven independent replicates, represented as boxplots. P-values were calculated using a two-way ANOVA and analyzed using PRISM/GraphPad version 6. Source data are available for this figure.

Figure S5.. (Related to Fig 5): Supplementary data to PP5 treatments.

(A) Schematic representation of the experimental procedure used for the dephosphorylation analysis of the p-site cluster of FLAG-BAG3 derived from HEK293 cells. Cells were either overexpressing FLAG-PP5 or subjected to in vitro PP5 incubation. This workflow served to decipher the dephosphorylation of the BAG3 p-site cluster (LC-MS/MS; Fig 5A). (B) HEK293 cells overexpressing FLAG-BAG3 were subjected to various modulations of PP5 activity including PP5 co-overexpression for 24 h, activation using arachidonic acid (2 h, 250 μM), or in vitro incubation with recombinant PP5 (1 h, 1 μM, 30°C) in lysate or on-beads (30 min, 1 μM PP5, 30°C). LC-MS/MS was used to quantitatively assess the dephosphorylation status of individual p-sites within the p-site cluster. (C) Related to Fig 5E. Immunoblots of lysates from BAG3-overexpressing HEK293 cells that were treated with PP5 to ensure equal amounts of proteins were present. (D) Related to Fig 5E. Immunoblots of FLAG-BAG3 co-IP samples from HEK293 cells subjected to dephosphorylation treatments with PP5. (E) A7r5 cells were transfected with PPP5C siRNA for the specified incubation time and protein levels were determined using immunoblots. Statistical significance in the changes of protein levels was determined based on data from five independent replicates, and the means are presented as bar plots. The nonparametric Spearman correlation between BAG3 and PP5 upon PPP5C siRNA treatment verifies the negative correlation between the protein levels. (F) Smooth muscle cells were subjected to treatment with control siRNA or PP5-targeting siRNA (siPP5) followed by treatment with BafA1 to inhibit CASA-mediated degradation. Statistical significance of changes in protein levels of SQSTM1 was assessed using data from seven independent replicates, represented as boxplots. P-values were calculated using a two-way ANOVA and analyzed using PRISM/GraphPad version 6. Source data are available for this figure.

Figure 6.. Regulation of BAG3 protein interactions through PP1 and PP5.

Dephosphorylation of pS136 on BAG3 by PP1 leads to loss of 14-3-3 protein binding. Dephosphorylation of the p-site cluster pS284-pS291 by PP5 enables HspB8 binding, and PP5 depletion increases BAG3 levels in a CASA-dependent manner.