Hammerhead ribozymes selectively suppress mutant type I collagen mRNA in osteogenesis imperfecta fibroblasts

Cell culture and transfection

Dermal fibroblasts from an OI patient, G85V (16), were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Gibco BRL, Grand Island, NY) supplemented with 10% fetal calf serum. Cells were plated in 6-well plates and transfected at ∼70% confluency. Cells were transfected for 3.5 h at 37°C with fresh serum-free medium containing 0.14 nM premixed plasmid DNA and 2 µg/ml cationic lipid (LipofectAMINE; Gibco BRL). Post-transfection the medium was replaced with 2 ml of DMEM supplemented with 10% fetal calf serum. Transfected cells were selected by addition of 0.4 mg/ml G418 (Gibco BRL) for 3 weeks. Stable cell lines were maintained in medium containing 0.2 mg/ml G418. Clonal cell lines were generated by plating stably transfected cells in 96-well plates and culturing in F10 medium (Zazo Diagnostic Laboratories, Lexington, KY) supplemented with 10% fetal calf serum and 0.4 mg/ml G418.

To determine cell growth rates, untransfected and stably transfected (pCI.neo vectors) cells were plated in triplicate at a density of 5 × 10³ cells/well of a 24-well plate in DMEM supplemented with 10% fetal calf serum. Cells were trypsinized and counted at various times over a period of 11 days. Linear analyses of growth rates were performed using a normal Z-test.

Construction of plasmids

For in vitro experiments we prepared plasmids encoding substrate RNA. Inserts of 226 bp containing 117 bp of α1(I) collagen cDNA derived from the G85V cell line and 109 bp of flanking vector sequence were amplified by PCR using the primer pair 85S1 and 85AS1 (Table 1). The PCR products were ligated into vector pCR2.1 (Invitrogen, Carlsbad, CA). Inserts were sequenced and the orientation verified. For plasmids encoding ribozymes, two complementary oligodeoxynucleotides were synthesized for each ribozyme construct. The sequences of the oligodeoxynucleotides, containing SalI sites on both the 5′- and 3′-termini (lower case), the catalytic core (bold sequence) and 13 nt binding arms, are shown in Table 1. Two base changes in the highly conserved catalytic core sequence (underlined bases) were introduced to generate inactive ribozymes (IR). Aliquots (4 µg) of each primer pair (sense and antisense) were mixed in 1× PCR Buffer II (Perkin Elmer-Cetus, Norwalk, CT) to a final volume of 50 µl. The primers were denatured by heating to 95°C for 4 min and annealed by slow cooling to 20°C over 1 h. The fragments were digested with SalI (Gibco BRL) and ligated into the SalI site of the mammalian expression vectors pCI.neo (Promega, Madison, WI) and pHβAPr-1-neo (a gift from Dr L. H. Kedes, Department of Biochemistry, University of Southern California) to produce plasmids pCI.neo/AR, pCI.neo/IR, pHβAPr-1-neo/AR and pHβAPr-1-neo/IR. In the pCI.neo vector, ribozyme sequences were placed downstream of the human CMV immediate early enhancer/promoter region and an intron element (comprising the 5′-donor site from the first intron of the human β-globin gene and the branch and 3′-acceptor site from the intron of an immunogloubulin gene heavy chain variable region) and upstream of the SV40 late polyadenylation site. The pCI.neo vector also contains T7 and T3 RNA polymerase promoters that flank the ribozyme sequences. In the pHβAPr-1-neo vector, ribozyme sequences were cloned downstream of the human β-actin promoter plus 5′-UTR and intervening sequence 1 (IVS1) and upstream of the SV40 late polyadenylation signal. The pCI.neo and pHβAPr-1-neo vectors are predicted to generate ribozyme-containing transcripts of 552 and 1123 nt, respectively, plus an additional poly(A) tail. Plasmid DNA was prepared using an EndoFree Plasmid Maxi Kit (Qiagen, Santa Clarita, CA). Sequence and orientation of the ribozyme templates was verified.

