To better understand proteasomal degradation of nuclear proteins and viral antigens we studied mutated forms of influenza virus nucleoprotein (NP) that misfold and are rapidly degraded by proteasomes. In the presence of proteasome inhibitors, mutated NP (dNP) accumulates in highly insoluble ubiquitinated and nonubiquitinated species in nuclear substructures known as promyelocytic leukemia oncogenic domains (PODs) and the microtubule organizing center (MTOC). Immunofluorescence revealed that dNP recruits proteasomes and a selective assortment of molecular chaperones to both locales, and that a similar (though less dramatic) effect is induced by proteasome inhibitors in the absence of dNP expression. Biochemical evidence is consistent with the idea that dNP is delivered to PODs/MTOC in the absence of proteasome inhibitors. Restoring proteasome activity while blocking protein synthesis results in disappearance of dNP from PODs and the MTOC and the generation of a major histocompatibility complex class I–bound peptide derived from dNP but not NP. These findings demonstrate that PODs and the MTOC serve as sites of proteasomal degradation of misfolded dNP and probably cellular proteins as well, and imply that antigenic peptides are generated at one or both of these sites.
Proteasomes are complex multisubunit proteases abundant in the nucleus and cytosol that function in all eukaryotic cells to degrade damaged or unwanted proteins. The major means of tagging proteins for proteasomal destruction is covalent modification by multiple chains of ubiquitin (Ub)1 (Ciechanover 1998). Incubation of cells with proteasome inhibitors results in the accumulation of polyubiquitinated proteins, and in most circumstances, cell death (Bogyo et al. 1997; Lee and Goldberg 1998).
Proteasomes participate in the generation of many of the peptides that are displayed on the cell surface in a complex with class I molecules of the major histocompatibility complex for surveillance by the immune system (York and Rock 1996). Peptides are generated indiscriminately from host gene products and gene products of viruses and other intracellular pathogens (Pamer and Cresswell 1998). Discrimination between self and nonself peptides is performed at the level of the antigen receptor expressed by CD8+ T lymphocytes. There is sufficient plasticity in the system to enable T cell recognition of some self antigens, which accounts for the generation of most defined T cell responses to tumors.
Although there has been rapid progress in understanding the function of the proteasome-Ub system in cellular metabolism, and more specifically, its role in generating antigenic peptides, many important questions remain, including the involvement of molecular chaperones in the process, the intracellular sites of ubiquitination and proteasomal degradation, and the nature of substrates that provide antigenic peptides. In the present communication we describe findings with a metabolically unstable form of influenza virus nucleoprotein (NP) that provide some initial answers to these questions.
Materials and Methods
143B human osteosarcoma cells lacking thymidine kinase and HeLa cells were obtained from the American Type Culture Collection and maintained in DMEM supplemented with 7.5% (vol/vol) fetal bovine serum in an air/CO2 (91%/9%) atmosphere at 37°C. L929 cells expressing Kb from a transfected gene were maintained similarly.
Recombinant vaccinia viruses (rVVs) were constructed by standard methodology (Chakrabarti et al. 1985) using a modified form of pSC11 with multiple cloning sites to express inserted genes under the control of the p7.5 promoter active early and late in VV replication. dNPpep was constructed by inserting an oligonucleotide encoding KEKEKNKLKRKKLENKDKKDEERNKIREE at position 333 in NPpep. Green fluorescent protein (GFP) constructs were created by fusing NPpep or dNPpep encoding constructs with a cDNA encoding EGFP (Clontech), a red-shifted version of GFP modified by two substitutions (Phe64→Leu, Ser65→Thr), and 190 synonymous coding alterations to match human codon usage. For cloning convenience, sequences encoding the spacer peptide ARDPPVAT were inserted between the COOH terminus of the NPpep constructs and initiating Met of GFP. Most VV infections were performed with media containing cytosine arabinoside at 40 μg/ml to limit expression to viral gene products expressed under the control of early promoters. This reduces the morphological alterations in cells induced by VV and also controls for the blockade in late VV gene expression induced by proteasome inhibitors (Antón et al. 1998).
Antipeptide antisera were prepared by immunizing rabbits with synthetic peptides conjugated to KLH. Peptides corresponded to PR8 NP sequences 2-12 +Cys and Cys+488-498; the extraneous Cys being used for conjugation and coupling to beads. The NH2-terminal–specific serum was affinity purified against the synthetic peptide disulfide coupled to SulfoLink® beads (Pierce Chemical Co.) and absorbed multiple times against uninfected fixed and permeabilized cells to remove antibodies specific for cellular proteins. The COOH-terminal antiserum could be used without further purification. Monoclonal antibodies specific for the proteasome (clones MCP20 and MCP21) were generously provided by K.B. Hendil (University of Copenhagen, Copenhagen, Denmark). Human sera from primary biliary cirrhosis (PBC) patients containing anti–promyelocytic leukemia (PML) oncogenic domain (POD) antibodies were generously provided by D.B. Bloch (Harvard Medical School, Charlestown, MA) and J. Liang (NIDDK, Bethesda, MD). The colocalization of human antibody staining of PODs in TK− cells with the anti-PML mAb was confirmed. The commercial sources of the following antibodies are as listed: poly Ub mouse mAb, clone FK2, Nippon Bio-Test Laboratories; PML mouse mAb, clone PG-M3, Santa Cruz Biotechnologies; γ-tubulin mouse mAb, clone gtu-88, Sigma Chemical Co.; HSP27 mouse mAb, clone G3.1, HSP40 rabbit Ab (SPA-400), HSP47 mouse mAb, clone M16.10A1, HSP56 mouse mAb, clone KN382/EC1, HSP60 mouse mAb, clone LK-1, HSC70 rat mAb, clone 1B5, HSP70 mouse mAb, clone C92F3A-5, HSP90, rat mAb, clone 16F1, and HSP110 rabbit Ab (SPA-1101), all from StressGen Biotechnologies.
