The structure of nucleosome assembly protein 1

Results

yNAP-1 Exhibits a Previously Uncharacterized Fold. The crystal structure of NAP-1 from S. cerevisiae (yNAP-1) was determined at 3.0 Å resolution (Table 1, which is published as supporting information on the PNAS web site). Of a total of 417 aa, well defined electron density was observed for the central region encompassing residues 70–370 (Fig. 1 A and B), whereas the N- and C-terminal regions were largely disordered. It has been shown earlier that residues 74–365 are sufficient for histone binding and nucleosome assembly. Additionally, limited proteolysis has demonstrated that this is the most stable structural entity (7, 8). An N-terminally truncated version of the protein (encompassing amino acids 74–417) crystallized in the same space group, indicating that the N-terminal region is not involved in crystal contacts. We have also determined the structure of a NAP-1 construct consisting only of the core region (74–365) and, with one minor exception, found no differences when compared with the structure obtained with full-length protein (see below). These observations suggest that the N- and C-terminal segments are normally disordered and do not contribute to the overall architecture of NAP-1, in agreement with in vitro solution studies (7).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Overall structure of yeast NAP-1. (A) Amino acid sequences alignment of members of the NAP protein family. yNAP1, xNAP1, dNAP1, and hNAP1 are shown. y, S. cerevisiae; x, Xenopus laevis; d, Drosophila melanogaster; h, Homo sapiens nucleosome assembly protein 1. Red or blue color indicates identical or conserved amino acid residues from all of the NAP family proteins. PKE*P**K*EE (* is any amino acid) repeat sequences of NAP-1L3 are not shown in this alignment as indicated with parentheses. The alignment was generated with clustalw. Subdomains A, B, C, and D are boxed in blue, yellow, green, and red, respectively. Secondary structure elements are indicated. Intervals of 10 aa for S. cerevisiae (black circles) are indicated. The two highly conserved hydrophobic sequence motifs ₁₈₅IPSFWLT₁₉₁ and ₃₁₁SFFNFF₃₁₆ are indicated with dotted lines. (B) The monomer structure of yeast NAP-1, shown as a ribbon diagram. Subdomains A (domain I), B (domain I), C (domain II), and D (domain II) are shown in blue, yellow, green, and red, respectively, as in A. (C) Topology diagram of the yeast NAP-1 monomer (helices are shown as circles, strands are shown as triangles, and loops are shown as pipes). Domains I and II are indicated.

Fig. 1B shows an overview of the yNAP-1 monomer; the corresponding connectivity diagram is shown in Fig. 1C. The monomer can be divided into two domains: domain I (the dimerization domain) consists of one long α-helix that spans 47 aa, flanked by two shorter α-helices. The main feature of domain II (the protein interaction domain) is a four-stranded antiparallel β-sheet that is protected by α-helices at its underside. A small antiparallel β-hairpin that functions as a nuclear localization sequence (NLS) extends from it (34, 35). Each of these domains can be divided into two subdomains. Subdomain A (shown in blue) is mainly responsible for dimerization, whereas subdomain B (yellow) is implicated in regulating access to a NES embedded in subdomain A. Subdomain C (green) forms an amphipathic β-sheet that is protected at its underside by subdomain D (red). In addition to the disordered C- and N-terminal regions (see above), residues 172–180 in the α3–α4 loop and 293–301 in the β5–β6 loop were disordered in our electron density maps and are not included in the model.

In the crystal structure of full-length NAP-1, one NAP-1 monomer represents the asymmetric unit (ASU), and the crystallographic twofold symmetry generates a homodimer with an extensive interaction interface (Fig. 2). The dimer has an overall ellipsoidal shape and approximate dimensions of 75 × 57 × 32 Å. It consists of three layers: the first layer is defined by the dimerization helix α2 (domain I, shown in blue in Fig. 2), the second is formed by the β-sheets of domain II (green), and the bottom layer is formed by α-helices (shown in red in Fig. 2) that span its entire width. Truncated yNAP-1 (residues 74–365) crystallized in P4₂2₁2 and contains one NAP-1 dimer per ASU (data not shown). Except for β5, β6, and two of the flexible loops (which are involved in different crystal contacts in the two lattices), the structures of the NAP-1 molecules in the noncrystallographic and crystallographic dimer are identical and superimpose with an rms deviation of 1.05 Å for 268 residues.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

The dimer structure of the yeast NAP-1. The two monomers are related by a crystallographic twofold axis. A, top; B, bottom; C and D, two side views. (A and B) A view is obtained by a 180° rotation around the vertical axis (as indicated). (C and D) The side views of the NAP-1 are obtained by rotation of 90° around the axis of crystallographic symmetry. The unstructured region of α3–α4 loop or acidic C-terminal domain that is not a part of this model is indicated with a yellow or red dotted line.

