Lineage-specific immune checkpoint dynamics define a context-dependent regulatory framework with translational relevance in human γδ T cells

Abstract

Background

Immune checkpoint receptors (ICRs) are widely used as markers of T-cell exhaustion, yet their interpretation remains context-dependent and is poorly described in γδ T cells. Here, we investigated the dynamics, functional impact, and differentiation-associated distribution of ICRs across human γδ and αβ T-cell subsets.

Methods

Peripheral blood mononuclear cells from healthy donors were stimulated in vitro to assess ICR dynamics in γδ and αβ T-cell subsets by flow cytometry. Functional effects of PD-1 and TIM-3 blockade were evaluated through proliferation, cytokine production, and degranulation assays. In parallel, previously published single-cell RNA-sequencing datasets of tumor-infiltrating γδ and CD8 T cells were subjected to secondary bioinformatic analysis to characterize differentiation-associated ICR expression patterns and their modulation following immune checkpoint blockade (ICB) therapy.

Results

ICR expression was regulated in a receptor-, lineage- and context-dependent manner. Polyclonal stimulation induced broad ICR upregulation, whereas phosphoantigen-driven Vδ2 T cell expansion resulted in a selective profile with sustained TIM-3 expression and transient modulation of PD-1 and TIGIT. Checkpoint distribution was structured across differentiation states, with TIGIT enriched in antigen-experienced subsets, LAG-3 and TIM-3 in naive compartments, and PD-1 broadly expressed. PD-1 blockade was associated with donor-dependent trends toward increased proliferation and cytokine production, particularly under IL-15 stimulation, whereas TIM-3 inhibition showed variable effects and combined blockade did not produce a consistent additive response. Single-cell transcriptomic analyses revealed that tumor-infiltrating γδ T cells displayed heterogeneous ICR expression across differentiation states, whereas CD8 T cells showed a more progressive pattern of checkpoint expression. Because classical exhaustion-associated transcriptional and epigenetic programs were not evaluated, these patterns neither establish nor exclude the presence of exhausted γδ T-cell subsets. ICB therapy was associated with persistence or upregulation of alternative checkpoints.

Conclusions

These findings indicate that ICR expression in γδ T cells is dynamically shaped by activation, differentiation, and environmental context. However, ICR expression alone is insufficient to define or exclude a state of T-cell exhaustion.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12967-026-09004-1.

Background

T cells play a central role in antitumor immunity and represent the primary targets of current immunotherapeutic strategies. Immune checkpoint blockade (ICB), particularly targeting the PD-1/PD-L1 and CTLA-4 pathways, has transformed the treatment of multiple malignancies by restoring T-cell effector function [1]. The approval of LAG-3 inhibitors [2–4] and the ongoing clinical development of targets such as TIGIT, and TIM-3 reflect the rapid expansion of this field [5–7]. In addition, combinatorial approaches, including multi-target blockade and integration with cell-based therapies, are increasingly recognized as promising means to overcome resistance and broaden the spectrum of responsive patients [8, 9]. However, clinical responses remain heterogeneous, and resistance frequently emerges, highlighting the need to better understand the biological role of immune checkpoint receptors (ICRs).

ICRs expression has traditionally been interpreted as a hallmark of T-cell exhaustion, a dysfunctional state arising during chronic antigen exposure. However, accumulating evidence indicates that checkpoint expression is highly dynamic and can also reflect physiological activation and differentiation processes rather than irreversible dysfunction. Dissecting these distinct contexts is critical for interpreting responses to immunotherapy [10, 11]. Moreover, immune regulation is increasingly recognized as a dynamic process shaped by systemic and pathological contexts, including variations in inflammatory status and disease-associated immune remodeling [12, 13].

γδ T cells represent a distinct T-cell lineage with innate-like properties and potent cytotoxic potential against malignant cells. Unlike αβ T cells, γδ T cells recognize stress-induced ligands and phosphoantigens in an MHC-independent manner and can be rapidly activated [14]. The Vδ2 subset, which predominates in human peripheral blood, can be pharmacologically expanded using aminobisphosphonates such as zoledronic acid (ZOL), a process involving members of the butyrophilin family and is currently being explored in adoptive cell therapies [15, 16]. Despite their increasing therapeutic interest [17–24], the mechanisms regulating γδ T-cell activation, as well as the functional significance and lineage-specific dynamics of ICR expression, remain incompletely understood [25]. Recent studies have demonstrated that ICR expression in γδ T cells is influenced by the tissue and disease context and may differ between Vδ1 and Vδ2 T-cell subsets [26, 27]. Moreover, intratumoral γδ T cells displaying phenotypic features of exhaustion may retain effector functions [28], further challenging the interpretation of ICR expression as an unequivocal hallmark of functional impairment. More recently, the dynamics of ICR expression in γδ T-cell subsets have been associated with clinical responses to anti-PD-1 immunotherapy [29], but the lineage-specific regulation and functional consequences of ICR expression across γδ and αβ T-cell subsets remain to be fully elucidated.

Recent single-cell transcriptomic studies have revealed that tumor-infiltrating αβ T cells occupy multiple differentiation states rather than a linear exhaustion trajectory [30–33]. However, whether similar principles apply to γδ T cells—and how these cells respond to ICB—remains largely unknown.

Here, we systematically investigated the dynamics, functional relevance, and differentiation-associated distribution of multiple ICRs across γδ and αβ T-cell subsets. By integrating in vitro activation models, functional checkpoint blockade assays, and single-cell transcriptomic analyses of tumor-infiltrating lymphocytes (TILs), we demonstrate that ICR expression in γδ T cells follows lineage-specific regulatory programs rather than a canonical exhaustion pathway. These findings reveal distinct checkpoint circuits and suggest adaptive, compensatory responses to immunotherapy.

Methods

Isolation of peripheral blood mononuclear cells

Peripheral blood samples (n = 14) were collected from healthy voluntary donors as either whole blood or buffy coat preparations. The healthy donors included in the present study were aged 30 to 50 years.

