Sequential Inhibitor Therapy in CML: In Vitro Simulation Elucidates the Pattern of Resistance Mutations after Second- and Third-Line Treatment Free

A drawback of imatinib therapy in chronic myelocytic leukemia (CML) and Ph+ acute lymphocytic leukemia (ALL) presented itself in emerging resistance toward imatinib, found to be caused by several molecular resistance mechanisms, in particular, by point mutations in the kinase domain of BCR-ABL (1). The second-generation tyrosine kinase inhibitors (TKI) dasatinib and nilotinib were shown to be active in imatinib-resistant disease including patients with BCR-ABL kinase mutations with exception of T315I (2, 3). Given first-line in chronic phase CML, both compounds compare favorably to imatinib with respect to response rates and progression events (4, 5). In vitro studies suggest a narrowed but partially overlapping spectrum of resistance mutations with the novel inhibitors compared with imatinib (6–10). Clinical studies showed that these mutations identified in vitro (Q252H, E255K/V, V299L, F317L, and T315I for dasatinib; Y253H/F, E255K/V, F311I, T315I and F359C/V for nilotinib) were associated with less favorable response rates and also emerged at the time of disease progression receiving second-line treatment (11–13). Dasatinib or nilotinib are given as second- and third-line therapies in patients with CML with resistance or intolerance to imatinib. Therefore, many patients are treated with a sequence of 3 or more ABL TKIs. In this work, we aimed at depicting the clinical approach of sequential inhibitor therapy in vitro. Our results reveal significant differences in response characteristics and evolution of individual mutations depending on the order and concentration of nilotinib and dasatinib, sequentially administered in identical replicates of imatinib-resistant cell lines.

The concentrations of imatinib, nilotinib and dasatinib correspond to the clinically attainable plasma mean trough levels (imatinib: 4,000 nmol/L, dasatinib: 100 nmol/L, nilotinib: 1,700 nmol/L; refs. 3, 14–16) or 50% of these concentrations. First-line therapy was simulated by plating Ba/F3 Mig p185 wild-type cells into 96-well plates at a density of 4 × 10⁵ cells/well and incubation with imatinib at 4 μmol/L over 28 days (17). Individual resistant cell colonies grew up, each in a different well of the 96-well plate. These individual colonies were picked and expanded in the presence of imatinib to individual imatinib-resistant cell lines and numbered 1 to 24 in chronologic order of appearance. Identical replicates of the 24 imatinib-resistant cell lines were replated and treated with dasatinib at 50 nmol/L, dasatinib at 100 nmol/L, nilotinib at 850 nmol/L, or nilotinib at 1,700 nmol/L, with 8 replicates per condition. Outgrowing, sequentially resistant wells or colonies were again picked, expanded, replated in 8 replicates per line, and incubated for 28 days, with nilotinib-resistant lines being switched to dasatinib (50 or 100 nmol/L) and dasatinib-resistant lines being switched to nilotinib (850 or 1,700 nmol/L). BCR-ABL mutation analysis was conducted after completing all 3 lines of treatment.

RNA isolation, reverse transcription (RT)-PCR, and sequencing were conducted as described (17). To avoid bias, sequencing was done after all 3 stages of treatment had been completed.

SDS-PAGE and immunoblotting were conducted as described previously (18). Used antibodies were ABL (Pharmingen 8E9; BD Biosciences), phosphotyrosine (Upstate Biotechnology 4G10; Biozol; and Transduction PY20; BD Biosciences), and β-actin (AC-15; Sigma-Aldrich).

