ABSTRACT
The length requirements of the antisense portion of hammerhead ribozymes for efficacy in living cells was investigated. The HIV-1 tat-directed asymmetric hammerhead ribozyme [alpha]YRz195 was used with a 195 nt 3'-antisense arm and a 3 nt 5'-antisense portion as well as a set of successively 3'-shortened derivatives thereof. In the 3'-antisense arm a minimum length of 20 complementary nucleotides was required for efficient association with a 645 nt target RNA transcript in vitro (for all constructs k ass ranged between 0.3 and 1.8 104/M/s). The cleavage rate constants (kcleav) were independent of the length of the antisense flank and ranged between 0.8 and 1.2*10-4/s. However, the length of the antisense arms, as well as the mode of delivery and the subcellular location of the ribozymes, had a dramatic effect on efficacy in HIV-1-producing human cells. When proviral HIV-1 DNA and ribozymes were co-microinjected into the nucleus of human cells, a minimum length of 51 nt in the antisense arm was necessary for antisense- and ribozyme-mediated inhibition of HIV-1 replication. Ribozymes with shorter antisense arms were almost ineffective. Conversely, short chain ribozymes, including those with chemical modifications, were superior to long chain ribozymes when co-microinjected into the cytoplasm. When transfected, all ribozymes showed an antisense effect as well as an additional ribozyme-mediated increase in inhibition. Consequences for the design and application of ribozymes are discussed.
Hammerhead ribozymes, originally found in some subviral plant pathogens (
1
), have been used successfully to inhibit gene expression and viral replication
in prokaryotic and eukaryotic cells (
2
-
6
). When designed to cleave a target RNA
in trans
, hammerhead ribozymes usually consist of two antisense stretches on either side
of a catalytic domain of 22 nt (
7
,
8
). Inhibition is thought to consist of at least two steps, binding of the
ribozyme to the target RNA via its antisense sequences and, subsequently, site-specific hydrolysis of a phosphodiester bond 3' of the cleavable motif NUX (
9
-
12
). For a complete catalytic cycle, dissociation of the cleaved products has to
occur before the ribozyme can enter another cycle. For efficient association as
well as fast dissociation
in vitro
, the optimal length of the antisense flanks was estimated to be in the range 6-8 nt on either side of the catalytic domain (
13
,
14
). Much longer antisense flanks are assumed to prevent the dissociation step,
resulting in only stoichiometric cleavage of the target. Surprisingly, when
hammerhead ribozymes with antisense flanks of several hundred nucleotides were
applied in living cells, the catalytic activity significantly improved the
antisense-mediated inhibitory effects (
15
,
16
). However, several studies suggest that there is a significant influence of the
length of antisense portions of hammerhead ribozymes on the inhibitory activity
(
17
-
20
). Thus, the optimal length of antisense flanks for ribozyme-mediated inhibition in living cells may differ from predictions derived
from
in vitro
studies.
Further, the co-localization of the ribozyme with its target, i.e. the subcellular
distribution of ribozymes, is assumed to be crucial for efficacy. Hammerhead
ribozymes can be delivered to living cells either endogenously by introduction
of recombinant ribozyme coding genes or exogenously by transfection or simply
by the addition to the cell culture medium, as with antisense oligonucleotides
(
3
). In the latter case, chemically modified ribozymes are generally used to
substantially increase the stability and efficacy of ribozymes in the cell (
6
,
22
,
22
). The size of chemically synthesized ribozymes is limited, however, to less
than ~50 nt. Therefore, the size of the antisense flanks is restricted to 14 nt
or less on either side of the catalytic domain.
In this work, we used conventional symmetric as well as asymmetric hammerhead
ribozymes to systematically analyze the length requirements of the antisense
flanks of hammerhead ribozymes for efficient association and cleavage
in vitro
as well as for efficacy in human cells. To address the influence of subcellular
localization on efficacy against HIV-1, the RNAs were microinjected into the nucleus or cytoplasm of SW480
cells or transfected together with infectious proviral HIV-1 DNA followed by quantification of virus production.
Plasmid pBS-Rz12 was used for the
in vitro
transcription of ribozyme 2as-Rz12 (Fig.
