Mass Spectrometry Reveals the Missing Links in the Assembly Pathway of the Bacterial 20 S Proteasome*♦

The 20 S proteasome is an essential proteolytic particle, responsible for degrading short-lived and abnormal intracellular proteins. The 700-kDa assembly is comprised of 14 α-type and 14 β-type subunits, which form a cylindrical architecture composed of four stacked heptameric rings (α7β7β7α7). The formation of the 20 S proteasome is a complex process that involves a cascade of folding, assembly, and processing events. To date, the understanding of the assembly pathway is incomplete due to the experimental challenges of capturing short-lived intermediates. In this study, we have applied a real-time mass spectrometry approach to capture transient species along the assembly pathway of the 20 S proteasome from Rhodococcus erythropolis. In the course of assembly, we observed formation of an early α/β-heterodimer as well as an unprocessed half-proteasome particle. Formation of mature holoproteasomes occurred in concert with the disappearance of half-proteasomes. We also analyzed the β-subunits before and during assembly and reveal that those with longer propeptides are incorporated into half- and full proteasomes more rapidly than those that are heavily truncated. To characterize the preholoproteasome, formed by docking of two unprocessed half-proteasomes and not observed during assembly of wild type subunits, we trapped this intermediate using a β-subunit mutational variant. In summary, this study provides evidence for transient intermediates in the assembly pathway and reveals detailed insight into the cleavage sites of the propeptide.

The 20 S proteasome is a macromolecular assembly designed for the controlled proteolysis of short-lived regulatory proteins as well as abnormal and misfolded proteins (1,2). The structure of this 700-kDa particle is highly conserved and can be found in eukaryotes, archaebacteria, and some eubacteria (3). It is composed of 28 subunits, arranged in a cylindrical architecture consisting of four heptameric rings: two outer ␣-type subunit rings embracing two central ␤-type subunit rings (␣ 7 ␤ 7 ␤ 7 ␣ 7 ). The ␤-subunits contain the active proteolytic sites. The main difference between the prokaryotic and eukaryotic 20 S proteasomes is that of complexity. Prokaryotic 20 S proteasomes (e.g. from Thermoplasma acidophilum (4,5)) are generally composed of identical copies of 14 ␣-subunits and 14 ␤-subunits; eukaryotic proteasomes recruit ␣and ␤-subunits from 14 different subfamilies (6,7). In mammals, complexity is enhanced further by three ␥-interferon inducible subunits, which are incorporated into the proteasome during the immune response (for a review, see Ref. 8). The bacterial proteasome from the actinomycetes Rhodococcus erythropolis is intermediate in complexity, consisting of two ␣-type and two ␤-type subunits. Any of the four possible combinations of ␣and ␤-type subunits assembles to fully active proteasomes in vivo and in vitro (9).
␤-subunits of 20 S proteasomes from both eukaryotes and prokaryotes are expressed in a precursor form, with a propeptide sequence at the N terminus. Interestingly, the propeptides from different organisms are highly divergent and are thought to play different roles in proteasome assembly and activation (10,11). During the course of assembly, they are post-translationally processed and removed (12). This autocatalytic mechanism exposes the catalytic nucleophile (i.e. the N-terminal threonine) and ensures that active sites are formed only after completion of the assembly process (13). In contrast to the short and often dispensable propeptides of the archaeal ␤-type subunits (14 -16), the Rhodococcus ␤1and ␤2-subunits are translated as precursor proteins with relatively long propeptides of 65 and 59 residues, respectively. In the absence of these propeptides, the rate of formation of the proteasome complex is retarded (13). In general, the propeptide sequence of eubacteria comprises a region of about 15 amino acids near the center of the propeptide (Ser-42 to Pro-27), which exhibits significant conservation (17). Designated the "central box," this region is assumed to be important for propeptide function. Similarly, propeptides from eukaryotic ␤-type subunits comprise more than 70 residues and have been shown to be essential for proteasome biogenesis and cell viability (18 -20).
