Structure of Vps75 and implications for histone chaperone function

Abstract

The vacuolar protein sorting 75 (Vps75) histone chaperone participates in chromatin assembly and disassembly at both active and inactive genes through the preferential binding to histone H3–H4. Vps75 is also one of two histone chaperones, along with antisilencing factor 1, that promotes histone H3-Lys-56 acetylation by the regulation of Ty1 transposition protein 109 (Rtt109) histone acetyltransferase. Here, we report the x-ray crystal structure of Vps75 and carry out biochemical studies to characterize its interaction with Rtt109. We find that the Vps75 structure forms a homodimeric “headphone” architecture that includes an extended helical dimerization domain and earmuff domains at opposite ends and sides of the dimerization domain. Despite the similar overall architecture with the yeast nucleosome assembly protein 1 and human SET/TAF-1β/INHAT histone chaperones, Vps75 shows several unique features including the relative disposition of the earmuff domains to the dimerization domain, characteristics of the earmuff domains, and a pronounced cleft at the center of the Vps75 dimer. These differences appear to correlate with the unique function of Vps75 to interact with Rtt109 for histone acetylation. Our biochemical studies reveal that two surfaces on the earmuff domain of Vps75 participate in Rtt109 interaction with a stoichiometry of 2:1, thus leaving the pronounced central cleft of the Vps75 dimer largely accessible for histone binding. Taken together, our data provide a structural framework for understanding how Vps75 mediates both nucleosome assembly and histone acetylation by Rtt109.

The genetic information of eukaryotic organisms is packaged into a compacted chromatin structure containing nucleosome core particles with DNA wrapped around two copies each of the four histone proteins H2A, H2B, H3, and H4 (1). DNA-mediated activities, including transcription, replication, recombination, and DNA repair use the concerted efforts of chromatin regulatory proteins that can be grouped into four broad categories: ATP-dependant chromatin remodeling enzymes that physically move the histone proteins along the DNA (2); posttranslational modification enzymes that covalently modify histones, predominantly on the N-terminal histone tail regions (3); chromatin targeting proteins or modules that recruit proteins to the DNA, histones, or modified histones (4); and histone chaperone proteins that assemble and disassemble histones and replace variant histones in chromatin (5). The latter class of proteins come in different structural forms including nucleoplasmin, chromatin assembly factor 1 (CAF-1), histone regulator (Hir), antisilencing factor 1 (Asf1), and nucleosome assembly protein (Nap) (6, 7).

The NAP family of histone chaperones are conserved from yeast to human and have been implicated in many biological functions including cell proliferation, cell-cycle regulation, transcription, replication, silencing, and apoptosis (7). The defining function of the NAP proteins is to bind histones. These proteins contain a conserved central domain (NAP domain), a nonconserved N-terminal region, and a highly acidic C-terminal tail. The NAP domain is sufficient for histone binding, although the acidic tail may enhance histone binding (8). The yeast Nap1 protein has been shown to preferentially bind the H2A/H2B heterodimer in vivo (7), although it can also bind a (H3–H4)₂ heterotetramer in vitro (9). Structures of the yeast Nap1 (10) and human SET/TAF-1β/INHAT (11) homologues reveal a symmetrical elongated homodimer formed by a long central helical dimerization domain with two α/β globular domains at each end.

A more recently identified member of the NAP protein family is vacuolar protein sorting 75 (Vps75), previously identified in a genomic screen for vacuolar protein sorting genes and for factors that affect telomere length (12, 13). Vps75 harbors several unique properties relative to other NAP proteins. First, Vps75 is one of two histone chaperones, along with Asf1, to form a complex with the regulation of Ty1 transposition protein 109 (Rtt109) (14, 15) histone acetyltransferase to stimulate its acetylation of histone H3 at Lys-56, a modification implicated in the maintenance of genome stability (16–18). Second, in contrast to the other NAP members, Vps75 preferentially binds histones H3–H4 over H2A–H2B (19).

To obtain molecular insights into the Vps75 structure and its mode of interaction with Rtt109, we determined a high resolution x-ray crystal structure of Vps75. The structure, together with biochemical studies, provides a framework for understanding how Vps75 mediates both nucleosome assembly and histone acetylation by Rtt109.