1
, left side), in which the hammerhead domain has been inserted into the HIV-1-directed antisense RNA 2as (
15
). The plasmids p[alpha]YRz195+
and p[alpha]YRz195-
(deletion of G12 in the catalytic core; numbering according to
23
) were described recently (
16
). The templates for the
in vitro
transcription of shortened derivatives of p[alpha]YRz195+
and p[alpha]YRz195-
respectively (Fig.
1
, right side) were generated by PCR. The 5'-DNA primer oligonucleotide for PCR contained a T7 promoter sequence
(5'-GCAGATCTCGAGTAATACGACTCACTATAGGGAACAAAAGCTTATCTC-3'). The sequences of the 3'-DNA primer oligonucleotides for PCR
(20mers) for the successively shortened hammerhead ribozymes correspond to the
last 20 nt of the respective 3'-antisense arm and can be derived from Figure
1
.
For the
in vitro
synthesis of 2as-Rz12, plasmid pBS-Rz12 (
15
) was linearized with
Sac
I, followed by trimming the ends with T4 DNA polymerase. The plasmids p[alpha]YRz195+
and p[alpha]YRz195-
(
16
) were linearized with
Eco
RI for the
in vitro
synthesis of [alpha]YRz195+
and [alpha]YRz195-
respectively. For the synthesis of the shorter ribozyme [alpha]YRz60+
and the control [alpha]YRz60-
the same plasmids were linearized with
Dde
I. For the synthesis of [alpha]YRz20/25/31/35/41/45/51 and the respective inactive ribozymes, PCR-generated templates were used directly for the transcription of RNA
in vitro
. Catalytically inactive ribozymes were obtained by deletion of G12 in the
central core. For the synthesis of SR6 RNA, used as the HIV-1 target for the determination of cleavage and association rates
in vitro
,
Not
I-linearized plasmid pRC-CMV-SR6 was used (
24
). Seven micrograms of template DNA were incubated in a reaction mixture
containing 18 mM Na2HPO4, 2 mM NaH2PO4, 5 mM NaCl, 20 mM dithiothreitol, 8 mM MgCl2, 4 mM spermidine and 1 mM NTPs in a total volume of 300 [mu]l. Reactions were started by adding 60 U T3 RNA polymerase in the case of
linearized templates or T7 RNA polymerase when the PCR-generated templates were used. After 2 h incubation at 37oC the reactions were stopped by adding 200 [mu]l 20 mM MgCl2 and 40 U DNase I. After a further incubation for 20 min at 37oC, the solutions were extracted with phenol and RNAs were precipitated by
adding a 1/10 vol. of 3 M sodium acetate and a 2.0* vol. of ethanol. The pellet was dissolved in 10 mM Tris-HCl, pH 8.0, 1 mM EDTA and the RNAs were further purified by
chromatography with Sephadex G-50 using the same buffer.
The short chain
gag
-directed symmetric ribozymes Rz12x12+
and Rz12x12-
, where guanosine at position 5 was replaced by adenine, and Rz7x7+
were chemically synthesized as described (
25
). In ribozyme Rz10x10mod all cytidines and uridines were replaced by their 2'-fluoro derivatives but uridine at position 4 was 2'-aminouridine and uridine at position 7 adenosine. The
last internucleotide phosphate at the 5'-end and the last three at the 3'-end were phosphorothioates, as described previously (
25
).
For the determination of association rate constants, ribozymes and the
complementary target RNA SR6 were annealed essentially under the conditions
described previously (
15
). Briefly, the annealing reactions with 32
P-labeled ribozymes and unlabeled target RNA SR6 were performed in a total
volume of 30 [mu]l in a buffer containing 100 mM NaCl, 20 mM Tris-HCl pH 7.4, 10 mM MgCl2 at 3 nM ribozyme and 20-100 nM SR6 (pseudo first order kinetics). Aliquots (3 [mu]l) were withdrawn from the reaction mixtures and added to 30 [mu]l pre-cooled (0oC) stop buffer (50 mM Tris-HCl, pH 8.0, 25 mM EDTA, 8 M urea). The single-stranded and double-stranded fractions were separated
on 5% polyacrylamide gels containing 8 M urea in a Tris-borate buffer (89 mM Tris-borate, pH 8.3, 25 mM EDTA). The temperature was kept below 25oC during electrophoresis. Under these experimental conditions,
RNA-RNA double strands of 16 bp or more remained stable (Homann, unpublished
results). The amounts of RNA contained in bands representing the single-stranded ribozyme or the duplex were quantified using a PhosphorImager
(Molecular Dynamics) and were used to calculate the association rate constants
(
k
ass).