The relatively low complexity of archaeal 20 S proteasomes has allowed the first studies of assembly. When archaeal proteasomal ␣and ␤-subunits are co-expressed in Escherichia coli, mature and fully active 20 S proteasomes are formed (16,21). In vitro studies of subunits individually produced demonstrate that the archaeal assembly appears to be conserved (12,14,15,22) with the formation of seven-membered ␣-subunit rings providing a platform or template onto which the ␤-sub-units are mounted (Fig. 1, pathway a). Nevertheless, it is noteworthy that there is no evidence for rings of ␣-subunits forming as intermediates in the in vivo assembly pathway of archaebacterial proteasomes. In contrast, in vitro assembly experiments of recombinant Rhodococcus proteasomes (9,23) showed that individual ␣and ␤-subunits remain monomeric and inactive, whereas they spontaneously form active proteasomes when allowed to interact. Therefore, it is likely that the first intermediates in the assembly pathway are ␣/␤-subunit precursor heterodimers ( Fig. 1, pathway b). However, this early intermediate species has not been identified experimentally. It is established, however, that the assembly proceeds via the formation of halfproteasomes composed of one ␣-subunit ring and one unprocessed ␤-subunit ring (␣ 7 ␤ 7 ). Two half-proteasomes assemble further to form a preholoproteasome, a species that still contains the unprocessed ␤-subunits (13,23,24). Finally, processing of the ␤-subunits occurs, and the mature holoproteasome (␣ 7 ␤ 7 ␤ 7 ␣ 7 ) is formed (9,20). Recently, the study of mutational variants from Rhodococcus proteasomes revealed that the dimerization and self-processing of the ␤-subunit precursors is accompanied by a conformational change of the ␤-subunit and that this is needed for the correct coordination of the active site residue (25).
The aim of the present study was to investigate the assembly pathway of the 20 S proteasome from Rhodococcus, using an approach not applied previously to the assembly of the proteasome, that of mass spectrometry (MS) 7 of protein complexes in their native state (26,27). We report here the in vitro assembly of ␣and ␤-subunits expressed separately, incubated together and monitored in real-time by MS. This approach enables timedependent mass changes to be recorded directly by continuous sampling into the gas phase where the reaction is quenched. This method has been applied previously to smaller oligomeric complexes than the proteasome (28,29). By recording mass spectra throughout the reaction time course, we were able to monitor the formation of an early assembly intermediate, namely the ␣/␤-heterodimer, the half-proteasomes, as well as mature 28-subunit 20 S proteasomes. In addition, we were able to determine in great detail the sites that are cleaved within the ␤-subunit propeptides during the different assembly states. More generally, besides providing valuable insight into the assembly process, the results highlight the power of this method for structural biology and biogenesis studies of macromolecular complexes in particular.

EXPERIMENTAL PROCEDURES
Plasmids and Mutagenesis-For reasons of simplicity and the availability of the crystal structure of a simplified ␣1␤1 20 S proteasome from R. erythropolis (24), only one type of ␣-subunit and one type of ␤-subunit was used for the in vitro and in vivo assembly experiments. The single gene constructs pT7-7 ␣1His 6 and pT7-7 ␤1His 6 encoding the C-terminal His-tagged ␣and ␤-subunit, respectively, where used for the in vitro assembly experiments. For the in vivo assays, the bicistronic expression plasmid (pT7-7 ␤1His 6 -␣1) was used for recombinant co-expression of the genes encoding the ␣1and the ␤1-subunit (tagged with a C-terminal His 6 tag) leading to active, mature proteasomes assembled in vivo. All plasmids were a kind gift of Dr. Frank Zühl.
The latter plasmid was also used as a template for site-directed mutagenesis experiments. The ␤1His 6 -subunit residues Asp-173 and Asp-176 were mutated to alanines in the single gene construct using the QuikChange site-directed mutagenesis kit from Stratagene. The experiments were performed according to the manufacturer's manual. The sequence of the mutant construct was verified by non-radioactive DNA sequencing.
