A genome-wide strategy for the identification of essential genes in Staphylococcus aureus

pT5X xylose-inducible promoter: RNAs stability and abundance

The stability and abundance of an antisense RNA can play a critical role in determining its ability to block the expression of a cognate mRNA (Case et al., 1989). To determine the stability of RNA expressed from the pT5X promoter, cells containing the plasmid pEPSA5 (Fig. 1) were treated with 2% xylose for 10 min followed by rifampicin (200 μg ml⁻¹) treatment. Total cellular RNA was isolated at various times, and the half-life of vector RNA transcribed from the pT5X promoter was determined to be >5 min by Northern analysis (data not shown), suggesting that, for a bacterial RNA, this molecule is relatively stable. In addition, the vector RNA transcribed from the pT5X promoter appears to be so abundant to be visible on ethidium bromide-stained agarose gels (data not shown). Given these observations, high steady-state levels of RNA from this promoter in S. aureus should be achievable.

Shotgun antisense identification of essential genes

An example of the process for shotgun antisense identification of essential genes in S. aureus is described in Fig. 2. Genomic fragments from S. aureus strain RN450 between Å200 and 800 bp were ligated into pEPSA5, downstream of the xylose-inducible promoter pT5X. The genomic library was amplified by passage through Escherichia coli to provide sufficient amounts of DNA, and DNA from the pooled library was then transformed into S. aureus strain RN4220. The inserts of the resulting 3117 xylose-sensitive clones were sequenced, and the identity of the gene source and fragment orientation were determined by BLAST analysis against the annotated genome sequence of the S. aureus MRSA strain, N315 (Kuroda et al., 2001). From this particular screen, 2169 clones were found to contain genomic inserts in an antisense orientation relative to pT5X, representing 658 unique genes. The remaining clones contained inserts representing several classes, including inserts in the sense orientation relative to the promoter, mixed convergent and divergent inserts and intergenic inserts.

Antisense induction resulted in various growth phenotypes ranging from a complete inability to grow to marginal reduction in growth rate (Fig. 3). In some cases, multiple clones targeting the same essential gene showed different levels of inhibition (data not shown). Only clones showing strong growth inhibition upon antisense induction were characterized further.

Saturation analysis

The average size of the inhibitory inserts was 221 basepairs (bp), with 43 bp being the smallest and 854 bp being the largest fragment. Overall, 50% of the recovered fragments were 200–500 bp and 40% were 100–200 bp. Using the average length of the clones found in this screen, 221 bp, and applying the formula n = ln(1-Φ^f)/ ln(1–f), in which n is the number of clones necessary to obtain a single specific clone, Φ is the probability of obtaining at least one of any possible gene sequence, and f is the fraction of the genome contained in an average-sized cloned insert (Zissel et al., 1992), we calculate that by screening 208 226 inserts, we ensured that we sampled the entire genome with greater than 99.99% confidence.

Nearly 50% of the genes inhibited by antisense in this screen were targeted two or more times (Fig. 4). On average, 3.2 genomic fragments expressed antisense to a given gene. In one extreme example, 31 different antisense fragments to the rpoB gene were recovered. These fragments cover discrete regions of the gene, with a notable absence of fragments mapping to the N-terminal portion of the coding region (Fig. 5). We observed this for other genes identified in this screen, suggesting that there may be preferred sites for antisense inhibitory activity, as well as sites that are not favourable substrates for antisense inhibition. Alternatively, such regions may not be amenable to cloning, possibly because transcription of the DNA in that region is deleterious to E. coli.

Demonstration of specific targeting of the antisense RNA

To gain confirming evidence that the primary effect of antisense induction is the specific perturbance of the target mRNA, Real Time PCR (RT-PCR) was used to monitor the fate of target mRNAs and a non-targeted control mRNA early in response to induction of antisense (Fig. 6). The rplQ gene encodes the ribosomal protein L17 of the 50S subunit, and lig codes for DNA ligase, an activity shown to be essential in Bacillus subtilis (Petit and Ehrlich, 2000). RT-PCR revealed that rplQ mRNA decreases by 85% within 15 min of rplQ antisense RNA induction. No effect is observed on the unrelated mRNA of the lig gene. Conversely, lig mRNA decreases nearly 60% after lig antisense RNA induction, with no significant change in the rplQ mRNA.

