Heat shock protein-chaperoned peptides but not free peptides introduced into the cytosol are presented efficiently by major histocompatibility complex I molecules.

The studies reported here bear on the events in the cytosol that lead to trafficking of peptides during antigen processing and presentation by major histocompatibility complex (MHC) I molecules. We have introduced free antigenic peptides or antigenic peptides bound to serum albumin or to cytosolic heat shock proteins hsp90 (and its endoplasmic reticular homologue gp96) or hsp70 into the cytosol of living cells and have monitored the presentation of the peptides by appropriate MHC I molecules. The experiments show that (i) free peptides or serum albumin-bound peptides, introduced into the cytosol, become ligands of MHC I molecules at a far lower efficiency than peptides chaperoned by any of the heat shock proteins tested and (ii) treatment of cells with deoxyspergualin, a drug that binds hsp70 and hsp90 with apparent specificity, abrogates the ability of cells to present antigenic peptides through MHC I molecules, and introduction of additional hsp70 into the cytosol overcomes this abrogation. These results suggest for the first time a functional role for cytosolic chaperones in antigen processing.

The studies reported here bear on the events in the cytosol that lead to trafficking of peptides during antigen processing and presentation by major histocompatibility complex (MHC) I molecules. We have introduced free antigenic peptides or antigenic peptides bound to serum albumin or to cytosolic heat shock proteins hsp90 (and its endoplasmic reticular homologue gp96) or hsp70 into the cytosol of living cells and have monitored the presentation of the peptides by appropriate MHC I molecules. The experiments show that (i) free peptides or serum albumin-bound peptides, introduced into the cytosol, become ligands of MHC I molecules at a far lower efficiency than peptides chaperoned by any of the heat shock proteins tested and (ii) treatment of cells with deoxyspergualin, a drug that binds hsp70 and hsp90 with apparent specificity, abrogates the ability of cells to present antigenic peptides through MHC I molecules, and introduction of additional hsp70 into the cytosol overcomes this abrogation. These results suggest for the first time a functional role for cytosolic chaperones in antigen processing.
Cellular proteins undergo degradation in the cytosol, and the resulting peptides are transported into the endoplasmic reticulum (ER), 1 generally through transporter associated with antigen processing (TAP). Within the ER, the peptides are charged onto MHC I molecules. One of the key unresolved questions in this scheme pertains to the mechanism through which peptides are channeled to the TAP or other transporters.
Although peptides are generated in the cytosol, there is little evidence that the cytosol harbors free peptides. It has been proposed that the peptides exist in association with peptidebinding proteins in the cytosol and the ER (1,2). Because heat shock proteins (HSPs) are known to chaperone a wide array of molecules (3) and because immunological and structural evidence exists that HSPs chaperone antigenic peptides (see Ref. 4 for review), it was suggested that HSPs are the peptidebinding proteins that transport peptides (1,2). This view has received little formal attention in the form of support or rejection, although no alternative mechanisms of peptide traffic have been suggested. Nonetheless, evidence has continued to accumulate that (a) HSPs are associated with peptides from a wide spectrum of antigens, including tumor antigens (5,6), viral antigens (7), model antigens (8 -10), and minor H antigens (8), and that (b) the repertoire of peptides associated with the HSP of the ER is dependent upon the functional status of TAP (9).
In this report, we address the issue functionally and ask if the chaperoning of peptides in the cytosol by HSPs confers on the HSP-chaperoned peptides any advantage not available to unchaperoned peptides in terms of their presentability by MHC I molecules.

EXPERIMENTAL PROCEDURES
Brefeldin A Treatment-EL4 cells were treated with brefeldin A (BFA) at two different concentrations in succession to respectively block the MHC I pathway of antigen presentation (6.0 g/ml for 3 h) and to maintain the block (0.6 g/ml for up to 12 h). Maintenance of the BFA block did not affect CTL function during the CTL assay. EL4 cells, untreated or treated with BFA at these concentrations, were analyzed by FACScan to show maximal decreases (ϳ40%) in surface expression of MHC I after 20 h (data not shown). BFA-treated cells were loaded with protein and used as targets in the CTL assay, as described, in the presence of BFA.
Cell Lines, Mice, and Reagents-The T-Ag-transformed cell lines SVB6 and PS-C3H were obtained from Prof. S. S. Tevethia and have been previously described (11) The VSVNP-transfected EL4 cell line, N1, was obtained from Dr. Lynn Puddington and has been previously described (12). EL4 cells and the TAP-dysfunctional cell line, RMA-S, were obtained from Prof. S. Nathenson. The RMA cell line has been previously described (13).
All chemicals were purchased from Sigma Chemical Co. unless otherwise specified. HL-1 and RPMI media, together with pyruvate, glutamine, penicillin-streptomycin, and non-essential amino acids were purchased from Life Technologies, Inc. RPMI containing 5% fetal calf serum (Intergen) and 1% each of pyruvate, glutamine, penicillin-streptomycin, and non-essential amino acids is subsequently referred to as complete RPMI.
