Immune Defects in 28-kDa Proteasome Activator γ-Deficient Mice¹

Proteasomes are large multicatalytic proteases found throughout the eukaryotic cell. They mediate the normal turnover of most intracellular proteins, including proteins tagged for destruction by the ubiquitin system, such as defective ribosomal products, signaling, and regulatory proteins (1, 2). In vertebrates, the proteasome is not only responsible for housekeeping functions, but is also responsible for generating peptide epitopes for MHC class I molecules for immune surveillance. The core 20S proteasome is assembled from 28 subunits arranged into four rings of seven subunits. It has a α₇β₇β₇α₇ organization, with three catalytic sites located inside each β ring at positions β₁, β₂, and β₅ (3, 4, 5). The α rings of the proteasome form a tightly closed pore at either end of the proteasome complex, regulating substrate access to the catalytic core of the enzyme (6, 7, 8). Due to this structure, it is thought that these 20S proteasomes are largely latent in vivo, and require activation for optimal function. Activation of the proteasome minimally includes opening this outer pore to allow greater substrate access to the catalytic sites located within the internal chamber, but may also involve allosteric effects on individual catalytic sites.

There are two classes of proteasome activator complexes found in eukaryotic cells, which bind to the α rings of the proteasome and enhance catalytic function. The PA700 (or 19S) activator is composed of ∼20 subunits and binds to proteasomes in an ATP-dependent manner to form the 26S proteasome complex (9), which is primarily responsible for the degradation of ubiquitinylated proteins. A second class of activators is ATP independent and is composed of a family of proteins known as PA28⁴(28-kDa proteasome activator). The three PA28 family members form two complexes in vivo: a hetero-oligomer composed of α and β subunits, and a homo-oligomer of γ subunits (10, 11, 12). The different PA28 complexes appear to have different subcellular distributions, with PA28γ restricted to the nucleus, and PA28α/β found throughout the cell (13, 14, 15). In addition, PA28α and -β expression is enhanced by treatment with IFN-γ, whereas PA28γ expression is not significantly affected (16, 17). However, both complexes bind to and activate proteasomes to digest peptides in vitro. Unlike PA700, the PA28 proteins do not enable proteasomes to digest full-length proteins or ubiquitinylated substrates in vitro (11).

Previously published data demonstrate that human PA28γ activates the individual catalytic activities of the proteasome unequally, with a strong bias toward cleavage after basic residues and only a weak to moderate enhancement in proteolysis after acidic or hydrophobic residues (18, 19). In contrast, PA28α/β has been reported to strongly activate all three catalytic activities of the proteasome. Therefore, PA28α/β and -γ differentially enhance the rate of proteolysis and may also differentially affect proteasome cleavage specificities. Thus, it seems likely that the PA28 family of proteasome activators may play a significant role in the flexibility of Ag processing in the cell. Both in vitro and in vivo evidence suggests that PA28α/β is essential for the processing of specific antigenic epitopes (20, 21, 22). However, there exists no evidence that PA28γ acts in a similar manner. The only observed phenotype of a PA28γ^−/− mouse was a reduction in body size coupled with defects in mitotic progression of cultured embryonic fibroblasts (MEFs) (23). As degradation of many of the components that regulate progression through the cell cycle is a proteasome-mediated process, this study implicates PA28γ as activating the proteasome in vivo. Therefore, the goals of this investigation were to assess the role of PA28γ in proteasome-mediated Ag processing in vivo.

We found that unmanipulated PA28γ^−/− mice have normal levels of surface MHC class I molecules, but have slightly reduced numbers of CD8⁺ T cells. The proportions of CD8⁺ T cells responding to a panel of influenza virus epitopes after influenza infection as well as the proportion responding to an SV40 T Ag (T Ag) epitope after infection with recombinant vaccinia virus are also normal. However, PA28γ^−/− mice demonstrate reduced numbers of CD8⁺ T cells and impaired clearance after infection with the intracellular fungal pathogen Histoplasma capsulatum. These results suggest that the loss of PA28γ expression results in Ag-specific defects in the processing and presentation of MHC class I-restricted T cell epitopes.