Results

Overall Structure of the Vps75 Protein.

To facilitate Vps75 crystallization for x-ray structure determination, we prepared full-length Vps75 and several deletion constructs by removing various length segments of the poorly conserved C-terminal acidic domain (Fig. 1). Although many of the Vps75 protein constructs produced crystals, only the Vps75(1–232) protein construct produced crystals that were suitable for high-resolution x-ray structure determination. These crystals diffracted x-rays to ≈2.0 Å resolution, formed in spacegroup P2₁2₁2₁, and contained a complete Vps75 homodimer per asymmetric unit. The structure of a Vps75(1–232)I80M/I160M double mutant (designed to facilitate phase determination) was determined by using selenomethionine multiwavelength anomalous dispersion (MAD) phasing. The structure was refined to an R_work/R_free of 0.209/0.240 at 2.0 Å resolution with excellent geometrical parameters (Table 1).

Fig. 1.

Sequence alignment and domain structure of Vps75 homologues with yeast Nap1 and human SET/TAF-1β/INHAT. (A) Secondary structure (numbered) of Vps75 is shown above the sequence alignment of the Vps75 homologues. A sequence alignment and corresponding secondary structure of Nap1 and SET/TAF-1β/INHAT are shown below the alignment of the Vps75 homologues. Residues involved in the dimerization-earmuff domain interface, implicated for histone H3–H4 tetramer binding, and critical for Rtt109 binding are indicated with closed ovals, triangles, and stars, respectively, above the sequence alignment. Abbreviations are as follows: Saccharomyces cerevisiae (Sc), Candida glabrata (Cg), Kluyveromyces lactis (Kl), Vanderwaltozyma polyspora (Vp), Debaryomyces hansenii (Dh), Candida albicans (Ca), Pichia guilliermondii (Pg), and Homo sapiens (Hs). (B) Domain structures of Vps75, Nap1, and SET/TAF-1β/INHAT, respectively.

Table 1.

Data collection, phasing, and refinement statistics

Vps75 (1–232)I80M/I160M
Data collection
Space group	P212121
Cell dimensions, Å	a = 79.01, b = 89.66, c = 94.07
	Peak	Inflection	Remote
Wavelength, Å	0.9789	0.9791	0.9537
Resolution, Å	50–2.0 (2.07–2.0)	50–2.2 (2.28–2.20)	50–2.4 (2.49–2.40)
No. reflections	45,182	34,344	26,643
Rsym*	0.096 (0.414)	0.093 (0.409)	0.087 (0.402)
I/σI	24.4 (5.48)	27.1 (6.8)	30.2 (6.7)
Completeness (%)	99.1 (95.9)	99.6 (99.3)	99.7 (99.3)
Redundancy	13.8 (11.0)	14.3 (13.3)	13.8 (13.1)
Refinement
Resolution, Å	50–2.0 (2.07–2.0)
Rwork/Rfreec	0.209/ 0.240
No. atoms
Protein (selenium)	3,440 (10)
Water	362
B factors
Protein	34.5
Water	43.0
R.M.S.D.
Bond lengths, Å	0.0054
Bond angles, °	1.07

Values in parentheses are for the highest-resolution shell.

*R_sym = 100 × Σ|I − 〈I〉|/Σ〈I〉, where I is the observed intensity and 〈I〉 is the average intensity from multiple observations of symmetry-related reflections. ^cR_free was calculated with 5% of the data.

The Vps75 protein forms a nearly symmetrical elongated homodimer with the same overall 2-domain “headphone” topology of the related yeast Nap1 and human SET/TAF-1β/INHAT proteins (Figs. 1 and 2A). In brief, domain I contains a long 40-residue extended helix (α1) “head” segment of the headphone that interacts in an antiparallel fashion with the helix of the opposing subunit via H-bond and van der Waals interactions to form a dimerization domain. Domain II contains the α/β globular “earmuff” segments of the headphone (Fig. 2B). These globular domains are on opposite ends and sides of the domain I dimerization element and contain a 4-stranded β-sheet (β1–β4) on the same plane as the domain I helices, with 6 helices (α2, α3, α4, α6, α7, and the N-terminal end of α8) arranged below the β-sheet that together surround a hydrophobic core region. Two helices project away from the core and at opposite ends (the C terminus of α8 and α5), and two 3₁₀ helices (η1 and η2) form between the β-hairpin region between β1 and β2 and between β4 and α5, respectively. Eleven N-terminal residues, 7 loop residues between β3 and β4 (130–136), and 9 residues of the C terminus (223–232) are disordered and not modeled in the structure. There is a pronounced cleft formed near the center of the homodimer, which is flanked on top by the domain I dimerization domain and on opposite sides by the domain II earmuffs of the homodimers (Fig. 2 A and C). The dimensions of the cleft are ≈55 Å high with a width of ≈18 Å at the narrowest point near the dimerization domain and ≈70 Å wide between the N-terminal tips of the α5 helices.

