Impact of clonal competition for peptide-MHC complexes on the CD8⁺ T-cell repertoire selection in a persistent viral infection

The αβ T-cell receptor (TCR) recognizes immunogenic peptides in association with the major histocompatibility complex (pMHC) on the surface of virus-infected cells.^1,2  Upon recognition of foreign antigen, the naive T-cell repertoire responds to produce either polyclonal, oligoclonal, or clonal immune responses. The diversity of the TCR repertoire is created primarily due to differences in the sequence of the hypervariable complementarity determining region 3 (CDR3) of each chain of the TCR.^3,4  The CDR3 loops are formed at the junction of the V(D)J gene segments as a result of random nucleotide addition and removal during the gene rearrangement process in the thymus.^5,6  Numerous factors shape the T-cell repertoire, including thymic selection of the naive T-cell repertoire, TCR avidity for the pMHC complex, antigen load, and duration of the pMHC-TCR interaction.^7-10  However, the relative contribution of these in shaping the TCR repertoire remains unclear

TCR repertoire selection is also complicated by MHC class I–restricted T cells interacting with peptides of noncanonical and canonical length. Typically, noncanonical peptides of more than 10 amino acids in length bulge from the peptide-binding cleft,^11-17  and this unusual feature can result in biased selection of the TCR repertoire, suggesting the structure of the pMHC complex plays an important role in shaping the TCR repertoire. Conversely, featureless canonical 8- to 9- amino acid long epitopes from both persistent and nonpersistent viruses can also induce a highly biased T-cell repertoire.^18-21  Collectively, these studies highlight that a variety of factors shape the TCR repertoire.

Here we have examined how endogenous antigen presentation and the structural landscape of the pMHC complex might together determine the character of CD8⁺ T-cell responses during a persistent viral (human cytomegalovirus; HCMV) infection in humans. As such, we have developed a model system based on 2 immunodominant HCMV epitopes derived from the structural protein, pp65. These epitopes, the noncanonical ¹⁰³CPSQEPMSIYVY¹¹¹ (CPS) and the canonical ¹⁸⁸FPTKDVAL¹⁹⁵ (FPT), are restricted through HLA B*3508.²²  Given that these 2 epitopes are highly immunogenic, derived from the same antigenic source, and restricted through the identical MHC alleles, this model provides an ideal opportunity to investigate the role of pMHC in shaping the TCR repertoire. Here, we present evidence that competing levels of endogenous epitope presentation is pivotal in shaping the T-cell repertoire during a persistent viral infection.

These studies were approved by the Human Research Ethics Committee of the Queensland Institute of Medical Research. Informed consent from all blood donors was obtained in accordance with the Declaration of Helsinki.

The human fibroblast cell line, MRC-5 was used as host for HCMV infection and was maintained in growth medium (10% fetal calf serum [FCS] in RPMI 1640 medium). Phytohemagglutinin blasts were generated as described previously.²³  Peripheral blood mononuclear cells (PBMCs), isolated from healthy donors after informed consent, were used to expand HCMV-specific CTL clones and polyclonal T-cell lines.²³  CTL cultures were generated by culturing PBMCs (2 × 10⁶ per 2 mL well) in 10% FCS/RPMI 1640 with autologous PBMCs that had been precoated with an HCMV peptide epitope (0.1 μM for 1 hour, responder:stimulator ratio of 2:1). Cultures were supplemented with rIL2 (20 U/mL) on day 3, split on day 7, and analyzed on day 10. CTL cultures were tested in for cytotoxicity in the standard 5-hour chromium release assay. HLA-peptide multimers (ProImmune, Oxford, United Kingdom) were also used to confirm the specificity of some CTL cultures.