Mouse PA28γ genomic clones were isolated from a 129 Sv/J liver DNA λ Zap2 library. The targeting vector was constructed as displayed in Fig. 1,A. The PA28γ targeting construct was linearized with NotI and inserted into 129Sv/J embryonic stem cells by electroporation. G418-resistant colonies were selected and screened for homologous recombination events. Chimeric animals were bred to C57BL/6 partners, and germline transmission of the mutant allele was identified by Southern blot analysis with either 5′ or 3′ probes (Fig. 1 A). Progeny containing the mutant allele were intercrossed to generate PA28γ^−/− mice. Gene disruption was confirmed by Southern and Western analysis. The mutant allele was then bred onto the C57BL/6 for 10 backcross generations, followed by intercrossing to generate homozygotes. The genotype of the offspring was monitored by PCR on genomic tail DNA with the following primers PA28γ knockout (KO)-specific (5′-TCGAGCGAGCACGTACT-3′), wild-type (WT)-specific (5′-CTAACATAACTTACCTTGCC-3′), and common PA28γ (5′-CACGATGGACTGGATGGT-3′).

MEFs were generated from 16- to 18-day-old embryos (24) and cultured at 37°C in 5% CO₂ in DMEM containing 10% FBS supplemented with 1 mM sodium pyruvate, 50 μM 2-ME, 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin (Life Technologies, Gaithersburg, MD).

All protein samples were separated on 11% polyacrylamide gels. Western blot analysis and quantitation using ImageQuant software (Molecular Dynamics, Sunnyvale, CA) were performed as previously described (25).

Infections with influenza virus (26) and recombinant vaccinia virus expressing SV40 T-Ag (rVV-941T) (27) were performed as previously described. Infections with 2 × 10⁶ live H. capsulatum were performed as previously described (25, 28).

For cell cycle analysis, the designated cells were harvested and washed in HBSS, and single cells were fixed and permeabilized at −20°C in ice-cold 70% ethanol for a minimum of 24 h. After fixation, cells were washed twice in ice-cold PBS containing 100 μM EDTA and then incubated in the presence of 25 μg/ml propidium iodide and 25 μg/ml RNase A for 15–20 min before analysis by flow cytometry (FACSCalibur; BD Biosciences, Mountain View, CA) (29). For analysis of apoptotic cells in mitogen-stimulated cultures, primary spleen cells were cultured for 72 h in the presence of either of 4 μg/ml Con A or 40 μg/ml Escherichia coli O55:B5 LPS, fixed, and stained with propidium iodide as described above, and sub-G₀ apoptotic and fragmenting cells were counted by flow cytometry.

For cell surface staining, lymphocytes were isolated from lung or spleen tissue by homogenization in RPMI 1640 containing 10% FBS, supplemented with 2 mM l-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, 1 mM sodium pyruvate, 1 mg/ml gentamicin, 50 μM 2-ME, and 0.1 mM nonessential amino acids (Life Technologies). After removal of debris by filtration through nylon mesh, cells were loaded onto a 40–70% Percoll gradient and centrifuged. Purified lymphocytes were washed twice with RPMI 1640 medium, counted, diluted in HBSS, and plated into 96-well, V-bottom plates. Cells were washed in PBS with 5% FBS and 0.02% sodium azide and stained with Ab against the indicated cell surface markers (BD PharMingen, San Diego, CA) in FACS medium. Cells were fixed with 2% paraformaldehyde in PBS and kept at 4°C until analysis by flow cytometry was performed (LSR and FACSCalibur (BD Biosciences) or EPICS XL (Coulter, Hialeah, FL)). Analysis of activated influenza-specific CD8⁺ T cells by intracellular cytokine staining was performed as previously described (30). Staining and analysis of SV40 T Ag-specific CD8⁺ T cells were performed as previously described (31).

Groups of H. capsulatum-infected animals were sacrificed at 7, 14, and 21 days after infection, and lungs and spleens were harvested. The organs were homogenized in HBSS to lyse host cells and release live H. capsulatum organisms. These organ lysates were serially diluted with HBSS and cultured on mycobiotic agar plates for 7–10 days at 37°C, after which colonies were counted (32).