For plasmids containing competitor sequences for quantitative PCR, oligonucleotides were synthesized (Table 1) to generate competitor templates of β-actin, ribozyme and α1(I) collagen cDNA using PCR. Each competitor template contains an additional 20 nt artificial sequence (lower case italic in Table 1) which allows the electrophoretic separation of amplified endogenous cDNA from competitor cDNA. All competitor templates were cloned into vector pCR2.1 (Invitrogen) and sequences and orientation were verified. For construction of the β-actin competitor template, primer pairs BA1.S + BA2.AS and BA3.S + BA4.AS were used to amplify 100 and 166 bp segments, respectively, of β-actin cDNA. Aliquots (1 µl) of each PCR product were added to a second round PCR to amplify a 246 bp competitor template using primers BA1.S and BA4.AS under the same conditions. For construction of the pCI.neo and pHβAPr-1-neo ribozyme competitor templates, primer pairs RZ1.S + RZ2.AS and RZ3.S + RZ4.AS were used to amplify 105 and 66 bp segments, respectively, of pCI.neo vector sequence using the conditions described for the β-actin competitor. Aliquots (1 µl) of the PCR products were combined and a 151 bp fragment was amplified using primers RZ1.S and RZ4.AS. The same procedure was used to generate a 132 bp competitor template of pHβAPr-1-neo/ribozyme using the pHβAPr-1-neo vector as template and primers RZ5.S, RZ6.AS, RZ7.S and RZ8.AS. For construction of the α1(I) collagen competitor template, primers 85S1, 85AS2, 85S2 and 85AS1 were used to PCR amplify a 138 bp α1(I) cDNA fragment derived from OI G85V cells. The α1(I) collagen competitor template includes the G907→T mutation. The competitor constructs are designated BAC (β-actin competitor), pCRZC (pCI.neo/ribozyme competitor), pβRZC (pHβAPr-1-neo/ribozyme competitor) and α1C [α1(I) collagen competitor for OI G85V cells].

In vitro transcription and ribozyme cleavage

Vectors containing substrate or ribozyme templates were linearized with BamHI or NotI, respectively. All transcripts were generated with T7 RNA polymerase (Promega). Substrate transcripts were labeled by incorporation of [α-³²P]UTP (800 Ci/mmol; Amersham Pharmacia Biotech, Cleveland, OH). Specific activity of the [α-³²P]UTP and the base composition of each substrate molecule were used to calculate the substrate concentration. Ribozyme transcripts were quantified spectrophotometrically.

Cleavage reactions contained 1 nM substrate RNA, increasing amounts (1–50 nM) of ribozyme, 20 mM MgCl₂ and 50 mM Tris–HCl, pH 8.0, in a final volume of 10 µl. Substrates and ribozymes were denatured at 95°C for 30 s, then chilled on ice. Cleavage reactions were initiated by the addition of MgCl₂ and were incubated for 2 h at 50°C. Reactions were stopped by addition of loading buffer (80% formamide, 10 mM Na₂EDTA, pH 8.0, and 1 mg/ml each bromophenol blue and xylene cyanol). Cleavage products were analyzed on 10% polyacrylamide–7 M urea gels. Product and substrate fragments were quantitated by densitometry analysis using a Fuji phosphorimager.

RNA extraction and amplification

Prior to RNA extraction cells were incubated in G418-free medium for 48 h. RNA was isolated using TriReagent (Molecular Research Center, Cincinnati, OH) according to the manufacturer’s protocol, then treated with 1 U RQ1 RNase-free DNase (Promega) for 15 min at 37°C and purified using an RNeasy Total RNA kit (Qiagen).

