Differential expression of miRNAs in acute myeloid leukemia quantified by Nextgen sequencing of whole blood samples

New approaches are needed for understanding and treating acute myeloid leukemia (AML). MicroRNAs (miRs) are important regulators of gene expression in all cells and disruption of their normal expression can lead to changes in phenotype of a cell, in particular the emergence of a leukemic clone. We collected peripheral blood samples from 10 adult patients with newly diagnosed AML, prior to induction chemotherapy, and 9 controls. Two and a half ml of whole blood was collected in Paxgene RNA tubes. MiRNA was purified using RNeasy mini column (Qiagen). We sequenced approximately 1000 miRs from each of 10 AML patients and 9 controls. In subset analysis, patients with NPM1 and FLT3 mutations showed the greatest number of miRNAs (63) with expression levels that differed from control with adjusted p-value of 0.05 or less. Some of these miRs have been described previously in association with leukemia, but many are new. Our approach of global sequencing of miRs as opposed to microarray analysis removes the bias regarding which miRs to assay and has demonstrated discovery of new associations of miRs with AML. Another strength of our approach is that sequencing miRs is specific for the 5p or 3p strand of the gene, greatly narrowing the proposed target genes to study further. Our study provides new information about the molecular changes that lead to evolution of the leukemic clone and offers new possibilities for monitoring relapse and developing new treatment strategies.


Introduction
One area of research that has grown explosively in the past 15 years is the study of non-coding RNA. MiRNAs are 19-22 nucleotide long non-coding RNAs which regulate the expression of genes by sequence-specific binding to mRNA to either promote or block its translation [1]. This is a powerful level of epigenetic control for gene expression that can influence the phenotype of a cell [2]. Several authors have examined the role of miRNA in the transformation of hematopoietic stem cells into leukemic cells [3][4][5][6][7][8]. It is now well established that miRNAs play a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 a role in blocking differentiation of leukemic cells and promoting their unchecked cell division [9,10].
To date, researchers have analyzed miRNA expression in leukemic cells using arrays or RT-PCR. Our work is the first quantitative sequencing of miRNAs found in peripheral blood from patients with newly diagnosed acute myeloid leukemia (AML). We demonstrate here proof of principle for this relatively simple method. We compare the expression levels to normal controls to find statistically significant increase or decrease in levels of expression of approximately 1000 miRNAs. Subset analysis of AML patients with FLT3 or NPM1 mutations yields further information regarding statistically significant changes in expression of specific miRNAs.

Materials and methods
This protocol was reviewed and approved by the Upstate Medical University Institutional Review Board. Blood was obtained following written informed consent from patients at University Hospital with newly diagnosed AML, prior to start of induction chemotherapy. Blood was also obtained following written informed consent from healthy volunteers involved in patient care at Upstate. 2.5 ml of whole blood was collected into Paxgene RNA preparation tubes and stored at -80C for batch processing. miRNA was purified using a Qiagen miRNA purification kit. The yield and quality of the RNA samples was assessed using the Agilent Bioanalyzer prior to library construction using the Illumina TruSeq Small RNA Sample Prep protocol (Illumina; San Diego, California). Multiplexed samples of RNA that exceed quality control metrics (RIN > 6.0) were run on an Illumina NextSeq500 instrument at a targeted depth of 10 million reads per sample. After filtering and trimming of index and adapter sequences, whole genome alignment of the miR FASTQ reads was performed using the Homo sapiens/hg21 reference genome in the SHRiMPS aligner included in the miRNAs analysis application available in BaseSpace (Illumina), as well as the sRNA Toolbox application suite.

Statistical method
The analysis of the RNA-seq data was performed following the pipeline available from the limma packages [11] in the Bioconductor project [12]. Log2counts per million (logCPM) transformation was applied before normalization and linear model fitting. Empirical Bayes moderation was carried out to obtain robust estimates of gene-wise variability and the final pvalues from the linear model with appropriately designed contrasts were adjusted by the Benjamin-Hochberg procedure for a targeted false discover rate of 0.05. Volcano plots and boxplots were used to graphically examine the differences between groups.