Fig. 2.

Overall structure of Vps75 and its relationship to Nap1 and SET/TAF-1β/INHAT. (A–C) Diagram representation of the Vps75 homodimer, with C rotated 90° along the x axis relative to A and secondary structural elements of the earmuff domain indicated in B. The view in (B) and (C) are similar. (D and E) Overall superposition of Vps75 with yeast Nap1 and human SET/TAF-1β/INHAT. Vps75 is colored in red, with Nap1 and SET/TAF-1β/INHAT both colored in cyan. (F–H) Surface electrostatic potential of Vps75, Nap1, and SET/TAF-1β/INHAT, respectively, with blue, red, and white representing electropositive, electronegative, and neutral areas, respectively. Views are the same as in D and E.

Comparison with Yeast Nap1 and Human SET/TAF-1β/INHAT.

In addition to a similar overall structural topology with the NAP homologues, yeast Nap1 and human SET/TAF-1β/INHAT, Vps75 shares other structural features with these proteins. Specifically, the domain I dimerization helices of the three proteins are of comparable length with ≈45 Å between the dimerization proximal surface of the earmuff domains, and they superimpose well with an rmsd of <1.78 Å between the Cα atoms of residues 12–53 of the Vps75 dimerization helices and the corresponding residues of SET/TAF-1β/INHAT (84 atoms) and Nap1 (72 atoms) (Fig. 2 C–E). The 4 β-strands and 6 α-helices, which form the hydrophobic core of Vps75, also show structural superposition with the Nap1 and SET/TAF-1β/INHAT proteins with an rmsd of <1.9 Å between the Cα atoms of residues 56–223 of a Vps75 monomer with the corresponding atoms of Nap1 (92 atoms) and SET/TAF-1β/INHAT (98 atoms) (Fig. 2 D and E).

Despite the similarities between the NAP proteins noted above, there are interesting differences that are likely to have functional significance. First, the relative positions of the earmuff domains to the dimerization domain differ among the three proteins, as evidenced by significant offsets of the earmuff domains when the dimerization domains are superimposed (Fig. 2 D and E). In both Nap1 and SET/TAF-1β/INHAT, the earmuff domains are rotated closer toward the helical dimerization domain and the center of the dimer than in Vps75. As a consequence, the pronounced cleft within the Vps75 dimer is more restricted in the Nap1 and SET/TAF-1β/INHAT proteins. This is particularly true for Nap1, which also has additional helices (α1, α3, and α10) and an extended loop (β3–β4) reaching into its central cleft. The respective surface electrostatic potential of the cleft region of the three proteins also show considerable divergence. Whereas Nap1 and SET/TAF-1β/INHAT show a highly electronegative surface in this region, Vps75 shows a more defined electronegative ring of charge only directly under the dimerization helices and electropositive patches at the bottom of the earmuff domains (Fig. 2 F–H). As will be discussed below, these differences in cleft size and surface electrostatic potential may reflect the different histone substrate specificities of these histone chaperones.

Within the earmuff domains, the α5 helix represents a unique secondary structural element of Vps75 (Fig. 2B). Interestingly, together with the α6-loop region, this represents the most conserved region within the Vps75 proteins (Fig. 1A), implicating its functional importance. In place of the α5 helix in Vps75, the Nap1 structure contains a β-hairpin region that overlaps with a nuclear localization sequence that has also been proposed to regulate oligomerization (20). Whereas the corresponding region in the SET/TAF-1β/INHAT structure is largely disordered in the crystal's lattice, mutations in this region have been found to impair its histone chaperone activity (11). Taken together, it seems that this region of the NAP proteins contains elements of functional importance.

