Furin-Processed Antigens Targeted to the Secretory Route Elicit Functional TAP1^−/−CD8⁺ T Lymphocytes In Vivo¹

Peptides produced by the cellular proteolytic machinery have been evolutionarily reused for immune surveillance. Cells display a sample of their own peptides associated with surface MHC class I molecules, but in pathological circumstances such as viral infection and tumoral transformation, cells change their protein expression pattern and display novel peptides. These can be recognized as antigenic epitopes by specific CD8⁺ CTL, which eliminate epitope-bearing cells (1).

Most of the studied MHC class I epitopes depend on direct cytosolic proteasome activity for their production. After peptide transport to the endoplasmic reticulum (ER)⁵ by the TAP (2), subsequent proteolytic trimming is frequently required (1). Considering the gorgeous molecular diversity contributed by infectious microorganisms and MHC class I alleles, nonproteasomal proteolytic activities could contribute to ensure pathogen control. However, their relevance might be concealed by the proteasome overwhelming activity. Proteasome-deficient cells or animals are not viable. In contrast, TAP-deficient mice (3) are functionally deprived of proteasome products for MHC class I Ag presentation and, in this respect, constitute a useful tool to explore the role of proteasome-independent proteolysis in MHC class I Ag processing. There is evidence for TAP-independent pathways of Ag presentation by MHC class I molecules (4, 5). The proteases involved include ER signal peptidase, trans-Golgi network furin and cathepsins for endogenously synthesized proteins (6, 7, 8, 9, 10) and for internalized Ags (11, 12, 13, 14).

A 10-to 20-fold smaller CD8 compartment is present in TAP-deficient humans (15) and mice, which have a reduced functional repertoire to some Ags and a reduced response to OVA (16). Surprisingly, given the key role of MHC class I presentation and CD8⁺ T lymphocytes in the control of viruses, TAP^−/− humans and mice are not abnormally susceptible to viral infections. The clinical phenotype of TAP-deficient human beings, who have a limited immunodeficiency with chronic respiratory bacterial infections, illustrates that the TAP-independent pathways may be sufficient to allow the control of viral infections. Indeed, in vivo TAP-independent priming of CD8⁺ T lymphocytes has been demonstrated in TAP-deficient mice (14, 17, 18, 19) and human beings (20). There is a need, however, for quantitative evaluation of the contribution of these supplementary pathways to Ag presentation in vitro and in vivo. This would help test the hypothesis that noncytosolic-processing mechanisms may compensate in vivo for the lack of proteasome contribution to Ag processing in these persons genetically deficient in TAP.

Furin, a proprotein convertase located mainly in the trans-Golgi network, mediates maturation of many proproteins by cleaving at precise stretches of three to four basic residues (21). Notably, when chimeric proteins containing a MHC class I epitope downstream of furin cleavage sites are directed to the secretory pathway in TAP-negative cells, so that there is no contribution from cytosolic epitopes, antigenic presentation is elicited in a strictly furin-dependent fashion (8, 9). Furin is also essential for generating MHC class I ligands from some exogenous Ags (11, 12). Considering these capabilities, we have explored the potential of the furin-mediated secretory pathway of Ag processing to generate in vivo a compensatory and effective CD8⁺ T lymphocyte response in TAP-deficient mice and the potential relevance of this mechanism to contribute to overall MHC class I Ag presentation also in TAP-positive individuals.

C57BL/6 mice (H-2^b haplotype) and TAP1-deficient B6.129S2-Tap1^tm1Arp/J mice (3) (stock no. 002944; The Jackson Laboratory) were bred in our colony in accordance with national regulations. Animal studies were approved by Instituto de Salud Carlos III’s Review Board. Mice were infected i.p. with 10⁸ PFU of recombinant vaccinia virus (rVACV).

T2/K^b is a human lymphoblastoid T2 cell line deficient in TAP and transfected with K^b (22). It was cultured at 37°C in RPMI 1640 medium supplemented with 10% FBS and 5 × 10⁻⁵ M 2-ME in a 5% CO₂ atmosphere and was provided by Dr. H. G. Ljunggren (Karolinska Institutet, Stockholm, Sweden). L/K^b derive from L murine fibroblasts, express TAP, and are transfected with K^b (23). They were cultured at 37°C in IMDM supplemented with 10% FBS in a 9% CO₂ atmosphere and were provided by Drs. J. W. Yewdell and J. Bennink (National Institutes of Health, Bethesda, MD).

