In vitro analysis of the relationship between endonuclease and maturase activities in the bi-functional group I intron-encoded protein, I-AniI

Authors


R. B. Waring, Department of Biology, Temple University, 12th & Norris Sts., Philadelphia, PA 19122, USA. Fax: + 1 215 204 6646, Tel.: + 1 215 204 8877,
E-mail: waring@temple.edu

Abstract

The AnCOB group I intron from Aspergillus nidulans encodes a homing DNA endonuclease called I-AniI which also functions as a maturase, assisting in AnCOB intron RNA splicing. In this investigation we biochemically characterized the endonuclease activity of I-AniI in vitro and utilized competition assays to probe the relationship between the RNA- and DNA-binding sites. Despite functioning as an RNA maturase, I-AniI still retains several characteristic properties of homing endonucleases including relaxed substrate specificity, DNA cleavage product retention and instability in the reaction buffer, which suggest that the protein has not undergone dramatic structural adaptations to function as an RNA-binding protein. Nitrocellulose filter binding and kinetic burst assays showed that both nucleic acids bind I-AniI with the same 1 : 1 stoichiometry. Furthermore, in vitro competition activity assays revealed that the RNA substrate, when prebound to I-AniI, stoichiometrically inhibits DNA cleavage activity, yet in reciprocal experiments, saturating amounts of prebound DNA substrate fails to inhibit RNA splicing activity. The data suggest therefore that both nucleic acids do not bind the same single binding site, rather that I-AniI appears to contain two binding sites.

Abbreviations
ORF

open reading frame

nt

nucleotide

Group I and group II introns frequently contain open reading frames (ORFs), which are either free-standing within the intron itself or are in-frame with the preceding 5′ exon [1,2]. While some of these encode essential host proteins, others have been shown to encode proteins that facilitate the splicing of their cognate introns. These are called maturases [3].

All known group I maturase proteins are characterized by two repeated LAGLIDADG amino acid motifs [4]. Interestingly, maturases are highly homologous to a class of intron-encoded DNA endonucleases (also found in inteins) [2,5–12] that are characterized by one or two copies of the same LAGLIDADG motif [7,8,13], reviewed in [7].

Intron-encoded DNA endonucleases catalyze the mobilization of their cognate intron (or intein) into the equivalent exon sequence of intron-less alleles of the same gene in a process called homing [14]. Significant progress has been made in the past decade in the characterization of homing endonuclease biochemistry and several crystal structures now exist, both with [15–17] and without bound DNA substrate [18–21].

Homing endonucleases containing one LAGLIDADG motif (e.g. I-CreI) are about half the size of those with two copies and structural analysis has shown that they function as homodimers [21]. Double motif-containing endonucleases, including the intein-encoded endonucleases PI-SceI and PI-PfuI as well as the archael intron-encoded protein I-DmoI, were crystallized as monomers [19,20]. Molecular modeling and crystal structure studies suggest that single-motif, homodimer and double-motif, monomeric LAGLIDADG homing endonucleases contain one DNA-binding site and share roughly the same extended overall structure with either two- or pseudo twofold symmetry [19–21], reviewed in [8,13].

The phylogenetic distribution of LAGLIDADG homing endonucleases is widespread [7] but that of maturases is less well known. Introns from Saccharomyces cerevisiae encode proteins with either maturase or endonuclease activity, but not with both activities [2]. However they are clearly closely related [22–25] and in vivo assays have shown that the Saccharomyces capensiscobi2 intron-encoded protein is both an endonuclease and a maturase [26].

There is evidence that endonuclease ORFs, acting as the minimal agent of mobility, invaded group I introns [27,28]. The parsimonious argument follows that this eventually conferred mobility upon the host introns and that maturase activity subsequently evolved from some endonucleases [2,5] thus facilitating intron transposition to new sites [9–12,14,29].

The AnCOB group IB intron from the apocytochrome b gene in Aspergillus nidulans self-splices in vitro, providing that the MgCl2 concentration is >25 mm[29]. We have shown that AnCOB encodes a maturase protein with two LAGLIDADG motifs that specifically and significantly facilitates AnCOB splicing in Mg2+ concentrations as low as 2 mm[30]. Previous genetic evidence indicated that the AnCOB intron is mobile [31] and we have shown since then that the A. nidulans AnCOB-encoded maturase is also a DNA endonuclease [30]. According to homing endonuclease convention, the protein is called I-AniI. One other in vitro maturase-assisted splicing assay has been developed recently for the yeast mitochondrial bI3 intron ORF, but in this interesting case, the maturase requires the assistance of a nuclear-encoded protein and lacks endonuclease activity [32]. To our knowledge, I-AniI is the only protein with which one can biochemically assay both DNA endonuclease and RNA maturase activities in vitro. This investigation provides the first step in the study of the relationship between these two distinct activities.

Experimental procedures

Expression of I-AniI

I-AniI was expressed and purified as described previously [30]. Purified I-AniI was stored at −20 °C in protein storage buffer [50 mm Tris, pH 8, 100 mm KCl, 1 mm dithiothreitol and 50% (v/v) glycerol]. Unless otherwise noted, all reaction components are indicated at their final concentrations. 1 nm I-AniI has been defined previously in our laboratory as the concentration of protein that gives a burst of 1 nm RNA products in an RNA-splicing reaction performed under multiple-turnover conditions [30]. This definition was instituted because similar assays resulted in RNA/protein ratios of 1 : 1 to 2 : 1 when the concentrations of different protein preparations were determined by the Bradford Assay. Subsequent precise calibration using multiple-turnover RNA splicing assays (see below) preserves uniformity between different protein preparations. Throughout this work, only the calibrated protein concentration was used.

