Preliminary work on intermolecular ribozyme cleavage
There have been numerous reports in which hammerhead ribozymes seem to function either intramolecularly or intermolecularly, to reduce levels of specific protein synthesis in eukaryotic [20–22] or bacterial cells [23,24]. Therefore, we first tried to develop an intermolecular assay for ribozyme cleavage in E. coli, which might permit the rapid evaluation of ribozyme activity against many possible RNA substrates. Yet preliminary studies using ribozymes as expressed from a T7 promoter in vivo showed little if any reduction of target β-galactosidase activity, even when ribozymes were expressed next to their target on the same plasmid (data not shown).
Why did this preliminary attempt fail? Further experiments showed that in vitro such transcribed ribozymes would cut a short RNA substrate to 90% of completion within a few minutes at 37 °C, but would not cut a long RNA substrate by more than 5–10% as mRNA even after several hours at 37 °C. Many studies have shown that long molecules of RNA tend to contain far more secondary structure than short ones, on account of fortuitous base-pairing interactions which increase with size. Hence, rates of cleavage for ribozymes added in trans are often much lower on long RNA substrates than on short ones, because secondary structure in the long RNA may hinder a ribozyme from finding its target.
Given such a serious problem in RNA accessibility, how might one develop an assay for ribozyme activity in vivo? One approach might be to search for highly exposed parts of any mRNA, that could be cut or spliced by a ribozyme without steric hinderance [26,27]. Alternatively, one could develop an intramolecular assay, that would not be influenced by the slow kinetics of locating and binding some intermolecular RNA target within a cell. Indeed, when studied intramolecularly, one can design the sequence of any ribozyme so that it will bind very efficiently to its cutting-site nearby within the same molecule. To test this intramolecular assay, we used strains HB2151 and XL1-Blue of E. coli, which do not contain T7 RNA polymerase.
Design of a model system for intramolecular ribozyme activity
To create a test plasmid for our intramolecular assay, we started from a derivative of pUC18 that contains a promoter for T7 polymerase in vitro(Fig. 1). Next we added a full-length gene for β-galactosidase along with its E. coli gpt promoter , and also a polylinker at KpnI within the 5′-end of a fusion protein for β-galactosidase.
Figure 1. Schematic drawing of plasmid pT74Kgal into which a new polylinker was inserted to create pKpn-right.The distance from the T7 promoter to the GTC ribozyme is 230 bp, while the distance from GTC to PvuII is 200 bp. Note that PvuII lies 80 bp within the α-peptide for β-galactosidase, whereas GTC and the Kpn polylinker lie 130 bp upstream.
The base sequence of this new polylinker is shown in Fig. 2. For the orientation pKpn-right as shown, it creates a series of restriction sites NotI–XhoI–BstEII–KpnI, which enable easy insertion of a ribozyme or hairpin in subsequent steps, within the nonessential upstream part of a fusion protein for β-galactosidase. That 27-bp polylinker, as well as all other DNA molecules used in our study, was designed to add 3n bases to the length of the DNA to maintain a correct reading-frame for β-galactosidase. All sequences were also designed to avoid stop codons except in certain controls.
Figure 2. Detailed base sequence of a small region of pT74Kgal near KpnI: (A) before insertion of a polylinker, and (B) after such insertion to create pKpn-right.
Cloning of ribozymes R, M, S, T and the J series
Next, in order to determine whether any ribozyme might reduce levels of specific protein synthesis, we cloned into NotI–BstEII a series of hammerhead ribozymes R, M, S and T as shown in Fig. 3. Those DNA insertions by themselves do not influence β-galactosidase activity, because they lie 120 bp or 40 amino acids upstream from the α-peptide of β-galactosidase.
Ribozyme R, as shown at the top of Fig. 3, is a standard hammerhead [6,7]. Its 5′-hybridizing arm is drawn on the left, its ribozyme core in the centre and its 3′-hybridizing arm on the right. Only bases within the upper strand are shown; the lower follows by Watson–Crick pairing. Next, M is a miniribozyme in which the core has been reduced in size from 22 to 16 nucleotides [29,30]. The third ribozyme S is a derivative of R which contains several base pair mismatches (shown as crosses) to give flexibility. Finally, T at the bottom of Fig. 3 is a mutant of S, with sequence GAG rather than GAA (bold type). It serves as an inactive control despite having a similar folded structure of RNA.
The intramolecular folding of each ribozyme onto a nearby RNA cutting-site GUC is shown in Fig. 4. Any cloned ribozyme R, M, S or T will fold back strongly onto its GUC target in practically all molecules of mRNA, because the base sequence has been designed to create a hairpin loop–double helix of 15 bp downstream of GUC, and a double helix of 14 bp upstream of GUC.
Figure 4. Schematic drawing of how ribozyme R folds back onto its cutting-site GUC within β-galactosidase mRNA. The large arrow indicates specific ribozyme cleavage at base C, while ‘ribo’ indicates a ribozyme domain of 22 nucleotides.