From two 96-well plates with Ba/F3 Mig p185 cells cultured in the presence of imatinib at 4 μmol/L, 24 resistant cell colonies were picked and expanded to imatinib-resistant cell lines (Fig. 1). In patients treated with imatinib at 400 mg daily, plasma C_max and trough concentrations were 5.26 and 2.46, respectively, at steady state (14). We therefore decided to use 4 μmol/L in our study. All 24 cell lines harbored one BCR-ABL kinase domain mutation, making a total of 10 mutations at 9 sites that corresponded to known imatinib resistance mutations in patients with CML (ref. 1; Fig. 2A). We next exposed 8 replicates of each imatinib-resistant line to dasatinib at 100 nmol/L or nilotinib at 1,700 nmol/L [“optimal” concentrations, corresponding to mean plasma trough levels with nilotinib 400 mg twice daily or dasatinib 70 mg twice daily; refs. (3, 15, 16)], and 50% thereof (“suboptimal” concentrations: dasatinib 50 nmol/L, nilotinib 850 nmol/L). Five of these 24 imatinib-resistant lines were able to give rise to dasatinib-resistant colonies in the presence of second-line dasatinib at 100 nmol/L and 6 of 24 at dasatinib 50 nmol/L (Fig. 2B). As one imatinib-resistant line (no. 10 in Table 1) delivered 2 independent dasatinib-resistant colonies, we were able to generate 9 second-line dasatinib-resistant cell lines. All kept their initial mutations and 4 of them developed additional mutations (Fig. 2C). In contrast, under second-line treatment with nilotinib, 23 of 24 imatinib-resistant lines were able to grow, 13 of 24 with nilotinib at 1,700 nmol/L, and 10 additional lines with 850 nmol/L (Fig. 3A). In total, 24 second-line nilotinib-resistant cell lines were recovered. All of them showed kinase domain mutations, with 22 keeping their initial mutation, 13 developing additional mutations, and 2 switching to a different mutation (Fig. 3B). Third-line treatment of imatinib-dasatinib–resistant cell lines with nilotinib showed growth of third-line–resistant cells in replicates of all 9 cell lines, at both concentrations (Fig. 2E). All lines kept their kinase domain mutations, and only one additional mutation was observed (Fig. 2F). Under third line treatment with dasatinib at 50 and 100 nmol/L, 18 and 12 of 24 second-line nilotinib-resistant cell lines developed dasatinib resistance, respectively (Fig. 3D). Twenty-one imatinib-nilotinib-dasatinib–resistant cell lines were generated, all of them kept their baseline mutations and 7 of them developed new mutations (Fig. 3E). Thus, with respect to the abundance of sequential resistance in vitro, dasatinib was superior to nilotinib at the concentrations used in this study.

When we treated replicates of imatinib-resistant cell lines with nilotinib or dasatinib, we noticed 3 specific, recurrent cellular growth patterns: (i) general growth of all identical replicates (designated as “all” in Tables 1 and 2); (ii) growth of single colonies within single wells (designated as “single”); and (iii) no growth (designated as “none”).

The first pattern implies immediate growth in all replicates despite the presence of a subsequent TKI. In this setting, we found the T315I mutation 3 times each with dasatinib and nilotinib second-line, and the Y253H and E255V P-loop mutations 4 times after nilotinib but not dasatinib (Tables 1 and 2). These findings show that general, immediate second-line resistance is linked to exchanges with high compound-specific IC₅₀ values (refs. 9, 10; see Supplementary Table S1).

