Identification of a public CDR3 motif and a biased utilization of T-cell receptor V beta and J beta chains in HLA-A2/Melan-A-specific T-cell clonotypes of melanoma patients

To analyze TRBV or TRBJ segment usage, the 95% confidence intervals of the respective proportions were calculated. "Preferentially used" were defined those segments whose lower limit of the respective 95% confidence interval was higher than the mean percentage of TRBV or TRBJ transcripts usage, obtained by arbitrarily hypothesizing a uniform distribution of all segments. When proportions were compared, Fisher's exact test was employed, while the differences between the means of CDR3 length distributions in the four groups of clonotypes were evaluated by Kruskal-Wallis test and Dunn's post-hoc test. Results were considered significant for p < 0.05.

We first investigated whether clonotypes identified in HLA-A2+ melanoma patients with CTL specificity against Melan-A (Mel/M-A group) had a preferential usage of particular TRBV chains and whether these preferential TRBV were also predominantly utilized in the control (Ctrl/M-A, Mel/noM-A and Ctrl/HLA-A2+ groups) clonotypes. As shown in Figure 1A, multiple transcripts covering the majority of the TRBV families were observed in the 4 groups of clonotypes, although some TRBV segments were preferentially used. In particular, while TRBV6 and TRBV27 were highly represented in all groups of clonotypes, TRBV4 was overrepresented in response to melanoma Ags but not to unrelated Ags, TRBV19 was preferentially used in clones of HLA-A2+ control individuals, and TRBV28 appeared to be preferentially selected only by Melan-A-specific CTL. TRBV usage comparison among the 4 groups suggested that the proportion of clonotypes using TRBV27 chains was higher in Mel/M-A, Ctrl/M-A and Mel/noM-A sequences compared to Ctrl/HLA-A2+ clonotypes (p = 0.03; p = 0.004; p < 0.001), while TRBV28 was significantly more frequent in Mel/M-A clonotypes than in Mel/noM-A and Ctrl/HLA-A2+ groups (p = 0.001 and p < 0.001).

Among Mel/M-A clonotypes there was a high number of clonotypes bearing the TRBJ2-1, TRBJ2-7 and TRBJ1-5 segments (Figure 1B). However, the first two TRBJ chains, however, were highly utilized also in other groups of clonotypes (Figure 1B), and had also been frequently observed among peripheral blood T-cells from healthy individuals [56].

The mean CDR3 length was highly similar (p = NS) in Mel/M-A, Ctrl/M-A and Mel/noM-A groups (mean ± SD: 12.37 ± 1.29, 12.32 ± 1.43 and 12.35 ± 1.71, respectively), but significantly lower in Ctrl/HLA-A2+ clonotypes (11.95 ± 1.50) in respect to Mel/M-A (p < 0.01) and Mel/noM-A sequences (p < 0.05). Furthermore, the majority of Mel/M-A and Ctrl/M-A CDR3 were 12 amino acid long (32.9% and 31.9% respectively), while most of CDR3 of Mel/noM-A and Ctrl/HLA-A2+ sequences were 13- and 11-amino acid-long, respectively (Figure 1C).

Collectively, the present analysis demonstrated that in melanoma patients there is a biased T-cell response to Melan-A, which is characterized by TR clonotypes using preferentially TRBV28 and TRBJ1-5 segments and containing a 12-amino acid-long CDR3.