We also cloned as a second set of molecules into our polylinker another cutting-site GUC, by adding the sequence IL shown in Fig. 5 (upper). This IL sequence comes from part of a gene for human interleukin-2, and has been used as a substrate for in vitro selection of ribozymes from a random RNA pool (J. Conaty, unpublished data). Five active ribozymes from that work were chosen as JB, JC, JD, JE or JF (Fig. 5, lower). All resemble a native hammerhead, but differ within nonconserved portions of a central N18 domain. Note that JC contains a stop codon TGA or TAA within each of its two orientations as a control. In order to direct those five J ribozymes against the cutting-site IL in E. coli, we cloned each member of the family downstream of BstEII.
Hence for the first cloning trial, all ribozymes R, M, S or T were inserted at NotI–BstEII, while all of JB to JF were inserted just downstream of BstEII. For each cell transformation, roughly equal numbers of blue and white colonies were obtained on plates containing 5-bromo-4-chloro-3-indolyl-β-d-galactoside. Further analysis showed that all authentic in-frame ribozymes would produce only blue colonies, whereas the white colonies were either frameshift deletions or some other DNA rearrangement. However, those white colonies serve as useful controls, for reduced levels of β-galactosidase when no active mRNA is present.
Intramolecular ribozyme cleavage does not alter gene activity in E. coli
To begin our study of intramolecular ribozyme cleavage, β-galactosidase activities were measured in cell extracts using the dye CPRG as listed in Table 1. RNA expression was driven by E. coli polymerase from a noninducible gpt promoter. Three extracts were prepared from each blue clone and one from each white. Transformed cells from two strains HB2151 or XL1-Blue were tested. When averaged over both strains, the ratio of β-galactosidase for blue versus white clones of R, M, S or T was found to be 29, 16, 14 or 19, respectively (upper right-hand column of Table 1). Similar ratios of 15 or 25 were found for blue versus white clones of JD or JE (lower right-hand column of Table 1). Hence there exists a wide range of activity between active and inactive, which may be used to measure possible ribozyme effects on protein synthesis.
Table 1. Activities of β-galactosidase in two strains of E. coli are not influenced by intramolecular ribozymes. All β-galactosidase activities were determined from extracts of E. coli and are listed on a relative scale. On an absolute scale at 0.5 mg·mL−1, β-galactosidase activities range from 80 to 2 × 10−3 U·mL−1. Concentrations of DNA were similar for all plasmids under conditions of ampicillin selection and logarthmic growth. Three cell extracts were tested for each blue clone (the mean is shown), while one was tested for each white. The error for measuring β-galactosidase within each group of blue clones was 20–30%. Both JC and JC-reverse provide for a stop codon as TGA or TAA, respectively, yet show nonzero activities of 0.38 or 0.16 perhaps due to a minor start-site downstream of KpnI.
| ||β-Gal activity |
|No gene||0.03||0.01||0.02||Negative control|
|No insert||1.00||1.00||1.00||Positive control|
|R-white mutant||0.01||0.04||0.03|| |
|M-white mutant||0.10||0.07||0.08|| |
|S-white mutant||0.09||0.03||0.06|| |
|T-white mutant||0.04||0.07||0.05|| |
|IL-no insert||0.87||1.35||1.11||Positive control|
|JDwhite mutant||0.05||0.04||0.05|| |
|JE-white mutant||0.03||0.02||0.03|| |
To our surprise, however, we found that none of these many self-cleaving ribozymes could reduce β-galactosidase synthesis in E. coli by any substantial amount. Thus, all of the colonies for active ribozymes R, M, S, JB, JE or JF show a dark blue colour on 5-bromo-4-chloro-3-indolyl-β-d-galactoside plates, even though 50% or greater reduction in β-galactosidase would be detectable by a pale-blue or almost-white colony. Indeed, JC in either of its two orientations contains a stop codon that produces a pale-blue colour and low β-galactosidase activities of 0.38 or 0.16 relative to a control 1.00 (Table 1, column 4). Such activities are not precisely zero, because some protein may be made downstream of the stop codon, or read-through a stop codon in certain cases. Similarly, JD-white as a frameshift deletion shows a residual activity of just 0.05 and an almost-white colour on plates.
Nor do β-galactosidase activities as measured quantitatively in cell extracts, show any substantial difference for active ribozymes R, M or S when compared with inactive T or a no-insert control, which together provide a reliable baseline for full enzymatic activity. As shown in column 4 of Table 1 (upper), active ribozymes R-blue, M-blue or S-blue show mean activities of 0.87, 1.26 or 0.85, close to 0.95 for inactive T-blue or 1.00 for a no-insert control.
