We and others have shown that influenza A nucleoprotein (NP) targeted to the secretory pathway cannot be processed to yield several cytotoxic T lymphocyte (CTL) epitopes in cell lines that lack the transporter associated with antigen processing (TAP). However, a large COOH-terminal fragment of NP is processed and presented in these cells. Full-length NP is cotranslationally glycosylated in the lumen of the endoplasmic reticulum at two sites distal to the major H2-Kk and H2-Db restricted CTL epitopes, and we show here that pharmacological or genetic inhibition of N-linked glycosylation, leads to the processing and presentation of both these epitopes in a TAP-independent way.

Cytotoxic T lymphocytes (CTLs) recognize complexes formed between MHC class I molecules and peptides generally derived from the breakdown of intracellular proteins. There is a large body of evidence indicating that proteolysis of these proteins begins in the cytosol, probably mediated by a multicatalytic cytosolic protease known as the proteasome (1), and that after these initial events, peptide fragments are transported into the endoplasmic reticulum (ER) by the transporter associated with antigen processing (TAP; for review see reference 2). The continued processing or trimming of peptide epitope precursors in the ER has been suggested by studies in which polypeptides of up to 150 amino acids have been shown to yield nonamer peptides in the ER of TAP-negative cell lines (3, 4). The release of epitopes from the COOH terminus of such precursors is more efficient than from their NH2 termini, suggesting the involvement of N-aminopeptidases (4, 5). Other studies in which peptide epitopes were liberated from the interior of a long polypeptide or from a protein with an inaccessible NH2 terminus also suggests the involvement of endoproteases in the ER (5). These activities are insensitive to inhibitors of the proteasome, lysosomal enzymes and calpains (3, 5).

The extent to which these proteases can act on secretory and membrane proteins that are co-translationally inserted into the ER is unclear. Peptides derived from glycoproteins can be presented by class I MHC, but several of these have been shown to be processed in the cytosol by a process of either mistranslation (6, 7) or dislocation (8, 9), which involves the retrograde translocation of glycoproteins from the ER to the cytosol (for example see reference 10). However, there are examples of MHC class I binding peptides derived both from signal sequences and from within secretory and membrane proteins found in association with MHC class I molecules in TAP-negative cells (11, 12), suggesting that these may be generated in the ER itself. Also, the demonstration that other ER-targeted proteins can be processed in a TAP-independent manner (3, 4, 1315) suggests the involvement of an alternative degradation pathway in the ER.

One issue that has not been addressed by any of the studies looking at the processing of full-length secretory and membrane proteins is the relationship between N-linked glycosylation and susceptibility to antigen processing in the ER. All the full length proteins studied thus far contained consensus sites for attachment of N-linked glycans, and regardless of other factors this could prevent their processing and presentation from within the ER.

As a way of investigating the role of glycosylation in the ER, we have used the observation that ER-targeted NP, which is cotranslationally glycosylated, is not processed in the ER to yield CTL epitopes (3, 16). Here we show that N-glycosylations at sites distal to CTL epitopes can protect them from processing in the ER and that the removal of both sites is necessary before the polypeptide is susceptible to ER processing.

Materials And Methods

Cell Lines.

Simian COS-1 cells were grown in DMEM. Murine L cells transfected with the H-2Db gene were a gift from Dr. A. Townsend (Institute of Molecular Medicine, Oxford, UK) and were maintained in RPMI 1640 medium. The human TAP-negative T2 cell lines, transfected with the genes encoding H-2Db (T2-Db) or H-2Kk (T2-Kk), were gifts from Dr. P. Cresswell (Yale University, New Haven, CT) and were maintained in RPMI 1640. LBL.174 was a gift from Dr. R. DeMars (University of Wisconsin, Madison, WI) and was transfected with the murine H-2 gene H-2Db (0.174Db) by Dr. V. Cerundolo (University of Oxford, Oxford, UK). The thymidine kinase–negative human fibroblast line 143 was maintained in DMEM. All media were supplemented with 10% FCS (Meldrum Ltd., Bourne End, UK), 50 U/ml penicillin, 50 mg/ml streptomycin, and 2 mM l-glutamine.

Antibodies.

