Mixed Proteasomes Function To Increase Viral Peptide Diversity and Broaden Antiviral CD8⁺ T Cell Responses

Abstract

The three proteasome subunits with proteolytic activity are encoded by standard or immunoproteasome genes. Many proteasomes expressed by normal cells and cells exposed to cytokines are “mixed”, that is, contain both standard and immunoproteasome subunits. Using a panel of 38 defined influenza A virus–derived epitopes recognized by C57BL/6 mouse CD8⁺ T cells, we used mice with targeted disruption of β1i, β2i, or β5i/β2i genes to examine the contribution of mixed proteasomes to the immunodominance hierarchy of antiviral CD8⁺ T cells. We show that each immunoproteasome subunit has large effects on the primary and recall immunodominance hierarchies due to modulating both the available T cell repertoire and generation of individual epitopes as determined both biochemically and kinetically in Ag presentation assays. These findings indicate that mixed proteasomes function to enhance the diversity of peptides and support a broad CD8⁺ T cell response.

Immunodominance (ID) is a central feature of T cell responses. T cell responses to multiple pathogen-derived peptides and other complex Ags form a hierarchy based on the number of responding cells. This hierarchy tends to be highly reproducible in individuals expressing the same restricting MHC allomorphs. The CD8⁺ T cell (T_CD8+) response to influenza A virus (IAV) in C57BL/6 mice is the most extensively studied mouse ID model with nearly 40 immunodominant (IDD) and subdominant determinants (SDD) reported derived from 10 of the 11 known open reading frames translated in IAV-infected cells (1-3) (Immune Epitope Database, http://www.iedb.org/).

Given that an IAV-encoded peptide binds class I with sufficient affinity, the most important factors in governing its position in the ID hierarchy are the numbers of naive T_CD8+ and the amount of peptide generated by APCs. Both of these factors are influenced by proteasome composition (4, 5). Proteasomes have three active subunits, β1, β2, and β5, that can be replaced by the immune versions, β1i, β2i, and β5i, to create immunoproteasomes. Immunoproteasomes are constitutively expressed by dendritic cells (DCs) and many other bone marrow–derived cells, and are induced in other cells by exposure to IFNs and other cytokines. Although immunoproteasomes are known to influence the generation of individual peptides and alter the T_CD8+ repertoire to certain viral peptides, the systematic impact of the immunoproteasome on ID hierarchies remains an important question.

Expanding on prior work with mice lacking one or two immunoproteasome subunits, Kincaid et al. (6) studied T_CD8+ responses to nine defined foreign determinants after knocking out all three immunoproteasome subunits. T_CD8+ responses are more broadly impaired in the absence of all immunoproteasome subunits than in individual subunit-deficient strains, and the self immunopeptidome is clearly greatly altered, as determined by mass spectrometry. It is also important, however, to understand the effects of missing one or two subunits, because “mixed” proteasomes with both standard and immunoproteasome subunits are commonly expressed at high levels constitutively (up to 50% of proteasomes) in human monocytes, immature and mature DCs, and liver, intestine, and kidney cells, and also following cytokine induction of immunoproteasome subunits in other cells (7).

Multiple studies have reported that recall responses to IAV amplify and even reshuffle the primary ID hierarchy (8-10). During primary T_CD8+ responses, the small numbers of naive T_CD8+ appear to have ample access to infected or cross-priming DCs. Upon recall, however, memory T_CD8+ must compete for APCs (8, 11-13), and they are at a disadvantage when their cognate peptides are generated with slower kinetics than other peptides (13). Given that immunoproteasome subunits can greatly influence epitope presentation, it is possible that they play a major role in shaping secondary ID hierarchies (14, 15).

To explore the role of mixed proteasomes on antiviral T_CD8+ responses, we examine the effect of knocking out β1i, β2i, and β5i/β2i on the generation of primary and recall immune response to 38 IAV peptides.

Materials and Methods

Mice

Wild-type (wt) C57BL/6 (B6) mice were purchased from Walter Eliza Hall Institute of Medical Research (Kew, Melbourne, VIC, Australia). LMP2^−/− mice were a gift from Dr. L.Van Kaer (Department of Microbiology and Immunology, Vanderbilt University, Nashville, TN). LMP7/MECL-1^−/− and MECL-1^−/− mice were a gift from Dr. John Monaco (Department of Molecular Genetics, Biochemistry, and Microbiology, University of Cincinnati, Cincinnati, OH). Mice were housed in specific pathogen-free isolators. Experiments were performed with animals aged at 6–12 wk and were conducted under the auspices of the Austin Health Animal Ethics Committee and conformed to the National Health and Medical Research Council Australian code of practice for the care and use of animals for scientific purposes.

