Báo cáo khoa học: potx

Design of hairpin ribozyme variants with improved activity for poorly processed substrates Irene Drude1,*, Anne Strahl2, Daniel Galla2, Oliver Mu¨ller1, and Sabine Mu¨ller2 1 Max Planck

Design of hairpin ribozyme variants with improved activity for poorly processed substrates

Irene Drude1,*, Anne Strahl2, Daniel Galla2, Oliver Mu¨ller1,  and Sabine Mu¨ller2

1 Max Planck Institute for Molecular Physiology, Department I, Dortmund, Germany

2 Ernst-Moritz-Arndt Universita¨t Greifswald, Institut fu¨r Biochemie, Germany

Introduction

In recent years, a number of ribozymes, particularly

the rather small hammerhead and hairpin ribozymes,

have been designed for cleavage of therapeutically

rele-vant targets [1–5] Cleavage occurs at conserved sites

that first need to be identified on the target; this is

followed by adapting the sequence of the ribozyme

substrate-binding domain to specifically recognize,

bind and cleave the chosen site on the target RNA

In vitroselection studies have allowed the identification

of hammerhead ribozymes for cleavage of sites with altered sequences [6,7] However, in the case of the hairpin ribozyme, to the best of our knowledge, all variants that have so far been designed for specific RNA destruction cleave their substrates within the consensus sequence 5¢-Y)2N)1*G+1U+2Y+3B+4-3¢

We have started an effort to design a hairpin ribozyme

Keywords

cleavage; hairpin ribozyme; kinetics; ligation;

RNA

Correspondence

S Mu¨ller, Ernst Moritz Arndt Universita¨t

Greifswald, Institut fu¨r Biochemie, Felix

Hausdorff Str 4, 17487 Greifswald,

Germany

Fax: +49 (0) 3834 864471

Tel: +49 (0) 3834 8622842

E-mail: sabine.mueller@uni-greifswald.de

*Present address

NOXXON Pharma AG, Max-Dohrn-Strasse

8-10, 10589 Berlin, Germany

 Present address

University of Applied Sciences

Kaiserslau-tern, Campus Zweibru¨cken, Amerikastraße 1,

66428 Zweibru¨cken, Germany

(Received 26 August 2010, revised

1 December 2010, accepted 6 December

2010)

doi:10.1111/j.1742-4658.2010.07983.x

Application of ribozymes for knockdown of RNA targets requires the iden-tification of suitable target sites according to the consensus sequence For the hairpin ribozyme, this was originally defined as Y)2N)1*G+1U+2

Y+3B+4, with Y = U or C, and B = U, C or G, and C being the preferred nucleobase at positions)2 and +4 In the context of develop-ment of ribozymes for destruction of an oncogenic mRNA, we have designed ribozyme variants that efficiently process RNA substrates at

U)2G)1*G+1U+2A+3A+4 sites Substrates with G)1*G+1U+2A+3 sites were previously shown to be processed by the wild-type hairpin ribozyme However, our study demonstrates that, in the specific sequence context of the substrate studied herein, compensatory base changes in the ribozyme improve activity for cleavage (eight-fold) and ligation (100-fold) In partic-ular, we show that A+3and A+4are well tolerated if compensatory muta-tions are made at posimuta-tions 6 and 7 of the ribozyme strand Adenine at position +4 is neutralized by G6fi U, owing to restoration of a Watson– Crick base pair in helix 1 In this ribozyme–substrate complex, adenine at position +3 is also tolerated, with a slightly decreased cleavage rate Addi-tional substitution of A7with uracil doubled the cleavage rate and restored ligation, which was lost in variants with A7, C7 and G7 The ability to cleave, in conjunction with the inability to ligate RNA, makes these ribozyme variants particularly suitable candidates for RNA destruction

Abbreviations

CPG, controlled pore glass; dNTP, deoxynucleoside triphosphate; ds, double strand; EDTA, ethylene diamine tetraacetic acid; lcaa, long chain amino alkyl; NHS, N-hydroxy succinimidl; NTP, nucleoside triphosphate; PAGE, polyacrylamide gele electrophoresis; RP-HPLC, reversed phase high performance liquid chromatography.

