Leukemia (2012) 26, & 2012 Macmillan Publishers Limited All rights reserved /12

Transcription

1 Leukemia (2012) 26, All rights reserved /12 ORIGINAL ARTICLE Prevalence and dynamics of bcr-abl kinase domain mutations during imatinib treatment differ in patients with newly diagnosed and recurrent bcr-abl positive acute lymphoblastic leukemia H Pfeifer 1, T Lange 2, S Wystub 1, B Wassmann 1, J Maier 2, A Binckebanck 1, A Giagounidis 3, M Stelljes 4, M Schmalzing 5,UDührsen 6, L Wunderle 1, H Serve 1, P Brück 1, A Schmidt 1, D Hoelzer 1 and OG Ottmann 1 Imatinib is highly effective in newly diagnosed, but not in relapsed, Philadelphia chromosome positive acute lymphoblastic leukemia (Ph þ ALL). BCR-ABL tyrosine kinase domain (TKD) mutations are associated with acquired imatinib resistance, but their role in primary resistance is uncertain. Using highly sensitive ligation-pcr and denaturing high-performance liquid chromatography (DHPLC), we identified baseline TKD mutations in 21% and 42% of imatinib-naïve patients with newly diagnosed (n ¼ 26) or recurrent (n ¼ 65) Ph þ ALL, respectively (P ¼ ns). Within 4 weeks of starting the imatinib treatment, absolute levels of mutant bcr-abl transcripts increased significantly in patients with advanced, but not with de novo, Phþ ALL. The net expansion of pre-existing mutant clones during imatinib treatment resulted in the rapid appearance of initially undetectable TKD mutations, which after 4 weeks were detectable in 70% of patients with advanced disease. There was a high degree of concordance between the type of mutations detected at relapse and during initial imatinib treatment. The profoundly different outgrowth dynamics of leukemic clones with bcr-abl mutations in imatinib-treated patients who differ in their disease history, provides clinical - translational evidence for a contributory role of non-mutational resistance mechanisms, possibly induced by prior chemotherapy. Moreover, the prevalence of pre-existing, clinically relevant TKD may have been underestimated in tyrosine kinase inhibitor-naïve patients with Ph þ ALL. Leukemia (2012) 26, ; doi: /leu Keywords: BCR-ABL; mutation; minimal residual disease; resistance; Ph þ ALL; tyrosine kinase inhibitors INTRODUCTION Complete remission (CR) rates exceed 90% when imatinib is used as frontline therapy for newly diagnosed Philadelphia chromosome positive acute lymphoblastic leukemia (Ph þ ALL), 1-6 whereas it has very limited clinical efficacy in patients who relapse after chemotherapy Unfortunately, the majority of patients who do not undergo allogeneic stem cell transplantation eventually relapse. 1,8,11 Previous studies identified bcr-abl tyrosine kinase domain (TKD) mutations in more than 80% of patients with Ph þ ALL who relapsed during treatment with a tyrosine kinase inhibitor (TKI), with predominance of mutations displaying high IC 50 values in biochemical and cellular assays Bcr-abl mutations may pre-exist at low levels in a subset of patients with relapsed Ph þ ALL before salvage therapy with imatinib, and they have also been detected with variable frequency in imatinib-naïve patients with newly diagnosed Ph þ ALL. 12,16,19 Although no mutations were detected in 12 patients using standard sequencing or pyrosequencing for the T315I and Y253H mutations, 14 a more sensitive cloning and sequencing approach used by Soverini et al. 19 showed the presence of low frequency TKD mutations in all imatinib-naïve patients analyzed. This study identified a broad spectrum of mutations, most of which have not been linked to clinical resistance. On the basis of a sensitive denaturing high-performance liquid chromatography (DHPLC) method, we previously demonstrated that TKD mutations are present in approximately 40% of elderly patients with newly diagnosed, imatinib-naïve Ph þ ALL, with a high degree of concordance between the type of mutation detected at baseline and at relapse. 16 Remarkably, the CR rate achieved with imatinib alone did not differ in patients with pre-existing TKD mutations and those in whom only unmutated BCR-ABL was detected, suggesting that bcr-abl mutations have an only limited role during the initial treatment for de novo Ph þ ALL. Moreover, it has not been established to what degree TKD mutations contribute to the primary imatinib resistance commonly observed in patients with Ph þ ALL recurring after chemotherapy. Conceptually, the inferior imatinib response observed in these patients with advanced Ph þ ALL could be due to (i) a higher frequency of pre-existing mutations, (ii) a preponderance of mutations with greater transforming activity, 20 (iii) larger initial mutant clones or (iv) non-mutational resistance mechanisms. To evaluate the potential contributions of these mechanisms to 1 Department of Medicine, Hematology/Oncology, Goethe University, Frankfurt, Germany; 2 Center for Internal Medicine, Department of Hematology and Oncology, University of Leipzig, Leipzig, Germany; 3 St. Johannes Hospital, Department of Medicine, Duisburg, Germany; 4 Department of Hematology and Oncology, University of Münster, Münster, Germany; 5 Department of Hematology and Oncology, Eberhardt-Karls University, Tübingen, Germany and 6 Center for Internal Medicine, Department of Hematology and Oncology, University of Essen, Essen, Germany. Correspondence: Dr OG Ottmann, Medizinische Klinik II, Goethe Universität Frankfurt, Theodor Stern Kai 7, Frankfurt D-60590, Germany. ottmann@em.uni-frankfurt.de Received 14 December 2009; revised 6 September 2010; accepted 29 September 2010; accepted article preview online 9 January 2012; advance online publication, 14 February 2012

