Gly-Ala Repeats Induce Position- and Substrate-specific Regulation of 26 S Proteasome-dependent Partial Processing*

Partial degradation or regulated ubiquitin proteasome-dependent processing by the 26 S proteasome has been demonstrated, but the underlying molecular mechanisms and the prevalence of this phenomenon remain obscure. Here we show that the Gly-Ala repeat (GAr) sequence of EBNA1 affects processing of substrates via the ubiquitin-dependent degradation pathway in a substrate- and position-specific fashion. GAr-mediated increase in stability of proteins targeted for degradation via the 26 S proteasome was associated with a fraction of the substrates being partially processed and the release of the free GAr. The GAr did not cause a problem for the proteolytic activity of the proteasome, and its fusion to the N terminus of p53 resulted in an increase in the rate of degradation of the entire chimera. Interestingly the GAr had little effect on the stability of EBNA1 protein itself, and targeting EBNA1 for 26 S proteasome-dependent degradation led to its complete degradation. Taken together, our data suggest a model in which the GAr prevents degradation or promotes endoproteolytic processing of substrates targeted for the 26 S proteasome by interfering with the initiation step of substrate unfolding. These results will help to further understand the underlying mechanisms for partial proteasome-dependent degradation.

Polyubiquitinated substrates are targeted for 26 S-dependent proteasomal degradation by their ubiquitin moiety that is recognized and bound to one of the 19 S regulatory complexes. Prior to degradation, the substrate has to be unfolded by an ATP-dependent mechanism in the 19 S cap structure. This allows the polypeptide to be translocated and threaded into the central chamber of the 20 S complex where three different types of proteolytic activity catalyze the cleavage of peptide bonds, leading to the complete breakdown of the substrate into small peptide fragments (1,2).
In addition to this classic model of 26 S proteasome-mediated degradation, the proteasome can also partially degrade specific substrates leading instead to the release of larger degradation products with distinct cell biological activity, the socalled regulated ubiquitin proteasome-dependent processing (3)(4)(5)(6). For example, the p50 subunit of NF-B is generated by the processing of the larger p105 protein (7,8), and it has been suggested that a glycine-rich region in p105 is responsible and necessary for this partial degradation (9,10). Partial processing also applies to other transcription factors like the yeast NF-B homologs Spt23 and Mga2 and the Drosophila Ci (3,4,11). The direct targeting of proteins carrying polyglutamine repeat sequences, which are characteristic for several neurodegenerative diseases including Huntington disease and spinocerebellar ataxias, for proteasomal degradation also results in partial degradation in vitro and in vivo (12). These different examples highlight the fact that regulated ubiquitin proteasome-dependent processing is an important physiological aspect of the proteasomal activity that might also play a role in the development of different diseases.
Mechanistically partial processing can be explained by two alternative models. The first one proposes that a substrate is threaded into the proteasome by one end and processed until a tightly folded structure disrupts further degradation (the endfirst model). The second model suggests that a loop structure is formed at an internal protein site that is then threaded into the proteolytic chamber and cleaved, leading to one part of the protein being degraded and the other being released (the endoproteolysis or loop model) (13). This model has been elegantly demonstrated by using circular polypeptides (14) or natural proteins, such as the NF-B p105 polypeptide (15). The main difference between these two models is whether processing ends, or starts, at the site of the released fragment and, thus, whether the effect is on the elongation or the initiation of substrate processing.
The function of the Gly-Ala repeat (GAr) 4 sequence of the EBNA1 protein of Epstein-Barr virus is to prevent endogenous presentation of EBNA1-derived antigenic peptides for the major histocompatibility complex class I pathway (16). This function of the GAr has been linked to its capacity to prevent its own synthesis (17) and proteasomal degradation (18). It was initially suggested that the GAr interferes with the proteasome-dependent degradation pathway and abrogates or severely inhibits EBNA1, or any other protein to which it is fused, from being degraded by the proteasome by a yet unknown mechanism downstream of the polyubiquitination process (18 -22). More recent reports have linked the GAr with partial proteasomedependent degradation when fused to the ornithine decarboxylase substrates (23,24). However, partial degradation of EBNA1 would impair its function, and this is not observed in Epstein-Barr virus-infected cells, indicating that this model cannot fully explain the function of the GAr. This prompted us to look further into the mechanism by which the GAr affects the proteasomal degradation process.
We studied the effects of GAr on proteasome-dependent degradation when fused to the p53 and IB␣ reporter proteins that are well known substrates for the ubiquitin-proteasome pathway as well as in its native EBNA1 context. In vivo results showed that GAr does not always act as an inhibitor of degradation, and it can either promote or prevent 26 S-dependent degradation of the same substrate, depending on where in the protein it is located. In all cases where the GAr was associated with preventing degradation, we observed a majority of the substrate being released intact from the 26 S proteasome and at the same time a fraction of the chimeras being partially degraded. These results cannot be explained by previous models in which the GAr would cause a substrate-and position-independent prevention of degradation or would inhibit the processing of the substrate through the 20 S chamber. We instead propose a model in which the GAr is disrupting the unfolding of the substrate. This model not only offers an explanation to the results presented here but also encompasses previous work on how the GAr affects proteasomal degradation and might help to shed light on the mechanisms that control regulated ubiquitin-dependent partial processing.

