Structural Insights into the Assembly and Function of the SAGA Deubiquitinating Module

See allHide authors and affiliations

Science  21 May 2010:
Vol. 328, Issue 5981, pp. 1025-1029
DOI: 10.1126/science.1190049


SAGA is a transcriptional coactivator complex that is conserved across eukaryotes and performs multiple functions during transcriptional activation and elongation. One role is deubiquitination of histone H2B, and this activity resides in a distinct subcomplex called the deubiquitinating module (DUBm), which contains the ubiquitin-specific protease Ubp8, bound to Sgf11, Sus1, and Sgf73. The deubiquitinating activity depends on the presence of all four DUBm proteins. We report here the 1.90 angstrom resolution crystal structure of the DUBm bound to ubiquitin aldehyde, as well as the 2.45 angstrom resolution structure of the uncomplexed DUBm. The structure reveals an arrangement of protein domains that gives rise to a highly interconnected complex, which is stabilized by eight structural zinc atoms that are critical for enzymatic activity. The structure suggests a model for how interactions with the other DUBm proteins activate Ubp8 and allows us to speculate about how the DUBm binds to monoubiquitinated histone H2B in nucleosomes.

The covalent modification of chromatin is a core feature of eukaryotic transcription (1). The SAGA complex regulates genes transcribed by RNA polymerase II by carrying out multiple functions that include histone acetylation and deubiquitination (2, 3). The 1.8-megadalton SAGA complex, which contains 21 proteins conserved from yeast to humans (4), plays a role in transcription activation and also couples transcription elongation with export of the nascent RNA through the nuclear pore complex (5). SAGA is recruited to promoter regions by activator proteins, where it catalyzes the cleavage of monoubiquitin from lysine-123 (K123) of histone H2B, as well as acetylation of histone H3 (6). Both of these activities are thought to be important for evicting nucleosomes from the promoter region and for facilitating transcription elongation (7). The yeast SAGA complex has been a model for studying SAGA function in the transcription cycle. Histone acetylation by SAGA is mediated by the Gcn5 histone acetyltransferase subunit, whereas histone H2B deubiquitination is mediated by a discrete subcomplex known as the deubiquitinating module (DUBm) (3, 8). The yeast DUBm comprises four proteins: the ubiquitin-specific protease (Ubp8) ubiquitin hydrolase, Sgf11, Sus1, and Sgf73 (9, 10). Sgf73 tethers DUBm to the rest of the SAGA complex through a central domain; the N-terminal domain forms an integral part of the DUBm (9). Sgf73, along with Sus1, also facilitates SAGA’s role in nuclear export by binding to components of the nuclear pore complex (11). Although Ubp8 contains a ubiquitin-specific hydrolase (Usp) domain (12), the protein is inactive unless it is in complex with the other three DUBm proteins (9, 13). The structural basis for Ubp8 activation is not understood, nor is its dependence on the other three DUBm proteins to form a stable complex that can specifically target H2B for deubiquitination.

Drosophila and human SAGA include a DUBm that functions analogously to the yeast DUBm and comprises homologs of the yeast proteins (10, 14, 15). In addition, the human DUBm, consisting of USP22, ATXN7L3, ENY2 and ATXN7 (homologs of Ubp8, Sgf11, Sus1, and Sgf73, respectively), also plays a role in telomere maintenance by regulating the stability of TRF1, a component of the telomeric shelterin complex (15). These additional functions suggest that the human DUBm may have other substrates that remain to be identified. The importance of DUBm function in humans is highlighted by the observations that USP22 is a cancer stem cell marker (16) and that ATXN7 is the affected protein in the polyglutamine expansion disease, spinocerebellar ataxia type 7 (17, 18).

