ALBERT

Description: Abstract 3383 There are three currently identified secondary resistance mechanisms observed in chronic myeloid leukemia (CML) patients receiving tyrosine kinase inhibitors (TKIs). These include overexpression of drug-efflux proteins (ABCB1 and ABCG2), increased BCR-ABL expression, and mutations in the kinase domain (KD) of BCR-ABL. We investigated the interplay between these three modes of resistance in vitro, as well as looking for other mechanisms. Three CML blast crisis cell lines (K562, its ABCB1-overexpressing variant, K562 Dox, and KU812) were cultured in gradually increasing concentrations of imatinib (IM) to 2μ M, or dasatinib (DAS) to 200nM. Two IM-resistant K562 lines were established, both with increased IC50s for IM (from 13.7μ M in naïve cells, to ~50μ M), as well as increased IC50s for DAS and nilotinib (NIL). No cell-surface expression of ABCB1 or ABCG2 was observed, nor were KD mutations present. However, BCR-ABL expression was seen to steadily increase in both lines from 178% in naïve cells, to ~380% and 1200% in the resistant lines, suggesting this was the major mode of resistance. However, when a DAS-resistant K562 culture was generated the T315I mutation emerged. Studies of the intermediate stages of resistance revealed that BCR-ABL overexpression occurred in a step-wise fashion, peaking at 1915% in the 3.5nM intermediate, but then dropping significantly to ~1000% in the 5nM intermediate (P=0.0003). BCR-ABL expression then stabilised at this level, and the T315I mutation was subsequently detected. Thus, it appears that BCR-ABL overexpression was the first mechanism of resistance detectable, but was followed by the emergence of a KD mutation which had a clear selective advantage. This sequential selection was observed a further four times: in a DAS-resistant K562 Dox culture, and in three IM-resistant KU812 cultures. BCR-ABL expression in the DAS-resistant K562 Dox culture increased from 186% in naïve cells to 540% in the final culture. Studies of intermediate cultures revealed that BCR-ABL expression peaked at 850% in the 55nM intermediate, but then dropped significantly to ~500% in the 75nM intermediate (P=0.004). This drop in BCR-ABL expression coincided with the appearance of the V299L mutation. Interestingly, the K562 Dox DAS-resistant line also displayed resistance to NIL and IM, likely conferred by BCR-ABL overexpression as the 55nM intermediate (with the highest BCR-ABL expression levels) had the highest IC50s for NIL and IM, while the 75nM intermediate (with the V299L mutation) had increased IC50DAS but lower NIL and IM IC50s. Thus, BCR-ABL overexpression was the primary event, followed by the KD mutation. Likewise in three IM-resistant KU812 cultures, BCR-ABL expression levels rose from 443% in naïve cells, to peak levels of 6210%, 10,448% and 990% respectively, followed by drops in expression which coincided with the appearance of compound KD mutations, and the F359C mutation respectively (Table). In contrast, three IM-resistant K562 Dox cells were not found to have any KD mutations, nor BCR-ABL overexpression. Instead, the primary cause of resistance in these lines appears to be a further increase in ABCB1 expression. All three lines had increased IC50s for IM (from 12μ M in naïve cells, to ~27μ M), NIL (from 400nM to ~1000nM) and DAS (from 100nM to 〉625nM). The addition of PSC833 (an ABCB1 inhibitor) decreased the IM, NIL, and DAS IC50s for all three resistant lines to the level of the naïve control (~3μ M, ~250nM and ~10nM respectively), indicating that ABCB1 expression, facilitating active efflux of the drugs, is the primary mechanism of resistance in these lines. We have demonstrated that KD emergence is a stochastic event, as the same mutation did not always occur twice, however BCR-ABL and ABCB1 overexpression were more likely to arise recurrently in predisposed lines. Notably, different TKIs elicited different resistant mechanisms, but all were BCR-ABL dependent. Furthermore, all resistant lines showed cross-resistance to the three TKIs tested (IM, DAS and NIL), suggesting that currently available TKIs share the same susceptibilities to drug resistance. Table 1. Summary of resistance mechanisms detected in three cell lines exposed to IM or DAS. ✓ = yes; × = no. Culture condition K562 K562 Dox KU812 IM IM DAS IM IM IM DAS IM IM IM Resistance to 3 TKIs ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ KD Mutation x x T315I x x x V299L E450QE459KE470K E459KE462KE466E F359C Increased BCR-ABL ✓ ✓ ✓ x x x ✓ ✓ ✓ ✓ Increased ABCB1 x x x ✓ ✓ ✓ x x x x Disclosures: White: Novartis Pharmaceuticals: Honoraria, Research Funding; Bristol-Myers Squibb: Honoraria, Research Funding. Hughes:Novartis Pharmaceuticals: Honoraria, Research Funding, Speakers Bureau; Bristol-Myers Squibb: Honoraria, Research Funding, Speakers Bureau.