for cleavage of the CTNNB1 mRNA encoding the

proto-oncoprotein b-catenin b-Catenin is an effector

of the canonical Wnt signaling pathway, which plays

essential roles in the regulation of cell growth, mobility

and differentiation High intracellular concentrations

of b-catenin can induce constitutive activation of Wnt

target genes, which has been proposed to be an

impor-tant oncogenic step in cancerogenesis [8,9] Therefore,

systematic suppression of b-catenin expression by

ribo-zyme-mediated destruction of CTNNB1 mRNA could

be a suitable way to counteract cancer development

and progression

The hairpin ribozyme is derived from the negative

strand of the tobacco ringspot virus satellite RNA,

and catalyzes the reversible cleavage of a

phosphodi-ester bond through an SN2-like mechanism, leading to

characteristic products with 2¢,3¢-cyclic phosphate and

5¢-OH termini [10–12] The minimal catalytic motif is

characterized by a two-stem structure, each stem

con-sisting of a central loop region flanked by two helices

For catalysis, the hairpin ribozyme has to undergo

conformational changes that bring the two loops into

close proximity [13–17] This docking process generates

a complex network of interactions between the bases

in the two loops, with a ribose zipper, hydrogen

bonds, noncanonical base pairs and a Watson–Crick

base pair between G+1in loop A and C25in loop B as

characteristic elements [18–21] The consensus sequence

of the hairpin ribozyme, determined by site-directed

mutagenesis [22–27] and in vitro evolution methods

[25–30], defines the helical regions as being highly

flexi-ble in sequence, provided that complementarity is

pre-served In contrast, base substitutions within the loops

strongly interfere with catalytic activity Therefore,

suitable RNA substrates were originally supposed to

fulfill the following sequence requirements: reversible

cleavage occurs between the conserved G+1 and N)1

within the 5¢-Y)N)1*G+1U+2Y+3B+4-3¢ motif located

in loop A, where N can be any base, B can be

cyto-sine, guanine or uracil (with cytosine being the

pre-ferred base), and Y can be uracil or cytosine (with

cytosine preferred over uracil) (Fig 1A) In the

wild-type hairpin ribozyme, each of these bases participates

in interactions with partner bases in the ribozyme

strand [12,21] Therefore, mutations at these positions

could disrupt essential interactions or enforce

alterna-tive ones, with a strong input on catalytic activity

However, previous results imply that base changes in

the ribozyme domain can compensate for changes in

the conserved bases in the substrate [31], indicating

that the hairpin ribozyme can flexibly respond to base

substitutions in the substrate The postulated

consen-sus sequence was later refined by Berzal-Herranz and

coworkers [32], who showed that previous studies failed to evaluate all possible combinations of nucleo-tides surrounding the cleavage site On the basis of the analysis of 64 substrate variants, Pe´rez-Ruiz et al [32] demonstrated that, in addition to the wild-type A*GUC substrate, H*GUC (H = A, C or U), G*GUN, G*GGR (R = A or G), A*GUU and U*GUA substrates were also sufficiently well cleaved, although with about five-fold lower activity When CTNNB1 mRNA was examined, no target site fully corresponding to the hairpin ribozyme consensus sequence YN*GUYB could be identified Therefore,

we decided to search for a site that keeps at least some

of the required nucleotides intact, and to design hair-pin ribozyme variants for cleavage at this specific site The major criterion for defining a suitable target site was the presence of a G+1immediately at the cleavage site, because substitution of G+1with any of the other natural RNA bases has been shown to completely abolish activity [26,28,33,34] Although substitution of

G+1 can be compensated for by corresponding substi-tution of C25 in loop B, regenerating the interdomain Watson–Crick base pair, the catalytic activity of the resulting double mutants was rather low [31] There-fore, we decided to retain the essential G+1, and chose

a site consisting of U)2G)1*G+1U+2A+3A+4 (Fig 1A), with U)2, A+3and A+4being distinct from the wild-type hairpin ribozyme substrate According to the nomenclature used in the study of Pe´rez-Ruiz et al [32] mentioned above, the chosen target site corre-sponds to a G*GUA substrate, which was found to be cleaved by the hairpin ribozyme with about three-fold lower activity In the context of the CTNNB1 substrate used in this study, cleavage activity was reduced by a factor of 60 as compared with cleavage of the typical A*GUC substrate [35] (Table 1) In order to improve activity to the level of that with the wild-type A*GUC substrate, we decided to search for hairpin ribozyme variants with base substitutions in the ribozyme strand that might restore full activity Furthermore, the bases

at positions)2 and +4 (not included in the study of Pe´rez-Ruiz et al [32]) also do not fully correspond to the consensus sequence (Fig 1A), and therefore require additional investigation In general, there are two possible ways of adapting the ribozyme sequence

to a specific target sequence Suitable ribozymes can be developed by selection of active species from a random library, or by rational design For the hairpin ribo-zyme, a number of crystal structures are available [20,21,36–40] Careful inspection of the crystal struc-tures reveals that nucleobases at positions +3, +4 and )2 of the substrate strand interact only with nu-cleobases in loop A of the ribozyme strand, without

