Variable coordination of cotranscribed genes in Escherichia coli following antisense repression

Background A majority of bacterial genes belong to tight clusters and operons, which complicates gene functional studies using conventional knock-out methods. Antisense agents can down-regulate the expression of genes without disrupting the genome because they bind mRNA and block its expression. However, it is unclear how antisense inhibition affects expression from genes that are cotranscribed with the target. Results To examine the effects of antisense inhibition on cotranscribed genes, we constructed a plasmid expressing the two reporter genes gfp and DsRed as one transcriptional unit. Incubation with antisense peptide nucleic acid (PNA) targeted to the mRNA start codon region of either the upstream gfp or the downstream DsRed gene resulted in a complete expression discoordination from this artificial construct. The same approach was applied to the three cotranscribed genes in the endogenously expressed lac-operon (lacZ, Y and A) and partial downstream expression coordination was seen when the lacZ start codon was targeted with antisense PNA. Targeting the lacY mRNA start codon region showed no effect on the upstream lacZ gene expression whereas expression from the downstream lacA gene was affected as strongly as the lacY gene. Determination of lacZ and lacY mRNA levels revealed a pattern of reduction that was similar to the Lac-proteins, indicating a relation between translation inhibition and mRNA degradation as a response to antisense PNA treatment. Conclusion The results show that antisense mediated repression of genes within operons affect cotranscribed genes to a variable degree. Target transcript stability appears to be closely related to inhibition of translation and presumably depends on translating ribosomes protecting the mRNA from intrinsic decay mechanisms. Therefore, for genes within operons and clusters it is likely that the nature of the target transcript will determine the inhibitory effects on cotranscribed genes. Consequently, no simple and specific methods for expression control of a single gene within polycistronic operons are available, and a thorough understanding of mRNA regulation and stability is required to understand the results from both knock-down and knock-out methods used in bacteria.


Background
Antisense agents are useful for functional genomics in bacteria as they can down-regulate expression of specific genes by binding mRNA and inhibit translation. As antisense agents may be titrated into cultures of wild type cells they are attractive for studies of any gene, even those stringently required for survival. Also, down-regulation through externally added antisense agents bypasses the need for genome modifications, thereby allowing gene specific studies of both clustered and overlapping genes. A remaining question, however, is whether antisense agents retain specific inhibitory effects when cotranscribed genes are targeted.
Prokaryotic genes are commonly expressed as polycistronic transcripts. For E. coli, several concordant studies predict that the >4400 genes are expressed as 2500-2800 transcriptional units (TUs) of which 70-75% are expected to be monocistronic [1][2][3]. Therefore, over half of the E. coli genes are expressed from multigene operons. Similarly, predictions performed on various other prokaryotic genomes show that between 27 and 78% of the genes are expressed within polycistronic transcripts, with an average well above 50% [2]. As a practical consequence, reliable single ORF genetic knock-out techniques are difficult or unavailable for functional genomics studies on a majority of prokaryotic genes.
We are interested in the use of antisense peptide nucleic acid (PNA) for functional genomics as well as antibacterial drug discovery. PNA is a nucleic acid mimic with the phosphodiester backbone replaced with pseudo-peptides and can bind to other nucleic acids with high affinity [4]. PNA targeted specifically to mRNA results in inhibition of gene expression, but unlike many other synthetic nucleic acids PNA does not induce RNase H mediated cleavage [5,6]. Instead, inhibition occurs through steric hindrance of the translation machinery [6], and potent antisense effects can be obtained by targeting the mRNA start codon region [7].
Here we examined the consequences of PNA-mediated inhibition of cotranscribed gene expression in E. coli. The separate mRNA start codon regions within both an artificial TU (gfp/DsRed) and the endogenous E. coli lac-operon were targeted by antisense-PNAs followed by protein quantifications for each ORF. Expression from the artificial gfp/DsRed construct was completely discoordinated, as only GFP levels were reduced when the gfp start codon region was targeted, and only RFP levels were reduced when the start codon region of DsRed was targeted. However, a directional downstream expression inhibition of the native lac-genes was observed. Interestingly, mRNAlevel measurements confirmed this expression pattern, although inhibition was less pronounced than at the pro-tein level. The main conclusion is that steric hindrance of translation initiation affects cotranscribed genes to a variable degree. Therefore, simple, specific methods for functional studies on most bacterial genes are lacking.