In order to gain insight into the basis for DUBm assembly, activity, and substrate binding, we determined the structure of the heterotetrameric DUBm with and without bound ubiquitin aldehyde (Ubal), an inhibitor that forms a reversible covalent bond with the active-site cysteine (19, 20). Soluble, enzymatically active complex was obtained by coexpressing the intact Ubp8 (4741 amino acids), Sgf11 (99 amino acids), and Sus1 (96 amino acids) proteins along with an N-terminal fragment of Sgf73 (residues 1 to 96) that is sufficient for enzymatic activity (9). The coexpressed DUBm used in the structural study cleaves ubiquitin C-terminal 7-amido-4-methylcoumarin (ubiquitin-AMC) with an enzymatic rate (kcat) of 0.17 s−1 and a Michaelis constant (Km) of 1.5 μM, giving a kcat/Km value of 1.1 × 105 M−1 s−1 (fig. S1). The activity is inhibited by Ubal with an apparent dissociation constant (Ki) of 0.2 μM (fig. S2). The coexpressed DUBm can also cleave ubiquitin from the free histone substrates, H2B and H2A (fig. S3), which demonstrates that the complex retains the reported deubiquitinating activity of the complete SAGA complex. The crystal structure of the DUBm bound to Ubal was determined by selenomethionine multiwavelength anomalous dispersion phasing and refined at 1.90 Å resolution to an R/Rfree of 15.5/20.3%. The 2.45 Å resolution structure of free DUBm was determined by molecular replacement with the high-resolution DUBm structure as a search model, and refined to an R/Rfree of 20.4/28.1%. Data collection and refinement statistics are shown in table S1. Sequence alignments for all four DUBm proteins, annotated with the positions of secondary structural elements, are shown in fig. S4. The description of the structure that follows focuses on the higher-resolution model of the Ubal-bound complex, except where noted.

The structure of the DUBm shows Ubp8, Sgf11, Sus1, and Sgf73 to be remarkably intertwined (Fig. 1, A and B). Each polypeptide contacts the other three proteins in the complex, which explains why active complex formation requires the presence of all four DUBm subunits (9, 13). With the exception of Ubp8, which contains two globular domains separated by a short linker, the conformations of the other three proteins are largely dependent on their interactions with other DUBm proteins. The complex is organized into two lobes, each containing one of the two domains of Ubp8: the N-terminal zinc finger–ubiquitin binding protein (ZnF-UBP) domain and the C-terminal enzymatic Usp domain (Fig. 1, A and B). The ZnF-UBP lobe (Fig. 1C) is organized around the long N-terminal helix of Sgf11, which binds the Ubp8 ZnF-UBP domain on one face and Sus1 on the opposing face. The 40 N-terminal residues of Sgf73 comprise the remainder of this lobe and primarily contact the Ubp8 ZnF-UBP domain and Sus1 with short helices (Fig. 1, A and C). The second lobe of DUBm contains the ~350–amino acid Usp domain of Ubp8 (Fig. 1, A and D). Ubiquitin aldehyde binds, as expected, to the “fingers” region of the Usp domain (21), which consists of an antiparallel β sheet and a zinc-binding domain (Fig. 1D), and forms a covalent bond between the modified C-terminal glycine and the active-site cysteine. To the opposite side of the active site lies the N-terminal zinc finger domain of Sgf11, which binds to the face of the Usp domain (Fig. 1D). An extended well-ordered linker region joins the Sgf11 zinc finger with its N-terminal helix (Fig. 1B) and forms a bridge between the two lobes of the DUBm while forming additional contacts with Ubp8. The interactions at the interface between the two lobes are mediated primarily by 50 C-terminal residues of the Sgf73 fragment (45 to 95), which are buried between the two lobes. This portion of Sgf73 contains a short α helix and a zinc finger domain that are connected by 20 residues that lack secondary structure, yet are well ordered and have low temperature factors. Sus1 helices α2 and α3 mediate additional interface contacts with Ubp8 on the back face of the Usp domain. Together, Sgf11, Sus1, and Sgf73 bury a total of 7910 Å2 of the Ubp8 surface. The overall features of the DUBm complex are preserved in the structure of apo DUBm, which crystallizes with different crystal packing. This indicates that the unusual arrangement of proteins in the DUBm complex is not an artifact of crystal contacts.