PBMC or HCMV-specific T cells from healthy virus carriers were incubated for 30 minutes at 4°C with an APC-labeled HLA-B*3508-CPS pentamer or with an APC-labeled HLA-B*3508-FPT pentamer (Pro-Immune). Cells were then washed twice and labeled for 30 minutes at 4°C with Tricolor-labeled antibody to human CD8 (Invitrogen, Carlsbad, CA). Cells were again washed and analyzed on a FACSCanto with FACSDiva software (BD Biosciences, Palo Alto, CA). In some experiments these T cells were also co-stained with anti-Vβ 1 (TRBV9), 2 (TRBV20), 3 (TRBV28), 5.1 (TRBV5), 5.2 (TRBV5), 5.3 (TRBV5), 6.7 (TRBV7), 7 (TRBV4), 8 (TRBV12), 11 (TRBV25), 12 (TRBV10), 13.1 (TRBV6), 13.6 (TRBV6), 14 (TRBV27), 16 (TRBV14), 17 (TRBV19), 20 (TRBV30), 21.3 (TRBV11), 22 (TRBV2) and 23 (TRBV13)-FITC (Serotec, Raleigh, NC); anti-Vβ4 (TRBV29), 9 (TRBV3), and 18 (TRBV18)-PE (Immunotech, Marseille, France). Briefly, these cells were initially stained with anti–CD3 PE-Cy7 and anti–CD8 PE-Cy5 for 30 minutes at 4°C and then stained with the relevant pentamer. After extensive washing with phosphate-buffered saline (PBS) (supplemented with 1% FCS) these cells were stained with a panel of monoclonal Vβ antibodies for 30 minutes at 4°C. Cells were washed and fixed in 1% paraformaldehyde and analyzed on a FACS Canto using FACS Diva software (BD Biosciences).

The Enzyme-linked Immunospot assay used in this study has been described previously.²³  Briefly, 96-well multiscreen plates were coated overnight at 4°C with 1 μg/mL anti–human interferon (IFN)–γ mAb 1-D1K (Mabtech, Mosman, NSW, Australia). The plates were washed with PBS and blocked with culture medium. Between 2.5 × 10⁴ and 2 × 10⁵ cells/well were incubated for 24 hours at 37°C with 1 μg/mL of the relevant peptide (Mimotopes, Clayton, Australia). The plates were washed to remove cells, and cytokine bound to the nitrocellulose was detected using biotinylated anti–human IFN-γ, 7-B6–1 (Mabtech), followed by streptavidin alkaline phosphatase (Sigma-Aldrich, Castle Hill, Australia). ELISPOTS were developed using the substrate 5-bromo-4-chloro-3-indolyl phosphate/nitroblue tetrazolium (Sigma-Aldrich) and counted automatically using image analysis software.

MRC-5 were infected with AD169 at an MOI of 5, 2.5, 0.5, 0.25, or 0.05 for 2 or 16 hours and then labeled with ⁵¹Cr (Amersham Biosciences, Pittsburgh, PA) for 1 hour at 37°C. After incubation, these cells were washed in growth medium and used as targets in a standard cytotoxicity assay.²⁴ 

2.5 × 10⁴ MRC-5 cells were seeded into a 48-well plate containing 500 μL growth medium, and 24 hours later were infected with the Ad5F35 HLA B*3508 construct at an MOI of 5:1. Twelve hours later these cells were infected with AD169 at an MOI of 0.5 for 1, 2, 16, or 24 hours. HCMV-specific CD8⁺ T cells (5 × 10⁵) were added to the HCMV-infected and HLA B*3508⁺ MRC-5 cells, along with GolgiPlug. After 4 hours' incubation at 37°C, the HCMV-specific CD8⁺ T cells were then analyzed for intracellular IFNγ production and CD107a mobilization using the Cytofix/Cytoperm Plus (with GolgiPlug) kit (BD Pharmingen, San Diego, CA) according to the manufacturer's instructions. Cells were analyzed using the FACS Canto and FACS Diva software (BD Biosciences) and expressed as a percentage of the positive control (peptide-pulsed target cells).

The HLA B*3508-CPS and HLA B*3508-FPT complexes were expressed and purified as described.²⁵  Crystals of the HLA B*3508-CPS and HLA B*3508-FPT complexes were formed at 4°C with the hanging-drop vapor-diffusion technique, in which an equal volume of HLA B*3508-CPS or HLA B*3508-FPT (10 mg/mL) was mixed with the reservoir (200 mM ammonium acetate, 16% PEG 3350, and 100 mM cacodylate, pH 7.6).

Crystals were soaked in reservoir solution containing increasing increments of glycerol as a cryoprotectant (5%, 10%, and 15%) and then flash frozen before data collection. Data were collected at Bio-CARS (Advanced Photon Source, Argonne, IL) and was processed and scaled using the HKL suite.²⁶  The HLA-B35 complex structures were refined from an HLA B*3508 structure that was previously determined in our laboratory.¹⁴  The model was built/refined using ARP/wARP and Refmac/COOT. The progress of refinement was monitored by Rfactor and Rfree values. Rigid-body refinement and simulated annealing were used in the first instance, but in later rounds energy minimization and B-individual refinement were used to improve the quality of the model. See Table S1 for the final refinement and model statistics (available on the Blood website; see the Supplemental Materials link at the top of the online article).

