Analysis of Drosophila 26 S Proteasome Using RNA Interference*

We have utilized double-stranded RNA interference (RNAi) to examine the effects of reduced expression of individual subunits of the 26 S proteasome in Drosophila S2 cells. RNAi significantly decreased mRNA and protein levels of targeted subunits of both the core 20 S proteasome and the PA700 regulatory complex. Cells deficient in any of several 26 S proteasome subunits ( e.g. d (cid:1) 5, dRpt1, dRpt2, dRpt5, dRpn2, and dRpn12) displayed decreased proteasome activity (as judged by hydrolysis of succinyl-Leu-Leu-Val-Tyr-aminomethyl-coumarin), increased apoptosis, decreased cell proliferation without a specific block of the cell cycle, and accumulation of ubiquitinated cellular proteins. RNAi of many individual 26 S proteasome subunits promoted increased expression of many non-targeted subunits. This effect was not mimicked by chemical proteasome inhibitors such as lactacystin. Reduced expression of most targeted subunits disrupted the assembly of the 26 S proteasome. RNAi of six of eight targeted PA700 subunits disrupted that structure and caused accumulation of increased levels of uncapped 20 S proteasome. Notable exceptions included RNAi of dRpn10, a polyubiquitin binding subunit, and dUCH37, a ubiquitin isopeptidase. dRpn10-deficient cells showed a significant increase in succinyl-Leu-Leu-Val-Tyr-aminomethylcou-marin

Most intracellular protein degradation in higher eukaryotes is catalyzed by 26 S proteasomes (1). With few known exceptions, proteins degraded by the 26 S proteasome must be covalently modified by a polyubiquitin chain (2,3). Ubiquitin is attached to the substrate protein via an isopeptide bond between the C-terminal carboxyl group of ubiquitin and an ⑀-amino group of a particular lysine of the substrate protein.
Ubiquitin monomers within the polyubiquitin chain are likewise linked by isopeptide bonds between the C-terminal carboxyl group and the ⑀-amino group of Lys-48 of the preceding ubiquitin (4 -7). Polyubiquitin chains of this structure bind specifically to the 26 S proteasome, which then degrades the protein and releases free ubiquitin monomers.
The 26 S proteasome is a 2,100,000-dalton protease complex composed of ϳ64 subunits representing at least 32 different gene products. It is assembled from two multisubunit subcomplexes, the 20 S proteasome and PA700 (19 S complex). The 20 S proteasome is a 700,000-dalton protease composed of 28 subunits arranged as a cylinder-shaped stack of four heptameric rings. Both exterior rings contain one copy each of seven different but homologous ␣-type subunits. Likewise, both interior rings contain one copy each of seven different but homologous ␤-type subunits (2,8,9). The crystal structure of yeast 20 S proteasome reveals that the complex contains three contiguous interior chambers sealed from the exterior by a polypeptide shell (9,10). The central chamber, formed by the two ␤-rings, is lined by two copies each of three catalytic centers with differing specificities for peptide bond hydrolysis (11). Thus, access of protein substrates to the catalytic sites of the proteasome is normally highly restricted.
PA700 is a 700,000-dalton complex composed of 16 -18 distinct gene products (12). PA700 binds to one or both of the outer ␣-rings of the 20 S proteasome and promotes protease activation. Binding of PA700 induces a conformational change in proteasome structure that breaks a seal formed by N-terminal portions of four subunits in each outer ring. This conformational change, roughly analogous to the opening of a ligandgated channel, creates an entry point through which substrates can pass to reach the catalytic sites in the interior of the proteasome (13,14). The opened pore has a diameter of about 13 Å, which allows unimpeded passage of short peptides, whose hydrolysis is greatly accelerated when PA700 binds to the proteasome. A pore of 13 Å, however, is too small to allow passage of folded proteins known to be degraded by the 26 S proteasome (9,10). Therefore, in addition to altering proteasome structure by conformational activation, PA700 probably also alters protein substrate structure by promoting substrate unfolding (15)(16)(17). Finally, PA700 may be involved in the processive translocation of the unfolded polypeptide substrate through the opened proteasome pore and into the central chamber for peptide bond hydrolysis. Protein substrate unfolding and/or translocation may be linked to ATP hydrolysis, an enzymatic activity inherent to PA700 and obligatory for the degradation of proteins by the 26 S proteasome (17,18).