Bone marrow-derived dendritic cells (DC) were generated from bone marrow progenitors as described previously (24). Briefly, freshly prepared bone marrow cells were cultured as the T2/K^b cells in the presence of 200 U/ml GM-CSF (PeproTech) and fed with fresh GM-CSF on days 3 and 6. After 7 days, the nonadherent cells were collected and shown to have a typical DC morphology with a myeloid DC phenotype (MHC class II⁺, CD11c⁺, CD8⁻).

Infection of T2/K^b with rVACV was performed at a multiplicity of infection of 30 PFU/cell at a concentration of 10⁷ cells/ml. Multiplicity was adjusted so that a rVACV expressing GFP infected >95% cells. After 1 h of adsorption, virus inoculum was thoroughly washed, and infection was allowed overnight. DC were infected at a multiplicity of 10 for 4–5 h. L/K^b cells were infected at a multiplicity of 5 for overnight infections and at a multiplicity of 30 for 4-h infections to study the effect of inhibitors. In the latter infections, cells were treated with the inhibitor 15 min before virus adsorption, and inhibitor was kept throughout adsorption and infection. Brefeldin A (BFA) (Sigma-Aldrich) was used at 10 μg/ml, the proteasome inhibitor lactacystin (LC) (Dr. E. J. Corey, Harvard University, Cambridge, MA) at 10 μM, and furin inhibitor decanoyl-RVKR-chloro-methyl-ketone (decRVKR-cmk) (Bachem) at 50 μM. Percentage-specific inhibition was calculated from formula 100 × {[(X − N) − (X_i − N_i)]/(X − N)}, where X represents the percentage of total CD8⁺ lymphocytes that were stained with mAb to IFN-γ when targets were infected with rVACV, X_i is the value when infection was conducted in presence of the inhibitor, and N and N_i are the equivalent values when targets were infected with negative control rVACV-hepatitis B virus secretory pre-core protein (HBe). As positive control, cells were infected with this virus and pulsed during the 1 h adsorption period with 10⁻⁸ M SIINFEKL peptide.

A series of rVACV based on the wild-type Copenhagen strain and encoding the chicken OVA octamer ²⁵⁷SIINFEKL²⁶⁴ that is presented by K^b in a chimeric protein context were produced. Chimeric proteins had insertions of the ²⁵⁷SIINFEKL²⁶⁴ sequence at position 179 at the carboxyl terminus of the natural hepatitis B virus cytosolic core protein (HBc) or HBe (see Fig. 1). The chimeric protein cC-Ova was based on HBc and was thus expressed in the cytosol, because it lacks a signal sequence. Chimeras sC-Ova and sC-OvaΔC were based on HBe and entered the secretory pathway. cC-Ova, sC-Ova, and sC-OvaΔC DNA constructs were produced from plasmids encoding cC-A9A or sC-A9A (9). DNA sequence encoding RAAAAAYPHFMPTNLAAAAASRESQC was removed by AvaI and BstXI restriction enzymes (Roche) and replaced by sequences encoding RASIINFEKLSRESQC and RASIINFEKL, respectively. All genes were cloned under the control of the vaccinia early-late promoter 7.5k.rVACV were produced as previously described (25) and DNA sequence confirmed. The recombinant encoding hepatitis B virus secretory protein, rVACV-HBe (26), was used as negative control. rVACV-OVA encoding full-length OVA, rVACV-OVA_257–264 encoding the miniprotein MSIINFEKL, and rVACV-ES-OVA_257–264 encoding the equivalent ER-targeted miniprotein were based on vaccinia WR strain (27) and were provided by Drs. J. W. Yewdell and J. Bennink (National Institutes of Health, Bethesda, MD).

Detection of K^b/SIINFEKL complexes was made with 25-D1.16 Ab (28). Purified Ab was provided by Drs. J. Yewdell and J. Bennink and labeled with Alexa Fluor 647 (Molecular Probes) according to manufacturer specifications. Infected or peptide-pulsed cells were incubated for 30 min on ice with the Alexa Fluor 647-labeled 25-D1.16 Ab and after two washes incubated 30 min on ice with an Alexa Fluor 647-labeled rabbit anti-mouse IgG Ab (Molecular Probes) to increase the specific signal. Cells were acquired using FACSCalibur or FACSCanto flow cytometers (BD Biosciences), and data were analyzed using CellQuest software (BD Biosciences). When results from several experiments were averaged, the mean and SEM were calculated throughout this report.

