BAG3 Proteomic Signature under Proteostasis Stress

Abstract

The multifunctional HSP70 co-chaperone BAG3 (BCL-2-associated athanogene 3) represents a key player in the quality control of the cellular proteostasis network. In response to stress, BAG3 specifically targets aggregation-prone proteins to the perinuclear aggresome and promotes their degradation via BAG3-mediated selective macroautophagy. To adapt cellular homeostasis to stress, BAG3 modulates and functions in various cellular processes and signaling pathways. Noteworthy, dysfunction and deregulation of BAG3 and its pathway are pathophysiologically linked to myopathies, cancer, and neurodegenerative disorders. Here, we report a BAG3 proteomic signature under proteostasis stress. To elucidate the dynamic and multifunctional action of BAG3 in response to stress, we established BAG3 interactomes under basal and proteostasis stress conditions by employing affinity purification combined with quantitative mass spectrometry. In addition to the identification of novel potential BAG3 interactors, we defined proteins whose interaction with BAG3 was altered upon stress. By functional annotation and protein-protein interaction enrichment analysis of the identified potential BAG3 interactors, we confirmed the multifunctionality of BAG3 and highlighted its crucial role in diverse cellular signaling pathways and processes, ensuring cellular proteostasis and cell viability. These include protein folding and degradation, gene expression, cytoskeleton dynamics (including cell cycle and transport), as well as granulostasis, in particular.

Keywords: BAG3; autophagy; interactome; protein quality control; proteostasis; stress response.

Figure 1

Methodological workflow to establish BAG3 interactomes under basal and proteostasis stress conditions via affinity purification combined with quantitative mass spectrometry (qAP-MS). Briefly, HEK293T cells were treated either with DMSO or 10 µM MG132 for 6 h. After extraction of proteins, endogenous BAG3 was immunoprecipitated and the eluates were separated by SDS-PAGE. Following reduction, alkylation and in-gel trypsin digestion, peptides were analyzed by LC-MS/MS. Raw files were processed using MaxQuant. Relative label-free quantification (LFQ) was performed with MaxLFQ algorithm integrated into MaxQuant. Statistical data analysis was conducted by Perseus software followed by functional annotation and PPI enrichment analysis via STRING database.

Figure 2

Multivariate statistical analysis of processed BAG3 interactome data set by Perseus software. (A) Principal component analysis (PCA) of LFQ intensities of all proteins identified in the respective samples. Replicates of DMSO_IP (DMSO_IP1-DMSO_IP3) are shown as green open circles, replicates of MG132_IP (MG132_IP1-MG132_IP3) are shown as blue crosses and replicates of the IP control MG132_IgG (MG132_IgG1-MG132_IgG3) are shown as red open squares (Number of components: 2; Cutoff method: p-value = 0.05). (B) Multi scatter plot of LFQ intensities of all proteins identified in DMSO_IP, MG132_IP and MG132_IgG with Pearson’s correlation coefficients (blue numbers). (C) Hierarchical clustering and heat map of LFQ intensities of proteins identified in DMSO_IP, MG132_IP and MG132_IgG. The three groups DMSO_IP, MG132_IP and MG132_IgG including their three biological replicates are clustered in columns and proteins are clustered in rows. Color scale reports log2-transformed LFQ intensity values; blue indicates a high LFQ intensity, yellow marks a low LFQ intensity.

Figure 3

Quantitative proteomic analysis of the BAG3 interactome under basal conditions (DSMO_IP). (A) Scatter plot generated by plotting the log2 ratios against the negative log10 p-values of the Student’s t-test DMSO_IP compared to MG132_IgG (DMSO_MG132IgG; p-value = 0.05; S0 = 0). Proteins with a p-value ≤ 0.05 and a ratio ≥ 1.5 (black dash lines) were considered as significant, labelled with the gene symbols of the respective proteins and are shown as green open circles. BAG3 as bait protein was marked with red. Class A BAG3 interactors with a p-value ≤ 0.01 and a ratio ≥ 2 (grey dash lines) are depicted as light green open circles. (B) Hierarchical clustering and heat map of LFQ intensities of all significant BAG3 interactors (p-value ≤ 0.05 and ratio ≥ 1.5) under basal conditions (DMSO_IP). The three groups DMSO_IP, MG132_IP and MG132_IgG including their three biological replicates are clustered in columns and proteins are clustered in rows. Color scale reports Z-scored log2-transformed LFQ intensity values; blue indicates a high LFQ intensity, yellow marks a low LFQ intensity. (C) Protein-protein interaction (PPI) network of class A BAG3 interactors (p-value ≤ 0.01 and ratio ≥ 2) under basal conditions (DMSO_IP). Network was clustered into 12 subnetworks using the k-means clustering method; nodes/proteins of the same cluster exhibit the same color. Thickness of connecting lines/edges correlates with the strength of the association. Further network parameters: number of nodes: 90, number of edges: 203 (expected: 89), average node degree: 4.51, avg. local clustering coefficient: 0.46, PPI enrichment p-value < 1.0e-16, statistical background: whole genome. Scatter plot including Student’s t-test and hierarchical clustering/heat map were performed by Perseus software and the PPI network was created by the STRING database.

