The Crystal Structure of Archaeal Nascent Polypeptide-associated Complex (NAC) Reveals a Unique Fold and the Presence of a Ubiquitin-associated Domain*

Nascent polypeptide-associated complex (NAC) was identified in eukaryotes as the first cytosolic factor that contacts the nascent polypeptide chain emerging from the ribosome. NAC is highly conserved from yeast to humans. Mutations in NAC cause severe embryonically lethal phenotypes in mice, Drosophila, and Caenorhabditis elegans. NAC was suggested to protect the nascent chain from inappropriate early interactions with cytosolic factors. Eukaryotic NAC is a heterodimer with two subunits sharing substantial homology with each other. All sequenced archaebacterial genomes exhibit only one gene homologous to the NAC subunits. Here we present the first archaebacterial NAC homolog. It forms a homodimer, and as eukaryotic NAC it is associated with ribosomes and contacts the emerging nascent chain on the ribosome. We present the first crystal structure of a NAC protein revealing two structural features: (i) a novel unique protein fold that mediates dimerization of the complex, and (ii) a ubiquitin-associated domain that suggests a yet unidentified role for NAC in the cellular protein quality control system via the ubiquitination pathway. Based on the presented structure we propose a model for the eukaryotic heterodimeric NAC domain.

In eukaryotes nascent polypeptide-associated complex (NAC) 1 is a very abundant heterodimeric cytosolic protein complex composed of ␣and ␤NAC, which show substantial homology with each other (1). It was originally characterized as the first ribosome-associated protein to contact the emerging polypeptide chain (2). Protease protection assays further suggested that NAC might function as a shield for newly synthesized polypeptides against inappropriate interaction with cytosolic factors (3). It was proposed that cycles of binding and releasing NAC would expose the polypeptide to the cytosol in "quantal units," rather than amino acid by amino acid. NAC would thus contribute to fidelity in cotranslational processes such as targeting and folding. NAC was also shown to play a role in regulation of ribosome access to the translocation pore in the endoplasmic reticulum membrane in cotranslational protein translocation (4 -6). Still, the cellular function of NAC seems to be diverse and is probably not restricted to translation, because transcription-related functions have been suggested for the human ␣NAC subunit (7). In addition, yeast NAC was shown to play a role in the import of proteins into mitochondria (8). The importance of the in vivo function of NAC is emphasized by early embryonically lethal phenotypes of NAC mutants in mice, Drosophila melanogaster, and Caenorhabditis elegans (9,10). Furthermore, intracellular levels of the individual NAC subunits change dramatically in the context of several human diseases, such as Alzheimer disease, Trisomy 21, AIDS, and malignant brain tumors, and a role for NAC in apoptosis was only recently proposed (11)(12)(13)(14).
The two subunits of eukaryotic NAC show substantial homology to each other. Two recent comparative studies of completed archaeal genomes revealed that all of them contain one gene with apparent homology to NAC (15).
Here we present the first functional archaeal NAC homolog from Methanothermobacter marburgensis. It forms a homodimer and as eukaryotic NAC it is ribosome associated, binds to ribosomal RNA, and is in contact with the nascent chain. We present the first crystal structure of a NAC protein that provides an explanation for the unusually stable dimer formation and also implies a yet unidentified role for NAC on the ribosome, namely involvement in the cellular protein quality control system via the ubiquitination pathway. On the basis of the presented novel structure we propose a model for the eukaryotic heterodimeric NAC domain.

EXPERIMENTAL PROCEDURES
Materials-Chemicals and secondary antibodies were purchased from Merck and Sigma, restriction enzymes and Vent polymerase from New England Biolabs, protease inhibitor mixture and DNaseI from Roche Applied Science, RNase inhibitor from Promega, and TRIzol from Invitrogen.
Strains-M. marburgensis and its genomic DNA were a kind gift from Dr. R. Hedderich, MPI Marburg. Cloning was performed in Escherichia coli XL1blue (Stratagene) and expression in E. coli ER2566 (New England Biolabs).
