Rapid clonal shifts in response to kinase inhibitor therapy in chronic myelogenous leukemia are identified by quantitation mutation assays

Abstract

Treatment of CML with the tyrosine kinase inhibitor (TKI) imatinib mesylate results in the emergence of point mutations within the kinase domain (KD) of the BCR‐ABL1 fusion transcript. The introduction of next‐generation TKIs that can overcome the effects of some BCR‐ABL1 KD mutations requires quantitative mutation profiling methods to assess responses. We report the design and validation of such quantitative assays, using pyrosequencing and mutation‐specific RT‐PCR techniques, to allow sequential monitoring and illustrate their use in tracking specific KD mutations (e.g. G250E, T315I, and M351T) following changes in therapy. Pyrosequencing and mutation‐specific RT‐PCR allows sequential monitoring of specific mutations and identification of rapid clonal shifts in response to kinase inhibitor therapy in CML. Rapid reselection of TKI‐resistant clones occurs following therapy switch in CML. (Cancer Sci 2010)

Mutations in the ABL1 kinase domain (KD) represent the most common mechanism of therapy resistance in Philadelphia chromosome (Ph)‐positive chronic myelogenous leukemia (CML), arising in response to targeted therapy of BCR‐ABL1 with the tyrosine kinase inhibitor (TKI) imatinib mesylate.^{( 1} , ² , ³ , ⁴ , ⁵ , ^{6 )} More than 90 different ABL1 KD mutations involving 50 different amino acids have now been described, but only a handful account for a cumulative 70% incidence.^{( 1} , ⁶ , ^{7 )} The most commonly detected mutations are those involving the ATP binding (P)‐loop (codons 248–255, particularly G250E, Y253H/F, and E255K/V together comprising approximately 50%), as well as those involving the imatinib binding region (particularly amino acids T315I and F317L), the catalytic domain (particularly amino acid M351T), and the activation (A)‐loop (particularly amino acid H396R/P).

With the availability of a range of new TKIs, including nilotinib and dasatinib,^{( 8 )} screening for ABL1 KD mutations in imatinib‐resistant CML is now routinely used to help guide therapy selection.^{( 1} , ^{7 )} Therefore, detection of changes in the relative levels of a particular mutated transcript following the switch to a new TKI can be helpful in tailoring treatment regimens for individual patients. Here we report the design and validation of rapid quantitative assays to monitor the levels of mutated BCR‐ABL1 transcripts, and show that rapid reselection of TKI‐resistant clones occurs following therapy switch in CML.

Materials and Methods

Patients’ characteristics.  The study was carried out according to an institutional review board‐approved laboratory protocol and followed the provisions of the Helsinki Accord. Informed consent was obtained from the subjects. Bone marrow aspirate or peripheral blood specimens were obtained from patients with imatinib‐resistant CML and Ph+ precursor B‐lymphoblastic leukemia/lymphoma.^{( 1} , ³ , ^{9 )} Sequential samples from patients with imatinib‐resistant CML following TKI switch were those previously reported by our group.^{( 1 )} Criteria for secondary imatinib resistance that triggered mutation screening included the re‐emergence of the Philadelphia chromosome following attainment of complete cytogenetic response and a rising BCR‐ABL1/ABL1 transcript ratio equivalent to 2‐log increase.

Sample preparation.  Total leukocytes were isolated by centrifugation following red blood cell lysis and total RNA was extracted using TRIzol reagent (Gibco‐BRL, Gaithersburg, MD, USA). cDNA was synthesized using random hexamer primers and Superscript II reverse transcriptase (Invitrogen, Carlsbad, CA, USA). RNA quality was assessed by agarose gel electrophoresis prior to cDNA synthesis.

