A human MAP kinase interactome

Abstract

Mitogen-activated protein kinase (MAPK) pathways form the backbone of signal transduction in the mammalian cell. Here we applied a systematic experimental and computational approach to map 2,269 interactions between human MAPK-related proteins and other cellular machinery and to assemble these data into functional modules. Multiple lines of evidence including conservation with yeast supported a core network of 641 interactions. Using small interfering RNA knockdowns, we observed that approximately one-third of MAPK-interacting proteins modulated MAPK-mediated signaling. We uncovered the Na-H exchanger NHE1 as a potential MAPK scaffold, found links between HSP90 chaperones and MAPK pathways and identified MUC12 as the human analog to the yeast signaling mucin Msb2. This study makes available a large resource of MAPK interactions and clone libraries, and it illustrates a methodology for probing signaling networks based on functional refinement of experimentally derived protein-interaction maps.

Figure 1

Functional properties of the MAPK network. (a) Breakdown of network proteins (columns 1–3) and protein interactions (columns 4–10) by functional category. Enrichment p-value is based on the probability of identifying an equal or greater number of proteins in the same category at random assessed via the hypergeometric test with a background of 30,000 genes. (b) The percentage of Positive Reference Set or Negative Reference Set pairs overlapping with each network is shown. Error bars are s.e.m. (see Methods). (c) Percentage of MAPK network genes which are able to knockdown AP-1 or NFκB activity. Random represents the percentage of forty-five randomly selected proteins which knockdown NFκB levels.

Figure 2

Protein subnetworks reveal known and putative MAPK scaffolds. Network neighborhoods are shown for (a) Filamin protein FLNA, (b) the Na-H exchanger NHE1, (c) RAN binding protein RANBP9, and (d) the kinesin family member KIF26A. Newly-identified Y2H interactions (red) and interactions from literature are shown (blue). Proteins are colored based on their annotation as membrane (grey), MAPK (blue), transcription factor (red), or phosphorylated under EGF stimulation (yellow border). (e) Binding of in vitro-translated MAP3K7 (TAK1) to GST tagged C-terminus of NHE1 or GST alone. N-moesin, a known NHE1 binding partner is used as a positive control. Expressed input proteins used for in vitro binding assays are marked with asterisks. (f) Phosphorylated (pp38) and total levels of p38, assayed with and without PMA stimulation and two different siRNA knockdowns of NHE1. The bars quantify the pp38 / p38 ratio by image analysis of western blot. As a negative control, we observed nearly equal amounts of pp38 to p38 in the absence of PMA stimulation (−PMA). As a positive control, PMA stimulation with a scrambled siRNA induced p38 phosphorylation (Scramble, +PMA). (g) Using a luciferase reporter fused to the AP-1 gene, we tested the ability of various siRNAs to reduce AP-1 activation when stimulated with PMA. As a negative control, scrambled siRNAs resulted in maximal AP-1 transcription (“Scramble”). As positive controls, siRNAs targeted to luciferase showed a large reduction of luminescence (“Luciferase”), and siRNAs directed directly to p38, upstream of AP-1, reduced signal intensity by 50%. Error bars represent s.e.m. with 6 replicates.

Figure 3

Functional modules in the core network. (a) Bird’s eye view of the core MAPK Y2H network. (b–g) High confidence conserved functional modules. Red edges correspond to core MAPK Y2H interactions which were conserved with yeast. Grey edges indicate core interactions not conserved with yeast. Thickness of the edge increases with the number of observations. (h) AP-1 luciferase activation assay for various siRNAs targeting members of conserved modules. (i) p38 phosphorylation levels are decreased with siRNAs targeting members of conserved modules. (j–l) Novel modules not conserved with yeast.