Antibodies-HSPs were detected by immunoblotting with specific antibodies: gp96 (rat monoclonal antibody SPA-850, clone 9G10); cytosolic hsp70 (mouse monoclonal antibody SPA-820, clone N27F3-4 recognizes constitutive hsp73 and inducible hsp72); hsp90 (rat monoclonal antibody SPA-845, clone 1R2D12p90). All these antibodies were purchased from StressGen Biotechnologies Corp., Victoria, Canada. Anti-K b , anti-D b , anti-D d , or anti-LFA-1 (clones AF6-88-5, KH95, 34-2-12, and 2D7, respectively)-fluorescein-conjugated monoclonal antibodies were obtained from PharMingen (San Diego, CA). Cellular Loading of Proteins or Peptides-To prepare proteins (gp96, hsp70, hsp90, or SA; complexed or not) for loading, the indicated amount of protein was incubated with DOTAP (N-[-(2,3-dioleoyloxy)propyl]-N,N,N-trimethylammonium methylsulfate (C 43 H 83 NO 8 S)) (Roche Molecular Biochemicals) at a 3:2 ratio (microgram amounts) for 15 min at room temperature. In all loading experiments, 1.5 ϫ 10 6 cells (EL4, RMA, or RMA-S) were washed three times with serum-free HL-1 media and then incubated in 1 ml of HL-1 media with a protein:DOTAP combination for 4 -4.5 h at 37°C. Control cells were either mock-loaded by incubating 1.5 ϫ 10 6 cells in the same amount of DOTAP alone or were incubated with protein alone in the absence of DOTAP (pulsed cells). After loading (or mock loading for controls), cells were washed three times with HL-1 media and once with complete RPMI. Where indicated, loaded, mock-loaded, or pulsed cells were used as targets in CTL assays. Loading efficiencies of gp96, hsp70, hsp90, or serum albumin alone were the same, and inter-experimental values did not vary significantly. Free peptides were loaded into cells using the same protocol.
CTL Assays-CTL assays were carried out as follows. Briefly 2 ϫ 10 3 51chromium (supplied as Na 2 CrO 4 ; ICN) labeled target cells in 100 l of complete RPMI were added to various dilutions of T-Ag or VSV8specific CTL in 100 l of complete RPMI, as indicated by the effectorto-target ratios. Effectors and targets were cultured in v-bottomed 96-well plates for 4 h. Supernatants (50 l) were harvested, mixed with scintillation fluid, and counted in a 1450 MicroBeta Trilux liquid scintillation counter (Wallac Inc. Maximum and spontaneous releases were measured by culturing 2 ϫ 10 3 labeled target cells in lysis buffer (0.5% Nonidet P-40, 10 mM Tris, 1 mM EDTA, 150 mM NaCl) and complete RPMI, respectively, for 4 h. VSV8-specific CTL were obtained by dual immunizations of C57BL/6 mice, 1 week apart, with N1 cells. Spleen cells were harvested 1 week after the second immunization, restimulated in culture with irradiated N1 cells, and cloned by limiting dilution (14). The specificity of the CTL clone was tested by cold target inhibition and antibody blocking experiments. This CTL clone was shown to be specific for the VSV8 peptide (NH 2 -RGYVYQGL-COOH) bound to K b molecules. A similar strategy, with SVB6 cells, was used to obtain the T-Ag-specific CTL clone. This clone was shown to be specific for the 9-mer peptide (NH 2 -AIN-NYAQKL-COOH), previously named epitope 1 (11).
Flow Cytometry Analysis of DSG-treated Cells-N1 cells were irradiated (5000 rads) and allowed to recover in AIM V medium with or without DSG for 48 h at 37°C or 25°C. Half of the cells incubating at 25°C were then placed at 37°C for an additional 8 h. One group of the cells not treated with DSG but incubated at 25°C for 48 h was placed at 37°C in the presence of DSG for 8 h. Cells (1 ϫ 10 6 ) were then stained at 4°C for 40 min with anti-K b , anti-D b , anti-D d , or anti-LFA fluorescein-conjugated antibodies and analyzed on a FACScan flow cytometer purchased from Becton Dickinson (San Jose, CA).
Infection of Cells with Vesicular Stomatitis Virus-Vesicular stomatis virus (VSV) was obtained from Advanced Biotechnology (Columbia, MD). Meth A or EL4 cells were incubated with 10 plaque forming units of VSV per cell for 1 h at 37°C in plain RPMI and allowed to recover in RPMI with 10% fetal calf serum for 4 h. Cells were washed three times in PBS (10 mM phosphate buffer, 150 mM NaCl, 2.7 mM KCL, pH 7.4). gp96 was then purified from these cells as described below.