Finally, let us consider small variations of reporter-gene activity among the ribozymes J, which include two possible orientations as forward (active) or reverse (inactive). These data are listed in Table 1 (column 4, lower). Both orientations of JB show high activities of 0.90 (forward) or 1.00 (reverse), which are similar to 1.11 for IL as another neutral-insert control, that provides a baseline of full activity for the J series. Both orientations of JC include stop codons, and show low activities of 0.38 (forward) or 0.16 (reverse). Ribozyme JD shows a control activity of 0.77 (reverse); no forward in-frame clone was obtained. Lastly, both JE and JF show fairly normal activities of 0.76 or 0.71 in their forward active orientation, close to a control of 1.00.
In principle, two different isomers of any self-cleaving RNA are possible: (a) one in which a cutting-site GUC lies to the 5′-side of a ribozyme, or (b) another where a ribozyme lies to the 5′-side of GUC. In all the above cases, we studied isomers of type a. Therefore, we also tested one isomer of type b. A newly synthesized ribozyme R was cloned into NotI upstream of IL, so that it would fold downstream onto GUC. When ribozyme (b) was grown in E. coli, all authentic colonies appeared dark blue on 5-bromo-4-chloro-3-indolyl-β-d-galactoside plates (data not shown). Hence, isomer (b) seems to be no more effective at reducing β-galactosidase activity than any of the isomers (a) already tested.
High ribozyme cleavage during T7 transcription in vitro
Because no hammerhead ribozyme produced any significant reduction of β-galactosidase protein in E. coli, we wondered whether such ribozymes really self-cleave their mRNA well enough to reduce protein synthesis in a living cell? As a control, we found that every cloned ribozyme R, M, S or those within the J series, would self-cleave a mRNA to 60–80% of completion during transcription in vitro by T7 RNA polymerase, over 30 min at 37 °C (Fig. 6 and Table 2).
Figure 6. Ribozyme self-cleavage during T7 transcription in vitro from linear PvuII DNA, as visualized on a 5% acrylamide-urea gel. Ribozymes R, S or JB, JC, JD, JE, JF self-cleave to leave two products of size rb (230 + 230) or else rb1 + rb2 (270 + 210) nucleotides, respectively, whereas controls IL and T do not self-cleave. ‘B’ stands for blue clones, and ‘W’ for white. Hpa II DNA markers (arrows on right) are of size 400 (doublet), 330 or 240 nucleotides.
Table 2. Efficiencies of RNA self-cleavage for intramolecular ribozymes in vitro versus in vivo. The per cent self-cleavage was determined in vitro by transcription from PvuII fragments. Each reaction contained 500 ng of DNA, T7 RNA polymerase, ATP, GTP, CTP, and a mixture of UTP with [α-32P]UTP. Determinations of full-length versus self-cleaved RNA were made by densitometry on a PhosphorImager. Per cent cleavage was determined also in vitro by S1-trimming for R-blue, giving 84% versus 72% from a gel. Per cent self-cleavage in E. coli was determined by hybridization of a DNA oligonucleotide to cellular RNA, followed by trimming with S1 nuclease.
| ||Self-cleavage in vitro||Self-cleavage in E. coli (%)|
|Insertion||(%)||As isolated||With Mg+2||Mg+2-DNAase I|
|R-blue active||72.2 (84)||–||–||–|
|R-white mutant||56.0|| || || |
|M-white mutant||33.7|| || || |
|S-blue active||70.2||44.5||79.4||78.8 (four clones)|
|S-white mutant||63.5|| || || |
|T-blue inactive||0.1 (or less)||1.0||1.0||1.0 (or less)|
|T-white mutant||2.0 (or less)|| || || |
|JB-forward||80.8||51.3|| || |
|JC-forward||86.8||51.6|| || |
|JD-white mutant||90.6||46.4|| || |
|JE-forward||83.6||54.7|| || |
|JF-forward||57.0||54.9|| || |
Thus, the left-hand side of Fig. 6 shows that a full-length RNA ‘RB’ or ‘RW’ of 460 nucleotides (top) is cut into two nearly equal products of 230 nucleotides (‘rb’), with efficiencies of 72.2% or 56.0%, respectively (Table 2, upper left); inactive ‘TB’ and ‘TW’ do not cleave. The right-hand side of Fig. 6 shows that all full-length RNAs ‘JB’ to ‘JF’ of 480 nucleotides (top) are cut into two products of 270 or 210 nucleotides (‘rb1’ or ‘rb2’), with efficiencies of 80.8% or 86.8%, respectively (Table 2, lower left). Ribozyme cleavage was seen also during T7 transcription from supercoiled DNA (not shown).
Only moderate ribozyme cleavage during transcription in E. coli
But how well might those same ribozymes self-cleave in E. coli? To address this question, we isolated total RNA from cultures of E. coli growing in logarithmic phase, by rapid spin–lysozyme–SDS/DEPC treatment at 4 °C . This procedure yields the best-quality RNA of several methods tested, although not all mRNA molecules will be obtained intact, owing to their short half-lives in E. coli of just 2–3 min. Next, we used two different methods to measure ribozyme cleavage in β-galactosidase mRNA: (a) primer extension by reverse transcriptase or (b) oligonucleotide trimming by S1 nuclease.