Antisense effects on genes within an artificial operon
To obtain a system producing polycistronic mRNA transcripts where expression from cotranscribed ORFs is easily quantified we inserted a stop codon between the gfp and DsRed parts in the pGRFP double reporter plasmid [8]. Also, to improve prospects for translational coordination between the genes, the intergenic distance was kept short and a strong SD sequence was included [9]. The resulting pGRFPsep construct (Figure 1, Figure 2A) was introduced into E. coli AS19. After treatment with varying concentrations of anti-gfp, anti-DsRed or two unrelated control PNAs (anti-lacZ and anti-lacY, see Table 1 for details) protein expression was determined. As seen in Figure 2B, treatment with the anti-gfp PNAs resulted in a reduction of GFP levels, whereas RFP levels remained largely unaffected (upper panel). Similarly, the anti-DsRed PNAs only reduced the expression of RFP ( Figure 2B, middle panel). Moreover, the control experiments with PNAs directed to unrelated mRNA targets did not inhibit the expression of GFP or DsRed ( Figure 2B, lower panel). Therefore, the results show a specific downregulation of expression even though the coding sequences belong to the same transcriptional unit.

Antisense effects on gene expression within the lac-operon
The above result describes the effect of antisense PNA targeted to cotranscribed genes within a system where translation is discoordinated. However, as this system is an Illustration of plasmid pGRFP-sep Figure 1 Illustration of plasmid pGRFP-sep. The parent greenred fusion plasmid pGRFP [8] was modified by insertion of a gfp stop codon, a strong Shine-Dalgarno sequence and start codon for DsRed. pGRFP-sep expresses gfp and DsRed as one transcript encoding two different proteins.
Expression of GFP and DsRed from an artificial polycistronic transcript after treatment with anti-gfp and anti-DsRed PNA artificial construct lacking naturally evolved regulatory components and sequences we decided to continue the study with an endogenous operon. We targeted the genes in the chromosomally expressed lac-operon and determined the effect on the expression of its three structural genes; lacZ, lacY and lacA (see Figure 3A). Treatment with anti-lacZ PNA resulted in a strong inhibition of the lacZ encoded protein ( Figure 3B). Also, expression from the two downstream genes lacY and lacA was markedly reduced, however not to the same extent as lacZ expression. When bacteria were incubated with anti-lacY PNA, lacY and lacA gene expression was reduced, whereas expression from the upstream lacZ gene was unaffected ( Figure 3C). Unfortunately, we did not obtain any data with anti-lacA PNA as this displayed strong growth inhibitory effects. Nevertheless, the results show that antisense targeted to an upstream gene in the lac-operon has a large impact on the expression from the downstream genes. In contrast, no upstream coordination was observed, suggesting a directional inhibitory effect in the lac-operon.
To ensure that the anti-lacZ and anti-lacY PNAs did not interfere with bacterial growth and examine whether expression of an unrelated protein was affected by the presence of antisense, we determined growth rate and Lux activity in cells carrying the plasmid pLux1 following treatment with PNAs that target the lac-operon. Neither growth nor Lux expression was altered by the PNAs when added in concentrations up to 1000 nM (see Additional file 1).

Antisense effect on mRNA levels
To better understand the expression pattern observed at the protein level, we used quantitative real time PCR (qPCR) to study the impact of antisense PNA inhibition at the mRNA level. Samples from bacteria treated with varying concentrations of either anti-lacZ (0, 200 and 400 nM) or anti-lacY PNA (0, 100 and 200 nM) were analysed for their lacZ and lacY mRNA content ( Figure 4). Treatment with anti-lacZ PNA significantly reduced lacZ mRNA at both 200 nM and 400 nM (p < 0.01 and p < 0.001, respectively) ( Figure 4A). LacY mRNA was less affected and a sig-nificant reduction was only obtained with the higher anti-lacZ PNA concentration (p < 0.05). Moreover, there was a significantly higher reduction of lacZ mRNA than lacY mRNA after treatment with the lower antisense dose (p < 0.05), but not at the higher dose (p = 0.066). Interestingly, a comparison between protein and mRNA levels after anti-lacZ PNA treatment reveals a coupled reduction (cf. Figure 3B with 4A). Anti-lacY PNA significantly reduced the level of lacY mRNA at both 100 and 200 nM concentrations (p < 0.05 and p < 0.005, respectively), whereas lacZ mRNA levels were comparable to untreated cells (Figure 4B). Again, a coupled reduction of transcript and protein levels is demonstrated as LacY but not LacZ protein concentrations were reduced by anti-lacY PNA (cf. Figure  3C with 4B).
The above results demonstrate a dose dependent gene expression inhibition at both the mRNA and protein levels, with a proportionally stronger reduction of protein.
To further examine this relationship we performed a more detailed analysis of lacZ gene expression with anti-lacZ PNA doses ranging from 0 to 400 nM concentrations at 50 nM intervals. Again, a coupled reduction of both protein and mRNA levels was observed ( Figure 5A). Supporting our previous observations ( Figure 3B and Figure 4A), increasing doses of anti-lacZ PNA resulted in a negative exponential decrease in LacZ protein activity whereas the decrease in lacZ mRNA was linear ( Figure 5B).

