Molecular Genetics and Genomics

, Volume 286, Issue 5, pp 395–410

RNA splicing and debranching viewed through analysis of RNA lariats

Authors

  • Zhi Cheng
    • School of Biological SciencesUniversity of Missouri-Kansas City
    • The Pediatric Surgery Laboratory, Department of SurgeryCedars-Sinai Medical Center
    • School of Biological SciencesUniversity of Missouri-Kansas City
Original Paper

DOI: 10.1007/s00438-011-0635-y

Cite this article as:
Cheng, Z. & Menees, T.M. Mol Genet Genomics (2011) 286: 395. doi:10.1007/s00438-011-0635-y

Abstract

Intron lariat RNAs, created by pre-mRNA splicing, are sources of information on gene expression and structure. Although produced equivalently to their corresponding mRNAs, the vast majority of intron lariat RNAs are rapidly degraded. However, their levels are enhanced in cells deficient for RNA debranching enzyme, which catalyzes linearization of these RNAs, the rate-limiting step in their degradation. Furthermore, RNA lariats are resistant to degradation by the 3′ exonuclease polynucleotide phosphorylase (PNPase), providing a means to enrich for lariat RNAs. Working with the yeast Saccharomyces cerevisiae as a model organism, our goal was to develop novel combinations of methods to enhance the use of intron lariat RNAs as objects of study. Using RT-PCR assays developed for detecting and quantifying specific lariat RNAs, we demonstrate the resistance of RNA lariats to degradation by PNPase and their sensitivity to cleavage by RNA debranching enzyme. We also employ sequential treatments with these two enzymes to produce characteristic effects on linear and lariat RNAs. We establish the utility of the methods for analyzing RNA debranching enzyme variants and in vitro debranching reactions and discuss several possible applications, including measuring relative rates of transcription and combining these methods with non-gene-specific RNA sequencing as a novel approach for genome annotation. In summary, enzymatic treatments that produce characteristic effects on linear and lariat RNAs, combined with RT-PCR or RNA sequencing, can be powerful tools to advance studies on gene expression, alternative splicing, and any process that depends on the RNA debranching enzyme.

Keywords

Intron RNA lariatsDebranching enzyme Dbr1Polynucleotide phosphorylasemRNA splicingYeast Ty1 retrotransposon

Introduction

Pre-mRNA introns play an important role in the regulation of gene expression for many eukaryotes because their presence allows for the occurrence of alternative splicing, which results in the creation of multiple proteins from a single gene, many of which are expressed in cell- or tissue-specific patterns (Hallegger et al. 2010; Nilsen and Graveley 2010). Introns are excised in a lariat conformation (Abelson 2008; Smith et al. 2008; Wahl et al. 2009) and, following excision, the 3′ tails of the lariats undergo exonucleolytic degradation up to the lariat branch point (Chapman and Boeke 1991; Salem et al. 2003). The predominant pathway for further exonucleolytic degradation requires cleavage of the 2′–5′ bond at the branch point by RNA debranching enzyme, a 2′–5′ phosphodiesterase found in all eukaryotes (Ooi et al. 2001).

Although intron RNA sequences contain information necessary for their removal from pre-mRNAs, some introns contain additional information. In most eukaryotes microRNAs (miRNAs) (Rodriguez et al. 2004; Kim and Kim 2007; Tang and Maxwell 2008) and small nucleolar RNAs (snoRNAs) (Filipowicz and Pogacic 2002; Lestrade and Weber 2006) are encoded within introns. In studies with human cells it has been found that the vast majority of intronic miRNAs are excised from pre-mRNAs by Drosha (Kim and Kim 2007; Kataoka et al. 2009). However, mirtrons, which have been found in fruit flies, worms, and vertebrates, are miRNAs processed from introns that have been excised from pre-mRNAs by the spliceosome (Berezikov et al. 2007; Okamura et al. 2007; Ruby et al. 2007). Intronic sno-RNAs are also processed from excised introns, as determined in baker’s yeast, humans, and other eukaryotes (Ooi et al. 1998; Filipowicz and Pogacic 2002; Lestrade and Weber 2006).

Debranching and subsequent degradation of most intron RNAs are rapid, resulting in low steady state levels of intron RNAs relative to the levels of the corresponding mRNAs, as determined by studies in yeast (Chapman and Boeke 1991; Salem et al. 2003). The exceptions are intron sequences corresponding to RNAs with additional functions (e.g., snoRNAs). Studies in many different organisms have determined that cleavage of the 2′–5′ bond by RNA debranching enzyme is important for maturation of intron-encoded snoRNAs and mirtrons (Ooi et al. 1998; Kiss et al. 2006; Berezikov et al. 2007; Okamura et al. 2007; Ruby et al. 2007). For mirtrons, RNA debranching enzyme acts instead of Drosha (Berezikov et al. 2007; Okamura et al. 2007; Ruby et al. 2007).

Genome-wide studies analyzing excised intron RNAs in fruit flies and yeast have identified introns and (for flies) alternative splicing patterns (Conklin et al. 2005; Juneau et al. 2007; Zhang et al. 2007). These analyses relied on creating cell populations that accumulate excised intron RNAs at elevated levels either due to mutation of the gene encoding debranching enzyme in yeast (Juneau et al. 2007; Zhang et al. 2007) or knock down of debranching enzyme expression with siRNA in flies (Conklin et al. 2005). Cells defective for RNA debranching activity accumulate excised introns in their lariat forms, with shorted 3′ tails, as described in studies with yeast mutants (Chapman and Boeke 1991; Salem et al. 2003). Therefore, information on the 3′ intron–exon junction is not obtainable from intron lariat RNA sequences. However, unlike other methods of identifying intron sequences, studies in plants and human cells have shown that the positions of RNA branch points can also be deduced from analyzing intron RNA lariats (Vogel et al. 1997; Gao et al. 2008).

Our interest in intron RNA lariats comes from the observation that RNA debranching enzymes are host factors for retroviruses and retrovirus-like elements. The human RNA debranching enzyme, Dbr1, is a host factor for the retrovirus HIV-1, acting to promote retroviral reverse transcription (Ye et al. 2005; Bushman et al. 2009). Previously, the S. cerevisiae RNA debranching enzyme, Dbr1p, had been found to be a host factor for both the Ty1 and Ty3 retrotransposons, which are related to animal retroviruses in their genomic structures and replication cycles (Chapman and Boeke 1991; Karst et al. 2000; Griffith et al. 2003; Irwin et al. 2005). For Ty1, Dbr1p also appears to promote reverse transcription (Karst et al. 2000; Griffith et al. 2003). In fact, we have presented evidence that Ty1 RNA forms a lariat and we proposed a model in which the formation of the lariat and its later debranching facilitate both the minus strand transfer reaction and extension of the minus strand that occur during retroviral reverse transcription (Cheng and Menees 2004; Goff 2007; Engleman 2010). However, subsequent work by others has raised concerns about the interpretation of our results and our model for the role of Dbr1p in retrotransposition (Coombes and Boeke 2005; Pratico and Silverman 2007). Although siRNA knockdown of human Dbr1 interferes with production of HIV-1 cDNA at the precise stage of reverse transcription predicted by our disputed model (Ye et al. 2005; Bushman et al. 2009), there is currently no further evidence that either HIV-1 RNA or Ty1 RNA forms a lariat.

