Vous êtes sur la page 1sur 17

BACTERIAL PROTEOMICS AND ITS ROLE IN ANTIBACTERIAL DRUG DISCOVERY

Heike Bro tz-Oesterhelt,1* Julia Elisabeth Bandow,2 and Harald Labischinski1 1 Bayer HealthCare AG, Anti-infective Research, Wuppertal, Germany 2 Pzer, Inc., PGRD Ann Arbor, Michigan
Received 26 February 2004; received (revised) 29 April 2004; accepted 9 May 2004
Published online 29 July 2004 in Wiley InterScience (www.interscience.wiley.com) DOI 10.1002/mas.20030

Gene-expression proling technologies in general, and proteomic technologies in particular have proven extremely useful to study the physiological response of bacterial cells to various environmental stress conditions. Complex protein toolkits coordinated by sophisticated regulatory networks have evolved to accommodate bacterial survival under ever-present stress conditions such as varying temperatures, nutrient availability, or antibiotics produced by other microorganisms that compete for habitat. In the last decades, application of man-made antibacterial agents resulted in additional bacterial exposure to antibiotic stress. Whereas the targeted use of antibiotics has remarkably reduced human suffering from infectious diseases, the ever-increasing emergence of bacteria that are resistant to antibiotics has led to an urgent need for novel antibiotic strategies. The intent of this review is to present an overview of the major achievements of proteomic approaches to study adaptation networks that are crucial for bacterial survival with a special emphasis on the stress induced by antibiotic treatment. A further focus will be the review of the, so far few, published efforts to exploit the knowledge derived from bacterial proteomic studies directly for the antibacterial drugdiscovery process. # 2004 Wiley Periodicals, Inc., Mass Spec Rev 24:549565, 2005 Keywords: 2D gel electrophoresis; proteomics; antibiotics; drug discovery; bacteria

I. INTRODUCTION
The term proteome, in analogy to the term genome, was coined to describe the complete set of proteins that an organism has produced under a dened set of conditions (Wasinger et al., 1995). The genome is static because it represents the blueprint for all cellular properties that a cell is able to develop. In contrast, the proteome is highly dynamic and much more complex than the genome. It is critical for survival that the protein composition of a cell is constantly adjusted to meet the challenges of changing environmental conditions. Already in 1975, the powerful method of two-dimensional-polyacrylamide gel electrophoresis (2DPAGE) was introduced that allowed one to separate highly complex cellular protein extracts into individual proteins on a single gel based on two properties of the proteins the isoelectric

tz-Oesterhelt, Bayer Pharma Research *Correspondence to: Heike Bro Center, Building 405, D-42096 Wuppertal, Germany. E-mail: heike.broetz-oesterhelt@bayerhealthcare.com

point (pI) and the molecular weight (MW), (Klose, 1975; OFarrell, 1975). Proteomics, and 2D-PAGE in particular, has been used from the beginning to study the bacterial proteome under different growth conditions (Linn & Losick, 1976; Reeh, Pedersen, & Neidhardt, 1977; Agabian & Unger, 1978) and various external stress factors (Young & Neidhardt, 1978; Krueger & Walker, 1984; Gomes et al., 1986). However, it was only after 1995 that a new era was opened to the study of the dynamic behavior of the bacterial proteome by the advent of the rst complete genome sequence of a bacterium, Haemophilus inuenzae strain RD KW20 (Fleischmann et al., 1995). Based on a well-annotated genomic sequence, it became possible to introduce large-scale mass spectrometry (MS) techniques to identify virtually every protein detected on a 2D gel. The increase in throughput, the partial automation, and the higher reproducibility of 2D-PAGE analysis recently made it a very attractive tool to study cellular functions on a molecular level. The complete genomic sequences of more than 120 bacteria are now publicly available (for an constantly updated list, see http:// www.tigr.org/tigr-scripts/CMR2/CMRGenomes.spl) that allow one to select among a variety of microorganisms for proteomic investigations according to the scientic question of interest. In parallel, MS techniques, advanced to identify many proteins from 2D gels and from alternatives to gel electrophoresis such as the Isotope-Coded Afnity Tag (ICAT) technology, have emerged to overcome some of the weaknesses of the 2D-gel approach (for recent reviews, see e.g., Godovac-Zimmermann & Brown, 2001; Hamdan & Righetti, 2002; Aebersold & Mann, 2003; Lill, 2003; Sechi & Oda, 2003). Compared to eukaryotic cells, bacteria are great model organisms to study regulatory networks, protein function, and even cell differentiation, because their genomes are relatively small and adaptation processes are less complex and involve smaller numbers of protein components. Some bacteria are easily genetically manipulated and are thus excellent models to study protein function. In addition, bacteria are commonly used in the food industry as well as in biotechnology. In both areas, it is desirable to understand bacterial metabolism in order to optimize production yields and quality. Bacteria also have an even more direct impact on human life in that a variety of species are indispensable for aspects as immune-system maturation, nutrition digestion, and vitamin production (a 70 kg human contains approximately 1 kg of bacteria, and thus more bacterial than human cells). On the other hand, interactions harmful to the human host occur when bacteria override the defense barriers and cause infections. In fact, infections by microorganisms cause some 17 million deaths each year according to WHO statistics.

Mass Spectrometry Reviews, 2005, 24, 549 565 # 2004 by Wiley Periodicals, Inc.

&

TZ-OESTERHELT ET AL. BRO

Although most of those deaths occur in the less-developed countries, death due to infectious diseases is back to rank number three even in the most developed countries such as the US (Armstrong, Conn, & Pinner, 1999). One important reason for that unpleasant development is the fact that bacteria that were previously susceptible to the large armory of antibiotics have now developed resistance against them (Hiramatsu et al., 2001; WHO, 2001; Appelbaum, 2002; Walsh, 2003). Another reason is ironically provided by the progress in medicine in general, because we are becoming older and more often subject to aggressive treatment regimens; for example, in surgery, transplantation, and cancer chemotherapy. All of those manipulations lead to a suppression of our immunological defense capabilities, and, thereby, to more serious and more difcult to treat infections. Thus, novel treatment options are urgently required, and the need for novel antibacterial agents without cross-resistance to existing antibiotics as well as the development of alternative treatment regimens should have high priority on any meaningful public health agenda. In that environment, it is not astonishing that proteome analysis of the consequences of antimicrobial treatment for bacteria has recently gained increasing interest. It can, on one hand, provide a deeper insight into how a bacterium responds to a certain antimicrobial treatment. In addition, benets are expected in many other aspects of modern drug development approaches such as the identication of novel target areas and the elucidation of the molecular mechanisms of action of novel drug candidates. Thus, the intent of this review is to present an overview of the major achievements in proteomic studies of adaptation networks that are crucial for bacterial survival with a special emphasis on stress that is induced by antibiotic treatment. A further focus will be the review of the, so far few, published efforts to exploit the knowledge derived from bacterial proteomic applications directly for the antibacterial drug-discovery process. We will also touch on some proteome studies that aim at a more general insight into the physiological exibility of bacteria as well as on some methodological pre-requisites. However, the reader interested in a full overview of the latter topics is referred rg et al., 2000; Nyman, to some excellent recent reviews (Go 2001; Lilley, Razzaq, & Dupree, 2002; Hecker, 2003).

variable growth conditions with respect to temperature, pH, osmolarity, nutrient availability, host interactions, etc. It should be noted that those stress situations, often regarded as naturally occurring, do not principally differ from the stresses induced by antibiotic attack. Antibiotics are a frequent encounter for many bacteria in their natural habitats, because many microorganisms produce them to suppress the growth of competitors. Actually, the capability of a microorganism to produce a substance that prevents the growth of another is eponymous for the term antibiotic, although it is nowadays used more broadly to include man-made compounds as well. Even antibiotic classes that stem from purely synthetic approaches and never experienced by bacteria during evolution can, to a certain extent, mimic natural processes for which bacteria have developed regulatory mechanisms. For instance, the oxazolidinones, which inhibit protein synthesis (Livermore, 2003), simulate a starvation-like situation. Also, the quinolones as topoisomerase-inhibitors (Drlica & Zhao, 1997) cause DNA-replication errors and repair system failures to which the bacteria react with their SOS response (Sutton et al., 2000). The evolutionary success of bacteria was strongly dependent on their ability to respond to such adverse conditions via a bewildering range of behavioral responses (Armitage et al., 2003). A large number of external and internal signal molecules and signal transduction processes are present in bacteria to adapt their protein composition to the changing requirements of their environment (Armitage et al., 2003). Several of the environmental challenges are experienced by many bacterial species, and are, therefore, met by somewhat conserved response mechanisms. However, it should be clear from the foregoing that the majority of the reactions are rather species-specic, and depend on the environmental and lifestyle preferences of the species. Proteomics technologies appear to be the natural tools to study the consequences of those regulatory processes on protein composition. Key to the physiological interpretation of proteome studies performed by 2D-PAGE, the most commonly used technology platform, is the determination of the identity of the proteins contained in the spots on the gel. Thus, we will start with some remarks on the process of proteome mapping.

A. Proteome Mapping II. THE ROLE OF PROTEOMICS TO DECIPHER THE BACTERIAL RESPONSE TOWARDS CHANGES IN ENVIRONMENTAL CONDITIONS AND ANTIBIOTIC ATTACK
The capability to grow many bacterial species in well-dened articial culture media has been a pre-requisite for our current indepth understanding of bacterial physiology. Very often, those culture media provide cockaigne-like growth conditions that allow for a maximal and uniform logarithmic bacterial growth behavior until some components of the medium become exhausted and logarithmic growth ceases. Under such optimal conditions, the protein composition of the cell is usually quite constant and tuned to support the special conditions of fast growth as, for example, support of several DNA-replication forks within a single cell and maximal protein biosynthesis. However, outside the laboratory bacteria face much less supportive and highly
550

The large collection of fully sequenced bacterial genomes includes those of important pathogens such as H. inuenzae, Staphylococcus aureus, Enterococcus faecium, Enterococcus faecalis, Streptococcus pneumoniae, Pseudomonas aerigunosa, Mycobacterium tuberculosis, or Escherichia coli, which are in the focus of antibacterial drug discovery (http://www.tigr.org/ tigr-scripts/CMR2/CMRGenomes.spl). The genome contains a wealth of information that helps an organism to survive, but this blueprint does not reveal which of the encoded molecules are relevant under any given condition. The majority of effector molecules in a cell that act and interact to make life possible are proteins. Although one can predict from the genome the number of encoding entities (open reading frames), one cannot directly deduce the number of different proteins that an organism is capable of generating. One needs to perform global protein analyses to dene the protein composition of a given cell under a certain circumstance.

