Evaluating Group I Intron Catalytic Efficiency in Mammalian Cells

Abstract

Recent reports have demonstrated that the group I ribozyme from Tetrahymena thermophila can perform trans-splicing reactions to repair mutant RNAs. For therapeutic use, such ribozymes must function efficiently when transcribed from genes delivered to human cells, yet it is unclear how group I splicing reactions are influenced by intracellular expression of the ribozyme. Here we evaluate the self-splicing efficiency of group I introns from transcripts expressed by RNA polymerase II in human cells to directly measure ribozyme catalysis in a therapeutically relevant setting. Intron-containing expression cassettes were transfected into a human cell line, and RNA transcripts were analyzed for intron removal. The percentage of transcripts that underwent self-splicing ranged from 0 to 50%, depending on the construct being tested. Thus, self-splicing activity is supported in the mammalian cellular environment. However, we find that the extent of self-splicing is greatly influenced by sequences flanking the intron and presumably reflects differences in the intron’s ability to fold into an active conformation inside the cell. In support of this hypothesis, we show that the ability of the intron to fold and self-splice from cellular transcripts in vitro correlates well with the catalytic efficiency observed from the same transcripts expressed inside cells. These results underscore the importance of evaluating the impact of sequence context on the activity of therapeutic group I ribozymes. The self-splicing system that we describe should facilitate these efforts as well as aid in efforts at enhancing in vivo ribozyme activity for various applications of RNA repair.

The observation that certain RNA enzymes can efficiently perform cleavage or splicing reactions upon RNA substrates in the test tube (7) has engendered much speculation about the potential utility of ribozymes for inhibiting gene expression (5, 26, 30, 37) or for repairing mutant RNAs in human cells (28, 36). Such therapeutic development has been hindered by the fact that it has been difficult to directly evaluate RNA catalysis in clinically relevant settings. Experimental approaches in which trans-splicing group I ribozymes may be evaluated for substrate specificity and RNA repair efficiency in mammalian cells have recently been described (15, 16). While a significant fraction of substrate RNAs were converted to products by trans-splicing in this experimental system, one significant limitation of these studies is that they did not allow one to determine what fraction of the ribozymes was folded into a catalytically competent conformation after expression from a cellular polymerase. Herein we set out to begin to directly evaluate this important reaction parameter, which we call catalytic efficiency, for group I ribozymes inside human cells.

The group I intron from Tetrahymena thermophila has served as a model system for studing both RNA catalysis and RNA folding (8). This intron catalyzes its own excision from precursor 23S rRNAs without the aid of proteins by a well-characterized two-step pathway (Fig. 1) (6). In the context of the pre-rRNA transcript, the intron folds into a complex tertiary structure, forming an active site for phosphodiester transfer. This structure is achieved by base-pairing interactions throughout the intron sequence forming RNA helices, which then fold into a specific conformation to create a catalytic core within the intron. The 5′ splice site is determined by the P1 helix, which is formed by base pairing between the internal guide sequence of the intron and the last 6 nucleotides (nt) of the 5′ exon. In the first step of splicing, P1 docks into the catalytic core, and the 5′ splice site is cleaved by an exogenous guanosine nucleophile. In the second step of splicing, the 5′ and 3′ exons are ligated together, and the intron is released (6). Since generation of functional ribosomes in T. thermophila depends on this rRNA processing event, it is not surprising that intron-containing precursor transcripts rapidly self-splice to completion in their native cellular environment (4).

FIG. 1.

Self-splicing reaction of the Tetrahymena group I intron. The exons are shown as dashed lines with capital letters, and the intron is depicted as a plain line with lowercase letters. The 5′ exon recognition site (CUCUCU) and the internal guide sequence (ggaggg) are indicated. The exogenous guanosine (G) nucleophile, the terminal guanosine (g) of the intron, and the 5′ (⧫) and 3′ (●) splice sites are also shown. The junction in the ligated exon sequences is noted (CUCUCU/UAAG).

The group I intron can also self-splice from full-length and truncated versions of Tetrahymena rRNAs in vitro. However, the rate of self-splicing even under optimal in vitro conditions does not reach the apparent rate achieved in Tetrahymena (1, 4). A similar rate enhancement has been observed in Escherichia coli when the Tetrahymena intron was expressed in the homologous position in the E. coli rRNA gene (40). Because group I introns are not naturally found in E. coli, this result strongly supports the hypothesis that species-specific cofactors are not required for efficient group I ribozyme activity in vivo (40). Other studies with E. coli indicate that the Tetrahymena intron can efficiently self-splice from non-rRNA transcripts expressed from the lacZ and malE genes (23, 25, 33). These E. coli studies suggest that the Tetrahymena ribozyme may be an attractive molecule for therapeutic development since this catalytic RNA can function efficiently in an unnatural cellular setting and form an active catalytic center even when the ribozyme is present in a variety of sequence contexts.

For the Tetrahymena ribozyme to become a useful therapeutic agent, however, it must be able to function efficiently in human cells. Two recent reports demonstrating that trans-splicing versions of the Tetrahymena ribozyme can be employed to repair clinically relevant mutant mRNAs in two different human cell types are encouraging in this regard (19, 22). However, in both of these studies the level of ribozyme-mediated splicing was apparently quite low, with repaired products being detected only after PCR amplification. One potential explanation for this limited activity in human cells is that the ribozyme may not readily adopt a catalytically competent conformation in this setting.