Since RNA debranching enzymes act predominantly on excised intron lariats, we are interested in developing tools for studying these lariat RNAs, which may hold the key to understand the role of these enzymes as host factors for Ty1 and HIV-1. S. cerevisiae cells lacking a functional DBR1 gene, which encodes the RNA debranching enzyme, accumulate excised intron RNA lariats (Chapman and Boeke 1991; Salem et al. 2003). Since S. cerevisiaedbr1 mutants exhibit nearly wild-type growth capabilities, the mutant cells are a robust source of excised intron lariats. Human and Drosophila cells, in which expression of RNA debranching enzyme has been knocked down with siRNA, have also been used as sources of intron RNA lariats (Conklin et al. 2005; Ye et al. 2005).

In this work we have developed methods for detecting and quantifying specific RNA lariat species and for treating RNA populations to enrich them for RNA lariats. We establish the utility of these methods by analyzing the functionality of Dbr1p mutants and to observe time courses of debranching reactions. The techniques described can be adapted to quantify transcription levels for specific genes as well as to perform genome-wide analyses to identify introns, including their branch point sequences, and analyze alternative splicing patterns.

Materials and methods

Yeast and bacterial strains, plasmids, and general procedures

The following yeast strains were used: TMY30 (MATa ura3-52 lys2-801 ade2-101 trp1-Δ63 his3-Δ200 leu2-Δ1), TMY60 (TMY30 dbr1::neor), TMY497 [=TMY30 mutated to dbr1 (D180Y allele)], TMY498 [TMY30 mutated to dbr1 (G84A allele)], TMY499 [=TMY30 mutated to dbr1 (Y68S allele)]. TMY453, a dbr1Δ::hisG version of sigma strain 10560-23C (Loeb et al. 1999), was used for FLO8 RT-PCR experiments (sigma strain 10560-23C = MATalpha ura3-52 his3::hisG leu2::hisG). The dbr1Δ::hisG allele was created using pTM513, a DBR1 gene blaster plasmid (Alani et al. 1987) containing dbr1Δ::hisG-URA3-hisG, and targeted to replace DBR1 chromosomal sequences by digestion with PvuII.

The following Escherichia coli strains were used: Rosetta DE3 (Novagen) [FompT hsdSB(rBmB) gal dcm (DE3) pLysSRARE (CamR)]; XL1 Blue [F′::Tn10 proA+B+lacIq Δ(lacZ)M15/recA1 endA1 gyrA96 (NaIr) thi hsdR17 (rkmk+) supE44 relA1 lac]; JM109 [F′ traD36 lacIqΔ(lacZ)M15 proA+B+/e14(McrA) Δ(lac-proAB) thi gyrA96 (NaIr) endA1 hsdR17 (rkmk+) relA1 supE44 recA1]; ES1301 [lacZ53 thyA36 rha-5 metB1 deoC IN(rrnD-rrnE) mutS201::Tn5]; and TOP10 [F-mcrA Δ(mrr-hsdRMS-mcrBC) Φ80lacZΔM15 ΔlacX74 recA1 deoR araD139 D(ara-leu)7697 galU galK rpsL (StrR) endA1 nupG].

The following plasmids were used for this study: pET16b-DBR1 was a gift of Beate Schwer and was used to express Dbr1p in E. coli (Martin et al. 2002). pRS306 (Sikorski and Hieter 1989) was used as a URA3 template for making a PCR fragment to create a dbr1Δ::URA3 allele at the DBR1 locus. YEp351 (LEU2) (Hill et al. 1986) was used in cotransformations with the PCR fragment that resulted in the creation of a dbr1Δ::URA3 strain. This strain was an intermediate in the creation of dbr1 point mutants (see section below on the creation of dbr1 mutant strains). pTM431, pTM432, and pTM435 were all created by random mutagenesis of pYES2/GS-DBR1 from Invitrogen and encode Dbr1p D180Y, Dbr1p G84A, and Dbr1p Y68S, respectively (Salem et al. 2003) (Salem and Menees, unpublished data). The DBR1 gene blaster plasmid pTM513 was created in three steps. First, the 3.8 kbp BamHI-BglII fragment from pNKY51 (Alani et al. 1987), containing hisG-URA3-hisG, was ligated into the BamHI site of pBluescript (Invitrogen) to create pTM509. Second, the 5′ UTR of DBR1 was amplified from genomic DNA using oligonucleotide primers 331 and 332, then the PCR product was trimmed with EcoRI and BamHI and ligated into EcoRI and BamHI sites of pTM509 to create pTM511. Third, the 3′ UTR of DBR1 was amplified from genomic DNA using oligonucleotide primers 333 and 336, then the PCR product was trimmed with XbaI and NotI and ligated into XbaI and NotI sites of pTM511 to create pTM513.

When not specifically described, general molecular techniques (Ausubel et al. 2003) as well as standard yeast media and general procedures (Kaiser et al. 1994) were used. Oligonucleotides are listed in Tables 1 and 2.
Table 1

Oligonucleotides

Primer

Sequence

Positiona

146

cactctcccataacctccta

ACT1 intron nt 100–119

215

ctcaaaccaagaagaaaaagaa

ACT1 nt −128 to −107

216

tgataccttggtgtcttggtct

ACT1 nt 130–109

331

aggatgtttccgtctttagaa

−761 to −741 upstream of DBR1 ORF

332

gaggatcctgataaatgtctgcccatctt

−10 to −30 upstream of DBR1 ORF; EcoRI site added at 5′ end

333

gctctagaacgaatgcagacggaattaga

16–30 after DBR1 stop codon; XbaI site added at 5′ end

336

ataagaatgcggccgcaaagggatccaatgtggtga

779–760 after DBR1 stop codon; NotI site added at 5′ end

363

gcaagcgctagaacatacttag

ACT1 intron nt 18–1, 265–262

372

agtgaatagttcgtatccagattc

FLO8 nt 12–35

373

catacaaaaagccttgaggtg

FLO8 nt 418–398

374

ggtagcaaatattctgggacatct

FLO8 nt 422–445

375

attctgggttggccctacattt

FLO8 nt 837–816

376

agtcaaaacgttactggctgg

FLO8 nt 841–861

377

tgcttgattgcggaagttag

FLO8 nt 1260–1241

378

ttggcgaggaagatatttattc

FLO8 nt 1268–1289

379

aagataatggactggatacagccg

FLO8 nt 1675–1652

380

ttcgatccagaaagtggcaa

FLO8 nt 1693–1712

381

ttttcctctggagtagataatgtg

FLO8 nt 2036–2013

382

atcaaggatatgattttgacgc

FLO8 nt 2054–2075

383

cagccttcccaattaataaaattg

FLO8 nt 2399–2376

408

taaatagcttggcagcaacagg

URA3 nt 67–46

417

ttgcgaattgctgtacaagg

DBR1 nt 10–29

418

caagtcatgaatttagagataaatgc

DBR1 nt 1217–1192

443

tgctgtcatggtcagctaaaccaaatttataaagaagtgt…

5′ 40 nt = DBR1 nt 31–70

…taactatgcggcatcagagc

3′ 20 nt = URA3 flank in pRS306

444

gataaatgctttagtttgtcgtacttcatctttctgaata…

5′ 40 nt = DBR1 nt 1200–1161

…cctgatgcggtattttctcc

3′ 20 nt = URA3 flank in pRS306

aFor the ACT1, FLO8, URA3 and DBR1 genes, the nucleotide positions are relative to the first nucleotide of the coding sequence, except for the ACT1 intron, where positions are relative to the first nucleotide of the intron