Description: Background: Majority of MDS cases appear to be sporadic in nature, but 10-15% have clear familial basis due to predisposing mutations in genes such as RUNX1, GATA2, CEBPA and DDX41. Contribution of germline variants in sporadic MDS is not studied. This study attempts to address the contribution of germline variants in MDS pathogenesis. Methods: We performed amplicon-based massively parallel sequencing (AmpliSeq custom panel adapted for Illumina HiSeq2500 sequencing) on all coding regions of 29 myeloid genes for 144 MDS samples. After identifying the variants in five genes (TET2, MET, GATA2, ASXL1, NOTCH1), we tested an additional 96 MDS samples including therapy-related myeloid neoplasm (T-MN) using a Sequenom assay. We also analyzed WES data for these variants in 178 AML samples and 758 normal controls and AmpliSeq data for ASXL1 and TET2 variants in 655 CML samples. Results: Collation of all coding variants in the 29 myeloid genes sequenced identified germline variants occurring in primary MDS at frequencies significantly higher than expected when compared to the normal population (ExAC and matched cohort were similar) (Table 1). These variants occurred in 5 genes (TET2, MET, GATA2, ASXL1 and NOTCH1) at increased frequencies of 1.5-16.6 fold. Numerous MDS samples had multiple variants (4 with 4 variants, 4 with 3 variants, 18 with 2 variants) while 70 had 1 variant. The 3 germline MET variants have been previously investigated in solid tumorigenesis and likely generate MET variant proteins that contribute to numerous cancer types including MDS. Interestingly, 7/17 (41%) MDS cases with germline MET variants also had other cancers including pancreatic, gastric and laryngeal cancers. Of the TET2 variants, Y867H and P1723S were concurrent in 5 MDS, 5 AML and 6 CML samples indicative of them being on the same allele (i.e. a haplotype). They were seen at higher than normal frequency in MDS and AML, but were not significantly enriched in CML. We are currently confirming their coexistence on the same allele and assaying for decreased TET2 activity to determine whether one or both variants contribute to the phenotype. Other variants identified in MDS include the rare GATA2 (P161A) variant which is present in 1% of the population and the nearby common GATA2 (A164T) allele (~20%). These were mutually exclusive in our cohort and were seen at 3.9 and 1.5-fold, respectively, above the expected population frequency. We generated the P161A variant using site-directed mutagenesis and assayed for GATA2 transactivation activity in HEK293 cells with a GATA2-responsive LYL1 promoter-Luciferase construct (Figure 1). We also included empty vector (EV), wildtype (WT) GATA2 and T354M which is the most common highly penetrant autosomal dominant mutation leading to familial MDS/AML. As expected, T354M displayed a marked decrease in transactivation ability when compared to WT. The P161A variant similarly displayed loss-of-function in this assay, but not to the same magnitude as T354M. This is consistent with the hypothesis that reduced GATA2 function predisposes to myeloid malignancy where decreasing GATA2 activity correlates with increasing risk of developing malignancy. In our study 10/36 (28%) cases harboring these variants were T-MN cases. Apart from MET (E168D) (11.4-fold), the 2 rare variants with highest frequency in MDS versus controls were ASXL1 (N986S) (16.6-fold) and NOTCH1 (R912W) (6.5-fold). ASXL1 is an epigenetic regulator often mutated in hematopoietic malignancy and aberrant NOTCH1 function has been associated with myeloid and lymphoid malignancies. Conclusions: We have identified common and rare germline variants in genes involved in myeloid malignancy that may contribute to MDS pathogenesis. It remains to be seen whether they contribute to initiation, maintenance and/or progression of MDS and other hematopoietic malignancies. This is the first study reporting higher frequency of germline variants in sporadic MDS cases. Table 1. Frequency of germline variants in MDS, AML and CML in comparison to ExAC. Table 1. Frequency of germline variants in MDS, AML and CML in comparison to ExAC. Disclosures Hiwase: Celgene Corporation: Research Funding.