NAP-1 Dimerizes via a Non-Coiled-Coil Motif. Domain I is mainly responsible for the dimerization of NAP-1 (Fig. 2) by close antiparallel pairing of the long α2-helices along their entire length of 47 aa (≈69 Å). The arrangement is unusual in that the two helical axes are aligned in an antiparallel “tram-track” fashion, rather than coiling around each other as in the widespread coiled-coil motif. The α2-helices are curved in both planes, with the helical axes following a sigmoidal path in one plane and forming a dome-shaped architecture in the other (Fig. 3A). More precisely, a ≈30° kink is introduced into α2 by the presence of a proline at position 130; a second, ≈25° kink is observed at Val-109 (Fig. 3A). As a consequence of the antiparallel pairing of the two helices, Pro-130 juxtaposes Val-109′ of the second NAP-1 molecule within the dimer. It is likely that the bend at Val-109 is caused by the need to maintain packing by keeping the two helical axes parallel. Fig. 1A demonstrates the complete amino acid conservation at these two positions within the entire NAP-1 family of proteins. The dimer is further stabilized by the α2–α3 loop, the α3-helix, and the α3–α4 loop that wrap around the base of the α2-helix of the dimerization partner (Fig. 2 A and B). In addition, α1 has a key role in orienting the C-terminal acidic domain of the dimerization partner by forming extensive hydrophobic interactions with α8′ (Fig. 2A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

The dimerization domain. (A) Top and side views of subdomain A show detailed interactions in the dimerization interface of yNAP-1. Val-109 and Pro-130 are indicated. The arrow denotes the location of Pro-130. (B) Stereo diagram of the final 2Fo-Fc electron density map (contoured at 2 sigma) of residues 112–119 of subdomain A and 109–126 of subdomain A′. Dotted lines indicates hydrogen bonds.

The dimer interface is characterized by exclusively hydrophobic interactions over the entire length of the involved α-helices (see, for example, Fig. 3B). It buries ≈22% of the overall surface of each subunit and involves 20 residues within α2 and four in α1. The hydrophobic nature of these residues is overall maintained between humans, frog, flies, and yeast (Fig. 1A), suggesting that dimer formation is strictly conserved. Indeed, the majority of conserved residues are clustered along the dimerization interface of α2 (Fig. 6A, which is published as supporting information on the PNAS web site). The extensive dimerization interface that spans the entire dimer diagonally (Fig. 6B) is consistent with a previous investigation showing that the NAP-1 dimer cannot be disrupted without denaturation and that this disruption is coincident with a global disruption in 2° structure (9). Domain II projects away from each side of the dimer and is therefore not involved in dimer stabilization, with the exception of α8.

The Nuclear Export Sequence (NES) Is Masked by the Accessory Domain. Residues 88–102 of yNAP-1 have been identified as a NES, and this activity has been confirmed for yeast NAP-1 in vivo (23, 34). Replacement of the hydrophobic residues Leu-96 and Leu-102 that are characteristic of NES sequences to alanine caused a radical redistribution of the protein from the cytoplasm to the nucleus (23). The NES of yNAP-1 initiates in the loop connecting α1 with α2 and includes the N-terminal 11 aa of α2, which are also involved in forming the dimerization interface. The α-helical character and the distribution of hydrophobic and hydrophilic residues in the present structure have also been found in other published structures containing a functional NES (38).

In yNAP-1, the NES is for the most part occluded by subdomain B (residues 141–180; Figs. 2A and 4A). We have termed this domain the accessory domain because of variability in length between different NAP forms (Fig. 1A). In yeast, it consists of an α-helix (α3) and two loops and effectively links domain I to domain II. By wrapping an extended α2–α3 loop around the base of α2′ of the dimerization partner, and by packing α3 next to α2′, it not only contributes to dimer formation but also effectively covers the upper part of the NES (Fig. 4). A conserved proline (Pro-145) in the α2–α3 loop restricts its conformation and positions it over α2. The loop connecting α3 and α4 (amino acids 161–180 in yNAP-1) is highly variable among proteins in the NAP family (Fig. 1A). In the present structure, it is rather flexible and characterized by poor electron density. Residues 172–180 were too disordered to be included in the model, although weak electron density was observed that covers the N-terminal segment of the NES (Fig. 4B).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

NES masking by the accessory domain. The accessory domain and NES region are shown in yellow and blue, respectively. The color code for the other subdomains is as in Fig. 1. (A) Molecular surface of the molecule, shown in the same view as in Fig. 2 A. (B) Detailed view of the masking of the NES by subdomain B, shown in two orientations. The region between amino acids 169 and 181 is missing from the model; these amino acids are circled in red.

