|Home | About | Journals | Submit | Contact Us | Français|
Detailed assessment of how the structural properties of T cell receptors affect clonal repertoires of antigen-specific cells is a prerequisite for a better understanding of human anti-viral immunity. Here we examined the α-TCR repertoires of CD8 T cells reactive against the influenza A viral epitope, M158-66, restricted by HLA-A2.1. Using molecular cloning, we systematically studied the impact of α-chain usage in the formation of T cell memory and revealed that M158-66-specific, clonally diverse VB19 T cells express α-chains encoded by multiple AV-genes with different CDR3 sizes. A unique feature of these α-TCRs was the presence of CDR3 fitting to an AGA(Gn)GG-like amino acid motif. This pattern was consistent over time and among different individuals. Further molecular assessment of human CD4+8− and CD4−CD8+ thymocytes led to the conclusion that the poly-Gly/Ala runs in CDR3α were a property of immune, but not naïve repertoires and could be attributed to influenza exposure. Repertoires of T cell memory are discussed in the context of clonal diversity, where poly-Gly/Ala runs in the CDR3 of α- and β-chains might provide high levels of TCR flexibility during antigen recognition while gene-encoded CDR1 and CDR2 contribute to the fine specificity of the TCR-pepMHC interaction.
CD8 T cells express αβ-T cell receptors (TCRs) that bind to immunogenic peptides loaded into class I MHC molecules (pMHC) and initiate formation of the supramolecular activation clusters between T cells and antigen-presenting cell (1-4). The T cells that have an identical clonal origin express a unique αβ-TCR that defines clonal fine specificity to antigen. Multiple clones with diverse, and to some extent overlapping, specificities provide protective immunity against viral infections. After viral clearance, a number of epitope-specific clones are retained, thus creating long-lasting memory TCR repertoires. It is currently agreed that clonal survival during and after viral clearance is a final result of multiple factors. Among these factors is the molecular nature of αβ-TCRs.
Antigen-driven clonally expressed β-TCR repertoires have been intensively examined in experimental animal models (5-7) and human diseases (8-11). These studies have concluded that repetitive antigenic challenges correlate with increased frequencies of antigen-specific, clonally diverse cells that share amino acid sequences within complementary-determining regions 3 (CDR3) of their expressed β-TCRs “fitting best” to epitope recognition. Since these memory cells are at high precursor frequencies and usually have lower TCR-mediated activation requirements than naïve cells, they provide rapid pathogen clearance in the case of re-infection. Although β-TCR-mediated selections in response to the immunogenic epitopes are well documented, little is known about α-TCR involvement in selection of human CD8 T cells.
Crystallization of TCR-pMHC has revealed that α-chains might provide a significant contribution to the interactive interface, varying from 37% to 74% of the total surface (12-16). This implies that α-TCR usage might be a critical element that defines whether Ag-reactive cells are saved in a memory compartment. In this study, we sought to investigate α–TCRs expressed by memory cells, and observed several previously unknown properties that might determine the clonal nature of memory repertoires.
Human CD8 T cell reactivity against the influenza A matrix M1 protein-derived epitope, M158-66, represents an exceptional system to understand the molecular properties of α-TCRs expressed by memory cells selected in humans. Because the M1 protein is highly conserved among influenza A viral strains, re-infections during a lifetime (17) result in formation of the strong CTL recall responses against the M158-66-epitope practically in all HLA-A2 (HLA-A*0201) individuals (10, 11, 18-20). For instance, by age 15 years HLA-A2 children possess a well-established M1-specific memory pool comprised of CD8 T cells expressing BV19 gene-encoded β-chains (formerly, BV17) (20, 21).
Our previous studies revealed that multiple VB19 clones specific to M158-66 co-exist in middle-aged individuals (22-24). Those clones were defined based on the uniqueness of the nucleotide composition in the V-NDN-J regions encoding β-chains. Therefore, they were referred to as VB19 clonotypes since the α-TCR usage remained unknown. These flu-specific clonotypes utilize BV19 gene-encoded β-chains with two CDR3 sizes fitting into IRSS- and IGS-like motifs. Although individual VB19 CD8 T clonotypes expressed structurally identical β-TCRs, they have different M158-66:HLA-A2.1 tetramer (M1-tetramer) binding capacity and peptide-concentration-dependent proliferation in cell cultures (23). This suggests that even if VB19 cells were selected due to the best CDR3β “fit” to M158-66:HLA-A2 recognition, their α-TCR usage could be different. Therefore, we reasoned that memory cells from a single family, VB19, could be used to examine the breath of the memory α-TCR repertoire and provide insight on why β-VB19 chains alone are not exclusive determinants in clonal selection (10, 23).
It has been reported that the optimal αβ-TCR interaction with pMHC requires that CDR3α and CDR3β should have similar sizes for proper engagement of Vα- and Vβ-domains (25). Our present study demonstrates that VB19 cells express AJ42 gene-encoded α-TCRs which contain multiple, up to five, poly-Gly/Ala runs in the long CDR3α allowing engagement of cells from different VA families, thus utilizing different CDR1α and CDR2β in epitope recognition. This observation occurred in five individuals whose M1 -specific cells were selected using M1-tetramer and VB19 mAbs. During the study period, we did not find pre-selection for poly-Gly/Ala runs in the CDR3α in human CD4−8+ thymocytes from the T cell subset (VA27-JA42) that was most prominent in response to flu-M158-66 epitope. Taken together, our study led to the conclusion that the selection of T cells possessing poly-Gly/Ala runs within their CDR3 was driven in response to the influenza A M158-66 epitope rather than by biased gene recombination or thymic selection. We propose that the presence of poly-Gly/Ala runs in the CDR3 of α- and β-TCRs contributes to conformational flexibility of antigen-specific receptors. This implies that a robust immune response associates not only with T cells whose CDR3α and CDR3β, generated during random gene recommendations, have the “fittest” complementarity to pMHC, but also with T cells whose antigen receptors utilize germline gene-encoded regions if CDR3s have high levels of flexibility.