Figure 4

Quantitative proteomic analysis of the BAG3 interactome upon proteasome inhibition (MG132_IP). (A) Scatter plot generated by plotting the log2 ratios against the negative log10 p-values of the Student’s t-test MG132_IP compared to MG132_IgG (MG132_MG132IgG; p-value = 0.05; S0 = 0). Proteins with a p-value ≤ 0.05 and a ratio ≥ 1.5 (black dash lines) were considered as significant, labelled with the gene symbols of the respective proteins and are shown as blue crosses. BAG3 as bait protein was marked with red. Class A BAG3 interactors with a p-value ≤ 0.01 and a ratio ≥ 2 (grey dash lines) are depicted as light blue crosses. (B) Hierarchical clustering and heat map of LFQ intensities of all significant BAG3 interactors (p-value ≤ 0.05 and a ratio ≥ 1.5) upon proteasome inhibition (MG132_IP). The three groups DMSO_IP, MG132_IP and MG132_IgG including their three biological replicates are clustered in columns and proteins are clustered in rows. Color scale reports Z-scored log2-transformed LFQ intensity values; blue indicates a high LFQ intensity, yellow marks a low LFQ intensity. (C) Protein-protein interaction (PPI) network of class A BAG3 interactors (p-value ≤ 0.01 and ratio ≥ 2) upon proteasome inhibition (MG132_IP). Network was clustered into 12 subnetworks using the k-means clustering method; nodes/proteins of the same cluster exhibit the same color. Thickness of connecting lines/edges correlates with the strength of the association. Further network parameters: number of nodes: 319, number of edges: 2160 (expected: 1216), average node degree: 13.5, avg. local clustering coefficient: 0.429, PPI enrichment p-value < 1.0e-16, statistical background: whole genome. Scatter plot including Student’s t-test and hierarchical clustering/heat map were performed by Perseus software and the PPI network was created by the STRING database.

Figure 5

Gene ontology (GO) functional annotation and enrichment analysis of the BAG3 interactome under basal conditions (DMSO_IP). Only class A BAG3 interactors with a p-value ≤ 0.01 and a ratio ≥ 2 were subjected to analysis. The 25 most significantly enriched GO terms in the category Biological Process (A), Molecular Function (B) and Cellular Component (C) are shown. For each GO term, the observed gene count is presented. Analysis was performed by STRING database.

Figure 6

Gene ontology (GO) functional annotation and enrichment analysis of the BAG3 interactome upon proteasome inhibition (MG132_IP). Only class A BAG3 interactors with a p-value ≤ 0.01 and a ratio ≥ 2 were subjected to analysis. The 25 most significantly enriched GO terms in the category Biological Process (A), Molecular Function (B) and Cellular Component (C) are shown. For each GO term, the observed gene count is presented. Analysis was performed by STRING database.

Figure 7

Quantitative proteomic analysis of alteration in BAG3 protein-protein interactions upon proteasome inhibition. (A) Scatter plot generated by plotting the log2 ratios against the negative log10 p-values of the Student’s t-test MG132_IP compared to DMSO_IP (MG132_DMSO; p-value = 0.05; S0 = 0). Proteins with a p-value ≤ 0.05 and a ratio ≥ 1.5 or ≤ 0.75 (black dash lines) were considered as significant, labelled with the gene symbols of the respective proteins, and are shown as purple open triangles. (B) Hierarchical clustering and heat map of LFQ intensities of proteins whose interaction with BAG3 was significantly changed upon proteasome inhibition. The three groups DMSO_IP, MG132_IP and MG132_IgG_IP including their three biological replicates are clustered in columns and proteins are clustered in rows. Color scale reports Z-scored log2-transformed LFQ intensity values; blue indicates a high LFQ intensity, yellow marks a low LFQ intensity. Scatter plot including Student’s t-test and hierarchical clustering/heat map were performed by Perseus software.

Figure 8

Gene ontology (GO) functional annotation/enrichment analysis and PPI enrichment analysis of proteins whose interaction with BAG3 was significantly altered upon proteasome inhibition. (A) Significantly enriched GO terms in the category Cellular Component are shown. For each GO term, the observed gene count is presented. (B) Protein-protein interaction (PPI) network of proteins whose interaction with BAG3 was significantly altered upon proteasome inhibition. Network was clustered into 12 subnetworks using the k-means clustering method; nodes/proteins of the same cluster exhibit the same color. Thickness of connecting lines/edges correlates with the strength of the association. Further network parameters: number of nodes: 39, number of edges: 31 (expected: 12), average node degree: 1.59, avg. local clustering coefficient: 0.382, PPI enrichment p-value: 2.39e-03, statistical background: whole genome. GO functional annotation/enrichment analysis and creating of PPI network were performed by STRING database.

Figure 9

The SRC tyrosine kinase YES1 was identified as a novel BAG3 interactor. (A) HEK293T cells were transfected with a human BAG3 (hBAG3) or a FLAG-tagged human BAG3 (FLAG-hBAG3) overexpressing plasmid and total proteins were extracted; FLAG immunoprecipitation was performed and the eluates were separated by SDS-PAGE. Expression of indicated proteins was analyzed by immunoblotting. (B) HEK293T cells were transfected with a FLAG-tagged human YES1 overexpressing plasmid. Fixed cells were immunohistochemically stained with anti-BAG3 (blue) and anti-FLAG (red) and analyzed by confocal microscopy. Scale bar: 10 μm.