Nucleic Acids Research

Construction of assay plasmids and expression plasmid libraries

Assay plasmids were constructed as previously described (28). Synthetic duplexes containing O₂₁, O_21S, O_21RC, O₁₄ and O₁₂ were assembled by annealing equimolar quantities of synthetic oligonucleotides in the presence of 250 mM NaCl. Duplexes were ligated between the KpnI and PstI sites of pNN396 (26) to yield plasmids containing the desired promoter/operator constructs. Promoter/operator constructs were confirmed by sequencing. The small HindIII-NotI fragments from these plasmids were isolated and ligated between the HindIII and NotI sites of pNN388 (26) to yield single-copy assay plasmids, or into pNN387 (26) to yield single-copy lacZ constructs. The assay plasmid containing O_lac was generated by removing a small HindIII fragment from pNN390 (26). Restriction digests, ligations and transformations using chemically competent cells were performed using standard procedures (29). Oligonucleotides were synthesized by phosphoramidite methodology using an ABI model 394 DNA synthesizer. A detailed description of plasmid vectors is available from the authors upon request.

DNA templates encoding RNA pools C, 7.4C, 7.2R and 7.4R were previously designed and/or selected in vitro from random sequence libraries (27). For expression in E.coli, DNA templates were first digested with HindIII and EcoRI. Digestion products were dephosphorylated using calf intestinal alkaline phosphatase (Boehringer Mannheim), and ligated between the HindIII and EcoRI sites of the RNA expression vector pJDC408 (30,31) to produce expression plasmid libraries L1, L2, L3 and L4 (Table 1). Ligation reactions (200 µl) contained ~2 µg HindIII-EcoRI-digested DNA pool, ~10 µg HindIII-EcoRI-digested pJDC408, 50 mM Tris-HCl pH 7.8, 10 mM MgCl₂, 10 mM dithiothreitol, 1 mM ATP, 25 µg/ml bovine serum albumin and 800 U T4 DNA ligase (New England Biolabs), and were incubated overnight at 16°C. Ligation products were digested with BamHI to linearize clones lacking insert. One-fourth of the products from each ligation reaction were used to transform 400 µl aliquots of electrocompetent DH5[alpha] cells (32) by electrotransformation using a Gene Pulser apparatus (Bio-Rad). Cells were electrotransformed in 0.2 cm cuvettes using a 2.5 kV pulse. The resistance and capacitance of the pulse circuit were set to 200 [Omega] and 25 µF, respectively. Following 1 h of non-selective growth in 4 ml yeast extract and nutrient broth (YENB) medium (32), transformants were grown overnight at 37°C in an incubator shaker (250 r.p.m.) in 500 ml Luria-Bertani (LB) medium containing 50 µg/ml carbenicillin (Cb). Plasmid libraries were purified from cultured cells using a modified alkaline lysis method followed by QIAFILTER and QIAGEN resin as suggested by the manufacturer (QIAGEN). Plasmid libraries were redissolved in water, and DNA concentrations were determined assuming that one absorbance unit at 260 nm corresponds to a concentration of 50 µg DNA/ml.

Table 1. Origin of expression plasmid libraries

Plasmid library	Source
	Poola	Typeb	Roundsc	pHd
L1	initial pool C	C	0	-
L2	7.4C	C	11	7.4
L3	7.2R	R	20	7.2
L4	7.4R	R	26	7.4

^aDNA pools encoding RNAs previously selected in vitro for binding to the duplex DNA target sequence (O₂₁) via triple helix formation (27).
^bType C pools encode RNAs that contain the O₂₁-specific 21 nt pyrimidine recognition sequence adjacent to a 42 nt random region. Type R pools contain pyrimidine recognition sequences that arose from a 60 nt random region through in vitro selection.
^cNumber of rounds of in vitro selection and amplification.
^dpH of final in vitro selection buffer.

In vivo selections

For the first round of in vivo selection, 400 µl aliquots of electrocompetent DH5[alpha] cells carrying the assay plasmid containing O₂₁ were electrotransformed with 10 µg of expression plasmid library L1, L2, L3 or L4. Following 90 min of non-selective growth in 1.8 ml LB at 37°C in an incubator shaker (300 r.p.m.), transformants were grown for 2 h in the presence of 50 µg/ml Cb, 40 µg/ml chloramphenicol (Cm) and 0.1 mM isopropyl [beta]-d-thiogalactopyranoside (IPTG). Cells from each transformation were collected by centrifugation for 10 min at 4000 g and plated on four LB agar plates containing 50 µg/ml Cb, 40 µg/ml Cm, 1 mM IPTG and 80 µg/ml Sp. After growth for ~48 h at 37°C, Sp^r cells from each library were harvested and pooled. Plasmids were isolated from pooled cells (QIAGEN Mini Prep) and expression plasmids were separated from assay plasmids by agarose gel electrophoresis. For subsequent rounds of in vivo selection, 400 ng of gel-purified expression plasmid library were used for each electrotransformation, and cells were plated on three agar plates containing selective media. Parallel transformations were also performed using 400 ng of pJDC408 (expression plasmid lacking insert) to determine the frequency of Sp^r colonies arising from spontaneous mutations in the assay plasmid.

