|Home | About | Journals | Submit | Contact Us | Français|
Aims of this chapter on the role of toxicoproteomics in assessing organ-specific toxicity are to define the field of toxicoproteomics, describe its development among global technologies, and show potential uses in experimental toxicological research, preclinical testing and mechanistic biological research. Disciplines within proteomics deployed in preclinical research are described as Tier I analysis, involving global protein mapping and protein profiling for differential expression, and Tier II proteomic analysis, including global methods for description of function, structure, interactions and post-translational modification of proteins. Proteomic platforms used in toxicoproteomics research are briefly reviewed. Preclinical toxicoproteomic studies with model liver and kidney toxicants are critically assessed for their contributions toward understanding pathophysiology and in biomarker discovery. Toxicoproteomics research conducted in other organs and tissues are briefly discussed as well. The final section suggests several key developments involving new approaches and research focus areas for the field of toxicoproteomics as a new tool for toxicological pathology.
Toxicoproteomics applies global protein measurement technologies to toxicology testing and research. Aims of the field are the discovery of mechanisms governing key proteins in critical biological pathways creating adverse drug effects, development of biomarkers, and eventual prediction of toxicity based upon pharmacogenomic knowledge [1–4].
An increasing number of proteomic applications to many established scientific disciplines have generated great interest and enthusiasm in basic biology and medicine, as well as toxicology. There are well over 10 000 publications relating to some aspects or applications of proteomics in the biosciences. However, the numbers of published proteomic studies is quite limited in reporting primary data for drug-mediated adverse reactions and biochemical toxicities that lead to undesirable phenotypes [5–7]. Toxicoproteomics was initially developed under the auspices of toxicogenomics  and proteomics , but it has emerged as its own discipline. Toxicoproteomics is defined by goals of furthering mechanistic understanding of how specific exposures alter protein expression, protein behavior and response to cause injury and disease, but has also been greatly influenced by a growing body of research focusing upon key organs such as liver and kidney. The field has been augmented by tools from proteomics, bioinformatics and other enabling high throughput technologies. Interestingly, it might be argued that the overt pursuit of defining biomarkers as a major objective in toxicoproteomics research may not be appropriate or right. Biomarkers should be a natural progression of excellence in research from elucidating toxic mechanisms or modes of chemical actions in response to acute exposure to toxicants or during the long-term development of diseases caused or influenced by these exposures. Compared with such an important mission, the identification of biomarkers might be or should be a comparatively smaller part of the whole picture of toxicoproteomics. However, as a motivating factor, biomarker and toxicity signature discovery is very high in the minds of those who use proteomics in toxicology [10, 11]. Major drivers in toxicoproteomics are the commercial need to discover markers associated with drug exposure, efficacy or toxicity in the pharmaceutical arena, and also the urgencies of environmental hazard evaluation for the protection of public health. Finally, an overarching principle among all discovery technologies is that eventual placement of protein changes within biochemical pathways and processes will result from a mechanistic understanding of larger biochemical systems and signaling networks. Systems biology has come to represent this wider integration of functional genomics disciplines such as transcriptomics, proteomics, interactomics and metabolomics among organisms [12, 13].
Issues for toxicoproteomics and toxicogenomics in pharmaceutical data submission have been recently reviewed regarding non-clinical safety testing to regulatory agencies . Several issues are still in development for data submission of genomics and proteomics studies to regulatory agencies. These would include data quality standards, wide differences in platforms and data formats, accepted criteria for data validation, relationships to traditional toxicological endpoints, added-value to established biochemical and molecular methods, animal-to-human extrapolation, mechanism of action, impact upon the NOAEL (“no observed adverse effect level”), early compared to adaptive or non-pharmacological responses, limits of bioinformatics algorithms, tools and available databases, and defined metrics of how and when genomic and proteomic data would influence regulatory decisions.
Toxicoproteomics and metabonomics have sometimes been called to task for their seemingly meager contributions to biomarker discovery compared to more well-established clinical chemistry and histopathology indicators. For example, a review of 13 toxicoproteomic and metabonomic studies with various nephrotoxic agents examined them for their respective abilities to determine specificity and sensitivity of renal toxicity. The review concluded that proteomics (and metabonomics) data compared very poorly with traditional methods of blood and urine chemistries and histopathology without significant improvements . However, it is the potential for discovery and new insights into pathobiology and therapeutics that fuels interest in Omics technologies. In defense of these new fields, the same criticism could be levied upon pathology, histology and clinical chemistry for the many years and countless studies required for them to develop from the beginning of the twentieth century to present day capabilities. Although the discovery potential is high for new biomarkers from toxicoproteomics studies, the strategies for conducting proteomic analysis and using such data in drug development, preclinical safety and regulatory submission are far from standardized. The complexity of protein expression, multiple technology platforms, and emerging technical standards are major challenges for continued growth of the field. Researchers are finding that no one platform is best suited for toxicoproteomics research, and that more than one platform may be required for suitable proteome coverage.
If a primary goal for toxicoproteomics is to translate identified protein changes into improved biomarkers and signatures of chemical toxicity , then care must be exercised in designating any protein change observed during toxicoproteomic studies as “new biomarkers”. Part of the challenge arises from an imprecise meaning of the term ‘biomarker’, accounting for its wide variation in use (and misuse) in scientific and regulatory communities [16, 17]. At a biochemical and molecular level, biomarkers can be narrowed down to “singular biological measures with reproducible evidence of a clear association with health, disease, adverse effect or toxicity”. This is a necessarily limited definition for quantitative biochemical or molecular measures. Historical and more current examples of biomarkers are the detection of a single protein such as C-reactive protein in cardiovascular disease , an enzyme activity like alanine aminotransferase activity in liver injury , gene transcription products such as Her2/neu  in breast malignancies, gene mutations/polymorphisms like slow acetylators that affect xenobiotic metabolism  or small molecules/metabolites such as serum glucose, insulin and urinary ketone bodies in pathological or drug-induced diabetes. Many subcategories of biomarkers are also in popular usage, including biological, surrogate, prognostic, diagnostic and bridging biomarkers [16, 22]. Importantly, a major development of the large datasets derived from Omics technologies is the possibility of greater molecular topography compared to a singular biomarker. One of the major tenets of toxicoproteomics and other Omics analyses are that specific patterns of protein changes can comprise a consistent “signature” of toxicity  or a “combinatorial biomarker”  that is robust enough to be observed in spite of variations in biology, experimental design, or technology platforms. This is a critical assumption first, because there is great potential for including nonspecific or indirect protein changes in such a signature, and, secondly, because of the inherent challenges in establishing a causal linkage of multiple protein changes to a toxic or adverse phenotype. Specific descriptions of such toxicity signatures and biomarkers are at an early stage in the field of toxicoproteomics [6, 10].
Proteomics in a global protein analysis mode generally links separation and identification technologies to create a protein profile or differential protein display. Although the focus of proteomics has been grouped in various ways, Figure 1 shows representative subdisciplines of proteomic analysis that provide a means to categorize toxicoproteomics research.