A C G G A

G C U C

G C

C G

U A

G C

A G

A

A C

A

C

A

U

A

U

A U G

G C

A

U G

C G

G U

A U

HP-CTNNB1 N7

WT

6

U = G

7

N = A

A C G G A A

A

U G C C U U

N

G C U C

G C

C G

U

G C

A G

A

A

A C

A

C

A

U

A

U

A U G

U G

A U

HP-CTNNB1 N7

A C G G A A

A

U G C C U U

N

GC U C

G C

C G

U

G C

A G

A

A C

A

C

A

U

A

U

A U G

A

U G

A U

HP-CTNNB1 N7

C C

A

A G G A G

N

G C

C G

U

G C

A G

A

A

A C

A

C

A

U

A

U

A U G

A

A U

HP-CTNNB1 N7

C C

A

G

C C U C

U

U G

A

A

A

A

A

A

A G G

G

G G

G G

U

U

U

C C

C

2 ′,3′cp-GUGAGUCUCUUCCUCG -5′

U

A U

U

C

A

G C

A

G C

A

C G

G U

U

A U

U

C

A

G C

G C

A

C G

G U

U

A U

U

C

A

G C

A

G C

A

C G

G U

U

A U

U

C

A

G C A

A

B

C

U

A G

C U

A G

C

G

C

G

C

C

A G C

3 ′ A

C C

A

A G G A G A G

A

C C

A

A G G A G A G C

S-CTNNB1-1

S-CTNNB1-2

S-CTNNB1-3

S-CTNNB1-2

+3 +4

6

7 11

U

N

UG

+3 +4

Y

A A

G A

U

+3 +3

+4

G

Hairpin ribozyme consensus sequence

Hairpin ribozyme

Fig 1 Hairpin ribozyme variants for knockdown of CTNNB1 mRNA (A) Sequences of the wild-type hairpin ribozyme and the consensus sequence according to [25], and the CTNNB1 target sequence (N = A, C, G, U; B = C, G, U; Y = C, U) Cleavage sites are marked by arrows (B) Two-way-junction hairpin ribozymes were used for analysis of the cleavage reaction In this process, the 20mer S-CTNNB1-1 sub-strate, corresponding to the CTNNB1 target site, is cleaved into a 15mer S-CTNNB1-2 fragment and a 5mer product, which should rapidly dissociate from the ribozyme strand WT refers to a wild-type hairpin ribozyme that was adapted for recognition of the CTNNB1 substrate Base changes in comparison with HP-CTNNB1 N7 are shown (boxed area) (C) Hairpin ribozyme constructs for analysis of the ligation reac-tion Binding of the 3¢-cleavage product ⁄ ligation substrate, S-CTNNB1-2, and a second ligation substrate, S-CTNNB1-3, leads to the formation

of three-way-junction ribozymes with favored ligation properties The substrates S-CTNNB1-1 and S-CTNNB1-2 are labeled with the fluores-cent dye ATTO680 (indicated by the gray dot).

being involved in interdomain interactions On the

basis of this analysis, it seemed most straightforward

to rationally design compensatory mutations in loop A

of the ribozyme strand, and thus to develop hairpin

ribozyme variants with improved activity for the

cho-sen UG*GUAA substrate

Results

Design of hairpin ribozymes targeting a CTNNB1

mRNA model substrate at a UG*GUAA site

Literature data show that substitution of wild-type

C+4 by adenine strongly decreases cleavage activity,

probably because of destabilization of helix 1 as a

result of the emerging G–A mismatch [25] However,

in vitro selection studies afforded hairpin ribozymes

with uracil instead of cytosine at position +4 in the

substrate strand, and, in addition, adenine instead of

guanine at position 6 of the ribozyme strand [25] This

result allows for the conclusion that, essentially, a

Watson–Crick base pair is required at this location

Thus, we replaced G6with uracil in the CTNNB1

ribo-zyme, assuming that the resulting A+4–U6 base pair

would restore activity

There are no literature data available on

compensa-tory mutations for substitutions at position C+3

On the contrary, it has been shown that the single

substitution C+3fi A strongly decreases activity [26]

Careful inspection of crystal structures of the hairpin

ribozyme–substrate complex as four-way-junction

[20,21] and minimal junction-less [36–40] structures,

however, shows that C+3 forms a noncanonical base

pair with the nucleobase at position 7 in the ribozyme

strand, which naturally is adenine Therefore, we

investigated whether a single base substitution at

posi-tion 7 in the CTNNB1 ribozyme can compensate for

C+3fi A in the substrate Accordingly, we designed

four hairpin ribozymes carrying any of the four bases

at position 7 (hence dubbed A7, C7, G7 and U7

vari-ants) and analyzed their cleavage and ligation proper-ties in comparison with those of a wild-type hairpin ribozyme that was adapted to recognize CTNNB1 RNA (Fig 1) To analyze the cleavage efficiencies of all hairpin ribozyme motifs, we chemically synthesized

a 20mer RNA substrate, S-CTNNB1-1, containing the target CTNNB1 mRNA sequence and a 3¢-terminal

NH2-linker for postsynthetic labeling of the substrate with ATTO680 This allows for detection and quantifi-cation of the cleavage event on a DNA sequencer, as shown previously [41] The substrate and the ribozyme form a two-way-junction structure with a 12-bp helix 1 and a 4-bp helix 2 (Fig 1B) According to the model, cleavage and subsequent rapid dissociation of the 5¢-cleavage product leads to destabilization of helix 2 Therefore, the reaction pathway for ribozyme-mediated cleavage of the RNA substrate can be described as:

Rþ S Ðkon

koffR S !kcleav

R 30Pþ 50P

Under these conditions, the time dependence of the cleavage product concentration typically follows [42]:

½50P ¼ ½50P1ð1  ðexpÞðkðobs;cleavÞ tÞÞ Hence, the kinetic parameters kcleav and Km can be calculated from the observed cleavage rate kobs,cleavat different ribozyme concentrations [R]o, using the following equation [40]:

kobs;cleav¼ kcleav½R0

Kmþ ½R0 with

Km¼koffþ kcleav

kon

In order to use the same constructs for ligation stud-ies, we extended the ribozyme by 14 additional nucleo-tides at the 3¢-end Thus, upon binding of ligation

Table 1 Kinetic parameters of HP-CTNNB1 N7-mediated cleavage of S-CTNNB1-1 and S-CTNNB1-1 U)2C under single turnover conditions Amplitude values refer to maximal product yield obtained with 50-fold ribozyme excess over substrate For ribozyme sequences compare Fig 1.

