Antigen Selection Shapes the T-cell Repertoire in Chronic Lymphocytic Leukemia

A pivotal role for microenvironmental interactions in the natural history of chronic lymphocytic leukemia (CLL) is strongly supported by (i) the remarkably skewed repertoire of the antigen-specific immunoglobulins (IG) expressed by the B-cell receptors (BcR) of the CLL clones (1, 2); (ii) the division of CLL cases into those with unmutated or minimally mutated IGs (“unmutated” U-CLL) who generally experience an aggressive disease course and those with a significant load of somatic hypermutation within their IG genes (“mutated” M-CLL) who display a considerably more indolent disease (3–5); (iii) the existence of subsets of patients with quasi-identical, “stereotyped” IGs within their BcRs, pointing to clonal selection by a restricted set of antigens (6–10); and (iv) the clinical efficacy of drugs interfering with signaling through the BcR (11–13). However, it is not yet clear at which stage of clonal evolution the putative antigen(s) may act or whether this stimulation is persistent. Furthermore, their possible role in the selection and activation of cognate T lymphocytes remains obscure, yet highly relevant given accumulating evidence about the intricate cross-talk of T cells with CLL B cells (14–20).

A number of T-cell defects have previously been characterized in patients with CLL, including elevated absolute counts (21), inversion of the CD4⁺/CD8⁺ ratio (22), altered cytokine production and chemokine receptor expression profile (23), downregulation of co-stimulatory molecules (24), and oligoclonal T-cell receptor β chain (TRB) gene repertoire (25–28); however, the mechanisms underlying such defects are not fully elucidated. Gene expression profiling studies showed that the CLL B cells are able to induce gene expression changes in healthy CD4⁺ and CD8⁺ T cells following direct contact, potentially accounting for the changes observed in T cells from patients with CLL (14). Furthermore, animal model studies have demonstrated that autologous CD4⁺ T cells may actually provide trophic signals to the CLL clone, which are critical for its survival, expansion, and CD38 expression (16).

Taking all the above into consideration, here we sought to obtain immunogenetic insight into the role of antigen selection in shaping the repertoire of T cells in CLL. To this end, we performed a large-scale molecular study of the TRB gene repertoire in patients with CLL, focusing on those cases where the malignant clone expresses stereotyped BcR IGs and thus is most evidently selected by antigen(s). Overall, our findings suggest that the T cell driving antigens could include epitopes implicated in the selection of the malignant B cells and, perhaps more intriguingly, CLL cell (tumor)–derived epitopes.

We analyzed 73 peripheral blood samples from 58 untreated patients with CLL (median age, 68; range, 45–85 years) and 2 control samples from age-matched (ages, 70- and 80-year-old) healthy individuals. Fifty-two of the 58 CLL cases were selected on the basis of their expression of stereotyped BcR IGs leading to their assignment to well-annotated CLL subsets (10). Detailed information regarding the immunogenetic characteristics of these subsets is provided in the Supplementary Data. More specifically, our CLL series was composed of cases belonging to subset #4 (n = 20), subset #1 (n = 16), and subset #2 (n = 16), as well as cases carrying BcR IGs not assigned to subsets (n = 6). Five subset #4 cases were analyzed over time (19 samples in total, sampling period ranged from 18–42 months; median 28 months—the patients remained untreated throughout the sampling period). In one subset #1 case, in addition to peripheral blood, a lymph node specimen from the same time point was also analyzed. No case had evidence of infection at sampling. The local Ethics Review Committees of the participating institutions approved the study.

Total cellular RNA was isolated from peripheral blood mononuclear cells (or, in one case, fresh lymph node tissue) and reverse transcribed to cDNA using the SuperScript II RT Kit (Invitrogen Life technologies). RT-PCR and heteroduplex analysis of TRBV–TRBD–TRBJ gene rearrangements was performed according to the BIOMED-2 collaborative effort (29). RT-PCR products were gel-purified with the QIAGEN DNA purification columns (QIAGEN), ligated into the pCR2.1 vector (Invitrogen Life technologies), and transformed into E. coli/TOP10F' competent bacteria (Invitrogen Life technologies). Colonies were chosen randomly and sequenced using the −20 universal primer or M13 primers.

Sequence data were analyzed using the IMGT/V-QUEST tool from the international ImMunoGeneTics information system (http://www.imgt.org). Only productive TRBV–TRBD–TRBJ gene rearrangements were included in the analysis.

