Response Frameshifting Give Rise to a Protective CTL Epitopes ...

The Journal of Immunology

Epitopes Derived by Incidental Translational Frameshifting Give Rise to a Protective CTL Response1 Matthew B. Zook,* Michael T. Howard,† Gomathinayagam Sinnathamby,* John F. Atkins,†‡ and Laurence C. Eisenlohr2* Aberrant gene expression can be caused by several different mechanisms at the transcriptional, RNA processing, and translational level. Although most of the resulting proteins may have no significant biological function, they can be meaningful for the immune system, which is sensitive to extremely low levels of Ag. We have tested this possibility by investigating the ability of CD8ⴙ T cells (TCD8ⴙ) to respond to an epitope whose expression results from incidental ribosomal frameshifting at a sequence element within the HSV thymidine kinase gene. This element, with no apparent functional significance, has been identified due to its ability to facilitate escape from the antiviral compound acyclovir. Using a recombinant vaccinia virus expression system, we find that in vitro and in vivo TCD8ⴙ responses to the frameshift-dependent epitope are easily discernible. Furthermore, the in vivo response is at a sufficient level to mediate protection from a tumor challenge. Thus, the targets of immune responses to infectious agents can extend beyond the products of conventional open reading frames. On a per-cell basis, responses to such minimally expressed epitopes may be exceedingly effective due to the selective expansion of high avidity TCD8ⴙ. The Journal of Immunology, 2006, 176: 6928 – 6934. he CD8⫹ T lymphocyte (TCD8⫹)3 is activated through recognition of Ag-derived peptides (epitopes), generally 8 –11 aa in length, displayed at the surface of the Agbearing cell in complex with MHC class I molecules (1, 2). This population serves to limit the spread of intracellular parasites, either by outright killing of the infected cell, or by release of factors that halt or limit replication (3, 4). A striking property of TCD8⫹ is their exquisite sensitivity to very low levels of Ag. TCD8⫹ can be triggered by tens to hundreds of copies of epitope at the cell surface, which can be derived from quantities of precursor protein that are undetectable by standard biochemical techniques (5, 6). Although some proteins are naturally expressed at very low levels, other trace species result from errors in gene expression during transcription, splicing, and translation. A number of variations on conventional translation are relevant in this regard. These variants include initiation at non-AUG codons (7–9), scanning over the primary AUG with initiation at an internal AUG (10, 11), and ribosomal frameshifting, in which the ribosome shifts into a new reading frame in midtranslation (12, 13). This last mechanism may

T

*Department of Microbiology and Immunology, Kimmel Cancer Center, Thomas Jefferson University, Philadelphia, PA 19107; †Department of Human Genetics, University of Utah, Salt Lake City, UT 84112; and ‡Bioscience Institute, University College Cork, Cork, Ireland Received for publication December 8, 2004. Accepted for publication February 24, 2006. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 This work was supported by National Institutes of Health Grant R01 GM48152 (to J.F.A.) and Grants R01 AI39501 and R01 AI46511 (to L.C.E.). L.C.E. is a Leukemia and Lymphoma Scholar. This work was also supported by a development grant from the Muscular Dystrophy Association (to M.T.H.) and Training Grant AI07492 from the National Institutes of Health (to M.B.Z.) and by the Dr. Alfred W. and Mignon Dubbs Fellowship Fund. 2 Address correspondence and reprint requests to Dr. Laurence C. Eisenlohr, Department of Microbiology and Immunology, Bluemle Life Sciences Building 730A, Thomas Jefferson University, 233 South 10th Street, Philadelphia, PA 19107. E-mail address: [email protected]

Abbreviations used in this paper: TCD8⫹, CD8⫹ T cell; TK, thymidine kinase; vac, vaccinia virus; NP, nucleoprotein.

3

Copyright © 2006 by The American Association of Immunologists, Inc.

have particularly rich potential for the generation of “cryptic” epitopes because, in contrast to the former two, it is not limited to early phases of translation. Translational frameshifting can be separated into two general categories. Programmed alternative translation (recoding) is orchestrated by cis-acting signals in mRNA and occurs at high levels (14, 15). The most common form of recoding, programmed ⫺1 frameshifting occurs at a specific location within the overlap of two open reading frames, the trans-frame products are generally functional proteins, and this mechanism is commonly associated with viruses that infect mammalian cells. A well-known example occurs during translation of the HIV gag-pol overlap. Frameshifting occurs at a shift prone site, a series of consecutive bases that allow for slippage of the A- and P-site tRNAs relative to the mRNA (16). High-level frameshifting is achieved by the presence of a downstream stimulatory RNA structure (17–19). Although the exact mechanism of frameshift stimulation is unknown (20, 21), it is generally believed that the RNA structure causes the translocating ribosome to pause, allowing for slippage of the tRNAs relative to the mRNA. Programmed ribosomal frameshifting at the HIV gag-pol shift site occurs at a frequency of ⬃5–10%, determining the ratio of the Gag to Gag-Pol polyprotein, which is critical for viral propagation (22–24). The second category of ribosomal frameshifting is incidental, directed by sites in coding sequences that fortuitously bear trace elements of highly evolved programmed recoding sites, and induce frameshifting at much lower frequencies. These include homopolymeric runs of nucleotides (slippery sites) and sequences that can stall the ribosome, due to slowly decoded codons in the A-site (25–28). A particularly interesting example provides the means for HSV escape from acyclovir treatment (29, 30). The thymidine kinase (TK) gene contains a run of seven consecutive guanines that allows for ⬃1% of translating ribosomes to shift into the ⫹1 reading frame, negligibly reducing the yield of functional TK. In contrast, the escape mutant contains eight consecutive G’s (31) effectively disrupting the reading frame beyond the point of nucleotide insertion. Faithful translation of the mutant results in functionless TK. 0022-1767/06/$02.00

The Journal of Immunology However, 1% of the time, the correct reading frame is regained during translation via the ⫹1 frameshift event and functional TK is produced. This level of expression is sufficient for viral propagation but insufficient for developing toxic levels of acyclovir monophosphate, the conversion product (30). In this study we asked whether incidental ribosomal frameshifting, as occurs during translation of the tk gene and ostensibly during the course of most, if not all, viral infections, can reveal “outof-frame” epitopes at a level that is functionally meaningful. Using a recombinant vaccinia virus (vac) expression system, several in vitro assays for epitope expression, as well as a tumor challenge model, we found that this is indeed the case.

