ABSTRACT
We report here the in vitro use of hammerhead ribozymes as an approach to the gene therapy of osteogenesis imperfecta (OI). Our strategy for the treatment of this dominant genetic disorder is based on selective reduction of the level of the mRNA transcripts from the mutant allele. We studied the in vitro cleavage activity of five different hammerhead ribozymes targeted against synthetic transcripts of two naturally occurring human collagen mutations and against a point mutation introduced into a construct containing a portion of the mouse COL1A1 gene. This is the first demonstration that ribozyme cleavage is absolutely dependent on the presence of the ribozyme cleavage site introduced by the disease-causing mutation. Cleavage specificity and activity were unchanged when the cleavage site was located in transcripts of progressively longer length. Cleavage efficiency depended directly on the ratio of ribozyme/substrate, as well as on the time and temperature of incubation. We investigated the competitive effects of both total RNA and normal synthetic transcripts on ribozyme cleavage activity. The ribozyme was able to localize and cleave its specific target even in the presence of a vast excess of total RNA. However, cleavage efficiency was linearly inhibited by the presence of a non- cleavable competitor substrate which contained a ribozyme binding site identical to the site present in the cleavable target. Although this competition could be eliminated by introducing a mismatch into one ribozyme binding arm, the presence of the mismatch decreased ribozyme cleavage efficiency. The mutation- specificity of ribozyme cleavage demonstrated in this work provides support for in vivo studies aimed at ribozyme development as a treatment for dominant negative genetic disorders.
The general approach to gene therapy for recessive genetic disorders is based on the goal of replacing the missing gene product in a specific tissue (1 ). Recessive disorders are most commonly due to an absent or inactive form of an enzyme or transport protein. The replacement of even a fraction of the normal protein level would be expected to restore adequate function for relief of clinical symptoms.
A different approach is required for the gene therapy of dominant genetic disorders, including defects of structural proteins of the extracellular matrix (2 ). These disorders are characterized by the presence of a defective protein, produced by the mutant allele, which negatively affects cellular metabolism or extracellular environment. The presence of the normal allele protein does not prevent the negative effect of the mutant one. Thus, introducing an additional copy of the normal allele into the cell would not be sufficient for gene therapy, although it might be useful as a supplemental treatment. One strategy for preventing a dominant negative effect consists of reducing the level of the mutant mRNA transcript using antisense techniques. This approach would result in the production of protein almost exclusively from the normal allele, although the total amount of normal protein would be reduced compared with an unaffected individual with two normal alleles.
In the present work, we continued the investigation of this laboratory on an antisense approach (3 ) to the gene therapy of osteogenesis imperfecta (OI). OI is a dominant negative genetic disorder caused by mutations in either of the [alpha] chains of type I collagen. It is a generalized disorder of connective tissue, which has as its most significant clinical feature the susceptibility of affected individuals to bone fractures from very mild trauma (4 ). Biochemically, OI is an excellent candidate for an antisense approach to gene therapy. The study of >150 naturally occurring mutations has shown that a clinically significant bone dysplasia results from the synthesis of structurally abnormal chains, while a null allele is associated with a mild clinical phenotype (5 ). The fact that the mildest form of the disease is associated with a null [alpha] chain (6 ) suggests that it would not be detrimental to patients to suppress mutant mRNA transcripts and that such treatment could ameliorate the severe clinical forms of OI.
Our previous investigation (3 ) focused on the use of linear thio- oligodeoxynucleotides. Linear oligonucleotides can significantly decrease the levels of mutant mRNA (7 ,8 ). However, in our experiments the maximum reduction of mutant protein chain was 50%. Furthermore, the linear oligonucleotides did not have adequate specificity to distinguish mutant transcripts and caused unacceptable decreases in the level of normal allele mRNA. Because the coding sequences for collagen are highly repetitive, a very high degree of specificity is required of the antisense agent.
