Background
Treatment of chronic myeloid leukemia (CML) has advanced with the introduction of tyrosine kinase inhibitors (TKI) that target the
BCR-ABL1 fusion protein such as imatinib, and furthermore with second line inhibitors such as dasatinib, nilotinib, bosutinib and ponatinib. To measure the effect of TKI therapy, real-time quantitative PCR (RQ-PCR) of the
BCR-ABL1 fusion transcript is routinely performed and transcript levels are followed longitudinally for each patient. However, in case of limited TKI response or of progression to accelerated phase or blast crisis, mutational analysis of the ABL1 kinase domain should be performed, as stated by the ELN (European Leukemia Net) recommendations [
1], since evolution of such mutations may lead to poor response to TKIs. One mutation of particular importance for clinical investigations is the multi-resistant substitution T315I, resulting in an amino acid change within the p-loop binding site. Furthermore, rare mutations within the regulatory domain of
ABL1 have also been reported to lead to TKI resistance in patients without kinase domain mutations [
2]. A further concern is the presence of concurrent
BCR-ABL1 mutations, which may also hamper successful therapy [
3-
5]. Ideally, mutations in both regulatory and kinase domains as well as co-existing mutations should therefore be detected as early as possible, prior to an expansion of resistant clones. In addition to point mutations, the
BCR-ABL1 protein can be affected by alterations in splicing where whole exons, or smaller parts of exons, are included or skipped from the main transcript [
6,
7]. The clinical significance of splice isoforms remains to be elucidated, mainly because their detection has until recently required time consuming cloning steps prior to sequencing.
Today, various assays including Sanger sequencing and quantitative RT-PCR are routinely applied for
BCR-ABL1 mutation detection. While Sanger sequencing has limited sensitivity, real time reverse transcription PCR requires mutation specific panels with separate standard curves and variable sensitivity. A further limitation is that these assays can typically not resolve the patterns of co-existing mutations. With the introduction of massively parallel sequencing (MPS) technologies it is now possible to study these mutations at an entirely new level of resolution. In recent studies performed on the Roche 454 system,
BCR-ABL1 mutations were detected at a higher sensitivity as compared to Sanger sequencing [
8,
9]. However, although the 454 system produces longer sequences than most other instruments, these still cannot span the complete transcript. Thus, MPS studies have until now mainly been based on sequencing of smaller fragments of
BCR-ABL1, often amplified in two successive rounds using a nested PCR approach. This strategy not only limits the analysis to a portion of the transcript but it is likely to introduce a bias in the resulting mutation frequencies.
Here we present for the first time an assay to directly investigate the entire 1,578 bp BCR-ABL1 major fusion transcript, amplified from a single PCR reaction and sequencing on the Pacific Biosciences (PacBio) RSII system. When comparing available MPS platforms, the PacBio instrument is particularly attractive for BCR-ABL1 analysis. In addition to enabling a rapid workflow at a relatively low cost, the PacBio system produces reads sufficiently long to span across a full length BCR-ABL1 molecule. This allows for an immediate detection of compound mutations and splice isoforms.
Methods
Patient samples
Six patients diagnosed with CML at Uppsala University Hospital, all receiving imatinib as first line treatment were included in this study. All six patients showed limited or no molecular response to TKI treatment. Two of the patients were included in a ponatinib study (PACE, study nr AP24534-10-201, phase 2 clinical trial, Ariad Pharmaceuticals, MA, USA). Samples at diagnosis, and following TKI therapy were tested. For a complete list of all patient samples sequenced in this study, see Table