TaqMan quantitative RT‐PCR.  Levels of BCR‐ABL1 fusion transcripts were quantitated in a multiplex real‐time RT‐PCR assay that simultaneously detects e13a2 (b2a2), e14a2 (b3a2), and e1a2 transcript types using 1 μg cDNA equivalent.^{( 10 )} BCR‐ABL1 levels were normalized to total ABL1 transcript levels. Quantitative PCR was carried out on 7700 or 7900 instruments from Applied Biosystems (ABI; Foster City, CA, USA) with post‐PCR sizing of transcripts by capillary electrophoresis on a 3700 or 3130 genetic analyzer (ABI).^{( 11 )} Positive controls for each transcript type (BCR‐ABL1+ cell lines KBM7, K562, and B15) and sensitivity controls (RNA from BCR‐ABL1+ cell lines diluted into HL‐60 RNA) were included on each run.

Direct sequencing of ABL1 KD.  A nested PCR sequencing approach was used for direct sequencing of the ABL1 KD, with a first‐round amplification of the BCR‐ABL1 transcript followed by two separate PCR reactions that covered codons 221–380 and codons 350–500 of the ABL1 KD, respectively.^{( 1 )} For the first‐round PCR, the forward primer for b2a2/b3a2 was 5′‐ACAGCATTCCGCTGACCAT‐3′, the forward primer for e1a2 was 5′‐CTCGCAACAGTCCTTCGAC‐3′, and the reverse primer for all transcripts was 5′‐TCCACTTCGTCTGAGATACTGGAT‐3′, with 25 cycles. Primers for the second‐round PCR were 5′‐M13 tag‐CGCAACAAGCCCACTGTCT‐3′ and 5′‐M13 tag‐GCTGTGTAGGTGTCCCCTGT‐3′ for codons 221–380, and 5′‐M13 tag‐CGCAACAAGCCCACTGTCT‐3′ and 5′‐M13 tag‐ACTCCAAATGCCCAGACGTC‐3′ for codons 350–500.

Standard dideoxy chain‐termination DNA sequencing was carried out using M13 primers and Big Dye chain terminator reagents with detection on 3700 or 3100 genetic analyzers (ABI) and analysis using Sequence Analysis and SeqScape software (ABI).^{( 3 )} All mutations were confirmed by sequencing of forward and reverse strands, with a sensitivity of 20% mutation‐bearing transcripts in the analyzed population established by periodic dilution studies.

Quantitative BCR‐ABL1 pyrosequencing.  A set of PCR amplicons with associated sequencing primers was designed for pyrosequencing of the four ABL1 KD mutation hotspots (Fig. 1). First‐round BCR‐ABL1 PCR was carried out, as above, followed by second‐round PCR for each KD mutation hotspot including one biotin‐labeled primer (Table 1). Prior to pyrosequencing, single‐stranded PCR products were isolated according to the manufacturer’s recommendations by binding to streptavidin‐labeled beads that were captured by filter probes, followed by fixation in 70% ethanol, denaturation in 0.2 M NaOH, washing in 10 mM Tris–acetate (pH 7.6), and releasing product into 12 μL annealing buffer containing 0.3 μM sequencing primers. Pyrosequencing was carried out on a PSQ HS 96 Pyrosequencer (Biotage, Uppsala, Sweden) using nucleotide dispensation tips and PyroGold or PyroMark reagents. Quantitation of mutated/unmutated transcripts was done by comparing mutated and unmutated base peak heights. Sensitivity for detecting the mutation status of the BCR‐ABL1 transcript was determined by dilution of RNA from samples with only the mutated sequence detected by direct sequencing into RNA from the BCR‐ABL1‐negative cell line HL‐60 (Fig. 2a–d).

Figure 1.

 Locations of pyrosequencing primers and ABL1 kinase domain mutations detected. The location of the pyrosequencing primer sets is mapped on a schematic of the ABL1 kinase domain. Eighty‐one imatinib‐resistant blood or bone marrow samples with mutation detected by direct Sanger sequencing were compared. Asterisks indicate that pyrosequencing and direct sequencing both detected the mutation. Black dots indicate a mutation not clearly detected by pyrosequencing but seen by direct sequencing. A, activation loop; C, catalytic domain; P‐loop, ATP binding loop (codons 248–255).

