Prognostic impact of somatic mutations on time to first treatment: Results of targeted next‐generation sequencing in 211 patients with early stage chronic lymphocytic leukemia

The clinical course of patients with CLL is highly heterogeneous reflecting an underlying biologic heterogeneity of the disease. In the last decade, whole exome/genome sequencing and targeted next-generation sequencing (NGS) techniques have revealed several genes recurrently mutated in CLL, involving a limited number of pathways, namely microenvironment-dependent signaling through NOTCH (NOTCH1, FBXW7), inflammatory receptors (MYD88), MAPK–extracellular signal-regulated kinase (BRAF, KRAS, NRAS, MAP2K1) and NF-κB pathways (BIRC3, TRAF3, NFKBIE), as well as intracellular programs such as DNA damage and cell cycle control (ATM, TP53, SAMHD1, POT1), chromatin modification (HIST1H1E, CHD2, ZMYM3), transcription (EGR2, IRF4, BCOR, MED12) and ribosomal processing (XPO1, SF3B1, RPS15).1, 2 In this study we analyzed by NGS the mutational status of 10 target genes in a homogeneous cohort of 211 patients with Binet stage A CLL diagnosed between 2002 and 2014, and correlated the mutational status with time-to-first-treatment (TTFT). Patients' characteristics are reported in Table S1. All patients were diagnosed according to the iWCLL 2008 criteria, retrospectively applied to patients diagnosed before 2008. The study was conducted according to the Declaration of Helsinki. Mononuclear cells (BMMNCs) were isolated by standard density gradient centrifugation (Lympholyte-H; CEDARLANE Laboratories Ltd) from peripheral blood (n = 182) or bone marrow (n = 29). Buccal cells as control tissue were available in 66 patients (31%). Genomic DNA and total RNA were extracted following standard protocols for human tissue. Targeted mutation analysis of 10 genes (ATM, BIRC3, FBXW7, KRAS, MYD88, NOTCH1, POT1, SF3B1, TP53, XPO1) was performed using a Truseq Custom Amplicon Sequencing Panel designed using Design Studio software (Illumina, San Diego, CA, USA). The oligo pool targets five full genes (coding exons and splice sites) and exonic hotspots of additional five genes (Table S2). Dual-barcoded TSCA libraries were created according to the manufacturer's protocol. Libraries were sequenced on a MiSeq system (Illumina). The resulting average depth of coverage for the 198 amplicons was 1287x (range: 503–3219x). The NGS data analysis relied on a robust workflow that implements GATK's pre-processing Best Practices. Functionally annotated variants were filtered based on the information retrieved from public databases (dbSNP, 1000Genome, dbNSFP, ESP6500) or in a set of in-house control tissue or healthy subiects. The remaining variants were finally tagged as oncogenic, based on the information retrieved from the literature, COSMIC and in silico prediction effect, as previously described.3 The variant analysis was limited to variants with allele frequency equal or greater than 1%. The variant allele frequency (VAF) in unselected mononuclear cells was adjusted for the percentage of the CLL cells quantified by immunophenotype. Mutations were considered subclonal when VAF was <12% and clonal when VAF was ≥12%. Methods used for the assessment of IGHV mutational status and detection of TP53 mutations with Sanger sequencing are described in the supplementary methods (Appendix S1). Qualitative variables were described as absolute and relative frequencies, while quantitative variables were summarized as median and interquartile range (IQR). Time-to-first-treatment, defined as the time between diagnosis and date of initiation of first treatment (event) or last follow-up (censored), was estimated by Kaplan–Meier product limit method. Log-rank test was used to compare outcome between two or more groups of patients. The effect of baseline characteristics of patients on TTFT was assessed by univariable Cox Proportional Hazard models. Variables with a p value lower than 0.2 at univariable analysis were included in multivariable model. Akaike's information criterion and Harrell's C index were used to compare multivariable models. Due to the high number of missing data, CLL-IPI and cytogenetics were not included in the multivariable model. The p values < 0.05 were considered statistically significant. All statistical analyses were conducted using the Stata 16 software (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC). Overall, 113 mutations were found in 74/211 patients (35%), co-occurrence of mutations in ≥ 2 genes was observed in 24/74 (32% of mutated cases). The median corrected VAF of all mutations was 0.18 (IQR: 0.05–0.34). The ATM mutations were observed in 20 patients (9%). Multiple mutations in the ATM gene were rather common with two ATM mutations in four patients, three ATM mutations in one patient and four ATM mutations in another one. Six of twenty patients with ATM mutations had 11q deletion by FISH. The SF3B1 mutations were detected in 14 patients (7%). Fourteen XPO1 mutations were observed in 13 patients (6%) and most of them (13/14) occurred at E571. All but one patient harboring the XPO1 mutation showed IGHV unmutated genes. The TP53 mutations were found in 11 patients (5%) and were located at the DNA-binding domain of the protein. Four of eleven patients with TP53 mutations by NGS also had a 17p deletion by FISH. All TP53 mutations found with VAF > 10% (10/11) were confirmed by Sanger sequencing. So, POT1 mutations were found in 12 patients (6%), who were IGHV unmutated in all cases but one. The NOTCH1 frame-shift mutations were found in 11 patients (5%), 10 of whom at P2514 in the PEST domain. Nine of 11 patients harboring NOTCH1 mutations at diagnosis showed unmutated IGHV. None of the patients with NOTCH1 mutation had a concomitant FBXW7 mutation. Ten FBXW7 mutations were detected in nine