The human alpha-type proteasomal subunit HsC8 forms a double ringlike structure, but does not assemble into proteasome-like particles with the beta-type subunits HsDelta or HsBPROS26.

The eukaryotic proteasome is a barrel-shaped protease complex made up of four seven-membered rings of which the outer and inner rings may contain up to seven different alpha- and beta-type subunits, respectively. The assembly of the eukaryotic proteasome is not well understood. We cloned the cDNA for HsC8, which is one of the seven known human alpha-type subunits, and produced the protein in Escherichia coli. Recombinant HsC8 protein forms a complex of about 540 kDa consisting of double ringlike structures, each ring containing seven subunits. Such a structure has not earlier been reported for any eukaryotic proteasome subunit, but is similar to the complex formed by the recombinant alpha-subunit of the archaebacterium Thermoplasma acidophilum (Zwickl, P., Kleinz, J., and Baumeister, W. (1994) Nat. Struct. Biol. 1, 765-770). The ability of HsC8 to form alpha-rings suggests that these complexes may play an important role in the initiation of proteasome assembly in eukaryotes. To test this, we used two human beta-type subunits, HsBPROS26 and HsDelta. Both these beta-type subunits, either in the proprotein or in the mature form, exist in monomers up to tetramers. In contrast to the alpha- and beta-subunit of T. acidophilum, coexpression of the human beta-type subunits with HsC8 does not result in the formation of proteasome-like particles, which would be in agreement with the notion that proteasome assembly in eukaryotes is much more complex than in archaebacteria.

The 20 S proteasome is a multicatalytic protease complex which functions as the catalytic core of the larger 26 S proteasome. This particle has been found in all eukaryotes analyzed to date and plays a major role in nonlysosomal proteolysis via the selective degradation of intracellular proteins, mostly by ubiquitin-dependent, but also ubiquitin-independent, pathways (1)(2)(3)(4). In the archaebacterium Thermoplasma acidophilum a particle is found with a similar quaternary structure, containing two different subunits named ␣ and ␤ (5). The complex consists of a stack of four rings, each containing seven subunits. The outer rings of the archaebacterial proteasome consist of ␣-subunits and the inner rings of ␤-subunits, resulting in an ␣ 7 ␤ 7 ␤ 7 ␣ 7 complex (6,7). In eukaryotes several different proteasomal subunits are known, which can be divided into two classes, ␣-type and ␤-type, based on the similarity to either the ␣or ␤-subunit of Thermoplasma (8,9). Seven different members of these ␣and ␤-type subunits may constitute, respectively, the heptameric ␣and ␤-rings of the eukaryotic proteasomes, each subunit residing possibly at a defined position (10 -13).
The 20 S proteasomes degrade unfolded proteins (14) and oxidized proteins (15,16) in vitro. Degradation is achieved by at least five distinct catalytic activities, which can be detected with chromogenic peptide substrates (17). Recently, the Nterminal threonine of the ␤-subunit of T. acidophilum was identified as the catalytically active amino acid by x-ray crystallographic studies of inhibitor-proteasome complexes (7) and extensive mutational analysis (18). The N-terminal threonine results from proteolytic processing of the N terminus, which is an essential step in the maturation of the proteasome. In eukaryotic proteasomes not all of the ␤-type subunits may be proteolytically active (18).
Little is known about the assembly of the eukaryotic proteasome particle. Several groups have identified 13-16 S particles containing most of the ␣-type subunits and some ␤-type subunits in their unprocessed form (19 -21). These so-called "preproteasomes" are converted into 20 S proteasomes, a process which is translation-dependent and is accompanied by the proteolytic processing of the ␤-type subunits. Expression of the T. acidophilum ␣-subunit in Escherichia coli revealed that this subunit forms rings of seven subunits by itself, whereas the pro-␤-subunit forms monomers (22). Coexpression of these subunits in E. coli resulted in the correct processing of the ␤-subunits and subsequent formation of functional proteasomes (22,23). Recently, it was shown that the proteolytic processing of these ␤-subunits is autocatalytic and that their folding and assembly is chaperoned by the ␣-subunits (24).
