Efficiency of nilotinib to target chronic phase-chronic myeloid leukaemia primary mature CD34− and immature CD34+ cells

Accumulation in target cells is an essential pharmacokinetic step of targeted therapies. Tyrosine Kinase Inhibitors (TKI) against the BCR-ABL fusion protein in Chronic Phase-Chronic Myeloid Leukaemia (CP-CML) cells constitute a unique model in terms of efficacy, specificity, and in vivo demonstration of response heterogeneity by target cells. The overall therapeutic response to nilotinib is heterogeneous with no satisfactory explanation. To better understand the patients’ response heterogeneity, we quantified nilotinib uptake by primary CP-CML cells in standardized conditions using flow cytometry, which allowed also distinguishing mature (polymorphonuclear cells) from immature (CD34+) cells. Nilotinib was undetectable in 13.3% of PMN and 40% of CD34+ cells. Moreover, in CD34+ cells, intracellular nilotinib did not completely abolish BCR-ABL activity (monitored by CrkL phosphorylation inhibition), although nilotinib accumulated well in most CD34+ cell samples. Intracellular nilotinib concentration was inversely correlated with disease burden parameters, Sokal score, and early haematologic response at day 6 ± 1 only in PMN, suggesting an intrinsic ability to limit nilotinib entry in the forms with higher tumor cell burdenat diagnosis. These findings suggest that nilotinib accumulation in CP-CML cells is influenced by individual characteristics and intra-clonal heterogeneity, and might be used for pharmacokinetic studies and to assess the therapeutic response.


Results
Measurement of nilotinib uptake by flow cytometry-based natural fluorescence detection. Like imatinib, nilotinib is a naturally fluorescent molecule under UV light 26 , thus making possible its detection by quantifying the UV fluorescence emission. Therefore, we exploited the flow cytometry method we developed to evaluate imatinib intracellular concentration 25 to monitor nilotinib intracellular accumulation. As each cell population is characterized by its own natural fluorescence (i.e. auto-fluorescence due to endogenous fluorophores that become fluorescent when excited by UV light exposure), we compared cell samples incubated or not with nilotinib and quantified nilotinib intracellular concentration as the difference between the UV fluorescence of control and of treated cells (i.e. Additional Fluorescence Units; AFU) ( Supplementary Fig. S1). In our hands, data for at least 500 target cells need to be acquired to calculate the mean fluorescence reliably. This number was reached for all samples used in this study.
We validated our approach by quantifying the UV fluorescence in K562 cells (derived from a patient with CML) at different time points (5,15,30,60, 120 and 240 min) after incubation with 1 and 5 µM nilotinib (Fig. 1A). Nilotinib intracellular concentration (mean AFU) increased rapidly and in a dose-dependent manner after only 5 min of incubation: 1.7 × 10 3 ± 0.38 and 5.0 × 10 3 ± 0.79 for 1 and 5 µM nilotinib, respectively. After 2 h of incubation with nilotinib, AFU values remained stable. On the basis of these kinetic data, we used 2 h of incubation for all the experiments described here.
Then, we compared nilotinib uptake in three CML cell lines (K562, LAMA84 and KCL22 ( Supplementary  Fig. S2). On the basis of the linear relationship (r 2 = 0.97) between nilotinib extracellular concentration and AFU values, we chose the K562 cell line for all subsequent experiments.

Correlation between fluorescence values and nilotinib intracellular concentration assessed using a physical-chemical assay.
