Review

Inhibition of toxin function

Many pathogenic bacteria cause damage to host tissues through the release of toxins, which are proteins that act to perturb host cell functions and may ultimately result in host cell death. Thus, an obvious approach to inhibiting bacterial virulence is disruption of toxin function, which can occur in a direct manner by inhibition of the toxin activity itself, or in an indirect manner, by modulating the host response to the toxin. In fact, direct inhibition is the basis for the historical use of antitoxins (antibodies) against toxins such as diptheria, botulinus, tetanus and other toxins. Recently, much effort has been focused on inhibiting the effects of the three proteins that comprise anthrax toxin: lethal factor (LF), edema factor (EF) and protective antigen (PA) (Fig. 3a).

Though each toxin component alone is nontoxic, the pairing of either LF or EF with PA results in characteristic toxin activity that is ultimately responsible for the pathology of the disease². Much is already understood about toxin entry into mammalian cells. First, individual PA monomers diffuse to the surface of mammalian cells where they are proteolytically cleaved by host proteases. Cleaved PA spontaneously oligomerizes into heptamers, which can bind either LF or EF. PA-EF PA-LF complexes are then endocytosed and trafficked to the endosome, where a drop in the pH triggers a conformational change in PA that converts it to a transmembrane pore. Ultimately LF and EF are translocated through PA's pore into the host cytosol where they exert their respective toxic effects (Fig. 3a). EF is a calmodulin-dependent adenylate cyclase whose action is known to result in prolonged increases in cyclic AMP levels, but it is otherwise poorly understood. The exact mechanism by which LF, a Zn²⁺ protease, exerts its cytotoxic effect in vivo is also not clearly understood. Genetic evidence suggests that LF-mediated cell death is dependent on susceptible alleles of the Nlrp1b gene, as well as the gene encoding caspase-1 in murine macrophages³. LF can also cleave the N termini of several mitogen-activated protein (MAP) kinase kinases^{4, 5}. However, it is unclear how LF-mediated MAPKK cleavage is related to cytotoxicity, though small molecules that activate MAP kinase cascades have been shown to protect mouse macrophages from LF-induced cell death⁶. Apart from causing cell death, LF can also paralyze actin-based motility in neutrophils—an effect that has been correlated with decreased levels of Hsp27 and p38 MAP kinase phosphorylation in LT-treated, cultured neutrophils⁷.

Regardless of the precise mechanism of activity of LF, it has certainly been the focus of great attention as a target for new antimicrobials⁸. In a search for small molecules that would protect the population from anthrax, Merck has identified a hydroxymate (LFI) that inhibits LF protease activity and promotes cellular survival in a macrophage cytotoxicity assay⁹. LFI (Fig. 3b) binds the active site of Bacillus anthracis LF and offers complete protection from spore infection when administered to mice in combination with ciprofloxacin 66 h postinfection in a model in which ciprofloxacin alone only offers 50% protection from lethality⁹. Along similar lines, a mixture-based peptide library was used to determine the optimal peptide substrate sequence for LF, which was then used to design peptide analogs that inhibit LF activity in vitro and protect macrophages from LF-induced cytolysis¹⁰.

Compounds that inhibit anthrax toxin are not limited to targeting LF or EF. PA is also a potential target that could be inhibited in multiple ways. Strategies include inhibition of binding of PA to its host receptor, processing of PA by host proteases, binding of processed PA to LF or EF, or translocation of LF or EF by physically blocking the PA heptamer pore⁸. Recent work indicates that a phenylalanine clamp controls protein translocation through the heptameric PA pore and that this clamp can be targeted with small molecules that block the pore¹¹. Alternatively, one could interfere with toxin translocation by inhibiting formation of the pore itself by inhibiting endosome acidification^{12, 13} or by inhibiting PA heptamer assembly. Cisplatin seems to prevent both LF and EF toxicity by inhibiting heptamer assembly, and simultaneous administration of cisplatin with a lethal dose of anthrax lethal toxin has been shown to be protective in rodent models, though a delay in cisplatin administration was shown to be ineffective¹⁴. This result should be contrasted with the result obtained in ciprofloxacin and LFI combination therapy, which works even in delayed administration. This difference underscores the concept that the specific mechanism of virulence inhibition may determine the efficacy of a particular small molecule in targeting virulence at different stages of infection (see below).