Table 1.

 Polymerase chain reaction primer sets and sequencing primers used for ABL1 kinase domain mutation quantification by pyro‐sequencing

PCR set	Codons	Common mutations	PCR primers†	Sequencing primers
1	236–274	M244V, L248V, G250E, Q252H, Y253H, E255V	F: 5′‐AGATGGAACGCACGGACAT‐3′ R: 5′‐B‐CTTCAAGGTCTTCACGGC‐3′	5′‐AGATGGAACGCACGGACAT‐3′ (same as F)
2	300–349	F311I, T315I, F317L	F: 5′‐AGCTCCTTGGGGTCTGCA‐3′ R: 5′‐B‐TGACGAGATCTGAGTGGC‐3′	5′‐AGCTCCTTGGGGTCTGCA‐3′ (same as F)
3	343–392	M351T, E355G, E355A, F359C, F359V	F: 5′‐TGGCCACTCAGATCTCGTC‐3′ R: 5′‐B‐TGTCCCCTGTCATCAACCT‐3′	5′‐TGGCCACTCAGATCTCGTC‐3′ (same as F)
4	343–422	L364I, H396R	F: 5′‐TGGCCACTCAGATCTCGTC‐3′ R: 5′‐B‐ACGTCGGACTTGATGGAG‐3′	5′‐ATGACAGGGGACACCTACAC‐3′ (for H396R only)

†The primers listed are for second‐round PCR. F, forward primer; R, reverse primer; B, biotin‐tagged.

Figure 2.

 Correlation of mutation detection using mutation‐specific PCR and pyrosequencing. (a) Sensitivity of assay for mutation detection by pyrosequencing for T315I was established by dilution study on involved patient sample and compared to expected dilution curve. RNA from a patient sample with predominantly mutated transcript (approximately 75%) was diluted into HL‐60 RNA. (b) Correlation of relative mutation levels in quantitative (q)RT‐PCR and pyrosequencing assays was tested in 14 patient samples. Dotted lines indicate 95% confidence intervals. (c) Sensitivity of assay for mutation detection by pyrosequencing for G250E was established by dilution study on involved patient sample and compared to expected dilution curve. RNA from a patient sample demonstrating only mutated G250E transcript was diluted into HL‐60 RNA. (d) Correlation of relative mutation levels in qRT‐PCR and pyrosequencing assays was tested in 18 patient samples. (e) Precision of the T315I pyrosequencing assay quantitation was assessed in 98 PCR reactions split at the cDNA step and separately amplified in first‐ and second‐round PCR and pyrosequenced.

Quantitative BCR‐ABL1 mutational analysis by mutation‐specific PCR.  Mutation‐specific quantitative (q)RT‐PCR assays were designed to detect the mutated and unmutated sequences for the G250E, T315I, and M351T KD mutations. Primers and probes were as in Table 2, with TaqMan probes labeled with 6‐carboxyfluorescein and a 3′ minor groove binder/non‐fluorescent quencher. Real‐time PCR reactions were carried out on the 7900HT Sequence Detection System (ABI) using 0.2 μg cDNA template, 0.4 μM primers, 0.4 μM probe, 1× universal master mix, and AmpliTaq Gold polymerase. Ratios of mutated to unmutated transcripts were calculated by relative quantification by the delta‐delta C_t method using the threshold cycle. A standard curve was generated on each run by diluting RNA from positive control samples into HL‐60 RNA.

Table 2.