Spleen cells from PA28γ^−/− and control C57BL/6 mice were isolated and plated at 3 × 10⁶/ml into a 96-well plate (4.5 × 10⁵ cells/well) in RPMI 1640 containing 10% FBS and supplemented with l-glutamine, penicillin, and streptomycin. Cells were cultured at 37°C and 5% CO₂ for 0, 24, 48, or 72 h alone or in the presence of 4 μg/ml Con A (Sigma-Aldrich, St. Louis, MO) or 40 μg/ml LPS (Roche, Indianapolis, IN) before addition of 1 μCi/well ³H-labeled thymidine. After 16- to 18-h incubation, the cells from each plate were harvested onto a membrane filter using a semiautomatic cell harvester (Skatron, Molecular Devices, Sunnyvale, CA). Each sample was then counted in a LS-3801 liquid scintillation counter (Beckman, Fullerton, CA) using ScintiSafe 30% scintillation mixture (Fisher Scientific, Pittsburgh, PA).

The targeting construct was designed to delete exons 5–8 of the PA28γ gene (Fig. 1 A) and results in a frameshift mutation in downstream exons. The deleted segment of the gene corresponds to the activation loop of the protein, known to be responsible for binding to and activating the proteasome (33, 34), and hence this mutation is expected to result in complete loss of PA28γ activity.

Segregation of the mutated allele displayed normal Mendelian ratios in intercrosses of PA28γ^+/− mice. PA28γ^−/− mice are fertile, develop normally, and display an essentially normal gross phenotype. Analysis of genomic DNA of PA28γ^+/− and PA28γ^−/− mice confirmed the gene disruption (Fig. 1,B), and Western blots were used to confirm the absence of protein expression in PA28γ^−/− animals (Fig. 1 C).

After weaning, PA28γ^−/− mice gain weight at a reduced rate compared with WT or heterozygous littermates. Representative growth curves for all three genotypes from weaning to 15 wk of age are shown in Fig. 2 A. From 9–15 wk of age, male WT and heterozygous animals gain weight at ∼0.60 g/wk, whereas homozygous PA28γ littermates gain weight at <0.50 g/wk. This tendency is similar to what has been reported in a different PA28γ^−/− mouse line (23). The difference in weight gain is probably caused by an overall reduction in cellularity or cell size, because mice of all three genotypes maintain constant organ to body mass ratios (data not shown). The ratios for heart, liver, lungs, kidney, and spleen remained constant regardless of age, sex, or genotype.

The size reduction witnessed in PA28γ^−/− animals may be partially explained by aberrant growth kinetics and rates of apoptosis observed in MEFs derived from homozygous KO animals. As previously described, PA28γ^−/− MEFS showed a reduced growth rate and lower saturation density than heterozygous counterparts (23). Furthermore, asynchronous populations of PA28γ^−/− MEFs consistently displayed a reduced number of actively mitotic cells in either S or G₂/M phase by flow cytometry (Fig. 2,B), and actively growing populations of PA28γ^−/− MEFs displayed a 2- to 3-fold increase in apoptotic cells over their heterozygous counterparts (Fig. 2 C).

PA28γ^−/− mice were examined for the expression of MHC class I pathway-associated proteins. Quantitative Western analysis demonstrated no consistent differences in the expression of PA28α and -β, total proteasomes, or immunoproteasomes between WT (C57BL/6) and PA28γ^−/− mice (Fig. 3,A). Furthermore, purified 20S proteasomes from PA28γ^−/− mice displayed normal levels of immunoproteasome subunits and comparable activity profiles to purified proteasomes from heterozygous or WT littermates (data not shown). PA28γ^−/− and C57BL/6 mice also express identical amounts of cell surface MHC class I molecules (mean fluorescence intensity: PA28γ^−/−, 56.5 ± 1.6; WT, 55.3 ± 1.4; Fig. 3,B). Interestingly, spleen T cell and B cell percentages are normal in knockout animals (data not shown), but CD4⁺/CD8⁺ ratios are slightly elevated in the spleen (Fig. 3 C), corresponding to a slight (∼15%), but significant, reduction in CD3⁺CD8⁺ T cells compared with WT controls. We were unable to detect significant differences in thymic MHC class I expression or proportions of thymic T cell subpopulations (data not shown).