2 1476 primary resistance, we compared imatinib-naïve patients with newly diagnosed and recurrent Ph þ ALL in terms of the frequency and pattern of baseline mutations, the initial level of mutant clones, their outgrowth dynamics during the first 4 weeks of imatinib monotherapy and the relation between mutational status and time to progression (TTP). PATIENTS AND METHODS Study design This is an analysis of 91 patients with Ph þ acute lymphoid leukemias who were enrolled in the initial phase II studies of imatinib for treatment of BCR-ABL positive leukemias after chemotherapy failure (n ¼ 65), 8,9 or in a prospective, randomized clinical trial assessing frontline imatinib therapy in elderly patients with newly diagnosed Ph þ ALL (n ¼ 26). 2 Results of mutational analyses performed in the latter study have been published previously, 16 but did not include data on the dynamics and absolute levels of bcr-abl mutations during imatinib monotherapy, the central elements of the present communication. The studies were approved by the responsible Institutional Review Boards, and all patients gave informed consent according to the Declaration of Helsinki. Patient samples We serially assessed the mutation status of patients with imatinib-naïve Ph þ ALL (n ¼ 64) or lymphoid blast crisis (n ¼ 1) who had failed prior chemotherapy and were treated with imatinib at 400 or 600 mg per day, 8,9 and of 26 elderly patients with de novo Ph þ ALL who received 4 weeks of single-agent imatinib as induction therapy as described recently. 2 These elderly patients subsequently received combined imatinib and repetitive cycles of consolidation chemotherapy. Patient characteristics are shown in Table 1. To ensure comparability of the two patient groups with respect to the impact of imatinib treatment on the outgrowth of TKD mutations, the focus of this current analysis is limited to the first 4 weeks of therapy. Patients randomized to the frontline induction chemotherapy arm in the latter study are not included in this analysis. Reverse transcription-pcr, direct sequencing and DHPLC Extraction and reverse transcription of mrna, hemi-nested PCR and sequence analysis of the ABL kinase domain were performed as described. 16 Table 1. Characteristics Patient characteristics Phase II studies n ¼ 65 GMALL elderly trial N ¼ 26 Median age (years; range) 48 (17-76) 66 (54-79) Sex (n (%)) Male 34 (52.3) 12 (44) Female 31 (47.7) 14 (56) Leukemia type Ph+ ALL CML in lymphoid blastic phase 1 - Disease status (n (%)) Newly diagnosed - 26 First relapse 24 (39) - X Second relapse 10 (15) - Primary refractory 23 (35) - CR1 2 (3) - Breakpoint (n (%)) M-Bcr-Abl (b2a2, b3a2) 20 (31) 10 (42) m-bcr-abl (e1a2) 45 (69) 16 (58) Abbreviations: CML, chronic myelogenous leukemia; Ph+ ALL, Philadelphia chromosome positive acute lymphoblastic leukemia. Direct sequencing of the PCR products encoding the BCR-ABL ATP binding pocket (exons 4-7) was performed by a commercial laboratory (MWG Biotech, Ebersberg, Germany). For high sensitivity detection of bcr-abl TKD mutations, a semi-automated, high-throughput DHPLC system (WAVE, Transgenomic Ltd, Omaha, NE, USA) was used as described. 16 Detection thresholds of DHPLC for different mutations ranged from 0.1 to 1%. Ligation PCR Allele-specific PCR for the Y253H/F, E255K, G250E, Q252H, F359V and T315I was used to confirm results and perform the quantitation as described by Pelz-Ackermann et al. 22 In addition to the already described hybridization, ligation and real-time quantitative PCR quantification procedures to detect bcr-abl mutations, further hybridization probes were used for the remaining mutations.. All mutation-specific assays achieved a sensitivity of 0.01% mutated (bcrabl MUT ) in total BCR-ABL after subtracting the cross reactivity for this specific mutation. The comparative Ct method was used to calculate the relative percentage of mutated cells in the positive samples directly according to the following equation: % mutant allele ¼ 2 (ct bcr ablmut ct bcr abl total) 100. BaF3 cell lines containing the respective bcr-abl MUT or patient samples with a known ratio of bcr-abl MUT in total bcr-abl were used as positive controls in all assays. Similarly, BaF3 cells containing unmutated bcr-abl only were used as a negative control, specifically to reveal any cross reactivity with the mutation-specific oligonucleotides. Samples were only scored as positive if the Ct value of the sample did not exceed the Ct value of the negative control in two independent runs. Additional negative controls were used in each step: one negative control without cdna in the bcr-abl accumulation PCR; one ligation negative control lacking target DNA in the ligation reaction and one real-time PCR control performed in the absence of the template. Quantitation of mutated bcr-abl level The size of the mutated clone depends on the bcr-abl load and the percentage of the mutation in relation to unmutated bcr-abl. To obtain a quantitative assessment of the size of the mutated clone, the bcr-abl MUT / GAPDH ratio was calculated by multiplying the proportion of mutated bcr-abl measured by ligation PCR with the bcr-abl/gapdh ratio measured at the same time point by quantitative real-time PCR. A detailed summary of the analyses performed by DHPLC, ligation PCR, sequencing or combinations of these three techniques is provided in two comprehensive Supplementary Tables. Statistical analysis TTP was plotted according to the methods of Kaplan - Meier, with differences between patient groups analyzed by the log rank test. The w 2 test was used for comparison of the incidence of mutations in the two patient groups. Statistical analyses were performed using the GraphPad Prism (GraphPad Prism Inc., La Jolla, CA, USA) and the BiAS (epsilon Verlag, Frankfurt, Germany) software packages. RESULTS Type and