Comparison of S. aureus antisense inhibited genes to the M. genitalium gene set

The 658 S. aureus genes identified in this particular screen were compared with the gene set of M. genitalium to examine the efficiency of essential gene identification by antisense. Hutchison and colleagues (Hutchison et al., 1999), mapped most of the non-essential genes in M. genitalium by nearly saturating the genome using transposon mutagenesis. Those genes for which no insertions were recovered were inferred to be essential, thus providing a potential minimal set of 256–350 essential genes in the 517 genes comprising the genome of these bacteria. Comparison of the antisense RNA-inhibited S. aureus genes to the full M. genitalium gene set resulted in 168 Staphylococcus genes with clear Mycoplasma homologues (Table 1). Out of these 168 genes, 146 were not disrupted in the exhaustive transposon mutagenesis screen of the Mycoplasma genome, suggesting that these genes are essential in both organisms. Not surprisingly, M. genitalium and S. aureus share essential genes in most major functional categories; cell envelope pro-cesses, metabolism, DNA replication, RNA synthesis, protein synthesis and a set of unknown genes. Mycoplasma genitalium lacks a functional cell wall, and genes involved in cell wall synthesis are not present in this organism. In contrast, many of these genes were represented in the S. aureus antisense hits (data not shown).

Antisense RNA expression leads to selective cell sensitization

We hypothesize that expression of an inhibitory antisense RNA results in a decrease in concentration of an essential protein, and that this may hypersensitize the cell to drugs that specifically inhibit that protein. The S. aureus clone expressing an antisense RNA to the gene, fab (fabF, yjaY), encoding β-ketoacyl-acyl carrier protein synthase, was examined for sensitivity to cerulenin (Schujman et al., 2001), a specific inhibitor of this enzyme. When the antisense RNA to the fab gene was induced with xylose, cells exhibited increased sensitivity to the inhibitor cerulenin (almost 12-fold, Fig. 7). In contrast, the same cells were not significantly sensitized to a variety of antibiotics that inhibit different protein targets, for example, targets involved in the synthesis of DNA, RNA, proteins, cell wall synthesis or even to triclosan (Slater-Radosti et al., 2001), an inhibitor of another fatty acid synthesis enzyme, enoyl-ACP reductase. All strongly inhibitory antisense clones to known antibiotic targets that we have tested show hypersensitivity to their respective antibiotics (data not shown).

Bacterial strains and plasmids

Staphylococcus aureus strains used include RN450 and RN4220 (Novick, 1990) and E. coli strains, DH5α and XL1Blue, which were obtained from Gibco-BRL and Stratagene respectively. Plasmids pRN5543 (Novick, 1991), pLEX5BA (Diederich et al., 1994) and pWH942 (Schnappinger et al., 1995) were described previously.

Construction of pEPSA5 containing the pT5X xylose-inducible promoter

The pEPSA5 S. aureus/E. coli shuttle vector (Fig. 1) contains elements that confer autonomous replication and Cm^R in S. aureus (obtained from the pC194-derived plasmid pRN5548; (Novick 1991). Also included are elements of the multiple cloning site, rrnB T1T2 terminators and the ampicillin resist-ance gene of the plasmid pLEX5BA (Krause et al., 1997), excluding the ColE1 origin of replication, which was exchanged for a NotI cassette containing the lower copy number p15a origin (Diederich et al., 1994). Upstream of the multiple cloning site and terminators is a Gram-positive optimized bacteriophage T5 P_N25 promoter (LeGrice, 1990) in context with the operator sequence for the Staphylococcus xylosis XylR repressor protein (Schnappinger et al., 1995), the gene of which is also included as indicated in the map of pEPSA5.

Media and growth conditions

LB, B2 and SOC media and M9 salts were prepared as described (Sambrook et al., 1989). Where indicated, LB was supplemented with 0.2% glucose (LBG). The concentration of antibiotics as selective agents was 100 μg ml⁻¹ for carbenicillin and 15 μg ml⁻¹ for chloramphenicol, unless otherwise indicated.

Electroporation

All electroporations were conducted using a Bio-Rad GenePulser™. Escherichia coli electroporation was carried out according to the manufacturer’s instructions. Staphylococcus aureus electroporations were as described (Schenk and Laddaga, 1992).

Library construction and screening

Genomic DNA was isolated from RN450 using a kit obtained from GENTRA Systems. A final concentration of 50 μg ml⁻¹ of lysostaphin was added during the lysis step, otherwise the isolation was carried out according to the manufacturer’s directions. Genomic DNA was fractionated by DNase I digestion and then blunt-ended with T4-DNA polymerase as described (Sambrook et al., 1989). The resulting genomic fragments were gel-isolated to enrich for fragments in the range of 200–800 basepairs (bp) using a Qiaquick Gel Extraction Kit (Qiagen) according to the manufacturer’s directions. Resulting genomic fragments were then ligated into the pEPSA5 vector, digested by SmaI and dephosphorylated with calf intestinal alkaline phosphatase (CIP). The resulting ligation was electroporated into E. coli strain DH5α to obtain greater than 1 × 10⁶ individual colonies, which were then combined and subjected to plasmid purification with the use of a Plasmid Maxi Kit (Qiagen). This plasmid library was electroporated into RN4220 to generate approximately 315 000 transformants that were recovered on LBG + 15 μg ml^–1 chloramphenicol agar plates. A total of 250 000 S. aureus transformants were arrayed into 384-well plates using the GeneMachine Gel-2-Well™ robot. Following overnight growth at 37°C, 384-well culture plates were replica-plated with a Genomic Solutions Flexys robot onto inducing (LBG + 2% xylose) and non-inducing (LBG) agar medium. Replica plates were incubated overnight at 37°C, and 4452 clones that did not form colonies in the presence of xylose were chosen for re-testing of sensitivity.

