Depletion of polyfunctional CD26^highCD8⁺ T cells repertoire in chronic lymphocytic leukemia

We recruited 55 patients with confirmed CLL for this study (Additional file 2: Table S1), along with 44 age-and-sex-matched healthy controls for comparison. We collected peripheral blood specimens and bone-marrow aspirates in EDTA-containing tubes. The clinical data including IGHV mutation status, FISH analysis, clinical staging (Rai staging system) [51], and treatment state/course were collected for further analysis.

The peripheral blood mononuclear cells (PBMCs) and the bone-marrow cells were isolated using Ficoll-Paque gradients (GE Healthcare). CD3⁺ T cells were enriched by a negative selection kit (EasySep isolation kit, Stem Cell Technologies) with a purity of > 97% (Additional file 1: Fig. S1a). For effector T cell (CD3⁺CCR7⁻) isolation, CD3⁺ T cells were stained with the PE-conjugated anti-CCR7 antibody followed by the anti-PE-conjugated microbeads (Miltenyl) with a purity of > 96% (Additional file 1: Fig. S1b). For B cell enrichment, B cells were stained with FITC-conjugated anti-CD19 antibody and then isolated by the anti-FITC microbeads (Miltenyl) with a purity of > 91% (Additional file 1: Fig. S1c).

The fluorochrome-conjugated antibodies were purchased from BD Biosciences, Thermo Fisher Scientific or Biolegend including human anti-CD3 (SK7), anti-CD4 (RPA-T4), anti-CD8 (RPA-T8), anti-CD26 (M-A261), anti-CD161(HP-3G10), anti-TV \(\alpha\) 7.2 (3C10), anti-IL-18R \(\alpha\) (H44), anti-CD5 (UCHT2), anti-CD19 (HIB19), anti-CD160(BY55), anti-2B4(eBioDM244), anti-TIGIT(MBSA43), anti-PD1 (EH12.1), anti-TIM-3 (7D3), anti-CD39 (TU66), anti-CD73 (AD2), anti-CD95 (DX2), anti-CD127 (HIL-7R-M21), anti-ROR \(\gamma \delta\)(Q21-559), anti-CD45RA (HL100), anti-CCR7 (3D12), anti-CD27 (G3H69), anti-CD28 (CD28.2), anti-ICOS (C398.4A), anti-CD57 (NK-1), anti-CD16 (B73.1), anti-CD56 (B159), anti-KLRG1 (2F1/KLRG1), anti-CD69 (N50), anti-CD107a (H4A3), anti-IL-2 (MQ1-17H12), anti-TNF-\(\alpha\)(MAB11), anti-IFN-\(\gamma\)(45.B3), anti-Perforin (dG9), anti-Granzyme-B (GB11), anti-Granzyme-K (G3H69), anti-CLA (HECA452, Miltenyi), anti-CCR4 (1G1), anti-CCR5 (2D7/CCR5), anti-CCR6 (11A9), anti-Integrin-\(\beta\) 7 (FIB504), anti-CXCR3 (1C6/CXCR3), anti-CXCR4 (12G5), anti-Galectin-9 (9M1-3), anti-Annexin-V (Annexin V), anti-TOX (TXRX10), anti-FOXP3 (150D/E4), and anti-T-bet (4B10). We also used mouse anti-CD26 (H194-112), anti-CD8 (53–6.7), and anti-CD3 (17A2) antibodies.

Surface staining was performed, as we have reported elsewhere [52, 53]. Data were acquired on an LSR Fortessa-SORP (BD Bioscience) and subsequently analyzed using Flow Jo software (V.10.8.1). Cell viability was analyzed using the LIVE/DEAD kit (Life technologies).

Isolated PBMCs were cultured and stimulated with soluble Purified NA/LE anti-human CD3 (UCHT1, 3 μg/ml)/CD28 (CD28.2, 1 μg/ml) or PMA (20 ng/ml)/Ionomycin (1 g\(\mu\)/ml) (Cell stimulation cocktail, Biolegend) in the presence of the protein transport inhibitor Brefeldin A (BD Biosciences, 1/1000) for 5 h. Intracellular cytokine staining was performed according to our protocols [54]. For cytokine-dependent cultures, PBMCs were treated with a cocktail of cytokines including recombinant human IL-12 (Cedarlane,100 ng/ml), IL-18 (Biolegend,100 ng/ml), and IL-15 (Biolegend, 100 ng/ml) for 18 h. Brefeldin A (1/1000) was added 5 h before the intracellular staining. In other experiments, the effects of different cytokines including TNF-α (50 ng/ml), IFN-γ (100 ng/ml), IL-10 (100 ng/ml), IL-16 (500 ng/ml), IFN-α (100 ng/ml), IL-2 (20 ng/ml), IL-6 (100 ng/ml), and TGF-β (20 ng/ml) on CD26 expression was analyzed.