Fig. 1

Overall structure of the SAGA DUBm. (A) View of DUBm bound to ubiquitin aldehyde (Ubal). The ZnF-UBP lobe and Usp lobe are labeled. The DUBm protein coloring scheme is shown at left; Ubal is yellow, with zinc atoms shown as red spheres. Disordered loop residues are indicated with dotted lines. Residues in Sgf73 that are ordered in the structure of the apo DUBm are shown in purple. (B) View of DUBm rotated 180° about the vertical as compared with (A). (C) View of the ZnF-UBP lobe. The Ubp8 ZnF-UBP domain and Sus1 bind on opposite faces of the Sgf11 N-terminal helix. The Sgf73 residues shown in purple are disordered in the Ubal-bound complex but are ordered in the apo complex. The insertion point for polyglutamine expansion in the human homolog, ATXN7, is indicated by an inverted triangle containing the letter, Q. (D) View of the Usp lobe. The fingers region to which ubiquitin (yellow) binds is at the bottom and the Sgf11 zinc finger (magenta) is at the top.

The binding of Sgf73 between the two lobes of the DUBm is consistent with its role in anchoring the DUBm to the rest of the SAGA complex (9, 10) and in helping to activate DUB activity. As compared with Sgf11 and Sus1, which are also nonglobular, Sgf73(1–96) appears to be the most dependent on other DUBm proteins for its observed conformation, because it contains the fewest intrachain contacts. In the intact protein, the remaining C-terminal residues of Sgf73 link the DUBm to the SAGA complex (Fig. 1A), with residues 227 to 402 mediating interactions with the other SAGA proteins (9, 10). There is a low-complexity acidic region (residues 130 to 162) between the SAGA-binding and DUBm-binding regions of Sgf73, which suggests that the DUBm might be flexibly tethered to the remainder of the SAGA complex. Sgf73(1–96) contains a zinc finger that is coordinated by two cysteines (C78, C81), and a histidine (H93) (6). The fourth ligand in the DUBm-Ubal complex is an aspartic acid (D95) that is conserved in the human homolog, ATXN7 (fig. S4); however, this residue does not coordinate the zinc in the apo DUBm structure and, instead, is part of an extended C-terminal helix. It is likely that, in the full-length protein, the histidine (H97) or cysteine (C98) residues that are missing from the crystallized fragment serve as the fourth zinc-coordinating ligand.

The incorporation of Sgf73 into the DUBm depends on the integrity of the zinc finger, because a slightly shorter fragment containing residues 1 to 92 is defective in binding to the other DUBm subunits (9). We therefore tested the effect of zinc chelation on the incorporation of Sgf73 and the other DUBm subunits into the complex. In the absence of EDTA, the DUBm sediments with an S20,w value of 4.9 (Fig. 2), as expected for a single globular 90-kD species with a somewhat asymmetric axial ratio. The addition of 2 mM EDTA causes the complex to dissociate into two principal species corresponding to Sgf73 and the Ubp8-Sgf11-Sus1 heterotrimer (Fig. 2). We note that the Sgf11-Sus1 heterodimer sediments as multiple species (fig. S5) that do not account for any of the observed peaks. On the basis of the DUBm structure, the stable association between Ubp8, Sgf11, and Sus1 is most likely maintained by contacts between the long N-terminal helix of Sgf11, Sus1, and the ZnF-UBP domain. The loss of complex integrity is accompanied by a loss of activity, which is observed upon addition of increasing concentrations of EDTA, as well as increasing incubation time (fig. S6). In addition to dissociating Sgf73, which is known to compromise DUBm activity (10), it is also possible that the EDTA further affects enzymatic activity by chelating the zinc atoms in the enzymatic domain of Ubp8 (Fig. 3, A and B). The importance of wild-type Sgf73 function is highlighted by the effect of polyglutamine expansions in the N terminus of the human homolog, ATXN7 (18). These insertions of up to 300 glutamine residues compromise SAGA activity and give rise to defects in transcription (22). The polyglutamine expansion occurs near the N terminus of Sgf73 (18), in the coil region between the first two helices (Fig. 1C). It is possible that large insertions in this region disrupt Sgf73 association with the other DUBm proteins, which compromises enzymatic activity or causes them to dissociate from the SAGA complex. Alternatively, polyglutamine tracts may cause Sgf73 to aggregate and, therefore, to fail to bind to the other DUBm proteins.