Generation of CTL lines has been described previously (29). Briefly, splenocytes were obtained from rVAVC-OVA-infected C57BL/6 mice. SIINFEKL-specific CTL lines were generated by stimulation with 10⁻⁸ M synthetic SIINFEKL peptide and restimulated weekly. Recombinant human IL-2 for their long-term propagation was generously provided by the National Cancer Institute Preclinical Repository.

OVA²⁵⁷SIINFEKL²⁶⁴ peptide was synthesized in an Applera peptide synthesizer model 433A, purified, and found homogeneous by HPLC analysis.

ICS assays to detect recognition by CTL of Ags presented by infected cells were performed as previously described (30) with modifications (31). CTL lines were stimulated for 4 h or overnight in the presence of 10 μg/ml BFA (Sigma-Aldrich) with target cells infected with rVACV. When exogenous peptide was used, it was added during infection and during ICS assay coculture. Coculture was performed at an E:T of 1:5. In ex vivo ICS assays, which were performed 7 days after infection of mice, splenocytes were stimulated with an excess 10⁻⁶ M peptide for up to 2 h, and stimulated for an additional 3 h in the presence of BFA. After stimulation during coculture, cells were incubated with FITC-conjugated anti-CD8α mAb (Proimmune), fixed, and incubated with PE-conjugated mAb to IFN-γ (BD Biosciences) during permeabilization (DakoCytomation). Events were acquired using a FACSCalibur flow cytometer (BD Biosciences), and data were analyzed using CellQuest software (BD Biosciences). An average of 5000 CD8⁺ cells was analyzed in each sample. Background activation obtained with nonpulsed cells or with cells infected with negative control virus (usually 0–5% for CTL lines and 0–0.1% for ex vivo assays) was subtracted.

In vivo cytotoxicity assays were performed as published (32, 33) with modifications. Spleens were obtained, erythrocytes were removed, and C57BL/6 and TAP1^−/− splenocytes were split in two populations and labeled with either a high concentration (5 μM) or a low concentration (0.5 μM) of CFSE. After washing excess CFSE, C57BL/6 and TAP1^−/− CFSE^high cells were pulsed with 10⁻⁶ M SIINFEKL for 30 min at 37°C. Excess peptide was washed at least twice, and CFSE^high peptide-pulsed cells were mixed with equal numbers of CFSE^low cells. A total of 8 × 10⁶ cells of the C57BL/6 mixed suspension were i.p. injected into each C57BL/6 mouse, which had been infected i.p. 7 days earlier. The same protocol was followed for TAP1^−/− cells and mice. Two days later, the peritoneal cavity was lavaged, and spleens were extracted, and the cells were analyzed by flow cytometry to measure in vivo killing, using a FACSCanto flow cytometer (BD Biosciences). Data were analyzed using FACSDiva software (BD Biosciences). Specific lysis was calculated as published (34) according to the formula: [1 − (ratio unprimed/ratio primed) × 100], where the ratio unprimed is %CFSE^low/%CFSE^high cells remaining in control HBe-infected recipients, and the ratio primed is %CFSE^low/%CFSE^high cells remaining in experimental infected recipients.

In our experimental system, rVACV encoded chimeric constructs of carrier proteins with an inserted CTL epitope (Fig. 1). Modified HBe was selected as carrier protein because it is a well-known furin natural target previously used to study furin-dependent Ag processing (8, 9). Cytosolic HBc was used as control carrier protein. The epitope is inserted in the chimeric protein downstream the furin sites (arrows in Fig. 1). After furin-specific cleavage, the epitope is contained within the carboxyl-terminal released peptides.

We selected SIINFEKL peptide, the broadly studied epitope from OVA, Ova_257–264, which is presented by the MHC class I molecule H-2K^b. It allows the use of 25-D1.16 mAb, which specifically detects SIINFEKL associated with H-2K^b (28), to quantify Ag processing, and of H-2^b TAP-deficient mice, for in vivo assays.

Three different constructs were generated: 1) an HBc-based cytosolic one (cC-Ova), which cannot gain access to the ER because it has no signal sequence and thus can only be processed in the cytosol; 2) an HBe-based secretory one (sC-Ova) identical to cC-Ova but with a signal sequence to enter the ER, allowing furin-dependent release of the epitope, which is flanked by amino- and carboxyl-terminal extensions; and 3) another secretory construct (sC-OvaΔC) identical to sC-Ova, except for the carboxyl-terminal extension flanking the epitope, which has been deleted.

