Hammerhead ribozyme is a versatile tool for down-regulation of gene expression in vivo. Owing to its small size and high activity, it is used as a model for RNA structure–function relationship studies. In the present paper we describe a new extended hammerhead ribozyme HH-2 with a tertiary stabilizing motif constructed on the basis of the tetraloop receptor sequence. This ribozyme is very active in living cells, but shows low activity in vitro. To understand it, we analysed tertiary structure models of substrate–ribozyme complexes. We calculated six unique catalytic core geometry parameters as distances and angles between particular atoms that we call the ribozyme fingerprint. A flanking sequence and tertiary motif change the geometry of the general base, general acid, nucleophile and leaving group. We found almost complete correlation between these parameters and the decrease of target gene expression in the cells. The tertiary structure model calculations allow us to predict ribozyme intracellular activity. Our approach could be widely adapted to characterize catalytic properties of other RNAs.

INTRODUCTION

The knowledge of how building blocks affect an assembly of cellular macromolecules into biologically active structures forms the basis for modelling and designing new objects [1]. To complement the results of various experimental studies, new approaches based on bioinformatics and molecular modelling have been pursued to make predictions of RNA tertiary structures and their biological activities possible. In silico design of definite structure and function of molecules is a major challenge for molecular biology.

The hammerhead ribozyme (HH or HHRz) is the smallest natural catalytic RNA that for many years has been a model for studies of RNA structure–function relationships [2]. It has been shown that the minimal hammerhead ribozyme is the simplest active form that catalyses the transesterification of the phosphodiesther bond in cis as well as in trans [3]. It consists of three helical arms (I, II and III), joined by a highly conserved catalytic core of 11 nucleotides. The cleavage specificity of the hammerhead ribozyme is generally described as the NUH↓ rule, where N can be any nucleotide and H cannot be G [4]. The substrate is located within an active site cleft, surrounded by nucleotides distant in the primary structure [5].

It was found that TSMs (tertiary stabilizing motifs) occurring in natural hammerhead ribozymes play an important role in acquiring the catalytically active conformation [6]. These so-called extended or full-length hammerhead ribozymes are highly active at physiological Mg2+ concentrations in contrast with minimal variants that often require 10 mM Mg2+ for efficient catalysis [7]. In contrast with highly conserved nucleotides of the catalytic site, TSMs of natural hammerheads show many sequence and topology differences [8]. Tertiary interactions are distal to the catalytic core, but alter the geometry of the two intervening helices [9].

A comparison of the X-ray structures of the minimal hammerhead ribozyme [10] and Schistosoma extended hammerhead ribozyme [11] shows that the loop–loop interaction leads to a substantial conformational rearrangement of the catalytic centre [12]. The catalytic core of the extended hammerhead ribozyme shows a Watson–Crick base pair between nucleotides C3 and G8. Although it is not present in the minimal hammerhead ribozyme, HHRz has to acquire a similar structure before cleavage [13]. Computer simulations of a ribozyme structure and its mutants suggested that Watson–Crick pair C3–G8, hydrogen bond network between C17 and G5 and base stacking of G8 and C1.1 are crucial for keeping the conformation active [14]. These interactions place the attacking and leaving groups in line. As TSMs are not conserved, we have shown previously that the TLR (tetraloop receptor) can be used as a stabilizing element that increases ribozyme catalytic activity [15].

The aim of the present study was to design and optimize new TLR-extended hammerhead ribozymes along with analysis of their properties in silico, in vitro and within the cell. We describe distinctive catalytic core geometry of six parameters concerning inter-atomic distances and angles. We call them the ribozyme fingerprint. This unique set of data characterizes the molecule's catalytic core and correlates it with its intracellular activity. We propose ribozyme fingerprint calculations as a general method of activity prediction.

EXPERIMENTAL

Ribozymes, RNA substrates and oligonucleotides

Fluorescein-labelled RNA substrates and ribozymes were synthesized by IBA and Future Synthesis. DNA oligonucleotides were synthesized by IBB PAN and Genomed S.A.

Ribozyme labelling

Ribozyme (2 μg) was labelled at the 5′-end with [γ-32P]ATP (ICN) and the T4 polynucleotide kinase (USB) at 37°C for 45 min in 10× reaction buffer. Labelled RNA was purified on SDS/PAGE (10% gel) containing 7 M urea. The radioactive band was cut off and RNA was eluted overnight at 4°C with an elution buffer containing 0.5 M sodium acetate and 0.1 mM EDTA, then precipitated with 2.5 vol. of 96% ethanol in the presence of 0.1 vol. of 3 M sodium acetate, pH 4.8, and dissolved in the RNase-free water (Ambion). The quantification of the labelled RNA was performed by scintillation counting.

RNA isolation and cDNA synthesis

Total RNA from HeLa cells was isolated using the TRIzol® Reagent (Invitrogen) according to the manufacturer's protocol. Extracted RNA underwent DNase treatment with the DNA-free kit (Ambion). The quality of RNA was estimated using agarose gel electrophoresis. Total RNA (2 μg) was used for cDNA synthesis with the RevertAid H Minus First Strand cDNA Synthesis Kit (Fermentas) according to the manufacturer's protocol using random hexamer primers.

In vitro assays

Designed ribozymes and fluorescein-labelled short targets, corresponding to transcripts of selected gene fragments, were synthesized chemically. Comparative trans-cleavage analysis of hammerhead variants were conducted and rate constants under single-turnover conditions using an excess of the ribozyme were determined. All ribozymes were analysed under the same conditions with respect to MgCl2 concentration, ribozyme/target ratio, reaction time and temperature, buffer and additional components.

For single-turnover reactions, 2 pmol of the 5′-fluorescein-labelled substrate T16 (16 nt) was mixed with 200 pmol of ribozyme in 50 mM Tris/HCl, pH 7.5, heated at 70°C for 2 min in a water bath, then allowed to cool slowly (overnight) to the reaction temperature. Cleavage reactions were initiated by addition of MgCl2 and spermine, both to a final concentration of 10 mM or solely MgCl2 to a final concentration of 1 mM. Total reaction volume was 10 μl. Reactions were carried out for 5 h at 37°C and stopped with 10 μl of a loading buffer at 0°C. For determination of the cleavage rate constants, after taking the zero-time aliquot, the reaction was triggered by adding MgCl2 and spermine. Aliquots (10 μl) were removed at appropriate time intervals and quenched with a loading buffer at 0°C. Reaction products were separated on a 20% denaturing (7 M urea) polyacrylamide gel and quantified using the ImageQuant software (Molecular Dynamics). Cleavage yield was estimated by treating the density of the control band as 100% and calculating the density of the product band as x%.

For analysis of complex forming, 2 pmol of the 5′-fluorescein-labelled substrate T16 (16 nt) was mixed with 200 pmol ribozyme in 50 mM Tris/HCl, pH 7.5, V1 buffer (10× RNA structure buffer, 100 mM Tris, pH 7.0, 1 M KCl and 100 mM MgCl2; Ambion), S1 buffer (10×, pH 4.5; Promega), TMN buffer (10×, 200 mM Tris/HCl, pH 7.5, 2 mM MgCl2 and 100 mM NaCl) or hybridization buffer (10×, 200 mM Tris/HCl, pH 7.5, and 500 mM NaCl) and heated at 70°C for 2 min in a water bath, then allowed to cool slowly (overnight) to the reaction temperature. Alternatively, 5′ radioactively labelled ribozyme (30000 c.p.m.) was mixed with 200 pmol of unlabelled ribozyme and 2 pmol of substrate T16 in 50 mM Tris/HCl, pH 7.5, heated at 70°C for 2 min in a water bath, then allowed to cool slowly (overnight) to the reaction temperature. Complexes of ribozymes and substrates were separated on a 15% native polyacrylamide gel and quantified using ImageQuant software.

Cell culture

HeLa cells were cultured in RPMI 1640 medium (Sigma) supplemented with 10% fetal bovine serum (Sigma), 1% antibiotic solution (Sigma) and 1% vitamin solution (Sigma). U118 cells were cultured in DMEM (Dulbecco's modified Eagle's medium; A.T.C.C.) supplemented with 10% fetal bovine serum (A.T.C.C.) and 1% streptomycin solution (A.T.C.C.).