The proliferative capacity of primary splenocytes from PA28γ^−/− and C57BL/6 mice was investigated by [³H]thymidine incorporation. The kinetics and magnitude of the responses to both Con A and LPS stimulation were virtually identical in PA28γ^−/− and C57BL/6 cells (Fig. 4,A). Furthermore, mitogen-stimulated PA28γ^−/− and C57BL/6 splenocytes demonstrated identical proportions of mitotic and apoptotic cells by flow cytometry (Fig. 4 B and data not shown). Thus, the proliferative defects seen in MEF cultures were not apparent in mitogen-stimulated lymphocyte cultures, suggesting that different mitotic controls (and, hence, differential dependence on PA28) are operative in these cell types.

The ability of the PA28γ^−/− mice to process and present a variety of epitopes from native proteins was investigated using an influenza model system. CTL from PA28γ^−/− mice infected with influenza virus were screened for reactivity against a panel of previously characterized, MHC-restricted epitopes. The percentages of CD8⁺ T cells expressing intracellular IFN-γ after in vitro restimulation with each individual epitope is shown in Fig. 5. No statistically significant differences were observed between PA28γ^−/− and control (heterozygous or WT) littermates in either the proportion of T cells responding to each epitope or the hierarchy of immunodominance among the various epitopes. PA28γ^−/− mice had no obvious defect (or enhancement) in their ability to process any of these defined epitopes regardless of whether the source protein was cytosolic or nuclear.

To further examine the processing and presentation of nuclear Ags, we measured the proportions of SV40 T-Ag responding cells after in vivo immunization with recombinant vaccinia virus. CD8⁺ T cells specific for the SV40 T Ag epitope IV were enumerated with an MHC class I tetramer as previously described (31, 35). Although the proportions of T cells responding to T Ag IV were normal compared with those in littermate controls, we did observe a reduction in total CD8⁺ splenic T cells in infected PA28γ^−/− animals compared with controls (Fig. 6), suggesting that the response to other T Ag or, more likely, vaccinia epitope(s) may be adversely affected in these animals.

PA28γ^−/− and control C57BL/6 animals were infected with H. capsulatum, and the ability of the animals to clear the infection was monitored. Spleens and lungs were harvested on days 7, 14, and 21 postinfection and cultured to quantify the number of surviving H. capsulatum organisms. The number of viable organisms recovered from the lungs on day 7 was slightly, but not significantly, elevated in PA28γ^−/− mice (1.5 × 10⁶ for controls; 5 × 10⁶ for PA28γ^−/−; Fig. 7 A). Between days 7 and 14, control animals demonstrated a 40-fold reduction in live organisms recoverable from lung (to 4 × 10⁴ organisms) and another 15-fold reduction between days 14 and 21 (to 2500 organisms). Although PA28γ^−/− animals also cleared the infection (1500 organisms recovered on day 21), the kinetics of clearance were slowed. Between days 7 and 14, only a 10-fold reduction was observed, resulting in statistically significantly more (10-fold) live organisms in the lung compared with the controls (400,000 vs 40,000). Significantly (p < 0.05) elevated numbers of live organisms were also recovered from PA28γ^−/− spleens on day 14 (data not shown).

Lungs (data not shown) and spleens (Fig. 7,B) were harvested from groups of infected animals on day 14, and lymphocytes were isolated, stained for cell surface marker expression, and examined by flow cytometry. Day 14 has previously been shown to correspond to the peak of the T cell-mediated immune response in the lungs of H. capsulatum-infected animals (36). PA28γ^−/− mice display normal CD4⁺ T cell (Fig. 7 B) and B cell (data not shown) proportions in both organs. However, PA28γ^−/− mice display a significantly (p < 0.05) reduced proportion of total CD3⁺ T cells in the spleen and a significantly (p < 0.05) reduced proportion of CD8⁺ T cells in both organs. Thus, the increased fungal burden seen in PA28γ^−/− animals early in the infection appears to result from a defective CD8⁺ T cell response.