Four factors often shape the manner in which researchers pursue their activities in toxicoproteomics; these include (1) the complex nature of proteins; (2) the particular portion of the proteome targeted for study; (3) the integrative relationship of toxicoproteomics studies with other Omics technologies; and (4) the driving forces behind specific toxicoproteomics projects. Each of these four factors should be considered. First, a primary objective in proteomics is the isolation and identification of individual proteins from complex biological matrixes. In toxicoproteomics analysis, the Tier I of proteomic analysis is to determine individual protein identities (fingerprint, amino acid sequence), their relative (or absolute) quantities and their spatial location within cell(s), tissues and biofluids of interest. Tier II of analysis globally screens for protein functions, protein interactions, three-dimensional structure and specific post-translational modifications. Tiers I and II of proteomic analysis encompass the seven intrinsic attributes of proteins that play a role in toxicoproteomic analysis  as shown in Figure 1. Proteomic platforms (Fig. 2) each vary greatly in their respective abilities to deliver data on all protein attributes simultaneously during one analysis.
Proteome mapping is the most descriptive of proteomic inquiries and usually focuses upon identifying all proteins in the sample or at a cellular location at hand. Profiling experiments necessarily require quantitation (relative to control, or absolute) to be comparative among samples. Implicit in protein mapping and profiling are considerations about the spatial “origin” of the sample. Often, sample origins are the same in profiling experiments for comparability; for example, serums are most comparable to serums, livers to livers and so on. Structural proteomics is usually defined as high-throughput determination of protein structures in three-dimensional space and is often determined by X-ray crystallography and NMR spectroscopy. This definition has been expanded in Figure 1 to included spatial location of proteins within the organism rather than continue to divide proteomic fields by specific levels of protein organization that might range from subcellular, cellular, organ, tissue, organism to species proteomics. A second factor to consider in toxicoproteomics is that the proteomes of most cells, tissues and organs are so vast that, unlike whole genome queries, proteomes cannot be completely analyzed by existing proteomic platforms. By default, toxicoproteomic studies most often analyze only a portion of the proteome contained in typical biological samples. A frequent strategy to broaden protein coverage is to take steps prior to analysis to reduce sample complexity (analyze a portion of the proteome or ‘subproteome’) by such procedures as subcellular fractionation, affinity or adsorptive chromatography or electrophoretic separation. Third, toxicoproteomic analysis may be conducted as an independent activity or alternately as a component of a large, formalized gene expression project for which the study design, type of experimental subjects and the availability or amount of biological specimens may greatly impact sample preparation procedures and proteomic platform selection . Fourth, the forces and individuals driving toxicoproteomics studies such as drug discovery, biophysical and chemical analyses, safety assessment, drug efficacy, absorption-distribution-metabolism-excretion (ADME) properties and clinical trials will greatly influence the study design, analysis and, importantly, biological interpretation of toxicoproteomic data.
The complexity of a “proteome” contained in a biological sample presents numerous challenges for comprehensively describing the seven attributes of protein expression during any single proteomic analysis . The primary aims of proteomic analysis are to (1) achieve maximal coverage of the proteome (i.e., Tier I analysis) in each sample; (2) complete analysis at high throughput; (3) produce an accurate quantitative protein measurement; (4) deliver data and interpretable results in a timely period; and (5) use of discovery-oriented, open platforms.
All proteomic platforms typically share two common capabilities: a means of global separation and a technology for identification of proteins. Identification usually means assignment to international gene (i.e., NCBI) or protein (i.e., Uniprot or Swiss-Prot) identification numbers. The following proteomic platforms represent a brief description of the principal technologies used for separating and identifying proteins during toxicoproteomic studies as summarized in Figure 2.
Two-dimensional (2D) gel electrophoresis systems have been combined with mass spectrometry (MS) in an established and adaptable platform since 1975, and it is the most commonly used proteomic platform to separate and comparatively quantitate protein samples . Current state-of-the-art 2D gels use immobilized pH gradient (IPG) gels to separate proteins first by charge and then subsequently by mass using SDS-polyacrylamide gel electrophoresis (PAGE) for effective separation of complex protein samples. Proteins are separated to sufficiently homogeneity on 2D gels to permit MS identification. Typical IPG gels of 18–24 cm fitted with similarly sized SDS-PAGE gels can separate between 2000 and 3000 proteins. Each spot does not represent a unique protein (i.e., gene product) but often occur as post-translationally modified forms of the same protein. Fluorescent staining is often the most sensitive means of protein detection (nano- to microgram range). After electronic alignment (registration) of stained proteins in 2D gels by image analysis software, intensities of identical protein spots are compared among treatment groups and a ratio (fold change) is calculated for each protein using specialized software. In 2D-difference gel electrophoresis (DIGE)-MS, protein samples to be compared are labeled with either Cy2-, Cy3- and Cy5-based linkers. Labeled samples are mixed together and electrophoresed on the same gel. This procedure minimizes image analysis errors from trying to electronically register different gels since each dye (sample) is read at a different wavelength on the same gel . Up to three or four samples can be run on the same 2D gel.
The combination of 2D-gel separation of proteins with MS provides a ready means of protein identification after protein excision, enzymatic digestion and MS analysis. The 2D-MS platform forms a versatile and discovery-oriented standardized approach for use in toxicoproteomic studies . A downside to this platform is the limited coverage of a proteome that can be realized on 2D gels by even the most sensitive fluorescent stains.
One-dimensional gel-based proteomics platforms, 1D-gel liquid chromatography-tandem MS (LC-MS/MS), may also be extremely effective for protein separation and identification using SDS-PAGE only (i.e., mass separation) with specially pre-processed samples such as immuno-depleted plasma  or cell secretomes . Such pre-processing sufficiently reduces the original protein complexity to allow small amounts of sample protein (micrograms) or serum (microliters) to be resolved to near protein homogeneity in stained protein bands. Bands are enzymatically digested to obtain diagnostic peptides for protein identification after amino acid sequencing by LC-MS/MS.
Multidimensional liquid chromatography (LC-LC) is used to separate protein digests (nano- to mirograms peptides) by charge (strong anion exchange) and hydrophobicity (C18) immediately prior to entry into a tandem mass spectrometer for protein identification . A premier representative of LC-MS/MS proteomics is the multidimensional protein identification technology or “MuDPIT” platform. This approach has also been called “shotgun proteomics” since entire protein lysates are trypsin-digested into thousands of peptide fragments without the need for any fractionation or processing prior to LC-LC separation and MS/MS identification. Advantages of this newer platform are the potential for detection and identification of low abundance proteins that may not be observed in gel-based protein separations. However, the MuDPIT platform is only semi-quantitative. The platform is very effective in proteomics mapping and discovery studies and should find great utility in toxicoproteomics.