Two representative Sp^r clones were selected from each library following the final round of selection. Clones 1-2 and 3-4 were isolated from L1 and L2, respectively, after two rounds of selection. Clones 5-6 were isolated from L3 after three rounds of selection. EP1 and EP2 were isolated from two Sp^r colonies arising from a transformation of pJDC408. Expression plasmids were purified from each clone.

Expression plasmids were used to transform chemically competent DH5[alpha] cells harboring assay plasmids containing either O₂₁ or O_lac. Several colonies representing each transformant were diluted in LB media. Dilutions were used to streak transformants onto sectors of LB agar plates containing 50 µg/ml Cb, 40 µg/ml Cm and 1 mM IPTG in the absence or presence of 80 µg/ml Sp. Plates lacking Sp were incubated for ~24 h at 37°C. Plates containing Sp were incubated for ~48 h at 37°C.

Sequence analyses

Selected expression plasmids were sequenced using an ABI prism model 377 DNA sequencer. Secondary structures of RNAs encoded by expression plasmids were predicted using the program Mulfold (33-35) running on an Apple Power Macintosh microcomputer.

The stabilities of putative sense:antisense RNA interactions were estimated by modeling them as intramolecular interactions using Mulfold. For this purpose, the sense and antisense sequences were appended to the 3[prime] and 5[prime] ends of a stable RNA hairpin. The free energies of intermolecular complex formation for predicted sense:antisense interactions were calculated according to published parameters (36).

Construction of mutants

The pyrimidine recognition sequence of RNA T21 was scrambled or deleted to create mutant RNAs, T21S and T21[Delta], respectively. Expression plasmids encoding T21S and T21[Delta] were constructed by insertion of a synthetic DNA duplex encoding the mutations between the EcoRI and HindIII sites of pJDC408. The sequences of the resulting expression plasmids were confirmed. Expression vector pJDC408 and expression plasmids encoding T21, T21S and T21[Delta] were used to transform chemically competent DH5[alpha] cells carrying assay plasmid containing O₂₁ or O_21S. Several colonies representing each transformant were diluted in LB media. Dilutions were used to streak transformants onto sectors of LB agar plates containing 50 µg/ml Cb, 40 µg/ml Cm and 1 mM IPTG in the absence or presence of 20 µg/ml Sp. Plates were incubated for 24 h at 37°C.

Expression plasmids that encode other T21 mutants in which the pyrimidine recognition sequence was reversed (A21), truncated (T14 and T12), or truncated and reversed (A14 and A12) were constructed and assayed in a similar manner.

[beta]-Galactosidase assays

[beta]-galactosidase assays were performed as previously described (37). Briefly, cultures of DH5[alpha] cells harboring expression and assay plasmid combinations were grown to saturation in LB media containing 50 µg/ml Cb and 40 µg/ml Cm. Subcultures were prepared by combining 100 µl of saturated culture with 5 ml of LB media containing 50 µg/ml Cb, 40 µg/ml Cm and 1 mM IPTG. Subcultures were grown by continuous aeration at 37°C to a density of 2-5 × 10⁸ cells/ml (absorbance at 600 nm in the range 0.28-0.7). Each subculture was placed on ice, and the absorbance at 600 nm was determined. A 50 µl aliquot of each subculture was combined with 950 µl Z buffer (100 mM sodium phosphate pH 7.0, 10 mM KCl, 1 mM MgSO₄, 50 mM 2-mercaptoethanol). Cells were lysed by addition of 50 µl chloroform and 25 µl of 0.1% SDS followed by vigorous agitation. Enzyme assays were initiated by addition of 200 µl Z buffer containing 4 mg/ml o-nitrophenyl [beta]-d-galactopyranoside (ONPG), incubated at 28°C for 7-8 min, and terminated by addition of 500 µl of 1 M Na₂CO₃. Cell debris was removed from each reaction by centrifugation, and the absorbances at 550 and 420 nm were determined. The units of [beta]-galactosidase activity present in each sample were calculated using the equation: where t is the reaction time (min) and v is the volume of subculture assayed (ml). A₆₀₀, A₅₅₀ and A₄₂₀ relate to the cell density of the subculture assayed, light scattering due to cell debris, and the concentration of the reaction product (o-nitrophenol), respectively.