Cloning, Expression, and Purification-Cloning experiments were performed following standard protocols (16). mth177, the gene coding for archaebacterial NAC (aeNAC), was amplified from genomic DNA and cloned into pET28a (Novagen), leading to a His tag fusion to the N terminus of aeNAC. The gene was expressed in E. coli ER2566 and the protein purified via nickel-nitrilotriacetic acid-agarose (Qiagen) according to the manufacturer's manual, followed by anion exchange chromatography (POROS 20 HQ; Applied Biosystems) and size exclusion chromatography (Superdex S75; Amersham Biosciences). If necessary, the thrombin cleavable His tag was removed.
Antibodies-Purified His-aeNAC was used for immunization of a rabbit (Davids Biotechnologie, Regensburg, Germany). Preimmune se-rum of the same rabbit did not react with any protein in a Western blot with total cell extract from M. marburgensis.
In Vitro Transcription, Translation, and Isolation of Ribosome Nascent Chain Complexes-Ribosome nascent chain complexes (RNCs) were generated in, and purified from, a cell-free rabbit reticulocyte lysate system (Promega) programmed with truncated synthetic mRNA coding for the first 77 amino acids of firefly luciferase in the presence of L-[ 35 S] methionine as previously described (17). Photo-cross-linking-Assays in which 4-(trifluoromethyl-diazirino) benzoic acid (TDBA)-modified lys-tRNA lys was added to a reticulocyte lysate translation system were according to previous studies (17,18).
Association of aeNAC with Ribosomes-10 g of M. marburgensis wet cells were resuspended in 25 ml of buffer (20 mM Hepes/KOH, pH 7.4, 50 mM Mg(Ac) 2 , 100 mM KCl, 5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 0.1 mg of DNaseI) and disrupted by French press. Cell debris was removed by centrifugation (25 min, 4°C, 11,000 rpm, SS34 rotor; Sorvall). To remove membranes and aggregates, the supernatant was spun again at 38,000 rpm for 40 min at 4°C in a 60Ti rotor (Beckman). This supernatant was called S100. Aliquots of S100 extract were adjusted to 150, 300, 500, and 1000 mM potassium acetate. 80 l of extract (100 g of protein) were loaded onto a 120-l sucrose cushion (20 mM Hepes/KOH, pH 7.4, 25 mM Mg(Ac) 2 , 2 mM dithiothreitol, 1 M sucrose, 100 M phenylmethylsulfonyl fluoride, and 150, 300, 500 or 1000 mM KAc) and spun for 1 h at 100,000 rpm in a TLA-100 rotor (Beckman) at 4°C. Ribosomal pellets were resuspended in H 2 O. One half of the supernatant and the pellet (0.45-0.48 A 260 units) were trichloroacetic acid precipitated, proteins were separated on SDS-PAGE, and supernatant and pellet were checked for the presence of aeNAC by Western blotting. From the other half of the supernatant and pellet, RNA was extracted with TRIzol following the manufacturer's protocol. Samples were run on an agarose gel to detect ribosomes by means of their rRNA.
Protein Preparation and Crystallization-The gene of ⌬1-18aeNAC was amplified by PCR, cloned into the vector pET28a (Novagen), expressed in E. coli, and purified by a combination of nickel-nitrilotriacetic acid affinity chromatography (Qiagen) with subsequent thrombin cleavage, anion exchange (SAX-PolymerLabs), and size-exclusion chromatography (Superdex S75; Amersham Biosciences). The selenomethionine-substituted ⌬1-18aeNAC was expressed using the procedure of Budisa (19) and purified as described above. Crystals were grown at 18°C in 1.6 M sodium malonate, pH 7.0, using the hanging drop vapor diffusion method. For both the ⌬1-18aeNAC and the selenomethionine derivative, well diffracting crystals were obtained by streak seeding.
