Clonal distribution of BCR-ABL1 mutations and splice isoforms by single-molecule long-read RNA sequencing

Abstract

Background: The evolution of mutations in the BCR-ABL1 fusion gene transcript renders CML patients resistant to tyrosine kinase inhibitor (TKI) based therapy. Thus screening for BCR-ABL1 mutations is recommended particularly in patients experiencing poor response to treatment. Herein we describe a novel approach for the detection and surveillance of BCR-ABL1 mutations in CML patients.

Methods: To detect mutations in the BCR-ABL1 transcript we developed an assay based on the Pacific Biosciences (PacBio) sequencing technology, which allows for single-molecule long-read sequencing of BCR-ABL1 fusion transcript molecules. Samples from six patients with poor response to therapy were analyzed both at diagnosis and follow-up. cDNA was generated from total RNA and a 1,6 kb fragment encompassing the BCR-ABL1 transcript was amplified using long range PCR. To estimate the sensitivity of the assay, a serial dilution experiment was performed.

Results: Over 10,000 full-length BCR-ABL1 sequences were obtained for all samples studied. Through the serial dilution analysis, mutations in CML patient samples could be detected down to a level of at least 1%. Notably, the assay was determined to be sufficiently sensitive even in patients harboring a low abundance of BCR-ABL1 levels. The PacBio sequencing successfully identified all mutations seen by standard methods. Importantly, we identified several mutations that escaped detection by the clinical routine analysis. Resistance mutations were found in all but one of the patients. Due to the long reads afforded by PacBio sequencing, compound mutations present in the same molecule were readily distinguished from independent alterations arising in different molecules. Moreover, several transcript isoforms of the BCR-ABL1 transcript were identified in two of the CML patients. Finally, our assay allowed for a quick turn around time allowing samples to be reported upon within 2 days.

Conclusions: In summary the PacBio sequencing assay can be applied to detect BCR-ABL1 resistance mutations in both diagnostic and follow-up CML patient samples using a simple protocol applicable to routine diagnosis. The method besides its sensitivity, gives a complete view of the clonal distribution of mutations, which is of importance when making therapy decisions.

Figure 1

Overview and evaluation ofBCR-ABL1mutation detection using the PacBio sequencing. A) Schematic overview of the workflow. Total RNA was used to generate a 1578 bp long BCR-ABL1 fusion transcript cDNA amplicon. PacBio adaptors were ligated to the amplicon and the resulting library was sequenced on a PacBio SMRT cell. The data analysis detected BCR-ABL1 mutations down to a frequency of at least 1%, as well as the different clones present in the sample. B) Alignment of reads to the BCR-ABL1 reference sequence. The grey area shows reads for a CML sample (patient P3,49 months) produced from one SMRT cell on the PacBio RSII instrument. The sequencing generates a uniform coverage of about 10,000X over the entire reference sequence. The red vertical line indicates the presence of a T315I mutation, present in 88.9% of the reads. The mutation F359C was also detected in this sample at a frequency of 4.2% and can be seen as a faint vertical line. C) Results of a dilution experiment of the CML sample in panel B) (P3, 49 m). The leftmost bars show mutation rates of T315I (red) and F359C (blue) for the undiluted sample. To the right are observed mutation frequencies for a dilution series where the expected T315I frequency reached 50%, 10%, 1% and 0.5%. The expected frequencies of T315I and F359C are shown in red and blue letters, respectively. Positions marked with ‘X’ indicate mutations not detected by the PacBio sequencing.

Figure 2

BCR-ABL1mutations and their composition in patient samples. Overview of BCR-ABL1 mutations detected in five CML patients (P1-P5) at the time of diagnosis and at subsequent follow-up examinations following TKI treatment. Samples taken at the time of diagnosis are labeled ‘Diag’. The follow-up samples are labeled with the number of months after diagnosis. The numbers next to the colored bars show the frequencies of all mutations observed by PacBio sequencing. Asterisks (*) indicate mutations that failed to be detected by Sanger sequencing.

Figure 3

Overview of treatments and PacBio results for patients with single mutations. A) Results for patient 1. The BCR-ABL1 IS% values measured by routine quantitative RT-PCR are shown in open circles. The sensitivity of this assay was measured for the BGUS reference gene and depicted by gray squares. The samples that were analyzed by PacBio sequencing are indicated by black arrows and their mutation composition showed in the circle plot diagrams above each time point. Vertical lines indicate the treatment periods. HU (Hydroxyurea). B) Results for patient 2. The T315I mutation was detected after nilotinib treatment, as indicated by the red cross. The mutation was detected at this time point using our allele specific quantitative PCR used in routine analysis.

Figure 4

Overview of treatments and PacBio results for patient 3. A) The BCR-ABL1 IS% values measured by routine quantitative RT-PCR are shown in open circles. The sensitivity of this assay was measured for the BGUS reference gene and depicted by gray squares. As indicated by the red cross, the T315I mutation was detected after eleven months of imatinib treatment. The mutation was detected at this time point using our allele specific quantitative PCR used in routine analysis. The samples that were analyzed by PacBio sequencing are indicated by black arrows and their mutation composition showed in the circle plot diagrams. Vertical lines indicate the treatment periods. HU (Hydroxyurea). B) This panel shows the mutational composition in the BCR-ABL1 transcript for the 49 m and 55 m follow-up samples. Horizontal lines gives a schematic representation of high-quality PacBio reads that were used for examining the mutational composition. At 49 m, 91.8% of the reads carried T315I mutation. 4.2% of the reads showed the presence of F359C and 3.9% of the reads contained none of the mutations. At 55 m, two new clones appeared, one containing D276G and T315I (2.0% of the reads) and one containing T315I and H396R (1.1% of the reads).

Figure 5

Overview of treatments and PacBio results for patients 4 and 5. A) Results for patient 4. The BCR-ABL1 IS% values measured by routine quantitative RT-PCR are shown in open circles. The sensitivity of this assay was measured for the BGUS reference gene and depicted by gray squares. As indicated by the red cross, the T315I mutation was detected after 58 months of dasatinib treatment. The mutation was detected at this time point using our allele specific quantitative PCR used in routine analysis. The samples that were analyzed by PacBio sequencing are indicated by black arrows and their mutation composition showed in the circle plot diagrams. Vertical lines indicate the treatment periods. HU (Hydroxyurea). B) Results for patient 5. Measurements shown were made as in A. AlloSCT: allogeneic stem cell transplantation. DLI: donor lymphocyte infusion.

Figure 6

Overview of treatments and PacBio results for patient 6. A)BCR-ABL1 IS% values measured by routine quantitative RT-PCR are shown in open circles. The sensitivity of this assay was measured for the BGUS reference gene and depicted by gray squares. The samples that were analyzed by PacBio sequencing are indicated by black arrows. The vertical line indicates the treatment period. B) This panel shows BCR-ABL1 isoforms in patient 6. At 7 months post diagnosis four different splice isoforms were identified. The most common isoform was the ‘wild type’ (WT) BCR-ABL1 transcript isoform, i.e. identical to the reference sequence used for mapping, present in 80% of the molecules. Two other isoforms contained insertions of entire exons, of lengths 35 bp and 154 bp, respectively and one contained a partial deletion of exon 7 of ABL1. At 13 months post diagnosis the WT isoform was present in 54% of the molecules whereas isoforms containing the 35 bp insertion between exon 8 and 9 in ABL1 was present in the other two isoforms.