To avoid ribozyme cleavage during isolation, we used excess EDTA and 4 °C except for a brief incubation with DEPC at 37 °C. Indeed, the use of 1–5 mm magnesium ion for RNA extraction, reverse transcriptase or DNAase I, may induce ribozyme cleavage outside a cell [31,32]. Yet 1 mm zinc ion and pH 5 as used for S1 nuclease ought not to produce any extra ribozyme cleavage [33,34]. Finally, our preparations were never dried or lyophilized .
Using method (a), we found strong and specific RNA cleavage for each active ribozyme tested (data not shown). Yet the efficiency of cleavage cannot be determined by this method, because the start-site for transcription lies far upstream of GUC. Also, the use of 5 mm MgCl2 for reverse transcriptase may induce extra ribozyme cleavage outside a cell .
For those reasons, we also analysed such RNA preparations using a better method (b), which involves trimming a DNA oligonucleotide with S1 nuclease . Trimming a hybridized oligonucleotide near the ribozyme cutting-site GUC, ensures that any general decay of RNA due to its short half-life in E. coli, will not interfere with the measurement of specific decay due to localized ribozyme cleavage. Following method (b), we first selected for β-galactosidase RNA by hybridization of a DNA oligonucleotide to the 3′-side of any ribozyme cutting-site GUC. Different DNA oligonucleotides were designed for different sequences, but all could typically be trimmed from 64 to 56 bases, to indicate the amount of intact mRNA within a local region. Next, if a hammerhead ribozyme has cleaved at GUC, the same oligonucleotide could be trimmed further from 56 to 45 bases, to indicate the amount of ribozyme-cleaved RNA.
Some S1-trimming results for ribozymes S and T are shown in Fig. 7, while all data are summarized in Table 2 (right). Ribozyme S shows a mean self-cleavage of 44.5% as an average over four clones of 56, 57, 33 or 32%. Detailed data for two of these S clones are shown on either side of Fig. 7, after 5 or 30 min of S1 action. There one can see roughly equal amounts of 56-mer (intact) or 45-mer (ribozyme-cleaved) product; this result does not depend much on the time of digestion. As a control, four clones of inactive ribozyme T show no self-cleavage, nor does a clone of IL. We deliberately avoided the use DNAase I to remove any slight DNA impurity from our RNA, because 1 mm MgCl2 as needed for DNAase I may induce ribozyme cleavage outside of a cell.
Figure 7. Ribozyme self-cleavage in E. coli RNA, as measured by trimming a hybridized DNA oligonucleotide with S1 nuclease. Such DNA will be trimmed by S1 nuclease from 64 to 56 nucleotides if it is bound to full-length RNA; or from 64 to 45 nucleotides if it is bound to ribozyme-cleaved RNA. Two time-points of digestion for 5 or 30 min are shown at pH 5.0 in 1 mm ZnCl2. Two clones of ribozyme S self-cleave to leave almost equal amounts of 56-mer and 45-mer, whereas controls IL and T leave only a 56-mer. Size markers are not shown.
In fact, subsequent treatment of cellular RNA with 1 mm MgCl2 and DNAase I prior to S1-trimming, increased the apparent self-cleavage for ribozyme S from 44.5 to 78.8%. Similarly, incubation of ribozyme S with 1 mm MgCl2 alone for 30 min at 37 °C increased the apparent self-cleavage from 44.5 to 79.4%. Therefore, the original self-cleavage of 44.5%, as measured for RNA in EDTA, appears to be correct as an upper limit. Other measurements of 79.4 or 78.8% after incubation with Mg2+ must be due to extra self-cleavage after leaving a cellular environment. We never found high levels of ribozyme cleavage near 80%, unless the isolated cellular RNA was incubated with magnesium.
Using a different DNA oligonucleotide, we found for ribozymes JB to JF a mean self-cleavage of 52% omitting MgCl2 (Table 2, right). As a control, β-galactosidase RNA when transcribed in vitro and trimmed by S1 nuclease, shows for ribozyme R a self-cleavage of 84%, which agrees well with 72% obtained by direct measurement (Table 2 left, upper).
In summary, the mean self-cleavage for our most active ribozyme S was measured as 45% after isolation of total RNA from E. coli. That seems to be significantly less than its self-cleavage of 70% during T7 transcription in vitro, or 79% after incubation with MgCl2, and is apparently not enough to produce a measurable reduction in β-galactosidase protein. S1-trimming of a hybridized DNA oligonucleotide may be the best technique to use in this case, because analyses of those same RNA preparations by gel electrophoresis alone (as a Northern blot, see below) show a broad range of sizes up to 5000 nucleotides, but do not reveal any ribozyme cleavage.
