Reconstitution of Nup157 and Nup145N into the Nup84 Complex

About 30 different nucleoporins (Nups) constitute the nuclear pore complex. We have affinity-purified 28 of these nuclear pore proteins and identified new nucleoporin interactions by this analysis. We found that Nup157 and Nup170, two members of the large structural Nups, and the Gly-Leu-Phe-Gly nucleoporin Nup145N specifically co-purified with members of the Nup84 complex. In addition, Nup145N co-enriched during Nup157 purification. By in vitro reconstitution, we demonstrate that Nup157 and Nup145N form a nucleoporin subcomplex. Moreover, we show that Nup157 and Nup145N bind to the heptameric Nup84 complex. This assembly thus represents approximately one-third of all nucleoporins. To characterize Nup157 structurally, we purified and analyzed it by electron microscopy. Nup157 is a hollow sphere that resembles a clamp or a gripping hand. Thus, we could reconstitute an interaction between a large structural Nup, an FG repeat Nup, and a major structural module of the nuclear pore complex.


INTRODUCTION
Nucleocytoplasmic transport occurs through nuclear pore complexes (NPCs), highly elaborate supramolecular assemblies within the double nuclear membrane (1,2). The core structure of the NPC is formed by a spoke complex, which is sandwiched between a cytoplasmic and a nuclear ring. The eight spoke units as part of the eightfold symmetrical NPC array surround the center of the NPC through which active nucleocytoplasmic transport is thought to take place. Moreover, a nuclear basket and short cytoplasmic fibrils can be visualized by high resolution electron microscopy, which are attached, respectively, to the nuclear and cytoplasmic ring of the NPC (3)(4)(5).
Nucleoporins exist in 8-, 16-and 32-fold copies per NPC and thus this ~50MDa structure is assembled from a relatively low number of components (6). Nucleoporins are grouped into two major classes, one class with FG repeats that directly functions in nucleocytoplasmic transport by binding the soluble transport receptors (8,9), and another class that is devoid of FG repeat sequences and is considered to represent the predominant structural constituents of the NPC.
Moreover, a few of the structural nucleoporins should mediate integration in the double nuclear membrane or organize the repeat containing nucleoporins to form the active transport channel (for review see Refs. 2,10).
In the past years, significant progress has been made in the biochemical analysis of the NPC composition and organization of individual nucleoporins within NPC subcomplexes (1,2,10). Moreover, immuno-EM has revealed the relative location of nucleoporins within the overall NPC framework. Thus, first models have emerged from these studies that discuss how nucleoporins could perform their distinct roles in NPC structure and nucleocytoplasmic transport (11)(12)(13).
One of the best characterized subcomplexes of the NPC is the Nup84 complex in yeast (14)(15)(16). This assembly is composed of seven subunits and was shown to perform several essential roles including functions in nuclear envelope organization, NPC biogenesis and mRNA export. Since none of the members of the Nup84 complex contain FG repeat sequences, which bind to the shuttling transport receptors, it is intriguing how this complex can specifically affect nuclear export of poly(A) + RNA. A vertebrate Nup107-Nup160 complex (17)(18)(19), which is 4 homologous to the yeast Nup84 complex, was also shown to be essential for the postmitotic nuclear pore assembly (19)(20)(21).
Recently, we have reported that the Nup84 complex could be reconstituted in vitro from its recombinant seven subunits and exhibits a Y-shaped structure in the electron microscope, which is undistinguishable from the native complex (16). Thus, the Nup84 complex can selfassemble in a modular way from distinct smaller nucleoporin construction units. Motivated by these findings, we sought to extend the Nup84 complex further by searching for new interaction partners and successively assemble them in vitro into the pre-existing complex. A nucleoporin interaction analysis has been recently performed (22). Protein interactions that occur at individual FG Nups were sampled using immobilized nucleoporins and yeast extracts. This study showed that several FG Nups also captured the hexameric Nup84 complex.