To study whether similar processes occur during the formation of eukaryotic proteasomes, we analyzed the assembly properties of the recombinant human proteasomal ␣-type subunit HsC8 (25). We found that, upon expression in E. coli, HsC8 forms by itself large complexes consisting of pairs of heptameric rings, which closely resemble the recombinant ␣-rings of T. acidophilum. To test whether simple proteasome-like complexes could be formed, we coexpressed HsC8 with the ␤-type subunit HsBPROS26 (also called HN3) (26) or HsDelta (also called Y) (27), but in contrast to T. acidophilum proteasome assembly, no such complexes could be isolated.

Cloning Strategies
Mutation and PCR 1 primers were purchased from Eurogentec. The NcoI and NdeI sites in the primer sequences are underlined and the * This work was supported by the Netherlands Technology Foundation (STW). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom correspondence should be addressed: Dept. of Biochemistry, University of Nijmegen, P. O. Box 9101, NL-6500 HB Nijmegen, The Netherlands. Tel.: 31-24-3616753/3614254; Fax: 31-24-3540525. 1 The abbreviations used are: PCR, polymerase chain reaction; mutations introduced in the proteasomal cDNAs double underlined. The pJG4-5HsC8 clone was selected via a yeast two-hybrid screening of a HeLa fusion cDNA library (28) with ␣B-crystallin (29) fused to LexA as the bait. The coding sequence of the HsC8 cDNA was amplified from the selected clone with Pwo DNA polymerase (Boehringer Mannheim) using the oligonucleotide 5Ј-ATCTGCCATGGCTCAATCG-GCACT-3Ј, introducing an NcoI site containing the ATG-start codon and the oligonucleotide BcoII, consisting of a sequence located downstream of the polylinker of the pJG4-5 vector. The PCR product was digested with NcoI and XhoI and ligated into the NcoI-XhoI sites of the pET16b expression vector (pET16bHsC8). For coexpression of HsC8 with ␤-type proteasomal subunits, the XbaI-BamHI fragment of pET16bHsC8 was subcloned into the XbaI-BamHI sites of pET24d expression vector (Novagen, pET24dHsC8). Both vectors encode the complete HsC8 protein with only a Ser to Gly mutation at position 2.
Cloning of the HsBPROS26 cDNA and subcloning into pET3c (pET3cHsB) has been described elsewhere (26). Recently, also longer human (HsN3) (30) and rat (RN3) (31) BPROS26 cDNA clones have been isolated with a putative translation start 93 nucleotides upstream of the initiation codon of HsBPROS26. To construct the vector for expression of the mature protein (HsBPROS26 mat ) in bacteria, an NdeI site comprising a new ATG start codon 39 base pairs downstream of the original start position the HsBPROS26 cDNA was generated by sitedirected mutagenesis using the oligonucleotide-directed in vitro mutagenesis system (Amersham Corp.) with the oligonucleotide 5Ј-CA-GAGGTCCAATCCATATGACCCAGAACCCC-3Ј as the mutagenic primer. The NdeI-XhoI(blunt) fragment was then cloned into the NdeI-BamHI(blunt) sites of pET3c (pET3cHsB mat ). The HsDelta cDNA was cloned from a cDNA library of the human cell line MV3 (kindly provided by J. van Groningen, Nijmegen) using PCR with Taq polymerase. The primers were based on the reported cDNA sequence of human subunit Y (EMBL data base accession no. D29012) (27). The 5Ј primer used created also an NdeI site at the start codon: 5Ј-AGAATTCATATGGCG-GCTACCTAACTAGCTGCT-3Ј. The 3Ј primer was: 5Ј-TAGGATCCAG-GATTCAGGCGGGTGGTAAGGT-3Ј. The amplified product was first ligated into the pCRII vector using a TA cloning kit (Invitrogen). The NdeI-BamHI fragment was then