Fig. 2

Dissociation of the Sgf73 fragment from the DUBm complex. Velocity sedimentation of the DUBm in the presence (blue) and absence (black) of 2 mM EDTA, superimposed with the sedimentation of Sgf73(1–106) alone.

Fig. 3

Structural details of the DUBm. (A) Structural zinc coordination in the Ubp8 Usp domain located between α9 and α10. (B) Structural zinc coordination in the Ubp8 Usp domain located between α12 and α13. (C) Superposition of the structure of Sus1/Sgf11(1–33) (copper) (PDB ID 3KIK) with the corresponding residues of Sus1 (blue) and Sgf11 (magenta) in the DUBm. The arrow indicates the additional helical residues in Sgf11 seen in the DUBm structure.

The structures of the individual DUBm subunits contain some unexpected features. The overall fold of the Ubp8 catalytic domain is very similar to that of other Usp proteins (12). However, the catalytic domain of Ubp8 contains two bound structural zinc atoms (Fig. 3, A and B) that have not been previously observed (12). One of the bound zinc atoms (Fig. 3A) is located between helices α9 and α10 (see fig. S4 for numbering) and is coordinated by H170, C174, C182, and C185, whereas the second bound zinc (Fig. 3B) is located in a nonconserved loop region between helices α12 and α13 and is coordinated by H250, C271, C273, and H276. Although these bound zinc atoms are not conserved among the vast majority of USP enzymes, sequence comparisons indicate that both of these sites are conserved in the Ubp8 homologs, human USP22 (fig. S4) and Drosophila Nonstop, which are SAGA DUBm subunits (14, 15). Another interesting feature of Ubp8 is that the ZnF-UBP domain—which, in other USP enzymes such as Usp5, binds ubiquitin (23)—has its ubiquitin-binding face occluded in the DUBm structure. That region of the domain, which is centered around β strands 2 to 5 (fig. S4), is largely buried at the interface with Sgf11. Although a role for ubiquitin binding to this domain cannot be ruled out when Ubp8 is not part of the observed complex, the absence of conserved ubiquitin-binding residues in Ubp8—as well as in the human homolog, USP22 (24)—makes this unlikely. Other ZnF-UBP domains that similarly lack conservation in the ubiquitin-binding residues and do not bind ubiquitin, such as in USP33 (25), may also play structural roles in those enzymes.

A structure of Sus1 that was previously determined in complex with residues 1 to 33 of Sgf11 (26) is virtually superimposable with the corresponding residues of Sgf11 and Sus1 in the DUBm complex (Fig. 3C). In the DUBm structure, however, which contains the intact 96–amino acid Sgf11 protein, the N-terminal Sgf11 helix is extended by an additional 10 amino acids at its C terminus; this presumably is due to the additional contacts with the Ubp8 ZnF-UBP domain (Figs. 1C and 3C). The ZnF-UBP domain also contacts helix α1 of Sus1 and likely helps to stabilize its interaction with Sgf11. In the structure of Sus1 bound to Sgf11(1–33), helix α1 of Sus1 adopts widely variable orientations in the four complexes found in the asymmetric unit (26), only one of which contacts Sgf11 and matches the closed conformation observed in the DUBm. The closed conformation of Sus1 is also observed when it binds to Sac3 as part of the TREX-2 complex, which functions in mRNA export (27). Thus, the N-terminal helix of Sgf11 plays a scaffolding role similar to that of Sac3, which also forms a long α helix to which multiple proteins bind.

