Reproducible selection of high avidity CD8⁺ T-cell clones following secondary acute virus infection

Reproducible Changes in TCRβ Clonotype Frequency in the Recall Response.

To measure changes in CTL clone representation from primary memory to acute recall and to determine whether such alterations were a result of stochastic or deterministic mechanisms, a parallel adoptive transfer approach was used. A small sample of splenocytes from a mouse primed with PR8 influenza virus 6–8 wk earlier was characterized with respect to D^bNP₃₆₆-specific TRBV13-1⁺ CD8⁺ TCRβ clonotype use (donor memory), whereas the majority of the population was split three ways and transferred into sublethally irradiated recipient mice (Fig. 1A). Recipient mice were then challenged intranasally (i.n.) 24 h later with the serologically distinct (H3N2) HKx31 influenza virus that shares the identical PR8 nucleoprotein (NP). Profiles of TCRβ use were then characterized for individual CD45.1⁺ CD8⁺ TRBV13-1⁺ D^bNP₃₆₆-specific CTLs recovered from the spleen at the acute phase (day 8) of the CTL recall response (Fig. 1B).

The transition in D^bNP₃₆₆-specific clonal prevalence from the resting memory to the d8 recall response is shown for the first experiment in Fig. 1C. A noticeable shift in TRBV13-1⁺ CD8⁺ D^bNP₃₆₆-specific clonotype representation was observed between the transferred, memory population, and the responders recalled following secondary challenge (Fig. 1C). In particular, we observed an increase in prevalence of one TCRβ clonotype (SGGGNTGQL) in the recall response (Fig. 1C, black bars, and Table S1). Intriguingly, this was remarkably consistent for all three recipients, with the same T-cell clonotype being preferentially expanded despite its relatively low frequency in the adoptively transferred memory population (P = 4.21E−18, comparing memory to recipients). This pattern was highly reproducible, with six separate experiments yielding similar results (Table S1). In addition, the selective increase in clonotype prevalence was compensated by a reduction in the frequency of one or two other prominent T-cell clonotypes, a pattern that was observed for all three of the mice analyzed in this initial experiment (Fig. 1C, white bars, and Table S1). Overall, the consistency with which clonotypes were enriched or diminished in three independent recall responses strongly suggests a deterministic bias in the recall of CTL-mediated immunity.

Although the TCRβ clonotypes that consistently changed in prevalence following virus challenge were not identical for all six experiments, in three of six experiments, the SGGGNTGQL CDR3β clonotype was reproducibly increased in prevalence following virus challenge, whereas CTLs expressing SGGANTGQL decreased in four of six experiments (Table S1). Despite this trend, TCRβ clonotypes identified as notably increased or decreased in prevalence in one, or a few, experiments were also found to remain stable in some instances. This variability could reflect differential TCRα pairing, although it is also possible that the context of the repertoire (i.e., the suitability of a clone in the context of all other clones) determines which T cells are selectively expanded in the recall response.

T-Cell Selection into the Recall Response Is Based on the TCRαβ Phenotype.

Given that the TCR CDR3β analysis indicated preferential recruitment/expansion of particular clones in the recall response, we questioned the extent to which the TCRα chain contributed to this selection. Using a multiplex RT-PCR approach (15) allowing simultaneous characterization of TCRα and β transcripts from a single lymphocyte, we found that donor memory and recipient recall responses (CD8⁺TRBV13-1⁺) showed reproducible enrichment of a TCRα (RVSGGSNAKL) in each of the three mice following secondary virus challenge, relative to donor memory (P = 7.6E−05; Fig. 2A). Moreover, a different TCRα (RANSGTYQ), present at a relatively high frequency in the donor, decreased in prevalence in all three recipients (Fig. 2A; P = 1.3E−04).