Five healthy blood donors: Donors A, B, C, D and E (50, 47, 40, 56, and 26 years old, respectively) were defined to be HLA-A2.1 (HLA-A*0201) positive based on MHC class I typing with the Biotest SSP System (Biotest Diagnostics, Nashville, TN).
Thymic tissue was collected as a discard during reconstructive surgical procedure on an HLA-A2.1 positive 3-months old child having a congenital cardiac defect, under a protocol approved by the Internal Review Board of The Children's Hospital of Wisconsin.
CD8 T cells were isolated from peripheral blood collected from donors A-E using anti-CD8 microbeads (Miltenyi Biotech, Auburn, CA), following manufacturer recommendations. The purity of isolated CD8 T cells usually exceeded 95%. To generate peptide-specific cell lines, CD8 T cells (0.25 × 106 cells/ml) were co-cultured with TAP1/TAP2-defective target T2 (174 × CEM.T2, ATCC) cells (0.05 × 106 cells/ml) in 4 ml complete culture media/well. The T2 cells were incubated overnight with M158-66 peptide (1 × 10−6 M). Prior to setting of M1-specific cultures, peptide-pulsed T2 cells were irradiated (3000 Rad) and intensively washed to avoid residual peptide. The complete culture AIM-V (Gibco, Invitrogen, Carlsbad, CA) media contained human rIL-2 (10 U/ml) (BD Pharmingen, San Diego, CA) and was supplemented with 14% supernatant from the IL-2-producing MLA 144 cell line (TIB 201, ATCC). The CD8 T cell lines were cultured in 12 well plates, and split with rIL-2-containing media every three days. Once a week cultures were re-stimulated with peptide-coated, irradiated T2 cells. The cultures where T2 cells were not coated with peptide served as controls.
The peptide M158-66 (GILGFVFTL) was synthesized on Pepsyn KA resin (BioSource, Hopkinton, MA) using a 9050 Pepsynthesizer (Millipore Corp.). Peptide was purified by reverse phase HPLC (> 90% purity) using a C18 column (Vydac, Hesperia, CA).
To examine the frequency of M158-66-specific VB19 cells, CD8 T cells were sampled from the cultured lines and co-stained with M158-66/HLA-A2.1 tetramer (M1-tetramer) (Beckman Coulter Inc., Fullerton, CA) and VB19 family-specific mAbs (Immunotech, Marseille, France) according to manufacturer recommendations. Initially cells were stained with APC-labeled M1-tetramer for 20 min at room temperature, and then FITC-labeled VB19 mAbs were added for an additional 20 min under the same conditions. The stained cells were washed in a copious volume of FACS buffer (PBS; 2% FCS; 0.2% sodium-azide). Flow cytometry was performed using a FACSVantage and FACScaliber (BD Bioscience, San Jose, CA). To isolate M1-specific cells, CD8 T cell line samples were stained with M1-tetramer and VB19 mAbs (Donors A, B and C), or M1-tetramer alone (Donor D), and then FACS-sorted using a MoFlo Sorter (DakoCytomation, Collins, CO). Purity of the FACS-sorted cells exceeded 95% (data not shown).
The thymus was disaggregated by passing through a wire mesh. Cells were suspended in RPMI medium (Life Technologies, Gaithersburg, MD), 0.1% sodium azide, and 2% FCS, and stained with mouse mAbs to human cell surface markers: CD3-FITC conjugate, CD4-Tri-color conjugate, and CD8-R-PE conjugate (Caltag, San Francisco, CA). A three-color sort was performed using FACStar (Becton Dickinson), and single-positive CD4+8− and CD4−8+ thymocytes were collected. Primary gating was set on the CD3 marker, which resolved the thymocytes into three populations, CD3neg, CD3low, and CD3high. The CD3high population was further divided on the basis of CD4 and CD8 expression. Cells were collected into 0.5 ml of FCS so that the final concentration in the tube was 10% (5 ml final volume).
CD8 T cells separated from peripheral blood, cultured lines, or FACS-sorting after M1-tetramer/VB19 mAbs staining, were used for RNA isolation and cDNA synthesis as previously described (22, 26). Genomic DNA was prepared from FACS-sorted thymocytes treated with nucleic acid lysis buffer, pH 8.2 (10 mM Tris, 0.4 M NaCl, and 2 mM EDTA), in the presence of SDS and proteinase (26). Then the cells were incubated overnight at 45°C to ensure the complete lysis. After the incubation, proteins were precipitated by adding 5.3 M NaCl, and DNA was precipitated from the supernatant with ethanol (27). Before generation of α-TCR specific CDR3 spectratypes, genomic DNA and cDNA samples were titrated using semi-quantitative PCR amplification of TCR-CB region, as described (26, 28). Briefly, serially diluted cDNA (or, genomic DNA) aliquots were amplified 24 cycles using forward and reverse CB-specific primers under non-saturating PCR conditions, and resolved on denaturing 50% Urea/ 5% polyacrylamide gels. The cDNA (or genomic DNA) aliquots of cDNA that contained an equal quantity of total β-TCR transcripts were used for sequential CDR3α spectratype-generating PCR amplifications.
