Quantitative Proteomic Approach for MicroRNA Target Prediction Based on O/O Labeling.

Motivation: Among many large-scale proteomic quantification methods, O/O labeling requires neither specific amino acid in peptides nor label incorporation through several cell cycles, as in metabolic labeling; it does not cause significant elution time shifts between heavy- and light-labeled peptides, and its dynamic range of quantification is larger than that of tandem mass spectrometry-based quantification methods. These properties offer O/O labeling the maximum flexibility in application. However, O/O labeling introduces large quantification variations due to varying labeling efficiency. There lacks a processing pipeline that warrants the reliable identification of differentially expressed proteins (DEPs). This motivates us to develop a quantitative proteomic approach based on O/O labeling and apply it on Kaposi sarcoma-associated herpesvirus (KSHV) microRNA (miR) target prediction. KSHV is a human pathogenic γ-herpesvirus strongly associated with the development of B-cell proliferative disorders, including primary effusion lymphoma. Recent studies suggest that miRs have evolved a highly complex network of interactions with the cellular and viral transcriptomes, and relatively few KSHV miR targets have been characterized at the functional level. While the new miR target prediction method, photoactivatable ribonucleoside-enhanced cross-linking and immunoprecipitation (PAR-CLIP), allows the identification of thousands of miR targets, the link between miRs and their targets still cannot be determined. We propose to apply the developed proteomic approach to establish such links.

Method: We integrate several O/O data processing algorithms that we published recently and identify the messenger RNAs of downregulated proteins as potential targets in KSHV miR-transfected human embryonic kidney 293T cells. Various statistical tests are employed for picking DEPs, and we select the best test by examining the enrichment of PAR-CLIP-reported targets with seed match to the miRs of interest among top ranked DEPs returned by statistical tests. Subsequently, the list of DEPs picked by the selected statistical test is filtered with the criteria that they must have downregulated gene expressions, must have reported as targets by an miR target prediction algorithm SVMcrio, and must have reported as targets by PAR-CLIP.

Result: We test the developed approach in the problem of finding targets of KSHV miR-K1. The RNAs of three DEPs are identified as miR-K1 targets, among which RAB23 and HNRNPU are novel. Results from both Western blotting and Luciferase reporter assays confirm the novel targets. These results show that the developed quantitative approach based on O/O labeling can be combined with genomic, PAR-CLIP, and target prediction algorithms for the confident identification of KSHV miR targets. The developed approach could also be applied in other applications.