Peptides and Abs

IAV peptides were synthesized by Mimotopes (Clayton, VIC, Australia) with purity >95%. A complete list of peptide sequences, elution times, and MHC class I restriction can be found in Table I. PE-labeled anti–IFN-γ and APC-labeled anti-CD8α were purchased from eBioscience (San Diego, CA). For flow cytometry, Abs were used at 1:300 dilution in PBS supplemented with 10% FCS. IL-2 was purchased from Roche (Pleasanton, CA) and Proleukin (San Diego, CA).

Table I.

Reported IAV H-2^b –restricted epitopes

Peptide	Sequence	Restriction	Elution Time (min)
HA243-250	MNYYWTLL	Kb	28.205
HA290–297	ECNTKCQT	Likely Kb	24.308
HA304–311	SSLPYQNI	Db Kb	21.623
HA332–340	TGLRNTPSI	Db	19.597
HA402–409	MNIQFTAV	Kb	23.873
HA468–476	SQLKNNAKEI	Db	17.824
HA475–482	KEIGNGCFEF	Db	27.652
Mll28–135	MGLIYNRM	Kb	23.540
NA46–54	TGICNQNII	Db	23.042
NA54–62	ITYKNSTWV	Db	21.078
NA181–189	SGPDNGAVAV	Db	18.559
NA335–343	YRYGNGVWI	Db	25.140
NA425–432	SSISFCGV	Kb	27.350
NA430–438	FCGVNSDTV	Db	20.160
NP17–25	GERQNATEI	Db	16.601
NP35–42	IGRFYIQM	Kb	25.570
NP55–63	RLIQNSLTI	Db	22.889
NP97–105	YRRVNGKWM	Db	19.709
NP217–225	IAYERMCNI	Kb	22.384
NP366–374	ASNENMETM	Db	17.998
NS1133–140	FSVIFDRL	Kb	27.377
NS2114–121	RTFSFQLI	Kb	27.377
PA224–233	SSLENFRAYV	Db	24.706
PA238–245	NGYIEGKL	Kb	21.279
PA300–307	GIPLYDAI	Kb	26.044
PA463–470	VYINTALL	Kb	24.576
PA648–655	SLYASPQL	Kb	21.502
PB1141–149	TALANTIEV	Db	22.788
PB1214–221	RSYLIRAL	Kb	23.737
PB1465–472	RFYRTCKL	Kb	20.183
PB1653–660	KNMEYDAV	Kb	18.350
PB1703–711	SSYRRPVGI	Kb	19.914
PB1F262–70	LSLRNPILV	Db	25.933
PB2196–206	CKISPLMVAYM	Kb	26.260
PB2198–206	ISPLMVAYM	Likely Db	26.420
PB2227–234	VYIEVLHL	Kb	26.173
PB2358–365	GYEEFTMV	Kb	24.443
PB2689–696	VLRGFLIL	Kb	29.134

Peptide sequences, reported MHC class I restriction, and RP-HPLC elution time on an Agilent 1100 instrument are shown. To determine the exact time that each naturally processed and presented IAV peptide elutes from the HPLC, the elution times of the 38 IAV-derived synthetic peptides were first determined (the elution times are indicated in Fig. 3 and Supplemental Fig. 2 as vertical lines). The data indicate distinct elution time points over a range of ~13 min under the selected running program and solvent gradient. Several peptides were observed to elute at similar time points, such as HA_332–340 and NP_97–105.

Viruses and mouse immunization

Mice were infected i.p. with 10⁷ PFU IAVs, including Puerto Rico/8/34 H1N1 (PR8) and its reassortant strain X-31 (H3N2).

Cell line culture

RMA-s and EL4 cells were cultured in RF-10: RPMI 1640 containing 10% FCS, 50 μM 2-ME, 2 mM L-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin (all from Invitrogen, Carlsbad, CA). Cells were passaged when reaching 80% confluent and used during the exponential growth phase.

Bone marrow–derived DCs, T_CD8+ cultures, and brefeldin A kinetics

The detailed methods were previously described (16, 17). Briefly, bone marrow cells were flushed from femurs and RBCs were lysed and cultured in RF-10 in the presence of GM-CSF for 8 d. For generating polyspecificity T_CD8+ cultures, bone marrow–derived DCs (BMDCs) were infected with PR8 at a multiplicity of infection (MOI) of 10 in FCS-free, acidified RPMI 1640 for 1 h at 37°C. RF-10 (10 ml) was then added and further incubated for 5 h. The infected cells were then washed, irradiated (10,000 rad), and used as APCs to stimulate PR8-specific memory splenic cells. For generating peptide-specific T_CD8+ cultures, 1 million memory splenic cells were pulsed with 1 nM peptide of interest for 60 min and then cocultured with 9 million unpulsed memory splenic cells in RF-10 containing 10 U IL-2/ml (T cell media). Beading of CD4⁺ and B220⁺ cells was performed if required from day 5 onward after removal of dead cells by Ficoll-Paque gradient. For BFA kinetics assay, BMDCs or EL-4 cells were infected with 10 MOI of PR8 for 1 h at 37°C, washed, and then added to monospecificity T_CD8+ cultures, with BFA added at various time points. Following 4 h exposure to BFA, cells were transferred onto ice and stained by intracellular cytokine staining (ICS) for IFN-γ (17).

