Identification of anti-infective targets through comparative proteomics

Efforts to control infectious diseases increasingly benefit from basic research into the mechanisms by which pathogens infect and persist in their hosts. However, there are still few examples of drugs that have been developed by rational approaches, and most compounds in current use derive from leads that were identified serendipitously or through high-throughput screening. Many drugs in current use act by mechanisms that are not understood or upon targets that are not identified. The emerging technology of comparative proteomics has the power to identify novel drug targets and to reveal the mechanisms by which existing drugs act.

The rate at which putative drug targets are identified and validated has increased with advances in molecular biology that culminated in the sequencing of complete pathogen genomes. The first key step in target validation is the demonstration of essentiality, and proteins that have a predictable function are a better bet to clear this hurdle. Given the choice between a protein with a likely essential function and a novel ‘hypothetical’ protein, researchers will usually select the safe bet. This bias at the level of basic research permeates all subsequent stages in drug development. Yet the ‘orphan’ proteins in a pathogen genome – hypothetical proteins that have no homologs in mammals and no known function – will include many of the best drug targets. These are the unique proteins that define pathogens by mediating specific interactions with the host that permit infection and persistence.

A large proportion of all sequenced genomes comprise hypothetical proteins. These are defined as open reading frames for which there is no obvious similarity to other proteins of known function, nor evidence of translation or expression. The great majority of hypothetical proteins are likely to prove to be bona fide, but their abundance presents an enormous challenge to genome annotation. Even for intensely studied model organisms such as yeast, the proportion of hypothetical proteins remains high (∼30% of predicted genes), and is higher still for many divergent microbial and protozoan genomes. Buried within this large number of presently unfathomable genes will be those key few that contribute to infectivity, virulence, immune evasion, transmission and all the other features that contribute to pathogenicity. To identify these essential, divergent proteins by functional characterization of the plethora of hypothetical proteins is impractical. An alternative is a nonhypothesis-driven approach to reveal those genes that play a role in pathogenesis by a global screen for the molecular differences that underpin pathogenic phenotypes such as virulence.

Comparative phenotype screens might focus on any aspect of the complex systems that are elaborated by cells, such as metabolites or signaling networks. Global analysis of differential RNA expression (transcriptomics) is greatly facilitated by the propensity of complementary nucleic acid sequences to hybridize with predictable kinetics. This has enabled the development and widespread use of subtractive hybridization and, particularly, gene array technologies. Such approaches have permitted rapid and quantitative comparison of changes in gene expression that accompany important changes in pathogen phenotype. Transcriptomic screens directly reveal the encoding genes, either by recourse to genome sequence or by facile cloning and DNA sequencing. However, the focus on gene transcription has a major limitation in that the steady-state levels of mature proteins may bear little quantitative or even qualitative resemblance to the levels of their cognate RNA transcripts. In contrast, global comparison of protein complement (proteomics) has the potential to report directly on differences amongst the effector molecules by revealing post-translational protein modifications or changes in protein abundance that are not predictable from DNA sequence or transcript levels.

Comparative proteomics is an emerging field, and many challenges remain to be overcome. Proteomes comprise heterogeneous populations of molecules, with a broad dynamic range of abundance and inherent instability. In contrast to nucleic acids, there are no tools for the amplification of rare proteins. Methods for determining protein sequence are laborious and easily confounded by amino acid modifications. The proteome of even the simplest organisms comprises thousands of proteins, and a proportion of these are so rare, insoluble or unstable that they will not feature in a proteomic analysis. To analyze and identify those proteins that are amenable, they must be resolved from the complex mixture that is produced when cells are lysed. The standard method that is employed is 2D gel electrophoresis (2DE) [1]. This has the advantages of high resolving power and direct relative quantification of protein abundance. Proteins are resolved by charge and mass into spots on a large format gel, and can be excised for analysis. 2DE is a cumbersome technique, and the search for an alternative has led to the development of gel-free systems for proteome resolution. Gel-free approaches have a number of advantages, notably the capacity to identify some of the more rare or less soluble proteins that do not resolve in 2DE. However, the major limitation of gel-free approaches for comparative proteomics is that quantification of relative protein abundance is only possible by differential tagging approaches. Regardless of the method of proteome resolution, mass spectrometry is almost universally employed for protein identification. Where genome data is available, protein identification can often be achieved by the comparatively high-throughput approach of peptide-mass fingerprinting. Complete or partial sequence is now available for more than 100 medical or veterinary pathogens [101]. Where the proteome in question derives from an organism for which genome sequence information is not available, tandem-mass spectrometry approaches are often more successful; however, they are slower and more technically demanding. Since purchase and maintenance of mass spectrometry instrumentation is costly, protein identification will typically be performed in a specialized facility.

2DE, the workhorse technology of comparative proteomics, is a well-established technique that can reveal changes in protein abundance or post-translational modifications that alter protein charge or mass. Combined with mass spectrometry, 2DE has the potential to reveal proteins that are involved in many aspects of pathogenicity. In the recent literature, pathogenic and vector stages of Leishmania have been compared, to identify proteins that are upregulated in the clinically relevant life-cycle stage [2]. Proteomic analysis of rhoptry organelles, implicated in host cell invasion of Toxoplasma gondii, identified a number of proteins whose organellar localization could not have been predicted from genome sequence [3]. Many of the proteins that were highlighted in these examples could not have been selected on the basis of transcriptomic analysis because they do not appear to be subject to transcriptional regulation or display a subcellular localization that is not currently predictable from gene sequence. In another recent example, the response of Campylobacter jejuni to iron limitation was assessed in parallel by both transcriptomic and proteomic approaches; the results of these analyses differed both quantitatively and qualitatively, highlighting how translational and other post-transcriptional control mechanisms influence the accumulation of proteins [4].

Comparative proteomic approaches can also be used to investigate host–pathogen interactions. Serological protein analysis (SERPA), for example, enabled identification of antigens from the parasitic flatworm, Schistosoma haematobium, that are recognized by the sera of infected individuals [5]. Critical mechanisms in disease control or pathogenesis may be dissected by comparing, for example, drug-sensitive and -resistant lines or attenuated and virulent lines. Identification of biomarkers for infectious diseases can also be achieved by comparative proteomics, potentially enabling the development of novel diagnostic approaches [6]. A variety of examples that illustrate the potential of comparative proteomic approaches have been reviewed [7].

High-throughput ‘omics’ approaches are revolutionizing discovery in the life sciences. Conventionally, experiments are designed to test hypotheses that have been formed through a combination of evidence and speculation. The results of such research will be subjectively interpreted in the context of the initial hypothesis and unexpected results may often be sidelined. In contrast, the current limited understanding of the complex systems that control phenotype expression means that surprising proteins are often highlighted in nonhypothesis-driven comparative proteomic screens. A stated goal of such studies is often the identification of novel proteins, and this is a predictable outcome. More problematic is the identification of proteins with known function in apparently irrelevant pathways. The results of such experiments may fit incompletely with existing data and may reveal changes in diverse areas of metabolism, in part because discriminating between primary and secondary effects on protein modulation is a major challenge. Many results of this type do not appear in the literature because they fit no coherent ‘story’. Nevertheless, they hint at unanticipated complexity and cooperativity in the biological systems that the proteome controls. The output of comparative proteomic screens may provide a shortcut to some of the essential and unique proteins that present the best drug targets, but to explain mechanisms of host–pathogen interactions will also require the application of many of the more conventional approaches of biochemistry and molecular biology.