Domain II Is Stabilized by an Extensive Network of Aromatic Residues. Domain II spans residues 181–370 of yNAP-1 and consists of two layers (subdomains C and D, shown in green and red, respectively, in Figs. 1 and 2). Subdomain C is a four-stranded antiparallel β-sheet. A single disulfide bond between Cys-249 and Cys-272 covalently links β3 and β4. Because the cysteine in β4 is only found in the yeast homolog but not in any of the other NAP proteins, this disulfide link is not essential for maintaining the stability of the protein. The β-sheet has a pronounced amphiphilic character. The solvent-exposed side is relatively hydrophilic with 53% of all residues being charged. In contrast, 87% of the amino acids that line the other side of the β-sheet are hydrophobic. This hydrophobic surface is spanned in its entirety by α6 of subdomain D (Fig. 1B), whose residues interact extensively with residues from all four β-strands (Fig. 7A, which is published as supporting information on the PNAS web site). The hydrophobic core is reinforced by a highly conserved sequence motif that folds back from the extended hairpin formed by β5 and β6, the ₃₁₁SFFNFF₃₁₆ motif. Ten aromatic residues, including those in the SFFNFF motif, are arranged in an unusual edge-to-face manner, forming an aromatic core that anchors the β-sheet to the underlying α-helix (Fig. 7A). This extensive hydrophobic core confers a high degree of stability to the β-sheet and explains why truncated versions of yNAP-1 missing subdomain D are unable to fold upon expression (7) and are thus inactive (23).

The second stretch of hydrophobic amino acids that is 100% conserved among the NAP family of proteins is the ₁₈₅IPSFWLT₁₉₁ motif that is also located in domain II (Fig. 1 A). Like the SFFNFF motif, its primary function appears to be the formation of a hydrophobic core, effectively connecting α2, α4, α7, and β1 (Fig. 7B).

A database search revealed no structural homologues for the entire yNAP-1 structure. However, several proteins with antiparallel β-sheets similar to subdomain C of NAP-1 were identified. Intriguingly, one of the closest matches to subdomain C was ASF1, despite the absence of amino acid homology. Subdomain C also exhibits some similarity to other histone chaperones, such as nucleoplasmin, ASF-1, and CAF-1 (Fig. 8, which is published as supporting information on the PNAS web site). Like yNAP-1, these proteins are implicated in histone binding and nucleosome assembly. The antiparallel β-sheet domain of the NAP-1 structure and the currently available structures of histone chaperones (or the homology model of dCAF1 p55) are compared in Table 2, which is published as supporting information on the PNAS web site.

Residues 290–295 of yeast NAP have been identified as a NLS (35). These residues are part of a short antiparallel β-sheet formed by β5 and β6 that protrudes from the main structure (Figs. 1 and 2). This region is highly basic (theoretical pI 10) as is characteristic for nuclear localization sequences.