The assembly and production of the recombinant Ad5F35-based adenovirus was completed in 3 stages using a highly efficient, ligation-based protocol of the Adeno-X System (Clontech, Palo Alto, CA). Firstly, the wild-type or mutated (with CD8-binding site modified) HLA B*3508 cDNA sequence was cloned into the pShuttle expression vector. Following amplification in Escherichia coli, the expression cassette from pShuttle was cloned into an Ad5F35 expression vector. The recombinant Ad5F35 vector was transfected into human embryonic kidney (HEK) 293 cells, and recombinant adenovirus (referred to as HLA B*3508 wt or mut) was harvested from the transfected cells by successive freeze-thawing cycles.

RNA was extracted from whole PBMCs or antigen-specific CD8⁺ T cells using TRIzol reagent (Invitrogen Life Technologies, Mount Waverley, Australia) as per the manufacturer's instructions. RNA was subjected to SuperScript III Reverse Transcriptase (Invitrogen Life Technologies) using antisense TCRα and TCRβ chain primers (5′-GTT GCT CCA GGC CAC AGC ACT G-3′ and 5′-TAT CTG GAG TCA TTG AGG GCG GGC T-3′, respectively). PCR was done in a 25-μL reaction containing a final concentration of 0.2 mM dNTP (Promega, Madison, WI), 2 mM MgCl₂ (Applied Biosystems, Foster City, CA), 10× buffer (Applied Biosystems), and 1.25 U AmpliTaq Gold (Applied Biosytems) with 0.2 μM TCR Cβ constant primer and one of the region-specific primers (0.2 μM). Thirty-five cycles were completed on a GeneAmp PCR System 9700 (Applied Biosystems) under the following conditions: 95°C 20 seconds, 58°C 40 seconds, and 72°C 1 minute. A 30-minute final extension was included. Samples were run on a 2% agarose gel at 90 V to confirm correct band size.

PCR product from the TRBV PCR was used in an elongation reaction using Platinum GenoTYPE Tsp DNA Polymerase (Invitrogen Life Technologies). A 5′-6-FAM-labeled C_βP primer was used (5′-TTC TGA TGG CTC AAA CAC-3′). After a 95°C 10-minute denaturation stage, 7 cycles of 95°C 20 seconds, 55°C 40 seconds, and 72°C 1 minute were completed. Products were diluted 1:10 in DEPC H₂O and added to GeneScan-500 ROX Standards (Applied Biosystems) and formamide. Samples were heat denaturated by boiling for 2 minutes before being run on an Applied Biosystems 3700 Sequencer. Results were analyzed using ABI PRISM GeneScan software (Applied Biosystems).

TRBV PCR product was purified using the Qiagen MinElute PCR Purification Kit (Qiagen, Doncaster, Australia) and was used to transform DH5α competent cells (Subcloning Efficiency DH5α Cells; Invitrogen Life Technologies) using the pGEM-T Vector System I kit (Promega) according to manufacturer's instructions. Cells were plated out onto agarose plates containing X-galactose and isopropyl-beta-D-thiogalactopyranoside (IPGT; Sigma-Aldrich) and left to grow overnight at 37°C. Colonies were picked and a PCR screen done using vector-specific primers to confirm correct product size. The DNA was purified as above and a sequencing reaction done using the ABI PRISM BigDye Termination Reaction kit (Applied Biosystems). Sequencing products were purified using the Qiagen DyeEx 2.0 Spin Kit according to the manufacturer's instructions. Sequences were analyzed on the Applied Biosystems 3700 Sequencer (Applied Biosystems) and using the IMGTV-Quest software.

Recent studies have shown that HLA micropolymorphisms can significantly impact on the selection of CD8⁺ T-cell epitopes from HCMV antigens.²²  This is illustrated by HLA B35-restricted CD8⁺ T-cell responses to the pp65 epitopes, CPSQEPMSIYVY and FPTKDVAL (herein referred to as CPS and FPT, respectively), which are exclusively recognized by HLA B*3508⁺ individuals, while HLA B*3501⁺, HLA B*3503⁺, and HLA B35⁻ donors showed no significant ex vivo response to these epitopes (Figure 1A). These observations were also confirmed by in vitro stimulation of T cells, which showed that these epitopes selectively expanded T cells from HLA B*3508⁺ individuals (Figure 1B).