The mAb 4.7.18 recognizes a region near the NH2 terminus of influenza A nucleoprotein (NP) and was a gift from J. Skehel (National Institute of Medical Research, Mill Hill, UK). For immunoprecipitation studies, it was used at a final dilution of 1:100. The mAb B22.249 is specific for a conformational epitope on the α1 domain of H-2Db (17) and was used for immunoprecipitation studies B22.949 at a final concentration of 15 μg/ml.

Peptides.

The nonamer ASNENMDAM corresponds to residues 366–374 of NP from strain A/NT/60/68 and is the naturally occurring epitope presented to CTL by H-2Db (18). SDYERGLI, corresponding to residues 50–57 of NP, is presented by H-2Kk (19). ILKEPVHGV corresponds to residues 476–484 of HIV pol, binds to HLA-A2, and is the naturally processed epitope (20). All peptides were synthesized by Research Genetics, Inc. (Huntsville, AL).

Site-directed Mutagenesis.

Oligonucleotide-directed mutagenesis was performed using the Muta-Gene Phagemid In Vitro Mutagenesis kit (Bio-Rad Labs. Ltd., Hemel Hempstead, Hertfordshire, UK) and single stranded pcDNA3 containing the cDNA for wild-type NP tagged with the hemagglutinin leader sequence (3) as a template. Single-site mutants for position 21 (NAT to DAT) and 144 (NDT to NDA) were generated, and a further round of mutagenesis with the reciprocal oligonucleotide was performed to generate a gene in which both glycosylation sequons had been deleted. Wild-type NP (L+NP+/+) contains both sites. The position 21 mutant is designated L+NP−/+, the position 144 mutant is designated L+NP+/−, and the nonglycosylated mutant is designated L+NP−/−.

Recombinant Vaccinia Viruses.

Recombinant vaccinias were made by homologous recombination into the thymidine kinase gene using the shuttle vector pSC11.30R.2 as previously described (3). Vaccinias encoding wild-type L+NP (L+NP+/+vacc), a nonglycosylated COOH-terminal fragment of NP containing residues 368– 498 tagged with the hemagglutinin leader sequence (L+IMPvacc) and influenza matrix protein (M1vacc) had been constructed previously (3, 18, 21, 22). All constructs were confirmed by dideoxy sequencing.

Expression of NP Glycoforms in Mammalian Cells.

COS-1 cells were transfected with 10 μg of sterile plasmid DNA by culturing in 500 μg/ml DEAE-dextran in RPMI, followed immediately by 20 μl of 100-mM chloroquine. Cells were incubated at 37°C for 4 h, after which the transfection mix was removed and replaced with 10% DMSO in RPMI for 3 min. This was replaced with 20 ml of medium. Assays were performed after 48 h. For expression from recombinant vaccinia, cells were infected at a multiplicity of 10 for 90 min at 37°C, after which they were washed twice, resuspended at 5 × 105/ml and incubated for a total of 3 h before further manipulation. 721.174Db at 107/ml were transfected by electroporation with 50 μg of DNA (400 V/cm, 960 μF) using a Cell Electroporator (Bio-Rad Labs., Ltd.). After 1 wk in culture, G418-resistant cells were examined for L+NP expression by immunoprecipitation.

CTL Production.

6-wk-old female C57Bl/6 (H-2b) and CBA (H-2k) mice (Harlan Olac, Bicester, Oxfordshire, UK) were immunized by intraperitoneal injection of 107 PFU of NP-vacc in 200 μl of sterile PBS. Splenocytes were prepared at day 8. Autologous, irradiated splenocytes pulsed for 1 h with 5 mM peptide were used as stimulators. Recombinant human IL-2 was added to the cultures at a final 10 U/ml after the third restimulation. CTL clone F5 was a gift from A. Townsend.

CTL Assays.

Standard 4-h 51Cr-release assays were performed using infected or peptide-pulsed target cells. Targets were labeled with 51Cr at the time of infection for 1 h. Virus was then removed and the cells were incubated for a further 2 h before their exposure to CTL. Where appropriate, cells were treated with 10 μg/ml tunicamycin 2 h after the start of virus infection for a further 2 h before the 4-h killing assay. This approach was taken to ensure that all H-2Db molecules at the cell surface at the start of the killing assay had been synthesized in the presence of tunicamycin.

Metabolic Labeling and Immunoprecipitation.

