Low frequency mutations independently predict poor treatment-free survival in early stage chronic lymphocytic leukemia and monoclonal B-cell lymphocytosis

Recent studies employing next generation sequencing (NGS) technologies have identified novel recurring mutations in monoclonal B-cell lymphocytosis (cMBL) and chronic lymphocytic leukemia (CLL).1 NOTCH1 and SF3B1 mutations are the most prevalent and are associated with reduced survival independent of established clinical and biological variables. However, we have limited understanding of the impact on clinical outcome of less prevalent mutations, especially when identified at diagnosis in patients with cMBL or Binet stage A CLL. To address this question, we have investigated the clinical significance of SF3B1, NOTCH1 and four ‘low frequency’ mutations: POT1, XPO1, MYD88, BIRC3 in a single center cohort of well-characterized patients. A fifth gene (FBXW7) was screened only in cases with trisomy 12, in view of the known strong association between the two abnormalities.2

This study included 206 previously untreated patients with a diagnosis of either Binet stage A CLL [n=116] or cMBL [n=−90] (Online Supplementary Table S1) according to the National Cancer Institute-Working Group (NCI-WG) criteria up-dated in 2008.3 Follow up ranged from one to 35 years with a median of ten years. One hundred and ten patients (53%) remained in stable Binet Stage A or cMBL while 29 of 90 (32%) of patients with cMBL evolved to Binet stage A during the observational period. Sixty-seven (33%) patients progressed to Binet stage B or C disease and required treatment. Of these, 21 (23%) progressed from cMBL and 46 (40%) from Binet stage A CLL. The indica tions for treatment and the treatment regimens were based on current guidelines at the time of treatment, with most patients receiving an alkylating agent or purine analog. Biomarker studies and mutational analysis (Online Supplementary Appendix) were performed on samples stored at or within six months of diagnosis. Mutational analysis was performed at a second time point in 84 patients with a median of 72 months (range 24–145 months) between initial and subsequent testing (Online Supplementary Table S2). This study was implemented in accordance with the Declaration of Helsinki and has been ethically approved by the Regional Ethics Committee (REC).

At diagnosis, mutations were detected in SF3B1 (16 of 199, 8%), NOTCH1 (11 of 203, 5%), POT1 (8 of 198, 4%), XPO1 (2 of 172, 1%) MYD88 (3 of 198, 1.5 %), BIRC3 (1 of 197, 0.5%) and FBXW7 (2 of 31 trisomy 12 cases, 6.5%). The majority of mutations have been previously observed in CLL, are annotated in COSMIC and predicted to have deleterious functional consequences based on PolyPhen and SIFT scores (Online Supplementary Table S3). These figures are broadly comparable to those in a recent large study of 1160 previously untreated patients of all stages of whom 82% were screened at diagnosis: NOTCH1, SF3B1, XPO1, FBWX7 and MYD88 mutations were found 12.3%, 9.0%, 3.4%, 2.5% and 1.5% of cases, respectively.2 The clinical and biological features of patients with NOTCH1, SF3B1 and low frequency mutations are shown in Figure 1A and Online Supplementary Table S4. The expected associations were apparent, such as both NOTCH1 and SF3B1 mutations being significantly associated with IGHV unmutated genes, high CD38 and ZAP70 expression.4 Three of the 8 POT1 mutated cases had mutated IGHV genes in contrast to previous data,5 in which POT1 mutations have occurred exclusively in cases with unmutated IGHV genes. In view of the reported increased incidence of telomere-containing chromosome fusions in patients with POT1 mutations, we examined the incidence of chromosomal complexity and instability in our 8 sequential cases with POT1 mutations. Based on cytogenetic and FISH data at diagnosis and later time points, we detected a del(13q14) at the second time point only in 2 cases, but found no evidence of genomic complexity (Online Supplementary Table S5). Larger studies in patients pre- and post therapy, using more sensitive methods to detect complexity and instability, will be required to determine the biological significance of POT1 mutations. The low incidence of the other mutations precluded any meaningful analysis of their associations with other biomarkers. We found no difference in the frequency of genomic mutations or any other biomarker, apart from a border-line higher incidence of CD38 expression in cMBL, between cases presenting with cMBL or stage A CLL. This is consistent with other recent data confirming the biological similarity between cMBL and Rai O CLL, including a similar incidence of NOTCH1 and SF3B1 mutations,6 and provides justification for analyzing the outcome of the combined cMBL and stage A cohorts.

