The biochemical and genetic discovery of the SAGA complex

Abstract

One of the major advances in our understanding of gene regulation in eukaryotes was the discovery of factors that regulate transcription by controlling chromatin structure. Prominent among these discoveries was the demonstration that Gcn5 is a histone acetyltransferase, establishing a direct connection between transcriptional activation and histone acetylation. This breakthrough was soon followed by the purification of a protein complex that contains Gcn5, the SAGA complex. In this article, we review the early genetic and biochemical experiments that led to the discovery of SAGA and the elucidation of its multiple activities.

1. Introduction

The role of chromatin structure in eukaryotic transcriptional regulation was uncertain for many years. While several studies revealed intriguing correlations between transcription and chromatin structure (for example [8, 9]), the causal relationship between the two was debated. Similarly, other studies suggested that post-translational modifications of histones, specifically histone acetylation, were required for the activation of transcription; however, this was again based upon an association between active transcription and acetylated histones [11].

The chromatin field moved forward dramatically in the 1990’s with important progress coming from biochemical studies in mammalian cell systems and from genetic studies in yeast. Several biochemical studies of mammalian transcriptional activators provided compelling evidence that nucleosomes were inhibitory to transcription in vitro and that the function of activators was to help overcome this repression (for example, [12–14]). Providing a strong complement to these biochemical studies, genetic studies in yeast provided compelling evidence that chromatin structure was dynamic in response to gene activation [15, 16], and furthermore, that altering histone levels caused transcriptional changes in vivo [17–19]. Importantly, a combination of genetic and biochemical studies demonstrated that chromatin structure itself was regulated by the Swi/Snf complex, a conserved chromatin remodeling complex [20–25]. This strengthened the connection between chromatin structure and transcriptional regulation, as Swi and Snf proteins were initially identified as transcriptional regulatory factors in yeast [26–28]; they are now known to be critical tumor suppressors in mammals [29].

A critical breakthrough in establishing the connection between histone acetylation and transcription came in 1996 when Gcn5, a conserved factor previously identified in yeast as a transcriptional regulatory protein, and proposed to function with a group of factors as a bridge between DNA-bound activators and the basal machinery [30–33], was demonstrated to be a histone acetyltransferase [34]. This discovery inspired a number of genetic and biochemical studies that coalesced on the identification of the SAGA (Spt Ada Gcn5 Acetyltransferase) complex [4]. SAGA was one of the first complexes shown to have coactivator activity in vivo, comprised of modular components that contribute diverse transcriptional regulatory functions [1–3, 35–38]. In this chapter, we provide a historical look at the studies that led to the identification of SAGA by our labs and others, and the subsequent studies that elucidated its functions in yeast.

2. The S of SAGA - the identification of SPT genes

SPT genes were identified via a yeast mutant screen whose initial goal was to understand the nature of eukaryotic transposable elements (Ty elements) with respect to their stability and effects on neighboring gene expression. At the time, it was not envisioned that the mutants would identify conserved transcription factors and chromatin proteins.

Ty transposable elements of S. cerevisiae are a family of heterogeneous elements that are structurally and genetically similar to mammalian proviruses [39–44]. Early studies revealed that Ty elements conferred mutant phenotypes when they transposed into the regulatory regions of genes [43, 45–47].

Two Ty insertions were isolated as spontaneous mutations at the HIS4 gene of S. cerevisiae [43]. Both Ty insertion mutations, his4–912 and his4–917, occurred in the 5’ noncoding region of HIS4 and caused histidine auxotrophy, although they differed by position, orientation, and sequence. Selection for His⁺ revertants of these two insertion mutations revealed many types of cis-acting rearrangements, including recombination between the long terminal repeats (δ sequences) of the Ty element, inversions, translocations, and gene conversions with other Ty elements in the yeast genome [44, 45, 48, 49].

Mutant hunts that focused on finding extragenic suppressors of Ty insertion mutations led to the identification of spt mutants [50, 51]. To avoid the genomic rearrangements isolated as suppressors of complete Ty insertions, these later selections started with single LTR derivatives of his4–912 and his4–917 [43], as well as a Ty-derived allele later isolated at the LYS2 gene [52]. The extragenic suppressors that were identified occurred in several unlinked genes, suggesting that several functions control the behavior of insertion mutations [50, 51]. The genes identified by these mutations were named SPT genes for Suppressor of Ty. Early studies showed that spt mutations suppressed at the level of transcription [51, 53–55]. Additional analysis showed that some, but not all spt mutations impaired transcription of Ty elements themselves [51, 55]. These results as well as additional phenotypic analyses elucidated two major classes of SPT genes, with both classes able to suppress overlapping but distinct classes of insertion mutations, as well as each class conferring distinct mutant phenotypes [56].