Materials and Methods Mice, cell lines, and chemicals Six- to 8-wk-old female C3FeB6F1/J, B6C3F1, or C57BL/6 mice were purchased from Taconic Farms or The Jackson Laboratory and maintained in the Thomas Jefferson University Animal Facilities (Philadelphia, PA) according to Institutional Animal Care and Use Committee guidelines. The following cell lines were maintained in DMEM (Cellgro; Fisher Scientific) supplemented with 5% FCS at 9% CO2: L-Kb (L929 cells transfected with the Kb gene); L-Db (L929 cells transfected with the Db gene); 143B (TK-1, CCL-1; American Type Culture Collection (ATCC)); and CV-1 (CCL-70, ATCC). The following lines were maintained in RPMI 1640 (Cellgro; Fisher Scientific) supplemented with 10% FCS and 50 ␮M 2-ME at 6% CO2: B3Z (OVA257–264/Kb-specific T cell hybridoma); DBFZ (NP366 –374/Db-specific T cell hybridoma); EL4 (TIB-39; ATCC); and E.G7-OVA (EL4 constitutively expressing OVA). All chemicals were purchased from Sigma-Aldrich unless otherwise noted. All enzymes were purchased from New England Biolabs.

Molecular constructs and viruses Construction of the NP/SIINFEKL gene has been described elsewhere (32). For insertion of frameshift-inducing or control sequences into NP/ SIINFEKL and wild-type NP, complimentary oligonucleotides (sense strand sequences shown in Fig. 1), designed such that when annealed they would have SphI compatible ends, were ligated into SphI-digested NP/ SIINFEKL (HSV TK) or NP (HIV) genes contained within modified versions of the pSC11 plasmid (33). All constructs were verified by DNA sequencing. Plasmid constructs were recombined into the vac genome (34) and confirmed by sequencing as described elsewhere (35). Briefly, CV-1 cells were infected with wild-type vac (CR-19) followed by transfection of molecular constructs. Recombinant viruses were triple plaque purified in 143B cells in the presence of 5 mg/ml BrdU (Boehringer Mannheim) and expanded and titered on 143 HuTK⫺ cells.

Radio immunoprecipitation A total of 1 ⫻ 106 L-Kb cells were infected with 5 PFU/cell recombinant vac for 1 h in 250 ␮l of PBS supplemented with BSA, and then transferred to complete medium for 3 h. Medium lacking methionine and cysteine was added for 45 min, followed by addition of 50 ␮Ci 35S Met/Cys (Amersham Biosciences) for 2–3 h. Cells were washed in PBS, lysed in 0.01 M TrisHCl, 0.14 M NaCl, 0.5% Nonidet P-40 with 1 mM PMSF, and the insoluble portion separated by centrifugation. Soluble cell lysate was precleared with protein A-Sepharose (RepliGen) for 1 h at 4°C and NP/SIINFEKL or NP protein purified using a 1:1 culture-supernatant mix of mAb HB-64 and H19-S24 (Dr. W. Gerhard, Wistar Institute, Philadelphia, PA) bound to protein A-Sepharose for 12 h at 4°C. Samples were boiled with 2⫻ SDS reducing buffer (0.125 M Tris-HCl, 4% SDS, 20% glycerol, and 10% 2-ME) and separated on a 10% polyacrylamide gel containing 0.1% SDS. Gels were treated in Amplify solution (Amersham Biosciences) dried (Hoefer GD2000; Amersham Biosciences) and exposed to film (X-OMAT LS; Kodak) for 24 h. Band density was quantified with a Typhoon 9410 imager and ImageQuant software (Amersham Biosciences).

Flow cytometry A total of 1 ⫻ 106 L-Kb cells were infected with 5 PFU/cell recombinant vac for 1 h in BSA supplemented PBS followed by addition of 9 ml of DMEM containing 5% FCS and incubated at 37°C for 7 h. Cells were washed two times with FACS buffer (PBS, 1% FCS, 0.08% sodium azide), resuspended in a 1/2 dilution of 25.D1.16 mAb culture supernatant and incubated at 4°C for 1 h. Cells were washed three times with FACS buffer and incubated with 5 ␮g/ml FITC-conjugated horse anti-mouse IgG (Vec-

6929 tor Laboratories) for 30 min at 4°C. Cells were washed three times in FACS buffer, fixed in 2% paraformaldehyde and analyzed using an Epics Profile flow cytometer (Beckman Coulter). Cell surface density of the KbSIINFEKL complexes was quantified by comparing the mean fluorescence of samples to that of FITC-labeled calibration beads (Beckman Coulter).

T cell hybridoma assays Assays for epitope expression based upon use of the B3Z or DBFZ hybridomas have been previously described (32). Briefly, 5 ⫻ 104 L-Kb or L-Db cells were infected in triplicate with 5 PFU/cell recombinant vac for 2 h in 250 ␮l of BSA supplemented PBS in the presence of 250 ␮g/ml arabinofuranosylcytosine. Cells were washed with PBS and cocultured in 96-well flat-bottom plates with 5 ⫻ 104 B3Z or DBFZ cells in complete media supplemented with 250 ␮g/ml arabinofuranosylcytosine overnight. T cell activation was assessed by addition of the fluorogenic substrate 4-methylumbelliferyl ␤-D-galactoside (4-MUG) at 33 ␮g/ml. To compare results across several independent experiments, the values obtained with the frameshift constructs were normalized to those of the controls using the following formula for normalized value: ((sample ⫺ Neg)/(Pos ⫺ Neg) ⫻ 100).