Preparation of nucleic acid substrates

The standard DNA substrate, pCOBLE, was generated previously [30]. A preparative amount of pCOBLE plasmid DNA was purified over a single CsCl centrifugation density gradient. Ten micrograms of BsaHI-linearized pCOBLE was end-labeled with 20–50 µCi (800–3000 Ci·mmol−1) [α-32P]dCTP (New England Nuclear, Boston, MA, USA) and unincorporated nucleotides were removed using a P-30 spin column (Biorad, Hercules, CA, USA) after organic extraction. The concentration of DNA was measured spectrophotometrically. The following pairs of oligonucleotides (Integrated DNA Technologies, Coralville, IA, USA) were annealed together to make recognition site variants containing 5′-EcoRI and 3′-HindIII sticky ends: AnI19R 5′-AATTCATGAGGAGGTTTCTCTGTAACA-3′; AnI19H 5′-AGCTTGTTACAG-AGAAACCTCCTCATG-3′; AnI17R 5′-AATTCACGAGGAGGTTTCTCTGTACTA-3′; AnI17H 5′-AGCTTAGTACAGAGAAACCTCCTCG TG-3′; AnI15R 5′-AATTCACAGGAGGTTTCTCT-GTC TA-3′; AnI15H 5′-AGCTTAGACAGAGAAACCTCC TGTG-3′. Each annealed oligonucleotide pair was subcloned into the equivalent sites of pIBI24 to generate LE19, LE17 and LE15, respectively. Point mutations were made by altering the sequence of the pair of AnI19 oligonucleotides as required. The plasmid construct, pCOBsal, used to transcribe the AnCOB RNA substrate, was generated previously [29]. Either PvuII- or HindIII-linearized pCOBsal run-off RNA transcripts were generated, purified by denaturing PAGE and their concentrations quantitated by liquid scintillation counting as previously described [33]. The pre-RNA derived from a PvuII-linearized pCOBsal DNA template was 632 nucleotides (nt) in length and contained 311 nt of intron sequence, 112 nt of 5′ exon and 209 nt of 3′ exon sequence. HindIII-derived pre-RNAs contained a shorter 3′ exon, 29 nt in length.

Endonuclease cleavage assays

The standard endonuclease reaction was performed in TK9 buffer (50 mm Tris, pH 9, 50 mm KCl, 1 mm dithiothreitol). Except where indicated, TK8 buffer (50 mm Tris, pH 8, 50 mm KCl, 1 mm dithiothreitol) was used for experiments containing both DNA and RNA. For single-turnover protein in excess cleavage reactions, 10 nm I-AniI were mixed with 1 nm end-labeled pCOBLE in TK9 buffer at 37 °C for 2 min. Reactions were started with MgCl2 at a final concentration of 10 mm. Using the Marquardt–Levenberg algorithm and nonlinear regression analysis (PSI-Plot), single-turnover endonuclease data were fit to a single exponential, Fpre = Aekt + (1 − A). Fpre represents the fraction of DNA remaining, A is the amplitude, k is the first-order rate constant, kobs, and (1-A) represents the fraction of unreacted DNA. All DNA cleavage reactions were quenched in 6 × stop buffer (1% SDS, 100 mm EDTA, 0.25% bromophenol blue, 30% glycerol) and reaction products were separated on 1% agarose gels, dried under vacuum at 90 °C and were visualized by autoradiography.

RNA splicing assays

Single- and multiple-turnover protein-assisted RNA splicing assays were performed in TK8 buffer containing 5 mm MgCl2, 0.2 U RNA-Guard (Pharmacia, Piscataway, NJ, USA) and 0.5 mm guanosine as described previously [30,33]. Note that all RNAs studied in this investigation were uniformly labeled. Therefore the yield of each RNA species in splicing reactions was corrected for the number of uridines present.

Endonuclease optimization experiments

All endonuclease optimization experiments were performed using a subsaturating concentration (4 nm) of I-AniI and 1 nm end-labeled pCOBLE. All reactions were quenched during a time frame in which product formation varied exponentially with time and no more than 50% of the starting material reacted to further ensure that each determination was sensitive to minor changes in reaction rate. Optimization experiments were performed in TK9 buffer containing 10 mm MgCl2 at 37 °C with one component varied as required. Fifty millimolar Mes replaced 50 mm Tris for pH optima experiments performed at pH < 7.

Nitrocellulose filter binding assays

To determine the degree to which DNA and RNA saturate I-AniI, 1 nm I-AniI was bound to varying concentrations of end-labeled pCOBLE or uniformly labeled AnCOB pre-RNA for 5 min at 37 °C in TK8 buffer containing either 2 mm CaCl2 or 5 mm MgCl2 for DNA or RNA substrates, respectively. Preliminary experiments indicated that equilibrium was reached during that time. Samples, in triplicate, were subsequently filtered through pre-wet nitrocellulose filters and were quantitated as described previously [33]. To estimate the stoichiometry of RNA and DNA binding, the data were analyzed as described [34]. To determine the dissociation rate constant of the DNA substrate, 2 nm I-AniI were preincubated at 37 °C for 10 min in TK8 buffer containing 2 mm CaCl2 (this inhibits DNA cleavage) and 1 nm end-labeled pCOBLE. Reactions were then diluted 20-fold in a similar buffer containing 28 nm linearized, unlabeled pCOBLE and the release of the labeled DNA was followed over time using a nitrocellulose filter binding assay as described above. A control reaction that did not contain a chase was also performed. Adding both DNAs simultaneously gave a negligible signal above background.

