In human cells, large multi-protein transcription co-activators, such as chromatin remodelers or histone acetyltransferases, play critical roles in gene-expression regulation and are often implicated in disease. However, interrogating their structures or analyzing their properties and interactions in different organs or in medically relevant cell-types is hindered by the difficulty in purifying them. We overcome this difficulty by applying an affinity-ligand composed of a small molecule that specifically recognizes a particular domain in a given co-activator multiprotein complex. This molecule is coupled to a desthiobiotin moiety, which allows binding to streptavidin beads and can be eluted using biotin. To demonstrate the universal utility of this idea to practically any co-activator complex from any cell source we synthesized a compound conjugating desthiobiotin to GSK4027, a molecule that targets the bromodomain in the GCN5/PCAF catalytic subunit of SAGA and ATAC acetyltransferase complexes. Employing this heterobifunctional affinity ligand and a novel purification scheme adapted to low-abundance complexes, we isolated the 1.6 MDa SAGA complex from two cancer cell lines to high degree of purity and activity. We then solved the structure of the isolated 20-subunit SAGA complex to high-resolution (2.3-3 Å) by cryo-EM, elucidating for the first time the molecular details of how its enigmatic splicing module anchors into SAGA. Analyzing these interactions raises the possibility that SAGA serves to relay this module to the splicing machinery. Our approach will be instrumental for characterizing many other multi-protein complexes from medically important sources.
Transcription regulation controls many cell decisions including growth, differentiation and response to external cues. DNA-sequence-specific transcription factors recruit co-activators, many of them are large multi-protein complexes, to regulate transcription by impacting chromatin structure, enlisting other co-activators or by a direct interaction with the transcription pre-initiation complex. In line with their pivotal role in transcription regulation many co-activators are implicated in human disease, notably cancers. A case in point is the SAGA complex that alters chromatin structure via two enzymatic activities: histone acetylation (HAT) and deubiquitination (DUB).
Human SAGA regulates both transcription initiation and elongation and is essential for normal embryonic development. It also has many non-histone targets and thus impacts gene expression at multiple levels from transcription to protein stability. The ∼1.6 MDa complex comprises 20 subunits organized functionally and structurally in five modules: a histone acetyltransferase (HAT) module, a histone deubiquitinase (DUB) module, the TRRAP subunit that serves as a docking surface for transcription activators that recruit SAGA to enhancers, a metazoan-specific splicing factors module (SPL) with an enigmatic role and a central module that scaffolds the complex by physically connecting to all other modules. Cryo-electron microscopy (EM) maps of human and yeast SAGA describe well the core and TRRAP parts but do not reveal at high-resolution the HAT, DUB or Splicing modules.
GCN5/PCAF is the catalytic subunit of the SAGA HAT module. This subunit is also incorporated into the essential 10-subunit co-activator ATAC, where the HAT module is identical to that of SAGA with the exception of one subunit being replaced by a homologue. SAGA and ATAC are regulating distinct sets of genes.
SAGA can be recruited to chromatin by the c-MYC oncoprotein whose deregulation leads to unfavorable variations in the expression of target genes in the vast majority of cancers. The HAT activity of SAGA is necessary for the maintenance of MYC oncogenic transcription program in many tumors. In addition, USP22, the catalytic subunit of the DUB module in SAGA, is over-expressed in aggressive cancers and was identified in microarray screens as part of an 11-gene ‘death from cancer’ signature for therapy-resistant tumors. Given the involvement of co-activators such as SAGA in human pathology there is an urgent need to characterize the composition, interactome, activity and structure of endogenous co-activators in cell lines derived from tumors or other human diseases. However, this remains almost impossible and thus a largely untapped source of information, due to the difficulty in purifying these complexes. Although affinity-tags fused to endogenous subunits of a desired complex via CRISPR/cas9 engineering could alleviate this issue, many cell-lines are not amenable to this technology. Moreover, the efforts that such tagging entails impede large comparative studies where complexes from several different cell-lines are required. Furthermore, optimized purification schemes are required due to the very low abundance of these complexes in cells. For example, a recent study reported only a few picomoles of isolated SAGA from 30 liters of HeLa cell culture.
Affinity-ligands, molecules that bind the desired complex and can be coupled to a solid support (e.g. agarose or magnetic beads) offer an attractive alternative to affinity-tags because they can be applied to many different cell-lines. For example, alprenolol is a ligand of beta-adrenergic receptors that was used to select only the functional active receptors during the purification of the over-expressed protein. In another instance, trapoxin, a strong inhibitor of histone acetylation, was covalently linked to a matrix to capture the first catalytic subunits of mammalian HDAC complexes. These examples also demonstrate however the difficulty in applying affinity-ligands to the endogenous co-activator case. The low concentration of co-activators requires very high affinity-ligands, unlike the alprenolol example. Furthermore, these multi-protein complexes need to be isolated intact, in contrast to the HDAC isolated subunit example.