It should be emphasized that our measured self-cleavage of 45% represents the maximal possible influence of the best intramolecular ribozyme on protein synthesis in E. coli, given no cellular RNA decay. If much of the RNA degrades naturally soon after it is made, then that time-averaged ribozyme cleavage of 45% would have much less effect on protein synthesis as observed.
Strong influence of RNA hairpin loops on protein synthesis in E. coli
Because our first round of experiments did not show any influence of ribozyme cleavage on protein synthesis in E. coli, we cloned four more types of insert to see whether any of those might reduce protein synthesis. The sequences of these four new inserts are shown in Fig. 8A as A, N, B and C; their folded structures once transcribed as RNA are shown in Fig. 8B.
Sequence A provides for an antisense loop of 30 bp (Fig. 8B, top). It is similar to R, except that the central ribozyme core of 22 nucleotides has been deleted (cross in Fig. 8A). Hence the residual hairpin loop should become more stable, owing to the deletion of several mismatches in the ribozyme domain. Sequence A was cloned into NotI–BstEII as previously for R. Sequence N is a slow-cutting ribozyme from the newt  cloned into the same location (Fig. 8B, middle).
Sequences B and C are spacers of 24 or 45 nucleotides, respectively, which may be inserted between any ribozyme and its target. For our purposes, such spacers were cloned into NotI between ribozymes R or T and their cutting site GUC. When spacers B or C are added as monomers, they will separate each ribozyme from its cutting site by an extra 24 or 45 nucleotides as a loop (Fig. 8B, lower middle). Yet when B or C are added as inverted dimers, they will add between any ribozyme and its cutting site a stable hairpin of size 24 or 45 base pairs (Fig. 8B, bottom).
For all authentic clones of A, B, C or N, we prepared cell extracts by which to measure β-galactosidase activities in E. coli. Three different strains were tested as shown in Table 3: HB2151, XL1-Blue and SØ3831 .
Table 3. Activities of β-galactosidase in three strains of E. coli for intramolecular ribozymes with spacers or hairpin loops. All β-galactosidase activities were determined from cell extracts of E. coli; those listed for HB2151 are the average of two separate extracts. Strain SØ3831 was grown in 50 µg·mL−1 ampicillin, with 100 µg·mL−1 streptomycin to enhance cell growth. Even after adding streptomycin, some cultures of SØ3831 could not be grown (dashed lines). Colony colours were examined after light growth on plates containing 40 µg·mL−1 5-bromo-4-chloroindol-3-yl β-d-galactoside. Relative growth rates for SØ3831 as measured on ampicillin plates were similar with or without streptomycin.
|Insertion||β-Gal HB2151||Colour HB2151 or XL1-Blue||β-Gal SØ3831||Growth SØ3831|
|A-blue hairpin||0.59||Pale blue||(0.84)||Medium|
|R-spacer B-dimer left||0.39||Almost white||(0.04)||Fast|
|R-spacer B-dimer right||0.29||Almost white||(0.93)||Fast|
|T-spacer B-dimer right||0.18||Almost white||(0.41)||Fast|
In Table 3 (left), one may examine β-galactosidase activities as measured from HB2151, or colony colours as measured from HB2151 and XL1-Blue. The majority of clones show no significant reduction of β-galactosidase (column 2) or colony colour (column 3). Yet in four cases: A-blue, R-spacer B-dimer left, R-spacer B-dimer right and T-spacer B-dimer right, β-galactosidase activities are reduced from 1.00 to 0.59, 0.39, 0.29 or 0.18 (column 2); while colony colours are reduced from blue to pale-blue or almost-white (column 3). Each of those four suppressed clones contains a long hairpin loop within the upstream region of β-galactosidase RNA, regardless of their ribozyme activities. Indeed, when a ribozyme is deleted on going from R to A, the stabilized hairpin loop reduces β-galactosidase activity by more than before.
In Table 3 (right) one may examine the results obtained from strain SØ3831, which grows at one-third the normal rate in the absence of streptomycin, or two-thirds the normal rate in the presence of streptomycin . The β-galactosidase activities as measured from SØ3831 (column 4) appear unreliable, and so are listed in parentheses. The main problem is that cells of SØ3831 grow poorly under ampicillin selection, even with streptomycin. Hence, liquid cultures of that strain grow irreproducibly, while some could not even be prepared (see dashed lines).
Chen et al.  reported both antisense and ribozyme-induced suppression of gene activity in SØ3831 after induction with isopropyl thio-β-d-galactoside (IPTG), but only under stationary-phase conditions and in the absence of streptomycin. By contrast, under the normal logarithmic-phase conditions used here, it seems that SØ3831 does not grow well enough to make such measurements possible.