In this study we used PA28γ gene-targeted mice to examine the role of PA28γ in complex immune responses. These PA28γ^−/− mice display a similar overall phenotype to a previously reported PA28γ^−/− mouse line (23). Both PA28γ^−/− strains manifest a normal gross developmental and anatomical phenotype, but they display an abnormally small body mass that appears to be due to miniaturization. Certain cell types, minimally MEFs, appear to have a cell cycle progression defect, previously characterized as a block in entrance to S phase (23), that may account for this miniaturization. However, we found no evidence for altered cell cycle progression in lymphocytes of the PA28γ^−/− animals. These cells respond and expand normally, i.e., with the same kinetics and magnitude, in response to both T and B cell mitogens in vitro (Fig. 4). We also found no differences between PA28γ^−/− and control lymphocytes in the extent of either mitosis or apoptosis in mitogen-stimulated cultures (Fig. 4 and data not shown). Moreover, CD8⁺ T cells to defined epitopes examined in the context of two different viral infections in vivo appeared to proliferate and expand to the same extent as those from WT animals (Figs. 5 and 6), further suggesting that there is no defect in the ability of the PA28γ^−/− lymphocytes to proliferate effectively in response to normal stimulation by Ag through the TCR. It is likely that different cell types, in particular MEFs compared with lymphocytes, use differentially expressed cell cycle regulatory proteins whose dependence on PA28γ for appropriate degradation may vary. Some potential candidates are p27kip1 (37), Sp1 (38), p53 (39), and members of the Myc family (40, 41).

Despite the apparently normal proliferative capacity of lymphocytes from KO animals, our observation of reduced body weight, but constant organ to body weight ratios, implies that the KO animals have fewer total lymphocytes than do normal control animals. However, these numbers are appropriate for the size of the animal and imply the operation of normal homeostatic mechanisms for controlling lymphocyte expansion in the KO animals. Nevertheless, when T cell subpopulations were examined, a slight, but consistent and significant, reduction in the proportion of CD8⁺ T cells (to ∼85% of normal levels) was observed in the PA28γ^−/− animals, resulting in a slightly higher overall CD4:CD8 T cell ratio. The most obvious potential explanation for this deficit is a defect in positive selection of CD8⁺ T cells in the thymus as a consequence of an inability to supply MHC class I molecules with a complete repertoire of peptide epitopes. This could result either from a direct requirement for PA28γ in the efficient generation of particular peptides by the proteasome (as has been postulated for the PA28α/β complex) or an indirect effect on the incorporation of differentially expressed proteasome catalytic subunits during proteasome assembly, as has been observed in PA28β^−/− (42) (but curiously not in PA28α^−/−β^−/− (43)) mice. We found no evidence to support the latter possibility. PA28γ^−/− mice express normal levels of PA28α and -β (Fig. 3) (23) and normal levels of the mature form of both constitutive and immunoproteasome subunits (Fig. 3). These data suggest that PA28γ does not play an essential chaperone role in proteasome assembly in vivo and are consistent with observations made in PA28α^−/−/β^−/− mice.

If the reduced numbers of CD8⁺ T cells in PA28γ^−/− mice reflects MHC class I peptide repertoire defects, we would expect to observe such defects in peptide presentation during the effector phase of immune responses to specific Ags. We first examined responses in two viral systems in which CD8⁺ T cell epitopes have been characterized. As PA28γ is predominantly a nuclear protein, these systems were also chosen in part because they involve responses to viral proteins localized to the nucleus of the infected cell.

PA28γ^−/− animals generated normal proportions of CD8⁺ T cells against all epitopes of influenza. Furthermore, there were no significant alterations in the hierarchy of the responses to the various epitopes (Fig. 5).