Table 2

Primers and probes for qRT-PCR

Target and primers/probe

Sequence

Positiona

ACT1 mRNA

 FWD primer

TCCCAAGATCGAAAATTTACTGAAT

−30 to −6

 REV primer

TTTACACATACCAGAACCGTTATCAAT

54 to 28

 TaqMan probe

VIC-TGAATTAACAAGGTTGCTGCT-MGBNFQ

−4 to –26

ACT1 intron

 FWD primer

ATTTTTCACTCTCCCATAACCTCCTATA

94 to 121

 REV primer

TTTCAAGCCCCTATTTATTCCAAT

173 to 150

 TaqMan probe

6FAM-TGACTGATCTGTAATAACCA-MGBNFQ

123 to 142

RPP1B mRNA

 FWD primer

AGGCCGCTGGTGCTAATG

89 to 106

 REV primer

TCCAAAGCCTTAGCGTAAACATC

146 to 124

 TaqMan probe

VIC-CGACAACGTCTGGGC-MGBNFQ

108 to 122

RPP1B intron

 FWD primer

AATGCAACCTAAAACGACTTTGTG

12 to 35

 REV primer

TTTCTCGGGACGATTGTTGTC

77 to 57

 TaqMan probe

6FAM-ACTACGAAGAGAAAGATT-MGBNFQ

38 to 55

YRA1 mRNA

 FWD primer

AGGTTTGCCAAGGGACATTAAG

249 to 270

 REV primer

ACACCACCTACTTGAGATGCAAAA

314 to 291

 TaqMan probe

VIC-AGGATGCTGTAAGAGAAT-MGBNFQ

272 to 289

YRA1 intron

 FWD primer

CGCATCGTCTCGTGTGGAT

42 to 60

 REV primer

GATCAAAAGCGTGTGCCATATC

107 to 86

 TaqMan probe

6FAM-CGAGAAATATTCTTTGTAAGGAA-MGBNFQ

62 to 84

The primers and probe for human GAPDH qPCR were designed by Applied Biosystems and obtained from the company

aRelative to start of coding sequence for mRNA primers and probes. Relative to start of intron sequence for intron primers and probes

RNA extraction

Yeast strains were grown to mid-logarithmic phase prior to isolating total cellular RNA. In some cases yeast cells were used directly for RNA preparation after cell growth was complete. In other cases yeast cells were pelleted and flash frozen in a dry ice ethanol bath and stored at −80°C prior to RNA preparation. We found no difference in results for RNAs prepared from cells processed in these two ways. Total yeast RNA was prepared by the hot acid phenol method (Ausubel et al. 2003) or by a column purification method (RNeasy kit, Qiagen) from small cultures (10 ml) grown to mid-logarithmic phase (OD600 = ~1). RNA samples were treated with RNase free DNase I (Fisher) to remove DNA contamination. RNA concentration was measured spectrophotometrically by reading OD260. The OD260/OD280 ratio was used as an RNA quality assessment.

Preparation of Dbr1p enzyme from E. coli

The pET16b-DBR1 expression plasmid encodes yeast Dbr1p as a N-terminal 10×-histidine-tagged protein (Martin et al. 2002). Expression and purification of the histidine-tagged Dbr1p were performed mostly as described previously (Martin et al. 2002). Rosetta DE3 E. coli cells (Novagen) were used for expression of Dbr1p instead of E. coli strain BL21-Codon Plus(DE3)RIL (Stratagene). Sonication of cells was performed on ice for 60 s, in 1 s pulses, with a large probe at 50% power. Triton X-100 was added after sonication to a final concentration of 0.1%. The tagged Dbr1p was purified from E. coli extracts by binding to and eluting from Nickel–nitrilotriacetic acid-agarose columns, as described previously (Martin et al. 2002), and fractions were assessed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Peak fractions from the elution were dialyzed against debranching buffer (Ooi et al. 2001) (20 mM HEPES KOH, pH 7.9; 125 mM KCl; 0.5 mM MgCl2; 1 mM DTT; 10% glycerol). In some cases, Dbr1p was concentrated by spinning through a Microcon YM-30 spin concentrator at 14,000×g for 40 min at 4°C in a Beckman Allegra 25R centrifuge (TA-15-1.5 rotor). The concentrations of Dbr1p preparations were 50–100 ng/μl.

Mass spectrometry of purified Dbr1p was performed at the UMKC School of Biological Sciences Proteomics Facility.

Enzymatic treatments of RNA

Bacillus stearothermophilus PNPase was acquired from Sigma (St. Louis, Missouri) and a stock of 3.5 units/ml was prepared by dissolving the protein in water, then adding Tris HCl, pH 8.5, to a final concentration of 50 mM. PNPase reactions were performed in PNPase buffer (50 mM Tris HCl, pH 8.5; 1 mM 2-mercaptoethanol; 1 mM EDTA; 20 mM KCl; 1.5 mM MgCl2; 10 mM Na2HPO4, pH 8.3) on 20–1,000 ng of total yeast RNA in 20 μl reactions for 1 h at 60°C, using 1 μl of the PNPase stock. Upon completion of reactions, samples were heated to 85°C for 10 min, then either used directly in RT-PCRs or ethanol precipitated. Mock treatments were performed in the same way, minus PNPase.

Approximately 50–100 ng of yeast Dbr1p prepared from E. coli was used for in vitro debranching reactions of 20–200 ng of RNA. Reactions were performed at 30°C for 45 min in a 20 μl volume containing 1× debranching buffer (20 mM HEPES–KOH pH 7.9, 125 mM KCl, 0.5 mM MgCl2, 1 mM DTT and 10% glycerol). Reactions were stopped by heating at 65°C for 10 min. Mock treatments were performed in the same way, minus Dbr1p.

For sequential enzymatic treatments, RNA samples were phenol/chloroform extracted and ethanol precipitated after the first treatment (PNPase or Dbr1p) then resuspended and treated with the second enzyme.

RT-PCR methods

RT-PCRs of lariat and linear RNAs were performed with QIAGEN one-step RT-PCR kit (Valencia, CA) under the following general conditions: 50°C, 30 min; 95°C, 15 min; nine cycles of 94°C for 30 s, 54°C for 30–60 s [touchdown to 46°C (−1°C per cycle)], 72°C for 30 s; X cycles (see below) of 94°C for 30 s, 46°C for 30 s, 72°C for 30–45 s; 72°C for 5–10 min; 4°C hold. The number of cycles in the post-touchdown phase of different RT-PCRs (X cycles above) varied with the experiment and are reflected in the following reaction profile names: ACT1-1, 29 cycles, post-touchdown; ACT1-2, 24 cycles, post-touchdown; ACT1-3, 19 cycles, post-touchdown; ACT1-4, 15 cycles, post-touchdown; and ACT1-5, 11 cycles, post-touchdown. RNA amounts between 2 and 50 ng were used in RT-PCRs. RT-PCRs were analyzed by either PAGE or agarose gel electrophoresis.

Real-time RT-PCR (qRT-PCR) of lariat and linear RNAs

Primers and probes for qPCR were designed using Sequence Detection Systems software from Applied Biosystems and are listed in Table 2. All probes and primers for qRT-PCR were purchased from Applied Biosystems. Validation experiments were performed that demonstrated that the efficiencies of target and reference PCRs were approximately equal.

For total RNA samples (untreated or treated with Dbr1p/PNPase, as described above), 20–1,000 ng of RNA was reverse transcribed into cDNA using random hexamers in a 100 μl reaction at 45°C for 60 min.

