Human nucleophosmin (NPM; also known as B23, NO38, or numatrin), encoded by the NPM1 gene, is an abundant phosphoprotein in nucleoli.1–3 It is primarily localized to nucleolus but has been shown to shuttle between the cytoplasm and nucleus.4 Two isoforms of NPM, designated B23.1 and B23.2,5 consist of 294 and 259 amino acid residues, respectively, with B23.1 being the prevalent form in all tissues.5, 6 The C-terminal 35 amino acids of B23.1 are missing from B23.2 and the N-terminal 257 residues of the two splice variants are identical.
NPM participates in numerous cellular activities, which include ribosome biogenesis, histone assembly, regulation of DNA integrity, cell proliferation, and regulation of tumor suppressors p53 and ARF.7–10 It has been shown that NPM is a downstream effector of phosphoinositol 3-kinase and is potentially important in malignant cells when the phosphoinositol 3-kinase pathway is altered.11 NPM functions as a nuclear phosphatidylinositol 3,4,5-triphosphate receptor that mediates the antiapoptotic effects of nerve growth factor by inhibiting the DNA fragmentation activity of caspase-activated DNase.11 NPM plays an important role in hematopoiesis via mechanisms involving modulation of hematopoietic stem/progenitor cell cycle progression and stress response.12 A recent study identified NPM as a novel polyadenylation mark left on messenger RNAs as a result of 3′-end processing.13
NPM contributes to oncogenesis through diverse mechanisms. Either an abnormal overexpression of NPM1 or its functional loss can lead to tumor.14 As highly proliferating and cancerous cells show increased NPM levels compared with quiescent cells,15 NPM has been regarded as a tumor marker for gastric,16 colon,17 ovarian,18 and prostate cancers.19 The NPM1 gene is frequently altered or disrupted in human cancer. Several tumor-associated chromosome translocations result in the fusion of the human NPM1 gene with the genes of anaplastic lymphoma kinase (ALK),20 retinoic acid receptor α,21 or myelodysplasia/myeloid leukemia factor 1.22NPM1/ALK fusion is derived from translocation of the N-terminal portion of the NPM1 gene (corresponding to residues 1−117) from chromosome 5 to the ALK gene locus on chromosome 2 t(2;5).20, 23
Crystal structures of Xenopus nucleoplasmin-core, Drosophila nucleoplasmin-like protein (NLP)-core, and Xenopus NO38-core have been determined.24–26 They revealed the family fold and showed how a pentamer is formed. The structures of Xenopus nucleoplasmin-core and Xenopus NO38-core also showed that two pentameric rings associate in a head-to-head fashion to form a decamer. It was proposed that the decamer binds core histones to form a large complex when they function as a chaperone.24 Human NPM preferentially interacts with histone H3/H4 complexes, binding specifically to histone H3.27 To provide the structural basis for understanding numerous functions by human NPM, we have determined the crystal structure of human NPM-core (Met9−Asp122). Our decamer structure reveals structural plasticity at the pentamer–pentamer interface, which is likely to have significant implications for the chaperone function of NPM.
MATERIALS AND METHODS
Protein expression and purification
Part of the human NPM1 gene encoding the nucleophosmin (NPM)-core (Met9–Asp122) was PCR-amplified and cloned into the expression vector pET-21a(+) (Novagen). This construct adds a hexa-histidine containing eight-residue tag to the C-terminus of the recombinant NPM-core, which was overexpressed in E. coli BL21(DE3) cells using Terrific-Broth culture medium. Protein expression was induced by 1 mM isopropyl β-D-thiogalactopyranoside, and the cells were incubated for additional 18 h at 30°C following growth to mid-log phase at 37°C. The cells were lysed by sonication in a lysis buffer (20 mM Tris-HCl at pH 7.9, 500 mM NaCl, and 5% (v/v) glycerol) containing 5 mM imidazole. The crude lysate was centrifuged at ∼18,000g for 60 min. The supernatant was applied to an affinity chromatography column of nickel-nitrilotriacetic acid-agrose (Qiagen). The protein was eluted with the lysis buffer containing 1M imidazole. The next step was gel filtration on a HiLoad 16/60 Superdex-200 prep-grade column (GE Healthcare), employing an elution buffer of 25 mMTris-HCl (pH 7.6), 100 mM NaCl, and 2 mM β-mercaptoethanol.