PROTEOMICS AND ANTIBACTERIAL DRUG DISCOVERY

&

Classical 2D-PAGE is still the method of choice for the analysis of the protein compartment of an organism. Proteome mapping utilizes different 2D gel formats to detect and index as many proteins of an organism as possible. Protein identication, typically by mass spectrometry but also N-terminal sequencing, links a protein spot on the gel to its coding sequence and knowledge on protein function, which has mostly been generated by classical molecular biology experiments. Ideally, each gene would be represented by one or more protein spot(s) on at least one of the gel formats. Unfortunately, not all proteins separate well in 2D-PAGE or are present in sufcient amounts for detection, but proteome maps for some organisms are very advanced. The map of Mycoplasma pneumoniae, for instance, covers up to 44% of the predicted open reading frames (Ueberle, Frank, & Herrmann, 2002). In the case of H. inuenzae, the authors identied the gene products of 33% of the open reading frames and extrapolated from the number of identied and non-identied proteins on the gels that they could visualize approximately 70% of the predicted open reading frames (Langen et al., 2000). Similar coverage was reported for E. coli (Tonella et al., 2001). Today, many bacterial protein maps are available; among those maps several are from pathogenic bacteria, including Chlamydia pneumoniae (Vandahl et al., 2001), M. tuberculosis (Schmidt et al., 2003), S. aureus (Cordwell et al., 2002), and Helicobacter pylori (Cho et al., 2002). Several of those maps are accessible via World-Wide-Web servers (e.g., http://microbio2.biologie. uni-greifswald.de:8880/2dnet.htm, http://www.mpiib-berlin. mpg.de/2D-PAGE/, http://us.expasy.org/ch2d/). One of the most comprehensive mapping studies was performed on H. inuenzae, a causative agent for respiratory-tract infections. Its genome encodes for approximately 1,742 gene products, and is comparably small. The proteome map of H. inuenzae is well-advanced because that organism is easily cultivated and has been used to study the proteome under different growth conditions, including treatment with antibiotics that inhibit DNA, RNA, and protein synthesis (Evers et al., 2001; Gmuender et al., 2001). Another proteome study of that organism aimed at the identication of potential vaccine candidates (Thoren et al., 2002). The 2D-gel proteome map of H. inuenzae published in 2000 (Langen et al., 2000) displays approximately 500 identied proteins, representing about 30% of all predicted open reading frames. Those proteins were mainly identied by peptide mass ngerprinting, using MALDI-TOF-MS, and some additional proteins were identied by amino acid composition analyses. On 2D images of crude protein extracts (soluble fraction pI 310), about 1,100 protein spots were detected. Different methods were studied to efciently enrich lowabundance proteins that were not visible on 2D gels obtained with crude cell extracts. In addition, maps were created for soluble basic proteins in the pI range of 811, and for envelopebound proteins in the pI range of 4.59.5. Great progress was achieved in the still-difcult separation of membrane proteins: 70 membrane proteins were identied from 2D gels in that study for the rst time. The membrane proteins were separated by an inverse, discontinuous electrophoresis system, using the cationic detergent benzyldimethyl-n-hexadecylammonium chloride and a separation gel at pH 2.1 in the rst dimension and a standard SDS gel (anionic detergent) in the second dimension. Thus, separation was performed according to molecular mass in both

dimensions. One-third of the 1,350 H. inuenzae proteins with known function have been identied on at least one of the gel formats applied in that mapping study. That 2D reference map provides an excellent basis for experiments designed to study the cellular response to external stimuli. Recently, a study on the same organism has been published that elegantly demonstrates the potential of non-gel-based technology for the purpose of proteome mapping and analysis (Kolker et al., 2003). Approximately 25% of all predicted open reading frames were detected by liquid chromatography (LC) coupled with ion-trap tandem mass spectrometry (MS/MS). Whereas the number of identied proteins was quite similar to that identied in the 2D gel-based study, there were some interesting differences in the type of proteins detected. The LC/MS/MS technology identied more ribosomal proteins than the 2D gel approach (80 vs. 34%) and also more membrane proteins. Vice versa, this difference obviously means that some proteins detected by 2D gel analysis were not found in the LC/MS/MS experiments. Future will tell, whether both types of methods will co-evolve, or whether the inherent advantages of one technique will attract attention predominantly. The mapping study of S. aureus shall be taken as a further example of proteome analysis of a crucial human pathogen. S. aureus is one of the major causes of community-acquired and nosocomial infections, and raises increasing public health concerns due to progressing multi-drug resistance (Hiramatsu et al., 2001; Sievert et al., 2002). Due to a variety of different virulence factors, including different cell wall-associated proteins and extracellular toxins, S. aureus causes a broad spectrum of infections, and the current drug-discovery programs of many pharmaceutical companies aim at targeting that pathogen. Despite its importance, information on the proteome of S. aureus emerged only recently. One reason is that the rst genome sequence of this species, which is indispensable for rapid protein identication by MS-techniques, was not publicly available until 2001 (Kuroda et al., 2001). However, because several groups are pursuing S. aureus proteomics, substantial progress has been made (for a review, see Hecker, Engelmann, & Cordwell, 2003). In a recent mapping study, 377 protein spots were analyzed from a cytoplasmic protein map (Fig. 1) that could be linked to 266 (approximately 12%) of the open reading frames (Cordwell et al., 2002). Together with the extracellular proteins described previously (Ziebandt et al., 2001), at least 20% of the theoretical S. aureus proteome has now been identied. The establishment of such protein reference maps is a crucial starting point for many physiological studies that may follow. However, because such proteome maps do not depict a real gel, but represent virtual compilations of all proteins ever detected or identied in an organism, they do not disclose which subset of proteins is expressed under the specic growth condition of interest. In order to obtain that information, proteinexpression proles must be generated as discussed in the following paragraph.

B. The Concept of Proteomic Signatures


The protein map links a protein spot on a 2D gel to its corresponding open reading frame and the respective knowledge on protein function. In contrast, the protein-expression prole is the
551

&

TZ-OESTERHELT ET AL. BRO

FIGURE 1. Example of a protein reference map. The proteome of Staphylococcus aureus 8325 was

separated by 2D-gel electrophoresis, using an immobilized pH gradient in the range of pI 47. Proteins were stained with silver, and were identied by MALDI-MS after tryptic digestion. The identity of selected proteins that serve as landmarks on the gel are indicated. Reproduced from Hecker, Engelmann, & Cordwell (2003), with permission from Elsevier, copyright 2003.

quantitative catalog of proteins made by a cell under a given circumstance (VanBogelen et al., 1999; VanBogelen, 2003). It is the protein-expression prole that indicates which particular subset of proteins is present under the growth condition studied. In spite of the progress in proteome mapping described in the previous section, it is still not technically feasible to obtain a complete expression prole from a single 2D gel, because not all proteins are equally well-separated by this technique. Extremely basic, acidic, small, or large proteins as well as those that are poorly soluble or appear in low-abundance still pose major challenges. Nevertheless, classical 2D gels still show a substantial portion of the protein expression prole and are widely used to study the cellular response to external stimuli. For all proteins with an altered expression in response to a particular stimulus, the expression stimulon was coined (Neidhardt, Ingraham, & Schaechter, 1990). For example, all proteins that are up- or down-regulated after a shift to highgrowth temperature belong to the heat-shock stimulon. The term stimulon describes the changes in protein expression on a phenotypic level, and does not provide any information on the underlying transcriptional regulation. A regulon, on the other hand, consists of proteins that are under the control of the same global transcriptional regulator. Even in bacteria, the leastcomplex organisms, a stimulon usually consists of more than one regulon, demonstrating the complexity of regulation required for adaptation. In our heat-shock example, three regulons contribute to the heat-shock stimulon in B. subtilis (Fig. 2): (1) class I heatshock proteins under the control of the global repressor HrcA,
552

including the chaperones of the GroEL and DnaK machines, (2) class III heat-shock proteins under the control of the global regulator CtsR, and (3) the general stress proteins that depend on the alternative sigma factor sB for transcription. In addition, a fourth class contains further heat-responsive proteins that could lker, not yet be assigned to any regulon (Hecker, Schumann, & Vo 1996; Hecker, 2003). By studying expression levels of a multitude of proteins under a variety of different growth conditions, specic proteins become indicative of a particular physiological state of the cell. Such a subset of proteins, whose expression levels are characteristic for a dened condition, was also designated proteomic signature (VanBogelen et al., 1999). To identify a proteomic signature, it is essential to recognize the connection between the expression levels of specic proteins and a particular physiological state. Knowledge of the identity or function of those proteins is not strictly required, although it helps to understand the molecular basis for their expression. Often, it is necessary to analyze several related and unrelated conditions to propose and verify the proteomic signature for a certain environmental factor of interest. However, once such protein signatures are established for a variety of different physiological states, that compilation can be extremely helpful in the interpretation of a protein expression prole obtained under an unprecedented growth condition. Some proteomic signatures published previously for E. coli (VanBogelen & Neidhardt, 1990; VanBogelen et al., 1999) are particularly illustrative to describe how that concept can be

PROTEOMICS AND ANTIBACTERIAL DRUG DISCOVERY

&

FIGURE 2. Heat-shock stimulon in Bacillus subtilis. Three regulons contribute to the response to heat stress in this organism. The rst regulon is under control of the HrcA repressor, and contains chaperones of the GroEL and DnaK machinery (marked by *), which are crucial for protein folding during heat stress. The second is the CtsR regulon (marked by #), which regulates the chaperones of the Clp family and the Clp protease, and the third is the sB regulon (marked by ). Whereas the former two regulons react specically to heat stress, the large regulon controlled by the alternative sigma factor sB is induced by various kinds of stress and starvation stimuli. The bar charts depict the expression levels of one representative member of the respective regulons under different stress conditions: C, control; H, heat; E, ethanol; S, salt; G, glucose limitation; Pm, treatment with the antibiotic puromycin; Ox, oxidative stress. Figure kindly provided by J. Bernhardt & M. Hecker, University of Greifswald.

applied to studies on bacterial physiology in the presence of external stress factors, including antibiotic treatment. In those studies, a clear correlation was demonstrated between the proteomic signatures for growth at high and low temperature on one hand, and the changes in protein expression proles in response to antibiotic inhibition of ribosomal function on the other hand. Between 23 and 378C, protein expression proles do not show specic signatures for growth temperature. Outside of that range, however, there are protein subsets characteristic for growth at low and high temperature. Some proteins seem to behave as cellular thermometers: their amount changes gradually with increasing/ decreasing temperature. Other proteins are regulated in an off/on fashion, and are highly induced specically at either high or low growth temperature. At high temperature, the folding of newly synthesized proteins is impaired, resulting in misfolded proteins that trigger the induction of chaperones and proteases. In contrast, at low temperature the proteins involved in the translation process (ribosomal proteins and elongation factors) are induced in addition to the specic cold-shock proteins, suggesting

that under this condition translation is the rate-limiting step for growth of E. coli. The ribosome is also the target of many antibiotics, that interfere with translation via different molecular mechanisms of action, and their effects on the proteome overlap with the signatures for growth temperature. Aminoglycosides such as streptomycin and kanamycin interfere with the ribosomal proofreading activity and cause an increase in mistranslation. The resulting accumulation of mistranslated and, therefore, misfolded proteins leads, just as the increase in growth temperature, to the induction of chaperons and proteases. Similarly, puromycina protein synthesis inhibitor that causes abortive translationleads to the accumulation of truncated and misfolded proteins, thereby also inducing the heat-shock signature. However, treatment with all three of those antibiotics also induces an additional response that is not observed during the shift to high growth temperature: the stringent response in E. coli, which is an adaptive response to limited availability of amino acids. The stringent response is triggered by an increase in ppGpp
553

&

TZ-OESTERHELT ET AL. BRO

and manifests in a down-regulation of many genes, including those that encode rRNA and proteins involved in translation. Thus, the combination of the signature for misfolded proteins and that for the stringent response results in the characteristic proteome expression prole for streptomycin, kanamycin, and puromycin in E. coli (VanBogelen et al., 1999). Another group of antibiotics impairs the efciency of the peptidyl transferase reaction: tetracycline, chloramphenicol, erythromycin, fusidic acid, and spiramycin. Although they differ in their exact binding sites and in the particular molecular mechanisms of action, they all have one thing in common with growth at low temperature: they slow down translation. In E. coli, those antibiotics, as growth in the cold, lead to an induction of cold shock proteins and of ribosomal proteins. Each organism is adapted to a particular ecological niche, which is reected on the genome level by differences in the types of proteins that are encoded and by variations in their amino acid sequences. That adaption is achieved by differences in posttranscriptional and post-translational regulation that mediate the adaptation on the protein level. Therefore, proteins that constitute a proteomic signature for a specic condition in one organism do not necessarily belong to the proteomic signature for the same physiological state in another organism. We take the treatment of E. coli and B. subtilis by antibiotic inhibitors of protein synthesis as an example to demonstrate how protein signatures may vary between bacterial species. When B. subtilis is treated with kanamycin or streptomycin, chaperons and proteases are induced as in E. coli, but in contrast to E. coli the stringent response is not triggered by those antibiotics. Similarly, treatment of B. subtilis with tetracycline, chloramphenicol, erythromycin, and fusidic acid leads to an induction of proteins, forming the translation apparatus; however in contrast to E. coli cold shock proteins are not induced (Bandow et al., 2003a).