B220/CD4 beading

CD4- and B220-expressing cells were removed from T_CD8+ cultures via magnetic separation. Briefly, Dynal beads (Invitrogen) coated with purified rat anti-mouse CD4 and B220 Abs (BD Biosciences, Franklin Lakes, NJ) were incubated with cultured cells for 10 min at room temperature. Particle-bound cells were depleted via magnetic separation, with cells left in supernatant resuspended in T cell media (18).

In vitro activation of ex vivo Ag-specific T_CD8+ and enumeration by ICS

Peritoneal exudates and spleens were collected in RF-10. The Ag-specific T_CD8+ were enumerated using ICS for the production of IFN-γ after being stimulated with 1 μM antigenic peptide unless otherwise stated (19). Briefly, samples were stained with anti-CD8 for 20 min, washed and fixed with 1% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA) in PBS for 30 min at room temperature, and then stained with anti-IFN-γ Ab in the presence of 0.4% saponin. Samples were then washed, acquired on a FACSCanto II (BD Biosciences), and analyzed by FlowJo software (Tree Star, Ashland, OR).

Peptide extraction, reverse-phase HPLC fractionation, and assessment by T_CD8+

The method was reported previously (19). Briefly, 8-d-cultured BMDCs were infected at 10 MOI with PR8 for 5 h and pellets were snap-frozen on dry ice. Pellets were resuspended in trifluoroacetic acid/H₂O, disrupted by douncing with a glass homogenizer, further sonicated for 30 s using a microtip, and then gently mixed for 30 min at 4°C on a rotating wheel before being pelleted at 2000 rpm. The peptide-containing supernatant was further cleared of large molecules with ultracentrifugation at 27,000 rpm for 45min. Supernatants were then passed through a 3 kDa cut-off filter and dehydrated to, <400 μl in a SpeedVac. Extracted peptides were then fractionated with reverse-phase HPLC (RP-HPLC; Agilent 1100) and fractions between 10.0 and 38.8 min were collected at 6-s intervals. Elution times of synthetic peptides were determined by passing 1 μl 1 mM synthetic peptide through a RP-HPLC machine under the same running condition. To detect naturally presented peptides in the HPLC fractions, RMA-s cells were low temperature induced overnight at 26°C and pulsed (10⁵) with either synthetic peptide at log fold dilutions or fractions for 1 h at 26°C in RPMI 1640. T_CD8+ (10⁵) were added to the pulsed RMA-s cells in RF-10 containing 10 μg/ml BFA for a further 5 h and Ag-specific T_CD8+ were enumerated by ICS.

Statistical analysis

All error bars are SEM using a sample size of more than three per group.

Results

Ex vivo IAV immune response in immunoproteasome subunit-deficient mice

To better understand the contribution of immunoproteasome subunits to antiviral immunity, we studied B6 wt and knockout strains lacking one or two immunoproteasome subunits (β1i^−/−, β5i/β2i^−/−, and β2i^−/−). Immunoproteasome subunit–deficient mice possess professional APCs expressing mixed proteasomes as shown by Western blots (data not shown), presumably very similar if not nearly identical to intermediate proteasomes (7, 20, 21). At day 7, the peak of the T_CD8+ response, we analyzed local and systemic T_CD8+ responses to 38 characterized H-2^b–restricted peptides following i.p. PR8 infection.

As reported, the local peritoneal response in wt mice focuses on two codominant epitopes, nucleoprotein (NP)_366–374 and acidic polymerase (PA)_224–233, atop reasonably robust subdominant responses to basic polymerase 1 (PB1)F2_62–70, Pβ1_703–711, and a newly identified epitope basic polymerase 2 (PB2)_196–206 (Fig. 1A and D. Zanker, J. Waithman, R. Lata, R. Murphy, and W. Chen, manuscript in preparation). Responses to other epitopes (Fig. 1A) are slightly but reproducibly above detection limits. β1i^−/− mice demonstrate diminished responses, with clear alterations in the ID hierarchy. Codominant responses to PB2_196-206 and nonstructural protein 2 (NS2)_114–121 led subdominant responses to Pβ1F2_62–70, NP_217–225, hemagglutinin (HA)_304–311, HA_402–409, along with the clearly demoted NP_366–374 and PA_224–233 epitopes (5, 16). β5i/β2i^−/− mice exhibit NP_366–374 dominance, with PA_224–233 completely absent. β2i^−/− mice exhibit less difference than other knockout (KO) mice, with a slightly reduced PA_224–233 response (Fig. 1A).