The ribozyme cleavage assays were performed under single turnover conditions as
described recently (
15
). Briefly, the ribozymes (20 nM) were incubated with 32
P-labeled target RNA SR6 (2 nM) at 37oC for 15 h to ensure complete annealing in a solution containing 100
mM NaCl, 20 mM Tris-HCl, pH 7.4. The cleavage reaction was started by the addition of Mg2+
to a final concentration of 10 mM. Aliquots were withdrawn at different time
points and added to stop buffer (see above). The reaction products were heated
to 96oC for 5 min and separated in polyacrylamide gels containing 8 M urea. From
quantification of the time-dependent decrease in substrate we derived the cleavage rate constants (
k
cleav) according to a single exponential decay.
Ribozymes were directed against the 5'-leader/
gag
portion of HIV-1 or the
tat
/
rev
coding exon respectively (Fig.
1
). The
gag
-directed ribozymes were `symmetric', i.e. both antisense arms had the same
length in the case of the short chain constructs Rz7x7+
, Rz10x10mod, Rz12x12+
and Rz12x12-
and >100 nt for the long chain ribozyme 2as-Rz12 (Fig.
1
). The
tat
/
rev
-directed ribozymes were derived from [alpha]YRz195+
(
16
), which was designed asymmetrically, i.e. when complexed with the target helix
I consisted of only 3 bp whereas helix III almost exclusively contributed to
binding of the target (Fig.
1
).
Replication of HIV-1 was measured by co-microinjection of
in vitro
synthesized ribozymes together with infectious proviral HIV-1 DNA pNL4-3 (
26
) into the nucleus or the cytoplasm of SW480 cells as described (
27
,
28
). SW480 cells were grown on glass coverslips (10 mm diameter) in DMEM medium
supplemented with 10% fetal calf serum. The DNA and RNA was dissolved in 10 mM
Tris-HCl, pH 8.0, 1 mM EDTA at concentrations of 10 ng/[mu]l pNL4-3 and 70 ng/[mu]l RNA. One day after microinjection, 2 * 105
MT-4 cells were added and the final volume was adjusted to 1 ml with RPMI
medium supplemented with 10% fetal calf serum. Virus production was measured 5
days post-injection and quantified by a commercial p24 ELISA (Organon). For each
ribozyme, 50 cells were microinjected and the experiment was repeated six to
nine times. As a positive control, we used a
cat
RNA transcript (850 nt), tRNAbulk or the HIV-1
tat
-derived `sense' RNA se100, a 100 nt long sense transcript (
29
), respectively. Replication in the presence of
cat
RNA was set to 100%.
Ribozymes were also co-transfected together with the proviral DNA pNL4-3 as described (
15
). Briefly,
in vitro
synthesized RNA was transfected together with 40 ng pNL4-3 into SW480 cells grown semi-confluent (~5 * 103
cells) in 48-well plates by calcium phosphate co-precipitation (
30
). One day after transfection, 2 * 105
MT-4 cells were added and production of HIV-1 was measured as described above.
The
tat
/
rev
-directed hammerhead ribozyme (
16
) is characterized by an asymmetric design of the antisense arms that flank the
catalytic domain (Fig.
1
, lower right panel). Whereas the 5'-antisense region consists of only three complementary nucleotides,
the 3'-terminal region of the largest ribozyme [alpha]YRz195+
contains 195 nt. Leaving the 5'-antisense region and the catalytic domain unchanged, derivatives of
[alpha]YRz195+
were made with truncated 3'-antisense flanks of 60, 51, 45, 41, 35, 31, 25 and 20 nt (Fig.