Expression and Purification-Cells of the E. coli strain BL21 (DE3) (Stratagene) were transformed with the respective expression plasmid and grown in 6 liters LB medium to mid-log phase at 30°C. After induction with a final concentration of 1 mM isopropyl-␤-D-thiogalactopyranoside for 5 h, cells were harvested and resuspended in sonication buffer (20 mM sodium phosphate, 50 mM NaCl, pH 7.4) containing a protease inhibitor mixture (Complete EDTA-free (Roche Applied Science) used according to the manufacturer's specifications), treated for 30 min on ice with lysozyme 1 mg/ml (Sigma) and a few grains of DNase I (Roche Applied Science), and sonicated for 15 min (Sonifier 250, Branson). The lysate was further fractionated first by a low speed spin (6000 ϫ g, 15 min) followed by a high-speed spin (30, 000 ϫ g, 30 min). The supernatant was filtered and directly loaded onto a 1-ml His-trap column (GE Healthcare), previously equilibrated with buffer A (20 mM sodium phosphate, 500 mM NaCl, 20 mM imidazole, pH 7.4). Protein bound nonspecifically was removed with 10 column volumes of buffer A. Assembled proteasomes as well as single subunits were eluted with a gradient of 20 -500 mM imidazole. Fractions were analyzed by 12% Schaegger SDS-PAGE and 4 -20% Novex Trisglycine native PAGE (Invitrogen). Fractions containing the designated proteins were pooled, concentrated, dialyzed against gel filtration buffer (20 mM Hepes, 150 mM NaCl, pH 7.5), and applied to a High Load 16/60 Superdex 200 (GE Healthcare) previously equilibrated with gel filtration buffer. Fractions were FIGURE 1. Two possible assembly scenarios of the Rhodococcus 20 S proteasome. In archaea, yeast and mammalian proteasome assembly is initiated with the formation of a seven-␣-subunit ring (pathway a). By contrast, it has been suggested that in the Rhodococcus proteasome assembly, an ␣/␤ heterodimer is formed first (pathway b). Subsequently, in both cases, assembly proceeds through the formation of a half-proteasome. Two half-proteasomes then dimerize to form a preholoproteasomes. This intermediate most likely triggers both autocatalytic removal of propeptides and formation of active holoproteasomes.
analyzed by 12% Schaegger SDS-PAGE and 4 -20% Novex Trisglycine native PAGE (Invitrogen). Those fractions containing 20 S proteasomes or the individual subunits were pooled, concentrated, and used for subsequent analysis.
Native PAGE Analysis-4 -20% Novex Tris-glycine native PAGE (Invitrogen) was used to probe the assembly state of in vitro assembled ␣and ␤-subunits. 1 nM of each individually purified subunit was mixed in a final total volume of 250 l of the assay buffer (20 mM HEPES, 150 mM NaCl, pH 7.5). The reaction mixture was incubated at 37°C for 30 s, 2 min, 5 min, 30 min, 1 h, 2 h, or overnight. Reactions were quenched by flash-freezing the samples in liquid nitrogen, and the samples were applied to the native PAGE system run at 4°C. The in vivo assembled wild type proteasome and the individual purified subunits were used as controls.
Proteolysis Assay-The synthetic fluorogenic peptide Suc-LLVY-AMC (Bachem, Heidelberg, Germany), dissolved in Me 2 SO, was used as a substrate for measuring the proteolytic activity (13). Fluorescence of the reaction mixture was assayed immediately. Reaction progress was monitored at 53-s intervals by measuring the relative fluorescence of each well using excitation and emission wavelengths of 320 and 460 nm, respectively. Each experiment was repeated at least three times.
Mass Spectrometry-Nanoflow electrospray ionization (ESI)-MS and tandem MS experiments were conducted on a high mass Q-TOF type instrument (30) adapted for a QSTAR XL platform (31). Conditions were carefully chosen to allow the ionization and detection of proteasome assemblies without disrupting non-covalent interactions. Prior to MS analysis, 0.3-0.6 mg/ml aliquots of ␣and ␤-subunits samples were concentrated 5-10 times and buffer-exchanged into 1 M ammonium acetate solution using Nanosep (Pall) columns. Subsequently, the concentrations of the ␣and ␤-subunit-containing solutions were determined by UV absorbance. The assembly reaction was analyzed at 4, 25, and 37°C after mixing equimolar quantities of the ␣and ␤-subunits at various time intervals until the reaction was complete. The first time point was taken 2 min after incubating the two subunits.