Cells were grown overnight on acid-cleaned, 0.17-mm-thick, 12-mm-diameter glass coverslips placed in 24-well plates, infected with virus, and incubated as indicated in the text and figure legends. At the appropriate time, cells were fixed by incubation with 3% (wt/vol) paraformaldehyde in PBS for 20 min and permeabilized by 2 min of treatment with 1% (vol/vol) NP-40 in PBS. After quenching of formaldehyde with 200 mM glycine/PBS, cells were incubated with a mixture of primary antibodies diluted in PBS supplemented with 5% (vol/vol) donkey serum, usually overnight, at 4°C, washed, and incubated for 2–8 h in the same diluent containing secondary donkey antibodies conjugated to DTAF, Texas red, or Cy5, specific for mouse, human, rabbit, or rat Ig (Jackson ImmunoResearch). Coverslips were mounted on glass slides with Fluoromount-G (Southern Biotechnology) containing 15-μm-diameter beads to prevent cell compression, and images collected with a Bio-Rad MRC1024 laser scanning confocal Zeiss Axioplan microscope, using a 63× planapochromat oil immersion objective. Controls established the specificity of fluorochrome-conjugated antibodies for their respective Igs, and that signals in green, red, and far red channels were derived from the respective fluor. Digital images were assembled using Adobe Photoshop software and printed with a Fujix Pictrography digital printer (Fuji). For cytofluorography, cells were incubated with mAbs for 30 min on ice, washed, and incubated with rabbit anti–mouse Ig conjugated to fluorescein (Dako). Cells were suspended in PBS containing ethidium homodimer (Molecular Probes), and analyzed using a FACScalibur® cytofluorograph (Becton Dickinson). Live cells were gated based on scattering properties and low ethidium homodimer staining.
Confluent 143B cells grown in 6-well plates were infected with rVVs and incubated for 150 min and then for 360 min in the presence or absence of 20 μM cbz-LeuLeuLeucinal (zLLL). Plates were transferred to an ice bath, and subjected to a sequential extraction to prepare the nuclear matrix, according to the protocol of Staufenbiel and Deppert 1984, with the exception that all buffers contained Complete® protease inhibitor cocktail (EDTA-free; Boehringer Mannheim) and 10 μM zLLL. Equivalent amounts of samples from each step of the procedure were acetone precipitated, and precipitates resuspended in boiling SDS-PAGE sample buffer (Laemmli 1970). Samples were also prepared from rVV-infected HeLa cells by suspending cells in ice-cold buffer containing 50 mM Tris-hydrochloride (pH 8.0), 5 mM EDTA, 100 mM NaCl, 0.5% (wt/vol) CHAPS ([3-[(3-cholamidopropyl)-dimethyl-ammonio]-1-propanesulfonate]), 0.2% (wt/vol) deoxycholate, and then mixed with an equal volume of boiling SDS-PAGE sample buffer. After electrophoresis, proteins were transferred to Immobilon P membranes (Millipore) in transfer buffer (Towbin et al. 1979) lacking SDS. Membranes were incubated overnight with TBS-Casein (Bio-Rad), and then with rabbit anti-NP antibodies, followed by peroxidase-labeled anti–rabbit IgG (Boehringer Mannheim). Blots were developed using the ECL system (Pierce Chemical Co.), and luminescence recorded by Biomax MR film (Kodak). Images were digitized by a flat bed scanner, assembled using Adobe Photoshop software, and printed with a Fujix Pictrography digital printer.
143 cells infected 3 h previously were incubated in Met-free DMEM with 100 μM lactacystin (LC) for 40 min, and then radiolabeled by 5 min of incubation with [35S]Met. After washing, cells were chased at 37°C for up to 120 min in Met containing DMEM. At appropriate times, 2 × 106 cells were removed to ice. Cells were incubated with 1% (vol/vol) Triton X-100 (TX100) containing buffer and centrifuged at 15,000 g for 10 min. Supernatants and pellets were suspended in boiling SDS-PAGE sample buffer boiled for 5 min, and analyzed by SDS-PAGE. Images of autoradiographs of the dried gels were digitized by a flat bed scanner, assembled using Adobe Photoshop software and printed with a Fujix Pictrography digital printer. The radioactivity present in dried gels was quantitated using a PhosphorImager (Molecular Devices) and the screens supplied by the manufacturer. The VV protein shown in Fig. 2 a was used as an internal standard for normalization of the amount of protein recovered from each sample.