1. The clinical characteristics of each patient are given in further detail in the results section.
Table 1
Characteristics of the patient samples included in this study
Patient 1
| 43/M | high risk | 46,XY,t(9;22)(q34;q11)[20] | CP CML | 0 | 37432 | wildtype | ND |
| | | ND | CP CML | 7 | 32703 | T315I (3.9%) | wildtype |
| | | ND | CP CML | 9 | 30251 | T315I (53,5%) | ND |
Patient 2
| 70/M | high risk | 46,XY,t(9;22)(q34;q11)[25] | CP CML | 0 | 23089 | wildtype | ND |
| | | 46,XY,t(9;22)(q34;q11)[19]/ | | 6 | 27633 | wildtype | ND |
| | | 46,XY,t(9;22)(q34;q11)[4]/46,XY,idem,del(11)(q14)[16] | AP CML | 43 | 34467 | wildtype | wildtype |
| | | ND but later sample shows 46,XY,t(9;22)(q34;q11)del(11)(q14)[20] | | 64 | 41963 | T315I (98%) | T315I |
Patient 3
| 65/M | high risk | 46,XY,t(9;22)(q34;q11)[25] | CP CML | 0 | 35377 | wildtype | ND |
| | | 46,XY,t(9;22)(q34;q11)[24]/46,XY[1] | CP CML | 3 | | wildtype | ND |
| | | 46,XY,t(9;22)(q34;q11)[6]/47sl,i(17)(q10),add(20)(p13),+mar[4]/46,XY[10] | AP CML | 49 | 39685 | T315I (88,9%), F359C (4,2%) | T315I |
| | | 46,XY,t(9;22)(q34;q11)[3]/46,XY,del(5q)[6]/46,XY[10] | AP CML | 55 | 42642 | T315I (94,8%), F359C (2,2%), D276G (1,8%), H396R (1%) | T315I |
Patient 4
| 61/F | high risk | 46,XX,t(9;22)(q34;q11)[25] | CP AML | 0 | 36658 | wildtype | ND |
| | | 46,XX[20] | CP CML | 111 | 41922 | F359I (83,1%), T315I (13,5%) | F359I, T315I |
Patient 5
| 66/M | | ND | CP CML | 0 | 24062 | wildtype | ND |
| | | 46,XY,del(6)(q2?1;q2?3),-7,t(9;22)(q34;q11)[20] | Blast crisis | 4 | 28446 | Y253H (94,8%), E255V (1,8%) | Y253H |
Patient 6
| 65/M | | 46,XY,t(9;22)(q34;q11)[20] | CP CML | 0 | 32982 | wildtype | ND |
| | | 46,XY,t(9;22)(q34;q11)[11]/ | CP CML | 7 | 29221 | 4 isoforms | wildtype |
| | | 46,XY[20] | CP CML | 13 | 34726 | 3 isoforms | ND |
Ethics statement
This study was performed in accordance with the Declaration of Helsinki. The Ethical Committee at Uppsala University, Dnr 00–623, approved this study. Written informed consent was obtained from the patients.
RNA extraction and cDNA synthesis
RNA was extracted from peripheral blood or bone marrow samples using a TRIzol®, (Thermoscientific, MA, USA) standard protocol and quantified by the nanodrop 2000 instrument (Thermoscientific, MA, USA). cDNA was synthesized using the SMARTer™ PCR cDNA synthesis kit (ClonTech, CA, USA), using 1000 ng total RNA.
Dilution series of BCR-ABL1 samples
Using a quantitative real time reverse transcription PCR assay, total BCR-ABL1 p210 transcripts were quantified in the 49 month post diagnosis sample for patient three (P3) and a wild type sample at diagnosis. The two samples where then diluted to contain the same amount of BCR-ABL1 copies/microliter. The P3 sample was then serially diluted into the sample wild type at varying amounts 50%, 10%, 1% and 0.5%.
Library preparation and PacBio sequencing
Long range PCR amplification of the BCR-ABL1 p210 transcript was performed using the Clontech Advantage PCR kit (Clontech, CA, USA). Using BCR-ABL1 specific primers (BCR exon 12, forward 5’- tga cca act cgt gtg tga aac tc – 3’ and ABL1 exon9/10, reverse 5’ - tcc act tcg tct gag ata ctg gat t - 3’) a 1578 bp cDNA amplicon ranging from exon 12 (including e13 and e14 breakpoint variants) in BCR to exon 9 in ABL1 was obtained. The cDNA reaction was diluted to 50 μl using TE buffer. 5 μl of diluted cDNA were used in a 50 μl PCR reaction, following the manufacturers instructions. Samples were placed into a preheated 95°C thermocycler and cycled as follows; initial denaturation for 1 minute at 95°C followed by 30 cycles of 15 seconds at 95°C, 30 seconds at 61°C and 3 minutes at 68°C. Following amplification, amplicon size was confirmed using the bioanalyser 12000 kit (Agilent, CA, USA) and the concentration confirmed by Qubit assay (Life Technologies, CA, USA).
