ORIGINAL ARTICLE
Global molecular changes in rat livers treated with RXR agonists: a comparison using transcriptomics and proteomics Peter T. Vedell1,a, Reid R. Townsend2,a, Ming You1, James P. Malone2, Clinton J. Grubbs3, Kirby I. Bland3, Donald D. Muccio4, Venkatram R. Atigadda4, Yang Chen5, Katie Vignola5 & Ronald A. Lubet6 1
Department of Pharmacology, Medical College of Wisconsin, Cancer Center, Milwaukee, Wisconsin, 53226 Department of Internal Medicine, Washington University School of Medicine, St. Louis, Missouri, 63110 3 Department of Surgery, University of Alabama at Birmingham, Birmingham, Alabama, 35294 4 Department of Chemistry, University of Alabama at Birmingham, Birmingham, Alabama, 35294 5 Department of Science Development, Metabolon Research, Triangle Park, North Carolina, 27709 6 Chemoprevention Agent Development Research Group, National Cancer Institute, Rockville, Maryland, 20892 2
Keywords Genomics, liver, metabolomics, proteomics, RXR agaonist Correspondence Clinton J. Grubbs, Chemoprevention Department, University of Alabama at Birmingham, VH-G78B, Box 800, 1670 University Blvd., 1530 3rd Avenue S., Birmingham, AL 35294. Tel: 205-934-6384; Fax: 205-975-5082; E-mail:
[email protected] Funding Information The PCL laboratory is supported, in part, with grants P41RR000954 and UL1 RR024992 from the National Center for Research Resources, a component of the National Institutes of Health (NIH) and NIH Roadmap for Medical Research. Additional funding for this work was provided by The National Cancer Institute (Breast SPORE CA089019) and contract HHSN261200433001C. Received: 6 December 2013; Revised: 5 June 2014; Accepted: 1 July 2014 Pharma Res Per, 2(6), 2014, e00074, doi: 10.1002/prp2.74 doi: 10.1002/prp2.74 a
These authors contributed equally.
Abstract The ability of the retinoid X receptors (RXRs) specific agonists (targretin [TRG] and UAB30) to alter rat liver gene and protein expression was determined using Affymetrix Exon arrays and high-performance liquid chromatography – tandem mass spectrometry (LC-MS/MS). TRG profoundly increases triglycerides levels while UAB30 does not. The expression patterns of transcripts or proteins from rat liver treated with TRG or UAB-30 were different from controls and each other. There were six times more gene transcripts identified than proteins. Differentially expressed RNAs or proteins were mapped into known gene ontology (GO) categories and GeneGo Metacore (KEGG) pathway maps. The GO categories which were highly overrepresented with differentially expressed RNAs (P < 1016) were also overrepresented at the protein level. This high concordance of GO Terms was achieved despite the fact that typically ≤1/3 of the elements identified by gene expression were identified by proteomics. Within these GO categories, the magnitude of alterations induced by RXR agonists at the transcript and protein levels were correlated. When GO categories with moderate overrepresentation (105 < P < 109) were examined, there was greater discordance between the transcript and protein data. Examination of KEGG pathway maps with highly significant changes at both the protein and the RNA levels showed that the individual proteins/ genes altered were often the same and changes were of similar magnitude; while KEGG pathways showed limited statistical significance and exhibited minimal overlap. Finally, metabolomics analysis of liver and serum identified altered expression of metabolites related to fatty acid oxidation and bile acid metabolism that were consistent with transcript/protein changes. We observed significant concordance between genomics and proteomics implying either can identify pathways modulated and can indirectly predict resulting physiologic effects. Abbreviations BHBA, b-oxidation acetylcarnitine and 3-hydroxybutyrate; FDR, false discovery rate; ESI, electrospray ionization; LIT, linear ion trap; GC/MS, gas chromatography/ mass spectrometry; UHPLC/MS/MS, ultrahigh-performance liquid chromatography/
ª 2014 The Authors. Pharmacology Research & Perspectives published by John Wiley & Sons Ltd, 2014 British Pharmacological Society and American Society for Pharmacology and Experimental Therapeutics. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
| Vol. 2 | Iss. 6 | e00074 Page 1
P. T. Vedell et al.
Effects of RXR on RNA and Protein Levels
tandem mass spectrometry; RARs, oid receptors; TRG, targretin; LC-MS/MS, liquid chromatography – tandem mass spectrometry; KEGG, Kyoto encyclopedia of genes and genomes.