The RNA was isolated from CD8⁺ T cells from HCs and CLL patients using the Direct-zol RNA MicroPrep kit (Zymo Research). cDNA was synthesis using the Quantitect Reverse Transcription Kit (Qiagen). RT-PCR was carried out using the Quantitect primer Kit (Qiagen) to measure the expression of CD26 mRNA. Each sample was run in duplicate, using the CFX96 Touch Real-Time PCR Detection System (BioRad). Beta-2-microglobulin was used as a reference gene and the relative fold change of the targeted genes was calculated by the ΔΔ CT method.

Isolated effector T cells (CD3⁺ CCR7⁻) were labeled with the CFSE dye (Life Technologies) before stimulation using the Dyna beads Human T-activator CD3/CD28 (Thermo Fisher Scientific) according to the manufacturer’s instruction and our protocols [53, 54]. After 72 h cells were stained and analyzed.

The migration assay was performed using the CytoSelect migration assay kit (Cell Biolabs), as we have reported elsewhere [55, 56]. PBMCs were starved overnight in FBS-free culture media. The next day, FBS (10%), recombinant human RANTES (CCL5) (R and D, 10 nM), and recombinant human IL-18 (Biolegend, 100 ng/ml) were used as chemoattractants. Cell suspension of starved cells (0.5 × 10⁶ cells/well) was added to the upper chamber and incubated in the incubator (37 °C, 5% CO25) for 24 h. Migrated cells in the lower chamber were harvested and quantified by flow cytometry according to the manufacturer’s instructions. The migration ratio was calculated compared to the wells lacking the chemoattractant.

The plasma concentration of cytokines/chemokines was measured using the Meso Scale Discovery (MSD) multiplex kit, as we have reported elsewhere [45]. Data were acquired on the V-plex^® Sector Imager 2400 plate reader and analyzed using the MSD Workbench 3.0 software. In addition, soluble CD26, IL-18, TGF-β, and Galectin-9 (Gal-9) were detected using the DuoSet ELISA kit (R&D) according to the manufacturer’s protocol. The microplate reader (Synergy H1 Biotek) was used for acquiring ELISA data and analyzed by Gen5 V.2.07 software.

To determine the frequency of CD26⁺CD8⁺ T cells, PBMCs from CLL patients (n = 55) and healthy controls (HCs) (n = 44) were subjected to CD26 expression analysis. (Additional file 1: Fig. S1d, the gating strategy). These studies revealed that the frequency of CD26⁺CD8⁺ T cells was significantly declined in CLL patients compared to HCs (Fig. 1A, B). While on average half of CD8⁺ T cells in PBMCs of HCs expressed CD26 (Mean \(\pm\) SD: 45.88 \(\pm\) 21.36) this was substantially lower (Mean \(\pm\) SD: 26.84 \(\pm\) 16.64) in CLL (Fig. 1A, B). Although the majority of CD26 expressing CD8⁺ T cells were CD26^low, the proportions of both CD26^high and CD26^low were significantly reduced in CLL patients compared to HCs (Fig. 1A, C, D). We also measured the cell number in both groups, which confirmed a significant reduction in the number of CD26^low and CD26^high T cells in CLL patients (Additional file 1: Fig. S1e). Moreover, the intensity of CD26 expression was significantly decreased in CD8⁺ T cells of CLL patients compared to HCs (Fig. 1E, F). As expected, the CD26^low subpopulation had a significantly lower intensity of CD26 expression than their CD26^high counterparts in CLL patients (Additional file 1: Fig. 1f).