frequency of mutations before imatinib treatment Evidence of baseline bcr-abl mutations will influence clinical decisions on the selection of effective, individualized treatment strategies, but to date only limited data on the prevalence of TKD mutations in imatinib-naïve patients are available. The present analysis was conducted to elucidate whether imatinib-based therapy results in selection and clonal expansion of pre-existing, low-level imatinib-resistance mutations or whether they occur for the first time during treatment. Baseline BCR-ABL TKD mutations were detected by ligation-pcr and/or DHPLC in 42% (25/60) of patients with recurrent and in 21% (5/24) of patients with newly diagnosed Ph þ ALL (P ¼ 0.07). In both groups combined, P-loop mutations predominated, being detected in 22 of the 24 patients (91%) who by ligation-pcr harbored at least one mutation before

3 21% Proportion of patients with a detectable TKD mutation at different timepoints RELAPSE 17% 33% De Novo Ph+ALL 88% 42% 38% 70% 66% Relapsed or refractory Ph+ALL Figure 1. Proportion of patients with detectable BCR-ABL mutations during 4-week imatinib induction in relation to disease stage. imatinib treatment. A T315I mutation was detected in three patients, one of whom showed a dual mutation (T315I; Y253H). This high frequency of P-loop mutations at baseline is consistent with the results of our previously published, prospective GMALL study of elderly patients with Ph þ ALL. 16 Proportion of patients with a mutation during treatment and at relapse Figure 1 depicts the proportion of patients in whom a TKD mutation was detected during the first 4 weeks of single agent imatinib and at relapse. Among elderly patients with newly diagnosed Ph þ ALL, 3 of 18 (17%) and 6 of 18 (33%) evaluable patients were found to have a TKD mutation after 2 and 4 weeks of imatinib, respectively. The corresponding values for patients with advanced disease were 25 of 60 (42%), 15 of 39 (38%) and 21 of 30 (70%) at baseline, after 2 and 4 weeks, respectively (Figure 1). TTP differed substantially between these two patient groups. Median TTP was 67 days (range days) in patients receiving single-agent imatinib as salvage therapy, as opposed to 452 days (range days) in patients with de novo Ph þ ALL who after the first 4 weeks received imatinib in combination with chemotherapy. Of the 65 patients, 44 with advanced disease relapsed or progressed on imatinib monotherapy; of the remaining patients, 10 underwent stem cell transplantation in CR, 8 were still in CR at the end of study (EOS), 1 died in CR, 2 had no data available. Samples collected at the time of treatment failure were available from 36 of the 44 progressing patients; 24 of these (67%) harbored a mutation, the other 12 patients (34%) revealed no mutation by direct sequencing. We identified five different types of TKD mutations at relapse, three of which (E255K/V, T315I and Y253H) comprised 93% of all mutations observed (Figure 2). Double mutations were present in four cases: T315I/E255K/V (n ¼ 2), T315I/Y253H (n ¼ 1) and Y253H/E255K/V (n ¼ 1). Analysis of relapse samples from patients with de novo Ph þ ALL revealed a TKD mutation in 14 of the 15 (93%) evaluable cases; unmutated bcr-abl was observed in only one patient. Notably, the spectrum of mutations differed from that of patients with advanced disease, with a total of eight different mutations and a less marked predominance of P-loop mutations (Figure 2). Outgrowth dynamics of bcr-abl mutants in relation to disease stage The rapid emergence of imatinib-resistant leukemic cells harboring TKD mutations in heavily pretreated patients with Advanced Ph+ ALL Y253H 13% Q252H 5% E255K/V 33% WT 31% T315I 18% E355G 6% G250E 25% F317L 6% De novo Ph+ ALL D276G 6% WT 12% E255K/V 13% T315I L384M 13% 13% Q252H 6% Figure 2. Frequency of bcr-abl TKD mutations determined by direct sequencing at the time of disease progression in the two patient cohorts. Double mutations were observed in four patients with advanced Ph þ ALL who relapsed on imatinib: T315I/E255K/V (n ¼ 2), Y253H/E255K/V (n ¼ 1) and Y253H/T315I (n ¼ 1). No compound mutations were observed in patients relapsing on firstline therapy for diagnosed Ph þ ALL. Ph þ ALL raised the possibility that the presence of baseline mutations could be underestimated largely because of the insufficient sensitivity of current PCR methodology. We hypothesized that treatment with imatinib would rapidly unmask preexisting, but initially undetectable mutant clones by selectively eliminating leukemic cells with unmutated bcr-abl. To test this possibility, we compared the outgrowth kinetics of TKD mutations during short-term treatment with imatinib in patients with newly diagnosed and with advanced, heavily pretreated Ph þ ALL. The pre-treatment level of the mutated allele in the subset of patients with a detectable baseline mutation was not significantly higher in patients with recurrent leukemia (median bcr-abl mut / GAPDH ratio 1.3E-09; range E-03) (Figure 3a) than that in patients with de novo Ph þ ALL (median bcr-abl mut /GAPDH ratio 0; range E-02 ) (Figure 3b) (P ¼ ns). During the first 4 weeks, a profound increase in the percentage of mutated clones occurred in a subset of patients with advanced Ph þ ALL, but not in patients receiving imatinib as firstline therapy (Supplementary Tables 1 and 2). Although this suggested that the dynamics of BCR-ABL mutated clones during imatinib treatment differed depending on disease stage, it did not discriminate between a net expansion of the mutated clone and an only relative increase resulting from the preferential elimination of cells with unmutated bcr-abl. To quantitatively assess the absolute rather than the relative clone size