Next we analyzed fresh PBMCs from HLA B*3508⁺ healthy virus carriers (donor 1: HLA A2, A2, B*3508, B57; donor 2: HLA A23, A32, B*3508, B49, and donor 6: HLA A30, A32, B*3508, B42) using pMHC pentamers in combination with a panel of monoclonal TRBV antibodies. Ex vivo staining of fresh PBMCs from donors 1 and 2 revealed that 5.9% and 3.71% of CD8⁺ T cells stained with CPS-pentamers, respectively, while 2.32% and 2.34% of CD8⁺ T cells stained with FPT pentamers. Co-staining of CPS-specific T cells with TCR-specific antibodies revealed a highly biased TCR repertoire with more than 80% of the T cells expressing TRBV28 (Figure 2A,B). Conversely, we could not detect a predominating TRBV usage for FPT-specific T cells (Figure 2C,D), indicative of an unbiased T-cell response. Next we expanded CPS- and FPT-specific T cells following stimulation with synthetic peptides and reanalyzed TCR usage by using pMHC pentamers in combination with the panel of monoclonal TRBV antibodies (Figure 2E-H). Although in vitro–expanded CPS-specific T cells showed no change in the TCR usage (Figure 2E,F); a donor-specific focusing of FPT-specific T cells was observed with unique TRBV expansions in each donor (Figure 2G,H).

To examine the repertoire diversity of the CPS- and FPT-specific CD8⁺ T cells in finer detail, we further analyzed the CDR3 region of the TCR using immunoscope technology. Figure 3A-C shows the immunoscope profile of CPS-specific CD8⁺ T cells from fresh blood or in vitro–expanded T-cell lines. In addition to strongly biased TRBV28 usage, CPS-specific T cells further showed a highly conserved CDR3 region, with all CDR3 regions being 9 amino acids in length (Figure 3J). By comparison, ex vivo analysis of the CDR3 region of the FPT-specific T cells revealed an oligoclonal profile of CDR3 lengths (Figure 3D), which is consistent with the data presented in Figure 2C,D. However, when these cells were cultured in vitro, the CDR3 profile remained either oligoclonal (Figure 3E) or showed a highly biased clonal CDR3 profile (Figure 3F-I) with an average length of 12 amino acids (Figure 3J).

To complement the immunoscope analysis, we next sequenced the CDR3 regions of the TCR. Analysis of the CPS-specific T cells from the fresh blood revealed a high level of conservation at the nucleotide level, which has been described as a “type 3 bias,” whereby both the TRBV usage and CDR3 sequence are identical or highly homologous¹⁰  (Figure 4A top). Furthermore, when the CDR3 region from the TRAV chain was sequenced, it too showed high levels of bias (Figure 4A bottom). Conversely, analysis of the FPT-specific T cells from fresh blood showed highly diverse CDR3 regions (Figure 4B). In vitro expanded clonal FPT-specific T cells showed donor-specific CDR3 sequences, while polyclonal cultures included a mixture of CDR3 sequences with varying length (Figure 4B). Taken together, these analyses clearly indicate that the noncanonical HLA B*3508-CPS complex drives a biased TCR selection, while canonical HLA B*3508-FPT complex favors oligoclonal TCR usage.

Considering the emerging evidence on the impact of pMHC-1 structural constraints on the binding of the TCR and its selection¹⁰  within an immune repertoire, we next determined whether the structural landscape of the HLA B*3508^CPS and HLA B*3508^FPT complexes correlated with the pattern of TCR repertoire. Crystal structures of HLA B*3508^CPS (PDB accession code 3BW9) and HLA B*3508^FPT (PDB accession code 3BWA) complexes were resolved to 1.75 and 1.30 Å, respectively (Figure 5 and Table S1). The HLA B*3508^CPS and HLA B*3508^FPT complexes were crystallized under the same conditions and thus, any structural differences observed were attributable to the peptide antigen, although the overall structure of HLA-B*3508 from both complexes was very similar (r.m.s.d. = 0.165 Å for all Cα atoms within the Ag-binding cleft).