Introduction The retinoid X receptors (RXRs) were identified as orphan receptors in the late 1980s based on sequence homology to isolated retinoic acid receptors (RARs) (Chambon 1996). 9-cis-Retinoic acid was identified as a high-affinity agonist for RXRs (Heyman et al. 1992; Levin et al. 1992; Wolf and Phil 2006). RXRs form heterodimers with various nuclear receptors (PPARa, PPARc, PPARd, CAR, LXR, FXR, VDR, TR, etc.). The resulting heterodimers are transcriptional activators of multiple genes (Bastien and Rouchette-Egly 2004; Szanto et al. 2004; Sussman and de Lera 2005). Signaling through RXR heterodimers with PPARs or LXRs modulate fatty acid, triglyceride, and cholesterol metabolism (Lalloyer et al. 2009). RXR agonists can initiate this signaling alone or synergize with an agonist of its NR partner. Thus, the PPAR agonists WY 1638 or clofibrate, or RXR agonists will initiate gene transcription in the PPAR a/RXR heterodimer (Viswakarma et al. 2010). In other RXR heterodimers, signaling occurs only with agonists from the RXR partner (VDR:RXR or TR:RXR) and not RXR agonists. Given these multiple gene changes, questions arise as to how to characterize the biological effects of these agonists in intact animals. The most common approach has been the use of gene array technologies (Wang et al. 2006; Vedell et al. 2013) to measure altered expression of specific transcripts and gene pathways. A second potential approach is quantitative proteomics (Neubert et al. 2008; Andley et al. 2013). A third newer approach toward understanding physiological changes is the use of metabolomics (Evans et al. 2009a,b; Ohta et al. 2009). We recently examined a series of RXR agonists employing gene expression arrays (Vedell et al. 2013). The agonists examined were Targretin (TRG) (the only clinically employed RXR agonist) (Farol and Hymes 2004; Rigas and Dragnev 2005; Gniadecki et al. 2007; Lansigan and Foss 2010) and two RXR-selective agonists, UAB30 and 4-Me-UAB30, (Muccio et al. 1998; Atigadda et al. 2003; Grubbs et al. 2006; Gorman et al. 2007; Kolesar et al. 2010). Recently the 3D structures of RXR homodimers were determined containing either TRG or UAB30 and a coactivator peptide (Boerma et al. 2014). Two interesting biologic characteristics of these RXR agonists are (1) all three agonists prevent the development of ER+ and ER breast cancers in standard rodent models; and (2) while TRG and 4-Me-UAB30 strikingly increase triglycerides levels, UAB30 does not. Prior array studies with these agonists show that hepatic gene expression signatures were
2014 | Vol. 2 | Iss. 6 | e00074 Page 2
different and only partially overlapping. In this study, we examined two of these three agonists (TRG and UAB30) using genomics, proteomics, and metabolomics to more fully analyze their effects and to compare results with the different omic approaches. We excluded 4Me-UAB-30 because it is not being developed clinically. In brief, we (1) examined differences in RNA or protein expression when comparing these two RXR agonists; (2) compared pathways altered using proteomic and RNA expression approaches; (3) measured RNA and protein level changes to determine if there was an overall correlation in the magnitude of the changes; (4) employed pathway databases (Metatcore-GeneGo and Kegg) and the GO biological process databases to compare the detection of expression differences by RNA and protein approaches at the pathway and biological process level; and (5) assessed whether certain physiologic changes observed by metabolomics were reflected in the gene and protein data.
