Thermostable group II intron reverse transcriptase fusion proteins and their use in cDNA synthesis and next-generation RNA sequencing

Abstract

Mobile group II introns encode reverse transcriptases (RTs) that function in intron mobility ("retrohoming") by a process that requires reverse transcription of a highly structured, 2-2.5-kb intron RNA with high processivity and fidelity. Although the latter properties are potentially useful for applications in cDNA synthesis and next-generation RNA sequencing (RNA-seq), group II intron RTs have been difficult to purify free of the intron RNA, and their utility as research tools has not been investigated systematically. Here, we developed general methods for the high-level expression and purification of group II intron-encoded RTs as fusion proteins with a rigidly linked, noncleavable solubility tag, and we applied them to group II intron RTs from bacterial thermophiles. We thus obtained thermostable group II intron RT fusion proteins that have higher processivity, fidelity, and thermostability than retroviral RTs, synthesize cDNAs at temperatures up to 81°C, and have significant advantages for qRT-PCR, capillary electrophoresis for RNA-structure mapping, and next-generation RNA sequencing. Further, we find that group II intron RTs differ from the retroviral enzymes in template switching with minimal base-pairing to the 3' ends of new RNA templates, making it possible to efficiently and seamlessly link adaptors containing PCR-primer binding sites to cDNA ends without an RNA ligase step. This novel template-switching activity enables facile and less biased cloning of nonpolyadenylated RNAs, such as miRNAs or protein-bound RNA fragments. Our findings demonstrate novel biochemical activities and inherent advantages of group II intron RTs for research, biotechnological, and diagnostic methods, with potentially wide applications.

Keywords: miRNA; next-generation sequencing; qRT-PCR; retrovirus; transcriptome.

FIGURE 1.

Thermostable group II intron RT fusion proteins. (A) Comparison of group II intron TeI4c and retroviral HIV-1 RTs. Group II intron RT domains: RT with conserved sequence blocks RT-1 to RT-7, corresponding to the fingers and palm of retroviral RTs; X/thumb, with predicted α-helices (top) corresponding to those in the HIV-1 RT thumb; DNA-binding (D), and DNA endonuclease (En). Group II intron RTs have an N-terminal extension (RT-0) and “insertions” between the conserved RT sequence blocks (RT-2a, RT-3a, etc.) that are absent in retroviral RTs (Blocker et al. 2005; Lambowitz and Zimmerly 2011). Some group II intron RTs (e.g., GsI-IIC in this work) lack the En domain. (B) Group II intron RT fusion proteins. MalE-RT constructs have a MalE tag fused to their N terminus via a flexible linker with a TEV protease-cleavage site (underlined). MRF or NRF constructs have MalE or NusA solubility tags, respectively, fused to their N terminus via a rigid linker containing five alanines (underlined). For rigid fusions, the MalE tag has charged amino acid residues changed to alanines (italics), and the NusA tag is missing the two C-terminal amino acid residues. (C) Temperature profiles of RT activity. Poly(rA)/oligo(dT)₄₂ and [³²P]dTTP substrates were incubated with TeI4c-MRF (50 nM, 90 sec) or other indicated RTs (100 nM, 5 min), and polymerization of [³²P]dTTP was quantified by binding to DE81 paper. Temperature profiles for additional group II intron RT fusion proteins in this assay are shown in Supplemental Figure S1. (D) RT activity of TeI4c-MRF RT constructs with different solubility tags and linkers. Assays were done as in C with 50 nM enzyme for 90 sec at 60°C. Bar graphs show the mean ± standard deviation (error bars) for three determinations.

FIGURE 2.

Thermostability and processivity of group II intron RTs. (A) Gel assays of cDNA synthesis at different temperatures. A 509-nt in vitro–transcribed RNA (pBluescript KS(+)/AflIII) with a 5′-³²P-labeled (star) primer (AflIIIR) annealed near its 3′ end was incubated for 30 min with TeI4c-MRF (2 μM), GsI-IIC-MRF (200 nM), or SuperScript III (10 units/μL) RTs, and the products were analyzed in a denaturing 6% polyacrylamide gel. Arrowheads to the right of the gel indicate the position of full-length cDNAs, and numbers to the left indicate positions of size markers (10-bp ladder). The regions of the gels containing the labeled DNA primer are shown in Supplemental Figure S2. (B) Taqman qRT-PCR. A 1.2-kb kanR RNA with primer P078 annealed near its 3′ end was reverse transcribed with TeI4c-MRF (200 nM) for 30 min at 60°C. The Table shows cDNA copies detected with primer sets 188–257 and 562–634, which detect cDNAs of 920 and 546 nt, respectively. (C) Capillary electrophoresis assays of cDNA synthesis. An 807-nt in vitro transcript containing an Ll.LtrB-ΔORF group II intron RNA with a fluorescently labeled DNA primer (5′ fluorophore WellRED) annealed near its 3′ end was reverse transcribed for 30 min with TeI4c-MRF RT (1 μM), GsI-IIC-MRF RT (200 nM), or SuperScript III (10 units/μL). cDNA lengths were determined relative to fluorescently labeled DNA markers (data not shown).