Detailed descriptions of CDR3 spectratyping conditions were published previously (22, 26). Briefly, cDNA samples were amplified in a PCR cocktail that contained forward VA-family specific primer and reverse CA-specific primer labeled from 5′-end with carboxyfluorescein (6-FAM). The nucleotide sequences of thirty-four VA-family specific and CA-specific primers were described elsewhere (29). To examine the α-TCR repertoires of human thymocytes, genomic DNA samples were amplified using VA27- and JA42-specific primers. The PCR cycle consisted of 0.5 min at 94°C for denaturation, 0.5 min at 58°C for annealing and 1.5 min at 72°C for elongation. The final elongation was extended to an additional 7 min at 72°C. After 31 cycles of PCR amplification, 10 μl of VA-CA amplified cDNA products were loaded and run on 50% urea / 5% polyacrylamide sequencing gels for 2 hours 15 min to 3 hours 45 min depending on VA-CA primer combination. This technique ensures CDR3α bands visualization after gel scanning on the fluorescence detection system FluorImager 595 (Molecular Dynamics, Sunnyvale, CA). Multiple CDR3α bands within given VA family reflect clonal complexity based on α-TCR transcript size heterogeneity.
cDNA samples generated from M1-tetramer/VB19 mAbs-stained, FACS-sorted CD8 T cells (Donors A-C), M1-tetramer stained (Donor D), and non-sorted cultured cells (Donor E) were amplified with unlabeled CA-specific primer and one of VA-family specific primers in separate non-saturated PCR reactions and immediately subcloned in to the plasmid vector pCR4-TOPO (Invitrogen, Carlsbad, CA). The following cDNA libraries were generated and screened: VA8 (includes VA8.1 and VA8.3), VA8.6, VA10, VA12 (includes VA12.1, VA12.2 and VA12.3), VA27, VA29, VA34 and VA35. VB19-CDR3β plasmid subclones of Donor A cultures were described elsewhere (23). From 48 to 96 plasmid subclones from each VA cDNA library have been sequenced using a Taq DyeDeoxy Terminator cycle sequencing kit (Applied Biosystems, Foster City, CA). Analysis of VA and JA regions flanking CDR3α nucleotide sequences indicated <0.25% divergence from the genomic sequences, which can be attributed to the MMLV Reverse Transcriptase and/or Taq DNA Polymerase infidelity. The assignment to AV and AJ gene families of the subcloned and sequenced CDR3α plasmid inserts is given according to IMDT. CDR3α size count includes C (from CASS) and F (FGXG) according to IMDT [ImMunoGenDeneTic]).
The VA and JA family origins of the influenza A M158-66-specific VB19 clonotypes from Donors A to E are shown in Table 1, Supplemental. Here we depict each clonotype number within VA cDNA libraries, their CDR3α sizes (aa), amino acid (highlighted in red) and nucleotide sequences within AV –JA gene junction. The data sets corresponding to VA27-JA42 repertoires of the CD4−8+ and CD4+8− thymocytes and CD8 T cells from M158-66-specific cell culture (Donor A, year 2004, 5 week culture) are given in Table II, Supplemental. The CDR3α loops are shown in red, non-template encoded amino acids in CDR3α are revealed in blue and underlined. Clonotype unique identifiers are labeled as ID.
To define the breadth of the α-TCR repertoire of flu-specific cells that express VB19 β-chains with high affinity to M158-66 epitope (19), we used CDR3α spectratyping. Lack of available VA family-specific mAbs and the relatively low precursor frequency of M158-66-specific cells varying in a range of 0.1-0.8% in the peripheral blood CD8 T cells (30) were the two major reasons for using cell lines. In order to generate a sufficient number of cells to screen α-TCR families responding to the flu-epitope, the CD8 T cells collected from five middle-aged HLA-A2 individuals were stimulated with irradiated T2 cells charged with the M158-66 peptide to induce T cell division in vitro. After three to five weeks in culture, cells were either FACS-sorted after co-staining with M1-tetramer and VB19 mAbs (Donor A-C), M1-tetramer alone (Donor D), or used directly (Donors E). In our experiments, the purity of FACS-sorted cells gated as M1-tetramer+ or M1-tetramer+/VB19+ populations usually exceeded 95% (data not shown).
We reasoned that by using RT-PCR amplification of cDNA samples with AV-gene specific primers, we would define α-TCR transcriptional profiles of the cells proliferating upon peptide stimulation and binding M1-tetramer. Therefore, cDNA samples, generated from non-stimulated CD8 T cells, FACS-sorted with M1-tetramer alone and M1-tetramer/VB19 mAbs combined, were amplified with VA-specific primers specific to 29 out of 52 human α-TCR genes (29). To examine the α-TCR repertoires of the peripheral blood CD8 T cells and cells FACS-sorted from the epitope-specific cultures, we run off the RT-PCR products on the CDR3α spectratyping gels assuming the M158-66-specific cells will be selected within VA family(ies). The increased fraction of M158-66-specific cells within each VA family was ascertained by increased densities of the selected CDR3a sizes detected on the gel images. To provide representative examples of the α-TCR repertoires of the unstimulated CD8 T cells and cells from the M158-66-specific cultures, we present CDR3α spectratypes from Donor A (Figure 1), whose α-TCR repertoire of peripheral blood CD8 T cells is shown in Figure 1A. The α-TCR transcriptional profiles of Donor A M1-tetramer+ (FACS-sorted, week 4 culture) and M1-tetramer+/VB19+ (FACS-sorted, week 5 culture) cells across and within VA families are shown in Figure 1B and C, respectively. The numbers on the X-axis indicate the VA family origin of α-TCR transcripts, while DNA bands resolved on the Y-axis and their intensities represent proportions of α-TCR transcripts with identical CDR3 sizes within the respective families. Expecting that a repertoire of the peripheral blood CD8 T cells is extremely diverse, we observed multiple CDR3α bands within each of VA families examined. While only 56% of the VA repertoire (29/52 families) was examined, we observed that CD8 T cells were derived from different VA families and utilized different CDR3α sizes. Representation of each VA family was skewed, with a high yield of VA24-, VA30- and VA12 -specific transcripts and a low yield of the VA14-, VA16-, VA40-specific transcripts (Fig. 1A). Interestingly, CDR3α spectratypes of the freshly isolated cells from Donors B-E have similar patterns to Donor A, namely: complexity of VA families, multiple CDR3a size usage within each of VA families and variation of these values among individuals (data not shown).
