CircRNAs: the transcriptional landscape of haematopoiesis at higher definition

Abstract

Cell states in haematopoiesis are controlled by complex circuits, involving master regulators transcription factors and a growing family of RNA species, shaping cell phenotype, its maintenance and plasticity. Amongst RNA species, circular RNAs (circRNAs) are rapidly gaining the status of particularly stable transcriptome members with distinctive qualities. Regarding molecular functions, circRNAs modulate host gene expression, compete for binding of microRNAs, RNA-binding proteins and translation initiation, and participate in regulatory circuits. RNA-seq studies identified thousands of circRNAs with developmental stage- and tissue-specific expression corroborating earlier suggestions that circular isoforms are a natural feature of the cell expression program. circRNAs are abundantly expressed and highly regulated also in the haematopoietic compartment, as described by recent and preliminary studies on circRNAs in blood cells.
\n
\nIn my PhD project we focused on the development of a bioinformatics pipeline to detect, quantify and characterize circRNAs from RNA-seq data, by combining both publicly available tools and custom scripts. Aiming to increase the discovery power of the pipeline as well as results robustness, we combined four programs for circRNA detection in parallel.
\n
\nThe pipeline was tested on a publicly available dataset of haematopoietic lineage cells such as Haematopoietic Stem Cells, Lymphoid progenitors, Myeloid Progenitors and Megakaryocyte–Erythroblast Progenitors. This pilot analysis allowed to retrieve a great number of circRNAs despite features of the data that were not optimal for circRNA detection. Major results from the pilot study were the identification of distinct sets of circRNAs specifically expressed in different cell types, and the feasibility and convenience of circRNA detection in published datasets to complement the original studies.
\n
\nIn parallel, we studied circRNA and linear RNA expression in differentiated cells of the haematopoietic compartment, specifically B cells T cells and Monocytes. We produced RNA-seq data of 12 samples, obtained by cell sorting from peripheral blood of healthy donors and using ribosomal RNA depletion for the library construction.
\nOut of the over 115000 detected backsplices supported by at least two reads, we selected putative circRNAs found by at least two methods, gaining also indirect support that most of them are truly circular forms thanks to independent evidence. This subset consists of 26211 circRNAs expressed by 7307 different genes, with 38.6% of genes expressing one circRNA each, and 40.7% of genes producing from 2 to 5 different circular isoforms and the remaining genes expressing a higher number of circRNAs. The large majority of circRNAs are exonic, 11.5% have backsplice ends falling into intronic regions and only a few (2.5%) probably derive from genomic regions annotated as intergenic. Comparison with the analysis of the linear transcriptome pointed out that the expression levels of linear and circular RNAs expressed from the same gene have only a very slight tendency toward positive correlation, with most of the pairs showing scarce or even negative correlations, suggesting specific regulatory mechanisms underlying the expression of circRNAs.
\n
\nThe comparison between B cells T cells and Monocytes indicated groups of circRNAs expressed in all the cell types and specific of each cell type. Unsupervised analyses of expression profiles showed for the first time specificities of circRNA expression associated to different blood cells. B cells and T cells circRNAomes are similar from quantitative and qualitative points of view, whereas Monocytes express a lower number of circRNAs and have a more specific circRNAome. Indeed, differential expression tests outlined sets of circRNAs with significantly variable expression in B cells compared to Monocytes (2589), B cells compared to T cells (168) and Monocytes compared to T cells (977). Differentially expressed circRNAs are associated to genes enriched in protein products involved in key blood processes and pathways. Finally, we focused on 74 circRNAs upregulated in B cells compared to both Monocytes and T cells, 40 upregulated in T cells and 159 upregulated in Monocytes, for a total of 273 circRNAs with differential expression and cell specificity. Additional criteria for circRNA prioritization selected circRNAs associated to genes with key functions in haematopoiesis, or altered/deregulated in haematologic malignancies. Prioritized circRNAs will undergo experimental validations. The sequence analyses for in silico prediction of possible circRNAs functions, as presence of multiple miRNA binding sites, protein binding motifs, or open reading frames, will be the starting point for experimental studies to better elucidate the functions of more promising circRNAs.
\n
\nIn conclusion we performed the first study of circRNAs in normal B cells T cells and Monocytes grounding on several biological replicates of each cell type being informative on circRNA differential expression. The integration of circular and linear RNA expression profiles with gene annotations and functions, in conjunction with differential expression data, produced new and original results. We showed that taking into account circRNA expression might add definition to the representation of transcriptome variations in normal haematopoiesis, posing the basis to better comprehend the role of circRNAs in the regulatory circuits of blood cells differentiation, which is a prerequisite for transferring this knowledge to research on haematological malignancies.