patients (4%), five of whom showed unmutated IGHV. Six of 11 patients with NOTCH1 mutation and five of 10 patients with FBXW7 mutations had trisomy 12 by FISH. A total of seven MYD88 missense mutations were found in seven patients (3%), all of whom were IGHV mutated. We found also two KRAS missense mutations and two BIRC3 mutations in two patients each (1%). The frequency, type and VAF of mutations are reported in Figure S1. With a median follow-up of 96 months (IQR: 65–138), 108 patients (51%) were treated per iwCLL criteria. The median TTFT was 6 years (95% CI: 5–12). In univariate analysis, the presence of one or more mutations by NGS was associated with shorter TTFT (p < 0.001) (Figure S2). The presence of mutations in the following genes were associated with shorter TTFT: ATM (p < 0.001), POT1 (p < 0.001), NOTCH1 (p < 0.001), XPO1 (p = 0.002), SF3B1 (p = 0.007), TP53 (p = 0.022), MYD88 (p = 0.041), FBXW7 (p = 0.045). Other variables associated with shorter TTFT were unmutated IGHV (p < 0.001), trisomy 12 (p = 0.007), 11q deletion (p = 0.004). Of note, TP53 mutations by NGS, 17p deletion and CK were not associated with significantly shorter TTFT in univariate analysis (Figure 1). A multivariable model identified unmutated IGHV (hazard ratio 2.2, 95% CI 1.4–3.6; p = 0.001), ATM mutation (hazard ratio 2.3, 95% CI 1.3–4.2; p = 0.005), FBXW7 mutation (hazard ratio, 3.1, 95% CI 1.3–7.3; p = 0.012), MYD88 mutation (hazard ratio 4.9, 95% CI 2.0–11.6; p < 0.001) and POT1 mutation (hazard ratio 3.3, 95% CI 1.6–6.6; p = 0.001) as being independently associated with shorter TTFT (Table S3). In this study we analyzed with NGS 10 genes, selected for their relevance in CLL, in a homogeneous cohort of 211 patients with Binet stage A disease and correlated the mutational status with TTFT. TTFT has recently emerged as an important end-point for untreated CLL patients, able to identify early drivers of disease which may be masked when investigating long-term and treatment-defined outcomes, such as progression-free survival and overall survival. We found that the presence of one or more mutations by NGS was predictive of significantly shorter TTFT, in line with a recent study reporting that both the presence and the total number of mutated genes (i.e., the tumoral mutational load) identify those individuals who are more likely to progress requiring therapy.4 As one could argue that the detrimental effect was mainly attributable to TP53 mutations, we evaluated the impact of having at least one mutation by NGS excluding TP53 mutations and confirmed the adverse prognostic impact on TTFT. In our study, mutations in POT1, ATM, FBXW7 and MYD88 genes were independently associated with shorter TTFT. While the prognostic impact of some of these mutations has been already reported in previous studies,5-7 to our knowledge this is the first study showing POT1 as an independent risk factor for progression in early stage CLL patients. The MYD88 mutations were detected exclusively in patients with mutated IGHV, as previously reported, and were associated with a significantly shorter TTFT in our series. The role of MYD88 mutations on clinical outcome of CLL patients remains controversial. Despite previous report of favorable outcome of MYD88 mutated CLL patients, subsequently observations showed that MYD88 mutations is associated with shorter TTFT in M-IGHV patients and may counteract the survival advantage of M-IGHV.7 Of note, neither TP53 mutations by NGS nor 17p deletions were associated with shorter TTFT. This is in keeping with prior studies where TP53 abnormalities did not predict a shorter TTFT, suggesting that these abnormalities, who are strong prognostic factors in advanced or relapsed CLL, have limited role in early untreated disease.6 The main strengths of this study are the homogeneity of the study population and the prolonged follow-up which seems adequate to assess long-term outcomes of patients with early stage CLL. On the other hand, the main limitations of this study include its single-center retrospective design, and the analysis of a limited number of genes with exclusion of non-coding regions. In addition, we did not analyze purified CLL cells, however we normalized the VAF by the actual CLL cell content. With these limitations, our study demonstrates that one third of patients with Binet stage A CLL harbor somatic mutations with prognostic relevance and confirms the value of NGS for the identification of predictors of time to first treatment in early stage CLL. The study was supported by Fondazione Cariplo & Regione Lombardia, Milan, Italy (Grant ID 42916996) to M.G.D.P. The authors declare no conflicts of interests. Ester Orlandi, Marzia Varettoni, Marianna Rossi designed the study, analyzed and interpreted data; Elena Flospergher, Sara Rattotti, Chiara Cavalloni, Fabio Bergamini, Caterina Cristinelli, Nicole Fabbri and participated to the collection and analysis of data; Silvia Zibellini, Anna Gallì and Ettore Rizzo analyzed sequencing data and performed bioinformatic analysis; Virginia Valeria Ferretti performed statistical analysis; Marzia Varettoni and Ester Orlandi wrote the manuscript; Massimo Gentile, Matteo Giovanni Della Porta and Luca Arcaini critically revised the manuscript. Appendix S1. Supporting Information. Figure S1. Frequency of patients with at least one somatic mutations in each gene (A), pattern of mutation of target genes (B), type of mutations (C) and variant allele frequency of mutations (D). Figure S2. Time to first treatment (TTFT) according to the number of mutations. Table S1. Baseline clinical and biologic characteristic of patients. Table S2. List of exons sequenced for each gene. Table S3. Multivariable Cox model for TTFT based on IPS-E and mutational status. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.