Data Collection and Structure Determination-Before data collection, crystals were transferred to 3.0 M sodium malonate and flash frozen in liquid nitrogen. Multiwavelength anomalous diffraction data were collected at three wavelengths ((1 ϭ 0.9795 Å, 2 ϭ 0.9797 Å, 3 ϭ 0.9117 Å) at beamline BL1 at BESSY-II, Berlin. The individual data sets were processed with programs DENZO/SCALEPACK (20) and merged/scaled by the CCP4 program suite (21). The positions of the anomalous scatterers were located with the HySS (Hybrid Substructure Search) software of the PHENIX-project (22). Heavy atom positional refinement, location of additional sites, and phasing was done with SHARP (Global Phasing, Cambridge, UK). Subsequent density modification and model building with RESOLVE (23, 24) produced a 70% complete model. The rest of the model was built manually in O (25) and refined with programs CNS (26) and REFMAC5 (27). After initial completion of the model the structure was refined against a data set of a different selenomethionine variant crystal, which showed better statistics and higher resolution up to 2.27 Å at a low remote wavelength. Final TLS refinement (28), followed by isotropic individual B-factor refinement using REFMAC5, gave a model with R and R free values of 19.4 and 24.1%.

RESULTS
Archaea Contain a NAC Homolog-In eukaryotes, NAC is a heterodimeric complex of ␣ and ␤ subunits that are highly conserved among eukaryotes. Furthermore, the ␣ and ␤ subunits of the complex share considerable similarity with each other. A search for conserved domains in NAC using the Pfam data base (29) revealed two distinct domains, a unique "NAC domain" that is exclusively found in both the ␣and ␤NAC proteins and a "UBA (ubiquitin-associated) domain" that is only present at the C terminus of ␣NAC. Two recent comparative studies of archaebacterial genomes (15, 30) revealed the possible existence of an apparent ortholog of ␣NAC in Archaea with the same conserved domain organization (Fig. 1A). For ␤NAC no ortholog was found. The close ␣NAC ortholog MTH177 (PDB accession number NP_275320) was identified in M. thermoautotrophicus. See supplemental Fig. 1 for amino acid sequence alignments of MTH177 with human ␣NAC and of archaebacterial NAC homologs representative for different taxonomic groups. MTH177 will be called from here on aeNAC for "archaebacterial NAC." aeNAC Is Expressed in M. marburgensis-Because, to date, only genomic sequencing suggested the presence of a NAC homolog in Archaea we cloned the gene mth177 from M. marburgensis and expressed it as a His tag fusion protein in E. coli. The purified recombinant protein was used to raise polyclonal antibodies in a rabbit. Fig. 1B shows an SDS-PAGE and a Western blot with purified His-aeNAC (lanes 1 and 5) and total cell extract from M. marburgensis (lanes 2 and 6). The antibodies clearly react only against one protein of the expected size in the archaebacterial extract. The recombinant protein is slightly larger because of the His tag. The antiserum also cross-reacts with the homolog of Methanococcus jannashii (MJ0280, acces- sion number NP_247253, 128 amino acids, lanes 3 and 7), which is slightly larger, and, more importantly, even with the ␣ subunit of human NAC (lanes 4 and 8). No cross-reactivity was observed with the mammalian ␤NAC subunit (lanes 4 and 8).
aeNAC Is Associated with Ribosomes in M. marburgensis-To address the question of whether aeNAC is associated with ribosomes we sedimented ribosomes from a Dnase-treated S100 extract of M. marburgensis through sucrose cushions at different salt concentrations. aeNAC was detected in one half of the pellet (0.45-0.48 A 260 units) and supernatant with the rabbit antiserum. Fig. 1C, upper panel, shows that at 150 mM salt about 60% of aeNAC sediments with ribosomes. At 300 mM salt less aeNAC (ϳ30%) sediments with the ribosomes, and at 500 mM salt ribosome association is almost abolished. As no antibody against an archaeal ribosomal protein is available yet, we extracted RNA from the other half of the pellet and the supernatant with TRIzol and ran samples on an agarose gel. Fig. 1C, lower panel, shows that all ribosomes sediment under these conditions, as 23 S and 16 S rRNA are only found in the pellet fractions.