FIGURE 1.

Loss of immunoproteasome subunits results in change of IAV ID hierarchy. β1i^−/−, β5i/β2i^−/−, β2i^−/−, and wt mice were infected with IAV and assessed for (A) primary or (B) secondary immune responses 7 d after infection. Intraperitoneal exudates (local site of infection) were taken and T_CD8+ responses measured using synthetic peptides to known IAV epitopes in an ICS assay for IFN-γ. Six mice per strain were collected individually. Data are pooled from two independent experiments.

Ranking responses to each epitope as a percentage of the total response to all epitopes tested (Supplemental Table I) reveals clear shuffling related to immunoproteasome composition. Highly similar patterns are apparent in the spleen despite a lower overall response (Supplemental Fig. 1A).

Ex vivo IAV recall response in immunoproteasome subunit–deficient mice

The ID hierarchy in recall B6 anti-IAV responses can vary significantly from primary responses (5, 22). To examine the influence of immunoproteasome subunits on recall responses, we challenged mice (>30 d after PR8 infection) with X31, a reassortant IAV with PR8 internal genes and genes encoding H3N2 glycoproteins, to avoid neutralization by anti-PR8 glycoprotein Abs (13). As above, T_CD8+ in peritoneal exudates and spleen were assessed at the peak of the secondary response (day 7).

NP_366–374 dominates recall responses in wt, β5i/β2i^−/−, and β2i^−/− strains. In β1i^−/− mice, decreased responses to NP_366-374 and PA_224–233 occur due to decreased naive T_CD8+ precursor frequencies and Ag presentation, respectively (5, 16). Concomitantly, SD responses to Pβ1F2_62–70 and PB2_196–206 rise to codominant status, although at lower levels than NP_366–374-specific response in other strains. Barely detectable responses in wt mice to HA_304–311, HA_332–340, HA_402–409, and NP_217–225 ascend to substantial SD status in β1i^−/− recall responses (Fig. 1B). Conversely, among β5i/β2i^−/− peritoneal responses, NP_366–374-specific T_CD8+ account for 75% of total anti-IAV T_CD8+ (Fig. 1B; note the scale break), that is, >200% of the wt response. Curiously, this trend was reversed in splenic wt and β5i/β2i^−/− T_CD8+, clearly demonstrating the potential for organ and condition-specific differences in ID (Supplemental Fig. 1B). Accompanying enhanced NP_366–374 responses in β5i/β2i^−/− mice, responses to SDDs (HA_304–311, NP_217–225, NS2_114–121, Pβ1_703–711, and Pβ1F2_62–70) were reduced in absolute terms relative to responses in wt mice. This is likely due to immunodomination, the active suppression of T_CD8+ specific to SDDs by those specific to IDDs (23, 24). Immunodomination was much less marked in other circumstances (Supplemental Fig. 1B).

Fig. 1B further shows the enhancement of various subdominant responses in a secondary infection in the β2i^−/− strain, for example, responses to matrix protein 1 (M1)_128–135, neuraminidase (NA)_430–438, and PB2_227–237, which were diminished to absent in other strains. These results indicate that enhanced T_CD8+ responses are likely due to enhanced local Ag processing and presentation in the peritoneum for those epitopes in the absence of β2i.

Interestingly, responses to some peptides are completely absent in certain strains, such as the missing PA_224–233 and PB2_196–206 response in β5i/β2i^−/− mice. The response to PA_224–233 was not efficiently recalled in wt mice, as previously reported (13), but even less efficiently recalled in all subunit-deficient mice tested (Fig. 1B, Supplemental Fig. 1B). Similar to the primary splenic response shown in Supplemental Fig. 1A, the secondary splenic response (Supplemental Fig. 1B) is largely paralleled by the peritoneal response, in this case with less dramatic differences.

To illustrate the ID hierarchy change, the ID hierarchies of the secondary responses to IAV in the four mouse strains are displayed in Supplemental Table I and ranked according to the magnitude, as well as the relative change, in ranking between primary and secondary immune responses in the various strains.

These data clearly show that the absence of immunoproteasome subunits diversely affects the specificity and magnitude of antiviral T_CD8+, often in disparate ways in local versus systemic T_CD8+ populations.