Here, we conducted a proteomic approach to unravel additional nucleoporin interactions.
We found that Nup157 and Nup145N are specifically associated with purified members of the Nup84 complex. In agreement with these findings, Nup157 and Nup145N could be reconstituted into the heptameric Nup84 complex. We examined in the electron microscope purified Nup157 and found that it has a characteristic clamp-like structure. Our data thus show the interaction of a large structural Nup with the Nup84 complex, a key structural module of the NPC.

EXPERIMENTAL PROCEDURES
Microbiological techniques and DNA manipulations  Microbiological techniques (growth and transformation of E. coli and yeast) and standard DNA manipulations (cloning, PCR amplification, ligation) were performed as described earlier (16,23).

Insertion of yeast nucleoporin genes into E. coli expression vectors, yeast 2µ plasmids
and genomic integration of the TAP tag manipulations  For expression in E. coli, the ORFs of Nup145N and Nup157 were amplified by PCR and cloned into pPROEXHT-GST-TEV (16). For 5 tagging the yeast nucleoporins, the cassette harboring the TAP tag was genomically integrated at the 3'-end of the genes as described (25).
Overexpression of proteins in E. coli and yeast  All recombinant protein expression in E. coli was done with the BL21 codon plus RIL strain (Stratagene) and performed as previously described (16), except that bi-cistronic expression of GST-Nup157/Nup120 was performed over night at 16°C and induced with 0.1 mM IPTG. Yeast cells overexpressing ProtA-tagged Nups were grown in 2-6 l SDC-leu medium at 30°C to OD of ~3.5. After pelleting the cells by centrifugation, the pellet was washed in cold water and NB buffer (150 mM NaCl, 50 mM KOAc, 20 mM Tris pH 7.5, 2mM Mg(OAc) 2 , 0.15% NP40). The cell pellet was frozen in liquid nitrogen and stored at -20°C.

Affinity-purification of chromosomally TAP-tagged yeast nucleoporins and MS analysis 
Affinity-purification from 2-6 l of yeast cultures grown at 30°C in YPD medium to OD of ~3.5 was performed as described (26). Only the first affinity-purification step of the TAP-method (TEV elution) was applied. The second step (i.e. Calmodulin-Sepharose) was omitted, since it was found to be disruptive for some nucleoporin interactions (e.g. Nup133 dissociates from the Nup84 complex upon incubation with Calmodulin-Sepharose; R.K and M.L., unpublished results). Affinity-purifications of Nup53 and Nup59 were not successful. TEV-eluted nucleoporins were analyzed on Novex SDS-4-12% gradient polyacrylamide gels (Invitrogen) and stained with colloidal or normal Coomassie R250 (both Sigma). Mass spectrometry using tryptic digests of bands excised from the Coomassie stained SDS-polyacrylamide gel was performed according to (27). Analysis was performed on a Bruker Reflex III MALDI-TOF instrument.
Identification of protein bands was achieved by peptide mass fingerprints using Mascot (Matrix Science) and the MSDB protein database.
Affinity-purifications of GST-and ProtA-tagged proteins  Purification of GST-tagged proteins from E. coli were performed in NB buffer containing 10% glycerol as described (16).
Electron microscopy and image processing  For negative staining 5 µl of sample were placed on a freshly glow-discharged carbon-coated grid, then washed 4 times with water and stained with uranyl formate (2% w/v) and dried. Electron microscopy was carried out on a Philips CM120 BioTwin electron microscope equipped with a tungsten filament operating at 100 kV.
Micrographs of selected areas were recorded with a wide angle CCD camera (Gatan, DV300W1) at a nominal magnification of 93,000.