Crystals were grown by the hanging-drop vapor diffusion method at 24°C by mixing equal volumes (2 μL each) of the protein solution (19 mg mL−1 concentration in 25 mM Tris-HCl at pH 7.6, 100 mM NaCl, and 2 mM β-mercaptoethanol) and the reservoir solution. To grow the best crystals, we used a reservoir solution consisting of 50 mM ammonium sulfate, 50 mM bis-tris (pH 6.5), and 30% (w/v) pentaerythritol ethoxylate (15/4 EO/OH). The crystals grew to approximate dimensions of 0.2 mm × 0.2 mm × 0.1 mm in a month.
X-ray data collection and structure determination
A crystal was frozen using a cryoprotectant solution containing 10% (v/v) glycerol in the crystallization mother liquor. X-ray diffraction data were collected at 100 K on an ADSC Quantum 210 CCD area detector system at the BL-4A experimental station of Pohang Light Source, Korea. For each image, the crystal was rotated by 1° and the crystal-to-detector distance was set to 270 mm. The raw data were processed and scaled using the program suite HKL2000.28 The crystal belongs to the space group P212121, with unit cell parameters of a = 99.26 Å, b = 107.84 Å, and c = 108.69 Å. Ten monomers are found in the asymmetric unit, giving a crystal volume per protein mass (VM) of 2.15 Å3 Da−1 and a solvent content of 42.8%. The structure was solved by the molecular replacement method, using the model of Xenopus NO38-core (PDB ID: 1XEO) as the probe. When we used the decamer model as the starting model, the molecular replacement trials were not successful. They were successful only with the pentamer model of Xenopus NO38-core. Cross-rotation search followed by translation search was performed using the program CNS.29 Subsequent manual model building was done using the program O.30 The model was refined with the program CNS and several rounds of model building, simulated annealing, positional refinement, and individual B-factor refinement were performed. Noncrystallographic symmetry restraints were relaxed in successive rounds of refinement. Water molecules were added using the program CNS, followed by visual inspection, positional refinement, and B-factor refinement.
RESULTS AND DISCUSSION
We have determined the crystal structure of human NPM-core (residues Met9–Asp122) (Table I). The model has been refined to crystallographic Rwork and Rfree values of 20.4% and 25.2%, respectively, for 20−2.75 Å data without a sigma cut-off. The refined model contains 1000 residues of the 10 monomers of human NPM-core and 86 water molecules in the asymmetric unit. Both N-terminal and C-terminal residues (9–14, 119–122, and the eight-residue C-terminal tag) as well as an internal loop between residues 35 and 38 are disordered in the crystal and are not visible in the electron density map. Ten independent monomers in the asymmetric unit adopt similar conformations. When we compare monomer A against other monomers B−J, the root mean square (r.m.s.) deviations range between 0.19 Å and 0.23 Å for 100 Cα atom pairs.
Table I. Data Collection and Refinement Statistics
Values in parentheses refer to the highest resolution shell (2.85–2.75 Å).
Rmerge = ∑h∑i | I(h)i − <I(h)> | / ∑h∑iI(h)i, where I(h) is the intensity of reflection h, ∑h is the sum over all reflections, and ∑i is the sum over i measurements of reflection h.
R = ∑ | |Fobs| − |Fcalc| | / ∑ |Fobs|, where Rfree is calculated for a randomly chosen 10% of reflections, which were not used for structure refinement and Rwork is calculated for the remaining reflections.
Each monomer folds into an eight-stranded β-barrel of jellyroll topology and forms a single compact domain [Fig. 1(A)]. The β-barrel can also be described as a β-sandwich built of two opposing, four-stranded antiparallel β-sheets. The strands of the two β-sheets are arranged in the order β5-β4-β7-β2 and β1-β8-β3-[β6/β6′], respectively. The two contiguous strands β6 and β6′ lie adjacent to the long strand β3 [Fig. 1(A)]. The β-strands within each monomer are aligned approximately parallel to the fivefold axis of the molecule [Fig. 1(B)].