Description: Abstract 1692 Introduction. BCR-ABL1 kinase domain mutations are the most common known cause of resistance to tyrosine kinase inhibitors (TKIs) in CML. Some imatinib resistant mutations also confer resistance to second generation TKIs nilotinib and/or dasatinib. Therefore, it is recommended that mutation analysis be performed before changing therapy. However, BCR-ABL1 mutant clones are often de-selected upon TKI cessation or change of therapy, and may become undetectable (Hanfstein et al, Haematologica 2011). It is not known whether treatment discontinuation or long term alternative TKI therapy leads to eradication of these mutant clones. If mutant clones persist at sub-clonal levels they have the potential to be re-selected and expand clonally given favorable conditions, such as change to a TKI for which they confer resistance. We examined longitudinal data of patients with imatinib resistant mutations that became undetectable by direct sequencing to determine whether these “long dormant” mutations could reappear, and the circumstances related to reappearance. Method. All chronic phase patients who had been monitored at our institution since starting imatinib, and had mutations detectable by sequencing during imatinib therapy were analyzed; 49 patients, median follow up since starting imatinib was 4.3 years (range 0.6–11.6 years). Sensitive mutation analysis using mass spectrometry (detection limit 0.2% mutant) was performed at selected times when the mutations became undetectable by direct sequencing (detection limit 10–20%). Results. Of the 49 patients with mutations detected by sequencing during imatinib therapy, mutations became undetectable by sequencing in 21 patients (29 mutations), at a median of 2 months after changing therapy (range 1–20 months). This was associated with increased imatinib dose (3 mutations), stopping imatinib (2), hematopoietic cell transplant (6), chemotherapy (1), switching to nilotinib (3), or switching to dasatinib (14). All mutations that became undetectable by sequencing when the patient switched to nilotinib or dasatinib were those known to be sensitive to the inhibitor received (e.g. F359V in a patient treated with dasatinib). In 16 of the 21 patients whose mutations became undetectable by direct sequencing, the mutations have not been detected again with 0.1 to 6.9 years of follow up since the mutations were last detected (median 1.1 years). Of these 16 patients, 15 maintained a stable complete cytogenetic response and 1 lost a major cytogenetic response. In the other 5 patients, the same mutations as those originally detected (identical nucleotide exchange) became detectable by sequencing between 1.7 and 5.4 years after last detection (median 4.4 years), Figure. The original mutations in 4 of these patients confer resistance to nilotinib as well as imatinib (Y253H and F359V), and their reappearance was associated with initiation of nilotinib therapy, Figure. Three of these 4 patients died of their disease, and 1 lost a major cytogenetic response. Sensitive mutation analysis could detect the mutation in 1 of these patients during the time of “dormancy” and before nilotinib therapy. The 5th patient received an autologous hematopoietic cell transplant upon detection of F359V, and the mutation became undetectable by sequencing. The patient subsequently received dasatinib for 3 years and the mutation remained undetectable. Dasatinib therapy was stopped due to intolerance and F359V rapidly reappeared while the patient was off TKI therapy, having been undetectable for 4.8 years. Using sensitive mutation analysis, F359V could be detected at low levels after the transplant, suggesting that the mutant clone had not been eradicated. Conclusion. The data suggest that some BCR-ABL1 mutations may persist at sub-clonal levels for many years after changing therapy. This could lead to clonal expansion under the selective pressure of a TKI for which the mutation confers insensitivity. Alternatively, the reappearance of the mutation could be a new occurrence of the same mutation. The study highlights the importance of knowing the mutation history of individual patients to enable informed therapy choices. Disclosures: Yeung: Novartis Pharmaceuticals: Honoraria, Research Funding; Bristol-Myers Squibb: Honoraria, Research Funding. Hughes:Ariad: Consultancy, Honoraria; Bristol-Myers Squibb: Consultancy, Honoraria, Research Funding; Novartis: Consultancy, Honoraria, Research Funding. Branford:Novartis: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding; Bristol-Myers Squibb: Honoraria, Research Funding; Ariad: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding; Cepheid: Consultancy.