subcloned into the NdeI-BamHI sites of pET3c (pET3cHsD). Our HsDelta cDNA clone encodes an HsDelta protein which has a Gln to Arg mutation at amino acid position 41 of subunit Y, corresponding to position 7 in the mature protein (27). To produce the vector for expression of the mature HsDelta protein (Hs-Delta mat ) part of the HsDelta cDNA in the pCRII vector was reamplified with Pwo DNA polymerase using the 5Ј oligonucleotide: 5Ј-AGAAT-TCATATGACCACTATCATGGCCGTGCAGTTT-3Ј and the HsDelta 3Ј oligonucleotide. The generated fragment contains a 5Ј NdeI site with the new ATG start codon 102 bp downstream the original start site. This fragment was cloned into the EcoRI-BamHI sites of pTZ18r. The NdeI-BamHI fragment was subsequently subcloned into the NdeI-BamHI sites of pET3c (pET3cHsD mat ). Mutated cDNAs and cloned PCR fragments were analyzed by sequencing with the Sequenase version 2.0 sequencing kit (U. S. Biochemical Corp.). A summary of the expression constructs and putative products is shown in Table I.

Single Expression and Coexpression of Recombinant Proteasomal Proteins
Single proteasomal subunits were expressed in the E. coli strain BL21(DE3) (32). Briefly, the pET vectors were transformed to compe-tent BL21(DE3) bacteria. A single colony of the transformation plate was grown at 37°C in 1.5 ml of LB (10 g/liter casein, enzymatic hydrolysate (Sigma), 5 g/liter yeast extract (Life Technologies, Inc.), 10 g/liter NaCl) containing 200 mg/liter ampicillin (Sigma, pET3c and pET16b) or 60 mg/liter kanamycin (Merck, pET24d) to an A 550 of 0.1-0.2. This culture was stored for maximally 20 h at 4°C or used directly to inoculate 50 or 500 ml of LB containing the appropriate antibiotics. At an A 550 of 0.5-1.0 expression of proteasomal proteins was induced by adding isopropyl-1-thio-␤-D-galactopyranoside (IPTG, Research Organics) to a final concentration of 1 mM. Bacteria were harvested 3 h after induction by centrifugation at 5000 ϫ g for 15 min.
For coexpression host bacteria were cotransfected with pET24dHsC8 and a pET3c construct containing a cDNA encoding a proteasomal ␤-subunit. Bacteria containing both vectors were selected on LB plates containing ampicillin and kanamycin. The expression procedure was as described above except that all growth media contained ampicillin and kanamycin.

Bacterial Fractionation
Bacterial pellets were resuspended in 0.04 volume of the culture volume in TEN300 (50 mM Tris/HCl, pH 8.0, 0.5 mM EDTA, 300 mM NaCl) and frozen at Ϫ80°C. The bacteria were thawed at 37°C, lysed with chicken egg white lysozyme (Sigma) at 0°C, and centrifuged for 30 min at 90,000 ϫ g at 4°C. The supernatant (water-soluble fraction) was used for determination of the molecular mass of the proteasomal ␤-type subunits.

HsC8 Purification
Recombinant HsC8 was purified from bacterial water-soluble fraction on a Sepharose Q Fast Flow column (Pharmacia Biotech Inc.). After loading the HsC8 extract, the column was washed with TEN300, and proteins were eluted with a linear gradient from 300 to 600 mM NaCl. HsC8 emerged as almost pure protein (Ͼ95%) at about 550 mM NaCl. HsC8-containing fractions were pooled and concentrated in an Amicon ultrafiltration cell using a filter (Filtron) with a cutoff of 100 kDa. For electron microscopy (see below) HsC8 protein was further purified on a Superose 6 HR 10/30 prepacked size exclusion column (Pharmacia-LKB) in 10 mM Tris/HCl, pH 7.5, followed by concentration of the HsC8 peak fractions in a microsep device (Filtron) with a cutoff of 100 kDa.