To test the correlation between UV fluorescence emission (AFU) and intracellular nilotinib concentration (pg/cell), we analysed the correlation between AFU values measured by flow cytometry and amount of nilotinib released after lysis of a known number of cells from the same K562 cell suspension incubated with different nilotinib concentrations for 2 h (see Material and Methods). We found a significant correlation between these parameters (Spearman's rho = 0.8115; P < 0.001; Fig. 1B). Specifically, 1 pg of nilotinib per cell was equivalent to 17.8 × 10 3 AFU detected by flow cytometry. We then used this equivalence to express all data as pg of nilotinib per cell. However, in cells incubated with the lowest concentration of nilotinib (0.02 µM), we could detect nilotinib inside the cells only by flow cytometry, suggesting that this is a more sensitive technique. www.nature.com/scientificreports/ Nilotinib uptake by primary CML cells. We then used our approach to analyse primary cells from patients with chronic phase-CML (Supplementary Table S1) at diagnosis before any TKI treatment (n = 92). All analyses were done only with cells isolated from peripheral blood. Nilotinib intracellular accumulation (AFU) was dose-dependent up to 5 μM (the highest tested concentration). Moreover, starting from 1 µM, nilotinib uptake was different in lymphocytes (Ly), monocytes (Mo) and polymorphonuclear cells (PMN) from the same sample (n = 60 patients; Fig. 2A and supplementary Table S2), as we previously showed for imatinib 25 . For instance, upon incubation with 2.5 µM of nilotinib, nilotinib intracellular concentration was significantly higher in PMN than in Mo (0.31 ± 0.003 vs 0.17 ± 0.02 pg/cell; P = 0.027) and Ly (0.03 ± 0.003 vs 0.05 ± 0.01 pg/ cell; P < 0.0001). These results validated our strategy to assess in vitro nilotinib uptake by CML primary cells. Moreover, flow cytometry allowed us to identify rare cell subsets without immunoselection, on the basis of the expression of specific cell surface markers. As in CML, LSC are in the CD34 + cell compartment, we could compare the in vitro uptake of nilotinib by mature CD34 -(PMN) and immature (CD34 + ) cells from 30 patients with CML (Fig. 2B). Nilotinib uptake by CML CD34 + cells was heterogeneous among patients, and was not correlated with the uptake by PMN. Overall, after 2 h of incubation with 1 µM nilotinib, its concentration in immature CD34 + cells was significantly lower than in mature PMN cells (0.08 vs 0.14 pg/cell respectively, P = 0.019). This difference was explained mainly by the undetectable level of nilotinib in CD34 + cells from 12 (40%) patients. Conversely, we could not detect nilotinib in PMN from four (13.3%) patients (this group included also two patients with undetectable nilotinib in CD34 + cells). In the 18 patients with detectable nilotinib in CD34 + cells, we did not observe any relationship between nilotinib uptake in CD34 + cells and in PMNs. Nilotinib concentration was higher in PMN than in CD34 + cells in 12 patients, and in CD34 + cells in 6 patients.

Relationship between nilotinib uptake and in vitro BCR-ABL inhibition.
We then studied the relationship between nilotinib intracellular concentration and its targeting efficiency in primary CML cells (n = 3) by assessing the inhibition of CrkL phosphorylation (pCrkL), as a molecular target of BCR-ABL TK activity, and cell survival after 30 h of incubation with increasing nilotinib concentrations (Fig. 3A,B). CrkL phosphoryla- Relationship between nilotinib intracellular uptake and patients' characteristics. We then evaluated the relationship between nilotinib intracellular uptake before treatment, Sokal prognostic score at diagnosis (low, intermediate and high risk), and features of disease burden (leucocytosis, number and percentage of circulating CD34 + cells) measured at diagnosis and at day 6 ± 1 days after treatment initiation. We first evaluated nilotinib intracellular concentration in PMN from 28 of the 33 patients who received nilotinib as first-line treatment (Supplementary Table S3). After incubation with 1 µM nilotinib, the median intracellular concentration was 0.10 pg/cell (0-0.51). Nilotinib intracellular concentration was significantly and negatively correlated with Sokal prognostic score (P = 0.02) (Fig. 4A), percentage of CD34 + cells in peripheral blood ( Fig. 4B; P = 0.018), and number of circulating CD34 + cells/µl (P = 0.03). Leucocytosis and percentage of CD34 + cells were lower in patients with higher nilotinib uptake by PMN (P = 0.04 and 0.05, respectively). We did not find any correlation between these parameters and nilotinib intracellular concentration in CD34 + cells.