Description: Background: Clinical scoring systems, such as Sokal risk, continue to have prognostic relevance for patients (pts) treated with tyrosine kinase inhibitors (TKI) and may have utility in combination with emerging biomarkers. The BCR-ABL value at 3 and 6 months (mo) of TKI are the strongest predictors of response. However, recent data demonstrate that the rate of BCR-ABL decline from the pre imatinib level adds significant predictive information (Hanfstein, Leukemia 2014; Branford, Blood 2014). Among poor risk pts with 〉10% at 3 mo in our cohort of first line imatinib, those with a slow rate of decline from their pre imatinib value, assessed by calculating the number of days over which BCR-ABL halved (halving time), predicted significantly poorer outcomes. Notwithstanding the importance of the 3 and 6 mo values, a prognostic biomarker obtained at an earlier timepoint may allow opportunities for therapy optimization. We therefore examined the prognostic significance of the rate of BCR-ABL decline at 1 mo in the context of other predictors of response. Aim: To determine whether baseline factors (age, gender, Sokal risk and imatinib starting dose: 400, 600 or 800 mg) and the BCR-ABL halving time at 1 mo of imatinib have predictive significance. Method: 528 first line imatinib treated pts were evaluated (median 45 mo of imatinib). Molecular assessment was performed pre imatinib and at 1 mo for 470/528. 453 of these 470 pts had a Sokal score available and were included in the analysis of outcome. Results: The median BCR-ABL halving time at 1 mo of imatinib was 17 days, quartiles 11, 29. An initial rapid BCR-ABL decline, indicated by halving times in the lowest quartile of ≤11 days (n=115), was associated with significantly superior rates of MMR by 12 mo, MR4.5 and failure-free survival (FFS) by 4 years compared with longer halving times of 〉11 days (n=338), Table. MMR by 12 mo was assessed since it represents an optimal response and is associated with subsequent deep molecular response. By univariate and multivariate regression analysis only the 1 mo halving time and Sokal risk significantly predicted MMR, MR4.5 and FFS. These factors were independent and there was no difference between the median 1 mo halving times among the Sokal risk groups, P = .36. The high Sokal risk pts had significantly poorer outcomes. To improve response prediction, these pts were divided into 2 groups according to their 1 mo halving time; ≤11 days (n = 28) and 〉11 days (n=90). A 1 mo halving time of ≤11 days was associated with significantly improved outcomes for these pts, Table and Figure. The responses equated to those of pts with low Sokal risk: high risk ≤11 days vs low risk: MMR by 12 mo 57% vs 59%, P = .95; MR4.5 by 4 years 36% vs 40%, P = .82; FFS by 4 years 79% vs 84%, P = .39. The high Sokal risk pts with the rapid initial BCR-ABL decline also had a lower probability of BCR-ABL 〉10% at 3 mo (early molecular response [EMR] failure), which is considered a warning or treatment failure; ≤11 days vs 〉11 days: 14% vs 33%, Table. Table 1 Outcome* by Sokal risk and BCR-ABL halving time at 1 mo of imatinib Factor No. of pts EMR % 3 mo MMR % 12 mo MR4.5 % 48 mo FFS % 48 mo Sokal Low 195 90 59 40 84 Intermediate 140 79 50 35 71 High 118 71 43 26 59 P value 29 days) was associated with a significantly lower cumulative incidence of MMR by 12 mo: low Sokal risk ≤29 days (n = 151) vs 〉29 days (n = 44) 65% vs 39%, P = .002; intermediate Sokal risk ≤29 days (n = 103) vs 〉29 days (n=37) 57% vs 31%, P = .004, Figure. Conclusion: Imatinib treated high Sokal risk pts have a higher rate of treatment failure and poorer molecular response. However, our data suggest their prognosis can be refined by taking into account the kinetics of BCR-ABL decline after only 1 mo of treatment. A rapid initial decline defined a subgroup of high Sokal risk pts with outcomes equivalent to those of low Sokal risk pts. Frequent molecular monitoring in the critical first months of treatment could enhance outcome prediction and limit the indication for a change of treatment. Figure 1 Figure 1. Disclosures Branford: Novartis: Consultancy, Honoraria, Research Funding; Bristol-Myers Squibb: Honoraria, Research Funding; Ariad: Honoraria, Research Funding; Otsuka: Honoraria, Research Funding. Yeung:Novartis: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding; BMS: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding. Ross:Novartis: Honoraria, Research Funding; BMS: Honoraria. Seymour:Novartis: Honoraria; Bristol-Myers Squibb: Honoraria. Hughes:Novartis: Honoraria, Research Funding; Bristol-Myers Squibb: Honoraria, Research Funding; Ariad: Honoraria, Research Funding.