 Mutation‐specific quantitative PCR primers and probes

Mutation	PCR primers	TaqMan probes (6FAM‐labeled)	PCR conditions
G250E	F: 5′‐ATGGAACGCACGGACATCA‐3′ R: 5′‐TCTTCCACACGCCCTCGTA‐3′	Unmutated probe: 5′‐6FAM‐GTACTGGCCCCCGCCCAGCTT‐MGB‐3′ Mutant probe:† 5′‐6FAM‐GTACTGGCCCTCGCCCAGCTT‐MGB‐3′	Start: 50°C 2 min/95°C 10 min Step 1: 95°C 30 s Step 2: 68°C 30 s Step 3: 72°C 30 s 40 cycles
T315I	F: 5′‐TCTGCACCCGGGAGCCCCCGT‐3′ R: 5′‐AGTCCAGGAGGTTCCCGTAGG‐3′	Unmutated probe: 5′‐6FAM‐ATCATCACTGAGTTC‐MGB‐3′ Mutant probe:† 5′‐6FAM‐ATCATCATTGAGTTC‐MGB‐3′	Start: 95°C 10 min Step 1: 95°C 15 s Step 2: 58°C 30 s 40 cycles
M351T	F: 5′‐TGCTGTACATGGCCACTCAGA‐3′ R: 5′‐TGTGGATGAAGTTTTTCTTCTC‐3′	Unmutated probe: 5′‐6FAM‐GTACTCCATGGCTGA‐MGB‐3′ Mutant probe:† 5′‐6FAM‐GTACTCCGTGGCTGA‐MGB‐3′	Start: 50°C 2 min/95°C 10 min Step 1: 95°C 30 s Step 2: 61°C 30 s Step 3: 72°C 30 s 40 cycles

†These probes can also be labeled with VIC for multiplex detection with unmutated sequence in the same well. 6FAM, 6‐carboxyfluorescein. The nucleotide that is different in mutant probe as compared with unmutated probe is highlighted in bold.

Results

Comparison of qualitative detection of ABL1 KD mutations by pyrosequencing with direct Sanger sequencing.  Table 1 summarizes the PCR and sequencing primers for pyrosequencing detection of the most common ABL1 KD mutations. Both pyrosequencing and direct Sanger sequencing used the same first‐round PCR amplification of the BCR‐ABL1 transcript, followed by different second‐round PCR reactions. As shown in Figure 1, targeted pyrosequencing detected identical KD mutations demonstrated by direct sequencing in 78/81 imatinib‐resistant CML/Ph+ precursor B‐lymphoblastic leukemia/lymphoma samples, with discordant cases being two with G250E, and one with E255K. All three discordant cases showed both mutated and unmutated amino acid (with the mutated much lower than unmutated) by direct sequencing. We also analyzed 56 imatinib‐resistant samples where no mutation was detected by direct sequencing, including 28, 27, 13, and 11 samples tested with PCR sets 1, 2, 3, and 4, respectively. Two mutations were detected by pyrosequencing (T315I at 67% and M351T at 29%) that were not detected by direct sequencing.

Comparison of pyrosequencing quantitation with mutation‐specific qRT‐PCR.  The dynamic range and sensitivity of the pyrosequencing assays for common mutations in the four primer sets was determined by dilution studies to be approximately 5% (Fig. 2a,c, and not shown). For all assays, replicates were done from cDNA through nested PCR to establish the reproducibility of quantitation (Fig. 2e). The few outlier discordant samples (with the difference in mutated/unmutated ratio in the replicate study >0.1) were those with either poor quality RNA, or with low BCR‐ABL1 transcript levels (<0.01), which may cause unequal amplification of the fusion transcript in the first‐round PCR.

For three selected KD mutations (G250E, T315I, and M351T), we compared the relative quantification of mutated transcripts obtained by qRT‐PCR using mutation‐ and unmutated‐specific TaqMan probes with the pyrosequencing method. All three qRT‐PCR assays showed good linearity down to 0.1% mutated transcript, as established by dilution studies (not shown). There was a good correlation for the %mutated/unmutated ratios for the 18 samples tested for T315I by both methods (Fig. 2b; R = 0.99, P < 0.0001). Concordance was also seen for the 24 samples with M351T tested by both methods, with qualitative discordance seen in only 2/24 (8.3%, both low level) and a strong correlation between quantitative levels (R = 0.83, not shown).