Unlike the 20 S proteasome, whose structure is understood in considerable detail, less is known about the detailed structure of PA700. Nevertheless, the relative arrangement of many subunits within the complex has been mapped. For example, six subunits are distinct but homologous members of the AAA ATPase family (2,3). They probably form a hexameric ring that binds directly to the outer rings of the proteasome and, with 2-3 additional subunits, form a subcomplex of PA700 termed the "base" (19). These subunits mediate ATP-dependent functions of the complex and appear sufficient to promote conformational opening of the substrate entrance pore described above (13,14). A second group of approximately eight subunits, termed the "lid," constitute a region of PA700 distal to the base. In yeast, the base and the lid appear to be physically linked by the Rpn10 subunit, which like its mammalian ortholog, S5a, binds to polyubiquitin chains (20). However, yeast that lack Rpn10 are viable and degrade most ubiquitinated proteins normally (19,21,22). These data indicate that other as yet unidentified subunits of 26 S proteasome also must participate in the selective recognition of polyubiquitinated proteins.
Despite the impressive progress in understanding the structure and function of the 26 S proteasome, relatively little is known about the functions of and interactions among individual 26 S subunits, especially within PA700. The use of conventional "knock out" technologies to address these issues has been limited by the large number of subunits to be studied as well as the fact that most of the subunits are essential for growth and development.
In this report we have used double-stranded RNA interference (RNAi) 1 to block expression of individual subunits of 26 S proteasome in Drosophila S2 cells. Unlike conventional methods, RNAi is simple, rapid, and amenable for the study of reduced expression of even those proteins essential for growth. RNAi is based on the process whereby specific dsRNAs induce a potent and specific interference in protein expression or posttranscriptional gene silencing in some invertebrates, where it functions as an antiviral/antitransposomal defense mechanism. This mechanism was originally exploited for experimental use in Caenorhabditis elegans (23) and more recently in cultured Drosophila cells (24,25). Because Drosophila 26 S proteasomes are well characterized (26 -28), we have applied RNAi to examine the relative roles of individual subunits in cultured S2 cells.

EXPERIMENTAL PROCEDURES
Nomenclature for 26 S Proteasome Subunits-The core 20 S proteasome subunits are designated by the ␣/␤ nomenclature, and the PA700 subunits are designated by the Rpn/Rpt nomenclature. The prefix "d" is used to indicate the respective Drosophila orthologs (29).
Chemical Reagents and Antibodies-Polyclonal antibodies against individual subunits of the human 26 S proteasome were prepared in rabbits. Synthetic peptides corresponding to unique sequences of the subunits were synthesized by standard solid phases methods and are as follows: Rpt2, VLYKKQEGTPEGYLY; Rpt5, EGILEVQAKKKA-NLQYYA; Rpt6, KVMQKDSEKNMSIKKLWK; Rpn12, TELAKQVIEYA-RQLEMIV. Peptides were conjugated to keyhole limpet hemocyanin as used as antigens for antisera production as described previously (30). All antisera were demonstrated to specifically recognize the corresponding subunits of purified bovine 26 S proteasome (not shown). The antiserum against Rpn12 reacted with two closely migrating bands of purified 26 S proteasome and purified PA700, indicating the existence of two closely related forms of this subunit. Preliminary experiments demonstrated that these antibodies cross-reacted with corresponding 26 S proteasome subunits from numerous species including Drosophila. Monoclonal antibodies MCP236 and MCP72 recognizing both human and Drosophila ␣2(C3) and ␣7(C8) proteasome subunits, respectively, were a generous gift of Dr. K. Hendil (Denmark, Copenhagen) (26,31). Ubiquitin antibody was purchased from Sigma.
Horseradish peroxidase-conjugated sheep anti-mouse whole IgG was from Amersham Biosciences. Horseradish peroxidase goat anti-rabbit whole IgG was from American Qualex.
Reverse Transcriptase-PCR-S2 cells were lysed in Trizol reagent (Invitrogen), and total RNA was extracted according to the manufacturer's instructions by a modified method of Chomczynski and Sacchi (32). RNA content was calculated by measuring A 260 . RT-PCR was carried out using OneStep RT-PCR kit (Qiagen) according to the manufacturer's instructions.
In order to design the primers, sequences of human proteins of interest were found at the NCBI web site. Protein sequences were used for a PSI-BLAST search (33) against the whole Drosophila genome (34). The PSI-BLAST hit with the highest score, and the lowest E value was chosen for the design of the primers. This approach did not yield consistent results only in the case of dCSN1, and the Drosophila CSN1 gene was found by direct data base search. At the 5Ј ends of the primers, the T7 RNA polymerase-binding site (TTA ATA CGA CTC ACT ATA GGG AGA) was added. All primers were purchased from Operon Technologies. The primers used are presented Table I (without the T7 sequence shown). After 28 cycles (annealing temperature of 58°C), products were assessed by agarose gel electrophoresis and purified by the use of QIAquick PCR Purification Kit (Qiagen) according to manufacturers' instructions. The product was quantified by measurement of the A 260 nm and adjusted to a concentration of 125 ng/liters.
