Antigen Bias in T Cell Cross-Priming

See allHide authors and affiliations

Science  28 May 2004:
Vol. 304, Issue 5675, pp. 1314-1317
DOI: 10.1126/science.1096268

This article has a correction. Please see:


Activated CD8+ T cells detect virally infected cells and tumor cells by recognition of major histocompatibility complex class I–bound peptides derived from degraded, endogenously produced proteins. In contrast, CD8+ T cell activation often occurs through interaction with specialized antigen-presenting cells displaying peptides acquired from an exogenous cellular source, a process termed cross-priming. Here, we observed a marked inefficiency in exogenous presentation of epitopes derived from signal sequences in mouse models. These data indicate that certain virus- and tumor-associated antigens may not be detected by CD8+ T cells because of impaired cross-priming. Such differences in the ability to cross-present antigens should form important considerations in vaccine design.

Cytotoxic CD8+ T cells recognize peptides that are presented on the cell surface by major histocompatibility complex (MHC) class I molecules. Two cellular pathways lead to this form of presentation. In classical antigen presentation, MHC class I molecules bind peptides that are generated via degradation of endogenous proteins (1). In an alternative pathway, professional antigen-presenting cells (APCs) internalize material derived from virally infected or transformed cells and then present the exogenously derived antigens to naïve CD8+ T cells, a process referred to as cross-presentation (2). T cell activation through cross-presentation (cross-priming), can be readily demonstrated for a number of antigens (35) but has not been observed for all antigens, including those that are efficiently presented through the endogenous pathway (6, 7).

Whether the parameters governing efficient exogenous and endogenous antigen presentation are equivalent is not clear. To investigate this issue, we generated two green fluorescent protein (GFP) fusion gene constructs that both encode two MHC class I Db-restricted epitopes. In the first construct (sNP366-GFP-E749), the epitope NP366 (derived from influenza A) was inserted into the N terminus of the signal peptide of a secreted GFP molecule. The second epitope, the antigen E749 (derived from human papilloma virus 16), was inserted close to the C terminus of GFP (Fig. 1A) (8). In the second construct, the same epitopes were introduced but in reverse order (sE749-GFP-NP366) (Fig. 1A). The introduction of each CTL (cytotoxic T lymphocyte) epitope into the hydrophilic fragment of the signal peptide did not affect signal peptide cleavage or intracellular transport of GFP when compared with a GFP molecule harboring an uninterrupted wild-type signal peptide (sGFP-NP) (Fig. 1, B and C) (figs. S1 and S2). This introduction of NP366 and E749 epitopes into the GFP fusion genes allowed us to monitor antigen presentation of two epitopes present in the same protein synthesis product but located in two different protein fragments. Assessing the presentation of signal peptide–encoded epitopes seemed particularly useful because signal peptides represent an important source of endogenously produced MHC class I–binding peptides (912).

Fig. 1.

Endogenous antigen presentation of T cell epitopes from signal peptides and mature protein segments. (A) Structure of the sNP366-GFP-E749 and sE749-GFP-NP366 fusion gene products. The arrowhead indicates the signal peptide cleavage site. (B) Secretion of GFP variants. RMA sE7-GFP-NP cells (a), RMA sNP-GFP-E7 cells (b), RMA cells (c), and RMA GFP-NP cells (cytosolic location; d) were treated with brefeldin A for 5 hours (white area) or left untreated (gray area). GFP expression was determined by flow cytometry analysis; the y axis represents cell counts. (C) The introduction of the NP366 or E749 epitope into signal peptides does not affect signal peptide cleavage. Cellular GFP products (c) were obtained by immunoprecipitation with polyclonal rabbit antibody to GFP from RMA cells expressing cytosolic GFP-NP (cy), sGFP-NP containing the wild-type signal peptide (s), sE7-GFP-NP (sE), or sNP-GFP-E7 (sN). Cells were starved in Met/Cys-free Dulbecco's modified Eagle's medium for 60 min before a 30-min pulse labeling with [35S]Met/Cys in DMEM containing 5% fetal bovine serum. In vitro transcription/translation (v) was performed with standard rabbit reticulocyte system (Promega) according to the manufacturer's protocol. Samples were analyzed by SDS–polyacrylamide gel electrophoresis. Compare lanes 6 and 7 (s), 9 and 10 (sE), and 11 and 12 (sN). (D and E) Analysis of NP366 antigen presentation of RMA sNP-GFP-E7 (▲), RMA sE7-GFP-NP (⚫), RMA (×), TAP-deficient RMA-S sNP-GFP-E7 (△), and RMA-S sE7-GFP-NP (◯) cells. Percentages of IFN-γ–producing NP366-specific CD8+ T cells derived from in vitro restimulated spleen cells of NP366-vaccinated mice (8) were determined by intracellular cytokine staining (Pharmingen), either upon incubation with the indicated cell transfectant (D) or upon peptide stripping and subsequent loading of serial dilutions on D1 target cells (E) (8).