For image processing, micrographs of Nup157 were taken at a calibrated magnification of 50000 and were scanned with the Zeiss Scai Scanner at a step size of 21 µm/pixel, which corresponded to 4.2 Å at specimen level. 5000 particle images were selected and boxed using the MRC-image-processing package (28). Boxed particle images were imported to the IMAGIC 5 package (29), which was used for all further steps. Particles were normalized and band pass filtered with a low frequency cutoff of 1/170 Å -1 and a high frequency cut-off of 1/9 Å -1 . The box-size of the extracted particle images was 50 x 50 Pixels. For initial alignment the alignment by classification strategy was chosen. After a few rounds of refinement the class averages stabilized. Relative spatial orientations of the class averages were determined by sinogram correlation. This process was started several times starting with different class averages for the initial determination of Euler-angles. Three-dimensional maps were calculated using the exact weighted back projection algorithm. Determination of Euler angles and calculation of threedimensional maps was repeated until the map converged to a stable shape, from which projections could be generated which were similar to all of the initially observed class averages.
A three-dimensional map was further refined by projection matching. This process was repeated several times. Resolution of the final map was estimated by Fourier-Shell correlation (FSC). FSC was 0.5 at a spatial frequency of 1/17.5 Å -1 and crossed the 3 σ curve at 1/14 Å -1 . 8

Proteomic analysis of nucleoporin interactions 
To obtain a more comprehensive picture of nucleoporin interactions in the NPC, we sought to affinity-purify all yeast Nups in a comparative way and identify stoichiometrically and sub-stoichiometrically co-purifying bands.
Therefore, NUP genes were chromosomally tagged with the TAP (Tandem Affinity Purification) construct (26). Functionality of the TAP-tagged nucleoporins, which are essential for cell growth or whose deletion causes a slow growth phenotype, was tested by growth analysis (Supplemental Fig. 1A). In the case of non-essential Nups, whose deletion does not cause a growth inhibition, growth analyses did not reveal whether the TAP tag impairs nucleoporin function; however, all these yeast Nups have been already tagged at the C-terminus by ProtA or GFP and were shown to be assembled into the NPCs (6,30).
Subsequently, TAP-modified nucleoporins were affinity-purified from whole cell lysates under standardized conditions (see Experimental Procedures). We applied only the first affinitypurification step of the TAP-method (i.e.proteolysis-mediated elution from IgG-Sepharose by the TEV protease), but omitted the second step (i.e. Calmodulin-Sepharose), which was found to be disruptive for some known nucleoporin interactions (R.K and M.L., unpublished results). The TEV-eluates from the various nucleoporin purifications were analyzed by SDS-PAGE and distinct bands discernable by Coomassie staining were excised and analyzed by MALDI-TOF mass spectrometry (see Experimental Procedures and Supplemental Fig. 2). Due to the omission of the second affinity purification step, all the nucleoporin purifications were contaminated by common impurities (e.g. heat shock proteins, fatty acid synthetase, translation factors, ribosomal proteins etc.) that tend to stick to IgG-Sepharose. Nevertheless, side-by-side comparison of the different nucleoporin preparations allowed us to identify specific co-purifying bands (Fig. 1).
Additionally, several members of the Nup84 complex co-enriched a specific set of nucleoporins not previously known to interact. Affinity-purified Nup84, Nup120 and Nup145C all contained substoichiometric amounts of Nup170, Nup157 and Nup145N ( Fig. 1B-C, Table 1 and Supplemental Fig. 2). In addition, we found Nic96 and Mex67 in the Nup84 preparation, and Mlp1 in the Nup120 purification (Supplemental Fig. 2). It was recently suggested that Mlp1 is bound via Nup60 and Nup145 to the NPC (33), and that the mRNA export receptor Mex67 interacts with Nup85 (23). Moreover, Nup98 (the vertebrate counterpart of Nup145N) was described to interact with the Nup107-Nup160 complex (the vertebrate counterpart of the Nup84 complex; Ref. 18,34). Further biochemical analysis and in vitro reconstitution are required to show that these observed protein interactions are direct. Consistent with the findings described above, affinity-purified Nup157 also revealed the presence of Nup120 and Nup145N (Fig. 1D). It is not clear, why only Nup120, but not the other members of the Nup84 complex, was found in the Nup157 preparation (see Discussion).