EXPERIMENTAL PROCEDURES
Plasmid Constructions-The GAr-carrying plasmid encoding the full-length 235-amino acid GAr and carrying an HA tag in the N terminus, EBNA1, and EBNA⌬GAr have been described previously (17). Each one of the constructs used was made in the following way using the pCDNA3.1 vector. For the GAr-p53 construct, p53 was amplified using the sense primer 5Ј-GCGCGAATTCTTGAGGAGCCGCAGTCAGATC-3Ј and the antisense primer 5Ј-GCGCTCTAGTCAAGACGTCTGA-GTCAGGCCCTTC-3Ј and cloned into the GAr plasmid using EcoRI and XbaI sites. For the p53-GAr construct, p53 was amplified using the sense primer 5Ј-GCGCAAGCTTGAGGA-GCCGCAGTCAGATC-3Ј and antisense primer 5Ј-GCGCTC-CGGACGTCTGAGTCAGGCCCTTC-3Ј and cloned into the GAr plasmid using HindIII and BspEI restriction sites. For p53F19A-GAr, p53F19A (25) was amplified using the same sense and antisense primers as for the p53-GAr construct and was then cloned in the GAr plasmid. For the GAr-IB␣ construct, IB␣ was amplified using the sense primer 5Ј-GCG-CAATTCTTCCAGGCGGCCGAGCGCC-3Ј and antisense primer 5Ј-GCGCTCTAGATCATAACGTCAGACGCTGG-3Ј and cloned into the GAr plasmid using EcoRI and XbaI restriction sites, and then it was cloned in the GAr-p53 plasmid in the place of p53. For the IB␣-GAr construct, IB␣ was amplified using the sense primer 5Ј-GCGCGAATTCATGTTCCAGGC-GGCCGAGC-3Ј and antisense primer 5Ј-GCTCCGGACTAA-CGTCAGACGCTGGCC-3Ј that introduced EcoRI and BspEI restriction sites, and then it was cloned in the p53-GAr plasmid in the place of p53. For the HA-p53-GAr-p53 construct, p53 was amplified using the same sense and antisense primers as for the p53-GAr construct, and then it was cloned in the GAr-p53 plasmid in front of p53. For the HA-IB␣-GAr-p53 construct, IB␣ was amplified using the sense primer 5Ј-GCGCAAGCT-TTTCCAGGCGGCCGAGCGCC-3Ј and antisense primer 5Ј-GCTCCGGACTAACGTCAGACGCTGGCC-3Ј that introduced HindIII and BspEI restriction sites, and then it was cloned in the GAr-p53 plasmid in front of p53. For GAr-p53-GAr-HA, the GAr plasmid was digested with HindIII and EcoRI, and GAr was inserted into the p53-GAr plasmid in front of p53. For poly(Q)-p53 and p53-poly(Q), 80-amino acid-long poly(Q) was amplified from mutant huntingtin (a gift from N. Deglon) using the sense primer 5Ј-GCGCGATATCTCCCTC-AAGTCCTTCCAG-3Ј and antisense primer 5Ј-GCGCGATA-TCCGGTGGCGGCTGTTGCTG-3Ј that introduced EcoRV restriction sites. GAr-p53 and p53-GAr plasmids were digested with EcoRV, and poly(Q) was inserted in the place of GAr. The Ub-EBNA1 construct was a gift from J. Tellam.
Transfection Assays and Western Blot Analysis-H1299 p53 Ϫ/Ϫ and HeLa cells were incubated in RPMI 1640 medium supplemented with 10% fetal calf serum and penicillin/streptomycin. Cells were cultured at 37°C in 5% CO 2 . 140 ϫ 10 3 cells were seeded on 6-well plates and transfected using GeneJuice Transfection Reagent (Novagen) according to the manufacturer's protocol. Cells were harvested 48 h post-transfection and lysed in lysis buffer (20 mM Tris, pH 7.5, 150 mM NaCl, 1% Nonidet P-40) containing a protease inhibitor mixture (Roche Applied Science). Protein concentrations were measured using a Bradford assay. Total cell extracts were fractionated by SDS-PAGE, transferred to BioTraceNT nitrocellulose blotting membranes (Pall Corp.) and probed with anti-p53 rabbit polyclonal antibody (CM-1), anti-IB␣ mouse monoclonal antibody (Imgenex Corp.), anti-GAr rabbit polyclonal antibody, anti-EBNA1 mouse monoclonal antibody (OT1X), or anti-HA tag mouse monoclonal antibody. After incubation with the appropriate peroxidase-conjugated secondary antibodies, proteins were visualized by ECL (Amersham Biosciences) and quantified with the CHEMI-SMART 5000 imaging system. Pulse-Chase Assay-Transfected H1299 cells were pulse-labeled with 90 Ci of EasyTag Expre 35 S 35 S Protein Labeling Mix (PerkinElmer Life Sciences) for 1 or 2 h after being cultured in methionine-free medium, and then chased in fresh medium containing 10 mM cold methionine for the indicated time points before harvesting. After centrifugation, cell pellets were lysed in buffer containing 20 mM Tris, pH 7.5, 150 mM NaCl, 1% Nonidet P-40 in the presence of Complete protease inhibitor mixture (Roche Applied Science). Lysates were precleared with mouse IgG and protein G-Sepharose and immunoprecipitated with anti-EBNA1 or anti-p53 monoclonal antibodies and protein G-Sepharose. The beads were washed with phosphatebuffered saline and lysis buffer four times and boiled in SDS loading buffer. Immunoprecipitates were analyzed by SDS-PAGE using 4 -12% precast gels (Invitrogen). Data were obtained using phosphorimager analysis.
