Novel antimicrobial targets from combined pathogen and host genetics

The identification of drug targets for a given human disease, whether it is mainly environmental or genetic in origin, rests on an understanding of the molecular chain of events that unfold in the disease process. Anatomic pathology, biochemistry, cellular physiology, and pharmacology constitute the main traditional approaches towards identifying potential therapeutic targets. Genetic approaches, by determining the phenotypic consequences of mutations in genes and ordering these genes into functional pathways, are uniquely powerful in identifying novel gene products involved in a disease process. By characterizing mutations that block or reverse the disease phenotype, genetics can provide a direct route to target identification. The wild-type versions of these “suppressor” gene products are potential therapeutic targets, because chemical compounds that phenocopy suppressor gene mutations should similarly block the disease phenotype and thus constitute candidate therapeutic drug leads (1). Such genetic approaches to target identification are most feasible in well-studied model organisms with short generation times that are easily maintained in the laboratory, principally the yeast Saccharomyces cerevisiae, the soil nematode Caenorhabditis elegans, and the fruit fly Drosophila melanogaster. Using model organisms to study homologues of genes causally mutated in human disease is relatively well established. This approach has not often been applied to human infectious diseases, however, as most human pathogens have a highly restricted host range. An exception is the Gram-negative bacterium Pseudomonas aeruginosa, strains of which are pathogenic not only to humans but also to C. elegans, Drosophila, and the genetically tractable model plant, Arabidopsis thaliana (2, 3). In the December 21 issue of PNAS, Darby et al. (4) describe a genetic approach toward the identification of potential therapeutic targets for P. aeruginosa infection.

P. aeruginosa is a common bacterium in soil and water worldwide and an opportunistic pathogen in humans, causing acute and chronic infections in patients with compromised immunity, severe burns, and cystic fibrosis. A variety of virulence factors, including secreted enzymes as well as toxic chemicals, contribute to the pathogenesis of P. aeruginosa infections (for review see ref. 5). The production and secretion of most known virulence factors is increased at high versus low bacterial cell density, through a cell-to-cell signaling mechanism known as “quorum sensing” that links cell density to gene expression (for review see refs. 6 and 7). In P. aeruginosa two quorum-sensing systems have been described in some detail, las and rhl. Each is composed of two components: an inducer locus, lasI or rhlI, that controls the synthesis of a diffusible, cell-permeant, pheromone and a responder locus, lasR or rhlR, that encodes a transcription factor that binds to and is activated by the pheromone. Activation of LasR enhances the expression of genes encoding secreted proteases, phospholipases, ADP-ribosylating enzymes, and genes that control secretion of virulence factors as well as LasI (thus providing positive feedback to LasR activation) and RhlR. Activation of RhlR controls some genes in parallel to LasR and, in addition, activates the expression of loci responsible for the production of toxic chemicals, including hydrogen cyanide, rhamnolipids, and phenazines. LasR and RhlR are further subject to higher-level positive and negative regulators (8, 9).

Two recent PNAS papers extend and expand our knowledge of quorum sensing. First, Whiteley et al. (10) report the isolation of 47 mutants in 39 different genes that are up-regulated by treatment with P. aeruginosa pheromones. The authors note that their screen is far from saturated and speculate that 1–3% of the total 5000–6000 genes of P. aeruginosa are controlled by quorum sensing. Second, Pesci et al. (11) describe a third cell-to-cell signaling molecule produced by P. aeruginosa, 2-heptyl-3-hydroxy-4-quinolone (a chemical unrelated to the las and rhl pheromones, which are acyl homoserine lactones), that regulates the expression of virulence factors. Synthesis of this new signaling molecule requires both LasR and RhlR. These studies highlight the potential value of targeting the quorum-sensing regulatory hierarchy for new anti-infective therapies (7, 12).

The complexity and long life cycles of mammalian models have limited the understanding of host factors involved in microbial pathogenesis. To overcome these limitations, Rahme, Ausubel, and colleagues pioneered the use of model organisms to identify “universal” virulence factors for P. aeruginosa pathogenicity. Infection of Arabidopsis thaliana with P. aeruginosa strain PA14, a pathogen derived from a human infection, caused soft-rot, chlorosis, and eventual leaf collapse, and the same strain caused lethal infections in a mouse full-thickness skin burn model (2, 13). Three recent papers from this same laboratory describe PA14 pathogenesis in C. elegans (3, 14, 15). In the nematode, two types of toxicity were observed. Exposure of L4 stage larvae, but not adults, to PA14 grown on nutrient-rich media at high osmotic strength killed the nematodes in 4 hr (“fast killing”), whereas exposure of either L4 or adults to PA14 grown on less rich or minimal media killed after 1 to 2 days (“slow killing”). Screening for Pseudomonas mutants with reduced pathogenicity in Arabidopsis or in slow or fast killing of C. elegans identified mutations in 23 genes, 19 of which also reduced virulence in the mouse burn model (refs. 13–15; for review see ref. 16). Because most of these genes were identified by only a single mutation, this screen also appears to be far from saturated.