Rather than directly inhibiting toxin function, one could block the downstream effects of the toxin by targeting host proteins. For example, in secretory diarrheas and cholera, Cl^- secretion is central to intestinal fluid secretion (the disease pathology); thus inhibitors of Cl^- channels may have therapeutic value in blocking these diarrheas. A screen for inhibitors of the cystic fibrosis transmembrane conductance regulator (CFTR) protein¹⁵, a cAMP-activated Cl^- channel that is defective in individuals with cystic fibrosis, found that compounds of the 2-thioxo-4-thiazolidinone class inhibit the CFTR protein. The most potent thiazolidinone compound was evaluated for its ability to inhibit mouse intestinal fluid secretion after oral administration of cholera toxin. A single intraperitoneal injection of the CFTR inhibitor decreased the level of fluid secretion by more than 90% after administration of cholera toxin. Although it is still unclear whether administration of this compound can positively influence the resolution of infection with Vibrio cholerae, this finding underscores the notion that new therapeutics aimed at targeting virulence need not be absolutely restricted to pathogen proteins.

Targeting bacterial toxin delivery

In addition to targeting bacterial toxin function, one could also inhibit virulence by interfering with appropriate delivery of the toxin to its site of action. This principle has been applied for several decades in the treatment of the antibiotic-associated diarrheal disease caused by Clostridium difficile¹⁶. Cholestyramine simply binds the clostridial toxins, preventing their delivery or accessibility to the intestinal epithelium and thus blunting their toxic effects. Cholestyramine binds toxin B (a cytotoxin) and likely toxin A (an enterotoxin) as well, though its use has been limited to relatively mild cases of the disease.

A more recent application of this principle is prevention of toxin (or effector) delivery by inhibiting bacterial secretion systems. Many proteins involved in bacterial secretion are specific to prokaryotes and are therefore rational candidates to target with novel antimicrobials. Recently, there has been interest in targeting the type III secretion system (T3SS) common to Yersinia spp., Pseudomonas aeruginosa, pathogenic Escherichia coli, Shigella spp., Salmonella spp. and Chlamydia spp. The T3SS is a syringe-like apparatus that facilitates the injection of bacterial effectors from these species directly into the host cytosol (Fig. 3a)¹⁷. Depending on the species, T3SS effectors have been implicated in perturbing a variety of host cellular processes, including cytoskeletal dynamics, gene expression, cell cycle progression and apoptotic cell death programs.

Chemical screens for inhibitors of the T3SS in Yersinia pseudotuberculosis identified acylated hydrazones of different salicylaldehydes^{18, 19}. One compound was found to directly target the Y. pseudotuberculosis T3SS, thereby preventing effector molecule translocation and attenuating the pathogen in a cell culture model of infection without interfering with in vitro growth¹⁹. Compounds of this class have also been found to inhibit intracellular replication and infectivity, and translocation of the T3SS effectors IncG and IncA during early and middle stages of infection of Chlamydia trachomatis (Fig. 3c)²⁰. They have also been shown to interfere with the life cycle of Chlamydia pneumoniae²¹. Very recently, acylated hydrazones of different salicyclaldehydes have also been shown to be effective at inhibiting both the secretion of T3SS effectors and the invasion of Salmonella enterica serovar Typhimurium into cultured epithelial cells²². However, compounds of this class were only able to suppress T3SS-dependent secretory and inflammatory responses in a bovine intestinal loop model of S. enterica serovar Typhimurium infection if the bacteria were first preincubated with the compounds. Simultaneous infection and administration of these compounds was ineffective in preventing fluid accumulation and neutrophil influx into the infected intestinal loops²². Thus, while it seems that acylated hydrazones of different salicyaldehydes may effectively target the T3SS of many different pathogens, more work will need to be done to evaluate whether this class of compounds has broad-spectrum therapeutic value.

Targeting the regulation of virulence expression

In addition to targeting toxin function or delivery, disrupting the regulatory mechanisms that result in virulence expression is attractive because it would prevent the formation of toxin. To some degree, the use of the protein synthesis inhibitor clindamycin, in addition to a cell wall antibiotic such as penicillin, in the treatment of streptococcal toxic shock syndrome to prevent the production of toxin (a phenomenon known as the "Eagle effect") is a clinical application of this principle²³. Recent advances in our understanding of virulence regulation have identified many different regulatory steps that could be targeted.