Transfection

Cells were transfected with the Lipofectamine™ 2000 Reagent (Invitrogen). For transfection, cells at 80–95% confluence were cultured in a six-well multidish (Nalge Nunc International) containing 1.5 ml of RPMI 1640 (Sigma) or DMEM. Ribozymes, deoxyribozymes 10–23, plasmids and, separately, Lipofectamine™ (6.5 μl per well) were diluted in 250 μl of Opti-MEM (Gibco) and incubated for 5 min at room temperature (22°C). Lipofectamine™ mixture was added to the transfection mixture and incubated together for 20 min at room temperature and afterwards added to each well. The final concentration of ribozymes and deoxyribozymes was 50 nM for the GFP (green fluorescent protein) series and 300 nM for others. Plasmid pEGFP-N3 (Clontech) was added at 0.655 nM (4 μg per well).

Real-time qPCR (quantitative PCR)

Real-time qPCR was performed to assess transcripts of GFP, GLI1 (glioma-associated oncogene 1) and CRYAB (αB-crystallin) expression relative to HPRT1 (hypoxanthine phosphoribosyltransferase 1) and ACTB (β-actin) (Table 1). The basic protocol used to determine the relative abundance of the gene transcripts is described in [16]. Each cDNA sample was analysed in triplicate using the thermocycler Stratagene Mx3005P (Agilent Technologies). The 25 μl reaction mixture was prepared with the DyNAmo HS SYBR Green qPCR Kit (Finnzymes) and included 1× MasterMix, 0.3× ROX reference dye, 0.3 μM of each primer, 2 μl of template cDNA and water to a final volume of 25 μl. The PCR conditions for all genes were as follows: initial denaturation (95°C, 10 min), a four-step amplification programme repeated 40–50 times [95°C for 15 s, Tm (melting temperature) for 30 s and 72°C for 30 s], a melting curve programme (95°C for 1 min, 55°C for 30 s, 55–95°C with a heating rate of 0.1°C/s and 95°C for 30 s). All standard curves were generated by amplifying series of 5-fold dilutions of cDNA. The quality of PCR products was checked by an analysis of the melting curve.

Table 1
Real-time qPCR assessment of transcripts of GFP, GLI1 and CRYAB expression relative to HPRT1 and ACTB

F, forward; R, reverse.

Gene Accession number/vector name Nucleotide sequences (5′→3′) PCR efficiency (slope) Tm (°C) Fragment size (nt) 
ACTB NM-001101.3 F: TCTGGCACCACACCTTCTAC 2.129 60 168 
  R: GATAGCACAGCCTGGATAGC    
HPRT1 NM-000194.2 F: CTGAGGATTTGGAAAGGGTG 2.068 60 155 
  R: AATCCAGCAGGTCAGCAAAG    
GFP pEGFP-N3 (Clontech) F: CCTGAAGTTCATCTGCACCA 2.094 60 120 
  R: AAGTCGTGCTGCTTCATGTG    
GLI1 NM-001167609.1 F: GTACCACTGTGTCCCGCCGC 2.105 59 189 
  R: GCAGGTAGTGCTGGGCAGGC    
CRYAB NM-001885.1 F: GATCCGCCGCCCCTTCTTTCC 2.097 60 184 
  R: CAGGCGCATCTCTGAGAGTCCAGT    
Gene Accession number/vector name Nucleotide sequences (5′→3′) PCR efficiency (slope) Tm (°C) Fragment size (nt) 
ACTB NM-001101.3 F: TCTGGCACCACACCTTCTAC 2.129 60 168 
  R: GATAGCACAGCCTGGATAGC    
HPRT1 NM-000194.2 F: CTGAGGATTTGGAAAGGGTG 2.068 60 155 
  R: AATCCAGCAGGTCAGCAAAG    
GFP pEGFP-N3 (Clontech) F: CCTGAAGTTCATCTGCACCA 2.094 60 120 
  R: AAGTCGTGCTGCTTCATGTG    
GLI1 NM-001167609.1 F: GTACCACTGTGTCCCGCCGC 2.105 59 189 
  R: GCAGGTAGTGCTGGGCAGGC    
CRYAB NM-001885.1 F: GATCCGCCGCCCCTTCTTTCC 2.097 60 184 
  R: CAGGCGCATCTCTGAGAGTCCAGT    

Western blot

Target protein level from total protein level of HeLa cells 24 h after transfection was estimated by Western blot with use of a ‘semi-dry’ transfer system with the PVDF membrane and monoclonal antibodies specific against GFP, CRYAB, GLI1 and GAPDH (glyceraldehyde 3-phosphate dehydrogenase) as reference. Cells (approximately 2.5×106 cells) were washed with PBS, scraped into 10 mM Tris, pH 7.5, and sonicated under the following conditions: 4×15 s with 1 min intervals 75% amplitude followed by centrifugation for 10 min at 4°C and 13.2×103 rev./min using an FA-45-24-11 rotor (centrifuge 5415, Eppendorf). The quality of proteins was checked by electrophoresis. Western blot analysis was quantified by treating the density of the control band as 100% and calculating the density of the product band as x%. Each sample was denaturated by heating at 95°C for 10 min. Total protein extract (μg of protein/sample: 30 μg of GAPDH, 30 μg of GFP, 150 μg of CRYAB and 150 μg of GLI1) were subjected to electrophoresis on SDS/PAGE. Separated proteins were transferred on to a PVDF membrane using Western Unit (Bio-Rad Laboratories) in Towbin buffer [25 mM Tris, pH 7.5, 190 mM glycine and 20% (v/v) methanol]. The membrane was blocked with 10% (w/v) non-fat dried skimmed milk powder in 1× PBS/0.05% Tween 20 at 4°C overnight and washed three times in 1× PBS/0.05% Tween 20 at room temperature. The membrane was incubated with the mouse monoclonal antibodies (Santa Cruz Biotechnology) against GAPDH (1:500 dilution, 0411, sc-47724), GFP (1:500 dilution, A00185.01, sc-81045), GLI1 (1:100 dilution, D-1, sc-271075) and CRYAB (1:100 dilution, 2E8, sc-53919) for 2 h at room temperature. After washing three times in 1× PBS/0.05% Tween 20, the membrane was treated with the secondary biotin-conjugated anti-(mouse Ig) antibody (1:2000 dilution, Sigma) for 2 h at room temperature. The membrane was washed three times and then incubated with the streptavidin–alkaline phosphatase conjugate for 15 min at room temperature. After washing, membrane was developed using the BCIP (5-bromo-4-chloroindol-3-yl phosphate) NBT (Nitro Blue Tetrazolium) liquid substrate system (Sigma). Bands were quantified using ImageQuant software.

Structure modelling

In the design strategy, the TSM motif based on TLR sequence was attached to the 5′-end of a hammerhead ribozyme. A selected 16 nt substrate strand was joined to the ribozyme 3′-end by the GGG sequence that allowed building a model of a type I cis-acting ribozyme, necessary for modelling of an entire complex. RNA secondary structure predictions were performed through the Vienna RNAfold server (http://rna.tbi.univie.ac.at/cgi-bin/RNAfold.cgi) with standard parameters to predict minimum free energy structures (Turner model, 1999 [16a]). Catalytic core sequences (CUGAUGA and GAA) were turned into NNNNNNN and NNN respectively, so these nucleotides did not affect the MFE (minimal free energy) structure prediction. The resultant dot-bracket notation was enriched with a catalytic core C3–G5 base pairing to resemble the ribozyme catalytic core active conformation. RNA tertiary structure modelling was performed through the RNAComposer system (http://rnacomposer.cs.put.poznan.pl) [17]. The program generates up to ten structure models, where only the first model is the same in all repeats. Additional models are chosen by chance. For each secondary structure, ten tertiary structures were given, five of which, possessing the lowest CHARMM force field energy, were selected for further calculations. Measurements were taken in PyMOL (http://www.pymol.org).