Analytical Procedures
Analytical ultracentrifugation was performed on an Optima XLA analytical ultracentrifuge (Beckman Instruments) equipped with an ultraviolet absorption system (33). The concentration of the recombinant proteasome particles was adjusted to A 277 ϭ 0.11, and the protein was analyzed at 20°C in 50 mM Tris/HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA. Sedimentation velocity runs were performed at 52,000 rpm in a 12-mm double-sector Kel-F cell, which was filled with 0.12 ml of sample solution in one sector and the same volume of buffer measured at 230 nm in the other. Sedimentation coefficients were corrected to H 2 O by a standard procedure (34). Sedimentation equilibrium runs were performed at a rotor speed of 4,400 rpm in an epon-charcoal 12-mm doublesector cell filled with 1.1 ml of sample solution. The average molecular mass was determined using a linear regression computer program that adjusts the baseline absorption so as to obtain the best linear fit of ln A versus r 2 (A, absorption; r, radial distance from the rotor center). A partial specific volume of 0.73 cm 3 /g was assumed.
N-terminal amino acid sequence analysis was performed at the Sequence Center of Utrecht, Institute of biomembranes, University of Utrecht.

Electron Microscopy and Digital Image Processing
Conventional Transmission Electron Microscopy-For negative staining, a 5-l aliquot of the HsC8 suspension diluted to about 25 g/ml was adsorbed for 60 s to a glow-discharged carbon-coated collodium film on a 400 mesh/inch copper grid. The grid was then washed sequentially on two drops of distilled water, stained for 15 s with 0.75% uranyl formate, pH 4.25, and finally air-dried after removal of excess liquid with filter paper followed by suction with a capillary applied to the edge of the grid. Specimens were examined in a Hitachi H-8000 TEM (Hitachi Ltd.) operated at 100 kV. Electron micrographs were recorded on Kodak SO 163 (Eastman Kodak Co.) electron image film at a nominal magnification of ϫ 52,000. Magnification calibration was performed according to Wrighly (39) using negatively stained catalase crystals.
Mass Analysis by Quantitative Scanning Transmission Electron Microscopy (STEM)-For STEM mass determination of the HsC8 particles, a vacuum Generators (East Grinstead) HB-5 STEM interfaced to a modular computer system (Tietz Video and Image Processing System GmbH) was employed (40). For this purpose, a 5-l drop of the HsC8 suspension diluted to about 25 g/ml was adsorbed for 60 s to a thin (i.e. 2.5 nm) hydrophilic carbon film which was supported by a thick fenestrated carbon film mounted on a gold-coated 200 mesh/inch copper grid. Without negative staining, the grids were then washed sequentially on four drops of quartz bi-distilled water with a blotting step between each wash. Finally, the specimens were freeze-dried in the STEM's pretreatment chamber at Ϫ80°C overnight. Low dose (i.e. ranging from 300 e Ϫ /nm 2 to 1000 e Ϫ /nm 2 ), 512 ϫ 512-pixel elastic annular darkfield images were recorded at 80 kV and a magnification corresponding to a sampling distance of 0.92 nm in the specimen plane. All microscope parameters, including the exact magnification and dose for each image, were recorded. Evaluation of the digital electron micrographs was carried out by the IMPSYS software package (40) running on a VAX 3100 workstation. The mass histograms of the evaluated HsC8 particles were fitted by Gaussian curves.