over time, we calculated the level of mutated bcr-abl transcripts based on their frequencies in conjunction with the concurrent level of minimal residual disease as determined by the bcr-abl/gapdh ratio. The mutation detection methods and all individual factors entering the analysis are provided in Supplementary Tables 1 and 2. During the first 4 weeks, the median level of mutated alleles increased profoundly in patients with advanced Ph þ ALL from 0 bcr-abl mut /GAPDH (range: E-02) after 2 weeks to 1.2E-03 bcr-abl mut /GAPDH (range: E þ 01) after 4 weeks (Figure 3a). The proportion of mutated alleles exceeded 10E-07 bcr-abl mut /GAPDH in 20 of 43 (23%) patients at start of imatinib treatment, in 11 of 38 (29%) patients after 2 weeks and in 17 of 29 (59%) patients after 4 weeks (Figure 3a) of the treatment. In contrast, mutated BCR-ABL levels did not increase throughout treatment weeks 1--4 in any of the patients with de novo Ph þ ALL, the median bcr-abl mut /GAPDH ratio remaining 0 (Figure 3b). Although relatively high bcr-abl mut /GAPDH ratios were observed in eight patients at some time during the initial 4 weeks of firstline imatinib treatment, these may be overestimated as they were based on DHPLC analysis and not ligation-pcr (Supplementary Table 2). Moreover, only a single sample was available in five of these patients. In three patients in whom at least two serial samples were available, the bcr-abl mut /GAPDH ratio decreased or was stable (Figure 3b and Supplementary Table 2). 1477

4 1478 (n=43) (n=38) (n=29) (n=20) (n=16) (n=18) EOS (n=16) # # * 1 * * 2 1 * 2 (n=16) (n=18) Expansion kinetics of bcr-abl mutants in patients with advanced Ph þ ALL A more detailed analysis of this patient cohort revealed three basic patterns of mutation kinetics. Pattern 1 represented detectability of only unmutated bcr-abl throughout the 4 weeks of treatment, and was observed in 12 of 44 (27%) patients. Pattern 2 reflects an early and rapid net expansion of mutated alleles as shown in Figure 4a, whereas pattern 3 represents initially decreasing or unchanging absolute levels of mutated alleles (Figure 4b). With disease progression, coinciding with EOS, neither mutant allele levels nor proportion of patients harboring a mutation differed significantly between these subsets (Figure 4a and b). At EOS, all but two patients demonstrated very high levels of mutated bcr-abl, irrespective of their initial short-term response to imatinib. # * (n=17) Figure 3. Level of BCR-ABL mutations during the initial 4 weeks of single-agent imatinib treatment. Comparative analysis of the median bcr-abl mut /GAPDH ratios as a quantitative measure of the mutation status of bone marrow and peripheral blood samples from all evaluable patients with relapsed or refractory Ph þ ALL who received imatinib for advanced Ph þ ALL (a) or with de novo Ph þ ALL treated with single-agent imatinib as induction therapy (b). 2,8 Only patients in whom samples from at least two of the three time points were available are included in this analysis. As a consequence, this graphical depiction of mutations omits four baseline mutations in recurrent and two baseline mutations in de novo Ph þ ALL. These mutations are, however, included in the other analyses. The median level of mutated BCR- ABL increased substantially in patients with relapsed or refractory Ph þ ALL, but not in patients with newly diagnosed Ph þ ALL. (b) # indicates one patient with samples available from all three time points. *Indicates five patients in whom a mutation was detected in only a single sample during the initial 4-week treatment period. y1 and y2 denote two patients in whom samples were available only from weeks 2 and 4, demonstrating that the mutant bcr-abl level did not increase during this time period. (n=12) (n=11) (n=9) EOS (n=13) Figure 4. Rapid expansion of mutated bcr-abl during imatinib therapy in patients with advanced Ph þ ALL. Median bcr-abl mut / GAPDH ratios before and during the first 4 weeks of imatinib mono-therapy. Patients who showed exclusively unmutated bcr-abl during this early observation period are not included in this graph. (a) The outgrowth dynamics showed a profound net increase of mutated bcr-abl in a subset of patients, which became apparent as early as 2 weeks after starting treatment. (b) A second group of patients with pre-existing TKD mutations showed an initial decline or constant levels of mutated bcr-abl during imatinib, despite predominance of high resistance mutations. When patients in this cohort were assessed separately according to their pre-treatment bcr-abl mut /GAPDH levels, no significant difference was observed with respect to the TTP between patients who displayed a high (Figure 5a) versus low or undetectable (Figure 5b) mutated clone (TTP 76.5 days and 64 days, respectively). Accordingly, two main types of outgrowth dynamics were observed, which may reflect a different biology of the leukemic cells. TKD mutations were identified by direct sequencing at the time of progression, and were found to be almost exclusively high resistance mutations (Figure 2). The spectrum of mutations was independent of the expansion kinetics of the mutated population. Early bcr-abl mutations do not predict response and TTP We hypothesized that patients with advanced disease in whom mutations were detectable already before imatinib treatment would have a lower probability of responding. Of the 24 (54%) patients with an upfront mutation, 13 achieved a CR or Cri, whereas 11 (46%) patients failed to respond. Of the 35 patients in whom no mutation was detected, 20 (57%) achieved a CR or Cri and 15 (43%) failed therapy (P ¼ ns). Thus, there was no correlation between detection of a mutation and response. To further evaluate whether the early detection of a mutation was predictive of response duration, we performed a Kaplan-- Meier