Clonotypes were defined as TRBV–TRBD–TRBJ gene rearrangements using a particular TRBV gene and carrying a distinct TRB complementarity determining region 3 (CDR3) amino acid sequence. Clonotypes were considered expanded (forming clusters) when they contained ≥2 sequences, with the most expanded clonotype within a sample referred to as the immunodominant clonotype. The relative frequency of each clonotype/sample was calculated as the number of rearrangements corresponding to the clonotype divided by the total number of subcloned productive rearrangements for that particular sample. Clonality analysis was performed for samples with >10 sequenced productive rearrangements. For repertoire analysis, clonotypes rather than single rearrangements were considered to avoid potential biases due to expansion following antigenic stimulation.

All CLL TRB sequences were aligned and compared with the TRB sequences of the healthy control samples, as well as a panel of 5,717 nonredundant, well-annotated, and productive TRBV–TRBD–TRBJ rearrangement sequences from various entities (normal, autoreactive, and malignant T-cell clones) available to our group (n = 1,715) or retrieved from the IMGT/LIGM-DB sequence database (http://www.imgt.org/IMGTindex/antibodies.html; n = 4,002).

Typing of the HLA-A, -B, -C loci for low resolution and HLA-DRB1 locus for allelic level high-resolution determination was performed on the 2 healthy controls and 8 CLL samples (more specifically, the 4 pairs of CLL cases that were found to share common clonotypes, as described in the Results) by reverse PCR-SSOP (sequence-specific oligonucleotide probe) using a commercially available kit (LABtype RSSO, One Lambda).

Descriptive statistics for discrete parameters included counts and frequency distributions. For quantitative variables, statistical measures included means, medians, and min–max values. The significance of bivariate/multivariate relationships between variables was assessed using the unpaired t test, and for all comparisons, a significance level of P = 0.05 was set. All statistical analyses were performed using the statistical Package GraphPad Prism version 5.03 (GraphPad Software, Inc.).

Circos graphs were produced using freeware, licensed under GPL, and available at http://circos.ca (30).

Overall, 1,378 productive CLL TRBV–TRBD–TRBJ gene rearrangements were included in the analysis, corresponding to 985 distinct clonotypes. Thirty-nine functional TRBV genes were identified. The most frequent gene was TRBV19, contributing to 108 clonotypes (11.0%), followed by TRBV27 (8.1%), TRBV12-3 (7.6%), TRBV7-9 (7.0%), TRBV6-5 (6.4%), and TRBV20-1 (5.1%). Thus, collectively, 6 TRBV genes accounted for almost half of all clonotypic rearrangements, highlighting a remarkable bias within the CLL TRBV gene repertoire. Importantly, analysis of the healthy control samples revealed a different pattern of TRBV gene usage, with clear overrepresentation of the TRBV10-3 and TRBV24-1 genes (P = 0.004 and P = 0.01, respectively), and underrepresentation of the TRBV27 and TRBV12-3 genes (P values did not reach statistical significance, likely due to the small size of the respective sequence dataset; Fig. 1A). Detailed immunogenetic information regarding gene frequencies and TRB CDR3 length and composition are provided in the Supplementary Data.

When comparing the TRBV gene usage among the four patient groups within our cohort (i.e., subsets #1, #2 and #4, as well as non-subset cases), several differences were noted. More specifically, (i) clonotypic rearrangements belonging to cases with heterogeneous BcR IGs were significantly enriched for TRBV27 and TRBV6-5 genes, whereas TRBV7-8, TRBV20-1, and TRBV7-9 were underrepresented; and (ii) clonotypic rearrangements belonging to subset #2 cases were significantly enriched for TRBV20-1, TRBV7-9, and TRBV7-8 gene usage (Fig. 1B).

In total, 73 samples from 58 patients with CLL were analyzed. The great majority of patients exhibited oligoclonality. More specifically, when samples with ≤10 sequenced productive rearrangements were excluded, 52 of the remaining 57 samples (91%), corresponding to 40 of 45 cases (89%), were found to carry clusters of identical rearrangements (≥2; expanded clonotypes). For the remaining 5 cases, the failure to detect expanded clonotypes may be attributed to their significantly (P = 0.02) lower median number of subcloned productive TRBV–TRBD–TRBJ rearrangement sequences.