Pulse Assay-Transfected H1299 cells were pulse-labeled with [ 35 S]methionine (90 Ci of EasyTag Expre 35 S 35 S Protein Labeling Mix (PerkinElmer Life Sciences)) after being cultured in methionine-free medium and 25 M proteasome inhibitor MG132 (Merck Biosciences). Then cells were harvested at the indicated time points. Proteins were immunoprecipitated and separated by SDS-PAGE as described above Quantitative Reverse Transcription-PCR-Total cellular RNA was extracted using TRIzol reagent (Invitrogen). cDNA synthesis was carried out using the Moloney murine leukemia virus reverse transcriptase (Invitrogen). Triplicate samples were subjected to quantitative PCR using LightCycler SYBR Green I and hybridization probe systems (Roche Applied Science). The relative abundance of gene target mRNA was calculated after normalization using the TATA box-binding protein.
In Vitro Degradation Assay-H1299 cells were transfected with p53 or the p53-GAr-HA and GAr-p53-GAr-HA constructs. Cell lysates were then immunoprecipitated using anti-p53 monoclonal antibody and protein G-Sepharose. The beads were washed with phosphate-buffered saline and lysis buffer and then incubated in 26 S proteasome buffer (20 mM Tris-HCl, pH 7.5, 20 mM NaCl, 10 mM MgCl 2 , 0.25 mM ATP, 1 mM dithiothreitol) with 400 ng of bacterially expressed hMdm2, 50 ng of E1 (Calbiochem), 50 ng of UbcH5 (Calbiochem), and 1 g of purified 26 S proteasome (BioMol) at 37°C for 3 h. The reaction was terminated by adding SDS loading buffer and boiling for 15 min at 85°C. The results were visualized by Western blot using anti-p53 or anti-GAr polyclonal antibodies.
Fluorescence Microscopy-Cells were grown on coverslips. 24 h after transfection cells were washed twice in phosphatebuffered saline and fixed with 20% acetone, 80% methanol for 20 min at Ϫ20°C. The fixative was removed by three phosphate-buffered saline washes, and cells were probed with anti-p53 mouse monoclonal antibody and anti-mouse Texas Red secondary antibody. The nucleus was localized by using 4Ј,6-diamidino-2-phenylindole staining. Cells were visualized using an LSM 510 META (Carl Zeiss) confocal microscope.

RESULTS
The GAr Does Not Affect EBNA1 Stability-Earlier studies have suggested that the GAr sequence of EBNA1 protein has the capacity to inhibit ubiquitin-dependent degradation of EBNA1 (18). However, when we compared the rate of degrada-tion of EBNA1 with that of an EBNA1 protein that lacks the GAr (EBNA⌬GAr) in vivo using cycloheximide pulse-chase we found that both proteins have a similar half-life over 8 h (Fig.  1a). It has been suggested previously that the effect of the GAr on EBNA1 turnover rate is not evident after 8 h (18), so to test the half-life of these proteins over a longer time period, cells expressing both constructs were labeled with [ 35 S]methionine and chased for 12 and 24 h (Fig. 1b). Surprisingly we found that the half-life of EBNA1 and EBNA⌬GAr are very similar, indicating that the GAr does not impose any significant changes to EBNA1 stability under these conditions. On the contrary, when ubiquitin was fused to the N terminus of EBNA1, which has been shown to target it for the 26 S proteasome (26,27), EBNA1 appeared to be efficiently and rapidly degraded (Fig. 1c). Together these data suggest that EBNA1 is a long lived protein but that the explanation for this is to be found outside the GAr sequence.
The GAr Affects Proteasomal Degradation in a Substrateand Position-dependent Way-The capacity of the GAr to mediate protein stability has mainly been studied using substrates other than EBNA1, so we wanted to investigate to what extent the reported effects of GAr on protein degradation are substrate-specific. We constructed a set of chimeras containing the full-length GAr fused to the N or C terminus of p53 (GAr-p53 or p53-GAr, respectively) ( Fig. 2a). We chose p53 because it is well known that binding of the Mdm2 E3 ligase to the N terminus of p53 promotes p53 polyubiquitination and its subsequent degradation by the 26 S proteasome (28 -30). Hence we could control the degradation of the Gly-Ala-p53 fusion constructs via the ubiquitin-dependent pathway by regulating Mdm2 expression levels.