Other variations on the LC-MS/MS approach, closely linking LC separation to MS/MS instruments, have incorporated isotopic labeling strategies for protein quantitation and in-depth proteomic profiling of samples. Examples of such platforms are isotope coded affinity tags (ICAT), isobaric tag for relative and absolute quantitation (iTRAQ) and stable isotope labeling with amino acids in cell culture (SILAC). These methods use “light” and “heavy” forms of isotopes in linkers that bind to functional groups of proteins (i.e., cysteines or amino groups) in lysates. SILAC and iTRAQ are particularly effective for metabolic incorporation of “light” and “heavy” forms of amino acids (i.e., 1H:2H/12C:13C/14N:15N) into cellular proteins during cell culture incubations. Although sample throughput is slow and analysis time is lengthy, the protein coverage has been greatly expanded with the development of these new multidimensional proteomic platforms. Careful sample, dose and time selection to a few samples appears to be a successful strategy in achieving the most value from multi-dimensional, quantitative LC-MS/MS platforms.
Mass spectral data derived from shotgun proteomics approaches (e.g., 1D-LC-MS/MS) can also be used for relative or absolute protein quantitation and sample comparison. This can be accomplished without stable-isotope labeling or tagging, in any of several ways that include integration of ion chromatogram intensities [32, 33], spectral counting [34–37], or selected reaction monitoring (SRM) [38, 39]. Comparison of label-free techniques suggests these approaches, like isotope labeling, have their own strengths and weaknesses .
Higgs et al.  developed a comprehensive, fully automated, and label-free approach to relative protein quantification using data from LC-MS/MS analysis of proteolytic protein digests. The platform includes de-noising, mass and charge state estimation, chromatographic alignment, and peptide quantification via integration of extracted ion chromatograms. One important advantage of this technique lies in its ability to identify large numbers of proteins and provide abundance information for all of them, in a statistically robust manner. This approach has been applied to the development of protein biomarkers of cisplatin resistance in human ovarian cancer , and recently to evaluate the effect in the rat nucleus accumbens of ethanol self-administration in the posterior ventral tegmental area of the brain  where 1120 proteins where identified and comparatively quantified. The same technique was used to assess the toxic effect of JP-8 jet fuel exposure on rat alveolar type II epithelial cells, at sublethal levels that are occupationally relevant . In that study 1135 unique proteins were identified with high confidence and quantified. Post hoc bioinformatic analysis of differentially expressed proteins suggested that the decreased cell viability of jet fuel-exposed cells corresponded to significant down-regulation of proteins involved in all manners of cell activity, but predominantly by declines in translational and protein synthetic machinery.
Spectral counting is a method for relative protein quantitation in MS-based experiments, and is based on the observation that the total number of detected peptides identifying a specific protein correlates strongly with the abundance of that protein. Simply put, one counts the total number of proteolytic peptide ions identified by MS/MS for a specific protein and on that basis relative comparisons can be made between samples. It is assumed that the more abundant a particular peptide, the more likely it will be selected by the mass spectrometer's operating software for MS/MS analysis, and the more it will be counted. Although this useful approach has yet to be exploited by toxicologists, it has found successful application in the analysis of protein complexes that yield comparatively small-scale datasets [44, 45] and in studying the effects of lipopolysaccharide treatment in initiating the cellular immune response .
SRM and its plural multiple reaction monitoring  enable “hypothesis-driven” or “candidate-based” analyses of protein expression, in contrast to the “discovery” orientation of most shotgun proteomics efforts. In SRM, only the current of ions with preselected mass to charge ratio (m/z) values are monitored. This improves detection sensitivity by decreasing the detector's response to other ions, thus decreasing the background noise. Inclusion of internal standards, specific isotopically (13C/15N) labeled peptides that are otherwise identical to the candidate peptides, enables absolute quantitation . Although its application in toxicoproteomics studies is just now emerging, SRM has been used toxicologically to detect and quantify the acrolein metabolite (3-hydroxypropyl)mercapturic acid in urine as a biomarker of cigarette-smoke-induced disease  and its potential utility in hypothesis-driven toxicoproteomics applications is outstanding.
Retentate chromatography-MS (RC-MS) is a high-throughput proteomic platform that creates a laser-based mass spectrum from a chemically absorptive surface. The principle of this approach is the adsorptive retention (pico- to nanograms protein) of a subset of sample proteins on a thin chromatographic support (i.e., hydrophobic, normal phase, weak cation exchange, strong anion exchange or immobilized metal affinity supports). The absorptive surfaces are placed on thin metal chips, which can be inserted into a specially modified matrix-assisted laser desorption/ionization (MALDI)-type mass spectrometer. The laser rapidly desorbs proteins from each sample on a metal chip to create a mass spectrum profile.
RC-MS can be performed upon any protein sample but thus far this platform has found greatest utility in the analysis of serum and plasma for disease biomarker discovery [6, 49]. The lead commercial platform of RC-MS proteomic platform is the surface-enhanced laser desorption/ionization time-of-flight (SELDI-TOF)-MS instrument . Analysis of samples is relatively rapid. Only a few microliters of biofluid sample are necessary and hundreds of samples can be screened in a few days. Downsides are that only a fraction of the proteome can be analyzed (i.e., that adsorbs to the particular chemical surface), there are sample reproducibility issues, and protein identification of peaks is not readily achieved without additional analysis . However, the RC-MS approach fits many problem areas as a proteomics discovery tool for defining drug or chemical exposure when rapid screening is needed for hundreds or thousands of pre-clinical or clinical samples.
Protein capture arrays (any mass parallel array of proteins, peptides, capture ligands or adsorptive surfaces for protein analysis) represent a promising new proteomic tool that closely emulates the design for parallel analysis of DNA microarray technology . Many different types of capture molecules can be arrayed (recombinant proteins, aptamers, peptides, drug libraries) but the most prevalent are antibody arrays that directly separate proteins from each other by affinity binding to specific protein targets. Generally, commercial antibody array platforms have widely varying sensitivities (pico- to microgram peptides) that fall into three classes based on targeted proteins: cytokine/chemokine arrays, cellular function protein arrays, and cell signaling arrays. However, antibody arrays are not presently available for any given cell type, biofluid or species. This platform provides a rapid screen for limited sets of proteins that may fit some applications in toxicoproteomics.
The liver is the major organ for biotransformation and elimination of pharmaceutics from the body . As a result, initial toxicoproteomics studies sought insights into drug-induced liver injury using rodent models of toxicity. Animal models of liver toxicity are often selected for prevalence of one phenotype such as necrosis, hepatitis, cholestasis, steatosis, fibrosis, cirrhosis or malignancy, but in fact many of these molecular processes occur simultaneously . The removal from the marketplace of several widely prescribed drugs due to hepatotoxicity has attracted considerable attention that highlight underlying susceptibility factors to drug-induced injury including age, sex, drug-drug interactions, and genetic polymorphism in metabolic pathways involved in activation or disposition of therapeutic drugs . Reactive intermediates produced during metabolism can be toxic or some compounds may dysregulate critical biochemical pathways or functions of the liver .