yNAP-1 Exhibits a Distinct Charge Distribution. The dome-shaped yNAP-1 dimer exhibits an uneven charge distribution: The convex surface of the NAP-1 dimer, mainly defined by the dimerization helix (subdomain A) and the upper surface of the β-sheet and the β1–β2 loop (subdomain C; Fig. 2B), exhibits a relatively nondistinct charge distribution. The exception is the subdomain encompassing the NLS (β5 and β6; Fig. 5A) (see above). In contrast, the concave underside of the yNAP-1 dimer is highly acidic (Fig. 5B). Acidic residues from α6 and from the loop connecting β3 and β4 delineate a cavity that is defined by conserved acidic residues from α2. This cavity (Fig. 5C) is too small to accommodate an α-helix, let alone an entire histone dimer; however, it may be involved in neutralizing and binding the basic N-terminal histone tails. Histone tail binding has been shown to contribute to the histone–yNAP-1 interaction. For example, the tails appear to be necessary to distinguish between H3-H4 and H2A-H2B (7). The acidic character of this region is enhanced by the presence of the acidic C-terminal domain that extends from this region but that is too disordered in the current structure to be included in the model. Although it is not essential for histone binding, this increase in acidic residues contributes to the stability of the NAP-1–histone complex. We show by fluorescence quenching that the yNAP-1–H2A/H2B complex is significantly stabilized by the presence of the N- and C-terminal domain (Fig. 5D), confirming the notion that histone binding utilizes the underside of the yNAP-1 dimer for interaction with histones.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Electrostatic potential distribution of NAP-1 dimer. (A–C) The scale of the surface potential (in K_BT, where K_B is Boltzmann constant) is shown at the top, with the blue color representing strong positive charge and the red color representing strong negative charge. The electrostatic potential on the underside of the yNAP-1 dimer is significantly negatively charged. Glu-370, the beginning of the acidic CTD, is indicated. (D) The salt-induced dissociation of NAP-1 with H2A-H2B dimer complex. Ten micromolar NAP-1 dimer was incubated with 10 μM fluorescently labeled [H2A-H2B CPM (CPM, 7-diethylamino-3-(4′-maleimidylphenyl)-4-methylcoumarin)] dimer. Under these conditions, no free NAP-1 is observed (data not shown). Tryptophan was excited at 300 nm, and fluorescence of the complex was monitored. NAP-1 (1–417) with fluorescently labeled (H2A-H2B CPM) dimer complex (red square) or NAP-1 (74–365) with fluorescently labeled (H2A-H2B CPM) dimer complex (blue circle) was dissociated by increasing the amount of salt concentration. Ratios of fluorescence intensities (470 nm/335 nm, acceptor/donor) are calculated, and then normalized data are plotted. All three independent measurements are plotted. The midpoints for the loss of FRET between NAP-1 and the (H2A-H2B) dimer are 494 mM (±8 mM) NaCl for full length NAP-1 and 414 mM (±11 mM) NaCl for N- and C-terminally truncated NAP-1 (residue 74–365).

Discussion

The structure of homodimeric yNAP-1 reveals a previously uncharacterized fold that is characterized by a long, non-coiled coil of two α-helices forming an extensive dimerization interface and a slightly offset α/β domain that is implicated in the interaction with histones and other cellular proteins. An exhaustive search of all available structures in the protein database identified only fragments of unrelated proteins containing β-strands or short helix bundles. The overall structure of NAP-1 is unrelated to any deposited structure, suggesting that NAP-1 represents a previously uncharacterized fold.

Because of the pronounced curvature of the dimerization helices, the structure assumes a dome-shaped architecture with an extended acidic region on its underside. This is the region from which the disordered C-terminal acidic domain extends (indicated by the dotted line in Fig. 2 C and D). Although not essential for histone binding or chromatin assembly, we find that the acidic domain contributes to the stability of the yNAP-1–histone complex. We have shown previously that the C-terminal region is essential for yNAP-1 to promote reversible histone H2A-H2B dimer removal on a nucleosomal template, thereby promoting nucleosome sliding (13). A recent report suggested that the C-terminal region of NAP in HeLa cells exhibits a high degree of polymorphism due to covalent posttranslational modifications by polyglutamylation. Roughly 50% of all NAP-1 molecules are glutamylated in HeLa cells. In contrast, only ≈4% of tubulin (the only other protein known to be glutamylated) is found in glutamylated form in the same cell (39). Two glutamylation sites (modified with up to 10 glutamyl units) were identified in the C-terminal region of NAP-1 and NAP-2 in HeLa cells. This modification may lead to the formation of a second or third acidic C-terminal moiety (39). Accumulation of glutamyl moieties would allow for modulation of the total negative charge of the C-terminal acidic domain, in a reversible manner, with the potential of switching between different NAP functions. The presence of glutamylated NAP-1 is not restricted to HeLa cells but is also observed in different tissues and in different organisms. Interestingly, this modification seems to occur in other acidic histone chaperones as well; for example, nucleoplasmin in Xenopus (40). Together, these data implicate the underside of the β-sheet (subdomain C) as a histone interaction surface.

The antiparallel β-sheet in subdomain C of yNAP-1 appears to be a recurring theme in histone chaperones, although the structural motif is embedded within different architectures in the various chaperones (Fig. 8). This structural similarity, despite the absence of overall sequence homology and differences in tertiary and quaternary structure, may be based on the conservation of function, and the variation in overall sequence may allow for distinct roles in chromatin assembly. With one exception (51), there is no experimental evidence that this domain is actually involved in histone binding; rather, it seems to provide the scaffold onto which histone binding elements are grafted.

