A high-throughput screening method for evolving a demethylase enzyme with improved and new functionalities

Abstract

AlkB is a DNA/RNA repair enzyme that removes base alkylations such as N1-methyladenosine (m1A) or N3-methylcytosine (m3C) from DNA and RNA. The AlkB enzyme has been used as a critical tool to facilitate tRNA sequencing and identification of mRNA modifications. As a tool, AlkB mutants with better reactivity and new functionalities are highly desired; however, previous identification of such AlkB mutants was based on the classical approach of targeted mutagenesis. Here, we introduce a high-throughput screening method to evaluate libraries of AlkB variants for demethylation activity on RNA and DNA substrates. This method is based on a fluorogenic RNA aptamer with an internal modified RNA/DNA residue which can block reverse transcription or introduce mutations leading to loss of fluorescence inherent in the cDNA product. Demethylation by an AlkB variant eliminates the blockage or mutation thereby restores the fluorescence signals. We applied our screening method to sites D135 and R210 in the Escherichia coli AlkB protein and identified a variant with improved activity beyond a previously known hyperactive mutant toward N1-methylguanosine (m1G) in RNA. We also applied our method to O6-methylguanosine (O6mG) modified DNA substrates and identified candidate AlkB variants with demethylating activity. Our study provides a high-throughput screening method for in vitro evolution of any demethylase enzyme.

Figure 1.

Establishment of the high-throughput screening system for AlkB. (A) The scheme of the Broccoli RNA fluorescence assay for high-throughput screening of AlkB activity towards modifications in RNA. RT: reverse transcription. DFHBI-IT is the fluorophore that the Broccoli RNA binds to. (B) The 51 nt Broccoli RNA sequence with conservative bases labeled in green. Thin lines with arrows denote chain connectivity and the Leontis-Westhof symbols denote canonical and non-canonical base pairs (26). The structure is based on the secondary structure of Spinach 1.2 and the high similarity of their core domains (12,13). The numbering scheme for Broccoli follows Filonov et al. (11) and is used throughout. The 51 nt consists of nucleotides numbering from 29 to 79, and the 33 mer substrate used in our assay consists of nucleotides numbering from 29 to 61, with m¹G located at site 46. (C) The chemical structure of m¹G. (D) D135S and the catalytically dead mutant are expressed at the same level in E. coli cells. (E) Demethylation activity of D135S versus the catalytically dead mutant from E. coli crude lysates on m¹G in the RNA. Error bar indicates SD, n≥3.

Figure 2.

Screening at positions D135 and R210 of AlkB. (A) Residues D135 and R210 are located in close proximity to the modified nucleobase in the crystal structure (3BIE) (16). (B) Under the optimized condition, a high dynamic range was achieved to allow for the identification of mutants with higher reactivity. Error bar indicates SD, n≥3. (C, D) Screening results for libraries D135X and R210X (D135S). Each dot represents one variant in the library. The absolute fluorescence signal generated with each member in the library was adjusted by subtracting it from the signal generated with the dead mutant. Variants with adjusted signals below zero were presented as zero to indicate that no activity was observed for these variants under the condition. (E) In vitro demethylation assay with purified D135T and D135S mutants using the fluorescence assay confirmed that D135T has higher activity than D135S to demethylate m¹G. The demethylation activity was judged by comparing with the signal of the positive control, which was the unmodified RNA incubated with the respective variant. For direct comparison between the two variants, data were adjusted so that the positive controls for the two variants had the same level of fluorescence signals. (F) Plot of fraction of product as a function of time for the reaction containing 4 μM enzyme and 4 μM 9-mer m¹G-RNA. (G) Plot of (μM P/μM E)/min as a function of RNA concentration for reactions containing 4 μM enzyme. The value of (μM P/μM E)/min was calculated as ([S]/[E])*A*k (see Materials and Methods). Data shown are the average and standard deviation of three independent reactions for each condition.

Figure 3.

D135T exhibited similar capacity as D135S to remove human tRNA modifications in tRNA-seq experiments. (A) Mutation rate in sequencing of HEK293T samples for tRNA^Arg(ACG) and tRNA^Arg(CCT) in samples of no enzyme, wild type AlkB (WT), wild type AlkB plus the D135S mutant, and wild type AlkB plus the D135T mutant. I34 has an A-to-G mutation signature under all circumstances and is not sensitive to AlkB treatment. (B) Heatmap showing mutation rates at m¹A58 for all cytosolic tRNAs under the four conditions in tRNA-seq. Mutation rates are shown for the highest expressed tRNA isodecoder among each isoacceptor family (27).

Figure 4.

Application of the screening assay to a DNA substrate. (A) The scheme of the Broccoli RNA fluorescence assay for high-throughput screening of AlkB activity towards O6mG in DNA. (B) The Broccoli DNA substrate with O6mG used in the assay. Conserved regions are labeled in green. O6mG is located at site 59. (C) The chemical structure of O6mG. (D) ssDNA with O6mG produced much lower level of fluorescence than the unmodified DNA. (E) Screening of the D135X library against O6mG identified positive hits. The absolute fluorescence signal generated with each member in the library was adjusted by subtracting it from the signal generated with the dead mutant. Variants with adjusted signals below zero were presented as zero to indicate that no activity was observed for these variants under the condition. (F) Reproducing positive signals of hits from the screening using the fluorescence assay. Error bar indicates SD, n ≥ 3. Statistical differences between the positive hits and the dead mutant were determined by Welch's t tests.