SMRTbell™ libraries were produced using the Pacific Biosciences 1.0 template preparation kit according to the manufacturer’s instructions. SMRTbells™ were constructed and sequenced following the recommended pacific biosciences 2kb template preparation protocol. In brief, cDNA amplicons (300–750 ngs) underwent end-repair and adaptor ligation processes to generate SMRTbell™ libraries for circular consensus sequencing. Libraries were then subjected to exo treatment and Ampure bead wash procedures for clean up. SMRTbell™ libraries were quantified using the Qubit assay and library size was confirmed using the bioanalyser 12000 kit. Following SMRTbell™ construction, v2 primers and P4 polymerase were annealed and enzyme bound complexes attached to magnetic beads for loading. Each SMRTbell™ amplicon library was loaded on to one SMRT cell and sequenced on the PacBio RS II instrument using C2 chemistry and a 120 minute movie time.
PacBio data analysis and mutation detection
Detection of mutations in the PacBio data was performed using the ‘Minor and Compound Variants’ plug-in available in v2.0.1 of the PacBio SMRT Analysis Portal. Custom R scripts were used to study the mutational composition in patients carrying several mutations. This was done by looping through all circular consensus (CCS) reads and recording the mutational composition in each individual read. For a read to be present in the analysis of compound mutations, 20 bases in a window surrounding each mutation were required to match perfectly to the BCR-ABL1 p210 transcript reference sequence. In this way only reads with relatively high quality were used, thereby reducing the effects of sequencing errors.
BCR-ABL1 splice isoforms were identified from full-length CCS reads spanning the length of the entire transcript. For a splice isoform to be reported, we required at least two independent CCS reads to contain identical nucleotide sequences over the entire length of the transcript.
Discussion
Our results show that mutations can be detected at a level of ~1% in a background of wild type
BCR-ABL1 making it a useful tool for screening of both high and low level kinase domain mutations. Further, it provides information on the clonal distribution of mutations as well as
BCR-ABL1 isoforms in a single assay. This feature is of major clinical relevance as compound mutations show different resistance profiles compared to individual mutants [
5]. Standard Sanger sequencing methods routinely used in diagnostic laboratories are unable to distinguish between independent or compound mutations. Until now, this information has only been available through time consuming cloning experiments [
5], underlining the potential clinical utility of our assay.
Although recent reports showing MPS-approaches are emerging [
8,
9], its usefulness in establishing the clonality of mutations has recently been debated. In a recent report, Parker et al. [
13] showed that compound mutations detected by MPS technologies might actually be artifacts due to PCR-mediated recombination. However, our assay has a somewhat different setup compared to previous studies. Instead of performing a two round nested PCR, as is required for shorter read technologies, the fusion transcripts were amplified in one single round. This could potentially reduce the rate of PCR recombination. We were able to evaluate the degree of
in vitro artifacts in our experiments. For example, in the 49-month sample from patient 3, 91.2% of the reads contained only T315I and 4.2% of the reads contained only F359C (see Figure
4B). In the same sample 0.1% of reads show the presence of both T315I and F359C. These results suggest that the recombination rate in this particular case is very low, well below the frequency of the individual clones. However, since the rate of chimeric reads can be influenced by the experimental conditions and may vary between samples, a more thorough investigation would be required to validate the PCR recombination rate under different circumstances.
On the practical level, the PacBio assay allows for a simple, efficient and streamlined workflow conducive to clinical routine. Our laboratory workflow provides a quick turnaround time of approximately two days, encompassing all steps from RNA isolation to report generation. A simple library preparation procedure, rapid sequencing and straight forward bioinformatics analysis enable this efficient workflow. The library preparation is performed during one day and the sequencing run takes approximately 2–3 hours per sample. Under the current set up, PacBio sequencing is more expensive compared to the more traditional Sanger sequencing and RQ-PCR based assays. However, the cost for sequencing of small target regions such as the BCR-ABL1 transcript is comparable to that of other available MPS technologies. In the present study, we obtained 10,000X coverage of BCR-ABL1 for each of the samples. In light of this rather extensive coverage, it is likely that a similar sensitivity for mutation detection could to be obtained when utilizing a reduced coverage, thus opening up the possibility of barcoding of two or more samples on one SMRT cell. Further, due to the continuous improvements of the PacBio system in terms of quality, read length and throughput, the potential for multiplexing is likely to increase, thus leading to substantial reductions in experimental cost.
