Multiplexed Digital Quantification of mRNA Abundance: a Highly Reproducible, PCR–independent Assessment of BAALC, ERG, and MN1 mRNA Levels in Acute Myeloid Leukemia (AML) Bone Marrow (BM) and Peripheral Blood (PB) Samples.

Abstract 1599

Poster Board I-625

The greatest obstacle to routine clinical testing of gene expression levels has been the lack of reproducibility of currently used methodologies, such as quantitative reverse transcriptase PCR (qRT-PCR) and microarray expression profiling. While these assays are useful for retrospective analyses of batched samples, they cannot be used for upfront evaluation of individual patients (pts) for molecular risk and treatment assignment. To overcome this barrier, we tested a recently developed, high throughput, PCR-independent, digital quantification technology, the nCounter system (Nanostring® Technologies). This system counts individual mRNA molecules, rather than measuring non-linear fluorescence generated by PCR-amplified targets (eg qRT-PCR). Using 72 AML samples and spike-in controls, we and collaborators demonstrated that the nCounter system is highly reproducible, sensitive, and accurate to femtomolar concentrations (Payton, J, et al. JCI 119:1714-26; Geiss, G, et al. Nat Biotech 26:317-25). Here we validated this technology using an independent set of 101 pts with a diagnosis of de novo cytogenetically normal AML. At diagnosis, pts presented with FAB subtypes M0, M1, M2, M4, M5(A, B), had a median age of 43 years (range 19-59), median white blood count of 28.5× 10³/μL (range 1.4-273.0), median of 69% BM blasts (range 22-95) and median of 65% PB blasts (range 0-97). Paired BM and PB specimens were available for 27 pts; blast percentages were ≥ 20% for all paired specimens. We used the nCounter system to measure mRNA abundance (‘counts‘) of 27 genes whose expression correlates with clinical and/or pathological criteria, including 3 genes associated with prognosis (BAALC, ERG, MN1), and control/housekeeping genes (GAPDH, ABL, Actin). Briefly, mononuclear cells from pretreatment BM or PB were enriched on Ficoll-Hypaque gradients and RNA was isolated using Trizol reagent; 100ng of total RNA was assayed in triplicate by nCounter according to the manufacturer's protocols. The nCounter results demonstrated substantial reproducibility, with a median CV [coefficient of variation, (standard deviation/mean *100)] <6% across replicates. In addition, the nCounter counts for BAALC, ERG, and MN1 normalized to ABL were highly correlated with the ABL-normalized qRT-PCR results. Significant correlation was observed for all 3 genes, with the following Spearman correlation coefficients: BAALC r = 0.9, ERG r = 0.7, and MN1 r = 0.8 (all p<0.001). Correlation of BAALC, ERG, and MN1 nCounter counts with the expression levels measured by Affymetrix® HG-U133 plus 2.0 microarrays were also tested. Summary measures of microarray gene expression levels were computed using the Robust Multichip Average method, which incorporates quantile normalization of arrays. Significant correlation of nCounter and microarray results was observed, with Spearman correlation coefficients as follows: BAALC r = 0.96, ERG r = 0.8, and MN1 r = 0.8 (all p<0.001). For the 27 sets of paired samples, nCounter results for BM and PB were also significantly correlated, with Spearman correlation coefficients of BAALC r = 0.9, ERG r = 0.7, and MN1 r = 0.6 (all p<0.001). Because RNA quickly degrades if not promptly isolated from PB or BM, and degraded RNA often fails qRT-PCR assays, we determined whether RNA quality affected nCounter performance by assessment of standard quality parameters, including ratio of absorbance at 260 and 280 nm (260:280, a measure of RNA purity, acceptable 1.8-2.0) and RNA Quality Index (RQI, which assesses 18S:28S rRNA ratio and RNA degradation, 7-10 acceptable). Quality ranged from very high, with 260:280 ratios >1.9 and RQI scores >9, to relatively low, with 260:280 ratios <1.8, RQI scores <4, and degraded RNA visible on the Experion® RNA chip. Such a range of RNA quality is consistent with our experience with clinical specimens, which may be delayed in transit to the laboratory. Nevertheless, fewer than 3% of nCounter assays failed to generate acceptable results (11/393 assays), likely because no PCR step is required. Our results show that the nCounter system is a rapid, relatively inexpensive ($0.72/assay), and highly reproducible methodology that will be very useful for routine diagnostic testing of prognostic gene expression and upfront molecular-risk assessment for treatment guidance in AML pts.