In a first set of experiments, TAP⁺ L/K^b fibroblasts were infected with rVACV expressing sC-Ova and cC-Ova. As shown in Fig. 2 A, left panels, K^b/SIINFEKL complexes were produced from both constructs and were readily detected by flow cytometry by using 25-D1.16 Ab over the background staining generated by a control rVACV that only encodes the carrier protein HBe.

In contrast, and as expected, in TAP-deficient T2/K^b cells cytosolic cC-Ova did not produce K^b/SIINFEKL complexes detected by 25-D1.16 Ab (Fig. 2,A, right panels), because these cells are not able to transport cytosolic peptides to the ER. Secretory sC-Ova did not yield complexes either, despite its potential for furin-mediated Ag processing in the secretory pathway. Considering the chimera design (Fig. 1), peptides released by furin would contain the epitope extended in both amino and carboxyl termini. Thus, they would require further exo- or endopeptidase activity to trim them to the minimal SIINFEKL. Therefore, in an attempt to improve Ag presentation, the new sC-OvaΔC construct was developed with no carboxyl-terminal extension after the epitope (Fig. 1). The sC-OvaΔC construct generated K^b/SIINFEKL complexes, detected by 25-D1.16 staining, not only in rVACV-infected TAP-competent L/K^b cells, as control (Fig. 2,A, left panels), but also in TAP-deficient T2/K^b cells (Fig. 2 A, right panels), suggesting that sC-OvaΔC is processed in the secretory pathway.

Ag presentation by rVAVC-expressed constructs was also assessed in primary bone marrow-derived DC prepared from TAP⁺ and TAP⁻ mice, allowing comparison of TAP-dependent and TAP-independent pathways in paired cells (Fig. 2,B). In TAP⁻ DC, only the strong sC-OvaΔC protein was presented, whereas the carboxyl-terminal extension in sC-Ova severely limited its presentation and precluded detection with the TCR-like Ab (Fig. 2 B), as shown before in fibroblast cell lines. None of the cytosolic constructs, including cC-Ova and full-length Ova, was detected in TAP⁻ DC. In TAP⁺ DC, where all constructs have also access to the classical proteasome-dependent pathway of Ag presentation, presentation from sC-Ova was weakly positive. The number of complexes produced by cC-Ova and by sC-OvaΔC was higher and similar at all times postinfection. Interestingly, there was a trend for a more efficient presentation from the secretory construct sC-OvaΔC in TAP⁻ DC than in TAP⁺ DC.

The presence of K^b/SIINFEKL complexes at the plasma membrane can also be detected by a SIINFEKL-specific CTL line. Detection of CTL activation by intracellular staining of IFN-γ accumulation is saturated at lower complex levels than complex detection by 25-D1.16, and it is at least 1000-fold more sensitive (Fig. 3,A). Despite the strong increase in sensitivity, the sC-Ova construct with the carboxyl-terminal extension, which was not detected in TAP-deficient T2/K^b-infected cells by 25-D1.16 staining (Fig. 3,B), remained undetected by specific CTL (Fig. 3 C). In contrast, the sC-OvaΔC protein was positive by both techniques. This suggests that sC-OvaΔC produces at least 1000-fold more K^b/SIINFEKL complexes than sC-Ova, although presentation from the sC-Ova construct of a related peptide of different size and marginal antigenicity cannot be excluded. Only when a 20-fold even more sensitive SIINFEKL-specific CTL line was used, sC-Ova-infected T2/K^b cells were finally recognized by specific CTL (data not shown), confirming the integrity of the construct.

Overall, these results show that the SRESQC carboxyl-terminal extension of SIINFEKL in the sC-Ova construct strongly limits Ag processing in the secretory pathway by an estimated factor of at least 1000-fold. Additional deletion of amino-terminal extensions of the epitope in secretory minigene-like constructs led to a 20-fold higher number of complexes than in sC-OvaΔC (data not shown).

Processing of sC-OvaΔC in the secretory pathway was further characterized taking advantage of 25-D1.16 staining. As shown in Fig. 4 B, sC-OvaΔC-infected TAP⁻ T2/K^b cells generated K^b/ SIINFEKL complexes in a process that was sensitive to BFA and was also blocked by the furin inhibitor decRVKR-cmk, pointing out that Ag processing depended both on Ag from an endogenous source, most likely endogenously expressed Ag, and on furin activity, respectively. Presentation of the SIINFEKL epitope from TAP⁻ cells to specific CTL was unaffected by proteasome inhibitors (data not shown), as demonstrated earlier for the CMV 9pp89 epitope (8).