The study was approved by the Ethical Committee (No. 25, dated 30/10/2024) and conducted in accordance with the Declaration of Helsinki. Peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation using Ficoll-Paque PLUS (Cytiva, Marlborough, Massachusetts, USA), according to the manufacturer’s instructions. Briefly, buffy coats were diluted 1:1 with phosphate-buffered saline (PBS; Sial, Rome, Italy) and layered over Ficoll-Paque PLUS, followed by centrifugation at 400 × g for 30 min at room temperature without brake. The PBMC layer was collected, washed twice in PBS, resuspended in StableCell™ RPMI-1640 containing stable glutamine and sodium bicarbonate (Sigma-Aldrich, St. Louis, Missouri, USA), supplemented with 25 mM HEPES buffer, 100 IU/mL penicillin, 0.1 mg/mL streptomycin (Euroclone, Milan, Italy), and 10% heat-inactivated fetal bovine serum (Dominique Dutscher, Bernolsheim, France), and used immediately for experiments.

Cell stimulation and immune checkpoint profiling

PBMCs (n = 8) were resuspended at a concentration of 1 × 10⁶ cells/mL and stimulated with anti-CD3/CD28 magnetic beads (Thermo Fisher Scientific, Waltham, Massachusetts, USA) at a bead-to-cell ratio of 1:1, in the presence of recombinant human interleukin-2 (IL-2; 30 U/mL; PeproTech, Cranbury, New Jersey). Cultures were maintained at 37 °C in a humidified atmosphere with 5% CO₂ for 9 days. ICR expression was evaluated at baseline (day 0) and on days 3, 5, 7, and 9 post-stimulation. To assess ICR expression in relation to T-cell phenotype, additional analyses on the same samples were performed on days 0, 3, and 6.

For the selective expansion of Vδ2 T cells, PBMCs (n = 8) were seeded at 1 × 10⁶ cells/mL and stimulated with zoledronic acid (ZOL; 2.2 μM; Mylan, Canonsburg, PA, USA), a synthetic aminobisphosphonate that induces intracellular accumulation of isopentenyl pyrophosphate (IPP) and selectively activates Vγ9 Vδ2 T cells through phosphoantigen recognition. IL-2 (100 U/mL; PeproTech) was added to support cell proliferation and survival. Cultures were maintained at 37 °C and 5% CO₂ for 12 days, with fresh medium containing IL-2 replenished every 2–3 days. No additional ZOL was added after the initial stimulation. Cell density and viability were monitored by trypan blue exclusion (Cytiva). At selected time points (days 0, 3, 6, 9, and 12), cells were harvested and processed for flow cytometric analysis.

At each time point, cells were harvested, washed, and stained with fluorochrome-conjugated monoclonal antibodies to identify Vδ1, Vδ2, CD4, and CD8 T-cell subsets, as well as to assess T-cell differentiation status (CD45RA and CD27) and the expression of selected immune checkpoint molecules (PD-1, TIM-3, LAG-3, TIGIT, and CTLA-4). ICR expression was quantified as both the percentage of positive cells and median fluorescence intensity (MFI). For each ICR, the positivity threshold was established based on the corresponding negative control. All events exhibiting fluorescence above this threshold were included in the ICR-positive gate, encompassing all positive peaks when heterogeneous or multiphasic expression patterns were observed. The MFI was then determined for the entire gated ICR-positive population, excluding negative cells.

To assess the functional impact of ICB, PBMCs (n = 3) were resuspended at 2.5 × 10⁵ cells/mL and stimulated with recombinant human interleukin-15 (IL-15; 10 ng/mL; Miltenyi Biotec, Bergisch Gladbach, Germany), either alone or in combination with IPP (3 μg/mL; Sigma-Aldrich), or alternatively with anti-CD3/CD28 magnetic beads (Thermo Fisher Scientific; 1:1 bead-to-cell ratio), in the presence of IL-2 (30 U/mL; PeproTech). Cells were cultured for up to 48 hours, followed by treatment with PD-1 and TIM-3 inhibitors (pembrolizumab and sabatolimab; both at 10 μg/mL; Selleck Chemicals, Houston, Texas, USA) for 48 hours. Cultures were then subjected to a short restimulation (4 hours) with anti-CD3/CD28 magnetic beads (1:1 bead-to-cell ratio), and monensin (10 μg/mL; Sigma-Aldrich) was added 1 hour later to inhibit cytokine secretion for intracellular staining. Finally, cells were harvested, washed, and stained with fluorochrome-conjugated monoclonal antibodies to assess Vδ1, Vδ2, CD4, and CD8 T-cell subsets, ICR expression (PD-1, TIM-3), cell proliferation (Ki67), and cytokine production (IFN-γ and TNF-α).

Cytotoxic degranulation was further examined through CD107a/b mobilization assays. To this end, IL-15 (10 ng/mL; Miltenyi Biotec)– or anti-CD3/CD28–stimulated PBMCs (n = 3) were cultured at 5 × 10⁵ cells/mL in the presence of IL-2 (30 U/mL; PeproTech) for up to 5 days. Following 48 hours of baseline culture, cells were exposed to PD-1 and TIM-3 blockade (pembrolizumab and sabatolimab; both 10 μg/mL) for an additional 48 hours. For functional readout, cells were restimulated for 4 hours with anti-CD3/CD28 magnetic beads (1:1 bead-to-cell ratio) in the presence of CD107a/b monoclonal antibodies, with monensin (10 μg/mL; Sigma-Aldrich) added after 1 hour. Cells were then harvested, washed, and stained with fluorochrome-conjugated monoclonal antibodies to identify Vδ1, Vδ2, CD4, and CD8 T-cell subsets.

Flow cytometry was performed using a BD FACSLyric™ Flow Cytometer (BD Biosciences, Franklin Lakes, New Jersey, USA), and data were analyzed with FlowJo software (version 10.10.0; FlowJo LLC, BD Biosciences) in accordance with the EJI guidelines for the use of flow cytometry and cell sorting in immunological studies [34].

A complete list of reagents and tools is provided in Table S1 and S2.