C. Snapshots of Protein Biosynthesis: Metabolic Labeling and Dual-Channel Imaging


For a given cell, it is crucial for survival to quickly adjust its protein composition and the activity of individual proteins in order to meet the challenges of ever-changing growth conditions. That adaptation is mediated on a number of levels: transcriptional, post-transcriptional, and translational regulation; protein stability also effects protein levels (the amount of protein present under a given condition at a certain time point), whereas posttranslational modication is often a means of regulating protein activity. 2D gel-based proteomics techniques are not only wellsuited to study protein levels and to detect protein modications, they also allow one to monitor changes in the relative protein synthesis rates and thus give a sensitive read-out on adaptation in progress. The set of proteins that is newly synthesized by an organism can change dramatically in response to modications in growth conditions or environmental-stress factors. When confronted with a new situation, the cell dedicates a large proportion of its translation capacity to the de novo synthesis of proteins needed at higher levels to adequately meet the challenges posed upon it. Pulse-labeling of the proteins with 35S-[L]-methionine is a very sensitive method to visualize by autoradiography specically the newly synthesized protein fraction. Short labeling
554

times allow one to capture snapshots of protein synthesis at any time point during adjustment to the new condition and in the new steady state. Dual-channel imaging, rst described by Bernhardt et al. (1999), was developed to facilitate the comparison of de novo protein synthesis detected on autoradiographs and protein amounts detected by silver staining. In the original protocol, the 2D gels of the 35S-labeled protein extracts were silver stained, dried, and exposed to phospho screens for autoradiography. The false color green was assigned to the protein spots on the silver image by means of a photo editor, and the false color red to the spots on the autoradiograph. When both false color images were overlayed, proteins that were newly synthesized during the pulse, but had not accumulated to amounts detectable by silver staining, appeared in red. On the other hand, proteins that had already been present prior to the pulse but were no longer synthesized appeared in green. Similar expression levels under both conditions resulted in a yellow color. Today, more sophisticated software packages are available that contain warping tools to overlay also independent 2D gels, and that provide a variety of color schemes from which to chose (Delta2D software, DECODON GmbH, Greifswald/Germany; Z3 and Z4000 software, Compugen Ltd., Tel Aviv, Israel). Dual-channel imaging was applied, for instance, to the identication of new stimulons. Proteins induced in response to the stimulus could be conveniently detected by their red color, whereas repressed proteins were colored in green. An impressive example of the utility of the dual-channel imaging technique was published recently (Bernhardt et al., 2003). Global changes in protein expression occurred in B. subtilis grown in synthetic medium, when, after a period of exponential growth, the primary carbon source, glucose, was exhausted. A transition phase was followed by a phase of glucose starvation and when, eventually, glucose was added to the starving culture, exponential growth resumed. The snapshots of protein synthesis taken at different time points during exponential growth and adaptation to starvation vividly show the changes in cellular resource allocation. At all time-points, the rates of de novo protein synthesis were compared to the total protein amounts as visualized by silver staining. During the transition from exponential growth to glucose starvation, the protein expression pattern changed dramatically: about 150 proteins not synthesized during exponential growth were induced, and the synthesis of nearly 400 proteins ceased. Most of the 150 induced proteins belonged either to regulons induced specically by glucose starvation, or to more general forms of stress response induced in response to various stimuli. The glucose-starvation specic proteins indicated a drop in glycolysis, or were involved in the utilization of alternative carbon sources and gluconeogenesis. The general stress and starvation proteins belonged to the sB-dependent general stress regulon, the stringent response stimulon, and the sporulation cascade.

D. Protein Modications
A great advantage of 2D gel-based proteomics as opposed to, for example, the newly developed ICAT technology is the detection of protein modications that cause a polypeptide to migrate to a different pI/Mr location on the 2D gel. Because such modica-

PROTEOMICS AND ANTIBACTERIAL DRUG DISCOVERY

&

tions are often linked to protein function or protein activity, that information is crucial for understanding of the physiological state of a cell. For example, the alternative sigma factor sB in B. subtilis governs a large regulon that comprises more than 150 general stress proteins that are induced under a great number of different stress conditions. sB activity is regulated by a complicated signaling cascade (Fig. 3), and is controlled eventually by the phosphorylation state of the anti-anti-sigma factor RsbV lker, (for a review on sB regulation in B. subtilis, see Hecker & Vo 2001). RsbV in its phosphorylated state has a reduced afnity for the anti-sigma factor RsbW, which is in turn free to capture sB in a stable complex, thereby preventing the transcription of the genes of the sB-regulon. In contrast, dephosphorylated RsbV binds RsbW and thus releases sB. The active alternative sigma factor competes with the housekeeping sigma factor sA for the polymerase core enzyme and induces transcription of the sBdependent genes. The phosphorylation state of the anti-antisigma factor RsbV is carefully regulated, and involves the activity of the two phosphatases, RsbU and RsbP. RsbP senses the energy status of the cell, and dephosphorylates RsbV upon glucose and phosphate starvation, whereas RsbU takes over this function after exposure to heat, acid, or ethanol. Not only was 2D gel-based proteomics instrumental in the identication of the members of the sB regulon, it also allowed the monitoring of the phosphorylation state of the anti-anti-sigma factor RsbV, which appears on 2D gels in two distinct isoformsone the phosphorylated and the other the dephosphorylated protein. Given the importance of the sB-response in B. subtilis for general stress adaptation, it was somewhat unexpected that, in a recent proteomics study where B. subtilis was exposed to sublethal concentrations of 30 antibiotics from various compound classes, only rifampicin induced the sB-response in that tz, & Hecker, 2002; Bandow et al., organism (Bandow, Bro 2003a). Even then, the general stress response was not induced immediately after exposure to the antibiotic, but occurred with a delay of about 1 hr during a drug-mediated growth arrest. A mutant, in which the sigB gene was deleted, responded to

rifampicin treatment with a considerably prolonged growth arrest compared to the wild-type, and it was, therefore, postulated that the sB response helped B. subtilis to overcome the growth arrest. To investigate which of the above-mentioned phosphatases is involved in sB activation during rifampicin treatment, 35S methionine pulse-labeling and 2D-PAGE were repeated with tz, & Hecker, rsbU and rsbP insertion mutants. (Bandow, Bro 2002). After rifampicin treatment, the active, dephosphorylated form of RsbV was induced in the wild-type and the rsbU mutant, but not in the rsbP mutant. This result indicates that, during rifampicin exposure, the energy-signaling pathway via RsbP is responsible for RsbV dephosphorylation and consequently sB activation. A further example of protein modications in bacteria stems from the analysis of protein-expression proles of a conditional deformylase mutant (Bandow et al., 2003b). In bacterial protein biosynthesis, formyl-methionine is always incorporated into nascent proteins as the rst amino acid. Peptide deformylase is needed afterwards to remove that N-terminal formyl residues from the polypeptide chains; that function is essential for bacterial survival. With respect to antibacterial drug discovery, the deformylase gained recent interest as a novel target, and the rst class of inhibitors has now reached phase I of clinical development (Johnson et al., 2003). B. subtilis encodes two functional peptide deformylases, Def and YkrB. The latter represents the major deformylase in this organism, although both enyzmes can at least partly substitute for each other, because single deletion mutants in both genes remain viable (Haas et al., 2001). A def deletion mutant in which the ykrB gene was placed under the control of a xylose promoter was constructed and was analyzed by 2D gel-electrophoresis (Bandow et al., 2003b). As long as xylose was present in the growth medium, ykrB was transcribed and the protein pattern of the mutant matched that of the isogenic wild-type. When xylose was depleted and glucose was added to the medium for efcient repression of the xylose promoter, the protein expression pattern of the mutant changed dramatically. A new protein spot accumulated next to almost every protein spot that had been present under non-repressing conditions. The newly accumulating proteins were more acidic than their counterparts, and were shown by ESI-Q-TOF-MS to still carry the N-terminally formylated start methionine, which under control conditions is usually removed from a large percentage of the proteins. The same shift of protein spots to a more acidic position was observed after treatment with the antibiotic actinonin, which acts as a deformylase inhibitor (Bandow et al., 2003b).

III. PROTEOMICS AND THE ANTIBACTERIAL DRUG DISCOVERY PROCESS


So far, we have discussed in this review the sometimes astonishing capacity of bacterial cells to adapt to environmental stress conditions, including antibiotic exposure, and also the utility of proteomic techniques to elucidate those adaptive responses. In the examples mentioned above, antibiotics were employed as a kind of tool to modulate the bacterial metabolism by directed inhibition of essential cellular functions. Treating a bacterium with an antibiotic from an established class with well-understood
555

FIGURE 3. Simplied scheme of the s activation cascade: Different

environmental signals stimulate the phosphatase RsbP and RsbU to desphosphorylate the anti-anti-sigma factor RsbV. Dephosphorylated RsbV sequesters the anti sigma factor RsbW, thereby releasing sB for its interaction with DNA polymerase. For a detailed review on further regulatory elements in the process, refer, for example, to Hecker & lker (2001). Vo

&

TZ-OESTERHELT ET AL. BRO

mechanism of action provides valuable insights into the physiological consequence of an impaired metabolic function or pathway. However, with the search for novel antibacterial agents in mind, more direct applications of proteomics in antibacterial drug discovery can also be envisaged. We will discuss this topic in more detail in The Potential Roles of Proteomics in Antibacterial Drug Discovery of the following chapter. Prior to that, we will outline in the next paragraphs some approaches and processes in antibacterial drug discovery to give an impression of the underlying aims and obstacles. In 1972, the US Surgeon General made the often-cited statement The book of infectious diseases can now be ultimately closed. The rational for thisas we know today clearly wrong statement was the enormous success in combating infectious diseases due to improved hygiene measures and the causal treatment of many bacterial pathogens by antibiotics. The role of that development can hardly be exaggerated with respect to the increase in life expectancy and the avoidance of serious complications of bacterial infections in well-developed countries. The success story rst started with Gerhard Domagks discovery of the sulphonamides (introduced in 1936) and was followed by the b-lactams (1940), the tetracyclines (1949), chloramphenicol (1949), aminoglycosides (1950), macrolides (1952), glycopeptides (1958), streptogramins (1962), and quinolones (1962). Although extreme progress was made in the chemical modication of those antibiotic classes, which led to much improved new subclasses, almost 40 years passed until the next truly new class, the oxazolidinones, was introduced into the market in 1999 (Strahilevitz & Rubinstein, 2002). Given the extraordinary adaptability of bacteria, it should come as no surprise that many of the formerly effective antibiotics have to a certain degree lost their ability to kill previously susceptible pathogens. Under antibiotic pressure, bacteria have developed various protective mechanisms such as additional barriers for antibiotic penetration, active pump systems to extrude the drug from intracellular compartment, enzymatic modication of the drug to render it ineffective, and mutation of the molecular targets to prevent successful interaction between target and drug (for review, see e.g., Walsh, 2003). As a consequence, there is an urgent need for novel antibacterial compounds that are devoid of cross-resistance to the antibiotics already in use. This need can be addressed by (1) structural modication of an existing antimicrobial compound class such that it is no longer prone to the inactivation mechanism (e.g., a b-lactam stable against b-lactamases), (2) a combination of an antibiotic and a compound that inhibits the resistance mechanism (e.g., b-lactamase inhibitors (Bush, 2002), already a clinically proven concept, or efux pump inhibitors (Lomovskaya & Watkins, 2001), or (3) most preferentially, a new compound class that would act on a target site that has not yet been exploited by any existing approaches. Several studies clearly show that only a subset of the essential genes and cellular functions (essential targets) of bacteria is hit by todays antibiotics (Fig. 4); thus, in principle, there should be ample opportunity to nd such novel antibacterial drugs. In order to understand the potential role of bacterial proteomics in that process, the present approaches of antibacterial drug discovery are outlined below in somewhat more detail.

