Fifty patients diagnosed with MPN [according to the 2016 WHO classification (31)] were included in this study. Four asymptomatic individuals harboring a low CALR-mutant allelic burden with clonal hematopoiesis of indeterminate potential (CALR-mutant CHIP), from the GESUS cohort (32), were also included. The detailed baseline clinical parameters from the patient population are summarized in Table 1 . For comparison purposes, 24 age-matched healthy controls (HC) with a median age of 57 years were included. The study was approved by the local ethics committee at Zealand Region (SJ-175, SJ-452, SJ-456 and SJ-585) and conducted according to the provisions of the Declaration of Helsinki. Written informed consent was obtained from all patients and healthy volunteers prior to the beginning of the study.

Peripheral blood was collected from MPN patients, age-matched HC and young healthy volunteers. The last group was used for the in vitro studies. Bone marrow aspirates were collected from seven JAK2-mutated MPN patients (ET = 1; PV = 5; PMF = 1).

Peripheral blood mononuclear cells (PBMC) and bone marrow mononuclear cells (BMNC) were isolated from venous blood and bone marrow aspirate, respectively, by density gradient as described elsewhere (33). The cells were then used immediately or cryopreserved.

Frequencies of PLT-binding to lymphocytes, as well as the activation status of bound PLT were detected using flow cytometry, an established technique to detect these aggregates (25–27). Platelet-bound T cells were evaluated in fresh and cryopreserved PMBC from MPN patients, and age-matched HC (Panel I, Supplemental Table 1 ). A similar setup was used to compare PLT-binding in isolated PBMC and BMNC from MPN (Panel II, Supplemental Table 1 ).

Ex vivo PLT-binding to antigen-specific T cells in MPN and HC was evaluated in 10 x 10⁶ PBMC stained with CMV- and FLU-loaded MHC-multimers and analyzed by flow cytometry (Panel III, Supplemental Table 1 ). See the Supplemental Material for the detailed methodology.

Blood was collected from young healthy volunteers who had not taken any drugs known to affect PLT function, for at least 14 days prior to the study. Venous peripheral blood was collected into acid citrate dextrose solution A tubes (Greiner Bio-One) and the PLT isolation procedure was adapted from Radomski et al. (34). Detailed methodology is available in the Supplemental Material .

CD8 T cells from young healthy volunteers were stimulated with different concentrations of an anti-OKT-3 antibody or CMV-peptide. For OKT-3 stimulation, CD8 T cells were first isolated from PBMC using the MagniSort™ Human CD8 T cell Enrichment Kit (Thermofisher) following the manufacturer instructions. The cells were rested overnight, at 37°C and 5% CO₂, and stimulated for three days with high (500 ng/ml), intermediate (20 ng/ml) or low (0.5 ng/ml) concentrations of plate-coated anti-CD3 (clone: OKT-3, Thermofisher), at a cell density of 1 x 10⁶ cells/ml. For CMV-peptide activation 4 - 5 x 10⁶ PBMC/ml from HLA-A2 positive healthy volunteers were stimulated with 20 nM CMV peptide and 120 U/ml IL-2 for seven days, followed by an additional overnight restimulation. After the activation period, the stimulated cells were co-cultured with allogenic PLT and PLT-binding was analyzed by flow cytometry (OKT-3: Panel IV; CMV-MHC-multimer staining: Panel V; Supplemental Table 1 ).

OKT-3 stimulated CD8 T cells, CMV-stimulated PBMC and gp100-transduced (gp100⁺) T cells (see transduction protocol in the Supplemental Material ) were co-cultured with allogenic PLT for one hour, at a PLT to lymphocyte ratio of 100: 1. The co-cultures were washed twice to remove any unbound PLT, before proceeding. To compare the sole effect of platelet-derived molecules on T cell functionality, gp100⁺ T cells were also co-cultured with PLT supernatant (sPLT), for one hour before removal by centrifugation.

PLT-binding to antigen-specific T cells was evaluated by tetramer staining, in MPN patients and after PLT-lymphocyte co-culture. Empty loadable-MHC multimers (HLA-A*02:01; The Tetramer Shop) were loaded with HIV, CMV or FLU peptides, as described by Sanai et al. (35). See the supplemental material for the detailed protocol.