Secondary analysis of previously published single-cell RNA-sequencing data

A single-cell RNA sequencing (scRNAseq) reference data of both CD8 and γδ T cells were previously reported [35]. These datasets were generated independently of the healthy-donor experiments described in the present study. Consequently, the healthy donors and the patients represented in the scRNA-seq reference datasets were not prospectively age-matched, and the analyses were not designed as an age-matched case-control comparison. The demographic and clinical characteristics of the patients are reported in the original publication. [35]. These data were obtained using the manufacturer’s instructions “Chromium Next GEM Single Cell 3′ Reagent Kit v3.1 (Dual Index)” and the Chromium X (10x Genomics, Pleasanton, CA, USA). Briefly, samples were thawed, and to improve viability Dead Cell Removal Kit (Miltenyi Biotec) was applied. Cells were washed with calcium and magnesium free phosphate buffered saline (PBS; Sial) supplemented with 0.04% bovine serum albumin and resuspended in cold PBS 1X. Cell suspensions of 10,000 cells were obtained and loaded to yield between 2,000 and 7,000 sequenced cells. Then, following reverse transcription, cDNA was extracted from each Gel Beads in Emulsion (GEM) and library’s quantity and quality were assessed. Finally, the sequencing was performed on a HiSeq 3000 (Illumina, San Diego, CA, USA) with a depth of 50,000 read pairs per cell.

Data processing

scRNAseq data of both CD8 and γδ T cell references were recovered. For each cell type we recovered a gene ×cell matrix already integrated and normalized [35]. In addition, we also integrated their respective metadata to these count matrixes (including: UMAP coordinates, pseudotime, Time clusters, differentiations states, etc.) which were also available [35]. Then, using the Seurat package v5 (CRAN) we reconstruct a Seurat object and extract gene expression information of target genes and the differentiation states for both CD8 and γδ T cell references.

Statistical analysis and graphical representation of data

Statistical analyses were performed using R (version 4.5.2). Longitudinal changes in ICR expression were assessed using linear mixed-effects models (lme4 package), with subject ID included as a random intercept to account for repeated measurements. Depending on the analysis, time was modeled either as a continuous or categorical variable, and model selection was based on data structure and goodness-of-fit criteria.

For analyses evaluating the distribution of differentiation states within checkpoint-positive populations, subset frequencies (Tn, Tcm, Tem, and Temra) were treated as compositional data. To account for the interdependence of proportions summing to 100%, values were transformed using a centered log-ratio (CLR) transformation. Linear mixed-effects models were then fitted including time, differentiation phenotype, and their interaction (Time × Phenotype) as fixed effects.

p-values for fixed effects were estimated using Satterthwaite’s approximation (lmerTest package). Post-hoc comparisons were performed using estimated marginal means with Benjamini–Hochberg correction for multiple testing (emmeans package). Adjusted p-values < 0.05 were considered statistically significant. For visualization purposes, delta (Δ) values shown in remodeling heatmaps represent mean percentage differences relative to baseline (day 0). These values are descriptive only and were not used for statistical inference, which was based exclusively on the mixed-effects models described above.

Data visualization was performed in R. For main figures, population-level kinetics of ICR expression and MFI are shown as mean ± standard error (SE) across donors for visualization purposes. To illustrate inter-individual variability, supplementary materials include donor-level facet plots displaying percentage and MFI values over time (Figure S2-S3).

ICR gene expression differences among CD8⁺ and γδ T-cell differentiation states (Tn, Tcm, Tem, and Temra) in the scRNA-seq dataset were assessed using a Kruskal–Wallis’s test followed by Dunn’s multiple-comparisons test in GraphPad Prism (version 11).

Results

Activation-driven checkpoint dynamics across γδ and αβ T cell subsets

Time-dependent expression of ICRs was analyzed in γδ (Vδ1 and Vδ2), and αβ (CD4 and CD8) T-cell populations following anti-CD3/CD28 stimulation by evaluating both the frequency of positive cells and median fluorescence intensity (MFI) within each ICR-positive T-cell population using linear mixed-effects models (Fig. 1A, Figure S1–S2). To account for potential non-linear dynamics, we modeled time both as a continuous variable and using quadratic effects when appropriate, selecting the best-fitting model based on likelihood ratio tests. Both LAG-3 and TIM-3 expression significantly increased over time (p < 0.001 for both percentage and MFI) in γδ T cells (Vδ1 and Vδ2). However, an early rise followed by stabilization—most evident in Vδ2 T cells—was observed in TIM-3. In agreement, PD-1 and TIGIT were also upregulated in γδ T cells, while PD-1 showed an early increase in both frequency and MFI (p < 0.001), TIGIT only increases in frequency of positive cells (especially Vδ2, p < 0.001 quadratic effect). In contrast, CTLA-4 expression remained low across γδ T cells, with only minor transient increases at intermediate time points (percentage p < 0.01 quadratic effect; MFI p < 0.01). Interestingly, αβ T cells showed similar trends, with time-dependent upregulation of LAG-3, PD-1, and TIM-3 (Figure S1–S2).

Fig. 1.

Time-resolved analysis of immune checkpoint receptor dynamics in T lymphocytes. (A) Kinetics of IC expression (LAG-3, PD-1, TIM-3, TIGIT, and CTLA-4) on Vδ1, Vδ2, CD8 and CD4 T cells over time following stimulation with anti-CD3/CD28 magnetic beads. Data are shown for n = 8 donors. (B) Kinetics of checkpoint expression on Vδ2 T cells following stimulation with zoledronic acid. Data are shown for n = 8 donors. Both the percentage of positive cells and median fluorescence intensity (MFI) are reported as mean ± SE. Shaded areas indicate the range of variation among individual samples. Statistical analysis was performed using linear mixed-effects models with patients as a random effect to assess temporal trends and population differences. Representative flow cytometry histograms (both Vδ1, and Vδ2 subsets) from a single donor are shown for each condition. Negative controls are included for comparison. Abbreviations: IC, immune checkpoint

Following zoledronic acid (ZOL)–mediated in vitro expansion, Vδ2 T cells exhibited distinct temporal patterns of ICR expression (Fig. 1B, Figure S3). As expected, both TIM-3 and LAG-3 were highly expressed; while TIM-3 increased rapidly in frequency with a marked rise in MFI, LAG-3 upregulated earlier and remained elevated over time. On the other hand, PD-1 and TIGIT exhibited a transient peak at day 3–6 and a modest transient increase, respectively. Whereas CTLA-4 remained overall low. In agreement, using linear mixed-effects models (patient as a random effect), we confirmed the significant temporal variation of TIM-3 (p < 0.001), LAG-3 (p < 0.001), and PD-1 (p = 0.004) receptor frequency, while CTLA-4 and TIGIT were not significantly modulated. Intensity expression analysis showed a significant increase of TIM-3 MFI (p = 0.002) and a modest late variation of CTLA-4 MFI (p = 0.038), whereas PD-1, LAG-3, and TIGIT MFIs did not change significantly.