Discussion
The aim of this study was to examine antisense PNA mediated inhibitory effects when genes in polycistronic operons are targeted. We demonstrate that protein expression from cotranscribed genes can be substantially inhibited by antisense and that there is a stringent coupling between protein and transcript levels when ribosomal binding is hindered.
We first demonstrate that antisense PNAs directed to the mRNA start codon region of either gene in an artificial gfp-DsRed dual reporter operon lead to a fully discoordinated reduction in gene expression. This was followed by target- mRNA target sites are shown relative to translation start sites of the respective coding sequences.
ing the separate genes in the chromosomally encoded lacoperon. Interestingly, in this case a different response to antisense PNA mediated translational inhibition was observed. Targeting the 5' lacZ resulted in a stronger inhibition of LacZ than the downstream LacY and LacA proteins, demonstrating a partial discoordination in protein expression, a phenomena previously observed by Pestka and co-workers when "anti-mRNA" directed to lacZ mRNA was expressed from a plasmid [10]. When lacY was targeted, downstream LacA expression was reduced as much as LacY, whereas LacZ was unaffected. The general differences in expression response between the gfp-DsRed transcript and the lac-operon transcript can likely be assigned to differences in expression regulation: For the naturally occurring lac-transcript, and also for many other endogenously expressed polycistronic transcripts, a variety of regulatory mechanisms have evolved to balance gene expression according to the cell's needs. The gfp-DsRed transcript, however, is not a natural component for bacteria and is therefore not likely to be affected by such specific regulatory mechanisms. Therefore we conclude that antisense PNAs targeted to genes belonging to polycistronic transcripts are very likely to affect cotranscribed genes, thereby rendering gene functional studies on such genes more difficult.
Several studies show that efficient ribosomal binding is necessary for bacterial mRNA stability [11][12][13], and that actively translating ribosomes protect mRNA from RNase E mediated decay [14]. Also, it has been shown that enhanced transcription rates of lacZ mRNA lead to decreased production of LacZ protein and lower transcript stability, probably due to prolonged exposure of unprotected mRNA behind the RNA polymerase [15]. Antisense PNAs inhibit gene expression by steric hindrance of the translation machinery and the effect is most potent when targeting mRNA ribosome binding sites [6,7]. By blocking ribosome loading, antisense PNA may have an effect on mRNA levels. Therefore, we investigated the impact of anti-lacZ and anti-lacY PNA on mRNA. We observed that the pattern of mRNA levels following antisense treatment was similar to that of protein levels within the lac-operon, suggesting that inhibition of translation by antisense PNA indeed has a destabilising effect on mRNA. Destabilisation of target mRNA is in accordance with that observed by Forsyth and co-workers when they used expressed antisense RNA to down-regulate gene expression [16]. Therefore, it appears that the activity of sterically blocking antisense agents, at least in some cases, can be assessed by examining the mRNA levels. The reason for this coupling of mRNA and protein levels is likely due to an increased exposure/decreased protection of the mRNA.
The altered expression pattern of the lac genes as a response to antisense treatment raises at least two new Protein expression from the lac-operon after treatment with antisense PNA directed to lacZ and lacY mRNA start codon regions questions: (i) "Why is expression from the downstream lacY and lacA less inhibited than lacZ with anti-lacZ PNA?" and (ii) "How can lacZ expression remain unaffected when cells are treated with anti-lacY PNA?" If one only considers protein expression levels it would be possible to explain the partly discoordinated downstream inhibition as being due to noise in the LacY and LacA assays. However, standard curves created with different mixtures of induced and non-induced cells gave linear responses (data not shown). Moreover, mRNA quantification experiments revealed a coupled pattern between transcript and protein levels when antisense was added. An alternative explanation could be that reinitiation at nearby start codons after a ribosome has reached the stop of a coding region is somehow affected in a polar manner. However, it has been demonstrated that reinitiation of ribosomes is a comparatively rare event during translation [9,17], an observation also supported here by our results from the double reporter experiment. It is therefore unlikely that altered reinitiation could explain the rather large expression differences between LacZ and LacYA. Also, the effect on mRNA suggests a different explanation. We believe the explanation may lie in direct inhibition of LacZ protein synthesis together with Rho-dependent premature transcription termination. Cessation of translation induces Rho-dependent termination, probably due to increased exposure of Rho protein binding sites behind the transcribing RNA polymerase [18]. As it has previously been shown that the lacZ coding region contains several Rho binding sites [19], antisense inhibition of ribosome binding at the lacZ mRNA should lead to increased Rho binding and therefore increased premature transcription termination. However, as translation inhibition of, in this case, LacZ is the primary event; premature transcription termination is expected to be slightly delayed. Indeed, such a tendency can be observed at both protein and mRNA levels following antisense inhibition ( Figure 3B and 4A). Moreover, the analysis using a more detailed anti-lacZ PNA dose range confirms this pattern ( Figure 5): While mRNA amounts decreased linearly with increasing concentrations of anti-lacZ PNA the decrease in LacZ protein activity was much more pronounced.
Why, then, is lacZ not affected by downstream targeting by the anti-lacY PNA? Even though the lac-operon is thought to be fully transcribed before being processed [20], it has been observed that lacZ transcripts are several folds more abundant than ZYA transcripts [20,21]. This imbalance can be explained by RNase E cleavage within the lacZY intergenic region that stabilises the lacZ portion of the transcript [22], an event that decouples the coding regions of lacZ and lacYA [23]. According to this model, Relative amounts of lacZ and lacY mRNA species after antisense PNA treatment most of the LacZ product is likely to be translated from processed monocistronic transcripts that can not be affected by antisense binding to lacY mRNA.