Typically, aliquots of 2 l of solution were electrosprayed from gold-coated borosilicate capillaries prepared in-house as described (32). The following experimental parameters were used: capillary voltage up to 1.2 kV, declustering potential 100 -150 V, focusing potential 200 -250 V, declustering potential two 15-55 V, and collision energy up to 130 V, MCP 2350 V. In tandem MS experiments, the relevant m/z value was isolated, and argon gas was admitted to the collision cell with an acceleration potential of up to 180 V. All spectra were calibrated externally by using a solution of cesium iodide (100 mg/ml). Spectra are shown here with minimal smoothing and without background subtraction.

RESULTS
Biochemical Analysis of the in Vitro Assembly-The formation of the fully assembled proteasome complex was monitored initially in a time-dependent manner by native PAGE analysis ( Fig. 2A). The individual ␣1and ␤1-subunits give rise to bands corresponding to trimeric ␣1-subunits and individual ␤1-subunits, which appear as more than one distinct band, indicating some proteolysis. Assembly was probed at several time points between 30 s and 16 h after introducing the ␣1and ␤1-subunits at 37°C (Fig. 2A). Two bands appear initially, one in the and applied to PAGE analysis. The "high molecular weight calibration kit" for native electrophoresis (Amersham Biosciences) was used: tyroglobulin (669 kDa), ferrodoxin (440 kDa), catalase (232 kDa), aldolase (140 kDa), and bovine serum albumin (67 kDa) (M). B, proteolytic activity occurring during the in vitro assembly process is assayed by the time-dependent relative increase in fluorescence caused by the release of the fluorescent dye during hydrolysis of the Suc-Leu-Leu-Val-Tyr-AMC. The reaction progress curve is referred to as ␣1ϩ␤1 in vitro. As a control, the reaction progress curve for the proteolytic activity of the in vivo assembled proteasome, referred to as (␣1␤1 in vivo), was probed under the same conditions. molecular mass range of a half-proteasome (ϳ440 kDa). The second band has a higher molecular mass (Ͼ669 kDa) and corresponds to the holoproteasome (␣ 7 ␤ 7 ␤ 7 ␣ 7 ). Despite the fact that aliquots of the assembly assay were flash-frozen, the only detectable intermediate by native PAGE is the half-proteasome. The appearance of fully assembled proteasomes is in accord with the disappearance of half-proteasomes, and depletion of the individual subunits is in agreement with previous results (13).
To compare the activity of proteasomes assembled in this in vitro assembly reaction with those isolated after assembly in vivo, we used a fluorescence-based assay. We compared the rate of proteolysis of a fluorogenic peptide substrate during in vitro assembly with that of active, mature proteasomes assembled in vivo. Prior to cleavage of the peptide substrate, fluorescence is quenched, whereas upon proteolytic cleavage, an increase in fluorescence can be monitored. For in vivo assembled proteasomes, this proteolysis assay is characterized by a steep initial increase in fluorescence. The slope can be fitted to a first order kinetic whose saturation (at ϳ250 s) is due to substrate depletion (Fig. 2B). This demonstrates that there is no lag phase in the cleavage reaction under these conditions. Rather, cleavage is initiated immediately after the in vivo assembled proteasome is introduced to the peptide. By contrast, for the in vitro assembly reaction, in which individual ␣-subunits are introduced to ␤-subunits together with the fluorogenic substrate, an initial lag phase is clearly observed. Moreover, activity is found to increase five times more slowly than for the corresponding reaction with in vivo assembled proteasomes. This shows that despite the assembly reaction producing high molecular weight species after ϳ120 s ( Fig. 2A), activity is gained at a much slower rate, with saturation occurring after ϳ1400 s. This difference between assembly rate and activity can be explained by the requirement for processing of ␤-subunits before full proteolytic function of the proteasome is achieved in vitro.