In vivo selection to identify potential repressor RNAs

RNA transcripts were selected in vivo on the basis of their ability to confer spectinomycin resistance to E.coli cells in a transcription interference assay (Fig. 1A) (25,26). The assay plasmid used in these selections included an operator element (O₂₁) containing a homopurine:homopyrimidine sequence for triple helix formation in the pyrimidine motif (Fig. 1B). Four expression plasmid libraries were constructed for in vivo selections (Table 1). Each library contained DNA sequences that encode RNAs that were previously designed and/or selected in vitro from random sequence pools for binding to the pyrimidine motif triplex target present in O₂₁ (27). Thus, each library encodes RNA transcripts that present pyrimidine recognition sequences previously shown to be capable of binding O₂₁ by forming triple helices.

Figure 1. In vivo genetic selection to identify RNAs that relieve transcriptional interference in E.coli. (A) Schematic diagram illustrating one mechanism by which RNAs might confer survival to E.coli cells in the presence of Sp. An expression plasmid library encoding potential repressor RNAs is transformed into E.coli cells harboring the assay plasmid. The aadA gene encodes resistance to Sp. Cells that carry the assay plasmid alone are Sp^s due to a transcriptional interference effect thought to involve physical collisions of transcribing polymerases that initiate from opposing sense and antisense promoters flanking the aadA gene (26). Agents that repress the antisense aadA promoter may therefore confer a Sp^r phenotype. An operator element is inserted adjacent to the antisense aadA promoter. Libraries of potential repressor RNAs are examined by in vivo selection for Sp^r cells. (B) Partial sequences of two antisense aadA promoter/operator regions. Assay plasmids contained either a 21 nt operator sequence for pyrimidine motif triplex formation (O₂₁), or the lac operator (O_lac) adjacent to the antisense aadA promoter. The transcription start site (+1) and the -10 and -35 promoter regions (underlined) are indicated.

RNA transcripts were selected from libraries L1, L2, L3 and L4 through two to three rounds of in vivo selection (Table 2). As expected, it was observed that the number of Sp^r transformants per microgram of expression plasmid library increased rapidly during the selection process. Libraries L1 and L2 were subjected to only two rounds of in vivo selection while L3 was subjected to three rounds. After two rounds of selection, L4 failed to yield Sp^r colonies at a frequency exceeding that of the expression vector lacking randomized insert. Interestingly, the sequence diversity of the pool from which L4 was derived had previously been shown to be much lower than for the pools used to produce L1, L2 and L3 (27). Because RNAs encoded by L4 had been subjected to extensive in vitro selection and amplification, we concluded that in vitro optimization of triple helix formation did not ensure the ability to confer Sp^r in vivo. Thus, pools with greater sequence diversity offered more solutions for overcoming transcriptional interference in vivo.

Table 2. Summary of in vivo selections

Plasmid library	Spr transformants per µg DNA
	Round 1	Round 2	Round 3
L1	450	>104	-a
L2	640	>104	-
L3	110	7750	>104
L4	100	90	-

^aA third round of in vivo selection was not performed.

Two clones from each library were isolated for further analysis after the final round of selection. Clones 1-2, 3-4 and 5-6 were selected from libraries L1, L2 and L3, respectively. In addition, clones EP1 and EP2 were isolated from two Sp^r colonies arising from a transformation of expression plasmid lacking randomized insert. To determine whether the Sp^r phenotype conferred by each RNA was operator-specific, expression plasmids were purified from each clone and used to transform cells harboring assay plasmids containing either O₂₁ or O_lac. Cells containing EP1 and EP2 failed to confer a Sp^r phenotype with either assay plasmid, suggesting that their initial Sp^r phenotype was due to mutation of the original assay plasmid. In contrast, cells containing plasmids 1-6expressed a Sp^r phenotype only in the context of the assay plasmid containing O₂₁ (data not shown). These data demonstrate that RNAs expressed from the selected expression plasmid clones require the presence of O₂₁ in the assay plasmid to confer Sp^r in E.coli.