The amino acid composition of TRB hypervariable regions of Melan-A-specific CTL from melanoma patients were subsequently analyzed in detail. Serine, Glycine, Alanine and Glutamine were by far the most frequently used residues in the IMGT-defined CDR3, and were almost equally represented in all groups of analyzed sequences (Figure 2A). However, while Alanine, Serine, and Glutamine were abundantly present because of their occurrence at positions 105, 106, 107 and 114 in the majority of canonical TRBV and TRBJ chains, Glycine, as reported for murine [57] and human sequences [56], was clearly predominant in the region created by N-D-N recombination events. Furthermore, in the N-D-N region of Mel/M-A and Ctrl/M-A sequences there was an increased Leucine usage (Figure 2A), and Glycine and Leucine were overrepresented at CDR3 positions 110, 112 and 113 (Figure 2B). Moreover, the overall percentage of non-polar amino acids at these CDR3 positions in the clonotypes carrying 12-amino acid-long CDR3s, which were the most commonly represented among the Melan-A-specific T-cell clones, was significantly higher in the Mel/M-A group (75%) compared to Ctrl/M-A (62%, p = 0.017), Mel/M-A (52%, p < 0.001) and Ctrl/HLA-A2+ (38%, p < 0.001) groups. This indicates that non-polar amino acids may be important for Melan-A-peptide-TR interaction. Furthermore, we found a public clonotype identified in two laboratories from cells of two melanoma patients: one was sequenced in our laboratory starting from a T-cell clone (ID 16) obtained from patient 22 [manuscript in preparation], the other from a T-cell clone (ID 27) obtained in the laboratory of Trautmann et al [6] employing melanoma-infiltrating lymphocytes of patient M180 (Figure 3). Both sequences contained identical 12-amino acid-long CDR3s, created by the joining of TRBV28 and TRBJ1-5 segments and containing a Glycine-Leucine-Glycine stretch at positions 110-112-113 of the CDR3. This motif was recurrent among other sequences derived from several patients, since it was found in 27 additional clonotypes sequenced in different laboratories and obtained from 15 melanoma patients. This peculiar motif rearranged only with members of TRBJ1 cluster, because 19 out of 29 clonotypes were joined with TRBJ1-5 segments, 7 with TRBJ1-1, 2 with TRBJ1-2 and one with TRBJ1-6 (Figure 3). TRBV usage was also restricted in these clonotypes since 16 of them were TRBV28, 7 were TRBV30 and 2 were TRBV20. The recurrent motif was found in Melan-A-specific CTL isolated from PBL and from tumor sites of HLA-A2+ melanoma patients, independently of the stage of disease and of the methodological approaches used for T-cell cloning. The same motif was identified in two Melan-A T-cell clones derived from cells of healthy donors [5, 19], but not in the remaining 504 clonotypes sequenced from T-cell lines or clones with specificity for other Ags. Similarly, the Glycine-Leucine-Glycine motif at position 110-112-113 was absent in the 219 clonotypes identified analyzing 353 sequences randomly obtained from CD8+ lymphocytes of healthy subjects (data not shown). Furthermore, no common motifs were found when Melan-A-specific sequences of melanoma patients were compared using particular BV or BVBJ combinations. Of clinical relevance, the Glycine-Leucine-Glycine motif was detected in lymphocytes obtained from untreated patients, representing spontaneous anti-tumor responses, as well as from patients having undergone vaccination with the natural or modified peptides (Figure 3). Interestingly, one clonotype sequenced in our laboratory (ID 4) was detected both in samples prepared before and after the vaccination [58]. Furthermore, all but one clonotype containing the Glycine-Leucine-Glycine motif were sequenced from T-cell clones whose specificity was identified using modified Melan-A peptide/multimers. The specificity of the remaining clone for natural Melan-A peptide was established by the analysis of the ability of Melan-A-transfected COS-7 cells to stimulate IFN-γ release. This last clonotype (ID 1E2), identified by Cole et al [10], bore TRBV28 and TRBJ1-1 chains and differed only by the amino acid at position 109 (Figure 3) from ID 57, ID CTL01 and ID 6E4 clonotypes [6, 7, 18], which were sequenced starting from 3 melanoma patients. Furthermore, the same motif was present, at slightly different positions of the CDR3, in 7 other Melan-A-specific clonotypes [5, 7, 10, 19], but never in non-Melan-A clonotypes. While the Glycine-Leucine-Glycine stretch is composed exclusively by non-polar or frankly hydrophobic amino acids, all the amino acids at position 114 and several of those at position 109 were hydrophilic (Figure 3). Finally, we looked for very similar sequences at the same CDR3 positions because it is conceivable that these sequences adopt equivalent structures in the recognition complex. We found a Glycine-Valine-Glycine stretch in 8 clonotypes, 5 of which were identified in melanoma patients [[4, 12, 14, 30] and manuscript in preparation] and 3 in controls [3, 5].

Since previous studies focusing on the analysis of shared TR amino acid sequences in humans did not address the extent to which TRB nucleotides are shared among public amino acid stretches, we identified the N-D-N regions of the 22 available nucleotide sequences of clonotypes with Glycine-Leucine-Glycine at position 110, 112 and 113. As summarized in Table 2, all N-D-N regions were different, with the only exception of those of ID D/a and ID 30 sequences, in which, however, the Adenine at the extreme 3'V region must be ascribed to the TRBV segment in clone ID D/a and to the D region in clone ID 30. Finally, the alignment of the 22 nucleotide sequences with the TRBV, TRBJ and TRBD germline gene segments allowed us to calculate the germline contribution and the number of nucleotide deletions (the so-called "nibbling") and additions during the VDJ recombination process. The exonucleolytic nibbling was highly heterogeneous: at 3' V end varied from 0 to 7 nucleotides, at 5' J end ranged from 4 to 9, at 3' D from 0 to 9 and at 5' D from 0 to 7. Similarly, N-addition was highly different at both sites ranging from 0 to 9 nucleotides at N1 and from 0 to 6 at N2 position. Finally, also TRBD region length is diverse since it varies from 3 to 8 nucleotides.