Moreover, we could not detect members of the Nup84 complex in the Nup170 preparation (Table   1). This may not be significant as Nup170-TAP could not be affinity-purified in satisfactory levels (K. S., unpublished data). Altogether, the data suggested that Nup170, Nup157 and Nup145N could be in physical contact with the Nup84 complex.
In vitro reconstitution of the Nup157-Nup145N heterodimer  To further study the interactions of Nup157 and Nup145N with the Nup84 complex, we performed in vitro reconstitution. First, we analyzed whether Nup157 and Nup145N can directly bind to each other. To perform this experiment, GST-tagged Nup145N was expressed in E. coli and purified to homogeneity by GST-affinity purification, TEV cleavage and gel filtration chromatography. Nup145N, which has a calculated M r of ~65 kDa, eluted from the gel filtration column in a relatively sharp peak with an apparent M r of ~200 kDa ( Fig. 2A).
Analytical ultracentrifugation was conducted to analyze the association state of Nup145N in greater detail (Supplemental Fig. 3). The apparent molecular weight of Nup145N was 10 determined to be 61.4 kDa from a fit to a single component model. No systematic deviations could be detected as evident from inspection of the residuals in Supplemental Fig. 3. This demonstrates that the model describes the data accurately and that only one component was present. The molecular weight of Nup145N calculated from its primary sequence is 64.8 kDa, which comes close to the value determined by analytical ultracentrifugation. Thus, Nup145N exists as a monomer under physiological buffer conditions and used protein concentrations. We assume that the unusual elution of purified Nup145N from the gel filtration column is due to the presence of GLFG repeats. It was recently reported that FG repeat domains within nucleoporins are natively unstructured and induce atypical elution during gel filtration (35).
In contrast to Nup145N, expression of Nup157 in E.coli yielded mostly insoluble protein (data not shown). Therefore, we sought to exploit yeast as expression system and overexpressed ProtA-tagged Nup157 from a high copy number (2µ) plasmid. The protein was affinity-purified on IgG-Sepharose and eluted by TEV protease cleavage. The eluate was then applied to a gel filtration column yielding highly purified Nup157 that eluted at ~200 kDa (Fig. 2B).
To analyze whether purified Nup145N and Nup157 proteins can form a complex in vitro, GST-Nup145N immobilized on GSH-beads was incubated with purified Nup157 from yeast. As shown in Fig. 2C, Nup157 bound to Nup145N in a ~1:1 stoichiometric ratio, as judged from the intensity of the Coomassie-stained bands. To show the specificity of this interaction, we tested whether Nup157 binds to GST alone. Only trace amounts of Nup157 were bound to GST-beads ( Fig. 2C). Moreover, we tested whether another large nucleoporin, Nup188, can interact with Nup145N in vitro. Similar to Nup157, ProtA-tagged Nup188 was expressed in yeast from a 2µ plasmid and purified by IgG-Sepharose and gel filtration chromatography. However, only very small amounts of Nup188 were seen to bind to GST-Nup145N (Fig. 2C).
To further characterize complex formation between Nup145N and Nup157, we mixed partially purified Nup145N and Nup157 in solution and monitored heterodimerization by gel filtration chromatography. The reconstituted Nup157-Nup145N complex eluted at ~400 kDa, which is 2 fractions earlier than the single subunits (compare Fig. 2D with 2A and 2B). Thus, Nup157 and Nup145N constitute a novel nucleoporin subcomplex.