Peptides were synthesized by Genemed to Ͼ95% purity as determined by high pressure liquid chromatography. The unextended MHC I binding 9-mer peptide is identical to epitope I of the T-Ag (11). The T-Ag 9-mer stabilized MHC I molecules on RMA-S cells and sensitized targets for lysis by the T-Ag-specific CTL. All three T-Ag 20-mer peptides failed to bind MHC H-2D b as determined by their inability to stabilize empty MHC molecules on the surface of RMA-S cells and their inability to sensitize target cells for lysis by T-Ag-specific CTL.
Purified gp96 or hsp90 was incubated, at the indicated amount, with peptide at a protein-to-peptide molar ratio of 1:50 in 700 l of PBS for 10 min at 50°C and incubated for a further 30 min at room temperature. The peptide concentration used for complexing was 10 Ϫ6 M. Approximately 1% of the gp96 or hsp90 molecules was loaded with the exogenous peptides by this method (15). The indicated amount of purified hsp70 or SA was incubated with peptide at a protein-to-peptide molar ratio of 1:50 in 300 l of PBS at 37°C for 1 h. Peptide concentration used for complexing was 10 Ϫ6 M. To remove free, uncomplexed peptides, complexes were washed extensively with PBS in an Ultrafree-4 centrifugal device, Biomax 10K NMWL membrane (Millipore Corp.). To determine the efficiency of complexing, peptides were labeled with 125 I (ICN) using IODO-BEADs (Pierce). In parallel with unlabeled peptides, 125 I-labeled peptides were complexed to proteins and checked by SDS-polyacrylamide gel electrophoresis and autoradiography (data not shown). The efficiency of gp96, hsp70, hsp90, or SA to complex peptides was comparable.
Inhibition of Proteasome Function-EL4 cells (10 7  Purification and Identification of HSPs-hsp70 and gp96 were purified from cells according to previously described methods (16,17). hsp90 was purified according to the protocol of Denis (18) with minor modifications. Briefly, 100,000 ϫ g supernatants were obtained from cell lysates and applied to a Mono Q column (Mono Q HR 16/10, purchased from Amersham Pharmacia Biotech and attached to the BIOCAD, Perseptive Biosytems).
Treatment of MLTC and N1 Cells with 15-Deoxyspergualin-15-Deoxyspergualin (DSG) was a gift from Dr. S. Nadler at Bristol-Myers Squibb Co. (Wallingford, CT). Lyophilized DSG was dissolved in PBS and stored in aliquots at a concentration of 10 mg/ml at Ϫ130°C. Twenty micrograms per ml of DSG, with or without peptide (final concentration of 1 M), was added to the MLTC of VSV CTL clones. After a 5-day incubation at 37°C, each well of the MLTC was harvested and tested for its ability to lyse 51 Cr-labeled N1 and EL4 cells in a 4-h 51 Cr release assay.

Demonstration of the Experimental System to Introduce Molecules into the Cytosol-
The cationic liposome, DOTAP, was used to introduce HSP⅐peptide complexes or free peptides into the cytosol. Distinct properties of the detergents Nonidet P-40 and Saponin were used to demonstrate that DOTAP-loaded gp96 enters the soluble, non-vesicular, cytosolic compartment of the cells (Fig. 1). Although the cytosolic HSPs are of primary interest in this study, gp96 was used as a test case, because its distinct non-cytosolic localization (in the ER) permitted determination of the compartment into which the HSP⅐peptide complexes were being introduced, as will become clear from the following. The gp96/DOTAP-loaded cells were lysed with each of two detergents. Lysis of live cells with 0.5% Nonidet P-40 leads to solubilization of all non-nuclear membranes, whereas lysis with 0.01% Saponin results in solubilization of plasma membranes but not internal membranes (19). The lysates were centrifuged to obtain the solubilized components, which were analyzed for the presence of gp96 by immunoblotting: The internal resident gp96 is detected in the Nonidet P-40-lysates of non-loaded cells (Fig. 1, lane 1) but not in the Saponin-lysed non-loaded cells (Fig. 1, lanes 2 and 3), because gp96 is a luminal component of the ER compartment, which remains impervious to Saponin. gp96 is not detected in cells treated with DOTAP without gp96 (lane 2) or gp96 without DOTAP (lane 3). The only instance where gp96 is detectable in the Saponin-solubilized cells is if it has been introduced along with DOTAP into cells (lane 4), i.e. from an exogenous source. As additional controls, all samples tested predictably positive for the cytosolic chaperone hsp70 (Fig. 1, bottom panel, lanes 1-4). Thus, DOTAP-mediated delivery of gp96 (and by deduction other proteins) into cells introduces them into the cytosolic compartment. Similar results were obtained with introduction of labeled peptides by DOTAP. Quantitative analysis of exogenously introduced radiolabeled proteins through DOTAP indicated that ϳ5% of the DOTAP-loaded protein is introduced into the cytosol and that Ͼ96% of this 5% is detected in a soluble, non-vesicular, cytosolic compartment of the cells (data not shown).