To begin to evaluate the potential effects of the cellular environment as well as the importance of RNA sequence context on group I catalytic efficiency in human cells, we chose to assess the self-splicing activity of the Tetrahymena intron when it is embedded in cellular transcripts. The self-splicing version of the ribozyme was used because, in contrast to the trans-splicing version, substrate binding would not be expected to limit ribozyme activity (16). Thus, self-splicing represents a more direct measure of the ribozyme’s ability to form a properly folded catalytic center. Our results confirm that self-splicing is supported from nuclear expressed genes in human cells and that intron excision proceeds with high fidelity. However, we also find that sequences flanking the group I intron can significantly affect the catalytic efficiency of the ribozyme in vivo. Moreover, we find that the intron’s ability to efficiently splice from a given transcript in vivo correlates well with the intron’s propensity to fold into an active conformation and splice from the same transcript in vitro.

MATERIALS AND METHODS

Plasmid construction.

pNI was constructed by inserting a DNA fragment containing the Tetrahymena group I intron from pβGST7 (2) into the retroviral vector N2A (13). A fragment of pβGST7 containing the Tetrahymena self-splicing intron, including the T7 promoter and 274 nt of 5′ exon sequences and 29 nt of the 3′ exon, was amplified by PCR and PCR primers (upstream primer, 5′-CGGGGATCCTAATACGACTCACTATA; downstream primer, 5′-GGGACGCGTGACCCCTTTCCCGCAATTTGACAAGCTTGTGACGAGGCA). The PCR product was digested with BamHI and MluI and inserted between the BglII and MluI sites in the polylinker of N2A. A catalytically inactive version of the intron was made by deleting a BglII to NheI fragment from pβGST7 to remove 94 bp from the intron core (29). A retroviral construct carrying the inactive intron was constructed in parallel to provide a negative control for splicing activity (pNId).

pMI was derived from the trans-splicing ribozyme construct pT7 L-21 (39). First, the T7 promoter was replaced with a Moloney murine leukemia virus (MMLV) promoter, and lacZ sequences in the 3′ exon were replaced with the open reading frame for green fluorescent protein (GFP) by standard cloning techniques. A fragment containing the first 43 nt of the self-splicing intron and ∼750 nt of 5′ exon sequences was amplified via PCR from pNI (upstream primer, 5′ - CCCGAGC TCAAAAAAAGAGCCCACAACCCCTCAC TCGGGGC TCGAGCTGGATACTTCC; downstream primer, 5′-TGACCCCTTTCCCGCAATTTGAG) and inserted between the SacI site of the MMLV promoter and the SphI site of the intron. Again, a catalytically inactive version of the ribozyme was inserted in this context (pMId) to serve as a control.

To generate pNI+gfp, the sequences between the BglII and MluI sites in pNI were replaced by sequences contained in a DNA fragment from pMI digested with BglII and NotI by using blunt-end cloning techniques. This sequence “swap” restores the 3′ half of the intron but introduces 3′ exon sequences from pMI, including the coding region for GFP, upstream of the long-terminal-repeat (LTR) sequences present in the 3′ exon of pNI. To generate pMI+ltr, the sequences between the BglII and AgeI sites in pMI were replaced by sequences contained in a DNA fragment from pNI digested with BglII and SmaI, which includes the 3′ half of the intron and ∼500 nt of the 3′ exon sequence.

Transcription of self-splicing control transcripts.

RNAs generated by in vitro transcription were used as controls to ensure that self-splicing was not occurring during RNA extraction and analysis. pNI and pMI were digested with XbaI and NotI, respectively, to serve as DNA templates for transcription by T7 RNA polymerase. Transcription reactions were carried out in a volume of 75 μl in reaction buffer (40 mM Tris, pH 7.5; 5 mM MgCl₂; 10 mM dithiothreitol [DTT]; 4 mM spermidine, 4 mM deoxynucleoside triphosphates) for 1 h at 37°C. These conditions almost exclusively yield full-length precursor transcripts with very little self-splicing product generated during transcription. After DNase treatment and phenol-chloroform extraction, the transcripts were purified by spin chromatography through a Chroma-100 column (Clontech).

Transfection and RNA isolation.

Phoenix amphotrophic cells (17) were grown in Dulbecco modified Eagle medium (DMEM) supplemented with 10% fetal calf serum (FCS). Cells were typically seeded in 60-mm-diameter plates at a density of 2.4 × 10⁶ cells per plate and allowed to settle for 18 to 24 h. Cells were transfected with 5 to 10 μg of intron-containing plasmid DNA, and cotransfected (where appropriate) with 0.2 to 0.5 μg of pEGFP-N1 (Clontech) (a plasmid expressing GFP used to determine transfection efficiencies) by using the calcium phosphate method of transfection (17). The medium containing the DNA precipitate was replaced with fresh medium 16 h posttransfection. Total RNA was isolated from cells 42 to 48 h posttransfection. Transfection efficiencies were monitored by flow cytometry to assess GFP expression (when applicable). To isolate total RNA from transfected cells, the cells were trypsinized and collected by centrifugation. The cell pellet was then solubilized in 1.5 ml of Tri-Reagent (Molecular Research Center, Inc.) supplemented with 10 mM EDTA. In control samples, in vitro-generated transcripts containing the intron were added to the Tri-Reagent before the mock-transfected cells were solubilized. The total RNA was extracted and precipitated according to the manufacturer’s protocol and was resuspended in 10 mM Tris (pH 7.5)–0.1 mM EDTA. Prior to RNase protection analysis, the RNA samples were DNase treated for 5 min at 37°C in reaction buffer (10 mM Tris, pH 7.5; 6.25 mM MgCl₂; 50 U of DNase per ml) containing 10 mM l-argininamide to inhibit splicing.

Isolation and sequencing of spliced products.