Given the reported impact of age and sex on T cell repertoire [57], we found these variables did not influence the frequency of CD26⁺CD8⁺ T cells (Additional file 1: Fig. S1g, h). Also, we did not observe any difference in the frequency of CD26⁺CD8⁺ T cells among treated (n = 18) versus untreated (n = 37) CLL patients (Additional file 1: Fig. S1i, j). To determine the potential correlation between the clinical staging of CLL patients with the frequency of CD26⁺CD8⁺ T cells, we stratified our patients according to the Rai staging system for CLL into three groups [51]; however, we did not observe any significant difference between them (Additional file 1: Fig. S1k). Moreover, our analysis revealed that the lymphocyte count in the whole blood did not correlate with the frequency of CD26⁺CD8⁺ T cells (Additional file l Fig. S1l). To determine whether CD26 downregulation was CD8⁺ T cells specific or a general phenomenon of CLL, we measured the expression of CD26 on other blood mononuclear cells. Although CD4⁺ T cells were the most dominant CD26 expressing cells, their frequency in CLL patients was also significantly reduced compared to HCs (Additional file 1: Fig. S1m–o). Of note, NK cells exhibited a very small proportion of CD26-expressing cells without any difference between HCs and CLL patients (Additional file 1: Fig. S1p). Also, we compared the mean fluorescence intensity (MFI) of CD26 in malignant B cells in CLL with non-malignant B cells. These studies revealed a significant increase in the intensity of CD26 in malignant B cells, which is consistent with previous reports [58, 59] (Additional file 1: Fig. S1q). These observations suggested that CD26 may get shed from the cell surface resulting in the elevation of soluble CD26 in the plasma. However, the plasma levels of CD26 in CLL patients did not support this hypothesis (Additional file 1: Fig. S1r). Moreover, we compared CD26⁺CD8⁺ frequencies in the bone marrow and blood of CLL patients. Despite a trend towards lower CD26⁺CD8⁺ T cells in the bone marrow, it was not significant (Additional file 1: Fig. S1s, t). Finally, to understand the stage of CD26 reduction, we quantified CD26 mRNA levels in CD8⁺ T cells from CLL and HCs. However, we did not find any significant difference between the groups at the gene level (Additional file 1: Fig. S1u). Overall, these observations support the notion that CLL is associated with a substantial decline in the frequency of CD26⁺ T cells, particularly CD26^lowCD8⁺ T cells, without any changes in the plasma levels of soluble CD26.

To better phenotype CD26⁺CD8⁺ T cells in CLL, we conducted a detailed ex vivo analysis of these cells. Based on CD45RA, CCR7, CD95, and CD27 markers, we characterized T cell subsets such as naïve (CD45RA⁺CCR7⁺CD95⁻), stem cell memory (CD45RA⁺CCR7⁺CD95⁺), central memory (CD45RA⁻CCR7⁺), transitional memory (CD45RA⁻CCR7⁻ CD27⁺), effector memory (CD45RA⁻CCR7⁻ CD27⁻), and effectors (CD45RA⁺CCR7⁻) [60‐62]. We found that CD8⁺CD26^low T cells were mainly naïve, stem cell memory, and central memory (Fig. 1G). As illustrated in this figure, the frequency of CD8⁺CD26^low expressing T cells declines as T cells differentiate to other subsets (e.g. transitional memory, effector memory, and effectors) in both HCs and CLL patients. In contrast, CD8⁺CD26^high T cells were uniquely populated in transitional and effector memory subsets with very low frequency in other subsets and absent in the naïve population (Fig. 1G). In particular, we observed that the frequency of CD26^low was significantly lower in stem cell memory, effector memory, and effectors but unchanged in other T cell subsets in CLL patients compared to HCs (Fig. 1H–N). However, the frequency of CD26^highCD8⁺ T cells was significantly lower in all T cell subsets except the naïve subset in CLL patients compared to HCs (Fig. 1H–N). Despite the expansion of total effector memory and effector CD8⁺ T cell subsets in CLL patients (Additional file 1: Fig. S2a, b), the frequency of those expressing CD26^high was significantly lower in CLL patients (Fig. 1M, N).

Moreover, to better characterize CD26⁺CD8⁺ T cells, we subjected them to CD27 expression analysis. CD27 is involved in CD8⁺ T cell activation and memory formation that augments anti-tumor activity [63, 64]. Our further studies confirmed the abundance of transitional and effector memory subsets in CD26^low and CD26^highCD8⁺ T cells (Additional file 1: Fig. S2c, d).

Overall, our results indicate that CD26⁺CD8⁺ T cells are in distinct stages of differentiation. Considering the substantial co-expression of CD27 and CD26, a considerable decline in the proportion of CD26⁺CD8⁺ T cells may deprive CLL patients of the potent anti-tumor activity of this T cell subset [64].