Materials and Methods Animals and liver collection for RNA, proteomic, and metabolomic analysis The animals used and procedures for liver collection were described in Vedell et al. (2013). Briefly, female Sprague– Dawley rats were obtained from Harlan Sprague–Dawley, Inc. (Madison, WI). At 50 days of age, the rats (five per group) were treated with diets supplemented with either TRG (150 ppm), UAB-30 (200 ppm), or vehicle control for a period of 7 days. The doses were based on efficacy of these agents in a mammary cancer chemoprevention rat model (Grubbs et al. 2006). At termination, rats were sacrificed with CO2, and blood immediately obtained from the vena cava and placed on ice. Serum was collected after centrifugation at 5°C. Livers were frozen in liquid N2 and placed in an ultralow freezer (85°C) until assayed.
RNA isolation and microarray analysis The RNA isolation and microarray processing, normalization, and probe filtering were described in Vedell et al. (2013). Of particular relevance for the current work, an F-like statistic, Fs (Cui et al. 2005), was computed for each exon and each RXR agonist to measure significance of evidence for differential expression (RXR-agonist treatment group vs. control). P-values were computed for each of these Fs by permutation (Cui et al. 2005). To obtain a P-value and fold changes at the gene level P-val-
ª 2014 The Authors. Pharmacology Research & Perspectives published by John Wiley & Sons Ltd, British Pharmacological Society and American Society for Pharmacology and Experimental Therapeutics.
P. T. Vedell et al.
ues and log2 fold changes were averaged across exons. Differentially expressed genes were defined by an average overall Fs P-value less than 0.15 and an average absolute fold change greater than 1.2. Measurements of expressions at the gene level for proteins were determined as a summation of quantification at the peptide level as described below. After applying these protein summations to obtain fold changes and P-values, the same criteria were used to determine differential expression for proteins.
Gene set enrichment and pathway analysis Gene set enrichment and pathway analysis was performed using the R/GO.db package (Carlson 2012). The differentially expressed genes were clustered using the R/WGCNA package (Langfelder and Horvath 2008). Heat maps were made by using the R/ctc package (Lucas and Gautier 2005) to convert the R/WGCNA cluster object to Java Tree View format and by using the tree manipulation and image capture features of JavaTreeView (Saldanha 2004). Pathway figures were generated from the pathway maps of the GeneGo Metacore (KEGG) software package (ThomsonReuters 2010).
Protein sample preparation, end protease digestion, and peptide preparation Protein was extracted from cryopulverized rat livers using mass spectrometric compatible anionic detergent (Rapidgest) (Zybailov et al. 2005). The protein from the solubilized tissue lysates were precipitated using a 2D clean-up kit (Cat. no. 80-6484-51; GE Healthcare, Pittsburgh, PA). The protein pellets were solubilized in 20 lL of Tris buffer (100 mmol/L, pH 8.5) containing 8 mol/L urea. The proteins were reduced with 1 mmol/L TCEP (2 lL of a 50 mmol/L solution) (Cat. no. 77720; TCEP bond breaker, 0.5 mol/L solution, Thermo Fisher, Waltham, MA) and placed at room temperature for 30 min. Alkylation of the cysteine residues was performed using iodoacetamide (2.2 lL of a 100 mmol/L solution). After 30 min at room temperature in the dark, the reaction was quenched with 10 mmol/L dithiolthione at room temperature for 15 min. The reduced and alkylated proteins (~30 lL) were digested in 8 mol/L urea with 1 lg of endoproteinase LysC (2 lL of a 0.5 lg/lL stock; Roche, Basel, Switzerland) after an overnight incubation at 37°C. The samples were diluted 1:4 with 100 mmol/L Tris, pH 8.5, trypsin (Cat. no. T6567; Sigma, St. Louis, MO) was added (~1:4 enzyme ratio), and the incubation was continued for 24 h at 37°C. The digests were acidified with aqueous 5% formic acid (3.3 lL) (Cat. no. 56302; Fluka, St. Louis, MO). The peptides were extracted with a conditioned Nutip carbon tip
Effects of RXR on RNA and Protein Levels
(Cat. no. NT3CAR; Glygen, Columbia, MD). The sample was loaded with 50 pipetting cycles. The peptides were recovered by 20 pipetting cycles with 25 lL of elution solution, followed by four washes (20 lL each) of elution solution. The extraction and wash solutions were combined in an autosampler vial (Cat. no. 200 046; SunSri, Rockwood, TN) and dried in a Speed Vac (ThermoSavant, Swanee, GA).
