Antisense Phosphorodiamidate Morpholino Oligomer Length and Target Position Effects on Gene-Specific Inhibition in Escherichia coli

ABSTRACT Phosphorodiamidate morpholino oligomers (PMOs) are synthetic DNA analogs that inhibit gene expression in a sequence-dependent manner. PMOs of various lengths (7 to 20 bases) were tested for inhibition of luciferase expression in Escherichia coli. Shorter PMOs generally inhibited luciferase greater than longer PMOs. Conversely, in bacterial cell-free protein synthesis reactions, longer PMOs inhibited equally or more than shorter PMOs. Overlapping, isometric (10-base) PMOs complementary to the region around the start codon of luciferase inhibited to different extents in bacterial cell-free protein expression reactions. Including the anti-start codon in PMOs was not required for maximal inhibition. PMOs targeted to 5′ nontranslated or 3′ coding regions within luciferase mRNA did not inhibit, except for one PMO targeted to the ribosome-binding site. Inhibition of luciferase expression correlated negatively with the predicted secondary structure of mRNA regions targeted by PMO but did not correlate with C+G content of targeted regions. The effects of PMO length and position were corroborated by using PMOs (6 to 20 bases) targeted to acpP, a gene required for viability. Because inhibition by PMOs of ∼11 bases was unexpected based on previous results in eukaryotes, we tested an 11-base PMO in HeLa cells and reticulocyte cell-free protein synthesis reactions. The 11-base PMO significantly inhibited luciferase expression in HeLa cells, although less than did a 20-base PMO. In reticulocyte cell-free reactions, there was a trend toward more inhibition with longer PMOs. These studies indicate that strategies for designing PMOs are substantially different for prokaryotic than eukaryotic targets.

Antisense drugs are short (10 to 25 bases) oligomers that mimic DNA or RNA and inhibit gene expression in a sequence-dependent manner (2,6,29,30). They differ from native RNA or DNA in the chemical structure that links the four common bases.
Most of the reported work on antisense drugs has been accomplished in eukaryotic systems. In eukaryotes, antisense compounds inhibit by two general mechanisms. Compounds such as phosphorothioates hybridize to mRNA and promote its degradation by RNase H (3,5,28). Other compounds, such as peptide nucleic acid (PNA) and phosphorodiamidate morpholino oligomer (PMO), hybridize to specific mRNA and block translation by an RNase H-independent mechanism (3,28).
The most effective region for targeting PNA and PMO is the 5Ј untranslated region and initiation codon of mRNA (3,10,14,15,18,20,28,30,32). The optimal length of a PNA or PMO depends on achieving a balance between specificity and efficacy and is estimated to be about 15 to 25 bases (2,14,16,21,30). However, experimental evidence suggests that many factors, such as cellular uptake or invasion of mRNA secondary structure, influence efficacy and can tip the balance in favor of shorter antisense oligomers (13,15,19,23,31).
Sequence-specific antibacterial drugs are a recent application of antisense technology (11,12,17,24,25,33). Nielsen and coworkers (13) have shown that PNA in the 9-to 12-mer range were more active inhibitors than larger oligomers in Escherichia coli. Antibacterial PNA targeted to the region near the AUG start codon inhibit their specific targets (12). A recent, systematic study with PNA showed that the region around the start codon is the most reliable target site for bacterial antisense inhibition (7). One objective of our studies was to compare and contrast the effects of PMO with that of PNA on the inhibition of gene expression in E. coli.
We show here the effects of varying the length and position of PMO on bacterial gene expression. Our results suggest differences between the mechanisms of antisense oligomer inhibition in bacteria compared to eukaryotic cells and between the effects of PMO and PNA on bacterial gene expression.

MATERIALS AND METHODS
PMOs. PMOs were synthesized and purified at AVI BioPharma, Inc. (Corvallis, Oreg.), as previously described (11), dissolved in water, filtered through a 0.2-m-pore-size membrane (HT Tuffryn; Gelman Sciences, Inc., Ann Arbor, Mich.), and stored at 4°C. Sequences of PMOs used in the present study are shown in Table 1. The concentration of PMO was determined spectrophotometrically by measuring the absorbance at 260 nm and calculating the molarity by using an extinction coefficient calculated as described previously (4). The base sequences of nonsense PMOs were chosen randomly from a pool of bases that approximated the relative abundance of bases in the corresponding PMOs that were targeted to either myc-luc or acpP.