Surface Conservation of the Vps75 Homodimer and Implications for Protein Binding.

A mapping of sequence conservation among the Vps75 homologues onto the surface of the Vps75 structure reveals that the bottom surface of the earmuff harbors the greatest degree of conservation. This conservation largely maps to several basic residues from the Vps75-specific α5 helix and the loop following the α6 helix, together with a number of acidic residues in the α2, α3, α7, and α8 helices within the cleft region of the dimer (Fig. 3 A–C). This observation suggests that Vps75 may use the cleft region and base of the earmuff domains to interact with one or both of its two associated protein partners, the Rtt109 histone acetyltransferase (16, 17) and/or the histone (H3–H4)₂ tetramer (19).

Fig. 3.

Conservation surface of the Vps75 homodimer. (A and B) Two orthogonal views of the conservation surface of the Vps75 homodimer. The views are as in Fig. 2 A and C, respectively. The conservation color scheme is indicated. (C) Highly conserved residues facing toward the central cleft are highlighted as sticks in CPK coloring on a diagram of the protein with a transparent molecular surface. The color coding is as described in Fig. 2B. (D) Highly conserved residues that participate in the Vps75 dimerization-earmuff domain interface are highlighted as sticks by using conservation color coding as in A. Refer to Fig. S1 for this conversed feature in Nap1 and SET/TAF-1β/INHAT.

There is also a significant degree of sequence conservation among the Vps75 homologues that maps to the interface between the dimerization and earmuff domains, suggesting that the relative orientation of these domains are important for Vps75 function (Fig. 3D). One particular interaction consists of a highly conserved hydrophobic network involving Phe-15, Leu-18, Ile-51, and the aliphatic regions of Glu-22, Tyr-44, and Arg-47 from the dimerization domain. These regions interact with several conserved residues from the top surface of the earmuff domain, evidently defining both the length of the two-helix bundle dimerization domain between the earmuff domains and the relative orientation of the earmuff domains to the dimerization domain. This in turn defines the shape and width of the central cleft of the dimer. This interface is also highly conserved in yeast Nap1 and human SET/TAF-1β/INHAT [supporting information (SI) Fig. S1 A and B]. However, predominantly because of a 7-residue longer dimerization domain between these two interfaces, the relative positions of the two earmuff domains are ≈1 helical turn further apart in Nap1 and SET/TAF-1β/INHAT relative to Vps75. Interestingly, despite this shorter dimerization domain gap between the earmuff domains of Vps75, the center-of-mass distance between the two earmuff domains of Vps75 is actually greater than the corresponding distances in Nap1 and SET/TAF-1β/INHAT, resulting in the larger central cleft that is observed in Vps75 (Fig. 3A). These clear differences are likely to be correlated with the different histone binding properties of these histone chaperone proteins.

Vps75 Determinants for Interaction with Rtt109.

To correlate the Vps75 structure to its interaction with the Rtt109 histone acetyltransferase, we carried out a series of biochemical experiments. We first coexpressed the full-length Vps75 and Rtt109 proteins in bacteria and found that they were both produced in the soluble fraction and copurified through MonoQ ion exchange and Superdex S200 gel filtration chromatography. The Vps75/Rtt109 complex eluted from gel filtration around a protein standard of 158 kDa, suggestive of a 2:2 Vps75/Rtt109 complex (162 kDa). To confirm that Vps75 indeed forms a complex with Rtt109 in solution, Vps75 alone or mixed stoichiometrically with Rtt109 was subjected to chemical cross-linking with dithiobis(succinimidylpropionate) (DSP) and resolved with SDS/PAGE. Vps75 cross-linking yielded two species that presumably correspond to a Vps75 dimer and a dimer of dimers, respectively (Fig. S2 Right), with the dimer (61.5 kDa) migrating with an apparent molecular mass around a protein standard of 74 kDa. The Vps75/Rtt109 mixture yielded two cross-linked species that are indicative of a Vps75/Rtt109 complex and dimers of the complex, respectively (Fig. S2 Left). The lower Vps75/Rtt109 cross-linked species migrates between protein standards of 173 and 117 kDa, consistent with a complex with a stoichiometry of either 2:2 (162 kDa) or 2:1 (112 kDa).