Ribozyme

k cleav (min)1) K m (n M ) Amplitude k cleav (min)1) K m (n M ) Amplitude

substrates, a stable ribozyme–substrate complex

orga-nized in a three-way junction results, with both

liga-tion fragments being tightly bound to the ribozyme, by

12 and 15 bp, respectively (Fig 1C)

Intermolecular cleavage kinetics

Intermolecular single turnover cleavage kinetics were

examined for all hairpin ribozyme variants at 37C in

standard buffer, containing 40 mm Tris (pH 7.5) and

10 mm MgCl2 The data are summarized in Fig 2 and

Table 1 The wild-type hairpin ribozyme cleaves the

CTNNB1 substrate with kcleav= 0.007 min)1 and a

final yield of 34%, showing a 60-fold reduction in the

cleavage rate constant as compared with the cognate

A*GUC substrate ( 0.42 min)1[35]) The U7 variant

showed the best activity among the other four variants,

cleaving about 85% of the CTNNB1 substrate within

100 min The determined cleavage rate constant of

0.057 min)1 indicates an eight-fold increase as com-pared with the wild-type ribozyme, and was only seven-fold lower than the cleavage rate constant obtained for wild-type cleavage of the cognate A*GUC substrate [35] The A7 and C7 variants showed similar cleavage properties, with a maximal product fraction of  60% after 4 h Both ribozymes catalyzed the cleavage reaction with a kcleav of about 0.025 min)1, indicating an increase in the cleavage rate

of only three-fold to four-fold as compared with the wild-type hairpin ribozyme The G7 variant did not show any improvement in activity Only 30% cleavage product could be detected after 4 h With a kcleav of 0.005 min)1, it showed equally low activity as the wild-type ribozyme for cleavage of the CTNNB1 substrate

As mentioned above, the CTNNB1 substrate RNA used has a uracil instead of a cytosine at position –2

In order to evaluate the sole influence of base substitu-tions in the ribozyme strand, a modified CTNNB1 sub-strate was synthesized, carrying the consensus cytosine

at position –2, and the activities of the four variants for this substrate were tested C)2 in the CTNNB1 substrate increased cleavage rate constants about 10-fold in each variant as compared with cleavage of the U)2 substrate (Fig 3; Table 1) The U7 variant showed a slightly increased cleavage activity (kcleav= 0.75 min)1) as compared with the activity of the wild type for the consensus A*GUC substrate [35] The A7 and C7 variants catalyzed the cleavage reaction with a

kcleav of  0.2 min)1, indicating that the substitutions

at positions +3 (Cfi A) and +4 (C fi A) are well tolerated if adequate compensatory mutations (G6 fi U; N7 fi A, C or U) are made The G7 vari-ant again showed the lowest activity, with kcleav= 0.05 min)1 All variants catalyzed cleavage of the C)2

A7 C7 G7 U7

0.00

0.01

0.02

0.03

0.04

0.05

0.06

kobs

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

A

B

Reaction time (min)

Fig 2 HP-CTNNB1 N7-mediated cleavage of S-CTNNB1-1 under

single turnover conditions (A) Time course of reactions at 30-fold

excess of ribozyme over substrate (B) Dependence of k obs values

on ribozyme concentration.

A7 C7 G7 U7

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

kobs

Fig 3 Dependence of kobs values on ribozyme concentration Kinetic plot of HP-CTNNB1 N7-mediated cleavage of S-CTNNB1-1

U)2C under single turnover conditions.

substrate with a product yield similar to that of the

U)2 substrate, although with clearly shorter reaction

times

Intermolecular ligation kinetics

In order to fully characterize the designed hairpin

ribo-zyme variants, we also investigated the ligation

behav-ior In-trans ligation kinetics were measured in reactions

with ribozyme, 3¢-cleavage product ⁄ ligation substrate,

termed S-CTNNB1-2, and 5¢-ligation substrate,

S-CTNNB1-3 or S-CTNNB1-3 U)2C, containing a

2¢,3¢-cyclic phosphate terminus Binding of ligation substrates

to the ribozyme resulted in the formation of a stable

ribozyme–substrate complex, forming a

three-way-junc-tion structure (Fig 1C) Because of the stability of this

complex, ligation should be favored over cleavage,

although cleavage cannot be neglected Therefore, the

determined ligation rate will reflect an approach to the

equilibrium between cleavage and ligation, provided

that cleavage is much faster than dissociation of the

ribozyme–product complex It has to be taken into

account that the observed ligation rate will be the sum

of the cleavage and ligation rates Kinetic parameters

were determined under single turnover conditions, with

increasing concentrations of ribozyme and

S-CTNNB1-3 or S-CTNNB1-S-CTNNB1-3 U)2C with respect to the 3¢-ligation substrate

First, we investigated ribozyme-supported ligation of S-CTNNB1-2 to S-CTNNB1-3 with uracil at posi-tion)2 In contrast to cleavage analysis, where all variants were found to be active, ligation product was detected only for the wild type and the U7 variant (Fig 4), with 17% or 30% yield, respectively (Table 2) The wild type showed very little ligation activity Therefore, ligation was studied only at ribozyme satu-ration (50-fold excess of ribozyme⁄ 5¢-ligation substrate over 3¢-ligation substrate) to determine the correspond-ing kobs,lig (Table 2) For the A7, C7 and G7 variants, the ligation product levels were too low to be quanti-fied Kinetic data for three-way-junction ribozymes are not available from the literature, but the obtained