For the 52 CLL samples with expanded clonotypes, the number of expanded clonotypes/sample ranged from 1 to 11 (median, 3), whereas the relative frequency of each clonotype/sample, ranged from 2.7% to 80% (median, 10%). The relative frequency of the immunodominant clonotype ranged from 4.9% to 70.4% (median, 16.2%) and the cumulative frequency of all expanded clonotypes/sample ranged from 10.5% to 88.9% (median, 44.8%). The respective values concerning all analyzed CLL samples (n = 57) are shown in Table 1. Of relevance, in the CLL case where peripheral blood and lymph node from the same time point were cross-examined, only the lymph node sample exhibited oligoclonality (cumulative frequency of expanded clonotypes 42.9%). This finding may also be relevant for the 5 CLL cases lacking expanded clonotypes in the peripheral blood: put differently, it cannot be a priori excluded that such cases carry expanded T-cell clones albeit with a distinct tissue localization.

We then assessed clonality separately for each subgroup of patients, that is, subset #1, subset #2, subset #4, and cases with non-subset BcR IG rearrangements. Unexpectedly, subset #2 was substantially less oligoclonal in comparison to subset #4, as well as patients with heterogeneous BcR IG rearrangements (median cumulative frequency of expanded clonotypes 23.8% vs. 48.3%, P = 0.03 and 51.1%, P = 0.01, respectively; Table 1; Figs. 2 and 3).

Importantly, the healthy control samples did not exhibit oligoclonality. Collectively, of the 35 clonotypes that were identified, 34 concerned “singletons” that is appeared only once whereas only a single clonotype was detected twice (median cumulative frequency, 5.0%; Fig. 2).

To assess clonal dynamics over time, we tested longitudinal samples from 5 subset #4 patients (2–6 samples per patient, median 3); the sampling period ranged from 18 to 42 months (median, 28 months). With the exception of one patient who displayed a relatively stable TRBV gene repertoire, all other patients exhibited variable TRBV gene frequencies at the different time points analyzed, suggestive of clonal drift (Fig. 4A); however overall, in 4 of 5 cases (80%), at least one clonotype was found to persist over at least two different time points (Fig. 4B). Interestingly, the immunodominant clone differed at each time point.

Of relevance, the TRBV gene repertoire was also found to differ substantially when both the peripheral blood sample and the lymph node sample from the same case and time point were examined, with the two samples presenting distinct TRBV gene frequencies.

Cluster analysis of the TRB CDR3 amino acid sequences of all cases identified 4 distinct TRB CDR3 amino acid sequences used by pairs of patients. In 2 of 4 such pairs of patients, the TRBV gene was the same, thus translating to a shared (“stereotyped” or “public”) clonotype; interestingly, these 2 pairs involved patients assigned to subset #4. The remaining 2 pairs carrying identical TRB CDR3 amino acid sequences exhibited discordant TRBV gene usage; one of these pairs involved a patient assigned to subset #1 and a subset #4 patient, whereas the other involved a subset #2 patient and a non-subset patient (Table 2).

Given that antigen recognition by T-cell receptors occurs in the context of the major histocompatibility complex proteins, we then investigated the respective gene loci (HLA typing) in these 8 cases (the 4 pairs of patients carrying identical TRB CDR3s) to search for possible matches and found that within each pair of patients at least one gene match was found among the 8 loci tested (Supplementary Data). For 2 of 4 CLL pairs, this “implicated,” shared HLA gene was also found to be expressed by the healthy controls (Supplementary Data); however, none of the aforementioned “public” clonotypes were detected in the latter.

We next compared all productive CLL TRBV–TRBD–TRBJ gene rearrangement sequences from our study (n = 1,378) against the TRBV–TRBD–TRBJ gene rearrangement sequences obtained from the two healthy control samples (n = 36) and also a panel of 5,717 nonredundant, productive TRBV–TRBD–TRBJ gene rearrangement sequences from various entities. Only a single TRB CDR3 sequence from our CLL dataset was found to have an identical match within the public database; notably, this particular TRB CDR3 sequence participated in one of the aforementioned “pairs” with identical TRB CDR3. More specifically, the match concerned a shared TRB CDR3 sequence between two patients with CLL (P1812 and Swe-1025) of our cohort (P1812 used the TRBV10-3 gene, whereas Swe-1025 used the TRBV19 gene) and a single entry in the IMGT/LIGM-DB sequence database (Accession number AM041155, using the TRBV19 gene). The latter corresponded to a CD8⁺ T-cell clone specific for a highly immunogenic Epstein-Barr virus (EBV) epitope (EPLPQGQLTAY). This clone is known to be HLA-B*35:01–restricted (31), and indeed, our two patients with CLL carrying this clonotype were both found to express HLA-B*35.