Presentation of the NP366 and E749 epitopes via the endogenous pathway was examined by introducing the GFP fusion genes into the RMA tumor cell line (13). Cells containing GFP with the NP366 epitope located in either the signal peptide or the mature protein were recognized equally well by NP366-specific CD8+ T cells (Fig. 1D) (8). Similarly, the E749 epitope was presented efficiently to antigen-specific T cells, whether located in the signal peptide or the mature protein (14). Because saturation of T cell recognition at high antigen densities could potentially mask differences in presentation efficiency, a more quantitative measure of the antigen processing of T cell epitopes was required. To achieve this, we isolated peptides from each transfected cell line by acid elution and used them to sensitize target cells for CTL recognition. In this system, the epitopes contained in the mature protein fragment or in the signal peptide were again presented with comparable efficiency (Fig. 1E). Hence, for this antigen, location within the signal peptide or GFP moiety apparently does not affect the efficiency of endogenous presentation. Endogenous presentation of the NP366 epitope present within the hydrophilic segment of the signal peptide was in large part dependent on the transporter associated with antigen processing (TAP), because presentation of NP366 by TAP-deficient RMA-S cells containing sNP366-GFP-E749 was inefficient (Fig. 1, D and E) (9, 15). Similarly, endogenous presentation of the NP366 epitope contained within the mature protein required TAP function (Fig. 1, D and E).

To investigate whether antigen location might influence T cell priming by cross-presentation, we challenged naïve mice with RMA-S cells that contain the sNP366-GFP-E749 gene construct. Efficient induction of T cell immunity against the E749 epitope located within the mature protein was made evident by staining of peripheral blood cells with MHC tetramers containing E749 (Fig. 2A). In contrast, the NP366 epitope contained in the signal peptide of the same fusion gene induced only low numbers of antigen-specific T cells. To determine whether this difference was a direct result of the epitope location within the protein or was due to other factors, such as a difference in T cell precursor frequency, we challenged mice with RMA-S cells containing the fusion gene with the two epitopes in reverse order (sE749-GFP-NP366). In this setting, mice developed a pronounced NP366-specific T cell response, whereas the E749 epitope located within the signal peptide failed to induce efficient T cell immunity, as judged by MHC tetramer staining and by ex vivo interferon-γ (IFN-γ) production (Fig. 2, A and B). In all mice examined (n = 40; P < 0.0001), T cells specific for the epitope located in the mature protein outnumbered those specific for the signal peptide–encoded epitope. Antigen-specific T cell responses to the C-terminal antigens E749 and NP366 were stronger, on average by factors of 30 and 16, respectively, than T cell responses against the same epitope when present in the signal peptide. Collectively, these findings indicate that T cell priming by cross-presentation is considerably more efficient for epitopes derived from mature proteins than for epitopes contained within signal peptides.

Fig. 2.