Chemical and drug-induced hepatic necroses produce reproducible phenotypes representative of compound families and common metabolic activation pathways in preclinical species. For example, a comprehensive determination of bioactivation pathways of organic functional groups on xenobiotics and pharmaceutical reagents has been extensively cataloged in an effort to guide drug design and avoid toxicity . Such considerations provide a rationale for exploring and testing the capabilities of emerging Omics technologies like proteomics and transcriptomics upon acute hepatic injury. Which agents might be worthwhile for toxicoproteomics studies? A recent toxicogenomics study for classifying hepatotoxicants evaluated a representative list of 25 well-known model compounds or substances showing hepatotoxicity during testing . The aim of this preclinical research report was to determine if biological samples from rats treated with various compounds could be classified based on gene expression profiles. Such model agents causing acute hepatonecrotic injury included acetaminophen, bromobenzene, carbon tetrachloride, hydrazine and others. Hepatic gene expression profiles were analyzed using a supervised learning method (support vector machines; SVMs) to generate classification rules. The SVM method was combined with recursive feature elimination to improve classification performance. The goal was to identify a compact subset of probe sets (transcripts) with potential use as biomarkers. DNA microarray data have been generated for each substance in this study . Their list of representative hepatotoxic agents for preclinical testing served as a basis for examining the literature for corresponding toxicoproteomics studies.
Table 1 summarizes available primary data from toxicoproteomics studies. Generally, these studies have been conducted upon representative, model liver and kidney damaging agents relevant to preclinical assessment of toxicity. The agent, proteomic analysis platform, tissue or preparation and brief results for each study are summarized in Table 1. Liver toxicants will be addressed first.
Acetaminophen (N-acetyl-p-aminophenol, APAP) has been one of the most commonly tested agents for inducing hepatic injury in toxicoproteomics studies of the liver. It produces centrilobular hepatic necrosis in most preclinical species. Acute hepatocellular injury from acetaminophen exposure is primarily initiated by CYP2E1 bioactivation to form reactive intermediates such as N-acetyl-p-benzoquinone imine (NAPQI) that deplete glutathione (GSH) and then bind to critical cellular macromolecules . Mitochondria are thought to be primary targets in acetaminophen toxicity with particular attention on the mitochondrial permeability transition . It is worth noting that mitochondrial dysfunction underlies the pathogenesis of several toxicities in preclinical species especially in liver, skeletal and cardiac muscle, and the central nervous system (CNS) . Evidence has also been accumulating for the contribution of non-parenchymal cells such as Kupffer cells, NK cells, neutrophils, and endothelial cells that secrete cytokines and chemokines during acetaminophen-induced liver injury [60–64].
Some of the earliest toxicoproteomics studies using 2D-MS platforms were conducted using standard 2D-MS  analysis as well as the 2D-DIGE-MS platform alone  or in combination with DNA microarrays . Proteomic analysis of livers from these studies in mice identified altered proteins that are known targets for adduct formation such as mitochondrial proteins, heat shock proteins (HSPs), and other structural and intermediary metabolism proteins. A different type of 2D gel separation using a non-equilibrium approach to charge separation of proteins (NEPHGE) found 100–200 differentially expressed proteins in rat liver and HepG2 cells, especially in enzymes involved in intermediate metabolism .
Studies using rat hepatocytes exposed to acetaminophen and analyzed by 2D-MS have found it helpful to concurrently evaluate other cytotoxic pharmaceutical agents such as tetracycline, amiodarone and carbon tetrachloride [69, 70]. These studies found alterations in several metabolic enzymes and identified GSH peroxidase, peroxiredoxins 1 and 2 (PRX1, PRX2), which serve as cellular responsive antioxidative enzymes during toxicant exposure.
One of the first LC-MS/MS studies using ICAT technology that involved acetaminophen toxicity in mouse liver was published in 2005 . It was preceded by an earlier optimization study for ICAT in mouse hepatocytes . This study combined the more comprehensive ICAT analysis procedure with an adept choice of resistant (SJL) and susceptible (C57BL/6) mouse strains to investigate potential susceptibility factors (proteins and pathways) in acetaminophen toxicity . Inherent differences in liver homogenate protein expression levels between resistant SJL and susceptible C57BL/6 mice were found by comparison of hepatic proteomics after vehicle (saline) treatment for 6 h. Of the 1236 proteins identified, 121 were differentially expressed between the two mouse strains. At 6 h after treatment with 300 mg/kg acetaminophen given intraperitoneally, 1632 proteins were identified from which 247 were different between the two strains and 161 proteins were more abundant in the SJL strain. Some of these naturally more abundant proteins (in the absence of toxicant) may have protective roles against toxicity including two- to fourfold increases in lactoferrin, galectin-1, tripeptidyl-peptidase II, proteasomal subunit β-type1 and DnaJ homolog A1. Upon administration of acetaminophen, comparative expression showed that SJL mice expressed from three- to tenfold higher levels of ubiquitin-like 2 (SUMO1) activating enzyme E1B, complement c5, cyclooxygenase 1 (COX-1), peroxiredoxin 1, glucose-regulated protein 170 (Grp170), heat shock protein 70 (Hsp70), glutathione-S-transferase μ-2 (GSTμ-2) and regucalcin. In addition to antioxidant enzyme functions, many of these up-regulated proteins may have a reparative role in degrading denatured and damaged proteins, cell proliferation and regeneration, and cellular stress response. A selective loss of several mitochondrial proteins from susceptible C57BL/6 mice suggested this organelle is particularly vulnerable to acetaminophen-induced hepatic injury.
Carbon tetrachloride produces acute centrilobular hepatic necrosis but has been frequently used in a repeated exposure regimen over several weeks to produce an animal model of liver fibrosis . Activation of hepatic stellate cells from a quiescent vitamin A-storing cell to a myofibroblast-like cell is a key event in excessive accumulation of fibril-forming extracellular matrix proteins and development of liver fibrosis. Proteomic analysis was performed on cellular and secreted proteins of normal and activated rat hepatic stellate cells either in vitro or in vivo after carbon tetrachloride for 8 weeks. Of the 43 altered proteins identified, 27 showed similar changes in vivo and in vitro including up-regulation of calcyclin, calgizzarin and galectin-1 as well as down regulation of liver carboxylesterase 10. These changes were confirmed in fibrotic liver tissues. A compendium of 150 stellate cellular and secreted proteins was identified.
Another carbon tetrachloride fibrosis study conducted a 2D-MS proteomic analysis upon liver tissues from rats exposed to carbon tetrachloride for a period of 4–10 weeks . During this exposure period, collagen deposition and hydroxyproline content of fibrotic livers increased continuously. Differentially expressed proteins from proteomic analysis were categorized as proliferation-related proteins/enzymes (proliferating cell nuclear antigen p120, p40 and cyclin F ubiquitin-conjugating enzyme 7 UBC7), and apoptosis-related proteins, mainly caspase 12, which was absent in the control rats. These researchers found that proliferation- and apoptosis-related proteins are dynamically expressed during different stages of rat liver fibrosis induced by carbon tetrachloride.