ELISPOT Mice were immunized via i.p. injection with 1 ⫻ 107 PFU of recombinant vac diluted to 250 ␮l with BSA supplemented PBS. Fourteen days postinfection spleens were harvested, RBC lysed, and plated at 2.5 ⫻ 105, 5 ⫻ 104, or 1 ⫻ 104 cells per well in 50 ␮l of RPMI 1640 containing 10% FCS and 50 ␮M 2-ME in 96-well ELISPOT plates (Millipore) coated with 5 ␮g/ml anti IFN-␥ (51-2525KZ; BD Pharmingen). Specific activation was determined by addition of 2.5 ⫻ 105 L-Kb or L-Db cells either untreated, pulsed with 10⫺8 M NP50 –57, SIINFEKL, or NP366 –374 synthetic peptide, or infected with nonrecombinant (wild-type, Cr-19) vac. Plates were incubated at 37°C for 18 h, washed two times with PBS and three times with wash buffer (PBS with 0.5% Nonidet P-40), and incubated with 4 ␮g/ml biotinylated anti-IFN-␥ Ab (51-1818KA; BD Pharmingen) for 2 h at room temperature. After washing, 10 ␮g/ml avidin-HRP (551950; BD Pharmingen) was added and incubated at room temperature for 1 h. Plates were washed first with wash buffer followed by PBS and spots visualized by addition of substrate solution (551951; BD Pharmingen). Spots were counted using Image-Pro software (Media Cybernetics), and values were obtained by subtracting background (i.e., unpulsed) from treated cells. Due to this manipulation, the SIINFEKL and 366 –374 response in the negative control became negative because the values were actually less than the background. To control for differences in priming between individual animals, individual SIINFEKL-specific responses were normalized to the NP50 –57 responses as described earlier (36). The average NP50 –57 value for each experiment was determined. The fold difference of individual NP50 –57 responses from the average NP50 –57 response was calculated using the formula ((average NP50 –57 response ⫺ individual NP50 –57 response)/(individual NP50 –57 response)). The fold difference was then multiplied by the appropriate SIINFEKL-specific response to obtain a normalized value.

Tumor challenge Six-week-old female C57BL/6 mice were primed i.p. with 1 ⫻ 107 PFU of recombinant vacs. Ten days later, they were anesthetized and shaved on the right flank. A total of 3 ⫻ 106 E.G7 or EL-4 cells maintained at 5 ⫻ 105/ml were washed three times with PBS and resuspended to 3 ⫻ 107/ml. They were then injected s.c. in 100 ␮l of total volume PBS. Tumor volume was measured every other day beginning when tumor growth became visible by measuring the diameter in three dimensions using a vernier caliper.

Results Generation of constructs In broad outline, our strategy was to place a frameshifting element between sequences encoding two MHC class I-restricted epitopes. Translation of the upstream epitope is not impacted by the insertion and can be used to normalize expression of the downstream epitope, which requires ribosomal frameshifting at the element to be expressed. As indicated in Fig. 1, our base construct for these studies was the influenza virus NP, which contains three welldescribed epitopes at residues 50 –57 (H2-Kk-restricted, NP50 –57, (37)), 147–155 (H2-Kd-restricted, NP147–155, (38)), and 366 –374 (H2-Db-restricted, NP366 –374 (39)). A unique SphI site between the NP147–155 and NP366 –374 epitopes served as an insertion point for

6930

INCIDENTAL RIBOSOME FRAMESHIFTING IN PROTECTIVE CD8 RESPONSE necessary for translation of the downstream epitopes. Similar positive and negative control constructs were included (Fig. 1). All of these constructs were recombined into the vaccinia genome according to standard methods (34). Frameshifting induced by the HSV TK sequence is marginally detectable with standard biochemistry but is readily detectable using a TCR-like mAb For direct assessment of frameshifting mediated by the HIV and TK elements in the NP-based system, L929 cells expressing the transfected H2-Kb gene (L-Kb) were infected with the recombinant vaccinia panel, radiolabeled with 35S, and immunoprecipitated with mAbs specific for the N terminus of NP. As seen in Fig. 2A, bands corresponding to full-length 56-kD protein products are visible in lanes from both the HIV positive and HSV TK positive controls. Additionally, 38-kD bands are present in samples from frameshift constructs (HIV FS, HSV TK 7G, HSV TK 8G, Fig. 2A) as well as the negative controls (HIV Neg, HSV TK Neg, Fig. 2A), due to a termination codon in the ⫺1 reading frame that is encountered by nonframeshifting ribosomes after traversing the point of insertion. A faint band with the same mobility as the full-length protein is present in the lane from the HIV construct (HIV FS, Fig. 2A) resulting from the well-characterized ⫺1 frameshift event. Scanning densitometry of multiple samples established

FIGURE 1. Generation of viral frameshift (FS) constructs. All constructs were based upon the Influenza PR/8 NP containing endogenous MHC class I epitopes 50 –57 (Kk-restricted), 147–155 (Kd-restricted), and 366 –374 (Db-restricted). For the HSV series of constructs, the 257–264 epitope from OVA (SIINFEKL, Kb-restricted) was inserted immediately upstream of the 366 –374 epitope. Frameshift constructs were created by cloning sequence corresponding to the HIV or HSV TK frameshift site and surrounding sequence into the SphI site of the parent gene in such a way that a translational shift into the ⫺1 or ⫹1 reading frame, respectively, was necessary for production of the downstream epitopes. For HSV, the shift sites are underlined and the nucleotide additions are in bold. For HIV, the shift site is underlined.

two different frameshifting elements, that of the HSV tk gene (both 7 and 8 consecutive G versions) and that at the HIV gag-pol interface. As mentioned, the gag-pol element directs recoding at a frequency of 5–10% and was included to contrast epitope production driven by the low level (ⱕ1%) incidental frameshifting of the TK element. For the TK element, we used a version of NP (NP/ SIINFEKL) that contains the widely used OVA-derived epitope (OVA257–264, H2-Kb-restricted, (40)) immediately upstream of the NP366 –374 epitope due to the existence of several useful OVA257–264-specific reagents. The HIV-1 frameshifting element at the gag-pol interface consists of a U UUU UUA (gag reading frame in triplicates) slippery site and a downstream secondary structure (18, 19, 41). This entire element was inserted into the SphI site of NP in a way that places the downstream sequence in the ⫺1 reading frame, thereby necessitating a ⫺1 ribosomal frameshift event for production of the NP366 –374 epitope (HIV FS, Fig. 1). A relevant negative control was produced in which the shift site was mutated, abolishing the opportunity for mRNA-tRNA re-pairing in the ⫺1 frame (HIV Neg, Fig. 1). Similarly, a positive control was generated in which the downstream sequence was placed in the zero reading frame, thus requiring no frameshifting for its expression (HIV Pos, Fig. 1). The incidental and less potent HSV TK frameshifting elements, either the wild-type run of 7 G’s (TK 7G, Fig. 1) or the mutant run of 8 G’s (TK 8G, Fig. 1) with surrounding 5⬘ and 3⬘ sequence, were similarly inserted into the SphI site of NP/SIINFEKL. Both constructs were generated in such a way that a ⫹1 frameshift was

FIGURE 2. Production of full-length protein as a result of ribosomal frameshifting at viral sequences. A, L-Kb cells infected with 5 PFU/cell of the indicated vac were labeled with 35S, and NP (HIV) or NP/SIINFEKL (NP/S) TK protein immunoprecipitated. Protein species were separated via PAGE and visualized by autoradiography with care being taken to avoid background bands migrating closely with the full-length protein product during analysis. B, L-Kb cells were infected with 5 PFU/cell of the indicated vac, and surface Kb-SIINFEKL expression quantified using mAb 26.D1.16. Quantification was achieved using calibration beads containing a known density of fluorescent probe. Background fluorescence obtained with the TK negative control sample was subtracted from the experimental samples, with absolute values representing from 1 to 22% increase over the negative control. Values represent the average of two independent experiments.