HSP-chaperoned Peptides Introduced into the Cytosol Become Ligands for MHC I-As discussed in the previous section, the cytosolic chaperones hsp90 and hsp70 are of primary interest for the studies described all through this report. However, the ER chaperone gp96 was also used in all studies, primarily because (i) gp96 was used for the demonstration that DOTAP introduces proteins into the cytosol, (ii) gp96 is highly homologous (protein sequence homology of 50%) (see Ref. 1) to the cytosolic chaperone hsp90, and (iii) considerable immunological and structural information on gp96-peptide interaction is al- gp96, purified from the T-Ag-transformed cell line SVB6, and chaperoning T-Ag-derived peptides was loaded into EL4 cells by DOTAP. Presentation of T antigen-derived peptides by MHC I molecules of EL4 cells was monitored by specific lysis of DOTAP-loaded cells using a CTL clone specific for epitope 1 of the T-Ag (11) as described under "Experimental Procedures." The T-Ag-derived peptides were present in gp96 preparations and were observed to be efficiently re-presented in this assay ( Fig. 2A). Percentage of lysis of the loaded cells increased with increasing amounts of gp96 loaded into the cells. No lysis was observed with less than 25 g of gp96. In parallel control experiments, EL4 cells were pulsed, in the absence of DOTAP (as opposed to loaded) with gp96, to determine if there is extracellular exchange of peptides between gp96 and surface MHC I molecules on EL4 cells. No surface charging was detected.
SVB6-derived hsp70 or hsp90 preparations were also loaded into EL4 cells in increasing doses. Antigen-specific recognition of the loaded cells by CTLs was observed when either hsp70 or hsp90 was loaded ( Fig. 2A), indicating that similar to gp96, hsp70 or hsp90 donate their chaperoned peptides to become ligands for MHC I molecules. Again, lysis of cells was dependent on the amount of HSP loaded by DOTAP. Although peptides chaperoned by all three HSPs could become ligands for re-presentation by MHC I molecules, the efficiency of doing so was different for each HSP. For comparable lysis (ϳ40%) of HSP-loaded cells, 100, 250, and 500 g of gp96, hsp70, and hsp90, respectively, were required. Approximate amounts of HSP, below which no lysis was detected were 25, 180, and 400 g for gp96, hsp70, and hsp90, respectively. A second, well characterized antigenic system, the Vesicular Stomatitis Virus (VSV) system, was used to test the generality of the observation in the T-Ag system. VSV nucleoprotein (VS-VNP) derived peptides chaperoned by gp96 or hsp70 (purified from the VSVNP-transfected cell line N1 (12) are effectively re-presented and recognized by VSVNP-specific CTL after the respective HSPs are introduced into the cytosol of EL4 cells by DOTAP (Fig. 2B). To demonstrate that lysis by VSVNP-specific  2-4). Supernatants of the cell lysates (100,000 ϫ g, 90 min) from each sample were resolved by SDS-polyacrylamide gel electrophoresis, and the blots for each sample were probed with antibodies against gp96 or the cytosolic hsp70.
FIG. 2. Peptides chaperoned by HSPs are re-presented by EL4 cells after cytosolic loading into the cytosol. A, gp96, hsp70, or hsp90 purified from the T-Ag-transformed cell line SVB6 was loaded into EL4 cells at different doses as indicated. Closed crosses indicate EL4 cells pulsed with HSPs without DOTAP. Cells were used as targets for T-Ag-specific CTL clones in a 51 Cr release assay. B, gp96 or hsp70 purified from the VSVNP-transfected cell line N1 or EL4 cells was loaded into EL4 cells at different doses as indicated. Open crosses indicate EL4 cells pulsed with HSPs without DOTAP. ATP treated hsp70 indicates the N1-derived hsp70 preparation treated with ATP to remove peptides. Loaded cells were used as targets for VSVNP-specific CTL clones in a 51 Cr release assay. CTL, of cells loaded with HSPs, is peptide-dependent, equivalent amounts of peptide-free hsp70, obtained by ATP treatment of N1-derived hsp70 preparations (15), were delivered into EL4 cells. No lysis of EL4 cells loaded with peptide-depleted hsp70 preparations was observed (Fig. 2B). Furthermore, HSP preparations not carrying VSVNP-derived peptides (EL4-derived HSPs) (Fig. 2B), did not render loaded cells susceptible to VSVNP-specific CTLs, with any amount of HSP loaded. The results imply that presentation and consequent cell lysis are both peptide-dependent and -specific and require intracellular processing of the HSP⅐peptide complexes.