The second pattern consisted in the growth of cell colonies in single wells within a series of replicates of imatinib-resistant cell lines in the presence of nilotinib or dasatinib. Sequencing lines that were established from these expanded colonies showed that sequential TKI treatment frequently leads to the acquisition of a second or even third mutation in addition to the original one, resulting in compound mutations, consistent with a gradual increase in inhibitor resistance (Tables 1–3). Examples include the sequence H396P (imatinib, no. 18 in Table 2)—H396P/E255V (nilotinib)—H396P/E255V/T315A (dasatinib), or the mutation sequence E255K (imatinib, no. 13 in Table 2)—E255K/A380S (nilotinib)—E255K/A380S/V299L (dasatinib; cellular IC₅₀ values: see Supplementary Table S1). Of note, second-line double mutations were often sufficient to confer third-line resistance, resulting in immediate growth of all replicates in the sequence imatinib-dasatinib-nilotinib, for example, Y253H/F317L, no. 1 or Y253H/V299L, no. 11 in Table 1, and also in the sequence imatinib-nilotinib-dasatinib, for example, L248V/E255V, no. 2 or G250E/E255K, no. 10 in Table 2. On the other hand, third-line resistance, like second-line resistance was found to be associated with development of double or triple mutations, again resulting in the growth of resistant colonies within single wells displaying additional mutations, for example, the sequence G250E/F317C/A380T, no. 10 in Table 1 or the sequence E255V/F317L, no. 3 in Table 2. In addition, we observed imatinib-resistant cell lines with single mutations that under second-line therapy at low inhibitor concentrations showed growth within single wells and general growth under third-line therapy, with no changes in mutational status, for example, no. 23 with E281K in Table 1; no. 5 and No. 6 with L248V, no. 7 with H396P in Table 2. This suggests that additional resistance mechanisms may be active in these cells, which complement weak mutations to full resistance. In contrast to the frequent appearance of double mutations, switches in mutational status only rarely occurred, with only 2 cases in our study, both under second-line nilotinib (no. 12 and 14 in Table 2), suggesting clonal heterogeneity after selection in the presence of imatinib and clonal selection during second-line treatment. The emergence of novel mutations was strongly associated with a growth pattern of single-cell colonies, suggesting step-by-step acquisition of additional mutations.

The third growth pattern (“no growth”) occurred whenever in vitro first-line–resistant cell lines were successfully inhibited by second-line inhibitors in all replicates, and when second-line–resistant cell lines were not able to give rise to third-line resistance in any of the replicates. Under dasatinib, the majority of imatinib-resistant cell lines were unable to grow, including lines harboring mutations of P-loop (L248V, E255K/V), C-helix (E281K, apart from one case with growth at 50 nmol/L dasatinib, no. 23 in Table 1), substrate-binding region (F359I) and A-loop (H396P). Some lines with imatinib-nilotinib–resistant single (Y253H) and double mutations (F359I/E255K, E281K/Y253H, E281K/F359C and H396P/F311I) were also successfully inhibited by dasatinib. With nilotinib at the lower concentration of 850 nmol/L, complete growth inhibition was achieved only in one case of E281K (no. 22 in Table 2). At the higher nilotinib concentration of 1,700 nmol/L, imatinib-resistant lines with L248V, G250E, E255K, E281K, and H396P mutations were effectively inhibited. However, imatinib-resistant lines expressing Y253H or E255V displayed growth in all replicates, and nilotinib was unable to suppress growth of imatinib-dasatinib–resistant cell lines in this study (Table 1).

We next analyzed the impact of in vitro second- and third-line treatment of imatinib-resistant lines. Of 24 first-line imatinib-resistant cell lines, 16 could be successfully inhibited by second-line dasatinib therapy, whereas 5 showed resistance to dasatinib at 100 nmol/L and an additional 3 were resistant to dasatinib at 50 nmol/L (Fig. 2B). Under second-line therapy with nilotinib, only 1 of 24 first-line imatinib-resistant cell lines was inhibited by nilotinib at both concentrations, whereas 13 were nilotinib-resistant at 1,700 nmol/L and 10 more at 850 nmol/L (Fig. 3A). Of the 24 cell lines already resistant to the sequence imatinib-nilotinib, 5 were effectively inhibited by third-line treatment with dasatinib, 12 were resistant to dasatinib at 100 nmol/L and another 7 were resistant at 50 nmol/L (Fig. 3D). Third-line nilotinib treatment of 9 cell lines already resistant to imatinib-dasatinib was unable to prevent outgrowth, with all 9 showing resistance to nilotinib at 1,700 nmol/L (Fig. 2E). In other words, efficacy of dasatinib was 79% (19 of 24) inhibition at the optimal (clinically achievable) concentration, and 75% (18 of 24) inhibition at the suboptimal concentration for second-line therapy after imatinib, decreasing to 50% (12 of 24) inhibition at the optimal concentration and 25% (6 of 24) inhibition at the suboptimal concentration for third-line treatment after imatinib-nilotinib. In contrast, efficacy of nilotinib was 46% (11 of 24) inhibition at the optimal (clinically achievable) concentration, 4% (1 of 24) inhibition at the suboptimal concentration for second-line therapy, and 0% (0 of 9) inhibition for third-line treatment at both concentrations. Thus, after imatinib resistance in the presence of BCR-ABL resistance mutations in vitro, dasatinib was superior to nilotinib at concentrations that correspond to clinically achievable concentrations, and the activity of second-line nilotinib was critically dependent on optimal inhibitor concentrations.