When we expressed different Gly-Ala-p53 chimeras in H1299 cells (expressing a small amount of endogenous Mdm2) and induced 26 S proteasome-dependent degradation by overexpressing Mdm2, we observed a difference in stability of the fusion constructs depending on the position of GAr. Although fusing the GAr to the C terminus of p53 (p53-GAr) led to protection from Mdm2-mediated proteolysis (Fig.  2b), fusion to the N terminus of p53 (GAr-p53) resulted in a more unstable product compared with p53 itself (Fig. 2b). Pulse-chase experiments confirmed that the half-life of p53-GAr is almost doubled in the presence of Mdm2 as compared with p53 itself (Fig. 2c). In the case of GAr-p53, the mRNA translation inhibitory effect of the GAr (17) resulted in a protein in which the rate of degradation exceeded the rate of synthesis in the presence of Mdm2, making pulse-chase quantifications impossible. The effect of the destabilization of the GAr-p53 was only observed in the presence of Mdm2, demonstrating that fusion of the GAr to the N terminus of p53 did not generate an intrinsic unstable protein; this suggests that the GAr-p53 chimera is properly folded and that the observed differences in the stability of the chimeras are 26 S-dependent. This was further supported by the observation that both GAr-p53 and p53-GAr induced the expression of the p53 response genes p21, Bax, and Puma, indicating that these p53 chimeras take an active conformation (supplemental Table  1). There was also no difference observed in the intracellular localization of the GAr chimeras compared with p53 wild type (supplemental Fig. 1a), further confirming that GAr-carrying chimeras are properly folded and do not form aggregates. We FIGURE 2. The GAr affects proteasomal degradation in a substrate-and position-dependent way. a, schematic representation of the various GAr-carrying p53 and IB␣ constructs. b, the E3 ubiquitin ligase Mdm2 interacts directly with p53 and promotes its polyubiquitination and subsequent degradation by the 26 S proteasome. p53, GAr-p53, and p53-GAr constructs were transfected in H1299 cells in the absence or presence of Mdm2. Transfection efficiency was assayed by co-transfecting cells with green fluorescent protein (GFP), and equal loading was confirmed by immunoblotting for actin. c, an autoradiograph of a 4-h chase experiment following [ 35 S]methionine metabolic pulse labeling of p53 and p53-GAr in the presence of Mdm2. The GAr causes a slowdown of synthesis, and less p53-GAr is therefore labeled (17). Data for the pulse-chase experiment presented in the graph below were obtained using phosphorimager analysis. d, TNF␣ treatment promotes IB␣ degradation by the 26 S proteasome. HA-IB␣ wild type (wt), GAr-IB␣, and IB␣-GAr constructs were transfected in HeLa cells. 36 h post-transfection TNF␣ was added at a final concentration of 50 M for 30 min to one of each duplicate. Immunoblotting using an ␣-IB␣ monoclonal antibody revealed the levels of endogenous IB␣ (IB␣ end) and the different exogenous IB␣ fusion constructs in the absence (Ϫ) or presence (ϩ) of TNF␣. Equal loading was confirmed by immunoblotting for actin. The right graph shows the relative amount of protein after TNF␣ treatment as determined by quantification of Western blot analysis using chemiluminescence and a charge-coupled device camera. Because of differences in expression levels between the constructs before TNF␣ treatment, the relative amount of each construct is given the arbitrary value of 100, and the relative decrease in expression levels after treatment is shown with S.D. All data are representative of at least three independent experiments. IP, immunoprecipitation; WB, Western blot. also investigated whether fusion of the GAr to the N or C terminus of p53 affects Mdm2-dependent ubiquitination. Treatment of cells expressing p53 or p53-GAr chimeras with proteasome inhibitors revealed no significant differences in the relative amounts of ubiquitination of the chimeras (supplemental Fig. 1b); this is in line with previous reports that the GAr acts downstream of the polyubiquitination process (19,22). Hence fusion of the GAr to either end of p53 altered 26 S-dependent turnover rate but did not affect the intrinsic stability, the localization of the fusion protein, or the capacity of Mdm2 to promote polyubiquitination. Importantly the fact that the entire GAr-p53 chimera was degraded by the proteasome shows that GAr does not impose any physical hindrances for the proteasomal catalytic activity nor does it prevent the interaction of sub-strates with the proteasome as has been suggested previously (19,31).
To see to what extent the positiondependent effect of GAr on protein stability observed with p53 is substrate-specific, we also fused the GAr to the N and C termini of IB␣ (GAr-IB␣ and IB␣-GAr, respectively) (Fig. 2a). Extracellular stimulation by a variety of sources, including TNF␣, results in the activation of the IB kinase complex whereupon IB␣ is phosphorylated and rapidly ubiquitinated and targeted for 26 S proteasome-mediated degradation (32). As shown in Fig. 2d, 30-min treatment with TNF␣ in HeLa cells, in which TNF␣-dependent degradation of IB␣ is known to be efficient, resulted in a sharp reduction in the expression levels of an HA-tagged IB␣. Quantification of the Western blot showed that under the same conditions the levels of GAr-IB␣ were reduced by approximately 40% and the levels of IB␣-GAr were reduced by approximately 15% compared with a 70% reduction in the levels of HA-IB␣ wild type. Thus, the effect of GAr on ubiquitin-dependent degradation of IB␣ is also position-dependent; however, unlike p53, the GAr does not promote the turnover of IB␣.
To further confirm that the results observed were proteasomedependent, the p53-GAr and IB␣-GAr transfectants were treated for 2 h with the proteasome inhibitor MG132 (supplemental Fig. 2). It should be noted that the effect of proteasome inhibitors on Mdm2dependent degradation of GAr-p53 constructs was lesser compared with GAr-IB␣ constructs where the inhibitors were added before or at the same time as the degradation signal.
GAr-carrying Chimeras Are Partially Processed by the 26 S Proteasome-The position-dependent effect of GAr on proteasomal degradation has not been observed previously, so we wanted to investigate this further. Interestingly immunoblotting using an anti-GAr-specific antibody revealed the existence of a band at about 40 kDa in cells co-expressing p53-GAr and Mdm2 ( Fig. 3a and supplemental Fig. 3a). This band corresponds to the approximate size of the free GAr and was not observed when using a p53 sequence carrying a single point mutation (F19A) that prevents binding to Mdm2 (p53F19A-GAr construct) (25) (Fig. 3a and supplemental Fig. 3a), underlying the fact that this effect is dependent on targeting the p53-  (25), and free GAr* was not detected in cells expressing p53F19A-GAr in the presence of Mdm2. The gel was overexposed to show the presence, or absence, of the free GAr product. b, in vitro degradation of p53-GAr using purified 26 S proteasome as described under "Experimental Procedures." c, fusion of the GAr to the N terminus of p53 (GAr-p53) does not lead to protection from Mdm2-dependent degradation or in the release of the free GAr*. d, HeLa cells expressing the IB␣-GAr fusion protein show an increase in the amount of free GAr* after TNF␣ treatment that was suppressed in the presence of 20 M of the proteasome inhibitor MG132. All data are representative of at least three independent experiments. * indicates that the polypeptide is a degradation product. The corresponding less exposed gels are shown in supplemental Fig. 3. wt, wild type; WB, Western blot; IP, immunoprecipitation.
