Main

Although the tyrosine kinase inhibitor (TKI) imatinib has revolutionised, the treatment of chronic myeloid leukaemia (CML) and most patients respond well, a significant minority acquire resistance. Mutations in the BCR-ABL1 kinase domain are the best characterised cause. Although second-generation TKIs nilotinib and dasatinib were designed to target most imatinib-resistant mutations, some also confer clinical resistance to either nilotinib or dasatinib or both (T315I; Soverini et al, 2011). Therefore, it is recommended that mutation analysis be performed at TKI failure to aid subsequent therapy selection (Soverini et al, 2011). Particular second-line TKIs are recommended when specific mutations are detected by conventional sequencing techniques (Soverini et al, 2011). For example, dasatinib is recommended for patients with F359V or Y253H, nilotinib for patients with F317L, and ponatinib for patients with T315I. Using sensitive mass spectrometry-based mutation analysis, we have shown that detection of low levels of these same mutations is also predictive of response and their rapid clonal expansion with inappropriate TKI therapy (Parker et al, 2011). Hence, the same precautions with TKI selection apply. Mutant BCR-ABL1 clones are often deselected upon TKI cessation or change of therapy, and may become undetectable, even by very sensitive techniques (Hanfstein et al, 2011; Gruber et al, 2012). In the absence of the selective pressure of TKI, BCR-ABL1 mutations appear to confer a proliferative disadvantage, allowing the unmutated clone to predominate (Hanfstein et al, 2011). It is not known whether treatment discontinuation or long-term alternative TKI therapy leads to eradication of these mutant clones. If mutant clones persist at undetectable levels they have the potential to be reselected and expand clonally, given favourable conditions, such as change to a TKI for which they confer resistance. We examined longitudinal data of patients with imatinib-resistant mutations that became undetectable by direct Sanger sequencing, to determine whether ‘long dormant’ mutations could reappear, and the circumstances related to their reappearance.

Materials and methods

Patients included in this retrospective study were (1) molecularly monitored at our institution since the starting imatinib (400–800 mg daily), (2) chronic phase at imatinib start, and (3) had mutations detectable using direct Sanger sequencing (Sanger-sequencing) during imatinib therapy. We identified 50 such patients with mutations monitored longitudinally for 0.3–11.9 years since the starting imatinib (median 3.2 years). The study was performed in accordance with the Declaration of Helsinki, and approvals were obtained from the relevant institutional review boards. BCR-ABL1 transcript levels were assessed by RT–QPCR (Branford and Hughes 2006b) and mutation analysis was performed by Sanger-sequencing (Branford and Hughes 2006a; mutations detectable when mutant clones represent 10–20% of the total leukaemic population). Sensitive mass spectrometry-based mutation analysis (detection limit ∼0.2% mutant; Parker et al, 2011) was performed on selected samples as indicated.

Results

Fifty patients were followed since the starting imatinib in chronic phase who acquired mutations detectable by Sanger-sequencing during imatinib therapy. Of these, 48 patients had additional molecular follow-up after their mutation(s) were first detected. Mutations became undetectable by Sanger-sequencing in 20 of the 48 patients (27 mutations), at a median of 2.4 months after changing therapy (range 1–29 months; Table 1). The disappearance of individual mutations was associated with increased imatinib dose (3 mutations), stopping imatinib (1), haematopoietic stem cell transplant (HSCT) (6), chemotherapy (1), switching to nilotinib (3), or switching to dasatinib (14). All mutations that became undetectable when the patient switched to nilotinib or dasatinib were those known to be sensitive to the inhibitor received (for example, F359V in a patient switched to dasatinib).

Table 1 Details of patients with imatinib-resistant mutant BCR-ABL1 clones that became undetectable by Sanger-sequencing after changing therapy
Full size table

In 15 of the 20 patients whose mutations became undetectable by Sanger-sequencing after changing therapy, the mutations have not been detected by Sanger-sequencing again with 0.1–7.5 years of follow-up since the mutations were last detected (median 1.7 years; Table 1, patients 1–15).