In vitro reconstitution of Nup157 and Nup145N into the Nup84 complex  Our next aim was to assemble the Nup157-Nup145N heterodimer into the heptameric Nup84 complex. For these reconstitution studies, the pre-assembled Nup84 complex, which consists of seven subunits 11 (Nup84, Nup85, Nup120, Nup133, Nup145C, Seh1, Sec13), was immobilized on GSH-Sepharose via the GST-tagged Nup145C subunit (see Ref. 16). Then highly purified Nup157 from yeast (see Fig. 2B) and Nup145N purified from E. coli (see Fig. 2A) were incubated with the immobilized Nup84 complex. After a washing step to remove unbound material, proteins were eluted and analyzed by SDS-PAGE and Coomassie staining. As shown in Fig. 3A, Nup157 and Nup145N bound in a ~1:1 stoichiometry to the Nup84 complex, yielding an assembly of nine subunits with a calculated molecular mass of 820 kDa. However, when this supercomplex was subjected to gel filtration chromatography, most of it dissociated into the separate subcomplexes (data not shown). This suggests that the interaction between the Nup157-Nup145N heterodimer and Nup84 complex is not very strong. These in vitro findings are consistent with the observation that only small amounts of Nup157 and Nup145N were co-enriched during Nup84 complex purification from yeast lysates (see Fig. 1B-D). In further reconstitution studies, we could show that Nup157 (Fig. 3B) or Nup145N alone (data not shown) can bind to the pre-assembled Nup84 complex. This suggests that both Nup157 and Nup145N have separate binding sites on the Nup84 complex.
Interaction of Nup157 with Nup120-containing subcomplexes  Our proteomic analysis showed that affinity-purified Nup157 contained Nup120, which is a member of the Nup84 complex (see Fig. 1D). Hence, we wanted to analyze whether Nup157 and Nup120 interact directly. To this end, GST-tagged Nup157 and untagged Nup120 were co-expressed from a bicistronic construct in E.coli. Significantly, Nup120 (identified by MS) co-enriched with affinitypurified GST-Nup157 (Fig. 3C). However, E. coli heat shock proteins were also bound to beads, which could mean that Nup157 and/or Nup120 are not completely folded, or expose hydrophobic binding sites. In contrast, bi-cistronic co-expression of another GST-tagged Nup (GST-Nup84) together with Nup157 in E. coli did not reveal any co-enrichment of Nup157 during GST-Nup84 affinity-purification (Fig. 3C). This suggests that Nup157 per se is not a sticky protein.
To obtain further evidence for a Nup157-Nup120 interaction, we tested binding of soluble Nup157 purified from yeast to different in vitro reconstituted subcomplexes of the Nup84 complex immobilized on GSH-Sepharose (see Ref. 16). These analyses revealed that Nup157 was bound to the pentameric Nup120-Nup85-Seh1-Nup145C-Sec13 complex and the trimeric 12 lack Nup120 (e.g. Nup85-Seh1, Nup145C-Sec13 and Nup133-Nup84 complexes). We conclude from these studies that Nup120 exhibits a binding site for Nup157 within the Nup84 complex.
Electron microscopic analysis of purified Nup157  Since Nup157 and Nup145N could be efficiently purified and obtained in pure form, we sought to study their morphology by transmission electron microscopy (TEM). For the purified Nup145N, we could not see a discernable structure in the EM (M.L, unpublished results). This is most likely due to the small size of the Nup145N molecule and the presence of GLFG repeats, which are not structured and thus may not be contrasted enough by negative staining.
For negative staining of Nup157, we used fractions of the final gel filtration column that contained essentially pure protein (Fig. 4A). In the electron microscope, Nup157 showed a globular morphology with a diameter of ~12 nm. A significant number of Nup157 molecules exhibited an indentation (Fig. 4A). To find out whether the structure of overexpressed Nup157 differs from endogenous Nup157, we affinity-purified Nup157 from yeast that expressed chromosomally TAP-tagged NUP157. However, authentic Nup157 was undistinguishable from overproduced Nup157 (Fig. 4A).