To gain insight about the M158-66-specific α-TCR repertoire, we ran off RT-PCR products from M1-tetramer+ CD8 T cells (FACS-sorted, week 4 culture, Donor A) as shown in Figure 1B. The M1-tetramer+ cells expressed α-TCRs encoded by multiple VA families (VA8, 10, 14, 19 - 24, 12, 27, 8.6, 22, 38.2, 41 and 29). Remarkably, M158-66-specific cells from a few families utilized α-TCRs with different CDR3α sizes, as shown on VA12, VA27, VA8.6, and VA29 spectratypes (Figure 1B, brackets). To link the mentioned α-VA transcripts to VB19+ cells, we sorted cells co-stained with M1-tetramer and VB19 mAbs. A representative example of α-TCR repertoire of the double-positive cells from Donor A is shown in Figure 1C. Again, we observed that M158-66-specific VB19+ cells co-expressed α-TCRs encoded by different VA families with different CDR3α sizes (Figure 1C). Interestingly, CDR3α spectratypes of M1-tetramer+ cells (Fig. 1B) were similar to those generated from M1-tetramer+ VB19+ subpopulation (Fig. 1C). We showed that RT-PCR products from M1-tetramer+ and M1-tetramer+/VB19+ cells were derived from VA8, VA10, VA12, VA27, VA8.6, VA22 and VA29 families. Though we did not align CDR3α bands within each VA spectratype, this observation reflects a complexity of the M158-66-specific recall α-TCR repertoire with engagement of different VA-domains in pMHC recognition.
This observation is evidence that flu-specific VB19+cells utilized α-TCRs from different families, and the diverse CDR3α sizes further confirm our previous observations that VB19+ cells have different clonal origins and avidities to the M158-66-epitope (22-24). Although we did not show α-TCR repertoires from Donors B-E whose CD8 T cells were isolated from the peripheral blood and cell cultures, there were multiple α-TCR transcripts from M158-66-specific cells whether they were, or they were not, M1-tetramer/VB19 mAbs positive. Of note, since HLA-A2.1 T2 cells were used as APCs in our cell cultures, their α-TCR transcripts could be misinterpreted as specific to M158-66 reactivity, and affect the general picture of the recall α-TCR repertoires. Therefore, we screened α-TCR transcripts from T2-cells and defined that neither one of the VA-specific primers used amplified α-TCRs. We concluded that α-TCRs from 4 to 5-week cultures were derived from M158-66-specific CD8 T cells.
Since allelic exclusion is not applicable to α-TCR gene rearrangement, the peripheral T cells express two α-chain mRNAs (derived from both chromosomes), and 25-30% of cells express two α-TCR proteins paired with a single β-TCR where only one αβ-TCR heterodimer binds self-MHC (31-33). Considering the previous report where M1-specific VB19 clones were found to express primarily α-VA27 (former, VA10.2) (11), it was somewhat surprising that VA27 α-TCR mRNA/cDNA transcripts did not dominate within the α-TCR transcriptional pools (Fig. 1B and C). We reasoned that if the recall response was mediated by numerous VB19+ clonotypes then detection of multiple VA transcripts would be expected. However, if clonal cells were selected due to the molecular nature of their α-TCRs, then similarity in α-TCR repertoires would be revealed even if they were collected from different individuals. Thus, we tested α-TCR repertoires from four HLA-A2 blood donors. We thought that if only the VA27 family was involved in M158-66-recognition, then the probability that flu-specific cells from different donors would have identical second α-TCRs (same VA families and CDR3α sizes) would be extremely low. However, if cells were selected based on α-TCR fitness to flu-M158-66:HLA-A2, then they would have similar CDR3α sizes and amino acid sequences of α-TCRs.
First, we re-examined whether flu-M1 reactive cells consistently utilized CDR3α of different sizes. Peptide-specific cultures from Donor A were re-generated 12 months apart and re-examined regarding α-TCR usage. The representative CDR3α spectratypes of VA27 and VA8.6 cells isolated from blood, mock and M158-66-specific cultures are shown in Figure 2. We found that immunodominant CDR3α bands had identical sizes in all CD8 T cell lines and properly aligned with those from M1-tetramer+ sorted cells (Figure 2, arrows). Based on spectratyping patterns, we concluded that VA27+ cells evenly utilize three different CDR3α sizes, while VA8.6 cells express mostly long CDR3α.
To further verify that flu-M1 reactive cells derived from different VA families express CDR3α of different sizes, we generated and screened VA spectratypes of the cultured cells from Donors B-E. The representative VA27 and AV8.6 spectratypes of M1-tetramer+ VB19+ (Donor B), M1-tetramer+ (Donor D) and non-sorted cultured (Donor E) CD8 T cells, including controls, are shown in Figure 3. After M1-tetramer+ cells were sorted from Donors' B and C cultures, we intentionally gated on M1-tetramerhigh VB19high T cell populations. Again, we observed that flu-specific cells expressed AV27, 10, 8.6, 12, 29 gene-encoded α-TCR transcripts (not shown).