Checkpoint-specific and lineage-restricted remodeling of T cell differentiation architecture

Analysis of ICR expression revealed checkpoint- and lineage-specific organization of differentiation states within LAG-3⁺, PD-1⁺, TIM-3⁺, TIGIT⁺, and CTLA-4⁺ T cell populations. The relative frequencies of naive (Tn), central memory (Tcm), effector memory (Tem), and terminally differentiated effector memory (Temra) subsets were quantified within γδ (Vδ1 and Vδ2), and αβ (CD4 and CD8) compartments at day 0, day 3, and day 6 (Fig. 2A, Figure S4).

Fig. 2.

Distribution of differentiation states within ICR⁺ T cells. (A) Heatmap showing the average frequency of IC-positive cells across T cell differentiation states over time. The vertical axis represents differentiation of phenotypes, and the horizontal axis represents time points (day 0, day 3, day 6). Each panel corresponds to a major T cell lineage (Vδ1, Vδ2, CD4, and CD8). Color intensity indicates the mean percentage of cells expressing each checkpoint (LAG-3, TIM-3, PD-1, TIGIT, and CTLA-4), with lighter colors representing higher frequencies and darker colors representing lower frequencies. (B) Heatmap showing remodeling of differentiation-state distribution within ICR⁺ populations, expressed as mean percentage change (Δ) relative to baseline (day 0). The vertical axis represents differentiation of phenotypes, and the horizontal axis indicates Δ day3–day0 and Δ day6–day0. Red shades indicate an increase relative to baseline, whereas blue shades indicate a decrease. Statistical inference was performed using linear mixed-effects models assessing time×phenotype interactions, as reported in the main text. Data are shown for n = 5 subjects. Abbreviations: IC, immune checkpoint; Tn, naive; Tcm, central memory; Tem, effector memory; Temra, terminally differentiated effector memory

LAG-3⁺ and TIM-3⁺ T-cell populations showed a substantial representation of the naive subset across all αβ and γδ T-cell compartments, although the magnitude and kinetics of this distribution varied among CD4, CD8, Vδ1, and Vδ2 T cells. PD-1⁺ populations also displayed a broad differentiation profile, encompassing both naive and memory subsets across the four T-cell lineages, with an increased representation of naive cells at later time points. In contrast, TIGIT⁺ populations were predominantly composed of antigen-experienced subsets, particularly Tem cells, in CD4, CD8, and Vδ2 T-cell compartments, whereas Vδ1 T cells displayed a more heterogeneous distribution. CTLA-4⁺ populations were mainly represented by naive cells at specific time points in both αβ and γδ T-cell compartments, although distinct subset- and time-dependent patterns were observed. Overall, these findings indicate that the differentiation profiles of ICR-positive populations are checkpoint-, lineage-, and time-dependent (Fig. 2A).

To determine whether the distribution of differentiation states within ICR-positive populations changed over time, we fitted linear mixed-effects models to centered log-ratio-transformed subset frequencies, including time, differentiation phenotype, and their interaction (Time × Phenotype) as fixed effects and subject ID as a random intercept. p values were calculated using Satterthwaite’s approximation. The corresponding changes in mean subset frequencies from Day 0 to Day 3 and Day 6 are visualized in Fig. 2B. We found that LAG-3 significantly redistributed over time in CD4 (p = 0.0066), Vδ1 (p = 5.9 × 10⁻⁵), and Vδ2 T cells (p = 0.018), but not in CD8 T cells. PD-1 displayed dynamic redistribution in CD4 (p = 4.3 × 10⁻¹²), CD8 (p = 0.0013), and Vδ1 T cells (p = 0.026), but not in Vδ2 T cells, indicating a time-dependent redistribution of PD-1 across most T cell compartments during activation. TIM-3 demonstrated significant interactions in CD4 (p = 1.0 × 10⁻⁹) and Vδ1 T cells (p = 1.7 × 10⁻⁴), remained stable in CD8 T cells (p = 0.247), and showed a trend in Vδ2 T cells. TIGIT remodeling was confined to αβ T cell compartments (CD4 p = 6.7 × 10⁻⁴; CD8 p = 0.0054), whereas γδ T-cell subsets remained stable. Finally, CTLA-4 redistribution was restricted to CD4 (p = 6.7 × 10⁻⁶) and CD8 cells (p = 4.3 × 10⁻⁵), with no significant remodeling detected in Vδ1 or Vδ2 subsets, indicating limited redistribution of CTLA-4⁺ cells across γδ T-cell differentiation states during activation.

Differential effects of PD-1 and TIM-3 blockade on γδ and αβ T-cell effector functions

To assess the functional impact of ICB on T-cell effector responses, proliferation, cytokine production, and cytotoxic degranulation were evaluated in healthy PBMCs under unstimulated conditions or following stimulation with IL-15, IL-15 plus IPP or anti-CD3/CD28 beads, following treatment with pembrolizumab (anti–PD-1), sabatolimab (anti–TIM-3), or their combination (Fig. 3, Figure S5-6).

Fig. 3.

Functional impact of immune checkpoint blockade on T-cell effector responses. (A-B) Box plots showing Ki67 expression and cytokine production (IFN-γ and TNF-α) across T-cell subsets following stimulation with IL-15 or anti-CD3/CD28 beads and treatment with immune checkpoint inhibitors (pembrolizumab, sabatolimab, or their combination, dual blockade). Each data point represents an individual donor. Data are shown for n = 3 donors. (C) Box plots showing CD107a/b expression as a measure of cytotoxic degranulation across Vδ1, Vδ2, and CD8 T-cell subsets under the indicated stimulation and treatment conditions. Each data point represents an individual donor. Data are shown for n = 3 donors. Abbreviations: ICB, immune checkpoint blockade; IFN-γ, interferon gamma; TNF-α, tumor necrosis factor alpha

In γδ T cells under IL-15 stimulation, pembrolizumab treatment was associated with a donor-dependent trend toward increased Ki-67 expression. A similar trend in Ki-67 expression was observed in both Vδ1 and Vδ2 T cells following sabatolimab treatment and dual PD-1/TIM-3 blockade. In contrast, intracellular IFN-γ responses were highly heterogeneous among donors. Following pembrolizumab or sabatolimab treatment, the frequency of IFN-γ-positive cells increased in two donors but decreased in the third. Consequently, these data fail to demonstrate a consistent enhancement of intracellular IFN-γ production following PD-1 or TIM-3 blockade in γδ T cells (Fig. 3A). A comparable inter-donor variability was observed in conventional CD4 and CD8 T cells. Although the median frequency of IFN-γ-positive cells was higher under IL-15 stimulation combined with pembrolizumab or sabatolimab, individual responses remained discordant, with increases in two donors and a decrease in the other. Similarly, anti-CD3/CD28 bead-stimulated cells exhibited heterogeneous, donor-specific changes in intracellular IFN-γ upon treatment with pembrolizumab or sabatolimab, without a consistent directional effect across donors or T-cell subsets (Fig. 3B).

