Portal de Periódicos da CAPES

Carta Acesso aberto Produção Nacional Revisado por pares

BIBTEX RIS

Long non‐coding RNA s: biomarkers for acute leukaemia subtypes

2015; Wiley; Volume: 173; Issue: 2 Linguagem: Inglês

10.1111/bjh.13588

1365-2141

Carolina Pereira de Souza Melo, Catharina Brant Campos, Juliana de Oliveira Rodrigues, Joaquim Caetano de Aguirre Neto, Ângelo Atalla, Mara Albonei Dudeque Pianovski, Edna Kakitani Carbone, Luciana B. Q. Lares, Hélio Moraes‐Souza, Shirlei Octacı́lio-Silva, Fabiano Pais, Alessandro Clayton de Souza Ferreira, Juliana Godoy Assumpção,

MicroRNA in disease regulation

The analysis of the human genome and transcriptome has revealed a diverse population of DNA sequences that are transcribed into RNAs but do not code for proteins, called non-coding RNAs (ncRNAs) (Kapranov et al, 2010). Deregulation of specific long ncRNAs (lncRNAs – ncRNAs that are longer than 200 nt) has been found in leukaemias (Trimarchi et al, 2014) and this new field of research is starting to be explored. Until recently, Genome-wide Expression Profiling (GEP) studies on ncRNAs in leukaemias focused mainly on microRNAs (miRNAs). However, reports on the study of lncRNAs have emerged in the last two years. Differential expression of lncRNAs among different acute lymphoblastic leukaemia (ALL) subtypes (Nordlund et al, 2012) and between KMT2A (MLL)-rearranged ALL and non-rearranged KMT2A (Fang et al, 2014) were detected. Furthermore, NOTCH-dependent lncRNAs deregulated in T lineage-derived ALL (T-ALL) were also identified (Trimarchi et al, 2014), as well as a significant correlation between the expression of CEBPA and the expression of several lncRNAs in acute myeloid leukaemia (AML) cell lines (Hughes et al, 2014). With the aim of identifying molecular biomarkers of acute leukaemia subtypes, we obtained RNA expression profiles from bone marrow samples of de novo cases of AML (n = 110) and ALL (n = 97), diagnosed according to World Health Organisation classification system (Swerdlow et al, 2008)] using the Sure Print G3 Human GE 8 × 60 K microarray (Agilent Technologies, Santa Clara, CA). Probes with specific expression signatures for each 2-class model were identified by machine learning supervised analysis and the 60 most informative for each model were selected to build a classifier (see Data S1). All programs were run within a local installation of the GenePattern suite (Reich et al, 2006). The classifiers that achieved sensitivity and specificity rates of more than 95% were: AML versus heterogeneous group (the latter group includes control individuals and patients with other haematological diseases); ALL versus heterogeneous group; acute promyelocytic leukaemia (APL) versus other AMLs; AML CBFB-MYH11 versus other AMLs and T-ALL versus B lineage-derived ALL (B-ALL) (Tables SI–SIII). Accuracy and precision rates were also high, except CBFB-MYH11 precision (80%). During this process, we realised that a significant number (5–15%) of the highly informative probes selected for each model were lncRNAs (Table SIV). Most of them belong to the long intergenic ncRNAs class (lincRNAs), three are antisense strand RNA (aRNAs) and four are pseudogenes. We next evaluated expression patterns of four of the ncRNAs by real time quantitative reverse transcription polymerase chain reaction (qRT-PCR), using Taqman Gene Expression Assays (Applied Biosystems, Waltham, MA, USA). The expression of three of the genes studied by qRT-PCR (LOC728743, XLOC_009378 and LOC101929333) was significantly different between groups of patients (Fig 1), although the presence of outliers emphasises sample heterogeneity, a common issue in leukaemia disease. Differential expression of PRKCQ-AS1, selected for the B-ALL/T-ALL model, was not validated by Q-PCR. The PRKCQ protein is a kinase important for T-cell activation and the PRKCQ gene was shown to be up-regulated in T-ALL (Yeoh et al, 2002). It is possible that the PRKCQ-AS1 high expression detected in the microarray is actually an experimental artefact caused by erroneous strand priming during cDNA synthesis and this will be further investigated. LncRNAs were more frequent in the gene signature built to discriminate B-ALL from T-ALL patients, representing 15% of the genes differentially expressed. A non-supervised analysis using only the lncRNAs genes from this signature showed that B-ALL samples clearly clustered separately from T-ALL samples (Fig 2). The importance of lncRNAs in T cell differentiation has been recognised (Xia et al, 2014). Evidence of their involvement in T-ALL development is also emerging: a NOTCH-regulated lncRNA, LUNAR1, was shown to be required for leukaemia cells growth in vitro and in vivo (Trimarchi et al, 2014). Some of the lncRNAs detected in this work have already been associated with cancer, such as Colorectal Neoplasia Differentially Expressed (CRNDE). Initially identified as an lncRNA whose expression is highly elevated in colorectal cancer, it is also up-regulated in solid tumours and in AMLs, particularly French-American-British (FAB) classification types M2 and M3 (Ellis et al, 2014). In our study, the median CRNDE expression was ten times higher in APL than in other AML subtypes. However, the majority of the lncRNAs described here are transcripts of unknown function and their role in leukemogenesis is still unclear. Our methodological approach did not target lncRNAs specifically, but results showed that they were an important fraction of the biomarkers detected. The data confirm the notion that lncRNA expression may discriminate acute leukaemia molecular subtypes, providing in an additional tool to the classical tests used to categorise leukaemias and stratify patients. Our findings also contribute to the knowledge that lncRNAs have an important biological function in haematological malignancies rather than constituting transcriptional noise. Further efforts to elucidate the genes that might be regulated by these lncRNAs and the respective regulatory mechanisms involved are needed. The authors thank the patients and their families, as well as the nursing and medical teams for participating in this research. We thank Cristiane Lommez de Oliveira and Valéria Matarelli Calijorne for the research project grants management. This work was supported by FINEP-Inovação e Pesquisa (03·10·0060·01); Fundação de Amparo à Pesquisa do Estado de Minas Gerais (FAPEMIG -00385/12); and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq - 551150/2011-4). JGA and ACSF designed the research study. JCAN, AA, MADP, EKC, LBQL, HMS and SOS collected bone marrow samples and provided patient's diagnosis. CPSM, CBC and JOR performed the research. CPSM, FSMP and JGA analysed the data. JGA and CPSM wrote the paper. All authors approved the final manuscript. The authors declare no competing financial interests. Data S1. Supplementary methods. Table SI. Confusion matrix and predictive values for classification of acute leukaemia subtypes. Table SII. Confusion matrix and predictive values for classification of acute lymphoblastic leukaemia (ALL) subtypes. Table SIII. Confusion matrix and predictive values for classification of acute myeloid leukaemia (AML) subtypes. Table SIV. Overview of the lncRNAs selected in the gene signatures used to classify acute leukaemias in this study. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.