Discovering Human RNA Aptamers by Structure-Based Bioinformatics and Genome-Based In Vitro Selection

Abstract

In vitro selection and structure-based searches have emerged as useful techniques for the discoveries of structurally complex RNAs with high affinity and specificity toward metabolites. Here, we focus on the design of a human genomic library that serves as the DNA template for in vitro selection of RNA aptamers. In addition, the structural solutions obtained from the in vitro selection can be used for structure-based searches for discovery of analogous aptamers in various genomic databases.

Introduction

Over the past two decades, in vitro selection (also known as SELEX) has served as a powerful tool for the discovery of novel DNA and RNA aptamers. Since then, extensive selection and structural studies have highlighted the structural diversity within a given random pool of DNA or RNA sequences (Stoltenburg, Reinemann, & Strehlitz, 2007). Initial in vitro selection studies utilized pools of synthetic random DNAs flanked by fixed, primer-binding sequences that were transcribed into amplifiable RNAs of similar diversity to determine the frequency of aptamers capable of binding a target molecule (Ellington and Szostak, 1990, Tuerk and Gold, 1990). More recently, modifications to the in vitro selection procedure have aimed to identify naturally occurring aptamers and other functional nucleic acids by using genome-derived DNA pools as templates for selections. In the case of adenosine, but not GTP, both synthetic and genomic DNA selections revealed a number of structurally conserved aptamer sequences (Burke and Gold, 1997, Curtis and Liu, 2013, Davis and Szostak, 2002, Sassanfar and Szostak, 1993, Vu et al., 2012). These adenosine-binding motifs are sequence-independent and represent a rare example of convergent molecular evolution spanning both genomic and synthetic sequence space.

Genomic SELEX was introduced by Singer and Gold using a genomic library for in vitro selection studies using a set of primers consisting of a fixed 5′ end sequence and a randomized 3′ tail. This allowed for amplification of fragmented human, yeast, and Escherichia coli genomic DNA in vitro, followed by size selection and primer extension to allow for transcription (Singer, Shtatland, Brown, & Gold, 1997). On the other hand, Salehi-Ashtiani et al. designed a genomic pool by partial digestion of human genomic DNA using DNase I. After digestion, hairpin sequences of known composition were ligated onto the genomic DNA, subjected to single-stranded digestion and then amplified by primer extension (Salehi-Ashtiani, Lupták, Litovchick, & Szostak, 2006).

Among the best-characterized aptamer structures is the adenosine-binding motif. Both synthetic and genomic selections reveal a conserved binding pocket consisting of an 11-nucleotide loop and a bulged G formed by two flanking helical motifs (Fig. 2.1). Nuclear magnetic resonance and mutation studies have shown that these conserved nucleotides and flanking helices are required for the formation of a binding pocket to allow base stacking and hydrogen-bonding interactions with the ligand (Dieckmann et al., 1997, Dieckmann et al., 1996, Jiang et al., 1996, Vu et al., 2012). Although the sequence compositions of the flanking helical motifs vary, the adenosine-binding loop is largely sequence conserved, and both of these properties are exploited with structure-based search algorithms.

Structure-based search algorithms are powerful tools in the discovery of functional RNAs. Their success comes from the ability to find sequences in unrelated, unprocessed sequence data that match complex motifs (Gautheret, Major, & Cedergren, 1990). Their appeal lies in the user accessibility: the ease of use, the flexibility in descriptor design, and the efficiency and speed of searches. Structure-based search programs are used to identify sequences capable of fitting into a given secondary structure. These programs match the patterns of base-paired and single-stranded regions as defined by the user in a descriptor file. Furthermore, the descriptor allows the user to specify regions of strict Watson-Crick base pairing, wobble pairs, mismatches, and single-nucleotide insertions in helices. Two user-friendly programs with similar syntax are RNABOB (ftp://selab.janelia.org/pub/software/rnabob/) and RNArobo (Jimenez, Rampasek, Brejova, Vinar, & Lupták, 2012); the implementation of neither of these programs requires extensive programming skills. The implementation of RNABOB is as previously described (Riccitelli & Lupták, 2010) and will be outlined briefly below.

Our approach here focuses on the design of a genomic DNA pool for use in in vitro selection. In principle, the pool can sample the entire genome of the target organism at single-nucleotide resolution (in both directions, with respect to the engineered RNA polymerase promoter), independent of expression of individual genes, but lacks sequences corresponding to spliced and otherwise processed transcripts. Structural characterization of the resulting aptamers can be used to generate structure descriptors for mapping sequences against a genome database.