From these findings, we concluded that frequencies of VA27 and other families were low in peripheral blood and control cultures, since we used equalized quantities of RNA/cDNA transcripts from mock and peptide-specific lines. However, stimulation with influenza derived M158-66 peptide rapidly induced peptide-dependent proliferation of T cells derived from diverse VA families. Although VA27 spectratypes with three CDR3α bands were remarkably similar between donors, we observed that patterns of VA8.6 and other VA usage varied between individuals.
To better understand how influenza specific cells might utilize α-chains encoded by different AV genes and having different CDR3α sizes, we used cDNA samples from M1-tetramer+ VB19+, M1-tetramer+ and bulk M158-66-specific cultured CD8 T cells to generate VA-family specific cDNA libraries and sequenced CDR3 inserted into plasmid vector. It should be pointed out that our methodology does not allow assigning TCR α- and β-chains to a single clone; however, we were confident that utilized cells were influenza M158-66-specific given that CDR3β spectratypings were over-represented by VB19 family ((23), and not shown). Hereafter CDR3α sequences are referred to as “clonotypes” based on the uniqueness of nucleotide composition in the AV-N-AJ gene recombination sites. We thought that the number of CDR3α sequences within VA-cDNA library might serve as a proxy of the clonotypes' relative frequencies within the respective VA family, if the α-TCR repertoires possess a polyclonal nature. This also could validate M158-66-driven clonotype selection mediated through CDR3α amino acid compositions. The complete data sets of VA and JA usage, CDR3α sizes and amino acid sequences of α-TCRs expressed by M158-66-reactive clonotypes from all five donors are available as the Supplemental Material (Table I, Supplemental).
We have demonstrated that the α-TCR repertoire reactive against influenza was complex and included multiple VB19+ clonotypes from multiple VA families with different CDR3α sizes. Hereafter, we examined repertoire structure where clonotypes from JA42 family were considered as de facto M158-66-specific. Thus, we tabulated the number of clonotypes that share CDR3α sizes within VA families and determined proportions of the JA42+ clonotypes across VA families and across CDR3α sizes. In Figure 4 we illustrated the clonotype distribution by VA families, by CDR3α sizes (from 18 to 10 aa) for each donor (Fig. 4A) and overall (Fig. 4B) in color-coded manner (Fig. 4C). The color-coding system represents the number of clonotypes classified into five categories (shown in rows) and the proportions of corresponding JA42+ clonotypes also classified into four categories (shown in columns in three degrees of shading). We also provided the summaries for the total number of all VA clonotypes, JA42+ clonotypes, and their contribution (%) in all VA-cDNA libraries by each CDR3α size and overall for each donor (Fig. 4A) and for all five donors (Fig. 4B). For example, Donor A had 105 clonotypes, out of which 67 (64%) clonotypes were JA42+; the dominant CDR3α size was 15 aa, found in 40 clonotypes, of which 33 (83%) were JA42+ (shown in bold).
We observed that the total number of detected clonotypes varied from 5 to 105, and JA42+ clonotypes represented from 23% to 64% of all α-TCR repertoires. Among all donors, the distribution of clonotypes reflects that M158-66-specific T cells express CDR3α of different sizes; however, CDR3α with 12, 14 and 15 amino acids were dominant (Fig. 4A). The dominance of specific VA families and specific CDR3α sizes was well pronounced in the overall summary (Fig. 4B). We found that all VA8.1+ clonotypes, 67% of VA27+ clonotypes, and VA12.3+ clonotypes expressed JA42-encoded α-TCRs. The JA42 usage associated with CDR3α sizes of 12 amino acids (aa) (80%), 15 aa (59%), and 14 and 16 aa (36% and 33%, respectively).
The high utilization of JA42+ clonotypes with different CDR3α sizes from different VA families (on average 53%) upon epitope stimulation suggested that expression of VA27 α-chains by VB19+ CD8 T cells is not an exclusive requirement to M158-66-specific reactivity. We also determined the following hierarchy of CDR3α sizes as a function of VA family origin, such as: 15 > 12 > 14 amino acid residues. Moreover, if the VB19+ clonotypes from the JA42 family are considered as M158-66 -specific, then the hierarchy of VA usage can be presented as the following rule: VA27 (three CDR3α sizes and JA42) > VA8.6 and VA35 (two CDR3α sizes and JA42) > VA8.1 to VA29 (only CDR3α of 15 a.a. and JA42).
To investigate whether proper combination of CDR3α sizes and amino acid sequences is a strict requirement in M158-66 recognition, while CDR1α and CDR2α usage is less stringent, we aligned α-chains of JA42+ clonotypes originating from different VA families from all five donors (Table II) involved in this study and Donor A (Table III) alone. Strikingly, many clonotypes contained poly-Gly/Ala runs where only two Gly were AJ42-gene encoded (Table III, underlined red characters). Remarkably, the M158-66 -specific clonotypes expressing long CDR3α chains (14 and 15 a.a.) contained non-template encoded Gly/Ala runs (Table III, underlined blue characters) with other non-polar, polar and charged amino acids. Importantly, although α-TCRs might have identical amino acid sequences in CDR3, they were encoded by different nucleotide sequences generated during AV-N-AJ42 gene recombination that serves as clonotype marker (available as Table I, Supplemental). Based on the collected and analyzed data sets, we defined the following rules in CDR3α usage: - AGAGGGG in CDR3α with 15 amino acids; AGAGGG in CDR3α with 14 amino acids; and AGGG in CDR3α with 12 amino acid residues. Although we generated and screened the VA-cDNA libraries of different sizes for each individual (Fig. 4A), this pattern was consistent between the studied subjects (Tables III).
