Circulating microRNAs in Cerebrospinal Fluid and Plasma: Sensitive Tool for Detection of Secondary CNS Involvement, Monitoring of Therapy and Prediction of CNS Relapse in Aggressive B-NHL Lymphomas

Abstract

Lymphoma with secondary central nervous system (CNS) involvement represents one of the most aggressive malignancies, with poor prognosis and high mortality. New diagnostic tools for its early detection, response evaluation, and CNS relapse prediction are needed. We analyzed circulating microRNAs in the cerebrospinal fluid (CSF) and plasma of 162 patients with aggressive B-cell non-Hodgkin's lymphomas (B-NHL) and compared their levels in CNS-involving lymphomas versus in systemic lymphomas, at diagnosis and during treatment and CNS relapse. We identified a set of five oncogenic microRNAs (miR-19a, miR-20a, miR-21, miR-92a, and miR-155) in CSF that detect, with high sensitivity, secondary CNS lymphoma involvement in aggressive B-NHL, including DLBCL, MCL, and Burkitt lymphoma. Their combination into an oncomiR index enables the separation of CNS lymphomas from systemic lymphomas or nonmalignant controls with high sensitivity and specificity, and high Receiver Operating Characteristics (DLBCL AUC = 0.96, MCL = 0.93, BL = 1.0). Longitudinal analysis showed that oncomiR levels reflect treatment efficacy and clinical outcomes, allowing their monitoring and prediction. In contrast to conventional methods, CSF oncomiRs enable detection of early and residual CNS involvement, as well as parenchymal involvement. These circulating oncomiRs increase 1-4 months before CNS relapse, allowing its early detection and improving the prediction of CNS relapse risk in DLBCL. Similar effects were detectable, to a lesser extent, in plasma.

Keywords: B-NHL; Burkitt; CNS; DLBCL; MCL; cerebrospinal fluid; lymphoma; microRNA; plasma; relapse.

Figure 2

Receiver Operating Characteristics (ROC) of oncomiR indices for discrimination of CNS-involving and systemic lymphomas. (A) CSF and (B) plasma of indicated CNS lymphoma subtypes. The X-axes indicate % of specificity and the Y-axes indicate 100—% of sensitivity. The OncomiR index (logistic regression model) combines the expression of individual oncomiRs into a single classifier, yielding higher specificity/sensitivity.

Figure 3

OncomiR indices are increased in plasma of B-NHL involving CNS. OncomiR indices (logistic regression model) combining expression of individual oncomiRs into one classifier in plasma (A) of lymphoma patients with indicated B-NHL diagnoses, with (SCNSL) and without secondary CNS involvement (systemic) and control patients (CTRL). (B) The SCNSL lymphomas are subdivided into lymphomas with secondary CNS involvement present at diagnosis (SCNSL dg) and newly detected CNS relapses (all BL-SCNSL are from dg). Systemic diagnoses (syst), controls (CTRL). The red line indicates the threshold for positive CNS lymphoma involvement. Log2 scale. Median ± interquartile range, * p < 0.05, ** p < 0.01, *** p < 0.001, Kruskal-Walliss. No star = non-significant.

Figure 1

OncomiR indices are increased in cerebrospinal fluid of CNS-involving lymphomas. OncomiR indices (logistic regression model) combining expression of oncogenic microRNAs (oncomiRs, miR-21/-19a/-20a/-92a/-155 as in Table 1) into single classifier in cerebrospinal fluid (CSF) of lymphoma patients with indicated B-NHL diagnoses. (A) Lymphoma patients with secondary CNS involvement (SCNSL), compared to systemic lymphoma patients and control patients (CTRL). (B) SCNSL lymphomas are subdivided into lymphoma with secondary CNS involvement presented at the time of diagnosis (SCNSL dg) and newly detected CNS relapses. (C) OncomiR indices in CSF of DLBCL stratified according to parenchymal, meningeal, and combined parenchymal and meningeal (P+mening) CNS involvement (all BL-SCNSL are from dg). qRT-PCR. The red line indicates the threshold for positive CNS lymphoma involvement. Log2 scale. Median ± interquartile range, * p < 0.05, ** p < 0.01, *** p < 0.001, Kruskal-Walliss. No star = non-significant.

Figure 4

Dynamics of oncomiRs during therapy of CNS lymphoma patients. Levels of indicated oncomiRs in cerebrospinal fluid (CSF, left panel) and plasma (right panel) of patients during treatment. (A) B-NHL-NOS patient with secondary CNS involvement responding to therapy; (B) BL patient with secondary CNS involvement resistant to therapy; (C) DLBCL systemic patient who relapsed/progressed to CNS. Below the x-axes the following are indicated: date of sampling and positivity/negativity of findings of flow cytometry/cytology of CSF, magnetic resonance imaging (MRI), and computed tomography (CT). (D) Median values of oncomiR levels of patients during the course of the disease of CNS-involving lymphomas, compared to systemic lymphomas and non-lymphoma controls. Abbreviations: CNS inv. = CNS involvement at the time of diagnosis; PR = partial remission; CR = complete remission.

Figure 5

Validation of predictive value of circulating oncomiRs in plasma for CNS relapse in DLBCL. Kaplan-Meier estimates of risk for CNS relapse stratified by: (A) oncomiR levels in plasma acquired at the time of diagnosis, (B) CNS-IPI, and (C) combined prediction model of CNS-IPI and oncomiR risk. Both risks = high microRNA + high CNS-IPI; One risk = either high microRNA or highCNS-IPI; No risk = neither microRNA nor CNS-IPI are high. For details on risk stratification, see the supplementary methods. Note: HR, 95% Cl and P in the tables below the charts were obtained from univariate models of indicated categories. Abbreviations: CNS = central nervous system; IPI = International Prognostic Index; Int. = intermediate; HR = hazard ratio; n (%) = number of patients.

Figure 6

Probability of overall survival in DLBCL according to oncomiR levels in CSF and plasma. Overall survival (OS) Kaplan-Meier estimates of DLBCL patients (with both systemic and concomitant systemic and CNS involvement), stratified by oncomiR index in CSF (left panel) or plasma (right panel). For details on OS stratification, see the supplementary methods.