Structure analysis

To generate a sequence ladder RNA (60000 c.p.m.), alkaline hydrolysis was performed at 95°C for 2 min in 10 μl of reaction mixture containing 50 mM NaOH, 1 mM EDTA and 4 μg of crude tRNA from Vigna angularis as a carrier. In order to perform nuclease cleavages, the radiolabelled RNA (30000 c.p.m.) was treated with T1, S1 and V1 RNases. To generate the lead-induced cleavage, the radiolabelled RNA, supplemented with 2 μg of tRNA carrier, was incubated in 50 mM Tris/HCl, pH 7.5, with different concentrations of Pb2+ (2.0, 1.4, 0.7 and 0.2 mM), at 25°C for 15 min. Digestion products were analysed by SDS/PAGE (20% gels). Percentage structure compatibility was calculated by summing the number of cleavage sites compatible with predicted structure, multiplying it by 100 and dividing by the total number of depicted locations. T1 RNase (Sigma) digestion was performed in a buffer containing 20 mM sodium citrate (pH 5.0), 7 M urea and 1 mM EDTA with 0.3 unit of the enzyme, at 55°C for 20 min. RNase V1 digestion was done under native conditions using 0.01 unit/μl RNase V1 (Ambion) in a buffer containing 50 mM Tris/HCl, pH 7.2, 100 mM NaCl and 10 mM MgCl2 (Ambion). The RNase S1 digestion under native conditions was done using 0.95 unit/μl RNase S1 (Promega) in the 10× reaction buffer (pH 4.5, Promega), supplied by the manufacturer (Promega). The reactions were carried out for 15 min (V1) at 25°C and 30 min at 37°C (S1). The footprinting results were analysed on SDS/PAGE (20% gels).

Statistics

Statistical analysis was performed using GraphPad Prism (GraphPad Software) using the Kruskal–Wallis test.

RESULTS

Design of new hammerhead ribozymes

We designed four hammerhead ribozymes targeting GFP mRNA (the GUC site): minimal hammerhead ribozyme (HHgfp-0) and three extended ribozymes with TSMs on the basis of a sequence of the TLR of Tetrahymena group I introns. The extended ribozymes differ in predicted TSM secondary structure as a result of the number of linker nucleotides between the 5′-end of the catalytic core and the 3′-end of the original TLR sequence: 5′-UCGCCG-3′ (HHgfp-1), 5′-CGCCG-3′ (HHgfp-2) or 5′-GCCG-3′ (HHgfp-3) (Figure 1). We wanted to see the effect of stem I–II self-positioning and choose optimal variants for further research.

Secondary structures of designed hammerhead ribozymes

Figure 1
Secondary structures of designed hammerhead ribozymes

HHgfp-0, wild-type hammerhead ribozyme. HHgfp-1, HHgfp-2 and HHgfp-3, extended hammerhead ribozymes with the TSM based on the sequence of the TLR at the 5′-end. Ribozymes sharing the same predicted secondary structure are placed in the circles (HHgfp-2, HHcry-2c and HHgli-2c). Changes in the HHcry-2c and HHgli-2c ribozyme sequence are shown on a grey background. Regions changed in HHgfp-2 mutants (HHgfp-2MUT1, HHgfp-2MUT2 and HHgfp-2MUT3) are shown on a black background. HHgli-2, HHgfp-2 derivative without changes in ribozyme sequence. Black/white letters, ribozyme; grey letters, target RNA; ▲, cleavage sites; ∆, putative cleavage sites.

Figure 1
Secondary structures of designed hammerhead ribozymes

HHgfp-0, wild-type hammerhead ribozyme. HHgfp-1, HHgfp-2 and HHgfp-3, extended hammerhead ribozymes with the TSM based on the sequence of the TLR at the 5′-end. Ribozymes sharing the same predicted secondary structure are placed in the circles (HHgfp-2, HHcry-2c and HHgli-2c). Changes in the HHcry-2c and HHgli-2c ribozyme sequence are shown on a grey background. Regions changed in HHgfp-2 mutants (HHgfp-2MUT1, HHgfp-2MUT2 and HHgfp-2MUT3) are shown on a black background. HHgli-2, HHgfp-2 derivative without changes in ribozyme sequence. Black/white letters, ribozyme; grey letters, target RNA; ▲, cleavage sites; ∆, putative cleavage sites.

In vitro analysis of ribozyme activity

In vitro assays showed that all ribozymes are active in the presence of 1 mM Mg2+ ions, but HHgfp-0 and HHgfp-1 are the most active ones (Figures 2a and 2b). Calculated kobs (min−1) values reached 0.42±0.1 (HHgfp-0), 0.06±0.03 (HHgfp-1), 0.005±0.002 (HHgfp-2) and 0.0002±0.0001(HHgfp-3) (Figure 2f).

In vitro analysis of ribozyme activity

Figure 2
In vitro analysis of ribozyme activity

(a and b) Substrate (T16) and reaction products (P8) of HHgfp ribozyme activity on a 20% polyacrylamide/7 M urea gel. C1, target in water; C2, target in water with 1 mM MgCl2 (a) or 10 mM MgCl2 (b). (c) Native polyacrylamide gel showing the elecrophoretic mobility of 5′-radiolabelled HHgfp-2 RNA in complexes with substrate in Tris, pH 7.5, S1, V1, TMN and H buffers. H, hybridization buffer; S1, S1 buffer; TMN, TMN hybridization buffer; Tris, 50 mM Tris/HCl, pH 7.5; V1, V1 buffer (10× structure buffer). (d) Native 15% polyacrylamide gel showing the elecrophoretic mobility of 5′-fluorescein labelled substrate (T16) RNA in the presence of four ribozymes under standard reaction conditions without Mg2+. 1–4, different substrate–ribozyme complexes; C, substrate T16 in water. (e) Analysis of HHgfp-2 and HHgfp-2MUT3 activity against target RNAs of 12 and 16 nucleotides (T12 and T16). (f) Calculated kobs (min−1) for all tested ribozymes under the same conditions. (g) Changes in reaction efficiency (cleavage yield) with time showing different catalytic properties of minimal and TLR-extended ribozymes over 0–512 min. P8, reaction product of eight nucleotides.

Figure 2
In vitro analysis of ribozyme activity

(a and b) Substrate (T16) and reaction products (P8) of HHgfp ribozyme activity on a 20% polyacrylamide/7 M urea gel. C1, target in water; C2, target in water with 1 mM MgCl2 (a) or 10 mM MgCl2 (b). (c) Native polyacrylamide gel showing the elecrophoretic mobility of 5′-radiolabelled HHgfp-2 RNA in complexes with substrate in Tris, pH 7.5, S1, V1, TMN and H buffers. H, hybridization buffer; S1, S1 buffer; TMN, TMN hybridization buffer; Tris, 50 mM Tris/HCl, pH 7.5; V1, V1 buffer (10× structure buffer). (d) Native 15% polyacrylamide gel showing the elecrophoretic mobility of 5′-fluorescein labelled substrate (T16) RNA in the presence of four ribozymes under standard reaction conditions without Mg2+. 1–4, different substrate–ribozyme complexes; C, substrate T16 in water. (e) Analysis of HHgfp-2 and HHgfp-2MUT3 activity against target RNAs of 12 and 16 nucleotides (T12 and T16). (f) Calculated kobs (min−1) for all tested ribozymes under the same conditions. (g) Changes in reaction efficiency (cleavage yield) with time showing different catalytic properties of minimal and TLR-extended ribozymes over 0–512 min. P8, reaction product of eight nucleotides.

To understand the reasons for the low activity of the extended ribozymes in vitro, we analysed ribozyme–substrate complex formation (Figures 2c and 2d). HHgfp-2 revealed a different efficiency of complex formation due to buffer composition (Figure 2c and Table 2). Buffers were selected because of their properties of enhancing hybridization (TMN and H) or usefulness in structural RNA studies (S1 and V1). Incubation in V1 buffer resulted in efficient transesterification and effective complex formation, which were inefficient in Tris/HCl pH 7.5 buffer. Monitoring of fluorescein-labelled target migration revealed several extended ribozyme–substrate complexes (Figure 2d).

Table 2
Different efficiency of ribozyme–target complex formation depending on buffer composition
Buffer Complex-forming efficiency (%) 
V1 45.6 
Tris/HCl 12 
S1 50 
TMN 52 
30 
Buffer Complex-forming efficiency (%) 
V1 45.6 
Tris/HCl 12 
S1 50 
TMN 52 
30 

Analysis of ribozyme activity in HeLa cells

In order to determine the ribozyme activity, HeLa cells were transfected with vectors expressing HHgfp-0, HHgfp-1, HHgfp-2, HHgfp-3 and pEGFP-N3 and examined 24 h after transfection. Co-transfection conditions and ribozyme activity were compared with the effect of antisense DNA, scrambled DNA, antisense RNA, mutated ribozymes within the catalytic core (HHgfp-2MUT1 and HHgfp-2MUT2) and TSM region (HHgfp-2MUT3) (Figure 1) and DNAzyme 10–23 (Dzgfp).