The Journal of Immunology the frequency of frameshifting to be 5–16%, which is in good agreement with previous results (16). In the lanes corresponding to the HSV TK constructs (TK 8G and TK 7G, Fig. 2A), the level of full-length product resulting from ⫹1 frameshifting was much lower, with a trace band being only sporadically appreciable. Thus, with the constructs supporting low-level frameshifting, the amount of protein available for production of the NP366 –374 and OVA257–264 epitopes is at the limit of detection by immunoprecipitation. One of the useful OVA257–264-specific reagents is a mAb, 26.D1.16, which is “TCR-like” in having specificity for the OVA257–264/Kb complex (42). L-Kb cells, capable of OVA257–264/Kb presentation, were infected with equal amounts of the TK series of recombinant vacs. MHC class I/ligand cell surface density was determined by incubating samples in culture supernatant containing the 25.D1.16 Ab followed by anti-mouse, FITClabeled IgG and analyzed by flow cytometry. Through the use of standardized microbeads loaded with a known concentration of FITC, we were able to quantify the absolute number of complexes present. As seen in Fig. 2B (TK 8G and TK 7G), both HSV constructs produce a signal that is significantly above background. Importantly, this result suggests that frameshifting occurs during translation of wild-type TK, as has been reported previously (31). Thus, despite the full-length protein being barely detectable by immunoprecipitation, frameshifting occurs at a level that provides appreciable cell surface epitope as determined with this immunebased approach. Viral frameshift-derived epitopes stimulate T cell activation in vitro Detection of presented epitope by the 26.D1.16 Ab is highly sensitive and may, in fact, surpass the sensitivity of T cells in vitro and/or in vivo. As a first step in assessing the relevance of the TK frameshifting element for T cell activation, we determined the ability of the vac panel members to activate NP366 –374/Db- and OVA257–264/Kb-specific T cell hybridomas that produce ␤-galactosidase upon activation (43, 44). L929 cells transfected with the H2-Db-encoding gene (L-Db) were infected with the HIV recombinant vac panel and cocultured with the DBFZ T cell line (specific for NP366 –374/Db molecule); similarly, L-Kb cells were infected with the HSV TK panel and cocultured with the B3Z T cell hybridoma (OVA257–264/Kb-specific), and activation was assessed using soluble 4-MUG. Results shown in Fig. 3 demonstrate that frameshifting at both the HIV and HSV sites produces sufficient epitope for hybridoma stimulation. For the HIV element, the level of activation resulting from the HIV frameshift construct ranged from 55 to 66% of that of the positive control over four indepen-

6931 dent experiments (HIV FS, Fig. 3). This finding was not especially surprising given the estimated frameshift frequency of 5–10%. For the HSV constructs (TK 8G and TK 7G, Fig. 3), the level of activation ranged from 10 to 26 and 13 to 22%, respectively, over four experiments. Thus, in an in vitro setting, even low-level frameshifting induced by the TK element results in clearly appreciable T cell responses as determined by a T cell hybridoma. In vivo response to frameshift-directed epitopes One of the main reasons for using the vac expression system is the documented ability of this virus to elicit significant TCD8⫹ responses both in vitro and in vivo. Thus, the same constructs used in the in vitro assays we have described can be used for assessing TCD8⫹ activation in vivo. H2-Kk/H2-Db F1 mice were primed with individual members of the HIV recombinant vac panel. Two weeks later, spleen cells were assessed by ELISPOT analysis for the presence of NP50 –57- or NP366 –374-specific population, which would be detectable only by expansion of the exceedingly low number of epitope-specific precursors. Because NP50 –57 precedes the frameshift element, responses to this epitope should be roughly equivalent in all cases if the immunizations were equivalent, whereas responses to NP366 –374 reflect frameshifting in vivo. As expected, the CD8⫹ T cell responses to target cells pulsed with NP50 –57 peptide are relatively uniform among all constructs tested (Fig. 4A). In contrast, the NP366 –374 response to peptide-pulsed targets is proportional to the expected frameshift frequency with negative values representing a response that was below the background for that experiment. Fig. 4B shows the normalized NP366 –374-specific response, which corrects for the minor variations in priming as determined by the differences in the NP50 –57 responses (36). The anti-NP366 –374 response associated with the HIV frameshift construct remains significantly higher than the response to the negative control construct after this transposition. Again, this result is not surprising given the level with which the HIV element mediates frameshifting. A similar approach was taken to assess the impact of the HSV TK element in vivo except that the resultant TCD8⫹ responses to the OVA257–264 epitope were measured instead of NP366 –374. As with the HIV panel, responses to NP50 –57 are essentially uniform among all constructs (Fig. 4C) with the normalized SIINFEKLspecific response shown for correction (Fig. 4D). Notably, both HSV constructs (TK 8G and TK 7G, Fig. 4C) are associated with a significant response to the OVA257–264 epitope. These results show that epitope production via incidental frameshifting during the translation of viral transcripts can induce a detectable TCD8⫹ response. Protective effect of TCD8⫹ stimulated by frameshift-derived epitopes

FIGURE 3. Activation of epitope specific T hybridomas by the frameshift (FS) panel. L-Db (for HIV) or L-Kb (for HSV) cells were infected in triplicate with 5 PFU/cell of the indicated recombinant vac and cocultured with DBFZ or B3Z hybridoma, respectively. Specific activation was determined by addition of enzymatic substrate and measurement of OD. Data shown are the average and SD of normalized values calculated using the formula ((sample ⫺ Neg)/(Pos ⫺ Neg) ⫻ 100). Nonnormalized values for the positive controls were roughly 50% greater than the background obtained with the negative controls. p ⫽ 0.001. As determined by ANOVA compared with respective positive and negative controls.