In the other 5 of 20 patients, the same mutations as those originally detected (identical nucleotide exchange) reappeared and became detectable by Sanger-sequencing between 1.7 and 5.5 years after last detection (median 4.4 years; Table 1, patients 16–20). Patient 16 (Figure 1) acquired the F359V mutation on imatinib, and subsequently received an autologous HSCT and the mutation became undetectable by Sanger-sequencing. F359V could be detected at low levels at relapse post-transplant by mass spectrometry alone, suggesting that the mutant clone had not been eradicated at that time. After relapse post-transplant, the patient was treated with dasatinib for 3 years and BCR-ABL1 transcripts became undetectable. F359V remained undetectable by Sanger-sequencing and became undetectable by mass spectrometry. Dasatinib therapy was stopped owing to intolerance, and while the patient was off all TKI therapy F359V rapidly (within 1 week) reappeared and was detected by Sanger-sequencing (100% mutant), having been undetectable by Sanger-sequencing for 4.9 years and sensitive mass spectrometry for 2.7 years. To our knowledge, the acquisition of BCR-ABL1 mutations while CML patients are off TKI therapy has not been reported previously. This suggests that the original F359V-mutant BCR-ABL1-positive clone detected in this patient persisted at very low levels, and was the predominant leukaemic clone when dasatinib therapy was stopped. Given that the in vitro sensitivity of F359V-mutated BCR-ABL1 to dasatinib is approximately two-fold less than that of unmutated BCR-ABL1 (O'Hare et al, 2005), it is conceivable that the F359V-mutant BCR-ABL1-positive clone was more resilient to the selective pressure of dasatinib therapy than the unmutated clone, and hence became the major leukaemic clone during this time.

Figure 1
figure 1

Longitudinal molecular analysis of a CML patient with an imatinib-resistant mutation that became undetectable by Sanger-sequencing after changing therapy, and then reappeared and was detected by Sanger-sequencing after stopping all TKI therapy. The graph plots BCR-ABL1 transcript levels from the time of commencing imatinib. Shaded symbols indicate that a mutation was detected by Sanger-sequencing and/or mass spectrometry. Asterisks (*) indicate the time points when mass spectrometry mutation analysis was performed. Patient 16; 41-year-old male, treated with IFN for 8 months before commencing imatinib in 2001. Before stopping all TKI therapy, the F359V mutation was not detectable by either mutation detection method.

Full size image

The original imatinib-resistant mutations in patients 17–20 (Figure 2) also confer resistance to nilotinib (Y253H and F359V), and their reappearance was associated with initiation of nilotinib therapy and subsequent nilotinib resistance. One of these four patients lost a major cytogenetic response (patient 17), and three died of their disease (patients 18–20). Mass spectrometry-based sensitive mutation analysis was performed on samples collected immediately before starting nilotinib to determine whether the mutations that subsequently reappeared could be detected at low levels. The mutations were below the level of mass spectrometry detection in all four patients at that time.

Figure 2
figure 2

Longitudinal molecular analysis of four CML patients with imatinib- and nilotinib-resistant mutations that became undetectable by Sanger-sequencing after changing therapy, and then reappeared and were detected by Sanger-sequencing after switching to nilotinib and were associated with nilotinib resistance. The graphs plot BCR-ABL1 transcript levels from the time of commencing imatinib. Shaded symbols indicate that a mutation was detected by Sanger-sequencing and/or mass spectrometry. Asterisks (*) indicate the time points when mass spectrometry mutation analysis was performed. (A) Patient 17; 57-year-old female, treated with hydroxyurea for 2.2 years before commencing imatinib in 2002; Abbreviation: AP, progression to accelerated phase. Before nilotinib therapy, the F359V mutation was not detectable by either mutation detection method. (B) Patient 18; 71-year-old female, treated with hydroxyurea for 17 months before commencing imatinib in 2001. Before nilotinib therapy, the Y253H mutation was not detectable by either mutation detection method. (C) Patient 19; 35-year-old male, treated with IFN for 4 years before commencing imatinib in 2003; Abbreviations: DLI, donor leucocyte infusion; BC, progression to blast crisis. Before nilotinib therapy, the F359V mutation was not detectable by either mutation detection method. (D) Patient 20; 61-year-old male, treated with first-line imatinib in 2003. Before nilotinib therapy, the F359V mutation was not detectable by either mutation detection method. Owing to the complex mixture of mutations detected in patient 20, the samples assessed by Sanger-sequencing and mass spectrometry-based mutation analysis have been labelled (a–e), and the mutations detected at these time points are indicated in the adjacent box.