Monitoring the in Vitro Assembly by MS-Nanoflow ESI mass spectra were recorded for the individual subunits of the proteasome, the ␣1and ␤1-subunits, prior to carrying out the assembly reaction, Fig. 3. For the ␣1-subunit, only one major series of peaks is observed (Fig. 3A), with a measured mass consistent with that calculated for an ␣1-subunit without the first Met residue and including a C-terminal His 6 ( Table 1). Additional series of peaks are assigned to minor populations of ␣/␣-homodimers and trimers, emphasizing the tendency of the ␣1-subunit to aggregate. This observation is in agreement with the native PAGE analysis of ␣1-subunits. In the spectrum recorded for the ␤1-subunits (Fig. 3B), we could not identify full-length ␤1-subunits (31,983 Da including the His 6 tag). However, a set of 12 degradation products was assigned, corresponding to removal of the first 18,22,23,25,28,29,32,36,40,42,48, and 50 residues of the propeptide, denoted ⌬18, ⌬22, etc. (Table 1). The predominant degradation product is the ⌬18 species. From the masses of these species, we can determine that all the cleavage sites in these degradation products are located within the propeptide sequence, residues Ser-48 to Thr-16 (Table 1, Fig. 3E). To examine the assembly pathway in detail, we used MS, initially monitoring the reaction at 37°C to enable comparison with results obtained by native gel analysis and the activity assay. Mass spectra were acquired at different time intervals after incubating equimolar ratios of the ␣1and ␤1-subunits. 12 min after initiating the reaction, a charge state series corresponding to an unprocessed half-proteasome is observed ( Fig.  4A and Table 1). Subsequently, after 25 min, we observed spectra in which both half-proteasomes and intact holoproteasomes were present, demonstrating the co-existence of both species in solution. After 29 min, half-proteasomes could no longer be detected, only holoproteasomes, in accord with the data from the native gel analysis (Fig. 2A). Since no further changes in mass spectra were observed after this time, we conclude that the assembly reaction is complete. Interestingly, although the reaction was monitored at various time points until completion, we could not identify the predicted preholoproteasome species (13). We therefore repeated the same experiments at lower temperatures, 25 and 4°C, in an attempt to capture this intermediate and to trap populations of the half-proteasome. However, although at 25°C we were able to trap half-proteasomes for further study, preholoproteasomes were not detected. Also, to our surprise, at these temperatures (4 and 25°C), assembly was not complete. Even after 5 days, unprocessed half-proteasome particles persist at 25°C (Fig.  4B). This observation indicates that the assembly processes is temperature-dependent.
If we now turn our attention to the low m/z region of the spectra recorded during assembly, we can monitor the proportions of the ␣1and ␤1-subunits that remain in solution during the progression of the reaction. Simultaneously, with the formation of unprocessed half-proteasome and holoproteasome, the intensity of the monomeric subunits was found to decrease. A set of seven ␤1-subunits remained in solution 12 min after the assembly reaction was initiated at 37°C (Fig.  3C). This subset of seven corresponds to those peptides with the shortest propeptide (from ⌬25 to ⌬42), implying that the ␤1-subunits with the longest propeptides identified prior to assembly (Fig. 3B), namely ⌬18, ⌬22, and ⌬23, are either promptly integrated within the assembling proteasome or rapidly degraded. Both ␣1-subunits and truncated ␤1-subunits could still be detected even after the assembly reaction was complete. We attribute this observation to properties of the ␤1-subunits. Specifically, we anticipate that extensive truncations may perturb the native state of the ␤1-subunits, and consequently, preclude their successful incorporation into the holoproteasome.
To confirm the incorporation of ␤-subunits with the longest propeptides within the assembled form of the half-proteasome, ions of this complex at m/z 9000 were isolated and subjected to a tandem MS experiment. This process results in the expulsion of individual ␤-subunits and the formation of a "stripped" halfproteasome complex (␣ 7 ␤ 6 ) (25). Surprisingly, a set of 10 ␤-subunits was identified and cleaved within the propeptide region to different extents (⌬18, ⌬25, ⌬28, ⌬29, ⌬32, ⌬36, ⌬40, ⌬42, ⌬48, and ⌬50) ( Table 1 and Fig. 3D). This group of ␤-subunits contains a range of truncated forms from ⌬18 to ⌬50 and is very similar to the set of monomeric ␤-subunits identified prior to assembly (Fig. 3B). The major difference is in the intensity ratio of the individual species. Specifically, prior to assembly, the ⌬25 species was predominant, whereas after assembly, the ⌬50 species gave the highest intensity peaks. Consequently, we have found that the ␤-subunits expelled from the half-proteasome have undergone more extensive processing to form shorter propeptide sequences than the ␤-subunits prior to assembly.