The sequences of expression plasmids 1-6 were determined. Plasmids 1 and 2 from L1 were identical and encode an RNA henceforth termed T21. The DNA pool used to construct L1 encoded RNAs that contain a 21 nt pyrimidine recognition sequence adjacent to a 42 nt region of random sequence (27). Thus, T21 contains a 63 nt region derived from library sequence flanked 5[prime] and 3[prime] by sequences that are derived from the lpp promoter and terminator sequences of the expression cassette. The predicted lowest energy secondary structure of the 266 nt transcript suggests that the pyrimidine recognition sequence of T21 is positioned in the loop of a stem-loop structure. This arrangement may enable presentation of the pyrimidine recognition sequence as a single-stranded domain within the folded RNA transcript. Similar presentation of pyrimidine recognition sequences has been predicted for related RNA transcripts that have been demonstrated to form triple helices with a duplex DNA target in vitro (27). Thus, it was plausible that T21 conferred Sp^r by a mechanism in which triple helix formation with O₂₁ mediates transcriptional repression. To test this hypothesis, a number of T21 mutants and operator variants were constructed.

Characterization of mutants

To examine the sequence specificity of the putative triplex interaction between T21 and O₂₁, mutations were made in both the pyrimidine recognition sequence of T21 and in the O₂₁ operator sequence (Fig. 2A). The pyrimidine recognition sequence of T21 was either scrambled or deleted to generate mutants T21S or T21[Delta], respectively. The global secondary structures predicted for the mutant RNAs did not differ from that predicted for T21. An assay plasmid was created bearing a scrambled operator sequence, O_21S, designed to accommodate a triple helix with T21S, but not with T21 (Fig. 2A). Expression plasmids encoding T21, T21S or T21[Delta] were used to transform cells carrying assay plasmids containing either O₂₁ or O_21S. Survival in the absence or presence of Sp was examined (Fig. 2B). Cells transformed with expression plasmids lacking randomized insert (EP) or encoding T21[Delta] exhibited a Sp^s phenotype in the context of assay plasmids containing either operator sequence. Consistent with sequence-specific triple helix formation, a Sp^r phenotype was observed only for O₂₁-T21, and O_21S-T21S assay-expression plasmid combinations.

Figure 2. RNA and operator mutagenesis studies. (A) Schematic diagram showing partial sequences of operators and inhibitor RNAs. T21S represents a T21 mutant in which the pyrimidine recognition sequence was scrambled to destabilize triplex formation at O21. Operator mutant O21S contains a homopurine:homopyrimidine sequence that was altered to restore the potential for triplex formation with T21S. (B) Growth of cells harboring the indicated combinations of assay and expression plasmids in the absence or presence of 20 µg/ml Sp. Cells carry expression plasmids lacking randomized insert (EP) or encoding T21, T21S or T21[Delta] as indicated, together with assay plasmids containing O21 (top) or O21S (bottom).

While these data show that the pyrimidine recognition sequences within the putative repressor RNAs exhibit operator specificities consistent with triple helix formation, we considered other possibilities involving sense:antisense interactions (Fig. 3). Besides binding to duplex DNA (Fig. 3A), the pyrimidine recognition sequence of an inhibitor RNA might partially hybridize to the purine-rich DNA strand of an operator via strand displacement (Fig. 3B), or to RNA transcribed from that operator (Fig. 3C). Inhibitor RNA binding by partial Watson-Crick hybridization to the transiently unpaired purine-rich DNA strand of an operator might inhibit transcription initiation, thus conferring the Sp^r phenotype observed in the transcription interference assay. An alternative possibility involves partial Watson-Crick hybridization of the inhibitor RNA to the purine-rich mRNA resulting from transcription across the operator. Such sense:antisense interactions might cause transcription termination or alter RNA metabolism. An antisense mechanism involving transcription attenuation has been proposed to occur in the natural regulation of the crp gene in E.coli (38). Apart from attenuation of transcription, it is difficult to understand how a short sense:antisense interaction at the 5[prime] end of an RNA transcript might relieve transcriptional interference.

Figure 3. RNA strand polarities in triplex versus antisense interactions. In the formation of a pyrimidine motif triple helical complex (A), the pyrimidine recognition sequence of the inhibitor RNA binds parallel to the homopurine strand of the operator. A sense:antisense interaction involving the pyrimidine recognition sequence of the inhibitor RNA and the homopurine sequence of the operator (B) or the RNA transcribed from the operator (C) occurs through antiparallel Watson-Crick hybridization.

A fundamental difference between the formation of a triple helix and a conventional Watson-Crick hybrid is nucleic acid strand polarity (Fig. 3). In the triple-helical complex, the pyrimidine recognition sequence of a repressor RNA binds parallel to the homopurine strand of a duplex DNA operator. In a sense:antisense interaction, the pyrimidine recognition sequence of an inhibitor RNA binds antiparallel to the homopurine sequence of the RNA transcribed from an operator. Thus, repressor RNAs designed to form triple helices could, in principle, act as inhibitor RNAs by binding RNA message in the opposite orientation to form partial (mismatched) hybrids.