Most efforts to inhibit the regulation of virulence factor expression have focused on interfering with quorum sensing (Fig. 3a), a mode of bacterial communication used by multiple bacterial species to regulate processes such as bioluminescence, antibiotic synthesis, biofilm formation and virulence factor expression as a function of population density²⁴. Quorum sensing is a phenomenon whereby bacteria can sense their population density by releasing diffusible signaling molecules that accumulate in the population and surrounding environment. These signaling molecules increase in concentration as the bacterial population expands until a critical threshold concentration is reached. At critical threshold concentrations (when the bacteria are 'quorate'), quorum sensing signaling molecules can influence the expression of a variety of different genes, including the expression of genes involved in virulence. Teleologically, the bacteria may begin to express virulence factors only when they have reached a significantly large population size to cause an 'effective' infection. Thus, by interfering with quorum sensing pathways, one could inhibit a bacterial population's ability to monitor the number of cells within that population and thereby interfere with quorum sensing activation of virulence factor expression.

In many Gram-negative bacteria, quorum sensing is mediated by acylhomoserine lactone molecules (AHLs) that are synthesized and recognized by quorum sensing circuits composed of LuxI and LuxR homologs (Fig. 3a)²⁴. LuxI homologs synthesize AHL molecules from the common metabolic intermediate S-adenosylmethionine (SAM) and an acyl-acyl carrier protein. At critical threshold concentrations, AHL molecules are recognized by their cognate transcriptional activator (LuxR homologs), which in turn regulates the transcription of genes associated with virulence. Thus, one could inhibit AHL-mediated quorum sensing by inhibiting the enzymes that synthesize quorum sensing molecules (LuxI homologs). In vitro, the activity of the P. aeruginosa LuxI homolog RhlI can be inhibited with analogs of SAM such as S-adenosylhomocysteine, sinefungin and S-adenosylcysteine²⁵, resulting in the inability to synthesize quorum sensing molecules. Though SAM is a common metabolic intermediate, it is possible that improved analogs of SAM could be generated that could interfere with LuxI-mediated AHL synthesis but not host metabolic processes that rely on SAM.

Alternatively, one could inhibit quorum sensing by interfering with the concentration of the AHL signaling molecules through degradation. For example, Gram-positive Bacillus species produce acylhomoserine lactonase, an enzyme that hydrolyzes the lactone ring of AHLs, thereby rendering them unable to mediate signaling^{26, 27}. Tobacco plants engineered to express AHL-lactonase show an enhanced resistance to Erwinia carotovora infection²⁷, which demonstrates that degradation of the AHL signal is a viable approach to attenuating infection with E. carotovora. Mammalian sera also show strong AHL lactonase activity that is similar to the activity of paraoxonases²⁸, which themselves have lactonase activi-ty against AHLs in vitro²⁹. Thus, interference with quorum sensing by increasing degradation of AHL molecules may be a natural phenome-non used by multiple species. Though it has not been exploited in antimicrobial discovery, it represents an alternate mechanism for targeting virulence.

Finally, one could inhibit AHL-mediated quorum sensing by interfering with the recognition of and subsequent transcriptional response to the signaling molecules by LuxR homologs. For example, a number of structural analogs of AHLs have been shown to inhibit the expression of quorum sensing regulated genes and virulence factor production as well as biofilm formation in P. aeruginosa^{30, 31, 32}. In particular, one class of structural analogs of AHLs, halogenated furanones, have been shown to have quorum sensing inhibitory properties³³ and seem to function by accele-rating the turnover of LuxR homologs³⁴. In E. carotovora, halogenated furanones have been shown to inhibit the production of carbapenem, which is regulated by the LuxR homolog CarR³⁵. In P. aeruginosa, halogenated furanones inhibit production of exotoxins, repress the expression of quorum sensing regulated genes, and increase the susceptibility of P. aeruginosa biofilms to tobramycin in vitro^{36, 37}. In vivo, administration of synthetic halogenated furanones promotes clearance of P. aeruginosa from the lungs of infected mice^{37, 38} and increases the survival time of mice in a lethal P. aeruginosa lung infection model³⁸. However, AHL molecules themselves as well as other structural variant lactones are base-labile and are substrates for mammalian paraoxonases^{28, 29, 39}; they are thus likely to be suboptimal for therapeutic use. Halogenated furanones are also too reactive to be used therapeutically. For this reason, the recent identification of a triphenyl compound that is a potent agonist of quorum sensing in P. aeruginosa is promising in that this compound is structurally unrelated to AHL molecules and therefore may be useful as a scaffold for developing future quorum sensing inhibitors that are more stable and perhaps less reactive in vivo⁴⁰.