In the total population (n = 92) (Supplementary Table S1), the median nilotinib concentration in PMN was 0.17 pg/cell (0-0.77) after incubation with 1 µM. We did not find any correlation between nilotinib intracellular concentration in PMN and Sokal prognostic score, whereas nilotinib intracellular concentration was inversely Data are expressed as the mean ± standard deviation; the vertical bars indicate statistical comparisons, *P < 0.05, **P < 0.001 (B) Nilotinib intracellular amount quantification after identification by flow cytometry of immature CD34 + cells and mature PMN cells within the same sample. Nilotinib concentration was significantly lower in CD34 + than PMN cells (n = 30; P = 0.019), and was undetectable in CD34 + and PMN cells from 12 (40%) and 4 (13.3%) samples, respectively. Moreover, in the patients with undetectable nilotinib in CD34 + cells (n = 12/30 patients; Fig. 2B), the Sokal prognostic score at diagnosis was low in seven, intermediate in three, and high in two. Interestingly, in 8/12 patients (67%), BCR-ABL transcript level was higher than 1% (i.e. indicative of residual disease) after one month of TKI treatment. Conversely, among the patients with detectable nilotinib in CD34 + cells, BCR-ABL transcript level was higher than 1% in 7/18 patients (39%) after one month of TKI treatment.

Scientific Reports
Due to the absence of correlation between nilotinib uptake in CD34 + cells and in PMNs, we tried to summarize nilotinib uptake capacity by these two cell subsets in the same patient by calculating the ratio of the difference of the mean nilotinib concentration in CD34 + cells and in PMNs to the sum of the two concentrations (considered as the overall uptake) (see Methods). In the patients in whom this ratio could be evaluated (n = 30), it was significantly and inversely correlated with leucocytosis (r 2 = 0.58), percentage of CD34 + cells in peripheral blood (r 2 = 0.50), number of circulating CD34 + cells (r 2 = 0.62), and Sokal prognostic score (r 2 = 0.45).

Discussion
In oncology, the ability of a targeted therapy to reach the target cell is an essential pharmacological step for effective targeting of malignant cells. CML is a model in this field with the development of anti-BCR-ABL TKIs and the possibility of identifying clone subpopulations by flow cytometry. We have adapted the original procedure described for imatinib 25 to the detection of nilotinib, a second-generation TKI. After validating the procedure with three CML cell lines, we demonstrated that nilotinib can be detected in primary chronic phase-CML cells. In this controlled system, we observed significant differences in nilotinib uptake by PMN, monocytes and lymphocytes, with the highest values in PNM (i.e. the main cells in the CML clone). In these experiments, nilotinib uptake was essentially dependent on the intrinsic properties of the cells. The two-phase curve of in vitro nilotinib accumulation in CML cell lines and in primary cells suggests a first phase of fast, possibly active entry into the cell followed by a slower flow until the plateau that represents the system saturation. We found differences between nilotinib and imatinib 25 . Specifically, nilotinib accumulated less in KCL22 cells, and its uptake by monocytes was significantly higher than in lymphocytes in which both TKIs accumulate little, confirming www.nature.com/scientificreports/ different mechanisms of trans-membrane passage of the two TKIs 27 . These results might be partly related to the heterogeneous expression of membrane transporters by different cell types and the specific affinity of each TKI for these transporters. They could be consistent with the apparently contradictory results on the role of the influx transporter OCT1 in nilotinib intracellular penetration that was partially explained by its involvement in the early phase of penetration 28 , followed by its rapid saturation and inhibition 22 . However, the system is quite complex because nilotinib could be both a substrate and an inhibitor of the transporters ABCG2 and ABCB1 [29][30][31][32][33] , and TKIs can modify also their membrane expression. Therefore the role of ATP-binding cassette transporters in the resistance to TKI, particularly nilotinib, remains debated and difficult to study 22,27,28,33,34 . For this reason, we decided to quantify TKI intracellular accumulation as the final outcome of these complex mechanisms. Due to CML cell intra-clonal heterogeneity, and LSC resistance to TKIs, we compared the intracellular uptake of nilotinib in both CD34 + cells and PMN from the same patient. Nilotinib intracellular concentration in the immature CD34 + cell population was very variable, and it was undetectable in 40% of samples compared with 13.3% in PMN. This suggests that this immature subpopulation may be more resistant to nilotinib through its active rejection via a yet unknown mechanism due to the reported lack of involvement of the usual membrane transporters, such as ABCG2 that is strongly expressed by this subpopulation 22 . This result is consistent with the observed resistance to TKIs, including nilotinib, of CD34 + progenitor cells and CD34 + CD38 -LSC 35,36 . Unlike the interpatient heterogeneity, nilotinib accumulated homogeneously within each individual CD34 + cell sample without identification of a specific subpopulation (data not shown). This observation suggests a similar uptake by cycling and quiescent cells, or a very low/undetectable number of quiescent cells. However, as a previous study indicated that 10-20% of CD34 + cells are quiescent 35 , our flow cytometry approach should have detected them and therefore, it is likely that nilotinib penetrates into quiescent cells as well.