Retroviral supernatants were collected 48 h posttransfection from Phoenix cells transfected with the retroviral vector-based constructs and used to infect 2 × 10⁵ NIH 3T3 cells. Infections were performed with 1 ml of undiluted supernatant in 2 ml of DMEM plus 400 μg of Polybrene per ml at 37°C. After 4 h of incubation, 2 ml of DMEM with 10% FCS was added. The virus-containing medium was replaced 48 h after infection with selective medium containing 0.8 mg of Geneticin (Gibco-BRL) per ml. G418-resistant cells were isolated, and genomic DNA was harvested by using a genomic DNA isolation kit (Qiagen, Inc.). Genomic DNA was digested with SphI to destroy unspliced intron sequences in the DNA sample before amplification of splice junction sequences by PCR by using primers specific for 5′ and 3′ exon sequences. The PCR products were digested with BamHI and HindIII and subcloned into pUC19. Inserts were sequenced by using the dideoxy method and a downstream primer (5′-GTGACGAGGCATTTGGC) complementary to 3′ exon sequence located immediately downstream of the expected spliced junction. Sequencing reaction mixtures were separated on a 10% denaturing polyacrylamide gel and visualized by use of a PhosphorImager (Molecular Dynamics).

RNase protection assay.

RNase protection probes were transcribed from DNA templates that were generated by PCR. Intron-containing plasmids were amplified by using a downstream primer that contains the T7 promoter sequence, followed by 30 nt of nonspecific sequences and 18 nt which are complementary to the 3′ exon of a given construct. Upstream primers were designed to anneal to primer sites chosen 40 nt upstream of the 3′ splice site in the intron DNA sequence or 40 nt upstream of the splice junction in DNA containing the spliced exons. PCR products were gel purified and isolated from the gel by using Promega PCR Preps as instructed by the manufacturer.

Radiolabeled antisense probes were generated by transcribing DNA templates for 1 h at 37°C with 40 mM Tris (pH 8); 6 mM MgCl₂; 10 mM DTT; 2 mM spermidine; 0.5 mM GTP, ATP, and UTP; 12.5 μM CTP; 6.25 μM [α-³²P]CTP; and 20 U of T7 RNA polymerase (Roche Molecular Biochemicals). Transcription reaction mixtures were DNase treated with 4 U of DNase for 20 min at 37°C, and products were separated on a 6% denaturing polyacrylamide gel. Full-length probes were excised from the gel and eluted overnight in probe elution buffer (Ambion). Eluted probe samples were analyzed by using a scintillation counter in order to quantitate the incorporation of radioisotope and calculate probe concentration.

RNase protection assays were performed by using an RPA II kit (Ambion) as described by the manufacturer. Briefly, total RNA samples (5 to 10 μg) were DNase treated, added to 2 fmol of radiolabeled probe, and allowed to hybridize overnight at 42 to 45°C. The RNAs were then digested with a 1:100 dilution of RNase A and RNase T₁ mixture (0.5 U of RNase A, 20 U of RNase T₁) (Ambion) at 37°C for 30 min. The protected RNA fragments were precipitated and separated on a 6% denaturing polyacrylamide gel. Quantitation of protected products corresponding to precursor and spliced RNAs was performed on a PhosphorImager by using ImageQuant software (Molecular Dynamics). Reported values have been adjusted to account for different specific activities of the protected products due to different lengths of the expected fragments.

In vitro splicing reactions.

Total cellular RNA samples were thermally renatured by incubation of the RNA in H₂O at 50°C for 20 min. The in vitro splicing reaction was initiated with an equal volume of cis-splicing buffer (2× concentration of 300 mM NaCl, 100 mM Tris [pH 7.5], 20 mM MgCl₂, and 4 mM GTP) or with an equal volume of low magnesium cis-splicing buffer (2× concentration of 300 mM NaCl, 100 mM Tris [pH 7.5], 4 mM MgCl₂, and 1 mM GTP). Reaction mixtures were incubated at 37°C for 20 min for single time point samples or for 0, 2, 5, 20, 60, and 150 min for splicing time courses. Reactions were quenched by the addition of an equal volume of 20 mM EDTA. Splicing efficiency was assessed by RNase protection analysis as described above. The observed fractions of spliced products were adjusted to determine the fraction of spliced product accumulated in vitro; any fraction of spliced products generated in vivo was subtracted from the spliced products observed at each time point, and that value was divided by the fraction of precursor transcripts present in the initial sample. Data were fit to two-phase exponential equations by using GraphPad PRISM software. To compare the rates of splicing for the active intron populations, the data were first normalized by dividing the fraction of in vitro-generated spliced products calculated at each time point by the fraction of spliced products calculated at the last time point taken.

RESULTS

The Tetrahymena intron self-splices from transcripts expressed in human cells.

To determine if the Tetrahymena ribozyme can mediate splicing in mammalian cells when expressed within an RNA polymerase II (Pol II) transcript, we introduced the Tetrahymena self-splicing intron into a retroviral vector called NI. To generate NI, the intron was inserted in the U3 region of the 3′ LTR of the retroviral vector N2A (Fig. 2) (13). In addition, a control vector called NId was generated that is identical to NI except that it contains an inactive version of the ribozyme (29). The plasmids containing NI (pNI) and NId (pNId) were transfected separately into the Phoenix packaging cell line (17), which is derived from human embryonic kidney 293T cells. Inside the packaging cells, NI and NId express viral transcripts that encode neomycin phosphotransferase and contain the Tetrahymena intron in the 3′ untranslated region (UTR).

FIG. 2.

Self-splicing intron expression constructs. The Tetrahymena self-splicing intron (black box) and flanking exon sequences derived from pβGST7 (wavy lines) are shown. The MMLV LTR promoter (MLV) and coding sequences for neomycin phosphotransferase (Neo^R) and GFP are indicated. RNA Pol II transcription start sites are denoted with arrows, and the location of the polyadenylation signals [p(A)] are labeled. Insertion sites of GFP and LTR sequences in NI+gfp and MI+ltr, respectively, are denoted by the dashed lines.

