Overall structure of the Ubp3/Bre5 hetero-tetramer complex
We used X-ray crystallography to determine the high resolution structure of the NTF2-like domain of Bre5 (Bre5-NTF2, residues 1–146) in complex with an N terminal high affinity binding domain of Ubp3 (residues 189–233, called Ubp3-Nterm to distinguish it from the C-terminal Ubp3 catalytic domain). The crystals form in the space group P212121
with 2 subunits each of Bre5-NTF2 and Ubp3-Nterm per asymmetric unit cell. The structure was determined by molecular replacement using Bre5-NTF2 as a search model 17
and refined to 1.7Å resolution with good refinement statistics and geometry ().
Data collection and structure refinement statistics.
Bre5-NTF2 forms a tightly associated homodimer with each Bre5-NTF2 subunit interacting with a domain of Ubp3-Nterm (). The Bre5-NTF2 dimer in the Ubp3/Bre5 complex adopts a mixed α/β fold that is similar to the free Bre5-NTF2-like domain and other NTF2-like domains. Briefly, each Bre5-NTF2 subunit contains a 6-stranded (β1–β6) anti-parallel β sheet that forms a curved platform for three α helices (α1–α3) that lie on the concave side of the β-sheet. The dimer interface is formed along the convex surfaces of the β-sheets from each of the subunits such that residues from β1 and β3–β6 mediate both hydrogen bond and van der Waals contacts.
The Ubp3-Nterm domain forms a 3-residue loop, followed by a 7-residue β1-strand, sharp turn and a 15-residue α1-helix (). The strand and helix are nearly perpendicular to each other forming an L-shape. Residues 190–205 of Ubp3-Nterm, N-terminal to the 3 residue loop could not be traced into the electron density map and are presumably disordered.
The Ubp3-Bre5 interface
Bre5-NTF2 and Ubp3-Nterm form nearly symmetrical interactions within the heterotetramer. These interactions are predominantly hydrophobic in nature and can be separated into two distinct regions of contact; between the Ubp3-Nterm β1-sheet-turn with the β3-sheet and α1- helix of Bre5-NTF2, and between the N-terminal half of the α1-helix of Ubp3-Nterm with the L2 loop (β4– β5 turn) and, to a lesser extent, the L1 loop (β1– β2 turn) of Bre5-NTF2 ().
Ubp3-Nterm β1 residues 208–210 form anti-parallel β-sheet hydrogen bonds with the Bre5-NTF2 β3 residues 81–83. The side chains of Ubp3-Nterm β1 residues Leu 208, Phe209, Ile 210, Asn 211 and Phe 212 also make a series of side chain van der Waals interactions with several residues in the β3-strand (Lys 80, Leu 81, Lys 82 and Leu 83) and α1-helix (Pro 10, Gln 14, Tyr 17, Glu 18 and Arg 21) of Bre5-NTF2 (). In the turn that proceeds the β1 sheet of Ubp3-Nterm, Phe 217 also makes a van der Waals contact with the aliphatic region of Lys 82 of Bre5-NTerm and Ubp3-term Glu 216 makes the only side chain hydrogen bond in this region with Bre5-NTF2 Lys 80 ().
The Ubp3-Nterm α1-helix also makes predominantly van der Waals contacts with the L1 and L2 loops of Bre5. Residues Ala 220, Ser 221, Gln 223 and Arg 224 of Ubp3-Nterm make extensive van der Waals interactions with the L2-loop residues Pro 112, Val 113 and Tyr 114 of Bre5-NTF2 (). The side chain Arg 224 nitrogens of Ubp3-Nterm also makes a hydrogen bond with the main chain of Val 113 within the Bre5-NTF2 L2 loop as well as hydrogen bond and van der Waals contacts with Asp residues 51 and 52, respectively, within the L1-loop of Bre5-NTF2 ().
A sequence alignment of Ubp3-Nterm with other yeast Ubp3 homologues shows high sequence conservation of residues in the β1-turn-α1 region that contact Bre5-NTF2 (). In particular, this region shows the highest conservation outside of the catalytic domain and the majority of the residues that mediate Bre-NTF2 contacts are either strictly or highly conserved within the yeast Ubp3 homologues (). Not surprisingly, the residues in Bre5-NTF2 that mediate Ubp3 contacts are also highly conserved within the yeast Bre5 homologues (). Taken together, these findings suggest that the yeast Ubp3 homologues adopt a similar β1-turn-α1 structure and mediate analogous contacts with their corresponding Bre5 homologues. Interestingly, only part of the yeast Ubp3 conservation extends to the Ubp10 human ortholog (). In particular, of the Ubp3 residues that contact Bre5, only Pro 207, Val 210, Glu 216 and Phe 217 show conservation, suggesting that the yeast-specific Ubp3-Bre5 interaction may differ in human.