MAIT cells express high levels of CD161 [40], IL-18R \(\alpha\), and CD26 [36, 41]. In particular, the most specific surrogate marker for MAIT cells is the co-expression of CD161^high and TV \(\alpha\) 7.2 [65]. As such, we decided to determine whether CD26⁺CD8⁺ T cells were MAIT cells. We observed that the majority (Mean \(\pm\) SD: 67 \(\pm 14.48\)) of CD26^highCD8⁺ T cells co-expressed TV \(\alpha\) 7.2⁺ & CD161^high in CLL patients (Fig. 2A–C). However, a portion of CD26^high did not express TV \(\alpha\) 7.2 and CD161^high (Fig. 2A, C). Of note, the frequency of MAIT-like cells expressing TV \(\alpha\) 7.2 and CD161^high was significantly lower among CD8⁺CD26⁺ T cells in CLL patients (Additional file 1: Fig. S2e–g). Moreover, we investigated the expression levels of IL-18R \(\alpha\) in three subsets of CD26^neg, CD26^low, and CD26^highCD8⁺ T cells, which elucidated that CD26^high cells were the dominant cells expressing IL-18R \(\alpha\) in both HCs and CLL patients. However, a portion of CD26^high CD8⁺ T cells lacked the expression of this cytokine receptor (Fig. 2D–F). These observations suggest that CD26^highCD8⁺ T cells are enriched with MAIT-like cells but they are a heterogeneous T cell subset. It is worth mentioning that we found a considerable frequency of MAIT-like cells that did not express CD26^high. As such, we speculate that CLL may impact the expression of MAIT surrogate markers. Therefore, CD26^high might not be a definite marker for MAIT cell identification in CLLs. However, the majority of CD26^highCD8⁺ T cells displayed the MAIT-like phenotype (CD161^high TV \(\alpha\) 7.2⁺) and significantly declined in CLL patients compared to HCs.

To better characterize CD26⁺ versus their negative counterparts, we subjected them to further analysis for the expression of co-stimulatory/inhibitory molecules. Previously, we have reported that the co-inhibitory receptor, CD160, was selectively overexpressed on CD8⁺ T cells in CLL patients [45]. In agreement, we found that CD26^negCD8⁺ T cells were significantly enriched with CD160 (Fig. 2G, H), 2B4 (Fig. 2I, J), TIGIT (Fig. 2K, L), and ICOS (Fig. 2M, N) expressing T cells than their CD26^low and CD26^high siblings. On the contrary, CD26^low and CD26^high T cells were significantly populated with CD28 and CD27 expressing CD8⁺ T cells (Fig. 2O–R). Interestingly, we did not observe any difference in the proportion of PD-1 expressing CD8⁺ T cells between CD26^neg, CD26low, and CD26^high subsets ((Fig. 2S, T). Although HCs, in general, have a lower frequency of T cells expressing co-inhibitory/stimulatory receptors, we observed a significant reduction in the proportion of CD160, 2B4, and PD-1 expressing cells in CD26^lowCD8⁺ T cell subset compared to their CD26^neg or CD26^high siblings (Additional file 1: Fig. S2h–j). The frequency of TIGIT-expressing T cells was significantly lower in CD26^low/CD26^high subsets in HCs (Additional file 1: Fig. S2k). Although ICOS and CD28 had similar expression patterns in CD26^−/+ subsets, CD27⁺CD8⁺ T cells were significantly abundant in the CD26^low subset in HCs (Additional file 1: Fig. S2l–n). Owing to the tandem contribution of ectonucleotidases CD39, CD73, and CD26 in the adenosine pathway [66], we measured the expression of these two ectoenzymes in different subsets of CD26-expressing CD8⁺ T cells. We found that CD39 was highly expressed in CD26^negCD8⁺ T cells (Fig. 2U, V), whereas CD73 was prominently expressed in CD26^low followed by CD26^high and CD26^neg CD8⁺ T cells (Fig. 2W, X). Similarly, we analyzed the expression of CD39 and CD73 in CD26⁺CD8⁺T cells of HCs, which showed no differential expression pattern for CD39 but we found a higher abundance of CD73⁺ T cells among CD26^low compared to their CD26^neg/CD26^high counterparts (Additional file 1: Fig. S2o, p). Our further analysis revealed that CD26^lowCD73⁺ T cells predominantly displayed naïve T cell phenotype (Additional file 1: Fig. S2q, r). However, the subpopulation of CD26⁺CD73⁺ with effector and effector memory phenotype (CD8⁺CCR7⁻ T cells) was significantly reduced in CLL patients versus HCs (Additional file 1: Fig. S2s). Overall, we found that CD160, 2B4, TIGIT, and PD-1 expressing CD26^lowCD8⁺ T cells were significantly enriched in CLL versus HCs (Fig. 2Y). In contrast, the proportion of CD27⁺CD73⁺ expressing cells was significantly reduced in the CD26^lowCD8⁺ T cell subset without any changes in the frequency of ICOS, CD28, CD39 expressing CD8⁺ T cells in CLL versus HCs (Fig. 2Y). However, this pattern was different for CD26^high T cells and they showed higher expression of TIGIT⁺ and CD27⁺ cells in CLL patients compared to HCs (Additional file 1: l Fig. 2t). Interestingly, we found that the proportion of 2B4⁺ T cells was significantly increased but CD28, CD27 and CD73-expressing cells were decreased among CD26^negCD8⁺ T cells in CLL patients (Additional file 1: Fig. 3a).

