• Open Access

Sustained upregulation of effector natural killer cells in chronic myeloid leukemia after discontinuation of imatinib

Authors


To whom correspondence should be addressed.

E-mail: yoshimot@tokyo-med.ac.jp

Abstract

A number of CML patients who achieve a sustained complete molecular response (CMR) for at least 2 years during imatinib (IM) therapy can discontinue IM without relapse. With the long-term goal of developing immunological criteria for managing IM therapy in CML patients, we compared the immunophenotypic profiles of three groups of CML patients: those who received IM and had a CMR for more than two consecutive years (CMR group); patients who received IM and did not have a sustained CMR but maintained a major molecular response for more than 2 years (fluctuating CMR group); and patients with a sustained CMR for more than 6 months after IM discontinuation (STOP-IM group), together with healthy controls. The percentages of effector populations of natural killer (NK) cells, such as interferon (IFN)-γ+CD3CD56+ cells, were significantly higher in the STOP-IM and CMR groups than in the fluctuating CMR and control groups. The elevated levels of these effector NK cells were sustained for more than 3 years after IM discontinuation. In contrast, the percentages of effector memory CD8+ T cells, such as IFN-γ+ CCR7CD45RO+CD8+ cells, were significantly higher in the STOP-IM and control groups than in the CMR and fluctuating CMR groups, possibly owing to IM intake. These results suggest that the immunological activation status of NK cells contributes to CMR maintenance. Higher activation levels of effector NK cells in CML patients being treated with IM might reflect minimization of BCR-ABL1 transcript levels and therefore could be additive information for determining whether to stop IM.

The introduction of tyrosine kinase inhibitors (TKIs), including imatinib (IM),[1] dramatically changed the treatment of CML[2, 3] and led to the establishment not only of new therapeutic guidelines but also of criteria for judging the effectiveness of CML management.[4-6] Imatinib targets the BCR-ABL oncoprotein, and the drug's therapeutic effectiveness can be monitored by qualitatively and quantitatively assessing BCR-ABL1 transcript levels,[7] although the current techniques might not be sensitive enough to estimate the exact amount of minimal residual CML burden at the clinical level.[8] Moreover, CML stem cells, for example, CD34+ CML cells, carry much less BCR-ABL1 transcript than CML non-stem cells, are insensitive to IM, and are considered to be quiescent.[9-11] Nevertheless, a goal of TKI therapy for CML is to obtain a major molecular response (MMR), but hopefully a complete molecular response (CMR, also referred to as undetectable molecular response), which is defined as a ≥4.5-fold reduction in the amount of BCR-ABL1 transcript (CMR4.5) and is the optimal target point for TKI treatment.[6]

In a study of CML patients with a sustained CMR4.5 lasting 2 years before discontinuation of IM, Mahon et al.[12] found that approximately 40% of the patients maintained a CMR4.5 after stopping IM. Ross et al.[13] also reported that a subset of CML patients maintained a CMR after IM cessation; however, these patients had persistent BCR-ABL DNA, as indicated by a highly sensitive patient-specific nested quantitative PCR assay. These two studies clearly indicate that some patients with persisting residual CML cells can stop IM without apparent molecular relapse. However, we do not know how long CMR patients should take TKIs, including IM.[14] In the study by Mahon et al.[12], most of the patients who experienced a molecular relapse did so within 6 months after discontinuation of IM, and the relapsed patients showed a molecular response after restarting IM. This evidence strongly suggests that TKI therapy may contribute to minimizing BCR-ABL-positive CML cells but that eradicating CML cells is difficult. Although the current techniques are not sensitive enough to detect the BCR-ABL1 transcript,[13] other factors might also be involved in maintaining the minimization of CML cells. One of these could be immunological surveillance. We have previously shown that CML patients who sustain a CMR after IM discontinuation have higher levels of natural killer (NK) cells than do normal subjects or CMR patients under IM therapy.[15]

In the current study, we characterized and compared the functional immunophenotypic profiles of CML-CMR patients in detail with or without IM, along with the profiles of healthy volunteers. Our results suggest that higher levels of functional NK cells contribute to maintaining a CMR after stopping IM and, therefore, that monitoring both the immune system and the BCR-ABL1 transcript levels could be useful for identifying CML patients who might be candidates for discontinuation of IM.