Validating clone sensitivity

The sensitivity of the 4452 clones to xylose induction was ranked according to the number of orders of magnitude of growth inhibition. Overnight cultures were diluted 1:100 and grown for 3 h in 96-well microtitre plates (Costar) with shaking at 37°C. Subsequently, cultures were serially diluted 10-fold eight times in 1× M9 salts in 384-well microtitre plates (Costar). Cultures were then robotically replica-plated (Genomic Solution Flexys) onto inducing and non-inducing (LBG + 2% xylose and LBG) media for overnight incubation. Comparative analysis of the number of dilutions that grew in the absence of induction to those that grew in the presence of induction yielded the log inhibition score. A total of 3117 clones had at least three logs of growth inhibition in the presence of xylose induction and was thus considered growth inhibitory.

Clones of interest were also evaluated for growth inhibition in liquid media. Overnight cultures were diluted 1:100 in LBG + 60 μg ml⁻¹ of chloramphenicol. At an approximate OD₆₀₀ = 0.2 cultures were diluted 1:500 into inducing and non-inducing media in duplicate wells of a 384-well microtitre plate. The OD₆₀₀ was then followed over time in a SpectraMax plus plate reader.

Bioinformatic analysis

The original genomic location of the sensitive clone inserts was determined by BLASTN analysis of the insert sequences against the recently published S. aureus N315 genome (Kuroda et al., 2001). The identity and orientation of the genes covered were determined by comparison of BLAST hit co-ordinates to the published annotation. The protein sequences of the genes revealed as essential by this shotgun antisense screen were compared with BLAST analysis to the Mycoplasma genitalium proteome (Fraser et al., 1995). Gene products that shared more than 25% identity over more than 70% of the length of an S. aureus polypeptide were con-sidered to be potential homologues. Homologues reported here are the ones with the highest degree of similarity.

RT-PCR analysis

RN4220 carrying either an rplQ antisense clone (S1–760 A) or a lig antisense clone (S1–396 A) were grown to mid-log phase and diluted 1:10 into prewarmed Luria–Bertani (LB) broth. Then, 1 ml was transferred to each of six wells of a 96-deep well plate. Cells were grown to 0.1 OD₆₀₀ by incubation on an orbital plate shaker at 37°C. Xylose was then added to three of the wells to a final concentration of 2%. After 15 min, the cells were pelleted, and RNA was isolated from each culture using a modified version of the RNeasy 96 RNA purification system (QIAGEN). TaqMan nucleotide primer/probe sets were designed specifically to measure a small segment (typically 100–200 bp) of either the rplQ or lig target mRNAs by Real Time PCR (RT-PCR) (ABI Prism 7700) using a standard curve method (Applied Biosystems). Data were normalized to a 16S rRNA loading control on a well- by-well basis, and +/− induction data was analysed by an unpaired, two-tailed t-test.

Cell-based assay for target-specific inhibitors

An overnight culture of RN4220 carrying plasmid S1- 1941 expressing antisense to the fab gene encoding β- ketoacyl-acyl carrier protein synthase was inoculated into fresh LB plus 34 μg ml⁻¹ of chloramphenicol, and incubated with shaking at 37°C. Exponential cultures were diluted to a final OD₆₀₀ of 0.002 into two flasks of the same medium containing 0 and 12 mM xylose. After 210 min at 37°C with shaking, the resulting cultures were diluted to an OD₆₀₀ of 0.00022 into 1.1× LB medium containing either 0 or 12 mM xylose, and 45 μl of each cell suspension was dispensed into 384-well microtitre plates (Matrix Tech Corp.) containing 5 μl of antibiotic compounds or solvent per well.

Each plate contained duplicate sets of antibiotics with six twofold dilutions starting from the highest antibiotic concentration listed (a 7-point series). Antisense-induced and non-induced cells were tested against each set within a microtitre plate to enable direct comparisons of the relative potencies of antibiotics. Plates were sealed, shaken and incubated at 37°C. Cell growth was monitored (OD₅₉₅) using an UltraMark microtitre plate reader (Bio-Rad Laboratories) for 15 h. IC₅₀ values (concentration of a compound that inhibits growth by 50%) were generated using GRAPHPAD/PRIZM 3.0 to analyse the 7-point dose–response curves. The fold increase in sensitivity was calculated by dividing the IC₅₀ value in the absence of antisense RNA induction, by the IC₅₀ value in the presence of antisense RNA for a given antibiotic.