However, there was a relatively poor correlation for G250E quantitation by the two methods (R = 0.75), with qRT‐PCR detecting higher levels of mutated product (mean 84.27) than pyrosequencing (mean 48.76; Fig. 2d).

Tracking shifts in mutated BCR‐ABL1 transcripts following shifts in therapy.  The pyrosequencing assays were used to track the level of mutated BCR‐ABL1 transcripts in sequential samples from patients with imatinib‐resistant CML who were switched to another TKI. Thirty‐one patients with KD mutations at the time of imatinib‐resistant disease were analyzed for levels of mutated/unmutated ratio just prior to TKI switch and then at 4–6 weeks and 3–6 months after TKI switch. Quantitative RT‐PCR was also carried out if available for the target mutation. If regression of the mutation was observed, resequencing of the entire kinase domain was done by the Sanger method to look for additional KD mutations.^{( 1 )}

As shown in Figure 3, pyrosequencing could detect rapid changes in the level of mutated BCR‐ABL1 transcript occurring within 4–6 weeks of the TKI switch. Among the 31 patients tested before and after TKI switch, several distinct patterns of response of the mutated clone were seen. In 15 patients, the imatinib‐resistant KD‐mutated clones were not detectable by pyrosequencing at 3–6 months post switch to new TKI (nilotinib, dasatinib, or bosutinib). However, second mutated clones were observed over 3–6 months in four patients switched from imatinib to dasatinib or nilotinib, which coincides with the re‐emergence or persistence of disease as measured by BCR‐ABL1 transcript levels (Fig. 4a, and not shown). This pattern was manifested by loss of one mutated transcript and emergence of another within 3–6 months of switch.

Figure 3.

 Pyrosequencing assays to quantitate relative levels of mutated transcript in sequential samples. (a) M351T (ATG to ACG) mutation detected in an imatinib‐resistant sample, and in two follow‐up samples, at 10 weeks (b) and 12 weeks (c) following switch to nilotinib. Quantification of mutated/unmutated ratios by mutation‐specific quantitative RT‐PCR carried out on the same samples is shown for comparison in parentheses.

Figure 4.

 Patterns of mutation switch in sequential samples tested by pyrosequencing. (a) A case of imatinib‐resistant CML in late chronic phase with M351T kinase domain mutation showed complete regression of mutation on switch to nilotinib. This was followed by emergence of a second resistant clone with F359V mutation. TKI, tyrosine kinase inhibitor. (b) A patient with imatinib‐resistant CML showed rapid regression of M351T mutation but minimal response of G250E after the switch to nilotinib. The pattern of response is consistent with the differential sensitivity to nilotinib of BCR‐ABL1 with M351T compared to G250E. (c) A patient with CML who presented in blast crisis and responded to imatinib before developing T315I 24 months after presentation. The patient was switched to dasatinib, then to another kinase inhibitor (MK‐0457)^{( 14 )} before having a response to homoharringtonine (HHT).^{( 15 )} Quantitation was done by pyrosequencing assays.

A second pattern of response was seen in four patients who had two predominant KD mutations detected at the time of imatinib resistance. Upon the switch to a new TKI, there was increase in the level of one mutation, and decrease in the other, consistent with outgrowth of one mutated CML clone in response to the reselection imposed by the new TKI (Fig. 4b). A third response pattern was seen in eight patients (four T315I, three Y253H, and one G250E) who had KD mutations that were predicted to be cross‐resistant to both imatinib and the new TKI, and showed either persistence or slow regression of the level of the mutated/unmutated clone (Fig. 4c, and not shown).

Discussion

We show that pyrosequencing provides a rapid method to detect and quantitate KD mutations in BCR‐ABL1 transcripts and track responses to TKIs following switch in therapy. For most mutations, the low level of mutation detection was 5% mutation‐bearing transcripts although in practice this was affected by a number of issues that are discussed below. The qualitative results were similar to the results previously obtained with direct sequencing in imatinib‐resistant samples, including those patients who had no mutation detected by the Sanger method. This suggests that in most patients with imatinib‐resistant CML, the dominant mutated transcript is present at levels above the 20% sensitivity of the direct Sanger sequencing method. Therefore, direct sequencing by the Sanger method appears sufficiently sensitive for routine screening.