This study presents a proof of principle for detection of BCR-ABL1 mutations and our results are based on just a handful of patients, limiting the generality of our conclusions. Nevertheless the analysis of individual patient samples illustrates important aspects and strengths of our approach. One main advantage is the sensitivity of the assay, as illustrated in one of the patients (patient 1) where we could detect the T315I mutation four months earlier than detected by Sanger sequencing. These results indicate that an NGS-screen could be informative when performed at earlier time points, possibly in patients with no or limited responses to TKI therapy already at the three months control. Further studies are needed to specifically address this question.
The sensitivity of the method can also be instrumental in excluding other BCR-ABL1 mutations as responsible factors for the observed TKI resistance. For example, in patient 2, we could only detect the T315I and despite an initial molecular response to ponatinib, the patient remains with a minor molecular response. Thus in this case BCR-ABL1-independent factors might explain the failed therapy. This information is of particular importance when looking for alternative TKI-resistance pathways.
The ability to discern between independent and compound mutations is a major advantage of this assay. For example, patient 3 carried both the T315I and F359C mutations, but present in independent clones. A recent study has shown that the compound mutation F359C and T315I is associated with in vitro profiles implicating mutant pairing of these two positions in moderate and high-level resistance to ponatinib and rebastinib, respectively [
5], while the individual mutants are instead sensitive to these substances. Thus, we can speculate that the molecular response observed in this case upon ponatinib treatment, can be explained by the fact that these mutations are located in different molecules in this patient. In contrast, the T315I positive clone (84%) acquired independently two extra mutations (D276G and H396R) upon ponatinib treatment. These low frequency compound sub clones did not seem to impede the molecular response (MR5) attained after 3 years of treatment. The results are somehow conflicting with recent results showing that the compound mutant H396R/T315I has an intermediate to high resistance
in vitro profile with an IC
50 for ponatinib of 90.8 ± 24.7 nM compared to an IC
50 of 20.1 ± 3.5 (H396R) and 29.1 ± 7.8 nM (T315I) for the individual mutants [
5]. However,
in vitro sensitivity does not necessarily always correlate with the clinical response to treatment.
Similarly, patient 5, harboring the non-compound Y253H/E255V mutations, would clearly benefit today from sensitive examination of the clonal composition of these two mutations prior to therapy decisions because of the high ponatinib-resistance of the Y253H/E255V compound mutant [
5]. The fact that in this patient the mutations occur in different molecules would remain indiscernible by routine Sanger sequencing.
Another advantage of our approach is that as we amplify and sequence almost the entire
BCR-ABL1 p210 transcript we are able to identify transcript isoforms. We identified elevated levels of transcripts isoforms in two of the patients. Among the variants that we detected, 35INS is best studied in the literature. Some studies have described 35INS as a possible mechanism of imatinib resistance [
14,
15], while biochemical data from a separate study shows that it does not contribute to TKI resistance in vitro [
16]. Although there is at present no basis for taking 35INS into consideration for treatment decisions, these conflicting reports highlight the need for routine screening in CML patients in order to gain more knowledge. Our results suggest that the PacBio assay can be used for screening of 35INS as well as other splice isoforms down to a frequency of at least 5%. Furthermore, it enables simultaneous detection of multiple different alternative isoforms present in a single sample. These results corroborate previous findings that propose alternative splicing as a common mechanism among CML patients undergoing TKI treatment [
14,
17]. However, to clarify the role of transcript isoforms in drug resistance and response a larger number of samples should have to be analyzed and our approach might simplify this kind of analysis.
Although this study is focused on BCR-ABL1 the same method can be readily applied to the analysis of other cancer-associated fusion transcripts, providing not only information on the clonal distribution of mutations but also on isoform frequencies. Isoform analysis is to this date not performed routinely on CML patient samples and therefore the knowledge on their impact on disease progression and treatment efficiency is very limited.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
AA performed the sequence analysis and wrote the paper. LC and MH designed the study, performed the experiments, analyzed the data and wrote the paper. IH, SH performed the experiments. NC performed the experiments and wrote the paper. UOS provided the patient clinical data, analyzed the results and wrote the paper. All authors read and approved the final manuscript.