Ribozymes (2 ) represent a good tool to selectively reduce the intracellular level of a specific RNA. They have been successfully used to cleave a number of viral and cellular RNAs both in vitro and in vivo, although allele specificity was not required in these situations (9 -11 ). Cleavage by trans-acting ribozymes requires three components (12 ). Two of these are part of the ribozyme itself: the central hairpin sequence which is the catalytic core for cleavage (13 ) and the binding arms which flank the catalytic core and can be designed to be complementary to any target sequence (14 ). The third component for ribozyme action is on the target RNA and consists of a three nucleotide cleavage site, most often GUC (15 ,16 ) in nature although other trinucleotides can be cleaved as well (17 ,18 ). The binding arms position the ribozyme on the target so that the catalytic core cleaves after the third nucleotide in the cleavage site. This combination of requirements for a binding site and a cleavage site enhances the specificity capabilities of ribozymes in comparison with linear oligonucleotides. In addition, the catalytic capacity of ribozymes provides the potential for increased efficiency and stability.
It was, therefore, crucial to determine whether ribozymes could achieve the mutation-specificity which would be required for therapy of a dominant disorder, in which the normal allele mRNA must remain undigested. As a first step in our ribozyme-based approach to the gene therapy of OI, we studied the in vitro cleavage activity of five different hammerhead ribozymes. They were targeted against synthetic transcripts of two naturally occurring collagen mutations which we have previously described in unrelated patients (19 ,20 ) and against a point mutation introduced into a construct containing a portion of the mouse COL1A1 gene. In all these cases, the genetic defect creates a new ribozyme cleavage site in the mutant transcript that is not present in the normal one.
We selected two unrelated patients whose naturally occurring mutations are each responsible for the formation of a new ribozyme cleavage site in the mutant allele. In patient A, a genomic deletion extending from the last three nucleotides of exon 33/34 to the middle of intron 36 in the [alpha]1 (I) chain (20 ) results in the retention of a portion of intron 36, which includes the ribozyme target site (GUC). In patient B, a G -> T substitution occurring at position 2432 of pro[alpha]2(I) cDNA (19 ) creates two new ribozyme cleavage sites (GUU and UUC). For the two patients studied, cDNA fragments from both normal and mutant alleles were subcloned in pGEM-3Z vector. For patient A, the insert spanned nt 2388-2687 of pro[alpha]1(I). A sense RNA target of 348 nt was generated by in vitro transcription from the HindIII-digested mutant subclone using T7 RNA polymerase. Antisense mutant RNA was generated from the EcoRI digested mutant construct using SP6 RNA polymerase. For patient B, the insert spanned nt 2101-3207 of pro[alpha]2(I). Normal and mutant RNA targets of 512, 689 and 920 nt were generated from StyI, AccI and PstI linearized pGEM-3Z construct, respectively, using T7 RNA polymerase.
For the mouse COL1A1 genomic construct (Forlino and Marini, unpublished data), DNA fragments were subcloned in pBSII plasmid (Stratagene). Two different targets were subsequently generated by site specific mutagenesis. The cleavable mutant target contains a G -> T transversion responsible for a Gly349 -> Cys substitution and a C -> T transition creating a new GUC ribozyme cleavage site. The non-cleavable RNA contains only the G -> T transversion while the normal RNA contains neither the transversion nor the transition (see Fig. 5 ). From each clone, template DNA for in vitro transcription was generated by PCR. The upstream primer contains 37 nt of the T7 RNA polymerase promoter (bold) and 17 nt corresponding to the sequence of mouse COL1A1 gene at the boundary between intron 22/exon 23 (5'CGGGATCCGGATCCTAATACGACTCACTATAGGGAGACTTCTCTCCACTTAGGG3"). The downstream primer is complementary to nt 1584-1603 of mouse COL1A1 cDNA (5'TGCCAGGACTGCCAGTGAGA3"). PCR cycles were: 1 min at 95oC; 1 min at 65oC; and 1.5 min at 72oC for 35 cycles. After PCR, the 270 bp DNA produced from each clone was purified on an 8% polyacrylamide gel and used for in vitro transcription.
Radioactively labelled RNA substrates were generated by in vitro transcription with T7 RNA polymerase (Promega) in the presence of 3 [mu]l of [32P]UTP (Amersham, 800 Ci/mmol). After digestion with RNase-free DNase I (Promega), RNAs were concentrated and desalted using Microcon microconcentrators (Amicon). To check the integrity and the specificity of the RNA product, an aliquot of each sample was electrophoresed on a 10% polyacrylamide-7 M urea gel. The bands were excised and counted in a Beckman-LS 9000 [beta]-counter. RNA concentration was calculated from the final specific activity of UTP and the maximum number of nucleotides incorporated/molecule. Non-radioactive RNAs were quantified spectrophotometrically.