A comparison of the Ubp3-Nterm/Bre5-NTF2 complex with the free Bre5-NTF2 domain shows that, except for the L1 and L2 loops, the Bre5 dimer is essentially superimposable with an RMSD of 1.1Å for all atoms of Bre5 (). Importantly, the conformational changes within the L1 and L2 loops of Bre5 appear to be correlated with Bre5 interaction with Ubp3. In particular, in the absence of Ubp3 binding, the L1 loop appears flexible with 12 residues (43 to 54) untraceable in the electron density map. In the presence of Ubp3-Nterm, this loop becomes more ordered with only 4 residues (46 to 49) untraceable in the electron density map of the complex and residues Asp 51 and Asp52 of the L1 loop make contacts to Arg 24 within the α1-helix of Ubp3.
Comparison between Ubp3-bound and free Bre5-NTF2 domain
The L2 loop of the Bre5-NTF2 domain undergoes the most dramatic structural shift upon Ubp3-Nterm binding and this movement appears to be directly coupled to complex formation. In particular, in the absence of Ubp3-Nterm, this loop and the ends of the β-sheets that form it are flipped away from the body of the rest of the Bre5-NTF2 protein. Indeed, this “flipped out” conformation would prevent Ubp3-Nterm binding due to steric occlusion (). In the Ubp3-Nterm/Bre5-NTF2 complex, this loop flips inward by about 5 Å toward the body of the protein and forms part of the binding pocket for Ubp3-Nterm (). The movement of this loop in the Bre5-NTF2 domain not only helps create a binding site for Ubp3-Nterm but residues Pro 112, Val 113 and Tyr 114 of the loop also directly participate in interactions with the α1-helix of Ubp3-Nterm, as described above. Conservation of Pro 112 and Tyr 114 of the L2 loop and Asp 52 of the L1 loop suggests that the L1 and L2 loops of Bre5 may play important roles in Ubp3 association.
Structure-based mutagenesis of the Ubp3-Bre5 interface
To probe the functional importance of the Ubp3-Bre5 interface observed in the crystal structure we carried out structure-based mutagenesis. We initially prepared single alanine replacements of several Bre5 and Ubp3 residues shown to mediate protein-protein interactions in the complex and used GST-pulldown studies to compare the protein binding properties of these mutants. Specifically, using wild-type and single alanine GST-tagged Ubp3-Nterm mutants (L208, F209, I210, N211, T212, E216, F217 and R214), we carried out pull down studies with the wild-type Bre5-NTF2 domain. We also carried out analogous pull down studies with wild-type GST-Ubp3-Nterm and several single alanine His-Bre5-NTF2 mutants (Y17, K80, K82 and E105). As can be seen in , only the single I210A Ubp3 mutation showed reduced binding relative to wild-type levels.
Bre5-Ubp3 binding studies using wild-type and mutant proteins
Since the I210A Ubp3 mutation was not sufficient to abolish binding, we next prepared proteins containing multiple mutations and assayed protein binding using pull down studies as described above. The results of these pull down studies are shown in with a summary of the mutants that were tested and their binding properties in . Mutations of residues I208, F209, I210, N211 and F217 in Ubp3 in any combination significantly reduce binding to Bre5. Together, these results reinforce the structural findings that the β1-α1 helix of Ubp3-Nterm plays an important role in Ubp3-Bre5 association.
Summary of GST pull-down studies with wild-type and mutant Bre5 and Ubp3 proteins.
To obtain more quantitative information underlying the interaction between Upb3 and Bre5 we carried out an Isothermal Titration Calorimetry study. For these studies we produced in bacteria a near full-length Ubp3 construct that contained both the Bre5 interaction domain and the catalytic domain (residues 189–912) and the NTF2 domain of Bre5. This analysis revealed that Bre5 and Ubp3 interact very strongly with a dissociation constant of 119 nM (). We also prepared an Ubp3 (189–912) mutant containing alanine substitutions at the following residues: L208, F209, I210, N211 and F217 for ITC studies with the wild-type Bre5-NTF2 and could not detect binding using analogous conditions (). This comparison, in combination with our structural findings, reported here, is consistent with our conclusion that the β1-α1 region of the Ubp3-Nterm is necessary and sufficient for tight Bre5 association.