Full size image

Patient 20 had additional evidence of selection, deselection, and persistence of mutations over time (Figure 2D). Sensitive mutation analysis was performed on all samples collected from this patient after mutations were first detected by Sanger-sequencing. On imatinib, F359V and Q252H were detected by Sanger-sequencing. Using mass spectrometry, low-level M244V could also be detected immediately before stopping imatinib. At the time of nilotinib commencement following relapse post-allogeneic HSCT, M244V was the only mutation detectable by Sanger-sequencing, and low-level Q252H was also detected by mass spectrometry. F359V became detectable by Sanger-sequencing after 1 month on nilotinib, and the patient subsequently died of their disease. This example demonstrates the complex dynamics of selection and deselection of mutant BCR-ABL1-positive clones that can occur in patients.

There was one other resistant mutation that had the potential to re-emerge on dasatinib therapy. Patient 6 acquired F317L and Y253H on imatinib, which disappeared after increased imatinib dose (F317L) and switch to dasatinib (Y253H). After switching to dasatinib for 4 months, the patient rapidly achieved a major molecular response, which was followed by an allogeneic HSCT. Neither of the mutations was detected by Sanger-sequencing after they disappeared. The patient died of graft-versus-host disease.

Discussion

We show that imatinib-resistant BCR-ABL1 mutations that became undetectable by Sanger-sequencing, and even by sensitive analysis, after changing therapy may persist at very low levels for many years and should be taken into consideration when selecting TKIs to avoid their re-emergence.

It is unknown whether the reappearance of the mutations in the five patients described is due to clonal expansion of pre-existing low-level mutants or a new occurrence of the same mutation. However, the fact that the identical imatinib- and nilotinib-resistant mutation reappeared in patient 16 (Figure 1) without selection by either of these inhibitors, and immediately upon release from kinase inhibition by dasatinib, suggests that this was indeed the re-emergence of the original mutant clone that had been undetectable by Sanger-sequencing for almost 5 years. To assess the likelihood that the reappearance of the four identical BCR-ABL1 mutations in patients 17–20 (Y253H x1, F359V x3) is because of a new occurrence of the same mutations under the selective pressure of nilotinib, we used the published relative frequencies of mutation acquisition (Branford et al, 2009; Hughes et al, 2009). We estimate that the probability of patients 17–20 acquiring their mutation on imatinib and then again independently on nilotinib is 1.5 × 10−12 (based on the frequency of F359V or Y253H in imatinib-treated patients with mutations, 0.07 and 0.064, respectively (Branford et al, 2009); imatinib-resistant patients with baseline mutation(s) acquiring newly detectable mutation(s) on nilotinib, 0.31 (Hughes et al, 2009); and new F359V/C or Y253H in patients with newly detectable mutation(s) on nilotinib, 0.21 and 0.18, respectively (Hughes et al, 2009)). Therefore, the likelihood is low that the reappearance of the identical mutations in all of these patients is due to a new occurrence. This agrees with several studies showing that leukaemic relapse is almost always due to re-emergence or evolution of the founder clone (Mullighan et al, 2008; Ding et al, 2012; Egan et al, 2012; Parkin et al, 2013), exemplified by a study by Egan et al (2012) who demonstrated the shifting dominance of tumour clones during evolution of multiple myeloma to plasma cell leukaemia.

Despite the low probability that the mutations that reappeared in our patients were a new occurrence of the same mutations, several studies have demonstrated convergent evolution within distinct tumour sub-clones of individual patients (Schaub et al, 2010; Anderson et al, 2011). Convergent evolution, the independent acquisition of the same biological trait in unrelated lineages, implies co-operativity among somatically mutated genes. A recent study of 47 CML patients who had ⩾2 BCR-ABL1 mutations found that the mutations could be both components of compound mutant transcripts and present as single mutations using a cloning and Sanger-sequencing technique (Khorashad et al, 2013).

Irrespective of whether the BCR-ABL1 mutations in the five patients described herein have re-emerged owing to reselection and clonal expansion of pre-existing, persistent ‘dormant’ clones or whether the identical mutations have arisen independently, our data suggest that selection of a TKI to which a patient has exhibited a previous pattern of kinase domain-dependent resistance is best avoided. This study also emphasises the importance of characterising mutations during TKI therapy and before drug cessation or change of therapy as mutations can be rapidly deselected, masking the presence of potentially resistant clones. Furthermore, our data suggest that even if a historical mutation is not detectable using sensitive mutation analysis, it is not a guarantee that is has been eradicated, and thus the potential for re-emergence still exists. Selection of an inhibitor for which the historical mutation is sensitive would be recommended to reduce the risk of resistance to subsequent TKIs.