When following a less stringent threshold (>85%) for amino acid identity, several additional “hits” were identified concerning similar TRB CDR3s, as listed in the Supplementary Data.

We investigated the T-cell receptor gene repertoire in the largest CLL cohort studied so far for this purpose. We focused on patients with CLL expressing stereotyped BcR IG and assigned to the most populated CLL-specific subsets, namely #1 (the largest within U-CLL, clinically aggressive), #2 (the largest overall, mix of U-CLL and M-CLL, clinically aggressive), and #4 (the largest within M-CLL, particularly indolent; refs. 9, 10, 32). We also investigated patients with CLL carrying non-subset BcR IG rearrangements, so as not to disregard the possibility of identifying T-cell immunogenetic features that are ubiquitous in CLL, irrespective of the particular IG receptor that is expressed. Our cohort was intentionally biased towards subset #4, where ample evidence suggests ongoing interactions with the selecting antigen (33–36). For this reason, we also performed longitudinal immunoprofiling of five subset #4 patients, aiming to explore the possibility of clonal drift. Clonotypes were considered expanded when they contained ≥2 sequences, and while this definition may seem arbitrary, the identification of a specific rearrangement more than once using a low-throughput method, such as subcloning, justifies considering this clone as expanded.

The findings reported herein provide solid evidence for TRB gene repertoire skewing and oligoclonality in the great majority of patients with CLL analyzed. This is in line with previous reports of oligoclonality within the T-cell repertoire in CLL; however, ours is the first molecular documentation of this phenomenon, as previous studies were based solely on flow cytometric analysis and spectratyping (25–28). Importantly, within our study, T-cell oligoclonality was found to be universal in CLL and not confined to cases carrying stereotyped BcR IGs. The differences in TRBV gene usage between CLL and age-matched healthy samples argue that the reported TRB gene frequency skewing is CLL-biased and not an age-related repertoire restriction phenomenon. Although the small number of analyzed healthy samples precludes definitive conclusions, the relative lack of clonality within these healthy controls (at least with this low-throughput approach) efficiently highlights the greater extent of oligoclonality that characterizes CLL samples as compared with age-matched healthy controls, thus providing reassuring evidence that the reported TRB repertoire skewing is indeed CLL-biased.

That notwithstanding, TRB gene repertoire analysis performed separately for each subgroup of patients, that is, subset #1, subset #2, subset #4 and non-subset cases revealed that each subgroup exhibited varying TRBV gene restrictions, alluding to different antigenic elements driving T-cell selection. This is in line with previous studies detailing distinct antigenic stimulation that drives the selection and expansion of the malignant B-cell clone in CLL subsets carrying different stereotyped BcR IGs (6–10, 33, 34, 37–39). Counter-arguing, if certain antigenic elements were responsible for the TRBV gene repertoire skewing within each subgroup, one would expect the skewing to remain stable over time. However, with a single exception, all cases profiled over successive time points exhibited fluctuating frequencies of TRBV genes, although certain clonotypes did persist.

Subgroup analysis also revealed that clonality is less evident in certain subsets, especially subset #2. The underlying reason for this remains elusive; however, it should be noted that subset #2 cases follow an aggressive clinical course, even when harboring mutated BcR IGs and were recently found to display a particularly high frequency of SF3B1 mutations that may somehow render the cells less dependent on T-cell interactions through, as yet, unidentified mechanisms (37, 40–42). Indeed, whether T-cell oligoclonality is associated with disease course and outcome will require the study of larger cohorts of patients.

A notable finding in the present study was the identification of shared (“public”) clonotypes among different patients. According to estimates of normal T-cell repertoire diversity, the chance that this may happen by serendipity is negligible (43, 44). The possibility that this resulted from cross-contamination was also ruled out by the fact that shared clonotypes were identified in subcloning experiments performed at different times. Therefore, we have to assume that “public” clonotypes present in different patients with CLL stem from selection by the same or closely similar antigenic elements. Could these be the same antigens that are selecting the malignant B-cell clone?