Next, we examined the evolution of individual mutation types. The prevailing mutational profiles were cross-resistant single mutations and sequentially acquired compound mutations. These 2 patterns emerged with both treatment sequences. T315I, as expected, showed immediate growth in all replicates independent of the treatment sequence. However unexpectedly, although cell lines harboring T315I were selected in the course of treatment, the incidence of T315I did not increase in the course of sequential treatment. There was only one case of T315I emerging during second-line dasatinib treatment, and one case emerging during second-line nilotinib treatment (Tables 1 and 2). In contrast, 4 of 9 dasatinib second-line–resistant cell lines and 13 of 24 nilotinib second-line–resistant cell lines acquired compound mutations. With nilotinib third-line, again 4 of 9 nilotinib-resistant cell lines showed the pre-existing compound mutations with only one newly acquired mutation. Of 21 dasatinib third-line–resistant cell lines, 15 featured compound mutations, and of these, 7 had acquired new mutations (Figs. 2C–G and 3B–F; Tables 1 and 2).

We observed a higher frequency of second-line–resistant lines with nilotinib as compared to dasatinib, particularly at the “suboptimal” concentration of 850 nmol/L (23 of 24 vs. 6 of 24 with dasatinib at 50 nmol/L). These mostly showed P-loop mutations, specifically E255K/V and Y253H that either pre-existed or newly emerged after nilotinib second-line treatment (Fig. 3B–C, Table 1). This observation is consistent with the finding that P-loop mutations are the most abundant cluster of mutations causing imatinib resistance in the clinic (1), and the observation that nilotinib at clinically achievable plasma concentrations displays borderline activity against Y253H and E255V (8). In contrast, mutations at positions V299 and F317, that were associated with resistance to dasatinib in the clinic (11, 12, 19, 20), pre-existed only in one imatinib-resistant cell line (no. 9 in Table 1), but in all the remaining cases occurred after dasatinib second- or third-line, predominantly as compound mutations, and preferentially with pre-existing P-loop mutations (Figs. 2D and 3F; Table 3). Although nilotinib has been shown to be active against V299L and F317C/L/V (8, 11, 21), nilotinib second- or third-line treatment was unable to prevent growth of resistant lines in this particular setting in our system, due to pre-existing P-loop mutations or T315I.

In summary, dasatinib resistance occurred less frequently than nilotinib resistance in previously imatinib-resistant cell lines, mainly due to pre-existing P-loop mutations. Second-line–resistant cell lines were prone to third-line resistance, often accompanied by the acquisition of drug-specific mutations to form compound mutations. With regard to third-line treatment, dasatinib at 100 nmol/L was active in a proportion of nilotinib-resistant cell lines, whereas imatinib- and dasatinib-resistant lines gave rise to nilotinib resistance in all cases. Inhibitor concentration was a critical determinant of resistance development especially with nilotinib.