The dimerization motif found in yNAP-1 is completely different from the commonly found coiled-coil or helix bundle dimerization motif. The coiled-coil is formed by component helices coming together to bury hydrophobic residues as they twist around each helix (41). In contrast, the two long α-helices in yNAP-1 promote dimerization by an antiparallel, tram-track-like pairing, with hydrophobic residues interdigitating every helical turn. Coiled-coil formation is likely prohibited by the strong bends in the dimerization helix (Fig. 3A). The large buried surface area and highly hydrophobic nature of the amino acids between the two paired helices is consistent with the finding that yNAP-1 exists as a dimer or multimer of dimers in solution (9). Sequence alignment suggests that the dimerization motif is a highly conserved feature of all members of the NAP-1 family of proteins. For example, the N-terminal region (residue 25–66; Fig. 1C) of TAF1/SET was predicted to be important for dimerization via a coiled-coil interface, and dimerization was shown to be important for TAF1/SET activity (42).

Members of the NAP family have been shown to bind to a variety of proteins involved in transcription, cell cycle regulation, protein shuttling, chromatin assembly, histone variant exchange, and chromatin remodeling function (12, 16, 34, 43, 44). Drosophila NAP-1 coprecipitates with core histones H2A-H2B from embryos (5). Similarly, human NAP-1 coprecipitates with H2A from HeLa cytoplasmic extract (11). Recombinant yeast NAP-1 binds to linker H1 and high mobility group (HMG) proteins (45). Several other proteins, for example chromatin-remodeling factor SWR1 (12), importin kap114p (34), casein kinase 2 CK2 (46), transcription activator p300 (16), and cyclin (43), also have a functional interaction with NAP family members in a variety of systems. It appears that the histone binding region of NAP-1 is spatially distinct from the region that interacts with other cellular factors. For example, the ability of NAP-1 to bind histones H2A-H2B was not compromised in the presence of p300 (16) or Kap114 (34).

NAP-1 participates in the active transport of histones from the cytoplasm to the nucleus by forming a ternary complex with histones and karyopherins (47). The previously identified NLS (35) is bound specifically by Kap114p from yeast, a karyopherin with multiple cargo binding domains. Our result that β5 (encompassing the NLS-forming residues) and β6 form a short β-sheet that is slightly offset from the main β-sheet (see Fig. 1C) is consistent with the in vivo finding that histones and karyopherin can be bound by yNAP-1 simultaneously.

The majority of yNAP-1 is present in the cytoplasm because of continuous NES-dependent export from the nucleus (23, 34). In the present structure, most of the NES consensus sequence contained within α2 is shielded from solvent by the accessory domain, whereas the α1–α2 loop region is partially exposed. The α2–α3 loop and α3–α4 loop have relatively high temperature factors compared with the remainder of the accessory domain, indicating that these loop regions are quite flexible and may be able to detach from the NES. It is possible that the sequence and length variations in subdomain B may be related to differences in regulation of nuclear localization. NES masking as a mechanism to regulate nuclear occupancy has been demonstrated for p53 and STAT-1 (48, 49). Inhibition of the exportin Crm-1 (chromosome maintenance region 1) leads to the accumulation of NAP-1 in the nucleus, suggesting that Crm-1 might be one of the nuclear exporters of yeast NAP-1 (34).

In many cases, nuclear transport is regulated by phosphorylation, as has been observed for p53 (48), STAT-1 (49), and Pho4 (50). In Drosophila, CKII phosphorylates NAP-1 (46), and there are conserved serine sites among mammalian NAPs (21). NAP-2 undergoes cell-cycle-dependent changes in the phosphorylation state, and these sites are located in α3 and the α3–α4 loop of the subdomain B. These results support the hypothesis that histone chaperone localization may be regulated by CKII-mediated phosphorylation (24). Putative phosphorylation sites in yNAP-1 were identified at positions 140, 159, and 177, all in subdomain C. Phosphorylation of these serine residues has the potential to alter the interaction of this subdomain with the NES, thus regulating NES accessibility and subcellular trafficking of NAP-1.

No structure for any member of the NAP family has been described to date; thus, the structure presented here represents a paradigm for this class of proteins. The topology found for yeast NAP-1 represents a previously uncharacterized fold with unique features, although one subdomain exhibits similarities to other known histone chaperones. In particular, our structure suggests a mechanism by which nuclear localization may be regulated. The transport of canonical or variant histones into or out of the nucleus is an essential step in the activation of chromatin assembly and the regulation of the gene expression in response to a cellular signal. The masking and unmasking of nuclear export and import signals may be a general mechanism for regulating subcellular localization of NAP-1 and histones. Posttranslational modifications and masking/unmasking of specific signal sequences responsible for nuclear import and export may be important for the coordinated control of the nucleocytoplasmic transport.