To investigate the stabilities of putative RNA-RNA hybrids between the pyrimidine recognition sequences of T21 and T21S and RNAs transcribed from operators O₂₁ and O_21S, the free energies of complex formation were calculated (Table 3). The first four entries in Table 3 show that the assay-expression plasmid combinations that were observed to confer Sp^r in the plating experiment corresponded to more stable (more negative [Delta]G) putative RNA-RNA hybrids (Fig. 2). Conversely, the assay-expression plasmid combinations that yielded Sp^s cells corresponded to less stable (less negative [Delta]G) putative RNA-RNA hybrids. It was therefore plausible that a sense:antisense interaction was somehow responsible for the Sp^r phenotype conferred by the selected RNAs.

Table 3. Calculated free energy changes for putative RNA-RNA interactions

^aInteractions between RNAs were predicted using the algorithm of Zuker et al. (33-35). In each hypothetical duplex, the upper strand corresponds to the RNA transcribed through the target operator, and the bottom strand corresponds to RNA from the expression plasmid.
^bPredicted free energy changes for RNA-RNA duplex formation at 37°C were calculated according to published parameters (36).
^cObserved spectinomycin sensitivity or resistance of cells carrying the corresponding combinations of assay and expression plasmids.

To resolve whether triple helix formation or a sense:antisense interaction was responsible for the Sp^r phenotype, several additional operator variants and T21 mutants were constructed (Fig. 4A). Truncated operator sequences 14 or 12 bp in length were created (termed O₁₄ and O₁₂, respectively). Due to their shorter lengths, only high-affinity interactions (i.e. few mismatches) would be expected to confer Sp^r phenotypes in cells carrying these operators. Five variant inhibitor RNAs were created by modifying T21. Mutants T14 and T12 contained appropriate truncated recognition sequences of 14 and 12 nt assuming a triplex mode of interaction with their cognate operators (Fig. 4A). Mutants A21, A14 and A12 contained appropriate 21, 14 and 12 nt pyrimidine recognition sequences assuming an antisense interaction with RNAs transcribed across their cognate operators (Fig. 4A). Using this approach, we were able to unambiguously determine the type of nucleic acid interaction responsible for Sp^r in the transcriptional interference assay. Sp^r phenotypes were observed only for O₂₁-T21, O₂₁-A21, O₁₄-A14 and O₁₂-A12 combinations (Fig. 4B). Notably, O₁₄-T14 and O₁₂-T12 combinations were Sp^s. These data clearly show that only RNA combinations capable of stable sense:antisense interactions confer Sp^r in this transcriptional interference assay.

Figure 4. Selected RNAs function through a sense:antisense interaction. (A) Sequences of operators and inhibitor RNAs. A21 is a T21 mutant in which the pyrimidine recognition sequence was reversed to maximize the stability of a sense:antisense interaction with RNA transcribed from O21. Operator variants O14 and O12 contain homopurine sequences 14 or 12 bp in length, respectively. T14 and T12 contain pyrimidine recognition sequences 14 and 12 nt in length, respectively, optimized for triplex formation with their cognate operators. A14 and A12 contain pyrimidine recognition sequences 14 and 12 nt in length, respectively, designed to hybridize to RNAs transcribed from their cognate operators by sense:antisense interactions. (B) Growth of cells harboring the indicated assay and expression plasmids in the absence or presence of 40 µg/ml Sp. Cells carrying expression plasmids lacking randomized insert (EP) are indicated.

The free energies of predicted sense:antisense RNA interactions were similarly calculated (Table 3). These data confirm the previous observation that more stable predicted sense:antisense interactions correspond to assay:expression plasmid combinations that confer Sp^r, while less stable predicted sense:antisense interactions correspond to assay:expression plasmid combinations that yield Sp^s cells. Furthermore, these data suggest that there is a threshold for this effect at any given spectinomycin concentration. Under the present conditions, a sense:antisense interaction with a predicted stability of -12.2 kcal/mol was insufficient to confer Sp^r, but an interaction whose predicted stability is -14.3 kcal/mol did confer Sp^r (Table 3). Thus, the transcriptional interference assay can display an exquisite sensitivity to the stability of a sense:antisense RNA interaction.