Materials and Methods

Patient characteristics

We arbitrarily categorized 42 CML patients into three groups as follows: 22 patients who had a CMR for ≥2 consecutive years with IM (CMR group); 10 CML patients who had a CMR at least twice in the previous 2 years but did not have a sustained CMR under IM treatment (fluctuating CMR group; all patients maintained MMR: 3-log reduction); and 10 patients who discontinued IM and had a sustained CMR for ≥6 months (STOP-IM group).[14, 15] We also included 18 healthy volunteers as controls (control group). The study was approved by the institutional review board of Tokyo Medical University (nos. 1655 and 2159) and Nippon Medical School (no. 223064). Written informed consent was obtained from all participants in accordance with the Declaration of Helsinki.

All patients studied were in the chronic phase of CML and had no clinical signs of immunosuppression or apparent infection. In addition, none of the patients had been treated with TKIs other than IM or had undergone allogeneic stem cell transplantation. Although some patients had previously been treated with interferon (IFN)-α, the overall proportion of these patients was approximately 30%, and the proportions in each group were similar (P = 0.981, Table 1). Overall, no patient selection bias was apparent among these groups, except in the case of the total IM dose (Table 1). The total IM dose, which was calculated as the cumulative amount of IM taken prior to blood sample collection, was significantly higher in the CMR group than in the STOP-IM group (P = 0.030). Because all patients had a CMR, the lymphocyte fractions in the patients analyzed in this study were of non-leukemic origin.

Table 1. Characteristics of CML patients who achieved a sustained complete molecular response (CMR) during imatinib (IM) therapy
 ControlFluctuating CMRCMRSTOP-IMP-value
  1. Data are presented as means ± standard error. P-values were obtained by *one-way anova or **χ2 test.

  2. CMR group, patients who had a CMR for ≥2 consecutive years with IM; Fluctuating CMR group, patients who had a CMR at least twice in the previous 2 years but did not have a sustained CMR under IM treatment; IFN, interferon; NA, not applicable; STOP-IM group, patients who discontinued IM and had a sustained CMR for ≥6 months.

No. of people18102210 
Age, years55.3 ± 1.856.7 ± 4.556.2 ± 2.656.5 ± 4.00.987*
Sex, male/female14/48/216/64/60.147**
Sokal Category, low/intermediate/highNA7/2/119/3/06/2/10.542**
Sokal score at CML diagnosisNA0.785 ± 0.0780.674 ± 0.0190.740 ± 0.0680.219*
Imatinib daily dose, ≥400 mg/<400 mgNA9/115/76/40.295**
Total IM dose, gNA852.3 ± 88.5960.0 ± 62.1636.9 ± 113.90.030*
Duration of IM therapy, monthsNA77.2 ± 9.489.5 ± 5.567.4 ± 11.00.133*
Time to achieve CMR from diagnosis, monthsNA56.1 ± 14.558.9 ± 8.859.3 ± 21.50.988*
Prior IFN-α, yes/noNA3/76/163/70.981**

Lymphocyte activation

Freshly obtained PBMCs were separated with a Ficoll density gradient (GE Healthcare, Uppsala, Sweden) and resuspended in RPMI-1640 medium supplemented with 10% FCS. PBMCs (5 × 106 cells/mL) were then stimulated with PMA (10 ng/mL) and ionomycin (1 μg/mL; Sigma-Aldrich, St. Louis, MO, USA) for 4 h in the presence of monensin (GolgiStop, 2 μM; BD Biosciences, San Diego, CA, USA) for intracellular staining of effector molecules.

Flow cytometry

Immunophenotyping was carried out with a five-color flow cytometry panel including antibodies against the following cell surface antigens and effector molecules: CD3, CD8, CD45RO, CD56, CCR7, IFN-γ, granzyme B, and perforin (eBioscience, San Diego, CA, USA and BioLegend, San Diego, CA, USA) (Fig. S1). We also analyzed NKG2D-positive NK cells using anti-NKG2D antibody (eBioscience). For intracellular staining of IFN-γ, granzyme B, and perforin, cell surface antigens of PBMCs stimulated with PMA and ionomycin for 4 h in the presence of monensin were first stained, fixed, and permeabilized with Cytofix/Cytoperm Fixation/Permeabilization Solution (BD Biosciences) according to the manufacturer's instructions. These cells were then intracellularly stained and analyzed with the FACSCanto II (BD Biosciences) flow cytometer and CellQuest software.

