Discovery and Quantification of Small RNA Molecules In Acute Myeloid Leukaemia Using Solexa High-Throughput Clonal Sequencing

Abstract 847

It was recently established by real-time PCR that the expression level of microRNAs (miRNAs) provides molecular signatures characteristic of the major translocation-mediated gene fusion events in acute myeloid leukaemia (AML). In particular we showed that t(15;17) leukaemias are associated with the up regulation of miRNAs located in the 14q32 imprinted domain. However results were restricted to a fraction of known miRNAs. We report the results of high-throughput clonal sequencing (Illumina) applied to systematically characterise and quantify the small non-coding (snc)RNA transcriptome of 38 libraries of size-fractionated RNA obtained from 36 cytogenetically and clinically distinct cases of AML and 2 normal bone marrow from healthy donors.

Sequences were aligned to the genome of reference and to miRNA and sncRNA databases using Novoalign. To account for imperfect DICER processing overhang at both 5′- and 3′-end was allowed. A location was recorded where >=2 reads mapped in any sample. Reads were excluded that did not map uniquely. In order to determine the association of the pattern of expression of miRNAs and sncRNAs with the karyotypic leukaemic subgroups, samples were first normalised. The mean of total number of reads was calculated for each sample. One profile was used as reference for scaling factor. The values of each of the other samples were divided by the scaling factors. ANOVA was applied to search for miRNAs and sncRNAs associated with AML cytogenetic subtypes. The genomic context of unidentified tags was screened with RNAfold from the Vienna package to identify potential new miRNA.

The sequencing approach proved highly reproducible, correlation coefficients as high as 0.99 were calculated for duplicate experiments, and showed unbiased quantitative features when compared with the real-time PCR measurements previously obtained for known miRNAs. The average number of reads per sample was 11,212,898. The distribution of sncRNAs and miRNAs was determined for each sample. A total of 621 miRNAs and 1525 different species of sncRNAs were expressed in the 38 samples. The miRNAs represented up to 70% of total reads. The C/D Box small nucleolar RNAs (snoRNAs) were highly represented among the remaining sncRNAs. An unsupervised hierarchical cluster analysis of the sncRNA reads for 38 samples revealed molecular signatures characteristic of the major translocation-mediated gene fusion events in AML. To find sncRNAs with statistically significant differences in expression level among the major cytogenetic groups, an ANOVA test was applied to the 38 AML samples, including the two normal bone marrows. One hundred seventy five sncRNAs passed a 0.5% false discovery rate (FDR) filter. We identified a set of sncRNAs, called CD/Box, differentially expressed in imprinted regions. Certain small RNAs showed expression only in one leukaemia subtype, e.g. the majority of miRNAs located at 14q32, 41 out of 51, were exclusively expressed in the t(15;17) leukaemia samples, thus confirming and extending previous findings from this laboratory. We also noted the high expression of a group of small nucleolar RNAs (snoRNAs) located in the same region. We identified 36 potential new miRNAs expressed in AML. We validated by real-time PCR the expression of one of the new miRNAs located on chromosome 11 which showed clear differential read counts between the samples. We also observed a high degree of variation in length in 80% of the miRNAs, particularly evident at the 3′ end. Some of the length variants were consistent with leukemia sub-typetype.

We have demonstrated the potential of using high-throughput sequencing to uncover novel aspects of the disease aetiology by combining the global miRNA expression pattern with the detection of multiple miRNA variants and new hairpin molecules.