Decreases in GFP fluorescence, real-time PCR and Western blot levels pointed to HHgfp-1, HHgfp-2 and HHgfp-3 being very active in HeLa cells. HHgfp-2 was the most efficient (Figure 3 and Supplementary Figure S1 at http://www.biochemj.org/bj/451/bj4510439add.htm). HHgfp-2 inhibited target gene expression to 0.27 and 0.25 of relative RNA and protein levels respectively. Other molecules decrease GFP expression as follows: HHgfp-0 (0.78 and 0.53), HHgfp-1 (0.69 and 0.4), HHgfp-3 (0.4 and 0.43) and control molecules (Dzgfp: 0.78 and 0.45). The results were statistically significant (P<0.05).

Intracellular activity of designed ribozymes

Figure 3
Intracellular activity of designed ribozymes

(a) Western blot analysis. Upper panel, GAPDH as reference gene; lower panel, target genes (GFP, GLI1 and CRYAB). (b) Relative target protein (GFP, GLI1, CRYAB) level was quantified using ImageQuant Software. (c) Relative target level of GFP, CRYAB and GLI1 RNA obtained by real-time PCR. asDNA, antisense DNA; asRNA, antisense RNA; C, untreated control; Dz, DNAzyme; scrDNA, scrambled DNA.

Figure 3
Intracellular activity of designed ribozymes

(a) Western blot analysis. Upper panel, GAPDH as reference gene; lower panel, target genes (GFP, GLI1 and CRYAB). (b) Relative target protein (GFP, GLI1, CRYAB) level was quantified using ImageQuant Software. (c) Relative target level of GFP, CRYAB and GLI1 RNA obtained by real-time PCR. asDNA, antisense DNA; asRNA, antisense RNA; C, untreated control; Dz, DNAzyme; scrDNA, scrambled DNA.

HHgfp-2 secondary structure in vitro

Secondary structures of ribozyme–substrate complexes were calculated by the RNAfold webserver (HHgfp-0, HHgfp-1, HHgfp-2 and HHgfp-3) (Figure 1). The 5′-radiolabelled HHgfp-2 was subjected to enzymatic and chemical structure analysis. Owing to a low efficiency of HHgfp-2–substrate complex formation and a high level of hydrolysis in the V1 buffer, we analysed only the TSM region of HHgfp-2, which was different in proposed models. The in vitro hydrolysis pattern was compared with predicted models. We calculated the HHgfp-2 structure and experimental results compatibility to a level of 94% (Figure 4).

HHgfp-2 secondary structure

Figure 4
HHgfp-2 secondary structure

Autoradiograms of HHgfp-2 ribozyme digestion by specific ribonucleases (a) and Pb2+ ions (b) of target-independent regions (right-hand side, shown on a grey background). Top of autoradiograms: C, untreated control; HH, ribozyme; HH+T16, ribozyme in the presence of substrate T16; L, alkaline hydrolysis ladder, T1, limited digestion with RNAse T1. HHgfp structure: arrows, S1 digestion; dots, Pb2+-cleavage sites; triangles, V1 digestion.

Figure 4
HHgfp-2 secondary structure

Autoradiograms of HHgfp-2 ribozyme digestion by specific ribonucleases (a) and Pb2+ ions (b) of target-independent regions (right-hand side, shown on a grey background). Top of autoradiograms: C, untreated control; HH, ribozyme; HH+T16, ribozyme in the presence of substrate T16; L, alkaline hydrolysis ladder, T1, limited digestion with RNAse T1. HHgfp structure: arrows, S1 digestion; dots, Pb2+-cleavage sites; triangles, V1 digestion.

To confirm these results, we carried out further analysis in vitro. The TSM region of HHgfp-2 does not depend on the target length, whereas that of HHgfp-2MUT3 does (Figure 1). HHgfp-2 activity in vitro was tested separately against two substrates of different length: T12 (12 nucleotides, shortened at the 3′-end) and T16 (16 nucleotides). The results showed that cleavage efficiency after 5 h for HHgfp-2 against T12 is very low (15.5%), whereas in the case of the T16 it reached 98%. Activity of HHgfp-2MUT3 was high and did not depend on a target length (Figure 2e).

Therefore we propose a hammerhead ribozyme, called HH-2, stabilized with the new TSM (Figure 5). HH-2 relates to the general MFE predicted structure regardless of substrate sequence. The secondary structure of the TSM consists of a hairpin (A) with a helix H4, a multi-branch loop (B) and a supplementary region of substrate recognition (H5) (Figure 5). The hairpin A possibly interacts with the H2 helix and is required for catalytic core active conformation. The multi-branch loop B has four unpaired nucleotides that supposedly allow flexible adjustment of elements A and H2. The length of helix H4 determines proper positioning of hairpin A towards H2. H5 improves ribozyme specificity and consists of at least three base pairs that preferably should be strong G–C pairs.

Models of the secondary (left) and tertiary (right) HH-2 structures

Figure 5
Models of the secondary (left) and tertiary (right) HH-2 structures

Regions of possible tertiary interaction are indicated by grey boxes. A, hairpin loop; B, multi-branch loop; H1–H5, helices 1–5; black lettering, hammerhead ribozyme; grey lettering, target RNA. The cleavage site is depicted with a triangle.

Figure 5
Models of the secondary (left) and tertiary (right) HH-2 structures

Regions of possible tertiary interaction are indicated by grey boxes. A, hairpin loop; B, multi-branch loop; H1–H5, helices 1–5; black lettering, hammerhead ribozyme; grey lettering, target RNA. The cleavage site is depicted with a triangle.

Design of the HH-2 ribozyme against GLI1 and CRYAB genes

We designed HH-2 ribozymes against CRYAB and GLI1 [18,19]. The sequences of the generated MFE ribozyme structures were changed to adopt HH-2 and labelled ‘c’ (‘changed’). Six residues were changed in HHgli-2, whereas four were changed in HHcry-2, providing HHgli-2c and HHcry-2c respectively. These substitutions were limited to the 5′-end of predicted TSM regions (helix H5 and bulge B) interacting with the target sequence (Figure 5). In vitro structural analysis of HHcry-2c and HHgli-2c confirmed that the structures are compatible with the proposed model by 86% and 78% respectively (Supplementary Figures S2 and S3 at http://www.biochemj.org/bj/451/bj4510439add.htm). For control experiments we designed minimal variant HHgli-0, extended variant HHgli-2b (no changes in TSM sequence) and 10-23 DNAzymes (Dzcry and Dzgli).

Analysis of HH-2 activity

All designed ribozymes have similar in vitro properties of minimal HH-0 with the highest kobs (min−1) values (HHgfp-0: 0.42±0.1 and HHgli-0: 1.1±0.05) and extended HH-2 with much reduced kobs values (HHgfp-2: 0.005±0.002, HHgli-2c: 0.02±0.008 and HHcry-2c: 0.18±0.009) (Figure 2f). In vitro studies of HH-2 activity showed that the cleavage rate started to increase 15–60 min after starting the reaction by adding Mg2+ ions and spermine (Figure 2g). It suggested that the initial time was necessary to convert inactive ribozymes into active conformers.

We transfected HeLa and U118 cell lines with HHgli-2c and HHcry-2c. Their activity was analysed in comparison with the effect of 10-23 DNAzymes HHgli-0 and HHgli-2. The results showed a high activity of HHcry-2c and HHgli-2c, leading to 0.65 and 0.54 of the relative RNA level and 0.78 and 0.48 of the relative protein level respectively. The results were statistically significant (P<0.05). In each case HH-0 was more active in vitro than HH-2. Comparison of HH-0, HH-2 and 10-23 DNAzymes against the same target indicated that HH-2 was the most efficient within the cells.

Tertiary structure modelling and computational analysis

Secondary structures of all ribozymes were verified by in vitro structural analysis and introduced to the RNAComposer [17]. Tertiary structure models of substrate–ribozymes complexes were compared (Figures 6g and 7). Three distances were measured: G12(N1) (general base) to C17(2′O) (nucleophile)-D1, G8(2′O) (general acid) to C1.1(5′O) (the leaving group)-D2 and G5(N1) to C1.1(5′O)-D3. As the reaction demands in-line positioning of the attacking group, attacked phosphorus and leaving group, we measured three angles showing how far the atoms diverge from the in-line position. These were: C1.1(5′O)–C17(2′O)-P (α), C17(2′O)–G12(N1)-P (β) and C1.1(5′O)–G8(2′O)-P (γ), where P is a phosphate in the cleavage site. Additionally we measured the distance between the nearest atoms of stem I–II, indicating a probability of their direct contact and the D1/D2 ratio.