This was performed as previously described (23). 107 cells labeled with 100 μCi per 107 cells of 35S Promix (Amersham Pharmacia Biotech, Little Chalfont, Buckinghamshire, UK) for 30 min at 37°C were used for each experimental point. Endoglycosidase H (Endo-H) treatment of immunoprecipitates was with 1 U overnight at 37°C. Immunoprecipitates were resolved on a 10% SDS-polyacrylamide gel and analyzed by autoradiography. Comparative experiments were exposed for the same time period. Autoradiographs were quantitated using the BioImage Analyzer and software (Millipore Corp., Bedford, MA) where, for pulse–chase analyses, optical density readings were converted into a percentage value based on the reading at the start of the chase being 100%. Where the effects of tunicamycin were studied, the drug was added (from a stock of 1 mg/ml in DMSO) 1 h before labeling and was present at the indicated concentration throughout the experiment. An equivalent volume of DMSO was added to controls.

Results

We have previously shown that full-length NP expressed in the ER is not processed to reveal two immunodominant CTL epitopes in TAP-deficient cells (3, 16). One possible reason for this is that its cotranslational glycosylation at residues 21 and 144 somehow blocks antigen processing in this compartment.

Blocking the Glycosylation of L+NP Promotes the Release of CTL Epitopes after Processing in the ER.

To test this hypothesis, we assessed the effect of the glycosylation inhibitor tunicamycin on the ER processing of L+NP in T2-Db. At a concentration of 10 μg/ml of tunicamycin, L+NP but not H-2Db glycosylation is substantially inhibited (data not shown). Fig. 1 shows that, under these conditions, T2-Db are able to present both the Db-restricted epitope 366–374 (Fig. 1,A) and T2-Kk the Kk-restricted epitope SDYEGRLI (50-57, Fig. 1,B) to CTLs. In the absence of tunicamycin treatment, no presentation occurs. Presentation of the NP fragment L+IMP (which contains the Db-restricted but not the Kk-restricted CTL epitope) is unaffected by tunicamycin treatment. In contrast to the cell lines T2-Db and T2-Kk, the TAP-competent cell line L-Db is able to present these epitopes independently of the glycosylation status of L+NP, indicating that tunicamycin exerts its effect on the alternative, TAP-independent antigen processing pathway. Tunicamycin treatment restores the presentation of NP366–374 from L+NP in a dose-dependent manner (Fig. 1 C). Thus, the presentation of NP366–374 by T2-Db, induced using doses of tunicamycin that inhibit the glycosylation of NP, suggests that processing of proteins within the ER for presentation through MHC class I is directly linked to the absence of N-linked glycans on the antigen. The decrease in the efficiency of presentation of L+IMP-derived NP366–374 by T2-Db in the presence of tunicamycin is probably due to its effect on the level of H-2Db, and serves to indicate that the restoration of presentation of NP366–374 from L+NP+/+ that was seen in the presence of tunicamycin is significant.

To observe the effect of deglycosylation on the ER processing of L+NP in isolation of vaccinia virus infection, presentation of the H2-Db–restricted epitope was investigated in the TAP-negative cell line LBL721.174 transfected with the L+NP gene. Fig. 2 shows that treatment of the transfected cells with 5 μg/ml tunicamycin restored the presentation of NP366–374 to CTLs to ∼70% of that seen when tunicamycin-treated T2-Db were pulsed with 10 nM of peptide.

To eliminate the possibility that tunicamycin restored presentation of class I epitopes from within L+NP by a mechanism separate from inhibition of antigen glycosylation, we constructed a series of mutants of L+NP by site-directed mutagenesis. The asparagine at position 21 is found in the nucleoproteins of the majority of known strains of influenza A virus, but aspartic acid is present in some strains. Similarly, threonine 146 is found in ∼50% of the known strains, with alanine present in the majority of the remainder. We used these two naturally occurring amino acid substitutions to remove one or both of the glycosylation sequons from L+NP+/+. Recombinant vaccinias encoding the three possible glycosylation variants were then constructed. Fig. 3 shows that in contrast to infection with L+NP+/+, T2-Db infected with L+NP−/−vacc is able to process and present the H2-Db–restricted epitope NP366–374 to CTLs as efficiently as from L+IMPvacc. This demonstrates a direct correlation between the absence of N-linked glycans and the ability to process L+NP within the ER to yield CTL epitopes. The presence of either of the two glycosylation sites alone also prevents the presentation of NP366–374 from L+NP when expressed in T2-Db (Fig. 3). In contrast, like L+NP+/+, both are readily processed in the cytosol of a TAP-competent cell. These results indicate that in addition to the cytosolic processing of secretory glycoproteins, deglycosylated glycoproteins can also be exposed to TAP-independent processing pathways that are located in a distinct compartment. The possibility that processing of the nonglycosylated form of NP also occurs in the cytosol with TAP-independent transfer of the epitope into the ER is not supported by the observation that presentation of the epitope by TAP-negative cells occurs only if it is preceded by an ER targeting signal (3).

