The study was conducted as a sub-study of the Argentina Stop Trial (AST) registered in the National Registry of Health Research (RENIS: ISO02688) (12). For this prospective, single-arm, open-label, nonrandomized trial, we enrolled 46 patients with CP-CML at 7 Argentinean centers in Buenos Aires city and 3 more provinces (Buenos Aires province, Mendoza and Entre Rios). PB samples were collected once a month for the first 6 months after discontinuation of TKI, then every 2 months until month 12, and every 3 months thereafter for at least 2 years. Molecular response was assessed with RT-qPCR at 2 designated standardized laboratories (“Programme for harmonization to international scale”) (22, 23), and reported as the ratio of BCR-ABL1 to ABL1 on the International Scale. Molecular recurrence was defined as loss of major molecular response (MMR), corresponding to expression of more than 0.1% BCR-ABL1 ^IS transcripts at any time. For patients with loss of MMR, treatment was restarted with the same TKI. PB samples from healthy donors (HD) (n=10) were included for comparative purposes. All patients and HD gave written informed consent.

Twenty-three patients (50%) were females and the Sokal score was low in 22 patients (47.8%). Patients were treated with imatinib (n = 37), dasatinib (n = 4) or nilotinib (n = 5) with branded (n = 35) or generic (n = 11) TKIs. Median duration of treatment was 128 months (range 52 - 244 months) and median duration of DMR (detectable BCR-ABL1IS ≤ 0.01%, or undetectable BCR-ABL1 with 10,000 or more ABL1 transcripts, i.e. at least MR4.0 or better) was 70 months (range 30 - 147 months).

PB samples were obtained before stopping TKI and then at months 3, 12 and at any time when MMR was lost, as shown in Figure 1A . Peripheral blood mononuclear cells (PBMCs) were isolated by density-gradient centrifugation (Ficoll-Paque PLUS, GE Healthcare Life Sciences) for 30 minutes at 1,500g, followed by a washing step in physiological solution, and centrifugation at low speed (10 minutes at 1,300g). 2x10⁶ PBMCs were separated to determine NK cell subpopulations by flow cytometry, and the remaining cells were cryopreserved in heat-inactivated Fetal Bovine Serum (FBS) (Gibco, Thermo Fisher Scientific) plus 10% dimethyl sulfoxide (DMSO).

The following anti–human monoclonal antibodies (mAbs) were used in this part of the study: BV421 anti-CD56 (NCAM16.2), APC-H7 anti-CD3 (SK7), APC anti-IgG1 (X-40, MOPC-21), PE anti-IgG2b (MPC-11), PE anti-IgG2a (X-39), BB515 anti-NKp44 (p44-8), V500 anti-CD16 (3G8), PE-Cy7 anti-CD25 (M-A251), APC anti-NKG2D (1D11), FITC anti-CD57 (NK-1), APC anti-PD-1 (MIH4), PE anti-CD158a (HP-3E4), PE anti-CD158b (CH-L), PE anti-NKp30 (p30-15), PE-Cy7 anti-NKp46 (9-E2), all from BD Biosciences (San Jose, CA, USA); PE anti-NKG2A (131411; R&D Systems) and AF647 anti-NKG2C (134591; R&D Systems). PBMC were stained for 30 min at 4°C, washed twice with staining buffer (PBS with 2% FBS) and acquired in a FACSCanto II flow cytometer (BD). Flow cytometric data were analyzed using FlowJo software (Tree Star).

PBMCs from patient samples at the time of discontinuation were thawed and incubated in RPMI 1640 medium supplemented with 10% heat-inactivated FBS (GIBCO, New Zealand). 2 mM L-glutamine and 100 U/mL penicillin/streptomycin (complete RPMI, all from Gibco, Carlsbad, CA, USA) plus 200U/ml DNase (Sigma Aldrich, St. Louis, Missouri, US) for 1 hour at 37°C in a humid atmosphere containing 5% CO₂. Then PBMC were washed with PBS (Gibco, Thermo Fisher Scientific) and resuspended in complete RPMI medium. Cells were cultivated in complete RPMI Medium with 0.3 ng/ml of human rhIL-15 (PeproTech, New Jersey, US) overnight (ON) at 37°C in a humid atmosphere containing 5% CO₂.