Additional studies of spt mutants connected these two classes of SPT genes to known proteins involved in transcription or chromatin structure. One study demonstrated that two SPT genes, initially called SPT11 and SPT12, were the same as two histone genes, HTA1 and HTB1, encoding histones H2A and H2B respectively [17]. These studies showed that either a reduced or increased level of this pair of histones caused suppression of his4–912δ by altering transcription, suggesting that altered histone stoichiometry was responsible. This was consistent with an earlier finding that altered levels of histone pairs impaired chromosome segregation [57]. The mutations of these histone genes caused phenotypes similar to mutations in four other SPT genes, SPT4, SPT5, SPT6, and SPT16, suggesting that all of these function in chromatin-mediated transcription, something now known to be true [16]. Spt4 and Spt5 were later identified in mammalian cells as the elongation factor DSIF (DRB-sensitivity inducing factor; [58, 59]), and Spt16 was later identified in mammalian cells as a component of FACT (Facilitates transcription on a chromatin template; [60]).

Second, the cloning of the SPT15 gene, a member of the second phenotypic class of SPT genes, revealed that it encoded the general transcription factor TATA-binding protein (TBP) [61]. This discovery occurred at the same time as studies from several other labs that used biochemical approaches to identify SPT15 as the gene encoding TBP [62–65]. Although SPT15 is essential for viability, the spt15 mutations isolated in the spt mutant hunt were viable and had genetic and transcriptional phenotypes similar to those caused by four other SPT genes: SPT3, SPT7, SPT8, and SPT20 [61]. The genetic connection of these genes to the gene encoding TBP suggested related roles in transcription initiation. As described below, the Spt3, Spt7, Spt8, and Spt20 proteins are all members of SAGA.

3. The A and G of SAGA – the identification of ADA genes, including the rediscovery of GCN5

ADA genes were identified using a genetic screen for mutations that reduce transcriptional activation. This genetic screen followed the discovery, via biochemical assays, of intermediary factors in transcriptional activation [66–68]. The previous prevailing model for transcriptional activation involved transcriptional activators bound to DNA at enhancer regulatory elements in mammalian cells or their viruses (defined in 1980’s) or at upstream activation sequences (UASs) in yeast, looping or interacting in cis with the basal transcriptional machinery (TFIIA, B, C, D, E, F, H; isolated in 1980’s) bound to DNA near the transcriptional start site [69–71]. However, a major gap in knowledge was the mechanism by which the DNA activators that were bound upstream interacted with the basal machinery: first, which basal factor is the direct target for activation and, second, does contact increase transcription, via basal factor localization, increased concentration, or allosteric change? Moreover, as discussed above, at that time there was a “split” between biochemists and molecular biologists working on transcription, and those working on the function of chromatin and histone modifications, in particular, whether chromatin alterations have a causal role in transcriptional regulation [72].

The in vitro transcription biochemical assay that eventually led to identification of ADA genes utilized a transcriptional activator specifically crippled for DNA binding but retaining an intact activation domain [66]. The concept was based on findings that highly expressed strong transcription factors in vivo can negatively affect transcription due to activation domain titration—or “squelching”—of an unknown critical transcription factor(s) [73]. Thus, in the in vitro assay, using a complete nuclear extract for the assay, which contained all components needed for gene activation, the mutant activator was seen to inhibit transcriptional activation but did not lower basal transcription. This observation suggested that the intact activation domain of the mutant activator bound to – and thus lowered accessibility within the extract – a proposed intermediary bridging factor(s) required specifically for activation. Hence, the direct target appeared to be an accessory factor rather than a basal factor. This factor was named “adaptor”, to reflect a proposed function as a bridge between DNA bound activator and the basal transcriptional machinery [66]. Contemporaneous with these experiments were additional discoveries pointing to accessory factors for transcriptional activation [67, 68], and taken together, these models launched the concept that intermediary factors (co-activators) are crucial for transcriptional activation [74].