The Rate of L+NP Turnover Is Glycan Regulated but Does Not Correlate with Antigen Presentation.

For cytosolic processing, there is a correlation between the rate of antigen degradation and the efficacy with which it is presented to CTLs. To determine whether the same correlation was true for glycoproteins processed in the ER, we measured the rate of degradation of all four L+NP constructs in COS-1 cells. Fig. 4,a shows that L+NP+/+ was long-lived with a half-life of >4 h. It remained sensitive to Endo-H during this time (Fig. 2 and data not shown), suggesting that it was retained in the early part of the secretory pathway. We could not detect any secreted NP even after prolonged (24 h) expression (data not shown). The half-life of NP when expressed in the ER is thus comparable to that of NP expressed in the cytosol (21). Treatment of L+NP+/+ expressing COS-1 cells with 10 μg/ml tunicamycin (a dose that completely inhibits glycosylation of the protein) causes it to be degraded more rapidly with only 40% remaining after a 4-h chase, as compared with 75% of untreated NP. This observation was not due to a nonspecific effect of tunicamycin on the cells, because L+NP−/− is degraded with similar kinetics to L+NP after tunicamycin treatment and is unaffected by tunicamycin treatment. The increased rate of degradation of deglycosylated NP appeared to be controlled by the presence or absence of an N-linked glycan at position 21. Thus, Fig. 4 b shows that, although L+NP+/− was degraded at the same rate as L+NP+/+, L+NP−/+ was degraded at the same rate as L+NP−/−. The slow degradation of L+NP+/− was accelerated when it was synthesized in the presence of tunicamycin. Since neither of the monoglycosylated forms are processed and presented by T2-Db, we conclude that, unlike in the cytosol, where induction of rapid degradation is sufficient to potentiate presentation of class I epitopes (21), antigen processing in the ER is limited by the presence of N-linked glycans on a protein and this effect is independent of the degradation rate of the protein. It is important to note that, in these experiments, the term ‘degradation' is used operationally, since the failure to detect NP by immunoprecipitation does not necessarily reflect its hydrolysis but could reflect other events leading to the loss of the anti-NP mAb epitope.

Discussion

Many secretory and membrane proteins have been shown to yield class I–restricted CTL epitopes. With respect to cytosolic antigen processing, this raises a topological paradox because these proteins are cotranslationally transported into the ER. However, they can be exposed to cytosolic proteases by either their mistranslation (6, 7), or dislocation to the cytosol (8, 9). Alternatively, they can be processed in the ER itself (3, 4, 1315, 24).

The majority of glycoprotein antigens studied to date are processed and presented in a TAP-dependent manner, as is L+NP+/+ described here and elsewhere (8). Using L+NP as a model antigen, we have shown that a controlling factor in determining whether ER-targeted proteins can be processed in the ER is N-glycosylation. Like the diglycosylated form of the antigen (L+NP+/+), neither of the two monoglycosylated proteins (L+NP−/+ and L+NP+/−) yielded CTL epitopes, although they were degraded with different kinetics. This suggested that the glycan-regulated rate of degradation was not the controlling factor in determining whether the protein was processed for presentation to CTLs. Our results are consistent with the idea that a failure to glycosylate might prohibit a nascent glycoprotein from entering the quality control procedure that normally culminates in export from the ER into either the secretory pathway or the cytosol (where they are degraded), allowing it to become exposed to proteases in the ER itself. The glycosylation of proteins targeted to the secretory pathway would therefore have the effect of preventing the generation of large numbers of class I binding peptides within the ER by enzymes that normally function to trim peptides delivered by TAP. Our results show that N-glycosylation can inhibit the generation of epitopes distal to the site of glycosylation, presumably by affecting the general pathway for degradation as outlined above. By ensuring that proteins targeted for degradation are disposed of at a site that is remote from peptide delivery and trimming, the system can ensure that under normal conditions the favored route by which peptides interact with class I MHC molecules is via TAP. This process may not be totally efficient since some epitopes (25), including one in L+NP itself (16), are apparently generated from glycoproteins in a TAP-independent manner. Alternatively, some epitopes may be able to gain entry to the ER in a TAP-independent manner, although this is not the case for the epitopes studied here.