3×10⁵ PBMC were cultured with K562 cell line (CCL-243, ATCC) with complete RPMI medium. The number of target cells was calculated based on the percentage of NK cells, so the co-cultures were performed at a ratio of 1:5 NK:K562 cells. Cells were incubated for 6 h at 37°C in 5% CO2, with anti-CD107a (clone H4A3, BD) added at the beginning of the assay and Protein Transport Inhibitor (Golgi Stop, BD) added after the first hour. Cells were harvested, stained with the viability marker FVS 510 (BD) for 15 min at RT, washed and labeled with BV421 anti-CD56, APC-H7 anti-CD3, AF647 anti-NKG2C, PE-Cy7 anti-NKp46 and PerCP-Cy5.5 anti-HLA-DR for 15 min at RT. Cells were then fixed and permeabilized using Fixation and Permeabilization kit (BD) according to manufacturer’s protocol, and then labeled with PE anti-IFNγ (4S.B3, BD Biosciences) for 30 min at 4°C. Cell acquisition and analysis were performed as detailed above. Basal degranulation and IFNγ production were determined in the absence of target cells.

Anti-HCMV IgG titer was determined quantitatively in plasma samples at the Alexander Fleming Institute Clinical Analysis Laboratory by the enzyme-linked fluorescence assay technique using the VIDAS CMV IgG kit (Biomerieux) according to the manufacturer’s instructions. Results were expressed in arbitrary units/mL (AU/mL) as negative (< 4.0 AU/mL) or positive (≥ 4.0 AU/mL) with no ambiguous range.

Analysis of molecular recurrence-free survival was performed using Graphpad Prism (GraphPad software, version 8.0.1). To compare variables between groups, the Mann Whitney test or the Wilcoxon test was performed when appropriate. Quantitative variables were dichotomized using the ROC curves, and cut-off points were established according to the highest likelihood ratio, maximizing specificity. Survival curves were compared using the log-rank test. Correlation was assessed with the Spearman test. P values < 0.05 were considered statistically significant.

Among the forty-six patients enrolled in this study, seventeen (37%) lost MMR ( Figure 1A ), resulting in a molecular relapse-free survival of 63% at 48 months ( Figure 1B ). For these patients, baseline characteristics and treatment information are shown in Figure 1C . Regarding clinical variables, only prior time in deep molecular response (DMR) and treatment duration were significantly associated with the probability of successful TFR in our cohort, as we have recently published (12). Since NK cell receptors play a key role in target recognition and transmission of activating or inhibitory signals after binding to their ligands, we thus decided to examine the expression of a large panel of NK cell receptors, both at the time of discontinuation as well as at specific subsequent time points (3 and 12 months).

We first compared NK cell proportions in AST patients and healthy donors (HD) ( Supplementary Table 1 ). The median proportion of NK cells (CD3^-CD56⁺) among lymphocytes was significantly increased in patients (p = 0.0043), together with CD16⁺ NK cells and CD57⁺ NK cells (p = 0.0046 and p = 0.0208, respectively), suggesting an immunomodulatory effect of TKIs, resulting in a more mature phenotype. On the contrary, PD-1⁺ NK cells were decreased in AST patients (p <0,0001), but with low percentages of expression ( Supplementary Figure 1 ). There were no other statistically significant differences.

To determine if the absence of the TKI could affect the expression of these markers, patient samples at the time of discontinuation were compared with samples obtained three months after treatment interruption and analyzed regardless of patient clinical status ( Supplementary Table 2 ). It should be noted that one patient relapsed at month 2 therefore we do not have the sample corresponding to month 3, thus resulting in a total of 45 patients for this time point. Ten out of 14 receptors resulted significantly changed, suggesting a very dynamic behavior of this cell compartment, with significantly increased levels of CD56^bright, NKp44 and CD25 but decreased levels of CD56^dim, CD16, CD57, NKG2D, NKG2C, NKp46 and PD-1 ( Figure 2 ).

Next, our aim was to analyze the same panel of NK receptors in relapsing and non-relapsing patients at the time of discontinuation. By comparing the mean expression values between groups, no statistically significant differences were found for any of the activating receptors (CD16, NKG2C, NKG2D, NKp30, NKp44, NKp46), the inhibitory receptor NKG2A, CD158a/b from the Killer cell immunoglobulin-like receptor (KIR) family, the interleukin-2 receptor CD25 and the maturation marker CD57 ( Table 1 ). Interestingly, although there were no statistically significant differences regarding the percentage of CD56^dim and PD-1⁺ NK cells between groups ( Figures 3A, C ), here we report that molecular relapse free survival (RFS) time was significantly longer for those patients showing higher frequencies of these subsets (Log-rank (Mantel-Cox) Test p=0.0322 and p=0.0130, respectively) ( Figures 3B, D ).