Based on these models, the ADA genetic screen, conceptually based on squelching in vivo as described above [73], was set up in yeast to identify genes encoding intermediary factors. The screen selected for mutations that suppressed growth toxicity caused by artificially high expression of a transcriptional activator [33]. The hypothesis was that toxicity resulted from the over-expressed transcriptional activator binding, via its activation domain, to critical adaptors and making them unavailable to genes required for normal growth, and thus was parallel to the in vitro assay described above. Mutations in ADA genes lowered the growth toxicity, presumably by freeing up sufficient adaptor molecules to allow gene activation of the critical growth genes. The ADA mutations also lowered transcriptional activation without lowering basal transcription, again parallel to the initial biochemical identification of intermediary factors discussed above. Genes emerging from this screen were ADA1-ADA5 [1, 32, 33, 75], including two that had been previously isolated. ADA4 was previously isolated as GCN5 and AAS104 in earlier genetic screens [30, 31]. ADA5 was previously isolated as SPT20, as discussed below [75, 76].

The general model that emerged, that of transcriptional activators recruiting intermediary factors to promoters, provided a mechanism for location- and gene-specific chromatin modifications, including histone acetylation, via GCN5. The model informed the Brownell/Allis discovery of GCN5 as the first nuclear histone acetyltransferase (discussed below), providing a mechanistic framework for postulating how acetylation would activate specific genes [34, 77] (Figure 1). Indeed, subsequent studies demonstrated that specific amino acid substitutions in the Gcn5 activation domain that reduced gene transcription in vivo also lowered histone acetylation in vitro [78], and chromatin immunoprecipitation demonstrated histone acetylation at specific genes [79], together confirming that histone acetylation via a specific enzyme at a specific location was required for gene activation.

Figure 1.

A model for SAGA activation as an adaptor, or intermediary complex, to localize histone acetylation. The recruitment of SAGA by gene-specific activators results in gene-specific histone acetylation and transcriptional activation. Figure adapted from [6].

4. Connecting SPT and ADA genes

The idea that SPT and ADA genes might be involved in a common function came from the striking discovery that SPT20 and ADA5, isolated in distinct mutant hunts, were the same gene [75, 76]. Furthermore, these studies showed that ada5 mutants had a broader range of phenotypes compared to the other ada mutants (ada2, ada3, and gcn5), demonstrating two phenotypic classes of ADA genes. Subsequent analysis of the ADA1 gene showed that, like SPT20/ADA5, mutations caused both ada and spt mutant phenotypes [1]. The common phenotypes of mutations in SPT7, SPT20/ADA5, and ADA1 strongly suggested functional overlap. Furthermore, biochemical studies showing physical association of Ada proteins [1, 75, 80, 81] and separate studies showing association of Spt proteins [2], suggested the possibility of a protein complex that might contain all of these proteins.

5. The biochemical identification and characterization of SAGA

The identification of the Tetrahymena thermophila protein p55 as the first nuclear histone acetyltransferase (HAT) linked to a single polypeptide was a remarkable breakthrough [34, 77]. Their development of HAT assays [82] would ultimately allow for the identification of numerous such HAT enzymes. It also provided a direct biochemical link between the process of histone acetylation with transcriptional activation. Strikingly the sequence of p55 clearly identified it as a homolog of yeast Gcn5, a conserved protein that was also isolated in the aforementioned ADA screen as an adaptor required for the full activity of certain transcriptional activators [30, 33, 83].

However, in contrast to the potent in vitro HAT activity of Gcn5 on recombinant or “free” histones, Gcn5 was unable to efficiently acetylate nucleosomal histones [4, 82, 84]. A potential clue that might explain this apparent discrepancy was the fact that genetic screens in yeast, and protein interaction studies, identified Ada proteins as functionally interacting with Gcn5 and that Ada and Spt proteins share certain genetic functions and in some instances are encoded by the same gene (discussed above).

