Schematic representation of the SELEX procedure. Genomic aptamers are selected via multiple rounds of selection and amplification. The procedure starts by transcribing a genomic DNA library into an RNA pool (1). The RNA pool can optionally be used for counter selection against unspecific ligands (2) and is then incubated with the bait under appropriate conditions to form the desired RNA–protein complex (3). Bait-binding RNAs are selected via filter binding or another chromatographic separation method and separated from non-binding species, which are discarded (4). The binding RNAs are then recovered, proteins are removed with PCI and urea is removed with ethanol precipitation (5). Finally, the pool is reverse transcribed and amplified via PCR, generating a new pool (6). This cycle is repeated until an appropriate amount of RNAs is enriched. The winning sequences can be cloned and sequenced, or alternatively, when a large number of sequences are expected deep-sequenced (7). In the Neutral SELEX control experiment, the selection step is omitted, and the cycle proceeds directly from transcription to reverse transcription (red arrow).

Quality control of the E. coli genomic library. The genomic library pictured was size-selected for lengths between 50 and 500 bp. (A) PCR analysis of single-copy genes. Different genomic segments, ranging from 133 to 368nt in length, were amplified with genome-specific primers from both E. coli genomic library (L) and genomic DNA (+). Products represent sections from the sodB, ompF, rpsB and rpoB genes. (B) Schematic representation of products resulting from PCR with one E. coli genome-specific and one library endpoint specific (fixFOR or fixREV) primer. (C) Analysis of size variation and endpoint distribution. PCR products of varying lengths result from amplification with different combinations of E. coli genome-specific (damFOR and damREV) and library specific (fixFOR and fixREV) primers.

Recovery and the resulting enrichment of genomic aptamers. The results of a deep-sequenced Genomic SELEX experiment for Hfq-binding aptamers in E. coli are shown here as an example. (A) Recovery rates of RNA after each cycle of Genomic SELEX steadily increase. Since the RNA was always in 10-fold excess to the protein, a recovery rate of 3.63% (H8) is relatively high. To increase stringency and further enrich the binding sequences after the 5th round of selection, the amount of RNA and protein were both reduced. (B) The enrichment levels drastically changed after increasing the stringency. The chart shows the cumulative frequency (y-axis) of bases at increasing enrichment depths (x-axis). The bar labeled “1” represents the fraction of bases in the enriched regions that have enrichment depth of 1 or greater (so it is 1.0), “2” represents 2 or greater, and so on. No sequences from H5 were enriched at a depth greater than 15, barely clearing the background levels from the library, which had a maximum depth of 10. In just 4 more rounds of selection with increased stringency, the maximum enrichment depth leapt up to 384. This provides a much clearer stratification of the tightly binding from the less competitive species.

Schematic representation of the E. coli genomic library. (A) Isolated DNA is used for the construction of the desired genomic library. Two hybrid primers, consisting of nine random nucleotides (red) followed by a specific sequence (light and dark blue), are used to synthesize the strands of the library. The hybrid primers anneal at 25 °C and are extended by the Klenow fragment of DNA polymerase I. After synthesis of both strands, reaction products are size-selected by denaturing gel electrophoresis. The fraction that contains library fragments of the desired size range is eluted and amplified by PCR with the fix-primer pair. The T7 promoter (dashed) for transcription of the library into RNA is introduced through the fixFOR primer. (B) Library fragments are overlapping and displaying a continuous endpoint distribution.

Melting temperatures of incorporated primers from the genomic library construction. The hyb-primers contain a fixed region and a randomized region, the latter of which binds the genomic DNA during the Klenow reactions. We evaluated the randomness of the library priming by repositioning the library primers at random positions in the genome, at the same distance apart from each other. As compared to this in silico-randomized library (yellow bars), the melting temperatures of the library primers (brown bars) are slightly lower, possibly due to the preference for stable DNA–DNA complexes in the reaction.

Using the enrichment sequence to facilitate visual interpretation and in silico boundary mapping. The enrichment sequences for two different clusters from the same deep-sequenced final pool of Hfq Genomic SELEX are visualized here (see Sections 3.2 and 3.3 for description of the analysis). Each read from deep sequencing represents a full-length aptamer recovered from the experiment. (A) The positional enrichment depth can be summarized using an “enrichment sequence”, a sequence of numbers representing the number of high-throughput reads aligned to each base position. Here we see a visualization of one such sequence, represented as a histogram. The first 15 bases occur in four reads, the next two, seven reads, and so on. The escalating number of reads at each position can give clues as to where the consensus sequence lies. In this case, the highest plateau of enrichment in this cluster, underlined in red, contains an instance of a winning motif from the Hfq Genomic SELEX experiment, which was 5′-AAYAAYAA-3′ [23]. This can be seen as an in silico version of the boundary mapping technique, previously used to find the minimal aptamer required for binding [26]. (B) The visualization of read alignments from high-throughput sequencing can take up more space than a screen can hold and is often difficult to interpret. Incorporating the enrichment sequence histogram into a genome browser compacts the visualization, as shown for Cluster 1238 rendered here by GBROWSE [39]. Additionally, the 2D analysis gives better insight into the genomic location of the cluster. In this example, we see the ygcN gene, whose start codon is at the left end of the beige bar, a cluster, represented with a green arrow and its corresponding enrichment sequence, visualized with a histogram. In a regular 1D analysis, the underlying region of the cluster would be deemed to be primarily upstream of the open reading frame. However, by comparing the areas under the histogram within and upstream of the open reading frame, the mass of the enrichment is clearly favoring the open reading frame, and furthermore is quite sharply increasing just downstream of the start codon.