We also analyzed CD8⁺ T cell responses to (nuclear) SV40 T Ag after immunization with a recombinant vaccinia virus expressing this protein (Fig. 6). Although PA28γ^−/− animals respond normally to the T Ag IV epitope, we observed a significant reduction in total CD8⁺ splenic T cells. These results suggest that PA28γ^−/− animals may have a defect in their ability to process and respond to other T Ag epitopes (either the subdominant epitopes I or II/III) or, more likely, to as yet undefined epitopes from the vaccinia virus vector.

Similar results were obtained when the response to a more complex eukaryotic pathogen, H. capsulatum, was examined. Control and clearance of an H. capsulatum infection in mammals are primarily dependent on CD4⁺ Th1 cells; CD8⁺ T cells are required for optimal clearance of the infection, but not for host survival (32, 44, 45). Although, H. capsulatum infection dramatically affects proteasome subunit composition and may therefore affect MHC class I presentation (25), we found no differences in proteasome composition between WT and PA28γ^−/− animals after H. capsulatum infection, as determined by Western analysis (data not shown). Nonetheless, in three independent experiments, on day 14 after intranasal infection, PA28γ^−/− mice had ∼10-fold higher fungal burden than WT controls, although both groups of animals eventually cleared the infection. Interestingly, in a fourth experiment (data not shown), significant mortality was observed in both groups, either because the animals unintentionally received a higher dose of pathogen or because the pathogen preparation itself was more virulent. In this experiment, mortality in the PA28γ^−/− group was 60%, compared with 20% in the control group, further suggesting that the differences in fungal burden observed in the other three experiments may have important implications for fitness and survival under certain conditions.

These results recapitulate a phenotype similar to that seen in H. capsulatum-infected β₂-microglobulin^−/− mice or CD8⁺ T cell-depleted C57BL/6 animals (32) and are consistent with the possibility of a defect in the MHC class I pathway and CD8⁺ T cell immunity. When the lungs of infected animals were examined for T cell infiltrates, we observed a significantly reduced (p < 0.05) proportion of CD8⁺ T cells in KO compared with control animals. Examination of the spleens also showed a similar reduction in CD8⁺ T cells as well as in the proportion of total CD3⁺ T cells. These data further support the conclusion that a defect in CD8⁺ T cell immunity is responsible for the delayed clearance of H. capsulatum observed in PA28γ^−/− animals. Whether this defect occurs primarily during selection of the T cell repertoire (failure of positive selection of T cell receptors against critical epitopes of the pathogen), during the inductive/effector phase of the immune response (failure to generate particular epitopes by proteasomes in infected cells), or both is unclear at present. The fact that the magnitude of the CD8⁺ T cell difference between KO and control animals (∼15% fewer in KO) is similar between uninfected and infected groups suggests that those CD8⁺ T cells present in the KO animals expand normally in response to the infection and may be interpreted in favor the former over the latter hypothesis. Unfortunately, we found no defects in the T cell responses to defined epitopes in the two viral infection systems tested, and the epitopes recognized by CD8⁺ T cells during H. capsulatum infection are currently completely uncharacterized. In either case, however, it is likely that the underlying defect is an inability to produce a particular subset of peptides (from either self proteins, in the case of positive selection of the T cell repertoire, or from pathogen proteins during the priming and effector phases of the response) in the absence of PA28γ. The fact that PA28γ^−/− mice express normal levels of total proteasomes and immunoproteasomes suggests that any effect on MHC class I Ag processing should be directly attributable to PA28γ’s role as a proteasome activator. Therefore, these data represent the first evidence that PA28γ must function in the MHC class I Ag processing pathway and suggest that it exerts an influence on proteasome cleavage specificity. Further analyses of other infectious disease models as well as investigations into the effect of PA28γ on proteasome cleavage specificity in vitro are currently underway.

We thank the University of Cincinnati gene-targeting core, Jeanna Guenther, Timothy Hubbell, David Ginsburg, Holly Allen, Mark Scheckelhoff, Dr. George Babcock, Jim Cornelius, Sandy Schwemberger, Jacob Johnson, Dr. Joan Cook-Mills, Dr. Jon Yewdell, and Dr. Jack Bennink for their assistance and advice.