dsRNA Synthesis-MEGAscript kit (Ambion) was used for dsRNA synthesis according to Dixon lab protocol (24). 1 g of linearized DNA template in 8 l of nuclease-free water was supplemented with 2 l of 10ϫ reaction buffer, 2 l of T7 enzyme mix, and ATP, CTP, GTP, UTP mixtures. They were gently mixed, centrifuged, and incubated at 37°C for 5 h. The reaction products were precipitated by ethanol, resuspended in RNase-free water, heated for 30 min at 65°C, and slowly cooled to room temperature to anneal the RNA. The quality of dsRNA was assessed by agarose electrophoresis, and the final concentration of dsRNA was adjusted to 3 g/l. dsRNA was stored at Ϫ20°C until use.
dsRNA Interference-S2 cells were collected by centrifugation, resus- 1 The abbreviations used are: RNAi, RNA interference; AMC, aminomethylcoumarin; Z benzyloxycarbonyl; RT, reverse transcriptase; dsRNA, double-stranded RNA; PBS, phosphate-buffered saline; MTT, 3-(4,5-dimethylthiazol-2-yl)-diphenyltetrazolium bromide; Suc, succinyl. 2 A. Kisselev and A. Goldberg, manuscript in preparation. AGA TGG CAT CAA TCT GGT CTG AAC ACC AGG CAG ATT CTC CA pended in the Drosophila expression system (DES) serum-free expression medium (Invitrogen), and plated into 6-or 24-well plastic plates at a concentration of 1 ϫ 10 6 cells/1 ml. Immediately after 15 g of dsRNA was added (5 l of 3 g/l stock), and the plates were gently swirled to mix RNA and cells and incubated at room temperature for 1 h. After the incubation, 2 ml of complete Schneider medium was added, and the plates were returned to the incubator for subsequent incubation.

SDS-PAGE and Western
Blotting-PBS-washed S2 cells were lysed for 30 min in RIPA buffer supplemented with Complete Mini protease inhibitor mixture (Roche Molecular Biochemicals). After determination of protein concentration (35), 5ϫ concentrated Laemmli sample buffer was added (36). Samples were subjected to SDS-PAGE, transferred to nitrocellulose, and blotted with respective antibodies using standard methods (see figure legends for details of individual experiments). Detection was achieved using ECL TM Western blotting reagents, and they were recorded on chemiluminescence Hyperfilm TM ECL TM (Amersham Biosciences).
Measurement of Cell Viability by the MTT Assay-S2 cells were plated in 24-well plates, and the dsRNA treatment was carried out as described above. Each dsRNA was applied to three different wells. After 4 days, 100 l of a freshly made 3-(4,5-dimethylthiazol-2-yl)-diphenyltetrazolium bromide (MTT) solution (2.5 mg/ml PBS) was added to each well (37). After incubation for 4 h, 500 ml of a lysis solution (11% SDS, 50% v/v isopropyl alcohol in water) was added, and the plates were shaken for 3 h. Reaction product was measured at A 570 and corrected against blanks, which consisted of media processed in the absence of cells. Reported values represent the means of three independent measurements. The relative viability of S2 cells was calculated as follows: relative viability ϭ (A e ) ϫ 100/(A c ), where A e is experimental absorbance, and A c is the absorbance of untreated controls.
Measurement of Proteasome Activity in Cell Lysates-Proteasome activity was assessed in lysates of S2 cells using synthetic peptide substrates linked to the fluorimetric reporter, aminomethylcoumarin (AMC). dsRNA-treated S2 cells cultured in 24-well plates were collected, washed in PBS, and lysed in Buffer H (20 mM Tris, 20 mM NaCl, 1 mM EDTA, 5 mM ␤-mercaptoethanol, pH 7.6) for 30 min on ice. Cell lysates were cleared by centrifugation, and the supernatants were used for determination of protein concentration and enzymatic activity. Lysates (25 l) were assayed by addition of 250 l of 50 M solution of indicated substrates (in 50 mM Tris-HCl, pH 8.0, and 1 mM ␤-mercaptoethanol) and incubation for 30 min at 37°C. AMC hydrolyzed from the peptides was quantitated in a BioTek FL600 plate reader using 360 nm excitation and 460 nm emission wavelengths. Enzymatic activity was normalized for protein concentration and expressed as percent of activity in control lysates. Each measurement was carried out using at least three independent dsRNA experiments.