Bacteria and growth conditions. E. coli AS19 (26) and SM101 (9), which have defects in lipopolysaccharide synthesis that result in outer membrane permeability to high-molecular-weight solutes, were grown aerobically in Luria-Bertani broth at 37 and 30°C, respectively. Transformants that expressed pSE380myc-luc were grown in Luria-Bertani medium plus 100 g of ampicillin/ml. Reporter gene. Standard molecular biology procedures (1) were used for all constructions. All constructs were sequenced. Two reporter systems (pT7myc-luc and pSE380myc-luc) for antisense inhibition were previously constructed, as described previously (11), by fusing 30 bp of the 5Ј end of human c-myc to all but The acpP-luc reporter (pCNacpP-luc) was made by ligating a SalI-NotI restriction fragment of luc with the SalI-NotI fragment of pCiNeo (Promega Corp.), removing the adenosine from the start codon by site-directed mutagenesis, and then directionally cloning a synthetic fragment of acpP (bp Ϫ17 to ϩ23, inclusive, where ϩ1 is adenosine of the start codon) between the NheI-SalI sites. pCNmyc-luc was made in the same way, except that the myc sequence from Ϫ14 through ϩ16 (inclusive; numbering adenosine of the start codon as ϩ1) instead of acpP was cloned into the NheI-SalI sites. Luciferase enzyme activity was measured in bacteria as described previously (11).
Cell-free protein synthesis. Bacterial cell-free protein synthesis reactions were performed by mixing reactants on ice according to the manufacturer's instruction (Promega Corp.). Reactions were programmed with either pT7myc-luc plasmid in a coupled transcription-translation reaction or mRNA synthesized in a cellfree RNA synthesis reaction (MEGAscript T7 High Yield Transcription Kit; Ambion, Inc., Austin, Tex.) programmed with pT7myc-luc. All acp-luc reactions were programmed with pCNacpP-luc. Where indicated, cell-free reactions were composed with rabbit reticulocyte lysate as described by the manufacturer (Promega Corp.). PMO was added to a final concentration of either 100 or 200 nM as indicated in the figure legends. After 1 h at 37°C, the reactions were cooled on ice and luciferase was measured as described previously (11).
Mammalian tissue culture. HeLa cells were transfected in T75 tissue culture flasks (Nalge Nunc, Inc., Rochester, N.Y.) with a luciferase reporter plasmid (pCNmyc-luc) by using Lipofectamine reagent (Gibco-BRL, Grand Island, N.Y.), according to user's manual, for 5 h in serum-free media (Opti-MEM1; Gibco, Inc., Carlsbad, Calif.) before re-addition of recovery medium (HyQ DME/F-12 supplemented with 20% fetal bovine serum and Gibco Antibiotic-Antimycotic 15240-062; HyClone, Inc., Logan, Utah) at 37°C in 5% CO 2 . After 24 h, the cells were pooled, and 10 6 cells were added to each well of a six-well plate (BD Biosciences, San Jose, Calif.) in 2 ml of growth medium. After an additional 24 h, PMO was added to a final concentration of 10 M in 2 ml of fresh growth medium, and the cells were scraped from the plate surface with a rubber policeman to deliver the PMO to the cells as previously described (22). After scrape loading, the cells were transferred to fresh six-well plates and incubated at 37°C until the time of assay. At 7 and 24 h the cells were examined by microscopy to verify that each culture had the same number of cells, harvested by centrifugation, and lysed in Promega cell culture lysis reagent. Luciferase was measured by mixing the cell lysate with luciferase assay reagent (Promega) and reading the light emission in a model TD-20e luminometer (Turner Designs, Inc., Mountain View, Calif.).
RNA secondary structure. The RNA folding algorithm M-Fold (35) was used to predict the secondary structure of bases 1 to 120 or bases 90 to 1745 of the mRNA transcribed from pT7myc-luc. The folded structure of bases 1 to 120 had a minimum ⌬G of Ϫ6.5 kcal/mol and that of bases 90 to 1745 had a minimum ⌬G of Ϫ452.28 kcal/mol. Each PMO was scored (we refer to this as the "2°score") by calculating the fraction of bases (in the PMO) that are complementary to double-stranded (duplex) regions within the folded target mRNA. For example, PMO 331 (Table 1) is 10 bases in length and complementary to a region of myc-luc mRNA that (according to M-Fold) forms duplex RNA at 4 of its 10 bases (the other 6 bases are not paired). The 2°score for PMO 331 would therefore be 4 bases/10 bases, i.e., 0.4. Statistical analysis. A Spearman rank-order correlation was used to analyze correlations between the inhibitory effects of PMO and either GϩC content or secondary structure score of each PMO. An unpaired t test (one-sided) was used to analyze differences in inhibition of luciferase, optical density at 600 nm, and CFU/milliliter.