A collaboration between the labs of Workman, Berger, Winston, and Allis was subsequently launched. Using some of the same HAT assays developed by Brownell and Allis, large scale purification of HAT activities from yeast were expected to reveal Gcn5 isolated in its native form, and that either modification of Gcn5 or its physical interaction with other proteins would enable the enzyme to modify nucleosomal histones. This collaboration led to the discovery that native yeast Gcn5 exists in at least two high-molecular-weight complexes—and, indeed, as predicted, multiple subunits of which enable Gcn5 to acetylate nucleosomes [4]. The most notable of these complexes at the time was named SAGA (Spt-Ada-Gcn5 Acetyltransferase), based on the composition and known activity of the complex. It was discovered that SAGA preferentially acetylates multiple lysine residues on the N-terminal tails of histones H3 and H2B [84]. Notably, western blotting revealed that SAGA is composed of certain Ada and Spt proteins, fulfilling the genetic predictions that these proteins functionally interact with Gcn5 and one another. Further analysis of the composition and function of SAGA and its orthologs in larger eukaryotes has revealed that Gcn5 resides in a complex of 18–20 proteins (reviewed by [85, 86]). SAGA has served as a paradigm for our understanding of the regulation of HAT activity and the coupling of histone modifications, in general, with transcriptional co-activation from yeast to humans.

It is interesting that, while the discovery of SAGA was rapidly made after the report of Gcn5 HAT activity, the isolation of SAGA was actually quite fortuitous. It was found that Nickel-NTA agarose resin served to enrich for SAGA and other yeast HAT activities, including the Gcn5-containing ADA complex and the NuA3 and NuA4 HAT complexes ([4]; Figure 2), despite the absence of any known proteins containing stretches of histidines in these complexes. The use of Nickel-NTA agarose resin was likely key in the ability to purify large amounts of SAGA and other complexes from crude extracts, and to enable the eventual mass spectrometric identification of the components of these complexes. Indeed, in follow-up studies using hundreds of liters of yeast whole cell extracts, it was discovered SAGA is composed of at least three protein families known to function in gene expression: the Ada and Spt protein families, and a subset of TATA-binding protein (TBP)-associated factors (TAFs [4, 87]); and also the ATM-related cofactor Tra1 [88] that had been revealed as a coactivator interacting with activation domains [36, 89]. Importantly, the discovery of SAGA clarified years of genetic and biochemical studies on transcriptional regulation by showing that Gcn5 exists in macromolecular complex with the Ada, Spt, and TAF proteins.

Figure 2.

Identification of the SAGA and SLIK/SALSA complexes. Schematic representation of the original isolation of SAGA and SLIK/SALSA from yeast whole cell extracts, adapted from [4, 5]. Lower panel depicts fluorograms of HAT assays performed with nucleosomal histones and Mono Q fractions, which helped isolate the SAGA, SLIK, ADA, NuA3 and NuA4 HAT complexes. SLIK/SALSA was independently purified by Sterner et al, 2002 from synthetic complete media under conditions where it preferentially bound Nickel-nitrilotriacetic acid (NTA) agarose.

6. Genetic and biochemical studies identified distinct functional modules within SAGA

Along with the functional distinctions among SPT and ADA genes, additional genetic studies suggested that SAGA contained multiple activities in addition to Gcn5 that were required for transcriptional activation. One group of SAGA proteins, Spt7, Spt20, and Ada1 appeared to be required for SAGA structure, and hence all function, as these mutants had the most severe growth phenotypes of SAGA mutants tested [1–3, 76]. Mutants lacking any one of these proteins failed to form a SAGA complex [3, 4, 90, 91]. In contrast, Spt3 and Spt8 appeared to form one module of SAGA, likely to be required for TBP recruitment, and Gcn5, Ada2, and Ada3 were thought to form another module, required for histone acetylation [3, 4, 81, 92, 93]. In the latter two classes of mutants, SAGA complexes still formed, showing that these proteins are not required for SAGA structure. Importantly, double mutants that combined null mutations from each module (for example, spt3 and gcn5 null mutations) had more severe phenotypes, similar to spt7, spt20, ada1 mutants, suggesting that the two modules contribute distinct activities [2, 3, 94]. In these double mutants, a SAGA complex still formed, supporting the view that these modules are involved in function, not structure. This led to a model suggesting at least three modules for SAGA, with more likely to be found (Figure 3; [2, 3]). Subsequently a third module with a distinct function, was discovered for SAGA, a histone deubiquitinase activity[7, 91, 95–97].

Figure 3.

SAGA then and now. A. The proposed SAGA functions based on genetic and biochemical studies [1–3]. The “?” represents a hypothesized activity of SAGA based on genetic studies, later shown to be the histone deubiquitination activity [7]. B. SAGA structure from cryo-EM studies (PDB structure 6TBM; [10]). The modules are colored to correspond to panel A. The TAFs, now known to be part of the structural core, are shown in yellow and Tra1 is shown in pink.