Flow Cytometry-For flow cytometric analysis of the cell cycle, dsRNA-treated and control S2 cells grown on 24-well plates were collected, washed in PBS, resuspended in 1 ml of PBS, and fixed in 10 ml of ice-cold 75% ethanol. After 24 h at Ϫ20°C, the cells were washed twice in PBS and resuspended in 300 ml of PBS supplemented with 0.1% Nonidet P-40 and RNase A (10 mg/ml). After 30 min of incubation at room temperature, the cells were stained overnight with propidium iodide (5 mg/ml). Apoptosis was analyzed using the annexin V-fluorescein isothiocyanate apoptosis detection kit (Sigma) (38). dsRNA-treated and control S2 cells grown on 24-well plates were collected, washed twice in PBS, and resuspended in binding buffer (10 mM HEPES/NaOH, pH 7.5, 140 mM NaCl, 2.5 mM CaCl 2 ). Annexin V-fluorescein isothiocyanate conjugate (1 g/ml) and propidium iodide (2 g/ml) were added to the cells in suspension exactly 10 min before collecting the data. Fluorescence was measured using a FACSscan flow cytometer (Becton Dickinson). Data were collected and analyzed using CellQuest software.
Glycerol Density Gradient Centrifugation-Control and dsRNA-treated S2 cells were collected, washed in PBS, and suspended in 300 l of 50 mM Tris buffer, pH 7.6, on ice. After a 10-min incubation, cell were disrupted by 10 strokes in a Dounce homogenizer on ice. The cell lysate was centrifuged at 16,000 ϫ g for 15 min at 4°C. Two hundred l of the supernatant were layered on top of 1.55-ml glycerol gradients (10 -40% glycerol in 50 mM Tris-HCl, pH 7.6, at 4°C, 1 mM dithiothreitol, 1 mM ATP, 5 mM MgCl 2 ) and centrifuged at 200,000 ϫ g for 3 h at 4°C in Beckman OptimaTL Ultracentrifuge using a TLS55 rotor. Fractions of 100 l were collected and assayed for Suc-Leu-Leu-Val-Tyr-AMC hydrolyzing activity, as described above, or subjected to Western blotting with appropriate antibodies. Calibration of the gradients was accomplished by centrifugation of purified samples of PA28, 20 S proteasome, and PA700 from bovine red blood cells or 26 S proteasome from rabbit skeletal muscle (18).

RNAi Decreases Expression of Targeted Subunits of the 26 S Proteasome in Drosophila S2
Cells-We selected nine representative subunits of the 26 S proteasome as targets for RNAi. These subunits, summarized in Table II, include the following: one catalytic subunit of the 20 S proteasome; four ATPases and one non-ATPase of the base portion of PA700; two subunits of the lid portion of PA700; and dRpn10, a polyubiquitin chain binding subunit that in yeast functions as a structural link between the base and the lid. To determine whether expression of individual subunits of the 26 S proteasome could be blocked by RNAi, we treated Drosophila S2 cells with dsRNAs corresponding to six selected subunits. The mRNA level of each subunit was assessed after 4 days using semi-quantitative RT-PCR. Each dsRNA greatly reduced the mRNA level of its respective targeted subunit (Fig. 1). Although none of these dsRNAs influenced mRNA levels of non-26 S proteasome subunits (e.g. dCSN1, Fig. 1; cathepsin D and TPPII, data not shown), most of the dsRNAs significantly increased the mRNA levels of the non-targeted 26 S proteasome subunits. For example, RNAi of dRpt2 greatly reduced the level of dRpt2 mRNA, but significantly increased mRNA levels of d␤5, dRpt1, dRpt6, dRpt10, and dUCH37. To establish whether changes in mRNA levels of 26 S proteasome subunits resulted in corresponding changes in levels of encoded proteins, we performed Western blotting on lysates of S2 cells treated with dsRNAs to subunits for which appropriate antibodies were available. Fig. 2 shows that the protein level of each targeted subunit was significantly and in some cases almost completely reduced. Furthermore, consistent with the alterations of mRNA levels, reduced expression of a given 26 S proteasome subunit promoted increased expression of many non-targeted proteasome subunits. For example, RNAi of dRpt2 reduced dRpt2 protein to undetectable levels, but significantly increased the levels of dRpt6, dRpn12, d␣7, and d␣2. Both the reduction of the targeted subunit and the corresponding increases in non-targeted subunits were specific for components of the 26 S proteasome, because the expression of an irrelevant protein, such as actin, was not affected (Fig. 2). These results establish that levels of individual subunits of the 26 S proteasome could be targeted for reduction by RNAi but that this procedure resulted in increased expression of many, but not all, non-targeted subunits. The latter effect may not be linked strictly to a decline in proteasome activity because proteasome inhibition with lactacystin had no effect on 26 S proteasome subunit levels. A direct comparison of the relative effects of RNAi and lactacystin may not be possible, however, due to the different time courses of the experiments (see "Discussion").

