Article pubs.acs.org/est
Application of a Label-free, Gel-free Quantitative Proteomics Method for Ecotoxicological Studies of Small Fish Species K. J. Ralston-Hooper,†,‡ M. E. Turner,§ E. J. Soderblom,§ D. Villeneuve,⊥ G. T. Ankley,⊥ M. A. Moseley,§ R. A. Hoke,‡ and P. L. Ferguson*,†,∥ †
Nicholas School of the Environment and ∥Department of Civil and Environmental Engineering, Duke University, Durham, North Carolina, United States ‡ Dupont Haskell Global Centers, Newark, Delaware, United States § Proteomics Core Facility, Duke University School of Medicine, Durham, North Carolina, United States ⊥ National Health and Environmental Effects Research Laboratory, Mid-Continent Ecology Division, United States Environmental Protection Agency, Duluth, Minnesota, United States S Supporting Information *
ABSTRACT: Although two-dimensional electrophoresis (2D-GE) remains the basis for many ecotoxicoproteomic analyses, newer non-gel-based methods are beginning to be applied to overcome throughput and coverage limitations of 2D-GE. The overall objective of our research was to apply a comprehensive, liquid chromatography− tandem mass spectrometry (LC-MS/MS)-based proteomic approach to identify and quantify differentially expressed hepatic proteins from female fathead minnows exposed to fadrozole, a potent inhibitor of estrogen synthesis. Female fathead minnows were exposed to 0 (control), 0.04, and 1.0 μg of fadrozole/L of water for 4 days, and proteomic analysis was performed. Proteins were extracted and digested, and proteolytic peptides were separated via high-resolution one- or two-dimensional (1-D or 2-D) ultrapressure liquid chromatography (UPLC) and analyzed by tandem mass spectrometry. Mass spectra were searched against the National Center for Biotechnology Information (NCBI) ray-finned fish (Actinopterygii) database, resulting in identification of 782 unique proteins by single-dimension UPLC. When multidimensional LC analysis (2-D) was performed, an average increase of 1.9× in the number of identified proteins was observed. Differentially expressed proteins in fadrozole exposures were consistent with changes in liver function, including a decline in concentrations of vitellogenin as well as other proteins associated with endocrine function and cholesterol synthesis. Overall, these results demonstrate that a gel-free, label-free proteomic analysis method can successfully be utilized to determine differentially expressed proteins in small fish species after toxicant exposure.
■
under a specific physiological condition.4 This technique has been used by environmental toxicologists to mine for potential biomarkers of exposure and/or effect on organisms exposed to various environmental contaminants such as endocrinedisrupting chemicals (EDCs),5 polycyclic aromatic hydrocarbons (PAHs) and polychlorinated biphenyls (PCBs),6 flame retardants,7 metals,8 and insecticides.9 The vast majority of these studies have relied on two-dimensional gel electrophoresis (2D-GE) for protein separation and mass spectrometry (MS) for protein identification.10 Two-dimensional electrophoresis remains the most frequently applied technique for protein separation in ecotoxicological studies, but significant limitations are associated with this method. It can be cumbersome and time-consuming, it has low sensitivity, and
INTRODUCTION Ecotoxicologists currently rely on apical end points such as growth, death, and other physiological changes to determine impacts of environmental pollutants on organism health and function. With the improvement of analytical technologies, environmental researchers now have the ability to examine toxicity on a molecular level using transcriptomics, proteomics, and metabolomics. These techniques can provide insight into effects that do not necessarily manifest themselves in physiological responses. Even though these molecular end points can provide new insights into toxicity, linking molecular events to apical end points still remains challenging. In order to better understand toxicity on a molecular level, integrated system biology approaches are now being employed1,2 that provide not only a broad, integrated understanding of toxicity but also a pivotal platform for predictive toxicology. First applied in ecotoxicology by Shepard and Bradley,3 proteomics aims to simultaneously identify and quantify large numbers of proteins expressed by a cell, tissue, or organism © 2012 American Chemical Society