Description: Epidermal growth factor receptor (EGFR) expression is elevated in peripheral blood (PB) cells of polycythemia vera (PV) patients (Skov et al, Eur J Haematol 2011;87:54-60) and EGFR inhibitors (AEE788, erlotinib) inhibit erythroid burst-forming units (BFUE) from PV patients but not normal donors. The mechanism underlying the effect of EGFR inhibitors on MPN progenitor growth has not been established but could be due to an off-target effect on JAK2 activity (Li et al, J Biol Chem 2007;282:3428-32; Gaikwad et al, Exp Hematol 2007;35:1647-56). Therefore, we investigated the growth of BFUE in the presence and absence of the EGFR inhibitor gefitinib (10µM), which does not inhibit JAK2, and observed inhibition of growth of both erythropoietin (Epo)-dependent and -independent colonies from PB mononuclear cells (PBMNC) from PV patients but not from normal individuals. These results suggest a potential role for EGFR signalling in supporting growth and/or survival of PV progenitors. Therefore, to evaluate the possibility of somatic genetic abnormalities leading to EGFR hypersensitivity in MPN, we performed targeted exon capture and massively parallel sequencing, and sensitive mass array screening of 155 MPN patient samples. We identified a low-frequency, recurrent somatic variant of EGFR (p. C329R) in 3/155 MPN patients. The human EGFR C329 residue is homologous to the residue C359 of the C. elegans gene let-23, target of a known gain-of-function mutation (Katz et al, Mol Cell Biol 1996;16(2):529-37); it also aligns with the cysteine residue affected in the highly-transforming mutant ErbB2 C334S found in lung cancer (Greulich et al, PNAS 2012; 109:14476–14481); and it lies within the extracellular cysteine-rich region of EGFR that is the target of frequent somatic mutations in glioma. To confirm the hypothesis that the EGFR C329R mutant leads to altered cytokine response, we transduced Ba/F3 cells with empty vector, EGFR wild type (WT) or mutant constructs (BaF3/MIG, BaF3/EGFR and BaF3/EGFRC329R, respectively). Both WT and mutant receptors showed constitutive activation and transforming ability when expressed at high levels. However, BaF3/EGFRC329R cells display increased levels of STAT activation associated with a slight proliferation advantage when compared to BaF3/EGFR. Given that gefitinib inhibited the growth of both BaF3/EGFR and BaF3/EGFRC329R but did not affect BaF3/MIG cells grown in IL-3, we next compared the effect of gefitinib (10µM) on the growth of BFUE from PV patient samples with and without the EGFR C329R mutation. We observed significant inhibition of Epo-independent BFUE from all PV samples but not of Epo-dependent BFUE from normal controls (Figure A). Furthermore, genotyping of JAK2 and EGFR from the individual colonies obtained in BFUE assay (treated or not with gefitinib, 10µM) for a PV patient that is positive for EGFR C329R showed that drug treatment significantly reduced the proportion of JAK2 V617F heterozygous BFUE compared to the vehicle-treated control (chi-squared test = 0.0002, Figure B). This suggests that signalling from EGFR contributes to proliferation and/or survival in JAK2 V617F heterozygous BFUE from this patient. The results presented here are consistent with an EGFR signalling role in supporting growth of PV progenitors, particularly in the context of a heterozygous JAK2V617F mutation. STAT5 signalling is essential for PV (Walz et al, Blood 2012;119:3550-3560; Yan et al, Blood 2012;199:3539-3549) and JAK2-independent activation of STAT5 through EGFR (Quesnelle et al, J. Cell. Biochem. 2007; 102:311–319) via various mechanisms may contribute to the level of STAT5 activation required for the PV phenotype. A recent study demonstrating a role for EGFR in hematopoietic stem cells (Doan et al, Nat Med 2013;19:295-304) also highlights the potential of aberrant EGFR signalling to contribute to altered properties of MPN stem cells. Finally, given that gefitinib is currently in clinical use for treatment of solid tumors, these findings raise the possibility that gefitinib may have clinical utility in the context of MPN. Figure 1 Figure 1. Disclosures Branford: Novartis: Consultancy, Honoraria, Research Funding; Bristol-Myers Squibb: Honoraria, Research Funding; Ariad: Honoraria, Research Funding; Otsuka: Honoraria, Research Funding.