aeNAC Is in Contact with Nascent Chains on the Ribosome-If aeNAC is the functional homolog of eukaryotic NAC in Archaea, it should be in very close vicinity to the nascent chain as it emerges from the ribosome. This characteristic can be tested in an approach where a photo-activable cross-linker (4-(3-trifluoromethyl-diazirino) benzoic acid (TDBA)-modified lysine) is incorporated into the nascent chain during in vitro translation. When using a truncated mRNA without a stop codon, the nascent chain remains on the ribosome and can be isolated as an RNC. Because no archaebacterial in vitro translation system using truncated mRNA has been described yet, we chose the rabbit reticulocyte lysate system to prepare RNCs using mRNA coding for the first 77 amino acids of firefly luciferase (77aaffLuc) and the TDBA cross-linker. 77aaffLuc RNCs were isolated by pelleting through a high salt sucrose cushion to remove ribosome-associated proteins (including endogenous NAC). RNCs were incubated with purified human NAC, yeast NAC, or aeNAC. Fig. 2A shows that upon irradiation both subunits of human NAC efficiently cross-link to the nascent chain (lane 3), whereas the yeast NAC subunits (as observed before) result in much weaker cross-links (lane 5). aeNAC also results in a strong cross-link to RNCs from rabbit reticulocyte lysate (lane 6).
When RNCs were sedimented through a low salt sucrose cushion (100 mM KOAc) prior to irradiation, both subunits of the human complex still cross-linked to the nascent chain consistent with previously published data ( Fig. 2A, lane 7). In contrast, no cross-link to aeNAC (lane 10) was obtained anymore. This result is not due to the archaebacterial origin of the protein, because yeast NAC also could not be cross-linked anymore after sedimentation of the RNCs (lane 9). Rather, the binding partners of yeast NAC or aeNAC on the rabbit ribosome are not conserved enough to allow for tight interaction that would enable co-sedimentation of aeNAC or yeast NAC with rabbit ribosomes. Eukaryotic NAC needs the context of the ribosome to interact with nascent chains and cannot be cross-linked to polypeptide chains that are released from the ribosome by puromycin treatment (2,17). Similarly, no crosslink was obtained with aeNAC when RNCs were pretreated with puromycin to release polypeptide chains prior to addition of aeNAC and irradiation (data not shown).
Human NAC Competes with aeNAC for Binding to the Nascent Chain-Because the archaeal NAC homolog gave such a strong cross-link when tested with RNCs from rabbit reticulocyte lysate, we further tested whether human NAC could compete with aeNAC for cross-linking to the nascent chain. Fig. 2B shows the result of an experiment in which 77aaffLuc RNCs derived from 3 l of rabbit reticulocyte lysate were incubated with either 0.18 nmol aeNAC (lane 1) or with the same amount of aeNAC followed by a second incubation with 0.033 nmol human NAC (lane 2) prior to irradiation. Human NAC clearly competes with a 5-fold excess of aeNAC for contact with the nascent chain. This result probably reflects the higher affinity of eukaryotic NAC for its appropriate binding partners on a eukaryotic ribosome.
Preparation, Crystallization, and Structure Determination of ⌬1-18aeNAC-Despite extensive screening, no conditions were found in which human NAC, yeast NAC, or full-length aeNAC would crystallize. A treatment of aeNAC with different proteases, however, resulted in an almost full-length fragment resistant to proteolysis. Met-19 was identified as the first amino acid of the proteolytic fragment (data not shown). Crystallographically suitable crystals of purified ⌬1-18aeNAC diffracting to ϳ2.3 Å were obtained in 1.6 M sodium malonate, pH 7.0, by using streak seeding. The crystals belong to tetragonal FIG. 2. aeNAC interacts with nascent chains on ribosomes. A, high salt-stripped 77aaffLuc RNCs prepared in rabbit reticulocyte lysate with TDBA-lys-tRNA and L-[ 35 S] methionine were incubated with human NAC, human NAC with an ␣ subunit in which the C-terminal UBA domain was deleted at Glu-173 (hNAC⌬UBA), yeast NAC, or aeNAC as indicated. Samples were irradiated and analyzed by SDS-PAGE and fluorography. Human and yeast ␣and ␤NAC, human ␣NAC⌬UBA, as well as aeNAC were cross-linked to the nascent chain on the ribosome. When RNCs incubated with the proteins were sedimented through a sucrose cushion prior to irradiation the cross-link to yeast NAC and aeNAC was lost. B, high salt-stripped 77aaffLuc RNCs carrying the photo-cross-linker were incubated with a 5-fold excess of aeNAC followed by addition of human NAC. Samples were irradiated and analyzed by SDS-PAGE and fluorography. Human NAC successfully competes with aeNAC for interaction with the nascent chain.