5 TTP = 76.5 d End of study TTP = 64 d End of study Figure 5. Lack of relationship between pre-treatment size of mutated bcr-abl clones and TTP in patients with advanced disease. The size of the leukemic burden-harboring mutant bcr-abl was quantitatively assessed by calculating the bcr-abl mut /GAPDH ratios in patients with relapsed or refractory Ph þ ALL before imatinib treatment and at the time of imatinib discontinuation because of disease progression or relapse. The outgrowth dynamics in patients with low or undetectable (a) or high (b) bcr-abl mut /GAPDH ratios at study start Concordance between TKD mutations at baseline and relapse To determine whether TKD mutations present at relapse corresponded to the type of mutations detected before first imatinib exposure, we examined paired samples collected at baseline and at relapse from patients with advanced Ph þ ALL who received only imatinib as salvage therapy. Relapse without a detectable TKD mutation by sequencing was observed in 12 patients. Screening for baseline mutations revealed unmutated bcr-abl in 11 (92%) and a TKD mutation in 1 (8%; G250E) of these patients. Concordance was slightly lower when we compared the results of ligation-pcr performed exclusively on the pre-imatinib samples with the sequencing results obtained at EOS (n ¼ 18); the same mutation was detected in 14 of 18 (78%) samples, unmutated bcr-abl in 4 (22%) baseline samples. Concordance increased further when the results of all mutational analyses performed during the first 4 weeks of imatinib were considered, rather than only baseline samples; 17 of the 18 (94%) patients with a mutation at relapse were found to have the same mutation in at least one sample collected within 4 weeks of starting imatinib treatment. Thus, our results indicate that in the setting of imatinib monotherapy, detection of a mutation either before or shortly after starting imatinib treatment has a high positive predictive value for the subsequent appearance of the same mutation at relapse. Comparison of DHPLC and ligation-pcr Methodologically, DHPLC represents a screening method rather than a method to accurately identify a specific type of mutation. Therefore, we used ligation-pcr to confirm the results of DHPLC when a conspicuous DHPLC profile indicated the presence of a low level TKD mutation. An overview over the comparison of DHPLC with ligation-pcr is provided in the Supplementary data Tables. Overall, concordance in this retrospective analysis is good. Ligation-PCR could be performed in 16 of 19 patients with advanced disease in whom DHPLC analysis was indicative of a low level TKD mutation at initial diagnosis. Ligation-PCR verified the mutation detected by DHPLC in 14 of these 16 (87.5%) cases. Conversely, to determine whether DHPLC is sufficiently sensitive to exclude the pre-existence of low level mutations, we used the more sensitive ligation-pcr to reanalyze diagnostic samples from 12 patients in whom a TKD mutation was found at the EOS, but the DHPLC pattern at baseline indicated only unmutated bcr-abl. Ligation-PCR identified a TKD mutation in 8 of the 12 patients (67%); notably, the mutation was the same as the one detected at relapse in all 12 patients. Figure 6. Lack of relationship between pre-treatment mutational status and TTP in patients with advanced Ph þ ALL. Kaplan -Meier analysis of TTP was calculated either on the basis of mutational status determined exclusively before imatinib, or on the presence or absence of a TKD mutation detected at any time within the initial 4-week treatment period. Median TTP in the four groups was 59, 65.5, 52.5 and 70 days, respectively. Patients were censored at the time of stem cell transplantation or death in CR. analysis of TTP in patients with and without detectable mutations before, or at any time during, the first 4 weeks of imatinib (Figure 6). There is no statistically significant difference in TTP by mutational status. In newly diagnosed Ph þ ALL, eight patients were shown to harbor a mutation during the first 4 weeks of treatment. Seven of these patients relapsed and one was in ongoing CR at last followup (CCR). Of the 18 patients without a mutation during the initial treatment period, 10 relapsed, 5 died in CR and 3 were in CCR. Likewise, TTP did not differ in patients with and without early mutations, with a median of 493 days (range days) and 417 days (range days), respectively. DISCUSSION The majority of patients with Ph þ ALL who relapse during the treatment with a TKI carry bcr-abl TKD mutations that display high IC 50 values in biochemical and cellular assays, 13,20,21 implying that they have a causal role in acquired imatinib resistance. Their role in primary resistance commonly observed in Ph þ ALL recurring after chemotherapy is uncertain, and it also remains to be determined when these mutations first develop. Low-level bcr-abl mutations have been detected with variable frequency in imatinib-naïve patients with Ph þ ALL both at initial diagnosis and at the time of recurrence after prior therapy. 12,16,19 The significantly lower response rate and shorter TTP in the latter group suggest that the relative contribution of TKD mutations to resistance may vary by disease history and extent of prior therapy. The inferior efficacy of imatinib in patients with advanced Ph þ ALL could thus be due to non-mutational resistance mechanisms induced by prior chemotherapy, or alternatively by a higher frequency of pre-existing mutations, a preponderance of mutations with greater transforming activity or a larger mutant clone size at the start of treatment with TKI.