If this were the case, one would expect that common clonotypes would only be identified among patients belonging to the same subset (and sharing the respective HLA alleles). Indeed, identical clonotypes were found only among subset #4 patients with at least one common HLA allele. However, if a more relaxed definition of “clonotype” was adopted, for example, involving the same amino acid TRB CDR3 sequence yet with different TRBV genes, then common clonotypes were also found among patients in different subgroups. Thus, the possibility exists that the common antigenic drive concerns tumor-associated antigens, that is, proteins specifically expressed by the CLL cells. Alternatively, it could be related to shared, persistent subclinical infections.

To further interrogate the nature of antigenic elements that interfere with T-cell selection in CLL, we performed extensive cross-entity comparisons. Notably, within a panel of 5,717 productive TRBV–TRBD–TRBJ rearrangement sequences, only one identical match was found, which corresponded to the TRB CDR3 sequence shared by two patients with CLL from our cohort and a CD8⁺ T-cell clone from the literature, highly specific for an immunogenic epitope from the BZLF1 antigen of EBV (31). Interestingly, this clone is part of a “public” T-cell receptor repertoire, that is, a highly constrained, HLA-restricted T-cell receptor repertoire that is usually elicited in response to bulged viral peptides and is common among HLA-B*35:01 individuals (31, 45). Given that both patients with CLL also expressed HLA-B*35, it is reasonable to say that, at least on this occasion, the identified shared clonotype stemmed from EBV antigenic stimulation. This is not altogether unexpected, given that the immunocompromise of CLL is a predisposing factor for reactivation of chronic herpes virus infections (46); furthermore, it is relevant to mention our previous finding that demonstrated that certain patients with CLL, especially those using the IGHV4-34 gene, have persistent subclinical infection by EBV, indicating that viral antigens/superantigens might facilitate clonal expansion (47).

In conclusion, the present study provides clear molecular evidence for TRBV gene repertoire skewing among T cells in patients with CLL belonging to major stereotyped subsets as well as non-subset patients, thereby strongly supporting antigen selection in disease pathogenesis. The finding of identical or highly similar clonotypes among different cases raises the possibility that shared antigenic epitopes may be relevant for clonal selection of T cells. Whether the antigens that drive T-cell repertoire restriction are identical/related to those implicated in the selection of CLL progenitors or the malignant cells themselves remains to be clarified.

K. Stamatopoulos reports receiving commercial research grants from GlaxoSmithKline, Novartis Hellas, Roche SA. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Agathangelidis, R. Rosenquist, K. Stamatopoulos

Development of methodology: E. Stalika, M. Karypidou

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Siorenta, A. Anagnostopoulos, L.-A. Sutton

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Vardi, A. Agathangelidis, R. Rosenquist, L.-A. Sutton, K. Stamatopoulos

Writing, review, and/or revision of the manuscript: A. Vardi, A. Agathangelidis, R. Rosenquist, P. Ghia, L.-A. Sutton, K. Stamatopoulos

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Karypidou, A. Anagnostopoulos, L.-A. Sutton

Study supervision: A. Agathangelidis, A. Hadzidimitriou, K. Stamatopoulos

This work was supported in part by the ENosAI project (code 09SYN-13-880) co-funded by the EU and the Hellenic General Secretariat for Research and Technology; the KRIPIS action, funded by the Hellenic General Secretariat for Research and Technology and the European Regional Development Fund of the EU under the O.P. Competitiveness and Entrepreneurship, NSRF 2007–2013; Associazione Italiana per la Ricerca sul Cancro AIRC (Investigator Grant and Special Program Molecular Clinical Oncology—5 per mille #9965) and Ricerca Finalizzata 2010—Ministero della Salute, Roma; the Swedish Cancer Society, the Swedish Research Council and H2020 ‘AEGLE, An analytics framework for integrated and personalized healthcare services in Europe’, by the European Commission, the Swedish Cancer Society and the Swedish Research Council. K. Stamatopoulos receives research support from Roche SA, Novartis Hellas, and Glaxo SmithKline. A. Agathangelidis is recipient of a fellowship by Associazione Italiana per la Ricerca sul Cancro AIRC (Triennial fellowship “Guglielmina Lucatello e Gino Mazzega”).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.