The CPS peptide adopted an unusual conformation, whereby the epitope displayed a helical conformation that bulged from the Ag-binding cleft. The single helical turn was observed at the C-terminus of the peptide between the Met⁷ and the Tyr¹⁰ (Figure 5A-B). In addition to the Pro² and Tyr¹² main anchor positions, Glu⁵ acted as a secondary anchor, salt bridging with Arg¹⁵⁷ and Arg⁹⁷ (Table S2). The CPS peptide bulged approximately 6 Å out of the HLA-B*3508 cleft such that the side chains at positions Gln⁴, Pro⁶, Met⁷, Ser⁸, and Ile⁹ of the peptide were surface exposed and available for interaction with the TCR (Figure 5D). By comparison, the FPT peptide bound in an extended conformation, adopting a flatter “featureless” conformation in the Ag-binding cleft (Figure 5C), with all positions of the peptide interacting with HLA B*3508 (Table S2). In addition to the main anchor positions at P2-Pro and Leu⁸ the Asp⁵ acted as a secondary anchor, forming a salt bridge with the Arg⁹⁷ of HLA B*3508 (Table S2) while the side chains at positions Phe¹, Lys⁴, and Val⁶ of the peptide were surface exposed and available for interaction with the TCR (Figure 5E). These observations suggested that the structural features of the pMHC-I complexes may have minimal impact on the architecture of the antigen-specific T-cell repertoire during a persistent viral infection.

To further investigate whether this contrasting TCR selection was influenced by the presentation differences between the CPS and FPT epitopes, we developed ex vivo T-cell assays that used virus-infected cells as APCs. These assays allowed an assessment of clonal competition for the respective epitopes on virus-infected cells. Firstly, MRC-5 cells were co-infected with HCMV strain AD169 (MOI 0.5:1) and a recombinant adenoviral construct encoding HLA B*3508 and then exposed to CPS- or FPT-specific T cells and analyzed at different time intervals. The levels of INF-γ expression by each of these T cells were measured using intracellular cytokine assays. Data presented in Figure 6A showed that CPS-specific T cells were consistently stimulated more efficiently when compared with the FPT-specific T cells. An 8- to 10-fold difference in the IFN-γ expressing antigen-specific T cells numbers was observed over a period of 2 to 24 hours. To determine whether the decreased functional response of FPT-specific T cells was primarily due to limited availability of FPT epitopes in the context of natural HCMV infection or to low TCR affinity/avidity of FPT-specific T cells, we assessed functional avidity of CPS and FPT-specific T cells through peptide titration studies. MRC-5 cells expressing HLA B*3508 were sensitized with serially diluted synthetic CPS and FPT peptide epitopes and then exposed to antigen-specific T cells. Data presented in Figure 6B showed that the FPT-specific T cell recognized peptide-sensitized target cells more efficiently when compared with the CPS-specific T cells. These observations were also confirmed by intracellular cytokine assays, which showed almost 1.75-fold higher levels of IFN-γ expression by FPT-specific cells when CPS- and FPT-specific T cells were exposed to peptide-sensitized MRC-5 cells (Figure 6C). These results clearly show that the decreased functional response of FPT-specific T cells was primarily due to limited availability of FPT epitopes in the context of natural HCMV infection rather than low TCR affinity/avidity.

To examine the relative dependence on the CD8 co-receptor, we examined the recognition of virus-infected cells expressing wild-type or CD8 mutant HLA B*3508 molecules. HCMV-infected MRC-5 cells were co-infected with adenoviral constructs encoding either wild-type HLA B*3508 or a mutant form of HLA B*3508 (with D227K/T228A mutations on the CD8-binding site on the α3 domain) for 4 hours and then exposed to CPS- or FPT-specific T cells at a responder to stimulator ratio of 20:1 or 80:1. The levels of INF-γ expression by each of these T cells were measured using intracellular cytokine assays. Representative data presented in Figure 6D showed that CPS-specific T cells produced 1.5- to 2.5-fold higher IFN-γ levels when compared with FPT-specific T cells following exposure to HCMV-infected cells. More importantly, the T-cell recognition of HLA B*3508-CPS complex was significantly affected when the CD8-binding site was mutated, while the loss of the CD8-binding site on the HLA B*3508 molecule had minimal effect on the T-cell recognition of the FPT epitope. Notably, following presensitization with synthetic peptide, both the wild-type and mutant HLA B*3508 resulted in comparable levels of IFN-γ production by CPS- and FPT-specific T cells (Figure 6E).