To establish that the Tetrahymena intron self-splices in human cells and to directly evaluate the fidelity of this reaction, we took advantage of our experimental system and the retroviral life cycle. The Phoenix packaging cells used in this study encapsidate retroviral vector-derived transcripts into virions, and this virus can be used to transduce other mammalian cells. During infection, retroviral transcripts are reverse transcribed into double-stranded DNA, which is then stably integrated into the host cell’s genome (Fig. 3). The in vivo reverse transcription step in the viral life cycle allowed us to directly identify and characterize the splicing reaction products by analyzing proviral DNA sequences isolated from infected NIH 3T3 cells. Spliced junctions were preferentially amplified from genomic DNA that had been digested with SphI, which selectively cleaves unspliced intron sequences present in the DNA sample. The amplified products were then cloned and sequenced. The sequences of five individual spliced junctions were determined to be exactly as anticipated (Fig. 3). A larger sample of spliced junctions was analyzed after PCR amplification by restriction enzyme analysis. The amplified products corresponding in size to the expected splice junction sequences were cleaved to completion by AflII, as would be anticipated following precise excision of the intron (data not shown). Such analyses demonstrated that the Tetrahymena group I intron had indeed self-spliced from retroviral transcripts before reverse transcription and that such splicing had proceeded with high fidelity.

FIG. 3.

Sequence of a group I intron-derived splice junction. A scheme for retroviral gene transfer is shown on the left. Viral RNAs containing the Tetrahymena intron are expressed from retroviral vectors transiently transfected into a viral packaging cell line. As described in the text, products of group I self-splicing from transcripts in vivo can be evaluated at the DNA level by PCR amplification of proviral DNA sequences integrated in the genomes of infected cells. A representative example of a sequenced splice junction is shown on the right, along with the expected RNA sequence for a properly spliced product. The junction between the 5′ and 3′ exons and the AflII restriction site generated by this junction are labeled.

To determine the steady-state fraction of transcripts from which the intron has excised itself in vivo, total RNA was extracted from Phoenix cells 2 days after transfection with pNI or pNId and subjected to RNase protection analysis. An RNA probe complementary to 3′ exon derived LTR sequences and spanning the 3′ splice site (designated 3′ss) was hybridized to samples of total RNA to simultaneously detect precursor and spliced transcripts (Fig. 4a). Moreover, a second probe complementary to the 3′-exon-derived LTR sequences and to the expected spliced junction of a self-processed transcript (designated sj) was also generated to detect fully spliced products (Fig. 4a). With the 3′ss probe for RNase protection analysis, RNA isolated from cells transfected with the active intron construct NI yielded two protected RNA fragments of the expected sizes that correspond to the unspliced precursor and spliced product transcripts (Fig. 4b). The identity of the protected fragment corresponding to the spliced product was confirmed by incubating the total RNA sample under splicing conditions in vitro prior to RNase protection analysis. As predicted, the shorter RNA product increased in intensity relative to the longer protected fragment (compare lanes 7 and 8). This in vitro splicing analysis also demonstrated that the unreacted fraction of remaining precursor transcripts isolated from the transfected cells were competent to undergo self-splicing even though they did not self-splice in the cells. Detection of the RNase protection fragment corresponding to the spliced transcript was dependent on ribozyme activity, since only one protected RNA fragment, corresponding to the unspliced precursor, was detected when total RNA from cells transfected with pNId was analyzed. This precursor transcript was not converted to spliced product even when incubated under splicing conditions in vitro (Fig. 4b).

FIG. 4.

RNase protection analysis of intron-containing retroviral transcripts. (a) Schematic view of RNase protection probes for NI. The two RNase protection probes, 3′ss and sj, are indicated. The 3′ splice site probe, 3′ss, spans the junction between the 3′ exon LTR sequences and the intron in the precursor transcript. The splice junction probe, sj, is complementary to the expected spliced junction product RNA which lacks the intron. Both probes contain 30 nt of nonspecific sequence at their 5′ ends which is not complementary to NI-derived transcripts. RNA generated by in vitro transcription from the indicated T7 promoter (T7 pro) served as a control transcript as described in the text. (b) RNase protection analysis with the 3′ splice site probe. The probe lane shows the undigested probe used in the assay (387 nt). T7 transcripts containing the intron were added to mock-transfected cells in control lanes. The input control transcript (ø) was added to buffer just prior to RNase protection analysis and served as a size marker for unspliced transcripts. RNA samples were analyzed before (−) and after (+) they had been incubated under in vitro splicing conditions. Protected RNAs of 357 and 315 nt detected in control transcripts incubated under splicing conditions correspond to unspliced and spliced transcripts, respectively, and are marked on the right with arrowheads. Total RNA samples isolated from cells transfected with pNI or pNId are labeled. (c) RNase protec- tion analysis with the splice junction probe. Lanes are labeled as described in panel b. Protected RNAs of 357 and 315 nt detected in control transcripts incubated under splicing conditions correspond to spliced and unspliced transcripts, respectively, and are marked on the right with arrowheads. (d) Evaluation of RNase protection assays. Increasing amounts of total RNA isolated from cells transfected with pNI were analyzed as described in Materials and Methods. Protected fragments corresponding to unspliced (U) and spliced (S) transcripts are marked with arrowheads. A linear increase in radioactive signal is observed over the range of RNA amounts used in these studies. Moreover, a constant fraction of spliced NI RNAs is observed from the range of input RNA analyzed.