RNAi of Most Subunits of the 26 S Proteasome Inhibits Cell
Proliferation-To study the functional consequences of reduced levels of individual subunits of the 26 S proteasome, we assessed the effects of RNAi on some known functions of the proteasome. First, we examined the effect on cell proliferation. Untreated S2 cells proliferated rapidly as determined by microscopic inspection and by the MTT assay, an index of cell number based on the measurement of the metabolic activity of viable cells (37). After 3-4 days in culture, proliferation of control cells slowed, probably due to high cell density (Fig. 3A). Cells treated with dsRNA for three representative proteasome subunits, d␤5, dRpt2, and dRpn10, proliferated at rates similar to those of untreated cells for up to 2 days, but then their growth rate and cell number decreased compared with untreated controls. These results are consistent with the known requirement of proteasome function for cell proliferation (39) but differed temporally compared with results of lactacystin treatment, which completely inhibited cell growth within 24 h. When cells treated with dsRNA for 4 days were returned to fresh media, they resumed growth and eventually restored levels of the targeted subunit proteins to control levels (data not shown). This effect is consistent with the transient nature of RNAi reported previously (24). Based on these results, we analyzed the effects of RNAi of nine selected 26 S proteasome subunits on cell growth after 4 days. Reduced expression of most but not all subunits significantly decreased cell growth. In contrast, RNAi of the dCSN1 protein, a non-26 S proteasome subunit, had no effect on growth (Fig. 3B). Thus, these effects do not appear to result from nonspecific toxicity of dsRNA.
Although the MTT assay is an index of the number of viable cells, it does not distinguish between processes that affect cell viability via cytostatic as opposed to cytotoxic mechanisms (37). Therefore, we used flow cytometry and annexin V-propidium iodide staining to determine whether RNAi caused blocks in the cell cycle and/or apoptosis of existing S2 cells (38). RNAiinduced decreases in the number of viable cells were not accompanied by changes in cell cycle distribution. Proteasome inhibitors are known to induce a cell cycle block either at the G 1 /S transition (40), at the G 2 /M transition (41), or at both (42), depending on the cell line and experimental conditions. A 24-h treatment with 5 M lactacystin caused in S2 cells a block in the cell cycle at the G 1 /S transition (data not shown).
RNAi of most PA700 subunits significantly increased the number of apoptotic cells (Table III), without affecting the percentage of necrotic cells, an effect also promoted by lactacystin and other proteasome inhibitors (43)(44)(45)(46). These results suggest that RNAi of most 26 S proteasome subunits slowed progression of S2 cells through the cell cycle without specifically blocking a particular phase. They also support the conclusion that dsRNAs targeted for 26 S proteasome subunits do not promote nonspecific toxic effects, but probably exert their effects via the inhibited functions of the 26 S proteasome.
To extend these studies we tested whether cells treated with dsRNA targeting 26 S proteasome subunits remained sensitive to the cytostatic/cytotoxic effects of proteasome inhibitors such as lactacystin, epoxomicin, and PS341. Fig. 3C shows the results for the RNAi of three representative proteasome subunits, d␤5, dRpt2, and dRpn10. Each inhibitor induced strong cytotoxic/cytostatic effects within 24 h of treatment, which were magnified after 48 h. Cells treated with dsRNA for dRpn10 and dRpt6 (not shown) for 72 h before exposure to inhibitors had the same sensitivity to proteasome inhibitors as control cells. In contrast, cells treated with dsRNA for d␤5 for 72 h, and to a lesser extent cells treated with dsRNA for dRpt2 for 72 h, were significantly more resistant to the cytotoxic/ cytostatic effects of proteasome inhibitors. RNAi of other 20 S proteasome subunits produced results similar to d␤5, whereas most PA700 subunits (with the exception of dRpt6) produced results similar to dRpt2 (not shown). These results suggest that S2 cells with severely impaired 26 S proteasome function due to the RNAi suppression of crucial subunits adapt proteasomeindependent compensatory mechanisms for survival.