CD8⁺ T cells as cytotoxic T lymphocytes play an essential role against virally infected and tumor cells [67, 68] via granule-mediated cytotoxicity, FAS-FASL interaction, and the release of cytokines (e.g. TNF-α and IFN-γ) [69]. Granzymes as cytolytic molecules and their harmonized action with perforin (pore forming protein) are required for effective granule-mediated cytotoxicity [70]. To determine the granule-mediated cytotoxic ability of CD26^±CD8⁺ T cells, we measured the intracytoplasmic expression of perforin and granzyme-B (GzmB) in CD26^neg, CD26^low, and CD26^high CD8⁺ T cells in CLL patients ex vivo. We found that in contrast to the CD26^neg subset, CD26^low and CD26^high cells were devoid of perforin and expressed very low levels of GzmB (Fig. 3A–E). Considering the heterogeneous nature of CD8⁺ T cells, we observed that even CD26^low/highCD8⁺ T cells with effector and effector memory phenotype showed substantial downregulation of perforin/GzmB compared to their CD26^neg counterparts in CLL patients (Additional file 1: l Fig. S3b–d). We found the same phenotype in terms of GzmB and perforin expression in CD26^low/highCD8⁺ T cells in HCs (Additional file 1: Fig. S3e–g). To better characterize the potential cytolytic role of CD26⁺CD8⁺T cells, we subjected them to GzmK expression analysis. Interestingly, we observed that CD26^highCD8⁺ T cells expressed substantial levels of intracytoplasmic GzmK compared to CD26^low and CD26^neg T cells (Fig. 3F, G). Moreover, we assessed the degranulation capacity of CD8⁺ T cells by measuring CD107a expression (Lysosomal-associated membrane protein I (LAMP-I)) [71] in response to the global stimulation with anti-CD3/CD28 antibodies. These studies revealed that CD26^negCD8⁺ T cells had higher degranulation capacity following in vitro stimulation than CD26^low/high CD8⁺ T cells (Fig. 3H, I). To characterize the functional properties of CD26⁺CD8⁺ T cells in response to stimulation, we stimulated them either via T Cell receptor (TCR) (anti-CD3/CD28) or a cytokine cocktail (IL-18 + IL-12 + IL-15) for 18 h. We found that TCR-mediated stimulation significantly increased GzmB expression in CD26^low and CD26^high CD8⁺ T cell populations, while GzmK expression levels remained unchanged (Fig. 3J–M). Of note, TCR stimulation did not change the expression levels of GzmB and GzmK in CD26^neg CD8⁺ T cells (Fig. 3J–M). However, cytokine-mediated stimulation significantly enhanced GzmB/GzmK co-expression in all T cell subsets, but was more pronounced in the CD26^highCD8⁺ T cell subset (Fig. 3J–M). The same expression pattern was observed for the upregulation of perforin in different T cell subsets after stimulation (Additional file 1: Fig. S3h, i). Overall, these observations revealed differential expression of GzmB and GzmK in CD26^±CD8⁺ T cell subsets in CLL patients. Collectively, while CD26^negCD8⁺ T cells exhibited higher GzmB and perforin expression at the baseline, CD26^high and CD26^low CD8⁺ T cells acquired a greater cytolytic molecules expression upon stimulation.