Now, Darby et al. (4) describe a third model of Pseudomonas pathogenesis in C. elegans. Exposure of adult C. elegans to P. aeruginosa strain PAO1 (a standard laboratory strain isolated some 50 years ago from a human infection) inhibited feeding of the nematodes within seconds to minutes, more gradually slowed and disorganized locomotory behavior and caused paralysis and death after several hours. These effects were not observed when C. elegans was exposed to PAO1 derivatives bearing a mutation in the quorum-sensing gene lasR or rhlR. Mutations in genes encoding secreted enzymes controlled by quorum sensing, however, had no effect in this model, suggesting that one or more of the toxic chemicals regulated by quorum sensing is responsible for neuromuscular toxicity. Several outstanding questions remain regarding the mechanism of PAO1-induced paralysis: First, does toxicity result from the action of a single toxin or is more than one component required? And what is the chemical nature of the toxin(s)? Second, what genes are responsible for production of the toxin(s)? And what are their functions? Given that the genome of P. aeruginosa (PAO1) is completely sequenced and that powerful screens for virulence factors and for genes controlled by quorum sensing are available, the identity of the genes responsible for toxicity in this model is likely to be forthcoming soon. The time and effort required to determine the chemical nature of the toxin(s), if there are no breakthrough insights from the genetic studies, are more difficult to predict.

Darby et al. (4) further describe nematode mutants that resist the lethal effects of exposure to PAO1. Two strongly resistant strains were identified from among 8,000 mutagenized genomes. Genetic analysis revealed that both strains carry recessive mutations in a known gene, egl-9, previously identified in screens for egg-laying-defective mutants (17). Two additional egl-9 alleles isolated on the basis of their egg-laying phenotype are also resistant to P. aeruginosa-induced paralysis. Darby et al. (4) cloned the egl-9 gene and showed that the molecular lesions in the four mutations most likely generate nonfunctional EGL-9 proteins. These results are consistent with a model in which exposure to P. aeruginosa PAO1 induces paralysis and death by excessive or inappropriate activation of EGL-9. Because egl-9 function appears to be nonessential for C. elegans (in the comfortable confines of the laboratory), strains that lack egl-9 activity are both viable and resistant to PAO1-induced paralysis. It would be interesting to know whether egl-9 is the only gene that can mutate to give resistance to Pseudomonas-induced toxicity. The experiments performed so far do not survey a sufficiently large number of genomes to reliably answer this question.

EGL-9 is a novel 723 amino acid protein with no obvious signal sequence or other readily identified functional motifs. EGL-9 is homologous to SM-20, a 355 amino acid protein identified as an “immediate early gene” induced upon treatment of rat aortic smooth muscle cells with a mixture of growth factors (18). SM-20 appears to be a cytoplasmic protein localized to multiple types of muscle cells, scattered epithelial cells, and neurons (19), which resembles the expression pattern of EGL-9 in C. elegans (4). The function of SM-20, however, remains unknown. Furthermore, the amino acid similarity is limited to a domain in the C-terminal half of the predicted EGL-9 protein. Whether alternative splicing of egl-9 could produce an SM-20-sized protein product, or whether SM-20 or another mammalian homologue contains sequences homologous to the N-terminal half of EGL-9, remain to be seen. Of note, a preliminary search (L.X.L) of the emerging but unannotated Drosophila genome reveals fly genes with homology to EGL-9 and SM-20. Clearly, the completion of genomic sequencing in metazoans besides C. elegans will ease the identification and assignment of homologous and orthologous genes.

The genetic approach to selection of human therapeutic targets involves, as a start, the identification of mutations that suppress a surrogate disease phenotype in a genetically tractable organism. Loss-of-function mutations of egl-9 block P. aeruginosa-induced paralysis of C. elegans. Might inhibition of an EGL-9 homologue, possibly SM-20, similarly block Pseudomonas-induced pathology in infected humans? Accumulated knowledge suggests that P. aeruginosa pathogenesis in human disease involves multiple virulence factors and pathogenic mechanisms. Accordingly, a therapeutic strategy focused on a single mechanism is not likely to provide a completely satisfactory treatment. Nonetheless, the possibility that it might provide valuable supportive or ancillary therapeutic effects remains to be tested.

Traditionally, antimicrobial drug discovery has entailed screening candidate compounds directly on target microorganisms. The clear scientific advance represented by the dual model organism infection systems described above is the capacity to dissect specific interactions by using both pathogen and host genetics. These studies can be expected to yield both novel and “broad host spectrum” bacterial virulence factors and toxins, and to help elucidate mechanisms of toxicity and identify targets of toxin action. On the basis of these studies, one can envision microtiter plate-based high-throughput screens using a pathogenic Pseudomonas strain and susceptible C. elegans nematodes to identify chemical compounds that “cure” host pathology regardless of whether the compound blocks bacterial virulence or a deleterious host response. Of more direct utility, however, may be an enhanced focus on identifying chemical suppressors of P. aeruginosa virulence factor expression or secretion as candidate novel antibiotics, with Arabidopsis, C. elegans, and the mouse as confirmatory models. Finally, numerous microbial toxins have been invaluable as selective probes of biological function, such as tetanus, botulinum, pertussis, cholera, and diphtheria toxins. Because egl-9 appears to regulate neuromuscular activity, dissection of the Pseudomonas PAO1–C. elegans egl-9 interaction may reveal as much, if not more, about the control of muscle contraction or neuronal signaling as it reveals about microbial virulence.