1
). For each of these RNA, we determined the rate of association (
k
ass) with the target RNA SR6 (645 nt) and its rate of cleavage (
k
cleav). The association ranged between 1.8 * 104
/M/s for the parental ribozyme [alpha]YRz195+
and 3.0 * 103
/M/s for [alpha]YRz20+
and [alpha]YRz25+
(Table
1
). The size of the 3'-antisense arm did not seem to correlate with the
k
ass values, since [alpha]YRz31+
, [alpha]YRz41+
and [alpha]YRz51+
showed relatively fast annealing (
k
ass = 1.5 * 104
/M/s), whereas the intermediate chain lengths ([alpha]YRz35+
and [alpha]YRz45+
) annealed ~3- to 4-fold slower (Table
1
). For each of the above mentioned constructs, a catalytically inactive control
RNA was prepared in which nucleotide G12 (
23
) was deleted (see Fig.
1
, arrow). This deletion had no influence on the association rates
in vitro
(data not shown).
Table 1
The cleavage rate constants for all ribozymes were determined under single
turnover conditions (
15
). The values of
k
cleav were not correlated with the length of the 3'-antisense arm and ranged between 0.4 and 1.2 * 10-4
/s (Table
1
). The factor of two to three between the
k
cleav values is almost within the error range of the measurements. These relatively
slow cleavage rates were similar to those found for other long chain ribozyme-substrate complexes and seem to reflect a slow step after annealing and
before product release (reviewed in
31
).
The influence of the length of the antisense arm of derivatives of [alpha]YRz195+
on the extent of inhibition of HIV-1 replication was measured in two different ways: by microinjection and by
calcium phosphate co-precipitation. First,
in vitro
active and inactive [alpha]YRz195-derived ribozymes were microinjected together with proviral DNA
into the nucleus of human SW480 cells. Regardless of whether catalytically
active or not, essentially no inhibition of HIV-1 replication occurred with ribozymes with a 3'-arm of 45 nt or shorter (data summarized in Fig.
2
). A weak reduction of virus production (46%) was observed with the
in vitro
inactive ribozyme [alpha]YRz51-
and virus production was further 2.5-fold decreased with the
in vitro
active ribozyme [alpha]YRz51+
(Fig.
2
). A stronger antisense-mediated inhibition was observed with the inactive ribozyme [alpha]YRz60-
(31% virus production) and this inhibition was 5-fold increased with the active ribozyme [alpha]YRz60+
(7% virus production). The additional ribozyme-mediated inhibition was thus stronger than for the less inhibitory
ribozyme [alpha]YRz51+
. The ribozyme-mediated increase in antisense effects could only be observed if a
significant antisense-mediated inhibition occurred; the antisense effect seemed to be necessary
for the additional ribozyme effect.
In view of the above described different inhibition potentials of ribozymes with
long and short helix III-forming regions, depending on the mode of delivery, we decided to compare
direct microinjection into the nucleus with into the cytoplasm. In addition to
the
tat
/
rev
-directed ribozyme [alpha]YRz195+
, we also used ribozymes directed against the 5'-leader/
gag
region of HIV-1. This included the long chain ribozyme 2as-Rz12 (Fig.
1
;
15
) and the short symmetric ribozymes Rz12x12+
, Rz7x7+
and Rz10x10mod, which recognize the same target motif but contain only 12, 10
or 7 nt in each of their symmetric antisense arms (Fig.
1
).
When directly delivered into the nucleus, both long chain ribozymes 2as-Rz12 and [alpha]YRz195+
respectively showed ~99% inhibition (Fig.
4
). Conversely, when the same constructs were microinjected into the cytoplasm,
weak inhibition was observed with ribozyme [alpha]YRz195+
, with no significant inhibition with 2as-Rz12 and the antisense control [alpha]YRz195-
(Fig.
4
). The short chain ribozyme Rz12x12+
as well as the inactive control Rz12x12-
completely abolished production of HIV-1 when microinjected into the cytoplasm, but did not show any inhibition
when directly delivered into the nucleus (Fig.
4
).
To systematically study the influence of the length of helix III on association
and cleavage rates
in vitro
and inhibitory activity in living cells, the antisense arm of [alpha]YRz195+
was systematically shortened from 60 to 20 nt. A set of
tat
/
rev
-directed ribozymes with antisense flanks of 60, 51, 45, 41, 35, 31, 25 and
20 nt was constructed and a minimum length of the 3'-antisense arm of 51 nt was found to be required for antisense-mediated inhibition of HIV-1 replication when microinjected into the nucleus (~50% virus production) with the catalytically
inactive ribozyme [alpha]YRz51-
. This inhibition was further increased with the active ribozyme [alpha]YRz51+
(Fig.