Given that these separate analyses of CDR3α and CDR3β prominence in the recall response both show reproducible and similar patterns of clonotype enrichment or loss (Figs. 1C and 2A), it is likely that these predominantly represent CTLs expressing a single TCRαβ. As observed for the analysis of TCRα and β chains alone, characterization of the combined TCRαβ signatures showed a consistent (between the three recipients) increase (P = 3.24E−04) in the frequency of one TCRαβ (RVSGGSNAKL/SGGGNTGQL)-defined clonotype (Fig. 2B), together with a corresponding and reproducible decrease (P = 4.4E−04) in the prevalence of another (RANSGTYQ/SGGANTGQL), relative to donor memory. The patterns of clonotype selection were remarkably similar, irrespective of the chain being analyzed. Interestingly, some TCRβ or TCRα clonotypes identified as being selectively enriched or diminished in the recall response were identified in partnership with other TCRα or β chains, respectively, with no notable alteration in frequency (Fig. 2B). Collectively, these data indicate that enrichment of a particular clonotype in the recall response represents active selection via the combined TCRαβ heterodimer.

Homeostatic Proliferation Does Not Notably Alter Clonal Prevalence.

To exclude the possibility that the alterations in clonotype prevalence were due to homeostatic proliferation in the ∼24 h after memory cell transfer before infection, splenocytes from a long-term influenza virus–primed animal were adoptively transferred into three irradiated recipients (Fig. 1A), and the splenic D^bNP₃₆₆⁺ TRBV13-1⁺ CD45.1⁺ CD8⁺ T cells were analyzed 24 h later (Fig. S1A). Comparison of TCR CDR3β use between donor memory and pooled recipient recall sets showed that all populations were dominated by the SGGSNTGQL TCRβ clonotype, with a prevalence of 68.2% and 73.3%, respectively (Fig. S1B). Furthermore, the relative contribution of this CDR3β signature in the recipients showed no substantial or consistent (across three recipients) difference from the donor population (P = 0.67). Although there was some variability between mice and a few low-frequency clonotypes were observed only in individuals, this was likely due to the low numbers of D^bNP₃₆₆-specific cells that could be retrieved from recipients in the absence of a secondary virus challenge.

Collectively, these data indicate that the distribution of CTL clonotypes is not substantially or reproducibly impacted by homeostatic proliferation during the 24 h after memory cell transfer.

Route of Influenza Inoculation in Recipient Mice Does Not Drive Changes in Clonal Prevalence.

The experiments described in Fig. 1 were performed using the well-characterized i.p. prime (nonreplicative infection) and i.n. challenge (replicative infection) protocol (12, 13, 16). To ensure that the different route of virus exposure was not responsible for the TCR selection observed, we performed the same transfer experiment (Fig. 1A) by both priming and challenging mice with virus via the i.p. route (Fig. S2A). Single cell analysis of CDR3β clonotype use showed that, similar to previous experiments, the prevalence of the familiar (Figs. 1 and 2 and Table S1) SGGSNTGQL increased from 22.2% in donor memory to an average of 93.3% in the pooled recipients (P = 5.27E−17). Thus, the differential selection of TCR clonotypes from the memory pool into the recall response is not due to route of virus inoculation.

Clonotype Prevalence in the Recall Response Reflects TCR/pMHCI Avidity.

To address whether the change in size for particular CTL clones in the recall response was based on differential TCR binding characteristics, we transiently transfected 293T cells with retroviral constructs encoding the designated TCRαβ heterodimers along with a fluorescent reporter gene (GFP or AmCyan). These cells were stained for TCRβ and divided according to TCR expression levels (Fig. 3A). Analysis of each of these populations (TCR^{lo, int, hi}) showed that the TCR that was found to be enriched in the in vivo recall responses (RVSGGSNAKL/SGGGNTGQL) bound the D^bNP₃₆₆ tetramer more effectively than was the case for the diminished TCR (RANSGTYQ/SGGANTGQL) (Fig. 3B). Further, the enriched TCR had a substantially slower off-rate of pMHC binding (Fig. 3C). Collectively, considering the absence of accessory molecules on 293T cells, these data indicate that the prevalence of CTL clones in the recall response correlates directly with the intrinsic affinity of the TCR.