The presence of non-template poly-Gly/Ala runs in TCR α-chains expressed by flu-specific T cells could either reflect the CDR3α amino acid sequence distribution in naïve T cells emerging from the thymus, or it could reflect a preferential selection of these sequences during influenza specific immune responses. To examine whether poly-Gly/Ala runs are a property of M158-66- specific memory, we examined AV27-N-AJ42 gene recombination in human CD4+8− and CD4−8+ thymocytes. We reasoned that if poly-Gly/Ala is an intrinsic property of the long CDR3α, then this motif could be defined in CD4+8− (4SP) and CD4−8+ (8SP) thymocytes. Therefore, thymic tissue was used for 4SP and 8SP thymocyte isolation, genomic DNA preparation and PCR amplification, using VA27- and JA42-family specific primers. The flow cytometry data and corresponding VA27-JA42 spectratypes are shown in Figure 5. In addition to thymic spectratypes, we used spectratypes of the M158-66-specific culture and sorted M1-tetarmer+ VB19+ T cells that served as the positive controls for CDR3α sizes.
As shown in Figure 5B, the proportions of 4SP and 8SP thymocytes that have CDR3α composed of 12 aa residues were underrepresented within the VA27+ JA42+ population, yet they were abundant in bulk epitope-specific culture and within M1-tetramer+ VB19+ populations. To further examine the presence of poly-Gly/Ala runs in conjunction with CDR3α sizes we cloned and sequenced 400 and 300 CDR3α plasmid subclones from the 8SP and 4SP thymocytes, respectively. Here, we defined 129 “in-frame” re-arrangements for 8SP thymocytes and 89 “in-frame” re-arrangements for 4SP thymocytes. This frequency was expected since 2-out-of-3 rearrangements generate non-productive CDR3α. As a control, we also used VA27-JA42 subclones from bulk M158-66–specific culture (Donor A, year 2004, 5-week culture) and M1-tetramer+VB19+ sorted populations (Donor A, year 2002, 4-week culture). The α-TCR repertoire profiles of the VA27-JA42+ clonotypes from the 8SP and 4SP thymocytes and cultured CD8 T cells are available as the Supplemental Material (Table II, Supplemental).
Although 8SP thymocytes utilizing CDR3α of 14 and 15 aa residues were dominant based on spectratyping results (Fig. 5B), more clonotype diversity was observed among cells using CDR3α of 12 aa residues (20%, 13/66 clonotypes) and 15 aa residues (31%, 21/66 clonotypes) (Table II Supplemental, “CD4−8+ thymocytes”). The similar weak association between CDR3α band intensity on spectratyping and clonotype diversity was seen for 4SP thymocytes (Table II Supplemental, “CD4+8− thymocytes”). Only one dominant clonotype detected among 8SP thymocytes represented 11% (14/129 sequences) of the population and used short CDR3α (11 aa) created by bland AV27 and AJ42 gene ligation.
Since we were interested to examine whether TCR interaction with HLA-A2 would preferentially select cells based on CDR3α sizes with non-template encoded Gly/Ala runs, we examined the occurrence of poly-Gly/Ala runs by plotting frequencies of these amino acids, for all clonotypes, as a function of the CDR3α sizes (Fig. 6). The Gly/Ala runs could be encoded if linking AV27 gene (CAG) with AJ42 gene segment (GGSQG…) occurs after deletion of one nucleotide from the 3′-end of AV27 gene and 7 nucleotides from the 5′-end of AJ42 gene. Therefore, we expected to observe Gly and Ala strings in a short CDR3α. However, we were interested whether the long, 14-15 aa, CDR3α were enriched by non-template Gly/Ala. As shown in Figure 6A, the frequency of non-template Gly/Ala in 8SP thymocytes with 14-15 amino acids in CDR3α was below detection level, considering the number of identified clonotypes, similar to what was observed for 4SP (0-5%) clonotypes (Fig. 6B). However, proportions of cells with non-template Gly and Ala were in the range 40-60% (12/17 clonotypes) for M1-tetramer+ VB19+ cells (Fig. 6C) and 40-90% (10/26 clonotypes) for CD8 T cells that proliferated in the peptide–specific culture (Fig. 6D). Therefore, we conclude that increased frequency of Gly/Ala runs is associated with flu M158-66-driven selection rather than with V-J recombination and thymic selection.
A detailed assessment of the TCR repertoires of antigen-specific T cells is a prerequisite for a better understanding of human anti-viral immunity. Here we systematically examined the α-TCR repertoires of memory CD8 T cells reactive against the influenza A viral epitope, M158-66, restricted by HLA-A2.1. The M158-66-specific, clonally diverse VB19 CD8 T cells expressed α-chains from several VA families with different CDR3 sizes. A unique feature of these α-TCRs was the presence of poly-Gly/Ala runs in the CDR3, fitting to an AGA(Gn)GG-like amino acid motif. These non-template-encoded poly-Gly/Ala runs in the CDR3 of the M158-66-specific memory pool were significantly enriched over that in naïve thymocytes, indicating that Gly/Ala runs provided a selective advantage in antigen-driven repertoire development in the periphery. These poly-Gly/Ala runs in the CDR3 of α- and β-chains might provide enhanced TCR flexibility during antigen recognition.