FIGURE 3.

Gel assay of processivity of cDNA synthesis. A 807-nt in vitro transcript containing an Ll.LtrB-ΔORF group II intron RNA with a 5′-³²P-labeled primer annealed near its 3′ end was incubated for 30 min with TeI4c-MRF (2 μM) or GsI-IIc-MRF (1 μM) at 60°C or SuperScript III (10 units/μL) at 55°C in the presence of excess poly(rA)/oligo (dT)₄₂ as a trap, and the products were analyzed in a denaturing 6% polyacrylamide gel alongside a 5′-labeled 10-bp ladder (M). The processivity (average length of template copied per initiation) was calculated by using the equation Σ(L_n.I_n)/Σ(I_n), where L_n is the length and I_n is the intensity of each analyzed cDNA fragment.

FIGURE 4.

Error rates of different RTs determined by using an M13-based lacZ forward mutation assay. A 269-nt in vitro–transcribed RNA (pBluescript KS(+)/PvuI) encoding a segment of the LacZ α-fragment with annealed primer pBluescript 550R was reverse transcribed with TeI4c-MRF, GsI-IIC-MRF, or SuperScript III RTs, as described in the Materials and Methods. The resulting cDNAs were annealed to uracil-containing phage M13 single-stranded DNA, electroporated into E. coli MC 1061 F+ cells (Lucigen), and scored by plaque assays to determine the numbers of blue and white plaques. The mutation frequency was calculated as the ratio of white plaques to the total number of plaques. The error rate was calculated by dividing the mutation frequency by the number of nucleotide residues in the reverse-transcribed region at which changes would give a lacZ missense mutation. The background error rate was determined by electroporation of purified single-stranded M13 DNA. Sequence errors detected in cDNAs synthesized by different RTs are summarized below. “−1” and “−2” indicate −1 and −2 frameshifts, respectively; sequences complementary to the primer are shown in red.

FIGURE 5.

RNA-seq with a group II intron and retroviral RT. HeLa or MCF-7 RNAs were annealed with an oligo(dT)₄₂ primer and incubated with TeI4c-MRF RT (1.24 μM) at 60°C or SuperScript III (10 units/μL) for 2 h at 50°C. The cDNAs were converted into RNA-seq libraries and paired-end sequenced on an Illumina HiSeq. (A) Distribution of reads per unit length for transcripts of different size classes. Reads were aligned using Eland-32 to a set of ∼7203 transcripts curated by selecting the longest isoform of each annotated gene from RefSeq (downloaded 11/2010), removing sequences containing ambiguous bases, and requiring that >95% of bases could be uniquely mapped to RefSeq and have mean base coverage >3X in standard brain and/or UHR mRNA data sets. (B) Error frequencies. Raw data were base-called using the Illumina Off-Line Basecaller (OLB version 1.9), and the resulting reads were aligned to human NCBI reference build 36 and splice junctions from UCSC refFlat (downloaded 02/2010) using Eland RNA (Casava 1.7) with default parameters. Potential RT errors were detected by looking for single-base mismatches relative to the reference sequence in overlapping sequence in both reads R1 and R2 of a paired-end cluster. Both R1 and R2 were required to have a base quality >25 and belong to a perfectly overlapping section of length ≥20 nt. Base mismatches common to both the TeI4c-MRF and SuperScript III libraries, which include sequence polymorphisms compared with the reference RNAs, were filtered out.

FIGURE 6.

Template-switching activity of group II intron and retroviral RTs. (A) Gel assay. The initial ³²P-labeled IA–P1 RNA/Pc DNA template-primer substrate (50 nM) and equimolar miRNAx were incubated with TeI4c-MRF RT (2 μM, 60°C) or SuperScript III (10 units/μL, 50°C; SSIII) for 15 min in the standard reaction medium for each enzyme (see Materials and Methods). The products were analyzed in a denaturing 20% polyacrylamide gel, which was scanned with a PhosphorImager. Lane “−RT” shows the IA–P1 RNA/Pc DNA substrate incubated under TeI4c-MRF RT conditions without RT. (Lane M) ³²P-labeled 10-bp ladder size markers. (B) Template and primer sequences. The miRNAx target RNA has two randomized nucleotide residues (NN; blue) at each end to assess template-switching biases (Supplemental Fig. S4). The initial IA–P1 template RNA has a 3′ aminomodifier (AmMO) to impede template switching to that RNA end, and the Pc DNA primer is 5′ ³²P-labeled and has an internal deoxyuridine (underlined) for relinearization of cDNAs after circularization with uracil–DNA excision mix (UDE; see below). (C) Protocol for the construction of cDNA libraries via group II intron RT template switching. In the first step, the group II intron RT template switches from the IA–P1 RNA/Pc DNA template/primer to miRNAx to generate a continuous cDNA that links the IA–P1 adaptor sequence to that of miRNAx. The products are then incubated with RNase H to digest the RNA template, gel-purified, and circularized with CircLigase. After digestion of unincorporated primers with exonuclease I, the cDNAs were relinearized with UDE at the deoxyuridine in the primer and amplified by PCR with primers that append adaptors and barcodes for next-generation sequencing.