PCR MasterMix reagents from Applied Biosystems were used for qPCR reactions, which were performed in triplicate for each sample. Reactions were prepared and run according to a standard protocol established by Applied Biosystems on an ABI 7500 real-time PCR machine. Briefly, reactions contained 2× PCR MasterMix, 900 nM forward primer, 900 nM reverse primer, 250 nM TaqMan probe, and cDNA (~20 ng). Reactions were incubated for 2 min at 50°C and then 10 min at 95°C and before proceeding through 40 cycles of a 30 s incubation at 95°C and a 60 s incubation at 60°C. Completed reactions were held at 4°C.

Relative quantification (RQ) of results was performed using the comparative CT method (ΔΔCT) (Schmittgen and Livak 2008). The amplification of each target intron sequence was compared to amplification of the corresponding mRNA sequence and a ΔCT was determined. To compare the different samples to each other, the wild-type sample was used as the calibrator sample. Therefore, the ΔCT of the wild-type sample was subtracted from the ΔCT for each sample to determine −ΔΔCT values. In Fig. 7, RQ = \( 2^{{ - \Updelta \Updelta C_{\text{T}} }} \) for each −ΔΔCT and represents the fold-difference in intron levels between a given sample and the wild-type sample (DBR1).

In vitro debranching time course

Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) cDNA, the exogenous control for qPCR in these experiments, was generated by reverse transcribing 600 ng of human RNA at 45°C for 1 h using the RT kit from Applied Biosystems. A debranching reaction mix was set up on ice and contained 5,600 ng of total RNA from TMY60 (dbr1) cells, ~6 ng GAPDH cDNA, 140 μl of purified Dbr1p, and 350 μl 2× debranching buffer in a final volume of 700 μl. Seven 100 μl aliquots of this mix were distributed to 0.2 ml PCR tubes. The debranching reaction was directly inactivated in one tube (0 min reaction time) by raising the temperature to 95°C, followed by phenol/chloroform extraction and ethanol precipitation. The remaining six tubes were incubated at 30°C and individual reactions were stopped after 2.5, 5, 10, 15, 30, and 60 min. Reactions were stopped by raising the temperature to 95°C, followed by phenol/chloroform extraction and ethanol precipitation. RNAs were then treated with PNPase, as described above, to degrade intron lariats linearized by Dbr1p. Reverse transcription of the RNAs remaining from the different debranching reactions was performed using the RT kit from Applied Biosystems and random hexamer primers. qPCRs using these cDNAs were performed as described above, amplifying a volume of cDNA roughly corresponding to ~20 ng of starting total RNA, using primers and probes for yeast ACT1, YRA1, and RPP1B introns as well as human GAPDH. GAPDH cDNA was the exogenous control because it is insensitive to PNPase and remained at a constant level in each reaction.

Creation of dbr1 point mutant strains

Mutants were created using modifications of the delitto perfetto method (Storici et al. 2001) and the site specific genomic (SSG) method (Gray et al. 2004). Initially, a dbr1Δ::URA3 strain was created to facilitate the introduction of point mutant alleles of dbr1 into the DBR1 locus. Yeast strain TMY490, containing a URA3-marked deletion of 1,090 bp of the 1,215 bp DBR1 coding sequence (nts 71–1,160 deleted), was constructed by transformation of TMY30 with a PCR fragment containing the URA3 gene from pRS306 flanked by ends corresponding to 5′ and 3′ segments of the DBR1 coding region.

The fragment used for making the dbr1Δ::URA3 allele was created by PCR of pRS306 with oligonucleotides 443 and 444, the 3′ 20 nt of which anneal to the ends of the URA3 gene on pRS306 and the 5′ 40 nt of which correspond to DBR1 sequences (see Table 1).

The dbr1Δ::URA3 disruption on yeast chromosome XI was created by homologous recombination between the DBR1 locus and the dbr1Δ::URA3 PCR fragment. Briefly, TMY30 was transformed with the dbr1Δ::URA3 PCR fragment and transformants were selected on SD-Uracil plates. Transformants were screened by PCR with primer pairs 401/402, which anneal within the DBR1 sequences that are deleted in the dbr1Δ::URA3 allele, and 417/418, which anneal outside the DBR1 sequences that are deleted in the dbr1Δ::URA3 allele. Transformants containing the dbr1Δ::URA3 allele template a 417/418 PCR product but not a 401/402 PCR product. DNA sequencing of PCR products was performed to verify the presence of the dbr1Δ::URA3 allele.

Replacement of the chromosomal dbr1Δ::URA3 allele with dbr1 point mutations was accomplished by transformation. TMY490 (dbr1Δ::URA3 strain) was co-transformed with YEp351(LEU2) and PCR fragments of dbr1 point mutants. The PCR fragments were generated from plasmids pTM431, pTM432, and pTM435 with PCR primer pairs 417/418. Transformants (with YEp351) were selected in SD-leucine liquid media during a 48 h incubation period at 30°C (with shaking). After this selection period, cells were spread onto 5-fluoroorotic acid plates to select for cells that lost function of the URA3 gene within the DBR1 locus. Recombinants within the FOAr population that have replaced the dbr1Δ::URA3 allele with a dbr1 point mutant allele were identified by PCR screening. Positive clones were identified as those that template a 417/418 PCR product but not a 417/408 PCR product (specific for the dbr1Δ::URA3 allele). DNA sequencing of PCR products was performed to verify the presence of a dbr1 point mutant allele.

Results

RT-PCR detection of lariat RNAs

S. cerevisiae ACT1, which encodes actin, is a robustly expressed gene that contains an intron of 308 nt. The first example of a spliceosomal intron discovered in yeast (Gallwitz and Sures 1980; Ng and Abelson 1980), the ACT1 intron contains all the canonical features of yeast introns and is efficiently spliced from pre-mRNA, producing an excised lariat with a 265 nt circle. We chose this well-characterized gene to assess intron levels as we developed and tested tools for detecting and enriching excised intron lariats. Primers were designed for use in RT-PCR to detect the lariat form of the ACT1 intron RNA and, as a control, ACT1 mRNA (Fig. 1a). RT-PCR of total yeast RNA using primers that flank the ACT1 exon–exon junction (primers 215 and 216) amplifies a 285 bp product from ACT1 mRNA. Primer 363 spans the ACT1 intron lariat branch point and is used in combination with primer 146, which anneals to sequences complementary to the ACT1 intron upstream of the lariat branch point, in an RT-PCR that amplifies a 184 bp product from the lariat form of the ACT1 intron RNA. As expected, when RT-PCRs are performed using total RNA samples from wild-type (TMY30) and dbr1 mutant yeast cells (TMY60), the amounts of ACT1 mRNA products are similar when using equivalent amounts of RNA from the two cell types (Fig. 1b, lanes 1 and 3 as well as lanes 5 and 7). However, the ACT1 intron RNA lariat product is much more readily produced from dbr1 cells, also as expected (Fig. 1b, lane 4 vs. 2 and lane 8 vs. 6). These data clearly show that a dbr1 mutant strain or, where appropriate, a Dbr1p knock-down strain contains a rich source of expressed intron sequences. It is also evident that the use of intron-specific RT-PCR could be used to detect excised introns from genes expressed at very low levels. For studies on alternative splicing, the use of RT-PCR on RNA from Dbr1p-deficient cells can allow detection of rare splice variants, an approach that has already been exploited (Conklin et al. 2005).
https://static-content.springer.com/image/art%3A10.1007%2Fs00438-011-0635-y/MediaObjects/438_2011_635_Fig1_HTML.gif
Fig. 1