Since micrographs of negatively stained Nup157 showed a homogeneous spread of particles, we could perform image processing to reveal more details of the Nup157 structure. By calculating the three-dimensional map of the protein, it became apparent that Nup157 was hollow and looked like a clamp or a gripping hand ( Fig. 4B; see also Discussion). The outer dimensions of the clamp were 12 nm in the long and about 7 to 8 nm in the shorter direction.

DISCUSSION
From the ~30 yeast nucleoporins, about two-third are organized in biochemically stable NPC subcomplexes. However, the direct interacting partners for remaining Nups are still unidentified (for review see Refs. 2,10,36). Moreover, it is not known how the stable NPC subcomplexes interact with each other and which nucleoporins bridge between them. By reinvestigation of nucleoporin interactions, we identified several possible new Nup interactions, and could reconstitute two of them in vitro. The observation that some of the found in vivo interactions were not reciprocal in the TAP-purifications could be due to the possibility that a substoichiometric association did not always reveal a clear-cut band on the SDS-polyacrylamide gel (please note that the TEV-eluates from the various nucleoporin purifications were analyzed by SDS-PAGE and only distinct bands discernable by Coomassie staining were excised and analyzed by MALDI-TOF mass spectrometry). The finding that only Nup120 was detected the Nup157 preparation by mass spectrometry, but not the other subunits of the Nup84 complex, is not clear. However, Nup120 was found only as second hit in the peptide mass fingerprint analysis (see Fig. 1D and Supplemental Fig. 2; the first hit was elongation factor eEF2, a possible contaminant) and thus its amount on the SDS-polyacrylamide gel cannot be estimated. Thus, we suspect that also the other smaller subunits of the Nup84 complex were present in purified Nup157, but due to a reduced Coomassie staining were not seen as distinct bands and therefore not analyzed by MS.
Due to the observed substoichiometric interactions between Nup157, Nup145N and the Nup84 complex, we performed in vitro reconstitution. In this way, we could reconstitute the interactions between nine nucleoporins, an assembly, which represent 30% of all yeast nuclear pore proteins. However, the nine subunit assembly, formed in vitro between the robust heptameric Nup84 and the dimeric Nup157-Nup145N complexes, is not stable enough to perform further biochemical purification. Thus, for a complete in vitro reconstitution of the NPC, low affinity interactions between NPC subcomplexes as described here could pose a problem. On the other hand, it is conceivable that several low affinity contacts as part of the entire NPC structure will stabilize the overall assembly.
We also wanted to analyze the structure of the Nup84 complex bound to Nup157 in the electron microscope. However, images of the Nup84 complex with bound Nup157 were not homogenous and only in a few cases contact between Nup157 molecules and the Y-shaped 14 Nup84 complex could be discerned (M.L., unpublished data). We suspect that during specimen preparation for electron microscopy the unstable assembly of the Nup84 complex with Nup157 largely dissociated. Further work is required to reveal the structural details of how Nup157 is bound to Nup84 complexes.
Taking our biochemical data together, it is tempting to speculate that Nup157 (together with other structural Nups like Nup170) could link Nup84 complexes together generating a higher structural assembly. A single Nup157 molecule, whose structure resembles a clip with two spars, might access two separate Nup84 complexes. The Nup84 complex was estimated to occur in 16 copies per NPC and was suggested to represent a major part of the octagonal spoke-ring complex (2). However, also other scenarios of interaction are conceivable. Thus, the Nup157-Nup145N dimer could form a dual interface between two Nup84 complexes, or Nup157 bridges between a Nup84 complex module and another structural Nup.
Besides the in vitro reconstituted interactions of Nup157 and Nup145N with the Nup84 complex, we also describe in this work that Nup170 (homologue of Nup157), Mlp1 and Nic96 exhibit physical connections to the Nup84 complex. Since all these nucleoporins except Nup145N lack FG repeat sequences, the identified contacts may be important for the structural organization of the NPC. The observation that Nic96 was found to be associated with purified Nup84 suggests that two key structural modules of the NPC -the stable Nic96 complex composed of Nic96, Nup57, Nup49 and Nsp1 (37) (see also Table 1) and the Nup84 complexcould come in direct contact.