HSP-chaperoned Peptides Are Presented Ͼ100-fold More Efficiently Than Free Peptides-It is difficult to monitor and quantify presentation of specific antigenic peptides in naturally derived HSP⅐peptide complexes. To quantitate the efficiency of re-presentation of specific HSP-chaperoned peptides, HSPs reconstituted in vitro with known quantity of antigenic peptides or their extended versions were used. The D b -restricted 9-mer epitope I of the SV40 T-Ag protein (NH 2 -AINNYAQKL-COOH), or 20-mer peptides extended on the NH 2 terminus, COOH terminus, or both termini (Fig. 3A) were complexed to HSPs gp96, hsp90, or hsp70, or a control peptide-binding protein serum albumin (SA) (15 and "Experimental Procedures"). Peptides thus complexed (ϳ10 Ϫ8 M with respect to peptide concentration) or free peptides (10 Ϫ6 or 10 Ϫ4 M) were loaded into EL4 cells with DOTAP. In parallel, experiments using radiolabeled HSPs, SA, and each of the peptides were used to determine how much of each moiety administered with DOTAP could be recovered in the cytosol of the cells. This exercise demonstrated that ϳ6 -8% of the quantity of each moiety introduced in the cells by DOTAP could be recovered from the cytosol (data not shown). The constancy of this number allows for valid comparisons among the results with each antigenic moiety. The cells into which the HSPs, SA, or peptides were introduced were then monitored for lysis by T-Ag-specific CTLs (Fig. 3B). It was observed that (i) a concentration of 10 Ϫ4 M free peptide was required for loading to observe lysis of the EL4 cells comparable to that observed for 10 Ϫ8 M concentration of peptide when FIG. 3. T-Ag-derived peptides chaperoned by HSPs, but not by SA or unchaperoned, can be re-presented. A, 20-mer peptides extended from the precise MHC I H-2D b binding epitope (gray) on the NH 2 , COOH, or both termini were synthesized and complexed to gp96, hsp90, hsp70, or SA. B, peptides complexed to gp96, hsp90, hsp70, or SA or unchaperoned peptide were loaded into EL4 cells. Control EL4 cells were pulsed with these complexes without DOTAP. Loaded or pulsed cells were used as targets for T-Ag-specific CTL in a 51 Cr release assay. chaperoned by HSPs, (ii) peptides chaperoned by SA, which binds peptides efficiently ("Experimental Procedures"), were not re-presented by MHC I molecules, suggesting that HSPs play a role different from simply carrying the peptides, and (iii) MHC I epitopes are generated from peptides chaperoned by HSPs regardless of whether they are extended on the NH 2 , COOH, or both termini.

Re-presentation of HSP-chaperoned Peptides Requires Functional Proteasomes, Is TAP-dependent, and Is Brefeldin A-sensitive-
The cytosolic proteasomes have been implicated as the primary producers of peptide ligands for MHC I molecules (for review see Refs. 20, 21). Because DOTAP-mediated loading of cells with the HSP⅐peptide complexes results in presentation the peptides by MHC I, we tested the requirement for proteasomal activity for re-presentation of HSP-chaperoned peptides. Because HSPs are purified from cells after the peptides have been generated through protease activity and also have been shown to chaperone precise MHC I peptide epitopes (6, 7, 10), we expected that re-presentation of HSP-chaperoned peptides would not require further proteasomal action. EL4 cells were treated with the proteasome inhibitor, N-acetyl-Leu-Leu-norleucinal (LLnL) for 1 h prior to and during loading with either the endoplasmic reticulum (ER) chaperone gp96 or the cytosolic chaperone hsp70 derived from N1 cells. Surprisingly, re-presentation of VSVNP peptides chaperoned by gp96 or hsp70 was inhibited by LLnL (Fig. 4A), suggesting that re-presentation of HSP-chaperoned peptides requires functional protease activity. Control, LLnL-untreated EL4 cells loaded with gp96 or hsp70 in an identical manner were able to re-present VSVNPderived peptides.
Because LLnL has been shown to have inhibitory effects on proteases other than the proteasome (22), we replaced LLnL with the proteasome-specific inhibitor, lactacystin. To examine the proteasome dependence of HSP-chaperoned peptide re-presentation more precisely, we used HSP⅐peptide complexes reconstituted in vitro instead of the naturally derived complexes. The four T-Ag-derived peptides used earlier (Fig. 3A) were complexed separately to gp96, hsp70, hsp90, or the non-HSP, SA. HSP⅐peptide complexes, reconstituted in vitro, were loaded independently but identically, into EL4 cells, not treated or treated with lactacystin prior to loading. It was observed ( Table  I) that (i) treatment with lactacystin inhibited re-presentation of all the extended peptides, (ii) surprisingly, treatment of cells with lactacystin inhibited re-presentation of even the precise unextended MHC I-binding peptides when chaperoned by hsp70 or hsp90; (iii) in another surprise, re-presentation of the precise MHC I binding peptide complexed to gp96 was not inhibited by lactacystin. These observations suggest that, during re-presentation, proteasomes may contribute function(s) other than proteolytic degradation of extended peptides. They also suggest that peptides chaperoned by the ER HSP, gp96, are processed by a different mechanism from that of peptides chaperoned by the cytosolic hsp70 and hsp90. The structural basis for this difference is not yet clear.