RT-PCR detection of lariat RNAs. a Annealing positions of primers for RT-PCR detection of ACT1 intron lariat RNA and mRNA. b Agarose gel analysis of RT-PCRs for ACT1 RNAs. Lanes14 contain reactions run 15 cycles after the touchdown phase of PCR; lanes58 contain reactions run 11 cycles after the touchdown phase of PCR. The different numbers of cycles were run to show the linearity of the PCRs. Primer pairs for PCRs are indicated on right. Reactions using wild type (DBR1) RNA samples are in lanes1, 2, 5, and 6; reactions using dbr1 mutant RNA samples are in lanes3, 4, 7, and 8

A previous report described the use of radiolabeled primers spanning intron RNA branch points for analyzing intron populations by primer extension (Spingola et al. 1999). The RT-PCR method we describe could be modified to survey intron lariats containing specific sequences at intron 5′ ends and branch points, as described previously (Spingola et al. 1999). RT-PCR has added utility because the products can be cloned and sequenced to identify the individual introns represented in a lariat population.

Insensitivity of lariat RNAs to the 3′ exonuclease PNPase

Linear and lariat RNAs have different sensitivities to 3′ exonucleases, including PNPase, a component of bacterial RNA degradation systems (Suzuki et al. 2006). PNPase degrades linear RNAs but does not proceed past the 2′ branch present in intron RNA lariats (Suzuki et al. 2006). Therefore, treatment of RNA samples with PNPase should result in a vast enrichment of excised intron lariats in the RNA that remains intact after treatment. This difference should be evident in the results of the RT-PCR assay described above when amplifying PNPase-treated RNA samples. Since RNA secondary structures reduce the efficiency of PNPases (Guarneros and Portier 1990; McLaren et al. 1991), reactions were performed at elevated temperature (60°C) using PNPase from B. strearothermophilus to circumvent this problem. Total RNA samples from a dbr1 mutant strain (TMY60) were treated with a range of PNPase concentrations and then subjected to RT-PCR to detect ACT1 intron RNA lariats as well as the linear mRNA (Fig. 2). Results are consistent with expectations, as observed previously (Suzuki et al. 2006) the use of PNPase selectively preserves RNA lariats.
https://static-content.springer.com/image/art%3A10.1007%2Fs00438-011-0635-y/MediaObjects/438_2011_635_Fig2_HTML.gif
Fig. 2

PNPase degrades linear RNAs but not lariat RNAs. Upper panel (lanes17). Agarose gel analysis of RT-PCRs for ACT1 intron lariat RNA following a series of PNPase treatments. PNPase reactions were performed with diminishing concentrations of enzyme as indicated by the wedge at the top of the gel image (RT-PCR of highest concentration PNPase reaction is in lane1; RT-PCR of lowest concentration PNPase reaction is in lane7). RT-PCRs for ACT1 intron lariat RNA were performed with primers 146 and 363 (see Fig. 1) and run for 15 cycles after the touchdown phase of the reaction. Lower panel (lanes 814). Agarose gel analysis of RT-PCRs for ACT1 mRNA (linear) from the same series of PNPase treatments described above. PNPase reactions were performed with diminishing concentrations of enzyme as described for upper panel. RT-PCRs for ACT1 mRNA were performed with primers 215 and 216 (see Fig. 1) and run for 24 cycles after the touchdown phase of the reaction. Lanes15 and 16 contain ACT1 intron lariat and ACT1 mRNA RT-PCRs, respectively, of RNA samples that did not undergo PNPase treatment. The products in these lanes serve as size markers for the intron lariat and mRNA products in lanes114

The high-temperature reaction using PNPase from a thermophile appears to be much more efficient than the reported reaction with the E. coli PNPase at 37°C (Suzuki et al. 2006). To eliminate the RT-PCR product from the ACT1 mRNA, PNPase must degrade, at the very least, the RNA corresponding to the binding site for the downstream primer (oligonucleotide 216). To accomplish this, PNPase must degrade all the RNA that lies to the 3′ side of the oligonucleotide 216 binding site, which includes 998 nt of the ACT1 coding sequence plus the 3′ UTR and the polyA tail. To further examine the processivity of B. strearothermophilus PNPase, the degradation of FLO8 mRNA was assessed. FLO8 mRNA is >2.4 kb in length. Primer pairs were designed to amplify different portions of this mRNA along its length (Fig. 3a). Total nucleic acid samples and RNA samples (DNased total nucleic acid samples) were treated with PNPase and subjected to RT-PCR to detect the various segments of FLO8. As shown in Fig. 3b, PNPase readily degrades every segment of FLO8 mRNA assayed. As expected, PNPase has no effect on FLO8 DNA present in the total nucleic acid samples (Fig. 3c). Other enzymes reported to work as well as B. strearothermophilus PNPase in our studies are Thermus thermophilus PNPase at 65°C (Falaleeva et al. 2008) and E. coli RNase R at 37°C (Suzuki et al. 2006; Vincent and Deutscher 2006), although direct comparisons have not been made.
https://static-content.springer.com/image/art%3A10.1007%2Fs00438-011-0635-y/MediaObjects/438_2011_635_Fig3_HTML.gif
Fig. 3

Processivity of PNPase on FLO8 mRNA. a Primer pairs for amplifying different segments along the length of FLO8 mRNA: 1/2 = primers 372 and 373; 3/4 = primers 374 and 375; 5/6 = primers 376 and 377; 7/8 = primers 378 and 379; 9/10 = primers 380 and 381; 11/12 = primers 382 and 383. Primers are listed in Table 1. b PAGE analysis of RT-PCRs for FLO8 mRNA segments following PNPase treatment (+ lanes) and mock treatment (− lanes) of a total cellular RNA sample that had been pretreated with DNase I. Lanes containing the various FLO8 RT-PCRs are indicated below the gel image; the FLO8 primer pairs are indicated above the gel image. RT-PCRs for ACT1 RNAs are in the four lanes under the ACT1 title and serve as controls that indicate the PNPase reactions proceeded as expected. The RT-PCRs for ACT1 mRNA and intron RNA are indicated below the corresponding lanes; these reactions used primer pairs 215/216 and 146/363, respectively. The lane marked “M” contains a DNA molecular weight standard (50 bp ladder). c PAGE analysis of RT-PCRs as described for b except that total cellular nucleic acid samples were not treated with DNase I prior to PNPase treatment and RT-PCRs. For all RT-PCRs in b and c, reactions were performed with 24 cycles after the touchdown phase

Sensitivity of lariat RNAs to Dbr1p

Linear and lariat RNAs also have different sensitivities to RNA debranching enzyme, which can be exploited to confirm that an RNA species has a lariat conformation. The RT-PCR strategy employing a primer that spans a lariat branch point, as described above for the ACT1 intron, can be used to demonstrate the cleavage of the 2′–5′ bond. This is due to the fact that after Dbr1p treatment the binding site for the primer that spans the ACT1 intron branch point (oligonucleotide 363) is split into two non-contiguous sections, with the section that anneals to the 3′ end of the primer being only 3 bp in length. After debranching of the lariat, the critical 3′ end of the primer will not effectively anneal to the intron RNA to prime RT-PCR. Dbr1p treatment has no effect on ACT1 mRNA, which should still be readily detected by RT-PCR.