FIG. 3. Crystal structure of ⌬1-18aeNAC. A, the homodimeric structure of ⌬1-18aeNAC, highlighting the topological run of the polypeptide chain of one monomer. Monomer B is shown in color, changing gradually from red to blue, beginning at the N terminus and ending at the C terminus. Monomer A is gray. All figures were prepared with MOLSCRIPT (46) and RASTER3D (47). B, structure of tetrameric ⌬1-18aeNAC as observed in the crystal. The individual chains are colored differently. C-E, comparison of the NAC domain and nucleic acid binding OB folds. C, structure of the NAC domain, illustrating characteristic features of the outer surface (which corresponds to the right side of Fig. 3A). D, structure space group P4 3 2 1 2 with cell dimensions of 90.8, 90.8, and 49.5 Å and a solvent content of 50.7%. The crystal structure was solved from an electron density map that was calculated by using data collected from ⌬1-18aeNAC protein containing selenomethionines and phased by the three-wavelength multiwavelength anomalous diffraction method. The free R of the refined structure to 2.27 Å resolution was 24.0%, whereas the working R factor was 19.4%. (See supplemental Table I summarizing the data collection, phasing, and refinement statistics and supplemental Fig. 2B for a stereo image of a portion of the electron density map.) The presented structure includes the residues 25-117 of aeNAC, as the N-terminal 6 amino acids of ⌬1-18aeNAC are disordered in the crystal structure.
Overall Structure of ⌬1-18aeNAC-The crystal structure shows that ⌬1-18aeNAC forms a homodimer (Fig. 3A). The monomer is composed of two distinct domains, a C-terminal compact three-helix bundle domain (residues 79-117) homologous to UBA domains (31,32) and an N-terminal all-␤ domain (residues 25-70), forming the novel homodimeric NAC domain. The two domains of each monomer are connected via a highly flexible linker of 8 amino acids (residues 71-78). The novel homodimeric NAC domain forms a flattened ␤ barrel with two distinct, but topologically similar, concave antiparallel ␤-sheet surfaces on each side of the flattened barrel (see supplemental Fig. 2A for a stereo image of the NAC domain). This unique domain can be described as a 6-stranded pseudo ␤-sheet barrel with the shear number S ϭ 12 (33) (supplemental Fig. 3). The homodimer interface of the NAC domain is tightly packed, showing a conserved pattern of alternating hydrophobic residues that build up the hydrophobic core of the dimeric NAC domain (see supplemental Fig. 4).
In the crystal structure, two of these dimers intercalate to form a tetramer (Fig. 3B). The two dimers interact with each other mainly via the linker region, forming a surprisingly large interface for a crystallographic contact, with a buried surface area of 1714 Å 2 and 11 hydrogen bonds. No contact is made by the two "inner" surfaces of the NAC domains. To test the biological relevance of the tetramer, analytical ultracentrifugation was performed with ⌬1-18aeNAC (in 20 mM TRIS, pH 7.5, 150 mM KCl), revealing an estimated equilibrium of monomer to dimer with a K d of 4 M and for dimer to tetramer of 300 M, suggesting that the dimer is the biologically relevant form.