The functional responses observed following IL-15 stimulation were further examined in the presence of IPP (Figure S5A). Under IL-15 plus IPP stimulation, Ki-67 expression showed a donor-dependent tendency to increase following checkpoint blockade in Vδ1, CD8, and CD4 T cells. Vδ2 T cells maintained high Ki-67 levels following pembrolizumab or sabatolimab treatment, whereas responses to dual blockade varied among donors. Intracellular cytokine response was also subset- and donor-dependent. Specifically, pembrolizumab or dual blockade led to an increased frequency of IFN-γ-positive Vδ2 T cells in a subset of donors, whereas no consistent pattern emerged in Vδ1, CD8, or CD4 T cells. Similarly, a trend toward higher frequencies of TNF-α-positive cells was noted in certain donors within the Vδ1, Vδ2, and CD8 T-cell subsets following pembrolizumab or dual blockade, a finding not evident in CD4 T cells. Overall, these limited and heterogeneous donor responses precluded the identification of a consistent effect of checkpoint blockade on intracellular cytokine production.

We also examined the frequencies of cells with detectable surface PD-1 and TIM-3 across the different stimulation and treatment conditions (Figure S5B). TIM-3 was strongly induced by both IL-15-based stimulation and anti-CD3/CD28 beads, although the extent of induction differed among T-cell subsets. Conversely, PD-1 induction was most prominent following anti-CD3/CD28 stimulation, showing more limited responses to IL-15-based stimulation. As expected, surface detection of PD-1 was markedly reduced following pembrolizumab treatment, whereas detectable TIM-3 was reduced following sabatolimab treatment; dual blockade expectedly diminished the detection of both receptors. These findings are consistent with antibody-mediated target engagement and/or competitive masking of the epitopes recognized by the flow-cytometry antibodies and should not necessarily be interpreted as receptor downregulation.

Analysis of CD107a/b expression revealed substantial inter-donor variability across all T-cell subsets (Fig. 3C). Compared with the positive control, pembrolizumab enhanced degranulation in two donors within both the Vδ1 and Vδ2 T-cell subsets, whereas the third donor showed no discernible change. In CD8 T cells, an increase was observed in all three donors, although with varying magnitudes. Responses to sabatolimab remained highly heterogeneous, yielding increased, unchanged, or reduced CD107a/b expression depending on the donor and T-cell subset. Similarly, dual PD-1/TIM-3 blockade produced divergent donor-specific effects: increased degranulation was observed in a single donor within the Vδ1 and Vδ2 subsets, whereas the remaining donors exhibited responses comparable to or lower than those observed with pembrolizumab alone. In CD8 T cells, responses to dual blockade mirrored response to pembrolizumab in two donors and were lower in the third. Overall, these results highlight marked inter-individual variability and fail to provide consistent evidence of an additive effect for combined PD-1/TIM-3 blockade (Fig. 3C). Although certain donor-dependent trends emerged, these differences did not reach statistical significance, likely reflecting the limited sample size.

A single-cell transcriptomic atlas reveals/uncovers differentiation-associated checkpoint expression in γδ and CD8 T cells across tumors

To investigate the distribution of ICR expression in γδ and CD8 TILs we reanalyzed a previously published transcriptomic atlas encompassing approximately 150 human PBMC and tumor-biopsy samples, to map ICR landscapes in γδ and CD8 T-cell [35]. We first examined the expression of ICR genes—PDCD1, TIGIT, HAVCR2, CTLA4, and LAG3—across individual γδ T cells (n = 10,770) and projected their expression along the established differentiation trajectory (Fig. 4A–B). Notably, both Tcm and Tem γδ T cells displayed significantly higher expression of PDCD1, TIGIT, and LAG3 (p < 0.0001), whereas increased expression of HAVCR2 and CTLA4 was restricted to Tem γδ T cells (p < 0.0001). In parallel, we assessed the proportion of γδ T cells expressing each ICR gene (expression > 0) and observed that a large fraction of Tcm and Tem γδ T cells expressed TIGIT and LAG3. Moreover, the proportion of TIGIT⁺ γδ T cells progressively increased from the Tn to the Temra stage (Fig. 4A).

Fig. 4.

Single-cell analysis reveals heterogeneous checkpoint expression across γδ and CD8 T-cell differentiation states. (A) Violin plots of gene expression of the IC markers (PDCD1, TIGIT, HAVCR2, CTLA4 and LAG3) along the different milestones of the reference inferred trajectory of γδ T cells (left axis). In addition, the proportions of γδ T cells expressing the IC marker in green (right axis). (B) Violin plots of gene expression of the IC markers (PDCD1, TIGIT, HAVCR2, CTLA4, and LAG3) along the different milestones of the reference inferred trajectory of CD8 T cells (left axis). In addition, the proportions of CD8 T cells expressing the IC marker in green (right axis). For γδ T cells M1–M2: tn, M3–M7: tcm, M8–M13: tem, M14: Temra. For CD8 T cells M1–M2: tn, M3–M7: tcm, M8–M14: tem, M15: Temra. For both Tn: sky-blue, Tcm: yellow, Tem: red, and Temra: light brown. Abbreviations: IC, immune checkpoint

We next performed a similar analysis on CD8 T cells (n = 48,854). In contrast to γδ T cells, both Tcm and Tem CD8 T cells exhibited significantly increased expression of all five ICR genes (p < 0.0001) along the differentiation trajectory. The proportion of CD8 T cells expressing these ICRs increased steadily from Tn to Tem and subsequently declined at the Temra stage. Nevertheless, a high proportion of Tcm CD8 T cells expressed PDCD1, TIGIT, and LAG3, suggesting early acquisition of checkpoint-associated features in this lineage (Fig. 4B). These analyses map selected ICR gene expression across γδ and CD8 T-cell differentiation states. However, in the absence of systematic evaluation of classical exhaustion-associated transcriptional and epigenetic programs, these patterns should not be interpreted as definitively defining or excluding an exhaustion trajectory.