kapp,lig value for the U7 variant of 1 min)1 (Fig 5B; Table 2) lies within the range of typical ligation con-stants for two-way-junction and four-way-junction hairpin ribozymes [41] As observed for the wild type with its cognate substrates [43,44], the U7 variant also catalyzed ligation about 18 times faster than cleavage Therefore, the observed ligation rate constant essen-tially reflects the ligation step, as the reverse cleavage

2 min 10 min 30 min 1 h 2 h 4 h 6 h 8 h 2 min 10 min 30 min 1 h 2 h 4 h 6 h 8 h 2 min 10 min 30 min 1 h 2 h 4 h 6 h 8 h 2 min 10 min 30 min 1 h 2 h 4 h 6 h 8 h

Fig 4 Qualitative analysis of HP-CTNNB1 N7-mediated ligation of S-CTNNB1-2 with S-CTNNB1-3 The lower band represents the ATTO680-labeled ligation substrate, S-CTNNB1-2, and the upper band represents the ATTO680-ATTO680-labeled ligation product.

Table 2 Kinetic parameters of HP-CTNNB1 N7-mediated ligation of S-CTNNB1-2 with S-CTNNB1-3 and S-CTNNB1-3 U)2C under single turn-over conditions Amplitude values refer to maximal product yield obtained with 50-fold excess of ribozyme and 5¢-ligation substrate turn-over 3¢-ligation substrate For ribozyme sequences, see Fig 1 ND, not determined.

Ribozyme

kapp,lig(min)1) Km(n M ) Amplitude kapp,lig(min)1) Km(n M ) Amplitude

a kobs,ligat ribozyme saturation.

reaction is negligible The slightly higher Km value

(380 ± 39 nm) may be a result of inactive ribozymes

in the solution [45]

Next, ligation activities of the four variants were

investigated on CTNNB1 substrates with cytosine

instead of uracil at position)2 (S-CTNNB1-3U)2C)

As observed for the cleavage reaction, ligation was

also considerably improved by this substitution: rate

constants and product yields were increased for all

four variants (Table 2) All variants showed ligation

activity, with maximal product yields of 65% (U7),

30% (A7 and C7) and 20% (G7) (Fig 6A) Although

the U7 and A7 variants catalyzed ligation with

differ-ent amplitudes, the rate constants were similar

(1.5 ± 0.08 and 1.3 ± 0.08 min)1, respectively)

Liga-tion by the C7 and G7 variants was less efficient, with

kapp,lig values of 0.68 ± 0.05 and 0.26 ± 0.01 min)1 (Fig 6B), respectively

Interestingly, ligation data for the wild type and the U7 variant were better fitted with a double exponential than with a single exponential equation, whereas the other variants showed monophasic kinetics, as expected Biphasic kinetics with a fast and slow phase were previ-ously described for minimal hairpin ribozymes [46], with the assumption that the slow phase results from inactive ribozymes that have to undergo structural rearrange-ment prior to cleavage⁄ ligation In investigations of ribozyme–substrate complexes in native polyacrylamide gels (data not shown), all ribozyme–substrate complexes showed a similar band pattern, indicating that there are

no significant differences in global folding Further experiments addressing this question might include

0.0

0.1

0.2

0.3

0.4

0.5

kobs

Ribozyme (nM)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

A

B

Reaction time (min)

Fig 5 HP-CTNNB1 U7-mediated ligation of CTNNB1-2 with

S-CTNNB1-3 under single turnover conditions (A) Time course of the

reaction at 30-fold excess of ribozyme and CTNNB1-3 over

S-CTNNB1-2 (B) Dependence of kobsvalues on ribozyme

concentra-tion.

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

kobs

Ribozyme (nM)

A7 C7 G7 U7

Reaction time (min)

0.0 0.1 0.2 0.3 0.4 0.5 0.6

0.7

A

B

Fig 6 Ligation of S-CTNNB1-2 with S-CTNNB1-3 U)2C catalyzed

by HP-CTNNB1 N7 under single turnover conditions (A) Time course of the reaction at 50-fold excess of ribozyme and S-CTNNB1-3 U)2C over S-CTNNB1-2 (B) Dependence of kobs val-ues on ribozyme concentration.

time-resolved folding analysis of individual hairpin

ribozyme variants to look for differences in the folding

kinetics For the purpose of the study presented here,

the ligation rate constant was assigned to the fast phase

of the reaction, being 0.01 ± 0.004 min)1for the wild

type, and 1.0 ± 0.07 min)1for ligation of S-CTNNB1-3

by the U7 variant and 1.5 ± 0.08 min)1for CTNNB1-3

U)2C (Table 2)