Furin is a rate-limiting enzyme in this pathway, because its inhibition fully prevents presentation from TAP-deficient cells to 9pp89-specific CTL (8), and its overexpression potentiates presentation of the 9pp89 epitope (9). Furin inhibition also blocked SIINFEKL presentation from TAP-deficient cells to specific CTL (data not shown). Furthermore, overexpression of furin by coinfection with a rVAVC-furin also increased complex formation from sC-OvaΔC in TAP-deficient cells by 2-fold (data not shown), further confirming furin involvement. These results show that sC-OvaΔC is processed to generate SIINFEKL in the furin-mediated secretory pathway.

Another point of interest was exploring if sC-OvaΔC also followed a cytosolic processing in TAP-competent cells and, if so, quantifying the relative contribution of cytosolic vs secretory pathway. Fig. 4 A displays a representative 25-D1.16 histogram of sC-OvaΔC-infected TAP-competent L/K^b cells in the absence or in the presence of the proteasome inhibitor LC or the furin inhibitor decRVKR-cmk. Staining by 25-D1.16 was strongly reduced by LC treatment, demonstrating the existence also of a proteasome-dependent cytosolic Ag-processing pathway for this kind of secretory constructs. Furin inhibition by decRVKR-cmk also reproducibly diminished 25-D1.16 staining to a lesser extent.

Next, L/K^b cells were infected with additional control rVACV in the presence of inhibitors and assayed for 25-D1.16 staining and for CTL activation. Fig. 4,C displays in a bar diagram the 25-D1.16 staining of L/K^b cells infected with cC-Ova, sC-OvaΔC, or a minigene-encoding rVACV that directly expresses MSIINFEKL peptide in the cytosol. Results represent the relative fluorescence intensity of each rVACV with respect to the control condition in the absence of inhibitor. Secretory sC-OvaΔC (Fig. 4,C, middle) showed an inhibition trend by LC and decRVKR-cmk similar to that seen in Fig. 4,A, with an inhibition of 39 ± 6% (n = 3) by the furin inhibitor. As expected, and taking into account the detection limit of the technique, LC completely blocked generation of K^b/SIINFEKL complexes from cytosolic cC-Ova in infected cells. This confirms the inhibitory capability of the LC batch used. A slight effect on cells infected with the rVACV expressing the proteasome-independent preprocessed cytosolic miniprotein (Fig. 4C, right) might be explained by a LC effect on vaccinia late gene expression (35). As expected, BFA abrogated K^b/SIINFEKL complex formation from all three constructs. Thus, the furin pathway contributed with roughly one-third of all surface K^b/SIINFEKL complexes.

CTL activation by the same infected cells is shown in Fig. 4,D. Presentation to CTL apparently was not affected by the furin inhibitor alone, but these data cannot be interpreted, because it cannot be confirmed that the level of Ag presentation lied below the saturation range for CTL. Complexes generated in sC-OvaΔC-infected TAP-competent L/K^b cells (Fig. 4,D, middle) by proteasome-independent processing, that is, those detected in the presence of LC, still sufficed to activate half the CTL. In this situation of nonsaturating CTL reactivity, the furin-mediated secretory pathway was also blocked by using both LC and decRVKR-cmk. This further diminished recognition by CTL, indicating that furin-dependent generation of K^b/SIINFEKL complexes also contributes to CTL activation, although to a lesser extent than the proteasome does. These results concur with the respective number of complexes quantified with 25-D.1-16 Ab above. As it is also shown in Fig. 4 D, inhibitors in control constructs functioned as expected: on the one hand, T cell activation by the cC-Ova chimera, unable to access the secretory pathway, was fully abrogated by LC but was not inhibited by decRVKR-cmk. On the other hand, cytosolic miniprotein presentation, independent of proteasome and furin, was not affected by any of them. Also as expected, BFA blocked the ability to activate CTL of L/K^b cells infected with the three constructs.

Contribution of different processing pathways was also assessed in TAP⁺ and TAP⁻ DC. To this end, DC were infected in the presence of proteasome inhibitor leucyl-leucyl-norleucinal (LLnL). When TAP⁺ DC were infected with sC-OvaΔC in the presence of proteasome inhibitors, a 60% reduction in complex formation was measured (Fig. 4,E). Thus, proteasome-independent pathways provided roughly one-third of complexes from sC-OvaΔC in TAP⁺ DC, where precursor Ags have access to both cytosolic and secretory pathways. This relative contribution is similar to that measured in TAP⁺ L/K^b fibroblasts for the furin-mediated pathway. The secretory pathway did not provide any complex from the cytosolic construct, as expected, and also no measurable complexes from the sC-Ova construct in proteasome-inhibitor-treated TAP⁺ DC (data not shown), consistent with its at least 1000-fold less efficient processing in the secretory pathway. In summary, all observations on TAP-positive cells (Fig. 4, C–E), where there is no known limitation of available MHC class I molecules, led to a similar contribution to complex formation of the furin-mediated secretory pathway, which was estimated to be approximately one-third.