To detect ribozyme expression in cells, 100 ng of total RNA was RT–PCR amplified using 0.25 µM each primer: RZ1.1S + RZ4.AS for pCI.neo/ribozyme; RZ5.S + RZ8.AS for pHβAPr-1-neo/ribozyme. Amplified pCI.neo/ribozyme and pCI.neo vector alone yield fragments of 341 and 287 bp, respectively. Amplified pHβAPr-1-neo/ribozyme and pHβAPr-1-neo vector alone yield fragments of 168 and 114 bp, respectively. PCR products were resolved on a 1.5% agarose gel in 1× TBE, stained with ethidium bromide and visualized under UV light.

A quantitative competitive RT–PCR protocol was designed for quantitation of ribozyme, β-actin and normal and mutant type I collagen expression. Competitor RNAs were co-amplified with each of the endogenous transcripts. They were transcribed using HindIII-linearized plasmid constructs and T7 RNA polymerase in the presence of [α-³²P]UTP. Their concentration was calculated from the specific activity and the number of UTPs per molecule.

Reverse transcriptase reactions were performed in a total volume of 10 µl using 10 ng total RNA, serially diluted competitor RNA (β-actin, 5 × 10⁵–10⁴ molecules; RZ, 5 × 10⁶–10⁵ molecules; α1(I) collagen, 10⁶–5 × 10⁴ molecules), 1× PCR Buffer II (Perkin Elmer-Cetus), 2.5 mM MgCl₂, 12.5 U MuLV reverse transcriptase, 0.25 mM each dNTP, 5 U RNase inhibitor and 0.625 µM antisense primer BA4.AS, RZ4.AS, RZ8.AS or 85AS1. PCRs were performed in a total volume of 25 µl using 10 µl cDNA, 2 mM MgCl₂, 1 µCi [α-³²P]dCTP (3000 Ci/mmol; Amersham Pharmacia Biotech), 1× PCR Buffer II and 0.5 U AmpliTaq Gold (Perkin Elmer-Cetus) and 0.5 µM each primer pair BA1.S + BA4.AS (β-actin), RZ1.S + RZ4.AS (pCI.neo ribozyme), RZ5.S + RZ8.AS (pHβAPr-1-neo ribozyme) and 85S1 + 85AS1 [α1(I) collagen]. Co-amplified β-actin and competitor cDNA yield fragments of 226 and 246 bp, respectively. Co-amplified pCI.neo/ribozyme and competitor yield fragments of 187 and 151 bp, respectively. Co-amplified pHβAPr-1-neo/ribozyme and competitor yield fragments of 168 and 132 bp, respectively. Restriction endonuclease digestions were performed to separate amplified normal and mutant type I collagen cDNAs. The amplified α1(I) product was digested for 16 h with BstNI. The digested samples were desalted, concentrated using a microcon-10 spin column (Amicon, Beverly, MA) and electrophoresed on 10% polyacrylamide gels. The normal α1(I) collagen cDNA was cleaved, yielding 78 and 39 bp fragments, whereas the competitor and mutant α1(I) collagen cDNA were not cleaved, yielding 137 and 117 bp fragments, respectively. The radioactivity of the bands was quantified by densitometry analysis using a Fuji phosphorimager. The amount of mRNA in each sample was calculated by plotting the logarithm of the ratio of the target band to the competitor band against the logarithm of the corresponding number of molecules of the competitor. When the ratio is equal to 1, the initial amount of unknown RNA is equivalent to the amount of competitor. Each quantitative-competitive (QC) RT–PCR assay was performed in triplicate and the P values were calculated using an unpaired t-test in the MS Excel program.

Collagen protein studies

Dermal fibroblasts were incubated with 3.75 µCi [¹⁴C]proline (250 mCi/mmol; Amersham Pharmacia Biotech) at 37°C for 3 h. Type I procollagen was precipitated from the cell layer (17). Collagens were prepared by digestion with pepsin (100 µg/ml) for 3 h at 15°C and analyzed by SDS–urea–PAGE (15). To quantitate the radioactivity of individual collagen chains, bands were excised from the gel, digested in 1 ml of tissue solubilizer (TS-2; Research Products Information Corp., Mount Prospect, IL) and the radioactivity measured by scintillation counting. This analysis was performed in triplicate on independent collagen preparations from OI cells.