Bromobenzene is another model liver toxicant whose metabolism, reactive intermediates, protein adducts and liver toxicity phenotype (centrilobular necrosis) have been well characterized . A transcriptomic and proteomic comparison of bromobenzene conducted after 24-h exposure to a single dose of bromobenzene showed alterations in transcripts and genes involved in drug metabolism, oxidative stress, sulfhydryl metabolism and acute-phase response . Of the 1124 proteins resolved from liver homogenates, 24 proteins were differentially expressed and identified as intermediary or drug metabolism enzymes.
The peroxisome proliferator-activated receptors (PPARs) are ligand-activated transcription factors that modulate lipid and glucose homeostasis . Wyeth 14643 (WY14643) is a hepatic metabolic enzyme inducer and acts as a potent agonist of PPARα, a member of the nuclear hormone receptor superfamily and a key transcriptional regulator of many genes involved in free fatty acid oxidation systems in liver. Global gene and protein expression changes were compared by cDNA microarray of mouse liver and 2D-MS of mouse liver subcellular fractions from B6C3F1 mice treated from 0.5 to 6 months with oxazepam and the peroxisome proliferator, WY14643 . Each compound produces hepatocellular cancer after a 2-year bioassay of dietary exposure. The hypothesis was that each compound would produce cancer by different biochemical pathways and that transcript and protein changes measured prior to tumor formation (up to 6 months) would provide mechanistic insights into carcinogenesis. After 6 months, only 36 transcripts were altered after oxazepam compared to 220 transcripts with the Wyeth compound. Notable genes up-regulated in the signature profile for oxazepam were CYP2B20, Gadd45β, TNFα-induced protein 2 and Igfbp5. Up-regulated genes with WY14643 were cyclin D1, PCNA, Igfbp5, Gadd45β and CideA. Altered expression of over 100 proteins by proteomic analysis showed up-regulation of the cancer biomarker, α-fetoprotein in cytosol, and cell cycle-controlled p38-2G4 protein in microsomes during both treatments. Both transcriptomic and proteomic analyses were deemed complimentary in distinguishing between two chemical carcinogens that appear to proceed through different mechanisms and eventually lead to liver cancer as the common phenotype.
Insights into the therapeutic action of PPARα and PPARγ agonists, WY14643 and rosiglitazone, respectively, were reported in proteomic analysis of the ob/ob animal model of obesity disease . Hepatic protein expression profiles were developed by 2D-MS analysis of lean and obese (ob/ob) mice, and obese mice treated with WY14643 or rosiglitazone. Livers from obese mice displayed higher levels of enzymes involved in fatty acid oxidation and lipogenesis compared to lean mice and these differences were further amplified by treatment with both PPAR activators. WY14643 normalized the expression levels of several enzymes involved in glycolysis, gluconeogenesis and amino acid metabolism in the obese mice to the levels of lean mice. Rosiglitazone only partially normalized levels of enzymes involved in amino acid metabolism. This study used an established mouse model of obesity disease to map metabolic pathways and discriminate between PPARα and PPARγ agonist effects by proteomic analysis.
Hydrazine is a model, cross-species hepatotoxicant used as an industrial reagent and found as a drug metabolite of the structurally related pharmaceuticals, isoniazid (anti-tuberculosis drug) and the anti-hypertensive agent, hydralazine. Hydrazine typically causes initial steatosis, macrovesicular degeneration followed by marked hepatic necrosis. Transcriptomic studies suggest hydrazine initiates a process whereby the production and intracellular transport of hepatic lipids is favored over the removal of fatty acids and their metabolites .
Proteomics studies using 2D-DIGE-MS on the hepatotoxic effects of hydrazine were conducted in rats from 48 to 168 h [81, 82]. In one study, 2D gel patterns from liver were analyzed by principal component analysis (PCA) and partial least squares regression. PCA plots described the variation in protein expression related to dose and time. Regression analysis was used to select ten up-regulated proteins and ten down-regulated proteins that were identified by MS. Hydrazine treatment altered proteins in lipid metabolism, Ca2+ homeostasis, thyroid hormone pathways and stress response. In a second study, low-density cDNA microarrays and 2D-DIGE-MS proteomics of liver tissue and metabonomics analysis of serum was performed from hydrazine-treated rats at 48–168 h . Their findings supported known effects of hydrazine toxicity and provided potential biomarkers of hydrazine-induced toxicity.
Thioacetamide is metabolically activated in liver to produce thioacetamide-S,S-dioxide as a reactive intermediate, which binds to liver macromolecules to initiate centrilobular necrosis . Repeated administration of thioacetamide is an established technique for generating rat models of liver fibrosis and cirrhosis, depending upon dose and length of administration (weeks). A 2D-MS proteomic approach was used to profile liver protein changes in rat receiving thioacetamide for 3, 6 and 10 weeks to induce hepatic cirrhosis . Expression of 59 proteins altered by thioacetamide were identified, including three novel, unannotated proteins. Down-regulation of enzymes were noted in pathways such as fatty acid β-oxidation, branched chain amino acids, and methionine breakdown, which may relate to succinyl-CoA depletion and affect heme and iron metabolism. Increased levels were found for enzymes responding to oxidative stress and lipid peroxidation such as GSTs. Finally, these proteomics data were integrated into a proposed overview model for thioacetamide-induced liver cirrhosis affecting succinyl-CoA and cytochrome P450 production combined with iron release and hydrogen peroxide generation.
In another model of thioacetamide-induced liver cirrhosis in rats, researchers searched for potential serum biomarkers using the SELDI proteomic approach . A weak cation exchange surface was used to analyze serum by SELDI-MS from control (normal) rats, thioacetamide-induced liver cirrhosis rats and rats with bile duct ligation-induced liver fibrosis. A consistently down-regulated 3495-Da protein in cirrhosis samples was one of the selected significant biomarkers. This 3495-Da protein was purified on-chip and was trypsin digested on-chip for MS/MS identification, and was found to be a histidine-rich glycoprotein. This new protein was proposed as a novel preclinical biomarker for the rat cirrhosis model and might eventually prove useful for early clinical detection of liver cirrhosis and classification of liver diseases.
An innovative study involving stellate cell activation by 8-week treatment with thioacetamide utilized a proteomic approach that led to the discovery of a novel protein named STAP for “stellate cell activation-associated protein” . Quiescent and thioacetamide-activated stellate cells were analyzed by 2D-MS [using electrospray ionization-(ESI)-MS/MS] to identify 43 proteins altered during the activation process. Up-regulation of collagen-α1 (I and III), γ-actin, neural cell adhesion molecule (N-CAM), calcyclin, calgizzarin and galectin-1 was detected. In particular, STAP was highly increased both in activated stellate cells and in fibrotic liver tissues induced by thioacetamide treatment. These researchers cloned the STAP gene and found it was a cytoplasmic protein, expressed only in stellate cells, with molecular mass of 21 496 Da and a 40% amino acid sequence homology to myoglobin. Biochemical characterization showed STAP is a heme protein exhibiting peroxidase activity toward hydrogen peroxide and linoleic acid hydroperoxide. These results indicate that STAP is a novel endogenous peroxidase catabolizing hydrogen peroxide and lipid hydroperoxides, both of which have been reported to trigger stellate cell activation and consequently promote progression of liver fibrosis. STAP was postulated to play a role as an anti-fibrotic scavenger of peroxides in the liver.