Visualizing IAV-specific T_CD8+ repertoire and immunoproteasome-processed IAV epitopes

The capacity of proteasomes to generate immunogenic peptides is a major factor in ID (19). Surprisingly little is known about the influence of immunoproteasome subunits on the generation of the viral peptide repertoire in APCs or in vivo. To measure generation of a wide variety of IAV peptides we generated polyspecific T cell cultures from splenic T cells derived from wt and KO mice, taking care to stimulate T_CD8+ with saturating numbers of autologous IAV-infected BMDCs to minimize immunodomination. This generated highly polyspecific cultures that matched in vivo ID hierarchies in an immunoproteasome subunit KO–specific manner, as determined with peptides corresponding to 21 IAV epitopes to quantitate T_CD8+ specificities (Fig. 2). Fig. 2A shows representative clear patterns of ICS for SDD responses in the wt culture. The relatively lower total Ag-specific T cell percentage in the β1i^−/− culture is partially due to the hidden PB2_196–206 response (D. Zanker et al., manuscript in preparation).

FIGURE 2.

Diverse IAV responses from in vitro–expanded polyspecific IAV T_CD8+ cultures derived from wt and immunoproteasome-deficient mice. (A) Representative FACS plots of subdominant T_CD8+ responses from in vitro–propagated T_CD8+ cultures used in Fig. 2, showing (Ai) Nil peptide, (Aii) M1_128-135 peptide (baseline level subdominant response), and (Aiii)) NS2_114-121 (subdominant response). (B) Wild-type, (C) β1i^−/−, (D) β5i/β2i^−/−, and (E) β2i^−/− mice were immunized with 10⁷ PFU PR8 i.p. Thirty days later, spleens were isolated and T_CD8+ restimulated using host-strain BMDCs infected with PR8 at 10 MOI. Following 10 d culture, responses were measured using synthetic peptides to known IAV epitopes in an ICS assay measuring the production of IFN-γ. Asterisks denote when definitive T_CD8+ responses were observed for the defined epitope.

We used these polyspecific T_CD8+ cultures from wt and KO mice to assess antigenic peptides present in fractions eluted from IAV-infected, immunoproteasome-competent, or subunit-deficient BMDCs and thus reveal immunoproteasome subunit–specific changes in generation of antigenic peptides and/or individual T_CD8+ specificities. Plotting the data from this analysis as an “activity landscape” of percentage of IFN-γ–producing T_CD8+ versus fraction number reveals clear differences associated with each KO tested (Fig. 3). Note that the height of the peaks represents mainly the abundance of T_CD8+ of that specificity in the polyspecific T_CD8+ culture, although it may also serve as a direct indication of peptide abundance when the peaks from different infected BMDCs vary clearly, such as those in Fig. 3A at 18 min. However, the peptide abundance is further addressed in the HPLC fraction titration experiments readout by monospecific T_CD8+ cultures in Fig. 4. Also, note that there is little nonspecific T_CD8+ activation for data shown in Fig. 3 judging by FACS plots of the control fractions (not shown).

FIGURE 3.

IAV-specific T_CD8+ demonstrate altered IAV epitope processing by BMDCs expressing mixed proteasome. Polyspecific IAV-specific T_CD8+ cultures raised from (A) wt, (B) β1i^−/−, (C) β5i/β2i^−/−, and (D) β2i^−/− mice were restimulated using eluted HPLC fractions derived from IAV-infected BMDCs from wt (black dash), β1i^−/− (pink), β5i/β2i^−/− (green), or β2i^−/− (dark blue) mice and assessed for the production of IFN-γ in an ICS assay. Elution time points of known synthetic IAV peptides are shown by a dotted line and the peptides are indicated below. Arrows indicate areas of interest for further investigation using monospecific IAV T_CD8+ cultures.

FIGURE 4.

Individual immunoproteasome subunits are critical for generation of IAV T_CD8+ epitopes. (A and B) Monospecific T_CD8+ cultures raised against known IAV-derived epitopes. Ten-day T_CD8+ cultures were measured for purity using a titration of synthetic peptide in an ICS assay for IFN-γ. (C) NS2_114–121, (E) PA_224–233, (G) PB1_703–711, (I) Pβ1F2_62–70, (K) M1_128–135, and (M) NP_366–374 cultures were assessed for IFN-γ production following incubation with RP-HPLC fractions of IAV-infected BMDCs from wt (◆), β1i^−/− (■), β5i/β2i^−/− (▲), or β2i^−/− (○) mice. Peak fractions for (D) NS2_114–121, (F) PA_224–233, (H) Pβ1_703–711, (J) Pβ1F2_62–70, (L) M1_128–135, and (N) NP_366–374 were further diluted to estimate relative peptide abundance.

Only wt and β2i^−/− T_CD8+ cultures contained T_CD8+ activated by peptides eluting in the 24.8-min fraction, corresponding to the PA_224–233 elution time (Fig. 3A, 3D; also refer to Table I for synthetic peptide elution times). Correspondingly, β5i/β2i^−/− and β1i^−/− polyspecific T_CD8+ cultures lack detectable PA_224–233-specific T_CD8+. Importantly, this is not due to the inability of β5i/β2i^−/− and β1i^−/− cells to generate PA_224–233, which is actually enhanced in all subunit-deficient relative to wt cells (Fig. 3A at 24.8 min), but instead points to immunoproteasome subunits shaping the T_CD8+ repertoire.