Hammerhead ribozymes were designed using 9 nt binding arms and a catalytic core which has been reported previously (11 ,12 ). The sequences of the five ribozymes are shown in Table 1 . All ribozymes were chemically synthesized and were the gift of Dr K. Bashar Mullah, Applied Biosystems, Inc. Ribozymes were quantified spectrophotometrically.
Table 1
Ribozymes and substrates were combined in 8 [mu]l, then heated at 95oC for 4 s in 62.5 mM Tris-HCl pH 8.0 and transferred to ice for 1 min. Reactions were initiated by adding 2 [mu]l 100 mM MgCl2 (final MgCl2 concentration 20 mM) and transferring the reaction tubes to 37 or 50oC. Reactions were terminated with the addition of 10 [mu]l of stop mix containing 80% formamide, 10 mM EDTA pH 8.0 and 1 mg/ml of both bromophenol blue and xylene cyanol. In the experiments carried out using non-radioactive RNA, a fixed amount of substrate (2 pmol) was used. For [32P]UTP labelled RNA experiments, 1 pmol of target was used. Reactions containing radiolabelled RNAs were electrophoresed on a 6 cm denaturing mini-gel (10% polyacrylamide-7 M urea) run at a constant power of 12 W. The gels were fixed, dried and the bands corresponding to undigested substrate (S) and products (P) were excised and counted in a Beckman-LS9000 [beta]-counter. The percent of cleavage was determined as (P/P+S) * 100. Reactions containing unlabelled RNA were electrophoresed as above, stained with ethidium bromide and photographed. Band intensity was determined by densitometric scanning, and the (P/P+S) * 100 value was calculated.
All competition experiments were performed at 37oC for 75 min in 10 [mu]l as described. We used 0.1 pmol (7 ng) of cleavable mutant RNA and a 1:1 molar ratio of ribozyme/cleavable target. The molar ratio of cleavable mutant RNA/competitor ranged from 16:1 to 1:8. The weight ratio of cleavable mutant RNA/total RNA was 1:1 to 1:500. Total RNA was extracted from normal human fibroblasts using the method of Chomczynski and Sacchi (21 ).
Ribozyme GUC2m carries a mismatch localized in the middle of its 3" arm, but is otherwise identical to GUC2. In separate reactions, 0.1 pmol of each ribozyme was incubated with 0.1 pmol of 32P-labelled cleavable mutant RNA.
P values were calculated using an unpaired t-test in the MS Excel program.
The eventual use of hammerhead ribozymes as therapeutic agents for dominant genetic disorders requires that their cleavage activity be allele-specific. We first investigated the specificity of cleavage of ribozyme GUC1 targeted against a GUC triplet, which is present in the retained intron 36 sequence in the mRNA transcript from the mutant allele of patient A. A fixed amount of an unlabelled 348 nt fragment of sense mutant RNA was incubated with increasing amounts of ribozyme, from a ribozyme/ substrate molar ratio of 1:1 to 125:1. As a negative control we used the 348 nt antisense mutant synthetic transcript. The completed reactions were electrophoresed on a denaturing polyacrylamide gel and stained with ethidium bromide. Figure 1 A shows the results obtained. Digestion of the mutant sense RNA target yielded two fragments of 256 and 92 nt, as expected from cleavage at the specific ribozyme cleavage site. These fragments were not detected when the negative control mutant antisense RNA was incubated with the same ribozyme. As controls for RNA background degradation, we included samples of RNA loaded directly after in vitro transcription and aliquots which had undergone all experimental procedures in the absence of ribozyme.
We next investigated the factors affecting the efficiency of cleavage of ribozymes GUC1, UUC and GUU. The digestion sets for patients A and B were done at different temperatures (37 and 50oC) and times of incubation (25, 50 and 75 min). The extent of cleavage by the ribozymes was greater with increasing ratio of ribozyme/substrate, longer incubation time and increased temperature (Fig. 2 ). Under our experimental conditions, the extent of cleavage approximately doubles with each 5-fold increase of ribozyme from the 1:1 to the 25:1 ribozyme/substrate ratio (Fig. 2 A and B). Cleavage is approximately linear with incubation times of up to 75 min (Fig. 2 C and D). As one would predict, the effect of increased temperature varied with the particular cleavage substrate. Cleavage of the GUC site almost doubled at 50oC (Fig. 2 B and D), while the extent of cleavage of the other substrate increased only slightly.