2
). The minimum length required for inhibition of ~50 nt does not correlate with association rates determined
in vitro
, which were comparable with those for shorter 3'-antisense flanks (Table
1
). The HIV-1 inhibition data show that the mode of ribozyme delivery has an
extraordinary influence on ribozyme effects. Secondly, the experimental data
demonstrate that the presence of at least one long antisense arm in a
hammerhead ribozyme is highly favourable and almost a requirement for activity
in the nucleus. Similar observations have also been made in the use of a
different set of HIV-1-directed hammerhead ribozymes (Thompson
et al
., submitted). It should be noted, however, that our work made use of two
defined sets of ribozymes that add significantly to the understanding of
ribozyme design and delivery but that may not allow the derivation of general
conclusions. For example,
in vitro
transcribed hammerhead ribozymes with short antisense flanks can also show
inhibition in the nucleus (Thompson
et al
., submitted).
It is noteworthy that earlier studies on HIV-1-directed antisense RNA showed that the association rates
in vitro
correlated with efficacy in living cells when transfected into the cytoplasm (
32
). However, in these earlier studies the
k
ass values for slow and fast annealing antisense species differed by >100-fold, whereas here differences were maximally 6-fold. Thus, small differences in
k
ass values may not be important for effectiveness in the cell. A similar
observation has been made recently by W. James (personal communication). More
importantly, the results shown in Figure
2
could be interpreted such that the length requirements in the nucleus reflect
another critical parameter which is not relevant when RNA is delivered to the
cytoplasm by transfection protocols. For example, the stability of duplex RNA
formed between the ribozyme and the target in the nucleus of human SW480 cells
could be over-estimated when compared with melting temperatures
in vitro
at 37oC and physiological ionic strength. This hypothesis is consistent with the
complete lack of inhibition by the short chain chemically synthesized ribozymes
(Fig.
4
) that were proven to be effective in the cytoplasm. Recent experiments on the
dissociation of long chain duplex RNA (56 bp) showed that strand exchange can
occur at 37oC and can be increased by up to 105
-fold in the presence of cetyltrimethylammonium bromide (
33
), a known facilitator of nucleic acid annealing (
34
). Functionally similar facilitators in the nuclei of mammalian cells, such as
the hnRNP proteins (
35
-
37
) or the tumor suppressor protein p53 (
38
,
39
), could weaken interactions even between long chain RNA duplexes in the nucleus
and make them more dynamic than expected from
in vitro
data. It has been consistently suggested that a `helix-destabilizing environment' in the nucleus could shift the optimal length
of the antisense arms from eight to 20 or more nucleotides (
40
). Accordingly, the observed minimal antisense length of ~50 nt for a ribozyme to be active
in vivo
could reflect the stability of interactions that are necessary for biological
effectiveness in the nucleus. This is consistent with the earlier finding that
a similar minimum length of ~60 complementary nucleotides was important for inhibition by endogenously
expressed HIV-1-directed hammerhead ribozymes (
17
).
A completely different picture emerges from the transfection experiments. Here,
the short ribozymes with antisense flanks of 20 and 25 nt were most effective
(Fig.
3
).
In vitro
active ribozymes showed greater efficacy than
in vitro
inactive derivatives in the cytoplasm. Consistently, the synthetic ribozymes
Rz12x12+
, Rz10x10mod and Rz7x7+
performed well in the cytoplasm after microinjection, whereas the long chain
constructs 2as-Rz12 and [alpha]YRz195+
were not effective when tested at the same concentration of 70 ng/[mu]l (Figs
4
and
5
). When the long chain ribozymes 2as-Rz12, [alpha]YRz195+
or [alpha]YRz195-
respectively were applied at the same molarity (4.6 [mu]M, e.g. 640 ng/[mu]l for 2as-Rz12) as the short ribozymes at 70 ng/[mu]l (e.g. Rz12x12+
, 4.6 [mu]M; see Fig.
4
), inhibition could not be measured unequivocally, since all controls also
showed inhibition (data not shown). This non-specific RNA-mediated inhibition of HIV-1 seems to be due to a general toxicity of RNA at high
concentrations (640 ng/[mu]l). Thus, it remains open whether long chain ribozymes can elicit specific
inhibition in the cytoplasm. A consequence for application is to direct long
chain inhibitory RNA to the nucleus of target cells and to deliver short
constructs to the cytoplasm. For endogenous expression from the chromosome especially, long constructs seem to
be favourable.