In vivo cross-priming is biased toward epitopes located within the mature protein. (A) Naïve C57/Bl10 mice were subcutaneously challenged with 2 × 106 cells of the indicated RMA-S transfectant. Percentages of NP366-specific (top) and E749-specific (bottom) CD8+ T cells in peripheral blood were determined at the indicated time points with the use of phycoerythrin-conjugated antibody to CD8β (Pharmingen) and APC-conjugated NP49- or E7366-containing MHC tetramers (27, 28). Each dot represents an individual mouse; bars denote average percentage. (B) Antigen-specific T cells induced by cross-priming are functional. Twelve days after tumor inoculation, IFN-γ production of CD8+ T cells was determined directly ex vivo upon a 4-hour stimulation with 0.1 μM NP366 or E749 peptide in the presence of brefeldin A (top) and interleukin-2, and was compared with MHC tetramer staining (bottom). (C and D) MHC tetramer staining of peripheral blood cells (C) or splenocytes (D) from H-2d × H-2b F1 mice that were challenged with 107 cells of the indicated p815 transfectant, measured directly ex vivo (C) or after 7 days in vitro restimulation with the appropriate peptide (D). Identical results were obtained with H-2d–restricted mouse embryonic cells that contained either construct (14).

We wanted to ascertain that the T cell responses induced by tumor inoculation were a result of cross-presentation and were not due to a putative residual capacity of RMA-S cells to present endogenous antigen in vivo. Therefore, we sought to establish whether the observed antigen bias would also be evident in a system in which tumor cells lacked the relevant MHC allele (2). Thus, we introduced the two GFP fusion gene constructs into p815 cells (H-2d background and therefore lacking the required H-2b–presenting allele) and used them to challenge C57/Bl6 × Balb/c F1 (H-2b × H-2d) mice. Efficient antigen-specific T cell induction to E749 occurred when this epitope was located within the mature protein but, again, not when located within the signal sequence (Fig. 2C). Ex vivo T cell responses to NP366 were undetectable in this setting (14), potentially because of strain-dependent differences in NP366-specific T cell precursor frequency (16). However, when spleen cells of the challenged mice were analyzed for NP366-specific T cell responses upon in vitro restimulation, a similar bias toward the NP366 epitope encoded within the mature protein was apparent (Fig. 2D).

The selective priming of T cell responses against epitopes contained within mature antigens could potentially be a consequence of immunoglobulin (Ig)–protein complex formation. Specifically, Igs recognizing the (secreted) mature protein might form Ig-protein complexes that could be taken up by dendritic cells via Fc receptor–mediated endocytosis (17, 18). To determine whether Ig-protein complex formation is required for the observed bias in antigen-specific T cell immunity, we challenged μ-deficient mice, which lack B cells (19), with RMA-S cells containing either GFP fusion gene construct. Again, T cell responses to the mature antigens were efficiently induced, whereas T cell immunity induced by signal peptide–encoded epitopes was marginal (Fig. 3A). Thus, Ig-protein complex formation does not measurably influence the antigen bias observed in this system.

Fig. 3.

Differential exogenous presentation of epitopes from signal peptides and mature proteins by dendritic cells. (A) In vivo T cell induction through cross-priming is independent of Igs. NP336-specific (top) and E749-specific (bottom) CD8+ T cell responses induced in μ-deficient mice upon tumor inoculation with 2 × 106 RMA-S sNP366-GFP-E749 cells (left) or with RMA-S sE749-GFP-NP366 cells (right). (B) Epitopes derived from mature antigens are more efficiently cross-presented than epitopes located within a signal peptide. Irradiated RMA-S sNP-GFP-E7 cells (▲), RMA-S sE7-GFP-NP cells (⚫), and RMA-S cells (×) were fed to D1 cells, NP336-specific fluZ hybridoma cells were added, and T cell activation was determined (8). (C) Antigen bias is dependent on signal peptide location. C57/Bl6 mice were inoculated with 5 × 106 RMA-S sE749-GFP-NP366 cells (left) or RMA-S s*E749-GFP-NP366 cells (right). Peripheral blood was analyzed at the indicated time points for NP366-specific (top) and E749-specific (bottom) CD8+ T cells by MHC tetramer flow cytometry.

In an alternative scenario, the difference in the capacity of the two protein fragments to induce T cell immunity by cross-priming might result from a difference in the ability of APCs to extract and present the two epitopes from cells or cellular remnants. To test this idea, we incubated immature murine D1 dendritic cells (20) with irradiated RMA-S cells containing either GFP fusion gene construct. Subsequent cross-presentation of the NP366 epitope was more efficient by a factor of 10 when derived from the mature protein than when located within the signal peptide (Fig. 3B). Thus, the superior level of T cell induction observed toward epitopes within mature proteins can, at least in part, be attributed to differences in the ability of APCs to cross-present the two classes of antigens.