In vitro competition assays

RNA splicing and DNA cleavage competition experiments both involved a prebinding step in which either DNA or RNA substrates were incubated with I-AniI in a binding reaction containing either 1.5× or 1.1× the final concentration of each reaction component, respectively. RNA splicing and DNA cleavage reactions were subsequently started with the missing reaction components in a volume sufficient to dilute all the reaction components to their final concentrations. When an RNA inhibitor was included in an endonuclease reaction, it masked the 1.025 kb cleavage product. Therefore, since the DNA substrate was end-labeled, endonuclease reaction products were quantified by multiplying the yield of the 1.912 kb cleavage product by two. To preserve uniformity, control endonuclease experiments, without competitor RNA, were quantified in the same way. All pre-RNAs were derived from PvuII-linearized DNA templates except for indicated experiments presented in Fig. 5C,D where HindIII-linearized DNA templates were used [33]. Aliquots removed from splicing reactions were analyzed as previously described [33].

Figure 5.

Pre-bound RNA substrate stoichiometrically inhibits DNA cleavage. (A) I-AniI (3 nm) was incubated with or without uniformly labeled AnCOB pre-RNA at the indicated concentrations for 5 min at 37 °C in TK9 buffer containing 10 mm MgCl2. DNA cleavage reactions were subsequently started with 1 nm end-labeled pCOBLE (see Experimental procedures for details). (B) Control experiments to demonstrate that cleavage of 1 nm end-labeled pCOBLE substrate DNA is sensitive to the concentration of I-AniI. The ordinate represents the fraction of substrate DNA that remained after a 2.5 min cleavage reaction. (C) Control experiments to demonstrate the specificity of endonuclease inhibition by prebound AnCOB RNA. The assay of 5A was repeated with and without 5 nm AnCOBΔP9 pre-RNA and with KCl at 50, 100 and 150 mm. Note that increasing KCl concentration slows the cleavage of the DNA substrate in the absence of RNA (see also Fig. 2D) and that the addition of competitor RNA to an endonuclease reaction masked the 1.025 kb DNA cleavage product (see Fig. 1B and Experimental procedures for details). (D) DNA cleavage reactions with and without prebound AnCOBΔP9 (data derived from panel C). (E) Substrate RNA does not inhibit endonuclease activity when the DNA substrate is prebound. I-AniI (8 nm) was incubated with 0.8 nm end-labeled DNA substrate at 37 °C in TK8 buffer containing 10 mm MgCl2. After 0.5 min the reaction was diluted 20-fold (arrow) into a chase buffer containing either 10 nm prelinearized DNA substrate (▪) or 12 nm unlabeled pre-RNA substrate (▴). A control reaction that was not diluted in a chase buffer, but was left to react to completion is indicated (●). No end-labeled DNA substrate reacted when the reaction was started under chase conditions.

Results

I-AniI recognition site determination

The standard endonuclease substrate, pCOBLE [30], consists of 162 bp of exon sequence spanning from −97 bp to +65 bp with respect to the AnCOB intron insertion site (Fig. 1A). When incubated at 37 °C with 10 nm I-AniI in TK9 buffer containing 10 mm MgCl2, end-labeled pCOBLE (1 nm) was specifically cleaved into two reaction products, 1.912 kb and 1.025 kb in length (Fig. 1B). There was no difference in reaction rates when the protein concentration was doubled indicating that the DNA substrate was saturated with protein (data not shown). Typically, >95% of the starting material reacted in 5 min under these conditions, yielding an average maximum single-turnover rate constant (kobs(max)) of 2.5 min−1 over several different protein preparations (Fig. 1B).

Figure 1.

I-AniI recognition site determination. (A) I-AniI recognition site. 30 bp of AnCOB exon sequence flanking the intron insertion site (arrow) are shown. The cleavage site is indicated with a staggered line. The boundaries of the three truncation mutants, LE19, LE17 and LE15 are indicated above. Residues that were mutated to the corresponding VinCOB sequence (Table 1) are indicated in lowercase. (B) Typical DNA cleavage reaction under single-turnover conditions. I-AniI (10 nm) was incubated with 1 nm end-labeled pCOBLE in TK9 buffer containing 10 mm MgCl2 at 37 °C. (C) Single-turnover, subsaturating endonuclease cleavage reactions with varying DNA substrates in TK9 buffer containing 10 mm MgCl2 and 10% glycerol. Reactions containing 6 nm I-AniI and 0.3 nm DNA are indicated (●, ◆, ▴, ▪). A control reaction with 33% less I-AniI (4 nm) and 0.2 nm pCOBLE (□) reacted 24% slower, indicating that protein concentration was subsaturating.

Previous dideoxynucleotide sequencing studies mapped the boundaries of the I-AniI recognition sequence to approximately 20 bp [30]. However, in those studies the recognition sequence was located at the end of the DNA substrate. In this study, we set out to more precisely determine the minimum sequence cleaved by I-AniI. We therefore generated three successively shorter DNA substrates in pIBI24 (LE19, LE17 and LE15) that contain 19, 17 and 15 bp of AnCOB exon sequence surrounding the I-AniI cleavage site (Fig. 1A). To avoid inadvertently extending the size of the desired recognition sequence the oligonucleotides were designed to ensure that 7 bp of sequence flanking the truncated recognition sequence had minimal similarity to the omitted native sequence.