Both pyrosequencing and mutation‐specific qRT‐PCR provide quantitation that allows tracking of mutation levels in sequential samples, providing an improvement over direct sequencing once a mutation was detected. We show that mutation‐specific qRT‐PCR, although requiring optimization of PCR conditions and expensive mutation‐specific probes, is a sensitive method for relative quantification of mutation levels. A major limitation of TaqMan qRT‐PCR assays is the requirement for design of a separate probe for each specific nucleotide shift observed in each of the over 90 different KD mutations, whereas pyrosequencing can detect and quantitate multiple mutations with a single PCR assay.

A problem encountered with quantitation of some mutations by pyrosequencing (particularly codons 250 and 255 in PCR set 1) was related to the difficulty in detecting and quantifying of mutations that arise at codons that span polynucleotide stretches. For example, the lower level of mutation detected by pyrosequencing in G250E was related to the polyguanosine track spanning codon 250. As these repetitive nucleotides will generate high pyrosequencing peak heights, the ability to recognize a mutation and accurately quantitate the level of mutated product is usually compromised. It is difficult to define the number of nucleotides in a polynucleotide stretch that will cause problems. It depends on the intrinsic background of the assay, which will differ for each primer pair. However, in our experience, a polynucleotide stretch with greater than four identical nucleotides in a row is usually problematic, and qRT‐PCR might be a better choice of quantitation for some of these sites.

Other issues occasionally encountered in pyrosequencing that limit sensitivity are sporadic/artifactual or background peaks that affect quantitation and limit assay sensitivity. There are a number of causes for such false signals, including sequencing primer dimers, template loops, dispensation failures due to tip clogging, bead carryover from adjacent wells, breakdown of nucleotides, and the background “A” (affect low level detection limit) signal intrinsic to the assay chemistry. For these reasons, it is important to run adequate controls. During the assay validation, a sequencing control without any PCR product is essential to control for signal simply due to template loops. For routine runs, we also include a non‐template PCR control, and dilution sensitivity controls at 1:2, 1:4, and 1:16.

These background issues in pyrosequencing assays are most critical when the amount of authentic PCR product is low. For this reason, we used a nested PCR reaction to boost the amount of product coming from the BCR‐ABL1 transcript (as opposed to the non‐translocated ABL1 allele). However, sequencing in cases with low levels of BCR‐ABL1 expression can lead to discordant results due to preferential amplification early in PCR. If a small subset of clones carries a mutation in the ABL1 portion of the BCR‐ABL1 mRNA it is possible only the unmutated mRNA will be amplified due to competition in early PCR cycles. False negatives can occur, as observed when duplicate samples with low BCR‐ABL1 expression were amplified (outliers in Fig. 2e).

We used the pyrosequencing assays developed here to show that several distinct patterns of response occur in imatinib‐resistant CML when a switch to a new TKI is carried out. For those CML cases where the imatinib‐resistant KD mutation was predicted to be responsive to the new TKI (usually nilotinib or dasatinib), rapid regression of the mutated clone could be observed within weeks, as we^{( 1 )} and others^{( 7} , ^{12 )} have shown previously. However, in a subset of patients, reemergence of a second mutated clone occurred within 3–6 months of the switch and was associated with re‐emergence of significant disease levels and therapy resistance. In patients with pan‐resistant mutations, such as T315I, quantitative pyrosequencing assays can be used to track responses to new agents, such as third generation kinase inhibitors and biologic agents. Thus, quantitative profiling of mutation levels can serve to confirm responses to new TKIs,^{( 1 )} assess the biological significance of an identified mutation,^{( 13 )} and to screen for new agents that may be effective.