Real-time quantitative RT-PCR

Molecular genetic analysis of BCR-ABL1 transcript was carried out by means of real-time quantitative RT-PCR. The molecular response was assessed at least every 3 months. An MMR was defined as a 3-log reduction in the BCR-ABL1 transcript (international scale; <0.1%), and a CMR was further confirmed as the disappearance of the BCR-ABL1 transcript in a nested quantitative PCR assay, as described previously.[14, 15]

Statistical analysis

The immunophenotyping results were statistically analyzed with GraphPad Prism 5 (GraphPad Software, La Jolla, CA, USA). The profiles were analyzed by one-way anova or the χ2-test for more than three groups and Student's t-test for two groups. Linear regression was used to assess the correlation between the percentages of IFN-γ+ NK cells and the duration times after stopping IM. The receiver–operating characteristic (ROC) curve and the area under the ROC curve were used to assess the cut-off level between two groups. A P-value of less than 0.05 was considered to indicate a statistically significant difference.

Results

Percentages of effector populations of NK cells significantly higher in STOP-IM and CMR groups than in fluctuating CMR and control groups

Those CML patients receiving IM treatment (CMR and fluctuating CMR groups) had significantly lower PBMC counts than the control or STOP-IM groups (P < 0.0001, Table 2, Fig. S2), although the STOP-IM group tended to show slightly lower PBMC counts than the control group; these results indicate that the low PBMC counts in the CMR groups might have been due to IM intake and thus it is hard to find a marker in the cell counts for discrimination between fluctuating CMR and STOP-IM groups (Figs S3,S4). Therefore, we represent subsets of lymphocytes as percentages rather than absolute cell counts.

Table 2. Immunophenotypic profiles of CML patients
 ControlFluctuating CMRCMRSTOP-IMP-value
  1. Data are presented as means of percentage (%) of total lymphocytes ± standard error. P-values were obtained by one-way anova among four groups. Bold text indicates P < 0.05. CM, central memory (CCR7+CD45RO+); CMR, complete molecular response; CMR group, patients who had a CMR for ≥2 consecutive years with imatinib (IM) therapy; Fluctuating CMR group, patients who had a CMR at least twice in the previous 2 years but did not have a sustained CMR under IM treatment; E, effector (CCR7CD45RO); EM, effector memory (CCR7CD45RO+); IFN, interferon; NK, natural killer (CD3CD56+); NKT, natural killer T (CD3+CD56+); N, naïve (CCR7+CD45RO); STOP-IM group, patients who discontinued IM and had a sustained CMR for ≥6 months.