Description: Abstract 917 Background: Secondary kinase domain (KD) mutations represent the most well-documented mechanism of resistance to tyrosine kinase inhibitors (TKIs) in chronic myeloid leukemia (CML). In CML, multiple TKIs with different mutation profiles are approved and the ability to detect KD mutations at the time of disease progression can impact therapy choice. To optimize clinical impact, second generation TKI selection must consider the majority TKI-resistant mutant population as well as smaller mutant sub-populations that may be selected with subsequent treatment. Sequential TKI therapy is associated with additional complexity: multiple mutations can coexist separately in an individual patient (“polyclonality”) or can occur in tandem on a single allele (“compound mutations”). Multiple mutations are associated with poor clinical outcome (Parker et al., Blood 2012). Compound mutations can cause in vitro resistance to ponatinib, the only TKI clinically active against the highly resistant T315I mutation (Eide, et. al, ASH 2012 abstract #1416). Currently, no clinically adaptable technology can distinguish polyclonal from compound mutations. Due to the size of the BCR-ABL KD, most next-generation sequencing platforms cannot generate reads of sufficient length to determine if mutations separated by ≥500 nt reside on the same allele. Pacific Biosciences RS Single Molecule Real Time (SMRT) sequencing technology is a third generation deep sequencing technology capable of achieving average read lengths of ∼1000bp and frequently 〉3000bp, enabling sensitive and accurate sequencing of the entire ABL KD on a single strand of DNA. Though allele-specific detection methods such as MassARRAY offer sensitivity as low as ∼0.5%, these assays are designed to detect a limited number (∼31) of mutations whereas SMRT sequencing offers an unbiased approach capable of detecting novel variants. We sought to (1) develop a potential clinically-applicable SMRT sequencing assay for the detection of BCR-ABLKD mutations capable of distinguishing polyclonal and compound mutations, and (2) compare the accuracy and sensitivity of this method to standard sequencing and MassARRAY. Results: We assessed 54 samples from 36 CML patients who had clinically relapsed on ABL kinase inhibitor therapy and were previously analyzed by standard sequencing, and in a subset, by MassARRAY. We amplified an 863bp area of the BCR-ABLKD from patient-derived cDNA with primers containing 5' barcodes, enabling sequencing of 6 to 8 patient samples on a single SMRT cell on a single run. On average, 2519 reads were obtained for each sample per run (range 330 to 10,240). All of 131 known mutations detected by MassARRAY were identified by SMRT sequencing using a p-value threshold of 1.03e–03. SMRT sequencing also identified all 107 known mutations detected by direct sequencing with a p-value threshold of 6.0e–08. In addition to these known mutations, SMRT sequencing detected an additional 1320 non-silent mutations across all patient samples using a strict p-value threshold cut-off of 6e–08, ranging in abundance from 0.2% to 17% (median 0.75%). Among 47 samples where 〉1 mutation was detectable by direct sequencing or MassARRAY, SMRT sequencing revealed that 40 (85%) had compound mutations detectable at a frequency of ≥1. In total, we detected 73 different compound mutations at a frequency of ≥1%. In all cases where compound mutations were detected and more than one treatment timepoint was available, at least one compound mutation clearly evolved from a mutation detectable at a prior timepoint. In the most complex case, 4 separate mutations yielded 8 different mutant alleles. Conclusions: Pacific Biosciences RS SMRT sequencing detects KD mutations in patient samples with sensitivity comparable to or better than MassARRAY and can distinguish compound from polyclonal mutant clones. Among patient samples with multiple mutations, compound mutations were detectable in the vast majority of samples by SMRT sequencing, revealing a complex mutational landscape not demonstrable by other clinically viable sequencing methods and previously unappreciated. Given the growing numbers of patients exposed to multiple TKIs in a sequential manner, the ability to accurately and sensitively characterize drug-resistant alleles by SMRT sequencing promises to further facilitate a personalized approach to patient management and inform models of disease evolution. Disclosures: Brown: Pacific Biosciences: Employment. Travers:Pacific Biosciences: Employment. Wang:Pacific Biosciences: Employment. Branford:Novartis : Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding; Bristol-Myers Squibb: Honoraria, Research Funding; Ariad : Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding; Cepheid : Consultancy. Shah:ARIAD, Bristol Myers-Squibb: Consultancy, Research Funding; Novartis: Consultancy.