To perform Dbr1p treatments, S. cerevisiae Dbr1p was expressed in E. coli and purified by metal affinity chromatography (Fig. 4a). Although histidine-tagged Dbr1p is expected to have a mass of ~50 kDa, the mobility of the main product in SDS-PAGE is ~45 kDa. Others have observed this anomalous mobility for histidine-tagged Dbr1p (Ooi et al. 2001) and have speculated that the protein may undergo limited proteolysis in E. coli. However, mass spectrometric analysis of the main band in the stained gel shows it to be the expected molecular mass of the histidine-tagged Dbr1p (50,062 Da) (Fig. 4b), indicating that the protein is intact and must run anomalously in SDS-PAGE because of its physical properties.
https://static-content.springer.com/image/art%3A10.1007%2Fs00438-011-0635-y/MediaObjects/438_2011_635_Fig4_HTML.gif
Fig. 4

Purification of Dbr1p. a Elution profile of histidine-tagged yeast Dbr1p purified from E. coli. Dbr1p bound to a nickel–nitrilotriacetic acid (nickel–NTA) column was eluted with increasing concentrations of imidazole (indicated at top). Six, ~1.5 ml fractions were collected for each imidazole concentration. Fraction numbers are listed above each lane. Protein molecular weight standards (Precision Plus Protein Kaleidoscope standards, Bio-Rad) were run in lanes indicated by “M.” Dbr1p runs between the 50 and the 37 kDa standards. b Matrix-assisted laser desorption/ionization-time-of-flight (MALDI-TOF) mass spectrometry analysis to assess the molecular mass of the main elution product in fractions 2–6 of the 100 mM imidazole elution

Using the Dbr1p enzyme preparation, debranching reactions were carried out on total RNA samples from a dbr1 mutant strain. RT-PCR analysis reflects the differential sensitivity of linear and lariat RNAs to Dbr1p. After Dbr1p treatment, RT-PCR detection of ACT1 RNA lariat is greatly decreased (Fig. 5, lane 4 vs. 2). On the other hand, the product indicative of ACT1 linear mRNA is still readily detectable after Dbr1p treatment (Fig. 5, lane 3 vs. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs00438-011-0635-y/MediaObjects/438_2011_635_Fig5_HTML.gif
Fig. 5

In vitro debranching reaction. Agarose gel analysis of RT-PCRs for ACT1 RNAs following treatment with Dbr1p (+ lanes) and mock treatment (− lanes) of a total cellular RNA sample. Lanes2 and 4 contain RT-PCRs for ACT1 intron lariat RNA; lanes1 and 3 are RT-PCRs for ACT1 mRNA. The lane marked “M” contains a DNA molecular weight standard (50 bp ladder). RT-PCRs for ACT1 intron lariat RNA were run for 19 cycles after the touchdown phase of the reaction; RT-PCRs for ACT1 mRNA were run for 24 cycles after the touchdown phase of the reaction

Combinations of PNPase and Dbr1p treatments

PNPase and Dbr1p treatments can be used in combination when exploring the properties of a particular RNA species. Sequential enzymatic treatments can also be used to enrich for RNA lariats and then linearize them for further manipulations. To demonstrate this, ACT1 RNA species present within a total RNA sample from a dbr1 mutant strain were analyzed by RT-PCR following sequential PNPase and Dbr1p treatments. As shown in Fig. 6a, lanes 1–4, initial treatment of the RNA sample with PNPase degrades the linear mRNA (lanes 1 and 3), but leaves lariat RNA intact (lane 2). Subsequent treatment with Dbr1p shows that the resistant RNA is a lariat (lane 4). As shown in Fig. 6a, lanes 5–8, skipping the initial PNPase treatment leaves the linear mRNA intact (lanes 5 and 7) as well as the lariat RNA (lane 6). The lariat RNA is then distinguished by its sensitivity to cutting with Dbr1p (lane 8). The order of the PNPase and Dbr1p reactions can be switched to generate a complementary set of predictable results (Fig. 6b).
https://static-content.springer.com/image/art%3A10.1007%2Fs00438-011-0635-y/MediaObjects/438_2011_635_Fig6_HTML.gif
Fig. 6

Combinations of PNPase and Dbr1p treatments. a Agarose gel analysis of RT-PCRs for ACT1 RNAs following treatment of a total cellular RNA sample from a dbr1 strain with Dbr1p (+ Dbr1p) and PNPase (+ PNPase) as well as mock treatment (− treatment). In this experiment, PNPase treatment preceded Dbr1p treatment for samples that were treated with both enzymes. Lanes1, 3, 5, and 7 contain RT-PCRs for ACT1 mRNA of a total cellular RNA sample. Lanes2, 4, 6, and 8 contain parallel RT-PCRs for ACT1 intron lariat RNA. b Agarose gel analysis of RT-PCRs for ACT1 RNAs following treatment of a total cellular RNA sample from a dbr1 strain with Dbr1p and PNPase as well as mock treatment. In this experiment, Dbr1p treatment preceded PNPase treatment for samples that were treated with both enzymes. Lanes14 contain RT-PCRs for ACT1 mRNA of a total cellular RNA sample. Lanes58 contain parallel RT-PCRs for ACT1 intron lariat RNA. For both a and b, RT-PCRs for ACT1 intron lariat RNA were run for 19 cycles after the touchdown phase of the reaction and RT-PCRs for ACT1 mRNA were run for 24 cycles after the touchdown phase of the reaction. The lanes marked “M” and “m” contain DNA molecular weight standards (“M” = λ phage DNA cut with HinDIII + EcoRI; “m” = 50 bp ladder)

Real-time RT-PCR measurement of lariat RNA levels

A real-time RT-PCR method (qRT-PCR), using the TaqMan detection system (Applied Biosystems), was developed to quantitatively compare the intron RNA lariat levels of different samples. We expanded our study to include not only the ACT1 intron but also the YRA1 and RPP1B introns to investigate the generality of the methods. YRA1 encodes an RNA binding protein involved in mRNA export from the nucleus (http://www.yeastgenome.org) and is moderately expressed, although less than ACT1 (Holstege et al. 1998) (2005 update at http://www.web.wi.mit.edu/young/pub/holstege.html). The YRA1 intron is 765 nt in length, which is larger than the 300 nt average for yeast introns, and contains a non-canonical branch point sequence (http://www.compbio.soe.ucsc.edu/yeast_introns.html; http://www.embl.de/ExternalInfo/seraphin/yidb.html) (Lopez and Seraphin 2000). Furthermore, the intron is inefficiently spliced from pre-mRNA, which is important for the autoregulation of Yra1p protein levels (Preker et al. 2002; Rodriguez-Navarro et al. 2002; Preker and Guthrie 2006). RPP1B encodes a ribosomal protein and is even more highly expressed than ACT1 (Holstege et al. 1998) (2005 update at http://www.web.wi.mit.edu/young/pub/holstege.html). The RPP1B intron is typical for yeast, 301 nt in length, with canonical sequences (http://www.compbio.soe.ucsc.edu/yeast_introns.html; http://www.embl.de/ExternalInfo/seraphin/yidb.html) (Lopez and Seraphin 2000).