Peptides generated in the cytosol are transported predominantly by TAP into the endoplasmic reticulum for association with MHC I molecules (23)(24)(25)(26)(27). The requirement for TAP in re-presentation of HSP-chaperoned peptides was tested by comparing peptide re-presentation in TAP-expressing cells RMA and in TAP-dysfunctional cells RMA-S. RMA-S cells were FIG. 4. Re-presentation of peptides chaperoned by HSPs requires functional proteasomes, is TAPdependent, and is brefeldin A-sensitive. A, EL4 cells were untreated or treated with N-acetyl-Leu-Leu-norleucinal prior to loading with N1-derived gp96 or N1-derived hsp70 preparations at the indicated concentration. Loaded or unloaded cells were used as targets in a 51 Cr release assay with VSVNP-specific CTL as described. B, RMA or RMA-S cells were loaded or pulsed with N1-derived gp96 or hsp70. RMA-S cells, pulsed with VSV8, and RMA were also used as controls. VS-VNP-specific CTL were used as the effectors. C, EL4 cells untreated or treated with brefeldin A were loaded with gp96, hsp70, or hsp90 derived from SVB6 cells. Brefeldin A-treated EL4 cells, allowed to recover for 3 h, were also loaded in parallel. Loaded cells were used as targets for T-Ag-specific CTL in a 51 Cr release assay. not lysed by VSVNP-specific CTL after being loaded with N1derived gp96 or hsp70 at any dose of HSP used (Fig. 4B). In comparison, RMA cells, expressing functional TAP molecules, did re-present the HSP-chaperoned peptides as measured by the effective lysis of HSP-loaded RMA cells (Fig. 4B). RMA cells pulsed with HSPs in the absence of DOTAP were not susceptible to lysis, indicating that lysis was not due to extracellular exchange of peptides.
A requirement for TAP for presentation of VSVNP may appear inconsistent with the earlier findings of Bevan and colleagues (28) who showed that VSVNP can be presented by MHC I in the absence of functional TAP2 molecules in RMA-S cells. However, a closer scrutiny of the previous data and our results shows that the differences are not inconsistent. Essentially, TAP2-negative cells such as RMA-S can still re-present VSVNP, whereas TAP1-negative cells cannot, thus suggesting that TAP1 homodimers may still be able to transport peptides into the ER. It is conceivable that, under limiting quantities of antigenic peptides, such as those created by introduction of HSP⅐VSVNP complexes, the relative efficiencies of the TAP1/ TAP2 heterodimer vis-à -vis the TAP1 homodimer, become more evident. A second possibility may be envisaged where the VSVNP peptides generated in N1 cells are transported by anomalous TAP-independent means, whereas direct introduction of the same peptides with HSPs introduces them into the classical TAP-dependent pathway.
After MHC I molecules are loaded with peptides in the ER, they are transported to the cell surface via the Golgi by vesicular traffic. Brefeldin A (BFA) is a known inhibitor of post-ER vesicular traffic (29). EL4 cells were not treated or treated with BFA for 1 h prior to and during DOTAP-mediated loading of SVB6-derived gp96, hsp70, or hsp90. The loaded cells were then tested for lysis by T-Ag-specific CTLs. It was observed that BFA completely inhibited re-presentation of peptides chaperoned by gp96, hsp70, and hsp90 (Fig. 4C). This inhibition was reversible by incubating BFA-treated EL4 cells in the absence of BFA for 3 h prior to loading with HSP⅐peptide complexes (Fig. 4C).

Sequestration of Endogenous Cytosolic HSPs Abrogates Presentation of Antigenic Peptides by MHC I Molecules-Deox-
yspergualin (DSG) is a small molecular weight immunosuppressive drug shown to interact specifically with hsp70 and hsp90 (30,31). The drug can enter cells and interact with endogenous hsp70. We sought to exploit the HSP-binding property of DSG to test whether binding of DSG will lead to sequestration of hsp70 and hsp90, which will now be unable to chaperone the newly generated antigenic peptides into the endogenous presentation pathway. This idea was tested in a series of experiments. As a first measure, DSG was added to mixed lymphocyte tumor cultures (MLTC) of N1 cells (EL4 cells transfected with the gene encoding VSV NP (12)) and anti-N1 CTL clones. The MLTCs generated in the presence and absence of DSG were tested in a cytotoxicity assay for activation and proliferation of antigen-specific CTLs. Treatment with DSG was observed to inhibit dramatically the activation/proliferation of VSV NP-specific CTLs (Fig. 5A). However, this inhibition could be reversed completely if VSV NP-derived peptide VSV8 were added to the MLTC. Addition of an irrelevant peptide (corresponding to an epitope from SV40 T antigen with the same restriction element as the VSV epitope), did not reverse the inhibition. These data indicate that treatment with DSG resulted in a limitation in the quantity of the VSV epitope on N1 cells. To determine if DSG was acting at the level of the CTLs or the antigen-presenting cell, the CTLs were purified and were cultured in medium with or without DSG and were tested for their ability to lyse N1 cells. DSG treatment for as long as ϳ100 h had no discernible effect on the CTLs (Fig. 5B).