GAr chimera for the 26 S-dependent degradation pathway. This was further confirmed in in vitro experiments where we observed the appearance of a GAr product derived from the p53-GAr in the presence of purified 26 S proteasomes (Fig. 3b). Notably the GAr band was not visible when GAr was fused to the N terminus of p53 (GAr-p53), showing that this construct is completely degraded when targeted for the proteasome by Mdm2 (Fig. 3c and supplemental Fig. 3b). Expression of the IB␣-GAr fusion protein that was shown to be protected from degradation (Fig. 2d) also resulted in the appearance of the free GAr ( Fig. 3d and supplemental Fig. 3c). The amount of free GAr was increased after TNF␣ treatment and was prevented by the addition of proteasome inhibitor, demonstrating that this effect is 26 S-dependent.
Our results taken together so far demonstrate the following. (i) The GAr can prevent 26 S proteasome-dependent degradation of substrates to which it is fused, and this is linked to a fraction of the substrate being partially processed (e.g. p53-GAr and IB␣-GAr); (ii) the GAr can promote 26 S-dependent degradation and be completely degraded along with the rest of the fusion protein (e.g. GAr-p53); or (iii) the GAr can impose little, or no, effect on protein stability (e.g. EBNA1).
The GAr Is a Specific Regulator of the 26 S Proteasome-Several amino acid repeat sequences have been shown to interfere with the ubiquitin-proteasome pathway. The most commonly known example is the polyglutamine repeats (poly(Q)), associated with Huntington disease (HD) or spinocerebellar ataxias SCA 1-7, which are dominantly inherited neurodegenerative diseases. These diseases are caused by abnormal expansions of long glutamine sequences found normally in certain proteins (e.g. huntingtin or ataxin) (33). The poly(Q) sequences in huntingtin from normal individuals are 6 -36 residues long; however, in typical Huntington patients, they exceed 36 residues, and the protein has a strong tendency to form aggregates containing large amounts of ubiquitin, components of the 26 S proteasomes, and molecular chaperones, indicating unsuccessful attempts by the cell to refold or destroy these toxic protein aggregates (34 -36).
To test the specificity of the effects of GAr on proteasomal degradation, we fused an 80-amino acid-long poly(Q) to the N or C terminus of p53 and expressed these constructs in H1299 cells. As shown in Fig. 4a, insertion of the poly(Q) sequence in either end of p53 resulted in the stabilization of the chimeras (poly(Q)-p53 or p53-poly(Q)). Unlike the GAr, the stabilization conferred by the poly(Q) repeat was not accompanied by partial degradation of the fusion proteins (Fig. 4b). Similar to the GAr, however, the poly(Q) repeat also led to the accumulation of polyubiquitinated products. Hence both repeat sequences act downstream of the ubiquitination process, but only the GAr seems to cause regulated ubiquitin proteasome-dependent processing in a position-and substrate-dependent fashion, demonstrating the unique features of the GAr as a specific regulator of the 26 S proteasome.
The 26 S Proteasome Degrades the N-terminal Part of GArcarrying Chimeras-The capacity of GAr to protect the entire chimera from 26 S-dependent degradation and at the same time induce partial degradation of a fraction of the substrates is puzzling and cannot be easily explained by any existing models for proteasomal degradation. So we next set out to investigate this phenomenon in more detail. We first made a triple construct coding for two p53 proteins separated by the GAr carrying an HA tag at its N terminus (HA-p53-GAr-p53). This construct allowed us to test whether GAr is able to exert its role on inhibiting degradation and provoking partial proteolysis when it is encircled by two proteins that consist of good substrates for proteasomal degradation. It also allowed us to test whether the difference in the rate of degradation of GAr-p53 and p53-GAr is because of effects related to an N-or C-terminal initiation of p53 degradation.
When we expressed HA-p53-GAr-p53 in H1299 cells, we detected the full-length triple construct and a band corresponding to the GAr-p53* (* indicates a degradation product) that was recognized by both anti-p53 (Fig. 5a, left panel) and anti-GAr antibodies (data not shown). Because this band was not recognized by the anti-HA tag antibody (Fig. 5a, right  panel) we can conclude that it corresponds to the GAr-p53* part of the triple construct. The GAr-p53* fragment appeared to be decreased when cells were co-transfected with Mdm2 (Fig. 5a, left panel), suggesting that this fragment is further degraded by the proteasome. To test whether this GAr-p53* is indeed a product of proteasomal degradation, we treated cells with the specific proteasome inhibitor epoxomicin. This resulted in a net increase in the expression levels of HA-p53- GAr-p53, whereas the amount of GAr-p53* was slightly decreased (Fig. 5a, left panel) despite the fact that GAr-p53* was also subject to proteasome-dependent degradation. Importantly when cells were treated with several other protease inhibitors we did not observe significant differences in the ratios of HA-p53-GAr-p53 to GAr-p53* levels (Fig. 5a, middle panel). A metabolic pulse labeling using [ 35 S]methionine in the presence or absence of proteasome inhibitors showed that the GAr-p53* was not derived from partial translation of the HA-p53-GAr-p53 message (Fig. 5b).
