Identification of Critical Staphylococcal Genes Using Conditional Phenotypes Generated by Antisense RNA

See allHide authors and affiliations

Science  21 Sep 2001:
Vol. 293, Issue 5538, pp. 2266-2269
DOI: 10.1126/science.1063566


Comprehensive genomic analysis of the important human pathogen Staphylococcus aureus was achieved by a strategy involving antisense technology in a regulatable gene expression system. In addition to known essential genes, many genes of unknown or poorly defined biological function were identified. This methodology allowed gene function to be characterized in a comprehensive, defined set of conditionally growth-defective/lethal isogenic strains. Quantitative titration of the conditional growth effect was performed either in bacterial culture or in an animal model of infection. This genomic strategy offers an approach to the identification of staphylococcal gene products that could serve as targets for antibiotic discovery.

Gene disruption/inactivation technology remains an important tool for identifying essential genes for bacterial growth and pathogenesis. Various strategies have been used in bacterial systems to achieve functional inactivation of gene products. Most of these involve gene knockout methods that use insertional techniques, deletions (e.g., allelic replacement), and point mutation (1–4). Strains carrying null mutations in genes essential for growth are generally not recoverable for further analysis; thus, conditional lethal mutants have been used successfully to identify and characterize genes essential for viability. Temperature-sensitive mutations allow maintenance of the mutant strain because of the conditional nature of the phenotype, and despite its limitations some success has been achieved (5, 6).

Controlled gene expression systems, which allow selective genes to be regulated (i.e., titrated up or down) and thereby functionally analyzed, have the potential to provide more quantitative data on the gene product. Regulated systems have been described forEscherichia coli (7, 8) andBacillus subtilis (9, 10), and are now being developed for bacterial pathogens (11–13), although this approach cannot be applied readily to comprehensive genomic analyses.

Antisense technology has been used effectively to inhibit gene expression in a variety of eukaryotic systems (14–18), but not extensively in bacterial systems, despite antisense regulation being a well-known phenomenon in bacteria (19–23).

Here, we have combined regulated expression and antisense technology for rapid identification and characterization of essential (and growth-defective) genes from the human pathogen, Staphylococcus aureus. We created many defined strains exhibiting conditional growth phenotypes. These strains, in which specific genes are regulated by antisense RNA, were used to analyze the effect of staphylococcal gene products on growth and/or virulence in bacterial culture and in animal models of disease.

A tetracycline (tet) transcription regulatory system that we have adapted for S. aureus was used to regulate gene and antisense expression (13, 22). Induced expression of an antisense RNA to the α-toxin gene essentially eliminates expression of the α-toxin virulence protein, in vitro and in vivo (22). We used the same gene control plasmid vector system to create a random cloned library of small staphylococcal DNA fragments (200 to 800 base pairs) derived by shearing genomic DNA. The blunt-ended, agarose gel–sized and purified fragments were inserted into the plasmid downstream of the tet regulatory region [Web fig. 1 (24)]. The library was transformed intoS. aureus and screened by replica plating colonies in the presence or the absence of anhydrotetracycline (ATc; a weakly antibacterial analog of tetracycline with excellent inducer properties) [Web fig. 1 (24)]. Colonies that grew in the absence of ATc (i.e., without induction) but that were absent or growth-defective in the presence of ATc were selected (Fig. 1A). From 20,000 transformants screened in this manner [∼500 to 1000 colony-forming units (CFU) per plate], 600 independent colonies (3%) were identified as growth-defective or lethal after induction with ATc. These phenotypes were confirmed by restreaking to single colonies, then purifying plasmid from each colony, transforming the plasmid DNA back into a wild-type S. aureus host, and retesting growth in the presence and absence of ATc (Fig. 1B). All the initially selected conditional growth phenotypes were reconfirmed.

Figure 1

(A) Screening growth-defective and lethal colonies. Colonies that grew normally on tryptic soy broth agar containing 5 μl/ml of Erm (TSA-Erm) plate, but appeared defective in growth or unable to grow on TSA-Erm with ATc, were identified and retested by streaking part of each colony onto the TSA-Erm-ATc and onto TSA-Erm after incubation overnight. L, lethal colony; D, defective colony. (B) Confirmation of growth defects and lethal events. Plasmid DNA was purified fromS. aureus strains that carried different antisense DNA fragments and was elec- troporated into WCHU29-competent cells. Electrotransformants were selected on TSA-Erm and duplicated onto TSA-Erm plates containing various doses of ATc (0, 0.5, and 1.0 μg/ml).

To identify the specific cloned DNA fragments that resulted in loss of cell viability, the plasmid inserts were amplified by polymerase chain reaction (PCR) with a common set of parent plasmid-specific primers (25) and DNA sequenced (26). Bioinformatic analysis (27) of the DNA sequences obtained indicated that about one-third of the selected clones (i.e., 1% of total clones plated) contained small, antisense-oriented fragments derived from different single open reading frame (ORF) regions. The remaining two-thirds of the clones contained sense-oriented ORF and non-ORF fragments, antisense non-ORF fragments, a mixture of sense- and antisense-oriented chimeric fragments, as well as sense and antisense fragments spanning multiple ORFs. It is likely that many of the sense-oriented and intergenic fragments represent either dominant-negative interfering sequences and/or important noncoding regulatory regions. However, we chose only those clones (∼200) that represent single ORF antisense constructs for further analysis.

Our data indicated that random insertion of sheared, size-selected, blunt-ended fragments resulted in an overall efficiency of ∼1 in 102 of generating conditional growth-defective events resulting from sub-ORF antisense induction. This efficiency is several orders of magnitude greater than that of any other conditional growth phenotypic selection procedure. Moreover, this procedure allows isolation and maintenance of the conditional strain, as well as rapid identification of the gene segment responsible for the conditional effect.

