Combinatorial depletion analysis to assemble the network architecture of the SAGA and ADA chromatin remodeling complexes

Abstract

Despite the availability of several large-scale proteomics studies aiming to identify protein interactions on a global scale, little is known about how proteins interact and are organized within macromolecular complexes. Here, we describe a technique that consists of a combination of biochemistry approaches, quantitative proteomics and computational methods using wild-type and deletion strains to investigate the organization of proteins within macromolecular protein complexes. We applied this technique to determine the organization of two well-studied complexes, Spt-Ada-Gcn5 histone acetyltransferase (SAGA) and ADA, for which no comprehensive high-resolution structures exist. This approach revealed that SAGA/ADA is composed of five distinct functional modules, which can persist separately. Furthermore, we identified a novel subunit of the ADA complex, termed Ahc2, and characterized Sgf29 as an ADA family protein present in all Gcn5 histone acetyltransferase complexes. Finally, we propose a model for the architecture of the SAGA and ADA complexes, which predicts novel functional associations within the SAGA complex and provides mechanistic insights into phenotypical observations in SAGA mutants.

Figure 1

Proteomic analysis of wild-type purifications. (A) Venn diagram of previous knowledge of SAGA/ADA complexes: Using information obtained from the literature, the SAGA and ADA complexes were represented in a Venn diagram to indicate shared and specific proteins for the respective complexes. The SAGA/ADA complexes consist of distinct modules as shown by previous work, which are the recruitment module (Tra1), the acetylation module (Gcn5, Ada3 and Ada2), the TBP interaction unit (Spt3 and Spt8), the DUB module (Ubp8, Sgf11, Sgf73 and Sus1), the architecture unit (Spt7, Spt20, Ada1, Taf5, Taf6, Taf9, Taf10 and Taf12), a single subunit (Sgf29), a single subunit (Chd1) and the ADA module subunit (Ahc1) (reviewed in Koutelou et al, 2010). The numbers inside of the diagram represent the number of the proteins shared between the complexes. (B) Hierarchical clustering on the wild-type purifications. Hierarchical clustering analysis using WARD algorithm and Pearson correlation as distance metric was performed on the relative protein abundances expressed as dNSAFs normalized on the subunits of the SAGA/ADA complexes. Each column represents an isolated purification, and each row represents an individual protein (prey). The color intensity depicts the protein abundance with the brightest yellow indicating highest abundance and decreasing intensity indicating decreasing abundance. Black indicates that the protein was not detected in a particular sample. The HAT module is colored in green, the DUB module colored in violet, the SA_SPT module in orange, the SA_TAF module in blue and the two proteins unique to the ADA module were colored in red.

Figure 2

Hierarchical clustering on different deletion strains and analysis of catalytic mutants. (A) Each column represents an isolated TAP in a different deletion strain, and each row represents an individual protein (prey). The color intensity represents protein abundance (dNSAF) normalized on the subunits of the SAGA/ADA complexes with the brightest yellow indicating highest abundance and decreasing intensity indicating decreasing abundance. Black indicates that the protein was not detected in a particular purification. The proteins of the modules were colored as in Figure 1. The clustering result leads to the formation of distinct modules (represented on the right side of the cluster). Relative abundance of the 21 subunits of the SAGA/ADA complexes obtained from (B) purifications of the Gcn5 catalytic mutant using Spt7 as bait and (C) Ubp8 catalytic mutant purified by the bait Ada2. In each case, three replicate purifications were performed. The catalytic mutants of Gcn5 and Ubp8 were generated by mutating amino acids 125–127 (KQL to AAA) and by substituting the two zinc-finger amino acids C46A and C49A, respectively (Wang et al, 1998; Ingvarsdottir et al, 2005). All data is represented as average dNSAF values+s.d.

Figure 3

Deletion interaction network of the Gcn5 HAT complexes. (A) The probabilistic protein network of the Gcn5 HAT complexes was generated by representing proteins as nodes (baits in the respective deletion strains by triangles and preys as circles), connected by weighted edges denoting the calculated probabilities. Black dashed lines symbolize interactions with high probability, cyan dashed lines interactions with moderate probability and red dashed lines interactions with low probability. The baits are depicted as triangles and colored based on the TAP subunit: orange for the SA_SPT module, green corresponds to the ADA module, violet to the DUB module and blue to the SA_TAF module. Preys are symbolized by circles and colored as in Figure 1 (A–D). Focused probabilistic protein networks for preys of each of the four modules of the Gcn5 HAT complexes in all the baits, i.e. (A) the SA_SPT module, (B) the HAT/Core module, (C) the SA_TAF module and (D) the DUB module. The Cytoscape software environment was used to generate the probabilistic protein networks. In each network, only the baits that have a link (i.e. pull down the prey) with a prey are represented.

Figure 4

Deletion interaction network and the macromolecular assembly of the Gcn5 HAT complexes. Based upon all deletion purifications, all proteins of the SAGA/ADA complexes were organized into modularity and consequently a macromolecular model was assembled (for details, see main text of the manuscript). In addition to our deletion purifications, we integrated existing data from yeast two-hybrid and gene deletion experiments to further refine our model. As a result, we allowed direct contacts only between protein pairs (i.e Ada2–Gcn5; Ada2–Ada3; Ada3–Sgf29; Taf5–Taf6; and Taf6–Taf9) for which yeast two-hybrid data exist. Genetic interaction data was also used to position some of the proteins from different modules in close proximity. In particular, components of the DUB module exhibit negative genetic effects with two components of the HAT/core module, which are Ada2 and Gcn5. Therefore, these proteins were placed in close proximity. The color code is in accordance with Figures 1B and 2A. The size of the inset circle correlates with the molecular weight of each illustrated protein.

Figure 5

Sgf29 exhibits the characteristics of other known ADA proteins, including Ada2 and Gcn5. (A) Deletion of SGF29 rescue Gal4-VP16-mediated toxicity in yeast, similar to the deletion of ADA2. (B) β-Galactosidase activation by VP16 in yeast is compromised by the deletion of SGF29. This phenotype is similar to what is seen for the deletion of ADA2 as seen in the graph. (C) Yeast lacking SGF29 is compromised for growth on alternative carbon sources, similar to what is observed for other SAGA subunits, including SPT7 (separated by black lines). Yeast were serially diluted on the indicated plates and imaged at the indicated times (see Materials and methods for details).

Figure 6

Ahc2 is a bona fide member of the ADA HAT complex. (A) Western blot analysis of calmodulin pull-down experiments indicates that both Ahc2 and Sgf29 precipitate Ada3, a known component of Gcn5 HAT complexes (see lanes 3 and 4). (B) Silver stain of the TAP tag purification of Ahc2 identified only the six components of the ADA complex. Each of the components are indicated on the gel. (C) In vitro HAT assay using Ahc2–TAP-purified ADA complex demonstrates that the ADA complex preferentially acetylates nucleosomes compared with histones.