In recent years we witness a surge in the number of designed small molecules/inhibitors that target with high affinity and selectivity the active site of co-activators, or their “reader” domains that stabilize interactions with chromatin through recognizing specific histone modifications. However, so far, these novel molecules were not exploited for developing a purification scheme that yields pure and active sample of these low-abundance co-activator, let alone enables their structure determination.
In this study, we obtained pure and highly active endogenous human SAGA from two different unmodified cell lines, K562 and HeLa, by employing an affinity-ligand. This compound is composed of an inhibitor for the bromodomain of the enzymatic subunit GCN5/PCAF in the SAGA HAT module and is coupled to desthiobiotin. We introduced several innovations in the design of the affinity-ligand as well as in the purification scheme that was adapted to low abundance complexes. Using this new method, we determined the structure of the purified SAGA by cryo-EM, revealed for the first time at high-resolution the splicing factor module in SAGA and elucidated in detail how this module anchors into the co-activator complex. Detailed comparison of these interactions to the docking of this module in the U2snRNP, suggests that SAGA serves to relay the splicing module to its assembly in the pre-spliceosome. Our innovations critically contributed to the results obtained and will guide future endeavors to characterize multi-protein complexes from many medically important sources.
Low abundance co-activators, such as SAGA are present at a concentration of roughly 1 nM in nuclear extracts. On the other hand, the majority of inhibitors for catalytic enzymes or for histone modification “readers” harbored by co-activators have unfortunately a K d higher than the very low nanomolar range. Hence, the first obstacle that we needed to overcome, in order to make the use of most available inhibitors feasible, was to develop a method for concentrating nuclear extracts. Our experience in yeast and human cells showed that differential precipitation using a high-molecular-weight polyethylene glycol (PEG) is a simple technique to achieve that goal. Indeed, due to PEG tendency to precipitate first the larger molecules, this technique can also be considered as a real first purification step because it increases the proportion of large complexes in the sample with minimal losses. We worked with nuclear extract derived from 3 L of K562 or 4.5 L of HeLa S3 cells in suspension due to sample requirements for high-resolution structural studies. However much smaller volumes can be used for other applications since PEG precipitation procedure can be easily adapted to practically any volume of nuclear extract.
Figure 1:Purification scheme of SAGA and ATAC using affinity-ligand. Schematic overview illustrating the stepwise purification strategy adapted to low-abundance complexes employed to isolate SAGA and ATAC (PEG: polyethylene glycol).
We then screened the literature for high affinity and selectivity binders for SAGA. We chose eventually GSK4027 which targets the bromodomain, a “reader” that recognizes acetylated lysines, in the catalytic HAT subunit GCN5/PCAF with a K d = 1.4 nM. This compound is highly selective to GCN5/PCAF, presenting much lower affinity to other bromodomains. One drawback of using GSK4027 as a ligand is that it cannot distinguish between the two human complexes that harbor GCN5/PCAF namely ATAC and SAGA. However, we posited that due to the predicted size difference between SAGA (1.68 MDa) and ATAC (0.9 MDa), in-silico classification of the cryo-EM dataset would be able to separate the two complexes.
Figure 2:Synthesis of the affinity-ligands. a, structure of GSK4027. b, Structure of affinity-ligands 1a and 1b. c, details of the synthetic route to prepare the compounds.
To mediate the binding of GSK4027 to a solid matrix we chose to conjugate it to desthiobiotin which can be attached to streptavidin beads (K d = 10−11 M) and efficiently eluted under native conditions with biotin (K d = 10−16 M). From X-ray crystallographic data on GSK4027 bound to the human GCN5/PCAF bromodomain, the 4-position on the pendant phenyl ring seems to point outside the binding pocket. This is indeed confirmed by results obtained by Bassi et al. who developed a GSK4027-based PROTAC approach for modulating GCN5/PCAF immune cell function. This position was thus selected for derivatization with desthiobiotin while preserving binding to the bromodomain. Consequently, we designed conjugates 1a and 1b as ligands for immobilization of SAGA/ATAC on streptavidin beads and purification by affinity chromatography. The GSK4027 and desthiobiotin moieties are connected together through an oligoethylene spacer of variable length to modulate access of each ligand to its respective target: GSK4027 to SAGA and desthiobiotin to streptavidin beads and alleviate possible steric hindrance.
Prior to our work the synthesis of GSK4027 was achieved by Humphreys et al. in 7 steps. The compound was obtained as a mixture of four enantiomers and final separation of the racemate by preparative chiral HPLC was necessary to get the desired single enantiomer in an overall yield of 1.2%. In this study, we developed a stereo-controlled synthetic route to prevent the formation of mixed isomers. This approach aligns with a methodology previous