Blood is one of the most accessible and informative biofluids for specific organ pathology in preclinical studies. Biomarkers that can be assayed in biological fluids from preclinical species may hold relevance to human subjects . The Human Proteome Organization (HUPO) is currently undertaking a comprehensive mapping of soluble human blood elements of the plasma proteome for an improved understanding of disease and toxicity . Results from an international survey of soluble human blood proteins by chromatographic and electrophoretic separation have revealed several thousand resolvable proteins for which MS has provided evidence for over 1000 unique protein identifications [87, 88]. Researchers are also mapping the mouse  and rat  serum and plasma proteomes for use in preclinical and experimental studies. An excellent review has been published for 2D gel mapping of rat serum and rat tissue proteomic studies .
The sensitivity of 2D gel proteomic approaches to detect and measure alterations in the mouse or rat plasma proteomes has only recently been tested by various labs. Researchers have examined changes in the mouse plasma proteome focusing upon inflammation after cutaneous burn injury with superimposed Pseudomonas aeruginosa infection . Up-regulations of inter-α-trypsin inhibitor heavy chain 4 and hemopexin were detected along with other mouse acute-phase proteins, including haptoglobin and serum amyloid A. In another inflammation study, reference maps of the mouse serum proteome were generated by 2D-MS from control animals and from mice injected with lipopolysaccharide (LPS) to induce systemic inflammation, and from mice transgenic for human apolipoproteins A-I and A-II . The greatest changes were noted for haptoglobin and hemopexin.
Finally, a comparative plasma proteome analysis has been reported in which investigators used 1D-Gel LC-MS/MS analysis upon a few microliters of plasma from lymphoma-bearing SJL mice experiencing systemic inflammation . After removal of albumin and immunoglobulins from plasma, these researchers identified a total of 1079 non-redundant mouse plasma proteins; more than 480 in normal and 790 in RcsX-tumor-bearing SJL mouse plasma. Of these, only 191 proteins were found in common. Many of the up-regulated proteins were identified as acute-phase proteins but several unique proteins, including haptoglobin, proteasome subunits, fetuin-B, 14-3-3ζ, and MAGE-B4 antigen, were found only in the tumor-bearing mouse plasma due to secretion or shedding by membrane vesicles, or externalized due to cell death. These results are very encouraging for the effectiveness of a proteomics approach for protein identification from small sample amounts, and for comparative proteomics in animal models of drug-induced toxicity or disease.
The application of serum or plasma protein maps in toxicoproteomics, such as those for serum profiling of liver injury, is just beginning to take shape. A recent study reported identification of serum proteins altered in rats treated with four liver-targeted compounds including acetaminophen, α-naphthylisothiocyanate (ANIT), phenobarbital and WY14643 at early, fulminant, and recovery periods of effect . Nineteen serum proteins were identified as significantly altered from the four studies and among them, five serum proteins were of special interest as serum markers for early hepatic toxicity or functional alterations in rats, including vitamin D-binding protein (group-specific component, Gc-globulin), purine nucleotide phosphorylase (PNP), malic dehydrogenase (MDH), paraoxonase (PON1) and retinol-binding protein (RBP). Some of these proteins may serve as early predictive markers of hepatotoxicity for new drug candidates or may be more sensitive than other conventional methods.
The soluble portion of blood, serum or plasma, is regarded as a complex biofluid tissue. While many organs contribute various proteins as blood solutes, the liver is by far the most productive member of all organs and tissues. The liver parenchyma are often primary targets of drug-induced toxicity, and they also secrete many plasma proteins, which can be measured in preclinical species. Therefore, researchers have studied the secreted proteome of hepatocytes. Secreted proteins were separated and identified from primary rat hepatocytes using a collagen gel sandwich system. Proteomic analysis was conducted using a 1D gel LC-MS/MS procedure. More than 200 secreted proteins were identified; these included more than 50 plasma proteins, several structural extracellular matrix proteins and many proteins involved in liver regeneration. Secretion of two proteins, α1-antitrypsin and α2-macroglobulin, was greatly reduced in aflatoxin B1-exposed hepatocytes. This study provides evidence that proteomic analysis of medium from hepatocyte sandwich culture might represent a new in vitro model and general approach for future discoveries of secreted biomarkers in drug-induced chemical toxicity.
Kidney is a primary organ for preclinical assessment in pharmaceutical development since its metabolic and excretory functions often render it susceptible to drug-induced toxicity . The kidney is a major organ for filtration, reabsorption and secretion to maintain homeostasis of water-soluble salts and small molecules. The organ also has a considerable capacity for biotransformation of drugs and xenobiotics. Specific physiological characteristics of the kidney are localized to specific cell types (i.e., vascular endothelial and smooth muscle cells, mesangial cells, interstitial cells, podocytes, proximal and distal tubular epithelial cells), each of which demonstrates selective susceptibility to toxicity. Renal damage can be due to several different mechanisms affecting different segments of the nephron, renal microvasculature or interstitium. The nature of renal injury may be acute and recoverable. However, other drugs with repeated exposure can produce chronic renal changes that may lead to end-stage renal failure. The ability to perform kidney transplants and other organ replacements have saved many lives but relies on immunosuppressive drug treatment to prevent organ rejection. However, immunosuppressive drugs also run the risk of renal toxicity over time. New nephrotoxic markers amenable for multiple preclinical models and high-throughput screening is a major goal for toxicoproteomic and toxicogenomic technologies (Tab. 1) [94–96].
Some of the groundbreaking studies that initiated the field of toxicoproteomics took place in the mid-1990s and involved investigating the side effects of the immunosuppressant drug, cyclosporine A. Cyclosporine A is a calcineurin inhibitor that has been a mainstay for immunosuppressive therapy following solid-organ transplantation. Cyclosporine A blocks immune responses by inhibiting the calcineurin-dependent dephosphorylation of the nuclear factor of activated T cells (NFAT). However, a dose-dependent nephrotoxicity occurs with high incidence that is characterized by non-histological functional deficits or functional decline, with calcium loss in urine (hypercalcinuria), vascular-interstitial lesions and calcification of renal tubules .
Initial 2D gel studies were conducted in rat liver and kidney samples that showed changes in 48 proteins in these tissues in rats treated with cyclosporine A. An unidentified protein present only in the kidney was uniquely down-regulated . A subsequent 2D gel study of kidney homogenates identified a decrease in the 28-kDa kidney protein as calbindin-D using protein microsequencing. Importantly, this same study, using an ELISA, validated a time-dependent decrease in calbindin expression for up to 28 days of cyclosporine treatment . These toxicoproteomic studies published a decade ago represented an important advance in understanding a part of cyclosporine-induced pathophysiology in kidney.