By this method we identified more than 150 critical staphylococcal genes where antisense ablation led to lethal or growth-inhibitory effects (Table 1). About 40% of these genes are orthologs or homologs of known essential bacterial genes.

Table 1

Characterization of antisense isolates. “Gene” denotes closest Bacillus subtilis homolog; “–” denotes no B. subtilis homolog; gene names in parentheses denote known S. aureus gene; asterisk denotes Haemophilus ducreyi 50S ribosomal protein homolog. Additional mutants can be found in Web table 1 (24).

View this table:

About 30% of the staphylococcal genes that we identified appear to be homologs of bacterial genes with proposed functions, including members of known key pathways (e.g., transcription, translation, and metabolism), but also including genes with presumed biochemical activities (e.g., guanosine triphosphate binding/hydrolysis and methyl/acyl transferases) whose pathway functions remain unclear.

The remaining 30% of staphylococcal genes identified represent apparent critical genes of unknown function. Many of these have homologs in other bacteria, but as yet, little or no functional information is known.

Two distinct phenotypic classes were identified by our selection criteria. One class exhibits a no-growth (i.e., essential) phenotype after antisense induction. The second class exhibits a slow- or reduced-growth phenotype, but does eventually give rise to definitive small colonies (i.e., defective, but not lethal). Several of these defective isolates are known virulence factors, such as fibronectin binding protein (YJ69-1), virulence extracellular factor (YJ15-9), and an ABC transporter (JSB162). The slower growth of these strains in vitro may not have been detected in previous studies or, alternatively, may result from a more generalized growth defect caused by the antisense mechanism of inhibition on secreted gene products. The procedure allows identification of certain important nonessential genes by virtue of their detectable growth-inhibitory effects during selection. We cannot rule out that some of the growth-defective isolates may represent essential genes, because a suboptimal antisense effect would not inhibit expression sufficiently to obtain the lethal phenotype. Likewise, definitive identification of essential genes is problematic for genes configured in polycistronic operons. We do not know the potential effects of defined sub-ORF antisense RNA induction on the expression of genes upstream or downstream of the mRNA region to which the antisense fragment is made. In no case was a known nonessential gene identified as a conditional lethal phenotype in our analysis.

The vector system used to generate the antisense RNAs is both inducible and titratible (22). The growth of the various S. aureus antisense isolates in vitro was induced by ATc in a dose-dependent way (Fig. 2) (relative to the positive control, wild-type strain carrying the parent plasmid YJ335). Hence, quantitative evaluation of the relevance of each gene product to cell viability can be obtained by selective titration of any gene product.

Figure 2

Titration of inhibition of growth by induced antisense RNA. Staphylococcus aureus strains carrying inducible antisense RNA were incubated in tryptic soy broth (TSB) containing 5 μg/ml of Erm with various doses of ATc at 37°C, and incubated with shaking overnight. The density of cells was measured at an absorbance of 600 nm.

To demonstrate the titration of essential genes in vivo, we chose a murine model of hematogenous pyelonephritis, because it represents a localized kidney infection from which bacteria can be readily recovered (22). As a control we used the wild-typeS. aureus strain carrying the parent vector, YJ335. In the presence or absence of ATc induction, infection persisted in the animal, and ∼5 × 105 CFU were recovered from the kidneys at 72 hours after infection (Fig. 3). Similarly, when the same S. aureus strain was used, but carrying different inducible antisense gene segments (YJ2-8 and YJ3-5), ∼5 × 105 CFU were recovered from infected kidneys of animals not treated with the inducer ATc. When the animals were treated with increasing concentrations of ATc, a dose-dependent effect was observed on recoverable bacteria. At the highest ATc dose (500 ng/g), no bacteria were recovered, and the infection completely resolved (Fig. 3). Thus, our defined set of growth-defective/essential gene functions can be studied in the context of a titratible, conditional phenotype in relevant models of infection. This presents the prospect of examining the importance of a gene product when it is switched off after infection has been established.

Figure 3

Recovery of S. aureus from infected kidneys. CD-1 female mice (25 g) obtained from Charles River Laboratories were used for in vivo titration. Staphylococcus aureus strains YJ335, YJ2-8, and YJ3-5 were harvested from 1 ml of stationary-phase culture, washed once with 1 ml of phosphate-buffered saline (PBS), and diluted to an absorbance at 600 nm (A 600) of 0.2. These bacterial suspensions were diluted and plated onto TSA-Erm plates for determination of viable CFU. Five mice per group were infected with about 107 CFU of bacteria through an intravenous injection of 0.2 ml of bacterial suspension into the tail vein by using a tuberculin syringe. Various doses of ATc were given orally in 0.2-ml doses (containing 5 μg of Erm per gram of mouse body weight) to infected mice on days 1, 2, and 3 after infection. The mice were killed by carbon dioxide overdose 2 hours after the last dose of ATc induction. Kidneys were aseptically removed and homogenized in 1 ml of PBS for enumeration of viable bacteria.

In conclusion, the antisense system described offers a comprehensive genomic approach to readily identify and characterize growth-critical gene functions in the clinically important human pathogen, S. aureus. Each gene identified is maintained as one of a collection of conditional growth-defective/lethal isogenic strains. This set of isolates allows titratible phenotypic control over the expression of the gene's function in bacterial culture and in relevant models of infection.

  • * To whom correspondence should be addressed at 1250 South Collegeville Road, UP1345, Collegeville, PA 19426, USA. E-mail: yinduo_ji-1{at}

  • Present address: Pharmacia Corporation, Peapack, NJ 07977, USA.


View Abstract

Stay Connected to Science

Navigate This Article