Description: Background Mutation of genes linked to hematologic cancer have recently been reported in CML and are associated with early progression and resistance (Reviewed in Branford, Kim Leuk 2019). The mutations comprise single nucleotide variants (SNVs) and small insertions/deletions (indels), plus gene fusions and large focal gene deletions. In 39 patients (pts) in blast crisis (BC), all had at least 1 cancer gene mutation, including fusions in 33%: partner genes MLL, RUNX1, IKZF1, MECOM and CBFB. 50% of the fusions were novel and some were present at chronic phase diagnosis. BCR-ABL1 mutations rarely occurred as the sole mutant. NGS offers critical information for resistance assessment. For many clinical purposes, targeted DNA sequencing (seq) using panels of specific disease related genes is the most cost effective screening choice. However, this strategy could miss relevant fusions and deletions. Aim To determine whether an RNA based approach is more informative than DNA for detecting a broad range of mutations. Method A hybridization capture NGS gene panel was developed to target 126 genes relevant for myeloid/lymphoid leukemia. In a pilot study, DNA and RNA derived from 5 leukemia cell lines with well characterized mutations, including fusions and deletions, were panel sequenced. An additional 6 cell lines were sequenced using RNA, plus 49 pt samples with RNA stored for up to 14.6 years: 45 at diagnosis and 4 at BC/resistance. Six of these pt samples had prior whole exome and/or whole transcriptome seq. We used total RNA that detected intronic splice region variants from pre-spliced RNA. SNVs/indels were called from DNA/RNA with FreeBayes. Manta called focal deletions from DNA. Known and novel RNA fusions and novel splice junctions were detected using the STAR aligner. Gene expression used edgeR. Results For the 5 cell lines with DNA versus (v) RNA seq, SNVs/indels were reliably called in RNA, with a strong positive correlation of mutant allele frequency: DNA v RNA, r = 0.93. Two TP53 small deletions of 26 and 46 bp were not called in RNA, but were instead detected as novel RNA splice junctions. Read counts were 5.2 fold higher for RNA than DNA at sites of clinically relevant mutants, consistent with enrichment of seq read depth proportional to expression. Overall, RNA revealed a higher number of relevant mutants than DNA: RNA = 49 v DNA = 37, Fig A-C. Notably, the functional effect of splice region disrupting mutants and large focal deletions were evident by novel RNA splicing, Fig D-F. In the total 11 cell lines tested with RNA, all 13 reported fusions were called, including BCR-ABL1 and RUNX1, MLL, ETV6 and CBFB fusions. For 7 cell lines with variants described in the COSMIC Cell Lines Project, 23/23 cancer gene SNVs/indels were called, plus 7 cancer gene SNVs/indels not reported. These were verified by DNA seq. 15 gene deletions were evident by atypical RNA splicing and verified by DNA seq: IKZF1, CDKN2A/B, PAX5, BTG1, RB1 and NCOR1. Five other cell lines had verified CDKN2A deletions that were evident by loss of gene expression, Fig G. Two BTG1 deletions were not detected. For the 6 pt samples re-sequenced by the RNA panel, 8/8 verified fusion transcripts were detected with a 31 fold enrichment of read counts. 11/11 cancer gene SNVs/indels were called and 3/4 gene deletions. The exception was a CDKN2A deletion not detected by novel splicing but evident as loss of expression, Fig G. Seven other cancer gene SNVs were found at low allele frequency, including a resistant BCR-ABL1 mutation at 1.7% in the oldest sample. Of the 43 diagnosis samples without prior NGS, BCR-ABL1 transcripts were detected in all. BCR-ABL1 genomic breakpoints were called at base pair resolution in 39, 91%. Two pts had mutated ASXL1 at diagnosis and both failed imatinib by 9 months with mutant BCR-ABL1. By gene expression analysis, all but 1 of the total 45 diagnosis samples clustered together. The exception was a pt who transformed to lymphoid BC at 6 months that clustered with the lymphoid cell