PBMC from MPN patients were stimulated with OKT-3 for five hours or five days, to compare, respectively, the cytokine release and proliferation of PLT-bound and PLT-free T cells. Cytokine release and proliferation were evaluated using intracellular staining and CellTrace^® violet (CTV) (Thermofisher), respectively. Furthermore, gp100⁺ T cells were co-cultured with a melanoma cancer cell line (FM3) in the presence or absence of PLT, and real-time tumor cytotoxic capacity of T cells was evaluated with the xCELLigence system (Agilent, USA). Detailed methodology is available in the Supplemental Material .

Statistical analyses were performed using unpaired or paired T tests for comparisons of two groups, and unpaired non-parametric Kruskal-Wallis test for comparison of more than two groups. Single and multiple linear regression were used in the correlation studies, after applying a logarithmic transformation to the frequencies of PLT-bound lymphocytes. R² and p values were calculated for each linear regression. All statistical tests were performed with a two-sided 95% confidence interval, at the 0.05 significant level, and the analyses were conducted using GraphPad software (version 8).

This study investigated the presence and frequencies of PLT-bound lymphocytes in MPN patients. The percentage of circulating PLT-bound lymphocytes in patients (N=50) and age-matched HC (N=24) was evaluated using flow cytometry. Since MPN patients display increased PLT activation, and act-PLT are thought to bind more avidly than resting PLT (36), we will focus on the act-PLT binding results.

Figure 1A shows a representative plot comparing the frequency of act-PLT-CD8 T cells aggregates in MPN and HC (gating strategy in Supplemental Figure 1 ). Significantly higher frequencies of act-PLT-bound T cells, NK cells, and CD3⁺/CD56⁺ cells were observed in the patients compared to the HC ( Figure 1B ). Furthermore, MPN presented with significantly higher percentages of act-PLT bound CD8, CD4, and DN T cells than the control group ( Figure 1C ). Similar results were obtained for the frequencies of total-PLT (tPLT)-bound lymphocytes ( Supplemental Figure 2 ). The results described above were obtained from cryopreserved samples. Therefore, the frequency of PLT-bound immune cells was also analyzed in freshly isolated and cryopreserved PBMC from MPN patients ( Supplemental Figure 3 ). This comparison showed no significant differences between the samples. Comparison of the parent immune populations revealed no biologically relevant differences in the frequencies of live cells (mean_(MPN) = 99.9% vs mean_(HC) = 99.4%; Supplemental Figure 4A ). The frequency of NK cells showed a decreasing trend (mean_(MPN) = 8% vs mean_(HC) = 14%; p = 0.095) in MPN patients, but no differences were observed in the frequencies of T cells, B cells, CD3⁺/CD56⁺ cells, or T cell subsets ( Supplemental Figure 4B ).

Analysis based on MPN diagnosis revealed that act-PLT-immune cell aggregates were elevated in advanced-stage patients (i.e., PreMF and PMF) compared to HC ( Figure 1D ), particularly the PreMF patients. Although no statistics were performed due to the small sample size. the four CALR-mutated CHIP individuals had frequencies of act-PLT -CD8 T cells and -CD3⁺/CD56⁺ cells comparable to patients. The frequencies of act-PLT- ( Supplemental Figure 5A ) and tPLT- ( Supplemental Figure 5B ) bound lymphocytes were also compared across the different mutations: CALR-mutated patients exhibited increased frequencies of PLT-bound lymphocytes compared to HC and JAK2-mutated patients, although significance was not reached for the CALR-JAK2 comparison. The two MPL-mutated patients presented with high levels in all PLT-bound populations, compared to other patient groups and HC. Lastly, triple-negative-MPN patients (n=3) exhibited frequencies PLT binding comparable to the HC population. However, a larger study population would be necessary to verify the results in MPL and triple-negative-MPN patients. Stratification according to cytoreductive therapy (CRT) revealed that patients receiving therapy other than IFNα (Not IFNα, n = 22) presented with the highest levels of act-PLT-lymphocyte aggregates ( Supplemental Figure 5C ).

Overall, these results point towards elevated frequencies of tPLT and act-PLT -bound lymphocytes, especially PLT -CD8 T cell, -CD3/CD56 cell and -NK cell aggregates, in MPN compared to the controls. Furthermore, CALR-mutated patients, patients receiving CRT other than IFNα, and patients with advanced disease tend to have the highest frequencies of these aggregates.