Total no. of PBMCs (×10 6  cells/mL) 2.038 ±0.179 1.180 ±0.103 1.106 ±0.103 1.579 ±0.160 <0.0001
CD56+ (%)13.0 ± 1.211.0 ± 2.217.9 ± 2.418.9 ± 3.00.085
CD3+CD56+ (%)4.1 ± 1.03.2 ± 1.03.7 ± 0.54.3 ± 0.90.863
CD3CD56+ (%) 8.7 ± 1.1 7.8 ± 1.3 14.3 ± 2.3 14.7 ± 2.1 0.036
IFN-γ+CD56+ (%) 10.3 ± 1.2 7.6 ± 1.6 13.8 ± 2.2 17.1 ± 3.0 0.043
Granzyme B+CD56+ (%)10.6 ± 0.98.7 ± 1.913.0 ± 1.615.9 ± 2.80.077
Perforin+CD56+ (%) 10.9 ± 1.2 7.9 ± 1.3 15.6 ± 2.4 16.8 ± 3.0 0.043
IFN-γ+ NKT (%)3.6 ± 1.02.2 ± 0.92.8 ± 0.43.8 ± 0.90.554
Granzyme B+ NKT (%)3.4 ± 0.92.3 ± 0.92.5 ± 0.53.2 ± 0.90.718
Perforin+ NKT (%)3.3 ± 1.01.1 ± 0.22.2 ± 0.42.9 ± 0.90.276
IFN-γ+ NK (%) 6.7 ± 1.0 5.4 ± 0.9 11.0 ± 2.0 13.5 ± 2.2 0.021
Granzyme B+ NK (%) 7.0 ± 0.9 6.3 ± 1.1 10.3 ± 1.5 12.9 ± 2.1 0.022
Perforin+ NK (%) 7.7 ± 1.1 6.8 ± 1.2 13.2 ± 2.2 13.9 ± 2.2 0.030
CD8+ (%)28.7 ± 2.422.1 ± 2.427.8 ± 1.828.6 ± 3.10.265
IFN-γ+CD8+ (%) 21.8 ± 2.0 13.5 ± 2.2 17.8 ± 1.3 22.2 ± 3.1 0.031
Granzyme B+CD8+ (%)14.7 ± 1.510.1 ± 1.812.3 ± 1.015.7 ± 2.30.117
Perforin+CD8+ (%)10.5 ± 1.56.1 ± 0.99.3 ± 1.110.7 ± 1.40.153
N CD8+ (%)5.0 ± 0.65.3 ± 1.16.0 ± 0.84.4 ± 0.30.635
CM CD8+ (%)2.6 ± 0.32.4 ± 0.33.4 ± 0.41.9 ± 0.30.090
EM CD8+ (%)7.6 ± 0.95.1 ± 1.06.6 ± 0.68.4 ± 1.20.117
E CD8+ (%)13.2 ± 1.49.4 ± 1.811.6 ± 1.213.9 ± 2.40.318
IFN-γ+N CD8+ (%)1.3 ± 0.20.8 ± 0.11.0 ± 0.20.6 ± 0.10.102
Granzyme B+N CD8+ (%)0.3 ± 0.10.4 ± 0.10.4 ± 0.10.2 ± 0.10.375
Perforin+N CD8+ (%)0.2 ± 0.00.2 ± 0.00.2 ± 0.00.1 ± 0.00.337
IFN-γ+CM CD8+ (%)1.5 ± 0.30.8 ± 0.11.3 ± 0.20.8 ± 0.10.071
Granzyme B+CM CD8+ (%)0.7 ± 0.10.7 ± 0.10.7 ± 0.10.5 ± 0.20.673
Perforin+CM CD8+ (%)0.5 ± 0.10.5 ± 0.10.5 ± 0.10.4 ± 0.20.825
IFN-γ+EM CD8+ (%) 7.2 ± 0.9 4.2 ± 1.0 5.4 ± 0.5 7.6 ± 1.2 0.034
Granzyme B+EM CD8+ (%)3.4 ± 0.52.2 ± 0.62.5 ± 0.33.6 ± 0.60.151
Perforin+EM CD8+ (%)1.6 ± 0.40.6 ± 0.21.0 ± 0.21.6 ± 0.40.067
IFN-γ+E CD8+ (%)11.9 ± 1.47.7 ± 1.710.1 ± 0.913.4 ± 2.50.168
Granzyme B+E CD8+ (%)10.2 ± 1.36.8 ± 1.58.7 ± 1.011.6 ± 2.10.264
Perforin+E CD8+ (%)8.2 ± 1.34.7 ± 0.87.5 ± 1.07.9 ± 1.20.264

Among the four groups, significant differences were observed in the percentages of the CD3CD56+ NK cell population but not the CD3+CD56+ NKT cell population, whereas no significant difference in NK cells between the STOP-IM and CMR groups was observed, as reported previously (Fig. 1a).[15] To further examine the activation level of these lymphocytes, we carried out intracellular staining of effector molecules, such as IFN-γ, granzyme B, and perforin, after activation with PMA and ionomycin in the presence of monensin, a protein transport inhibitor. This is the most commonly used method to simply and efficiently activate many types of cells, including NK cells and CD8+ T cells, to produce effector molecules such as cytokines. Natural killer cells play important roles in innate immunity and also in linking the innate and adaptive immune responses. Their main functions are the release of cytokines and the direct cell-mediated cytotoxicity. Activated NK cells produce effector molecules such as IFN-γ, perforin, and granzyme B. IFN-γ is critically important for subsequent activation of macrophage and differentiation of naive CD4+ T cells into Th1 cells. Perforin and granzyme B are necessary for direct killing of target cells. Similar to the percentages of the NK cell population, the percentages of the IFN-γ+, granzyme B+, and perforin+ NK cell populations were significantly higher in the STOP-IM and CMR groups than in the fluctuating CMR and control groups (Fig. 1b). Among these distinct effector NK cell populations and total NK cell population, the IFN-γ+ NK cell population showed the lowest P-value. No significant differences were observed in the percentages of IFN-γ+, granzyme B+, or perforin+ NKT cells among these four groups (Fig. 1c). Note that the percentages of IFN-γ+, granzyme B+, and perforin+ NK cells in the CMR group were higher than the percentages in the fluctuating CMR group (P = 0.039, 0.051, and 0.034, respectively), and approximately half of the patients in the CMR group showed high levels of effector NK cells, which were similar to those in the STOP-IM group.