To determine whether these differences in the endogenous presentation were related to the binding efficiency of the CPS and FPT epitopes to HLA B*3508 molecules, we conducted MHC stabilization assays based on T2.B*3508 cells and thermal stability assays, which demonstrated that these epitopes bound with similar affinity to HLA B*3508 (data not shown). Although these experiments indicated that both the FPT and CPS epitopes bind to HLA B*3508 with equal efficiency, our recent studies have shown that the dissociation half-life for B*3508^CPS and B*3508^FPT complexes was 37.1 hours and 21.1 hours, respectively.²²  Taken together, these observations suggested the differences in the disassociation kinetics of the CPS and FPT epitopes for HLA B*3508 directly impacts on their antigen presentation efficiency in HCMV-infected cells, which was consistent with our previously documented evidence on the impact of pMHC binding on determinant selection.²⁷ 

Previous studies have proposed that competition for cognate epitopes can impact on the clonal differences in the maturational phenotype of antigen-specific T cells.²⁸  To explore this possibility, we conducted ex vivo phenotypic analysis of CPS- and FPT-specific T cells from fresh PBMCs. For these experiments cells were co-stained with HLA B*3508^CPS or HLA B*3508^FPT pentamers and mAbs specific for CD27, CD57, and CD62L. CPS- and FPT-specific T cells showed dramatic differences in the expression of these markers (Figure 7A). The majority of the CPS-specific T cells displayed a terminally differentiated (CD27⁻, CD57⁺, and CD62L⁻) phenotype, while FPT-specific T cells were predominantly CD27⁺ CD57⁻, and CD62L⁺, a phenotype typical of central memory T cells (Figure 7A).

To determine whether clonal antigen competition was also associated with functional heterogeneity of CPS- and FPT-specific T cells in response to cognate antigen, we examined the IFN-γ expression and perforin degranulation reflected by CD107a mobilization by in vitro–expanded T cells following stimulation with HCMV-infected cells. Representative data presented in Figure 7B shows that co-expression of IFN-γ and CD107a mobilization in CPS-specific T cells was approximately 2.6- to 3.6-fold higher when compared with the FPT-specific T cells. This difference in the levels of CD107a mobilization following antigen stimulation was observed over 2 different responder-to-stimulator ratios (Figure 7B). Taken together, the contrasting phenotypic profile of CPS- and FPT-specific T cells and increased mobilization of CD107a and/or IFN-γ by CPS-specific T cells correlated well with the data presented in Figure 6 and suggests that enhanced antigen stimulation of CPS-specific T cells may be driving selective expansion of terminally differentiated effector cells and a public TCR response.

We have attempted to dissect the relative roles of antigen presentation kinetics and the structural landscape of pMHC-I in dictating the T-cell repertoire selection using stoichiometrically equivalent peptide determinants presented by same MHC allotypes. Our model, based on 2 immunodominant HCMV epitopes derived from the structural protein, pp65 (the noncanonical CPS and the canonical FPT) antigen, has the advantage of being independent of any virus or host-associated variables. A number of conclusions were drawn from these studies.

First, structural analysis revealed that the CPS epitope bound to the HLA B*3508 molecule adopted an unusual helical conformation that bulged from the Ag-binding cleft. Thus, the CPS-HLA B*3508 structural data were consistent with the previous finding that the bulged peptide epitopes select highly restricted public TCR.^10,11,16,29  On the other hand, the FPT epitope, which stimulated a highly diverse TCR repertoire, adopted a flatter “featureless” conformation. By comparison, the FPT peptide was deeply embedded in the MHC groove with fewer exposed residues for TCR interaction. This was an unexpected finding, as previous studies with influenza virus, which indicated that viral epitopes that have been termed “featureless” and presumably too similar to self, select a biased TCR repertoire.^18,30  Similarly, noncanonical epitopes that adopt bulged conformation when bound to the MHC class I molecules also exhibited highly biased TCR usage.^13,17  These studies highlight that it is likely that in addition to the pMHC-I landscape, other factors also contribute toward the selection of TCR repertoire in persistent viral infection.