Modification of NP Results in Enhanced Generation of Antigenic Peptides
The NP from the PR8 influenza virus is a 498-residue protein that is transported to the nucleus via multiple nuclear localization sequences (Wang et al. 1997). We genetically engineered NP to contain a 29-residue sequence nearly identical to that from JAK1 kinase proposed to enhance the generation of antigenic peptides by targeting the protein to proteasomes (Realini et al. 1994). In addition, we appended to the COOH terminus a peptide corresponding to residues 257–264 from chicken ovalbumin (OVA). This peptide binds tightly to the H-2 Kb MHC class I molecule, and Kb–Ova257-264 complexes can be easily quantitated cytofluorographically using a mAb (25-D1.16) specific for this complex (Porgador et al. 1997). As a control, the peptide was also expressed at the COOH terminus of wild-type NP (this is termed NPpep and the other construct dNPpep).
After 6 h of infection of L-Kb cells with rVVs expressing NPpep or dNPpep, approximately threefold more Kb–Ova257-264 complexes were present on the surface of VV-dNPpep–infected cells as determined cytofluorographically after indirect immunofluorescence (Fig. 1, top histogram). Incubation of cells with the highly specific irreversible proteasome inhibitor LC resulted in the nearly complete inhibition of complex expression from the chimeric proteins and from OVA, the parent protein (Fig. 1, bottom histogram). There was only a slight effect on cells infected with a rVV expressing Ova257-264 as a cytosolic minigene product (a single Met is appended to the NH2 terminus to enable efficient translation), consistent with the interpretation that LC acts by preventing proteasome liberation of Ova257-264 (or a proteolytic intermediate) from NPpep, dNPpep, and OVA, and not by interfering with VV gene expression or delivery and loading of peptides onto Kb molecules.
Metabolic Stability of dNPpep and NPpep
Increased protein degradation is associated with enhanced generation of antigenic peptides (Tevethia et al. 1983; Townsend et al. 1988). To investigate the more efficient production of Ova257-264 from dNPpep, we examined the metabolic stability of dNPpep and NPpep in the presence and absence of LC. rVV-infected cells were labeled for 5 min with [35S]Met and chased for up to 2 h at 37°C. Proteins present in TX100-soluble and insoluble material were separated by SDS-PAGE and the amounts of NP present in gel migrating with the expected mobility were determined by PhosphorImager analysis (Fig. 2). The total amount of NPpep recovered remained nearly constant throughout the chase period, with the solubility decreasing in a time-dependent manner to a plateau value. This corresponds with the transport of NPpep into the nucleus where it is partially TX100 insoluble. As expected, the process was unaffected by LC. By contrast, in the absence of LC, recovery of both soluble and insoluble dNPpep decreased with time. Importantly, LC selectively increased the recovery of insoluble dNPpep, without affecting soluble dNPpep. We interpret this data to indicate that, first, insertion of the JAK1 sequence into dNPpep greatly enhances its degradation by proteasomes, and, second, that the form digested by proteasomes is insoluble in TX100. These findings predict that incubation of cells with proteasome inhibitors should result in the accumulation of dNPpep in cells.
Intracellular Localization of dNPpep
This prediction was first confirmed by immunofluorescence of fixed and permeabilized rVV-infected cells using rabbit antibodies raised to the NH2 terminus of unmodified NP (anti-NH2). In the absence of proteasome inhibitors, staining of dNPpep observed using a laser scanning confocal microscope (LCSM) was only slightly above background autofluorescence levels (data not shown). In the presence of either LC (data not shown) or the reversible proteasome inhibitor zLLL, dNPpep was detected in three locations: weak staining of the nuclear body (excluding the nucleoli), and strong staining of small nuclear substructures (Fig. 3, first row) and, in some cells, a cytoplasmic structure that was often juxtanuclear (Fig. 3, third and fourth rows, arrows). This differs markedly from the staining pattern of NPpep (Fig. 3, second row), which like wild-type NP (not shown) strongly stains the nuclear body in the presence or absence (not shown) of proteasome inhibitors.
We colocalized the focal staining of dNPpep with antibodies specific for defined cellular structures. The nuclear structures colocalized with those stained by a mouse mAb (Fig. 3) or human autoimmune antibodies (see Fig. 5) specific for proteins present in PODs. PODs are enigmatic 0.3–1.0-μm-diameter macromolecular complexes comprised of >20 different proteins that are attached to the nuclear matrix (Sternsdorf et al. 1997). The cytoplasmic structure surrounded the staining obtained with a mAb specific for γ-tubulin, which identifies the pair of centrioles present at the microtubule organizing center (MTOC), the site where microtubules originate.