Correlation Averaging of Negatively Stained Proteasome Particles-From digitized electron micrographs of negatively stained proteasome particles 64 ϫ 64-pixel frames of 190 top views and 110 side views were selected interactively using the SEMPER 6 image processing package (41) installed on a VAX 3100 workstation. For both top views and side views, the respective particle images were aligned relative to a single particle reference using cross-correlation techniques. After alignment, an intermediate reference particle image was computed by summing up those particles yielding the highest cross-correlation peak. This intermediate reference particle image was then used for a second round of alignment and averaging. The final average of the top views included 43 particles having a cross-correlation coefficient of Ն0.85 with the intermediate reference particle. To further enhance the 7-fold symmetry of the averaged top view, it was also 7-fold symmetrized. Similarly, the final average of the side views included 25 particles with a cross-correlation coefficient of Ն0.85. To further enhance the 2-fold symmetry of the averaged side view, it was also 2-fold symmetrized.

Recombinant HsC8 Protein Forms Large Complexes-HsC8
was isolated during the screening of a HeLa cDNA library with the yeast two-hybrid system in a search for proteins interacting with the small heat shock protein ␣B-crystallin. 2 Since little is known about the structural properties of the eukaryotic proteasomal subunits, we decided to study the HsC8 protein in more detail. We therefore cloned the coding region of the HsC8 cDNA into a pET16b expression vector (32) to produce the recombinant protein. Upon induction with IPTG, an E. coli BL21(DE3) strain transformed with the pET16bHsC8 construct expresses a protein of about 28 kDa (Fig. 1A). The amount of recombinant protein constituted up to 50% of total bacterial protein. The monoclonal antibody MCP72 directed to HsC8 reacts with this protein (Fig. 1B), confirming that the induced protein is HsC8.
To isolate the recombinant HsC8 protein, we extracted freeze-thawed and lysozyme-treated bacteria expressing the HsC8 protein with buffer containing 300 mM NaCl. The amount of recombinant protein which remains water-soluble is over 80%. We purified the HsC8 protein on a Sepharose Q anion exchange column, resulting in Ͼ95% pure HsC8 preparations. Peak fractions were pooled and subjected to size exclusion chromatography on a Superose 6 column. The fractions were analyzed by SDS-PAGE. HsC8 is eluted as a large complex in a single homogeneous peak (Fig. 2) as has also been observed for the ␣-subunit of T. acidophilum (22). By comparing the elution volume of HsC8 protein with that of calibration proteins, we estimated the molecular mass of the HsC8 complexes to be 540 Ϯ 30 kDa. We also analyzed the molecular mass of Sepharose Q-purified HsC8 complexes using analytical ultracentrifugation. The complex sediments with a sedimentation coefficient s 20,w of 12.5 S, and by sedimentation equilibrium it reveals a molecular mass of 526 kDa, which is in good agreement with the gel filtration data (see above).
Electron Microscopy-The Thermoplasma proteasomal ␣-subunit appeared to consist of a double ringlike structure (22). We wondered whether the HsC8 protein forms a similar complex. Therefore we analyzed purified HsC8 complexes by electron microscopy. Electron micrographs of negatively stained HsC8 complexes (Fig. 3A) reveal two different images, namely donut-or ringlike structures with a diameter of 14.3 Ϯ 0.5 nm and double-layered rectangular structures, 14.2 Ϯ 0.8 nm ϫ 12.0 Ϯ 0.8 nm in size. They most likely represent, respectively, the top view and side view of a particle consisting of a pair of interacting rings. These images are very similar to electron micrographs of the ␣-subunit of T. acidophilum (22,42). Detailed analysis of enlarged top and side views (Fig. 3B), and correlation averaged top and side projections (Fig. 3, C and D) revealed that, similar to the ␣-rings of T. acidophilum (Ta), the HsC8-rings consist of seven subunits. As can be seen in Fig.  3C, each subunit seems to contain two domains, a large outer domain and a smaller inner domain surrounding a low mass density center. The center of the HsC8 rings seems to be "plugged," contrary to Ta-␣-rings, which have a small central pore (22). The resolution of structural detail of the correlation averaged the side view is too low to decide whether the two stacked rings are out of register (Fig. 3D), like in Ta-␣-rings (22), or in a parallel orientation, possibly caused by different rotational orientations of the complexes.