There are other possible immunological consequences of this glycan-regulated antigen presentation. For example, it is possible that some glycoprotein-derived epitopes may be hidden from CTL surveillance through a combination of their failure to be generated in the cytosol and the protection afforded by N-linked oligosaccharides. The possible existence of “hidden” epitopes within glycoproteins has implications for the maintenance of tolerance; exposure of these epitopes by affecting the glycosylation of the source antigen provides a potential route for the initiation of CTL responses to novel epitopes. This could also open up the possibility of pharmacological intervention to modulate presentation of viral glycoproteins in vivo.

Acknowledgments

The authors would like to thank Dr. Enzo Cerundolo and Dr. Veronique Braud for helpful discussions. We would especially like to thank Dr. Tom Wileman for his help in constructing L+NP+/+.

This work was supported by the Wellcome Trust.

References

References
1
Howard
JC
,
Seelig
A
Antigen processing. Peptides and the proteasome
Nature
1993
365
211
212
[PubMed]
2
Elliott
T
Transporter associated with antigen processing
Adv Immunol
1997
65
47
109
[PubMed]
3
Elliott
T
,
Willis
A
,
Cerundolo
V
,
Townsend
A
Processing of major histocompatibility class I–restricted antigens in the endoplasmic reticulum
J Exp Med
1995
181
1481
1491
[PubMed]
4
Snyder
HL
,
Yewdell
JW
,
Bennink
JR
Trimming of antigenic peptides in an early secretory compartment
J Exp Med
1994
180
2389
2394
[PubMed]
5
Snyder
HL
,
Bacik
I
,
Bennink
JR
,
Kearns
G
,
Behrens
TW
,
Bachi
T
,
Orlowski
M
,
Yewdell
JW
Two novel routes of transporter associated with antigen processing (TAP)-independent major histocompatibility complex class I antigen processing
J Exp Med
1997
186
1087
1098
[PubMed]
6
Yewdell
JW
,
Anton
LC
,
Bennink
JR
Defective ribosomal products (DRiPs): a major source of antigenic peptides for MHC class I molecules?
J Immunol
1996
157
1823
1826
[PubMed]
7
Townsend
AR
,
Bastin
J
,
Gould
K
,
Brownlee
GG
Cytotoxic T lymphocytes recognize influenza haemagglutinin that lacks a signal sequence
Nature
1986
324
575
577
[PubMed]
8
Bacík
I
,
Snyder
HL
,
Antón
LC
,
Russ
G
,
Chen
W
,
Bennink
JR
,
Urge
L
,
Otvos
L
,
Dudkowska
B
,
Eisenlohr
L
,
Yewdell
JW
Introduction of a glycosylation site into a secreted protein provides evidence for an alternative antigen processing pathway: transport of precursors of major histocompatibility complex class I–restricted peptides from the endoplasmic reticulum to the cytosol
J Exp Med
1997
186
479
487
[PubMed]
9
Mosse
CA
,
Meadows
L
,
Luckey
CJ
,
Kittlesen
DJ
,
Huczko
EL
,
Slingluff
CL
,
Shabanowitz
J
,
Hunt
DF
,
Engelhard
VH
The class I antigen-processing pathway for the membrane protein tyrosinase involves translation in the endoplasmic reticulum and processing in the cytosol
J Exp Med
1998
187
37
48
[PubMed]
10
Wiertz
EJ
,
Tortorella
D
,
Bogyo
M
,
Yu
J
,
Mothes
W
,
Jones
TR
,
Rapoport
TA
,
Ploegh
HL
Sec61-mediated transfer of a membrane protein from the endoplasmic reticulum to the proteasome for destruction
Nature
1996
384
432
438
[PubMed]
11
Wei
ML
,
Cresswell
P
HLA-A2 molecules in an antigen-processing mutant cell contain signal sequence- derived peptides
Nature
1992
356
443
446
[PubMed]
12
Henderson
RA
,
Michel
H
,
Sakaguchi
K
,
Shabanowitz
J
,
Appella
E
,
Hunt
DF
,
Engelhard
VH