More subtle, but still clear, β1i^−/− T_CD8+ exclusively detect a peak at 20.3 min (Pβ1_465–472) (Fig. 3B, Supplemental Fig. 2) exclusively present in the β1i^−/− fractions. Less absolutely, wt T_CD8+ selectively detect activity at 19.9 min (PB1_703–711) that is diminished in all fractions from all of the KO DCs. In these cases, the immunoproteasome subunits are required for peptide generation, regardless of their possible effects on the T_CD8+ repertoire.

All major responses observed in the four T_CD8+ activity landscapes can be attributed to previously described T_CD8+ specificities with a lone exception: a major activity is detected at 22.0 min, where no known immunogenic peptide elutes (Fig. 3, Supplemental Fig. 2). This peptide is generated by all DCs, although at lower levels by β2i^−/− DCs (Fig. 3B at 22.0 min, blue line); remarkably, all mice mount a similarly vigorous response (note that y-axis scales vary). Owing to the lower overall responses in β1i^−/− mice, this peptide scores as an IDD.

Immunoproteasome subunits control peptide generation: biochemistry

The magnitude of each peptide-specific T_CD8+ component in polyspecific T_CD8+ cultures from wt versus KO mice varies dramatically. To firmly establish the basis for these observations, we directly compared the amounts of defined peptide present in each fraction by the ability of dilutions to activate T_CD8+ raised to individual peptides by in vitro culture.

We focused on the six most interesting peptides identified by the T_CD8+ activity landscape analysis corresponding to elution times 18.0 (NP_366–374), 19.9 (PB1_703–711), 23.6 (M1_128–135), 24.8 (PA_224–233), 26.0 (PB1F2_62–70) and 27.4 min (NS2_114–121). All T_CD8+ populations tested exhibited high sensitivities, with half-maximal responses between 10⁻¹¹ and 10⁻¹² M (Fig. 4A, 4B). None of the monospecific cultures exhibited reactivity with non-cognate major immunodominant epitopes (including NP_366–374 and PA_224–233; data not shown), consistent with monospecificity for their nominal peptides. We first used the monospecific cultures to show that naturally presented peptides match the elution times of the synthetic peptide (Fig. 4), providing the initial evidence that NS2_114–121, PA_224–233, PB1_703–711, and M1_128–135 are naturally generated epitopes (PB1F2_62–70 and NP_366–374 were demonstrated in Refs. 1 and 25, respectively, to be naturally presented).

Peptide titrations of the HPLC fractions (Fig. 4, right panels) reveal clear differences bestowed by immunoproteasome subunits on Ag processing efficiency. For example, PB1_703–711 (Fig. 4H) is generated 25-fold less efficiently in BMDCs lacking β1i or β2i, and nearly 125-fold less efficiently in BMDCs deficient for β5i/β2i. Loss of immunoproteasome subunits also impairs presentation of NP_366–374 and Pβ1F2_62–70. In contrast, the absence of immunoproteasome subunits uniformly enhances M1_128–135 activity (Fig. 4L).

Immunoproteasome subunits control peptide generation: presentation kinetics

Although peptide recovery from cells typically requires the peptide to be protected by MHC class I molecules (26), there are exceptions to this rule (27). To more firmly establish the role of immunoproteasome subunits in directly altering the presented IAV peptide repertoire, we examined the kinetics of peptide presentation. We infected DCs from wt and KO mice with IAV and examined their capacity to activate nine single specificity T_CD8+ cultures (characterized in Supplemental Fig. 3) at various times after infection by adding BFA to block delivery of peptide class I complexes to the cell surface (Fig. 5).

FIGURE 5.

The Ag presentation of specific epitopes is altered by APCs with different protein degradation machinery. BMDCs from β1i^−/− (■), β5i/β2i^−/− (▲), β2i^−/− (○), and wt (◆) mice were infected with IAV at an MOI of 10. Following 60 min infection (indicated as 1 h in the figures), cells were washed and incubated with monospecific T_CD8+ cultures specific to defined IAV epitopes, (A) M1_128–135, (B) PB1F2_62–70, (C) NP_366-374, (D) NS2_114–121, (E) PB2_196–206, (F) PB1_703–711, (G) NA_181–189, (H) PB2_227–237, and (I) PA_224–233. BFA was added for a total of 4 h at the indicated time points and recognition of naturally presented peptide was measured using ICS assay for IFN-γ.