By careful observation, we noted also a surprising result; when all of our plasmids were transfected into SØ3831 and spread onto plates containing ampicillin (with or without streptomycin), then the same four plasmids which show reduced levels of β-galactosidase in HB2151, show an enhanced rate of colony growth in SØ3831: A-blue, R-spacer B-dimer left, R-spacer B-dimer right and T-spacer B-dimer right (column 5 of Table 3). A similar effect was seen in solution, where cells that contain slow-growth plasmids R-spacer B-right or T-spacer B-right could not be propagated (column 4). The molecular mechanism of this growth enhancement remains unclear and requires further study. It cannot plausibly be due to spontaneous genetic reversion of SØ3831, because the effect may be seen with or without streptomycin and because it affects all colonies on the plate. Furthermore, it seems specific for RNA structure as a long hairpin loop, which might act analogously to streptomycin as a ribosome-stalling agent [16,39].
In summary, one can detect a clear influence of RNA hairpin loops on: β-galactosidase activity in HB2151, colony colour in HB2151 or XL1-Blue and growth rate in SØ3831. No hammerhead ribozyme exerts any influence on protein synthesis or growth rate by the same experiments. The observed inhibition of protein synthesis by hairpin loops appears to increase in proportion to the stability of any loop: long hairpins RB-dimer or TB-dimer show the strongest inhibition; short hairpin A provides a moderate amount; while ribozyme R (where hairpin A is interrupted by mismatches) provides none at all.
Why should long RNA hairpin loops inhibit the expression of a reporter gene in E. coli? In principle, those hairpin loops could block either transcription or translation. Yet we found that such RNA hairpin loops would exert only a minor influence on T7 transcription in vitro (data not shown). Therefore, we next examined the influence of RNA hairpin loops on translation in vitro.
Strong influence of RNA hairpin loops on protein synthesis in a transcription–translation extract
In order to see whether long RNA hairpin loops or ribozymes might reduce protein synthesis in a cell extract, we chose the E. coli-T7S30 extract for circular DNA, which makes large amounts of RNA using T7 RNA polymerase, and large amounts of protein using E. coli ribosomes. All of our plasmids contain a T7 promoter upstream of gpt (Fig. 1), so the start sites for RNA should be similar in this extract versus E. coli.
Furthermore, our plasmids contain just two major start sites for protein synthesis by E. coli ribosomes. One is located upstream of KpnI and initiates the synthesis of a fusion protein for β-galactosidase, while the other is located downstream of β-galactosidase and initiates the synthesis of β-lactamase. The β-galactosidase fusion protein is a long polypeptide of 120 000 Da, which shows two stable products of almost equal size. Alternatively, the β-lactamase protein is a medium-length polypeptide of 40 000 Da, which shows a stable product of single size.
In preliminary experiments, we found that the weight-ratio of β-galactosidase to β-lactamase as synthesized was typically 12%. Yet as shown in Table 4, certain inserts which contain long RNA hairpin loops: A-blue, R-spacer B-dimer left, R-spacer B-dimer right and T-spacer B-dimer right, reduce that ratio to 1% or less, or specifically to 0.11, 0.14, 0.03, 0.06 of a control 1.00. No other inserts show any effect. These results agree well with studies in E. coli (Table 3).
Table 4. Synthesis of β-galactosidase versus β-lactamase in a cell extract from E. coli, as modified by ribozymes or hairpin loops. The ratio of newly synthesized protein as β-galactosidase to β-lactamase was determined for various DNA plasmids, which contain a ribozyme or hairpin loop upstream of essential coding sequences for β-galactosidase. Plasmid DNA was transcribed by T7 RNA polymerase, then translated by ribosomes in a cell extract from E. coli. Newly synthesized proteins were labelled with [35S]l-methionine and separated on a 4–10% acrylamide–SDS gel. Autoradiography of the dried gel allowed for determination of the ratio β-galactosidase to β-lactamase by densitometry on a PhosphorImager. Each plasmid DNA was transcribed and translated several times with similar results; the clearest gel was then used for densitometry.
|Insertion||Ratio β-galactosidase/ β-lactamase|
|R-spacer B-dimer left||0.14|
|R-spacer B-dimer right||0.03|
|T-spacer B-dimer right||0.06|
A few specific examples are shown in Figs 9 and 10. Figure 9 compares the synthesis of β-galactosidase (small double peak) with β-lactamase (large single peak) for R-blue and A-blue. One can see that a 30 bp hairpin as A-blue reduces β-galactosidase strongly, by loss of the double peak; whereas a ribozyme as R-blue shows no effect. Figure 10 compares the synthesis of β-galactosidase with β-lactamase for R-spacer B-right and R-spacer B-dimer right. By changing a loop-spacer of 24 nucleotides between R ribozyme and its target, to a hairpin-spacer of 24 bp, the synthesis of β-galactosidase is eliminated. As a control, no truncated proteins of 5000–10 000 Da were detectable on peptide-SDS or acid-urea gels.
Figure 9. Transcription–translation of plasmids R-blue and A-blue in an E. coli–T7 cell extract, in which newly synthesized proteins are visualized by [35S]l-methionine on a 4–10% SDS/acrylamide gel. Taking β-lactamase as an internal control, one can see that antisense A reduces β-galactosidase synthesis by far more than does ribozyme R. Several minor peaks which appear just to the left of β-lactamase represent truncated forms of β-galactosidase as made in the extract.
