Risk-based classification of leukemia by cytogenetic and multiplex molecular methods: results from a multicenter validation study

Abstract

Modern management of leukemia and selection of optimal treatment approaches entails the analysis of multiple recurrent cytogenetic abnormalities with independent diagnostic or prognostic value. We report the first multicenter validation of a multiplex molecular assay for 12 relevant fusion transcripts relative to cytogenetic methods. Performance was evaluated using a set of 280 adult and pediatric acute or chronic leukemias representative of the variety of presentations and pre-analytical parameters encountered in the clinical setting. The positive, negative and overall agreements were >98.5% with high concordance at each of the four sites. Positive detection of cases with low blast count or at relapse was consistent with a method sensitivity of 1%. There was 98.7% qualitative agreement with independent reference molecular tests. Apparent false negatives corresponded to rare alternative splicing isoforms not included in the panel. We further demonstrate that clinical sensitivity can be increased by adding those rare variants and other relevant transcripts or submicroscopic abnormalities. We conclude that multiplex RT-PCR followed by liquid bead array detection is a rapid and flexible method attuned to the clinical laboratory workflow, complementing standard cytogenetic methods and generating additional information valuable for the accurate diagnosis, prognosis and subsequent molecular monitoring of leukemia.

Keywords: RT-PCR; diagnosis; leukemia; molecular classification; multiplex; prognosis.

Figure 1

Representative MMA output. Each row represents the results from a single multiplex PCR amplification hybridized onto 12 target-specific probes in a single reaction. The resulting MFI signals generated by each probe-bound PCR product are shown for three control samples, a total RNA sample isolated from a translocation-negative cell line (HL60), 12 different synthetic fusion transcripts prepared by in vitro transcription and spiked in a background of HL60 RNA (400 ng input) and eight total RNA samples purified from translocation-positive cell lines (400 ng input). Target-specific positive signals above the qualitative cutoff (350 MFI) are highlighted. The three controls are designed to assess the validity of the multiplex amplification, hybridization and detection steps in every batch/run.

Figure 2

Evaluation of analytical sensitivity. (a) Total RNA purified from the indicated translocation-positive cell lines was tested with the MMA at 1000, 100, 10 or 1 ng per RT reaction. (b) The same total RNA samples were tested either undiluted (100%) or diluted at 10, 1% or 0.1% in a background of total RNA isolated from the translocation-negative cell line HL60 at a final input of 400 ng per RT reaction. (c) Same experiment as in (b) at 100, 10 or 1% dilution and 100 ng input. (d) BCR–ABL1 sensitivity controls tested in duplicate with the MMA at 600 ng input. The graphs show the average MFI signals generated by the target-specific probes (black bars) and by the GAPDH endogenous control probe (white bars) relative to the 350 MFI cutoff value (dash lines). The complete data set is presented in Supplementary Table 2.

Figure 3

Quantitative analysis of signal output. The box plot shows the distribution in the log space of the positive (Pos), negative (Neg) and endogenous control (GAPDH) signals generated at each of the four sites for the 280 samples tested with the MMA. The boxes represent the 25th, 50th (median) and 75th percentiles of the signal distributions for each category. The tails of the distributions are indicated by whiskers corresponding to 1.5 times the interquartile range (IQR=75th percentile value minus the 25th percentile value). The median MFI values for each signal distribution and the qualitative 350 MFI cutoff value (dash line) are also shown.

Figure 4

Limit of blank study. (a) Summary of the MFI probe signals obtained from repeat testing with the MMA of a no RNA control sample, a total RNA sample purified from the translocation-negative HL60 cell line, and eight total RNA samples purified from asymptomatic control donors' white blood cells (WBC control RNA). Minimum (MIN), maximum (MAX), median, mean and s.d. (STDEV) values for the 11 fusion-transcript-specific probes combined are shown for each sample type and overall. (b) Results by probe type for all sample types combined. The graph shows the mean, maximum (MAX) and twice the limit of blank (2 × LOB) values for each of the 11 fusion-transcript-specific probes relative to the qualitative 350 MFI cutoff value (dash line). The error bars represent the s.d. of each probe-specific distribution. The complete data set is presented in Supplementary Table 2.

Figure 5

Panel expansion. Representative example of results (MFI) with two prototype assays detecting 23 different targets prepared by in vitro transcription and spiked in a background of translocation- and mutation-negative HL60 RNA (400 ng input). (a) Specific detection of 13 fusion transcripts commonly found in CML and ALL. Results for the two samples false negative for MLL–AFF1 with the MMA at site 3 (study ID no. 146 and no. 184) are also shown. (b) Specific detection of three NPM1 mutant transcripts and seven fusion transcripts commonly found in AML. For this assay, the NPM1 wild-type sequence (NPM1 WT) is used as an endogenous control. Target-specific positive signals above the qualitative cutoff (350 MFI) are highlighted.