RESULTS
Various PMO lengths. The efficacy of PMO inhibition of target gene expression was tested as a function of the length of PMO. Two series of PMOs (Table 1) were synthesized by sequentially truncating subunits from either the 3Ј or 5Ј end of a 20-base PMO that is complementary to the region around the start codon of a luciferase reporter gene (myc-luc). myc is a eukaryotic gene without a similar sequence in E. coli and contributes the nontranslated 5Ј end and start codon of the reporter transcript. Previous results have shown that the 20-base myc PMO inhibits expression of this reporter gene (11). Each of the various-length (7-to 20-base) PMOs was added to growing cultures of E. coli SM101(pSE380myc-luc). After 8 h in culture, luciferase was measured. Compared to a culture without PMO, 3Ј truncated PMOs inhibited luciferase from 16 to 95% and 5Ј truncated PMOs inhibited luciferase from 6 to 89% (Fig. 1). There is a trend toward more inhibition with shorter PMOs, with a maximum at about nine bases. The seven-and eight-base, 3Ј truncated PMOs inhibited less than the same length 5Ј truncated PMOs. Control PMOs 371 and 62-8 with nonsense sequences inhibited 0 to 1%. The same pattern of inhibition was found when luciferase was measured at earlier time points (data not shown).
Cell-free protein synthesis. PMOs from the same two series were added individually to bacterial cell-free protein synthesis reactions programmed to express myc-luciferase. These experiments were designed to eliminate the effects of entry of PMOs into the cell and to test the PMOs at 37°C instead of the permissive growth temperature (30°C) of the conditional mutant SM101. PMOs truncated at the 3Ј end from 10 to 20 bases in length inhibited about the same (Fig. 2). Inhibition decreased sharply with 3Ј truncated PMOs less than nine bases in length. PMOs truncated at the 5Ј end from 12 to 20 bases in length inhibited luciferase about the same (Fig. 2). A sharp decrease in inhibition occurred with 5Ј truncated PMOs less than 12 bases in length. Nonsense sequence controls did not inhibit significantly.
Various PMO positions. A series of isometric (10-base) PMOs, which varied by 1 base at each end (Table 1) and was targeted to the region around the AUG start codon of myc-luc, was added to bacterial cell-free reactions programmed to synthesize myc-luc. All PMOs inhibited luciferase expression (Fig.  3). A trend toward more inhibition was apparent as the target position moved downstream of the start codon. There was no  Table 1); ■, PMOs truncated at the 5Ј end (the series from PMO 126 through 334 shown in Table 1). Nonsensical sequence control PMOs (62-8 and 371) are indicated (14c and 20c, respectively). Error bars show the SD (n ϭ 2 to 5). correlation between inhibition and inclusion of the anti-start codon within the PMO sequence.
Another series of PMOs with various lengths was targeted to various positions within the transcript of the myc-luc, including the extreme 5Ј end of the transcript (PMOs 341 to 344), the ribosome-binding site (PMOs 345 and 347), the region upstream of the ribosome-binding site (PMO 346), the region around and immediately downstream of the start codon (PMO 126), and the 3Ј coding region of luciferase (PMOs 214 and 215). Each PMO or a nonsense PMO (371 or 373) was added (200 nM) to a bacterial cell-free protein synthesis reaction programmed to synthesize luciferase. After 1 h at 37°C, luciferase light production was measured. The results were that PMO 126 inhibited luciferase 79%, which was significantly (P Ͻ 0.001, n ϭ 4) more inhibition than that caused by either of the nonsense PMOs 371 or 373 (6 or 4% inhibition, respectively). PMOs 345 and 347 that overlapped the ribosome-binding site inhibited luciferase 24 and 16%, respectively, which for PMO 345 was barely different (P ϭ 0.016, n ϭ 3) but for PMO 347 was not significantly different (P ϭ 0.092, n ϭ 6) than the nonsense control PMO 371. Inhibition by each of the other PMOs in this series ranged from 9 to 14% but was not significantly different (P Ͻ 0.1, n ϭ 3 or 5) than either of the nonsense controls.