Having demonstrated in vivo expansion of TCD8⫹ specific for an epitope expressed as a result of incidental frameshifting, we sought to determine whether this expanded population has any functional significance. Toward this end, we determined whether the frameshifting-dependent CD8⫹ population also protects against a tumor challenge. Mice were primed as described for the ELISPOT assay, and then challenged s.c. at 10 days post-vac injection with EL-4 tumor cells (naturally expressing the Kb class I molecule) or E.G7 tumor cells (EL-4 cells stably transfected to express OVA). As seen in Fig. 5A, tumor burden was significantly delayed in mice primed with either HSV construct (TK Pos or TK 8G) compared with growth in mice primed with the HSV negative construct (TK Neg, Fig. 5A). As expected, no difference was noted in mice challenged with the parental tumor line (Fig. 5B).

6932

INCIDENTAL RIBOSOME FRAMESHIFTING IN PROTECTIVE CD8 RESPONSE

FIGURE 4. In vivo priming of a CD8⫹ T cell response by viral frameshift-derived epitopes. Mice were primed with recombinant vac and the resulting magnitude of the epitope specific T cell population analyzed by ELISPOT. Splenocytes were cocultured with L-Db cells (A and B) or L-Kb (C and D) cells unpulsed or pulsed with 10⫺8 M 50 –57 peptide, 10⫺8 M 366 –374 (A and B), or 10⫺8 M SIINFEKL (C and D) peptide. Following overnight incubation, epitope-specific cells were quantified based on IFN-␥ production. Values were obtained by subtracting the spots per well in unpulsed samples from the 50 –57 and SIINFEKL pulsed wells. Data in A and C are epitope-specific cells averaged from three mice per group. Normalized values (B and D) were calculated by correcting the NP366 –374 or SIINFEKL response as described in Materials and Methods. Raw values ranged from several hundred to several thousand above background for experimental and positive control constructs. Error bars represent SD.

Discussion We have demonstrated that the incidental ribosomal frameshifting directed by a fortuitous frameshift element within the tk gene is significant with respect to CD8⫹ T cell responses. Not only are such responses readily detectable both in vitro and in vivo, they also afford protection against a tumor challenge. These findings should be distinguished from publications that have described the production of “neo-epitopes” as a result of frameshift mutations (45– 47). Those cases involve nucleotide insertion/deletion, leading to an actual change in the reading frame. In our case, unconventional translation (i.e., ribosomal frameshifting) by the ribosome is responsible for the shift to an alternative reading frame. Ribosomal frameshifting within the human IL10 gene has been previously shown to produce a cryptic epitope as read out by T cells that had been extensively expanded in vitro following isolation from Reiters syndrome patients (12). Indeed, in vitro systems have elucidated many different mechanisms of aberrant gene expression that lead to cryptic epitope production. These include splicing errors (48, 49), translation initiation at a non-AUG (50), translation initiation at an internal AUG (10), and ribosomal frameshifting (12). Yet, in few sites has significance in vivo been demonstrated. In a report by Schwab et al. (51), a cryptic epitope produced from a non-AUG initiation site was expressed from a murine transgene. Spleen cells from the transgenic mouse were observed to elicit a TCD8⫹ response to the epitope upon transfer to nontransgenic counterparts, and the transgenic mice were shown to be tolerant to the epitope. Similarly, Wang et al. (52) showed that tumor infiltrating lymphocytes from patients with melanoma recognized an Ag produced from an alternative reading frame of the gp75 protein. In this case, the mechanism was felt to be initiated at an internal start codon. In a murine viral system, Mayrand and colleagues (53, 54) previously also showed that internal initiation was potentially capable of producing a cryptic epitope that was the target of a CD8 T cell response. Finally, a recent report by Cardinaud et al. (55) has shown that several cryptic epitopes are produced during the course of HIV replication and are capable of eliciting a T cell response. In these cases, insertion of a premature stop codon in the alterative reading frame abrogated epitope expression; however, the exact mechanism of epitope production remains to be elucidated. The work we have reported in this study demonstrates protection from tumor challenge as a result of ribosomal frameshifting, a process that we consider to have substantial potential for the production of cryptic epitopes.

Ribosomal frameshifting can ostensibly occur at any point during translation because sequences that induce frameshifting are not restricted to any region of an open reading frame. In contrast, aberrant initiation of translation is confined to upstream regions of the open reading frame because ribosomes lose the ability to initiate translation soon after engaging mRNA, presumably due to dissociation of initiation factors from the translation complex. Thus, ribosomal frameshifting would appear to be a more potent mechanism for the generation of cryptic epitopes. However, several other parameters, many poorly defined, need to be considered. First is the relative frequency with which these distinct events occur. Initiation at conventional start codons is influenced by surrounding sequence in predictable ways (11). The same rules appear to apply for internal initiation at conventional start codons (6). It is not yet known whether non-AUG initiation is similarly restricted and, therefore, how promiscuous this event is (8). Algorithms for predicting sites that stimulate ribosomal frameshifting are at very early stages of development and so the general frequency of lowlevel frameshifting is also not known, especially in mammals. Interestingly, although some sequences, such as homopolymeric runs of nucleotides, appear straightforward, there are many sequences that induce frameshifting in ways that are not immediately obvious. Indeed, a recent search of the yeast genome for under-represented sequences that might undergo translational frameshifting identified a subset of sequences, some of which do not resemble known frameshift motifs, which are prone to nonprogrammed error frameshifting (56). Identified in this group were heptamers that may direct frameshift levels of between 0.5 and 8% with no obvious gene expression function. Likewise, an analysis of Escherichia coli coding sequences found known error prone sequences, which can promote higher than 1% frameshifting errors in coding sequences (57). Although these sites were under-represented in highly expressed genes, they were readily identified in many other genes. The unexpected conclusion is that error frameshifting, at least in some cases, is not sufficiently detrimental to trigger strong negative selection pressure. Consequently, error frameshifting at significant levels may be more widespread than is readily appreciated. Also important to consider are the levels of expression that lead to meaningful T cell activation and how frequently they are exceeded for various aberrant species. The critical level will vary for each epitope due to factors such as affinity for the restricting class I molecule and efficiency of processing. With respect to this latter

The Journal of Immunology

6933 shift-derived epitopes would be skewed toward even higher levels of avidity. Although our observations are physiologically relevant, with expression of the constructs being driven by a natural vac promoter, there are undoubtedly cases in which gene expression is so low that fortuitous frameshifting events are insignificant even for the immune system. It remains to be seen what fraction of the TCD8⫹ response to various pathogens and cancers is specific for cryptic epitopes but the results reported suggest that simply counting by tetramer staining, intracellular cytokine staining, or ELISPOT may not reflect physiological effectiveness. It will be interesting to determine in future studies the actual contribution of such specificities and the extent to which they can be exploited. Finally, external influences and even aging may change the accuracy with which cellular genes are translated, leading to first-time production of some self epitopes. How such changes alter the definition of “self” and whether they play a role in precipitating autoimmunity are important questions for future studies.