Structural Homologs of the NAC Domain and Structure-Function Implications-A structural similarity search using the DALI algorithm (34) and the homodimeric NAC domain as search model yields no significant homologous or similar structures. Nevertheless, the oligonucleotide/oligosaccharide binding (OB) folds and, in particular, their "fold-related ligand binding surface" show substantial similarity to the concave ␤-sheet surfaces of the NAC domain (Fig. 3C). The nucleic acid binding superfamily is the largest within the OB fold containing proteins exhibiting a similar concave-shaped ␤-sheet surface as ligand binding surface showing a characteristic conserved amino acid distribution (35,36).
In comparison with the OB fold, the NAC domain is a 6-stranded antiparallel ␤ barrel and is composed of two polypeptide chains. Unlike OB folds, this homodimeric fold exhibits two topologically similar concave ␤-sheet surfaces that both show similarity to the canonical OB fold ligand binding surface. Both sides show characteristic features that could be responsible for binding to rRNA as the ligand. The "outer" surface of the NAC domain in Fig. 3C (which corresponds to the right side of the NAC domain in Fig. 3A) shows solvent-exposed aromatic and hydrophobic residues in positions that are equivalent to those of the aromatic or hydrophobic OB fold residues that were shown to be essential for nucleotide binding in aspartyl-tRNA synthetase (Fig. 3D) (37,38) and Protection of Telomere 1 Protein (Fig. 3E) (39). In contrast, the other "inner" NAC domain surface, in Fig. 3F, which corresponds to the left side of the NAC domain in Fig. 3A, with the two UBA domains pointing to the same side, exposes a unique stretch of positively charged residues that winds around the concave surface and could, for example, interact with the negatively charged backbone of nucleic acids.
UBA Domain-The C-terminal residues 79 -117 of aeNAC form a three-helix bundle. Structural similarity search using the DALI method (34) reveals strong homology to UBA domains of different proteins. In eukaryotic NAC only the ␣ subunit contains such a UBA domain. The aeNAC UBA domain shows a conserved characteristic hydrophobic patch on its surface (Fig. 3G) that is also found in other UBA domains and is thought to mediate protein-protein interactions (32). Deletion of this UBA domain in human ␣NAC (at Glu-173) results in a protein that still forms a complex with ␤NAC (h␣(⌬UBA)-␤NAC) and behaves in all described activity assays for NAC like the wild type protein complex. In particular, h␣(⌬UBA)-␤NAC interacts with nascent chains on the ribosome (see Fig.  2A, lane 4); even after sedimentation of the RNCs (lane 8), it binds to nucleic acids and prevents binding of RNCs to high salt-stripped microsomal membranes (data not shown).

DISCUSSION
Eukaryotic NAC is a heterodimeric protein composed of ␣and ␤NAC. NAC binds reversibly to ribosomes, where it is in contact with nascent chains as they emerge from the ribosome.
Here we report that all completed archaeal genomes contain one gene homologous to NAC. aeNAC forms a homodimer and exhibits the same domain structure as eukaryotic ␣NAC. aeNAC is also associated with ribosomes and contacts the nascent polypeptide chain emerging on the ribosome. None of the sequenced archaebacterial genomes contains a ␤NAC homolog (30). On the other hand, eukaryotic ␣and ␤NAC share homology with each other. Both contain a NAC domain indicating that they might have originated from one gene but developed differently after gene duplication during evolution. It is therefore likely that the aeNAC homodimer is the functional analog to the heterodimeric NAC in eukaryotes.