The four loops connecting β-strands at the pentamer–pentamer interface are well ordered (Fig. 1). In contrast, the distal face contains the flexible N- and C-termini, which lie close to each other, and a disordered loop between strands β2 and β3. The disordered β2–β3 loop is part of the first acidic A1 tract [residues Asp34–Glu39, Fig. 1(C)].
Comparison of monomer and pentamer structures
The monomer structure of human NPM-core is similar to those of Xenopus NO38-core, Xenopus nucleoplasmin-core, and Drosophila nucleoplasmin-like protein (NLP)-core.24–26 Between monomers of human NPM-core (monomer A) and Xenopus NO38-core (monomer A), the r.m.s. difference is 0.56 Å for 99 Cα atoms of structurally aligned residues (sequence identity 80% for residues 15–118 of human NPM-core). Between human NPM-core (monomer A) and Xenopus nucleoplasmin-core (monomer A), the r.m.s. difference is 0.83 Å for 87 Cα atoms of structurally aligned residues (sequence identity 51% for residues 15–118 of human NPM). Between human NPM-core (monomer A) and Drosophila NLP-core (monomer A), the r.m.s. difference is 0.96 Å for 85 structurally aligned residues (sequence identity 29% for residues 15–118 of human NPM). Between human NPM-core and Xenopus NO38-core (or nucleoplasmin-core), an outermost loop between strands β1 and β2 (Lys24–Lys27 of NPM) is most different. Between human NPM-core and Drosophila NLP-core, a β-hairpin between strands β4 and β5 (Ala64–Ile72 of NPM) is most different. The β-hairpins in human NPM-core [Fig. 1(B)] and Xenopus NO38-core26 are latched, while the β-hairpins of Xenopus nucleoplasmin-core24 are extended.
The pentameric unit of human NPM-core is also similar to those of Xenopus NO38-core, Xenopus nucleoplasmin-core, and Drosophila NLP-core. Between pentamers of human NPM-core (monomers A–E) and Xenopus NO38-core (monomers A–E), the r.m.s. difference is 0.60 Å for 466 Cα atoms of structurally aligned residues. Between human NPM-core (monomers A–E) and Xenopus nucleoplasmin-core (monomers A–E), the r.m.s. difference is 1.15 Å for 404 Cα atoms of structurally aligned residues. Between human NPM-core (monomers A–E) and Drosophila NLP-core (monomers A–E), the r.m.s. difference is 2.40 Å for 366 structurally aligned residues.
At the interface between the two neighboring monomers within the pentamer, a solvent accessible surface area of ∼960 Å2 (∼17% of the monomer surface area) is buried per monomer and many segments of the monomer are involved in intersubunit contacts. Only a few segments of the monomer do not contribute to monomer–monomer contacts within a pentamer. They are the residues 1–15 (disordered N-terminal loop), residues 22–39 (encompassing the outermost strand β2 and the acidic A1 tract), residues 96–102 (covering most of the nuclear export signal motif), residues 105–109 (covering the GSGP loop), and residues 118–122 (disordered C-terminal loop). 36.9% of the atoms in this interface are polar.
Decamer structure and comparisons
Two pentameric rings of human NPM-core associate in a head-to-head fashion to form a decamer of 522 symmetry, with approximate overall dimensions of 55 Å × 55 Å × 88 Å [Fig. 1(B)]. The two pentameric rings are eclipsed so that a monomer of a pentameric ring contacts only a single monomer of the other pentamer. At the interface between the two pentameric rings, a solvent accessible surface area of ∼560 Å2 is buried per pentamer and only two segments of the monomer contribute to the interface between the pentamers. They are residues 54–56 (part of the AKDE loop) and residues 80–82 (part of the K loop). 40.4% of the atoms in this interface are polar. The pentamer–pentamer interface has a slightly higher percentage of polar atoms than the monomer–monomer interface within a pentamer.