FIGURE 7.

Template-switching of group II intron RTs from 3′-overhang substrates. (A) Template-switching reactions were done with miRNAxs having different 3′-nucleotide residues (lanes A, C, G, U) and initial ³²P-labeled RNA template/DNA primer substrates (IA–P1 RNA/Pc 3′-overhang DNA) having different single nucleotide 3′ overhangs (A, C, G, T, or an equimolar mixture of all four nucleotides [N]; shown schematically below gel). Reactions were with 2 µM TeI4c-MRF RT for 10 min at 60°C in a high-salt reaction medium (450 mM NaCl, 5 mM MgCl₂, 20 mM Tris-HCl [pH 7.5], 1 mM DTT, 1 mM dNTPs), which reduces nontemplated nucleotide addition by the RT. The products were analyzed in a denaturing 20% polyacrylamide gel, which was scanned with a PhosphorImager. Numbers to left of the gel indicate positions of labeled size markers (10-bp ladder). (*) ³²P-label at the 5′ end of primer. (B) Template switching from IA–P1 RNA/Pc DNA with equimolar single-nucleotide 3′ overhangs to an miRNAx with a 3′ phosphorylated C-residue before and after dephosphorylation with T4 polynucleotide kinase (P and DP, respectively); a DNA oligonucleotide of identical sequence (miDNAx); or an miRNAx with a 2′ O-methyl group (CH₃) at its 3′ end.

FIGURE 8.

Cloning and sequencing of miRNAs by using group II intron RT template switching. Template-switching reactions were done with TeI4c-MRF RT (2 µM) to a miRNA reference set (963 equimolar miRNAs, 110 nM; Miltenyi miRXplore) from an initial IA–P1 RNA template/Pc DNA primer substrate (100 nM). The latter had single A, C, G, or T 3′-overhangs mixed at an equimolar ratio (TS1) or at 2:0.5:1:1 (TS2) to adjust the representation of miRNAs with 3′ U- or G- residues. Reactions were done as in Figure 7 and cDNAs were cloned as in Figure 6C. Parallel RNA-seq libraries were prepared from equal aliquots of the miRNAs by using either a Total RNA-Seq kit (Applied Biosystems; ABI) or a small RNA sample prep set 3 kit (New England BioLabs; NEB). These kits ligate adaptors for SOLiD sequencing to the miRNA 3′ and 5′ ends simultaneously (ABI) or sequentially (NEB) and reverse transcribe with ArrayScript or SuperScript II using a DNA primer complementary to the 3′ adaptor. (A) Plots showing counts for a subset of 898 miRNA with uniquely identifiable 16-bp core sequences (nucleotides 4 through 20) ranked from the least to most abundant, median normalized, log₂ transformed, and plotted to compare variance introduced by the library preparation method. To ensure no ambiguity of sequence mismatch across the miRNA reference panel while allowing for possible method-specific biases at the 3′ or 5′ ends, the distal sequencing adaptor sequence was concatenated to each mature miRNA sequence (the “concatenated reference”), and nucleotides 4 through 20 from each concatenated sequence were tested for occurrence within the concatenated reference anywhere in colorspace. Only concatenated sequences with no overlap to any other 16-bp core sequence were chosen for quantitation. (B) Template-switching junctions between the 3′ end of the miRNA and adaptor (IA) sequence of the 20 most frequent sequence reads from the TS1 library. (C) Venn diagrams showing overlap between under- and over-represented miRNAs in the RNA-Seq libraries prepared by the different methods. The 5% least and most abundant miRNAs in each library were identified using R and plotted using the VennDiagram R package (Chen and Boutros 2011). (D) Representation of miRNA 3′-terminal nucleotide residues in RNA-seq libraries. The bar graphs compare the percentage of miRNAs ending in each of the four bases in the miRXplore reference set (black) with the percentage of that base at the 3′ end of miRNAs in the RNA-seq libraries (TeI4c-MRF/TS1, blue; TeI4c-MRF/TS2, green; ABI Total RNA Seq, gold; NEB Small RNA Sample Prep, purple). (Left) The 3′-nucleotide residue of miRNAs in the RNA-seq libraries was identified as the base prior to the Internal Adaptor. To avoid primer-dimer, adaptor-only, and low-quality sequences, a perfect match to eight bases of the Internal Adaptor no closer than 15 bp from the start of each sequence was required when determining the terminal base in each sample. (Right) The distribution of 3′-nucleotide residues of the miRNAs in the RNA-seq libraries was inferred from the abundance-adjusted distribution of the set of 898 miRNAs identified by their unique core sequences. Similar trends were seen for both methods of identifying the 3′-terminal residue of the miRNA.