6 1480 Pre-existing TKD mutations were identified in a substantial proportion of patients from both treatment groups (21% and 42%, respectively; P ¼ ns), but the frequency of the mutated allele was low and detectable only by methods much more sensitive than direct sequencing. In contrast to our expectations, the prevalence of TKD mutations at baseline was not significantly greater in patients with disease recurring after multi-agent chemotherapy than with newly diagnosed leukemia; however, the number of patients may have been too small to identify a small, but statistically significant difference between the two groups. Alternatively, the sensitivity of the PCR techniques we used may have been insufficient to detect very low-level mutations. Ligation-PCR was our most sensitive method for detecting mutations, but our analysis was restricted to the relatively small panel of mutations that comprise the vast majority of mutations at relapse, and its use was furthermore directed towards those mutation found at relapse or suggested by the DHPLC pattern during the first 4 weeks of treatment. Accordingly, we may considerably underestimate the frequency of baseline TKD mutations, although not necessarily of those that became clinically relevant. Our results differ from data reported by Jones et al. 14 who did not identify baseline bcr-abl mutations in any of the 21 examined patients. This difference may be attributable to the lower sensitivity of the direct sequencing and pyrosequencing techniques used by Jones et al. when compared with the highly sensitive DHPLC and ligation-pcr-based methodologies used in our study. Supporting this explanation is a recent report in which baseline TKD mutations were identified in all 15 patients studied using a highly sensitive cloning and sequencing approach. 19 Interestingly, this study revealed two to five point mutations for each patient, with multiple point mutations co-existing in the same cloned fragment in some cases. In contrast to our observations showing a preponderance of P-loop and T315I mutations with high IC 50, the majority of the point mutations detected by Soverini et al. 19 have never been reported in association with TKI resistance. One possible explanation for the different distribution pattern of mutations observed by ourselves and by Soverini et al. is the analysis of different patient cohorts, whereas we analyzed primarily the patients who had relapsed after chemotherapy and Soverini et al. examined only the patients with newly diagnosed Ph þ ALL. An alternative explanation involves methodological differences, as our DHPLC screening covered only the exons known to harbor mutations, and our ligation-pcr approach was restricted to the most frequent TKD mutations. The high prevalence of TKD mutations at relapse and the speed with which patients with recurrent Ph þ ALL became resistant to imatinib prompted us to investigate whether the outgrowth of pre-existing, but subpopulations of mutant bcr-abl clones that were below the detection threshold contributed to acquired resistance. We hypothesized that short-term imatinib treatment would unmask minor pre-existing mutant clones by selectively eliminating leukemic cells harboring un-mutated bcr-abl, and that this would be more pronounced in recurrent as opposed to de novo Ph þ ALL. This hypothesis was confirmed by comparing the outgrowth kinetics of TKD mutations during 4 weeks of imatinib monotherapy in these two patient groups, whereas the median levels of mutant bcr-abl transcripts as well as the proportion of patients with detectable TKD mutations increased profoundly during short-term imatinib treatment in patients with recurrent Ph þ ALL, and no increases were observed in patients with newly diagnosed Ph þ ALL. In order to discriminate between a net expansion as opposed to an only relative enrichment of the leukemic clones during the brief imatinib exposure, we performed a quantitative assessment of the level of mutated bcr-abl transcripts (as determined by the bcr-abl MUT /GAPDH-ratio) and thus of the size of the mutated clone by multiplying the proportion of mutated bcr-abl measured by ligation-pcr with the bcr-abl/gapdh ratio measured at the same time point by quantitative real-time PCR. Our data clearly show a net increase in the size of the leukemic clone within the first 2-4 weeks after starting imatinib treatment in a substantial subset of patients with recurring Ph þ ALL. This extremely rapid expansion of leukemic clones with a TKD mutation strongly supports the selection of preexisting mutant clones rather than induction of new mutations. A striking finding is the difference in outgrowth kinetics when patients with de novo and recurrent Ph þ ALL were compared. We focussed our analysis on this very short time window to be able to compare patient cohorts who received identical treatment, that