The two opposing pentamers interact through similar motifs in human NPM-core, Xenopus NO38-core, and Xenopus nucleoplasmin-core. However, two pentameric rings of the human NPM-core decamer are closer to each other by ∼2 Å than those of Xenopus nucleoplasmin-core.24 This situation is similar to the case of Xenopus NO38-core.26 Even though human NPM-core and Xenopus NO38-core have highly similar monomer and pentamer structures, and similar separations between the pentamers, relative orientations of the two pentameric rings within a decamer are strikingly different [Fig. 2(A)]. Obviously this kind of structural plasticity cannot be predicted by homology modeling but it can only be detected by comparing experimental structures.
When we superimpose bottom pentamers of human NPM-core and Xenopus NO38-core, there is a large rotational offset (∼20°) between the top pentamers of human NPM-core and Xenopus NO38-core [Fig. 2(A)]. This rotational offset is much larger than ∼8° observed previously between the decamers of Xenopus NO38-core and Xenopus nucleoplasmin-core.26 Furthermore, the directions of the rotational shift are opposite in human NPM-core and Xenopus NO38-core relative to Xenopus nucleoplasmin-core26 [Fig. 2(B,C)]. We believe that the small interface area between the pentamers and the involvement in the pentamer–pentamer interaction of only a limited number of flexible loop segments enable such structural plasticity at the pentamer–pentamer interface.
Because of the considerable difference in the relative orientations of the pentamers, the role of Lys80 of human NPM-core is different from that of the equivalent Lys82 of Xenopus NO38-core (Fig. 3).24 Lys80 of human NPM makes a direct hydrogen bond to Asp55 of the opposing subunit (2.41–2.83 Å between Asp55 Oδ2 atom and Lys80 Nζ atom) [Fig. 3(A)], whereas Lys82 of Xenopus NO38-core makes a water-mediated hydrogen bond to Asp57 (equivalent to Asp55 in human NPM-core) of the opposing subunit [Fig. 3(B)]. Additional pentamer–pentamer interactions are present in human NPM-core. Lys54 and Glu56 of human NPM-core have close contacts with Glu56 and Lys54 of the neighboring subunit [Fig. 3(A)]. In Xenopus NO38-core, Lys56 and Glu58 (equivalent to Lys54 and Glu56 in the AKDE loop of human NPM-core) are similarly located at the pentamer–pentamer interface but they are not involved in pentamer–pentamer interactions. Different relative orientations of pentameric rings in the decamers of human NPM-core, Xenopus NO38-core, and Xenopus nucleoplasmin-core result in significantly different surface features of their sides [Fig. 3(E)]. This observation could have a significant implication for the histone chaperone function of human NPM, since the histones were proposed to bind to the side surface of Xenopus nucleoplasmin-core.24
ARF binding site
A recent study showed that mutation of the conserved residues (L102A, G105A, and G107A) prevented human NPM from interacting with the tumor suppressor ARF, and that the GSGP loop mutants were defective for oligomerization.33 In our decamer structure of the human NPM-core, the GSGP loop is located on the molecular surface and is not directly involved in intersubunit interactions (Fig. 1). Thus the GSGP loop is readily available for binding ARF. The G105A and G107A mutations may change the conformation of the AKDE loop indirectly via an interaction between the GSGP loop and the AKDE loop [Fig. 3(D)]. As a consequence, these mutations may adversely affect the oligomerization of NPM, since the AKDE loop is located at the pentamer–pentamer interface [Fig. 3(A)]. Leu102 makes a local hydrophobic core with neighboring residues (Leu23, Val60, Val109, and Ile111), of which Leu23 and Val60 are strictly conserved, while Val109 and Ile111 are semiconserved [Fig. 3(D)]. L102A mutation would destabilize this hydrophobic core, possibly causing a conformational change or an improper folding of the NPM subunit. Ser106 of the GSGP loop makes a hydrogen bond with Glu56 of the AKDE loop within a monomer [Fig. 3(D)]. However, the S106A mutation did not affect the ARF binding and oligomerization of NPM.33 This suggests that the hydrogen bonding between Ser106 and Glu56 makes only a minor contribution to the interactions between the GSGP and AKDE loops.
We thank the staff at beamline BL-4A of Pohang Light Source, Korea for assistance during X-ray diffraction experiments. BIL thanks Korea Research Foundation for the support of his overseas training (KRF-2005-214-C00145).