Received: Revised: Accepted: Published: 1091
August 3, 2012 November 21, 2012 November 26, 2012 November 26, 2012 dx.doi.org/10.1021/es303170u | Environ. Sci. Technol. 2013, 47, 1091−1100
Environmental Science & Technology
Article
it results in identification of a limited number of proteins,11 as it is challenged in the analysis of “extreme” proteins (proteins with high/low MW and pI, as well as hydrophobic proteins). A typical 2D-GE gel can visualize 30−50% of the proteome and has a limited dynamic range (2−3 orders of magnitude versus 3−4 orders for gel-free techniques).4 As a result, researchers are beginning to apply non-gel-based methods in place of or in addition to 2D-GE. Various chemical labeling techniques such as isobaric tagging for relative and absolute quantification (iTRAQ) have been used to examine the impacts of EDCs on reproductive function of the fathead minnow (Pimephales promelas).12,13 The major limitation associated with iTRAQ is that available reagents are limited to four or eight unique tags, which limits experimental design by reducing the number of treatments. Moreover, this approach is also challenged for the analysis of complex mixtures due to near-isobaric coeluting peptides, which may confound the ability to use a single peptide to quantify a given protein. In order to reduce the limitations of each method, researchers have employed both iTRAQ (nongel-based) and 2-D difference gel electrophoresis (2D-DIGE; gel-based) to improve proteome identification.14 Implementation of a gel-free, label-free, high-throughput technique will provide another useful approach to gain a system biology perspective on toxicity. Shotgun (bottom-up), label-free proteomics approaches are an attractive alternative to gel-based approaches for ecotoxicological studies. Such techniques involve protein extraction from the sample of interest (cells, tissue, or whole organism) followed by protein digestion, resulting in peptide fragments that are analyzed via one-dimensional (1-D) or 2-D liquid chromatography (LC) coupled with tandem mass spectrometry (MS/MS). These analytical platforms, coupled with sequence databases and state-of-the-art bioinformatics software, provide a direct and high-throughput analysis of complex protein mixtures when compared to 2D-GE. Furthermore, this highthroughput method fills a critical gap within the systems biology approach. Genomics and metabolomics provide a wealth of information and have established high-throughput protocols, unlike ecotoxicological proteomics, which currently does not utilize high-throughput proteomics methodologies that have been widely used during biomedical studies. Endocrine-disrupting chemicals are a comparatively wellstudied class of environmental contaminants due to their potential impacts on reproduction and development in both humans and wildlife. The drug fadrozole (FAD) is a model EDC that is a potent inhibitor of steroidoestrogen synthesis.15 It specifically inhibits aromatase (cytochrome P450 CYP19A), which converts androgens to estrogens in vertebrates.16 FAD has been well-characterized with respect to transcriptomic responses and impacts on 17β-estradiol and vitellogenin concentrations. Studies have shown FAD to be an effective inhibitor of steroid synthesis in fish and have linked it with decreases in estrogen production to effects on reproduction and development.17−21 Since FAD is a relatively specific inhibitor of CYP19A and its effects on fish species have been well characterized, it provides an excellent model compound to evaluate and validate a high-throughput toxicoproteomic method. The goal of the present study was to establish the use of a gel-free, label-free approach as a high-throughput, highcoverage technique to facilitate proteomic studies in a model small fish species such as the fathead minnow (P. promelas) for ecotoxicology research and regulation. An additional goal was
to implement this approach to better understand the impact of EDCs such as FAD on steroidogenesis of small fish species and to complement existing gene and metabolite data.