A comparison of the ubiquitin aldehyde (Ubal) DUBm complex with the apo DUBm complex reveals that a number of structural changes occur in Ubp8 when ubiquitin binds to the DUBm. The overall features of ubiquitin binding, including the arrangement of the catalytic residues, C146, H427, N443, and D444 (see fig. S7 for view of active-site residues in electron density), are similar to those seen in structures of other Usp enzymes bound to Ubal (21, 28, 29). Ubiquitin binds to the fingers domain of Ubp8, with its hydrophobic face (including I44) interacting with the β sheet, whereas the C-terminal tail of ubiquitin binds in a conserved groove that leads to the active site (Fig. 1, A and D). In the absence of ubiquitin, a number of structural differences occur in Ubp8. The fingers region that binds ubiquitin in the complex remains in an open conformation in apo DUBm (Fig. 4A), but the zinc-binding module and several β strand residues become less well ordered, which results in weak electron density that could not be completely traced (fig. S8A). Nonetheless, the fingers domain does not appear to collapse and occlude the ubiquitin-binding pocket, as is seen in the structure of the human USP8 apoenzyme (a Usp family member that is not an ortholog of yeast Ubp8) (30) (Fig. 4A). Most striking, however, are the changes that occur near the enzyme active site (Fig. 4B; electron density shown in fig. S8). Seven Ubp8 residues located adjacent to the ubiquitin C terminus, 228 to 233, are disordered in the absence of ubiquitin binding (Fig. 4B). In addition, loop residues 421 to 426 move inward by up to 2.0 Å toward the ubiquitin tail–binding groove (Fig. 4B). The loop and the disordered region contain residues that contact ubiquitin directly; this suggests that binding of the ubiquitin C-terminal tail triggers the observed conformational change. Despite these conformational differences, the active-site residues (C146, H427, N443, and D444) are in their catalytically competent orientation in both the presence and absence of ubiquitin (Fig. 4B and fig. S8B). This is in contrast to deubiquitinating enzymes, such as HAUSP/USP7 (21), in which the active-site residues are not in a catalytically competent arrangement in the apoenzyme.

Fig. 4

Interaction between the DUBm and substrate. (A) Alignment of apo Ubp8, Ubp8-Ubal, and USP8. In the apo USP8 structure, the fingers region is collapsed and cannot accommodate ubiquitin without a conformational change. (B) Structural changes in the region of the Ubp8 active site. Residues 228 to 233 of the Ubp8 apoenzyme (red) are disordered in the absence of ubiquitin and a loop, 421 to 426, shifts into the groove where the C-terminal tail of ubiquitin (yellow) would bind. A second loop, which contains the active-site cysteine, C146, maintains the same conformation in the presence and absence of ubiquitin and is stabilized by contacts with the Sgf11 zinc finger (magenta). The other active-site residues, H427, N443, and D444, also remain in the same orientation in both structures. (C) Structure of the DUBm showing a surface representation of Sgf73. The color scheme used is the same as in Fig. 1. (D) Electrostatic surface potential of the ubiquitin-binding face of the DUBm. The region of strong electropositive surface potential (blue) is due to the Sgf11 zinc finger.