Conclusion
In summary, our experiments show that antisense inhibition of genes within polycistronic operons can strongly affect cotranscribed genes. We further show that inhibition is followed by a decreased stability of the transcript that is coupled with translational reduction. This decreased mRNA stability can not be explained by RNase H activity because antisense PNA do not induce such cleavage. Instead, we believe that the presence of antisense unmasks the mRNA from translating ribosomes and thereby increases exposure to RNase cleavage and intrinsic decay mechanisms that regulate RNA stability. A practical issue associated with the stringent correlation between transcript and protein levels is that mRNA quantification may provide an additional measure of specificity for antisense experiments, at least for targets susceptible to rapid degradation when unprotected. Also, the antisense based methodology described here can be helpful for investigations of bacterial mRNA translation and stability. Nevertheless, the dependence of protein expression on mRNA structure and stability indicate that simple and specific methods for functional studies on the majority of bacterial genes are still lacking. Therefore, we suggest that antisense and other strategies for bacterial gene expression inhibition as a means for gene functional studies [24][25][26] and drug screening [16,27] should be performed with extra caution with respect to cotranscribed genes.

Bacterial strain and PNAs
E. coli strain AS19 was used throughout this study. Due to its hyper permeable nature lower antisense concentrations are required for gene expression inhibition in this strain. PNAs were designed to target the start codon regions of gfp, luc, lacZ and lacY (and lacA), with the peptide (KFF) 3 K attached for improved cellular uptake [28] and were obtained from Panagene (Daejeon Metropolitan City, Korea). Peptide-PNA sequences are listed in Table 1. Importantly, none of the peptide-PNAs (except anti-lacA PNA) displayed any alteration in bacterial growth under the conditions and with the concentrations used in this study.

Construction of a polycistronic pGRFP-sep plasmid
Cloning vector pGRFP [8] expressing GFP and DsRed as a fusion protein was kindly provided by Stephanie Mohr at Dana-Farber/Harvard Cancer Center DNA Resource Core, US. A GFP stop codon, a BglII cleavage site, a Shine-Dalgarno sequence and a unique PNA binding site were introduced between gfp and DsRed using upstream primer 5'gcgagatctcatttgtatagttcatccatgccatgtgtaatcccagcagc-3' and Dose effect of antisense PNA on LacZ protein activity and lacZ mRNA abundance downstream primer 5'-tatagatctaaggaggaaacagaatggcctcctccgaggacgtcatca-3'. Cleavage with BglII and ligation yielded pGRFPsep, expressing gfp and DsRed as a single mRNA encoding two different proteins (Figure 1). pGRFPsep was transformed into E. coli AS19 where the double reporter system was constitutively expressed. Beta-galactosidase content of the samples was determined by addition of 160 µl Z-buffer and 20 µl ONPG solution [29], followed by kinetic spectrophotometric measurements of colour conversion at 420 nm every 30 s with 5 s shaking. Curve slopes were used to determine relative levels of beta-galactosidase activity in the samples.