6.1. The Ada2/Ada3/Gcn5 module

Five ADA genes were discovered in the original screen, as described above. Of these five, mutations in ADA1 and ADA5 cause both Ada⁻ and Spt⁻ phenotypes [1, 75, 76], whereas mutations in ADA2, ADA3, and ADA4/GCN5 manifest only Ada⁻ phenotypes [1, 32, 33]. These similar phenotypes led to investigation of whether Ada2, Ada3, and Gcn5 interact in regulating transcription. Multiple genetic and biochemical approaches, including yeast two-hybrid interaction assays, lexA transcription reporter assays, and in vitro transcription/translation assays, demonstrated physical association and interdependent transcriptional activation between the three gene products [80, 93, 98]. Subsequently the three proteins were shown to be co-dependent in effecting histone acetylation [81]. These results suggested that they might assemble into a single module within SAGA, and this was confirmed by in vitro biochemistry using purified proteins; moreover, additional experiments showed that histone and nucleosome acetylation required all three proteins in the module [92]. A fourth subunit of the Gcn5 module, Sgf29, was uncovered through mass spectrometry of protein complexes binding to methylated H3K4 – a transcriptional promoter modification – with interesting implications for cross-talk between histone methylation and histone acetylation/ubiqutination (see below for ubiquitination) in SAGA [99, 100]. Later studies showed that Sgf29 helps to activate transcription and to recruit TBP to promoters [101, 102]. It is important to note that Gcn5 and acetyltransferases in general acetylate numerous substrates well beyond histones [103, 104].

6.2. The connection of Spt3 and Spt8 with TBP

Among the TBP-related set of SPT genes, several genetic results connected SPT15, encoding TBP, with two other members of this group, SPT3 and SPT8. First, starting with a particular allele of SPT15, spt15–21, extragenic suppressors were isolated and all of the suppressor mutations were in SPT3, with the spt3 suppressor mutations found in clusters [105]. In addition, there was allele-specific mutual suppression: spt15–21 and spt3–401 each caused an Spt⁻ phenotype and the double mutant had a wild-type phenotype, suggesting a direct interaction between the two proteins. Additional screens for spt15 suppressors of spt3–401 identified a region encoding one face of TBP as the likely interface with Spt3 [106]. These studies also demonstrated co-immunoprecipitation between these two proteins, suggesting a direction interaction [105, 106]. These genetic studies were supported by subsequent biochemical and structural studies [10, 107, 108], which confirmed and greatly extended our understanding of the Spt3-TBP interaction. In addition to Spt3, other studies provided strong evidence for Spt8-TBP interactions [3, 109, 110]. A less direct role for Spt8 was inferred by the ability of an spt3 mutation to suppress a complete deletion of the SPT8 gene, suggesting that Spt8 served to promote the Spt3-TBP interaction [111]. Interestingly Spt3 and Spt8 were also suggested to have TBP-inhibitory activity at the HIS3 and TRP3 genes [3, 112], consistent with a later study that showed that Spt3 acts negatively in regulation of the HO gene [113]. Thus, while SAGA is generally viewed as a coactivator, it can also repress in some contexts.

With the advent of chromatin immunoprecipitation, later studies of the Spt3/SAGA-TBP connection tested whether SAGA and specific components of SAGA, were required to recruit TBP to promoters. The first study showed that two SAGA components, Spt20 and Spt3, were strongly required to recruit TBP to the GAL1 promoter, while there was only a modest requirement for a third SAGA component, Gcn5 [114]. As SAGA structural studies showed that Spt20 was required for an intact SAGA complex, while Spt3 was not [3, 115, 116], this was consistent with earlier results suggesting a specific role for Spt3 in TBP recruitment. Subsequent studies established a role for SAGA and Spt3 for activation at the GAL1 promoter: Gal4 was required to recruit SAGA and Spt3 of SAGA was required to recruit TBP [35, 37, 90]. These studies, then, provided some of the first in vivo evidence for the hypothesized role for transcriptional co-activators [66–68, 117].

6.3. The histone deubiquitination module within SAGA

Soon after the initial purification of the SAGA complex, a new breakthrough in histone post-translational modification was made with the discovery that histone H2B is ubiquitinated in yeast [118]. The finding of H2B ubiquitination (H2Bub) was closely followed by the identification within SAGA of a putative protein deubiquitinating enzyme, which was named Ubp8 [95, 119, 120]. Deletion of the UBP8 gene revealed elevated H2B ubiquitination in vivo without activity on other core histones; furthermore, Ubp8 within purified SAGA was shown to deubiquitinate H2B in vitro, demonstrating intrinsic activity [7, 91, 121]. These findings ushered in a new paradigm of reversible histone ubiquitination, with SAGA as a key player.