Acknowledgments We thank Gabriela Plesa for generation of preliminary data and helpful discussions. FIGURE 5. CD8⫹ T cells specific for frameshift-derived epitopes are able to protect against a heterologous tumor challenge. Top, Female C57/ BL6 mice were primed with recombinant vac and rested for 10 days. A total of 3 ⫻ 106 E.G7 cells were then injected s.c. into the shaved right flank in 100 ␮l of PBS and tumor volume measured in three dimensions every other day beginning on day 4. Data represent the average of four mice per group and error bars represent SD. For HSV TK 8G, p ⫽ 0.01 at days 6, 8, 10, and 12 and p ⫽ 0.05 at day 14. For HSV TK Pos, p ⫽ 0.001 at days 6, 8, 10, 12, and 14. Bottom, As described, except non-SIINFEKL expressing EL-4 cells were used for challenge. Data are the average of three mice per group and error bars represent SD. Significance for HSV (p ⬎ 0.05) TK 8G and TK Pos vs TK Neg at all time points.

parameter, it is worth a mention that some products of aberrant gene readout are sequestered in aggregates that are resistant to proteasomal degradation (58) a consequence that certainly deserves exploration in the future. It is clear, however, from the work reported in this study and previously (5, 59, 60) that the necessary expression levels are minimal due to the exquisite sensitivity of CD8⫹ T lymphocytes. Furthermore, in vivo responses to the control and experimental constructs (see Figs. 3 and 4, B and D) show that the magnitude of a primary CD8⫹ T cell response is proportional to the level of Ag expressed, an observation that we and others have made previously using different systems (6, 10, 36). This is a relationship that is often unappreciated but potentially very important in the shaping of effective immune responses. Our findings would appear to be at odds with the concept that meaningful TCD8⫹ responses are focused on a small number of the potential epitopes within an Ag or pathogen (61, 62) and that such epitopes are those expressed at the highest levels (63). Vac encodes ⬃200 proteins and the recombinant TK-frameshifted product that contains the OVA257–264 epitope is an exceedingly small part of the infection. Nevertheless, the OVA257–264-specific population is detectable and protective. Under the conditions of low epitope availability, the response will likely be comprised of only high avidity T cells (64) that, “pound-for-pound,” are more effective than a population raised to a higher level of epitope. In the system used in this study, the expression of frameshift-derived epitopes is driven by a viral vector (vaccinia), which produces high but obviously relevant levels of protein. Thus, it seems likely that a T cell population responding to endogenously produced frame-

Disclosures The authors have no financial conflict of interest.

References 1. Doherty, P. C., D. J. Topham, and R. A. Tripp. 1996. Establishment and persistence of virus-specific CD4⫹ and CD8⫹ T cell memory. Immunol. Rev. 150: 23– 44. 2. Yewdell, J. W., and J. R. Bennink. 2001. Cut and trim: generating MHC class I peptide ligands. Curr. Opin. Immunol. 13: 13–18. 3. Guidotti, L. G., and F. V. Chisari. 2000. Cytokine-mediated control of viral infections. Virology 273: 221–227. 4. Berke, G. 1995. The CTL’s kiss of death. Cell 81: 9 –12. 5. Sykulev, Y., M. Joo, I. Vturina, T. J. Tsomides, and H. N. Eisen. 1996. Evidence that a single peptide-MHC complex on a target cell can elicit a cytolytic T cell response. Immunity 4: 565–571. 6. Bullock, T. N., A. E. Patterson, L. L. Franlin, E. Notidis, and L. C. Eisenlohr. 1997. Initiation codon scanthrough versus termination codon readthrough demonstrates strong potential for major histocompatibility complex class I-restricted cryptic epitope expression. J. Exp. Med. 186: 1051–1058. 7. Malarkannan, S., T. Horng, P. P. Shih, S. Schwab, and N. Shastri. 1999. Presentation of out-of-frame peptide/MHC class I complexes by a novel translation initiation mechanism. Immunity 10: 681– 690. 8. Schwab, S., J. Shugart, T. Horng, S. Malarkannan, and N. Shastri. 2004. Unanticipated antigens: translation initiation at CUG with leucine. PLoS Biol. 2: e366. 9. Touriol, C., S. Bornes, S. Bonnal, S. Audigier, H. Prats, A.-C. Prats, and S. Vagner. 2003. Generation of protein isoform diversity by alternative initiation of translation at non-AUG codons. Biol. Cell 95: 169 –178. 10. Bullock, T. N., and L. C. Eisenlohr. 1996. Ribosomal scanning past the primary initiation codon as a mechanism for expression of CTL epitopes encoded in alternative reading frames. J. Exp. Med. 184: 1319 –1329. 11. Kozak, M. 2002. Pushing the limits of the scanning mechanism for initiation of translation. Gene 299: 1–34. 12. Saulquin, X., E. Scotet, L. Trautmann, M. A. Peyrat, F. Halary, M. Bonneville, and E. Houssaint. 2002. ⫹1 Frameshifting as a novel mechanism to generate a cryptic cytotoxic T lymphocyte epitope derived from human interleukin 10. J. Exp. Med. 195: 353–358. 13. Baranov, P. V., R. F. Gesteland, and J. F. Atkins. 2002. Recoding: translational bifurcations in gene expression. Gene 286: 187–201. 14. Atkins, J. F., P. V. Baranov, O. Fayet, A. J. Herr, M. T. Howard, I. P. Ivanov, S. Matsufuji, W. A. Miller, B. Moore, M. F. Prere, et al. 2001. Overriding standard decoding: implications of recoding for ribosome function and enrichment of gene expression. Cold Spring Harb. Symp. Quant. Biol. 66: 217–232. 15. Namy, O., J.-P. Rousset, S. Napthine, and I. Brierley. 2004. Reprogrammed genetic decoding in cellular gene expression. Mol. Cell 13: 157–168. 16. Jacks, T., M. D. Power, F. R. Masiarz, P. A. Luciw, P. J. Barr, and H. E. Varmus. 1988. Characterization of ribosomal frameshifting in HIV-1 gag-pol expression. Nature 331: 280 –283. 17. Dulude, D., M. Baril, and L. Brakier-Gingras. 2002. Characterization of the frameshift stimulatory signal controlling a programmed ⫺1 ribosomal frameshift in the human immunodeficiency virus type 1. Nucleic Acids Res. 30: 5094 –5102. 18. Gaudin, C., M. H. Mazauric, M. Traikia, E. Guittet, S. Yoshizawa, and D. Fourmy. 2005. Structure of the RNA signal essential for translational frameshifting in HIV-1. J. Mol. Biol. 349: 1024 –1035.