Furthermore, our sensitive single-cell technique showed that in some patients, nilotinib accumulates in CD34 + cells, sometimes at high level, and this should contribute to its targeting efficiency. This suggests individual patient's mechanisms of regulation of nilotinib uptake that could partly explain why in some patients, TKI treatment can eliminate also immature CML cells, leading to long-term remission after treatment withdrawal. However, the analysis of the impact of intracellular nilotinib on BCR-ABL activity showed that for similar nilotinib intra-cellular concentrations, inhibition of TK activity was partial in CD34 + cells and almost complete in PMN. This is in agreement with previous findings 36 , but we could also demonstrate that a residual activity of the adaptor protein CrkL persisted despite significant nilotinib accumulation in the target cells. This could be explained by i) a limited efficacy of nilotinib, indicating the interest of combinatorial therapies with TKI molecules that target another pathway, such as asciminib, which binds to the myristoyl pocket of ABL in BCR-ABL instead of the www.nature.com/scientificreports/ ATP-binding domain 37 ; ii) the existence in these more immature cells of BCR-ABL-independent CrkL activation pathways. For example, it has been shown that CrkL is engaged in type I interferon receptor signalling 38,39 , and is implicated in the signalling pathways of lymphoid cells (B-and T-cell receptors, IL-7…) 40,41 . However, the role of similar molecular mechanisms in CML cells is unknown. More, interestingly, CrkL is a TGF-β target in other cancer models 42 , but TGF-β is implicated in CML tumorigenicity and TKI resistance 43,44 . Finally, in this small series, we analysed the relationship between intra-cellular nilotinib concentration and tumour burden in patients with chronic phase CML who received nilotinib (n = 33) or other TKI (n = 59) as first-line treatment. We found that nilotinib concentration in CD34 + cells was not correlated with the patient characteristics. Conversely, the relationship with the therapeutic response at month 3, 6 and 12 was almost significant, possibly due to lack of statistical power explained by the high number of patients in whom CD34 + cells did not accumulate nilotinib. On the other hand, nilotinib concentration in PMN was inversely correlated with leucocytosis, the percentage of CD34 + cells, the number of CD34 + cells per microlitre of blood, and the Sokal score. As at diagnosis, almost all CD34 + cells are Ph1-positive 45 and their number correlates with the tumour mass and therapeutic response 46 , our findings indicate that nilotinib uptake by mature malignant cells is inversely correlated with the tumour mass. This observation was confirmed by the correlation of the intracellular nilotinib ratio with the Sokal score, leucocytosis and proportion of CD34 + cells in the clone. Specifically, in samples from patients with high Sokal score and tumour mass, nilotinib intracellular concentration was lower in PMN than in CD34 + cells. Overall, these results suggest that poor prognostic parameters at chronic phase-CML diagnosis are correlated with lower nilotinib uptake by mature cells in the clone that represent the vast majority of malignant cells at diagnosis. Therefore, significant uptake by this population could be an important parameter of nilotinib targeting efficiency related to the early therapeutic response. Indeed, the initial CML tumour mass is a parameter of disease aggressiveness and poor prognosis, which is partially described by the Sokal score. Overall, the initial tumour mass could be an indicator of the clone proliferative capacity and of the cellular properties that interfere with nilotinib intracellular uptake and consequently cell targeting efficacy, which is confirmed by the decrease in the circulating malignant clone after day 6 ± 1 of treatment. Thus, nilotinib penetration in PMN and not in CD34 + CML cells could play an important role in the early phases of the therapeutic response.