We have described here our continuing development of antisense methodologies as a therapeutic approach to the gene therapy of dominant negative disorders. In these experiments, we have turned our attention to the hammerhead ribozyme (22 ) as the antisense agent. Hammerhead ribozymes are being developed as trans-acting antisense vehicles for HIV and cancer therapeutics. In these fields, the goal of ribozyme action is to suppress all transcription from a particular disease-causing gene. For dominant negative genetic disorders, which are characterized by the presence of one normal and one mutant allele, two approaches are possible. In a two-step approach, the ribozyme could be used in the first step to down-regulate expression from both the mutant and normal alleles by targeting a common region of the gene transcript. This approach would require a second step in which gene expression was replaced using a construct that did not contain the ribozyme target site. The feasibility of the first step in this scheme has been demonstrated (23 ) in cultured cells using transient transfection of ribozymes targeted to the fibrillin mRNA, the defective molecule in Marfan syndrome. This two-step approach has the drawbacks of requiring efficient and sustained suppression of the endogenous message, expression of the exogenous constructs, and a balanced co-ordination of their respective levels. The alternative approach is a one-step procedure. In this scheme, the ribozyme is targeted directly to the novel ribozyme cleavage site generated by the disease-causing mutation. This plan requires that the ribozyme be able to achieve allele-specific cleavage.
We present here our in vitro studies of both the cleavage specificity and binding affinity of synthetic ribozymes using RNA substrates which differ by only a single nucleotide. The present study is the first demonstration, in the field of genetic diseases, of point mutation-specific cleavage by hammerhead ribozymes. We tested ribozyme specificity against four cleavage sites located in three transcripts, and their corresponding controls. In the first substrate, the ribozyme was targeted to a novel cleavage site located in a retained intron from a patient with type III OI (20 ). This was a modest test of specificity since the entire 156 nt of retained intron is exclusive to the mutant transcript. In the second synthetic target, two different ribozymes were directed against two novel ribozyme cleavage sites generated by one naturally occurring point mutation in an OI type IV patient (19 ). This specificity test was more stringent since the only difference between the synthetic sense mutant target and the synthetic sense normal control was the single nucleotide mutation causing the disease. Only the mutant target was cleaved at the two novel sites; the normal transcript remained intact in spite of the presence of the complete ribozyme binding site. A similar result was obtained with a synthetic transcript from a murine construct with a novel ribozyme cleave site introduced by site-directed mutagenesis. This demonstrates that the mutation specificity of cleavage of hammerhead ribozymes is virtually absolute in vitro.
We tested the effects of ribozyme/substrate ratio, length of target RNA and temperature of reaction on the extent of ribozyme cleavage. The extent of cleavage was most influenced by the ribozyme/substrate ratio. We verified that the reaction was in the linear range for all ribozymes up to 75 min (Fig. 2 C). Then we varied ribozyme/substrate ratio and temperature of reaction. These assays were performed using relative ribozyme excess because of the slow ribozyme turnover in vitro. The conversion of substrate to product was demonstrated to have the expected hyperbolic kinetics. There was an approximate doubling of the percent of product obtained with each 5-fold increase in the ribozyme/substrate ratio from 1:1 to 25:1 after which the ribozyme concentration became saturating (Fig. 2 A and B). The ribozyme concentrations were at 50% of the saturation level at a ribozyme/substrate ratio of ~5-10:1, suggesting that the overall catalytic efficiency of the ribozymes in vitro was low.