■
EXPERIMENTAL SECTION Experimental Design and Sample Collection. Experimental design, exposures, and chemical analysis are similar to those previously published.58 Please refer to the Supporting Information for more specific details. Briefly, sexually mature (4−6 month old) female fathead minnows (FHMs) were randomly transferred to aquaria receiving a continuous flow (45 mL/min) of 0, 0.04, 0.2, 1.0, or 5.0 μg of FAD/L. In total, there were three replicate tanks per treatment group for a total sample size of n = 12 females per treatment. After 96 h of exposure, fish were anesthetized in a buffered solution of tricaine methanesulfonate and tissues were collected. Protein Extraction and Digestion. As a compromise between dose coverage and available resources, three experimental treatments were selected for proteomic analysis [0 (control), 0.04, and 1.0 μg of FAD/L; n = 5 for each treatment]. Prior to analysis, liver tissues were removed from −80 °C storage and placed in tubes. Fresh, cold TRIzol solution (Invitrogen, Grand Island, NY) was added to each sample (1 mL/0.1 g of tissue) and the sample was homogenized. TRIzol is a monophasic solution of phenol and guanidine isothiocyanate that can be used to simultaneously isolate DNA, RNA, and proteins from cells and tissues.22 Lysis buffer is the preferred method for the recovery of proteins, especially since TRIzol extractions have demonstrated potential loss of nuclear proteins.23 However, to gain a systems biology perspective on toxicity, which is the ultimate goal of ecotoxicologists, the ability to extract DNA, RNA, and proteins from an individual cell or tissue is highly advantageous. Samples were incubated on ice for 30 min, with vortexing every 10 min. This step was followed by additional incubation while shaking at 750 rpm at 30 °C for 5 min. Afterward, the samples were cooled on ice and centrifuged at 12 000 relative centrifugal force (RCF) for 5 min at 4 °C. The supernatants were removed and placed in new tubes. Chloroform (0.2 mL/0.8 mL of sample) was added to each lysate and the mixtures were vortexed. The new mixtures were centrifuged at 12 000 RCF for 5 min to perform phase separation. The top aqueous layers were removed from the bottom layers (phenol/chloroform) and discarded. Methanol (1.2 mL/0.8 mL of lysate), chilled by placement in a −80 °C freezer for 30 min, was added to the chloroform/phenol layers, and the samples were mixed, incubated for 10 min at 30 °C, and then cooled on ice. The samples were then centrifuged at 12 000 RCF for 10 min. The supernatants were removed and discarded, leaving behind protein pellets that were washed with 1.0 mL of cold methanol. After washing, the methanol was removed and 500 μL of cold methanol was added to each sample, followed by sonication to generate a fine protein particle suspension. The protein pellets were allowed to settle by gentle centrifugation (2000 RCF for 1 min at 4 °C) and the methanol supernatants were removed and discarded. Two hundred microliters of 0.25% Rapigest (Waters, Milford, MA) in 50 mM ammonium bicarbonate was added to each pellet and the samples were vortexed thoroughly and heated to 60 °C for 20 min to aid protein solubilization. Protein content of each sample was quantified by the Bradford assay (Bio-Rad, Hercules, CA). Samples were normalized on the basis of the lowest protein concentration extracted from the liver tissues and prepared for digestion (48 μg of protein). 1092
dx.doi.org/10.1021/es303170u | Environ. Sci. Technol. 2013, 47, 1091−1100
Environmental Science & Technology
Article
Figure 1. (A) Venn diagrams of the total number of proteins identified by a nonlabeled, gel-free proteomics method (LCMSe). (B−D) Principal component analysis (PCA) to visualize treatment effects as well as biological variability of (B) control vs low dose, (C) control vs high dose, and (D) all treatments.