Second, dominant CPS-specific clonotypes with a “type 3” biased TCR¹⁰  usage consistently showed much higher IFN-γ expression following exposure to virus-infected cells when compared with FPT-specific T cells expressing diverse TCR. This difference in their ability to produce IFN-γ was not due an inherent disparity, as these T cells produced comparable levels of this cytokine when stimulated with synthetic peptides. Although the CPS and FPT epitopes showed similar efficiency in stabilizing HLA B*3508, our recent studies have shown that the B*3508^CPS complexes display a much longer dissociation half-life when compared with B*3508^FPT complexes.²²  These observations suggested that the more efficient endogenous loading of the CPS epitope on HLA B*3508 molecules in virus-infected cells may drive increased expansion of CPS-specific T cells and thus clonal selection of a public TCR. Indeed, previous studies have shown that chronic stimulation of antigen-specific T cells results in continuous selection and accrual of high avidity clonotypes during the latent phase of infection.^31,32  Interestingly, CPS-specific T cells showed higher dependency on pMHC-CD8 interaction when compared with FPT-specific T cells, which is consistent with ongoing competition for antigen in vivo. Indeed, these observations are consistent with an earlier hypothesis that high levels of viremia during viral infection might lead to an expansion of clonotypes, which are highly dependent on the pMHC-CD8 interaction.²⁸  Although previous studies have argued that higher dependency on pMHC-CD8 interaction indicates lower avidity and may reflect a function of virus biology,^28,33,34  we propose that CD8 dependency is an intrinsic property of the epitopes and this interaction may provide increased stability for the binding of TCR for noncanonical peptide-MHC conformations.

Third, ex vivo phenotypic analysis of CPS- and FPT-specific T cells revealed that the CPS-specific T cells displayed a terminally differentiated effector phenotype (CD27^+/−CD57⁺ and CD62L⁻), while FPT-specific T cells were predominantly CD27⁺ CD57⁻ and CD62L⁺, a phenotype typical of central memory T cells.^35,36  This phenotypic difference was also reflected in their functional ability. The CPS-specific polyclonal T cells showed 2.6- to 3.6-fold higher IFN-γ expression and CD107a mobilization when compared with the FPT-specific T cells. These 2 different sets of experiments clearly showed that enhanced in vivo stimulation of CPS-specific T cells is driving selective expansion of terminally differentiated effector cells and a public TCR response. These studies clearly demonstrate that in vivo clonal competition for pMHC-1 complexes also plays a crucial role in the selection of TCR repertoire.

Taken together, data presented here further emphasize that TCR selection is a highly complex process and additional selection forces other than structural constraints would also have significant impact on this process.^7,9,28,37-39  Indeed, after a decade of intense interest into what dictates TCR selection, there is an increasing realization that many different parameters, all operating to varying degrees, seem to be at play. No single or even set of several parameters seem to universally apply for every pMHC combination. In fact, quite the opposite appears to be true. In latent or otherwise benign viral infections there is an apparent flexibility/plasticity inherent within the sum total T-cell response. This is particularly exemplified here with the selection of equally opposed repertoire responses despite the co-immunodominance and shared HLA restriction. The data provide further evidence that selection is a highly ordered yet incredibly faceted process that ultimately operates to ensure a hugely diverse collective response to a particular pathogen. To get such a model operational, one could imagine that clonal competition for endogenously processed epitopes could provide one facet of such a dynamic environment. Similarly, the pMHC-1 structural landscape, avidity, and precursor frequency would also interplay. More importantly, it would be incorrect to view TCR selection as a 2-dimensional model and that rather a model of integrating and competing variables need to be explored. It will be important to expand these observations to other epitopes (within individual viral infections) to provide further support for this proposition.

The authors wish to thank Dr Judy Tellam for assistance in preparation of the adenovirus constructs.

This work was supported by the National Health and Medical Research Council (NH&MRC), Queensland Cancer Fund (QCF), and the Australian Research Council (ARC). R.K. and S.R.B. are supported by NH&MRC Principal Research Fellowship and Senior Research Fellowship, respectively. J.R. is supported by an ARC Federation Fellowship. K.K.W. is supported by a scholarship from the QCF.

Contribution: K.K.W., Z.F., L.C., S.G., J.K.A., F.E.T., and J.J.M. performed the research; S.L.S., S.R.B., J.M., and J.R. analyzed the data and contributed toward the design of some of the experiments. R.K. designed the study and wrote the manuscript with J.R. and K.K.W.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Rajiv Khanna, QIMR, 300 Herston Rd, Herston (Qld) Australia 4006; e-mail: rajiv.khanna@qimr.edu.au; or Jamie Rossjohn, Department of Biochemistry, Monash University, Clayton, Australia 3800; e-mail: jamie.rossjohn@med.monash.edu.au.