Inhibition of [beta]-galactosidase expression

Transcriptional interference is thought to involve collisions between converging RNA polymerases that are producing RNA transcripts (26). It is remarkable that a 12 bp RNA-RNA hybrid formed at the 5[prime] terminus of one of these transcripts appears to be capable of relieving transcriptional interference. We investigated whether the same short RNA-RNA interactions are capable of directly inhibiting expression of a target gene. The promoter/operator sequences containing O₂₁, O₁₄ and O₁₂ were placed ~1.3 kb upstream of the E.coli lacZ gene encoding [beta]-galactosidase (Fig. 5A). An additional operator variant termed O_21RC, containing the reverse complement sequence of O₂₁, was also examined (Fig. 5). Homopyrimidine-containing RNAs acting through a sense:antisense mechanism would be expected to have no effect on gene expression from a promoter/operator containing O_21RC because the operator-encoded portion of the mRNA contains a homopyrimidine sequence rather than a homopurine sequence. However, inhibition of lacZ expression from this construct would still be predicted in the case of triple-helical complex formation or if inhibitor RNA hybridized to the purine-rich DNA strand of the operator via strand displacement.

Figure 5. RNAs complementary to the 5[prime] terminus of lacZ mRNA inhibit [beta]-galactosidase expression. (A) Promoter/operators sequences containing O_21RC, O₂₁, O₁₄ or O₁₂ were placed ~1.3 kb upstream of the start codon of the lacZ gene. (B) Operator variant O_21RC contains the reverse complement of O₂₁. (C) [beta]-galactosidase activities measured in extracts from cells harboring the indicated expression and assay plasmids. For each lacZ construct, [beta]-galactosidase activity is expressed relative to cells harboring expression plasmids lacking randomized insert (EP). Results represent averages from at least six experiments.

The ability of various RNAs to inhibit [beta]-galactosidase expression from these promoter/operator constructs was examined in vivo (Fig. 5C). [beta]-galactosidase activities present in cells harboring assay-expression plasmid combinations were normalized to those in cells containing the same assay plasmid with expression vector lacking randomized insert (EP). As expected, [beta]-galactosidase activity was not significantly inhibited in cells containing O_21RC-A21 or O_21RC-T21 combinations (Fig. 5C). These data confirm that inhibitor RNAs do not act by triple helix formation or hybridization to the operator via strand displacement. In contrast, [beta]-galactosidase activity was inhibited 30-50% in cells containing O₂₁-T21, O₂₁-A21, O₁₄-A14 and O₁₂-A12 combinations (Fig. 5C). These results are consistent with data previously obtained in transcriptional interference assays, demonstrating that only inhibitor RNAs capable of relatively stable sense:antisense interactions with the 5[prime] terminus of the target mRNA inhibit gene expression. Furthermore, the data demonstrate that ~40% inhibition of gene expression may result in E.coli from the formation of a very short RNA-RNA interaction (as little as 12 bp in the O₁₂-A12 combination) at the 5[prime] terminus of an mRNA. Inhibition of lacZ expression is particularly notable because the RNA-RNA hybrid forms >1 kb upstream of the lacZ translation start codon. These data demonstrate that this short sense:antisense interaction inhibits gene expression in a manner independent of the transcriptional interference assay format.

Perspective

We have performed in vivo genetic selections based on a transcriptional interference assay and have obtained RNA transcripts that relieve transcriptional interference in E.coli. While the selected RNA transcripts were designed to form triple-helical complexes with a duplex DNA operator element to mediate transcriptional repression of one promoter, genetic and biochemical data show that the selected RNA transcripts relieve transcriptional interference through a mechanism involving a sense:antisense interaction. Transcriptional interference in E.coli is shown to be reduced if a short intermolecular RNA-RNA hybrid can be made to form at the 5[prime] terminus of the inhibitory transcript.

Additional biochemical studies will be needed to determine how a short RNA-RNA hybrid at the 5[prime] terminus of an RNA relieves transcriptional interference thought to be mediated by the RNA polymerase that produced it. In fact, at least three mechanisms may contribute to transcriptional interference. First, collisions between converging RNA polymerases may cause transcription termination. Second, transcription termination due to collisions between RNA polymerases might require the participation of accompanying ribosomes if open reading frames are present on both transcripts. Third, transcriptional interference may be facilitated by hybridization of the complementary RNA transcripts that originate from the same template (26).

In the context of these possible mechanisms, it remains difficult to understand how transcriptional interference is relieved by the formation of a short RNA-RNA hybrid at the 5[prime] terminus of a transcript. Perhaps the RNA-RNA hybrid induces transcriptional attenuation analogous to the scheme originally proposed for crp gene regulation (38). On the other hand, a short RNA-RNA hybrid may inhibit translation of the target RNA or affect its metabolism, either of which might impact transcriptional interference. Northern blot analyses may provide evidence for mechanisms that affect the amount of transcript present in cells. However, the mechanism by which a sense:antisense interaction affects transcriptional interference or gene expression may not alter mRNA levels. In fact, results of reverse transcription assays suggest that the level of [beta]-galactosidase mRNA is unaffected in cells where [beta]-galactosidase expression is inhibited by sense:antisense interaction (data not shown). Further studies will therefore be required to elucidate the precise mechanism by which an RNA-RNA hybrid mediates relief of transcriptional interference.