The effect of DSG on N1 cells was monitored directly. As N1 cells already contain a population of specific MHC I⅐peptide complexes, which have a certain half-life, and because even a very small number of MHC I⅐peptide complexes are capable of stimulating activated CTLs (32), a system was sought where no preformed specific MHC I⅐peptide complexes exist. EL4 cells were treated (or not treated) with DSG for 24 h so as to allow sequestration of hsp70 and hsp90 molecules. Cells were then infected with VSV. The virus-infected and viral antigen-expressing cells, which had or had not been exposed to DSG pre-infection, were used to stimulate anti-VSV NP CTLs, as described in a previous experiment (see Fig. 5A). The DSGtreated cells were observed to be unable to stimulate the CTLs at all, whereas the control cells stimulated them as expected (Fig. 6A). As an additional control in these studies, EL4 cells,

Influence of inhibition of functional proteasomes by lactacystin on representation of precise and extended peptides chaperoned by HSPs
Data are percent inhibition of re-presentation. EL4 cells were treated with lactacystin prior to loading with HSP ⅐ peptide complexes reconstituted in vitro as described under "Experimental Procedures." Representation was monitored in a 51 Cr release assay by a T-Ag-specific CTL clone. Data were obtained in a cytotoxicity assay done at an effector:target ratio of 5:1. Percent inhibition was calculated, assuming as 100% the specific cytotoxicity obtained in the absence of lactacystin. This was approximately 60%. with or without prior treatment with DSG, were pulsed with VSV8, and these were tested for the ability to stimulate CTLs. Treatment with DSG was found to have no effect on the antigen-presenting ability of VSV8 pulsed cells (Fig. 6A). Prior treatment of cells with DSG had no effect on viral infection and expression of viral proteins as determined by staining of infected cells with anti-G protein antibody coupled to a photochrome (Fig. 6B). These results show clearly that treatment with DSG interferes with a step in the antigen-presenting cell, which is required for generation of the specific MHC I⅐peptide complex, although the block is not in generation of MHC I molecules per se.
Although DSG has been shown to interact specifically with hsp70 and hsp90 (30), the possibility that the effects observed are not due to the HSPs but due to interaction of DSG with an unknown intracellular pathway, the role of hsp70 was tested more directly. Experiments shown in some of the previous figures (Figs. 1-3) demonstrate how it is possible to introduce molecules into the cytosol of living cell with the help of DOTAP. This method was now used to introduce hsp70, or as a control, SA (which has been shown previously to bind peptides efficiently (15)), into the DSG-treated cells 1 h after infection with VSV. The cells were used to stimulate the anti-VSV CTLs as before. The experiment showed (Fig. 7) that introduction of hsp70 but not SA could completely relieve the inhibition in antigen-presenting ability in DSG-treated cells.
The ability of DSG to block the trafficking of peptides destined for loading the cell surface MHC I molecules was tested by an independent assay. MHC I⅐␤2 microglobulin complexes devoid of peptides are unstable on the cell surface at 37°C but are stable at 25°C (33). The MHC I⅐peptide complexes can also be detected by conformation and K b -specific antibodies. These tools were used to examine the presence of stable MHC I molecules on the cell surface of DSG-treated and untreated cells. It was observed that treatment of EL4 cells with DSG at 37°C lead to a nearly 5-fold reduction in the number of K bpeptide complexes as determined by the specific antibody Y3 (Fig. 8). At 25°C, no such inhibition was observed. Interestingly, if the EL4 cells kept at 25°C were now shifted to 37°C, the DSG-treated cells showed a nearly 5-fold less quantity of K b moieties than the DSG-untreated cells, indicating that a large proportion of K b molecules of DSG-treated cells at 25°C were devoid of peptides and hence labile in the DSG-treated cells. EL4 cells also express the LFA molecule whose expression in DSG-treated and untreated cells was monitored and found to be unaffected by treatment with DSG, indicating that DSG was not affecting the secretory pathway per se, as also indicated by the experiment carried out at 25°C.