The observed antigen bias could be a consequence of a specific inability to cross-present T cell epitopes present in signal peptides, or it could reflect a negative effect of N-terminal location of the epitope. As a direct test, we generated a fusion gene construct in which the signal peptide was rendered dysfunctional by removal of the hydrophobic segment that is bound by the signal recognition particle (s*E749-GFP-NP366, figs. S3 and S4) (8). Mice that were challenged with RMA-S cells transduced with the s*E749-GFP-NP366 construct developed an E749-specific T cell response whose magnitude was equal to or greater than that of the NP366-specific T cell response (Fig. 3C). These data indicate that poor cross-priming is a direct consequence of location of the epitope within a functional signal peptide.

Prior experiments have sought to determine the contribution of direct priming and cross-priming in the induction of tumor- and virus-specific T cell responses by disrupting either endogenous or exogenous antigen presentation (37). Our data show that epitopes derived from signal peptides are efficiently presented through the endogenous pathway but not through the exogenous pathway. Hence, we can analyze the contribution of these two pathways in T cell induction in a setting where both are operational. We challenged naïve mice with RMA cells that contained either GFP fusion gene construct. Despite the clear efficiency in endogenous presentation of both epitopes seen in in vitro assays (Fig. 1, D and E), antigen-specific T cell responses induced in vivo were skewed toward the epitopes contained in the mature antigen, whether NP366 or E749 (Fig. 4). Thus, exogenous antigen presentation proved to be the dominant mechanism for T cell induction, at least in this model system.

Fig. 4.

T cell responses induced after challenge with MHC-proficient RMA cells are biased toward epitopes contained within mature antigens. C57/Bl10 mice were subcutaneously challenged with 5 × 105 RMA sNP366-GFP-E749 cells (left) or RMA sE749-GFP-NP366 cells (right). Peripheral blood was analyzed at the indicated time points for NP366-specific (top) and E749-specific (bottom) CD8+ T cells by MHC tetramer flow cytometry.

Our data show a marked divergence in the efficiency of endogenous and exogenous antigen presentation. The fact that epitopes that are efficiently presented through the endogenous pathway are not in all cases efficiently presented upon exogenous presentation may help to explain the lack of cross-priming previously observed in several model systems (6, 7). In this regard, it is useful to note that the lymphocytic choriomeningitis virus (LCMV)–derived GP33 epitope, which has been shown to fail in inducing T cell immunity via cross-priming, is derived from the LCMV GP signal peptide (9). Another interesting issue to address will be whether these findings extend to the efficiency of tolerance induction against peripheral antigens that are not expressed within the thymus. Thus, epitopes derived from peripheral antigens that fail to efficiently enter the cross-presentation pathway may be less likely to be exposed to naïve T cells and may therefore not induce peripheral tolerance (21, 22). Such “ignored” peripheral antigens may prove particularly interesting targets for the induction of tumor-specific T cell immunity, as the T cell repertoire against these antigens may not have been affected by cross-tolerance (23).

How might the bias against signal peptide epitopes be explained at the molecular level? Potentially, signal peptides assemble with cellular factors that reduce their accessibility to the exogenous presentation pathway. Alternatively, and perhaps more likely, if exogenous presentation involves transfer of unprocessed antigens (24), this is likely to be inefficient for molecules that are degraded rapidly after synthesis (25). In this latter model, exogenous presentation would depend on the steady-state antigen levels in the donor cell, as opposed to the protein synthesis rate for endogenous antigen presentation (26).

The current data suggest a fundamental difference between endogenous presentation and exogenous presentation, a divergence that is likely to influence immunogenicity and immunodominance in virus- and tumor-induced T cell responses. In addition, because vaccine-induced T cell responses often rely on cross-presentation, understanding the rules that determine the efficiency of exogenous presentation should be important in helping to optimize vaccine design.

Supporting Online Material

Materials and Methods

Figs. S1 to S4


References and Notes

View Abstract

Navigate This Article