Comparative in silico analysis of distances and angles between definite atoms in 3D HHgfp ribozyme tertiary structure models

Figure 6
Comparative in silico analysis of distances and angles between definite atoms in 3D HHgfp ribozyme tertiary structure models

(ac) Comparison of distances C17(2′O)–G12(N1) (D1, a), C1.1(5′O)–G8(2′O) (D2, b) and C1.1(5′O)–G5(N1) (D3, c) in Å. (df) Atom divergence from in-line position for three major reactions during ribozyme catalysis in degrees: C1.1(5′O)–C17(2′O)-P (α, d); C17(2′O)–G12(N1)-P (β, e); and C1.1(5′O)–G8(2′O)-P (γ, f). (g) Ratio between distances D1 and D2. (h) Distances between the nearest neighbours (grey boxes in tertiary structure model in j) of stem I and II in Å. (i) Schematic representation of ribozyme catalytic core. Nucleotide numbering adopted from [11]. (j) Tertiary structure model of the ribozyme–substrate complex with regions of possible tertiary interaction indicated.

Figure 6
Comparative in silico analysis of distances and angles between definite atoms in 3D HHgfp ribozyme tertiary structure models

(ac) Comparison of distances C17(2′O)–G12(N1) (D1, a), C1.1(5′O)–G8(2′O) (D2, b) and C1.1(5′O)–G5(N1) (D3, c) in Å. (df) Atom divergence from in-line position for three major reactions during ribozyme catalysis in degrees: C1.1(5′O)–C17(2′O)-P (α, d); C17(2′O)–G12(N1)-P (β, e); and C1.1(5′O)–G8(2′O)-P (γ, f). (g) Ratio between distances D1 and D2. (h) Distances between the nearest neighbours (grey boxes in tertiary structure model in j) of stem I and II in Å. (i) Schematic representation of ribozyme catalytic core. Nucleotide numbering adopted from [11]. (j) Tertiary structure model of the ribozyme–substrate complex with regions of possible tertiary interaction indicated.

The hammerhead ribozyme fingerprint

Figure 7
The hammerhead ribozyme fingerprint

(a) Selection of target site (GUC or AUC) and ribozyme design. (b) Ribozyme–substrate secondary structure. The TSM motif of interest is attached to the 5′-end of a hammerhead ribozyme. The substrate strand is virtually attached to the ribozyme 3′-end by the GGG sequence, that allows the building of a model of the type I cis-acting ribozyme, necessary for modelling the entire complex. X, any nucleotide. (c) RNA secondary structure predictions by Vienna RNAfold. Catalytic core sequence (CUGAUGA and GAA) is turned into NNNNNNN and NNN respectively, so these nucleotides do not affect the MFE structure prediction. The resultant dot-bracket notation is manually enriched with catalytic core C3–G5 base pairing to resemble ribozyme catalytic core active conformation. (d) RNA tertiary structure modelling is performed through RNAComposer [25]. For each secondary structure, ten tertiary models are given, and the five of which that possess the lowest CHARMM force field energy are selected for further calculations. (e) Measurements are taken with PyMOL software. For each molecule a unique ribozyme fingerprint is obtained. (f) The ratio of D1/D2 is established and designed molecules can be compared. The highest D1/D2 value will result in the most efficient down-regulation of target gene expression.

Figure 7
The hammerhead ribozyme fingerprint

(a) Selection of target site (GUC or AUC) and ribozyme design. (b) Ribozyme–substrate secondary structure. The TSM motif of interest is attached to the 5′-end of a hammerhead ribozyme. The substrate strand is virtually attached to the ribozyme 3′-end by the GGG sequence, that allows the building of a model of the type I cis-acting ribozyme, necessary for modelling the entire complex. X, any nucleotide. (c) RNA secondary structure predictions by Vienna RNAfold. Catalytic core sequence (CUGAUGA and GAA) is turned into NNNNNNN and NNN respectively, so these nucleotides do not affect the MFE structure prediction. The resultant dot-bracket notation is manually enriched with catalytic core C3–G5 base pairing to resemble ribozyme catalytic core active conformation. (d) RNA tertiary structure modelling is performed through RNAComposer [25]. For each secondary structure, ten tertiary models are given, and the five of which that possess the lowest CHARMM force field energy are selected for further calculations. (e) Measurements are taken with PyMOL software. For each molecule a unique ribozyme fingerprint is obtained. (f) The ratio of D1/D2 is established and designed molecules can be compared. The highest D1/D2 value will result in the most efficient down-regulation of target gene expression.

Comparison of four HHgfp 3D (three-dimensional) in silico models confirmed that HHgfp-2 has significantly different parameters in every aspect, being the lowest or the highest in a defined group (Figures 6a–6f). The results of D1, D3, α, H1-H2 and D1/D2 calculations were statistically significant (P<0.05). The correlation coefficient for pairs of measured parameters and experiment results were calculated separately for ribozymes targeting GFP and together for all tested ribozymes (Table 3). Almost complete correlation was estimated for D1, D2, D1/D2 ratio and stem I–II distance to real-time PCR results in the case of ribozymes targeting solely GFP. Considering all ribozymes, a very high correlation was found for D1, D1/D2 ratio and stem I–II distance to transcript level as well.

Table 3
Correlation coefficients for pairs of calculated parameter and experiment results separately for ribozymes targeting GFP and for all tested ribozymes

D1, C17(2′O)–G12(N1); D2, C1.1(5′O)–G8(2′O); D3, C1.1(5′O)–G5(N1); α, C1.1(5′O)–C17(2′O)-P; β, C17(2′O)–G12(N1)-P; γ, C1.1(5′O)–G8(2′O)-P.

  Correlation coefficient 
Parameters Experiment GFP group All ribozymes 
D1 Real-time PCR −0.93668 −0.71725 
D1 Western blot −0.76932 −0.28656 
D1 Stem I–II −0.81065 −0.79884 
D2 Real-time PCR 0.973649 0.359495 
D2 Western Blot 0.887022 −0.26215 
D2 Stem I–II 0.86943 −0.08938 
D1–D2  −0.98447 0.133075 
D1/D2 Real-time PCR −0.94556 −0.82253 
D1/D2 Western blot −0.77933 −0.24188 
D1/D2 Stem I–II −0.82633 −0.83054 
D3 Real-time PCR −0.7121 −0.17405 
α Angle Real-time PCR 0.650545 0.516135 
β Angle Real-time PCR 0.757735 −0.047638 
γ Angle Real-time PCR −0.81378 −0.41569 
Stem I–II Real-time PCR 0.959146 0.751522 
Stem I–II Western blot 0.856755 −0.057469 
kobs Real-time PCR 0.775472 0.297158 
kobs Stem I–II 0.897635 0.318746 
kobs D1/D2 −0.52785 −0.33978 
  Correlation coefficient 
Parameters Experiment GFP group All ribozymes 
D1 Real-time PCR −0.93668 −0.71725 
D1 Western blot −0.76932 −0.28656 
D1 Stem I–II −0.81065 −0.79884 
D2 Real-time PCR 0.973649 0.359495 
D2 Western Blot 0.887022 −0.26215 
D2 Stem I–II 0.86943 −0.08938 
D1–D2  −0.98447 0.133075 
D1/D2 Real-time PCR −0.94556 −0.82253 
D1/D2 Western blot −0.77933 −0.24188 
D1/D2 Stem I–II −0.82633 −0.83054 
D3 Real-time PCR −0.7121 −0.17405 
α Angle Real-time PCR 0.650545 0.516135 
β Angle Real-time PCR 0.757735 −0.047638 
γ Angle Real-time PCR −0.81378 −0.41569 
Stem I–II Real-time PCR 0.959146 0.751522 
Stem I–II Western blot 0.856755 −0.057469 
kobs Real-time PCR 0.775472 0.297158 
kobs Stem I–II 0.897635 0.318746 
kobs D1/D2 −0.52785 −0.33978 

Comparing models of the extended and minimal variants, we observed the highest differences ranging from 1.04 to 1.28 Å (1 Å=0.1 nm) for D1. Efficient activity of HHgfp-2 in the cells strongly correlates with bringing G8(2′O) closer to C1.1(5′O) (D2) and withdrawing G12(N1) from C17(2′O) (D1). This relationship was described by the D1/D2 ratio. Its increase characterizes ribozymes more active in the intracellular environment (HHgfp-0: 0.923 and HHgfp-2: 1.284). Measurement of angles confirmed the necessity of bringing in-line atoms engaged in major ribozyme reaction α as well as for β. The highest differences between extended and minimal ribozyme structure angles were 2.46° α and 2.54° β and diverge approximately 10° from the in-line position. In the case of γ, a larger angle seemed to slightly increase ribozyme activity up to 28.44° for HHgfp-2 with the highest difference of 1.48° between models.