Figure 1.

Upregulation of functional natural killer (NK) cells in CML patients after stopping imatinib (IM). Immunophenotypic profiling was carried out by five-color flow cytometry analysis of samples from healthy controls and three patient groups: those who received IM and had a complete molecular response (CMR) for more than two consecutive years (CMR group); patients who did not have a sustained CMR but maintained a major molecular response for more than 2 years (fluctuating CMR group); and patients with a sustained CMR for more than 6 months after IM discontinuation (STOP-IM group). The percentages of cell populations of CD56+, CD3+CD56+ NKT, and CD3CD56+ NK cells (a) effector (interferon [IFN]-γ+, granzyme B+, and perforin+) cell populations of CD3CD56+ NK cells (b) and CD3+CD56+ NKT cells (c) were compared. Data are means of percentages of total lymphocytes ± standard error. *< 0.05. The percentage of IFN-γ+ NK cells in the CMR group is significantly higher than the percentages in the control or fluctuating CMR groups, whereas there was no difference between the STOP-IM and CMR groups (P = 0.2388). This tendency was also noted in granzyme B+ and perforin+ NK cells.

Percentages of effector populations of CD8+ T cells were significantly higher in STOP-IM and control groups than in CMR and fluctuating CMR groups

CD8+ T cells are functionally divided into four subpopulations: naïve (CCR7+CD45RO); central memory (CCR7+CD45RO+); effector memory (CCR7CD45RO+); and effector (CCR7CD45RO), according to the cell surface expression of CCR7 and CD45RO.[16] We observed no significant differences among the groups in the percentages of CD8+ T cells or in the percentages of the four CD8+ T cell subpopulations, including the effector memory population (Fig. 2a, Table 2). In contrast, the percentages of effector populations, such as IFN-γ+ CD8+ T cells, were significantly higher in the STOP-IM and control groups than in the CMR and fluctuating CMR groups (Fig. 2b); differences that were more clearly apparent were observed in the IFN-γ+ effector memory CD8+ T cell population among the four subpopulations (Fig. 2c, Table 2).

Figure 2.

Immunophenotypic profiles of CD8+ T cells in CML patients. Immunophenotypic profiling was carried out using five-color flow cytometry analysis of samples from healthy controls and and three patient groups: those who received imatinib (IM) and had a complete molecular response (CMR) for more than two consecutive years (CMR group); patients who did not have a sustained CMR but maintained a major molecular response for more than 2 years (fluctuating CMR group); and patients with a sustained CMR for more than 6 months after IM discontinuation (STOP-IM group). The percentages of cell populations of CD8+ T cells and CCR7CD45RO+ effector memory CD8+ T cells (a) effector (interferon [IFN]-γ+, granzyme B+, and perforin+) cell populations of CD8+ T cells (b) and CCR7CD45RO+ effector memory CD8+ T cells (c) were compared. Data are means of percentages of total lymphocytes ± standard error. *< 0.05.

