Introduction of mismatches in a random shRNA-encoding library improves potency for phenotypic selection

Abstract

RNA interference (RNAi) is a mechanism for interfering with gene expression through the action of small, non-coding RNAs. We previously constructed a short-hairpin-loop RNA (shRNA) encoding library that is random at the nucleotide level [1]. In this library, the stems of the hairpin are completely complementary. To improve the potency of initial hits, and therefore signal-to-noise ratios in library screening, as well as to simplify hit-sequence retrieval by PCR, we constructed a second-generation library in which we introduced random mismatches between the two halves of the stem of each hairpin, on a random template background. In a screen for shRNAs that protect an interleukin-3 (IL3) dependent cell line from IL3 withdrawal, our second-generation library yielded hit sequences with significantly higher potencies than those from the first-generation library in the same screen. Our method of random mutagenesis was effective for a random template and is likely suitable, therefore, for any DNA template of interest. The improved potency of our second-generation library expands the range of possible unbiased screens for small-RNA therapeutics and biologic tools.

Figure 1. Introduction of mismatches by random mutagenesis.

(A) Three steps were used to introduce mutations into the random template. Step 1: Extension reaction, minus one of the four dNTPs; in this example, minus dGTP. The extension should in theory stop at the first C. Step 2: Error prone reverse transcription forcefully incorporate a mismatched base opposite the C, still minus dGTP, but with different ratios of dATP, dCTP, and dTTP to compensate for their different paring affinities with C. Depending on the length of incubation, different lengths of stalled fragments will result. Step 3: After mutations are introduced, the extension reaction is completed with all four dNTPs present. (B) Abbreviated depiction of the rest of the library synthesis (described in detail previously [1]). Briefly the (single-stranded) DNA is nicked near the 5′ end, the hairpin is opened with an extension reaction using a strand-displacing polymerase, the ends are digested for cloning, the loop is digested asymmetrically and re-ligated to form a final loop sequence of 6 nucleotides (5′-CTAAAC’-3). For comparison, a non-mutagenized hairpin is also shown.

Figure 2. Sampling of 50 sequences from the second-generation library.

Out of 50 sequences sampled randomly, 35 (numbered in red, 70%) have mismatches, 12 (numbered in blue, 24%) have no mismatches, and 3 (numbered in gray, 6%) have non-complementary stem sequences and would not be expected to form a hairpin structure.

Figure 3. shRNAs selected from the second-generation library better protect FL5.12 cells from IL3 withdrawal.

(A) FL5.12 cells were transduced with different shRNA clones isolated from the side-by-side screens of the first-generation (300K) and second-generation (3M) libraries. The cells were subjected to an overnight IL3 withdrawal. Survival percentages (percentages of GFP+/To-Pro-3- cells) are shown, relative to the beginning of IL3 starvation. The six clones offering the most protection, relative to a control shRNA, were clones 3M-3 (p<0.0001), 3M-4 (p = 0.10), 3M-6 (p<0.0001), 3M-9 (p = 0.019), 3M-10 (p<0.0001) and 300K-5 (p = 0.011). Three clones from the second-generation library (3M-3, -6, and -10) were all significantly more protective than clone 300K-5 (p<0.0001 for all three). (B) Clones 3M-3, -6, and -10 were compared to two hit shRNAs (1p and 3p) isolated in our previous study from the first-generation library. The improved survival was highly statistically significant, with p<0.0001 by Student’s t-test in pair-wise comparisons between any of the three clones (3M-3, -6, or -10) versus either 1p or 3p. (C) Sequences of clones 3M-3, -6, and -10 from the second-generation, mismatched library, and of clones 1p, 3p and 300K-5 from the first-generation, non-mismatched library.