Initially we used a strategy like the one used for RT-PCR of ACT1 intron lariats described above, with one primer spanning the lariat branch point and serving as both the RT primer and the reverse primer for PCR. However, we switched to using random primers for the RT step to allow amplification of the different target sequences from a common pool of cDNA. Consequently, both PCR primers anneal upstream of the branch point for each target gene, with a TaqMan probe annealing between them (see Fig. 7a). Since these types of primers will also prime amplification of genomic DNA we ran control PCRs for each sample without a prior RT step to ensure that DNA contamination was not contributing to the PCR product. The mRNA for each target gene served as the endogenous control for qRT-PCR (Fig. 7a, top). Using this strategy, intron sequences for ACT1, RPP1B, and YRA1 were amplified from dbr1 and wild-type yeast strains (TMY60 and TMY30). As shown in Fig. 7b–d [DBR1 (wild type) vs. dbr1 null mutant], the real-time method generated the expected results: the different intron RNAs accumulate at higher levels in the dbr1 null mutant strain than in wild type.
https://static-content.springer.com/image/art%3A10.1007%2Fs00438-011-0635-y/MediaObjects/438_2011_635_Fig7_HTML.gif
Fig. 7

Real-time RT-PCR (qRT-PCR) measurement of lariat RNA levels. a Annealing positions of primers for RT-PCR detection of mRNA and intron lariat RNA species. FWDm and REVm primers are designed to amplify the segment of mRNA spanning an exon–exon junction (indicated by vertical line). A TaqMan probe is designed to span the same exon–exon junction. The star and the triangle at opposite ends of the TaqMan probes represent the fluorescent reporter molecule and the quencher that are bound to the 5′ and 3′ ends, respectively. The TaqMan probes that anneal to a particular mRNA and lariat RNA pair contain different fluorescent reporter molecules, indicated by solid and stippled stars. Note that lariat RNA detection does not involve annealing of PCR primers or TaqMan probes across lariat branch points. b Relative quantification of ACT1 intron lariat RNA in total RNA samples from different yeast strains. The allele at the DBR1 locus for each strain is indicated beneath the bar graph. RQ, the relative quantification, is the ratio of intron RNA to mRNA for a particular sample relative to the ratio of intron RNA to mRNA for the DBR1 (wild-type) sample at the left end of the bar graph (which sets the RQ for DBR1 itself to 1). Quantification experiments were repeated three times and the qPCRs were performed in triplicate each time. The standard error bars display the calculated maximum (RQMax) and minimum (RQMin) expression levels that represent standard error of the mean expression level (RQ value), as described in the Applied Biosystems 7300/7500 product literature. c Relative quantification of RPP1B intron lariat RNA for the same RNA samples presented in b. d Relative quantification of YRA1 intron lariat RNA for the same RNA samples presented in b

qRT-PCR was also used to analyze mutant variants of Dbr1p. Previously, a set of point mutants had been created by random PCR mutagenesis and analyzed for intron RNA levels by an RNase protection assay (Salem et al. 2003) (Salem and Menees, unpublished data). In those experiments the dbr1 mutant alleles were under the control of a strong, inducible promoter (pGAL1) and carried on a high copy plasmid. The yeast strain carried a dbr1Δ mutation [open reading frame (ORF) deletion] at the DBR1 locus so the plasmid-borne dbr1 mutant alleles were the only sources of Dbr1p. For the current study, three dbr1 point mutants (D180Y, G84A, and Y68S) were analyzed by qRT-PCR to determine their levels of intron lariat RNA relative to wild-type (DBR1) and dbr1Δ. To make the analysis more biologically relevant, each of the dbr1 mutant alleles was placed at the DBR1 locus, replacing the wild-type allele, and was under the control of the native DBR1 promoter. After log-phase growth of cells, RNA samples from wild-type and mutant strains were harvested and subjected to qRT-PCR to amplify intron and messenger RNA sequences from ACT1, RPP1B, and YRA1. The three dbr1 alleles tested show strong intron RNA accumulation phenotypes, comparable to the dbr1Δ knockout allele (Fig. 7b–d), which fits with information from previous experiments (Salem et al. 2003; Khalid et al. 2005) (Salem and Menees, unpublished data).

qRT-PCR analysis of a debranching time course

Using a combination of Dbr1p and PNPase treatments, in vitro debranching reactions of total cellular RNA from a dbr1 strain were followed over time courses of 30 min. Debranching reactions were stopped at different times and the reaction products were treated with PNPase to degrade linearized intron RNAs. The remaining intron lariats were detected by qRT-PCR as described above. Because the PNPase treatment step degrades all linear RNAs, human GAPDH cDNA was added to the yeast RNA samples as an exogenous control. The GAPDH cDNA is insensitive to both Dbr1p and PNPase, remaining at the same level in the various samples (data not shown). Debranching of the ACT1 and RPP1B intron lariats was almost complete within the first 5 min of the reactions (Figure S1). However, the debranching rate of the ACT1 intron lariat appears to be only two-thirds the initial rate of the RPP1B intron lariat.

Discussion

The results reported here describe the development and application of (1) RT-PCR methods for sensitively and specifically detecting and quantifying lariat RNA species and (2) enzymatic treatments with characteristic, distinguishing effects on lariat and linear RNAs. We have explored some applications of these methods, such as testing the activity of Dbr1p mutants and following debranching reactions over time courses.

Our results using qRT-PCR to follow in vitro debranching reactions match well with the results reported previously using pure, synthetic substrates (Nam et al. 1994; Khalid et al. 2005). Nam et al. (1994) performed an in vitro debranching time course of a 32P-labeled lariat RNA purified from an in vitro splicing reaction, using an extract from E. coli cells expressing yeast Dbr1p. Khalid et al. (2005) performed in vitro debranching time courses with a 32P-labeled synthetic branched substrate and yeast Dbr1p purified from an E. coli expression system, using the same plasmid that we used in the current study. Khalid et al. (2005) reported that Dbr1p acted very quickly in reactions, with 79% of the branched substrate converted to linear product within 13 s (the first time point of the experiment). Our reactions, using actual yeast intron lariat RNAs within a total RNA sample, show that the debranching rates can vary from one intron lariat to another. The ACT1 intron is debranched at only two-thirds the initial rate at which the RPP1B intron lariat is debranched. These data suggest the possibility that different intron lariats are debranched at different rates in vivo, which may be of functional significance. Slower rates of debranching may occur for introns that contain snoRNAs or mirtrons, reflecting the binding of additional factors to intron sequences or specific folding properties of the RNA. If so, the rate of debranching of introns could be used to predict which introns may contain additional information. Relative debranching rates can be inferred from quantitative analysis of intron RNA levels relative to mature mRNA levels for a given gene compared to a standard, rapidly debranching intron RNA. For these types of experiments RNA samples would have to be taken from a wild-type strain (DBR1), where lariat RNAs are not stabilized. Inefficient splicing would have to be ruled out before further study of candidate slow debranchers. As described above, YRA1 is an example of a gene that uses splicing inefficiency to regulate protein levels.