Next, the contribution of proteasome-independent pathways was also estimated by comparing the only two situations where only these pathways can provide peptides to the complexes, namely the sC-OvaΔC-infected TAP⁻ DC and the LLnL-treated sC-OvaΔC-infected TAP⁺ DC. When both pathways reached plateau expression of complexes, it was found that the efficiency of presentation via the secretory pathway in DC was ∼5-fold more efficient in the absence of TAP (Fig. 4 E).

Once demonstrated in vitro in TAP-deficient T2/K^b cells that the secretory construct sC-OvaΔC could be processed to generate enough K^b/SIINFEKL complexes and activate specific CTL (Figs. 2,A and F33, B and C) and that Ag presentation by the secretory pathway was efficient in TAP-deficient DC (Figs. 2,B and F44E), it was tempting to explore if this secretory pathway was competent in developing an effective CD8⁺ T lymphocyte response in vivo. To this end, TAP1-deficient mice constitute a valuable tool to explore this possibility because their cytosolic peptides cannot gain access to the ER to bind MHC class I molecules. TAP-competent and TAP1-deficient mice were i.p. immunized with several rVACV: Ova (full-length native OVA), cC-Ova, sC-Ova, and sC-OvaΔC. After 7 days, mice were sacrificed, and splenocytes were harvested and directly tested in ex vivo ICS assays to quantify the frequency of SIINFEKL-specific CD8⁺ T lymphocytes.

As shown in Fig. 5, left panel, all constructs elicited a specific CD8⁺ T lymphocyte response in TAP⁺ mice. On the other hand, as it is also shown in Fig. 5, right panel, all control cytosolic rVACV were unable to immunize TAP1-deficient mice. Not unexpectedly, the sC-Ova construct, which was at least 1000-fold less efficient in TAP-deficient T2/K^b in vitro than sC-OvaΔC, was not able to induce SIINFEKL-specific CD8⁺ T lymphocytes in vivo in TAP1-deficient mice. In contrast, as also control secretory minigene (data not shown), sC-OvaΔC did induce SIINFEKL-specific CD8⁺ T lymphocytes.

The accurate correspondence between the Ag processing results in vitro and the induction of CD8⁺ T lymphocytes in vivo in this particular setting illustrates the potential of the secretory pathway of Ag processing and presentation to induce an immune response.

The percentage of specific activated CD8⁺ T lymphocytes induced by sC-OvaΔC in the TAP-deficient setting was similar to that of the TAP-competent animals. This underscores the in vivo relevance of the secretory pathway when the cytosolic peptide supply is abrogated. Nonetheless, the efficiency of the CD8⁺ T lymphocyte response to control and eliminate epitope-bearing cells in both kinds of mice should also be evaluated. In this regard, it is important to note that, although the percentage of specific activated CD8⁺ T lymphocytes is similar, the absolute size of the CD8⁺ subset is ∼10- to 20-fold lower in TAP1-deficient than in TAP-competent infected mice. Therefore, the absolute number of Ag-specific CD8⁺ T lymphocytes is also 10- to 20-fold lower in TAP1-deficient mice. To compare the efficiency of the CD8⁺ T lymphocyte response in both types of mice, an in vivo killing assay was developed. Such assays best correlate with protection in mouse models of infection (36). They are more physiological than the widely used adoptive transfer of TAP⁺ TCR-transgenic OT-I T cells into TAP-deficient mice and, in our hands, are more sensitive than ex vivo ICS assays. Furthermore, they directly measure the functional in vivo activity of TAP-deficient CD8⁺ T lymphocytes, which has not been done before to our knowledge. TAP-competent and TAP1-deficient splenocytes were brightly and dimly stained with CFSE dye. Bright cells (CFSE^high) were pulsed with SIINFEKL peptide whereas dim ones (CFSE^low) were not. A 50% mix of syngeneic CFSE^high and CFSE^low cells was i.p. injected in TAP-competent and TAP1-deficient mice, which had been immunized previously with HBe, cC-Ova or sC-OvaΔC constructs. After two days the remaining cells were recovered from the peritoneum and analyzed by flow cytometry. Results of a representative experiment are represented by CFSE staining histograms in Fig. 6. TAP-competent mice (Fig. 6, upper panels) eliminated SIINFEKL-pulsed CFSE^high cells when they had been previously immunized with cC-Ova or sC-OvaΔC but not with control HBe rVACV. In contrast, TAP1-deficient mice (Fig. 6, lower panels) were unable to clear epitope-bearing cells when they had been immunized with cytosolic cC-Ova but eliminated them when they had been immunized with secretory sC-OvaΔC. Interestingly, TAP1-deficient mice required 2 days for eliminating infused epitope-bearing cells, whereas wild-type mice already cleared a substantial amount after 1 day (data not shown). Overall, in vivo killing results are concurring with those from ex vivo ICS assays from immunized mice. Collectively, they strongly support the idea that the single contribution of the secretory pathway of Ag processing suffices to elicit a SIINFEKL-specific TAP1-deficient CD8⁺ T lymphocyte response that is ∼20-fold lower in absolute numbers but efficient enough to selectively eliminate the majority of the challenging epitope-bearing cells.