The mechanisms that shape T cell memory through α-TCR selection have been difficult to delineate due to the technical restraints associated with the lack of VA-family specific mAbs and T cell ability to co-express two α-chains (31, 33, 34). Nevertheless, our molecular cloning techniques demonstrate that the influenza A M158-66-specific T cell memory contains a number of additional features contributed by α-TCR diversity. These TCR α-chains that paired with the VB19 β-chains were of eleven VA families with three remarkably different sizes in CDR3α (Fig. 4B). Given that the M1-specific clonotypes from different VA families express different CDR1α and CDR2α (Table II), proper accommodation of different CDR1α and CDR2α to the M158-66-:HLA-A2 might occur if the CDR3α could undergo conformational adjustment. In this regard, enrichment of Gly and Ala might provide increased structural flexibility in CDR3α and satisfy this criterion.
It is commonly accepted that the fine specificity of epitope recognition is due to structural complementarity of CDR3α and CDR3β to MHC-presented immunogenic peptides, under conditions where CDR1 and CDR2 orient the TCR α- and β-chains to MHC molecules. In contrast to antibodies that usually have large surfaces with complementarity to their cognate antigens (35), only 21-34% of the αβ-TCR's surfaces are in direct contact with pMHC complexes (16, 36). Moreover, the contribution of CDR3α and CDR3β is relatively small, representing 21% and 24% on average of Vα and Vβ domains, respectively. These properties of TCR-pMHC interactions impose strict requirements on α- and β-chains. For example, side chains of amino acids located within the CDR3 must have optimized sizes and charges to interact with the foreign peptide, and CDR3 of α- and β-chains ought to have the similar sizes (25). However, we show examples where short CDR3α (12 amino acids) pair with one amino acid longer VB19-CDR3β (counting from C (CAS) to F (FGXG)), similar to crystallized αβ-TCR expressed by M158-66-specific clone JM22 (37). The CD8 T cells from VA27 and VA8 families in five studied individuals mostly expressed these short CDR3α loops. In contrast, longer CDR3α sequences (14 and 15 aa) were found with VA families other than VA27 and VA8. These “non-VA27” VB19+ T cells share α-JA42 chains with VA27 T cells (Fig. 4B), but express different CDR1 and CDR2 of their α-chains (Table III). Given that VB19 cells express rigid CDR3β fitting into the “IRSS” or “IGS”-like motifs, a plausible explanation that structurally different VA-domains are used to recognize influenza epitope is that the CDR3α loops undergo significant conformation (38) associated with poly-Gly/Ala runs. We therefore suggest that the CDR3α bearing long poly-Gly/Ala strings allow CDR1α and CDR2α encoded by different VA families to be used during influenza antigen recognition, and we discuss the theoretical basis for this suggestion below.
AJ42-gene encoded products are not unique in the sense of containing two Gly runs, since 7 out of 51 AJ genes encode two, and even three, Gly. However, 59% (37/63) of the clonotypes with CDR3α with 15 aa belong to this particular JA42 family (Fig. 4B). It seems that the combination of long CDR3α with poly-Gly/Ala runs provides flexibility for αβ-TCR to bind to the M158-66:HLA-A2 epitope, since we defined only eight clonotypes from “non-VA27” families (namely, VA10, 8.6, VA34 and VA35) that expressed short CDR3α of 12 amino acids (Fig. 4B). Gly is a unique amino acid because it lacks a side chain, and Ala contains only a methyl group as a side chain. It has been shown that proteins whose functions depend on adjustment to ligands often contain flexible loops. Usually Gly is located within these loops, providing flexibility in protein-protein or protein-ligand interactions. For example, poly-Gly strings have been found in the HIV protease flap region, in β1,4-galactosyltransferase-I, fructose-1,6-bisphosphate aldolase and other enzymes (39-41) . TCR contact with the pep:MHC molecule also follows the same rule. Thermodynamic studies of three TCR-pMHC binding, including αβ-TCR from M158-66-specific clone JM22, revealed that this process correlates with considerable conformational adjustment in CDR3α and CDR3β (42, 43). For instance, Malissen and co-workers reported that KB5-C20 (TCR specific to pKB1/H-2Kb) exhibits large conformational alteration in the CDR3β (6 amino acids longer than CDR3α) for proper accommodation to pMHC (15). Recently, the same group reported that similar structural flexibility might be observed in CDR3α (BM3.3 specific to two peptides with low sequence similarity presented by H-2Kb) (13). Based on these studies, the authors concluded that αβ-TCR propensity to modify its complementarity surface, mostly in the CDR3, might be the origin of αβ-TCR intrinsic ability to interact with the different epitopes. In line with these studies is the observation of structural flexibility of the αβ-TCRs expressed by human T cell clones reactive against Tax11-19 peptide (from HTLV) presented by HLA-A2.1(44, 45). Remarkably, these clones expressed β-TCRs (VB13.1) that contained a PG×G motif in the CDR3β and efficiently recognized the original epitope and its variants. Another proof outlining the importance of TCR structural flexibility in epitope recognition comes from the crystallization of the αβ-TCRS expressed by clone LC13 specific to EBNA-3339-347 peptide presented by HLA-B8. In this case, AlaGlyGly runs were contained in the CDR3β (46).