To more quantitatively establish the stoichiometry of the complex, we used sedimentation equilibrium analysis. These studies were carried out by using three different protein complex concentrations (0.8, 0.4, and 0.2 mg/ml) and three different centrifugation speeds (18,144, 26,127, and 39,030 × g). The data were best fit to a single-species model containing two molecules of Vps75 and one molecule of Rtt109 (Figs. 4A and S3A), whereas a single-species 2:2 complex fit considerably more poorly to the data (data not shown). We were also able to obtain a good fit to the data with small residuals to a Vps75/Rtt109 equilibrium model containing a Vps75 dimer and two molecules of Rtt109 with dissociation constants of K_d1 = 0.01 and K_d2 = 2.7 μM, respectively (Fig. S3B). Taking these sedimentation equilibrium data together suggest that the Vps75 dimer associates primarily with one molecule of Rtt109 with high affinity and with a second Rtt109 molecule at a lower affinity (by ≈2 orders of magnitude).

Fig. 4.

Vps75 interaction with Rtt109. (A) A representative sedimentation equilibrium of the Vps75/Rtt109 complex (OD₂₈₀ = 0.5, 26,127 × g) fitting to a (Vps75)₂/(Rtt109)₁ single-species model. The global fit with three centrifugation speeds and three protein concentrations were shown in Fig. S3A. (B and C) GST-Rtt109 pull-down studies are shown with wild-type and deletion and substitution mutants of Vps75. (D) Mapping of designed Vps75 mutants. A diagram presentation of the Vps75 homodimer (in cyan) highlighting the selected residues for mutation (stick with CPK coloring) examined in the GST-Rtt109 pull-down assay. Mutations that show decreased and unaffected Rtt109-binding properties by pull-down experiment are colored in red and green, respectively. The solid and dashed ovals depict two alternative Rtt109-binding modes on the Vps75 homodimer, with histone H3–H4 (as white and gray triangles, respectively) positioned in the Vps75 dimer central cleft to depict a hypothesized Vps75/H3–H4-binding mode. The view is the same as in Fig. 2C.

To establish what region(s) of Vps75 interacts with Rtt109, we carried out Vps75 pull-down studies with GST-Rtt109 (Fig. 4 B and C). We first addressed whether the Vps75 core (residues 12–222), as observed in our structure, was sufficient for Rtt109 binding. As shown in Fig. 4B, a Vps75 construct containing a deletion of 8 N-terminal and 42 C-terminal residues Vps75 (9–222) still shows interaction with Rtt109. This result confirms that essential Rtt109-binding elements are located within the Vps75 core fragment as seen in the Vps75 structure reported here.

Within the Vps75 core fragment, the α5 helix of the earmuff domain, not present in Nap1 or SET/TAF-1β/INHAT, contains a basic patch of residues that are conserved exclusively among the Vps75 homologues; therefore, we hypothesized that this region might be involved in Rtt109 interaction. To test this hypothesis, we prepared a Δ(167–178) construct and the K169E/K170E and K173E/K177E Vps75 mutants (Fig. 4D) for pull-down studies with GST-Rtt109. As can be seen in Fig. 4C, whereas the K169E/K170E mutant binds Rtt109 at wild-type levels, the Δ (167–178) and K173E/K176E Vps75 mutants show significantly reduced Rtt109 binding. Based on these observations, we propose that the α5 helix of the Vps75 earmuff domains, a unique feature that is not shared with other NAP proteins, is a critical determinant for Rtt109 association.

Interestingly, in the pull-down experiments described above, the K173E/K176E charge reversal mutant showed poorer binding than the Δ (167–178) deletion mutant. This observation suggests that removing the α5-mediated Vps75-Rtt109 interaction is not as detrimental as introducing destabilizing reverse charges. The fact that both mutations did not abolish Vps75-Rtt109 interaction further suggests that another region of Vps75 might also be used for Rtt109 binding. To address this possibility, we prepared three additional charge-reversal double mutants that target other conserved regions of the earmuff domain of Vps75 (Fig. 4D) and carried out pull-down studies with GST-Rtt109. As can be seen in Fig. 4C, the D198K/S199Y and E206K/E207K mutations, at opposite ends of the Vps75 α7 helix and deeper within the central cleft, show no detectable effect on GST-Rtt109 pull-down. In contrast, the E218K/D222K mutant, located at the C-terminal end of the α8 helix, shows no detectable interaction with GST-Rtt109. This result demonstrates that the C-terminal end of the α8 helix of Vps75 also contributes to Rtt109 interaction. Taken together, we have identified two regions of the earmuff domain, the α5 and α8 helices of Vps75 that participate in Rtt109 interaction.