FIGURE 4. Targets of antibiotics in clinical application. Only a limited number of cellular processes/metabolic reactions are, so far, targeted by marketed antibiotics. Most compounds are derived from natural products, and only a few stem from purely synthetic approaches (marked by italics). p-AB, para-aminobenzoic acid; DHF, dihydrofolate; THF, tetrahydrofolate.

A. Current Approaches in Antibacterial Drug Discovery


With respect to the strategies pursued in the search for novel antibiotics, we will restrict ourselves here primarily to the discussion of therapeutic (as opposed to prophylactic) approaches that aim at hitting bacterial pathogens by interfering with their essential prokaryotic genes and functions. Antibacterial agents derived from such a strategy will act in a somewhat classical way by inhibiting bacterial growth even under standardized culture conditions. It should be mentioned that several other approaches are also under investigation, such as targeting genes that are not essential for bacterial survival per se, but indispensable under infection conditions such as virulence and pathogenicity factors important for infection initiation, disease progression, or persistence, as well as strategies that try to exploit eukaryotic defense mechanisms to control infections (Alksne, 2002; Suga & Smith, 2003; Weidenmaier, Kristian, & Peschel, 2003). It is important to note that all antibiotics in clinical use or in any phase of current clinical development stem from the traditional approach of measuring their inhibitory activity on bacterial growth in vitro. Only after their antibacterial potential was discovered did a detailed evaluation follow to assess all of the other properties that are required by a clinically useful drug (e.g., efcacy in animal infection models, pharmacokinetic properties, and toxicological prole). Accordingly, without exception their molecular targets and mechanisms of action were determined much later than their original discovery, and an in-depth understanding of the molecular basis of their activity often took the work of several laboratories over a considerable number of years. The situation has changed dramatically due to the advent of technologies that operate on the scale of complete microbial genomes. Todays approaches for the discovery of novel antibiotic classes can be categorized as being either directed against a specic molecular target or based on reverse genomics (Fig. 5). In the former case, a certain molecular target is carefully selected on the basis of a theoretical and experimental rational, and compound libraries are screened specically for inhibitors of its function. In the latter process, antibacterial compounds are selected somewhat more classically by their promising inhibition of bacterial growth, but are examined immediately with respect

556

PROTEOMICS AND ANTIBACTERIAL DRUG DISCOVERY

&

FIGURE 5. Antibacterial drug-discovery process. Current strategies for the discovery of novel antibacterial agents can be grouped into two major categories. The target-based approach starts with the selection of a suitable target, followed by the development of an assay to search specically for inhibitors of its function. In contrast, in the reverse-genomics approach, a compound is selected for its promising antibacterial activity, and the target is determined in a second step. Later in the hit-and-lead proling cascade, both strategies follow the same procedure.

to their selective activity against a spectrum of (often predened) molecular targets or processes. Obviously, a list of desired molecular targets for antibiotic attack is a pre-requisite for both approaches, and target selection and validation are of paramount importance. Initial target selection can be based on a variety of considerations, including its proven occurrence and essentiality in the desired spectrum of bacterial species, selectivity for microbial versus eukaryotic counterparts, amenability for screening, and percentage of reduction of target function needed to prevent bacterial growth. In addition, more difcult to generalize criteria play a role, such as technical and scientic experience in certain target areas, presumed drugability of the target, or availability of further information like, for example, availability of the 3D structure of the target (Abergel et al., 2003). In the following, we will discuss to what extent proteomics can be helpful in target selection, identication, and validation, and we will illustrate this process with some of the still few examples reported in the literature.

B. The Potential Roles of Proteomics in Antibacterial Drug Discovery


As outlined in The Role of Proteomics to Decipher the Bacterial Response Towards Changes in Environmental Conditions and

Antibiotic Attack, proteomic studies have been successfully applied to study bacterial adaptation to various stress situations, including antibiotic drug action. In fact, one would expect any antibacterial agent to induce a certain response in the bacterial proteome that reects its effects on the microbial physiology, at least as long as the drug concentration is low enough to not instantly kill and lyse the cells. Thus, the application of proteomics to the antibiotic-discovery process, technically spoken, requires the same methodological approaches as those applied to study the physiological response to environmental stresses (outlined above). Nevertheless, there are many potential questions to be asked that are specic for drug-discovery applications. Antibiotics exert their antibacterial activity via binding to and inhibition of certain molecular targets, thereby usually blocking a function essential for microbial survival. Therefore, one application of proteomics in drug discovery, that is easy to imagine, is the identication of novel antibacterial targets. In non-infectious diseases proteomic-based target identication approaches rely on the analyses of healthy versus diseased human or mammalian tissue to identify differentially expressed proteins as valid starting points for a detailed investigation of their disease-related role and their suitability as potential targets for therapeutic intervention (Yoshida, Loo, & Lepleya, 2001; Graves & Haystead, 2002). One might also expect that proteins
557

&

TZ-OESTERHELT ET AL. BRO

differently expressed in the bacterium after antibiotic attack could serve as novel targets to either enhance the activity of the drug under study, for example, in a combination therapy or for independent attack, if the novel target proves to be suitable for that purpose. Although this approach has been theoretically considered in several publications (e.g., Allsop, 1998; Schmid, 2001; Tang & Moxon, 2001), we are not aware of any published demonstration in the area of classical, broad-spectrum antibiotic research, probably because knowledge about essential targets and target selection in the antibacterial area is well-advanced and not so much a bottleneck as in other therapeutic areas (see e.g., Payne et al., 2000). However, examples for the exploitation of protein expression data in target nding exist for preventive approaches such as vaccination as well as for narrow-spectrum organism-specic therapeutic strategies (H. pylori, M. tuberculosis, P. aerigunosa) that aim either at essential or virulenceassociated targets (see e.g., Glass, Belanger, & Robertson, 2002; Kornilovska et al., 2002; Mollenkopf et al., 2002; Zhang & Amzel, 2002; Guina et al., 2003a,b; Lee, Almqvist, & Hultgren, 2003; Mathesius et al., 2003). Most of the few available studies, in which protemics was performed with clear emphasis on antibacterial drug discovery (sometimes in combination with transcriptional proling), focus on either target validation or mode of action studies, including those studies that aim at a better molecular understanding of the mechanisms of action of existing drugs (Gray & Keck, 1999; Apfel et al., 2001; Evers et al., 2001; Gmuender et al., 2001; Singh, Jayaswal, & Wilkinson, 2001; Bandow et al., 2003a,b; Ng et al., 2003). Although those studies differ in important details, the general procedure of all of them is similar. The proteome of bacteria grown in vitro under standardized conditions in the presence and absence of the antibiotic of interest is analyzed with respect to changes in the protein-expression pattern. Data analysis in most cases concentrates on listing the proteins with signicantly altered expression levels, which are subsequently discussed with respect to the current knowledge of the antibiotics mode of action. If several antibiotics with known activity in a certain metabolic pathway are investigated (e.g., antibiotics such as b-lactams, glycopeptides, D-cycloserine, and fosfomycin, which all act at different stages of bacterial cell wall synthesis (Singh, Jayaswal, & Wilkinson, 2001), or compounds such as quinolones and novobiocin, that inhibit DNA gyrase although by quite distinct molecular mechanisms (Gmuender et al., 2001)), then the data can be exploited to dene a pathway-specic stimulon or a proteomic signature that is indicative of the inhibition of a specic target, which might prove useful later in identifying and characterizing novel antibiotics that act within that pathway. In addition, protein-expression proles for compounds synthesized within a lead-optimization program can be used to investigate whether the modied compounds still act against the intended target, or whether they have lost their specic mode of action during chemical derivatization. Another application for proteomic studies within the drug-discovery process is the verication that a compound, which inhibits the activity of a desired isolated protein in a biochemical target assay, acts indeed as expected when tested against whole bacterial cells, and does not kill the cell due to other, not target-related, possibly undesired and non-specic activities such as general membrane perturbation or intercalation into nucleic acids. Thus, mode of
558

action determinations as well as validations are important and expected outcomes from such studies. Most publications cited above can be categorized as proof of principle studies limited to a certain subclass of antibacterial compounds. A broader exploitation of such proteomic mode of action analyses for the antibacterial drug-discovery process requires a large compilation of protein-expression proles for as many different compound classes with known or suspected modes of action as possible, to, ideally, represent all of the potential responses that bacteria are capable of inducing under various types of antibiotic attack. In order to allow a direct comparison between the proteomic signatures obtained for different antibiotics, it is important that highly reproducible experimental conditions are applied during bacterial growth and antibiotic treatment, as well as during 2DPAGE and data analysis. It is straightforward to build up such a data set for one selected bacterial species as a model organism, because all parameters apart from the various antibacterial agents under investigation are kept constant and, thus, a relatively large number of compounds may be analyzed by a limited number of gels. However, in a second step it is also desirable to obtain information on the responses of additional bacterial species to complete the picture. Furthermore, because there are molecular targets for which there is no inhibitory compound available, conditional mutants in such targets should, ideally, also be included. Finally, because many compounds that are active against bacteria act by mechanisms too non-specic to be exploited for antibacterial drug-discovery purposes (e.g., DNA alkylation or intercalation, detergent-like membrane damage, etc.), a comprehensive database should also include proteome data that are characteristic for such undesired activities for an early rejection of such compounds. The value of a database of that scope for major strategies, target-based drug discovery as well as reverse genomics methodologies, can hardly be overestimated and would nicely complement similar approaches that use alternative methods such as genome-wide mRNA-expression proling (Shaw & Morrow, 2003; Shaw et al., 2003; Fischer et al., 2004) and rapid phenotypic approaches such as conventional radioactive precursor incorporation techniques (limited to some rough pathway identication scope) or whole-cell FT-IR spectroscopy (Gale et al., 1981, Naumann & Labischinski, 1990; Kaderbhai et al., 2003). Although such a comprehensive database is not yet available, a recent publication shows that a realization is within reach. In the study from Bandow et al. (2003a), some 30 antibiotics have been analyzed by proteomics under uniform conditions, comprising examples for almost all known marketed antibacterial compound classes, several experimental drugs with novel mechanisms that rank high on current priority lists of pharmaceutical companies, and examples of drugs with undesired modes of action. The model organism chosen for that study was B. subtilis, the workhorse for molecular biology studies in the Gram-positive arena. Whereas at a rst glance that choice of a non-pathogenic species appears somewhat illogical for drug-discovery purposes, it was based on the fact that most major pharmaceutical companies for medical and economical reasons search for broad-spectrum antibiotics, targeting at least all of the most frequently isolated Gram-positive pathogens as staphylococci, enterococci, and streptococci and, therefore, restrict their research to targets common to all of these bacteria. Genomic comparisons have shown that such targets are present