The accumulation of dNPpep in PODs and the MTOC is not simply the result of prolonged overexpression. If cells were infected with rVV for 2–4 h in the absence of proteasome inhibitors, dNPpep was detected by anti-NH2 antibody staining in PODs and the MTOC as early as 30 min after adding zLLL in a small percentage of cells (data not shown), and by 90 min, in a high percentage of cells (see Fig. 8). In both circumstances, this represents the rapid accumulation of newly synthesized dNPpep, since it did not occur if protein synthesis inhibitors were added with zLLL.
dNPpep Present in PODs and MTOC Is Resistant to Detergent and High Salt Extraction and Present in High Mr Ubiquitinated Forms
PODs can be partially purified by progressive extraction designed to isolate the nuclear matrix (Staufenbiel and Deppert 1984). This was performed biochemically and cytoimmunochemically. rVV-infected 143B cells incubated with zLLL were subjected sequentially to NP-40, DNase I, high salt, DNase I/RNAase, and then fixed with paraformaldehyde and examined using the LCSM after staining with anti-NH2 and anti-POD antibodies. Under these conditions dNPpep was easily detected in PODs and the MTOC (Fig. 4 C, arrow), whereas the low level staining of remaining NPpep was in a pattern not clearly related to PODs.
Material recovered in the supernatant at each step of the extraction procedure (and a final step with Empigen BB to extract nuclear matrix–associated proteins, including those in PODs) was characterized by Western blotting using the anti-NH2 antibodies. Almost all of the NPpep expressed in the presence (Fig. 4 A) or absence (data not shown) of zLLL was recovered in the first three steps of the fractionation process. The major species recovered migrated with the expected mobility (arrowhead). In addition, faster migrating species were present that were probably generated by proteolysis during the extraction procedure.
In the absence of zLLL, the small amounts of dNPpep that were present behaved similarly to NPpep, being recovered in the first and third steps of the fractionation (Fig. 4 A). In the presence of zLLL, similar amounts of dNPpep were recovered in these fractions, but now a large amount of dNPpep was recovered from the final Empigen BB extraction step. Notably, in addition to dNPpep that migrated with the expected mobility (arrowhead), several lower mobility species were recovered at this stage, as well as the higher mobility species that probably represent proteolytic fragments.
The nature of the lower mobility forms of dNPpep was examined by Western blotting of whole cell lysates of zLLL-treated HeLa cells infected with rVVs expressing dNPpep, NPpep, or a control protein (OVA). Using anti-NH2 antibodies, the ladder-like nature of the higher Mr forms of dNPpep could be easily appreciated (Fig. 4 B). By contrast, only a few higher Mr forms were specifically detected (and at much lower levels) in NPpep-expressing cells (compare to OVA-expressing cells). Calculation of the Mrs of the lower mobility dNPpep bands revealed an 8.1-kD difference, close to the expected Mr of Ub (8.5 kD), indicating that dNPpep is ubiquitinated. In the same experiment we used an antiserum from a rabbit immunized with a synthetic peptide comprising the COOH terminus of NP. This reacted strongly with unmodified dNPpep or NPpep, but failed to detectably bind to any of the higher Mr forms of dNPpep or NPpep. As there are no Lys residues in the COOH-terminal NP peptide to serve as targets for ubiquitination, the inability of the antibody to bind the higher Mr forms may be due to either cleavage of a short segment of the COOH terminus, or to steric effects of ubiquitination on antibody access to the COOH terminus.
Based on these findings we conclude first that the bulk of dNPpep that is normally degraded by proteasomes accumulates in TX100/NP-40–insoluble forms concentrated in PODs and the MTOC when proteasomes are inhibited, and second, that a fraction of this material is present in modified high Mr forms resulting at least in part from ubiquitination. Due to uncertainties associated with efficiencies of recovering and detecting antigens in Western blots, the ratio of ubiquitinated to nonubiquitinated dNPpep cannot be determined by this method. It is worth noting, however, that at least some of the loss in the total amount of [35S]Met labeled dNPpep in LC-treated cells recovered over the 2-h chase period (Fig. 2 b) is due to ubiquitination with its attendant alteration in electrophoretic mobility.
Transport of dNPpep to PODs/MTOC in the Absence of Proteasome Inhibitors
The failure to detect dNPpep in PODs/MTOC in the absence of proteasome inhibitors raises two possibilities: the delivery of dNPpep to PODs/MTOC occurs only when proteasomes are blocked; and dNPpep is delivered to PODs/MTOC in the absence of proteasome inhibitors but is degraded too rapidly for cytochemical detection.
To address this issue, we labeled rVV-infected cells expressing NPpep or dNPpep with [35S]Met for 5 min, chased for 40 min, sequentially fractionated cells as above, and analyzed the fractions by SDS-PAGE, again taking advantage of the shut down of host proteins to visualize NP and dNPpep in an antibody-independent manner (Fig. 4 D). Consistent with the prior results, more NPpep (∼2.3-fold) is present in total cell lysates than dNPpep, and the bulk of NPpep is recovered in the first three fractionation steps. By contrast, less dNPpep is recovered from the DNase and high salt extracts, whereas its recovery in the Empigen BB extraction step is enhanced approximately sevenfold relative to NPpep. This finding is consistent with the idea that dNPpep is delivered to PODs in the absence of proteasome inhibitors.