To better delineate the effector functions of CD26⁺CD8⁺ T cells, we analyzed their cytokine production capacity (e.g., IFN-\(\gamma\) and TNF-\(\alpha\)). We found that CD26^high exhibited significantly higher IFN-\(\gamma\), TNF-\(\alpha\), and IFN-γ/TNF-α expression than their CD26^neg counterparts following 5 h stimulation with anti-CD3/CD28 antibodies in vitro (Fig. 4A–D). However, stimulation of PBMCs with PMA for the same period induced higher IFN-\(\gamma\), TNF-\(\alpha\), and IFN-γ/TNF-α in CD26^neg and CD26^high compared to CD26^lowCD8⁺ T cells, while the magnitude of cytokine response was more pronounced in CD26^highCD8⁺ T cells (Additional file 1: Fig. S3j–l). Next, we stimulated PBMCs via TCR-dependent (anti-CD3/CD28) or cytokine-dependent (IL-18 + IL12 + IL-15) manners for 18 h. Interestingly, we found that CD26^highCD8⁺ T cells exhibited a greater cytokine production capacity than their other counterparts (e.g., CD26^neg and CD26^low) within the same CLL patients (Fig. 4E–H).

We also measured IL-2 expression following 5 h PMA stimulation and found that CD26^high followed by CD26^low T cells had a superior capacity for IL-2 production compared to CD26^neg CD8⁺ T cells in CLL patients (Fig. 4I, J). However, we found that the proliferative capability of different CD8⁺ T cell subsets did not correspond with their cytokine production capacity. As such, CD26^lowCD8⁺ effector T cells displayed a significantly greater proliferation compared to their CD26^neg/high siblings following 72 h TCR stimulation in vitro (Fig. 4K, L). Moreover, we assessed the expression of cytokines in different CD8⁺ T cell subsets in HCs and noted that CD26^neg and CD26^high cells had higher levels of IFN-\(\gamma\) and TNF-\(\alpha\) compared to CD26^lowCD8⁺ T cells (Additional file 1: Fig. S3m–o).

Although CTLs are the best-characterized subpopulation of CD8⁺ T cells to kill infected cells or tumor cells, CD8⁺ T cells are highly heterogeneous. For example, it has been reported that environmental cues induce transcriptional factors to differentiate CD8⁺ T cells into Tc1, Tc2, Tc17, and Tc1/Tc17 cells that can be stratified based on the surface expression of CXCR3, CCR6, and CCR [72]. Therefore, we further characterized CD26⁺ and their CD26^neg counterparts in CLL patients, which showed CD26^high cells were enriched with Tc17 (CCR4⁺CCR6⁺) and Tc1/Tc17 (CCR6⁺CXCR3⁺) CD8⁺ T cell phenotype but Tc2 (CCR4⁺CCR6⁻) CD8⁺ T cells were more abundant in CD26^neg and CD26^low T cell subsets (Additional file 1: Fig. S3p–u). In addition, we evaluated the expression of different transcriptional factors in CD26-expressing subsets. We found a higher ROR \(\gamma \delta\) expression [72] in CD26^high CD8⁺ T cells, which supports the Tc17 skewed phenotype of this T cell subset in CLL patients (Fig. 4m, n). However, we did not find any significant difference in T-bet and FOXP3 expression among CD26 subsets in CLL patients (Additional file 1: Fig. S4a, b). Interestingly, we noted a higher expression of TOX transcriptional factor in CD26^high than CD26^neg T cells in CLL patients (Additional file 1: Fig. S4c, d). Although TOX may be involved in T cell exhaustion, recent studies suggested that it is expressed by most effector memory/polyfunctional CD8⁺ T cells and not exclusively exhausted T cells in humans [73]. Therefore, our observations support this notion that TOX is not linked to exhaustion but polyfunctionality.

Overall, our observations support the heterogenous nature of CD26^highCD8⁺ T cells with a greater cytokine production capacity compared to their CD26^neg/low counterparts.