The complex of two heptameric HsC8 rings contains 14 polypeptides of 28.5 kDa and has a calculated mass of 400 kDa. This is significantly smaller than the molecular mass determined with either gel filtration or ultracentrifugation. Therefore, we also performed STEM mass measurement of unstained freeze-dried HsC8 particles (Fig. 4A) as a third mass determination method. The STEM analysis (Fig. 4B) yielded a molecular mass of 530 Ϯ 90 kDa, thus confirming the mass values determined by gel filtration (i.e. 540 Ϯ 30 kDa) and analytical ultracentrifugation (i.e. 526 kDa). The discrepancy between the estimated molecular mass and the calculated mass is not clear but might be caused by a relatively large space between the two ␣-rings (Fig. 3D).

Expression of Recombinant Proteasomal ␤-Type Subunits-
Coexpression of the ␣with either the pro-␤-subunit (23) or the mature ␤-subunit (22) of T. acidophilum in the host E. coli resulted in the generation of functional proteasomes. To investigate whether such simple eukaryotic proteasome-like particles could be produced we performed coexpression of HsC8 with two human proteasomal ␤-type subunits. To our best knowledge it is not known which ␤-type subunit(s) bind to HsC8. Therefore, we chose arbitrarily the two ␤-type subunits HsB-PROS26 and HsDelta. At first, we expressed and characterized the assembly properties of the ␤-type subunits alone in their proprotein (HsBPROS26 and HsDelta) and mature form (HsBPROS26 mat and HsDelta mat ) from constructs with, respectively, the natural start codon or a start codon placed just upstream of the codon for the N-terminal threonine of the processed forms.
Induction of the ␤-type subunits encoded by the different pET constructs resulted in the expression of polypeptides in amounts of 10 -50% of total bacterial protein (Fig. 5A). Only HsDelta mat expression, even in different E. coli BL21(DE3) host strains and using several growth conditions, was undetectable by SDS-PAGE and Western blotting (data not shown). The expressed proteins were separated by SDS-PAGE and identified on Western blots with monoclonal antibodies and compared with the homologous subunit present in human placenta proteasomes (Fig. 5B). The bacteria expressing HsB-PROS26 contain a protein which reacts with the corresponding monoclonal antibody and is, as expected for the HsBPROS26 proprotein, larger than the mature proteasomal homologue. HsBPROS26 mat has about the same mobility on an SDS-polyacrylamide gel as its proteasomal homologue. To ascertain that the recombinant HsBPROS26 mat indeed contains the wild-type threonine at the N terminus, the N-terminal amino acid sequence has been determined. It appeared that 20% of HsBPROS26 mat has a free N-terminal threonine, the rest still having the translation start methionine at the N terminus. The protein expressed from the pET-vector encoding the HsDelta proprotein, which reacts with the anti-HsDelta monoclonal antibodies, has almost the same mobility on an SDS-polyacrylamide gel as the mature proteasomal homologue. Only a faint band at the expected position of the unprocessed HsDelta form can be detected in some lysates (data not shown). The Nterminal amino acid sequence of the recombinant protein is Ala-Val-Arg-Phe-Asp-Gly, which corresponds with the positions 5-10 in the mature protein. This means that the N terminus is proteolytically cleaved in the bacterium or, alternatively, that the second AUG codon, which codes for methionine at position 4 in the mature protein, is used as a translation start site.
The HsBPROS26 proprotein and our mature-like HsDelta, both of which do not have a threonine at the N terminus, are thus very likely proteolytically inactive. Furthermore, HsBPROS26 mat having a threonine at the N terminus, but lacking other for proteolytic activity essential amino acid residues located near the active center (18), may also be inactive.