DNA cleavage reactions were performed under single-turnover conditions with protein in excess, but limiting concentrations. Under such conditions, reduction in either binding or catalytic proficiency should be reflected by a concomitant decrease in reaction rate. The LE19 construct supported significant DNA cleavage activity yielding a corresponding rate constant approximately 24% of that observed with the standard 162 bp DNA substrate, pCOBLE (Fig. 1C, Table 1). Only trace DNA cleavage activity was observed when the LE17 construct was evaluated as substrate (Fig. 1C). The LE15 construct showed no detectable activity even in the presence of a 30-fold increase in protein concentration.

Table 1. I-AniI recognition site sequence specificity. Experiments were performed using end-labeled pCOBLE variants (lowercase letters in Fig. 1A) as substrates for DNA cleavage reactions under single-turnover conditions with subsaturating protein concentrations, as described in the legend to Fig. 1C. The primary data were fit to a single exponential as described in Experimental procedures. Relative activity (with respect to LE19) reflects the average of two independent determinations.
ConstructRelative activity
pCOBLE4.10
LE191.00
LE19A-8G<0.02
LE19T-2C0.25
LE19C+2T0.17
LE19T+5A0.12
LE19A+8G0.10

In general, homing endonucleases typically have large recognition sequences and consequently tolerate a wide variety of point mutations. However, bifunctional maturase/endonuclease proteins such as I-AniI might display a different pattern of tolerance compared to other homing endonucleases if they evolved to accommodate both RNA and DNA in the same binding site, as suggested previously [6,23]. Therefore, to address the sequence specificity of I-AniI, several individual point mutations were generated that correspond to sequence found in the homologous VinCOB gene from Venturia inaequalis[35] (Table 1).

To increase the sensitivity toward decreased binding affinity of AnCOB/VinCOB chimeric point mutants, I-AniI sequence specificity was characterized within an LE19 background as that construct is already partially impaired in binding. The observed reaction rates for most point mutants were reduced four- to 10-fold, compared to the LE19 reference construct (Table 1). By contrast, the identity of the base pair at position −8 is critical; <2% relative activity was observed when LE19A-8G was evaluated as a DNA substrate. The data therefore indicate that I-AniI is typical in its tolerance to mutation in its recognition site.

The in vitro biochemical properties of I-AniI

In this study we evaluated the biochemical properties of the I-AniI endonuclease reaction to first address whether acquiring maturase activity had significantly altered the endonuclease characteristics of I-AniI, compared to other homing endonucleases, and secondly to establish conditions whereby maturase and endonuclease activities could be studied simultaneously.

To determine optimal conditions for DNA cleavage by I-AniI, MgCl2 concentration, pH, temperature and ionic strength were varied systematically and their effects on pCOBLE cleavage were assessed under single-turnover conditions with a limiting concentration of protein. As with all known homing endonucleases, Mg2+ is an essential cofactor for I-AniI endonuclease activity. I-AniI activity was optimal in approximately 12.5 mm MgCl2, but when MgCl2 was omitted from the reaction, no cleavage was observed (Fig. 2A). Two additional group IIa divalent cations (Mn2+ and Ca2+) were evaluated. Mn2+ substituted for Mg2+ with similar optima trends, although the absolute activity was lower than with Mg2+. In contrast, 2 mm Ca2+ did not support cleavage and was completely inhibitory in 10 mm MgCl2[36].

Figure 2.

Optimization of I-AniI endonuclease cleavage reaction. Four parameters (A) MgCl2 concentration, (B) pH, (C) temperature and (D) KCl concentration, were evaluated for their effects on pCOBLE cleavage by I-AniI at 37 °C under single-turnover conditions with subsaturating protein concentrations. Relative activity in each panel was normalized to a reaction showing the greatest product accumulation at a single time point. Unless otherwise noted, all experiments contained TK9 buffer with 10 mm MgCl2 and were performed at 37 °C.

The pH optimum for DNA cleavage activity was approximately 9 (Fig. 2B). Interestingly, at the physiological pH of 7.5 in yeast mitochondria [37], the relative activity was only around 30%. No DNA cleavage activity was observed below pH 6.

I-AniI exhibited a broad temperature optimum between 45 and 60 °C (Fig. 2C). In the physiological temperature range, roughly 55% activity relative to the maximum was observed. I-AniI was completely inactivated after 2 min at 65 °C [36].

Although not absolutely necessary for catalytic activity, monovalent cations are required for efficient DNA cleavage activity. I-AniI activity was optimal in 25 mm KCl (Fig. 2D). The relative endonuclease activity was only about 10% when KCl was omitted from the reaction mix. Other monovalent salts were also evaluated (NH4Cl, NaCl) and similar trends were observed (data not shown) although lower relative activities were observed compared to KCl.

These studies revealed that the buffer used previously to characterize I-AniI maturase-assisted RNA splicing [30,33,38] was unsuitable for DNA cleavage by I-AniI. Therefore, TK8 buffer was chosen to simultaneously study maturase and endonuclease activities, apart from experiments the results of which are shown in Fig. 5A–D.

Many group I intron-encoded endonucleases are unstable under standard assay conditions but some can be stabilized either by target site DNA [39] or by nonspecific DNA [37,40,41]. Moreover, Mg2+ has been shown to partially stabilize at least one intron-encoded endonuclease [39]. Therefore, we evaluated I-AniI endonuclease stability in TK9 buffer at 37 °C.