Partial Reversion of Clonal Prevalence from Recall to Secondary Memory.

We next investigated the fate of these reproducibly enriched or diminished clones following contraction into secondary memory. To address this, the adoptive transfer system described above was modified to analyze both the acute recall and secondary memory profiles. The three recipients were bled (day 8) for analysis of the acute response, and memory phase TCR use was analyzed later (days 60–80) from the spleen. As observed previously in the acute recall repertoire, there was a significant diminution in prevalence of the SGGANTGQL CDR3β clonotype from 45.9% in donor memory to an average of 6.3% in the recipients (P = 1.9E−09), and a significant increase (P = 6.8E−11) in the abundance of a SGGGNTGQL-expressing TCR clone from 0% in donor memory to an average of 33.7% in the recipients (Fig. 4, compare A and B). Interestingly, these TCRβ chains paired with different TCRα chains to those previously observed (Fig. 2), reflecting a greater diversity of TCRα, relative to TCRβ, use. Furthermore, the TCR that was enriched in the recall response (RGSYQGGRAL/SGGANTGQL) bound significantly more tetramer than the diminished TCR (RSAYQGGRAL/SGGANTGQL) after retroviral expression in 293T cells (as previously described), indicative of higher TCR affinity/avidity (Fig. S3A), although the rate of tetramer dissociation appeared similar for the two TCRs (Fig. S3B).

Analysis of the secondary memory pool showed that the RSAYQGGRAL/SGGANTGQL clone regained prevalence (average of 19.8%) relative to that found for the acute recall response (P = 0.01), whereas the RGSYQGGRAL/SGGGNTGQL clone lost prevalence (average of 27.04%), although the latter was not significant (Fig. 4, compare B and C). Thus, based on the profiles for these two clonotypes, the secondary memory repertoire appears, in part, to revert toward the primary memory distribution, suggesting that preferential expansion is matched by enhanced contraction of these clones in secondary memory. This finding was further demonstrated by calculation of the prevalence of the RSAYQGGRAL/SGGANTGQL (diminished) and RGSYQGGRAL/SGGGNTGQL (enriched) clones in the total CD8⁺ T-cell pool in acute recall compared with secondary memory. The enriched clone showed a notably reduced prevalence (2.4-fold) in secondary memory, whereas the diminished clone showed a slightly elevated prevalence (1.7-fold). Given that all clones would be contracting to some extent and none would undergo a net expansion from acute recall to secondary memory, these data therefore indicate that the RGSYQGGRAL/SGGGNTGQL clone undergoes more pronounced contraction than the more stable RSAYQGGRAL/SGGANTGQL clone. This experiment was repeated with similar results (Fig. S4).

CTL Clonal Diversity and Distribution Across the Stages of the Antiviral Response.

Prior analyses of CTL responses demonstrated a focusing (or narrowing) of TCR β repertoire use with successive antigen encounters (17–19). Given that we observed reproducible shifts in clonal prevalence from memory to the recall response that (in secondary memory) appeared to trend back toward that observed following initial priming, we investigated the extent to which such changes in the repertoire impacted its overall diversity. Fig. 5 shows Simpson’s diversity index (SDI; a function of clone number and distribution) values (20–22) for the TCRβ repertoires analyzed at the indicated stages, with related repertoires connected. Of the 20 repertoires analyzed in the acute recall response (those shown in Table S1 and three recipients from experiment in Fig. 4), only 11 showed a narrowing in diversity, with 2 remaining unchanged, and in 7 populations TCRβ diversity increased relative to the donor memory population (Fig. 5). Overall, there was no statistically significant change in clonal diversity (P = 0.45). Similar results were obtained when analysis was restricted to evenness of clonal distribution using the Gini coefficient (21, 23) (Fig. S5). Intriguingly, in all five of the repertoires that were analyzed in secondary memory, there was a shift toward increased diversity (and increased evenness), although the sample size was too low to provide statistical significance at a P < 0.05 level with a two-sided nonparametric test. Nevertheless, the data suggest that, where selective clonal expansion in the recall response results in narrowing of TCR diversity, that diversity is regained following contraction into secondary memory. In summary, we find no evidence of a global narrowing of the virus-specific CTL repertoire subsequent to secondary virus infection, either acutely or in the secondary memory phase.