Discussion

We have developed hairpin ribozyme variants targeting

UG*GUAA sites on suitable RNA substrates The

UG*GUAA site was chosen in the context of the

development of a hairpin ribozyme for downregulation

of CTNNB1 mRNA, encoding b-catenin, which is an

essential player in the Wnt signaling pathway, and

which, if present at high cellular concentrations, may

support cancer development and progression [8,9]

We have analyzed the cleavage and ligation properties

of a variety of hairpin ribozymes targeting short model

substrates derived from CTNNB1 mRNA In

particu-lar, we searched for compensatory mutations in the

ri-bozyme part that are able to counterbalance the effects

of nucleobase substitutions in the substrate, with the

major focus on analysis of C+3fi A The other two

changes were assumed to be less detrimental: U)2 is

still within the frame of the consensus sequence, and

C+4fi A should be easily compensated for by

replace-ment of G6 with uracil in the ribozyme strand,

restor-ing a Watson–Crick interaction in helix 1 [25]

As known from the crystal structure, C+3 in the

sub-strate interacts with A7 in the ribozyme strand

There-fore, we speculated whether substitution of A7 would

compensate for C+3fi A Four hairpin ribozymes

car-rying any of the four bases at position 7 have been

prepared (N7 ribozymes) and studied in cleavage and

ligation assays

A wild-type hairpin ribozyme that recognizes the

CTNNB1 substrate showed 60-fold lower cleavage

activity and 100-fold lower ligation activity than with

the wild-type A*GUC substrate [35,41] A+4 in the

substrate is well tolerated if the ribozyme contains a

uracil at position 6, restoring a Watson–Crick base

pair This result is somewhat surprising, as an A–U

base pair at this position never seemed to have

emerged from in vitro selection experiments [25–30],

such that the nucleobase at position +4 was included

in the consensus sequence as B = C, G or U, but not

A A substrate with A+3 in addition to A+4 was

accepted by all variants, with the U7 variant being the

most active Interestingly, and in contrast to what we

had expected, C)2fi U showed the strongest effect on

activity The U)2substrate was cleaved by all variants Activity decreased in the order U7 > A7 = C7 > G7 The cleavage rate constant of the U7 variant, however, was still reduced 10-fold as com-pared with the C)2 substrate, which was cleaved more rapidly by all four variants, although in the same activity order: U7 > A7 = C7 > G7 The effect of

U)2 on ligation was even more pronounced: whereas ligation of the C)2 substrate was observed with all variants, with U7 = A7 > C7 > G7, apart from the wild-type ribozyme, only the U7 variant could ligate the U)2substrate

These results expand hairpin ribozyme consensus rules in different ways On the basis of previous stud-ies, it was concluded that the hairpin ribozyme accepts any base except adenine at position +4 [25] The results presented here indicate that A+4 is toler-ated without loss of activity, if the complementary base is located at position 6 in the ribozyme strand, allowing the essential Watson–Crick base pair to be formed Furthermore, as previously shown by Ander-son et al [26], the C+3fi A substitution almost com-pletely abolished the cleavage activity of the wild-type hairpin ribozyme We did not observe such a strong effect of C+3fi A on the ribozymes tested here, in good agreement with the cleavage of G*GUA sub-strates reported by Pe´rez-Ruiz et al [32] Apart from the different substrate sequence, our A7 variant corre-sponds to the sequence of the wild-type hairpin ribo-zyme, with the only difference at position 6 being uracil instead of guanine (Fig 1) Apparently, the

G6fi U substitution, together with the altered sub-strate sequence, not only compensates for the replace-ment of C+4 with adenine, but also neutralizes the change from C+3 to adenine A possible explanation for this observation may be found in the spatial situa-tion around the A+3⁄ A+4 site In the wild-type hairpin ribozyme complexed to its A)1*G+1U+2C+3C+4 substrate, the C+4–G6 Watson–Crick base pair stacks upon a noncanonical base pair formed between C+3 and A7, in which, according to the crystal structure, the excocyclic amino group of C+3 donates a proton

to ring nitrogen N1 of A7 [20,36–40] In the A7 ribo-zyme, the Watson–Crick base pair is formed between

A+4 and U6 followed by the noncanonical base pair

A+3–A7 Adenine provides a similar Watson–Crick edge as cytosine, and the function of the exocyclic amino group of C+3 as a hydrogen donor can be basically retained by adenine The larger nucleobase may be tolerated because of the different nature of the neighboring Watson–Crick base pair, which is A–U instead of G–C An A–U base pair is less stable than G–C, and thus might allow the neighboring A+3 to

squeeze in the site originally harboring a cytosine This

interpretation is further supported by the observation

that the U7 variant increased the cleavage rate by

another factor of two, and was the only ribozyme

among the variants with ligation activity The spatial

situation around the A+3⁄ A+4 site becomes more

relaxed in the U7 variant, because A+3 now interacts

with the smaller uracil instead of adenine U7 still

pro-vides two hydrogen acceptor sites, and thus allows the

noncanonical hydrogen bond with A+3 to be formed

Taken together, as compared with the wild-type

hair-pin ribozyme–substrate complex, the situation has

changed from pyrimidine+4–purine6 (Watson–Crick)