To confirm this, we tested the proteolytic activity in crude extracts of bacteria expressing ␤-type subunits using the fluorogenic test peptide substrates Suc-Leu-Leu-Val-Tyr-MCA, Bz-Phe-Val-Arg-MNA, and Cbz-Leu-Leu-Glu-␤NA. Such a test was valid to detect the proteolytic activity of the archaebacterial ␤-subunit (22). As a negative control, we used lysates of HsC8-expressing bacteria. We could not detect a significantly elevated activity in bacterial extracts containing any of the ␤-type subunits, indicating that they indeed are proteolytic inactive.
Molecular Mass of the Recombinant ␤-Type Subunits-The molecular masses of HsBPROS26, HsBPROS26 mat , and Hs-Delta present in the water-soluble fraction of crude bacterial lysates were determined using size exclusion chromatography. About 50% of the HsBPROS26 subunit and only a small frac-tion (less than 10%) of the HsBPROS26 mat and mature-like HsDelta proteins are soluble in buffer containing 300 mM NaCl. The pro-␤-subunit and mature form of T. acidophilum behave also differently (22). Fig. 6 displays Western blots of fractions obtained with the molecular mass determination on a sizecalibrated Superose 12 column. The soluble ␤-type subunits have much smaller molecular masses than the HsC8 subunit. Both the ␤-type HsBPROS26 and HsBPROS26 mat emerged at the monomer position of about 30 kDa. In contrast the HsDelta subunit is eluted at the tetramer position (Fig. 6). Thus, both ␤-type subunits do not form ring structures upon recombinant expression in the absence of other proteasomal subunits.
Coexpression of HsC8 with ␤-Type Subunits-To test whether the human ␤-type subunits, like in T. acidophilum, can assemble into proteasome-like particles with the HsC8 complex, we coexpressed HsC8 with HsBPROS26, HsBPROS26 mat or the mature-like HsDelta. The presence of both HsC8 and a ␤-subunit in IPTG-induced bacteria was confirmed on Western blots (data not shown). It appeared that HsDelta mat also could not detectably be expressed in combination with HsC8. Coexpression of the ␤-type subunits with HsC8 does not result in a change in solubility of the recombinant subunits (data not shown). The water-soluble fraction of bacteria coexpressing HsC8 with one of the ␤-type subunits was separated on a Superose 6 column and the fractions were analyzed by Western blotting (Fig. 7). Elution profiles of HsC8 obtained with single expression of HsC8 and coexpression with the ␤-type subunits were similar (data not shown). Upon coexpression with HsC8 a significant amount of HsBPROS26 is present in the HsC8-containing higher molecular mass fractions. There is, however, still HsBPROS26 present in the monomer region, although HsC8 is present in at least 5-fold excess (data not shown). The molecular mass of the HsC8-HsB-PROS26 complex is not increased up to the size of proteasomes (which are eluted in fraction 4, data not shown), indicating that HsBPROS26 cannot form a ring structure on the HsC8 complex. The ␤-type subunit is also not processed into the mature form and no proteasomal activity is present in the HsC8-HsB-PROS26 fraction. Furthermore, we did not observe a change in the elution profile of HsBPROS26 mat and mature-like HsDelta on a Superose 6 column when coexpressed with HsC8, compared with single expression. Thus coexpression of the ringforming ␣-type subunit HsC8 with two different ␤-type subunits does not result in the formation of proteasome-like complexes, suggesting that for their formation other ␣and ␤-type subunits are needed. DISCUSSION In this report we present the first detailed investigation of the assembly properties of isolated eukaryotic proteasome subunits. We show that the recombinant eukaryotic proteasomal ␣-type subunit HsC8 forms ringlike structures. Accordingly, the HsC8 subunits assemble into double-ring type structures with two rings, consisting of seven subunits each, stacking on top of each other. Rings containing seven subunits were also found in proteasomes of eukaryotes (43,44) and T. acidophilum (7). Recombinant Ta-␣-subunits also form complexes consisting of two heptameric rings. Correlation averaged top views of both complexes appear very similar (compare Fig. 3C with Fig. 4 in Zwickl et al. (22)). A remarkable difference between the human and archaebacterial complexes is that the Ta-␣-complex reveals a small, stain-filled hole in the center, which is apparently far less pronounced or absent in HsC8 rings. Unfortunately, the resolution in the side views of HsC8 double rings is not sufficient to distinguish if the two rings are staggered, like the Ta-␣-rings, or in register relative to each other.