Similarly we constructed a chimera where GAr is inserted between the IB␣ and p53 proteins (HA-IB␣-GAr-p53). This fusion protein could be targeted for 26 S proteasome-dependent degradation both by Mdm2 or TNF␣. Immunoblotting using a p53 antibody revealed the presence of bands corre-sponding to the HA-IB␣-GAr-p53 and the GAr-p53* (Fig. 5c). Overexpression of Mdm2 resulted in a decrease of both products (Fig. 5c,  left panel), whereas treatment with TNF␣ resulted in a decrease in the amounts of HA-IB␣-GAr-p53 and an increase in GAr-p53* (Fig. 5c,  right panel). Hence in either case, the appearance of GAr-p53* was a result of 26 S proteasome activity, whereas only Mdm2 targeted GAr-p53* for further degradation. This was also confirmed by the observation that treatment with epoxomicin in the presence of Mdm2 or TNF␣ prevented the degradation of the HA-IB␣-GAr-p53 and at the same time reduced the amount of GAr-p53* (Fig. 5c, left and right panels). In neither of the triple constructs tested could we detect any free HA-p53 or HA-IB␣ as a result of the partial degradation, indicating that the N-terminal part of the chimeras is completely degraded.

Partial Processing of GAr-carrying Chimeras by the 26 S Proteasome Is Preceded by an Endoproteolytic
Cleavage-The degradation results for the various GAr constructs raise the possibility that partial degradation observed with GAr may be due to the proteasome starting to degrade the substrate from either end (N or C terminus) and then stopping once it reaches the GAr (end-first model). Alternatively GAr may trigger a proteasome-dependent endoproteolytic cleavage of the substrate near the GAr (the loop model (4,13)). To test which one of these models applies to GArinduced partial processing, we constructed a model p53 protein sandwiched in between two GAr sequences, GAr-p53-GAr-HA. The GAr sequence does not contain any degradation signals and when expressed by itself was very stable (data not shown), indicating that the 26 S proteasome does not recognize GAr as a substrate itself. Hence if the GAr-p53-GAr-HA construct is degraded when expressed in cells, this would imply that the proteasome initiates proteolysis from an internal p53 site in this substrate.
We found that GAr-p53-GAr-HA was degraded in the presence of Mdm2 and accumulated in the presence of proteasome inhibitors, whereas the GAr-p53F19A-GAr-HA construct that cannot bind Mdm2 was resistant to degradation (Fig. 6a), confirming that its processing is 26 S-dependent. Thus, these results suggest that in vivo 26 S-dependent The HA-GAr-p53 polypeptide expressed from the HA-GAr-p53 mRNA was detected using an anti-HA tag antibody, but the GAr-p53* that is derived from degradation of the HA-p53-GAr-p53 was not (right panel). b, an autoradiograph showing a 30-min [ 35 S]methionine pulse label of cells expressing HA-GAr-p53 and HA-p53-GAr-p53 in the presence or absence of MG132 followed by an immunoprecipitation (IP) using an anti-HA tag monoclonal antibody. c, HA-IB␣-GAr-p53 fusion protein is partially degraded by the 26 S proteasome. Both the triple HA-IB␣-GAr-p53 and the GAr-p53* product are targeted for degradation by Mdm2 (left panel). When cells are treated with TNF␣ only the triple construct is targeted for degradation, leading to an accumulation of GAr-p53* (right panel). Treatment with epoxomicin prevents Mdm2-or TNF␣-dependent degradation of HA-IB␣-GAr-p53 and results in less GAr-p53*. The graphs show the quantification of the corresponding Western blots (WB), and the HA-IB␣-GAr-p53 in the absence of Mdm2 or TNF␣ has been given the arbitrary value 100. All data are representative of at least three independent experiments. * indicates that the polypeptide is a degradation product. AEBSF, 4-(2-aminoethyl)benzenesulfonyl fluoride. Error bars represent S.D.
degradation of the GAr chimeras can be initiated at an internal protein site. After expressing the GAr-p53-GAr-HA in cells, we also observed the appearance of GAr-HA* and GAr-p53* fragments as determined using anti-HA tag and anti-GAr antibodies, respectively (Fig. 6b, left and right panels). When cells were treated with proteasome inhibitors (MG132 or epoxomicin) (Fig. 6b, right panel), the amount of free GAr-HA*, which was the more stable fragment, was decreased indicating that the appearance of these products was proteasome-dependent. To further confirm this, we performed in vitro degradation assays using purified 26 S proteasomes (Fig. 6c). GAr-p53-GAr-HA protein cannot be purified from bacteria because GAr has the tendency to aggregate nor can it be in vitro translated due to the fact that GAr strongly decreases translation efficiency; therefore we were obliged to use a modified in vitro degradation assay where proteins are immunoprecipitated using p53 antibodies and then used for ubiquitination and degradation in the presence of purified 26 S proteasomes, Mdm2, and ubiquitin. The modification of the standard in vitro protocol could explain the relatively inefficient degradation of the p53 control protein observed in vitro. Nevertheless we were still able to confirm that it was ubiquitinated and slightly degraded (Fig. 6c, left panel). Most importantly, in the case of the GAr-p53-GAr-HA construct, we observed the release of GAr products after the addition of 26 S proteasome (Fig. 6c,  right panel). Taken together, these observations suggest that the 26 S proteasome initiates proteolysis of GAr-p53-GAr-HA from an internal site next to the C terminus of p53, thus giving rise to the GAr-p53* that is further degraded by the proteasome and the GAr-HA* that is resistant to degradation. Accordingly the effect of the GAr would not be on the elongation process but on the initiation phase of degradation.