To verify these observations we compared all structure models of HH-0 and HH-2 ribozymes. The question was whether target sequence changes of HHgli-2c and HHcry-2c affect parameters of the catalytic core (Figures 8a–8f). For D1 and D3 distances as well as the β angle, we observed a similar tendency of increasing distances and decreasing angles as function of changes between HH-0 and HH-2 (Figures 8a, 8c and 8e). For the other three parameters, the conclusions were less clear (Figures 8b, 8d and 8f), probably due to the role of the flanking sequence. To test this hypothesis we created 61 in silico HH-0 ribozymes against artificial targets such as poly(A), poly(C), poly(G) and poly(U) oligonucleotides with insertion of GUC sequence as a target site (Supplementary Table S1 at http://www.biochemj.org/bj/451/bj4510439add.htm). They all had the same secondary structure. Computational analysis showed that catalytic core parameters depend upon the amount of purines and pyrimidines and their location. It is not only a result of Watson–Crick pairing, but also depends on which strand (ribozyme or substrate) purines and pyrimidines are placed.

Comparative in silico analysis of distances and angles between definite atoms in 3D ribozymes structural models

Figure 8
Comparative in silico analysis of distances and angles between definite atoms in 3D ribozymes structural models

Designed ribozymes (dark grey boxes) act against different targets (GFP, GLI1 and CRYAB) and differ in secondary structure: HHgfp-0, HHgli-0 and HHcry-0 (not tested in cells), minimal hammerhead ribozymes; HHgfp-2, HHgli-2c and HHcry-2c, extended ribozymes sharing the same secondary structure. Artificial in silico ribozymes (light grey boxes) are designed against mononucleotide targets with a GUC cleavage site: Rz1C, monocytosine; Rz2G, monoguanosine; Rz3A, monoadenosine; and Rz4U, monouridine. (ac) Comparison of distances C17(2′O)–G12(N1) (D1, a), C1.1(5′O)–G8(2′O) (D2, b) and C1.1(5′O)–G5(N1) (D3, c) in Å. (df) Atom divergence from in-line position for three major reactions during ribozyme catalysis in degrees: C1.1(5′O)–C17(2′O)-P (α, d), C17(2′O)–G12(N1)-P (β, e) and C1.1(5′O)–G8(2′O)-P (γ, f). (g) Ratio between D1 and D2 distances.

Figure 8
Comparative in silico analysis of distances and angles between definite atoms in 3D ribozymes structural models

Designed ribozymes (dark grey boxes) act against different targets (GFP, GLI1 and CRYAB) and differ in secondary structure: HHgfp-0, HHgli-0 and HHcry-0 (not tested in cells), minimal hammerhead ribozymes; HHgfp-2, HHgli-2c and HHcry-2c, extended ribozymes sharing the same secondary structure. Artificial in silico ribozymes (light grey boxes) are designed against mononucleotide targets with a GUC cleavage site: Rz1C, monocytosine; Rz2G, monoguanosine; Rz3A, monoadenosine; and Rz4U, monouridine. (ac) Comparison of distances C17(2′O)–G12(N1) (D1, a), C1.1(5′O)–G8(2′O) (D2, b) and C1.1(5′O)–G5(N1) (D3, c) in Å. (df) Atom divergence from in-line position for three major reactions during ribozyme catalysis in degrees: C1.1(5′O)–C17(2′O)-P (α, d), C17(2′O)–G12(N1)-P (β, e) and C1.1(5′O)–G8(2′O)-P (γ, f). (g) Ratio between D1 and D2 distances.

Measurements of distance between the adjacent atoms of stem I and II indicated, for HH-2, that these regions are in the nearest position (2.58 Å for HHgfp-2, 2.62 Å for HHgli-2c and 2.92 Å for HHcry-2c) (Figure 6i). In the case of other molecules, these distances were: HHgfp-0, 10.64 Å; HHgfp-1, 7.16 Å; HHgfp-3, 5.1 Å; and HHgli-0, 4.96 Å (Figure 6i). The stem I–II distance and the D1/D2 ratio are highly correlated with the results obtained in cell lines. It indicates that the TSM motif is in the best position in HH-2, where stem I and stem II are always closer in comparison with HH-0. We suggest that TSM introduction causes bending of the ribozymes, resulting in changes in the catalytic core geometry. The difference between HH-0 and HH-2 stem I–II distances is the biggest for the GFP series (8.06 Å), whereas it is smaller for GLI1 (2.34 Å) and CRYAB (0.66 Å), which suggests an influence of target sequence.

DISCUSSION

The aim of the present study was to optimize stem I–II interaction along with its influence on the catalytic core of the hammerhead ribozyme. Differences between the primary sequence of HHgfp-1, HHgfp-2 and HHgfp-3 (one or two nucleotides) results in other TSM structures that arrange otherwise towards stem II and modify the ribozyme structure and activity. Our approach towards designing effective ribozymes is based on an analysis of the tertiary structure models on the basis of MFE secondary structures. Even single nucleotide addition dramatically changes RNA structure and should be considered during functional molecule design. In the present study we used confirmed secondary structures as a starting point to obtain reliable tertiary structure models.

All of the HH-2 ribozymes described in the present paper showed low efficiency in vitro in comparison with their minimal counterparts, often requiring 10 mM Mg2+, 10 mM spermine and a long renaturation step (16 h) for catalytic activity. Despite the differences in ribozyme activity, our in vitro analysis confirmed cleavage specificity already in 1 mM MgCl2. Ribozyme activity in vivo depends not only on Mg2+ concentration, but also on ionic strength, temperature, substrate accessibility, interaction with cellular components and localization [20]. A variety of macromolecules even at very low concentrations can induce conformational changes of the hammerhead ribozyme by collisions, non-specific weak interactions or molecular crowding effect [2125]. It has been shown that spermine can improve in vitro the catalytic activity of the hairpin ribozyme [26] and has a general influence on conformation of nucleic acids [27].

We showed that HHgfp-2 misfolds into alternative secondary structures or interacts with additional substrate or ribozyme molecules. Such conformational heterogeneity of ribozyme–substrate complexes could explain its low activity in vitro. It suggests that in vitro conditions (Tris/HCl buffer, pH 7.5) are insufficient for effective folding of TLR-extended ribozymes. Other buffers (S1, V1, TMN and H) could not be used for determination of the catalytic properties of ribozymes because of the presence of divalent ions. A similar observation was also made for other extended hammerhead ribozymes [28]. Two-phase folding can be observed in HH-2 kinetic analysis, suggesting the existence of populations of molecules with different catalytic efficiencies (fast-acting and slow-acting) [29]. It implies that an initial time is necessary to change inactive ribozymes into active conformers in the presence of Mg2+ ions and spermine. It seems that inactive conformers are kinetically trapped and do not revert to the active conformation even after long periods of time [21,30]. It has been suggested that at least two-step folding, if inhibited, might become the rate-limiting step explaining low in vitro efficiency [31].

Analysis of GFP expression in HeLa cells revealed that HHgfp-2 is the most efficient gene expression inhibitor. HHcry-2c and HHgli-2c also showed a high efficiency within the cells. To check a possibility of diverse transcript accessibility we used DNAzymes 10–23 [32]. They are more efficient in cleaving target RNA than ribozymes [33]. In all cases, DNAzymes confirmed mRNA accessibility, but were less active than HH-2. HHgli-2, designed without changes within the primary sequence and having a different MFE secondary structure, showed lower cellular activity in comparison with HHgli-2c. Antisense and scrambled oligonucleotides (RNA and DNA) did not show any significant decrease in mRNA or protein target level, functioning probably as space obstacle during translation. Catalytic core mutants (HHgfp-2MUT1 and HHgfp-2MUT2) that did not show any activity in vitro, caused no significant loss of protein level. Interestingly, TSM mutant (HHgfp-2MUT3) that includes hairpin (GC pairs in the stem) and five base pairs in helix I of predicted secondary structure was equally efficient in vitro as well as in the cells. Its intracellular activity was similar to other extended HHRz targeting GFP. It suggests that coaxial stacking and a possible tetraloop interaction with G–C pairs within the minor groove of the TSM can cause catalytic core bending that result in high intracellular activity. This could also be the result of coaxial helical stacking by the introduced second three-way junction in HHgfp-2 [34].