Given the already reported persistence of a small CD34 + cell population during treatment, the lack of correlation between nilotinib accumulation in CD34 + cells and the patient's clinical features is surprising. The sample size in our study might have been too small to investigate this relationship due the high heterogeneity of nilotinib concentration in this population. However, therapeutic targeting of immature cells might be more complex and intracellular nilotinib concentration is only one of the involved parameters, which is generally accepted but has www.nature.com/scientificreports/ never been documented. The inter-patient variability of the intrinsic properties of CML CD34 + cells remains poorly explained. Our study has some limitations. It was an in vitro study that evaluated only the intrinsic properties of cells in standardized conditions. It is likely that nilotinib uptake in vivo is not perfectly identical; however, the identification of a correlation between nilotinib concentration in PMN and patient prognostic criteria is in favour of a role of the intrinsic nilotinib uptake capacities of cells in its therapeutic efficacy. An in vivo study would be desirable, but technically difficult because of the low number of malignant cells that persist during treatment and the difficulty of distinguishing malignant cells from normal cells despite advances in immunophenotyping of immature cells using, for example, CD26 and CD93 47,48 .
In conclusion, the original procedure used to evaluate the intracellular accumulation of nilotinib allowed us to detect nilotinib in mature PMN and CD34 + cell subsets of the primary CP-CML clone in standardized conditions. For the first time, we could show the inter-individual variability and the intra-clonal heterogeneity of nilotinib accumulation, but no relationship between nilotinib intracellular concentration in the two subpopulations. We also revealed the absence of nilotinib uptake by CD34 + target cells in some patients and much more frequently (3 times more) than in PMN. Furthermore, we confirmed that BCR-ABL inhibition is partial in chronic phase-CML CD34 + cells that accumulated a significant amount of nilotinib, suggesting a lower targeting efficacy of nilotinib in this subset through yet unknown mechanisms. However, nilotinib uptake by CD34 + cells was not related to the CML characteristics, Sokal prognostic score, or clone decrease after 6 ± 1 days of therapy. Conversely, nilotinib uptake by PMN was inversely correlated with these parameters, suggesting a negative relationship between the intrinsic capacity of PMN to accumulate nilotinib and the CML clone proliferative capacity. Nilotinib accumulation in mature cells rather than in CD34 + cells might influence the early therapeutic response. Given the long-term persistence of immature cells, assessing the possible relationship between nilotinib concentration in CD34 + cells and long-term therapeutic response would require specific studies in this subpopulation. The ability of chronic phase CML cells to accumulate nilotinib is probably a crucial step in its targeting efficiency and additional studies are needed to understand the underlying mechanisms.

Cell lines and primary cells.
The BCR-ABL-positive K562, KCL22 and LAMA84 cell lines were derived from patients with CML in blast crisis. The K562 and LAMA84 cell lines were purchased from ATCC (Molsheim, France), and the KCL22 line was kindly provided by Dr V. Maguer Satta (UMR INSERM 1052 CNRS 5286 Centre Léon Bérard, Lyon). K562 and LAMA84 cells were grown in Iscove's modified Dulbecco's medium, and KCL22 cells in RPMI 1640 (Lonza, Verviers, Belgium). All cell lines were cultured in a humidified incubator at 37 °C in an atmosphere of 5% CO 2 . All experiments were done with cells in the log phase of growth.
Blood samples from patients with CML in chronic phase were collected in lithium heparin tubes at diagnosis, before any treatment (n = 92 patients). Nucleated cells were isolated by collecting the buffy coat, and erythrocytes were lysed using ammonium chloride (Stemcell Technologies, Vancouver, Canada). Cells were counted and plated at 1 × 10 6 cells per millilitre in minimal essential medium (Lonza) supplemented with 4% foetal calf serum. All experiments were carried out with fresh cells, within 24 h of sampling.
Clinical and laboratory data were collected at diagnosis for all patients, and hemogram parameters at day 6 ± 1 after NIL initiation. Residual disease levels were available at 3 (n = 72 for all patients, n = 30 for patients with NIL), 6 (n = 75 and n = 31 for all patients and patients with NIL, respectively), and 12 months (n = 70 and n = 27 for all patients and patients with NIL, respectively). Written informed consent was obtained from all patients, and the study was approved by the local ethics committee (C.P.P. Ouest V, CHU Pontchaillou, 9 Avenue Bataille Flandre-Dunkerque 35,033 Rennes Cedex 9). Nilotinib solubilization. Nilotinib (Sequoia Research Product, Pangbourne, UK) was dissolved in sterile DMSO. Stock solutions were prepared at 10 mM, aliquoted, and kept at -20 °C until use.