CD8⁺ T cell migratory ability is crucial for accessing tumor sites, and distinct homing receptors are involved in this process [74]. To better characterize the migratory capacity of CD26⁺CD8⁺ T cell subpopulations, we subjected them to further analysis for the expression of various homing receptors. We found that CD26^highCD8⁺ T cells had significantly higher frequency and intensity of CCR5 expression than CD26^neg and CD26^lowCD8⁺ T cells in CLL patients (Fig. 5A–C, and Additional file 1: Fig. S4e, f). We made similar observations for the proportion/intensity of CCR6 (Fig. 5D–F, Additional file 1: Fig. S4g, h), and \(\beta\) 7 Integrin in CD26^highCD8⁺ T cells (Additional file 1: Fig. S4i, j). While CCR5 directs migration along RANTES, CCL3, and MIP-1 \(\alpha /\beta\) to the secondary lymphoid organs and inflammation sites[75], CCR6 and integrin-\(\beta\) 7 traffic T cells towards mucosal tissues such as the gut in response to CCL20 and MadCAM-1 [76, 77]. In contrast to CD26^high, we found significantly a higher proportion of CCR7 expressing cells with greater intensity among CD26^lowCD8⁺ T cells (Fig. 5G–I, and Additional file 1: Fig. S4k, l), which implies these cells tend to home to lymph nodes in response to CCL19 [78]. Similar observations were made for the Common Lymphocyte Antigen (CLA), a skin homing-receptor via binding to E selectins [79], in CD26^lowCD8⁺ T cells (Fig. 5L, and Additional file 1: Fig. S4m, n). It is worth mentioning that we did not find any significant difference in the frequency/intensity of T cells expressing either CXCR3 or CXCR4 among CD26^neg/CD26^low/CD26^high CD8⁺ T cells in CLL patients (Additional file 1: Fig. S4o–r). Finally, we investigated the expression of CCR4 and found a higher advantage of CD26⁺ versus their CD26^neg CD8⁺ T cells counterparts for the expression of this skin homing receptor[75] (Additional file 1: Fig. S4s). To better delineate the migration capabilities of CD26^neg/CD26⁺CD8⁺ T cells, we performed a trans-well migration assay on PBMCs from CLL patients. We observed that CD26^neg and CD26^high exhibited equally but significantly higher migratory capacity than their CD26^low counterparts toward a general chemoattractant (Fetal bovine serum 10%) when examined after 18 h (Fig. 5M, N). However, CD26^high cells displayed an enhanced migration ability towards RANTES and IL-18 compared to CD26^neg/CD26^low CD8⁺ T cells, possibly due to higher expression of CCR5 and IL-18R \(\alpha\) (Fig. 5M, O, P). We also measured the proportion of CD69-expressing cells among different CD26^± subsets. CD69, as an early activation marker, is reported to regulate the retention of T cells from the periphery into tissues to generate tissue-resident memory T cells [80]. Interestingly, we noted that CD26^high subset was significantly enriched with CD69 expressing cells compared to CD26^neg/low CD8⁺ T cells in CLL patients (Additional file 1: Fig. S5a, b). Collectively, our results suggest that CD26^highCD8⁺ T cells have a greater migratory trait to peripheral organs such as the gut, mucosal surfaces, and inflamed tissues. In contrast, the CD26^low subset has a higher homing capacity to lymph nodes and skin.

To evaluate the survival capacity of these three subpopulations of CD8⁺ T cells, we subjected them to KLRG1 and CD127 (IL-7R \(\alpha )\) expression analysis because long-lived T cells express high levels of CD127 but low levels of KLRG1 [81]. Our studies show that CD26^lowCD8⁺ T cells have higher levels of CD127 but lower levels of KLRG1 expression in CLL patients (Fig. 6A, B). Given these observations, we next examined CD26^neg/low/high cells in terms of apoptosis. This apoptotic assay confirmed the longevity of CD26^lowCD8⁺ T cells as they showed lesser Annexin-V expression than their CD26^neg and CD26^high counterparts (Fig. 6C, D). In parallel, we measured the frequency of T stem cell memory (TSCM) [63] (CD45RA⁺CCR7⁺CD95⁺) in CD26^neg, CD26^low, and CD26^high subsets of CD8⁺ T cells. These analyses revealed that the CD26^low subset had a significantly higher proportion of TSCM cells compared to their CD26^neg/high counterparts, supporting a higher self-renewal propensity (Additional file 1: Fig. S5c).

As CD8⁺ T cells acquire terminal differentiation phenotype, they upregulate the expression of CD57 and/or CD16 along with GzmB and perforin. As a result, we observed a higher expression of CD57 and CD16 among CD26^neg CD8⁺ T cells (Fig. 6E, F, and Additional file 1: Fig. S5d). This is consistent with our previous observation of a higher GzmB and perforin expression in this subpopulation (Fig. 3A, B).

