A complementary role of multiparameter flow cytometry and high-throughput sequencing for minimal residual disease detection in chronic lymphocytic leukemia: an European Research Initiative on CLL study

Abstract

In chronic lymphocytic leukemia (CLL) the level of minimal residual disease (MRD) after therapy is an independent predictor of outcome. Given the increasing number of new agents being explored for CLL therapy, using MRD as a surrogate could greatly reduce the time necessary to assess their efficacy. In this European Research Initiative on CLL (ERIC) project we have identified and validated a flow-cytometric approach to reliably quantitate CLL cells to the level of 0.0010% (10(-5)). The assay comprises a core panel of six markers (i.e. CD19, CD20, CD5, CD43, CD79b and CD81) with a component specification independent of instrument and reagents, which can be locally re-validated using normal peripheral blood. This method is directly comparable to previous ERIC-designed assays and also provides a backbone for investigation of new markers. A parallel analysis of high-throughput sequencing using the ClonoSEQ assay showed good concordance with flow cytometry results at the 0.010% (10(-4)) level, the MRD threshold defined in the 2008 International Workshop on CLL guidelines, but it also provides good linearity to a detection limit of 1 in a million (10(-6)). The combination of both technologies would permit a highly sensitive approach to MRD detection while providing a reproducible and broadly accessible method to quantify residual disease and optimize treatment in CLL.

Figure 1

(a) 8-CLR 1-tube panel dilution analysis. Data from serial dilution analysis of 5 × 1:10 dilutions on five CLL cases were analyzed using a single-tube eight-marker panel. Markers with a dark gray fill indicate results above the limit of quantitation; markers with a light gray fill indicate results below the limit of quantitation but above the limit of detection; and markers with no fill indicate results below the limit of detection. For log-transformed data above the LOQ, linearity=1.01, correlation coefficient Pearson R=0.99. (b) Confirmation that six markers are sufficient for detection of MRD: Bland–Altman plot comparing MRD level calculated using the single-tube eight-marker combination against the MRD level calculated using a six-marker core panel, that is, excluding CD3 and CD22. For log-transformed data above the LOQ, linearity=1.00, correlation coefficient Pearson R=1.00, average difference=−0.0026 log, 95% limit of agreement ±0.012 log. LOQ, limit of quantification.

Figure 2

(a) Six-marker core panel dilution analysis. Data from serial dilution analysis of eight CLL cases diluted into normal peripheral blood leukocytes in serial 1:10 (n=5) or 1:5 dilutions (n=3). Markers with a dark gray fill indicate results above the limit of quantitation; markers with a light gray fill indicate results below the limit of quantitation but above the limit of detection; and markers with no fill indicate results below the limit of detection. For log-transformed data above the LOQ, linearity=1.02, correlation coefficient Pearson R=1.00. (b) Acceptable interoperator variation in analysis of the six-marker core panel: analytical variation was tested using 19 operators with experience of flow cytometry but not direct experience of MRD analysis in CLL using the six-marker core panel. The results showed good concordance at the 0.010% threshold with and acceptable 95% limit of agreement of ±0.27 log for results above the limit of quantitation. For log-transformed data above the LOQ, linearity=1.02, correlation coefficient Pearson R=0.99, average difference=0.013 log, 95% limit of agreement ±0.27 log.

Figure 3

Identification of the optimal reagent specification required for MRD analysis. CLL cells and normal B cells were separately labeled with CD19 PE-Cy7 and CD19 PerCP-Cy5.5, respectively, prior to washing and mixing to create five samples which were then incubated with serial dilutions antibodies at varying concentrations (neat to 1:243, serial 1:3 dilutions). This permitted calculation of the degree to which CLL cells overlapped with normal B cells in fluorescence intensity for each marker across a range of signal intensities. The signal intensities were calculated using internal positive and negative controls and plotted against the proportion of cases with suboptimal separation of CLL cells from normal B cells, where suboptimal separation was defined as an increase in overlap of 10% or more compared with the lowest overlap for each dilution series.

Figure 4

ClonoSEQ high-throughput sequencing (HTS) shows good linearity to one CLL cell in one million leukocytes. Analysis of three CLL cases diluted into leukocytes from leucodepletion filters in serial 1:10. Each CLL clone was tagged with two sequences, one productive and one non-productive. The plot shows the CLL sequence as a percentage of nucleated genomes. Each case has a different marker shape (square for case 1, circle for case 2 and diamond for case 3) with no fill for the productive sequence data, gray fill for the non-productive sequence and black fill for the average. For log-transformed data above the limit of detection, linearity=1.12, correlation coefficient Pearson R=0.98.

Figure 5

Comparison of the six-marker core MRD flow assay with the 4-CLR 4-tube MRD flow assay and ClonoSEQ HTS. Data from serial dilution studies (n=18) and from patient samples after FCR-based therapy (n=12) were analyzed using the harmonized 4-CLR ERIC panel, the six-marker core panel and ClonoSEQ high-throughput sequencing. (a) Comparison of the six-marker core MRD flow assay with the ERIC 4-tube 4-CLR panel: For log-transformed data above the LOQ, linearity=0.99, correlation coefficient Pearson R=1.00, average difference=−0.044 log, 95% limit of agreement ±0.17 log. (b) Comparison of the six-marker core MRD flow assay with ClonoSEQ HTS: For log-transformed data above the LOQ, linearity=0.89, correlation coefficient Pearson R=0.75, average difference=−0.12 log, 95% limit of agreement ±1.3 log.