It should be noted that the consistency of results obtained by calcium phosphate transfection and by cytoplasmic microinjection does not necessarily
mean that transfection delivers the RNA in an active form into the cytoplam. One could also imagine that endogenous
transport of transfected ribozymes guides them into the nucleus in a form that
is different from direct nuclear microinjection.
The stronger inhibition observed with Rz10x10mod than with Rz7x7+
and even Rz12x12+
shows that increased biological stability of a ribozyme by chemical
modification can substantially increase efficacy in living cells. The increased
efficacy of Rz10x10mod versus Rz7x7+
and Rz12x12+
could also reflect a stabilizing influence of the 2'-modification on stability of the ribozyme-substrate complex in the living cell. Since stabilization of
a chemically synthesized short ribozyme (Rz10x10mod; Fig.
5
) led to an increase in inhibition, this finding strongly suggests the
employment of therapeutic concepts based on exogenous delivery of chemically
modified ribozymes. This concept has been shown to be successful in a number of
applications (
41
,
42
). However, properties such as cellular uptake, prolonged half-life and high efficacy will have to be improved.
When delivering ribozymes into the nucleus, long antisense arms (>50 nt) seem to
be required for effectiveness. Such ribozymes are most easily expressed
endogenously in transduced cell lines from recombinant ribozyme genes or in
transgenics. Assuming that the locus of ribozyme action is the nucleus, one
could think of improving the efficacy of ribozyme transcripts by preventing
their nuclear export.
Regardless of the mode of delivery, once an antisense effect was observed,
inhibition was further increased by a factor of two to 10 (for most RNA pairs)
when an
in vitro
active ribozyme was used. Thus, based on the assumption that antisense effects
are a necessary prerequisite for ribozyme-mediated inhibition, one should aim to combine effective antisense species
with the catalytic hammerhead domain. The asymmetric design is well suited to
this approach (
16
). The increased efficacy measured with ribozymes versus
in vitro
inactive controls has to be explained formally by their cleavage capability,
although the cleavage rate constants measured
in vitro
seem to be too slow to play a role in the cell (
43
). Much higher cleavage rates were observed when higher Mg2+
concentrations were used (
44
), indicating that faster cleavage rates can in principle be achieved.
In vivo
, faster catalysis could be facilitated by cellular factors (
45
-
47
).
We thank H.zur Hausen for continuous support and W.Nedbal for critical comments
on this work. We acknowledge financial support by the Deutsche
Forschungsgemeinschaft (SC14/2-1), by the Bundesministerium für Bildung Forschung und Wissenschaft (01KV9517) and by HCM contract
ERBCHRXCT930162 of the European Union.
*To whom correspondence should be addressed. Tel: +49 6221 424939; Fax: +49 6221
424932; Email: sczakiel@dkfz-heidelberg.de
+
Present address: Genzentrum am Max-Planck-Institut für Biochemie, Am Klopferspitz 18a, D-82152 Martinsried, Germany
Ribozyme
Length of 3'-antisense arm (nt)
k
ass
(per M/s)
k
cleav
(per s)
[alpha]YRz20+
20
0.3 * 104
0.4 * 10-4
[alpha]YRz25+
25
0.3 * 104
0.5 * 10-4
[alpha]YRz31+
31
1.4 * 104
0.8 * 10-4
[alpha]YRz35+
35
0.5 * 104
1.0 * 10-4
[alpha]YRz41+
41
1.5 * 104
0.9 * 10-4
[alpha]YRz45+
45
0.4 * 104
1.2 * 10-4
[alpha]YRz51+
51
1.6 * 104
1.1 * 10-4
[alpha]YRz60+
60
1.6 * 104
1.2 * 10-4
[alpha]YRz195+
195
1.8 * 104
1.2 * 10-4
Rz7x7+
7b
n.d.c
0.5 * 10-4
Rz10x10mod
10b
n.d.c
0.1 * 10-4
Rz12x12+
12b
n.d.c
0.5 * 10-4
REFERENCES
Return