The occurrence of poly-Gly runs in β-TCRs with a long CDR3 is attributed to the D region where VDJ-gene transcription in three “open-reading-frames” would encode multiple Gly. This rule, however, cannot be applied to α-TCRs lacking D-encoded regions. The existence of VB19 clones specific to flu-M158-66 that utilize α-TCRs from different VA families provide an interesting example of epitope recognition where germ-line encoded segments of α-chains (i.e. SQG from AJA42 gene) contact M158-66, while segments created by AV/AJ42 recombination are positioned outside M158-66:HLA-A2 and could be flexible due to Gly/Ala enrichment (37, 42). A recent study of the αβ-TCR (from clone JM22 specific to M158-66-epitope) before and after binding to M158-66:HLA-A2 revealed that the CDR3α loop swiveled and made about 5Å outward shift (37, 42). It should be mentioned that CDR3α loop from JM22 contains only 12 amino acids. Therefore, we feel confident that the CDR3α with 14 and 15 aa defined in our study might have considerably more rotation and movement during M158-66:HLA-A2 binding and allow the CDR1α and CDR2α from different VA families to interact with HLA-A2 α1-helix.
A less explored field in human immunology is the analysis of the molecular nature of the pre-immune α-TCR repertoire. In our study, we could not exclude the possibility that poly-Gly/Ala runs might be a result of preferential AV27/AJ42 recombination where long CDR3α might have increased frequencies of Gly and Ala. If this were the case, we would have expected to see increased observations of Gly and Ala in CD4−8+ and CD4+8− thymocytes regardless of class I and II HLA-restrictions. It is important to point out that we examined the transcriptional profiles of α-TCRs where VA27/JA42 transcripts might encode functional (restricted by HLA-A2.1) and non-functional TCR α-chains. Following extensive sequencing analysis of AV27/AJ42 gene recombination and considering CDR3α sizes, we concluded that Gly and Ala have similar frequencies with other amino acids encoded by non-template segments of the VA27/JA42 rearranged genes (Fig. 6 and Supplemental, Table II). Although two Gly are derived from AJ42-gene as was expected since we examined VA27/JA42 transcripts, the non-template segments of long CDR3α (i.e. 14 and 15 aa) were not Gly- and Ala-enriched. Therefore, we conclude that selection for poly-Gly/Ala runs was driven in response to M158-66 epitope during influenza exposure rather than by a gene recombination.
The flexibility of αβ-TCR structure might be an important factor in the fate of memory T cells. In the case of influenza M158-66-specific cells, the poly-Gly/Ala runs do not contact M158-66 or HLA-A2 directly, based on the crystallization of the representative M158-66-specific JM22 clonal αβ-TCR and its variants (37, 42). Although the recognition M158-66:HLA-A2 is a function of VB19 β-chain (~70% of interactive interface), the α-TCRs with a highly flexible CDR3 might be used to recognize structurally different antigens, thus contributing to the pattern of T cell cross-reactivity which we have observed in “heterologous immunity” (47). If this is the case, influenza-specific VB19 T cells with IRSS in CDR3β might engage CDR1α and CDR2α during recognition of other p:HLA-A2, perhaps further enhancing the immunodominance of VB19 T cell clones.
Here we propose a three-step model explaining TCR interaction with the M158-66/HLA-A2.1 complex (Fig. 7). In the first step, CDR1β and CDR2β (β-VB19) contact the α2-helix of HLA-A2.1, pivoting the CDR3β to the peptide wherein R98 anchors the β-chains to HLA-A2 and S99 interacts with the M158-66 peptide. In the second step, the SQG (AJ42-gene encoded) anchors the CDR3α loop to Gly61 (M158-66), and the long CDR3α (poly-Gly/Ala) undergoes conformational change. Since AGA(G/A)G in CDR3α imposes minimal energy requirement to change shape, this leads to the third step, where engagement of different CDR1,2α (from any of VA-domain) is sufficient for final TCR:M158-66/HLA-A2.1 docking. The key element of this model is a long poly-Gly/Ala moiety, which allows CDR3 to be extremely flexible and adjust different TCR V-domains to the same pMHC complex. If long CDR3α- and/or CDR3β are more flexible and accommodate αβ-TCRs to different pMHC shapes and charges, then cross-reactivity against different antigens might be a major factor in memory formation. It is noteworthy mentioning that in our studies we also defined, based on tetramer binding, “non-VB19” cells that were able to recognize M158-66/HLA–A2 epitope. Remarkably, they also expressed long, 15 aa, CDR3β with a GXGG motif (Naumov, unpublished).
The early studies of the cytotoxic CD8 T cell lines and clones reactive against M157-68-expressing targets cells revealed that M157-68 peptide modifications in positions 58-60 were well-tolerated, while modifications in positions 61-65 could abrogate CTL response. In the last case, however, diminished cytotoxicity was not absolute and depended on amino acid substitutions. Interestingly, the T cell clones derived presumably from VB19 family have different patterns of epitope dependency in the CTL assay (10). Our own study demonstrated that small populations of the VB19 T cells are able to proliferate and to produce IFN-α in response to influenza M158-66 and EBV-BMLF1280-288 epitopes (30). Although we did not define yet the structure of the cross-reactive αβ-TCRs, these observations suggest that conformational flexibility of αβ-TCRs and clonal diversity of reactive T cells might be the best way to cope with different antigens.
In conclusion, we would like to suggest that the immune response evolves in a way where it engages T cells with structurally different αβ-TCRs specific to cognate antigens leading to an intrinsic capacity of these T cells to interact with different p:MHC shapes and charges. It is tempting to speculate that presence of multiple memory CD8 T cell clones of diverse specificities due to adjustable antigen receptors is the best way to optimize immune memory to ever-changing antigenic environment.
We would like to thank Dr. Jack Gorski and Dr. Martin Hessner for scientific discussion, and K. Bateman for editorial assistance with the manuscript.
This work was supported by NIH grants U19-AI057319 (YNN), U19-AI062627 (YNN, ENN) and AI45751 (LKS). The contents of this publication are solely the responsibility of the authors and do not represent the official view of the National Institutes of Health.