A number of TAP-independent Ag-processing pathways, which can provide endogenous viral peptides for presentation by MHC class I molecules to CD8⁺ CTL, have been identified in the past years (5). A well-defined one involves the trans-Golgi network protease furin (8, 9). After these qualitative descriptions, there is a need for quantitative evaluation of the contribution of these supplementary pathways to Ag presentation and antiviral responses in vitro and, especially, in vivo.

Our report is the first functional and quantitative report on any of these pathways, showing that the furin-mediated secretory pathway provides approximately one-third of all surface peptide/MHC class I complexes in wild-type TAP-competent cells. Generation of the epitope carboxyl terminus is a dramatic rate-limiting step in this route, because bypassing it increases efficiency by at least 1000-fold. In TAP-deficient primary DC, the secretory pathways are even 5-fold more efficient than in the presence of TAP. The in vivo relevance of secretory routes is demonstrated in infected TAP1^−/− mice, where specific TAP1^−/− CD8⁺ T lymphocytes, although ∼20-fold less abundant, accomplish elimination of epitope-bearing cells in vivo.

The results show that sC-OvaΔC is processed to generate the OVA epitope in the furin-mediated secretory pathway in infected cells. Indeed, SIINFEKL is the third epitope assayed in this pathway, and all three tested so far have been presented. However, there is a need to further develop high-affinity TCR-like Abs that may allow generalization of the quantitative findings. Also, K^b is the third MHC class I allotype tested, because the furin pathway also provided ligands for L^d and D^d (8, 9). Thus, trimming enzymes in this pathway are versatile, because they can deal with all epitopes tested so far. Although the amount of furin expressed in cells is limited, the real rate-limiting step in the pathway is production of the carboxyl terminus by endo- or carboxypeptidases. The limitation can be easily overcome, and it should be considered in construct design and in possible vaccination strategies. Altogether, these data support the furin-mediated secretory pathway as a general mechanism in Ag processing, independently of the epitope and of the presenting MHC class I molecule.

Both furin (11, 12) and cathepsin S (14) have been involved in TAP-independent cross-priming of immunogens that originate from an exogenous source. However, as the cytosolic proteins Ova and cC-Ova did not induce CD8⁺ T lymphocytes in TAP1-deficient mice, the evidence suggests that the potential of constructs like sC-OvaΔC for in vivo immunization depends on direct priming by professional APC rather than on cross-priming. Considering that the furin-mediated pathway contributed to most endogenous Ag presentation in TAP-deficient cells in vitro, it is expected that induction of specific CD8⁺ T lymphocytes in vivo in TAP1-deficient mice by sC-OvaΔC is also mainly mediated by furin. Unfortunately, available furin-deficient mice to further test it are unsuitable (37, 38).

The furin-dependent secretory pathway can generate around one-third of K^b/SIINFEKL complexes in TAP-competent cells, including cell lines and primary DC, where the classical proteasome-TAP-mediated pathway is also available. This quantity suffices to induce CTL activation in vitro. The fraction of complexes contributed by the secretory pathway may differ for different antigenic proteins or epitopes. Actually, for the original CMV 9pp89 chimeras (8, 9), the secretory pathway may be more efficient, since we could not even detect the contribution of the proteasomal pathway, because their presentation to CTL from TAP⁺ cells was unaffected by proteasome inhibitors, unlike that of sC-OvaΔC. The data suggest that this kind of mechanism may contribute to the in vivo CD8⁺ T lymphocyte response to some natural epitopes, perhaps those contained in viral proteins that, like HBe, undergo proteolytic maturation in the secretory pathway.