A close inspection of the mass spectra recorded while monitoring the assembly, in the range of 3000 -3700 m/z at all three temperatures (4, 25, and 37°C) ( Table 1 and Fig. 5), identified the appearance of two additional species. The predominant species is consistent with the ␣/␣-homodimers, emphasizing the tendency of the ␣-subunit to aggregate as observed in the native gel and mass spectra ( Figs. 2A and 3A). Interestingly, all other species correspond to ␣/␤-heterodimers in which the first 25, 28, 29, 32, and 40 residues of the ␤1-subunit propeptide are truncated. This set of ␣/␤heterodimers corresponds closely to the monomeric ␤1-subunits that were not incorporated during the assembly of the half-proteasome and holoproteasome (Fig. 3C). A possible explanation for this observation is that the reduced activity of these ␤1-subunits results in their becoming trapped within the ␣/␤-heterodimers, which are not able to assemble into the full proteasome. Mass Spectrometry Analysis of the in Vivo Assembled D173A,D176A Mutant-The formation of a wild type preholoproteasome in which propeptides are retained could not be detected in our real-time MS experiments, although this intermediate must form prior to the active proteasome particle. Therefore, a mutant of the Rhodococcus proteasome, altered in its ability to assemble, was used to characterize this late-assembly intermediate. Previously, we have reported that the ␤1-subunit mutational variant, D173A,D176A, expressed recombi-nantly in E. coli by using a bicistronic expression system, is trapped at the preholoproteasome stage (25). The mass spectrum recorded for this mutant indicates that the major charge state series corresponds in mass to preholoproteasomes, whereas the minor charge state is assigned to half-proteasomes (Fig. 6, A and B).
To examine in detail the composition of this late assembly intermediate, preholoproteasomes formed using the D173A, D176A mutant were subjected to in-source dissociation (Fig.  6C). Under these conditions of higher acceleration, series of charge states are observed at higher m/z values than the preholoproteasome. Two of these series are assigned to "stripped complexes" that have lost one or two ␣1-subunits (Table 1). At low m/z values, only ␣1-subunits are dissociated from the proteasome, consistent with the architecture in which the two ␣ 7 -ring structures are exposed and in accord with previous analyses of 20 S proteasomes (31,33,34). Expansion of the peaks assigned to the stripped complexes, minus ␣1-subunits or minus two ␣1-subunits, reveals fine splitting, indicating the incorporation of various truncated forms of either the ␣1-subunits or the ␤1-subunits (Fig. 6, D and E). We could exclude the possibility that the ␣1-subunits are cleaved as their mass measured from the low m/z region is consistent with the theoretical mass. We therefore conclude that the ␤1-subunits are truncated. By measuring the mass of the stripped complexes, we could identify the statistical incorporation of the various  Shown is a spectrum recorded 12 min after starting the assembly reaction at 37°C. The same spectrum is shown in Fig. 3C but at a lower m/z range. The charge states of dimers, labeled with stars, are indicated. Charge states of the monomers are given alongside their labels (circles). (Table 1). The spectra reveal that the intact preholoproteasome is composed of a minimum of 10 full-length ␤1-subunits. Up to four of the ␤1-subunits in any intact preholoproteasome, however, correspond to truncated ␤1-subunits (⌬18). The fact that similar ratios of truncated subunits are observed within both stripped complexes implies that truncated subunits are not stripped preferentially. In summary, we have demonstrated that the in vivo assembled preholoproteasome contains predominantly full-length ␤1-subunits but that four of the 14 ␤1-subunits correspond to truncated forms (⌬18).

