Immunoprecipitation and mass spectrometry defines an extensive RBM45 protein-protein interaction network

Abstract

The pathological accumulation of RNA-binding proteins (RBPs) within inclusion bodies is a hallmark of amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTLD). RBP aggregation results in both toxic gain and loss of normal function. Determining the protein binding partners and normal functions of disease-associated RBPs is necessary to fully understand molecular mechanisms of RBPs in disease. Herein, we characterized the protein-protein interactions (PPIs) of RBM45, a RBP that localizes to inclusions in ALS/FTLD. Using immunoprecipitation coupled to mass spectrometry (IP-MS), we identified 132 proteins that specifically interact with RBM45 within HEK293 cells. Select PPIs were validated by immunoblot and immunocytochemistry, demonstrating that RBM45 associates with a number of other RBPs primarily via RNA-dependent interactions in the nucleus. Analysis of the biological processes and pathways associated with RBM45-interacting proteins indicates enrichment for nuclear RNA processing/splicing via association with hnRNP proteins and cytoplasmic RNA translation via eiF2 and eiF4 pathways. Moreover, several other ALS-linked RBPs, including TDP-43, FUS, Matrin-3, and hnRNP-A1, interact with RBM45, consistent with prior observations of these proteins within intracellular inclusions in ALS/FTLD. Taken together, our results define a PPI network for RBM45, suggest novel functions for this protein, and provide new insights into the contributions of RBM45 to neurodegeneration in ALS/FTLD. This article is part of a Special Issue entitled SI:RNA Metabolism in Disease.

Keywords: ALS; Immunoprecipitation; Mass spectrometry; Protein–protein interaction; RBM45.

Figure 1. Identification of RBM45 interacting proteins

(A) Diagram of the immunoprecipitation-mass spectrometry approach to identify the RBM45 interacting proteins. (B) Triplicate immunoprecipitates from HEK293 cells stably expressing FLAG-RBM45, HA-RBM45 or empty vector were separated by SDS-PAGE and stained with Coomassie blue to visualize proteins. Immunoprecipitations with FLAG (sample 1, 2, 3, 4), HA (sample 5) antibody or IgG (sample 6, 7) were performed. Crosslinking IP (sample 1, 2) and regular IP (sample 3, 4, 5) were performed in parallel. For crosslinking IP, live cells were treated with 0.1% formaldehyde to cross-link proteins prior to cell lysis and immunoprecipitation. The crosslinking was reversed by heating in SDS-sample buffer prior to SDS-PAGE. The proteins along the entire length of the gel were extracted (excluding the IgG-heavy chain and IgG-light chain that are denoted by red *) and analyzed by LC/MS-MS. (C) Pie-chart representation of functional classes of RBM45 interacting proteins.

Figure 2

Verification of RBM45 interacting proteins. (A) Pull-down of selected proteins with FLAG-RBM45 expressed in HEK293 cells. Crosslinking immunoprecipitations with FLAG antibody or IgG were performed. The IP fractions were immunoblotted with hnRNP-L, hnRNP-A1, hnRNP-A2B1, Matrin-3, hnRNP-A3 and RBM14 antibodies. The same IP fractions were also immunobloted with FLAG (FLAG-RBM45) and GAPDH (negative IP control) antibodies. (B) Immunoprecipitations of endogenous candidate proteins in FLAG-RBM45 expressing cells shows that FLAG-RBM45 co-purified with the tested endogenous proteins. The endogenous candidate proteins were immunoprecipitated with hnRNP-L, hnRNP-A1, Matrin-3 or RBM14 antibody, while IgG pull-down was used for IP control. The immunoblots were detected with FLAG antibody (FLAG-RBM45), tested endogenous protein specific antibodies, and GAPDH antibody (negative IP control). Similar validation studies for TDP-43 and FUS were previously reported (Li et al., 2015). (C) In-cell RNase treatment and crosslinking-IP were performed on cells expressing FLAG-RBM45. The amount of hnRNP-L, hnRNP-A1, the lower-molecular-weight band of hnRNP-A2B1, Matrin-3 and RBM14 that co-purified with FLAG-RBM45 was reduced upon the RNase treatment. (D) The RBM45-(Δ286-318) construct, i.e. the homo-oligomerization assembly (HOA) domain deficient construct (Li et al., 2015), exhibits significantly reduced binding to the tested candidate proteins when compared to full-length-RBM45. Full-length FLAG-RBM45 or FLAG-Δ(286-318) construct were expressed in HEK293 cells. FLAG-IP was performed as described previously. Immunoblot analysis shows that all the tested candidate proteins displayed reduced co-IP% with FLAG-Δ(286-318) as compared with the full-length FLAG-RBM45.

Figure 3

Enriched GO Biological Process Terms. RBM45-interacting proteins were tested for GO Biological Process enrichment using the right-sided hypergeometric test with Benjamini-Hochberg post-hoc p value correction. Terms with a p value of 0.001 or less were visualized in a network layout, where node size corresponds to term p value. The proportion of shared proteins between terms was evaluated using the kappa statistic and nodes with a kappa score (κ) of at least 0.4 were connected with edges on the graph, with edge width proportional to kappa score. Leading terms, those terms with the highest number of proteins, are colored for emphasis.

Figure 4

Leading Terms with Associated Proteins. Leading terms from Figure 3 were placed into a separate network and all associated proteins from the list of RBM45-interacting proteins were visualized as nodes and connected to the appropriate term. Where a protein is associated with multiple terms, multiple edges emanate from that protein and edges are color-matched to their associated terms.

Figure 5

Co-localization analysis. (A–G) The co-localization of RBM45 and the indicated proteins were evaluated using immunocytochemistry together with image deconvolution and co-localization analysis. Representative images and pixel intensity scatter plots are shown with cutouts at higher magnification to highlight detail. (H) Statistical analysis of protein co-localization. M1 = RBM45 overlap with indicated protein. M2 = indicated protein overlap with RBM45. ICQ = intensity correlation quotient. p ICQ = p value of ICQ. Manders coefficients (M1 and M2) measure the proportion of co-localizing proteins in each channel of a two-channel image and are shown as mean ± SEM. The intensity correlation quotient (ICQ) has a range of −0.5 (perfect segregation) to 0.5 (perfect co-localization), with random intensity variation resulting in a value ~0. The statistical significance of each ICQ value is shown at far right. SMN staining was used as a negative control.