In all APCs tested, the NP_366–374 epitope is presented most efficiently and is detectably presented at the cell surface even 1 h after infection. The M1_128–135 peptide is also presented rapidly, with a slight enhancement in BMDCs lacking the β1i subunit (Fig. 5A, Supplemental Fig. 3D). The PB1F2_62–70 peptide shows enhanced presentation in BMDCs lacking the β5i or β1i subunit compared with β2i^−/− and wt BMDCs (Fig. 5B). The NS2_114–121, PB2_196–206, and PB1_703–711 epitopes are presented with medium kinetics (Fig. 5D-F). In the absence of both the β5i and β2i subunits, PB2_196–206 presentation is nearly absent (Fig. 5E). Conversely, PB2_196–206 presentation is enhanced in the absence of β1i. The NA_181–189, PB2_227–237, and PA_224–233 epitopes are presented with overall slower kinetics (Fig. 5G-I). The NA_181–189 and PB2_227–237 presentation kinetics are slightly enhanced in the absence of the β5i and β2i subunits.

Finally, and most remarkably, PA_224–233 presentation is the slowest and requires a complete immunoproteasome for most efficient presentation (Fig. 5I). The slow presentation is in stark contrast to the clear abundance of the peptide in cell lysates, and, as discussed below, extends the findings of Lev et al. (27) made with minigene products to physiological Ag processing.

Taken together, these findings demonstrate that knocking out individual proteasome subunits greatly influences ID to IAV by modulating the repertoire of responding T_CD8+ and by modulating the specificity of peptide generation. These findings strongly imply that mixed proteasomes function to diversify the antiviral repertoire.

Discussion

In this study, we provide the most comprehensive study to date of ID in an animal model system: we characterize the primary and recall B6 T_CD8+ responses to 38 defined IAV determinants (Immune Epitope Database, http://www.iedb.org/). Most of these determinants were predicted through computer algorithms based on class I binding and proteasome cleavage motifs and were assessed by class I binding and T cell activation (2, 28, 29). For four of these peptides we provide the initial evidence that they are naturally processed by IAV-infected cells, as determined by the coelution of synthetic peptides. For five of these peptides, we further show that T cells raised to the peptides recognize infected cells. This falls just short of the ultimate confirmation of epitope immunogenicity, which entails correlating lost peptide reactivity of antiviral T cells following infection with virus mutated to eliminate peptide binding to the restricting class I molecules (30).

To our knowledge, this study provides the first comprehensive analysis of the ability of mixed proteasomes to generate a large panel of viral peptides. We use polyspecific anti-IAV T_CD8+ cultures from immunoproteasome subunit KO mice to assess the peptides generated by IAV-infected BMDCs from these strains. This revealed subunit-specific shuffling of ID hierarchies including the enormous secondary response to PB2_196–206 in the β1i^−/− strain, as well as enhanced responses to PB2_196–206, M1_128–135, and Pβ1_465–472 in mice lacking immunoproteasome subunits. Some of the effects can be attributed to differences in repertoire, whereas others are clearly related to differences in peptide generation as shown both biochemically by peptide isolation and functionally by measuring presentation kinetics of infected cells. These findings point to a major contribution of mixed proteasomes to ID, and they raise the important issue of reconciling our findings with recently described triple immunoproteasome KO mice, which are incapable of generating mixed proteasomes (6).

The difference in activated naive and memory T_CD8+ effector function likely contributes to the changes in primary versus recall ID hierarchies. Naive T_CD8+ take days to kill following activation whereas memory T_CD8+ kill within hours (31) and could kill APCs quickly (11, 32). Rapid APC depletion could increase T_CD8+ competition for APCs, favoring proliferation of T_CD8+-targeting epitopes with faster presentation kinetics. Competition in recall immune responses has been demonstrated previously, although only with comparison of a limited number of epitopes (10). Our results indicate that ID hierarchy to nine epitopes in a recall response largely correlates to the particular presentation kinetics, arguing strongly for this being an important factor in establishing recall ID hierarchies.

The strength of our biochemical approach is illustrated by several findings. First, we detect a major response in all strains examined to an undefined IAV peptide eluting at 22.0 min in HPLC analysis. It should be possible in future studies to identify the peptide either directly by mass spectrometry or by traditional genetic mapping, starting with individual IAV gene products and truncating genes until the peptide can be pinpointed, aided by algorithms.

Second, we show a remarkable disconnect between the amounts of peptides recovered from infected cells and their kinetics of presentation. The influence of immunoproteasome subunits on peptide presentation and recovery is summarized in Table II, in which APC data from KO mice are shown as the percentage of data from wt APCs. Calculating the ratio of peptide recovery percentage to BFA kinetics percentage revealed a number of striking discrepancies. Setting a 10-fold difference as clearly significant, the HPLC assay underestimates the kinetics of presentation of PB1_703–711 in all KO APCs and PB1F2_62–70 and NP_366–374 in β5i/β2i KO APCs. Conversely, M1_128–135 is recovered at higher levels in β1i and β2i KO APCs than predicted by kinetics.