When incubated alone in the standard reaction buffer without MgCl2, around 50% of endonuclease activity was lost within approximately 15 min (Fig. 3). The addition of 10 mm MgCl2 to the preincubation mix stabilized endonuclease activity, with 50% activity lost in about 45 min. The substitution of either Na+ for K+ or acetate for Cl ions, as well as the inclusion of 0.1 mg·mL−1 BSA did not increase the stability of the protein and slowed the reaction rate (data not shown). The inclusion of 5–10% glycerol increased stability two- to fivefold, depending on the preparation of protein, although its inclusion slows the reaction rate by about 30% (data not shown). Strikingly, preincubation with linearized pCOBLE DNA substrate (without MgCl2) preserved, upon extrapolation, 50% activity for approximately 2 h. Pre-incubation with linearized nonspecific (vector) DNA did not detectably stabilize the protein (data not shown).

Figure 3.

I-AniI stability in endonuclease reaction buffer. A subsaturating concentration (3 nm) of I-AniI was incubated at 37 °C alone in TK9 buffer with no additional component, with 10 mm MgCl2 or with 1 nm end-labeled pCOBLE. DNA cleavage reactions were subsequently started with the addition of the missing reaction component. Ordinate values were calculated by normalizing the amount of product formed to a control reaction that was not preincubated.

RNA and DNA substrates bind I-AniI with the same stoichiometry

To determine the relative stoichiometry of RNA and DNA binding, nitrocellulose filter binding assays were performed (Fig. 4A). Both DNA and RNA substrates saturated limiting amounts of protein in the same manner with a stoichiometry close to 1 : 1. Two independent determinations were performed for each nucleic acid substrate and yielded an average ratio of DNA/protein of 1.09 : 1 and an average ratio of RNA/protein of 1.12 : 1.

Figure 4.

DNA and RNA substrates saturate I-AniI with the same stoichiometry. (A) Nitrocellulose filter binding assay in TK8 buffer with 1 nm I-AniI and varying concentrations of end-labeled pCOBLE or uniformly labeled AnCOB pre-RNA. RNA binding reactions contained 5 mm MgCl2 and DNA binding reactions contained 2 mm CaCl2. Both determinations were performed in triplicate and were made using the same diluted aliquot of the same protein preparation. (B) Multiple-turnover endonuclease cleavage assay in TK9 buffer containing 10 mm MgCl2. Reactions contained 3 nm I-AniI and either 30 nm or 60 nm end-labeled pCOBLE.

Multiple-turnover (substrate in excess) DNA cleavage reactions were also performed to estimate the stoichiometry of DNA binding (Fig. 4B). In those experiments, two different concentrations of end-labeled pCOBLE were incubated with 3 nm I-AniI. When cleavage reactions were started with MgCl2, a small rapid burst was observed. This was followed by a much slower phase, which is believed to result from slow release of the cleavage products from the protein. The amplitudes of the initial burst (2.78 and 2.83 nm) gave a ratio of DNA/protein of 0.94 : 1. As will be discussed further, these data indicate that both RNA and DNA bind I-AniI with a ratio of 1 : 1.

Pre-bound RNA substrate inhibits endonuclease activity

It has been hypothesized that bifunctional maturase/endonuclease proteins utilize the same binding site for DNA and RNA substrate binding [2]. Indeed the RNA helices that flank the splice sites of group I introns (P1 and P10) are similar in sequence to part of the endonuclease recognition sequence [42]. However, chemical mapping studies [38] as well as RNA mutational analysis [33] indicated that I-AniI binds multiple RNA domains within the AnCOB intron. This suggests that the overall tertiary structure of I-AniI may differ significantly from that of other characterized homing endonucleases and raises the possibility that I-AniI may actually contain more than one nucleic acid binding site.

We previously developed a novel system to evaluate the binding of AnCOB RNA mutants to I-AniI by using them as specific inhibitors of protein-assisted native AnCOB splicing at low Mg2+ concentrations [33]. In this study, we hypothesized that if both nucleic acid substrates share one binding site, then maturase or endonuclease activity could be blocked by a specific nucleic acid inhibitor. For the experiments discussed below, varying amounts of either AnCOB pre-RNA or linearized pCOBLE DNA were prebound to I-AniI and either DNA cleavage or protein-assisted RNA splicing activity was subsequently evaluated.

With increasing (subsaturating) concentrations of prebound, wild-type, competitor AnCOB pre-RNA, DNA cleavage by I-AniI was stoichiometrically inhibited (Fig. 5A), reaching maximal inhibition with a saturating (3 nm) concentration of AnCOB RNA. The sensitivity of the assay was such that 0.1 nm residual activity would have been detected (Fig. 5B). Inhibition was not due to nonspecific binding of degraded RNA as RNAs from endpoint aliquots were completely intact when resolved on denaturing polyacrylamide gels. Inhibition was also not due to trapping of the DNA substrate by the RNA, because increasing the DNA concentration from 1 to 6 nm (with 3 nm RNA) did not lead to any DNA cleavage. Tight stoichiometric binding of RNA to protein is consistent with a Kd < 10 pm under similar conditions, but at pH 7.5 [38].