PROTEOMICS AND ANTIBACTERIAL DRUG DISCOVERY

&

also in B. subtilis almost without exception. Of course, that means vice versa that projects that aim at small spectrum drugs (e.g., targeting staphylococci only) must rely on other adequate organisms, even though experimental difculties will be higher, because molecular-biology tools and organism-specic databases, etc. will not be as easily available or as rich in information. For each antibacterial agent investigated in that study, samples of a B. subtilis culture were collected for 2D-PAGE after exposure to compounds at two different concentrations at one or more different time-points after addition of the antibiotic. The proteins that were newly synthesized in response to antibiotic treatment were visualized by pulse-labeling with 35S-methionine followed by autoradiography, and compared to the proteins that were newly synthesized by an untreated control culture during the same labeling period. The two autoradiographs of the untreated and antibiotic-treated sample were superimposed and analyzed by the red-green dual-channel imaging technology described in Snapshots of Protein Biosynthesis: Metabolic Labeling and Dual-Channel Imaging. Several interesting observations could be obtained by that approach: rst and quite importantly, it turned out that the differential-expression patterns obtained, although never completely predictable, were in general consistent with the respective mode of action of the antibiotic as far as known. For example, protein-synthesis inhibitors clearly led to a reduction in overall translation, as expected. However, looking at the different protein-synthesis inhibitors in more detail, the proteomic signatures of those antibiotics, which simply reduce the rate of protein-synthesis, such as for example, chloramphenicol or tetracycline, were quite distinct from the signatures of compounds such as aminoglycosides and puromycin, that led to the production of mistranslated or truncated proteins. In addition, both of these groups could be distinguished from mupirocin, which interferes with protein synthesis via inhibition of isoleucine-t-RNA synthetase (Ile-RS) and which shows a protein-expression prole that is characterized by the induction of the classical stringent response (Eymann et al., 2002; Bandow et al., 2003a). Second, it was obvious, that irrespective of the overall consistency of the protein-expression data with the known mode of action of a given antibiotic, there were always some proteins with a rather unexpected induction, indicating that our present knowledge about the detailed mechanisms of antibiotic action and the cellular response to antibiotics is still limited. A special example was provided by the analysis of nitrofurantoin, an antibiotic introduced in the 1950s and still used frequently in the therapy of urinary tract infections. In earlier studies, its activity was attributed to such distinct target areas as DNA and/or RNA synthesis, carbohydrate metabolism, or an inhibition of other metabolic enzymes (Guay, 2001). The proteinexpression prole of nitrofurantoin showed a remarkable similarity to that of diamide (Bandow et al., 2003a), an agent that causes oxidative damage by inducing non-native disulde bonds (Kosower & Kosower, 1995). That result led the authors to propose that protein inhibition due to non-native disulde formation may be the primary antibacterial mode of action of nitrofurantoin; that proposal would explain nicely the pleiotropic effects reported earlier and is also compatible with studies that attributed its toxic side-effects on eukaryotic cells to the rapid formation of, for example, glutathione disuldes, glutathione-

protein disuldes, and proteinprotein disuldes (Hoener et al., 1989; Silva, Khan, & OBrien, 1993). Third, and most relevant for the use of proteomics data for drug-discovery purposes, it was demonstrated that crucial hints on the molecular mechanisms of novel antibacterial compounds can be obtained when the new mechanism is similar to that of a reference antibiotic already included in the protein-expression database. One example provided in the study of Bandow et al. (2003a) was the novel pyridiminone antibiotic BAY 50-2369, which is structurally related to the natural compound TAN 1057 A/B (Brands et al., 2003). Even by mere visual inspection, the proteomic pattern was almost identical to that of chloramphenicol and other peptidyltransferase inhibitors, leading to the suggestion that BAY 50-2369 as well as TAN 1057 inhibited the same target, although in a slightly distinct manner because no cross-resistance to other peptidyltransferase inhibitors was observed (Limburg et al., 2004). That interpretation has been proven correct in the meantime by direct mechanistic studies (Boeddecker et al., 2002; Limburg et al., 2004). A second recently published example (Beyer et al., 2004) proved the mode of action of a novel class of phenyl-thiazolylureasulfonamides as phenylalanyl-t-RNA-synthetase (Phe-RS) inhibitors by demonstrating that the proteomic signature was very similar to that of mupirocin. In fact, proteins that belong to the stringent response were similarly overexpressed after exposure to both antibiotics. Interestingly, both antibiotics led to an induction of their respective direct targets: the alpha-subunit of Phe-RS was induced in cells treated with the novel compound class, whereas mupirocin induced the corresponding Ile-RS subunit (Fig. 6). It should be noted that those conclusions could have been reached by a mere visual comparison of the respective dualchannel images of the 2-D gels, an identication of the differentially expressed protein spots by peptide mass ngerprinting, and a comparison with the well-annotated master gels. However, because visual observation is always inuenced by the personal impression of the respective observer, Bandow et al. (2003a) applied a marker-protein-based concept, which allows one to draw conclusions that are independent of the researcher who analyzed the data. In addition, as the reference database of proteomic signatures grows, a direct side-by-side evaluation of the gels becomes a very tedious task, and the marker-protein approach substantially reduces the evaluation efforts. In short, marker proteins were dened as such proteins that were overexpressed at least twofold under antibiotic inuence in two independent experiments, and that made up at least 0.05% of the total protein synthesized during the pulse-labeling period. The number of such marker proteins for each of the 30 antibiotics varied between 0 and 34 with an average of 13.3, and can be used in, for example, cluster analyses to obtain a rst hint to a potential mode of action. In spite of the progress reached and documented in that study, it should still be mentioned that the general applicability of the method for drug-discovery purposes in routine fashion is limited (i) by the time and effort needed to study a novel compound by the experimentally still demanding 2D gels and the evaluation of the massive data sets obtained, and, (ii) due to the number of conditions/compounds/mutants studied so far. Both of those bottlenecks will continue to benet greatly from further
559

&

TZ-OESTERHELT ET AL. BRO

FIGURE 6. Cytoplasmic protein-expression prole of B. subtilis after treatment with a phenyl-

thiazolylurea-sulfonamide (PTU). The autoradiograph of the antibiotic-treated sample (red) was warped by the dual-channel imaging technique onto the untreated control (green). Proteins induced during antibiotic exposure appear in red, and repressed proteins appear in green. The proteomic signature of the PTU included the induction of many proteins previously identied during norvaline (Eymann et al., 2002) or mupirocin (Bandow et al., 2003a) treatment of B. subtilis; for example, Ald, MinD, Spo0A, SpoVG, YurP, and YvyD. Those proteins are known to be positively controlled by the stringent response in this organism (Eymann et al., 2002). However, there were also differences in the protein-expression proles of the two antibiotics. The direct target, the a-subunit of Phe-RS (PheS), was induced in phenyl-thiazolylurea-sulfonamide-treated cells, whereas mupirocin, as expected, did not induce PheS, but rather the corresponding IleS and additional proteins of isoleucine/valine biosynthesis (Bandow et al., 2003a).

technological advances in 2D gel-based and non-gel-based technologies, which will be discussed briey in the following section.

IV. REMARKS ON TECHNOLOGICAL PROGRESS A. Progress in Two-Dimensional Gel-Based Technologies


The introduction of 2D-PAGE in 1975 (Klose, 1975; OFarrell, 1975) marked a major breakthrough in the analysis of the complex protein mixture of whole cells and tissues. Resolution and sensitivity were already high in those original studies, where polyacrylamide tube gels with Ampholines were employed for protein separation according to pI. After isoelectric focusing, proteins were reduced and alkylated in an equilibration step before separation on an SDS gel according to Mr, which is still the standard procedure to this day. Autoradiographs of dried 2D gels of E. coli crude cell extracts allowed the detection of about 1,100 protein species. For direct detection on the gel, proteins were stained with Coomassie Blue. Since its introduction, 2D gelbased proteomics has come a long way. Major technological developments were aimed at enhancing the reproducibility of the separation, increasing sensitivity and resolution, and addressing proteins with physico-chemical properties unfavorable for 2DPAGE. However, only when identication of the proteins from the 2D gel became easier, faster, and cheaper, did the popularity of proteomics begin to increase rapidly.
560

The introduction of immobilized pH gradient (IPG) strips rg, Postel, & Gu nther, 1988) for use in the rst dimension (Go certainly enhanced reproducibility. Different silver-staining protocols (Switzer, Merril, & Shifrin, 1979; De Moreno, Smith, & Smith, 1985; Rabilloud, 1999) offer high sensitivity but do not have a broad linear dynamic range that would allow reliable protein quantication. Fluorescent dyes such as Sypro Ruby (Berggren et al., 1999; Steinberg et al., 1999) are at least as sensitive as silver, are more sensitive than colloidal Coomassie Brilliant Blue staining methods, and offer a linear dynamic range of three orders of magnitude (Patton, 2000). The 2-Dimensional Fluororescence Difference Gel-Electrophoresis (DIGE) technology from Amersham Biosciences (Piscataway, NJ) relies on covalently labeling a small fraction (about 1%) of the proteins in the sample with Cy2, Cy3, or Cy5 dye. Because these dyes have distinct excision and emission wavelengths, up to three samples can be labeled, mixed prior to IEF, and separated in a single 2D gel to further enhance cross-sample comparison by decreasing any gel-to-gel variation. In addition, narrow pI gradient IPG strips were successfully employed to increase protein resolution rg et al., 2000). (Cordwell et al., 2000; Corthals et al., 2000; Go However, although some progress has been made in the separation of basic and poorly soluble proteins such as membrane rg et al., 1999; Molloy et al., 2001; Ohlmeier, Scharf, proteins (Go & Hecker, 2000), those protein species as well as extremely small and large proteins (<15 and >120 kDa) still pose major challenges to the 2D gel technology. Much progress has also been achieved with respect to throughput and automation. The newly launched ZOOM IPGRunner system from Invitrogen

PROTEOMICS AND ANTIBACTERIAL DRUG DISCOVERY

&

(Carlsbad, CA) allows 2D-PAGE separation in 24 hr from rehydration of the sample into the IPG strip for the rst dimension to protein staining. Furthermore, NextGen Sciences Ltd. (Cambridgeshire, UK) recently introduced a fully automated robot that is capable of analyzing three 2D gels at a time under highly reproducible conditions. Finally, although protein quantication is still one of the bottlenecks in the 2D gel-based workow, image analysis software packages have evolved to facilitate the quantication of all protein spots within large sets of 2D gels, requiring different levels of user interaction to ensure data quality. As mentioned above, it was the progress in protein identication from 2D gels that made proteomics attractive to the broader research community. Although a protein patternmatching approach can be successful in nding similarities between protein expression proles for a large number of tested growth conditions, any detailed physiological understanding of the changes in protein composition relies heavily on the identication of the differentially expressed proteins. Protein identication in the early days was time-consuming and expensive, because Coomassie stained proteins had to be sequenced by repeated cycles of Edman degradation (Edman & Begg, 1967). In each cycle phenylisothiocyanate was added to the N-terminal amino acid, and the cyclic amino acid derivative was removed under mild acidic conditions and identied by HPLC. Addition and removal reactions were repeated until the length of the analyzed protein sequence was sufcient to allow the identication of the protein by comparison with available protein- or DNAsequence information. A major drawback besides the lack of sensitivity was that about one-third of the bacterial proteins are N-terminally blocked and, therefore, eluded identication by this method. Two factors arose in the mid-1990s that substantially simplied proteomic analyses. For the rst time, DNA sequences of whole bacterial genomes became available and allowed the prediction of the approximate total number of encoded open reading frames. At the same time, progress in mass spectrometry facilitated the analysis of peptides and small proteins, and the mass accuracy of the measured peptide masses was sufcient to allow peptide mass ngerprinting. Experimentally obtained peptide masses of a digested protein spot were compared to a database that contained all theoretical peptide masses derived from an in silico digestion of the proteins predicted from the genome sequence (Henzel et al., 1993). In addition, recent automation of protein identication signicantly increased throughput (for review, see Godovac-Zimmermann & Brown, 2001; Mann, Hendrickson, & Pandey, 2001; Yarmush & Jayaraman, 2002). Robots have been developed that excise protein spots from 2D gels and transfer the gel plugs into microtiter plates. A digest-robot performs the in-gel tryptic digests directly in the microtiter plate, and a spotting robot applies peptide samples to MALDI targets. MALDI-TOF mass spectrometers acquire the data, and software packages are available to automatically extract peptide masses from the derived spectra, which are submitted to the database search. Ideally, the scientist is left only to do a quick quality check to ensure that the hits from the database match the predominant peaks on the spectra. The same level of automation is also available for mass spectrometric approaches that involve MS/MS peptide sequence elucidation. Although MS/MS is particularly

useful when working with highly complex genomes, where peptide mass ngerprinting is not reliable enough, it is also the method of choice for the identication of those bacterial proteins (especially small ones) that do not yield a sufcient number of peptides after tryptic digestion to ensure an unambiguous identication by peptide mass ngerprinting. In addition, MS/MS de novo sequencing of peptides is extremely useful when working with organisms with incomplete genomic information. Another important application of mass spectrometry-based sequencing is the identication of amino acid residues that carry protein modications (e.g., phosphorylation), which are often crucial for protein activity.