Immune checkpoint expression in γδ and CD8 TILs before and after immune checkpoint blockade therapy

In a separate secondary analysis, we leveraged the previously published scRNA-seq datasets [35] to evaluate ICR expression before and after ICB therapy (Fig. 5). We digitally extracted γδ and CD8 T cells from two published scRNAseq datasets of eleven basal cell carcinoma (BCC) and ten melanoma (MEL) samples [33, 36]. As described above, ICR marker expression was projected along the reference differentiation trajectories for both γδ and CD8 T cells.

Fig. 5.

Differentiation-associated checkpoint expression in γδ and CD8 TILs before and after immune checkpoint blockade. Violin plots of gene expression of the IC markers (PDCD1, TIGIT, HAVCR2, CTLA4 and LAG3) along the different milestones of the reference inferred trajectory of γδ T cells (left axis) and proportions of γδ T cells expressing the IC marker in green (right axis) in basal cell carcinoma (A,B) and melanoma (C,D), before and after anti-PD1 combined with anti-CTLA4 treatment. Data are shown for n = 11 BCC and 10 MEL samples. Abbreviations: IC, immune checkpoint; BCC, basal cell carcinoma; MEL, melanoma

In BCC samples prior to anti-PD-1 therapy, γδ T cells exhibited a trend toward higher ICR expression in the Tcm and Tem states (Fig. 5A). However, because only 52 γδ T cells were recovered, the estimates obtained after stratification by differentiation state were based on a limited sample size and should therefore be interpreted cautiously.

In contrast, CD8 T cells before treatment (n = 3,198) exhibited significantly increased expression of most ICR markers (PDCD1, TIGIT, CTLA4, and LAG3) in Tcm and Tem cells (p < 0.0001) (Fig. 6A). Nonetheless, only PDCD1, TIGIT, and LAG3 were expressed in a substantial proportion of Tcm and Tem CD8 T cells. As observed in the global CD8 T-cell atlas, the proportion of cells expressing HAVCR2 and CTLA4 increased from Tn to Tem and declined at the Temra stage (Fig. 6A). As expected, anti-PD-1 therapy was associated with an increase in total T-cell counts, both for γδ T cells (n = 109) (Fig. 5B) and CD8 T cells (n = 5,858) (Fig. 6B). In γδ T cells, both TIGIT and LAG3 expression increased globally, compared to pre-treatment levels. Although higher ICR expression remained enriched in Tcm and Tem γδ T cells, limited cell numbers prevented the detection of statistically significant differences. Nevertheless, the proportion of γδ T cells expressing ICR markers remained highest in Tcm and Tem states across all ICs (Fig. 5B). For CD8 T cells following anti-PD-1 treatment, a similar pattern was observed; however, only LAG3 expression increased relative to pre-treatment levels. This increase was most pronounced in Tem CD8 T cells, although both Tcm and Tem subsets showed higher proportions of LAG3⁺ cells (Fig. 6B).

Fig. 6.

Pre- and post-treatment exhausted CD8 T cells in basal cell carcinoma and melanoma. Violin plots of gene expression of the ICR markers (PDCD1, TIGIT, HAVCR2, CTLA4 and LAG3) along the different milestones of the reference inferred trajectory of CD8 T cells (left axis), together with the proportion of CD8 T cells expressing each ICR marker (green; right y-axis), before (A) and after (B) anti-PD-1 treatment in basal cell carcinoma. Violin plots of gene expression of the ICR markers (PDCD1, TIGIT, HAVCR2, CTLA4 and LAG3) along the different milestones of the reference inferred trajectory of CD8 T cells (left axis), together with the proportion of CD8 T cells expressing each ICR marker (green; right y-axis), before (C) and after (D) anti-PD-1 combined with anti-CTLA-4 treatment in melanoma. Abbreviations: ICR, immune checkpoint receptor

Finally, in melanoma samples treated with combined anti-PD-1 and anti-CTLA-4 therapy, the number of γδ T cells increased (pre-treatment n = 5; post-treatment n = 44), but remained insufficient for robust statistical analysis. Despite this limitation, post-treatment γδ T cells displayed higher ICR expression in the Tem state (Fig. 5C–D). In CD8 T cells, both before and after treatment, ICR expression was significantly elevated in Tcm and Tem subsets, although only Tem CD8 T cells consistently showed higher proportions of ICR-expressing cells. Unexpectedly, this overall profile remained largely unchanged following treatment (Fig. 6C–D).

Overall, these analyses delineate treatment-associated changes in selected ICR genes but do not ascertain whether the γδ T-cell populations acquired, maintained, or lost an exhaustion-associated program.

Discussion

ICRs are commonly regarded as a hallmark of T-cell exhaustion; however, their functional significance is increasingly recognized as context-dependent.

In this study, our data indicate that ICR expression in human γδ T cells reflects dynamic, activation-associated regulatory states rather than a fixed or terminally dysfunctional phenotype. These findings argue against interpreting checkpoint expression in isolation and support a more nuanced model in which activation, differentiation, and environmental context collectively shape ICR landscapes [37, 38].

A key aspect emerging from our data is the need to distinguish between transient activation states and chronic dysfunction programs. Our results suggest that, under the experimental conditions examined, ICR expression in γδ T cells can occur as part of physiological activation responses. However, these findings do not determine whether chronic stimulation may also induce exhaustion-associated states in specific γδ T-cell subsets. Accordingly, the applicability of the conventional exhaustion framework developed for αβ T cells to γδ T cells cannot be established from ICR-expression patterns alone [37–40].

Notably, the absence of a uniform progressive co-expression pattern among the ICRs examined, together with their heterogeneous distribution across differentiation states, indicates that checkpoint expression alone is insufficient to assign a uniform exhaustion phenotype to γδ TILs. Nevertheless, these observations neither demonstrate preserved functional competence nor exclude the presence of exhausted γδ T-cell subsets within the tumor microenvironment. Evidence that intratumoral Vδ2-negative γδ T cells displaying phenotypic features of exhaustion can retain effector functions similarly indicates that an exhaustion-associated phenotype does not necessarily correspond to complete functional impairment [28]. This interpretation aligns with emerging evidence highlighting the remarkable functional plasticity of γδ T cells across both tumor and inflammatory settings [11, 22, 41].