Additional control experiments were performed to determine the specificity of endonuclease inhibition by prebound RNA. AnCOBΔP9 pre-RNA lacks the short (18 nt) stem loop P9 and acts as a weak competitive inhibitor of RNA splicing [33], but when 5 nm AnCOBΔP9 pre-RNA was prebound to 3 nm I-AniI as described above in TK9 buffer, it significantly inhibited DNA cleavage activity (Fig. 5C,D). As RNA splicing inhibition studies were performed in 100 mm KCl [33], we re-evaluated AnCOBΔP9 under more stringent conditions. When the concentration of KCl was increased from 50 mm to 150 mm, there was only limited inhibition of DNA cleavage by AnCOBΔP9 pre-RNA (compare open symbols in Fig. 5D). By contrast, there was complete inhibition by the intact AnCOB pre-RNA in 150 mm KCl (data not shown), indicating that the inhibition was specific. Another deletion mutant preRNA, AnCOBΔP9.1, with similar binding affinity to AnCOBΔP9 and a trace of splicing activity [33], also only slightly inhibited endonuclease cleavage in 150 mm KCl whereas an inactive deletion mutant AnCOBΔP5aiib (described in [33]), which bound less tightly to I-AniI [33], inhibited DNA cleavage poorly even in 50 mm KCl. The second group I intron from the A. nidulans cytochrome oxidase gene (NOX2), which does not bind I-AniI [30], did not detectably inhibit DNA cleavage in 50 mm KCl; nor did a 224 nt RNA transcribed from the transcription vector (pSP65) (data not shown). In general, when analyzed at a suitably stringent concentration of KCl, there was a correlation between inhibition of endonuclease activity and inhibition of splicing activity indicating that RNA binding can be monitored by measuring inhibition of the endonuclease reaction [36].

In the above experiments, the RNA was prebound to I-AniI before addition of the DNA substrate. To determine whether the RNA substrate can inhibit the cleavage of prebound DNA substrate, a single-turnover DNA cleavage reaction was incubated at 37 °C (Fig. 5E) for 0.5 min, during which time 15% (0.12 nm) of the DNA reacted. The reaction was subsequently diluted into a chase buffer containing a large excess of either unlabeled AnCOB pre-RNA or unlabeled prelinearized DNA. The control reaction, performed with excess DNA, which prevents any subsequent binding of labeled DNA, showed that a significant fraction of labeled DNA remains bound long enough to react. This fraction is not detectably reduced by the presence of excess RNA indicating that the RNA cannot inhibit the cleavage of the DNA if the DNA is already bound. By contrast, when 10 nm end-labeled DNA and 3 nm AnCOB pre-RNA were simultaneously added to a limiting concentration of protein (2 nm) in the same buffer conditions, <1% DNA cleavage was detected (data not shown). The significance of this will be discussed later.

Pre-bound DNA substrate does not inhibit maturase activity

The data derived from the endonuclease competition experiments were consistent with the simplest hypothesis that both nucleic acid substrates compete for the same single discrete binding site, as the fraction of cleaved substrate DNA varied inversely with competitor RNA concentration. However, in striking contrast to the results described above, when 50-fold excess (over protein) pCOBLE DNA was prebound to I-AniI, the rate of protein-assisted splicing of AnCOB RNA was not significantly different from the rate of a reaction that did not contain competitor DNA (Fig. 6A,B). Within the first 0.25 min, around 42% (0.1 nm) of the pre-RNA reacted with or without prebound DNA. We investigated the robustness of this experiment by performing it independently a total of three times using two different preparations of protein and three different preparations of DNA and RNA substrates. Each time the splicing reaction was minimally affected by prebinding the DNA to protein.

Figure 6.

Pre-bound DNA substrate fails to inhibit RNA splicing. (A) I-AniI (2 nm) was incubated at 37 °C with or without 100 nm pCOBLE prelinearized DNA for 5 min in TK8 buffer containing 2 mm CaCl2, which inhibits endonuclease activity. RNA splicing assays were subsequently started with the addition of 0.25 nm uniformly labeled AnCOB pre-RNA, 5 mm MgCl2 and 0.5 mm guanosine (see Experimental procedures). To confirm that reaction rates were dependent on protein concentration, control protein-assisted RNA splicing assays, without competitor DNA, were performed with gradually decreasing concentrations of I-AniI. A control reaction with 0.5 nm I-AniI is shown. Separate control experiments demonstrated that self-splicing (in the absence of protein) does not occur in this reaction buffer (data not shown). (B) Protein-assisted RNA splicing reactions with and without competitor DNA substrate (derived from panel A). (C) Control protein-assisted RNA splicing reactions, without competitor DNA substrate (derived from panel A). Note that the reaction containing 2 nm I-AniI is the same as in panel B. (D) The DNA substrate dissociates very slowly from I-AniI. Nitrocellulose filter binding assays were performed as described in Experimental procedures. Binding reactions with and without a DNA chase are indicated as ● and ○, respectively.

As the concentration of protein used, 2 nm, yields a rate of splicing which is about 25% that of the maximal rate [33] and is therefore subsaturating, any reduction in reaction rate caused by prebinding the DNA was expected to be clearly detectable. Control protein-assisted RNA splicing assays, without competitor DNA, confirmed that the reaction rate was indeed dependent on protein concentration under the conditions of the competition assays (Fig. 6C). A comparison of Fig. 6B,C indicates that if protein-dependent splicing requires dissociation of bound DNA, then in order for the reaction with prebound DNA to react at the observed rate >1 nm free protein must have become available within <0.5 min (▪, Fig. 6B).