Proteins were reduced by adding 200 mM dithiothreitol (DTT) in 50 mM ammonium bicarbonate to each sample to achieve a final DTT concentration of 10 mM. The samples were then heated to 80 °C for 15 min and allowed to cool for 5 min at room temperature and centrifuged at 12 000 RCF for 30 s to separate the condensate. Iodoacetamide (IAA) (400 mM) was added to the solubilized protein pellet samples to achieve a final IAA concentration of 20 mM. The samples were incubated in the dark for 30 min. Trypsin (Promega, Madison, WI) was added to achieve a 1:50 trypsin:protein ratio (w/w), and the samples were digested overnight at 37 °C. After 18 h, samples were centrifuged at 12 000 RCF for 30 s to separate the condensate, and 10% trifluoroacetic acid (TFA) and 20% acetonitrile were added to achieve 1% and 2% solution concentrations, respectively, followed by the addition of an internal standard (yeast alcohol dehydrogenase; Waters Corp., Milford, MA) to each sample at a concentration of 50 fmol/μg of protein. The sample was shaken for 2 h at 60 °C and then centrifuged at 12 000 RCF for 5 min to remove the precipitated Rapigest formed by the acidification and heating steps. The associated supernatants were removed and placed into labeled autosampler vials (Waters) for ultrapressure liquid chromatography (UPLC) data independent analysis LC-MS analysis. On the basis of protein concentration calculated by the Bradford assay, 1 μg of protein was injected onto the LC-MS/MS instrument. One microgram was selected because it is the optimal amount that maximizes sensitivity and chromatographic peak shape for a 75 μm i.d. reverse-phase (RP) UPLC column. One- and Two-Dimensional Nanoscale Capillary LCMS/MS and Data Analysis. Resulting peptide mixtures were separated via UPLC followed by MSe analysis. Additionally, a
subset of samples was subjected to 2-D LC analysis (n = 1 for each treatment). Both were followed by mass spectral analysis on a Waters Synapt G2 quadrupole time-of-flight (QToF) mass spectrometer operating in a data-independent (DIA or “MSe”) mode of acquisition.24 Protein identifications were performed by searching against the National Center for Biotechnology Information (NCBI) ray-finned fish (Actinopterygii) forward− reverse protein database (downloaded June 2011). Protein quantification (area under the curve, AUC), x-fold-change values, and associated p-values were calculated as a sum of the peak areas of all peptides. Alignment and initial peak area quantitation were performed on the individual isotopomer (i.e., “feature”) level by generating selected ion chromatograms. Following database searching and annotation, all of the feature intensities that belong to a peptide were summed and then all of the peptides that belong to a protein were summed. Once a protein level quantity was determined, all the intensities of that protein from each sample in a treatment group were averaged and then a ratio was calculated against the average of a second treatment group for comparison. p-values were calculated for each protein across the different treatment groups by a Benjamini−Hochberg false discovery rate (FDR) corrected error-weighted analysis of variance (ANOVA) calculation. Principal component analyses (PCA) were performed on zcorrected values by use of JMP statistical software (version 9.0, SAS Institute, Cary, NC). Please refer to Supporting Information for more specific details of LC, MS, and database search conditions.
■
RESULTS AND DISCUSSION Qualitative Proteomic Analysis. During sample analysis, 782 unique proteins and 3419 peptides were identified. Four 1093
dx.doi.org/10.1021/es303170u | Environ. Sci. Technol. 2013, 47, 1091−1100
Environmental Science & Technology
Article
Figure 2. Scatter plot of the log transformed HPLC-MS peak areas of tryptic peptides from proteins of Fathead minnows exposed to 1.0 μg/L fadrozole (y-axis) versus control (x-axis). Red crosses (+) indicate peptides associated with proteins that did not show significant change in expression whereas dark blue crosses (+) indicate those proteins that were up-regulated (greater than 2.0 fold mean change) in fadrozole-exposed fish relative to control. Green crosses (+) represent proteins that were down-regulated in the fadrozole-exposed fish. Extracted ion chromatograms are shown for selected peptides to illustrate the reproducibility and alignment of peptides that are expressed at high (C − Predicted protein LOC 100126107) and low levels (A− Vitellogenin 6) in the fadrozole-exposed fish (light blue peaks) relative to controls (magenta peaks), as well as peptides that that did not show changes in expression (B − 40 S ribosomal protein S8).