Our in vitro analyses further showed that ICR expression is rapidly induced upon activation in both γδ (Vδ1 and Vδ2) and αβ (both CD4 and CD8) T-cell compartments, consistent with previous reports showing that checkpoint receptors are integral components of physiological activation programs. However, distinct activation modalities generated divergent checkpoint profiles. While polyclonal stimulation induced coordinated upregulation of multiple ICRs, phosphoantigen-driven Vδ2 expansion resulted in a more selective and temporally structured pattern, characterized by sustained TIM-3 expression and transient modulation of PD-1 and TIGIT. These findings suggest that checkpoint expression is not merely a consequence of activation per se, but is shaped by the nature, strength, and duration of upstream signals. Furthermore, the distinct checkpoint profiles observed in Vδ1 and Vδ2 T cells are consistent with the possibility of subset-specific regulatory programs, consistent with their reported differential susceptibility to immunosuppressive signals within the tumor microenvironment [27].

The signaling and transcriptional mechanisms underlying these differences were not directly investigated in the present study. Nevertheless, the observed subset- and stimulus-dependent patterns raise the hypothesis that differences in signaling downstream of the γδ TCR and cytokine receptors may contribute to the differential regulation of ICR expression.

This possibility may be relevant to TIM-3, whose sustained expression in phosphoantigen-expanded Vδ2 T cells could reflect functions beyond canonical inhibition. However, the present data do not establish the signaling or functional consequences of sustained TIM-3 expression. Indeed, accumulating evidence indicates that checkpoint receptors can function as modulators of activation thresholds and immune homeostasis rather than exclusively as inhibitory brakes [42–44].

Accordingly, ICR expression might contribute to the fine-tuning of γδ T-cell activation, although dedicated mechanistic experiments will be required to determine its functional role.

The distribution of ICRs across differentiation states further supports a structured and non-uniform organization of checkpoint expression. TIGIT was preferentially enriched in antigen-experienced populations, whereas PD-1 showed a broader distribution across differentiation stages. In contrast, LAG-3, PD-1, TIM-3, and CTLA-4 showed substantial representation within phenotypically naive compartments, although their magnitude and temporal distribution varied among T-cell lineages. These patterns underscore the importance of interpreting checkpoint expression within the context of differentiation dynamics rather than as static lineage markers.

Functional analyses reinforce this context-dependent interpretation of checkpoint signaling. Specifically, the functional experiments provided exploratory evidence that checkpoint blockade may differently influence T-cell responses depending on the stimulation condition and T-cell subset. However, these assays were performed using only three healthy donors, and most comparisons did not reach statistical significance. The observed changes must therefore be interpreted as preliminary trends rather than definitive functional effects. In cetain donors and experimental conditions, PD-1 blockade was associated with trends toward increased proliferation, cytokine production, and degranulation, particularly under IL-15 stimulation.

Nevertheless, the magnitude and direction of these responses varied among donors and T-cell subsets, preventing the conclusion that PD-1 blockade consistently enhances γδ T-cell function. TIM-3 blockade showed more variable effects, and combined PD-1/TIM-3 inhibition yielded no consistent additive effect.

Given the limited statistical power, the absence of a reproducible additive response cannot establish whether the two pathways operate independently, redundantly, or hierarchically. Larger functional studies will be necessary to test these possibilities. Such behavior is consistent with models of adaptive resistance to immune checkpoint blockade, in which inhibition of one pathway leads to compensatory upregulation of alternative inhibitory receptors [45–48]. The cytokine milieu has emerged as a critical determinant of checkpoint responsiveness. IL-15 was selected because of its established ability to promote T-cell survival and effector functions, as well as its potential relevance in combination with PD-1 blockade [49–51]. However, its effects are context-dependent, and prolonged or dysregulated IL-15 signaling may also promote inflammatory and counter-regulatory responses [52]. Our short-term experiments therefore primarily reflect its acute immunostimulatory activity. These preliminary observations raise the possibility that cytokine stimulation may influence the functional response to checkpoint blockade in γδ T cells. Nevertheless, additional biological replicates and mechanistic experiments will be required before conclusions can be drawn regarding combinations of ICB with cytokine-based approaches or adoptive cell therapies.

Integration with single-cell transcriptomic data provided complementary, context-dependent information regarding checkpoint regulation in vivo.

In TILs, γδ T cells displayed heterogeneous ICR expression across differentiation states, without a uniform progressive increase in the checkpoint genes examined. This heterogeneous distribution is consistent with the tissue-specific expression of TIGIT, PD-1, TIM-3, and CD39 reported in γδ T cells from patients with ovarian cancer and suggests that the local microenvironment may contribute to shaping checkpoint-receptor profiles [26]. In contrast, CD8 T cells exhibited a more progressive increase in checkpoint expression along differentiation trajectories, consistent with established models of T-cell exhaustion [16–19]. These observations identify lineage-associated differences in ICR-expression patterns but do not establish that γδ T cells lack a canonical exhaustion trajectory. Classical exhaustion-associated transcriptional and epigenetic programs were not systematically evaluated; hence, the presence or absence of exhausted γδ T-cell states cannot be determined from these analyses.

Importantly, analysis of TILs before and after immune checkpoint blockade revealed that PD-1 inhibition was not associated with a uniform reduction in the expression of the ICR genes examined. Instead, treatment was accompanied by persistence or upregulation of alternative checkpoints, including TIGIT and LAG-3. Although this pattern may be compatible with compensatory regulation described in other T-cell contexts, the present analysis does not demonstrate a compensatory mechanism or establish a causal relationship with PD-1 blockade [45–47]. Consistent with our observations, ICR expression in circulating γδ T-cell subsets was recently shown to change dynamically during anti-PD-1 immunotherapy and to be associated with clinical response [29]. Together, these findings suggest that ICR expression in γδ T cells can be remodeled during treatment and may reflect treatment-associated immune adaptation or shifts in the representation of distinct γδ T-cell states. Prospective studies with matched functional and clinical data will be required to determine the biological and clinical relevance of these changes.