To estimate the amount of free protein in the above experiment, we determined the rate at which the pCOBLE DNA inhibitor dissociated from the protein using a nitrocellulose filter binding assay. For those experiments, 1 nm labeled DNA substrate was prebound to 2 nm protein under the reaction conditions of the inhibition experiment (Fig. 6). The rate of the release of bound, labeled DNA was then followed after adding excess unlabeled, pCOBLE DNA (Fig. 6D). A koff of 0.07 min−1 was measured, nearly two orders of magnitude lower than the splicing reaction (Fig. 6B) arguing that very little of the RNA reacts with free protein and that the majority reacts with protein bound to DNA. The data are therefore consistent with the hypothesis that both nucleic acid substrates do not bind the same discrete binding site, but suggest that I-AniI contains either two discrete, or overlapping binding sites. This will be discussed below.

Discussion

The biochemical properties of I-AniI are typical of homing endonucleases

The recognition site length requirement as well as sequence specificity of other homing endonucleases have been determined [39,43,44]. In this study, the minimum sequence cleaved by I-AniI (19 bp) is nearly two turns of the double helix, which is consistent with the X-ray crystal structure for I-CreI [15,16]. Despite that extended length requirement, I-AniI tolerates substitution in its recognition sequence (Table 1). The sample of I-AniI DNA substrate mutations show a similar trend to those reported elsewhere, with the majority having a moderate effect and the occasional mutation (e.g. LE19A-8G) very significantly inhibiting cleavage. This indicates that I-AniI has not lost the ability to relax specificity, the hallmark of a homing endonuclease, in order to function as a maturase. Moreover, the elevated pH optimum, Mg2+-dependence and instability of I-AniI in the reaction buffer are typical properties of many group I intron-encoded endonucleases [36,39,40,43]. Furthermore, multiple-turnover DNA cleavage experiments yielded a rate constant (0.25 min−1) (Fig. 4B) approximately 10-fold lower than that observed under single-turnover conditions with a saturating protein concentration, suggesting that product release is rate-limiting for the cleavage reaction. This is consistent with observations made with other homing endonucleases that remain bound to either the 3′ exon cleavage product (e.g. I-SceI [45]), or remain bound to both 5′ and 3′ exon cleavage products (e.g. I-CreI [40]). Together, the available data indicate that I-AniI acquired the ability to function as an RNA-binding protein without compromising basic homing endonuclease properties.

The high pH reaction profile (Fig. 2B) displayed by many, if not all, of the LAGLIDADG enzymes may arise from the use of bulk solvent rather than specific enzyme side chains as a direct general base for activation of the nucleophilic water [13,15]. A highly diverse collection of basic side chains surround a large solvent pocket which, in turn, surrounds the active site. With such a site, activation of the water nucleophile would be expected to require a higher pH than in the case of a canonical restriction endonuclease that directly deprotonates the same group. One possible advantage for homing endonucleases of such a reaction mechanism would be an enhanced ability to retain activity during evolution (because relatively few side chains are absolutely essential for catalysis) [13]. The cost to the enzyme of this evolutionary advantage would be a pH optimum that is elevated relative to the surrounding cellular environment.

RNA and DNA substrates bind I-AniI with the same stoichiometry

Nitrocellulose filter binding assays, performed side by side for RNA and DNA (Fig. 4A) as well as burst size measurements from multiple-turnover DNA cleavage experiments (Fig. 4B) and stoichiometric inhibition of DNA cleavage by prebound RNA substrate (Fig. 5A) all argue that both nucleic acid substrates bind I-AniI with the same stoichiometry. As X-ray crystallography studies indicate that a homing endonuclease with two LAGLI DADG motifs binds a single DNA molecule [17], our data argue that RNA binds I-AniI with a stoichiometry of 1 : 1. This is consistent with Solem et al. (2001) who assayed I-AniI RNA binding using a nitrocellulose filter binding assay and measured protein concentration spectrophotometrically at 280 nm [46]. Together, these data justify the multiple-turnover RNA splicing calibration that we made to each preparation of protein (see Experimental procedures) to give a 1 : 1 stoichiometric burst of spliced RNA. We note however, that the active splicing complex could theoretically consist of two RNA and two protein molecules with each RNA using two separate domains to bind to a different region on the two protein subunits.

Relationship between endonuclease and maturase activities

It is generally believed that LAGLIDADG proteins were originally DNA endonucleases [2,5] (see Introduction). The fundamental question then arises as to how I-AniI acquired maturase activity and how in general, proteins might acquire a new function. It is apparent that as I-AniI acquired the property to function as an RNA maturase, it maintained properties common to other homing endonucleases suggesting that there have been no dramatic structural changes in the vicinity of the DNA-binding site that constitutes a significant portion of LAGLIDADG proteins [16]. Even though the protein-assisted splicing reaction rate is at least 10 times the rate of self-splicing at the optimal concentration of Mg2+[30] it would seem most likely that the maturase activity consists primarily of an RNA binding site but not a catalytic site. Because the 7 bp helix spanning the 5′ splice site (P1) shares partial sequence similarity with the DNA endonuclease recognition site, one might predict that I-AniI co-opted a pre-existing nucleic acid (DNA) binding site to function as an RNA binding protein [23]. However, previous RNA deletion analysis argues that the protein is recognizing considerably more than just a single short helical region [33]. As will be discussed below, the available data are consistent with a two binding site model.

To explore the relationship between endonuclease and maturase activities, we utilized a novel in vitro competition model system developed in our laboratory [33]. The competition assays described in this investigation demonstrated that prebound RNA substrate efficiently inhibited DNA cleavage activity only when prebound to I-AniI (Fig. 5A,E). The failure of pre-RNA to interrupt a DNA cleavage reaction when added shortly after it begins (Fig. 5E) argues that DNA cleavage can still take place in the presence of pre-RNA. Control RNA splicing reactions showed that the RNA at a concentration of 12 nm would have had sufficient time to bind in this experiment (data not shown). The rate of the splicing reaction in Fig. 6B, in which only 2 nm RNA is present, supports this statement and suggests that the RNA binding site is still available and its binding affinity is not altered by the presence of DNA in the DNA binding site.