is single-agent imatinib, and differed only in their disease history. Neither the proportion of patients with a detectable mutation nor the absolute levels of mutant bcr-abl increased in the group of patients with de novo Ph þ ALL. This implies that the proliferative capacity of leukemic clones harboring a TKD mutation was less pronounced during imatinib treatment in patients with newly diagnosed than with advanced disease. In addition, we cannot exclude that the frequency of minute leukemic clones with mutated bcr-abl is indeed lower in de novo Ph þ ALL, although our comparison of the two patient groups revealed only a trend towards a higher incidence of baseline mutations in patients with advanced disease. To unequivocally clarify this issue, methods for mutation detection that are even more sensitive than the ligation- PCR used in our study are needed. Nevertheless, on the basis of the outgrowth dynamics of mutated clones in the two patient cohorts under study, it can be speculated that prior genotoxic therapy and increasing genetic instability may have contributed to this difference in biological behavior. Different absolute levels of mutant bcr-abl before starting imatinib treatment do not explain the different speed with which clinical resistance developed in the two patient groups; rapid expansion kinetics of mutant bcr-abl clones were observed even among those patients with recurrent disease in whom no baseline mutations could be detected, with an essentially identical time to clinical progression as in patients with relatively high starting levels of mutant bcr-abl. Taken together, these data demonstrate (i) that the frequency of baseline mutations particularly in patients with advanced and heavily pre-treated Ph þ ALL is considerably underestimated by current PCR techniques, (ii) that the rapid emergence of imatinib resistance in these patients is associated primarily with the selection of pre-existing mutations rather than the induction of new mutations during TKI treatment, (iii) that the unmasking of mutations during short-term imatinib treatment results from a net expansion of pre-existing mutant clones in a substantial subset of patients and (iv) that the kinetics with which mutant clones expand in patients treated with imatinib for newly diagnosed Ph þ ALL is considerably slower than that in patients with advanced disease receiving imatinib as salvage therapy. Detailed investigation of the non-mutational resistance mechanisms that contribute to rapid clinical progression in the latter patients may provide insight into basic resistance mechanisms that are also involved in relapse during firstline therapy with TKI. Moreover, our data provide a rationale for early high sensitivity testing for mutations in Ph þ ALL, and for frontline use of agents that are effective against the BCR-ABL KD mutations associated with imatinib resistance. CONFLICT OF INTEREST OG Ottmann received research support and honoraria for advisory board activities and scientific presentations from Novartis and Bristol-Myers-Squibb. The remaining authors declare no conflict of interest. ACKNOWLEDGEMENTS We thank Doreen Badowski, Brigitte Gehrke, Tamara Klan and Sandra Markovic for excellent technical assistance.

7 Supported by grants from the Deutsche Jose Carreras Leukämiestiftung (R06/25v), BMBF Competence Network Acute Leukemias Grant No. 01G19971, the German Genome Research Network (NGFN), the Wilhelm Sander Stiftung and the Adolf- Messer Foundation, Germany, and from Novartis Pharma AG, Nürnberg, Germany. AUTHOR CONTRIBUTIONS HP designed the study, performed experiments, analyzed data and wrote the manuscript. BW conducted the clinical study, analyzed data and wrote the manuscript. AS, SW, JM and TL performed experiments and analyzed data. AB conducted the clinical study and analyzed data. AG, MS, MS, UD, PB, LW and HS enrolled and treated patients. DH designed the study and wrote the manuscript. OGO designed the study, conducted the clinical trial, analyzed data and wrote the manuscript. REFERENCES 1 Vignetti M, Fazi P, Cimino G, Martinelli G, Di Raimondo F, Ferrara F et al. Imatinib plus steroids induces complete remissions and prolonged survival in elderly Philadelphia chromosome-positive acute lymphoblastic leukemia patients without additional chemotherapy: results of the Gruppo Italiano Malattie Ematologiche dell Adulto (GIMEMA) LAL0201-B protocol. Blood 2007; 109: Ottmann OG, Wassmann B, Pfeifer H, Giagounidis A, Stelljes M, Dührsen U et al. Imatinib compared with Chemotherapy as Front-Line Treatment of Elderly Patients with Philadelphia Chromosome positive Acute Lymphoblastic Leukemia (Ph+ ALL). Cancer 2007; 109: Thomas DA, Faderl S, Cortes J, O Brien S, Giles FJ, Kornblau SM et al. Treatment of Philadelphia chromosome-positive acute lymphocytic leukemia with hyper-cvad and imatinib