RNAi of Most 26 S Proteasome Subunits Inhibits Proteasome Activity-To examine more specifically the functional consequences of RNAi of 26 S proteasome subunits, we measured proteasome activity in extracts of control and dsRNA-treated cells (Fig. 4). RNAi of most of the nine targeted subunits significantly reduced the hydrolysis of Suc-Leu-Leu-Val-Tyr-AMC and Ac-Gly-Pro-Leu-Asp-AMC, synthetic peptide substrates of the chymotryptic-like and peptidyl-glutamyl activities of the proteasome, respectively. The magnitude of reduced activity for most targeted subunits was comparable with or slightly less than that achieved by lactacystin. RNAi of several subunits, most notably dUCH37, had no detectable effect on proteasome activity. In contrast, RNAi of dRpn10 significantly increased the activity against each substrate. Little effect of RNAi of most subunits was observed using Z-Val-Val-Arg-AMC, a substrate hydrolyzed by the trypsin-like site of the proteasome. Unlike the other substrates, however, this peptide is hydrolyzed appreciably by non-proteasome proteases in crude cell extracts (47). Therefore, inhibitory effects of RNAi on this proteasome activity might be masked. Nevertheless, a significant inhibition of hydrolysis of this substrate was achieved with RNAi for the d␤5 subunit, and the magnitude of the effect was similar to that achieved with lactacystin. In sum, these results demonstrate that RNAi-promoted reduction of most 26 S proteasome subunits significantly decreased proteasome activity.
RNAi of Most 26 S Proteasome Subunits Interferes with 26 S Complex Assembly-The 26 S proteasome is formed by the ATP-dependent assembly of two multisubunit subcomplexes, the 20 S proteasome and PA700 (19 S complex). We analyzed the effect of RNAi of selected subunits of each subcomplex on the integrity and activity of the 26 S proteasome by glycerol density gradient centrifugation (Figs. 5 and 6). Purified 26 S and 20 S proteasomes, and purified PA700, sediment to distinctly different and characteristic positions in these gradients (fractions 15-16, fractions 11-13, and fractions 13-14, respectively), as determined by activity assays and Western blotting of gradient fractions (Ref. 48 and data not shown). Extracts from control S2 cells displayed a peak of proteasome activity at a sedimentation position indistinguishable from that of purified 26 S proteasome (Fig. 5, control). This activity was coincident with the distribution profile of component subunits of the 26 S proteasome from both the 20 S proteasome and PA700 (Fig. 6, control). In addition, a significant portion of 20 S proteasome subunits sedimented at a position characteristic of "free" 20 S proteasome ( Fig. 6 and data not shown). These fractions displayed low proteasome activity, consistent with the latent nature of 20 S proteasome in the absence of bound PA700. These results indicate that extracts from untreated S2 cells contain both 26 S and 20 S proteasomes and that most proteasome activity in these extracts is accounted for by the 26 S proteasome. Interestingly, significant portions of some subunits of PA700, including the ATPases dRpt2 and dRpt5, sedimented much more slowly than they did as components of free PA700 or 26 S proteasome. The significance of these distribution profiles is unclear, although similar results have been Glycerol density gradient centrifugation also was performed on extracts of S2 cells treated with dsRNA for the nine selected 26 S proteasome subunits. Because this entire analysis could not be conducted in a single experiment, data for different experiments are not always precisely comparable with one another due to minor experiment-to-experiment variability inherent in the centrifugation and assay methods. Nevertheless, the results shown in Figs. 5 and 6 permit clear general conclusions regarding the effects of RNAi of these selected subunits on 26 S proteasome structure and function. For example, RNAi of d␤5, a catalytic subunit of the 20 S proteasome, greatly reduced proteasome activity in the gradient fractions, consistent with the effect observed in crude extracts (Fig. 5, d␤5). Residual proteasome activity was present in the form of 26 S proteasome. RNAi of d␤5 redistributed other 20 S proteasome subunits to low molecular weight complexes that probably represent unassembled intermediates of the 20 S proteasome. For example, d␣7 (Fig. 6, d␤5) and d␣2 (data not shown) sedimented slowly at positions that may represent isolated ␣-rings that are normal intermediates of 20 S proteasome assembly (50). These results indicate that decreased d␤5 expression significantly disrupts normal 20 S proteasome assembly. RNAi of d␤5 also significantly redistributed PA700 to a position indicative of the unbound complex. Thus, reduced assembly of 20 S proteasome increases the amount of PA700 not associated with the proteasome.
RNAi of six different subunits of PA700, including four AT-Pases of the base (dRpt1, dRpt2, dRpt5, and dRpt6), a non-ATPase of the base (dRpn2), and a subunit of the lid (dRpn12), produced generally similar patterns of proteasome distribution (Figs. 5 and 6). As expected from assays of crude exacts, RNAi of each subunit reduced activity recovered in the gradient fractions relative to controls. The residual activity displayed a bimodal distribution. One peak of activity remained in the sedimentation position characteristic of 26 S proteasome, whereas a second peak emerged at a position characteristic of 20 S proteasome. The large increase in this latter activity relative to controls was accompanied by a large increase in the amount of 20 S subunits detected in corresponding fractions. Thus, RNAi of these PA700 subunits appears to disrupt assembly of the 26 S proteasome, resulting in increased levels of free 20 S proteasome. This increase could result from either a greater proportion of uncapped preexisting 20 S proteasomes and/or increased expression of 20 S proteasomes as a consequence of RNAi. The effect of RNAi of most base subunits examined here also greatly disrupted PA700 structure, as judged by the altered distribution of selected PA700 subunits. These results indicate that RNAi of many PA700 subunits decreased proteasome activity by impeding assembly of 26 S proteasome.