Quorum sensing is not a phenomenon that is unique to Gram-negative pathogens. Staphylococcus aureus strains make autoinducing peptides (AIP) that function as quorum sensing molecules in these bacteria. In gene-ral, Gram-positive bacteria synthesize a precursor protein that is proteolytically cleaved to form the processed quorum sensing signaling peptide. This peptide is then actively transported extracellularly and can then be recognized by a histidine-sensor kinase protein of a two-component regulatory cascade. Upon recognition of the peptide autoinducer, the two-component regulator activates a phosphorelay response, ultimately resulting in activation of a quorum sensing transcriptional regulator. Different S. aureus strains produce peptides whose sequences vary slightly from those produced by other strains. Thus, a given peptide sequence can activate its own histidine kinase two-component regulatory cascade while antagonizing the histidine kinase two-component regulatory cascades in other strains of S. aureus⁴¹. This has been exploited in a mouse model of S. aureus–mediated abscess formation, in which administration of inhibitory AIP to mice during the first 3 h of S. aureus infection inhibited the pathology^{42, 43}. Clearly, such a therapeutic approach would require a small molecule that could inhibit the quorum sensing systems of all S. aureus strains.

Apart from quorum sensing, transcriptional regulators that coordinate the expression of genes involved in adhesion, toxin production and secretion are themselves potential targets for future antimicrobials. A recent example is the small molecule virstatin (Fig. 3d), which was identified in a screen for small-molecule inhibitors of the V. cholerae cholera toxin promoter⁴⁴. Virstatin inhibits the transcriptional regulator ToxT, thereby preventing the expression of two critical virulence factors in V. cholerae: the toxin-coregulated pilus, which is involved in attachment to the intestinal epithelium, and cholera toxin. In vivo, administration of virstatin protected infant mice from intestinal colonization with V. cholerae⁴⁴. This work also demonstrated that late administration of virstatin (12 h after infection) still results in a 3-log drop in the ability to recover bacteria from the mouse intestine, which shows that inhibition of virulence expression, even in established infection, can still have potential therapeutic value.

Another example of inhibition of virulence regulation and expression was found in a screen for inhibitors of the T3SS in enteropathogenic E. coli (EPEC)⁴⁵. In this study, the imine adduct from the condensation of a halogenated salicylaldehyde and 3-aminoacetophenone seemed to decrease the transcription of T3SS effectors and components of the T3SS machinery without affecting growth or motility. Thus, toxin delivery can be inhibited by novel therapeutics targeting not only structural components of the secretion apparatus but also transcriptional regulators of secretory systems.

Inhibition of adhesion

Equally important for establishing infection is adhesion of the bacterial cell to the host. The human body has evolved multiple lines of defense to prevent the majority of bacteria from adhering to host tissues, including continual shedding of epithelial cell surfaces, mucus lining epithelial cell surfaces of the respiratory, intestinal and reproductive tracts, and direct inhibition of adhesion of bacteria to host surfaces by antibody molecules that line these tracts and perfuse host tissues. These multiple lines of defense erected by the body underscore the importance of preventing bacterial adhesion to the host. Moreover, because proteins involved in bacterial adhesion are specific to prokaryotes, like bacterial secretion systems, they are rational targets for novel antimicrobials.

A potentially broad class of antimicrobials that target adhesion are called 'pilicides'. Pilicides are aimed at inhibiting the formation of pili or fimbriae (Fig. 3a), which are hair-like projections protruding from bacterial cells that facilitate adhesion. Pili consist of repeating subunits of immunoglobulin-like domains wherein the N terminus of one subunit is donated to complete the immunoglobulin fold of its neighbor subunit; these structures are formed by donor-strand complementation via the chaperone usher pathway⁴⁶. In this pathway, the chaperone protein binds pilin subunits by donating its edge -strand to complete the folding of the pilin subunit's immunoglobulin fold. Chaperone–pilin complexes then traffic to outer-membrane usher channels where the pilin fiber is formed by donor strand exchange between the chaperone and the adjacent subunit in the growing pilin fiber. So far, efforts at developing pilicides have focused on inhibiting the periplasmic chape-rone proteins of the chaperone usher pathway in uropathogenic E. coli (UPEC). Bicyclic 2-pyridones (Fig. 3e) and N-substituted amino acid derivatives have been shown to competitively inhibit binding of chaperones to pilin subunits by surface plasmon resonance⁴⁷. In vitro, bicyclic 2-pyridones have also been shown to inhibit hemagglutination and biofilm formation in laboratory and clinical E. coli strains, and ex vivo they have been shown to inhibit adhesion of the bacteria to bladder carcinoma cells by 90%⁴⁸. It has been suggested that pilicides may have broad-spectrum activity due to the conservation of both chaperone structure and the chaperone-usher pathway⁴⁹.