Comparison of our in vitro and cellular results shows that the test tube conditions are only an approximation of intracellular environment. Extended ribozymes are thought to be very efficient in vitro, often better than a minimal variant [6,7,35,36]. However, some reports showed that ribozymes of poor in vitro activity can be effective tools for inhibition of gene expression in vivo [28,37,38]. Proposed ribozymes are designed using structural elements of different origin (HHRz and TLR). In such a case, RNA folding is not a result of evolutionary selection and can proceed in its own way [39].

To investigate ribozyme differences, we used tertiary structure models of substrate–ribozyme complexes in active conformational states. They did not possess any introduced pseudoknot formation between helices H1 and H2. The H1–H2 distance tells the probability of such interaction. Computational studies of all tested ribozymes and an additional 61 artificial molecules were taken. A similar approach has been used for computational mutagenesis studies of a hammerhead ribozyme catalysis [14,40].

The defined catalytic core parameters are specific distances and angles between atoms, describing ribozyme active conformation. A selection of geometrical parameters was based on certain generally accepted mechanistic assumptions, with regard to the role of G12 and G8 as the general base and acid respectively [11]. We used structural assumptions for hammerhead ribozyme catalytic activity: Watson–Crick base pairing between G8 and C3, the hydrogen bond network between C17 and G5, and the stacking interactions between G8 and C1.1 [14]. The active conformation can be examined by the distribution of nucleophile in-line attack angle and distance [41]. Interestingly, comparison of HHgfp ribozymes indicated that the distance between C17(2′O) and G12(N1) was the largest in HHgfp-2, which was the most efficient ribozyme within the cells. The D1/D2 ratio describes directional changes between these parameters within the catalytic core. It is not entirely consistent with the mechanism of general acid–base catalysis. Our results suggest inverted order of reaction phases, with regard to 5′O protonation as initial step. It has been proposed that a departure of the leaving group can be rate-determining and that protonation of the 5′O would have to occur prior to the formation of the pentaco-ordinated dianion [42]. That explains why HHgfp-2 shows the highest activity by having an advantage in the form of predestined first step of transesterification reaction. Computational simulations, especially the D1/D2 ratio. are consistent with the experimental data.

The comparison of HH-0 and HH-2 parameters and in silico tested artificial minimal ribozymes suggested an influence of target sequence on catalytic core constraints. A sequence-dependent effect on ribozyme efficiency was postulated previously to be a result of differences in product dissociation [43]. However, it was also suggested that it may be due to slight changes in the structure [4]. The tolerance to certain core nucleotide variations provided evidence for a complex interplay of secondary and tertiary structure elements that lead, mediated by long-range effects, to an individual modulation of the local structure in the catalytic core [44]. It is probable that such sequence dependency is also important in the case of other catalytic RNAs.

Unique distribution of geometrical properties is specific and can describe the molecule, so can be called the ribozyme fingerprint. Such a ribozyme profile is a set of data that reflect its catalytic core, which can be used as an identifier and help to predict its activity in vivo (Supplementary Table S1). Among the proposed parameters, the ratio of D1 and D2 distances seems the most promising. There is also a high correlation between stem I–II adjacent atom distances and ribozyme intracellular activity. The closer the stems I and II are to each other, the higher is the D1/D2 ratio together with ribozyme efficiency. As for now, we can state that this method of activity prediction is consistent for hammerhead ribozymes acting in trans against GUC or AUC sites in animal cell lines, modelled as type I in cis ribozymes. It is a very helpful system for in silico selection of highly active molecules against a specific target. It presents an interesting and reliable alternative for traditional ribozyme design. Close correlation between structure modelling and intracellular experiments results shows that improved RNA design supported by computational analysis could be a way to predict activity.

Abbreviations

     
  • 3D

    three-dimensional

  •  
  • CRYAB

    αB-crystallin

  •  
  • DMEM

    Dulbecco's modified Eagle's medium

  •  
  • GAPDH

    glyceraldehyde 3-phosphate dehydrogenase

  •  
  • GFP

    green fluorescent protein

  •  
  • GLI1

    glioma-associated oncogene 1

  •  
  • MFE

    minimal free energy

  •  
  • qPCR

    quantitative PCR

  •  
  • TLR

    tetraloop receptor

  •  
  • Tm

    melting temperature

  •  
  • TSM

    tertiary stabilizing motif

AUTHOR CONTRIBUTION

Marta Gabryelska, Eliza Wyszko, Maciej Szymański and Jan Barciszewski designed the research. Marta Gabryelska performed the research. Marta Gabryelska, Eliza Wyszko and Mariusz Popenda analysed the data. Marta Gabryelska and Jan Barciszewski wrote the paper.

FUNDING

This study was supported by the Polish Ministry of Science and Higher Education [grant numbers N N302 220535 and N N301 411238]. M.M.G. was a scholarship holder within the project ‘Scholarship support for Ph.D. students specializing in majors strategic for Wielkopolska's development’, Sub-measure 8.2.2 Human Capital Operational Programme, co-financed by the European Union under the European Social Fund.

References

References
1
Westhof
E.
Masquida
B.
Jaeger
L.
RNA tectonics: towards RNA design
Fold Des.
1996
, vol. 
1
 (pg. 
R78
-
R88
)
2
Perreault
J.
Weinberg
Z.
Roth
A.
Popescu
O.
Chartrand
P.
Ferbeyre
G.
Breaker
R. R.
Identification of hammerhead ribozymes in all domains of life reveals novel structural variations
PLoS Comp. Biol.
2011
, vol. 
7
 pg. 
e1002031
 
3
Uhlenbeck
O. C.
A small catalytic oligoribonucleotide
Nature
1987
, vol. 
328
 (pg. 
596
-
600
)
4
Amarzguioui
M.
Prydz
H.
Hammerhead ribozyme design and application
Cell. Mol. Life Sci.
1998
, vol. 
54
 (pg. 
1175
-
1202
)
5
Ferré-D’Amaré
A. R.
Sott
W. G.
Small self-cleaving ribozymes
Cold Spring Harbor Perspect. Biol.
2010
, vol. 
2
 pg. 
a003574
 
6
Khvorova
A.
Lescoute
A.
Westhof
E.
Jayasena
S. D.
Sequence elements outside the hammerhead ribozyme catalytic core enable intracellular activity
Nat. Struct. Biol.
2003
, vol. 
10
 (pg. 
708
-
712
)
7
Saksmerprome
V.
Roychowdhury-Saha
M.
Jayasena
S.
Khvorova
A.
Burke
D. H.
Artificial tertiary motifs stabilize trans-cleaving hammerhead ribozymes under conditions of submillimolar divalent ions and high temperatures
RNA
2004
, vol. 
10
 (pg. 
1916
-
1924
)
8
Dufour
D.
de la Pena
M.
Gago
S.
Flore
R.
Gallego
J.
Structure-function analysis of the ribozymes of chrysanthemum chlorotic mottle viroid: a loop-loop interaction motif conserved in most natural hammerheads
Nucleic Acids Res.
2009
, vol. 
3
 (pg. 
368
-
381
)
9
Shepotinovskaya
I.
Uhlenbeck
O. C.
Enhanced product stability in the hammerhead ribozyme
Biochemistry
2010
, vol. 
49
 (pg. 
4494
-
4500
)
10
Scott
W. G.
Finch
J. T.
Klug
A.
The crystal structure of an all-RNA hammerhead ribozyme: a proposed mechanism for RNA catalytic cleavage
Cell
1995
, vol. 
81
 (pg. 
991
-
1002
)
11
Martick
M.
Scott
W. G.
Tertiary contacts distant from the active site prime a ribozyme for catalysis
Cell
2006
, vol. 
126
 (pg. 
309
-
320
)
12
Boots
J. L.
Canny
M. D.
Azimi
E.
Pardi
A.
Metal ion specificities for folding and cleavage activity in the Schistosoma hammerhead ribozyme
RNA
2008
, vol. 
14
 (pg. 
2212
-
2222
)
13
Nelson
J. A.
Uhlenbeck
O. C.
Minimal and extended hammerheads utilize a similar dynamic reaction mechanism for catalysis
RNA
2008
, vol. 
14
 (pg. 
43
-
54
)
14
Lee
T.
York
D. M.
Computational mutagenesis studies of hammerhead ribozyme catalysis
J. Am. Chem. Soc.
2010
, vol. 
132
 (pg. 
13505
-
13518
)
15
Fedoruk-Wyszomirska
A.
Szymański
M.
Wyszko
E.
Barciszewska
M. Z.
Barciszewski
J.
Highly active low magnesium hammerhead ribozyme
J. Biochem.
2009
, vol. 
145
 (pg. 
451
-
459
)
16
Livak
K. J.
Schmittegen
T. D.
Analysis of relative gene expression data using real-time quantitative PCR and the 2−ΔΔCT method
Methods
2001
, vol. 
25
 (pg. 
402
-
408
)
16a
Mathews
D. H.
Sabina
J.
Zuker
M.
Turner
D. H.
Expanded sequence dependence of thermodynamic parameters improves prediction of RNA secondary structure
J. Mol. Biol.
1999
, vol. 
288
 (pg. 
911
-
940
)
17
Popenda
M.
Szachniuk
M.
Antczak
M.
Purzycka
K. J.
Lukasik
P.
Bartol
N.
Blazewicz
J.
Adamiak
R. W.
Automated 3D structure composition for large RNAs
Nucleic Acids Res.
2012
, vol. 
40
 pg. 
e112
 