To further explore potential clinical factors affecting PLT-binding in MPN patients, we investigated the association between these factors and the frequencies of act-PLT bound cells. To reduce the high variability seen for small subgroups, only CALR- or JAK2- mutated MPN were included.

PLT count accounted for 19% of the variability in the frequency of act-PLT bound NK cells (R² = 0.19; p = 0.004), while also correlating with act-PLT binding to T cells (R² = 0.15; p = 0.012), CD8 T cells (R² = 0.14; p = 0.013), CD4 T cells (R² = 0.16; p = 0.009) and DN T cells (R² = 0.14; p = 0.013) ( Figure 2A ). The driver mutation strongly correlated with a higher act-PLT binding to CD3⁺/CD56⁺ (R² = 0.24; p=0.001) and CD8 T cells (R² = 0.19; p = 0.004), with JAK2-mutated patients displaying lower frequencies of aggregates compared to CALR-mutant MPN (regression models in Supplemental Table 2 ). Finally, antithrombotic therapy (ATT), CRT and diagnosis showed some degree of correlation with act-PLT-binding to lymphocytes. However, the small sample sizes within each group prevents strong conclusions.

Multivariate linear regression analyses showed that PLT count and mutation type account for over 30% of the variability seen in the frequency of act-PLT-bound CD8 T cells (R² = 0.30; p = 0.001). Interestingly, PLT count did not correlate with the frequency of act-PLT-CD8 T cell aggregates in patients with CALR mutation (R² = 0.07; p = 0.209) but had a strong association in JAK2-mutated MPN (R² = 0.37; p = 0.013) ( Figure 2B ). Separating patients by mutation and CRT showed that patients receiving IFNα revealed higher act-PLT-bound CD8 T cells in CALR- than JAK2- mutated patients (23% vs. 6%; p = 0.029) ( Figure 2C ). These results should be interpreted with caution due to the limited sample sizes in the IFNα-receiving CALR-mutated group (n = 4). Lastly, MPN patients with and without a complete hematological response showed similar levels of act-PLT immune cell aggregates ( Figure 2D ).

The regression analysis was also performed for the frequencies of tPLT-bound lymphocytes, which yielded similar results ( Supplemental Figures 6A–B ; Regression models in Supplemental Table 2 ). Interestingly, the mutant allele burden showed a significant association with tPLT-binding to CD8 T for JAK2- (R² = 0.28; p = 0.033) but not CALR- (R² = 0.02; p = 0.540) mutated patients ( Supplemental Figure 6C ). Although, this correlation did not stem from an association between the two characteristics (R² = 0.03; p = 0.265; Supplemental Figure 5D ), the allele burden was significantly higher in CALR- compared to JAK2- mutated MPN (36% vs 10%; p = 0.016; Supplemental Figure 6E ).

Taken together, our results show that JAK2 mutation strengthens the correlation between PLT count and the frequency of PLT-CD8 T cell aggregates. Additionally, JAK2 mutations also increases the correlation between the mutant allele burden and tPLT-binding to CD8 T cells.

Since the transformed cells in MPN reside in the bone marrow and PLT-bound T cells in the bone marrow may affect the local tumor-specific immune response, we analyzed the frequencies of PLT-lymphocyte aggregates in freshly isolated PBMC and BMNC from MPN patients. No significant differences in the frequencies of act-PLT-bound ( Figure 3A ) or tPLT-bound ( Figure 3B ) lymphocytes were observed between the two populations. Nevertheless, some patients displayed twice the frequencies of act-PLT-bound BMNC compared to act-PLT-bound PBMC. No differences were registered in the parent immune populations between the groups (data not shown). Therefore, our results suggest that peripheral blood and bone marrow of MPN patients display similar frequencies of PLT-immune cell aggregates.