The function of histone ubiquitination/deubiquitination exhibits an interesting twist compared to reversible acetylation, where histone acetylation correlates with transcriptional activation and deacetylation associates with repression of transcription. Instead, H2B ubiquitination and deubiquitination are required in dynamic sequence during the early stages of transcription for full gene activation [7, 122–126]. Mechanistically, Ubp8 within SAGA functions as a gatekeeper during transitions in the transcription process: H2B ubiquitination acts as a barrier to gene association of Ctk1, an RNA polymerase II kinase, and subsequent H2B deubiquitination by Ubp8 triggers Ctk1 recruitment to switch RNA polymerase II into transcription elongation mode [127].

Genetic analysis demonstrated that acetylation by Gcn5 in SAGA and deubiquitination by Ubp8 in SAGA are distinct mechanisms, since loss of one was well-tolerated but loss of both was highly detrimental to gene activation [7]. The genetic separation of histone modifying functions was strengthened by structural investigations uncovering a protein DUB (deubiquitinating) module. Biochemical approaches revealed that Ubp8 associates with the SAGA complex via Sgf11 and Sgf73, with Sus1 as an additional subunit [91, 96, 97, 128–130]. Ubp8 and Sgf11 in the context of SAGA have important roles in physiology of the fly visual system [131] while Sgf73 is the homologue of human Sca7, which in its polyglutamine expanded pathological form, leads to a neurodegenerative disease [132–134]. X-ray crystallography of the DUB module revealed a strikingly interconnected complex, wherein numerous interactions of Sgf73 and Sgf11 allosterically activate the Ubp8 enzyme [135, 136].

The DUB module of SAGA thus emerges as a full functional partner of the Gcn5 acetylation module, and demonstrates the importance of modularity in the coordination of chromatin modification in transcriptional regulation.

7. The discovery of TAFs and Tra1 in SAGA

TAFs 5, 6, 9, 10 and 12, originally discovered as part of TFIID, were also subsequently shown to be integral components of SAGA and SAGA-related complexes in both yeast [87] and mammalian [137] cells. In fact, mammalian SAGA was initially named STAGA because of this finding of TAF subunits [138]. Moreover, and independently identified TBP-free TAF complex (TFTC) in mammalian cells turned out to be SAGA [139, 140]. TAF proteins, along with TBP, make up the general transcription initiation factor, TFIID (reviewed in [141]. TFIID is required for promoter recognition in RNA polymerase II (Pol II)-catalyzed transcription at most genes (reviewed in [142]. The TAF proteins are highly conserved, however metazoans have paralogous proteins for certain TAFs, which are instead incorporated into SAGA, including mammalian TAF5L, TAF6L and TAF9B [86]. The identification of TAFs in SAGA was the first, striking discovery that these proteins function outside the context of TFIID, and physically linked the members of three distinct families of gene products known to be involved in transcriptional activation, namely Ada, Spt, and TAF proteins [87].

The yeast pseudokinase protein Tra1 and the human counterpart TRRAP, which are related to the ataxia telangiectasis mutated (ATM) family of proteins, were also identified as a conserved component of SAGA [88, 89, 143]. Tra1 was subsequently demonstrated to physically interact with transcriptional activators such as VP16, Gal4, Gcn4 and Rap1 [36, 144–147], while TAF12 was found to mediate interactions with Gal4 [147] and Spt3 directly interacts with TBP [10, 107]. The ability of multiple subunits within SAGA to directly interact with the transcriptional machinery allows for the recruitment of SAGA to genes, the facilitation of chromatin acetylation and the assembly of the transcriptional apparatus during gene activation [148–150].