Discussion

Analysis of the Vps75 mutational data described here in the context of the Vps75 dimer structure reveals that the distances between the two Rtt109-binding surfaces of Vps75 within the same earmuff domain and between two earmuff domains across the central cleft of the Vps75 dimer are comparable (Fig. 4D). Therefore, one Rtt109 protein may bind to either one or two earmuff domains (as schematized in Fig. 4D), with both models consistent with the high-affinity Vps75/Rtt109 stoichiometry of 2:1 that we observe in the equilibrium sedimentation studies. Interestingly, our equilibrium sedimentation analysis also indicates that a 2:2 Vps75/Rtt109 complex can form at higher concentrations of Rtt109, and it is possible that the formation of a Vps75/H3–H4 complex might promote the formation of such a complex, possibly through structural changes.

Either of the Vps75/Rtt109 binding modes described above leaves the pronounced Vps75 central cleft (Fig. 2 A and C) largely accessible for another binding partner. Given the fact that this central cleft contains a high degree of conservation among Vps75 homologues (Fig. 3 A and B) and Vps75 is known to bind H3–H4 complexes, we propose that this central cleft might be used for histone binding. Several observations support this hypothesis: (i) the size of a (H3–H4)₂ tetramer approximately corresponds to the size of the Vps75 dimer central cleft, and (ii) the negatively charged nature of the Vps75 dimer central cleft (Fig. 2F) would largely complement the electropositive (H3–H4)₂ tetramer as seen in the nucleosome structure (1). Indeed, a preliminary docking exercise reveals that the Vps75 central cleft could accommodate either a symmetrical (H3–H4)₂ tetramer or a H3–H4 dimer, without steric clash (results not shown). Participation of the Vps75 central cleft in histone binding is consistent with the observation that mutations of two highly conserved patches of residues within the central cleft, D198K/S199Y and E206K/E207K (Fig. 4D), have no detectable effect on Rtt109 binding (Fig. 4 C and D). Further experimentation is clearly required to establish the mechanism of H3–H4 complex binding by Vps75 and how this binding is coordinated with Rtt109 acetylation of H3K56.

Asf1, another histone chaperone with a different fold than the NAP proteins, has also been shown to promote Rtt109-mediated acetylation of histone H3K56 (17, 18, 21). In contrast to Vps75 that has been reported to bind histone (H3–H4)₂ tetramers (22), Asf1 can only bind histone H3–H4 dimers (23). Recent in vitro studies demonstrate that the Rtt109/Vps75 complex acetylates H3K56 in the Asf1/(H3–H4) complex and that these complexes form a direct interaction (24). Therefore, we propose that, for histone H3K56 acetylation, Asf1 either hands off histone H3–H4 dimers to the Vps75/Rtt109 complex or that Vps75 hands off Rtt109 to the Asf1/(H3–H4) complex. Both possibilities are consistent with the fact that Rtt109 complexes with both Vps75 and Asf1 have not been observed, and the latter is consistent with the observation that Asf1 can promote Rtt109 acetylation in the absence of Vps75 (21). Given these observations it is possible that Vps75 and Asf1 may recognize the same binding surface on Rtt109, similar to how the CAF1 and HIRA histone chaperones recognize the same binding surface of ASF1a (25).

Despite the observation that Vps75 can promote histone H3K56 acetylation by Rtt109 in vitro, in vivo experiments reveal that unlike Rtt109 deletion cells, VPS75 deletion cells still retain histone H3K56 acetylation and are insensitive to perturbations in replication (19), suggesting that the Vps75/Rtt109 complex may function in other pathways via acetylating other histone, or possibly nonhistone, substrates. Clearly, additional studies will be required to address the mode of histone chaperone and chromatin regulatory activities of Vps75.