Figure 10. Transcription–translation of plasmids R-spacer B-right and R-spacer B-dimer right in an E. coli–T7 cell extract, in which newly synthesized proteins are visualized by [35S]l-methionine on a 4–10% SDS gel. Taking β-lactamase as an internal control, one can see that dimeric hairpin spacer B reduces β-galactosidase synthesis by far more than does monomeric unstructured spacer B. Three minor peaks which appear just to the left of β-lactamase represent truncated forms of β-galactosidase.
Why should the hammerhead ribozyme show no effect on protein synthesis in this extract? Recall that the self-cleavage of a ribozyme was measured as 45% in E. coli versus 70–79% in vitro. In this extract, we find that ribozyme self-cleavage becomes almost undetectable, based on labelling of newly synthesized RNA with [32P]UTP, and examination of products on a denaturing gel . By contrast, long RNA hairpin loops seem to reduce protein synthesis strongly, by a translational effect rather than by some reduced stability of the RNA (as seen in gels). Finally, antisense oligomers also reduced the synthesis of β-galactosidase, although not by as much for hairpin loops [1,2].
Because hammerhead ribozymes show practically no effect on protein synthesis in E. coli or its cell extract, we finally decided to test both ribozymes and hairpin loops in a eukaryotic cell system, involving CHO cells. In E. coli, transcription and translation are closely coupled, so that a ribozyme may not have much time to cleave its substrate before the RNA makes protein. Yet in mammalian cells, any mRNA will be transcribed in the nucleus before it is transferred to the cytoplasm to make protein. Hence, an assay in mammalian cells might provide more opportunity for a slow-cutting ribozyme to have an impact on gene expression.
Influence of ribozymes versus hairpin loops on protein synthesis in CHO cells
In order to generate substantial levels of β-galactosidase in CHO cells, we added an SV40 early promoter to the gene for β-galactosidase . All ribozymes remain fully active to self-cleavage in these new constructs. To prepare plasmids for transfection, we purified them on a small scale from E. coli and measured accurately their concentrations using agarose gels stained with ethidium bromide (EtBr) or acrylamide gels labelled with [32P]dATP.
Five independent sets of transfections were performed using a lipid transfection reagent CSO66 [41,42]. Variation in the lipid-to-nucleotide ratio of 1.4 : 1.0 or 2.8 : 1.0 did not influence the results. Higher ratios of 3.5 : 1.0 produced toxicity, while lower ratios of 0.7 : 1.0 gave low transfection. Duplicate measurements agreed to within 10%.
The right-hand column of Table 5 provides an average over trials 1–4, which may be inspected for a final result. It seems that active ribozymes R-blue and S-blue reduce the synthesis of β-galactosidase in CHO cells to 20–32% of a no-insert control. However, inactive ribozyme T-blue reduces such synthesis by almost half to 51% by means of its folded RNA structure alone. Hence, the moderate suppressive effect of ribozymes R or S may be due largely to tight folding of their RNA, as seen also for inactive ribozyme T.
Much stronger suppressions of protein synthesis are measured in CHO cells for RNA hairpin loops A-blue, R-spacer B-dimer right and T-spacer B-dimer right, which show consistently suppressed β-galactosidase activities of just 22, 13 or 16%, respectively. Also included in Table 5 is a frameshift deletion R-white, which serves as a control for minor start sites downstream.
Four additional inserts were tested in trials 3 and 4 only, in which M-blue is a miniribozyme and N-blue is a newt ribozyme. Theo-single is a hairpin of 25 bp which contains a single binding site for theophylline, while Theo-double is a hairpin of 50 bp which contains a double binding site for theophylline . Those latter two clones were made to see whether theophylline might provide for an inducible suppression of protein synthesis, by binding to an RNA hairpin loop in the cell which already suppresses β-galactosidase moderately.
In these additional trials, ribozymes M-blue and N-blue reduce β-galactosidase activity to just half (50–53%) of a no-insert control, almost as for inactive ribozyme T; while hairpin loops Theo-single and Theo-double reduce β-galactosidase more strongly to 38 or 15%, respectively. The long loop Theo-double shows greater suppressive activity than the short loop Theo-single. Unfortunately, neither hairpin loop shows any extra suppression of protein synthesis, on adding theophylline to the growth medium. Hence our preliminary attempt to design an inducible system, by which to control protein synthesis at the level of RNA, was unsuccessful.
In summary, only two of four active ribozymes R, S, M or N, when expressed intramolecularly in CHO cells, can reduce protein levels to 20–32%, which is less than the 51% measured for inactive ribozyme T. By contrast, all five hairpin loops A, RB-dimer, TB-dimer, Theo-single and Theo-double reduce protein levels to a low 10–40%. The long loops RB-dimer, TB-dimer or Theo-double as 13–16% repress more than the short ones A or Theo-single as 22–38%.