Table II.

Comparison of epitope presentation assessed by HPLC elution and BFA kinetics

Epitope	β1i−/−			β5i/β2i −/−			β2i −/−
	HPLC (%)	BFA (%)	Ratio	HPLC (%)	BFA (%)	Ratio	HPLC (%)	BFA (%)	Ratio
NP366–374	40	67		4	67	0.06	8	50
PA224–233	90	29		90	11		100	50
PB1703–711	5	100	0.05	1	88	0.01	5	58	0.09
PB1F262–70	4	15		8	225	0.04	12	100
M1128–135	2500	225	12	300	69		1000	90	11
NS2114–121	75	120		50	133		25	120

Peptide production and presentation of six IAV Tcd8+ epitopes were measured from wt, (β1i^−/−, (β5i/β2i^−/−, and β2i^−/− BMDCs and quantified relative to wt half maximal results (which were considered as 100%). Any difference >10-fold between the HPLC fraction titration and BFA kinetics assessment for the same epitope is shown in boldface italic type.

There are several potential explanations for these discrepancies. First, the BFA assay is much better suited for studying epitopes such as PB1_703–711, PA_224–233, and NA_181–189 whose slow presentation is more likely to linearly reflect the generation of peptide class I complexes than for rapidly presented epitopes such as NP_366–374 and PB1F2_62–70, which rapidly saturate T cell activation. Second, M1_128–135 showed an unusually wide elution peak (Fig. 4K), consistent with structural heterogeneity of the epitope that T cells might not distinguish, whereas we only measured peptide in the peak fraction. Third, and most importantly, HPLC elution experiments might be biased by intracellular peptide pools that form independently of their cognate class I molecule and therefore do not truly represent presented molecules.

The pioneering work of Rammensee and colleagues (25) established that antigenic peptide recovery is dependent on the presence of MHC class I molecules, presumably owing to their rapid destruction by proteasomes and many other cytosolic and endoplasmic reticulum proteases. But how general is this rule? We previously reported that PA_224–233 when expressed as a vaccinia virus-expressed minigene product forms a large intracellular pool independently of class I. Indeed, the pool was rather dependent on active heat shock protein (HSP)90, likely through binding to a HSP90-dependent client because we could not directly demonstrate association of PA_224–233 with HSP90 (27).

In this study we provide the initial evidence that PA_224–233 naturally generated from an IAV infection may form a large class I-independent pool, because its recovery from APCs is completely independent of its presentation kinetics. Class I–independent pools could also explain the clear discrepancy in PB1_703–711 recovery from wt versus KO APCs if only complete immunoproteasomes can generate peptides that populate the class I–independent pool. Although this idea may seem far-fetched, peptide generation can be complex, with different rules, for example, applying to ostensibly identical proteins encoded by viral versus cell mRNA as well as defective ribosomal products versus retirees, and even within retiree subsets (33). The possibility must be entertained that even identical peptides generated by different sorts of proteasomes may have different fates.

Although our experiments do not differentiate direct priming versus cross-priming, an exciting implication of cytosolic pools of PA_224–233 in IAV-infected cells is that its robust immunogenicity in primary responses might reflect the importance of cross-priming in activating naive anti-IAV T_CD8+. Lev et al. (27) showed that unlike other peptides, owing to its cytosolic pool, PA_224–233 was potently cross-presented and could prime T_CD8+ responses. Could it be that PA_224–233 dominates primary responses due to the importance of cross-priming in this process along with an advantage of PA_224–233 in cross-priming due to its stable pool? Conversely, the loss of ID status in secondary responses implies that direct presentation may play a more important role in the recall response.

The discrepancies we observe between the generation of PA_224–233 (and other peptides in an immunoproteasome subunit–dependent manner) and their presentation kinetics and ID provide a launching point for future studies on the exact contribution of direct versus cross-presentation in IAV and other infectious models, and despite intense research during the last three decades, remains a central issue for understanding ID and rational vaccine design.

Glossary

Abbreviations used in this article

B6

C57BL/6

BFA

brefeldin A

BMDC

bone marrow–derived dendritic cell

DC

dendritic cell

HA

hemagglutinin

IAV

influenza A virus

ICS

intracellular cytokine staining

ID

immunodominance

IDD

immunodominant determinant

KO

knockout

M1

matrix protein 1

MOI

multiplicity of infection

NA

neuraminidase

NP

nucleoprotein

NS2

nonstructural protein 2

PA

acidic polymerase

PB1

basic polymerase 1

PB2

basic polymerase 2

RP-HPLC

reverse-phase HPLC

SDD

subdominant determinant

T_CD8+

CD8⁺ T cell

wt

wild-type.