Virulence inhibitors as therapeutic candidates

Novel therapeutics that target virulence rather than simply in vitro cell growth would both supplement and add diversity to our current antimicrobial armamentarium, with the caveat that this type of strategy may not be effective in immunocompromised individuals who lack the ability to clear the 'disarmed' pathogen. However, for the appropriate patient population, a tantalizing advantage of targeting virulence is the potential of this approach to impose weaker selective pressure that would be less likely to foster the growth of antibiotic resistance, in contrast to therapeutics targeted at straightforward viability. If the function of a given virulence factor is unrelated to its viability within the host, there should be no in vivo selection for the rare, resistant mutant to an inhibitor of that factor. The difficulty is that it is unclear which virulence factors fit this criterion, and thus each individual case will need to be examined carefully.

Even if an inhibitor targets a virulence factor required for in vivo viability, the circumstances under which resistance could arise against the specific inhibitor relative to conventional antibiotics are more limited, because antivirulence inhibitors are often pathogen-specific and selection must occur in vivo. Thus, the window of opportunity during which resistance could be selected for is more narrow, equaling the length of an individual infectious cycle. Additionally, use of antivirulence drugs could also have an impact on the development of resistance to our current broad-spectrum antibiotics by offering an alternative therapy in certain cases, thus reserving our current broad-spectrum agents to a more limited set of cases in which they are most needed. Finally, because of their narrow spectrum and different mechanisms of action, therapeutics that act to inhibit virulence rather than cell growth also have a significant advantage in helping to preserve normal and potentially beneficial members of the normal human microbiota. This type of strategy would likely have a significant impact on the morbidity associated with shifts in host normal flora after antibiotic treatment.

The exact therapeutic role of antimicrobials that target virulence is as yet unclear. Antivirulence therapies have the potential to function effectively when used alone, when used in combination therapy with antibiotics, or simply as a prophylactic treatment. It is premature to rule out the use of antivirulence drugs simply because they are not bacteriocidal, as this would underestimate the role of host immunity. A long-standing dogma suggests a preference for bacteriocidal over bacteriostatic antibiotics. However, there is actually little clinical evidence to suggest that this dogma has significant bearing on the resolution of infection, with other variables such as tissue penetration and bioavailability potentially having a greater impact. Thus, a new paradigm for antimicrobial therapy redefines the goal to be simply the tipping of the balance in favor of the host, thus enabling it to control infection, rather than complete in vivo killing of a pathogen by the drug itself.

Antivirulence drugs could have a role in prophylactic treatment during an epidemic, a bioterrorist threat, or in select populations (for example, travelers) for several reasons. They would not engender resistance in the population toward conventional antibiotics, they would not alter the natural microbiome of the host, and they may work optimally to prevent infection, as evidenced by cases in which the drugs work best when administered with the pathogen rather than in established infection. It is likely that the unique mechanism of action for each antivirulence therapeutic will ultimately determine its utility as a prophylactic, mono- or combination therapeutic agent.

Because virulence factors are required at different times during infection, the definition of the time window in which a given inhibitor will be effective will likely be mechanism- and thus inhibitor-specific. For example, as described above, in V. cholerae infection, inhibition of the transcriptional regulator ToxT, which results in the downstream inhibition of cholera toxin synthesis, has therapeutic efficacy in established infant mouse colonization 12 h after infection⁴⁴. On the other hand, cisplatin inhibition of LF translocation into the host cytosol by blocking PA heptamer assembly is completely ineffective in mice if the inhibitor is administered after anthrax toxin inoculation¹⁴. In contrast, if the mechanism of protection is due to inhibition of the proteolytic activity of LF, delayed administration of LFI is effective (granted, in combination with ciprofloxacin) in rescuing infected mice⁹. Thus, the therapeutic efficacy of inhibiting virulence factors needs to be evaluated on an individualized, mechanistic basis, and efforts will need to be invested in determining the in vivo time window in which a protein involved in virulence may be effectively targeted. This analysis would provide insight into the timing of different virulence programs and their requirement throughout the course of infection.