DISCUSSION
In this study, we have followed the assembly of the ␣1and ␤1-subunits from R. erythropolis into the 20 S proteasome, FIGURE 6. Analysis of the preholoproteasome. A, nanoflow ES of the D173A,D176A mutant. The dominant charge state distribution corresponds to a fully assembled proteasome ␣ 14 b 14 , whereas the low abundant charge series in the 7900 -8900 m/z region corresponds to a half-proteasome complex, ␣ 7 ␤ 7 . Both species contain full-length ␤-subunits (␤ pro ) and partially degraded propeptides (␤ ⌬18 ). Inset B, an expansion of the half-proteasome particle. C, in source dissociation, mass spectra of the intact holoproteasome result in stripping of one and two ␣-subunits. The broad charge state distribution also corresponds to the intact holoproteasome that has undergone extensive charge stripping (39). The first ␣-subunit dissociation step (␣ 13 b 14 ) is expanded in D, and the region corresponding to the second dissociation step (␣ 12 ␤ 14 ) is expanded in inset E. Different stoichiometries of the full-length and partly degraded propeptide (⌬18) are labeled. using real-time MS. We were able to capture a transient intermediate along the in vitro reaction pathway, demonstrating that this early assembly intermediate is an ␣/␤-heterodimer. Such a heterodimer was predicted to exist (13); however, it has not been observed experimentally. Despite the high similarity in mass of the ␣/␤-heterodimer and the ␣/␣-homodimer, which prevents their distinction using biochemical methods, the high resolution of MS with the rapid time frame for recording spectra enabled the identification of this species. In the course of assembly, we could also follow the formation of unprocessed half-proteasomes and the gradual appearance of fully assembled holoproteasomes. This MS approach also allowed us to study in great detail the composition of the ␤-subunit propeptide during different stages of assembly. Interestingly, we could not detect preholoproteasomes that form from the docking of two unprocessed half-proteasomes during the in vitro assembly reaction. To investigate the composition of this late assembly intermediate, we used a mutational variant assembled in vivo. Overall, therefore, in this study, we were able to reveal the formation of both early and late intermediates and to demonstrate differential incorporation of truncated ␤-subunits within various stages of the assembly.
The observation that the first assembly intermediate for the Rhodococcus 20 S proteasomes is an ␣/␤-heterodimer is in contrast to that found for assembly of the Thermoplasma proteasome. The latter study showed that ␣-subunits spontaneously assemble into seven-membered rings when they are expressed alone in E. coli (16). Detailed comparison between the structures of proteasomes from Rhodococcus and Thermoplasma, however, suggests that there are fewer specific interactions and smaller contact areas between the ␣-subunits in the Rhodococcus proteasome (24). This may explain why extensive ␣-subunit interactions are observed in Thermoplasma but not in Rhodococcus. However, it remains to be established whether or not ␣-rings are assembly intermediates in vivo. This will depend upon the kinetics of the formation of the ␣/␣ homodimer versus that of the ␣/␤ heterodimer (35). If the formation of ␣/␤ heterodimers is much faster than assembly of ␣-rings, the assembly process of archeabacterial proteasomes would be expected to be similar to those of the eubacterial system studied here.
The preholoproteasome intermediate that forms from the dimerization of two half-proteasomes was not observed in our experiments, even at the lower temperatures. A previous study suggested that the conversion of preholoproteasomes into mature holoproteasomes is the rate-limiting step of the assembly process (13). However, the fact that this assembly intermediate was not detected in our analyses indicates that it is shortlived. A similar conclusion was reported in a study that used electron microscopy to characterize late events in the Rhodococcus proteasome assembly (23). In an attempt to characterize further this intermediate, we analyzed a ␤1-subunit mutational variant. Using this mutant, we found that the preholoproteasome consists primarily of the full-length ␤1-subunit but that up to four of these ␤1-subunits are substituted by truncated ␤1-subunits (⌬18) (Fig. 6). This result demonstrates that the first 18 residues of the ␤1-subunit propeptide are not necessary for docking of two half-proteasomes.