DISCUSSION
The GAr sequence of EBNA1 protein is known to have the capacitytoprotectproteinstowhichitisfusedfromproteasomedependent degradation. It was initially proposed that the effect of the GAr was substrate-unspecific and caused by a direct inhibition of the proteolytic activity of the proteasome (18 -20, 22). Thereafter GAr was characterized as a "stabilization signal" for the ubiquitin-proteasome pathway (37), and it was further proposed that it may have therapeutic applications in gene transfer settings for stabilizing specific proteins (22,38).
However, there have been many conflicting reports concerning the capacity of GAr to act as a stabilization signal, the physiological reason for this effect, and the underlying molecular mechanism. For example, when a 24-amino acid-long GAr was fused near the N terminus of ErbB-2 protein, a member of the epidermal growth factor receptor family, the chimera was shown to be efficiently degraded by the proteasome (39). Interestingly when GAr was further inserted between amino acids 55 and 56 of ErbB-2, the GAr-containing chimera was even more unstable than the wild type protein, suggesting that the previously reported protective effect of GAr is not valid in this case. Furthermore experiments with N end rule and ubiquitin fusion degradation-targeted green fluorescent protein reporters containing GAr have shown that GAr is not always efficient in protecting chimeras from proteasomal degradation (26). Even EBNA1 itself when targeted for the 26 S proteasome by the fusion of ubiquitin appears to be efficiently and quickly degraded (26,27).
To further investigate the effect of GAr on proteasomedependent degradation, we constructed p53 and IB␣ chimeras containing the full-length 235-amino acid-long GAr in their N or C termini (Fig. 2a). To ensure that the GAr is not causing aggregates under the conditions used here, we performed immunohistochemistry analysis (supplemental Fig. 1a) and tested the activity (supplemental Table 1) and ubiquitination (supplemental Fig. 1b) of the fusion constructs. When we looked at the degradation of p53 and IB␣ chimeras in vivo, we found that depending on the substrate FIGURE 6. Partial processing of GAr-carrying chimeras by the 26 S proteasome is preceded by an endoproteolytic cleavage. GAr was fused to either end of p53 (GAr-p53-GAr-HA) to protect the substrate from end first-mediated degradation by the proteasome. a, GAr-p53-GAr-HA or GAr-p53F19A-GAr-HA expression constructs were transfected in H1299 cells in the absence or presence of Mdm2 and treated with epoxomicin where indicated. b, coexpression of Mdm2 targets the triple chimera for degradation and results in the appearance of GAr-p53* and GAr-HA* degradation products (left and right panels). Immunoblotting using an anti-HA tag antibody confirms the identity of the degradation products (left panel). Treatment with the proteasome inhibitors MG132 or epoxomicin leads to a decrease in the amount of GAr-HA* (right panel). c, in vitro degradation of p53 or GAr-p53-GAr-HA using immunopurified substrates, 26 S proteasome, and Mdm2 as described under "Experimental Procedures." All data are representative of at least three independent experiments. * indicates that the polypeptide is a degradation product. WB, Western blot; IP, immunoprecipitation. and on the position where GAr is inserted, we obtained different results on protein degradation. Although fusion of the GAr to the C terminus of p53 or to both ends of IB␣ resulted in stabilization of the chimeras, insertion of the GAr to the N terminus of p53 strongly destabilized the chimera, leading to its accelerated degradation in the presence of Mdm2 (Fig. 2). Thus, contrary to what was initially suggested, these data demonstrate that the effect of the GAr on protein stability is substrate-and position-dependent. By fusing another repeat sequence in the N or C terminus of p53, the 80-amino acid-long poly(Q) repeat that like the GAr has been suggested to affect proteasome-dependent degradation (12,40), we were able to show that despite the fact that both repeats act downstream of ubiquitination they affect 26 S-dependent degradation via different mechanisms, highlighting the specificity of the GAr on regulating 26 S proteasome degradation (Fig. 4, a and b).
What is surprising, considering what has been published previously on the GAr, is our observation that when GAr was in its natural context, the EBNA1 protein, it had no major effect on protecting EBNA1 from degradation compared with a mutant lacking the GAr (EBNA⌬GAr) (Fig. 1, a and b). The turnover of EBNA1 and EBNA⌬GAr has been tested previously in vivo by Masucci and co-workers (18). By using a recombinant vaccinia virus system in pulse-chase experiments performed in CV1 cells, the authors observed that EBNA1⌬GAr has a half-life of ϳ20 h compared with EBNA1 that is stable over a 20-h chase. This difference in the turnover of EBNA1 and EBNA1⌬GAr is bigger than the 20% difference we observed after a 24-h chase in H1299 cells (Fig.  1b). The reason for this discrepancy is not clear, but it is possible, in view of the fact that strong overexpression of EBNA1 can lead to aggregates, that the vaccinia expression system used previously (18) could result in EBNA1 aggregates that would appear on an immunoblot as a "stabilization." But in either case, it is unlikely that such a limited effect on protein stability could explain how the GAr can prevent EBNA1 peptides from being presented to the major histocompatibility complex class I-restricted pathway. Instead the results presented here support the idea that the capacity of the GAr to suppress major histocompatibility complex class I antigen presentation of EBNA1 is due to its capacity to prevent EBNA1 mRNA translation (17). The reason why EBNA1 turns out to be the only protein tested for which the GAr has little effect on stability might suggest that EBNA1 is simply not targeted for degradation via the 26 S proteasome. This notion is supported by the fact that EBNA1 has never been shown to be a substrate for the ubiquitinproteasome pathway or to interact with any putative E3 ligases, and any attempts to demonstrate ubiquitination of both EBNA1 and EBNA1⌬GAr have failed (18,41). In contrast, when EBNA1 was artificially targeted for the proteasome by fusing a Ub to its N terminus, it was efficiently and completely degraded (Fig. 1c).