On the other hand, when the classical proteasome-TAP-mediated pathway contribution was absent, in TAP-deficient DC, the efficiency of presentation via the secretory pathway was ∼5-fold higher. It is likely that TAP⁺ DC have less peptide-receptive or empty MHC class I molecules, which can be filled to greater numbers with SIINFEKL produced from sC-OvaΔC in the secretory pathway in TAP-deficient DC. Binding of this peptide would also be favored by a likely reduced competition with other high-affinity ligands. In addition, these results indicate that proteasome-derived peptides are not needed for MHC class I molecules to bind the furin-produced peptides.

In the cytosol, proteasomes generate most correct carboxyl termini of epitopes (39). In contrast, peptidases producing the carboxyl terminus in the secretory pathway have only been indirectly implicated by our results, where a minimal size CMV 9pp89 peptide was naturally produced from a construct similar to sC-Ova (8). This suggests that there is little carboxypeptidase activity available for Ag trimming in the secretory pathway (40). From the rat cim model, it was known that the carboxypeptidase activity that removes hydrophobic residues from the carboxyl termini of antigenic peptides is essentially lacking in the ER (41). Our results now show that carboxy- or endopeptidases that remove some polar and charged residues are also extremely rate limiting for Ag presentation in the secretory pathway. In addition, we quantify this limitation for the first time, as we show that the carboxyl-terminal extension of the minimal epitope in the sC-Ova construct strongly limits Ag processing in the secretory pathway by an estimated factor of at least 1000-fold.

On the other hand, aminopeptidases are extensively documented in Ag processing in the secretory pathway (42). In line with these reports, we now find that additional elimination of the requirement for generating the correct amino terminus results in only additional 20-fold higher efficiency of Ag presentation.

A good correlation was found between the Ag processing results in vitro and the induction of CD8⁺ T lymphocytes in vivo. The in vivo functional results strongly support the idea that the single contribution of the secretory pathway of Ag processing suffices to elicit a specific TAP1-deficient CD8⁺ T lymphocyte response that is ∼20-fold lower in absolute numbers but efficient enough to selectively eliminate epitope-bearing cells.

Altogether, our data support the capacity of the secretory pathway of Ag processing to elicit a complete functional CD8⁺ T lymphocyte response in vivo. In addition to furin, cathepsins, or signal peptidase, other endoproteases contribute to this secretory MHC class I ligand pool in vitro (5, 10), depending on the natural epitope-bearing protein. It is likely that, besides furin, these undefined secretory endoproteases also contribute to eliciting a functional CD8⁺ T cell response in vivo, which warrants further investigation. Overall, Ag processing in the secretory pathway would be a reasonable mechanism to explain the surprisingly low rate of pathology associated with viral infections in TAP-deficient mice and humans (15). The suitability of this supplemental pathway to explain the clinical presentation is supported by the quantification of its contribution to Ag presentation both in vitro as well as functionally in vivo. Our data support the hypothesis that noncytosolic processing mechanisms may compensate in vivo for the lack of proteasome contribution to Ag processing in these persons genetically deficient in TAP and thus contribute to pathogen control.

Moreover, considering that in our TAP-competent setting at least one out of every three SIINFEKL epitope/MHC class I complexes depend on furin processing, when proper carrier proteins are used, a general contribution of the secretory pathway could be expected also in TAP-competent normal organisms. This contribution might be not only a quantitative but perhaps also a more qualitative one, used for those epitopes in secretory proteins that may not be so easily generated by the proteasome, as a mechanism that extends the repertoire of functional CD8⁺ T lymphocyte responses. In addition, these pathways may provide peptides derived from viruses that try to evade the CD8⁺ immune response by blocking TAP (43). Although several endoproteases might be contributing to this general Ag-processing secretory pathway, furin remains as an attractive executor: adopted by viruses during evolution for maturation of viral proteins, it might have been reused in a host counterattack to neutralize the pathogen. In this respect, a next goal would be finding natural epitopes in viral proteins generated by furin.

We thank Dr. H. G. Ljunggren for cell lines and Drs. J. W. Yewdell and J. Bennink for rVACV and reagents. Human rIL-2 was provided by the National Cancer Institute Preclinical Repository. Help with techniques by Natalia Romero and María Lorenzo, technical assistance by C. Mir, Y. Laó, and S. Sánchez as well as peptide synthesis from our central facility are gratefully acknowledged.

The authors have no financial conflict of interest.