Higher percentages of effector populations of NK cells and/or effector memory CD8+ T cells clustered in patients with a CMR after stopping IM

To estimate a cut-off value for the percentage of IFN-γ+ NK cells by which to distinguish between the STOP-IM and fluctuating CMR groups, we carried out ROC curve analysis (Fig. 3a). The area under the ROC curve was 0.9600 with 90% sensitivity (95% confidence interval [CI], 55.50–99.75%) and 90% specificity (95% CI, 55.50–99.75%). The cut-off value was thus determined to be 8.750. Furthermore, we also generated a ROC curve by using the percentages of IFN-γ+ effector memory CD8+ T cells in the STOP-IM group and in the fluctuating CMR group (Fig. 3b). The area under the ROC curve was 0.7600 with 60% sensitivity (95% CI, 26.24–87.84%) and 90% specificity (95% CI, 55.50–99.75%). The cut-off value was determined to be 6.150. Therefore, in the following analyses, we used 8.75% and 6.15% as cut-off values for the IFN-γ+ NK cells and the IFN-γ+ effector memory CD8+ T cells, respectively.

Figure 3.

Immunological characterization of CML patients after stopping imatinib (IM). (a) Differences in the percentage of interferon (IFN)-γ+ natural killer (NK) cells between patients with a sustained complete molecular response (CMR) for more than 6 months after IM discontinuation (STOP-IM group) and patients who did not have a sustained CMR but maintained a major molecular response for more than 2 years (fluctuating CMR group) were compared by receiver–operating characteristic (ROC) curve analysis. The analysis revealed that the cut-off value for IFN-γ+ NK cells was 8.75%. (b) Differences in the percentage of IFN-γ+ effector memory CD8+ T cells between the STOP-IM and fluctuating CMR groups were compared by ROC curve analysis. The analysis revealed that the cut-off value for IFN-γ+ effector memory CD8+ T cells was 6.15%. (c) The correlation between the percentages of the IFN-γ+ NK cell population and the IFN-γ+ effector memory CD8+ T cell population was analyzed for patients belonging to four different groups: healthy controls (n = 18); the fluctuating CMR group (n = 10); those who received IM and had a CMR for more than two consecutive years (CMR group; n = 22); and the STOP-IM group (n = 10). The plot was then divided into four quadrants at the 6.15% mark along the x-axis and at the 8.75% mark along the y-axis. Patients in the quadrant below both cut-offs are considered to be at high risk for relapse after stopping IM therapy. Patients in the CMR group located outside of this high-risk area are considered to be candidates for discontinuation of IM therapy.

The correlation between the percentages of the IFN-γ+ NK cell population and the IFN-γ+ effector memory CD8+ T-cell population was then analyzed (Fig. 3c). We used the cut-off values (at the 6.15% mark along the x-axis and at the 8.75% mark along the y-axis) to divide the plot into four quadrants: one quadrant in which the percentages were higher than both cut-offs; two quadrants in which the percentages were higher than one of the cut-offs; and one quadrant in which the percentages were lower than both cut-offs. Almost all the patients in the fluctuating CMR group fell into the quadrant lower than both cut-offs. In marked contrast, almost all the patients in the STOP-IM group fell outside the high-risk area, especially into the area higher than the cut-off of the IFN-γ+ NK cell population.

Time-course analyses of immunophenotypic profiles of patients in the CMR group after stopping IM

Finally, to examine how long these higher activation levels of effector NK cells such as IFN-γ+ NK cells were sustained after stopping IM, we used linear regression analysis to assess the correlation between the levels of IFN-γ+ NK cells in CMR patients who stopped IM and the time after stopping IM (Fig. 4a). The mean percentage of IFN-γ+ NK cells more than 3 years after stopping IM was sustained at approximately 13%, which was much higher than the percentages for the control and fluctuating CMR groups (6.7 ± 1.0% and 5.4 ± 0.9%, respectively, Table 2) (95% CI, 9–17%). Intriguingly, one patient sustained an IFN-γ+ NK cell level of 9.2% without relapse 111 months after stopping IM. Similar analysis of the levels of IFN-γ+ effector memory CD8+ T cells was also carried out (Fig. 4b). The mean percentage more than 3 years after stopping IM was sustained at approximately 7%, which is similar to the percentage for the control group (7.2 ± 0.9%, Table 2) (95% CI, 5–10%). These results suggest that the elevated level of IFN-γ+ NK cells was sustained for a long time in patients who did not relapse after stopping IM, and the upregulation of effector NK cells may have had some role in maintaining a CMR in CML patients with minimized BCR-ABL1 transcript levels.

Figure 4.