Mice and Virus Infections.

Female CD45.1⁺ C57BL/6J (B6; H-2^b) and CD45.2⁺ B6 mice were bred and housed in specific pathogen-free conditions in the animal facility of the Department of Microbiology and Immunology, University of Melbourne (Parkville, VIC, Australia). Donor mice were inoculated i.p. 6–8 wk previously with 1.5 × 10⁷ plaque forming units (PFUs) of the A/Puerto Rico/8/34 (PR8, H1N1) influenza virus. Recipient mice were sublethally irradiated with 550 rad before adoptive transfer and, following transfer, were either inoculated i.p. with 1.5 × 10⁷ PFU of the PR8 influenza virus or infected i.n. with 1 × 10⁴ PFU of the A/Hong Kong/X31 (X31, H3N2) influenza virus. All experimental procedures were reviewed and approved by the University of Melbourne Animal Experimentation Ethics Committee.

Adoptive Transfer.

Splenocytes (typically 2–3 × 10⁷ cells per mouse) were transferred via i.v. injection into sublethally irradiated CD45.2⁺ B6 recipient mice. Recipient mice were infected i.n. with 1 × 10⁴ PFUs or i.p. with 1.5 × 10⁷ PFU of ×31 influenza virus 24 h after cell transfer and killed 8 d thereafter for analysis of the acute recall response from splenocytes. In some instances, recipient mice were killed 24 h after cell transfer for splenic repertoire analysis.

Isolation of Individual CD8⁺ T Cells by Sorting.

Enriched splenocytes or peripheral blood lymphocytes (PBLs; SI Materials and Methods) were stained with phycoerythrin (PE) or allophycocyanin (APC)-conjugated tetrameric complexes of the influenza virus H-2D^b MHC class I glycoprotein and NP_366–374 (ASNENMETM) peptide for 1 h at room temperature. Cells were then stained with various combinations of conjugated antibodies specific for Vβ8.3, CD8α (53-6.7), and CD45.1-PE (A20) (BD Pharmingen and BioLegend). Individual D^bNP₃₆₆⁺ TRBV13-1(Vβ8.3)⁺ CD8⁺ T cells or D^bNP₃₆₆⁺ TRBV13-1⁺ CD45.1⁺ CD8⁺ T cells were sorted using a BD FACSAria (BD Biosciences).

RT-PCR and Sequencing.

Single-cell RT and PCR amplification of TRBV13-1⁺ TCR CDR3β regions was performed as previously described (12). Multiplex single-cell RT and PCR amplification of TCR CDR3α and CDR3β regions was performed for amplification of cells expressing TRBV13-1 and unknown TCR Vα (TRAV), using TRBV13-1–specific oligonucleotides and a panel of TRAV specific oligonucleotides, as previously described (15). PCR products were purified with ExoSAP-IT reagent (USB Corporation) and sequenced with corresponding TRAC or TRBC internal primers (12, 15).

Statistics.

Statistics were performed as described in SI Materials and Methods.

Transfection of 293T Cells and Analysis of TCR Binding Characteristics.

293T cells were transfected with a mouse stem cell virus (MSCV)-based retroviral vector containing αβTCR sequence + internal ribosome entry site (IRES)-GFP (pMIG) or IRES-AmCyan (pMIC), as well as pMIG encoding the CD3γδε and ζ subunits, according to ref. 42. Transfected 293T cells were labeled with D^bNP₃₆₆-PE tetramer 48 h later, followed either by conjugated antibodies specific for CD8α and TCRβ or incubated for designated times at 37 °C with anti-H-2D^b antibody (28-14-8; BD PharMingen) (50 μg/mL) and then with CD8α Ab (43).