New strategies will also be needed to further explore the possibility of RNA-directed triple helix formation in E.coli. As described above, the selected T21 RNA did not repress lacZ expression when the O_21RC operator was placed upstream of the transcription start point. Because this operator placement should permit triple helix formation, while eliminating the potential for sense:antisense interactions, we conclude that the T21 transcript does not form a triplex with this DNA target in vivo. Other RNA transcripts might be capable of triplex formation at this operator, and it would be desirable to perform new selections in the context of the O_21RC operator (Fig. 5).These experiments are presently impossible, however, because (as has been observed in many other experimental operators in this system) the O_21RC operator itself weakens the anti-aadA promoter to an extent that transcriptional interference is not observed.

The data presented in this study reinforce the notion that claims of triplex-dependent inhibition of gene expression must be evaluated with caution.

ACKNOWLEDGEMENTS

We gratefully acknowledge S. Elledge for assay plasmids, M. Ferber for technical assistance, and the Mayo Molecular Core Facility for synthesis of oligonucleotides and DNA sequencing. This work was supported by the Mayo Foundation and NIH grants GM47814 and GM54411. L.J.M. is a Harold W. Siebens Research Scholar.

REFERENCES

1. Felsenfeld, G., Davies, D. R. and Rich, A. (1957) J. Am. Chem. Soc., 79, 2023-2024.

2. Britten, R. J. and Davidson, E. H. (1969) Science, 165, 349-357. MEDLINE Abstract

3. Miller, J. H. and Sobell, H. M. (1966) Biochemistry, 55, 1201-1205.

4. Chubb, J. M. and Hogan, M. E. (1992) Trends Biotechnol., 10, 132-136. MEDLINE Abstract

5. Cohen, J. S. and Hogan, M. E. (1994) Sci. Am., 271, 76-82. MEDLINE Abstract

6. Helene, C. (1991) Anticancer Drug Des., 6, 569-584. MEDLINE Abstract

7. Maher, L. J. (1996) Cancer Invest., 14, 66-82. MEDLINE Abstract

8. Moser, H. E. and Dervan, P. B. (1987) Science, 238, 645-650. MEDLINE Abstract

9. Le Doan, T., Perrouault, L., Praseuth, D., Habhoub, N., Decout, J.-L., Thuong, N. T., Lhomme, J. and Hélène, C. (1987) Nucleic Acids Res., 15, 7749-7760. MEDLINE Abstract

10. Beal, P. A. and Dervan, P. B. (1991) Science, 251, 1360-1363. MEDLINE Abstract

11. Semerad, C. L. and Maher, L. J. (1994) Nucleic Acids Res., 22, 5321-5325. MEDLINE Abstract

12. Maher, L. J., Wold, B. and Dervan, P. B. (1989) Science, 245, 725-730. MEDLINE Abstract

13. Maher, L. J., Dervan, P. B. and Wold, B. (1992) Biochemistry, 31, 70-81. MEDLINE Abstract

14. Maher, L. J. (1992) Biochemistry, 31, 7587-7594. MEDLINE Abstract

15. Skoog, J. U. and Maher, L. J. (1993) Nucleic Acids Res., 21, 2131-2138. MEDLINE Abstract

16. Ing, N. H., Beekman, J. M., Kessler, D. J., Murphy, M., Jayaraman, K., Zendegui, J. G., Hogan, M. E., O'Malley, B. W. and Tsai, M. (1993) Nucleic Acids Res., 21, 2789-2796. MEDLINE Abstract

17. Singleton, S. F. and Dervan, P. B. (1992) Biochemistry, 31, 10995-11003. MEDLINE Abstract

18. Olivas, W. M. and Maher, L. J. (1995) Biochemistry, 34, 278-284. MEDLINE Abstract

19. Cheng, A. and Van Dyke, M. W. (1993) Nucleic Acids Res., 21, 5630-5635. MEDLINE Abstract

20. Neckers, L., Whitesell, L., Rosolen, A. and Geselowitz, D. A. (1992) Crit. Rev. Oncogen., 3, 175-231.

21. Schwab, G., Duroux, I., Chavany, C. and Hélène, C. (1994) Ann. Oncol., 5, 55-58. MEDLINE Abstract

22. Wang, G., Seidman, M. M. and Glazer, P. M. (1996) Science, 271, 802-805. MEDLINE Abstract

23. Havre, P. A., Gunther, E. J., Gasparro, F. P. and Glazer, P. M. (1993) Proc. Natl. Acad. Sci. USA, 90, 7879-7883. MEDLINE Abstract