More recently, the contribution of calbindin-D28k has been clarified by the generation of genetically modified mice. Cyclosporine A-induced hypercalciuria represents two pathophysiological processes: a down-regulation of calbindin-D28k with subsequent impaired renal calcium reabsorption, and a cyclosporine A-induced high turnover bone disease . In addition, there is evidence that one biochemical mechanism underlying cyclosporine A and other calcineurin inhibitors may be a drug-induced mitochondrial dysfunction .
The effects of cyclosporine A on gene up-regulation were advanced by a 2D gel proteomic analysis of newly synthesized [35S]methionine-labeled proteins in murine T cells activated in the absence or presence of cyclosporine A . Remarkably, these investigators found more than 100 proteins not present in resting or activated T cells that could be induced by cyclosporine A exposure. It is important to emphasize that the discovery nature of this proteomics study was capitalized upon (same researchers) with the identification of the corresponding genes under the same treatment conditions using a transcript enrichment technique called “representational difference analyses” . Among the up-regulated transcripts, a new gene was found named CSTAD, for “cyclosporine A-conditional, T cell activation-dependent” gene. CSTAD encodes two proteins of 104 and 141 amino acids that are localized in mitochondria . CSTAD up-regulation is observed in mice after cyclosporine A treatment, suggesting that up-regulation of CSTAD and perhaps many other genes are implicated in cyclosporine A toxicity. Thus, toxicoproteomics has played an important role in furthering the understanding of the critical proteins and biological pathways in cyclosporine A toxicity that should lead to better biomarkers for this important class of pharmaceutics.
The regionally specific structure and function of the kidney renders specialized areas more susceptible to toxicity from exposure to certain pharmaceutical agents. For example, puromycin aminonucleoside is an antibiotic that causes glomerular podocyte necrosis, nephrosis and proteinuria in rodent models . Gentamicin is an aminoglycoside antibiotic that accumulates in proximal tubular epithelia and inhibits cell lysosomal function, producing phospholipidosis and tubular degeneration . Some studies have begun to proteomically characterize specific regions such as the medulla and cortex  or subcellular structures of kidney cell types such as the nucleus . In one study, the nephrotoxic effects of gentamicin on protein expression were studied in rat kidney. Results revealed the identities of more than 20 proteins involved the citric acid cycle, gluconeogenesis, fatty acid synthesis, and transport or cellular stress responses . The authors believe that impairment of energy production and mitochondrial dysfunction were involved in gentamicin-induced nephrotoxicity.
Another approach to studying nephrotoxicity is by proteomic characterization of urine. Proteomic mapping of rat urine proteins studied by 2D-MS resolved 350 protein spots from which 111 protein components were identified including transporters, transport regulators, chaperones, enzymes, signaling proteins, cytoskeletal proteins, pheromone-binding proteins, receptors, and novel gene products . One toxicoproteomics study examined urinary protein expression profiles to gain insight into puromycin-induced kidney toxicity . Nephropathy and proteinuria caused by puromycin aminonucleoside in rats was studied by metabonomics and a 2D-MS proteomic analysis of urinary proteins from 8 to 672 h after dosing. Prior to exposure, major urinary protein (MUP), α2-microglobulin and glial fibrillary acid protein isoforms were the major urinary proteins found in addition to many other unidentified low-mass urinary proteins. Following puromycin treatment, a gradual increase in higher mass proteins was observed on 2D gels, particularly albumin, at 32 h after dosing. By 120 h, albumin, transthyretin and vitamin D-binding protein (Gc) were identified as major urinary proteins from puromycin-induced kidney damage. After 672 h, the urinary protein profile in 2D gels had largely returned to normal. Many of these plasma-derived proteins appearing in the urine over 0–672 h following puromycin were consistent with loss of glomerular integrity and major leakage of plasma protein in urine. This study suggests that urinary proteomics in conjunction with these other techniques, has the potential to provide significantly more mechanistic information than is readily provided by traditional clinical chemistries, and may be a productive means for biomarker discovery of nephrotoxic agents in preclinical species.
The nephrotoxin 4-aminophenol produces severe necrosis of the pars recta of the proximal tubules in the rat, which is thought to occur through formation of a toxic metabolite 1,4-benzoquinone imine . d-Serine, an enantiomer of l-serine, is another model nephrotoxicant that selectively damages the pars recta of proximal tubules in the kidney, which may involve formation of toxic oxidative metabolites . The chemotherapeutic agent, cisplatin is also a model nephrotoxicant and targets different portions of the kidney. It is metabolized to cytotoxic intermediates in proximal tubular epithelial cells and induces necrosis in distal tubules and collecting ducts along with causing mild glomerular toxicity . Proteomic profiling using 2D-MS was used to investigate plasma protein changes in rats treated with 4-aminophenol, d-serine and cisplatin compared to saline controls . Nontoxic isomers, l-serine, and transplatin, were also studied. Many plasma proteins were found that displayed dose- and temporal-dependent response to toxicants. Several isoforms of T-kininogen protein were identified as increasing in plasma at early time points and returning to baseline levels after 3 weeks with each nephrotoxicant but not with nontoxic compounds. In addition, inter-α inhibitor H4P heavy chain was increased in the 4-aminophenol and d-serine studies. A further set of proteins correlating with kidney damage was found to be a component of the complement cascade and other blood clotting factors, indicating a contribution of the immune system to the observed toxicity. It was proposed that T-kininogen may be required to counteract apoptosis in proximal tubular cells to minimize tissue damage following a toxic insult.
In a related study, plasma samples from 4-aminophenol and d-serine treated rats were profiled by 2D-MS, and showed dose- and time-dependent effects of various plasma proteins in response to these nephrotoxicants . One toxicity-associated plasma protein was identified as the cellular enzyme, fumarylacetoacetate hydrolase (FAH), a key component of the tyrosine metabolism pathway. FAH was elevated in the plasma of animals treated with 4-aminophenol and d-serine at early time points and returned to baseline levels after 3 weeks. The protein was not elevated in the plasma of control animals or those treated with the non-toxic isomer, l-serine. The investigators raised the possibility that FAH might serve as a marker of kidney toxicity in preclinical species.
Dichlorovinyl-l-cysteine (DCVC) is a model nephrotoxicant taken up by renal proximal tubular epithelia, where it is bioactivated by renal cysteine conjugate β-lyase to form reactive, cytotoxic intermediates . DCVC is a metabolite of trichloroethylene but can be chemically synthesized for use in experimental studies .