6934

INCIDENTAL RIBOSOME FRAMESHIFTING IN PROTECTIVE CD8 RESPONSE

19. Staple, D. W., and S. E. Butcher. 2005. Solution structure and thermodynamic investigation of the HIV-1 frameshift inducing element. J. Mol. Biol. 349: 1011–1023. 20. Brierley, I., and S. Pennell. 2001. Structure and function of the stimulatory RNAs involved in programmed eukaryotic-1 ribosomal frameshifting. Cold Spring Harb. Symp. Quant. Biol. 66: 233–248. 21. Giedroc, D. P., C. A. Theimer, and P. L. Nixon. 2000. Structure, stability and function of RNA pseudoknots involved in stimulating ribosomal frameshifting. J. Mol. Biol. 298: 167–185. 22. Biswas, P., X. Jiang, A. L. Pacchia, J. P. Dougherty, and S. W. Peltz. 2004. The human immunodeficiency virus type 1 ribosomal frameshifting site is an invariant sequence determinant and an important target for antiviral therapy. J. Virol. 78: 2082–2087. 23. Hung, M., P. Patel, S. Davis, and S. R. Green. 1998. Importance of ribosomal frameshifting for human immunodeficiency virus type 1 particle assembly and replication. J. Virol. 72: 4819 – 4824. 24. Shehu-Xhilaga, M., S. M. Crowe, and J. Mak. 2001. Maintenance of the Gag/ Gag-Pol ratio is important for human immunodeficiency virus type 1 RNA dimerization and viral infectivity. J. Virol. 75: 1834 –1841. 25. Weiss, R. B., D. M. Dunn, J. F. Atkins, and R. F. Gesteland. 1987. Slippery runs, shifty stops, backward steps, and forward hops: ⫺2, ⫺1, ⫹1, ⫹2, ⫹5, and ⫹6 ribosomal frameshifting. Cold Spring Harb. Symp. Quant. Biol. 52: 687– 693. 26. Gramstat, A., D. Prufer, and W. Rohde. 1994. The nucleic acid-binding zinc finger protein of potato virus M is translated by internal initiation as well as by ribosomal frameshifting involving a shifty stop codon and a novel mechanism of P-site slippage. Nucleic Acids Res. 22: 3911–3917. 27. Gallant, J. A., and D. Lindsley. 1998. Ribosomes can slide over and beyond “hungry” codons, resuming protein chain elongation many nucleotides downstream. Proc. Natl. Acad. Sci. USA 95: 13771–13776. 28. Atkins, J. F., B. P. Nichols, and S. Thompson. 1983. The nucleotide sequence of the first externally suppressible–1 frameshift mutant, and of some nearby leaky frameshift mutants. EMBO J. 2: 1345–1350. 29. Hwang, C. B., B. Horsburgh, E. Pelosi, S. Roberts, P. Digard, and D. M. Coen. 1994. A net ⫹1 frameshift permits synthesis of thymidine kinase from a drugresistant herpes simplex virus mutant. Proc. Natl. Acad. Sci. USA 91: 5461–5465. 30. Griffiths, A., S. H. Chen, B. C. Horsburgh, and D. M. Coen. 2003. Translational compensation of a frameshift mutation affecting herpes simplex virus thymidine kinase is sufficient to permit reactivation from latency. J. Virol. 77: 4703– 4709. 31. Horsburgh, B. C., H. Kollmus, H. Hauser, and D. M. Coen. 1996. Translational recoding induced by G-rich mRNA sequences that form unusual structures. Cell 86: 949 –959. 32. Wherry, E. J., K. A. Puorro, A. Porgador, and L. C. Eisenlohr. 1999. The induction of virus-specific CTL as a function of increasing epitope expression: responses rise steadily until excessively high levels of epitope are attained. J. Immunol. 163: 3735–3745. 33. Chakrabarti, S., K. Brechling, and B. Moss. 1985. Vaccinia virus expression vector: coexpression of ␤-galactosidase provides visual screening of recombinant virus plaques. Mol. Cell. Biol. 5: 3403–3409. 34. Eisenlohr, L. C., J. W. Yewdell, and J. R. Bennink. 1992. Flanking sequences influence the presentation of an endogenously synthesized peptide to cytotoxic T lymphocytes. J. Exp. Med. 175: 481– 487. 35. Yellen-Shaw, A. J., E. J. Wherry, G. C. Dubois, and L. C. Eisenlohr. 1997. Point mutation flanking a CTL epitope ablates in vitro and in vivo recognition of a full-length viral protein. J. Immunol. 158: 3227–3234. 36. Wherry, E. J., M. J. McElhaugh, and L. C. Eisenlohr. 2002. Generation of CD8⫹ T cell memory in response to low, high, and excessive levels of epitope. J. Immunol. 168: 4455– 4461. 37. Cossins, J., K. G. Gould, M. Smith, P. Driscoll, and G. G. Brownlee. 1993. Precise prediction of a Kk-restricted cytotoxic T cell epitope in the NS1 protein of influenza virus using an MHC allele-specific motif. Virology 193: 289 –295. 38. Eberl, G., A. Sabbatini, C. Servis, P. Romero, J. L. Maryanski, and G. Corradin. 1993. MHC class I H-2Kd-restricted antigenic peptides: additional constraints for the binding motif. Int. Immunol. 5: 1489 –1492. 39. Schild, H., M. Norda, K. Deres, K. Falk, O. Rotzschke, K. H. Wiesmuller, G. Jung, and H. G. Rammensee. 1991. Fine specificity of cytotoxic T lymphocytes primed in vivo either with virus or synthetic lipopeptide vaccine or primed in vitro with peptide. J. Exp. Med. 174: 1665–1668. 40. Dick, L. R., C. Aldrich, S. C. Jameson, C. R. Moomaw, B. C. Pramanik, C. K. Doyle, G. N. DeMartino, M. J. Bevan, J. M. Forman, and C. A. Slaughter. 1994. Proteolytic processing of ovalbumin and ␤-galactosidase by the proteasome to a yield antigenic peptides. J. Immunol. 152: 3884 –3894.