B. Progress in Non-Gel-Based Proteomics


Although 2D gel-based proteomics is the method of choice for many proteomic studies, there are certain limitations to that technique that are mainly based on the wide diversity of physicochemical properties of proteins. It is still difcult to achieve the separation of hydrophobic, or of extremely small or large, proteins. Proteomics applications in the eld of antibacterial research would greatly benet from closing these gaps, and being able to analyze the whole proteome of bacterial cells. Evolving mass spectrometry-based technologies circumvent some of the limitations of protein separation by 2D-PAGE. Protein extracts are subject to tryptic digest, and the complex peptide mixtures are separated by liquid chromatography coupled to mass spectrometric analysis (LCMS). Either 1D gel electrophoresis can be used to reduce the complexity of the protein mixture prior to digestion (Lasonder et al., 2002; Li, Steen, & Gygi, 2003) or in the case of multi-dimensional protein identication technology (MudPit) sample complexity is reduced after digestion by separating the peptides on strong cation-exchange resins and subsequent reversed-phase liquid chromatography (Washburn, Wolters, & Yates, 2001). Multiple approaches utilize heavyisotope labeling for quantication; for instance, 15N-labeling (Oda et al., 1999; Lahm & Langen, 2000), stable-isotope labeling with amino acids in cell culture (SILAC) (Ong et al., 2002), isotope-coded afnity tag (ICAT) technology (Gygi et al., 1999), or enzymatic labeling with 18O during protein digestion (Mirgorodskaya et al., 2000), to name the most prominent. Most of those technologies are still in the proof-of-concept stage, and are currently being compared to 2D-PAGE (e.g., Schmidt et al., 2003) and to each other regarding their benet for proteomics studies. The reader interested in those technologies is referred to some excellent recent reviews (Hamdan & Righetti, 2002; PasaTolic et al., 2002; Lill, 2003; Sechi & Oda, 2003; Tao & Aebersold, 2003; Wu & Yates, 2003; Ong, Foster, & Mann, 2003). Those heavy-isotope labeling techniques combined with chromatography and mass spectrometry hold extreme promise for the proteomics research community because they are capable of qualitative and quantitative analysis of protein samples with no obvious bias towards high solubility or a certain pI. However, high molecular weight proteins were over-represented in a study that compared ICAT technology and the classical 2D-gel approach (Schmidt et al., 2003). MS-based technologies seem to be more sensitive than the classical 2D-gel approach, and they have been shown to yield a good coverage of predicted open561

&

TZ-OESTERHELT ET AL. BRO

reading frames (Florens et al., 2002; Washburn et al., 2002). Different shortcomings are associated with the quantitative analysis of peptides rather than whole proteins: the analysis of the huge number of MS spectra obtained from a mixture of peptides derived from two complex protein samples poses a great challenge to developers of analysis software, and to chromatography. Furthermore, and somewhat in contrast to 2D-gel-based proteomics, the quantication of protein modications is extremely difcult because it requires the modied peptide to be detected and recognized as being modied. In addition, it should be kept in mind that protein quantication with ICAT technology requires the presence of cysteine residues in the protein sequence, because it relies on labeling these cysteines with alkylating agents of different isotope composition. In summary, enormous technical progress has been made in the past decade in gel-based and non-gel-based proteomics technologies, and further progress is still to be expected, that will contribute greatly to the popularity and usefulness of proteomics in the area of drug discovery.

V. OUTLOOK
The increasing resistance development of pathogenic bacteria requires the counteractive development of new antibiotics with novel modes of action and free of cross-resistance to presently applied drugs as an important public health priority. The drugs in use today stem with no exception from traditional approaches of random screening of chemical and natural compound libraries for antibacterial activity. Whereas still in its early days, there is reason to believe that more target-directed, molecular approaches will be instrumental in nding new antibacterial drugs and will help to facilitate the rational selection of compound classes stemming, for example, from the classical screening for antibacterial activity as well as target-based screening. The technical basis for this scenario was laid by the deciphering of the genomes of more than 120 bacterial strains, and on the evolving technologies of gene-expression analysis, in particular transcriptome and proteome technologies. The latter two techniques were themselves crucially dependent on progress in even more basic methodologies such as chip production and MS spectrometry as well as software tools to effectively deal with the large datasets produced by such approaches. There have often been debates whether proteomics or transcriptomics should be the most relevant technique for drug-discovery purposes. For example, proteomics appears to be preferred by many, because proteins are often the direct drug targets, and they also happen to be the effector molecules that mediate and regulate the basic cellular functions. On the other hand, present proteomic technology still does not offer to study the full genomic equivalent of all proteins, whereas transcriptome analyses cover the whole genomic sequence and are also able to produce data at a much higher pace. Nevertheless, transcript expression proling is unable to distinguish between different gene products derived from the same coding region on the genome (due to, e.g., modications, truncations, splice variants). It should also be kept in mind that none of these technologies will deliver novel drugs on their own. As many of such technologies as possible should be applied in combination to provide a deeper biological understanding of a
562

compounds action against a living microorganism. That knowledge will be instrumental in selecting from the many antibacterial molecules available those drugs with a desired and promising biological prole, thereby reducing the target-based attrition rates in later, more costly, stages of development. With respect to proteomics, substantial progress has already been made in elucidating the basic regulatory networks that form the basis for the extraordinary capacity of bacteria to adapt to a diversity of lifestyles and external stress factors. The application of this method for antibacterial drug-discovery purposes, however, is still in its early days. One reason for this phenomenon is the fact that the discovery of novel targets, which is one of the most important applications of proteome studies in other areas of drug discovery, is not so much a bottleneck in antibiotic research, because the pathophysiology of most bacterial infections is relatively well-understood and simple: killing the bacterium or interfering with its growth and, possibly, its virulence is usually all it takes. However, it has become obvious that proteome applications, alone or in concert with transcriptome analysis and other more phenotypic methods, play an increasing role in target validation and mode of action determination of novel compounds and variants of existing compound classes. In particular, those methodologies are very helpful in reducing the time needed to obtain that information, which is important in every drug discovery project. Successful exploitation of those technologies for the antibacterial drug discovery process depends on further progress in three main areas: 1. The data collection, which should be expanded to comprise as many antibacterial compounds with diverse mechanisms of action as possible, to cover, ideally, all relevant targets. Because for novel targets, such reference antibiotics are not always available, the analysis of conditional mutants in such targets should be included. 2. The data analysis tools, which should be optimized or developed to handle the enormous datasets efciently and to facilitate data evaluation in terms of mechanism-specic signatures; for example, by including clustering, chemometric, and articial intelligence approaches. 3. Last but not least, further methodological progress in order to increase the speed, throughput, and reproducibility of 2D gel-based as well as non-gel-based techniques.

REFERENCES
Abergel C, Coutard B, Byrne D, Chenivesse S, Claude JB, Deregnaud C, Fricaux T, Gianesini-Boutreux C, Jeudy S, Lebrun R, Maza C, Notredame C, Poirot O, Suhre K, Varagnol M, Claverie JM. 2003. Structural genomics of highly conserved microbial genes of unknown targets in search of new antibacterial targets. J Struct Funct Genomics 4:141157. Aebersold R, Mann M. 2003. Mass spectrometry-based proteomics. Nature 422:198207. Agabian N, Unger B. 1978. Caulobacter crescentus cell envelope: Effect of growth conditions on murein and outer membrane protein composition. J Bacteriol 133:987994.

PROTEOMICS AND ANTIBACTERIAL DRUG DISCOVERY

&

Alksne LE. 2002. Virulence as a target for antimicrobial chemotherapy. Expert Opin Investig Drugs 11:11491159. Allsop AE. 1998. New antibiotic discovery, novel screens, novel targets, and impact of microbial genomics. Curr Opin Microbiol 1:530534. Apfel CM, Locher H, Evers S, Takacs B, Hubschwerlen C, Pirson W, Page MG, Keck W. 2001. Peptide deformylase as an antibacterial drug target: Target validation and resistance development. Antimicrob Agents Chemother 45:10581064. Appelbaum PC. 2002. Resistance among Streptococcus pneumoniae: Implications for drug selection. Clin Infect Dis 34:16131620. Armitage JP, Dorman CJ, Hellingwerf K, Schmitt R, Summers D, Holland B. 2003. Micro meeting: Thinking and decision making, bacterial style: Bacterial Neural Networks, Obernai, France, 7th12th June, 2002. Mol Microbiol 47:583593. Armstrong GL, Conn LA, Pinner RW. 1999. Trends in infectious disease mortality in the United States during the 20th century. JAMA 281:6166. tz H, Hecker H. 2002. Bacillus subtilis tolerance of moderate Bandow JE, Bro concentrations of rifampin involves the sigma(B)-dependent general and multiple stress response. J Bacteriol 184:459467. tz H, Leichert LIO, Labischinski H, Hecker H. 2003a. Bandow JE, Bro Proteomic approach to understanding antibiotic action. Antimicrob Agents Chemther 47:948955. fe F, Freiberg C, Bro tz H, Hecker ttner K, Hochgra Bandow J, Becher D, Bu M. 2003b. The role of peptide deformylase in protein biosynthesis: A proteomic study. Proteomics 3:299306. Berggren K, Steinberg T, Lauber W, Carroll J, Lopez M, Chernokalskaya E, Zieske L, Diwu Z, Haugland R, Patton W. 1999. A luminescent ruthenium complex for ultrasensitive detection of proteins immobilized on membrane supports. Anal Biochem 279:129143. ttner K, Scharf C, Hecker M. 1999. Dual channel imaging of Bernhardt J, Bu two-dimensional electropherograms in Bacillus subtilis. Electrophoresis 20:22252240. Bernhardt J, Weibezahn J, Scharf C, Hecker M. 2003. Bacillus subtilis during feast and famine: Visualization of the overall regulation of protein synthesis during glucose starvation by proteome analysis. Genome Res 13:224237. Beyer D, Kroll H-P, Endermann R, Schiffer G, Siegel S, Bauser M, Pohlmann tz-Oesterhelt H. J, Brands M, Ziegelbauer K, Haebich D, Eymann C, Bro 2004. Discovery of a new class of bacterial phenylalanyl-tRNA synthetase inhibitors with high potency and broad spectrum activity. Antimicrob Agents Chemother 48:525532. Boeddecker N, Bahador G, Gibbs C, Mabery E, Wolf J, Xu L, Watson J. 2002. Characterization of a novel antibacterial agent that inhibits bacterial translation. RNA 8:11201128. Brands M, Endermann R, Gahlmann R, Kruger J, Raddatz S. 2003. DihydropyrimidinonesA new class of anti-staphylococcal antibiotics. Bioorg Med Chem Lett 13:241245. Bush K. 2002. The impact of b-lactamses on the develoment of novel antimicrobial agents. Curr Opin Invest Drugs 3:12841290. Cho MJ, Jeon BS, Park JW, Jung TS, Song JY, Lee WK, Choi YJ, Choi SH, Park SG, Park JU, Choe MY, Jung SA, Byun EY, Baik SC, Youn HS, Ko GH, Lim D, Rhee KH. 2002. Identifying the major proteome components of Helicobacter pylori strain 26695. Electrophoresis 23: 11611173. Cordwell SJ, Nouwens AS, Verrills NM, Basseal DJ, Walsh BJ. 2000. Subproteomics based upon protein cellular location and relative solubilities in conjunction with composite two-dimensional electrophoresis gels. Electrophoresis 21:10941103. Cordwell SJ, Larsen MR, Cole RT, Walsh BJ. 2002. Comparative proteomics of Staphylococcus aureus and the response of methicillin-resistant and methicillin-sensitive strains to Triton X-100. Microbiology 148:2765 2781. Corthals GL, Wasinger VC, Hochstrasser DF, Sanchez JC. 2000. The dynamic range of protein expression: A challenge for proteomic research. Electrophoresis 21:11041115.