Dasatinib and nilotinib recently were approved for first-line treatment of chronic phase CML. Nevertheless, the majority of patients with CML in chronic phase still receive first-line treatment with imatinib, and dasatinib and/or nilotinib are given as second- and third-line therapy when imatinib resistance or intolerance occurs. Clinical data suggest that sequential TKI treatment in CML is associated with the emergence of specific BCR-ABL mutations, for example, the pan-resistant T315I mutation and acquisition of drug-specific mutations. In some cases, more than one mutation emerges at the same time as compound mutations, that is, several mutations on the same BCR-ABL transcript, or multiple mutations, that is, in different disease clones (12, 22, 23). In vitro simulation of sequential TKI therapy enabled us to correlate mutational status with cellular growth patterns and to elucidate the evolution of individual mutations depending on the order and concentration of nilotinib and dasatinib sequentially administered in identical replicates of imatinib-resistant cell lines.

In CML, sequential acquisition of kinase domain mutations during TKI therapy poses a challenge in the clinical setting (12, 24). Specific mutations have been classified as second-generation inhibitor clinically relevant mutations, such as F317L/I/C/V, V299L, T315A, and T315I for dasatinib, and the mutations Y253H, E255K/V, F359V/C, and T315I for nilotinib (24). Most of these mutations were observed as newly acquired mutations in a clinical study with 95 patients receiving dasatinib or nilotinib as second or third TKI, and disease progression in 83% of cases was associated with emergence of newly acquired mutations (23). A study of 17 patients who were treated sequentially with imatinib and dasatinib reported newly acquired mutations in all patients; mutations at codons 299 and 317 were observed in 4 and T315I appeared in 12 patients after second-line dasatinib (22). In our in vitro study, serial treatment did not result in a markedly increased incidence of T315I after second- or third-line treatment, with T315I as a new mutation occurring in one case after dasatinib and in one case after nilotinib. As expected, cells harboring the T315I mutation (3 of 24 imatinib lines) showed sustained resistance behavior under sequential TKI treatment. The frequent appearance of additional exchanges at position F317 or V299 after dasatinib second- or third-line treatment in our study is consistent with the appearance of these mutations upon dasatinib exposure after mutagenesis in vitro (6, 7). It is also compatible with the clinical finding that F317L mutation compromises cytogenetic response rates in patients treated with dasatinib after imatinib failure and that F317L and V299L in addition to T315I are the most commonly detected new mutations after dasatinib failure (11). A parallel relationship can be established for nilotinib and P-loop mutations, as the frequent occurrence of P-loop mutations after second-line treatment with nilotinib in our study is compatible with development of P-loop mutations in vitro (8, 25). P-loop mutations (E255K/V, G250E, Y253H), along with T315I and F359C/V, were the most frequently detected new mutations in patients at the time of disease progression receiving therapy with nilotinib (13).

Activity of third-line TKI treatment in CML chronic phase is limited, with progression-free survival of less than 1 year (26). Also, it has been suggested that compound mutations modulate the response to inhibitors in unexpected ways, thereby making prediction of response more complex (24). In this context, our findings may help to make predictions for response to second- and third-line ABL inhibitor therapy. All cases of compound mutations emerging after second-line therapy with dasatinib at the same time caused third-line nilotinib resistance (Table 3). All but one of them consisted of a mutation at position F317 or V299, explaining dasatinib resistance, and a P-loop mutation, explaining the subsequent nilotinib resistance. In contrast, third-line dasatinib was able to suppress growth of imatinib-nilotinib–resistant lines with compound mutations in 4 of 13 cases. Translated to the clinic, application of third-line dasatinib after imatinib-nilotinib failure in patients with compound mutations might be a sustainable option. It has been shown that nilotinib does have clinical activity in CML following failure of imatinib and dasatinib, including patients with pre-existing mutations such as F317L (27). Thus, our in vitro system might underestimate the activity of nilotinib in the third-line setting after imatinib and dasatinib failure in patients with CML. However, our results strengthen the importance of individual, mutation-based decisions for second- and third-line treatment and optimal nilotinib dosing in imatinib-resistant, BCR-ABL mutant CML.