Figure 3. Bar graph demonstrating increased protein expression levels for 165 proteins matched across the 1.0 vs 0.04 μg/L FAD doses. Those proteins with negative x-fold mean changes indicate downregulation, while those with positive x-fold mean changes indicate upregulation.
hundred seventy-three proteins had greater than one peptide match per protein, while 309 had only one peptide match (Figure 1A). The coefficient of variation (CV) for the internal standard (yeast alcohol dehydrogenase) across the four quality control (QC) pools during the 2-day analysis was an average of
6.9%. Furthermore, the internal standard expression level across all 15 samples was an average 8.9%, suggesting excellent analytical reproducibility. Figure 2 demonstrates the excellent reproducibility and alignment of peptides across treatment based on extracted ion chromatograms of selected peptides 1094
dx.doi.org/10.1021/es303170u | Environ. Sci. Technol. 2013, 47, 1091−1100
Environmental Science & Technology
Article
and higher-abundance peptides can suppress the signal of lower-abundance peptides. However, ionization suppression can be reduced if the complex mixtures are fractionated prior to MS analysis to minimize coeluting peptides. This is evident based on results from this study. Peptides undergoing further separation via multidimensional analysis (i.e., 2-D) provided a 1.9× average increase in peptide and protein identification over those not undergoing further separation. A 2-D analysis may also identify proteins not typically detected during 2D-GE or other analysis. For example, GTP-binding nuclear protein was identified in all three 2-D analyses but not in the 1-D analyses. Previous studies30,31 reinforce the advantages of incorporating 2-D versus 1-D separation to increase peptide detection, which results in improved protein coverage and identification. To date, however, few proteomic studies in ecotoxicology have performed LC-MS-based analysis, let alone multidimensional separation. Comparison of Proteomic Methods. Researchers in other fields have begun to implement shotgun approaches in lieu of labeled techniques. In this study, a label-free dataindependent method (LCMSe) was utilized. This method was first introduced for biomarker identification by Plumb et al.24,32 and has been used to identify differentially expressed proteins in human fibroblasts,33 rat brain,34 plants,35 platelets,36,37 and human serum.38−40 Even though LCMSe has been previously utilized and validated,41 to our knowledge, it has not yet been applied in the field of ecotoxicology. The fundamental difference between data-independent (DIA, or MSe) and data-dependent (DDA) MS analyses is that a precursor ion is not isolated and selected prior to MS/MS analysis. Precursor and product ions are generated by alternating scans of low and high collision energies, resulting in a low duty cycle.32 This results in greater accuracy in quantification since more data points are acquired across a chromatographic peak. The instrument acquires a data point every 0.8 s versus every 3 s for a typical DDA experiment. Furthermore, research has shown LCMSe to be the most accurate of the label-free techniques,42 and it provides the best sequence coverage and higher number of average peptides per protein identification.43 For example, LCMSe analysis was used to identify lowabundance proteins in the celery secretome because initial DDA analysis was not able to detect proteins of interest.35 More importantly, during this experiment, LCMSe demonstrated high rates of analytical reproducibility (average 6.9%) compared to other proteomic studies.44 The majority of proteomic studies performed in the environmental sciences have focused on 2D-GE, 2D-DIGE, or iTRAQ approaches. Martyniuk et al.45 present a recent review of these as applied in ecotoxicology. A recent review of 2D-GE use in ecotoxicology was presented by Sanchez et al.10 Even though 2D-GE has the ability to isolate upward of 2000 proteins via a single polyacrylamide gel, technical issues such as staining, sensitivity, gel matching, and reproducibility are drawbacks for this technique.4 A shotgun approach is more efficient and effective since the entire proteome, independent of protein molecular weight or isoelectric point, is analyzed. Furthermore, 2D-GE normally takes 10−20 injections per sample whereas LC-MS/MS needs only one injection per sample. Furthermore, 782 proteins were identified during the present study. When compared to previously published ecotoxicological proteomic reports on small fish species, this method was comparable. For example, approximately 2000 proteins were identified and 1200 were quantified by iTRAQ to
expressed at low (Figure 2C) and high levels (Figure 2A) as well as peptides that were unchanged due to treatment (Figure 2B). Average protein intensity CVs for the individual treatments were 37.6%, 57.1%, and 44.3% for the control, 0.04, and 1.0 μg/L treatments, respectively. For the control versus 0.04 μg/L FAD dose comparison, 312 proteins had greater than an absolute 2-fold mean change value when compared to the controls. Of these 312 proteins, 242 were also significant (p-value 2 and a p-value