The new crystal structure of aeNAC reveals two structural domains, the NAC domain and a C-terminal UBA domain. The six-stranded ␤ barrel-like NAC domain, although structurally distinct, shows similar surface features as OB folds, which are composed of five ␤-strands. The barrel-like structure is stabilized by interacting conserved hydrophobic side chains on the inside of the barrel, while variable side chains point toward the outside and probably build the specific binding surface for the ligand or binding partner. OB folds are thought to be old folds in an evolutionary sense, and several core ribosomal and associated proteins such as L2, S12, S17 and IF1 interact via OB folds with the ribosomal RNA (35). It is therefore imaginable that the NAC domain accounts at least partially for the tight of the OB fold domain of aspartyl-tRNA synthetase from E. coli, which represents the canonical nucleic acid binding OB fold. E, structure of the N-terminal domain of Protection of Telomeres 1 Protein from Schizosaccharomyces pombe. The side chains of the OB fold residues identified as essential for nucleotide binding and the corresponding residues of the NAC domain are rendered in gray. F, surface charge distribution of the dimeric ⌬1-18aeNAC. The dimer is oriented equally to the NAC domain as shown in panel A and rotated 90°clockwise, respectively. Negative charges are colored red and positive charges blue. G, hydrophobic surface representation of the UBA domain of aeNAC, showing the surfaceexposed hydrophobic residues on the UBA domain. The respective residues are conserved between archaeal and human NAC. interaction of aeNACs with the ribosome. Strong evidence for this hypothesis comes from the experiments depicted in Fig. 2B with the human NAC complex where we deleted the UBA domain of ␣NAC. Because deletion of the UBA domain does not influence NAC activity in the described assays, we conclude that these features are localized at least partially on the NAC domain.
The homodimer formation of aeNAC via the NAC domain indicates that eukaryotic ␣and ␤NAC also dimerize via their NAC domains. Homology modeling using the homodimeric NAC domain of aeNAC as template for the yeast and human heterodimeric NAC domain supports this idea (see supplemental Fig. 5). In particular, conserved hydrophobic residues are found at the dimerization interface, emphasizing the conservation of the fold. Furthermore, the presence of solvent-exposed aromatic residues on the modeled heterodimeric NAC domain surface in corresponding positions to the ones illustrated in Fig. 3, D and E, underline the importance of this surface for NAC function. Finally, deletion variants of human NAC, which in the most extreme case consist almost only of the NAC domains of human ␣and ␤NAC (h␣71-146␤38 -130NAC), still form a complex (data not shown).
On the other hand, the dimerization via the NAC domain offers a possible explanation for the involvement of human ␣NAC in other processes such as transcription. Deletion of ␤NAC in yeast leads to a soluble ␣NAC that is no longer associated with ribosomes (40), and recombinant human ␣NAC forms homodimers 2 that could perform different functions in the cell than those in the context of the ribosome. However, the heterodimeric ribosome-associated NAC appears to be the default state in the cell, which could be concluded from the following observations. The transcription coactivator activity described for human ␣NAC was only observed when ␣NAC was tested alone or overexpressed in cell lines. Concomitant overexpression of ␤NAC together with ␣NAC abolished the transcription coactivator activity of overexpressed ␣NAC alone (7), which suggests that in the presence of ␤NAC the heterodimer is formed. In this context it is interesting to mention that ␤NAC is a substrate for caspase 3 (41). 2 Cleavage by the caspase could potentially lead to dissociation of the complex, which would set ␣NAC free for other functions.
The presence of a UBA domain in NAC has interesting implications. UBA domains are found in very diverse proteins involved in protein degradation, cell cycle control, and DNA repair and have been shown to bind ubiquitin or polyubiquitin. NMR studies have revealed that UBA domains interact via a hydrophobic patch with ubiquitin (42). Such a solvent-exposed hydrophobic patch is also present on the aeNAC UBA domain (Fig. 3G), and the hydrophobic residues are conserved in all NAC homologs.
The UBL/UBA domain protein Rad23 was thought to shuttle polyubiquitinated substrates to the proteasome. However, only recently the first biochemical evidence was provided that UBA domains of Rad23 might compete with the proteasome for these substrates and therefore inhibit ubiquitin-mediated degradation and stabilize proteins (43). NAC, which has contact to emerging not yet properly folded polypeptide chains, may have a similar function. These results together with the observation of cotranslational ubiquitination (44,45), which implies the existence of a ribosome-associated ubiquitin ligase, may indicate a yet unidentified function for NAC: whether NAC inhibits ubiquitination of newly synthesized polypeptide chains on the ribosome. Future experiments will focus on this finding.