lines and lymphoid BC pts, Fig H. Conclusion RNA gene panel seq demonstrated enhanced sensitivity and an increased yield of clinically relevant mutations compared with DNA panel seq. A single RNA assay has the capacity to detect SNV/indels, known and novel gene fusions, focal deletions and the likely functional effect of splice disrupting mutations. RNA panel seq is a valuable tool for the comprehensive assessment of mutations that drive CML treatment failure and drug resistance. Disclosures Branford: Novartis: Consultancy, Honoraria, Research Funding, Speakers Bureau; Bristol-Myers Squibb: Honoraria, Speakers Bureau; Qiagen: Consultancy, Honoraria; Cepheid: Consultancy, Honoraria. Shanmuganathan:Gilead: Other: Travel Support; Janssen: Other: Travel Support; Amgen: Other: Travel Support; Bristol-Myers Squibb: Honoraria, Other: Travel Support; Novartis: Honoraria, Other: Travel Support. Scott:Celgene: Honoraria. Hughes:Novartis, Bristol-Myers Squibb: Consultancy, Other: Travel; Novartis, Bristol-Myers Squibb, Celgene: Research Funding.

Description: Key Points The association between multiple BCR-ABL1 mutations and inferior response to nilotinib/dasatinib was not seen with ponatinib therapy. However, chronic phase patients with T315I plus additional mutation(s) did have poorer responses to ponatinib than those with T315I only.

Description: Specific imatinib-resistant BCR-ABL1 mutations (Y253H, E255K/V, T315I, F317L, and F359V/C) predict failure of second-line nilotinib or dasatinib therapy in patients with chronic myeloid leukemia; however, such therapy also fails in approximately 40% of patients in the chronic phase of this disease who do not have these resistant mutations. We investigated whether sensitive mutation analysis could identify other poor-risk subgroups. Analysis was performed by direct sequencing and sensitive mass spectrometry on 220 imatinib-resistant patients before they began nilotinib or dasatinib therapy. Patients with resistant mutations by either method (n = 45) were excluded because inferior response was known. Of the remaining 175 patients, 19% had multiple mutations by mass spectrometry versus 9% by sequencing. Compared with 0 or 1 mutation, the presence of multiple mutations was associated with lower rates of complete cytogenetic response (50% vs 21%, P = .003) and major molecular response (31% vs 6%, P = .005) and a higher rate of new resistant mutations (25% vs 56%, P = .0009). Sensitive mutation analysis identified a poor-risk subgroup (15.5% of all patients) with multiple mutations not identified by standard screening.

Description: Introduction: Therapy-related myeloid neoplasms (T-MN; inclusive of T-MDS and T-AML) are aggressive neoplasms occurring after exposure to chemo (CT) and/or radiotherapy (RT) and are characterized by poor prognosis. Unlike primaryMDS and AML, the mutational architecture of T-MN has not been well elucidated. Most published studies in T-MN are small in size with a restricted panel of genes tested. Here we compare the cytogenetic and mutation profile of T-MN and primary MDS patients from the South Australian Myelodysplastic Syndrome (SA-MDS) registry. Methods: Demographic, clinical and laboratory data including cytogenetic profile of 147 T-MN and 744 primary MDS patients were analysed. Targeted Massively Parallel Sequencing of a custom panel of 27 myeloid genes (all coding regions) was performed on bone marrow samples from 60 T-MN and 142primary MDS patient samples. Results: Median age of T-MN and primary MDS was 71 (20-91) and 73 years (19-98), respectively. In T-MN patients the most frequent primary neoplasms were lymphoproliferative neoplasms (n=56, 38%) and prostate cancer (n=22, 15%). Sixty-three (43%) patients had received CT only, 34 (23%) patients RT only, and 48 (33%) patients had received both CT and RT. The T-MN group consisted of 104 T-MDS (71%) and 43 T-AML (29%). Poor risk cytogenetics were more frequent in T-MN cases following CT than after RT alone, and more frequent than in primary MDS cases (52% vs 25% vs 13%; P