Ribozyme cleavage in vitro

In order to develop a gene therapy approach for OI, we selected a type I collagen mutation which not only causes an OI phenotype, but also generates a novel ribozyme cleavage site. The mutation selected is a 907G→T change at codon 85 (Gly85→Val) in α1(I) collagen mRNA which creates a novel GUA ribozyme cleavage site. Thus, the mutation itself provides the target for allele-specific cleavage of mutant collagen mRNA. We designed hammerhead ribozymes with 13 nt binding arms to target this site, containing an active or inactive catalytic core (Fig. 1). To confirm the activity of the ribozyme and the accessibility of the target site, preliminary in vitro experiments were performed (data not shown). Synthetic ³²P-labeled RNA transcripts, containing 117 nt segments of either normal or mutant α1(I) mRNA and 109 nt of flanking vector sequence, were incubated with active or inactive ribozyme as described in Materials and Methods. Only the active ribozyme cleaved mutant substrates into the expected 111 and 115 nt fragments. In this assay 17% of mutant substrate was cleaved at an active ribozyme:substrate ratio of 1:1 and >75% of target was cleaved at a 50:1 ratio.

Ribozyme cleavage in OI cells

For the studies in cultured cells we cloned inactive and active ribozyme templates into the mammalian expression vectors pCI.neo and pHβAPr-1-neo. Using selection with G418, we were able to generate cells stably expressing active ribozymes, inactive ribozymes or vector sequences alone for both vectors. Expression of ribozymes and vector sequence alone was detected by RT–PCR analysis (Fig. 2, lanes 5–7). Primer pairs used in this PCR were located within the vector outside the cloned ribozyme sequence and are designed to amplify both vector alone and vector containing ribozyme. Products of 341 and 168 bp were detected in samples expressing ribozyme from pCI.neo and pHβAPr-1-neo, respectively. Amplification of these empty vectors yielded 287 and 114 bp products, respectively. The absence of PCR products in the samples not containing reverse transcriptase (Fig. 2, lanes 1–4) confirmed that the amplified products were derived from an RNA template and not DNA contamination. The expression of active ribozymes from the pCI.neo vector was detected in G85V cells for multiple passages following transfection. There was a decrease in the level of amplified ribozyme cDNA in cell passages higher than P18 (Fig. 3), presumably due to either loss of expression cassette or CMV promoter attenuation.

Our next goal was to demonstrate the efficiency and specificity of the ribozymes expressed in stably transfected cells. The ability of ribozymes to specifically cleave mutant α1(I) collagen mRNA in cells was evaluated using a QC RT–PCR assay (Fig. 4). The α21(I) collagen mutation selected for this study not only generates a GUA ribozyme cleavage site, but also deletes a BstNI site. Thus, BstNI digestion of PCR-amplified α1(I) cDNAs allows the electrophoretic separation of normal and mutant products. Levels of normal and mutant type I collagen mRNA and ribozyme were normalized to β-actin mRNA levels. The G85V cells stably expressing active ribozyme from either the pCI.neo or pHβAPr-1-neo vector had substantially decreased levels of mutant α1(I) mRNA levels relative to untransfected cells or cells transfected with vector alone (Table 2). The G85V cells expressing pCI.neo/active ribozyme had a 53 ± 5% decrease in mutant α1(I) level and those expressing pHβAPr-1-neo/active ribozyme had a 57 ± 8% decrease relative to untransfected cells. These levels of mutant α1(I) mRNA suppression were achieved with ribozyme:total α1(I) mRNA ratios of 3–5:1 and 12–14:1, respectively.