De Moreno MR, Smith JF, Smith RV. 1985. Silver staining of proteins in polyacrylamide gels: Increased sensitivity through combined Coomassie blue-silver stain procedure. Anal Biochem 151:466470. Drlica K, Zhao X. 1997. DNA gyrase, topoisomerase IV, and the 4-quinolones. Microbiol Mol Biol Rev 61:377392. Edman P, Begg G. 1967. A protein sequenator. Eur J Biochem 1:8091. Evers S, Di Padova K, Meyer M, Langen H, Fountoulakis M, Keck W, Gray CP. 2001. Mechanism-related changes in the gene transcription and protein synthesis patterns of Haemophilus inuenzae after treatment with transcriptional and translational inhibitors. Proteomics 1:522 544. Eymann C, Homuth G, Scharf C, Hecker M. 2002. Bacillus subtilis functional genomics: Global characterization of the stringent response by proteome and transcriptome analysis. J Bacteriol 184:25002520. Fischer HP, Brunner NA, Wieland B, Paquette J, Macko L, Ziegelbauer K, Freiberg C. 2004. Identication of antibiotic stress-inducible promoters: A systematic approach to novel pathway-specic reporter assays for antibacterial drug discovery. Genome Res 14:9098. Fleischmann RD, Adams MD, White O, Clayton RA, Kirkness EF, Kerlavage AR, Bult CJ, Tomb JF, Dougherty BA, Merrick JM, McKenney K, Sutton G, FitzHugh W, Fields W, Gocayne JD, Scott J, Shirley R, Liu LI, Glodek A, Kelley JM, Weidman JF, Phillips CA, Spriggs A, Hedblom E, Cotton MD, Utterback TR, Hanna MC, Nguyen DT, Saudek DM, Brandon RC, Fine LD, Fritchman LJ, Fuhrmann JL, Geoghagen NSM, Gnehm CL, McDonald LA, Small KV, Fraser CM, Smith HO, Venter JC. 1995. Whole-genome random sequencing and assembly of Haemophilus inuenzae Rd. Science 269:496512. Florens L, Washburn MP, Raine JD, Anthony RM, Grainger M, Hayness JD, Moch JK, Muster N, Sacci JB, Tabb DL, Witney AA, Wolters D, Wu Y, Gardner MJ, Holder AA, Sinden RE, Yates JR, Carucci DJ. 2002. A proteomic view of the Plasmodium falciparum life cycle. Nature 419:537. Gale EF, Cundliffe E, Reynolds PE, Richmond MH, Waring MJ. 1981. The molecular basis of antibiotic action. 2nd edition. London, UK: Wiley. Glass JI, Belanger AE, Robertson GT. 2002. Streptococcus pneumoniae as a genomics platform for broad-spectrum antibiotic discovery. Curr Opin Microbiol 5:338342. Gmuender H, Kuratli K, Di Padova K, Gray CP, Keck W, Evers S. 2001. Gene expression changes triggered by exposure of Haemophilus inuenzae to novobiocin or ciprooxacin: Combined transcription and translation analysis. Genome Res 11:2842. Godovac-Zimmermann J, Brown LR. 2001. Perspectives for mass spectrometry and functional proteomics. Mass Spectrom Rev 20:157. Gomes SL, Juliani MH, Maia JC, Silva AM. 1986. Heat shock protein synthesis during development in Caulobacter crescentus. J Bacteriol 168:923930. Graves PR, Haystead TAJ. 2002. Molecular biologists guide to proteomics. Microbiol Mol Biol Rev 66:3963. Gray CP, Keck W. 1999. Bacterial targets and antibiotics: Genome-based drug discovery. Cell Mol Life Sci 56:779787. Guay DR. 2001. An update on the role of nitrofurans in the management of urinary tract infections. Drugs 61:353364. Guina T, Purvine SO, Yi EC, Eng J, Goodlett DR, Aebersold R, Miller SI. 2003a. Quantitative proteomic analysis indicates increased synthesis of a quinolone by Pseudomonas aeruginosa isolates from cystic brosis airways. Proc Natl Acad Sci USA 100:27712776. Guina T, Wu M, Miller SI, Purvine SO, Yi EC, Eng J, Goodlett DR, Aebersold R, Ernst RK, Lee KA. 2003b. Proteomic analysis of Pseudomonas aeruginosa grown under magnesium limitation. J Am Soc Mass Spectrom 14:742751. Gygi SO, Rist B, Gerber SA, Turecek F, Gelb MH, Aebersold R. 1999. Quantitative analysis of complex protein mixtures using isotope-coded afnity tags. Nat Biotechnol 17:994999.

563

&

TZ-OESTERHELT ET AL. BRO

rg A, Postel W, Gu nther S. 1988. The current state of two-dimensional Go electrophoresis with immobilized pH gradients. Electrophoresis 9: 531546. rg A, Obermaier C, Boguth G, Weiss W. 1999. Recent developments in Go two-dimensional gel electrophoresis with immobilized pH gradients: Wide pH gradients up to pH 12, longer separation distances and simplied procedures. Electrophoresis 20:712717. rg A, Obermaier C, Boguth G, Harder A, Scheibe B, Wildgruber R, Weiss Go W. 2000. The current state of two-dimensional electrophoresis with immobilized pH gradients. Electrophoresis 21:10371053. Haas M, Beyer D, Gahlmann R, Freiberg C. 2001. YkrB is the main peptide deformylase in Bacillus subtilis, a eubacterium containing two functional peptide deformylases. Microbiology 147:17831791. Hamdan M, Righetti PG. 2002. Modern strategies for protein quantication in proteome analysis: Advantages and limitations. Mass Spectrom Rev 21:287302. Hecker M. 2003. A proteomic view of cell physiology of Bacillus subtilis Bringing the genome sequence to life. Adv Biochem Eng Biotechnol 83:5792. Hecker M, Engelmann S, Cordwell SJ. 2003. Proteomics of Staphylococcus aureusCurrent state and future challenges. J Chromatogr B Analyt Technol Biomed Life Sci 787:179195. lker U. 1996. Heat shock and general stress Hecker M, Schumann W, Vo proteins in Bacillus subtilis. Mol Microbiol 19:417428. lker U. 2001. General stress response of Bacillus subtilis and Hecker M, Vo other bacteria. Adv Microb Physiol 44:3591. Henzel WJ, Billeci TM, Stults JT, Wong SC, Grimley C, Watanabe C. 1993. Identifying proteins from two-dimensional gels by molecular mass searching of peptide fragments in protein sequence databases. Proc Natl Acad Sci USA 90:50115015. Hiramatsu K, Cui L, Kuroda M, Ito T. 2001. The emergence and evolution of methicillin-resistant Staphylococcus aureus. Trends Microbiol 9:486 493. Hoener B, Noach A, Andrup M, Yen TS. 1989. Nitrofurantoin produces oxidative stress and loss of glutathione and protein thiols in the isolated perfused rat liver. Pharmacology 38:363373. Johnson KW, Loand D, Taylor S, Burli R, Gross M, Ayscough A, Moser H, Waller A, East S, Keavey K, Hu W, Girish S, Difuntorum S, Chen H, Garcia M, Hoch U, Clements J. 2003. Second generation PDF inhibitors for respiratory tract infections. Abstract F-1481, 43rd ICAAC, Chicago, IL, 2003. Kaderbhai NN, Broadhurst DI, Ellis DI, Goodacre R, Kell DB. 2003. Functional genomics via metabolic footprinting: Monitoring metabolite secretion by E. coli tryptophan metabolism mutants using FT-IR and direct injection electrospray mass spectrometry. Comp Funct Genome 4:376391. Klose J. 1975. Protein mapping by combined isoelectric focusing and electrophoresis of mouse tissues. Humangenetik 26:231243. Kolker E, Purvine S, Galperin MY, Stolyar S, Goodlett DR, Nesvizhskii AI, Keller A, Xie T, Eng JK, Yi E, Hood L, Picone AF, Cherny T, Tjaden BC, Siegel AF, Reilly TJ, Makarova KS, Palsson BO, Smith AL. 2003. Initial proteome analysis of model microorganism Haemophilus inuenzae strain Rd KW20. J Bacteriol 185:45934602. Kornilovska I, Nilsson I, Utt M, Ljungh A, Wadstrom T. 2002. Immunogenic proteins of Helicobacter pullorum, Helicobacter bilis, and Helicobacter hepaticus identied by two-dimensional gel electrophoresis and immunoblotting. Proteomics 2:775783. Kosower NS, Kosower EM. 1995. Diamide: An oxidant probe for thiols. Methods Enzymol 251:123133. Krueger JH, Walker GC. 1984. groEL and dnaK genes of Escherichia coli are induced by UV irradiation and nalidixic acid in an htpR-dependent fashion. Proc Natl Acad Sci USA 81:14991503. Kuroda M, Ohta T, Uchiyama I, Baba T, Yuzawa H, Kobayashi I, Cui L, Oguchi A, Aoki K, Nagai Y, Lian J, Ito T, Kanamori M, Matsumaru H,