An important control supports the conclusion that all of the observed group I intron self-splicing activity detected by RNase protection analysis occurred in the transfected cells and not during RNA extraction or during in vitro analysis. Control RNAs were generated by in vitro transcription that contain the same intron and 3′ exon flanking sequences as NI. These transcripts were introduced into the cell lysis reagent added to a mock-transfected plate of cells at the time of cell lysis. The control RNA sample was then analyzed in parallel with the RNAs isolated from transfected cells. As demonstrated by RNase protection analysis, the group I intron-containing control transcripts did not self-splice during the processing and analysis of RNA samples (Fig. 4b, lane 2), even though introns in these control transcripts are capable of self-splicing when incubated in vitro under appropriate conditions (Fig. 4b, lane 3).

To verify that both steps of splicing have occurred accurately, rather than hydrolysis of the 3′ exon or intron mis-splicing, the RNase protection probe that spans the expected splice junction sequence, probe sj, was also employed to evaluate the splicing products present in RNAs from cells transfected with pNI or pNId. RNase protection analysis with this probe detected a fragment of the size expected for a transcript containing this splice junction in addition to a shorter fragment that corresponds to unspliced precursor transcripts (Fig. 4c, lane 6). The identities of these products were confirmed by comparison with the protected fragments generated from in vitro-transcribed RNAs, which had been incubated under splicing conditions (Fig. 4c, lane 4). Again, no protected fragments corresponding to spliced products were detected when RNA from cells transfected with the inactive ribozyme were analyzed (Fig. 4c, lane 5) or when the unspliced in vitro-generated control transcript containing the group I intron was added to mock-transfected cells during RNA extraction (Fig. 4c, lane 3).

Ribozyme catalytic efficiency was measured by quantitating the relative abundance of the two protected RNA fragments generated by RNase protection analysis. Because spliced and unspliced transcripts are simultaneously detected in the RNA sample by using a single probe, the fractions of spliced transcripts from independent samples can be quantitated and compared. To verify that such RNase protection analysis yielded an accurate measure of spliced and unspliced transcripts over the working range of RNA, increasing amounts of cellular RNA containing NI transcripts (2.5 to 20 μg) were analyzed. As shown in Fig. 4d, the expected changes in radioactive signal were observed with increasing input RNA, while a constant fraction of spliced transcripts were detected.

The steady-state fraction of spliced products generated in cells transfected with pNI was calculated as the average fraction detected by RNase protection analysis of total RNA samples isolated from multiple, independent transfections. When the 3′ss probe was used, the intron had excised itself from 33% ± 5%, (range, 24 to 49% for seven independent experiments) of the retroviral transcripts. Similar splicing efficiencies were observed in these RNA samples when the sj probe, which spans the splice junction, was employed. These results indicate that a significant fraction of group I introns can form catalytic centers and promote self-splicing from nuclearly expressed transcripts in human cells. The observation that the 3′ss and sj probes detect the same level of splicing suggests that most group I splicing events that initiate go to completion inside cells. Similar self-splicing efficiencies were obtained from transcripts in two murine fibroblast cell lines, a viral packaging cell line called E86 and the corresponding parental cell line NIH 3T3, that had been stably transfected with pNI and pNId (19a). The fraction of introns which had self-spliced from the NIH 3T3 cells did not significantly differ from the fraction observed in mouse E86 cells or human Phoenix cells, which constitutively express viral genes. Therefore, the efficiency of group I intron splicing observed in Phoenix cells from intron-containing transcripts (Fig. 4) is not specific to the transient expression of self-splicing constructs or the expression of constructs in a specific cell type.

Transcript context affects ribozyme catalytic efficiency in vivo.

A second intron-containing construct was designed to assess self-splicing activity when the intron is placed in a different transcript context. MI is a construct which expresses a 1.5-kb transcript encoding GFP (Fig. 2). In contrast to the intron’s location in the 3′ UTR of NI, the intron is located upstream of the coding region in pMI. Approximately 800 nt of 5′ exon sequence immediately upstream of the intron is common to both NI and MI constructs. Downstream of the intron, only 23 nt of the 3′ flanking sequence is common to both expression cassettes. To assess self-splicing from MI transcripts, pMI was transiently transfected into Phoenix cells, and total RNA was isolated and analyzed by using RNase protection analysis. As shown in Fig. 5, this analysis demonstrated that the intron does not self-splice from MI transcripts at a detectable level inside the transfected cells. However, the intron-containing RNAs isolated from the cells are able to self-splice when incubated under standard in vitro reaction conditions (Fig. 5b, compare lanes 1 and 2). These results suggest that the intron is largely inactive in vivo within the MI transcript context but that this inactivity is not simply due to a defect in the intron sequence. Dideoxy sequencing of the group I intron within pMI confirmed that no mutations were introduced into the intron during plasmid construction (data not shown).

FIG. 5.

RNase protection analysis of MI transcripts. (a) Schematic view of an RNase protection probe for MI. The RPA probe spans the 3′ splice junction between the 3′ exon GFP sequences and the intron in the precursor transcript. The probe contains 30 nt of nonspecific sequence at the 5′ end which is not complementary to MI-derived transcripts. RNA generated by in vitro transcription from the indicated T7 promoter (T7 pro) served as a control transcript to monitor for potential self-splicing during RNA extraction and analysis. (b) RNase protection analysis of transcripts from MI-transfected cells. The probe lane shows the undigested probe used in this assay (322 nt). T7 transcripts containing the intron were added to mock-transfected cells in control lanes. RNA samples were analyzed before (−) and after (+) they had been incubated under in vitro splicing conditions. Protected RNAs of 292 and 250 nt detected in control transcripts incubated under splicing conditions correspond to unspliced and spliced transcripts, respectively, and are marked on the left with arrows. Total RNA samples from cells transfected with pMI are labeled.