We next analyzed RNAi of dRpn10, a subunit shown previously to bind polyubiquitin chains and to serve as a structural link between the base and lid regions of yeast PA700 (19,20). Gradient fractions of extracts of cells subjected to RNAi for dRpn10 displayed the same increased proteasome activity observed in crude extracts. Surprisingly, this increased activity was accounted for by the 26 S proteasome. Western blotting for several different PA700 subunits, including components of both lid and base, showed that nearly all subunits cosedimented in the region of the gradient characteristic of 26 S proteasome. Similar results were obtained with altered gradient conditions (e.g. 100 mM NaCl and lack of ATP, data not shown). Thus, deletion of dRpn10 did not appear to promote disruption of PA700 structure within the 26 S proteasome under the condition of these gradients.
Finally, we analyzed RNAi of dUCH37, a subunit previously shown to function as an ubiquitin isopeptidase (28,51,52). RNAi of dUCH37, which had no effect on proteasome activity in crude extracts, also had no detectable effect on proteasome activity in gradient fractions (Fig. 5, dUCH37). Thus, deficiency of dUCH37 has little apparent effect on the structure or peptidase activities of the 26 S proteasome (see "Discussion").

RNAi of Most 26 S Proteasome Subunits Promotes Accumulation of Polyubiquitinated Cellular
Proteins-Most physiological substrates of the 26 S proteasome are cellular proteins covalently modified with a polyubiquitin chain (2). To determine the effect of RNAi of 26 S proteasome subunits on the degradation of such proteins, we blotted extracts of control and dsRNA-treated S2 cells with an anti-ubiquitin antibody (Fig.  7). Control cells displayed a smear of high molecular weight immunoreactive material characteristic of, and expected for, the many cellular proteins modified by polyubiquitin chains. Consistent with previous reports (43,53), inhibition of proteasome activity by proteasome inhibitors, such as lactacystin, increased the intensity of this smear and is diagnostic of accumulation of non-degraded cellular proteins. RNAi of most 26 S proteasome subunits also greatly increased the level of polyubiquitinated S2 cell proteins. This effect was particularly prominent for RNAi of the catalytic subunit, d␤5, three of the four tested base ATPases, dRpt 1, dRpt2, dRpt5, a base non-AT-Pase, dRpn2, and a lid subunit, dRpn12. Little or no effect was detected for RNAi of the base ATPase, dRpt6, or dUCH37. Thus, these results generally mirrored corresponding effects of RNAi on proteasome activity against peptide substrates. In surprising contrast, RNAi of dRpn10, which increased proteasome activity against peptide substrates, produced significant accumulation of ubiquitinated proteins. These results indicate that this subunit is required for normal degradation of a population of S2 cell proteins. DISCUSSION We have utilized RNAi to examine the relative roles of selected subunits of the 26 S proteasome with respect to structure and function of this large protease complex. RNAi significantly and selectively blocked expression of targeted subunits of the 26 S proteasome in S2 cells as judged at the mRNA and protein levels. In some cases, inhibition of expression was probably much greater than was apparent from Western analysis of protein levels because of the relatively long half-life of most proteasome components. Nevertheless, RNAi of proteasome subunits produced significant phenotypic changes in S2 cells, including reduced growth, increased apoptosis, reduced proteasome function, and increased cellular levels of polyubiquitinated proteins, an indicator of inhibition of intracellular protein degradation via the ubiquitin-dependent pathway (43,53,54).
RNAi of most, but not all, 26 S proteasome subunits interfered with the structure of the complex. RNAi of subunits of the 20 S proteasome reduced the amount of 26 S proteasome by blocking assembly of the 20 S core particle. In contrast, RNAi of most PA700 subunits disrupted PA700 and promoted an increase in the cellular content of free 20 S proteasome. Notable exceptions to the latter effects included RNAi of dRpn10 and dUCH37.