Our results also demonstrate that the assembly process is temperature-dependent. Interestingly, however, assembly of the half-proteasome occurs at similar rates, even at the lower temperatures, whereas final docking of the two half-proteasomes is significantly retarded at the lower temperatures. To explain this observation, we assume that some structural perturbation within the individual ␤1-subunits is necessary for assembly to the full proteasome. More generally, it is established that protein-protein interfaces in obligate complexes, such as the proteasome, in which the monomers do not form stable structures in vivo, are generally more hydrophobic than non-obligate associations (36,37). It may be speculated, therefore, that at the relevant physiological temperature, hydrophobic stretches are exposed. The accessibility of these hydrophobic regions may well be necessary for docking of the two half-proteasomes, explaining the favorable reaction at 37°C and the hindered assembly at lower temperatures.
Detailed analysis of spectra of the ␤1-subunits has allowed us to gain insight into the mechanism of autocatalytic removal of the propeptide and its role in assembly. Previous observations suggested that the propeptide is degraded in a processive manner by undergoing multiple cleavages. In this study, we were able to identify clearly specific cleavage sites. In an attempt to find a common property of these 12 cleavage sites, we have examined in detail the location and relative abundance of the various degradation products (Fig. 7). A common feature of these residues is that they are all on the exposed face of the complex, which might explain the mechanism of their degradation. It is also noteworthy that 10 of the residues are polar, with the exception of Leu-43 and Gly-34, both of which are flanked by polar amino acids. This could also imply that in addition to exposed residues, processive degradation targets polar residues. Moreover, although the propeptide contains only one defined structural element, an ␣-helix positioned between Gly-34 and Asp-38, the cleavage sites are distributed both within and outside of this ␣-helix. We propose, therefore, that polarity and accessibility may well be important factors in determining favorable cleavage sites within the propeptide.
It is interesting to consider how the extent of cleavage affects the incorporation of truncated subunits into the intact proteasome and various assembly intermediates. The x-ray structure of a processing-incompetent Rhodococcus proteasome indicated the importance of the central box region (Ser-42 to Pro-27) to assembly (24) (Fig. 3E). Within this region, there are two highly conserved serine residues, Ser-42 and Ser-41, which form hydrogen bonds to Arg-85 of the ␣-subunit and Arg-82 of the ␤1-subunit, respectively. These interactions position a short proceeding ␣-helix (Gly-34 to Asp-38) and allow docking of the N-terminal end of the propeptide linking the ␤-subunit to two adjacent ␣-subunit. In accord with this structural information, our results indicate that once residues within the central box region are degraded, assembly is much slower or even abolished. We found heavily truncated subunits undergoing mul-tiple cleavage (18,23,38) trapped within ␣/␤ heterodimers or remaining monomeric at the end of the assembly reaction. Overall, it seems that longer propeptide sequences increase ␣/␤-interfaces. This, in turn, promotes efficient assembly.
The observation that even the heavily truncated species (⌬50, -⌬48, ⌬42) were incorporated into the half-proteasome was surprising (Fig. 3D). The fact that these truncated variants were not incorporated into the full proteasome, however, implies that these mutations affect the correct docking of the two half-proteasomes. In addition, it is noticeable that the degree of propeptide degradation is different between in vivo and in vitro assemblies. Although all the intermediates analyzed here for the in vivo assembly are degraded up to the first 18 residues (C-terminus to Ser-48) ( Fig. 6 and Ref. 25), in vitro, the propeptide is more heavily truncated, and degradation can proceed up to the first 50 residues (C-terminus to Thr-16). This observation may imply that in vivo the ␣and ␤-subunits are expressed, folded, and assembled in concert, thus preventing substantial degradation of the propeptide.
In summary, the study presented here reveals many of the different steps along the assembly pathway of the 20 S proteasome. The unique ability of MS to identify in great detail the modifications of individual ␤-subunits, differing by only one residue, has enabled us to monitor those that are trapped within the various intermediate states. We are therefore able to identify specific cleavage sites along the propeptide and to demonstrate that in vivo, the propeptide is significantly more protected from degradation than in vitro. More generally, this study highlights the tremendous potential for using real-time mass spectral acquisition and the phase transition from solution to gas phase to capture transient intermediates along assembly pathways of macromolecular complexes.