A proteomic study of DCVC toxicity was conducted in LLC-PK1 porcine renal epithelial cells by 2D-DIGE-MS to determine early changes in stress-response pathways preceding focal adhesion disorganization linked to the onset of apoptosis . DCVC treatment caused a greater than 1.5-fold up- and down-regulation of 14 and 9 proteins, respectively, prior to apoptosis. These included aconitase and pyruvate dehydrogenase, and those related to stress responses and cytoskeletal reorganization, such as cofilin, Hsp27, and αβ-crystallin. Most noticeable was a pI shift in Hsp27 on phosphorylation at Ser82. Only inhibition of p38 with SB203580 reduced Hsp27 phosphorylation, which was associated with accelerated reorganization of focal adhesions, cell detachment, and apoptosis. Inhibition of active JNK (JUN N-terminal kinase) localization at focal adhesions did not prevent DCVC-induced phosphorylation of Hsp27. Overexpression of a phosphorylation-defective mutant Hsp27 acted as a dominant negative form and accelerated DCVC-induced focal adhesion changes and onset of apoptosis. Early p38 activation appears to rapidly phosphorylate Hsp27, to maintain cell adhesion and to suppress renal epithelial cell apoptosis. This toxicoproteomics study combines both protein identification and post-translational modification to elucidate critical proteins (Hsp27) and protein attributes (phosphorylation) in critical pathways (p38 stress pathway) to gain insight into mechanisms of renal epithelial cell death.
More recently, de Graauw et al.  used phosphotyrosine proteomics (2-DE plus Western blotting) in cultured rat renal proximal tubule cells to demonstrate that DCVC-induced apoptosis is preceded by changes in the tyrosine phosphorylation status of actin-related protein 2 (Arp2), cytokeratin 8, t-complex protein 1 (TCP-1), chaperone containing TCP-1, and gelsolin precursor. It was concluded that the observed alterations are involved in the regulation of the F-actin reorganization and lamellipodia formation that precede renal cell apoptosis caused by DCVC.
Korrapati et al.  identified proteins indicative of DCVC-induced acute renal failure and autoprotection in mice using conventional, large-format 2-DE, Coomassie brilliant blue-based visualization, and MS/MS. Low-dose exposure (15 mg DCVC/kg i.p.) altered 30 proteins (9 up-regulated; 21 down-regulated) by 1.5-fold or more (at p < 0.01), while the lethal, high-dose exposure (75 mg DCVC/kg i.p.), altered the expression of 210 proteins (84 up-regulated; 126 down-regulated). As expected, when the low-dose exposure preceded the administration of the lethal dose by 72 h (autoprotection), the number and extent of differential protein expression was significantly reduced. The authors examined the 18 most radically altered proteins (>10-fold) and concluded that the DCVC-induced differential expression in proteins (involved in the biochemical mechanisms of renal injury and tissue repair) are implicated in the irreparable loss of renal structure and function after a high dose of DCVC, rather than being involved in recovery of structure and function in autoprotected mice.
This chapter has focused upon the development of toxicoproteomics in liver and kidney injury because of their respective roles in xenobiotic biotransformation and excretion. However, protein profiling studies are being conducted in many other organs and tissues to profile adverse effects of therapeutics. Proteomic approaches are revealing new blood serum and tissue biomarkers in animal models of human neurodegenerative diseases like Parkinson's disease, Alzheimer's disease, and amyotrophic lateral sclerosis [121, 122]. Proteomic studies are being conducted in cardiotoxicity models with doxorubicin  and renin-angiotensin models of hypertension . Comparative protein expression studies are systematically examining testicular toxicity in rats with several reproductive toxicants such as cyclophosphamide, sulfasalazine, 2,5-hexanedione and ethylene glycol monomethyl ether . The effects of formaldehyde on rat lung  and protein adduct formation of 1-nitronaphthalene metabolites in rat lung [127, 128] are being examined by proteomic techniques to provide insights into pulmonary pathology by these agents. Thus, protein mapping and profiling studies are exploring a variety of preclinical animal assessment models of toxicity and disease.
The expectations of Omics technologies in pharmaceutical development are very high but the breakthroughs in drug discovery and improvements over traditional measures in preclinical assessment have not proceeded as quickly as anticipated. This situation is understandable since the platforms for proteomics continue to be in dynamic development. Furthermore, applications to toxicology settings are still being explored to match platform sensitivity for differential protein expression with preclinical biological samples. Many of the published toxicoproteomics reports reviewed here have served as proof-of-principle studies using Tier I proteomic analysis (Fig. 1). The approach has been to examine a well-characterized toxicant(s) and compare proteomics data output with known toxicological endpoints (i.e., serum and urine chemistries, histopathology). These efforts might be described as the “discovery phase” of toxicoproteomics where differential protein expressions are determined in response to compound exposure. However, many of these initial studies have often not been accompanied by any confirmation analysis using ELISA, Western blot, immunohistochemistry or functional assay (i.e., enzymatic activity). Two other areas show a slow progress in toxicoproteomics research. One is in the follow-up “hypothesis-driven research” that further characterizes discovery findings and establishes causal-linkage of toxicant exposure and effect. The other area is in “validation studies” of proposed biomarkers using independent and blinded study samples. However, the full cycle of discovery, focused confirmation analysis and hypothesis-testing for causality is achievable [86, 102, 103, 118]. Extensive validation studies of biomarkers represent a lengthier process.
Future trends in toxicoproteomics studies will see developments in several areas where special attributes of proteins can be exploited by proteomics in preclinical assessment. First, further refinements of MS/MS with intimately integrated multi-dimensional separation schemes will continue to dominate proteomic analysis for identification and quantification. MS instruments and software will become more user-friendly and accessible, such as the recently introduced orbitrap MS/MS instruments. Second, “reduction of sample complexity” or any pre-purification strategy prior to toxicoproteomics analysis will be very useful upon innovative application to appropriate biological samples and problem areas (i.e., immunodepletion of albumin, immunoglobulins in plasma) or research problem areas (i.e., phosphoprotein enrichment in protein signaling). Third, Tier II proteomics will begin to be applied to toxicoproteomics problem areas such as global and targeted protein phosphorylation [129–131] and chemoproteomics  using pharmaceutics or enzyme substrates like ATP  as mass capture-ligands for proteins. Fourth, toxicoproteomics is readily positioned to exploit accessible biofluids (i.e., serum/plasma, urine and cerebral spinal fluid) for biomarker development  and could be combined with transcriptomic analysis of blood leukocytes for a parallel approach in biomarker discovery . Fifth, the astute use of genetically altered animals and cell models will enhance discovery of protein targets and mechanistic insights into adverse drug reactions. Finally, continued efforts for integration of proteomics, transcriptomics and toxicology data to derive mechanistic insight and biomarkers will be a continuing goal to maximize return on the investment in Omics technologies [25, 136, 137].
Challenges for toxicoproteomics in preclinical risk assessment are: use as a discovery tool for specific proteins affected by drug and toxicant action; better understanding of biochemistry and cell biology; and biomarker development. The discipline of proteome mapping will be a different and more complex enterprise from the high-throughput, linear-sequencing activities that have been so useful in mapping of the human genome. While the immensity of mapping and measuring the attributes in any one proteome is a large undertaking, biofluid proteomes such as serum/plasma, urine and cerebrospinal fluid hold the most immediate promise for preclinical assessment in terms of better biomarkers.
Although there are many challenges for toxicoproteomics in preclinical assessment, the opportunities are also close at hand for a greater understanding of toxicant action, the linkage to accompanying dysfunction and pathology, and the development of predictive biomarkers and signatures of toxicity.
This review was supported by the Intramural Research program of the NIH, National Institute of Environmental Health Sciences.