Results
We sequenced all miRNAs in peripheral whole blood from ten patients with newly diagnosed AML and nine normal controls. Table 1 shows the characteristics of the ten patients who entered the study. There were five males and five females, ages ranged from 42 to 87. Initial white blood cell (WBC) count ranged from 1.1 x 10 3 /ul to 88 x 10 3 /ul, with blast percentage of 3.8 to 73. Phenotype was determined by standard hematopathology staining for surface markers. For all ten patients, we recorded presence (pos) or absence (neg) of NPM1 or FLT3 mutations and cytogenetic analysis.
We obtained sequence for 996 miRNAs and averaged the amount of each miRNA for the ten AML patients and the nine controls. Levels of expression were compared with a t-test and the adjusted P-value was calculated for each miRNA. In our first analysis of all ten patients versus controls, we identified only thirteen miRNAs with expression levels that showed a statistically significant difference (adjusted P-value less than 0.05) [13].
We then performed subset analysis of our data and found that patients who were double negative for mutations (NPM1-and FLT3-) had only one miRNA with adjusted P-value less than 0.05, while patients who were double positive for mutations (NPM1+ and FLT3+) had 63 miRNAs with adjusted P-value less than 0.05 [14]. Results are shown in Fig 1 (volcano plots).
In Fig 1, the volcano plot for the comparison of all ten AML patients versus control is shown in panel A. Log2 Fold Change is shown on the x-axis and-log10 adjusted p-value is shown on the y-axis. Points in dark font indicate miRNAs with statistically significant log fold change and adjusted p-value (four are increased in AML, nine are decreased in AML). Panel B shows the volcano plot for the subset of patients who were double positive (NPM1+/FLT3+, n = 3), with 63 miRNAs having statistically significant change in expression. Similarly, panel C shows the volcano plot for the subset of AML patients who were double negative (NPM1-/FLT3-, n = 5) with only one statistically significant change in miRNA expression. Clearly, the subset analysis identified a greater number of miRNAs with statistically significant change in expression in the AML patients with NPM1 and FLT3 mutations (double positive). Table 2 shows the first twenty (out of 996) miRNAs that were identified, listed by rank order of adjusted P-value, lowest to highest, in comparison of AML versus control samples. The center list is from all ten AML patients versus nine controls. The first 13 show statistically significant difference with adjusted p-value less than 0.05. The list on the left is from the   Table 3 shows all 63 miRNAs that had statistically significant differences in expression between NPM1+/FLT3+ patients and controls, listed in decreasing order of log2Fold Change (log2FC). Positive values of log2FC indicate increased expression in the AML patients, while negative values of log2FC indicate decreased expression in the AML patients. MiRs that have not previously been associated with AML are indicated with asterisks.

Discussion
Rather than analyze by microarray which would identify only 300-400 miRNAs, we chose to sequence all miRNAs that were obtained in whole blood samples from patients with newly diagnosed AML and controls. We quantified expression of 996 miRNAs from each patient and control and performed statistical analyses to determine which miRNAs were increased or decreased in AML patients versus normal controls. Subset analysis revealed the most differences in patients with double positive AML (NPM1+/FLT3+ mutations). NPM1 mutations are the most common genetic abnormalities in AML (50-60% of cytogenetically normal AML and 30% of all AML) [15]. Up to one-third of NPM1+ patients also have a mutation in FLT3, which counteracts the favorable prognosis of the NPM1 mutation [16]. The NPM1 gene encodes a 32-kDA protein that is involved in numerous cellular processes. It can function as an oncogene and a tumor suppressor gene depending on expression levels, interacting proteins, and cellular compartmentalization. Dozens of mutations have been described in NPM1 and how they contribute to leukemogenesis is not known. Finding 63 distinct miRNAs either upregulated or downregulated in NPM1+ /FLT3+ AML suggests that there may be multiple molecular pathways disrupted that contribute to the leukemic phenotype. On our list of miRNAs associated with NPM1+/FLT3+ AML, miR-10a-5p showed the most statistically significant adjusted p-value as well as the highest fold change. This miRNA has been described in patients with NPM1 mutations and high expression levels are associated with good response to induction chemotherapy [17]. More recent studies demonstrate a role for miR-10a/ b in regulating the proliferation and differentiation of HL-60 leukemic cells in vitro [18].
One miR family that we identified with statistically significant change was the miR-181 family (181a-3p, 181a-2-3p and 181b-5p). This family has been shown by others to be consistently increased in AML patients [8,19,20]. There are numerous possible target mRNAs for the miR-181 family and it has been proposed as a therapeutic target. Another report has shown that miR-199b is consistently decreased in AML and we also found that to be the case in our patients [21]. We found increased expression of miR-223-3p and this has been shown to regulate granulopoiesis in mice [22].
We compared our list of miRNA differences determined by Nextgen sequencing to lists of miRNA differences found by hybridization arrays or RT-PCR [10,[23][24][25][26][27]. Many of the miR-NAs identified by these methods were the same as ours, confirming our ability to generate meaningful data with our approach. Differences in miRNAs identified can be attributed to differences in technique of sample processing (whole blood versus isolated cells or serum). Identification of novel miRNAs in association with AML are likely due to the greater sensitivity of our approach, i.e. sequencing all miRNAs obtained in whole blood. New therapeutic approaches are being created to target specific miRNAs in leukemic patients [28,29]. As we continue to investigate the role of miRNAs in leukemogenesis, the approaches of diagnosis, treatment, and post-treatment monitoring will be greatly improved.
Supporting information S1