HLA-A2.1–associated peptides from a mutant cell line: a second pathway of antigen presentation
Science
1992
255
1264
1266
[PubMed]
13
Zhou
X
,
Glas
R
,
Momburg
F
,
Hammerling
GJ
,
Jondal
M
,
Ljunggren
HG
TAP2-defective RMA-S cells present Sendai virus antigen to cytotoxic T lymphocytes
Eur J Immunol
1993
23
1796
1801
[PubMed]
14
Hammond
SA
,
Bollinger
RC
,
Tobery
TW
,
Silliciano
RF
Transporter-independent processing of HIV-1 envelope protein for recognition by CD8+ T cells
Nature
1993
364
158
161
[PubMed]
15
Gueguen
M
,
Long
EO
Presentation of a cytosolic antigen by major histocompatibility complex class II molecules requires a long-lived form of the antigen
Proc Natl Acad Sci USA
1996
93
14692
14697
[PubMed]
16
Elliott
T
,
Bodmer
H
,
Townsend
A
Recognition of out-of-frame major histocompatibility complex class I–restricted epitopes in vivo
Eur J Immunol
1996
26
1175
1179
[PubMed]
17
Lemke
H
,
Hammerling
GJ
,
Hammerling
U
Fine specificity analysis with monoclonal antibodies of antigens controlled by the major histocompatibility complex and by the Qa/TL region in mice
Immunol Rev
1979
47
175
206
[PubMed]
18
Townsend
AR
,
Gotch
FM
,
Davey
J
Cytotoxic T cells recognize fragments of the influenza nucleoprotein
Cell
1985
42
457
467
[PubMed]
19
Cossins
J
,
Gould
KG
,
Smith
M
,
Driscoll
P
,
Brownlee
GG
Precise prediction of a Kk-restricted cytotoxic T cell epitope in the NS1 protein of influenza virus using an MHC allele–specific motif
Virology
1993
193
289
295
[PubMed]
20
Tsomides
TJ
,
Walker
BD
,
Eisen
HN
An optimal viral peptide recognized by CD8+ T cells binds very tightly to the restricting class I major histocompatibility complex protein on intact cells but not to the purified class I protein
Proc Natl Acad Sci USA
1991
88
11276
11280
[PubMed]
21
Townsend
A
,
Bastin
J
,
Gould
K
,
Brownlee
G
,
Andrew
M
,
Coupar
B
,
Boyle
D
,
Chan
S
,
Smith
G
Defective presentation to class I–restricted cytotoxic T lymphocytes in vaccinia-infected cells is overcome by enhanced degradation of antigen
J Exp Med
1988
168
1211
1224
[PubMed]
22
Gotch
F
,
McMichael
A
,
Rothbard
J
Recognition of influenza A matrix protein by HLA-A2–restricted cytotoxic T lymphocytes. Use of analogues to orientate the matrix peptide in the HLA-A2 binding site
J Exp Med
1988
168
2045
2057
[PubMed]
23
Lewis
JW
,
Neisig
A
,
Neefjes
J
,
Elliott
T
Point mutations in the alpha 2 domain of HLA-A2.1 define a functionally relevant interaction with TAP
Curr Biol
1996
6
873
883
[PubMed]
24
Cerundolo
V
,
Benham
A
,
Braud
V
,
Mukherjee
S
,
Gould
K
,
Macino
B
,
Neefjes
J
,
Townsend
A
The proteasome-specific inhibitor lactacystin blocks presentation of cytotoxic T lymphocyte epitopes in human and murine cells
Eur J Immunol
1997
27
336
342
[PubMed]
25
Hammond
SA
,
Johnson
PR
,
Kalams
SA
,
Walker
BD
,
Takiguchi
M
,
Safrit
JT
,
Koup
RA
,
Siliciano
RF
An epitope selective transporter associated with antigen processing (TAP)-1/2–independent pathway and a more general TAP-1/2–dependent antigen processing pathway allow recognition of the HIV-1 envelope glycoprotein by CD8+ CTL
J Immunol
1995
154
6140
6156
[PubMed]

Author notes

Address correspondence to Tim Elliott, Institute for Molecular Medicine and Nuffield Department of Clinical Medicine, University of Oxford, John Radcliffe Hospital, Oxford OX3 9DU, UK. Phone: 44-1865-221949; Fax: 44-1865-222901; E-mail: tim.elliott@ndm.ox.ac.uk