Genome-wide expression analysis indicates that SAGA and TFIID, despite sharing a number of common subunits, are responsible for expression of different subsets of genes. Initially it was suggested that SAGA functions at about 10% of yeast genes and mostly at TATA-containing, highly regulated loci that respond to environmental stresses, such as metabolic starvation, DNA damage and heat, while TFIID plays a more general housekeeping role [151] However, the specific roles of SAGA in transcription is still an active area of investigation [151–157]. More recently SAGA has been found to map to the promoter regions and modify the chromatin of most yeast genes and deletion of key SAGA subunits reduces de novo transcription of nearly all genes [152, 154]. As such it has been suggested that SAGA and TFIID are substantially redundant at most genes. However rapid depletion of TFIID leads to a loss of transcription at 87% of genes analyzed. In contrast, a rapid depletion of Spt3/Spt20 (which does not lead to an immediate disruption of SAGA-dependent histone modifications) leads to a modest decrease in only 13% of genes analyzed [154]. Simultaneous depletion of both TFIID and SAGA leads to a severe transcription defect at these latter locations, termed “coactivator-redundant genes”. This observation is consistent with earlier studies that a combination of SAGA and TFIID mutations can result in greater transcription defects [151, 158]. A model has now emerged of a genome-wide role of SAGA in chromatin modification that regulates global gene transcription, and a gene-specific role for SAGA at a subset of genes, largely overlapping with TFIID, to promote TBP binding [154]. These studies further highlight how the conserved modular nature of SAGA contributes to different aspects of gene regulation and chromatin modification.

8. The alternative complex SLIK/SALSA

Subsequent to the discovery of SAGA, Gcn5 was shown to exist in another chromatographically distinct multiprotein complex, similar in size and composition to SAGA. This complex was independently named SLIK, for SAGA-like or SALSA for SAGA altered, Spt8 absent [5, 159]. Despite containing the vast majority of polypeptides found in SAGA and sharing overlapping functions, SLIK composition is notably distinct in a number of ways. SLIK lacks Spt8 and contains a C-terminally truncated version of Spt7 in a region required for Spt8 interaction [115, 159]. Notably a counterpart of Spt8 has not been identified in metazoan SAGA [86]. At certain gene promoters (HIS3 and TRP3), SAGA was found to have an inhibitory role, mediated at least in part by Spt8, while SLIK/SALSA is associated with activation of these genes [159]. Yeast Pep4 was subsequently identified as a protease required for Spt7 cleavage and formation of SLIK in a study which supported a role for SLIK in response to nutrient starvation [160]. The protein Rtg2, which is a core component of SLIK but not SAGA, is required for the stability of SLIK and links histone acetylation to the retrograde response pathway and yeast longevity [5]. This signaling is responsible for communicating to the nucleus the need to make metabolic adjustments in times of mitochondrial dysfunction and induces expression of genes whose products compensate for defects in the tricarboxylic acid cycle and allows use of acetate as a carbon source (reviewed in [161]. A role for SLIK or SAGA components in anaerobic respiration is also supported by observations that Gcn5 and Ubp8 are required for this process [162]. Despite these observations, it is likely there is much redundancy in the function of SAGA and SLIK/SALSA and the full significance of any different activities of these complexes waits to be determined.

9. SAGA today

Since the discovery of SAGA and the initial genetic and biochemical characterizations discussed in this review, labs around the world have continued to study its functions. Today, SAGA is known to be conserved throughout eukaryotes, with critical functions demonstrated in Drosophila [163, 164], plants [165], and humans [166]. In humans, there are close connections between SAGA and human disease [167, 168].

A major advance in our understanding of SAGA has come from structural studies (reviewed in [169–171]). It has been particularly gratifying that these studies have confirmed the model for distinct functional modules within SAGA, first proposed by the early studies reviewed in this article (Figure 3; [2, 3]). Two of the first structural studies, one using electron microscopy [116] and the other using mass spec/crosslinking studies [145], produced a similar picture of SAGA, with a structure corresponding to the modules defined by the early genetics and biochemistry. Since then, higher resolution studies by crystallography or cryo-EM have provided further insights into the structural organization, as well as elucidating the likely mechanisms by which the deubiquitinating (DUB) module [172], the TBP-recruitment module [10], and the histone acetylation module [108] function.

Highlights.

• SAGA was identified by purification of protein complexes that contained the histone acetyltransferase Gcn5.

• Characterization of SAGA fulfilled genetic predictions of a protein complex containing groups of factors identified by mutant hunts in yeast, including Spt and Ada proteins.

• Biochemical studies identified other SAGA components, including TAF proteins, Ubp8 and Tra1.

• Genetic and biochemical studies suggested that SAGA is a coactivator complex with multiple modules for histone acetylation, histone deubiquitylation, and TBP recruitment.