Sustained upregulation of effector natural killer (NK) cell population in CML patients who received imatinib (IM) and had a complete molecular response (CMR) for more than two consecutive years (CMR group) after IM discontinuation. The percentages of interferon (IFN)-γ+ NK cells (a) and IFN-γ+ effector memory CD8+ T cells (b) in patients with a sustained CMR for more than 6 months after IM discontinuation (STOP-IM group) plus patients who were in the CMR group and stopped IM were plotted along the y-axis, and the number of months after stopping IM was plotted along the x-axis for each patient. The correlation was examined by linear regression analysis; the mean value is indicated by the solid line, and the 95% confidence interval (CI) values are indicated by the dashed lines. One patient retained an IFN-γ+ NK cell level of 9.2% and an IFN-γ+ effector memory CD8+ T cell level of 5.6% 111 months after stopping IM.

Discussion

In the present study, we showed that elevated levels of effector NK cells were associated with minimization of CML cells and that sustained high levels of NK cells might be informative in determining whether a CMR can be maintained after discontinuation of IM. The population of effector NK cells (such as IFN-γ+ NK) was significantly higher in patients who maintained a CMR under IM therapy than in the fluctuating CMR group, and some of the CMR patients showed elevated effector NK cell levels, similar to the elevated levels observed in the STOP-IM patients. Recently, it has been reported that some CML patients who achieve a sustained CMR on IM can stop therapy and maintain a CMR.[12-14, 17, 18] However, even when molecular remission is sustained, CML stem cells may persist after discontinuation of IM,[13] partly because the cells are insensitive to IM.[9-11] Therefore, understanding the mechanism of drug-induced cure and determining which factors are necessary for the maintenance of residual tumor cell dormancy are important. To evaluate whether our immunological criteria is a good predictive factor for sustained CMR even after IM discontinuation, the time-course analyses of immunophenotypic profiles of patients before and after stopping IM in the CMR group are required.

Sustained elevation of effector NK cells after discontinuation of IM is considered to be linked to immunological surveillance.[19] As the elevation of effector NK cells appears to persist more than 3 years after IM discontinuation, these NK cells might be memory-type cells. Although NK cells do not rearrange the genes encoding their activating receptors, recent evidence has indicated that NK cells experience a selective education process during development, undergo a clonal-like expansion during virus infection, generate long-lived progeny, and mediate more efficacious secondary responses against previously encountered pathogens like T and B cells.[20, 21] At diagnosis, CML cells are reported to show upregulated expression of NKG2D ligand, and expression of this ligand is associated with BCR-ABL by translational regulation involving the phosphoinositide 3-kinase/mammalian target of the rapamycin pathway.[22] In addition, serum levels of MHC class I chain-related molecules (for example, one of the partners of the NKG2D activating receptor on NK and T cells) are reported to be elevated in untreated CML patients, whereas NKG2D expression on NK and CD8+ T cells is downregulated, owing to immune escape; IM therapy may restore NKG2D-mediated NK cytotoxic activity in CML patients.[22] This suggests that upregulation of effector NK cells observed in the current study may be the result of minimization of CML cells, and the maintained elevation of effector NK cells after stopping IM may be linked to a sustained CMR. Although we did not observe any significant differences in the NKG2D+ NK population among the fluctuating CMR, CMR, and STOP-IM groups (P = 0.8404, data not shown), the continuous upregulation of effector NK cells may have some immunological effects on CML cells.

Imatinib is a specific inhibitor of tyrosine kinase receptors, not only BCL-ABL but also KIT, that are required for the malignant transformation of stromal cells of the gut in gastrointestinal stromal tumors (GISTs). As IM enables disease control, including objective responses and stable disease in >80% of GIST patients,[23] IM has also became the standard treatment of advanced GISTs. However, several lines of evidence indicated that IM might mediate antitumor effects by an alternate mode of action instead of having a direct effect on tumoral c-kit mutations in GISTs.[24-26] Intriguingly, it was previously reported that IM acts on host dendritic cells to promote NK cell activation and NK cell-dependent antitumor effects in mice.[25] In addition, most GIST-bearing patients that were treated with IM acquired NK cell activation, which positively correlated with clinical outcome,[25] and the IFN-γ production level of NK cells after 2 months of treatment is considered a possible independent predictor of long survival in advanced GISTs treated with IM.[26] Whether such IM-mediated off-target effects trigger NK cell activation in CML patients remains to be clarified.