These observations provide strong support for the idea that HSPs are necessary for transport of antigenic peptides in the cytosol and that DSG interferes with this step. DISCUSSION The studies reported here shed light on trafficking of peptides in the cytosol, leading to presentation of peptides by the MHC I molecules. First, free peptides introduced into the cytosol are presented quite inefficiently as compared with HSPchaperoned peptides. This observation supports the idea that the trafficking of peptides in the cytosol does not occur by passive diffusion but by active mechanisms, including chaperoning (1,2,4). This is particularly relevant, because of quantitative considerations. The quantity of peptides available naturally in a cell (in the order of sub-femtograms/cell for an epitope derived from a moderately expressed protein) is too low to allow presentation by MHC I molecules, if the peptide were to diffuse passively. The same quantity of peptide has a significantly higher likelihood of getting presented if it were chaperoned by an HSP molecule, as shown here. Second, chaperoning (i.e. being carried by a larger molecule) is necessary but insufficient for peptide re-presentation, because SA-chaperoned peptides are not presented any more efficiently than free peptides. The structural rules that define the requirement for and efficiency of chaperoning in presentation for each HSP, such as binding affinities, number of peptide binding sites, peptide-dissociation rates, association with other peptide-recipient proteins are not yet known; however, our studies provide an assay through which they could be divined. These structural rules may account, at least partially, for the differences in proteasome requirements observed for hsp90 and hsp70 versus gp96 (Table I). In addition, because hsp70 and hsp90 are cyto-

FIG. 6. Pretreatment of cells with DSG blocks antigen presentation by MHC I molecules.
A, EL4 cells were irradiated and incubated without (open symbols) or with 40 g/ml DSG (closed symbols) for 24 h. EL4 cells were washed prior to infection with VSV (squares) or pulsing with VSV8 (circles) and were tested for their ability to be recognized by VSV CTLs in a 4-h 51 Cr release assay. The asterisk denotes background lysis of EL4 cells in the absence or presence of DSG. B, pretreatment of cells with DSG does not effect infection of cells by VSV. EL4 cells were fixed, permeabilized, and incubated with anti-VSV G protein antibody conjugated to cyt3 photochrome. Cells were analyzed by a confocal fluorescence scanning microscopy (Zeiss LSM 410 invert). Dual channel imaging was performed, and both images were overlaid. Channel 1 (green), transmittance image; channel 2 (red), fluorescence image. solic proteins, their delivery, with DOTAP, places them in the appropriate environment to mediate their chaperoning function. In contrast, the normally ER-resident gp96 may direct its peptides into an alternate route when delivered into the cytosol, for example, by docking with different peptide-recipient proteins. Third, and complementary to the first two sets of results, our studies show that sequestration of intracellular, native hsp70 (and hsp90) interferes with transport of antigenic peptides to MHC I molecules. DSG binds to cytosolic HSPs, hsp70 and hsp90 via the COOH-terminal EEVD sequence (34). Because DSG localizes almost exclusively in the cytosol (35) and because the ER hsp gp96 does not have the EEVD sequence, the effects of DSG appear primarily restricted to the cytosolic chaperones. Binding of DSG to HSPs does not affect the ability of peptides to bind to or be released from these HSPs and thus may mediate its effect on HSPs indirectly (Ref. 34 and data not shown). To the best of our knowledge, this is the first demonstration of an obligatory role for hsp70 molecules in antigen presentation by MHC I molecules. Our results in this regard explain the observations of Wells et al. (36) who showed that transfection of antigen presentation-defective B16 melanoma cells with hsp70 renders the cells presentation competent. Finally, our results hint at a surprising role for proteasomes, beyond cleavage of proteins or precursor peptides into final-sized MHC I epitopes. HSP-chaperoned peptides, even though generated from the processing events, including proteasomal activity in the cells from which they are purified, require further proteasomal action for re-presentation. Even when the peptides chaperoned by hsp70 or hsp90 are the exact MHC I-binding size, their re-presentation requires the presence of functional proteasomes. This observation is consistent with the proposal that, in addition to generating peptides from intact proteins, proteasomes are involved in delivery of the peptides into the TAP pathway through a multimolecular assembly, the presentosome (4). Treatment of cells with proteasome inhibitors may abolish the ability of the proteasomes to dock or undock with various members of the presentosome such as TAP and block peptide transport. The proteasomes may, in addition or alternatively, be actively involved in the release of peptides from the chaperone, a process that may require protease activity. This however remains to be demonstrated in a direct manner.
Our results may be viewed in light of some recent developments on the cytosolic events in antigen processing. Bercovich et al. (37) have shown recently that recruitment of hsc70 or hsp27 is required for ubiquitination of certain protein substrates in studies in vitro. Using a metabolically unstable form of the influenza virus nucleoprotein (NP) as the antigen, Anton et al. (38) have recently identified specific sites in the cytosol where antigenic NP peptides are generated. They propose that NP is chaperoned to these sites by hsc70 and that the polyubiquitinylated NP undergo degradation by proteasomes in situ at these sites, leading to generation of antigenic peptides. In view of our results that free antigenic peptides in the cytosol are presented extremely inefficiently and that HSP-chaperoned peptides are presented effectively, we propose to extend the model envisaged by Anton et al. (38) to suggest that hsc70 (and hsp90) is involved not only in the afferent end of this process by chaperoning partially or fully unfolded polypeptide chains to this site, but also in chaperoning the resulting antigenic peptides to the TAP complex. We have suggested previously that all of these processes, including the transport of peptides to TAP, occur in a single dynamic multimolecular assembly, which we termed the presentosome (4).