24. Postel, E. H., Flint, S. J., Kessler, D. J. and Hogan, M. E. (1991) Proc. Natl. Acad. Sci. USA, 88, 8227-8231. MEDLINE Abstract

25. Elledge, S. J., Sugiona, P., Guarente, L. and Davis, R. W. (1989) Proc. Natl. Acad. Sci. USA, 86, 3689-3693. MEDLINE Abstract

26. Elledge, S. J. and Davis, R. W. (1989) Genes Dev., 3, 185-197. MEDLINE Abstract

27. Soukup, G. A., Ellington, A. D. and Maher, L. J. (1996) J. Mol. Biol., 259, 216-228. MEDLINE Abstract

28. Huang, L., Sera, T. and Schultz, P. G. (1994) Proc. Natl. Acad. Sci. USA, 91, 3969-3973. MEDLINE Abstract

29. Sambrook, J., Fritsch, E. F. and Maniatis, T. (1989) Molecular Cloning: A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.

30. Coleman, J., Green, P. J. and Inouye, M. (1984) Cell, 37, 429-436. MEDLINE Abstract

31. Hirashima, A., Sawaki, S., Mizuno, T., Houba-Herin, N. and Inouye, M. (1989) J. Biochem., 106, 163-166. MEDLINE Abstract

32. Sharma, R. C. and Schimke, R. T. (1996) Biotechniques, 20, 42-44. MEDLINE Abstract

33. Jaeger, J. A., Turner, D. H. and Zuker, M. (1989) Proc. Natl. Acad. Sci. USA, 86, 7706-7710. MEDLINE Abstract

34. Jaeger, J. A., Turner, D. H. and Zuker, M. (1990) Methods Enzymol., 183, 281-306. MEDLINE Abstract

35. Zuker, M. (1989) Science, 244, 48-52. MEDLINE Abstract

36. Serra, M. J. and Turner, D. H. (1995) Methods Enzymol., 259, 242-261. MEDLINE Abstract

37. Miller, J. H. (1972) Experiments in Molecular Genetics. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.

38. Okamoto, K. and Freundlich, M. (1986) Proc. Natl. Acad. Sci. USA, 83, 5000-5004. MEDLINE Abstract

---

*To whom correspondence should be addressed. Tel: +1 507 284 9041; Fax: +1 507 284 2053; Email: maher@mayo.edu
⁺Present address: Department of Molecular, Cellular and Developmental Biology, 442 Kline Biology Tower, Yale University, PO Box 208103, New Haven, CT 06520-8103, USA

---

This page is run by Oxford University Press, Great Clarendon Street, Oxford OX2 6DP, as part of the OUP Journals
Comments and feedback: www-admin{at}oup.co.uk
Last modification: 19 May 1998
Copyright©Oxford University Press, 1998.
CiteULike    Connotea    Del.icio.us    What's this?

This article has been cited by other articles:

				H. S. Zaher and P. J. Unrau Selection of an improved RNA polymerase ribozyme with superior extension and fidelity RNA, July 1, 2007; 13(7): 1017 - 1026. [Abstract] [Full Text] [PDF]

				J. C. STRIGGLES, M. B. MARTIN, and F. J. SCHMIDT Frequency of RNA-RNA interaction in a model of the RNA World. RNA, March 1, 2006; 12(3): 353 - 359. [Abstract] [Full Text] [PDF]

				M. LEVY, K. E. GRISWOLD, and A. D. ELLINGTON Direct selection of trans-acting ligase ribozymes by in vitro compartmentalization RNA, October 1, 2005; 11(10): 1555 - 1562. [Abstract] [Full Text] [PDF]

				J. M. Zimmerman and L. J. Maher III In vivo selection of spectinomycin-binding RNAs Nucleic Acids Res., December 15, 2002; 30(24): 5425 - 5435. [Abstract] [Full Text] [PDF]

This Article Abstract Print PDF (179K) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Add to My Personal Archive Download to citation manager Search for citing articles in: ISI Web of Science (6) Request Permissions Commercial Re-use Guidelines for Open Access NAR Content Google Scholar Articles by Soukup, G. A. Articles by Maher, J. J. Search for Related Content PubMed PubMed Citation Articles by Soukup, G. A. Articles by Maher, J. J., 3rd Social Bookmarking          What's this?