and pyrimidine+3–purine7(noncanonical) to purine+4–

pyrimidine6 (Watson–Crick) and purine+3–pyrimidine7

(noncanonical) in the U7 variant This may be well

tolerated, owing to the ability of the modified base

pairs to provide the required base-pairing interactions

with similar spatial characteristics

The most significant effect was observed upon

replacement of C)2with uracil Both cleavage and

liga-tion activities suffered from this substituliga-tion, probably

because of the emerging wobble base pair between U)2

and G11 closing helix 2 This G–U wobble base pair,

located next to loop A, is presumably less capable of

stabilizing the required loop A conformation than the

regular Watson–Crick base pair that normally occurs

at this site, hampering active site chemistry The

obvi-ous compensatory mutation of G11to adenine in order

to restore the Watson–Crick base pair at this position

was shown to be unable to rescue ribozyme activity

[25] This is not surprising, as G11is involved in

forma-tion of the ribose zipper connecting the loop A domain

with the loop B domain The destabilization brought

about by the G–U wobble pair influences ligation more

strongly than cleavage (the A7, C7, G7 and U7 variants

were cleavage active, but only the U7 variant showed

ligation activity), because an even more rigid

conforma-tion is required for ligaconforma-tion This interpretaconforma-tion is given

further support by the recent finding that nucleobase

substitutions that exhibit significant levels of

interfer-ence with tertiary folding and interdomain docking

have relatively large inhibitory effects on ligation rates

while showing little inhibition of cleavage [47]

In conclusion, these results demonstrate that hairpin

ribozymes can be designed for cleavage of sites differing

from the consensus sequence, and thus extend previous

results on hairpin ribozyme specificity [32] Moreover,

the study shows that, on the basis of careful analysis of

the available structural data, rational design can be a

straightforward and effective strategy for the

develop-ment of catalysts with changed specificity As compared

with a full in vitro selection experiment, our rational

design study delivered functional ribozymes with less time, material and costs Moreover, our results demon-strate that several changes in the subdemon-strate sequence can

be advantageous over just one base substitution, owing

to the cooperative effect of two or more base changes The discrepancy between cleavage and ligation activities observed for the A7, C7 and G7 variants is a useful property with regard to the use of ribozymes for mRNA knockdown Here, only cleavage is required, and liga-tion activity is undesirable Altogether, the results of our study enlarge the window for application of tailor-made ribozymes in molecular biology and medicine

Experimental procedures

Substrate preparation

All substrates used for cleavage and ligation analysis were chemically synthesized on a solid phase as described previ-ously [48], with the use of phenoxyacetyl-protected phos-phoamidites (ChemGenes, Wilmington MA, USA) and a Gene Assembler Special synthesizer (Pharmacia, Freiburg, Germany) For postsynthetic labeling of the 20mer cleavage substrate and the 3¢-ligation substrate, 3¢-Amino Modifier C-3 lcaa CPG (ChemGenes) was used as the solid phase Phenoxyacetyl and cyanoethyl protecting groups were removed with a 1 : 1 mixture of 32% ammonia and 8 m methylamine in ethanol at 65C for 30 min, followed by lyophilization Tert-butyldimethylsilyl protecting groups were removed for 1.5 h at 55C in a 3 : 1 mixture of trieth-ylamine trihydrofluoride and dimethylformamide, and the reaction was stopped with 25% (v⁄ v) water RNA was pre-cipitated with butanol and purified by PAGE on a 15 % denaturating polyacrylamide gel Substrates were obtained

by elution from the gel with 0.3 m sodium acetate (pH 7.0), followed by ethanol precipitation

For postsynthetic labeling, 10 nmol of amino-modified oligonucleotide in 50 lL of 0.2 m sodium bicarbonate (pH 8.0) was mixed with 100 lg of ATTO680-NHS ester (ATTOTEC, Siegen, Germany) in 50 lL of dimethylforma-mide The reaction was performed for 3 h at room tempera-ture After ethanol precipitation, labeled oligonucleotides were purified by RP-HPLC on a Vario Prep 250-10 Nucleo-dur 100-5 C18 EC column (Macherey-Nagel, Du¨ren, Germany), with 0.1 m tetraethylammonium acetate (pH 7.5) and an acetonitrile gradient from 5% to 30% Product frac-tions were concentrated and desalted over NAP columns

Generating RNA fragments with 2¢,3¢-cyclic phosphate termini

The 16mer 5¢-ligation fragments were obtained by DNAzyme 8-17-mediated (5¢-AAG AGG ATT CCA GCG GAT CGA AAC TCA GAG AAG GAG C-3¢; Purimex,