LacY assay
Aliquots of 150 µl culture were mixed with 20 µl chloramphenicol to give a final 150 µg/ml antibiotic concentration. Samples were incubated on ice for 10 min and 150 µl were transferred to 96-well plates containing 50 µl 14

LacA assay
Aliquots of 50 µl culture were transferred to 50 µl lysis buffer and incubated at room temperature for 20 minutes. Samples were kept at 70°C for 5 min before a 15 min centrifugation at 20800 g in RT. Supernatants (50 µl) were transferred to 10 µl assay media (5 mg/ml acetyl CoA and 250 mg/ml IPTG in 0.05 M Tris and 0.01 M EDTA, pH 7.9) followed by a one hour incubation at 25°C. Addition of 200 µl DTNB solution (250 µg/ml in 0.05 M Tris pH 7.9) was followed by OD-measurement at 412 nm to determine relative amounts of lacA gene product.
Controls and standard curves were provided for all assays using un-induced AS19 cells and/or mixtures of uninduced and induced cells.
An additional control for anti-lacZ and anti-lacY PNA specificity was performed on AS19 cells carrying plasmid pLux1 [31] that contains the luxA and luxB genes from Vibrio harveyi. Cells were grown under the same conditions as in the lac-assays with increasing concentrations (0-1000 nM) of anti-lacZ and anti-lacY PNA. Cultures were either allowed to grow continuously for generating growth curves or harvested after an increase in OD 550 of 0.1 cells for quantification of luciferase expression following a previously described protocol [32]. Plasmid-free AS19 were used to remove background luminescence.

mRNA quantification by reverse transcription and real time PCR
E. coli AS19 treated with different concentrations of anti-lacZ or anti-lacY PNA in the presence or absence of 100 µM IPTG were incubated until the OD 550 increased bỹ 0.1. Cells with equal treatment were pooled to get two separate 1 ml samples for each treatment. Total RNA was extracted with TRIZOL ® Reagent following the manufacturer's instructions (Invitrogen). Samples were dissolved in 34 µl RNase-free water (Invitrogen) and treated with RNase-free DNase (Sigma Aldrich). RNA concentrations were determined by OD-readings on an Ultrospec 3000 (Pharmacia Biotech, Uppsala, Sweden). Reverse transcription was performed using random hexamers with 1 µg RNA in 50 µl reactions following the manufacturers protocol (Applied Biosystems). Aliquots derived from 50 ng RNA were analyzed by qPCR on an ABI PRISM 7000 (Applied Biosystems) according to the manufacturers protocol. Samples were run in triplicates and the data obtained were analyzed with ABI Prism SDS Software (Applied Biosystems). The ∆∆Ct method was used to calculate lacZ and lacY gene expression using rRNA as an internal standard [30]. Sequences for the lacZ and lacY primers were selected to amplify regions near the 5' end of the coding regions to increase the chances of detecting functional transcripts as lac mRNAs are degraded in a 5' to 3' direction [20,33], and were: lacZ forward 5'-ccctggcgttacccaacttaa-3'; lacZ reverse 5'-gcgggcctcttcgctattac-3'; lacY forward 5'-cgcaaatacctgctgtggatta-3' and lacY reverse 5'tggcccgaagataaaaataaagaa-3'. The primer binding sites did not interfere with antisense PNA target sites (see Figure  3A). Primer sequences for rRNA were taken from [34] and were: rRNA forward 5'-catgccgcgtgtatgaagaa-3' and rRNA reverse 5'-cgggtaacgtcaatgagcaaa-3'. RNA isolation and quantification was repeated at four different occasions giving eight samples and 24 data points per PNA concentration. Data comparisons were made to a normalised control value obtained by calculating the mean at each experimental occasion. Statistical calculations were performed in Excel using unpaired t-tests assuming equal variance among groups and data are presented as the mean with error bars reflecting the standard deviation. Significance was set at p < 0.05.
Parallel analysis of LacZ activity and lacZ mRNA abundance E. coli AS19 treated with anti-lacZ PNA concentrations ranging from 0 to 400 nM at 50 nM intervals were grown in six double replicas in the presence or absence of 100 µM IPTG. Cultures were incubated under the same conditions as described above and harvested when OD 550 had increased by ~0.1. Samples of 20 µl were removed from each well to be used for separate LacZ activity measurements as described above, and the remainder of the cultures were pooled into groups with equal treatment. Total RNA was extracted and treated with DNase I using RiboPure Bacteria kit (Ambion), according to the manufacturer's protocol. Quantification of lacZ mRNA was performed as described above. Mean values from the six replicates of each treatment were used to determine relative LacZ activity and the averages of three replicates derived from each sample pool were used to estimate mRNA abundance. The experimental procedure was repeated at three different occasions, and mean values with error bars were plotted as a histogram. Regression analysis was conducted using VassarStats [35].