Statistical analysis of all PMOs targeted to myc-luc, or only the 10-base isometric series indicated no correlation (r ϭ Ϫ0.098 [P ϭ 0.54] and r ϭ Ϫ0.46 [P ϭ 0.19], respectively) between inhibition in the cell-free reactions and percent GϩC content. However, an analysis of 37 myc PMOs (Fig. 4), excluding those shorter than nine bases (327, 328, 333, and 334) and those in the 3Ј coding region of Luc (214 and 215), revealed a significant negative correlation (r ϭ Ϫ0.85 [P Ͻ 0.001]) between inhibition of reporter expression and 2°score of the PMO ( Table 1). The 2°score is the fraction of bases in the PMO that are complementary to double-stranded secondary structure within the folded target mRNA (35). An analysis of all 10-base PMOs targeted to myc-luc also showed a significant negative correlation (r ϭ Ϫ0.91 [P Ͻ 0.001]) between inhibition and 2°score.
PMOs targeted to acpP. The effect of PMO was tested on an endogenous bacterial gene, acpP, which is essential for viability (34) and has been used previously to inhibit bacterial growth (11,13). Ten PMOs, which varied in length (6 to 20 bases) and were complementary to the region around the start codon of acpP (Table 1), were added to cultures of AS19, and growth at 37°C was monitored by determining the optical density and viable cell counts. Growth curves were normal for all cultures except the one with the 11-base PMO, which grew significantly (P Ͻ 0.001) slower than the others (Fig. 5A). Viable cells were significantly (P Ͻ 0.001) reduced at 8 h in cultures grown with a PMO of 10, 11, or 14 bases (Fig. 5B). No reduction in CFU was apparent in cultures treated with a PMO of Ͻ10 or more FIG. 2. Effect of PMO length on myc-luciferase expression in cellfree protein synthesis reaction. The same PMOs used for the experiments shown in Fig. 1 were added individually (100 nM) to bacterial cell-free protein synthesis reactions programmed to make myc-luc. After 1 h of synthesis, luciferase was measured. Bars: p, 3Ј truncated PMOs; ■, 5Ј truncated PMOs. Error bars indicate the SD (n ϭ 2).
FIG. 3. Effect of PMO position near start codon. Overlapping, isometric (10-base) PMOs, which are complementary to the region around the start codon of myc-luc, were added (100 nM) to bacterial, cell-free protein synthesis reactions programmed to make myc-luc. After 1 h of synthesis, luciferase was measured. PMO identification number (Table 1)  Cell-free inhibition of acpP. PMOs of various lengths (from 6 to 20 bases) and targeted to acpP (PMOs 62-1, 62-2, 62-3, 62-4, 169, and 379 to 383) were added to bacterial, cell-free protein synthesis reactions programmed to express an acpP-luc reporter. The results were that PMOs 11 to 20 bases in length significantly (P Ͻ 0.01, n ϭ 3) inhibited reporter expression to about the same extent (ranged from 35 to 50% inhibition). PMOs shorter than 11 bases in length inhibited luciferase expression less than 10% (range, 6 to 9%), which was not significantly (P Ͼ 0.1) different than nonsense sequence control PMOs 62-5, 62-7, and 384 (1 to 7% inhibition).
PMO effects in reticulocyte lysate. The 3Ј truncated series of PMOs targeting myc-luc were tested for inhibition of luciferase in a cell-free protein translation reaction made with eukaryotic (rabbit reticulocyte) components. The 20-base PMO inhibited significantly more than the shorter PMO (Fig. 6). There was a sharp decrease in inhibition between the 20-base and the 18base PMO. There was a trend of inhibition that generally favored the longer PMO.

DISCUSSION
Rationales and rules for designing and positioning antisense oligomers within a target mRNA are based on experimental data, theoretical calculations, and bioinformatic screening in eukaryotic systems (8,18,15,28,30,32,33). For PMO and PNA, which inhibit by blocking ribosome function (15), important design considerations include (i) oligomer length and base composition, which affect melting temperature; (ii) target secondary structure, which can interfere with oligomer hybridization; and (iii) position relative to the start codon, which may play a role in blocking ribosome function. Nevertheless, laborious methods of screening remain the rule for designing antisense drugs and identifying target positions (7,27), particularly in bacteria, which is a relatively new application of antisense drug therapy (13, 21; B. L. Geller, J. Deere, and P. Iversen, submitted for publication).