Though some virulence inhibitors such as pilicides and T3SS inhibitors have the potential to target a wider spectrum of bacteria, other virulence inhibitors that disrupt mechanisms specific only to single pathogens or that target bacterial proteins that do not have structurally similar orthologs in many clinically relevant bacterial species will be exquisitely narrow-spectrum by definition. Although this new paradigm could be the solution to the resistance crisis we currently face, it raises two problems that are not insurmountable but that need to be addressed. One problem is diagnostic; the other is economic.

The utility of therapeutic intervention with narrow-spectrum antivirulence antimicrobials will be very dependent on the clinician's ability to precisely diagnose the genus if not the species of bacterial infection in a patient in order to select the correct antimicrobial therapeutic agent. Thus, antimicrobials that target virulence will need to be developed hand in glove with new diagnostic tests that will allow the clinician real-time diagnosis. Certainly these technological problems are challenging, but they are currently the focus of great interest, with efforts intensified particularly in the area of biodefense. In addition to real-time identification of the culprit pathogen, a related issue is determining the susceptibility of the culprit pathogen to virulence inhibitors. Definition of conventional antibiotic susceptibilities has become a well-tuned science that includes Kirby-Bauer (disc diffusion) tests and broth dilution measurements. In contrast, determining susceptibilities to a virulence inhibitor cannot use simple cell growth as a sensitivity indicator. Assays exist and can be developed for determining susceptibility in vitro, such as the screening assays used to identify these inhibitors or ones engineered to include various gene reporter readouts. However, these assays too will need to be individualized for each specific inhibitor.

Perhaps the greatest challenge of changing to a new paradigm is not technological but economic. If compounds that target virulence are effective against a much more narrow range of bacteria, the economic incentive for a pharmaceutical company to develop that compound into a drug, with the accompanying diagnostic and susceptibility tools, is even lower than current economic incentives to develop wide-spectrum antibiotics. If we are to avoid plummeting into a postantibiotic era in the near future, clearly we must attack not only the technological hurdles facing new antimicrobial development but also the economic hurdles.

Future approaches to new antimicrobial development

To supplement our dwindling antimicrobial arsenal, a broader range of targets than the set of in vitro essential genes that are typically targeted by conventional antibiotics must be considered. We propose that a new paradigm for antimicrobial development should be based on targeting gene functions that are required in vivo to cause disease. This strategy includes targeting genes that are essential for causing virulence as well as those essential for in vivo viability. As the ability of pathogenic bacteria to both cause damage and survive within the host is required for causing disease, disrupting either of these processes could be exploited therapeutically.

In vitro and in vivo bacterial essential gene functions are distinct. The growth environment within a host is unlikely to be the same as the artificial ones induced in a laboratory, and therefore the genes required for viability will likely also differ. This concept was illustrated in a transposon site hybridization study conducted in Mycobacterium tuberculosis–infected mice. The study revealed that a different set of M. tuberculosis genes, representing 5% of the M. tuberculosis genome, were required for in vivo survival than were required for in vitro growth⁵⁰. Current antibiotic discovery approaches have focused only on in vitro essentials, and thus the realm of in vivo essentials that could be targeted has not yet been sufficiently explored.

One approach that has the capacity to identify new therapeutics that target either bacterial in vivo essential gene functions or the host itself is chemical screening of whole-organism infection models. Whole organism–based screening has the advantage of being able to identify small molecules that are more likely to be permeable to the cell, effective at elici-ting the desired phenotype (such as attenuation of infection), unlikely to have gross toxic side effects, and have acceptable pharmacokinetic profiles, at least in the model host, which may or may not translate from host to host⁵¹. Finally, whole organism–based screening has the capacity to define which proteins are, in fact, 'drug-targetable'. Thus, whole-organism screening has the potential to leapfrog over some of the major hurdles commonly encountered after typical in vitro target-based screens. Practically, of course, in order to identify candidate novel antimicrobials in a whole-organism infection model, the model host must be of a size suitable for chemical screening. The zebrafish (Danio rerio) has successfully been used in high-throughput chemical screens^{52, 53} and has also been used to model mycobacterial, streptococcal and Salmonella infections^{54, 55, 56}. Thus, the marriage of chemical screening to zebrafish infection models may be a new way of rapidly identifying compounds that target in vivo essential gene functions in a vertebrate model host⁵⁷.