Flow cytometry analysis of intracellular nilotinib level. Nilotinib intracellular level was measured by flow cytometry using a BD FACSAria SORP -flow cytometer (Becton Dickinson, Le Pont de Claix, France) equipped with a Genesis G2 355 nm laser (Coherent, Orsay, France), used at a power supply of 100 mWatt. UV fluorescence was detected using a 450/50 Band Pass filter.
As nilotinib is naturally fluorescent under UV light, for its intracellular level quantification it was assumed that in a controlled system, the UV fluorescence difference between control (untreated) and cells incubated with nilotinib was directly correlated to the additional amount of fluorescent nilotinib taken up by the cell. As many cellular components have intrinsic fluorescence and each cell has a weak natural fluorescence under UV light, it was essential to pre-determine the amount of fluorescent light emitted by each cell population. Therefore, nilotinib intracellular concentration was defined as the additional fluorescence units (AFU) relative to control cells.
BD FACSDiva CS&T Research calibrated beads were used to monitor the cytometer performance each day in order to generate reproducible data. www.nature.com/scientificreports/ Correlation between nilotinib intracellular levels measured by flow cytometry and by physical-chemical assay. To evaluate the correlation between flow cytometry and high-performance liquid chromatography (HPLC; a standard analytical method) data, nilotinib-related UV fluorescence in each cell and the nilotinib amount released after lysis of a known number of cells from the same cell suspension sample were quantified. To this aim, a defined number (5 × 10 6 ) of K562 or LAMA84 cells were incubated with different concentrations of nilotinib (0.02, 0.5, 1 and 2.5 µM) at 37 °C in a saturated humidified atmosphere of 5% CO 2 for 2 h. After stopping nilotinib uptake by diluting the cell suspension with cold medium, samples were washed twice and kept on ice. After the last wash, 150 µL of cell suspension was used for measuring nilotinib intracellular level flow cytometry. The other cell fraction was used for the physical-chemical assay. Specifically, cells were counted and viability was evaluated by trypan blue exclusion. After removal of as much supernatant as possible, cell pellets were stored at -80 °C until analysis (Dr MC Gagnieu's Laboratory, Lyon, France). After cell lysis (liquid-based homogenization), nilotinib was quantified (pg/cell) by HPLC with a UV diode array detector, using three wavelengths (264, 240 and 290 nm). A spectral analysis was performed to ensure the purity of the chromatographic peaks.

ICNIL ratio calculation.
To take into account the differences in nilotinib uptake between CD34 + cells and mature cells (PNM) from the same clone, each cell type was considered to be representative of a cell compartment. Therefore, the intra-cellular NIL (ICNIL) ratio was calculated as the difference in mean intracellular nilotinib concentration between CD34 + cells and PMN divided by the sum of the mean intracellular nilotinib concentration in CD34 + cells  Statistical analyses. Statistical analyses were performed using the Stata software, version 15 (StataCorp, College Station, US). Tests were two-sided, with a type I error set at 5%. Continuous data were expressed as the mean and standard deviation or median and [interquartile range] according to their statistical distribution. The assumption of normality was assessed with the Shapiro-Wilk test. Continuous variables were compared between groups with the Student's t-test or the Mann-Whitney test, when the assumptions of the t-test were not met. Homoscedasticity was analysed using the Fisher-Snedecor's test. Categorical parameters were compared between groups using the chi-square or Fisher's exact test. Relationships between continuous variables were assessed by estimating the Pearson or Spearman correlation coefficients, according to their statistical distribution, with the Sidak's type I error correction due to multiple comparisons. Random effects models were used to measure the correlation between intracellular and extracellular nilotinib concentrations, taking into account the between and within patient variability. The normality of residuals from these models was assessed as aforementioned. When appropriate, data were logarithmically transformed to achieve the normality of dependent outcomes.