Effects of zLLL and dNPpep Expression on the Distribution of Cellular Constituents of the Degradation Machinery
We next examined the effects of zLLL and dNPpep expression on the distribution of cellular proteins in rVV-infected cells expressing dNPpep or dNPpep with GFP added to the COOH terminus, as well as cells transiently transfected with plasmids encoding the GFP-fusion proteins. dNPpep was detected using either antibody staining of fixed and permeabilized cells or by GFP autofluorescence in fixed or live cells. Similar results were obtained from each of these vector/detection systems, demonstrating that the findings are not limited to VV-infected cells, and that the intracellular localization of dNPpep by antibodies is not biased by fixation/permeabilization/penetration artifacts. In the interest of brevity, results will be shown only for rVV-expressed dNPpepGFP. PODs were detected using sera from patients with PBC which contain Abs to PML and other POD proteins, poly Ub with a mAb (clone FK2) nonreactive with free Ub (Fujimuro et al. 1994), proteasomes using a mixture of two mAbs (MCP20 and 21) reactive with native proteasome subunits (Hendil et al. 1995), and molecular chaperones with various monoclonal and polyclonal antibodies.
In both live and fixed dNPpepGFP-expressing cells treated with zLLL for 4 h, fluorescent GFP was highly concentrated in PODs and the MTOC (Fig. 5, arrows point to the MTOC). The autofluorescence of dNPpepGFP indicates that the GFP domain is properly conformed, demonstrating that dNPpepGFP need not be completely denatured to localize to these structures. The accumulation of dNPpepGFP in these sites was accompanied by recruitment of poly Ub, proteasomes, and HSC70 from their normal diffuse distribution in the nucleus and cytoplasm, often to the extent that staining was reduced elsewhere in the cell (compare to Fig. 6; the distribution of proteasomes, not shown, is similar to poly Ub). While dNPpep and poly Ub filled the MTOC, in many cells proteasomes and HSC70 formed a ring around MTOC. The redistribution of cellular proteins is a specific effect of inhibiting proteasomes, as similar results were obtained with LC (data not shown). A survey of mAbs specific for other molecular chaperones (data not shown) revealed HSP27 recruited PODs similarly to HSC70 and somewhat less strongly to the MTOC, and HSP70 was recruited weakly to both sites. The distribution of a number of other cytosolic chaperones (HSP110, HSP90, HSP60, HSP56, HSP47, and HSP40) was not noticeably affected by dNPpep expression, and none were concentrated in either PODs or the MTOC.
In parallel experiments we examined the effects of zLLL on the distribution of the same cellular proteins in uninfected cells (Fig. 6), or VV-infected cells expressing NPpep, NPpepGFP, or GFP (data not shown). Infection with these rVVs had no major effects on the distribution of cellular proteins in untreated or zLLL-treated cells. In untreated uninfected cells, low levels of poly Ub were concentrated in a few PODs in some cells, but neither HSC70 nor proteasomes (data not shown) were concentrated in PODs. None of these cellular proteins were concentrated in the MTOC. After 6 h of zLLL treatment, poly Ub and HSC70 were clearly recruited to PODs and the MTOC. We did not detect proteasome recruitment to either the MTOC or PODs after zLLL treatment (not shown).
These data demonstrate that exposure of cells to proteasome inhibitors results in the accumulation of HSC70, HSP27, and poly Ub at PODs and the MTOC. Expression of dNPpep in the presence of proteasome inhibitors accelerates and enhances these effects, and also results in recruitment of proteasomes to these structures.
Subcellular Localization of Altered Conformational States of dNPpep
The conformational status of various forms of NP in cells was examined using other NP-specific antibodies. IC5-1B7 and HB65 are NP-specific mAbs that react with native NP and do not react with SDS-denatured NP in Western blots or when denatured virus is adsorbed to polyvinyl (Yewdell et al. 1981; Yewdell, J., unpublished results). In cells expressing NP or NPpep, IC5-1B7 and HB65 colocalized nearly perfectly with anti-NH2 antibodies, and the intensities of staining with the mAbs and the polyclonal serum were closely parallel, indicating that NP detected by the anti-NH2 antibodies is largely in a folded conformation by this criterion (data not shown). The small quantities of dNPpep present in cells not treated with proteasome inhibitors stained equally with the mAbs and the anti-NH2 antiserum, indicating that conformed molecules are preferentially spared from degradation (data not shown).
In dNPpep-expressing cells incubated with proteasome inhibitors, the mAbs failed to stain PODs, while intensely staining the MTOC (Fig. 3). This indicates that dNPpep rescued by proteasome inhibitors exists in multiple conformations, and that most or all dNPpep in PODs is at least partially unfolded.
In additional experiments (data not shown), we studied the intracellular distribution of NP constructs using the anti-COOH antiserum for immunofluorescence. When tested against VV-NP– or VV-NPpep–infected cells, staining with this serum closely paralleled staining with HB65 or IC5-1B7 as detected by double immunofluorescence. When used to stain dNPpep rescued by proteasome inhibitors, it strongly stained both the MTOC and PODs. Since this antiserum does not bind to ubiquitinated forms of dNPpep (Fig. 4 B), this extends the biochemical data to demonstrate that nonubiquitinated dNPpep is present in PODs and the MTOC.