The observation that the eukaryotic HsC8 forms rings in the absence of other ␣-type proteasomal subunits was surprising, since in eukaryotic proteasomes the ␣-rings may consist of seven different subunits (8,36). Indications for this are that in purified proteasomes from human placenta (36,37) and yeast (45) all seven ␣-type subunits are found, and that in yeast all ␣-type subunits except Y13 are essential for life (see Heinemeyer et al. (8) and references therein). Furthermore, immuno-EM localization studies of the ␣-type subunits HsPROS30/HC2 and XAPC7 (10) and the ␤-type subunit HsB-PROS26/HN3 (11) in human placenta proteasomes indicate that these subunits were present in two copies per proteasomal particle. The property of HsC8 to form rings may, however, indicate that this subunit can actually occupy several positions in ␣-rings of proteasome(-like) particles, which would result in proteasome subpopulations differing in the content of ␣-type subunits. Such compositional variation could explain why the monoclonal antibody p31k (46), which is directed to HsC8 (47), 3 stained mainly areas in rat liver cells that are close to the bile canaliculi, whereas a much more general staining was obtained with a polyclonal antiserum directed to proteasomes (48,49). Also in C2.7 myoblasts (50), LCLC103 lung carcinoma cells and the adherent fraction of NCI-H524 lung carcinoma cells, 4 and cells in some stages of early embryonic development of the newt Pleurodeles waltl (51), the p31k antibody gave different staining patterns compared with the monoclonal antibodies directed to other ␣-type subunits. On the basis of the present results it is not clear whether putative proteasomal complexes with a high HsC8 content are assembly intermediates or functional proteasome complexes. Proper folding, assembly, and processing of the ␤-subunits of T. acidophilum is chaperoned by the ␣-rings (22,24). The formation of an ␣-subunit ring is therefore the first event in assembly. Coexpression of the Ta-␣ and Ta-␤-subunits in E. coli produces functional proteasomes. Since HsC8 forms similar rings, these complexes may be important for proteasome assembly in eukaryotes. However, coexpression of the ringforming HsC8 and the ␤-type subunits HsBPROS26, HsBPROS26 mat , or HsDelta did not yield proteasome-like complexes, although a fraction of HsBPROS26 does bind to HsC8. Moreover, coexpression of ␤-type subunits with HsC8 did not change their solubility either (data not shown) indicating that 3  HsC8 does not chaperone the proper folding of these proteins. The three human ␤-type subunits tested thus apparently cannot bind properly to HsC8 rings. This means that the HsC8 rings and Ta-␣-rings behave differently since the Ta-␣-rings can assemble into proteasome-like particles with both processed and unprocessed Ta-␤-subunits (24). A possible explanation is that eukaryotic ␤-subunits do interact with two different ␣-type subunits in the proteasome, which define a binding pocket for a certain ␤-type subunit. This is possible since the ␣and ␤-rings are about 25°out of register, which is approximately half a subunit (6,7,52). Thus, the HsC8 ring may be an important complex for proteasome assembly, but before ␤-type subunits can bind properly to the ␣-ring, the latter needs to consist of different ␣-type subunits. In this respect it is of interest whether the ability of HsC8 to form rings by itself is a general feature of human ␣-type subunits. If so the variability in the subunit composition of proteasomes may be more extended than is believed presently. Alternatively, if only HsC8 can form rings, these rings may be a start point for proteasome assembly. As can also be concluded from the recent work on processing and assembly of ␤-type subunits in yeast (13), our data confirm that proteasome assembly in eukaryotes is much more complex than in archaebacteria.