It has also been shown that, in a small proportion of patients, IFN-α therapy can induce complete cytogenetic remission with prolonged survival without disease relapse after therapy is stopped.[27, 28] Patients who discontinue IFN-α therapy but remain in remission for more than 2 years reportedly show significantly higher NK cell levels than either patients still being treated with IFN-α or controls.[29] The immunophenotypic profiles of CML patients treated with IFN-α strongly agree with our results for patients treated with IM. Although IFN-α therapy has been suggested as a way to increase the likelihood that IM therapy can be safely stopped,[12] we observed no significant differences in the proportions of patients who previously used IFN-α among our three CML patient groups (P = 0.981, Table 1).

Effector memory (CCR7CD45RO) CD8+ T cells mediate protective memory by migrating to inflamed peripheral tissues and showing immediate effector function. In contrast, central memory CD8+ T cells mediate reactive memory by homing to T-cell areas of secondary lymphoid organs with little or no effector function, and readily proliferate and differentiate to effector cells in response to antigenic stimulation.[16] In CML, CD8+ T cells reportedly recognize leukemic cells,[30, 31] and autologous activated NK cells have been shown to efficiently suppress the growth of malignant hematopoietic progenitors in vitro.[32] Some patients with CML and Philadelphia-positive acute lymphoblastic leukemia who were treated with the second-generation TKI dasatinib showed elevation of circulating large granular lymphocytes and developed numerical expansion of NK cells, CD8+ T cells, or both, which is associated with superior therapeutic responses.[33, 34] In contrast, dasatinib has also been shown to suppress NK cell cytotoxicity.[35, 36] Therefore, the dissociation between the immune reactivation in vivo and immunosuppressive effects in vitro might exist in dasatinib.[33, 37] Salih et al.[38] showed that dasatinib impairs expression of NKG2D ligand on CML cells and reduces the production of IFN-γ and the cytotoxicity of NK cells derived from CML patients, whereas IM has no direct influence on NK cell activity. These results indicate that understanding of the immunological profiles of patients treated with TKIs is important and that caution should be exercised in stopping treatment with these agents just after a CMR has been sustained for 2 years. Further studies are necessary to examine whether or not the present criteria is applied to CML patients undergoing dasatinib therapy.

Previously reported results indicate that NK cells of patients newly diagnosed with CML are reduced in numbers or in proportion among lymphocytes and have limited cytolytic capacity at diagnosis of CML, and that these abnormalities persist during IM-induced remission.[39-41] The present results suggest that NK cells are more sensitive to the residual BCR-ABL-positive cells than are NKT cells and CD8+ T cells. However, the patterns in CD8+ T cells were different from those in NK cells among the four groups, implying that these two types of immune cells play different roles in CML tumor surveillance in different situations. It is possible that the effector population of CD8+ T cells, which were suppressed by IM, was simply restored to the normal level after stopping IM. Nevertheless, compared to patients in the fluctuating CMR group, some CMR patients treated with IM show either upregulation of effector memory CD8+ T cells or effector NK cells, or both (Fig. 3c), thus indicating a possible immunological mechanism for the minimization of BCR-ABL-positive cells that is linked to discontinuation of IM, while other underlying factors for stopping IM may exist.[42, 43]

In conclusion, our results suggest that monitoring of functional activation levels of NK cells, and perhaps of CD8+ T cells as well, in CML patients could be important for indirectly determining the level of residual CML cells, which contribute to immunological surveillance.

Acknowledgments

We thank Ms C. Kobayashi and A. Hirota for their technical assistance. This study was supported in part by the Private University Strategic Research Based Support Project: Epigenetics Research Project Aimed at General Cancer Cure using Epigenetic Targets (S0801020) and by a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology, Japan, and Tokyo Medical University Research Grant.

Disclosure Statement

Kazuma Ohyashiki received research support from Bristol-Myers Squibb KK and Novartis KK, and served as consultant and advisor of Novartis KK, Bristol-Myers Squibb KK and Ariad, honoraria for lecture fees from Novartis KK and Bristol-Myers Squibb KK.

Ancillary