The level of normal α1(I) mRNA was unchanged when active or inactive ribozyme was expressed from the pHβAPr-1-neo vector. The same ribozymes expressed from the pCI.neo vector did cause a small (10–20%) but significant reduction in normal mRNA levels. These data suggest that the expressed sequences surrounding the ribozyme may play a role. For the pCI.neo transcript this includes 184 nt of pGEM-3Zf sequence, the T7 promoter and vector cloning site at the 5′-end of the transcript and the T3 promoter and SV40 polyadenylation signal. These sequences contain no significant homology to the region of α1(I) cDNA flanking the cleavage site. The transcript from pHβAPr-1-neo is twice as long as that from pCI.neo and the presence of longer tails on both sides of the ribozyme might result in faster cycling of ribozyme off target, minimizing an antisense effect.

Ribozyme cleavage in clonal cell lines

Since the experiments described above used a mixed population of stably transfected cells, the calculated values are averages of the cell pool. To test for potential variations in the level of pCI.neo/active ribozyme expression and to determine if such variation would correlate to the degree of mutant α1(I) mRNA cleavage, G85V clonal lines were generated. RNA was collected from single wells of a 96-well plate and used for RT–PCR analyses (Fig. 5). Three of the clonal lines expressed higher levels of ribozyme (Fig. 5, lanes 5–7). These lines also showed marked reduction in the ratio of mutant to normal α1(I) mRNA (0.33–0.41). Two clonal lines (Fig. 5, lanes 3 and 4) had both undetectable ribozyme expression and unreduced levels of mutant collagen mRNA [ratio mutant:normal α1(I), 0.95–0.96]. We noted that all transfected cells have a reduced amount of collagen RT–PCR products as compared to untransfected cells. Since the clonal lines were derived with vector pCI.neo, there may also be some reduction in the relative amount of normal collagen transcript. Clonal lines were not derived from transfected normal cells and the potential antisense effect in this experiment was not quantitated.

Effect of ribozyme expression on type I collagen protein

Finally, the capacity of the pCI.neo/active ribozyme to reduce the level of mutant α1(I) collagen protein was investigated. The G85V substitution generates a pepsin-sensitive site in the type I procollagen helix. Following digestion with pepsin, truncated α1 chains are generated and detected by SDS–urea–PAGE (Fig. 6). In triplicate samples of untransfected OI cells and OI cells transfected with active ribozyme, we examined the effect of ribozyme on the level of truncated α1 chain. Only cells expressing active ribozyme were tested, because cells with inactive ribozyme had unreduced levels of α1(I) transcripts. Since the pepsin-sensitive site is not quantitatively cleaved, the overexposed gels required to detect the mutant chains radiographically were unsuitable for densitometry (Fig. 6B). Chains were instead quantitated using direct counts of solubilized bands. Cells expressing active ribozyme had a ratio truncated α1:normal α1 that was reduced to 70 ± 12% of the ratio in untransfected cells. The lack of significant homology of the ribozyme to α2 mRNA sequences makes a decrease in α2 mRNA unlikely. Furthermore, because of the fixed α1(I)₂α2(I) composition of type I collagen, this assay is only sensitive to the relative amounts of the normal and mutant forms of α1. As expected, the α1:α2 ratio was 1.97 in untransfected OI cells and 2.07 in transfected cells.

Effect of ribozyme on cell growth

To examine whether stably expressed active ribozyme affects cell viability, we investigated the effect of the pCI.neo/active ribozyme on the growth rate of G85V cells. Under subconfluent conditions the transfected cells displayed an initial delay in growth (days 2–5). However, after this lag time there was no significant difference in cell growth rate between untransfected cells (cell doubling 3.5 ± 0.3 days) and cells stably transfected with pCI.neo vector (cell doubling 4.4 ± 0.7 days,), pCI.neo/IR (cell doubling 3.2 ± 0.3 days) or pCI.neo/AR (cell doubling 4.4 ± 0.7 days), indicating that expression of the G85V-specific active ribozyme is not detrimental to cell viability.