To better investigate the mechanism associated with decreased CD26 frequency in CLL patients, we treated PBMCs from HCs with plasma (10%) obtained from CLL patients. After overnight culture, we found a significant reduction in the intensity of CD26 expression (Additional file 1: Fig. S5e, f). This observation suggested the presence of potential soluble mediator (s) in reducing CD26 expression in CLL patients. To identify the potential soluble factor, we performed multiplex ELISA and quantified 20 different cytokines/chemokines (Additional file 1: l Fig. S5g, h). We tested the effects of some of the most abundant cytokines on CD26 expression, however, these cytokines exhibited no effects or increased the intensity of CD26 expression in CD8 T cells (Additional file 1: Fig. S5i–n). Next, we measured the levels of TGF-β as a potential contributing factor in the attenuation of CD26, as reported in human breast cancer [82]. Although total TGF-β levels were the same in HCs and CLL patients, we noted a significant decrease in the plasma levels of active form of TGF-β in CLL patients (Additional file 1: Fig. S5o, p). In agreement with the other report in breast cancer, we observed a significant decline in CD26 expression upon treatment with TGF-β in vitro (Additional file 1: Fig. S5q). Notably, we discovered a moderate but inverse correlation between the plasma free TGF-β level with the percentages of CD26-expressing CD8⁺ T cells in CLL patients (Additional file 1: Fig. S5r). Although TGF-β may contribute to the reduction of CD26 levels, CLL patients had significantly lower levels of this cytokine in their plasma than HCs.

Considering the apoptotic effects of Gal-9 on highly activated CD8⁺ T cells [83, 84] and our previous studies [53] [85] we hypothesized that Gal-9 might contribute to the depletion of highly polyfunctional CD26^highCD8⁺ T cells in CLL. Therefore, we found a significant elevation in the plasma Gal-9 levels in CLL patients versus HCs (Fig. 6G).

To test our hypothesis, we treated total CD8⁺ T cells from HCs with recombinant human Gal-9 (0.02 \(\mu g/ml,\) physiologically relevant to the plasma levels) for 18 h. We observed a significant decrease in the frequency of the CD26^high subset (Fig. 6H, I), which was consistent with more robust apoptosis of CD26^highCD8⁺ T cells Fig. 6J, K). Overall, we discovered that Gal-9 exhibited a more pronounced apoptotic effect on the CD26^high population than the CD26^low and CD26^neg subsets (Additional file 1: Fig. S5s). To identify the possible source of Gal-9, we cultured PBMCs from CLL and HCs overnight in vitro and subjected their culture supernatants to Gal-9 quantification. This study revealed a significantly higher Gal-9 shedding in PBMCs from CLL patients (Additional file 1: Fig. S5t). Our further analysis confirmed B-CLLs as the major source of Gal-9 when compared to their non-B-CLL counterparts (Fig. 6L). Moreover, we found significantly higher levels of intracytoplasmic Gal-9 in B-CLL compared to healthy B cells (Fig. 66M, N). These results suggest that malignant B cells are a significant source of increased Gal-9 in CLL. Also, due to the elevated levels of plasma IL-18, IL-12, and IL-15 in CLL patients (Additional file 1: Fig. S5u–w), we assessed the potential effects of these cytokines on CD26 expression. We found that this cytokine cocktail, at physiological concentration detected in the plasma, significantly enhanced apoptosis of CD26^highCD8⁺ T cells (Fig. 6O, P, and Additional file 1: Fig. S6a). Overall, these observations suggest that B-CLL cells as a major source of elevated Gal-9 in CLL plasma could contribute to the depletion of CD26^highCD8⁺ T cells. Alternatively, IL-18 + IL-12 + IL-15 may promote apoptosis of CD26^highCD8⁺ T cells in CLL. However, these cytokines individually do not impact the expression of CD26.

To investigate further the role of CD26 cells in an animal model, we measured the frequency of CD26 in CD8⁺ T cells of BALB/c mice. Surprisingly, we found that nearly 100% of CD8⁺ T cells in mice regardless of their niche expressed CD26 (e.g. thymus, blood, and spleen) (Additional file 1: Fig. S6b, c). In addition, we did not see CD26^high/low subpopulations in CD8⁺ T cells of mice, therefore, the CD26 expression pattern is completely different in mice than humans. More importantly, our observations show that CD26^highCD8⁺ T cells are enriched with MAIT cells in humans (Fig. 2); however, mice MAIT cells have a CD44^highCD62^LOW phenotype [86]. These observations demonstrate that CD26-expressing cells have a different phenotype in humans than mice.