Extension of Findings to Other Forms of NP
We examined the behavior of two other forms of rapidly degraded PR8 NP, one consisting of the first 168 residues of the protein (NP1-168), the other full length NP with amino acid substitutions at residues 148 (Y→H) and 282 (G→R) (NPDM). Both colocalized to PODs and the MTOC in a proteasome inhibitor–dependent manner as demonstrated using anti-NH2 Abs, and recruited the same array of cellular proteins as dNPpep (data not shown). In contrast to dNPpep, NPDM was detected in PODs by the HB65 mAb, demonstrating that a more conformed form of NP can localize to PODs.
It was even possible to induce wild-type NP to localize to PODs and the MTOC by exposing VV-NP–infected to cells to canavanine, an amino acid analogue of Arg that induces protein misfolding (Fig. 7). NP in PODs and the MTOC was detected by anti-COOH but not anti-NH2 Abs, possibly due to the replacement of Arg in the NH2-terminal peptide with canavanine (the COOH peptide used for Ab generation does not contain Arg). In addition, large amounts of NP were now detected in the cytosol, an effect possibly related to canavanine modification of the nuclear localization signal. NP in PODs and MTOC recruited poly Ub, proteasomes, and HSC70. Notably this occurred in the absence of proteasome inhibitors, suggesting that the degradation machinery (which could also be affected by canavanine) was compromised under these conditions, either as a result of having to cope with vast quantities of proteins misfolded by incorporation of canavanine, or canavanine-induced modifications in the machinery.
Together, these findings indicate that, first, that the effects with dNPpep are not due to unique features of the protein conferred by the JAK1 sequence but are a general feature of misfolded PR8 NP, and, second, that similar effects can occur in the absence of proteasome inhibitors.
Proteasome-dependent In Situ Degradation of dNPpep
The accumulation of dNPpep in PODs and the MTOC in proteasome-inactivated cells suggested that these structures serve as sites for proteasome-mediated destruction of dNPpep. To test this idea, TK− cells were infected with VV-dNPpep GFP for 4 h to maximize the rate of dNPpep translation, incubated with zLLL for 90 min to accumulate a small but detectable amount of dNPpep in a high percentage of cells, and then for 4 h in the absence of zLLL but in the presence of protein synthesis inhibitors to shut off additional dNPpep synthesis (Fig. 8). After 90 min in zLLL, dNPpep GFP was detected in PODs and the MTOC in most cells. Over the 4-h reversal period, dNPpepGFP nearly completely disappeared (similar results were obtained in other experiments in which dNPpep was detected using anti-NH2 Abs; data not shown). This process was dependent on active proteasomes, since dNPpepGFP persisted in similar quantities in PODs and the MTOC if cells were incubated with zLLL and the protein synthesis inhibitors. Based on these findings, we conclude that PODs and the MTOC serve as sources of substrates for proteasomes, and given the recruitment of proteasomes to these sites, are likely to represent sites of proteasome digestion.
We next related these findings to antigen processing. Kb was expressed in TK− cells by coinfection with a rVV expressing Kb and mouse β2-microglobulin, and the expression of cell surface Kb–Ova257-264 complexes quantitated cytofluorographically after indirect staining with the 25-D1.16 mAb (Fig. 9). Levels of background staining were controlled for by infection with VV-NP. As with L-Kb cells (Fig. 1), Kb–Ova257-264 complexes are produced more efficiently from dNPpep than NPpep; in this case the difference is even more pronounced (six- versus threefold increase in mean fluorescence). To correlate the disappearance of dNPpep from PODs and the MTOC with proteasome mediated generation of Ova257-264, cells were infected for 4 h in the continuous presence of zLLL, washed, and incubated for 4 h in the presence of protein synthesis inhibitors without (EC) or with zLLL (zLLL EC). In the continued presence of zLLL (zLLL EC) no complexes were generated from dNPpep or NPpep, since levels of staining of VV-NP–, VV-NPpep–, and VV-dNPpep–infected cells were identical. As described above (Fig. 1), zLLL had little effect on the generation of complexes by cells expressing the cytosolic minigene product, OvaM257-264. Removal of zLLL in the presence of protein synthesis inhibitors was accompanied by the generation of a signal in dNPpep-expressing cells above the staining of NP-expressing cells (EC). Although the shift in the curve is relatively small, it represents 31% of the signal obtained in the continuous absence of inhibitors (no inhibitor), and in absolute terms, roughly 1,000 Kb–Ova257-264 complexes, which is more than sufficient for triggering most T cells. The effectiveness of the protein synthesis inhibitors is clearly demonstrated by the background staining of cells treated with inhibitors from the initiation of the infection (EC start), even if cells were infected with the rVV expressing the cytosolic minigene product. In contrast to results with dNPpep, Kb–Ova257-264 complexes were not generated from NPpep upon removal of zLLL. These findings indicate that removal of zLLL from cells allows proteasomes to generate peptides from the dNPpep that accumulates in the cells (but not from NPpep), and is consistent with the idea the peptides (or their precursors) are generated at PODs, the MTOC, or at both of these locations.