Interestingly, when 3 nm RNA and 10 nm DNA were added simultaneously to a limiting amount of protein in TK8 buffer containing 10 mm MgCl2, DNA cleavage was not detected (data not shown). The rate constant, kon, for the binding of RNA to protein is equal to 3 × 109 m−1·min−1 under similar conditions [38] to these. We measured the rate constant for the binding of DNA to protein kinetically and obtained a value of at least half this in TK8 buffer containing 10 mm MgCl2[33]. Therefore, given that the DNA·protein complex is more likely to react than dissociate (Fig. 5E), theoretically the DNA substrate should have competed effectively with the RNA when added simultaneously if both nucleic acids share a single binding site. However the protocol used to measure kon would capture, in a given period of time, every binding event to any region of DNA that eventually leads to cleavage if the protein were to scan the DNA for the recognition site. If DNA cleavage involves a two-step process such as this, the pre-RNA could inhibit the former, but not the latter step by binding to a separate, RNA specific site. It is possible that when RNA binds (to its own site) immediately after I-AniI has bound nonspecifically to DNA, it prevents the protein from scanning to its recognition site, but does not displace or inhibit cleavage of the DNA if I-AniI has already reached the recognition site (Fig. 5E). When a saturating amount of RNA substrate was bound before the DNA substrate, DNA cleavage was completely inhibited (Fig. 5A). Assuming a two site model, this could be because DNA sliding is blocked as just discussed or it may suggest that the RNA sterically interferes with DNA binding or allosterically inhibits DNA cleavage. Further experiments are warranted to address those important issues.

The discussion above is based on the assumption that both substrates can bind simultaneously to I-AniI providing that the DNA substrate binds first. This is supported by the fact that prebound DNA substrate failed to inhibit RNA splicing, even when in 50- and 400-fold excess over protein and RNA, respectively (Fig. 6B). Interpreting the data as being inconsistent with a single binding site model assumes that the 50-fold excess of DNA nearly saturates the protein prior to the addition of pre-RNA. It further assumes that either the dissociation of bound DNA is slow or the 400-fold excess (over RNA) free DNA effectively competes with the labeled pre-RNA to prevent it from binding.

Previous nitrocellulose filter binding assays gave an apparent Kd for DNA and protein of 0.1 nm in the same reaction conditions as Fig. 6[36]. The control experiment using a chase of excess unlabeled DNA as presented in Fig. 5E can be used to estimate the rate of dissociation and shows that labeled DNA dissociates slowly from I-AniI, with a kobs of 0.17 min−1, which is similar to the value of koff = 0.07 min−1 as determined by nitrocellulose filter binding assays (Fig. 6D). Therefore, these two observations would seem sufficient to suggest that the assumptions outlined above are valid and argues that RNA splicing does not occur as a result of rapid dissociation of the protein·DNA complex. Furthermore, the fact that the protein was saturated with the same amount of either substrate (Fig. 4) argues against the trivial explanation that only a fraction of protein was competent to bind DNA, leaving another fraction free to react with the RNA.

The control experiment with unlabeled DNA chase as performed in Fig. 5E as well as experiments performed to determine the apparent Kd by filter binding were carried out with protein in excess. This was not the case for the experiments in Fig. 4A and Fig. 6D. When the protein is in excess, this can lead to an overestimation of the strength of binding because protein molecules bound nonspecifically to DNA adjacent to the recognition sequence can replace a protein molecule that has dissociated from the recognition site. These limitations do not seem sufficient to explain the complete inability of a saturating amount of DNA substrate to inhibit RNA splicing (Fig. 6B) and therefore it seems reasonable to make the working hypothesis that the RNA and DNA binding sites are not the same, rather that they are separate. This does not exclude the possibility that both nucleic acid binding sites may share some degree of overlap. Careful experimentation to test those hypotheses will need to be performed to determine the binding characteristics of the nucleic acid substrates, both alone and in the presence of one another.

Although the structural details of four endonucleases are known (see Introduction) and none of these are believed to have maturase activity, we are not aware of any study which makes it possible to identify a maturase by examination of amino acid sequence. We attempted to predict the secondary structure of I-AniI and to look for some distinctive feature but this was unproductive. LAGLI DADG homing endonucleases bind their DNA recognition sequence in an extended groove formed by curved antiparallel β-pleated sheets stabilized on their outer surface by α-helices [13,15–21]. Our results suggest that I-AniI will have a similar DNA binding site and that the RNA will bind primarily outside the groove but in such a way that it directly or indirectly inhibits some step in DNA cleavage. Previous results [33,46] indicate that I-AniI recognizes a tertiary component of AnCOB that is dependent upon a relatively intact RNA intron structure. The degree to which the intron and protein had to adapt to fit one another remains to be seen.

Acknowledgements

We thank Dr Sarah Woodson and Dr Barry Stoddard for advice. We also thank Dr Karen Palter for reviewing the manuscript and Dr Mark Caprara for the gift of a preparation of I-AniI. This work was supported in part by a grant (MCB-0130991) from the National Science Foundation, by a Research Incentive Fund award from Temple University and by a grant from the Howard Hughes Medical Institute through the Undergraduate Biological Sciences Education Program.

Ancillary