To probe into the phenotype of PLT-bound T cells, PBMC from healthy volunteers were stimulated with different concentrations of OKT-3, co-cultured with allogenic PLT and analyzed by flow cytometry (N = 6, Figure 4 ). Figure 4A shows the mean distribution of the differentiation stages within the PLT-free and PLT-bound CD8 T cells, after low OKT-3 stimulation. PLT-bound CD8 T cells had a significantly higher frequency of memory cells (CD45RO⁺) than PLT-free CD8 T cells (26% vs 40%, p = 0.036) ( Figure 4B ), while PLT-free CD8 T cells harbored primarily a naïve (CCR7⁺/CD45RO^-) phenotype (58% vs 44%, p = 0.042). Similar trends were observed in higher dose OKT-3, as well as in unstimulated cells (data not shown). No differences were observed in the frequencies of activated CD8 T cells (CD137⁺) within the PLT-free or PLT-bound populations ( Figure 4C ).

Next, we examined the PLT capacity to bind antigen-experienced CD8 T cells. PBMC from healthy volunteers were stimulated in vitro with a CMV peptide, co-cultured with PLT and analyzed by flow cytometry (N=6) (gating strategy in Supplemental Figure 7 ). We compared the frequencies of PLT-bound naïve, non-naïve cells not expressing MHC-tetramer (non-naïve tetramer^-) and CMV-specific (tetramer⁺) CD8 T cells ( Figure 4D ). PLT-binding was highest in CMV-specific CD8 T cells (mean = 63%) compared to the non-naïve tetramer^- (mean = 42%; p = 0.026) and naïve (mean = 14%; p = 0.002) groups. Additionally, differences were observed between the naïve and the non-naïve tetramer^- populations (p = 0.0001), and a significant increase was found in PLT-binding to activated CD8 T cells compared to non-activated cells (mean of differences = 30%; p = 0.009; Figure 4E ).

Lastly, circulating PLT-bound CMV- and influenza- specific CD8 T cells were evaluated in ex vivo unstimulated cells from MPN patients and age-matched HC ( Figure 4F ). The frequency of PLT-bound virus-specific CD8 T cells (tetramer⁺) was significantly increased compared to tetramer^- CD8 T cells, in MPN (mean of differences = 13%; p = 0.0096) and HC (mean of differences = 14%; p < 0.0001) groups. PLT-binding to virus-specific CD8 T cells was also significantly increased in MPN patients compared to the HC group (MPN = 52% vs. HC = 25%; p = 0.0018).

Taken together, these results show that PLT have a clear preference to bind antigen-experienced CD8 T cells. Importantly, similar analyses in MPN patients show this feature too.

To investigate the impact of PLT-binding on CD8 T cell function, proliferation and cytokine release were evaluated in PBMC from MPN patients, after stimulation with OKT-3. Figure 5A shows a representative plot of PLT-bound and PLT-free CD8 T cell proliferation, after a 5-day OKT-3 stimulation. CD8 T cell proliferation is significantly reduced in PLT-bound compared to PLT-free cells in MPN patients (N=6, Figure 5B ), with PLT-bound cells undergoing less divisions than PLT-free CD8 T cells. Conversely, after five-hour OKT-3 stimulation, the expression of IFNγ, TNFα and CD107a was increased in PLT-bound CD8 cells compared to the PLT-free population ( Figure 5C ).

In a pilot study, we co-cultured gp100⁺-transduced T cells with PLT, before assessing the cytokine release and cytotoxicity of these aggregates against the gp100-expressing melanoma cell line FM3. Over an 80-hour period, CD8 T cells showed a reduction in cytolysis in the condition containing PLT compared to T cells alone ( Figure 6A ). Since this system cannot distinguish between PLT-bound and PLT-free within the same culture, a condition containing gp100⁺ T cells with PLT supernatant (sPLT) was added, which showed the highest cytolysis levels. Killing time 40 (KT40) and KT50 provided information on the number of hours necessary for effector cells to kill 40-50% of FM3. While KT50 for gp100⁺ T cells alone was 50.3 hours, the conditions containing PLT and sPLT registered 63.5 and 74.7 hours, respectively ( Figure 6B ). The cytokine release analysis after a 5-day co-culture with PLT and FM3 showed that PLT-bound CD8 T cells decreased the expression of IFN-γ and CD107a and increased Granzyme B, compared to PLT-free cells ( Figure 6C ).

These in vitro results point towards a decrease in the proliferation, cytokine release and cytotoxic capacity of CD8 T cells when exposed to prolonged/chronic PLT-binding.