The strong reduction of protein synthesis by RNA hairpin loops in CHO cells seems analogous to that seen already in E. coli. Yet the weak extra reduction of protein synthesis as seen for ribozymes R and S in CHO cells was not seen in E. coli. Could such weak effects be due to partial inactivation of a mRNA, by ribozyme cleavage in vivo? Possibly, but we cannot detect any substantial ribozyme cleavage in CHO cells despite the use of three different methods (see below). In any case, the suppressive action of a hairpin loop seems greater than that of any ribozyme. For example, adding ribozyme R to hairpin A actually increases the amount of protein made, from 20 to 30% of a control.
Attempts to measure ribozyme cleavage in CHO cells
In an attempt to measure how much any ribozyme R or S cleaves in CHO cells, we isolated total RNA and performed two kinds of experiment: primer extension by reverse transcriptase, or S1-trimming of a hybridized DNA oligonucleotide. As a control, β-galactosidase activities were measured from the same cell cultures (Table 5, column 5). However, neither method revealed any specific ribozyme cleavage despite repeated trials. Some workers have proposed reverse transcriptase-PCR to detect tiny amounts of ribozyme-cleaved RNA [22,44]. However, 5 mm MgCl2 as used for reverse transcriptase may induce further cleavage outside of a cell [31,32,35].
Analysis of full-length mRNA by Northern blot
In order to measure the size and stability of full-length mRNA for β-galactosidase, total RNA samples from E. coli and CHO cells were applied to agarose-formaldehyde gels then transferred to nitrocellulose membranes, where they were probed using fragments from the β-galactosidase coding region to detect mRNA, or a ribosomal RNA gene to quantify total RNA. The gel results (not shown) are summarized below and in Table 6.
Table 6. Measurement of full-length β-galactosidase mRNA by Northern blot. Total RNA was extracted from E. coli or CHO cells and applied (10 µg) to agarose-formaldehyde gels. After electrophoresis, such RNA was transferred to nitrocellulose membranes, where it was probed with the coding region of β-galactosidase to detect mRNA, or with a gene for ribosomal RNA (28S, 18S) to quantify the amount of total RNA. Sizes of full-length mRNA were found to be 6 kb in CHO cells or 5 kb in E. coli, by comparison with markers. Two determinations of mRNA relative to total RNA were made from CHO samples, while three determinations were made from E. coli (only the mean is shown in column 2). Insertions S-blue and T-blue list data for two independent clones.
|Cell type||Insertion||β-Gal RNA||β-Gal protein||Normalized protein/RNA|
|E. coli||IL-no insert||1.00||1.11||1.0|
|S-blue active||0.92 (0.4)||0.85||0.8|
|S-blue active||0.87 (0.3)||0.85||0.9|
|T-blue inactive||0.80 (0.1)||0.95||1.1|
|T-blue inactive||0.58 (0.2)||0.95||1.5|
|R-blue active||0.42 (0.09)||0.32||0.8|
|T-blue inactive||0.38 (0.06)||0.51||1.3|
|R-spacer B-dimer right||0.16 (0.06)||0.13||0.8|
|T-spacer B-dimer right||0.10 (0.03)||0.16||1.6|
No sample from E. coli or CHO cells shows any specific degradation due to ribozyme cleavage, even though RNA prepared by T7 polymerase in vitro shows a ribozyme product of the expected size of 0.5 kb. In E. coli, the maximal reduction of full-length RNA by an active ribozyme is just 10% for S-blue as 0.92 or 0.87 versus no-insert control IL as 1.00.
A clear band for full-length β-galactosidase mRNA can be seen in all cellular samples of size 6 kb in CHO cells or mainly 5 kb in E. coli (where a second band of 6 kb is also present). T7 RNA prepared in vitro is much larger and remains near the top of any gel track.
Quantification of these data by densitometry reveals that the amount of full-length β-galactosidase RNA remains roughly constant in E. coli, regardless of whether ribozymes (active or inactive S or T) have been inserted into it (Table 6, upper). By contrast, the amount of full-length RNA is reduced dramatically in CHO cells as ribozymes (active or inactive R or T) or especially long hairpin loops (RB or TB) are inserted (Table 6, lower).
A strong suppression of β-galactosidase protein, as produced by ribozymes or hairpin loops in CHO cells (Table 5), may be attributed entirely to their strong degradation of β-galactosidase RNA. Thus, the ratio of protein/RNA remains roughly constant near 1.0, even while individual values vary from 1.00 to 0.10 (Table 6, lower).
To conclude, the moderate reduction of β-galactosidase protein, as produced by RNA hairpin loops in E. coli or its cell extract, may be attributed plausibly to an attentuation of translation, because most RNA there remains intact. However, the more substantial reduction of β-galactosidase protein, as produced by RNA hairpin loops in CHO cells, must be attributed to some loop-induced degradation of the RNA, probably by certain nucleases which have not yet been identified.