The high frequency of compound mutations we detected suggests that the scenario of “multiple mutations” emerging during sequential TKI therapy in a proportion of patients might be explained by compound mutations in sequentially resistant disease clones, making sequential monotherapy a much less effective choice. In line with this, a recent study identified compound mutations in sequentially imatinib-dasatinib–treated patients with Ph+ leukemia (22). A different study addressed the frequency of compound mutations in the clinic (28). In samples from patients treated with various ABL-TKIs, as much as 70.2% (33 of 47) of double mutations detected by direct sequencing were identified as compound mutations. Moreover, imatinib-resistant patients with CML with multiple mutations were identified as a poor-risk subgroup (29). The presence of multiple mutations adversely affected response to second-line nilotinib or dasatinib and favored the emergence of new mutations as detected by standard sequencing, with 10 of 25 cases harboring more than one mutation. Thus, sequential TKI monotherapy predisposes for acquisition of compound mutations. Our data show that compound mutations are acquired in a step-by-step manner and that specific treatment sequences determine the specific composition of compound mutations. Therapeutic strategies minimizing the occurrence of sequentially acquired compound mutations may consist in upfront combination therapy of several TKIs, which is limited by toxicity, or first-line therapy with nilotinib or dasatinib. In imatinib-resistant, BCR-ABL mutant CML, therapy should be individualized according to mutation status, taking into account the effectiveness of dasatinib against P-loop mutations and of nilotinib against mutations at codons 299 and 317. This might well include sensitive methods of detection, as multiple mutations might emerge at any time, may coexist at low level and might subsequently be selected during second- or third-line treatment. Our findings suggest that during sequential TKI treatment beginning with imatinib and followed by nilotinib or dasatinib, the risk of second- and third-line resistance due to the acquisition of nilotinib- and dasatinib-resistant compound mutations composed of P-loop or F359 together with F317 or V299 exchanges might be a more common problem than acquisition of T315I.

In our study, TKI resistance showed to be strongly associated with BCR-ABL kinase domain mutations. It is known that other mechanisms such as BCR-ABL overexpression, gene amplification, or upregulation of transport proteins can contribute to TKI resistance (1, 30). Our in vitro system might overemphasize mutation-dependent resistance. Although we intended to preferentially depict mutation-dependent TKI resistance, we did not incorporate a mutagenesis step and thus allow mutation-independent mechanisms to contribute to resistance (31). Also, in a recent analysis sequential resistance to a second or third TKI was associated with the emergence of BCR-ABL kinase domain mutations in 83% of the cases (23), and it was shown that patients who already harbor BCR-ABL kinase domain mutations are prone to develop additional mutations with sequential TKI resistance (12, 23).

In conclusion, our results illustrate that in vitro growth patterns of sequentially TKI-resistant cell lines correlate with distinct mutation profiles. The composition of compound mutations depended on inhibitor type and order of application. Transferred to the clinic, sequential TKI treatment beginning with imatinib and followed by nilotinib or dasatinib might more often be associated with stepwise acquisition of compound mutations composed of P-loop or F359 together with F317 or V299, which might be a more common problem than acquisition of T315I. The activity of third-line TKI treatment is strongly limited by both compound mutations and T315I.

J. Duyster has honoraria from speakers' bureau and is a consultant/advisory board member of Novartis. N. von Bubnoff is a consultant/advisory board member of Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Duyster, N. von Bubnoff

Development of methodology: R.C. Bauer, J. Duyster, N. von Bubnoff

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.C. Bauer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.C. Bauer, J. Duyster, N. von Bubnoff

Writing, review, and/or revision of the manuscript: R.C. Bauer, C. Peschel, N. von Bubnoff

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Sanger

Study supervision: J. Duyster, N. von Bubnoff

This work was supported by a grant to N. von Bubnoff and J. Duyster from the Bundesministerium für Bildung und Forschung (NGFNplus) and by a grant to N. von Bubnoff by the Kommission für Klinische Forschung (KKF), Technische Universität München.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.