From a hypothesis-generating translational perspective, these findings may have potential relevance for the development of γδ T cell–based immunotherapies. Given their rapid activation kinetics and MHC-independent tumor recognition, γδ T cells represent attractive candidates for cellular therapies [18, 24, 53]. The rapid development of gene-engineering strategies to enhance persistence, specificity, and resistance to tumor immune escape further supports the clinical potential of γδ T-cell–based therapies [54]. Our results suggest that checkpoint expression should not be used in isolation as a surrogate for functional impairment, since it may reflect a dynamic regulatory state shaped by activation history, differentiation status, and environmental cues. Accordingly, strategies based solely on ICR expression for the selection or modulation of γδ T cells may be insufficient. However, the present study does not identify a predictive biomarker, establish a relationship with clinical outcome, or demonstrate the therapeutic efficacy of a specific intervention. Its translational relevance therefore lies primarily in generating hypotheses and identifying variables that should be considered in future clinical and experimental studies.

More broadly, our results suggest that therapeutic approaches optimized for αβ T cells may require specific evaluation before being applied to γδ T cells. Lineage-specific differences in differentiation programs, functional polarization, and checkpoint regulation [35, 55–57] highlight the need for tailored immunotherapeutic strategies. The possibility that adaptive or sequential checkpoint targeting may be more effective than fixed combinations represents a testable hypothesis for future preclinical studies rather than a conclusion of the present work.

Several limitations should be considered. The in vitro stimulation systems used here do not fully recapitulate the complexity of the tissue microenvironment, including stromal interactions, antigen heterogeneity, and metabolic constraints that are known to influence checkpoint regulation. In addition, the limited number of γδ T cells recovered from tumor samples reduced statistical power in some analyses, reflecting their relatively low abundance in certain tumor contexts.

Most importantly, functional experiments were performed using only three healthy donors. This limited number of biological replicates precluded adequately powered statistical comparisons and does not allow definitive conclusions regarding the functional effects of PD-1 or TIM-3 blockade, either alone or in combination. These results should therefore be considered exploratory and require validation in larger and independent donor cohorts. Moreover, cytokine production was assessed exclusively by intracellular staining, whereas cytokine release into culture supernatants was not measured. Therefore, we cannot determine whether reduced intracellular cytokine levels in some activated samples reflect lower cytokine production or enhanced secretion. Future studies combining intracellular cytokine staining with the parallel quantification of secreted IFN-γ and TNF-α will be required to clarify this aspect.

Additionally, the transcriptomic analyses were based on publicly available datasets that differed in tumor type, sampling strategy, and clinical context, and were not designed to establish direct associations with treatment response or patient outcome. The study did not include an independent prospective patient cohort, predictive-biomarker analysis, or direct clinical validation. Finally, the signaling pathways responsible for the observed checkpoint-expression patterns were not experimentally investigated; consequently, the proposed regulatory mechanisms remain hypothetical.

Furthermore, our transcriptomic analyses primarily focused on ICR-expression patterns and did not systematically evaluate established exhaustion-associated transcriptional programs or exhaustion-associated epigenetic features. Consequently, the present study cannot establish whether γδ T cells follow or diverge from a canonical exhaustion trajectory, nor can it exclude the existence of exhausted γδ T-cell subsets.

While our data demonstrate that ICR expression can accompany activation-associated states, they do not establish that checkpoint expression exclusively reflects activation rather than exhaustion. Therefore, the presence of dysfunctional or exhausted γδ T-cell subsets, particularly within tumor tissues cannot be excluded.

Collectively, our study underscores a descriptive and hypothesis-generating framework for interpreting immune-checkpoint expression in γδ T cells. The findings emphasize the importance of considering activation state, differentiation, lineage, and environmental context when evaluating checkpoint profiles. They further indicate that ICR expression alone is insufficient to define or exclude γδ T-cell exhaustion. Rather than establishing definitive functional or clinical effects, our results identify patterns and hypotheses that warrant validation in larger functional studies, mechanistic models, and clinically annotated patient cohorts.

Conclusions

Despite these limitations, our findings establish a framework in which checkpoint expression in γδ T cells is viewed as a dynamic and context-dependent process, providing the basis for the rational design of γδ T cell–based immunotherapies.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

ICB

Immune checkpoint blockade

HDs

Healthy donors

ICR

Immune checkpoint receptor

MFI

Median fluorescence intensity

ZOL

Zoledronic acid

Tn

Naive

Tcm

Central memory

Tem

Effector memory

Temra

Terminally differentiated effector memory

TILs

Tumor infiltrating lymphocytes

BCC

Basal cell carcinoma

MEL

Melanoma

PBMCs

Peripheral blood mononuclear cells

IL-2

Interleukin-2

IPP

Isopentenyl pyrophosphate

IL-15

Interleukin-15

scRNAseq

Single cell RNA sequencing

PBS

Phosphate buffer saline

GEM

Gel beads in emulsion

Author contributions

A.M.C., M.D.S., and J.P.C. conceptualized the manuscript. A.M.C., M.D.S., J.P.C., E.L.P., G.P., C.A., C.D., S.M., and L.D.P. conducted the experiments. AMC and JPC prepared the figures. F.D. and S.M. reviewed and provided critical feedback on the manuscript. All authors contributed to the drafting and revision process and have read and approved the final version of the manuscript.

Funding

This work was supported by grants from the European Union – NextGenerationEU through the Italian Ministry of Health, under the National Recovery and Resilience Plan (PNRR), Mission 6 Component 2 – Investment 2.1. Specifically, funding was provided for Project Code PNRR-TR1-2023-12377714, entitled “Multidisciplinary and multiomic approach to dissect the cellular network in the glioma microenvironment: translational perspective to improve patient’s management”, awarded to Francesco Dieli (CUP I73C24000220006), and for Project Code PNRR-MCNT2-2023-12378122, entitled “Improving cancer immunotherapy responsiveness by inducing tertiary lymphoid structures via the activation of innate lymphocytes (ICI-TeL)”, awarded to Serena Meraviglia (CUP I73C23000480006). This research was also funded by the European Union – NextGenerationEU through the Italian Ministry of University and Research, under the PNRR, Mission 4 Component 2 – Investment 1.3, Project PE00000019 “HEAL ITALIA” (CUP B73C22001250006).

Data availability

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to partecipate

The study was approved by the Local Ethical Committee (CEL) Palermo 1 (No. 25, dated 30/10/2024) and conducted in accordance with the Declaration of Helsinki. Informed consent was obtained from all donors involved in the study, and samples were processed in anonymized form.

Competing interests

The authors declare no competing interests.