We attempted to determine what differences between the active (NI) and inactive (MI) transcripts accounted for the striking difference in the ability of the group I intron to self-splice in the cellular environment. One possible explanation is that the local sequence context of the intron influences the proper secondary and tertiary folding of the ribozyme in vivo (24, 27). Since the 3′ exon flanking sequences are significantly more divergent than the 5′ exon sequences in NI and MI transcripts (Fig. 2), we sought to determine whether 3′ exon sequences could influence self-splicing efficiency. Two new expression cassettes were constructed from plasmids pNI and pMI to test this hypothesis. In the first, pNI+gfp, the 3′ exon GFP sequence flanking the intron in pMI was inserted downstream of the group I intron present in pNI. In the second, pMI+ltr, the 3′ flanking LTR sequence from pNI was inserted downstream of the intron present in pMI (see Fig. 2). Thus, pNI and pMI+ltr were constructed to contain identical 3′ exon sequences immediately flanking the intron, which is located either in the context of the 3′ UTR (NI) or the 5′ UTR (MI+ltr) of the expressed transcript. Similarly, pNI+gfp and pMI plasmids contain the identical flanking 3′ exon sequences in two different contexts. The effects of these 3′ exon sequences on group I intron splicing were then assessed. Each construct was transfected into Phoenix cells, and splicing efficiency was analyzed by RNase protection analysis by using the appropriate 3′ splice site probes and total RNA isolated from transfected cells. The resulting percentage of spliced products detected for the each of these transcripts is shown in Fig. 6. The presence of GFP sequences downstream of the intron in NI transcripts reduced in vivo self-splicing efficiency from ∼35% splicing to ∼5% splicing. In contrast, LTR sequences present downstream of the intron in MI transcripts enhanced self-splicing from below detection (<2%) in MI transcripts to ∼15% in MI+ltr transcripts. These results indicate that sequence context, and in particular the 3′ exon sequences, may significantly affect the efficiency with which the group I intron splices in human cells.

FIG. 6.

Effects of 3′ exon sequence on splicing efficiency in vivo. A schematic comparison of various 3′ exon sequences from the four constructs tested is shown on the left. The constructs containing these sequences are indicated, along with the corresponding splicing efficiency observed in vivo.

A difference in transcript accumulation does not account for the observed differences in catalytic efficiency of related constructs.

It is possible that different mRNA stabilities could indirectly affect the fraction of self-spliced transcripts detected from the various constructs analyzed in vivo if self-splicing rates were low relative to the half-life of the intron-containing transcripts. To address this issue, we compared the relative levels of accumulation of two transcripts from which group I introns self-spliced to significantly different extents. Equal molar amounts of pNI and pNI+gfp (Fig. 2) were cotransfected into Phoenix cells, and total RNA was isolated in order to assess the relative abundance of transcripts expressed from the two retroviral vectors by RNase protection analysis. Since the two transcription units within pNI and pNI+gfp are the same, with the exception of the 800-bp insertion of GFP coding sequences in the 3′ UTR of pNI+gfp, we assumed that the rates of transcription associated with the retroviral promoters would be equivalent for the two constructs. A single RNase protection probe, which spans the GFP-LTR sequences in pNI+gfp, was employed to detect transcripts expressed from both constructs (Fig. 7a). With this probe design, both unspliced and spliced transcripts contribute to the RNase protection signal detected for each construct. Thus, if the low abundance of spliced transcripts detected in NI+gfp (Fig. 6) resulted from a significant decrease in RNA transcript stability, we would anticipate that a decrease in NI+gfp transcripts relative to NI transcripts would also be evident. Results from RNase protection analysis of total RNA samples from singly or cotransfected cells with this probe strategy are shown in Fig. 7b. Quantitation of the relative abundance of the two transcripts present in the RNA isolated from cotransfected cells demonstrated that nearly equal amounts of NI and NI+gfp transcripts were reproducibly found in the cells. In contrast, RNase protection analysis with appropriate 3′-splice-site probes to assess splicing efficiencies in these RNA samples confirmed that there is a vast difference in the levles of accumulation of spliced products within cells cotransfected with NI and NI+gfp (Fig. 7c). These observations suggest that a difference in RNA half-lives does not significantly contribute to the differences in self-splicing efficiency observed between NI and NI+gfp transcripts in vivo.

FIG. 7.

Accumulation of transcripts from related constructs. (a) RNase protection analysis probe that distinguishes between the 3′ exon sequences following the intron in pNI and pNI+gfp and detects both unspliced and spliced transcripts. (b) RNase protection analysis of total RNA extracted from cells transfected with pNI and pNI+gfp. The probe lane shows the full-length undigested probe used in the assay (330 nt). Lanes indicate total RNA samples isolated from singly and cotransfected cells (performed in triplicate and labeled samples 1 to 3). Protected RNA fragments of 260 and 300 nt correspond to NI and NI+gfp transcripts, respectively, and are indicated by the arrows. The band marked with the asterisk is present in all cells transfected with retroviral vectors derived from N2A. The signal corresponds in size to 5′ LTR sequences which may result from read-through transcription of transfected plasmids. (c) Self-splicing efficiencies of NI and NI+gfp transcripts in RNA samples isolated from cotransfected cells. RNase protection analysis was performed by using appropriate 3′ splice site probes. Percent splicing of NI and NI+gfp transcripts is reported for RNA samples isolated from cotransfected cells and analyzed for relative accumulation of retroviral transcripts in panel b.

In vivo catalytic efficiency correlates with the ability of the intron to fold properly and self-splice from a given sequence context.