Some rules for designing antisense antibacterial agents have recently begun to emerge. For PNA, a length of 10 to 12 bases is optimal (7); however, optimization of PNA length did not correlate with affinity for target (7). Other factors, such as transfer across the outer membrane of E. coli, may have accounted for much of the improved efficacy of short PNA. It was not apparent that PMOs as short as 10 to 12 bases would be effective, given the uncertainty of the underlying factors that determined the optimal length of PNA.
We have found that PMOs 9 to 12 bases in length effectively inhibit bacterial gene expression, both in pure culture and in a bacterial cell-free protein expression system. At a reduced temperature of 30°C, PMOs as short as seven bases also caused significant inhibition (Fig. 1), probably because the melting temperature is between 30 and 37°C. The sieving effect of the outer membrane and/or the (unknown) mechanism of transport across the plasma membrane probably accounts for the reduction in efficacy of the longer PMOs (Fig. 1). Indeed, longer PMOs inhibited significantly better in cell-free expression systems than in culture. These results differ from PMO inhibition of eukaryotic gene expression, which is not significant with PMOs less than ϳ16 bases in length (30). Although the basis for this difference is unknown, we speculate that differences in eukaryotic and prokaryotic ribosome structure, mechanism of subunit assembly on the mRNA, and/or uptake may be important.
Hybridization position of the PMO was also a factor. Recent evidence shows that peptide-PNA inhibited ␤-lactamase expression only when targeted to either the Shine-Dalgarno sequence (ribosome binding site) or the region around the start codon but not to anywhere else along the entire length of the mRNA (7). Our results show that the region around the start codon is an effective target site. Interestingly, it was not necessary to include the anti-start codon itself within the sequence of the PMO, as exemplified by the inhibition with PMOs 326, 208, or 340 (Fig. 2). In fact, the most effective PMOs that we tested (356, 208, and 169) were targeted slightly downstream of the start codon.
Our results also suggest that the ribosome binding site was a less effective target for PMO inhibition than the region around the start codon, at least for myc-luc. The marginal inhibition of myc-luc expression by PMOs targeted to the ribosome binding site may suggest that these PMOs poorly block binding of the 30S subunit. This result is different than that shown for peptide-PNA inhibition of expression of ␤-lactamase or acpP (7). Perhaps differences in secondary structures of the three targets, differences in association constants of PNA and PMO, or contributions from the peptide used to facilitate entry of PNA would account for this difference in efficacy at the ribosome binding site.
Other characteristics of PMO were analyzed to detect a correlation with inhibition of myc-luc. The GϩC content of PMOs did not correlate with inhibition; however, a significant correlation between inhibition and theoretical secondary structure of the targeted region suggests that base pairing within the targeted region may reduce efficacy of the PMO (Fig. 4).
PMOs targeted to sequences far downstream of the AUG start codon did not inhibit expression. This result is consistent with a previous report that showed PNA lack efficacy at downstream sites (7). Each of our downstream PMOs targeted a different predicted secondary structure. One PMO (PMO 215) was targeted to a region predicted to form a stem-like structure with 10 of its 11 bases paired with a contiguous stretch of complementary bases further downstream. The other downstream PMO (PMO 214) was targeted to a region predicted to form a single-stranded region with all 11 of its bases unpaired. Target secondary structure does not appear to be a factor in the lack of efficacy of PMOs targeted to sequences well downstream of the start codon. However, this interpretation relies upon the limitations of secondary structure predictions based on one algorithm.
The efficacy of short (9-to 12-base) PMOs is somewhat unexpected for two reasons. First, in some reports, PMOs shorter than ϳ16 bases are ineffective in eukaryotic systems (30), whereas in others, PMOs as short as 12 bases cause significant inhibition (15). Second, PMOs have a lower binding affinity than equivalent PNAs and would lose efficacy at a longer length than the equivalent PNA as the length decreased and the T m fell.
The unexpected efficacy of short PMOs toward bacterial targets prompted us to test them in eukaryotic systems. Although the 20-base PMO was more effective in HeLa cells, the 11-base PMO inhibited target expression. This is consistent with the previous results (15) and shows that a PMO as short as 11 bases can cause a significant target-specific inhibition in eukaryotic cells. However, the 11-base PMO was only marginally effective in the cell-free reticulocyte expression system, whereas the 20 base PMO caused a level of inhibition that was comparable to that found in HeLa cells. This suggests that physiological factors not present in the cell-free reactions favored the shorter PMO in HeLa cells. We speculate that cellular uptake and transport to the cytoplasm likely is the factor that favored the shorter PMO.