18
Rossi
M.
Magnoni
L.
Miracco
C.
Mori
E.
Tosi
P.
Pirtoli
L.
Tini
P.
Oliveri
G.
Cosci
E.
Bakker
A.
β-Catenin and Gli1 are prognostic markers in glioblastoma
Cancer Biol. Ther.
2011
, vol. 
11
 (pg. 
753
-
761
)
19
Goplen
D.
Bougnaud
S.
Rajcevic
U.
Bøe
S. O.
Skaftnesmo
K. O.
Voges
J.
Enger
P. Ø.
Wang
J.
Tysnes
B. B.
Laerum
O. D.
, et al. 
αB-crystallin is elevated in highly infiltrative apoptosis-resistant glioblastoma cells
Am. J. Pathol.
2010
, vol. 
177
 (pg. 
1618
-
1628
)
20
Chen
X.
Denison
L.
Levy
M.
Ellington
A. D.
Direct selection for ribozyme cleavage activity in cells
RNA
2009
, vol. 
15
 (pg. 
2035
-
2045
)
21
Nashimoto
M.
Correct folding of a ribozyme induced by nonspecific macromolecules
Eur. J. Biochem.
2000
, vol. 
267
 (pg. 
2738
-
2745
)
22
Nakano
S.
Karimata
H. T.
Kitagawa
Y.
Sugimoto
N.
Facilitation of RNA enzyme activity in the molecular crowding media of cosolutes
J. Am. Chem. Soc.
2009
, vol. 
131
 (pg. 
16881
-
16888
)
23
Kilburn
D.
Roh
J. H.
Guo
L.
Briber
R. M.
Woodson
S. A.
Molecular crowding stabilizes folded RNA structure by the excluded volume effect
J. Am. Chem. Soc.
2010
, vol. 
132
 (pg. 
8690
-
8696
)
24
Bertrand
E. L.
Rossi
J. J.
Facilitation of hammerhead ribozyme catalysis by the nucleocapsid protein of HIV-1 and the heterogenous nuclear ribonucleoprotein A1
EMBO J.
1994
, vol. 
13
 (pg. 
2904
-
2912
)
25
Sioud
M.
Jespersen
L.
Enhancement of hammerhead ribozyme catalysis by glyceraldehyde-3-phosphate dehydrogenase
J. Mol. Biol.
1996
, vol. 
257
 (pg. 
775
-
789
)
26
Earnshaw
D. J.
Gait
M. J.
Hairpin ribozyme cleavage catalyzed by aminoglycoside antibiotics and the polyamine spermine in the absence of metal ions
Nucleic Acids Res.
1998
, vol. 
26
 (pg. 
5551
-
5561
)
27
Real
A. N.
Greenall
R. J.
Influence of spermine on DNA conformation in a molecular dynamics trajectory of d(CGCGAATTCGCG)2: major groove binding by one spermine molecule delays the A→B transition
J. Biomol. Struct. Dyn.
2004
, vol. 
21
 (pg. 
469
-
488
)
28
Roychowdhury-Saha
M.
Roychowdhury
S.
Burke
D. H.
Conformational heterogeneity and the determinants of tertiary stabilization in the hammerhead ribozyme from Dolichopoda cave crickets
RNA Biol.
2011
, vol. 
8
 (pg. 
1
-
11
)
29
Kaddour
H.
Vergne
J.
Hervé
G.
Maurel
M. C.
High-pressure analysis of a hammerhead ribozyme from Chrysanthemum chlorotic mottle viroid reveals two different populations of self-cleaving molecule
FEBS J.
2011
, vol. 
278
 (pg. 
3739
-
3747
)
30
Herschlag
D.
RNA chaperons and the RNA folding problem
J. Biol. Chem.
1995
, vol. 
270
 (pg. 
20871
-
20874
)
31
Buskiewicz
I. A.
Burke
J. M.
Folding of the hammerhead ribozyme: pyrrolo-cytosine fluorescence separates core folding from global folding and reveals a pH-dependent conformational change
RNA
2012
, vol. 
18
 (pg. 
434
-
448
)
32
Santoro
S. W.
Joyce
G. F.
A general purpose RNA-cleaving DNA enzyme
Proc. Natl. Acad. Sci. U.S.A.
1997
, vol. 
94
 (pg. 
4262
-
4266
)
33
Kurreck
J.
Bieber
B.
Jahnel
R.
Erdmann
V. A.
Comparative study of DNA enzymes and ribozymes against the same full-length messenger RNA of the vanilloid receptor subtype I
J. Biol. Chem.
2002
, vol. 
277
 (pg. 
7099
-
7107
)
34
Laing
C.
Wen
D.
Wang
J. T.
Schlick
T.
Predicting coaxial helical stacking in RNA junctions
Nucleic Acids Res.
2012
, vol. 
40
 (pg. 
487
-
498
)
35
Burke
D. H.
Greathouse
S. T.
Low-magnesium, trans-cleavage activity by type III, tertiary stabilized hammerhead ribozymes with stem 1 discontinuities
BMC Biochem.
2005
, vol. 
6
 pg. 
14
 
36
Carbonell
A.
Flores
R.
Gago
S.
Trans-cleaving hammerhead ribozymes with tertiary stabilizing motifs: in vitro and in vivo activity against a structured viroid RNA
Nucleic Acids Res.
2011
, vol. 
39
 (pg. 
2432
-
2444
)
37
L’Huillier
P. J.
Davis
S. R.
Bellamy
A. R.
Cytoplasmic delivery of ribozymes leads to efficient reduction of in α-lactalbumin mRNA levels in C127I mouse cells
EMBO J.
1992
, vol. 
11
 (pg. 
4411
-
4418
)
38
Crissel
P.
Thompson
S.
James
W.
Inhibition of HIV-1 replication by ribozymes that show poor activity in vitro
Nucleic Acids Res.
1993
, vol. 
21
 (pg. 
5251
-
5255
)
39
Schultes
E. A.
Spasic
A.
Mohanty
U.
Bartel
D. P.
Compact and ordered collapse of randomly generated RNA sequences
Nat. Struct. Mol. Biol.
2005
, vol. 
12
 (pg. 
1130
-
1136
)
40
Chi
Y.
Martick
M.
Lares
M.
Kim
R.
Scott
W. G.
Kim
S.
Capturing hammerhead ribozyme structures in action by modulating general base catalysis
PLoS Biol.
2008
, vol. 
6
 (pg. 
2060
-
2068
)
41
Lee
T. S.
Giambaşu
G. M.
Sosa
C. P.
Martick
M.
Scott
W. G.
York
D. M.
Threshold occupancy and specific cation binding modes in the hammerhead ribozyme active site are required for active conformation
J. Mol. Biol.
2009
, vol. 
388
 (pg. 
195
-
206
)
42
Emilsson
G. M.
Nakamura
S.
Roth
A.
Breaker
R. R.
Ribozyme speed limits
RNA
2003
, vol. 
9
 (pg. 
907
-
918
)
43
Fedor
M. J.
Uhlenbeck
O. C.
Substrate sequence effects on hammerhead RNA catalytic efficiency
Proc. Natl. Acad. Sci. U.S.A.
1990
, vol. 
87
 (pg. 
1668
-
1672
)
44
Przybilski
R.
Hammann
C.
The tolerance to exchanges of the Watson-Crick base pair in the hammerhead ribozyme core is determined by surrounding elements
RNA
2007
, vol. 
13
 (pg. 
1625
-
1630
)

Supplementary data