mesylate. Blood 2004; 103: Lee KH, Lee JH, Choi SJ, Lee JH, Seol M, Lee YS et al. Clinical effect of imatinib added to intensive combination chemotherapy for newly diagnosed Philadelphia chromosome-positive acute lymphoblastic leukemia. Leukemia 2005; 19: Yanada M, Takeuchi J, Sugiura I, Akiyama H, Usui N, Yagasaki F et al. High complete remission rate and promising outcome by combination of imatinib and chemotherapy for newly diagnosed BCR-ABL-positive acute lymphoblastic leukemia: a phase II study by the Japan Adult Leukemia Study Group. J Clin Oncol 2006; 24: Wassmann B, Pfeifer H, Goekbuget N, Beelen DW, Beck J, Stelljes M et al. Alternating versus concurrent schedules of imatinib and chemotherapy as frontline therapy for Philadelphia-positive acute lymphoblastic leukemia (Ph+ ALL). Blood 2006; 108: Druker BJ, Sawyers CL, Kantarjian H, Resta DJ, Reese SF, Ford JM et al. Activity of a specific inhibitor of the BCR-ABL tyrosine kinase in the blast crisis of chronic myeloid leukemia and acute lymphoblastic leukemia with the Philadelphia chromosome. N Engl J Med 2001; 344: Ottmann OG, Druker BJ, Sawyers CL, Goldman JM, Reiffers J, Silver RT et al. A phase 2 study of imatinib in patients with relapsed or refractory Philadelphia chromosome-positive acute lymphoid leukemias. Blood 2002; 100: Wassmann B, Pfeifer H, Scheuring UJ, Binckebanck A, Gökbuget N, Atta J et al. Early prediction of response in patients with relapsed or refractory Philadelphia chromosome positive acute lymphoblastic leukemia (Ph+ ALL) treated with imatinib mesylate (Glivec). Blood 2004; 103: de Labarthe A, Rousselot P, Huguet-Rigal F, Delabesse E, Witz F, Maury S et al. Group for Research on Adult Acute Lymphoblastic Leukemia (GRAALL). Imatinib combined with induction or consolidation chemotherapy in patients with de novo Philadelphia chromosome-positive acute lymphoblastic leukemia: results of the GRAAPH-2003 study. Blood 2007; 109: Delannoy A, Delabesse E, Lheritier V, Castaigne S, Rigal-Huguet F, Raffoux E et al. Imatinib and methylprednisolone alternated with chemotherapy improve the outcome of elderly patients with Philadelphia-positive acute lymphoblastic leukemia: results of the GRAALL AFR09 study. Leukemia 2006; 20: Hofmann WK, Jones LC, Lemp NA, de Vos S, Gschaidmeier H, Hoelzer D et al. Ph(+) acute lymphoblastic leukemia resistant to the tyrosine kinase inhibitor STI571 has a unique BCR-ABL gene mutation. Blood 2002; 99: von Bubnoff N, Schneller F, Peschel C, Duyster J. BCR-ABL gene mutations in relation to clinical resistance of Philadelphia-chromosome-positive leukaemia to STI571: a prospective study. Lancet 2002; 359: Jones D, Thomas D, Yin CC, O Brien S, Cortes JE, Jabbour E et al. Kinase domain point mutations in Philadelphia Chromosome positive acute lymphoblastic leukemia emerge after therapy with bcr-abl kinase inhibitors. Cancer 2008; 113: Branford S, Rudzki Z, Walsh S, Parkinson I, Grigg A, Szer J et al. High frequency of point mutations clustered within the adenosine triphosphate-binding region of BCR/ABL in patients with chronic myeloid leukemia or Ph-positive acute lymphoblastic leukemia who develop imatinib (STI571) resistance. Blood 2002; 99: Pfeifer H, Wassmann B, Pavlova A, Wunderle L, Oldenburg J, Binckebanck A et al. Kinase domain mutations of BCR-ABL frequently precede imatinib-based therapy and give rise to relapse in patients with de novo Philadelphia-positive acute lymphoblastic leukemia (Ph+ ALL). Blood 2007; 110: Soverini S, Colarossi S, Gnani A, Rosti G, Castagnetti F, Poerio A et al. Contribution of ABL Kinase domain mutations to imatinib resistance in different subsets of Philadelphia-positive patients. Clin Cancer Res 2006; 12: Hofmann WK, Komor M, Wassmann B, Jones LC, Gschaidmeier H, Hoelzer D et al. Presence of the BCR-ABL mutation Glu255Lys prior to STI571 (imatinib) treatment in patients with Ph+ acute lymphoblastic leukemia. Blood 2003; 102: Soverini S, Vitale A, Poerio A, Gnani A, Colarossi S, Iacobucci I et al. Philadelphiapositive acute lymphoblastic leukemia patients already harbor BCR-ABL kinase domain mutations at low levels at the time of diagnosis. Haematologica 2011; 96: Griswold IJ, MacPartlin M, Bumm T, Goss VL, O Hare T, Lee KA et al. Kinase domain mutants of BCR-ABL exhibit altered transformation potency, kinase activity, and substrate utilization, irrespective of sensitivity to imatinib. Mol Cell Biol 2006; 26: von Bubnoff N, Peschel C, Duyster J. Resistance of Philadelphia-chromosome positive leukemia towards the kinase inhibitor imatinib (STI571, Glivec): a targeted oncoprotein strikes back. Leukemia 2003; 17: Pelz-Ackermann O, Cross M, Pfeifer H, Deininger M, Wang SY, Al-Ali HK et al. Highly sensitive and quantitative detection of bcr-abl kinase domain mutation by ligation PCR. Leukemia 2008; 22: Supplementary Information accompanies the paper on the Leukemia website (