The utility of whole organism–based chemical screening for new antimicrobial candidates has already been demonstrated using Caenorhabditis elegans as the model host. In this study, Moy and colleagues screened 6,000 synthetic compounds and 1,136 natural product extracts for compounds that promote nematode survival of Enterococcus fecalis infection⁵⁸. They identified several compounds that rescue nematodes from the lethality of infection at therapeutic concentrations that are well below the minimal inhibitory concentrations (MICs) of the compounds against E. fecalis. Unless one posits that C. elegans is able to concentrate the compounds above the therapeutic index, one must assume that these compounds are rescuing the nematodes either by targeting an E. fecalis in vivo essential gene function, inhibiting virulence, or targeting the nematode immune defense. Though the targets of these compounds have not yet been identified, and though it is unclear as yet whether these compounds target the worm or the bacterium, this method highlights a novel approach for identifying compounds that target proteins essential for virulence in vivo.

Inhibitors of in vivo essential genes as therapeutics

Inhibitors of in vivo essential gene functions can target (i) functions required for viability in vitro (including conventional antibiotic targets), (ii) functions required for viability only in the host, (iii) functions required for virulence or (iv) some combination of the above. The advantages and disadvantages of a therapeutic will vary based on the functional class or classes to which the target belongs. Advantages and disadvantages are manifested in the spectrum of activity against different pathogens, the window of time in which a drug is effective, and the level of selective pressure for resistance.

Therapeutics that target gene functions essential for viability both in vivo and in vitro (that is, conventional antibiotics) often target proteins that are well conserved among species. Thus, this functional class has the potential to be broad-spectrum, and the time window during which it should be effective is relatively large. The disadvantage, of course, is that this class of therapeutics can impose strong selective pressure that fosters the growth of antibiotic resistance. At the other end of the extreme are therapeutics that target gene functions required for virulence. The advantages and disadvantages of this functional class have been discussed above, and they include the imposition of less selective pressure and a narrower window of efficacy and spectrum of activity. Therapeutics that target in vivo essential gene functions will have advantages and disadvantages that fall between these two extremes. They should exert some selective pressure, but only within the context of the host, and they may have a larger efficacy window and (potentially) a broader spectrum of activity compared to antivirulence drugs. Therapeutics that fall into a combination of these functional classes will have even more complex profiles.

Conclusion

This review highlights a number of novel strategies for developing new therapeutics against infection. There have been very few new classes of antibiotics discovered in the past 40 years, with efforts dwindling in large pharmaceutical companies. The effort that has been invested has been disappointing, resulting in a rather grim picture for the current state of antibiotic development. GlaxoSmithKline recently undertook a genomic, target and whole cell–based approach over a period of seven years to identify compounds that inhibit genes thought to be essential for viability in a number of pathogens⁵⁹. Disappointingly, only five leads were identified from a total of 70 high-throughput screens using the GSK compound collection, which suggests that in vitro target-based screening of traditional compound libraries biased to follow Lipinski's 'rule of five'⁶⁰ is an ineffective way to identify new antimicrobials. However, GlaxoSmithKline's experience does suggest that future efforts at antimicrobial identification will need to include screening of libraries that sample a wider diversity of chemical space with potentially differing physicochemical properties (possibly diversity oriented synthesis libraries⁶¹) and screening of natural products, which as yet have not been exhaustively mined for new antimicrobials. In addition, the GlaxoSmithKline experience also suggests that it is easier to find the target of a given lead compound than to engineer a lead compound to have greater permeability; thus identification of new antimicrobials is likely to be more fruitful in whole cell–based screens rather than in in vitro target-based screens. We suggest that this approach should be expanded to include whole-organism screening, which has the advantage that one can identify small molecules that target either the host or in vivo essential gene functions of the pathogen. Given the dearth of antimicrobials with novel modes of action against Gram-negative hospital pathogens currently in phase 1 clinical trials, it is estimated that it may be 10 to 15 years before antimicrobials against this class of pathogen are available for therapeutic use⁵⁹. We believe that this makes the argument for pursuing therapeutics that target in vivo essential gene functions all the more compelling, as antibiotic resistance continues to evolve and the need for new antimicrobials continues to grow.