Rpn10 is a polyubiquitin chain binding subunit of mammalian, yeast, and plant 26 S proteasomes (20). Deletion of Rpn10 from Saccharomyces cerevisiae has no effect on cell viability but promotes increased steady state levels of ubiquitin conjugates (21). Accordingly, RNAi of dRpn10 in S2 cells had only modest effects on cell growth but significantly increased accumulation of ubiquitinated S2 cell proteins. In surprising contrast, we have observed an increase in the proteasome-specific peptidehydrolyzing activity measured in whole cell lysates. This unexpected effect could occur from either increased levels of 26 S proteasomes (perhaps as a consequence of the increased expression and assembly of dRpn10-deficient 26 S proteasomes) or from increased activity of the dRpn10-deficient 26 S proteasomes. Support for the latter possibility was obtained by normalizing proteasome activity in the glycerol gradients to the amount of 26 S proteasomes in corresponding fractions (as determined by quantitative Western blotting). These results indicate that dRpn10-deficient 26 S proteasomes display ϳ5fold higher specific activity compared with controls (data not shown). Nevertheless, an unambiguous conclusion about this issue cannot be made until the purified complexes are compared directly.
dUCH37 is an isopeptidase that removes ubiquitin monomers from the distal end of polyubiquitin chains (51,52). It is a bona fide 26 S proteasome subunit in mammals, Drosophila, and Schizosaccharomyces pombe but is not present in S. cerevisiae (28,52,55). Electron microscopy suggests that dUCH37 may be located on the periphery of the lid portion of PA700 in Drosophila (28). A peripheral location could explain why its deletion has no effect on proteasome activity or structure. In addition, its lack of effect on growth or accumulation of polyubiquitinated proteins suggests that it does not play an essential role in the cellular function of the proteasome and that its activity can be compensated for by other isopeptidases of S2 cells.
Reduced expression of many individual subunits of the 26 S proteasome by RNAi resulted in increased expression of most other subunits both at the mRNA and protein levels. This phenomenon is similar to that reported previously in yeast, where Rpn4, a protein that associates loosely with the 26 S proteasome, is itself a rapidly degraded proteasome substrate (56). Rpn4 functions as a transcription factor for many subunits of the proteasome as well as for other components of the ubiquitin pathway (57,58). Thus, inhibition of proteasome function increases Rpn4 levels and leads to an increased production of proteasome complexes. Interestingly, the Drosophila genome does not contain a protein with significant overall sequence similarity to Rpn4 (34), although it may contain an as yet unidentified protein with an equivalent function. In any case, treatment of S2 cells with lactacystin did not promote increased expression of 26 S proteasome subunits, suggesting that inhibition of proteasome activity per se does not account for this response. Thus, the molecular mechanisms involved in the up-regulation of proteasome subunits as a consequence of RNAi remain to be defined. The fate of up-regulated subunits within cells appears to vary from subunit to subunit. For example, the increased 20 S proteasome subunits produced in response to RNAi of PA700 subunits appeared to assemble into functional 20 S proteasomes. In contrast, the increased 20 S proteasome subunits (e.g. d␣7) produced in response to RNAi of another 20 S proteasome subunit, d␤5, accumulated as low molecular weight complexes that probably correspond to intermediates of 20 S proteasome assembly (50). The fate of overexpressed PA700 subunits is less clear, although some appear to assemble into heteromultimeric complexes. For example, in response to RNAi of dRpt2, dRpn12 is detected in both the 26 S proteasome and in a low molecular weight complex that sediments more slowly than PA700. Obviously, a comprehensive analysis of the precise cellular fate of all subunits expressed during RNAi for any other subunit will require considerable additional work. dUCH37 is the only examined 26 S proteasome subunit whose targeting by RNAi did not induce upregulation of at least some other subunits. This is consistent with the lack of effect of this target on other parameters of proteasome structure and function.
Although RNAi of many 26 S proteasome subunits examined here had qualitatively similar effects to one another on multiple parameters, the relative magnitudes of these effects differed among the subunits. Such differences could result from different physiological roles of these subunits, but might also reflect differences in the magnitude of reduced expression of different subunits by RNAi. Combined with the functional contribution of residual proteasome activity due to the long halflife of most proteasome components, it is likely that even small differences in proteasome content could account for appreciable differences in given effects. For example, our failure to detect defined blocks in the cell cycle with RNAi may result from residual proteasome activity sufficient to promote cell cycle progression. The observed growth suppression may result from a nonspecific slowing of cell cycle progression in all phases combined with induction of apoptosis. In contrast, lactacystin inhibits proteasome activity rapidly and more completely than RNAi of most subunits, resulting in a clear block of the cell cycle and rapid induction of apoptosis.
Despite its general impairment of the degradation of ubiquitin conjugates, RNAi of 26 S proteasome subunits does not completely prevent cell survival and proliferation. Therefore, compensatory mechanism(s) may be activated during RNAi, as indicated by increased resistance of dsRNA-treated cells to proteasome inhibitors. Preliminary experiments have failed to indicate changes in lysosomal function or in the activity of tripeptidylpeptidase II. 3 The latter protease has been reported to be up-regulated in a leukemic cell line, adapted to grow in the presence of a proteasome inhibitor (59,60).