The observation that GAr was completely degraded in the context of the Ub-EBNA1 construct (Fig. 1c) or when fused to the N terminus of p53 (Fig. 2b) shows that GAr does not cause any problem for the proteolytic activity of the 26 S proteasome. This contradicts previous reports suggesting that GAr can inhibit Ub proteasome-dependent proteolysis in trans (31). It has also been proposed that GAr may act by preventing the interaction of the substrate with the 26 S proteasome (19). This is challenged by our observation that a fraction of the fusion proteins that were protected from proteolysis was actually partially degraded by the 26 S proteasome, leading to the release of GAr-containing fragments (Fig. 3). Hence the GAr does not affect the targeting of substrates to the 26 S proteasome and instead acts downstream of the recognition and interaction of the substrate with the proteasome. More recent reports have shown that a 30amino acid GAr when fused to the ornithine decarboxylase causes partial degradation of the chimera in yeast (23,24). The authors proposed a model where the ATPases of the 19 S regulatory particle "slip" over the GAr, thereby hindering translocation of tightly folded domains of the protein into the 20 S core particle. Hence this model predicts that GAr acts during the elongation process of degradation. Although GAr has been suggested to be unstructured and to act as an independent domain in chimeric proteins (42), the model proposed by Hoyt et al. (24) cannot explain our results demonstrating the release of the full-length 235-amino acid-long GAr from several double and triple chimeras (Figs. 3, 5, and 6). In particular, the partial degradation of the triple chimera containing GAr on both ends of p53 (GAr-p53-GAr-HA) (Fig. 6) cannot be explained by a similar model that claims that GAr prevents the elongation of the substrate through the chamber of the 20 S core by acting as a "slippery rope." If this was the case, we would not expect degradation of GAr-p53-GAr-HA to take place nor the release of the GAr domain. Finally the observation that only a fraction of the protected substrates was partially degraded is also against the idea that GAr acts on the elongation process of degradation because in this case it would be expected that all substrate should be partially degraded.
So what can the explanation be for the puzzling positionand substrate-dependent effect of GAr on 26 S-dependent degradation, and what is the link between protection from degradation and partial processing? The effects of GAr on causing partial processing and also protecting the entire substrate from degradation are likely linked to each other. It is recognized that binding of a protein to the proteasome by means of polyubiquitination is not sufficient to ensure the degradation of a protein, and it is becoming increasingly evident that protein degradation includes several steps that depend on different signals or tags within the protein sequence (43). The tethering, unfolding, and translocation steps are mediated by degradation signals (e.g. polyubiquitination) and unfolding tags (e.g. unstructured domains) that determine where initiation of degradation of the substrate will occur. In terms of degradation by the 26 S proteasome, the 19 S ATPases initiate unfolding of the substrate in close proximity to the degradation signal (43,44). This might offer an explanation for the different effects of the GAr on 26 S-dependent degradation. Fusion of the GAr to the C terminus of p53 suppressed degradation in an Mdm2-dependent fashion (Fig. 1c). Mdm2 promotes polyubiquitination on p53 C-terminal residues (45), and we speculate that this occurs in close proximity to the p53 unfolding tag, which is in accordance with the model that the unfolding tag is close to the polyubiquitination site. Similarly PEST sequences in the C terminus of IB␣ can be used as degradation signals (46,47). Thus, fusion of the GAr to the C terminus of p53 or IB␣ would position it near the degradation signal, and this might interfere with the natural unfolding of the substrate, leading to its protection or partial degradation. Instead fusion to the p53 N terminus could promote the unfolding due to the fact that the N terminus of p53 is known to control the folding of its C terminus (48,49).
Taken together, our results support a model of the GAr impairing unfolding of substrates by the 19 S ATPases of the regulatory particle when fused next to the unfolding tag of a substrate (Fig. 7). This leads to a complete failure of the proteasome to initiate degradation of the main pool of substrate (protection from degradation) or, alternatively, to the formation of a loop structure in a small amount of protein that is endoproteolytically cleaved by the 20 S core particle, resulting in the release of GAr or GAr-carrying parts of the substrate (partial degradation). This model offers an explanation to the unique capacity of GAr to protect the majority of the polyubiquitinated substrate from degradation and to induce at the same time partial degradation of a small fraction of the chimeras.
To date only a few examples, including NF-B (15), Mga2 and Spt23 (13), YB-1 (50), and Gli2/Gli3 (51), have been reported to be endoproteolytically processed by the proteasome, and in most of the cases studies have been carried out in vitro, and cleavage was shown to be mediated by the 20 S proteasome alone. Moreover it still remains unclear what are the properties a protein domain should have to allow specific recognition, endoproteolytic cleavage, and partial processing. Our findings that GAr caused in vivo 26 S-dependent endoproteolytic processing in a substrate-and position-dependent manner indicate a specific role of GAr on protein unfolding and degradation depending on the protein context. The model we propose could help to explain previous confusing reports regarding the role of GAr on protein stability and also how repeat or other sequences, such as the glycine-rich region sequence in NF-B (15) or the processing determinant domain in Gli2/Gli3 (51), can work as transferable elements and induce partial degradation. Finally our data may be relevant for a better understanding of how substrates are degraded by the 26 S proteasome and may provide some ideas as to how proteins can be manipulated to be selectively stabilized/destabilized or partially cleaved.