How does complex assembly activate Ubp8? In the absence of the other DUBm proteins, we find that Ubp8 is only about 1/200,000th as active as the intact DUBm (fig. S9) and shows substantial aggregation (fig. S10). A comparison of the DUBm structures with and without Ubal suggests that interactions among the DUBm proteins may stabilize a conformation of Ubp8 that is catalytically competent and able to bind ubiquitin. In the active-site region, the Sgf11 zinc finger directly contacts the loop in Ubp8 that contains the active-site cysteine, C146 (Fig. 4B). These contacts with Sgf11 may help to maintain the active-site residues in a catalytically competent conformation even when the neighboring residues (229 to 235) become disordered. Another way in which the other DUBm proteins may activate Ubp8 is by stabilizing a conformation of the fingers domain that favors binding of the globular portion of ubiquitin. The human USP8 apoenzyme (not an ortholog of yeast Ubp8), for example, has a partially occluded ubiquitin-binding pocket that would require a conformational change to accommodate ubiquitin (Fig. 4A). Because the β sheet of the Ubp8 fingers region is in an open conformation even in the absence of ubiquitin (Fig. 4A), it is possible that interactions between Ubp8 and the other DUBm subunits help to maintain a conformation that favors ubiquitin binding. Interactions of the upper portion of the fingers domain with Sgf73 helix α2 and with Sus1 (Figs. 1A and 4C) form extensive contacts that may favor the open conformation of Ubp8. Sgf73 may also play a general role in stabilizing the conformation of the Usp domain. Ubp8 forms the most extensive interface with Sgf73, with a total area of 3075 Å2, as compared with the other pairwise domain interactions in the DUBm. Note that Sgf73 is the “mortar” that holds together the two lobes of the DUBm complex (Fig. 4C); it promotes interactions between the two lobes and aligns Sgf11 and Sus1, which also contact the Usp domain at the interface between the two lobes. Although the structure of the Ubp8 Usp domain on its own is not known, it is possible that these extensive interactions with Sgf73 may also help to stabilize the overall USP fold or to affect protein dynamics in a way that favors the catalytically competent structure.

The structure of the SAGA DUBm suggests how this module interacts with its natural, in vivo substrate, monoubiquitinated histone H2B within a nucleosome. The electrostatic surface potential of the DUBm (Fig. 4D) reveals a basic region that could favor interactions with the negatively charged DNA when the nucleosome is positioned with ubiquitinated K123 of H2B in the active site of Ubp8. This region of the DUBm is positively charged because of the zinc finger module of Sgf11, which contains four basic residues (R78, R84, R91, and R95) (fig. S11) that are conserved in ATXN7L3, the human homolog of Sgf11. Two additional C-terminal arginine residues, R98 and R99, which are disordered in the structure but are also conserved in the human homolog, would further contribute to the strong positive charge in this region. Figure S12 shows a model for how a yeast nucleosome (31) monoubiquitinated at K123 of H2B can dock on the DUBm, with ubiquitin in the active site as seen in the Ubal-bound structure (Fig. 1D). This arrangement brings the basic patch on the DUBm in close approach with the DNA, which favors interactions with the sugar-phosphate backbone. The interdependent structural and functional roles of the four SAGA DUBm proteins in mediating biochemical activity, substrate binding, and incorporation into the larger SAGA complex is likely an example of how other subcomplexes of SAGA, and other coactivator and corepressor complexes, cooperate to regulate transcription.

Supporting Online Material

Materials and Methods

Figs. S1 to S12

Table S1


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.
  3. We thank A. DiBello for helping with the ubiquitin aldehyde synthesis, E. Hurt for providing the Sus1 clone, S. Corcoran at the General Medicine and Cancer Institutes Collaborative Access Team (GM/CA-CAT) beamline at the Advanced Photon Source for assistance with data collection and processing, and M. Bianchet for advice and discussions. C.E.B. is supported by a Ruth Kirchstein Fellowship from the National Institute of General Medical Science, NIH (F32GM089037), and T.Y. is a Special Fellow of the Leukemia and Lymphoma Society. GM/CA-CAT has been funded in whole or in part with federal funds from the National Cancer Institute (Y1-CO-1020) and the National Institute of General Medical Science (Y1-GM-1104), NIH. Use of the Advanced Photon Source was supported by the U.S. Department of Energy, Basic Energy Sciences, Office of Science, under contract no. DE-AC02-06CH11357. Coordinates and structure factors have been deposited in the Protein Data Bank with accession codes 3MHS (DUBm bound to Ubal) and 3MHH (apo DUBm).
View Abstract

Navigate This Article