Figure 1.The associations between gene mutations, established biomarkers and time to first treatment in our series of cMBL and stage A CLL patients. (A) Shows the mutual relationship between gene mutations and other genetic lesions and biomarkers in CLL, sorted by IGHV mutational status. Rows correspond to specific lesions/bio-markers, and columns represent individual patients (only patients with mutations in the genes tested are shown). Boxes colored black and gray show the presence and absence of a lesion/biomarkers. A white box denotes that no data are available. (B) TFS for patients with ‘low frequency’ mutations compared to wild-type controls. The P value is derived from Kaplan-Meier analysis with a log rank test and median survival times with 95% confidence intervals.

We then assessed the prognostic significance of mutations in our cohort. In view of the low frequency of all mutations other than NOTCH1 and SF3B1, and the known favorable outcome of cases with an MYD88 mutation,7 we hypothesized that the collective analysis of the other low frequency mutations might provide insight into their biological importance and clinical utility in early stage disease. Accordingly, we grouped mutations of POT1, XPO1, BIRC3 and FBXW7 together and compared the outcome of cases with any of these ‘low frequency’ mutations with those that were wild type for NOTCH1, SF3B1 and the four ‘low frequency’ genes. In support of this hypothesis, we identified an enrichment of NOTCH1 and SF3B1 mutations in Stage A CLL and cMBL patients who ultimately developed progressive disease, and also showed that the collective presence of a low frequency mutation was significantly associated with subsequent disease progression [Odds Ratio (OR) 17.4, P=0.002] and need for treatment (OR: 18.0 and P=0.002) (Online Supplementary Table S4).

Univariate analysis confirmed recent data8 showing that the presence of NOTCH1 [(median 58 vs. 105 months; Hazards Ratio (HR) 3.9, P<0.001)], and SF3B1 mutations (median 44 vs. 103 months; HR 2.8, P=0.001) were significantly associated with reduced treatment-free survival (TFS) but not overall survival (OS) (Table 1). However, the presence of a low frequency mutation was also associated with reduced TFS (median 42 vs. 113 months; HR 8.6, P<0.001) (Figure 1B).

Table 1.Univariate Cox proportional hazard analysis of treatment-free and overall survival.

We then estimated the impact of these ‘low frequency’ mutations on TFS and OS after controlling for confounding variables in multivariate Cox proportional hazard analysis depicted in Table 2. Along with the ‘low frequency’ mutations, other variables included in the analysis were SF3B1 and NOTCH1 status, trisomy 12, del(11q) and the presence or absence of mutated IGHV genes. Loss of chromosome 17p was omitted due to the very low frequency of del(17p) cases, all of which exhibited mutated IGHV genes.9 As expected, unmutated IGHV genes remained the strongest predictor of poor treatment free (HR: 5.7, 95%CI: 2.7–11.9, P<0.001) and overall survival (HR: 3.3, 95%CI: 1.9–5.7, P<0.001). Loss of chromosome 11q was confirmed as an adverse prognostic factor in treatment-free survival (HR: 3.7, 95%CI: 1.5–8.6, P=0.003) and SF3B1 mutations showed border-line significance as an independent predictor of reduced overall survival (HR: 1.3, 95%CI: 1.02–5.3, P=0.045). In addition, our ‘low frequency mutations’ variable retained significance for reduced treatment-free survival (HR: 3.7, 95%CI: 1.3–10.5, P=0.016). While we recognize that the apparent poorer outcome of cases with low frequency mutations might reflect the presence of other undetected mutations or genomic instability rather than the mutations we detected, none had del(17p) or cytogenetic evidence of genomic complexity and only one patient showed a del(11q).

Table 2.Multivariate Cox proportional hazard analysis of treatment-free and overall survival.

Finally, 84 patients were screened sequentially for mutations compared to those screened only at diagnosis in order to document clonal evolution in CLL. The sequential and single time point cases are shown in Online Supplementary Table S2 and no significant differences were observed between the two groups, apart from a border-line higher frequency of trisomy 12 in the cases tested only at diagnosis. Among the sequential cases, 47 patients (56%) remained stable during observation time, but 37 (44%) progressed and 36 patients received treatment between diagnosis and second testing. We detected mutations of SF3B1 in 4 cases and XPO1 in one case, not found on screening at diagnosis. Of these, 3 presented with cMBL, and 2 with Stage A CLL and 3 out of 5 had progressive disease. Additional characteristics of these patients are shown in Online Supplementary Table S6. Although the number of patients screened was small, these results were consistent with those of a recent large multinational study in which the incidence of SF3B1, but not NOTCH1, mutations rose with increasing time from diagnosis to the date of sampling.8

In summary, our study suggests that screening for these low frequency mutations may have utility in the clinical management of early stage CLL and cMBL, and future larger studies should evaluate the incidence and clinical significance of low frequency potential driver mutations in early disease to assess their relevance in new molecular prognostication systems.