We studied the fate of dNPpep as a model protein and class I–restricted antigen with a nuclear localization sequence that is ubiquitinated and degraded by proteasomes. Blocking proteasomal digestion results in the accumulation of dNPpep in a highly insoluble form, a portion of which is ubiquitinated. Immunofluorescence with a peptide-specific antiserum that does not detectably react with ubiquitinated substrate reveals that nonubiquitinated dNPpep is present at both the MTOC and PODs. The substrate-dependent enhanced recruitment of poly Ub to these structures clearly indicates that Ub-conjugated dNP is also present. This, together with the substrate-dependent recruitment of HSC70, which is required for the in vitro polyubiquitination of some proteins (Bercovich et al. 1997), is consistent with the following model: denatured dNPpep is chaperoned by HSC70 (and/or HSP27) to PODs and the MTOC, where it becomes an insoluble substrate for polyubiquitination and is degraded in situ by proteasomes. Several lines of evidence clearly indicate that complete denaturation of NP is not necessary for its delivery to PODs and the MTOC.
The idea that polyubiquitination of misfolded NP occurs at these sites is favored by several considerations. First, given the low solubility of ubiquitinated dNPpep, its destruction at its site of ubiquitination would bypass the need for special mechanisms to transport it to another cellular site for disposal. Second, the rapid diffusion of GFP-tagged proteasomes visualized in viable cells (Reits et al. 1997) is fully consistent with a “search and destroy” capability for proteasomes. Third, and most directly, we have found that prolonged treatment with proteasome inhibitors (>2 h) reduces the ability of cells to polyubiquitinate proteins (our unpublished results), presumably due to a decrease in the free Ub pool (Mimnaugh et al. 1997). Limiting levels of free Ub would account, first, for the failure of cells to completely ubiquitinate dNPpep at PODs and the MTOC, and, second, for the plateau observed in the level of unmodified dNPpep recovered from LC-treated cells after pulse radiolabeling (Fig. 2 b).
We believe that a subset of cellular proteins behaves similarly to dNPpep, since polyubiquitin and HSC70 are recruited to PODs and the MTOC of uninfected cells treated with proteasome inhibitors. Clearly, however, the effect is much less dramatic. This is not surprising given that in VV-infected cells infected for 2 h onwards, dNPpep represents ∼5–10% of all newly synthesized protein, and that most of it denatures rapidly after synthesis. This quantitative difference probably accounts for the failure of proteasome inhibitors to recruit proteasomes to PODs and the MTOC in cells not expressing dNPpep.
There is a considerable body of work relevant to these findings. Wojcik et al. 1996 reported that treating cells with proteasome inhibitors results in the accumulation of proteins (including proteasomes and Ub) at the MTOC: sufficient in fact to enable preferential staining of the MTOC in fixed cells with the protein stain amido black. Terming these structures “proteolysis centers,” they proposed that normal degradation of proteins occurs in this location. These findings were extended recently by Johnston et al. 1998, who demonstrated the presence of ubiquitinated proteasome substrates (misfolded forms of integral membrane proteins exported from the ER) at the same location (termed “aggresomes” by these authors) in the absence of proteasome inhibitors.
We extend these findings by demonstrating that: proteins with a nuclear localization sequence can also be degraded in proteolysis centers/aggresomes; proteins destined for destruction in the MTOC can retain at least portions of their native structure as indicated by binding to conformation-specific antibodies, or in the case of GFP fusion proteins, maintenance of autofluorescence; specific molecular chaperones are involved in the process; polyubiquitin may be added to substrates at this site; and substrates present in the MTOC are degraded by proteasomes.
The last four points apply also to PODs, providing the first conclusive evidence that these structures serve as a site for proteasomal degradation of ubiquitinated proteins, which as we argue above, is probably secondary to its serving as a site ubiquitination of denatured proteins. PODs have been implicated in a number of cellular processes, including tumorigenicity, apoptosis, and viral replication (Sternsdorf et al. 1997). Herpesviruses, adenoviruses, and papilloma viruses all encode proteins that localize to PODs, in some cases, disrupting the PODs. In addition to PML (the “P” in PODs), other examples of the 20+ known inhabitants of PODs include Sp100, HAUSP, and SUMO-1 (also known as PIC1). Significantly, both SUMO-1 and particularly HAUSP are related to the Ub-proteasome pathway. SUMO-1 is a Ub homologue that covalently modifies both PML and Sp100. Unlike Ub however, SUMO-1 modification appears to mainly affect the localization of its substrates and not their degradation, as both PML and Sp100 are localized to PODs only in their modified forms which are metabolically stable (as are other SUMO-1–modified proteins) (Saitoh et al. 1997). When used as an alternative for Ub, SUMO-1 is even known to prevent proteasomal degradation of proteins (Desterro et al. 1998). HAUSP is a Ub-dependent hydrolase, that removes Ub, but not SUMO-1 from substrates (Everett et al. 1998), and its presence in PODs is consistent with the idea that PODs serve as a center for protein ubiquitination and deubiquitination.