Maruyama A, Murakami H, Hosoyama A, Mizutani-Ui Y, Takahashi NK, Sawano T, Inoue R, Kaito C, Sekimizu K, Hirakawa H, Kuhara S, Goto S, Yabuzaki J, Kanehisa M, Yamashita A, Oshima K, Furuya K, Yoshino C, Shiba T, Hattori M, Ogasawara N, Hayashi H, Hiramatsu K. 2001. Whole genome sequencing of methicillin-resistant Staphylococcus aureus. Lancet 357:12251240. Lahm HW, Langen H. 2000. Mass spectrometry: A tool for the identication of proteins separated by gels. Electrophoresis 21:21052114. Langen H, Takacs B, Evers S, Berndt P, Lahm HW, Wipf B, Gray C, Fountoulakis M. 2000. Two-dimensional map of Haemophilus inuenzae. Electrophoresis 21:411429. Lasonder E, Ishihama Y, Andersen JS, Vermunt A, Pain A, Sauerwein RW, Eling WM, Hall N, Waters AP, Stunnenberg HG, Mann M. 2002. Analysis of the Plasmodium falciparum proteome by high-accuracy mass spectrometry. Nature 419:537542. Lee YM, Almqvist F, Hultgren SJ. 2003. Targeting virulence for antimicrobial chemotherapy. Curr Opin Pharmacol 3:513519. Li J, Steen H, Gygi SP. 2003. Protein proling with cleavable isotope-coded afnity tag (ICAT) reagents: The yeast salinity stress response. Mol Cell Proteomics 2:11981204. Lill J. 2003. Proteomic tools for quantitation by mass spectrometry. Mass Spectrom Rev 22:182194. Lilley KS, Razzaq A, Dupree P. 2002. Two-dimensional gel electrophoresis: Recent advances in sample preparation, detection and quantitation. Curr Opin Chem Biol 6:4650. Limburg E, Gahlmann R, Kroll HP, Beyer D. 2004. Ribosomal alterations contribute to bacterial resistance against the dipeptide antibiotic TAN 1057. AAC 48:619622. Linn T, Losick R. 1976. The program of protein synthesis during sporulation in Bacillus subtilis. Cell 8:103114. Livermore DM. 2003. Linezolid in vitro: Mechanism and antibacterial spectrum. J Antimicrob Chemother 51(Suppl 2):ii9ii16. Lomovskaya O, Watkins WJ. 2001. Efux pumps: Their role in antibacterial drug discovery. Curr Med Chem 8:16991711. Mann M, Hendrickson RC, Pandey A. 2001. Analysis of proteins and proteomes by mass spectrometry. Annu Rev Biochem 70:437473. Mathesius U, Mulders S, Gao M, Teplitski M, Caetano-Anolles G, Rolfe BG, Bauer WD. 2003. Extensive and specic responses of a eukaryote to bacterial quorum-sensing signals. Proc Natl Acad Sci USA 100:1444 1449. Mirgorodskaya OA, Kozmin YP, Titov MI, Korner R, Sonksen CP, Roepstorff P. 2000. Quantitation of peptides and proteins by matrix-assisted laser desorption/ionization mass spectrometry using (18)O-labeled internal standards. Rapid Commun Mass Spectrom 14:12261232. Mollenkopf HJ, Mattow J, Schaible UE, Grode L, Kaufmann SH, Jungblut PR. 2002. Mycobacterial proteomes. Methods Enzymol 358:242256. Molloy MP, Phadke ND, Maddock JR, Andrews PC. 2001. Two-dimensional electrophoresis and peptide mass ngerprinting of bacterial outer membrane proteins. Electrophoresis 22:16861696. Naumann D, Labischinski H. 1990. Process and device for rapid testing of the effects of agents on micro-organisms. International Patent WO 90/ 09454. Neidhardt FC, Ingraham JL, Schaechter M. 1990. Physiology of the bacterial cell: A molecular approach. Sunderland, MA: Sinauer Publishing. pp 351388. Ng WL, Kazmierczak KM, Robertson GT, Gilmour R, Winkler ME. 2003. Transcriptional regulation and signature patterns revealed by microarray analyses of Streptococcus pneumoniae R6 challenged with sublethal concentrations of translation inhibitors. J Bacteriol 185:359370. Nyman TA. 2001. The role of mass spectrometry in proteome studies. Biomol Eng 18:221227. OFarrell PH. 1975. High resolution two-dimensional electrophoresis of proteins. J Biol Chem 250:40074021.

564

PROTEOMICS AND ANTIBACTERIAL DRUG DISCOVERY

&

Oda Y, Huang K, Cross FR, Cowburn D, Chait BT. 1999. Accurate quantitation of protein expression and site-specic phosphorylation. Proc Natl Sci USA 96:65916596. Ohlmeier S, Scharf C, Hecker M. 2000. Alkaline proteins of Bacillus subtilis: First steps towards a two-dimensional alkaline master gel. Electrophoresis 21:37013709. Ong SE, Foster LJ, Mann M. 2003. Mass spectrometric-based approaches in quantitative proteomics. Methods 29:124130. Ong SE, Blagoev B, Kratchmarova I, Kristensen DB, Steen H, Pandey A, Mann M. 2002. Stable isotope labeling by amino acids in cell culture, SILAC, as a simple and accurate approach to expression proteomics. Mol Cell Proteomics 1:376386. Pasa-Tolic L, Lipton MS, Masselon CD, Anderson GA, Shen Y, Tolic N, Smith RD. 2002. Gene expression proling using advanced mass spectrometric approaches. J Mass Spectrom 37:11851198. Patton WF. 2000. A thousand points of light: The application of uorescence detection technologies to two-dimensional gel electrophoresis and proteomics. Electrophoresis 21:11231144. Payne DJ, Wallis NG, Gentry DR, Rosenberg M. 2000. The impact of genomics on novel antibacterial targets. Curr Opin Drug Discov Dev 3:177190. Rabilloud T. 1999. Silver staining of 2-D electrophoresis gels. Methods Mol Biol 122:297305. Reeh S, Pedersen S, Neidhardt FC. 1977. Transient rates of synthesis of ve aminoacyl-transfer ribonucleic acid synthetases during a shift-up of Escherichia coli. J Bacteriol 129:702706. Schmid MB. 2001. Microbial genomicsNew targets, new drugs. Expert Opin Ther Targets 5:465475. Schmidt F, Donahoe S, Hagens K, Mattow J, Schaible UE, Kaufmann SH, Aebersold R, Jungblut PR. 2004. Complementary analysis of the Mycobacterium tuberculosis proteome by two-dimensional electrophoresis and isotope coded afnity tag technology. Mol Cell Proteomics 3:2442. Sechi S, Oda Y. 2003. Quantitative proteomics using mass spectrometry. Curr Opin Chem Biol 7:7077. Shaw KJ, Morrow BJ. 2003. Transcriptional proling and drug discovery. Curr Opin Pharmacol 3:508512. Shaw KJ, Miller N, Liu X, Lerner D, Wan J, Bittner A, Morrow BJ. 2003. Comparison of the changes in global gene expression of Escherichia coli induced by four bactericidal agents. J Mol Microbiol Biotechnol 5:105122. Sievert DM, Boulton ML, Stoltman G, Johnson D, Stobierski MG, Downes FP, Somsel PA, Rudrik JT, Brown W, Hafeez W, Lundstrom T, Flanagan E, Johnson R, Mitchell J, Chang S. 2002. Staphylococcus aureus resistant to vancomycin-United States, 2002. MMWR Morb Mortal Wkly Rep 51:565567. Silva JM, Khan S, OBrien PJ. 1993. Molecular mechanisms of nitrofurantoin-induced hepatocyte toxicity in aerobic versus hypoxic conditions. Arch Biochem Biophys 305:362369. Singh VK, Jayaswal RK, Wilkinson BJ. 2001. Cell wall-active antibiotic induced proteins of Staphylococcus aureus identied using a proteomic approach. FEMS Microbiol Lett 199:7984. Steinberg T, Lauber W, Berggren K, Kemper C, Yue S, Patton W. 1999. Fluorescence detection of proteins in sodium dodecyl sulfatepolyacrylamide gels using environmentally benign, nonxative, saline solution. Electrophoresis 21:497508. Strahilevitz J, Rubinstein E. 2002. Novel agents for resistant Gram-positive infections: A review. Int J Infect Dis 6(Suppl 1):S38S46. Suga H, Smith KM. 2003. Molecular mechanisms of bacterial quorum sensing as a new drug target. Curr Opin Chem Biol 7:586591. Sutton MD, Smith BT, Godoy VG, Walker GC. 2000. The SOS response: Recent insights into umuDC-dependent mutagenesis and DNA damage tolerance. Annu Rev Genet 34:479497.

Switzer RC, Merril CR, Shifrin S. 1979. A highly sensitive silver stain for detecting proteins and peptides in polyacrylamide gels. Anal Biochem 98:231237. Tang CM, Moxon ER. 2001. The impact of microbial genomics on antimicrobial drug development. Annu Rev Genomics Hum Genet 2:259 269. Tao WA, Aebersold R. 2003. Advances in quantitative proteomics via stable isotope tagging and mass spectrometry. Curr Opin Biotech 14:110 118. Thoren K, Gustafsson E, Clevnert A, Larsson T, Bergstrom J, Nilsson CL. 2002. Proteomic study of non-typable Haemophilus inuenzae. J Chromatogr B Analyt Technol Biomed Life Sci 782:219226. Tonella L, Hoogland C, Binz PA, Appel RD, Hochstrasser DF, Sanchez JC. 2001. New perspectives in the Escherichia coli proteome investigation. Proteomics 1:409423. Ueberle B, Frank R, Herrmann R. 2002. The proteome of the bacterium Mycoplasma pneumoniae: Comparing predicted open reading frames to identied gene products. Proteomics 2:754764. VanBogelen RA. 2003. Probing the molecular physiology of the microbial organism, Escherichia coli using proteomics. Adv Biochem Eng Biotechnol 83:2755. VanBogelen RA, Neidhardt FC. 1990. Ribosomes as sensors of heat and cold shock in Escherichia coli. Proc Natl Acad Sci USA 87:5589 5593. VanBogelen RA, Schiller E, Thomas JD, Neidhardt FC. 1999. Diagnosis of cellular states of microbial organisms using proteomics. Electrophoresis 20:21492159. Vandahl BB, Birkelund S, Demol H, Hoorelbeke B, Christiansen G, Vandekerckhove J, Gevaert K. 2001. Proteome analysis of the Chlamydia pneumoniae elementary body. Electrophoresis 22:1204 1223. Walsh C. 2003. Antibiotic resistance. In: AntibioticsActions, origins, resistance. Washington: ASM Press. pp 89155. Washburn MP, Wolters D, Yates JR III. 2001. Large-scale analysis of the yeast proteome by multi-dimensional protein identication technology. Nat Biotechnol 19:242247. Washburn MP, Ulaszek R, Deciu C, Schieltz DM, Yates JR III. 2002. Analysis of quantitative proteomic data generated via multi-dimensional protein identication technology. Anal Chem 74:16501657. Wasinger VC, Cordwell SJ, Cerpa-Poljak A, Yan JX, Gooley AA, Wilkins MR, Duncan MW, Harris R, Williams KW, Humphrey-Smith I. 1995. Progress with gene product mapping of the Mullicutes Mycoplasma genitalium. Electrophoresis 16:10901094. Weidenmaier C, Kristian SA, Peschel A. 2003. Bacterial resistance to antimicrobial host defensesAn emerging target for novel antiinfective strategies? Curr Drug Targets 4:643649. WHO. 2001. Global strategy for containment of antimicrobial resistance. WHO/CDS/CSR/DRS/2001.2. World Health Organization, Geneva, who.int/emc/amrpdfs/WHO_Global_Strategy_English.pdf. Wu CC, Yates JR III. 2003. The application of mass spectrometry to membrane proteomics. Nat Biotechnol 21:262267. Yarmush ML, Jayaraman A. 2002. Advances in proteomic technologies. Annu Rev Biomed Eng 4:349373. Yoshida M, Loo JA, Lepleya RA. 2001. Proteomics as a tool in the pharmaceutical drug design process. Curr Pharm Des 7:291310. Young FS, Neidhardt FC. 1978. Effect of inhibitors of elongation factor Tu on the metabolic regulation of protein synthesis in Escherichia coli. J Bacteriol 135:675686. Zhang Y, Amzel LM. 2002. Tuberculosis drug targets. Curr Drug Targets. 3:131154. Ziebandt AK, Weber H, Rudolph J, Schmid R, Hoper D, Engelmann S, Hecker M. 2001. Extracellular proteins of Staphylococcus aureus and the role of SarA and sigma B. Proteomics 1:480493.

565

Vous aimerez peut-être aussi