In vitro activity assays have frequently been used to assess the rate and extent of proper folding of the group I ribozyme from Tetrahymena (9, 11, 12, 38). Since intron-containing transcripts generated from the four self-splicing constructs described in this study could readily undergo self-splicing under in vitro conditions regardless of their in vivo efficiency, we examined more carefully the in vitro splicing reactions mediated by these introns from their Pol II transcript contexts. The reactions were performed with RNA samples isolated from cells transfected with intron-containing constructs, and the accumulation of spliced products in these reactions was assayed by RNase protection analysis. As expected, the data fit best to a two-phase exponential equation (11). These phases are interpreted as an initial population of properly folded introns which react quickly after initiation of the reaction and as a second slow-reacting population of introns which must undergo refolding to an active structure before catalysis can occur (11). The extent of in vitro splicing for introns embedded in the various cellular transcripts described is shown in Fig. 8a. In Fig. 8b, the data were normalized in each reaction to the final fraction of spliced products in order to compare the levels of progress of active intron populations. Similar curves suggest that introns which fold into an active conformation during the course of the in vitro reaction also self-splice at essentially the same rate, regardless of which transcript context the intron is found (Fig. 8b). However, the extent of group I intron self-splicing varies when it is present in different Pol II transcripts (Fig. 8a). This presumably reflects a difference in the ability of the intron sequence to fold properly within a given cellular transcript under the in vitro conditions used. Interestingly, the transcript context with the greatest self-splicing efficiency observed in vivo, NI, was also the context from which the largest fraction of introns self-spliced in vitro. Moreover, the extents of the reactions observed in vitro for all four transcript contexts correlate exactly with the trend of relative efficiencies observed for group I self-splicing from these transcripts inside cells (Fig. 8a).

FIG. 8.

In vitro self-splicing reactions of various cellular transcripts. Total RNA samples were preincubated and self-spliced in vitro as described in Materials and Methods. The splicing reactions were stopped at the time points indicated, and the reaction products were subsequently analyzed by RNase protection analysis. The fraction of spliced products that had accumulated in vitro at each time point was quantitated and calculated as described in Materials and Methods. Data were fit to a two-phase exponential equation. Closed symbols are used for constructs containing LTR sequences downstream of the intron. Open symbols are used for constructs containing GFP sequences downstream of the intron. Circles correspond to retroviral (NI)-based vectors, while squares correspond to MI-based vectors. (a) Progress of in vitro self-splicing reactions. (b) Data normalized for each reaction in panel a such that the fraction spliced (norm.) is equal to the fraction spliced at each time point divided by the maximum fraction spliced during the reaction.

DISCUSSION

Our results demonstrate that group I introns are able to self-splice from transcripts expressed in the nuclei of mammalian cells. To establish a system which allows for isolation and characterization of bona fide in vivo splicing products without worrying that splicing was occurring during in vitro analysis, we inserted the intron into a retroviral vector. By taking advantage of the retroviral life cycle to recover spliced products that were reverse transcribed into DNA inside cells, we were able to verify that splicing does proceed in vivo and that the intron precisely ligates its flanking exon sequences together (Fig. 3). Moreover, by using quantitative RNase protection analysis, we determined that up to 50% of the introns have self-spliced with high fidelity from steady-state levels of the retroviral transcript, NI, in vivo (Fig. 4). These results suggest that a significant fraction of group I ribozymes, expressed within cellular transcripts, could potentially react in human cells. In addition, we compared the catalytic efficiencies of group I self-splicing from a variety of transcript contexts. Our inability to detect group I self-splicing from one of these transcripts, MI, indicates that transcript context can have dramatic effects on the in vivo catalytic efficiency of group I ribozymes (Fig. 5). The observation that self-splicing efficiencies from cellular transcripts increased or decreased when new sequences were inserted downstream of the ribozyme indicates that group I ribozyme activity may be enhanced or inhibited by the presence of certain flanking RNA sequences in vivo (Fig. 6). In the case of NI and NI+gfp transcripts, which are transcribed from closely related constructs yet differ greatly in catalytic efficiency, we show that differences in RNA stabilities do not contribute significantly to the markedly reduced level of spliced NI+gfp transcripts detected at steady state (Fig. 7). Finally, we obtained intriguing results when we compared the in vitro splicing efficiencies of the various intron-containing transcripts. A reduced extent of splicing in vitro, which is believed to indicate that a greater fraction of the introns are misfolding, correlated with reduced catalytic efficiency in cells (Fig. 6 and 8). Thus, although we cannot provide direct evidence for misfolding of ribozymes in vivo, our results suggest that the ability of the ribozyme to fold into an active conformation in vitro may significantly influence the catalytic efficiency of the ribozyme containing transcript inside cells.

The steady-state level of in vivo spliced products depends upon the relationship between the rate of self-splicing inside cells and the half-life of the expressed transcripts. If intron-containing transcripts have a relatively short half-life, then the intron may not have sufficient opportunity to self-splice before the transcript is degraded. Moreover, a transcript from which the intron has self-excised may become less stable than the precursor transcript and thus under-represent the true fraction of transcribed RNAs which undergo splicing. Alternatively, the intron’s propensity to fold into an active versus inactive conformation in vivo, within the context of the expressed transcript, may significantly contribute to differences in self-splicing efficiency. It has been demonstrated in vitro that misfolded ribozymes may result from unfavorable interactions within the intron and with flanking RNA sequences, as discussed below. In addition to such intrinsic folding properties, RNA folding inside cells may be affected by RNA binding proteins which may favorably or unfavorably influence the intron’s ability to adopt a catalytically competent conformation. It has been demonstrated in vitro that RNA binding proteins, such as ribosomal peptides, viral nucleocapsid proteins, and hnRNPs, may act as RNA chaperones to increase the rate of ribozyme reactions, presumably by enhancing ribozyme folding to achieve the catalytic structure (3, 10, 14, 32). Similarly, RNA chaperone activity may influence splicing rates inside cells. However, it is unclear if the intracellular concentrations of such RNA binding proteins would positively or negatively impact on the ability of ribozymes to correctly fold within transcripts and if those influences would be transcript context dependent. Therefore, a variety of factors can potentially influence the efficiency with which a group I intron self-splices from a Pol II transcript inside mammalian cells.