To assess how durable the immune profile could be, these comparisons were repeated with samples at 3 months after TKI cessation. Again, there were no statistically significant differences regarding the percentage of CD56dim (Figure 3E). Both NK subsets (CD56^dim and PD-1⁺) maintain significance in terms of RFS ( Figures 3F, H ) and in addition for the PD-1⁺ subset statistically significant difference was reported between mean groups ( Figure 3G ). Finally, at this time point analyses showed significant alteration only in the expression of NKp44, after TKI discontinuation. Both Mann Whitney ( Figure 3I ) and Log rank tests ( Figure 3J ) showed a significant difference at 3 months after discontinuation between relapsing and non-relapsing patients.

Considering that TKIs exert off-target effects on NK cells, and given that we have observed certain differences in the expression of several markers between the time of discontinuation and three months after, we thus wondered if the absence of the inhibitor affected relapsed and non-relapsed patients differently and if those differences could help to identify future relapses. Remarkably, several markers were down or upregulated between the time of discontinuation and the first three months off-TKI ( Table 2 ), but only CD57, NKp44, NKG2D and PD-1 showed the same tendency and significance for both groups ( Supplementary Figure 2A ). To address whether this change was maintained over time, immunophenotyping was also repeated 12 months after TKI cessation for the non-relapsing group ( Supplementary Table 3 ). Interestingly, those changes remained after one year without treatment ( Supplementary Figure 2B ) but also changes in NKp46 and NKG2C expression were observed.

After human cytomegalovirus (HCMV) infection, there is an expansion of a certain NK cell subset characterized by the expression of CD57 and NKG2C among CD56^dim NK cells (HCMV-NK) (16, 17, 21). To determine the presence of this subpopulation, HCMV status was assessed in plasma samples of TFR patients, resulting in 42 seropositive (15 relapsed and 27 non-relapsed) and 4 seronegative (2 relapsed and 2 non-relapsed) patients. Taking into account only seropositive patients, there were no statistically significant differences among HCMV-NK cells between groups ( Figure 4A ). A great dispersion of data was observed, thus to deepen the characterization of this subset, we evaluated the coexpression of two NCRs according to our cytometry panel design, NKp30 and NKp46. In order to have more reproducible results, only seropositive patients whose HCMV-NK cell compartment was 5% or higher were considered for this analysis. This is because in those patients with HCMV-NK cells greater than 5%, the analysis is more robust, as can be seen in the gating strategy from Figure 4B , compared to the ones with less than 5%, with fewer events to analyze ( Supplementary Figure 3 ). Consequently, only 17 patients met the inclusion criteria for this analysis, 7 in the relapsing group and 10 in the non-relapsing group. Although there is a tendency of higher frequencies of CD56^dimNKG2C⁺CD57⁺NKp30⁺ (NKp30+ HCMV-NK) cells and CD56^dimNKG2C⁺CD57⁺NKp46⁺ (NKp46+ HCMV-NK) cells in non-relapsing patients, no significant differences could be reported ( Figures 4C, D , respectively). Remarkably, the simultaneous coexpression of NKp30 and NKp46 on CD56^dimCD57⁺NKG2C⁺ subpopulation (NKp30+NKp46+HCMV-NK cells) was increased in non-relapsing patients ( Figure 4E ), and this subset was significantly associated with longer RFS ( Figure 4F ). At the 3- month and 12-months’ time points after TKI discontinuation there are no statistically significant changes over time (data not shown), suggesting that these subsets are not affected by the absence of TKI.

Finally, to better understand the relevance of this subpopulation, functional assays were performed using K562 cell line as target and analysis was conducted as shown in Figure 5A . No statistically significant differences were found between response groups in terms of degranulation or IFNγ production in NK cell population (data not shown). Taking into account that NKp30 and NKp46, show a very strong correlation ( Supplementary Figure 4 ) and that unfortunately we could not use all the original markers in the functional assays due to panel design and technical limitations of the cytometer, we sought to determine the functionality of CD56^dimNKG2C⁺NKp46⁺ NK cells, considering that this subset is analogous to that defined before. There were no differences between groups when analyzing the percentage of CD107a⁺ among this subpopulation ( Figures 5B, C ). However, when comparing the NKG2C⁺NKp46⁺CD107a⁺ NK cells as a percentage of CD56^dim, the non-relapsing group presented higher levels of degranulation ( Figure 5D ) and also showed better chances of survival ( Figure 5E ) (Log-rank (Mantel-Cox) Test p=0.0056). Finally, IFNγ production showed no statistically significant differences between groups ( Supplementary Figure 5 ).