We hypothesize that qRT-PCR of lariat RNAs can be used to determine the relative rates of transcription for different intron-containing genes, although there are many caveats. Currently, nuclear run-on assays are considered to be the best method for estimating a gene’s transcription rate (Garcia-Martinez et al. 2004; Smale 2009). Rate estimates have also been calculated from combining information on an mRNA’s steady state level and its half-life (Holstege et al. 1998). A new method, called dynamic transcriptome analysis (DTA) (Miller et al. 2011) uses metabolic labeling of cellular RNAs to measure RNA synthesis and decay rates. All these methods have different strengths and weaknesses. These differences are highlighted by the fact that there are differences in the relative transcription rates they predict (Table 3).
Table 3

ACT1, YRA1, and RPP1B mRNA expression

Gene

Transcriptional frequencya

DTAb

Relative intron levelsc

ACT1 (YFL039C)

45.5d (1)

7.2e (1)

63.2 (1)

1.0

YRA1 (YDR381W)

16.2d (0.4)

80.6e (11.2)

88.9 (1.4)

1.1

RPP1B (YDL130W)

120.0d (2.6)

23.0e (3.2)

192.7 (3)

28.1

a mRNAs/cell/hr; numbers in parentheses are levels normalized to ACT1 level

bDTA = dynamic transcriptome analysis (Miller et al. 2011), measured as mRNAs/cell/cell cycle time (150 min); numbers in parentheses are levels normalized to ACT1 level

cDerived from data in Fig. 7 for the dbr1 null strain versus wild type for each gene and normalized to ACT1 level

dEstimated from RNA expression levels and mRNA half-lives (Holstege et al. 1998)

eEstimated from genomic run on experiments (Pelechano and Perez-Ortin 2010)

The use of intron RNA lariats as a novel data source for estimating relative levels of transcription for pre-mRNAs limits the utility to intron-containing genes, a notable limitation for S. cerevisiae. Furthermore, a Dbr1p-deficient strain would have to be used for introns lariats to be a stable record of transcription. Work with yeast dbr1 mutants over the years has not found any significant perturbation of cellular physiology other than the accumulation of intron RNA lariats. In the experiments reported in Fig. 7b–d, the level of RPP1B intron RNA in a dbr1Δ strain relative to the level in wild type is much greater (~330-fold) than the corresponding levels of ACT1 and YRA1 intron RNAs (~13-fold). These data indicate that the transcription rate for RPP1B is almost 30-fold greater than the rates for ACT1 and YRA1 (summarized in Table 3). These relative transcription rates are very different from estimates based on nuclear run-on assays, mRNA steady state levels plus half-lives, and DTA (Table 3). Therefore, development of new methods will provide additional insights into the issues related to determining transcription rates and contribute to the development of ideas about this important aspect of gene expression.

An area where the utility of excised introns is clearer is in determining relative rates of alternative splicing for a particular gene. Variable stabilities of different mRNAs confound estimates of their rate of synthesis, whether the synthesis that produces the mRNAs in question is transcription, as discussed above, or alternative splicing. The use of a Dbr1p-deficient strain, which stabilizes the alternatively excised intron lariats equivalently, results in intron RNA lariat levels that directly reflect the rate of alternative splicing.

The methods reported here can also be applied to genome-wide analysis of introns themselves and are an improvement on previous analyses that also directly analyzed intron RNA lariats. In one study (Zhang et al. 2007), researchers sought to identify introns on a genome-wide scale by analyzing lariat RNA populations within S. cerevisiae cells. They accomplished this by identifying RNA sequences that accumulate at elevated levels in a dbr1 mutant strain compared to a wild-type strain. These researchers used a genomic tiling array with probe sequences spaced an average of 5 bp apart to map the origins of the elevated RNAs. The vast majority of RNAs with enhanced expression in the dbr1 strain corresponded to the lariat portions of annotated introns but several new introns were also identified. However, introns of genes that exhibit low levels of transcription were not detected. In fact, <150 yeast introns were detected in their experiments (about 50% of yeast introns), even though almost all the undetected introns were in expressed genes (i.e., expressed in the growth condition used for the study). In another study of yeast introns that also compared RNAs accumulating in dbr1Δ and wild-type strains using high density tiling arrays (Juneau et al. 2007), the sensitivity of the method was enhanced by optimizing signal analysis by training the software on known intron and exon sequences. In this study, 76% of known introns were detected and several new introns were discovered.

High-throughput sequencing has recently been applied to the analysis of whole transcriptomes (RNA-seq) and has proven to be a powerful method for deducing the presence of introns and assessing alternative splicing patterns (Cloonan et al. 2008; Lister et al. 2008; Mortazavi et al. 2008; Nagalakshmi et al. 2008; Pan et al. 2008; Sultan et al. 2008; Wilhelm et al. 2008). Analysis of RNA-seq data from the S. cerevisiae transcriptome detected 60% more introns than Zhang et al. (2007), identifying 240 of the 306 known yeast introns (Nagalakshmi et al. 2008).

RNA-seq of intron RNA lariat populations prepared using PNPase can provide complementary information to RNA-seq of whole transcriptomes and may reveal new lariat sequences not evident from transcriptome analysis alone. Intron RNA lariat levels can be greatly enhanced by blocking the RNA debranching reaction, which increases the likelihood of detecting even rare splicing events. Because cells defective for RNA debranching activity accumulate excised introns in their lariat forms, with shorted 3′ tails (Chapman and Boeke 1991; Salem et al. 2003), information on the 3′ intron–exon junction is not obtainable from intron lariat RNA sequences. Nevertheless, lariat sequences provide information about branch points that is not obtainable from whole transcriptome sequencing (Vogel et al. 1997). Such information is especially useful for studies of introns in organisms whose branch point sequences are not as highly conserved as those in S. cerevisiae [e.g., humans (see Gao et al. (2008) and references therein)]. Finally, the absence of known intron sequences from an RNA population enriched for RNA lariats can indicate that a gene is not expressed under the growth regimen employed. However, if an intron-containing gene is known to be expressed during the experiment, absence of intron sequences from the RNA lariat population could be an indication that the intron is removed by the hydrolytic splicing pathway observed for self splicing group II introns rather than the predominant branching pathway (Daniels et al. 1996; Vogel and Borner 2002).

A recent study of human branch points was performed to better establish their sequence characteristics (Gao et al. 2008). To this end, RT-PCR and sequencing of 367 individual intron lariat RNA branch points were performed. The method relied on the characteristic misincorporation of a single nucleotide during reverse transcription across the 2′–5′ bond as an indicator of the position of the branch point. However, branch points could be established for less than half of the introns (181 of the 367 introns) because RT-PCR products for the others lacked an apparent misincorporated nucleotide. Even though the work of Gao et al. (2008) represents an important advance, their study only analyzed introns from 20 housekeeping genes. The effect of this limited sampling on the resulting branch point consensus sequence is not known. High-throughput sequencing of enriched lariat RNAs from human cells will be useful for much more detailed analysis of human branch point sequences.

Recently, a computational model for identifying mammalian branch point sequences has been developed that relies, in part, on splicing factor 1 (SF1) binding affinity data (Pastuszak et al. 2010). However, in a test of the model, it only predicted 36% of the actual branch points identified by Gao et al. (2008). The study’s authors concluded that other criteria are needed to enhance branch point sequence predictions. Even with better models, it is clear that validation of any predicted branch point will be required. The RNA-seq method proposed here will be of great utility in efforts to identify or validate branch point sequences.

We are currently applying the methods described here to test our model that Ty1 RNA forms a transient lariat during the reverse transcription step of retrotransposition. We are also interested to see if these methods can be used to identify any novel lariat or other non-linear RNAs in S. cerevisiae. Such RNAs, whether Ty1 RNA or another RNA, may be the key to unlock the mystery of how Dbr1p acts as a host factor for Ty1 retrotransposition.

Acknowledgments

We thank Beate Schwer for the gift of pET16b-DBR1 and Haoping Liu for yeast sigma strain 10560-23C. Support was provided by National Science Foundation, the University of Missouri Research Board, and the University of Missouri-Kansas City School of Biological Sciences.

Supplementary material

438_2011_635_MOESM1_ESM.pdf (20 kb)
Supplementary Figure S1 (pdf 19.5 kb)

Copyright information

© Springer-Verlag 2011