Targeted Sequencing for Clinical Investigation in Chronic Lymphocytic Leukemia

Introduction

Next generation sequencing (NGS) studies have documented the highly heterogeneous landscape of somatic mutations in CLL. Beside the mere genetic diversity of the disease, frequency and clinical impact of mutations vary between analyzed cohorts since samples may be collected at variable clinical stages. To address this point, we undertook an integrative systematic genomic study of patients included in a prospective clinical trial.

Material and methods

We performed conventional karyotype (K), fluorescence in situ hybridization (FISH) and investigated 31 genes and 2 non-coding regions, in pretreatment samples obtained from 146 patients enrolled on a phase 3 trial comparing fludarabine (F)-cyclophosphamide (C)-rituximab vs. F-C-alemtuzumab (Lepretre et al, Blood 2012). All patients were Binet stage B (n=115) or C (n=31) and devoid of 17p deletion. All gave their informed consent according to the declaration of Helsinki. We studied by targeted deep sequencing exonic regions of 31 genes including TP53, SF3B1, NOTCH1, ATM, BIRC3, FBXW7, NFKBIE, MED12, MYD88 and POT1. DNA was extracted from total blood mononuclear cells (fraction of tumor cells determined by flow cytometry >80%). Targeted regions were PCR-amplified using an AmpliSeq (Life technologies) approach and PCR products sequenced in a MiSeq (Illumina) instrument with a mean depth of 1000X. We used a variant allele frequency (VAF) minimal threshold of 5% to detect mutation. EGR2, PAX5 enhancer and NOTCH1 3'UTR regions were analyzed by Sanger sequencing. For each identified mutation, clinical and biological variables were compared between mutant and wild-type patients using either Wilcoxon or Fisher exact test. For each mutation present in a sufficient number of samples, prognostic impact on overall survival (OS) and progression free survival (PFS) was assessed using log-rank test. For important clinical and molecular prognostic variables, adjusted prognostic effect was assessed using a multivariate Cox model for OS and PFS with a stepwise-forward selection based on Akaike criteria.

Results

Chromosomal abnormalities were observed by K in 63% of 126 patients, 11% showing complex K and 23% carrying translocations. FISH revealed 78/144 (54%) 13q deletion, 25/144 (17%) trisomy 12 (tri12), 30/146 (21%) 11q deletion. Unmutated (UM) IGHV status was observed in 84/146 (58%) cases. At least one mutation was observed in 118/146 (81%) patients, with 1, 2, 3 and 4 or 5 mutations in, respectively, 34%, 23%, 17% and 7%. The most frequently mutated genes were SF3B1 (23%), NOTCH1 (16%), ATM (14%), NFKBIE (10%), XPO1 (10%), MED12 (7%), EGR2 (6%), TP53 (5%), BIRC3 (5%), FBXW7 (5%), PAX5 enhancer (5%), CHD2 (5%), PCLO (5%), MYD88 (4%), POT1 (4%), KMTD2 (4%) and NOTCH1 3'UTR (4%). MYD88 and SAMHD1 mutations were always observed as clonal (VAF>40%) whereas BIRC3 and BRAF mutationsalways observed as sub-clonal. NOTCH1 mutations were associated with high CD38 expression (P<.0001), UM IGHV (P<.0001) and tri12 (P=.04); FBXW7 with tri12 (P=.009); NOTCH1 3'UTR with high CD38 (P=.01) and UM IGHV (P=.04). UM IGHV were associated with XPO1 (P=.002) and MED12 mutations (P=.04); mutated IGHV with MYD88 (P=.005) and PAX5 enhancer mutations (P=.01). PAX5 enhancer and POT1 mutations were associated with normal K (P=.02 and .03, respectively). No individual mutant gene carried prognostic significance, but 3 or more mutations were associated with a shorter PFS (HR=2.4 [1.1-5.7]), as UM IGHV (HR=2.2 [1.2-3.9]) and Binet stage C (HR=3.7 [2.1-6.5]). Three or more mutations were associated with a shorter OS (HR=2.3 [1.1-4.7]), as Binet stage C (HR=4.4 [2.1-9.1]), whereas 13q deletion was associated with a longer OS (HR=0.5 [0.2-0.9]). By multivariate analyses, UM IGHV and Binet stage C retained prognostic significance for PFS, and 13q deletion, Binet stage C and 3 or more mutations for OS.

Conclusion

These results show that our approach is amendable to routine CLL analyses. They also confirm, in pretreatment CLL samples, the frequency of mutations in NFKBIE, MED12 and EGR2 genes and in PAX5 enhancer and NOTCH1 3'UTR non-coding regions. Some associations were confirmed (eg NOTCH1 and tri12, MYD88 and mutated IGHV). Eighty-one percent of the patients (patients with 17p deletion being excluded) exhibited at least 1 mutation. No individual mutation is associated with a poor prognosis, but the presence of 3 or more mutations is associated with shorter PFS and OS.