Methods

Protein Preparation.

DNA encoding full-length Vps75 and Rtt109 proteins and truncation mutants of Vps75 were subcloned into the plasmid pRSF-GST to produce N-terminal GST-fusion proteins for expression in BL21(DE3) Gold cells. Protein expression was induced at an OD₆₀₀ of 0.8 with 0.8 mM isopropyl β-d-thiogalactoside and overnight cell growth at 20°C, lysed by sonication, and purified to homogeneity by using a combination of GST affinity, GST cleavage by TEV, MonoQ anion exchange, and Superdex S200 gel filtration chromatography. The proteins were concentrated to ≈30 mg/ml and flash frozen until further use. Selenomethionine-derivatized Vps75 protein was prepared essentially as described in ref. 26 and purified as described for the wild-type proteins. Further details on the protein preparation are supplied in SI Methods.

The Rtt109/Vps75 complex was prepared by cotransforming the pRSF-GST-Vps75 and pCDF-6His-Rtt109 plasmids into BL21(DE3) Gold cells. Protein coexpression was induced at an OD₆₀₀ of 0.8 with 0.8 mM isopropyl β-d-thiogalactoside followed by growth at 37°C for 3 h. The complex was captured with nickel resin followed by purification as described for GST-Vps75 above. The protein complex was concentrated to ≈10 mg/ml in PBS-βME buffer and flash frozen until further use.

Crystallization and Structure Determination of Vps75(1–232)I80M/I160M.

Crystals of Vps75 (1–232) I80M/I160M were obtained by using the hanging drop method from a reservoir solution containing 25% PEG4000, 100 mM Mes (pH 6.3) buffer, and 0.2 M malic acid; and cryoprotected for x-ray data collection by soaking them for ≈5 min each in reservoir solutions supplemented with 0, 10, and 20% glycerol, respectively, and flash frozen in liquid nitrogen. Three-wavelength anomalous (multiwavelength anomalous dispersion) (27) diffraction datasets were collected, and the data were processed and scaled with the HKL2000 suite (28). Ten out of 14 selenium atoms were located and used to calculate phases to 2.4 Å using the program SOLVE/RESOLVE (29). The protein model was built with COOT (30) by using the REFMAC library (31) and refined with CNS (32) with iterative model building, simulated annealing, and positional and B-factor refinement strategies. The complete model, checked for errors with composite simulated annealing omit maps, was further refined to a final resolution of 2.0 Å with the addition of 362 water molecules. The final model was evaluated by PROCHECK (33), revealing excellent stereochemical parameters and Ramachandran plot statistics (93.2% in most favored region, and 6.8% in additionally allowed regions) for all residues.

In Vitro Binding Assay.

GST-Rtt109-fusion protein was retained on glutathione Sepharose beads and aliquoted into 1.5-ml Eppendorf tubes for equilibration with 3×-washes with PBS-βME buffer. Samples were quantitated by SDS/PAGE analysis before addition of stoichiometric amounts of purified full-length or truncation/substitution mutants of Vps75. Binding reactions were incubated at 4°C for 1 h with gentle rotation before washing three times with PBS-βME buffer. Samples containing protein that remained bound to the glutathione Sepharose beads after the last wash were resolved by SDS/PAGE and visualized by Coomassie blue staining. As a negative control, Vps75 proteins were also incubated with GST alone, showing no detectable binding.

Equilibrium Sedimentation.

Analytical ultracentrifugation of the Vps75/Rtt109 complex was performed at 4°C with absorbance optics by using a Beckman Optima XL-I analytical ultracentrifuge, using a 4-hole rotor. The partial specific volume and viscosity were estimated by using Sedenterp (34). Analysis was performed by using six-channel centerpieces with quartz windows, spinning at 18,144, 26,127, and 39,030 × g. Protein samples were analyzed at three different protein complex concentrations of 0.8, 0.4, and 0.2 mg/ml in PBS-βME buffer. A global fit of the data fit best to a single species A/B complex or the equilibrium model A+B ↔ AB+B ↔ AB₂ with A as an Vps75 dimer and B as an Rtt109 monomer by using the program HeteroAnalysis. The quality of the fit was assessed from the examination of rmsd.