The in vitro finding that the Nup157-Nup145N heterodimer forms an assembly with the Nup84 complex, which can be easily dissociated, could reflect an in vivo requirement for a dynamic NPC biogenesis and/or organization. Notably, the conserved Nup145 is made as a precursor, which is post-translationally cleaved into two functionally separated domains, Nup145N and Nup145C in yeast (38,39), or Nup98 and Nup96 in mammals (34,40). The autoproteolytic cleavage of Nup145 is not essential in yeast, but becomes essential in cells lacking Nup188 (38,39). Moreover, self-cleavage of the Nup98-Nup96 precursor in mammals is crucial for NPC assembly (34,40). Our work has revealed that both Nup145C and Nup145N have physical contact to the Nup84 complex, but the way they interact with this NPC structural module is different. Whereas the Nup145C domain is a stable subunit of the Nup84 complex, the 15 Nup145N domain is only loosely associated. It is possible that Nup145N has to be cleaved off from Nup145C, in order to allow a dynamic interaction with the Nup84 complex.
Interesting in this context are recent studies, which showed that the binding dynamics between structural nucleoporins govern NPC permeability and affect channel gating (41).
Notably, Goldfarb and colleagues observed that yeast cells lacking Nup170 exhibit a reversible dissociation of several structural nucleoporins (including members of the Nup84 complex) from NPCs upon addition of aliphatic alcohols or chilling (41). Thus, dynamic rearrangements of structural NPC modules may not only be important for NPC assembly, but also during translocation of large cargo through the transport channel, which may not be possible with a rigid NPC structure. Linker nucleoporins, which connect structural modules, could regulate this structural flexibility.
Finally, we report in this study the first structural analysis of a so-called large Nup (Nup157), which is thought to be a major structural constituent of the spoke-ring complex of the NPC (42)(43)(44)(45). Isolation of this and other large Nups (e.g. Nup188, Nup192) in E. coli was limited due to their low solubility. Thus, we developed a method to overexpress these large Nups in yeast followed by subsequent affinity purification and conventional chromatography. Since these three large Nups could be obtained in reasonable (µg) amounts and pure form, we could perform the first biochemical and EM analyses (not shown in this study is the EM structure of Nup188 and Nup192).
The observation that Nup157 is not an elongated molecule, but forms a hollow sphere, is interesting in the structural context of the subunits of the Nup84 complex. Within the Nup84 complex, Nup120, Nup145C, Nup85 exhibit elongated structures as shown by electron microscopy (15,16). Notably, using computational and biochemical methods several subunits of the yeast Nup84 complex (Nup120, Nup145C, Nup85) and homologous vertebrate Nup107-160 complex were suggested to be elongated because they contain β-propeller/α-solenoid arrangements (46). Similarly, human Nup133 contains two domains: a carboxy-terminal domain responsible for its interaction with its subcomplex through Nup107, and an amino-terminal domain whose crystal structure reveals a seven-bladed β-propeller (47). The N-terminal domain of Nup159 had been crystallized and shown to contain an unusually asymmetric seven-bladed βpropeller (48). X-ray analysis of Nup157 or domains of it should reveal, which structural motifs occur in these large Nups. Protein bands were identified by mass spectrometry (see Table 1). Marked on the right are prominent contaminants (e.g. Hsp70, Rpl3) and the TEV protease. Interestingly, Rea1 was detected in the Gle2 and Nup116 preparations. Affinity-purification of Nup84-TAP. B, Nup120-TAP (C) and Nup157-TAP (D). Only nucleoporin bands identified by mass spectrometry, but not contaminants, are labeled on the right. Shown are also molecular weight markers.