In an earlier study, we showed that the same NTF2 domain of Bre5 interacts with an N-terminal Ubp3 (181–282) construct with a dissociation constant of 187 nM, simlilar to that reported here but with a Bre5-NTF2/Ubp-Nterm stochiometry of 2:1 instead of the 1:1 stociometry reported here 17
. Although the Bre5-NTF2 domain used in both studies are the same, the Ubp construct differs. While the earlier studies employed the shorter Ubp3 (181–282) construct, the present study employed the longer Ubp3 (189–912) construct that also includes the C-terminal catalytic domain of Ubp3 (). Although, we do not know the reason for this discrepancy, we hypothesize that other regions of Ubp that are outside of the N-term region that is shown in the crystals to contact the Bre5 NTF2-like domain, contributes to 1:1 stoichiometric Bre5/Ubp association. The high concentration of Bre5-NTF2 and Ubp-Nterm in the crystals may shift the equilibrium to the 1:1 stochiometry, despite the absence of other regions of Ubp3 that might contribute to Bre5/Ubp3 complex formation. Alternatively, a subset of the shorter Ubp3 (181–282) pool of the earlier study might have been improperly folded (consistent with our inability to produce a recombinant Ubp3 (189–233) construct in soluble form, data not shown), thus reducing the pool of Ubp3 available for Bre5-NTF2 binding. Also in an earlier study, we showed that Y42R and R139F mutations in the NTF2-like domain of Bre5 disrupted Upb3 binding in vitro
and Ubp3 function in vivo
and we had proposed that these residues might be directly involved in Ubp3 interaction. In the current structure, we see that these residues are not directly involved in Ubp3 interaction and thus hypothesize that since they are partially buried in the current structure (Y42 making intraatomic contacts and R139 making interatomic contacts) that their mutational properties may be due to a partial disruption of the Bre5 dimer structure that is required for Ubp 3 binding. Taken together, the structure of the Bre5/Ubp3 complex reported here now provides a scaffold from which to understand these earlier results.
In vivo analysis of Bre5p/Ubp3 interaction and function
To confirm that the Bre5 and Ubp3 interface characterized in vitro
is also responsible for in vivo
interaction and function of the Bre5/Ubp3 complex, we analyzed the ability of wild-type and interaction mutants of Bre5 and Ubp3 to interact and mediate deubiquitylation in cells. For this purpose, His-tagged versions of wild type or mutant Bre5 were introduced into Ubp3-HA/bre5Δ
cells as previously described 11; 17
. The ability of wild-type and mutant Bre5 protein to interact with endogenous Ubp3 was measured by co-immunoprecipitation using anti-HA antibodies followed by western blotting (, top panel). The function of the complex between Ubp3 and wild-type or mutant Bre5 was also followed by the capacity of the protein complex to mediate deubiquitylation of Sec23, a substrate for the Bre5/Ubp3 deubiquitylation complex (, top panel). In addition, His-tagged versions of wild type or mutant Ubp3 were introduced into Bre5-GFP/ubp3Δ
cells (, lower panel) and the extent of Ubp3-Bre5 interaction was also followed both by co-immunoprecipitation using anti-GFP antibodies and the capacity of the complex to mediate deubiquitylation of Sec23. As shown in , expression of wild type Bre5 or Ubp3 was able to form a complex with their endogenous counterpart and restore the deubiquitylation of Sec23 in bre5Δ
cells, respectively. However, in agreement with the in vitro
interaction studies (), a Bre5 mutant harboring both Y17A and K82A mutations displayed a modest (2–3 fold) but reproducible decrease in its binding to Ubp3 (, top panel) and showed a fairly small, but detectable, decrease in sec23 deubiquitylation (, top panel). In addition, the Ubp3 mutant harboring the L208A, F209A, I210A, and N211A substitutions was severely deficient in its interaction with Bre5 (, bottom panel) and led to a more dramatic decrease in sec23 deubiquitylation (, bottom panel). Significantly, these in vivo
results mirror the in vitro
binding studies showing that both the Bre5 Y17A/K82A but mainly the Ubp3 L208A/F209A/I210A/N211A mutants decrease Bre5-Upb3 association ( and ). Taken together, these results demonstrate that association between Bre5 and Ubp3 through the interface defined in the crystal structure reported here is important for Ubp3 function in vivo
Comparison of the Bre5/Ubp3 interface to other protein complexes with NTF2 domains
It is somewhat striking that the NTF2-like domains adopt highly homologous dimeric structures given the rather low but significant sequence homology (). One potential reason for the low sequence homology may be due to the fact that different NTF2-like domains exploit this sequence divergence to interact with different cognate proteins. A comparison of the Ubp3/Bre5 complex reported here with other protein complexes with NTF2 and other NTF2-like domains reveals that this is indeed the case. Specifically, while Ran GTPase interacts with its cognate NTF2 dimer on the same “top” surface as Ubp3 of the NTF2 dimer, Ran GTPase uses two loop regions to interact with loop regions on the opposite face of the top surface of the NTF2 dimer () 18
. Interestingly, the TAP subunit of the TAP/P15 heterodimer contains an extra C-terminal helix that sits at the dimer interface of the top surface of the NTF2-like dimer contributing to specificity of the NTF2 domain heterodimer 15; 16; 19
. In contrast, the nucleoporin FG repeat loop interacts with loops in the “bottom” surface of the TAP/P15 dimer () 13; 15
. Taken together, it appears that the NTF2-like domains employ sequence divergence, predominantly within loop regions, on different surfaces to specifically associate with their cognate proteins. The Ubp3/Bre5 complex represents yet another example of how different NTF2-like domains form protein-specific recognition modules to modulate distinct biological processes.
Comparison of the Bre5/Ubp3 complex with other protein complexes with NTF2-like domains