Figure 2 also shows that there is some dependence of reaction efficiency on the particular target sequence. At 37oC, the GUC target is cleaved to approximately the same extent as the UUC/GUU target. We found that cleavage of the GUC site was almost doubled at 50oC (Fig. 2 ), while the cleavage of the UUC/GUU template increased only slightly. We speculate that this variation with temperature may be due to different effects on the secondary structure surrounding the cleavage sites. We further speculate that such unfolding may in part mimic the intracellular status of the mRNA, in which it is coated with binding proteins and translational machinery. A few RNA binding proteins have been demonstrated to increase ribozyme binding in vitro (24 ,25 ) presumably by modifying secondary structure. Such a role could be hoped to enhance ribozyme cleavage in the cell.
Our investigation of the effect of increasing target length (Fig. 3 and Table 0 2 ) was also encouraging for applications to intracellular cleavage. The longer substrates are more representative of the in vivo situation where the ribozyme must cleave a long, highly structured mRNA. We detected no significant differences in the efficiency of cleavage for ribozymes GUU and UUC as the template length was increased. This data is encouraging for the in vivo cleavage of these particular sites, but does not necessarily extrapolate to sites in other positions of the same mRNA.
These reactions demonstrating mutation-specific cleavage occurred under non-competitive conditions for the target transcripts. We sought to estimate the effect of competing binding specificity by reactions in which more than one potential substrate was mixed together. To partially model the intracellular environment, ribozyme digestion of a cleavable target was carried out in the presence of increasing amounts of total RNA. There was no decrease in cleavage from this non-specific competition, confirming the ribozyme's ability to localize and bind its target even in the presence of a vast excess of unrelated RNA. There was a clear and direct competition effect from a non-cleavable target with an identical binding site. This situation simulates the presence of the normal allele transcript in the cell. When equal amounts of mutant and normal transcripts are present, cleavage of the mutant transcript is decreased by 50%. This competition raises the possibility of two in vivo problems for ribozymes in the treatment of genetic disease. First, the normal allele may effectively sequester ribozyme, making it unavailable for cleavable target. Since the kinetics of spontaneous ribozyme release from a non-cleavable binding site may be unfavorable, it would be necessary to determine experimentally the extent to which the translational machinery of the cell will remove the ribozyme from the non-cleavable target. The second potential problem in vivo is that a bound ribozyme might have an antisense effect on the normal transcript, making it susceptible to digestion by cellular enzymes. However, these RNA:RNA structures would not be good targets for RNase H, which recognizes a DNA:RNA hybrid target (26 ).
If the translational machinery does not release the ribozyme from the binding site on the normal allele transcript, then our experiments suggest that cycling of the ribozyme from the binding site can be further enhanced by the introduction of a mismatch between the target and one ribozyme arm. This decrease in binding affinity does, however, decrease the efficiency with which the mutant allele transcript is cleaved. In order to simultaneously decrease binding efficiency to the normal transcript and maintain cleavage efficiency of mutant transcript, the kinetic balance between binding, cleavage and release must be adjusted for each individual ribozyme. Factors such as the position of the mismatch in the binding arm and the nucleotide composition surrounding the mismatch (27 ,28 ) will play a significant role. We have initiated experiments on cultured fibroblast from OI patients to resolve these and other issues about ribozyme action. The virtually absolute mutation-specificity of cleavage which was achieved in vitro is encouraging for the future use of ribozymes as therapeutic agents in dominant genetic disorders.
We thank Dr K.Bashar Mullah for providing all the ribozymes used in this work, Dr J.Rossi for stimulating discussions on ribozyme design and action and Dr Q.Wang for preliminary experiments on patient A.
*To whom correspondence should be addressed. Tel: +1 301 496 6683; Fax: +1 301 402 0234; Email: oidoc@helix.nih.gov
5" binding arm
Catalytic core
3" binding arm
Ribozyme GUC1
5"AGGAUGGCU
CUGAUGAGUCCGUGAGGACGAA
ACGCCUUUG3"
Ribozyme UUC
5"GACCAUUUG
CUGAUGAGUCCGUGAGGACGAA
AACAGCUGG3"
Ribozyme GUU
5"ACCAUUUGG
CUGAUGAGUCCGUGAGGACGAA
ACAGCUGGG3"
Ribozyme GUC2m
5"CUUCACCGG
CUGAUGAGUCCGUGAGGACGAA
ACGACAAGC3"
Ribozyme GUC2
5"CUUCACCGG
CUGAUGAGUCCGUGAGGACGAA
ACGACCAGC3"
REFERENCES