In our studies, we demonstrate that sequence context can significantly influence the catalytic efficiency of the group I intron self-splicing inside cells even when two similar intron containing transcripts, the NI and NI+gfp RNAs, accumulate to similar levels and thus appear to have similar half-lives inside cells. One simple interpretation of these results is that in the NI context the intron tends to fold into an active conformation more often than in the NI+gfp context. The observation that a greater fraction of the ribozymes in NI transcripts adopt active conformations in vitro compared to that in NI+gfp transcripts strongly supports this theory.

The observation that exonic sequences can influence self-splicing activity is well established for the natural sequence context of the Tetrahymena intron. Inclusion of different lengths of exon sequences in precursor rRNA transcripts can affect the rate of intron self-splicing in vitro (12, 34). In these studies, relative splicing rates are more closely related to the predicted structure of the exon sequences than to the length of the exons. In very short rRNA precursor transcripts, exon sequences that sequester the 5′ exon recognition site from the P1 helix into an alternative helix, called P(−1), dramatically reduce the rate of group I cis-splicing (35). The stabilization or destabilization of P(−1) by point mutations (35) or additional rRNA exon sequences (34) alters the equilibrium between active and inactive intron conformations. Exon sequences have also been shown to limit the activities of catalytic introns of the group I td intron in bacteria (27) and a model group II intron precursor transcript in vitro (20). Local folding of sequences flanking the Tetrahymena intron which enhance or inhibit formation of the catalytically competent intron structure may similarly contribute to the efficiency of self-splicing from the various transcript contexts that we observe in mammalian cells.

Interestingly, such detrimental effects resulting from flanking exon sequences that reduce group I ribozyme catalytic efficiency in E. coli have not been apparent in the few studies that have been performed. The Tetrahymena intron self-splices from the homologous position in E. coli rRNA with a rate rivaling that of Tetrahymena rRNA processing (40). In addition, a number of studies have demonstrated that the Tetrahymena intron self-splices readily from sequences unrelated to the natural rRNA context in bacteria (23, 25, 33). The intron self-splices from the lacZ transcripts efficiently enough to fully complement β-galactosidase activity in E. coli (23, 33). The efficiency of splicing from the lacZ transcript was not directly quantitated in these studies, but RNA blot analysis demonstrated that the free intron was found to be much more abundant than precursor lacZ RNAs in E. coli. In a more recent study, the intron was expressed downstream of the malE gene (25). From this position the intron splices extremely efficiently, as evidenced by the fact that unspliced introns were detected in only 0.2% of the steady-state population of malE transcripts in the bacteria. Efficient self-splicing from these unnatural sequence contexts in E. coli demonstrates that rRNA exon sequences are not a prerequisite for efficient in vivo catalysis. In our mammalian cell studies, we observed that the intron self-excised from different transcript contexts with a range of catalytic efficiencies, including undetectable levels of self-splicing from certain transcripts. While these results cannot be compared directly to those of existing studies of E. coli since it is unknown how relative transcript stabilities affect the different steady-state levels of spliced transcripts, it is somewhat surprising that transcript context has not been shown to significantly impact upon self-splicing efficiency in the few studies undertaken with nonribosomal RNA transcripts in the prokaryotic cellular environment.

As established for the Tetrahymena ribozyme in vitro, kinetically trapped inactive conformations of the folded intron is one plausible mechanism by which self-splicing activity may be limited in human cells. A number of ribozyme derivatives containing point mutations have recently been made that modulate ribozyme folding properties in vitro (21, 31). These include mutations which stabilize the P3 helix of the intron core to increase the fraction of active intron conformers after in vitro transcription (21). Other mutations have been identified that allow the ribozyme to resolve inactive conformations rapidly in order to adopt the active ribozyme conformation more quickly in vitro (31). Evaluation of these and other derivatives of the Tetrahymena intron in mammalian cells by using the experimental system presented here should greatly enhance our understanding of the factors which influence the ribozymes propensity to fold into a catalytically competent conformation in a therapeutically relevant setting.

Recently, a few studies have reported that trans-splicing versions of the Tetrahymena ribozyme can be employed to revise targeted RNAs in E. coli (18, 29) and in mammalian cells (15, 16, 19, 22). Moreover, two of these reports describe the use of such trans-splicing to repair clinically relevant transcripts associated with myotonic dystrophy (22) and sickle cell disease (19) in mammalian fibroblasts and in erythrocyte precursors from sickle cell patients, respectively. Although these two reports are extremely encouraging with regard to the potential utility of RNA repair, since they demonstrate that ribozymes can revise endogenous transcripts, the levels of trans-splicing and RNA repair in these studies were apparently very low. Unfortunately, information about the catalytic efficiency of the ribozymes utilized in these experiments is very difficult to obtain. Nevertheless, for trans-splicing ribozymes to become useful in the clinic, the efficiency with which ribozymes can repair mutant RNAs will need to be assessed and will likely have to be increased. One parameter of the trans-splicing reaction that may significantly impact on a group I ribozyme’s ability to repair target RNAs in vivo is the propensity of the ribozyme to fold into a catalytically competent conformation when it is expressed as part of a longer transcript inside human cells. Unfortunately, we are currently unable to predict if group I ribozymes will tend to form a catalytically active or inactive structure when the intron is embedded in flanking exon sequences. However, the strategy that we employed herein to assess the catalytic efficiency of self-splicing in vivo should prove to be valuable for evaluating and potentially enhancing the catalytic potential of a variety of group I ribozyme-containing transcripts for therapeutic applications in mammalian cells.