The study included 210 chronic phase CML patients on TKI therapy. The on-treatment cohort comprised 6 newly diagnosed cases and 204 cases treated with imatinib (n = 101), nilotinib (n = 59), or dasatinib (n = 44) for treatment median durations of 4.8, 3.3, and 1.5 years, respectively. Demographic data and molecular outcomes for each molecular category are summarized in Table 1.

The study of CD26 + LSC and the molecular response was conducted on random samples during the follow-up period (Table 2). The CD26- and CD26 + stem cell populations were analyzed by MFC, as shown in Fig. 1. CD34 + /CD38-/CD26 + cells were detected in 67 (31.9%) of the 210 analyzed patients, with a median of 0.29 cells/µL (range: 0.001–890.010). As expected, all patients at diagnosis exhibited a BCR::ABL1 ratio greater than 10%, with only 1/6 (16.6%) lacking detectable CD26 + LSC. When analyzing the presence of CD26 + LSC across the different molecular response categories according to the IS, a significant decrease in the proportion of cases with CD26 + LSC was observed alongside the improvement of the molecular response (χ2, p < 0.001) (Fig. 2A).

Molecular Response and CD26 LSC in TKI treated CML patients. A. The X-axis represents the molecular response on the International Scale (IS), and the Y-axis represents the percentage of CML patients with presence (CD26 + LSC: CD45 + /CD34 + /CD38-/CD26 +) or absence (CD26-HSC: CD45 + /CD34 + /CD38-/CD26-) of LSC in each molecular category. The level of BCR::ABL1 transcripts in IS was classified as a discrete variable in: ≥ 10% (Null MR), > 1% (Minimal MR), > 0.1% (Minor MR), ≤ 0.1% (Major MR) and ≤ 0.01% (including MR4.0, MR4.5 and MR5.0). Chi Square Test p < 0.001. B. The X-axis represents CD26 + or CD26- and Y-axis the ratio BCR::ABL1/ABL1 as a continuous variable. U Mann–Whitney test for non-parametric variables p < 0.001. The green color represents CD26- HSC and the orange CD26 + LSC. CML, chronic myeloid leukemia; HSC, hematopoietic stem cell; LSC, leukemia stem cell

The same pattern was observed when analyzing CD26 + LSC in relation to BCR::ABL1/ABL1 ratio, with a significant decrease in CD26 + LSC number as transcript levels diminished (MW, p < 0.001) (Fig. 2B), showing an association between LSC burden reduction and the improvement on molecular response. When comparing the frequency of CD26 + LSC across molecular categories for the three different treatments, no significant differences were identified (Table 2). However, among patients who achieved a DMR, nilotinib showed a more pronounced reduction in CD26 + LSC compared to imatinib and dasatinib (χ2, p < 0.001) (Table 2). This greater reduction observed with this TKI was not influenced by treatment duration, as no significant differences were found when comparing the length of treatment between nilotinib and either imatinib (MW, p = 0.127) or dasatinib (MW, p = 0.063).

For patients in DMR, the analyzed cohort included 73 cases: 32 were in MR4.0, 37 in MR4.5 and 4 in MR5.0 (Md: 58.6 years, range: 24.7–85.1). Treatment distribution was as follows: 56.2% on imatinib (41/73), 24.7% on nilotinib (18/73) and 19.2% on dasatinib (14/73), with median durations of 8.2 years (range: 0.3–21.1), 5.8 years (range: 2.4–11.1), and 2.9 years (range: 0.2–14.5), respectively for each TKI (Table 2). Despite the depth of the molecular response achieved, 18.8% (6/32) of patients in MR4.0 and 19.5% (8/41) in ≥ MR4.5 showed persistence of CD26 + LSC, suggesting a potentially quiescent state for these cells.

To determine whether BCR::ABL1 transcripts levels are associated with the absolute number of CD26 + cells/µL, we performed a Kendall rank correlation test. Although the analysis revealed statistically significant differences (p = 0.046), the low correlation coefficient (r = 0.187) indicated a weak relationship between these variables, reflecting a poor predictive value (Fig. 3).

Kendall rank correlation coefficient and scatter plot between the percentage of BCR::ABL1 transcripts and CD26 + cells/µL. Both variables are transformed to the logarithmic scale for a better visualization of the data. Linear regression (black line). r: Kendall rank correlation coefficient. MR: Molecular response. ≥ MR4.5 includes all cases with ≤ 0.0032 BCR::ABL1IS

The analysis of the absolute number of CD26 + cells/µL over time, from diagnosis to 24 months of treatment, showed substantial reductions at 3 and 6 months (0.5 and 0.04 cells/µL, respectively) compared to diagnostic values (7.2 cells/µL). Follow-up at 9, 12 and ≥ 24 months showed minor fluctuations, suggesting variability in the counts of CD26 + LSC during treatment. However, these reductions were not statistically significant (KW, p = 0.102). Nevertheless, when analyzing the frequency of CD26 + LSC at different time points from diagnosis to 24 months, a significant decrease was observed (χ2, p < 0.001) (Table 3).

Levels of CD26 + LSC and BCR::ABL1 transcripts were analyzed on a small group of 16 patients from the Argentine Stop Trial (NCT05926128), which included only 12 females and 4 males, with a median age of 60.1 years (range: 49.5–85.0) at the time of treatment withdrawal. Prior to discontinuation, first-line TKI therapy consisted of 11 cases on imatinib, 2 on nilotinib, and 3 on dasatinib. All these cases were monitored during 2 years in TFR according to the AST protocol; after this period, they were analyzed in our laboratory, which means that only late relapses could be evaluated in this group. The study of LSC dynamics during follow-up revealed their reappearance in only 2 patients at different follow-up points (2.9 and 4.7 years in TFR). However, no LSC were observed in subsequent studies, suggesting that the fluctuations detected without loss of MMR do not imply loss of TFR. In this group, only 6% (1/16) of cases experienced molecular relapse after 4.2 years of follow-up (Table 4).

The 15 patients in TFR in the real-life settings comprised 6 females and 9 males, with a median age of 57.8 years (range: 24.3–87.6) at the time of treatment discontinuation. Twelve patients had received imatinib, 2 nilotinib, and 1 dasatinib prior to discontinuation. During follow-up, 3 cases showed molecular relapse between 2 to 5 months after discontinuation; however, only 1 case exhibited the presence of CD26 + LSC. The remaining 12 patients continue in TFR, and in 2 of these cases, LSC were observed without molecular relapse at 1.1 and 1.8 years into TFR (Table 4).