A chromosomal translocation characteristic for acute PML (APL) results in the creation of a fusion protein comprised of PML and the retinoic acid receptor α (RARα). The PML-RARα fusion protein acts as a dominant negative mutant, disrupting the integrity of PODs. Exposure of APL cells to retinoic acid returns the cells to a nontransformed phenotype concomitantly with the reformation of PODs and the degradation of the fusion protein (Daniel et al. 1993; Dyck et al. 1994; Weis et al. 1994). Alternatively, As2O3 treatment of normal or APL cells results in the recruitment of both PML and PML-RARα to PODs as well as their degradation (Zhu et al. 1997). The retinoic acid–induced degradation of PML-RARα is blocked by LC, implicating proteasomes in the process (Yoshida et al. 1996). Our findings are consistent with the idea that these proteins are degraded in PODs.
Everett et al. 1998 have shown that herpes simplex virus–induced destruction of PODs is mediated by the viral protein Vmw110 and is blocked by proteasome inhibitors. Vmw110 induces the proteasome-mediated destruction of PML and nuclear protein kinase. Vmw110 binds to HAUSP, but this is not required for its localization to PODs, or POD disruption of destruction of the kinase (Parkinson et al. 1999). These findings again support our conclusion that PODs serve as a general site of proteasome degradation, but Vmw110 probably induces proteasome degradation in multiple cellular sites, since Vmw110 mutants that do not localize to PODs can induce kinase degradation (Everett et al. 1999).
The involvement of PODs in the degradation of misfolded proteins also helps explain findings regarding mutant alleles of ataxin 1 that encode multiple copies of a polyglutamine domain present in the normal protein. These alleles are associated with a variety of inherited diseases of the nervous system. Ataxin 1 is normally present in small nuclear dots distinct from PODs. Mutant forms of ataxin 1 expressed in the absence of proteasome inhibitors are present in POD-like structures that contain PML and recruit Ub, proteasomes, HSP70, and HSP40 (Skinner et al. 1997; Cummings et al. 1998). It is uncertain to what extent ataxin versus other polyglutamine-containing proteins recruited into ataxin-initiated structures accounts for the recruitment of Ub, proteasomes, and chaperones (Perez et al. 1998). Our findings suggest that one (or more) of these misfolded proteins is recruited to PODs in association with HSP40 and HSP70, where it is polyubiquitinated but for some reason cannot be degraded by proteasomes, similar to what we describe for NP synthesized in the presence of canavanine.
Finally, we have linked the destruction of dNPpep in PODs and the MTOC to the generation of an antigenic peptide present in the protein. Given that all protein synthesis occurs in the cytosol, and that nuclear and ER proteins are commonly transported to the cytosol for degradation (Ciechanover 1998), the MTOC is probably a more general site of proteasome-mediated peptide generation, whereas peptide generation at PODs is expected to be limited largely to the subset of proteins located in the nucleus. It should be noted that the inner portion of the nuclear membrane forms part of the ER, and that peptides generated in the nucleus would not necessarily need to be delivered to the cytosol to access TAP (the MHC-encoded transporter that delivers class I ligands to the ER).
Due to the low efficiency of antigen processing, we cannot be certain that peptides are generated from dNPpep accumulated at PODs/MTOC and not from lesser amounts of antigen present elsewhere in the cell. Given the function of the MTOC as proteolytic centers/aggresomes, however, it would be surprising if this were not a common site of peptide generation. Regarding PODs, there are several published findings that would support a role in antigen processing, perhaps even a specialized role in regulating the process. First, the expression of PML and other POD constituent proteins is enhanced by exposure of cells to interferons, which increase the expression of genes encoding class I molecules and the other dedicated components of the class I–processing pathway (Sternsdorf et al. 1997). Second, PML itself has been directly implicated in the regulation of antigen processing, as modifications in PML that disrupt PODs result in decreased transcription of antigen-processing genes (Zheng et al. 1998). Together with our findings, these observations suggest the following hypothesis: a signal emanating from ubiquitination/proteolysis occurring at PODs is involved in a positive feedback loop that regulates antigen processing gene transcription.
We are grateful to Drs. Bloch, Hendil, and Liang for their generous gifts of antibodies. Bethany Buschling provided outstanding technical assistance.
1.used in this paper: APL, acute PML; dNP, misfolded nucleoprotein; GFP, green fluorescent protein; LC, lactacystin; LCSM, laser confocal scanning microscope; MTOC, microtubule organizing center; NP, influenza virus nucleoprotein; NPpep, NP with SIINFEKL appended to the COOH terminus; OVA, ovalbumin; PBC, primary biliary cirrhosis; PML, promyelocytic leukemia; POD, PML oncogenic domain; RAR, retinoic acid receptor; rVV, recombinant vaccinia virus; Ub, ubiquitin; zLLL, cbz-LeuLeuLeucinal