Grebenstein, Germany) cleavage of a chemically synthesized

25mer substrate fragment, S-FR-CTNNB1 (5¢-GCU CCU

UCU CUG AGU GGU CCU CUU U-3¢), containing the

16mer S-CTNNB1-3 fragment (underlined) Complementary

parts between DNAzyme and substrate are in italics To

generate the ligation fragments S-CTNNB1-3 and

S-CTNNB1-3U)2C with 2¢,3¢-cyclic phosphate termini, RNA

substrate, DNAzyme and Tris (pH 7.5) were mixed to give a

final concentration of 2 lm RNA, 400 nm DNAzyme and

40 mm Tris, heated for 2 min at 95C, and incubated for

15 min at 37C After addition of magnesium chloride to a

final concentration of 90 mm, the reaction was performed

for 2 h at 37C RNA was purified on a 10% denaturating

polyacrylamide gel, eluted from the gel with 0.3 m sodium

acetate (pH 7.0) at 4C, and precipitated with ethanol

Ribozyme synthesis

Hairpin ribozymes were transcribed in vitro from a dsDNA

template The DNA template was obtained from a Klenow

polymerase-mediated fill-in reaction of two synthetic

prim-ers (5¢- CTG TAC TAA TAC GAC TCA CTA TAG GGA

GAT GCC TTN GAA GCT CAG CTG AGA AAC ACG

AAT C-3¢ and 5¢-GCT CCT TCT CTG GGT AGC TGG

TAA TAT ACC GAA TGC GAA GAT TCG TGT TTC

TCA GCT GAG C-3¢; biomers.net, Ulm, Germany)

over-lapping at their 3¢-ends by 22 nucleotides (underlined) Both

primers and 10· KFI buffer (500 mm Tris, pH 7.6,

100 mm MgCl2and 500 mm NaCl) were mixed to give final

concentrations of 2 lm each primer and 1· KFI buffer,

heated for 2 min at 90C, and incubated for 15 min at

37C After addition of dNTPs (Fermentas, St Leon-Rot,

Germany) to a final concentration of 500 lm and Klenow

fragment (Fermentas) to a final concentration of

0.05 U lL)1, the reaction was performed for 30 min at

37C DNA was purified on a 10% native polyacrylamide

gel, eluted from the gel with 0.3 m sodium acetate (pH 7.0),

and precipitated with ethanol Transcription was performed

in a reaction mixture with final concentrations of 1 lm

DNA template, 1· transcription buffer (Fermentas), 2 mm

each NTP (Fermentas) and 0.6 U lL)1 T7-RNA

polymer-ase (Fermentas) for 3 h at 37C After phenol ⁄ chloroform

extraction, RNA was precipitated with ethanol, purified on

a 10% denaturating polyacrylamide gel, and eluted from

the gel with 0.3 m sodium acetate (pH 7.0) Salt was

removed by ethanol precipitation

Cleavage kinetics under single turnover

conditions

For kinetic characterization of cleavage events, reactions

were carried out in 40-lL reaction volumes with final

con-centrations of 25 nm substrate, 50–750 nm ribozyme,

40 mm Tris (pH 7.5) and 10 mm MgCl2 Substrate and

ribozyme were mixed separately in a 20-lL volume in Tris

and MgCl2 respectively, denaturated at 90C for 2 min, and incubated for a further 15 min at 37C Reactions were started by mixing substrate and ribozyme solutions

At suitable time intervals, aliquots of 1 lL were taken, and the reaction was immediately stopped by addition of 19 lL

of stop mix (7 m urea and 50 mm Na-EDTA) Samples were stored on ice before analysis All reactions were repeated at least twice Samples were analyzed on a 15% poly-acrylamide gel with a DNA Sequencer Long ReadIR 4200 (LI-COR Bioscience Bad Homburg, Germany); data were processed with gene imagir 4.05 The fraction of substrate cleaved was plotted versus time, and fitted to the single exponential equation

½30P ¼ Að1  ektÞ where [3¢P] is the product concentration, A is the ampli-tude, k = kobs,cleave, and t is the time Standard deviations were less than 20% in each case To determine the enzyme specific constants, kcleav and Km, the obtained kobs,cleav

values were plotted versus ribozyme concentration [R]0and the curve was fitted to the following equation:

kobs;cleav¼ kcleav½R0

Kmþ ½R0

Ligation kinetics under single turnover conditions

Ligation reactions were carried out in 40-lL reaction vol-umes with final concentrations of 10 nm 3¢-ligation sub-strate, 20–500 nm 5¢-ligation substrate, 20–500 nm ribozyme, 40 mm Tris (pH 7.5) and 10 mm MgCl2 First, ribozyme was denaturated in Tris buffer for 2 min at

90C, and this was followed by incubation for 15 min at

37C After addition of MgCl2, the solution was incubated for another 15 min at 37C The 5¢-ligation substrate was then added, and the solution was incubated again for

15 min at 37C The reaction was started by addition of the 3¢-ligation substrate After suitable periods of time, aliquots of 1.5 lL were taken and immediately added to 8.5 lL of stop mix, and samples were stored on ice until analysis Analysis of the ligation reaction was performed on

a DNA sequencer as described for cleavage reactions The fraction of ligation product was plotted versus time and fit-ted to single-exponential or double-exponential equations The single-exponential equation was:

½P ¼ Að1  ekobs;ligtÞ where [P] is the product concentration, A is the amplitude, and t is the time The double-exponential equation was:

½P ¼ A0þ A1ð1  ek1 tÞ þ A2ð1  ek2 tÞ

where A1 and A2 are the amplitudes of the biphasic time course and A0is the starting signal; k1and k2represent the corresponding ligation rates of the fast phase and the slow