41. Parkin, N. T., M. Chamorro, and H. E. Varmus. 1992. Human immunodeficiency virus type 1 gag-pol frameshifting is dependent on downstream mRNA secondary structure: demonstration by expression in vivo. J. Virol. 66: 5147–5151. 42. Porgador, A., J. W. Yewdell, Y. Deng, J. R. Bennink, and R. N. Germain. 1997. Localization, quantitation, and in situ detection of specific peptide-MHC class I complexes using a monoclonal antibody. Immunity 6: 715–726. 43. Shastri, N., and F. Gonzalez. 1993. Endogenous generation and presentation of the ovalbumin peptide/Kb complex to T cells. J. Immunol. 150: 2724 –2736. 44. Sanderson, S., and N. Shastri. 1994. LacZ inducible, antigen/MHC-specific T cell hybrids. Int. Immunol. 6: 369 –376. 45. Huang, J., M. El-Gamil, M. E. Dudley, Y. F. Li, S. A. Rosenberg, and P. F. Robbins. 2004. T cells associated with tumor regression recognize frameshifted products of the CDKN2A tumor suppressor gene locus and a mutated HLA class I gene product. J. Immunol. 172: 6057– 6064. 46. Ripberger, E., M. Linnebacher, Y. Schwitalle, J. Gebert, and M. von Knebel Doeberitz. 2003. Identification of an HLA-A0201-restricted CTL epitope generated by a tumor-specific frameshift mutation in a coding microsatellite of the OGT gene. J Clin. Immunol. 23: 415– 423. 47. Elliott, T., H. Bodmer, and A. Townsend. 1996. Recognition of out-of-frame major histocompatibility complex class I-restricted epitopes in vivo. Eur. J. Immunol. 26: 1175–1179. 48. Hedley, M. L., K. Ozato, J. Maryanski, P. W. Tucker, and J. Forman. 1989. Expression of mutant H-2d proteins encoded by class I genes which alternatively process the 5⬘ end of their transcripts. J. Immunol. 143: 1026 –1031. 49. Slager, E. H., M. Borghi, C. E. van der Minne, C. A. Aarnoudse, M. J. Havenga, P. I. Schrier, S. Osanto, and M. Griffioen. 2003. CD4⫹ Th2 cell recognition of HLA-DR-restricted epitopes derived from CAMEL: a tumor antigen translated in an alternative open reading frame. J. Immunol. 170: 1490 –1497. 50. Ronsin, C., V. Chung-Scott, I. Poullion, N. Aknouche, C. Gaudin, and F. Triebel. 1999. A non-AUG-defined alternative open reading frame of the intestinal carboxyl esterase mRNA generates an epitope recognized by renal cell carcinomareactive tumor-infiltrating lymphocytes in situ. J. Immunol. 163: 483– 490. 51. Schwab, S. R., K. C. Li, C. Kang, and N. Shastri. 2003. Constitutive display of cryptic translation products by MHC class I molecules. Science 301: 1367–1371. 52. Wang, R. F., M. R. Parkhurst, Y. Kawakami, P. F. Robbins, and S. A. Rosenberg. 1996. Utilization of an alternative open reading frame of a normal gene in generating a novel human cancer antigen. J. Exp. Med. 183: 1131–1140. 53. Mayrand, S. M., D. A. Schwarz, and W. R. Green. 1998. An alternative translational reading frame encodes an immunodominant retroviral CTL determinant expressed by an immunodeficiency-causing retrovirus. J. Immunol. 160: 39 –50. 54. Ho, O., and W. R. Green. 2006. Cytolytic CD8⫹ T cells directed against a cryptic epitope derived from a retroviral alternative reading frame confer disease protection. J. Immunol. 176: 2470 –2475. 55. Cardinaud, S., A. Moris, M. Fevrier, P. S. Rohrlich, L. Weiss, P. Langlade-Demoyen, F. A. Lemonnier, O. Schwartz, and A. Habel. 2004. Identification of cryptic MHC I-restricted epitopes encoded by HIV-1 alternative reading frames. J. Exp. Med. 199: 1053–1063. 56. Shah, A. A., M. C. Giddings, J. B. Parvaz, R. F. Gesteland, J. F. Atkins, and I. P. Ivanov. 2002. Computational identification of putative programmed translational frameshift sites. Bioinformatics 18: 1046 –1053. 57. Gurvich, O. L., P. V. Baranov, J. Zhou, A. W. Hammer, R. F. Gesteland, and J. F. Atkins. 2003. Sequences that direct significant levels of frameshifting are frequent in coding regions of Escherichia coli. EMBO J. 22: 5941–5950. 58. Lelouard, H., E. Gatti, F. Cappello, O. Gresser, V. Camosseto, and P. Pierre. 2002. Transient aggregation of ubiquitinated proteins during dendritic cell maturation. Nature 417: 177–182. 59. Sykulev, Y., R. J. Cohen, and H. N. Eisen. 1995. The law of mass action governs antigen-stimulated cytolytic activity of CD8⫹ cytotoxic T lymphocytes. Proc. Natl. Acad. Sci. USA 92: 11990 –11992. 60. Kageyama, S., T. J. Tsomides, Y. Sykulev, and H. N. Eisen. 1995. Variations in the number of peptide-MHC class I complexes required to activate cytotoxic T cell responses. J. Immunol. 154: 567–576. 61. Berzofsky, J. A. 1988. Immunodominance in T lymphocyte recognition. Immunol. Lett. 18: 83–92. 62. Yewdell, J. W., and J. R. Bennink. 1999. Immunodominance in major histocompatibility complex class I-restricted T lymphocyte responses. Annu. Rev. Immunol. 17: 51– 88. 63. Gallimore, A., H. Hengartner, and R. Zinkernagel. 1998. Hierarchies of antigenspecific cytotoxic T-cell responses. Immunol. Rev. 164: 29 –36. 64. Alexander-Miller, M. A., G. R. Leggatt, and J. A. Berzofsky. 1996. Selective expansion of high- or low-avidity cytotoxic T lymphocytes and efficacy for adoptive immunotherapy. Proc. Natl. Acad. Sci. USA 93: 4102– 4107.