High-resolution nano-LC-MS Peptide mixtures were analyzed using high-resolution nano-LC-MS on a hybrid mass spectrometer consisting of a linear quadrupole ion trap and an Orbitrap (LTQ-Orbitrap XL, Thermo Fisher Scientific, Inc., Waltham, MA). Chromatographic separations were performed using a nanoLC 2D PlusTM (Eksigent, AB-SCIEX, Redwood City, CA) for gradient delivery and a cHiPLC-nanoflex (Eksigent) equipped with a 15 cm 9 75 lm C18 column (ChromXP C18-CL, 3 lm, 120 A, Eksigent). The liquid chromatogram was interfaced to the mass spectrometer with a nanospray source (PicoView PV550; New Objective, Woburn, MA). Mobile phases were 1% FA in water (A) and 1% FA in 60% ACN (B). After equilibrating the column in 98% solvent A (aqueous 1% FA) and 2% of solvent B (ACN containing 1% FA), the samples (5 lL) were injected from auto sampler vials using the LC-system’s auto sampler at a flow rate of 500 nL/min followed by gradient elution (250 nL/min) with solvent B: isocratic at 2% B, 0–5 min; 2–25% B, 5–110 min; 25–80%, 110– 170 min; 80–2%, 170–175; and isocratic at 2% B, 175– 190 min. Total run time, including column equilibration, sample loading, and analysis was 217 min. The maximum injection times for the MS1 scan in the Orbitrap and the LTQ were both 500 msec, and the maximum injection times for the MSn scan in the Orbitrap and the LTQ were 800 and 5000 msec, respectively. The automatic gain control targets for the Orbitrap and the LTQ were 5 9 105 and 3 9 104, respectively, for the MS1 scans and 2 9 105 and 1 9 104, respectively, for the MSn scans. The MS1 scans were followed by six MS2 events in the linear ion trap with collision activation in the ion trap (parent threshold = 10,000; isolation width = 4.0 Da; normalized collision energy = 30%; activation Q = 0.250; activation time = 30 msec). Dynamic exclusion was used to remove selected precursor ions (0.20/+1.0 Da) for 90 sec after MS2 acquisition. A repeat count of 3, a repeat duration of 45 sec, and a maximum exclusion list size of 500 was used. The following ion source parameters were used: capillary temperature 200°C, source voltage 2.5 kV, source current 100 lA, and the tube lens at 79 V. The data were acquired using Xcalibur, version 2.0.7 (Thermo Fisher).
ª 2014 The Authors. Pharmacology Research & Perspectives published by John Wiley & Sons Ltd, British Pharmacological Society and American Society for Pharmacology and Experimental Therapeutics.
2014 | Vol. 2 | Iss. 6 | e00074 Page 3
P. T. Vedell et al.
Effects of RXR on RNA and Protein Levels
MS data processing and protein quantification The LC-MS data processing pipeline is detailed in Table S3. For protein identification, the liquid chromatography – tandem mass spectrometry (LC-MS/MS) files that were acquired using X caliber were processed using Mascot Distiller software (ver 2.0.3). A UNIPROT mouse and rat protein database (downloaded May 2011, with 135,387 sequences) was searched using Mascot software (ver. 2.2.04) with the parameters in Table S5. The protein database searches were further processed using Scaffold software (ver. 3_00_07) and proteins identified using the Protein Prophet Algorithm (Keller et al. 2002). The identified peptide sequences and mass spectrometric data are given in Table S3. For relative protein quantification, the LC-MS unprocessed files were imported into Rosetta ElucidatorTM (ver 3.3; Rosetta Biosoftware, Seattle, WA) for retention time alignment of the peptide ion currents across the chromatographic time window using previously described parameters (Perkins et al. 1999; Karpievitch et al. 2009). The aligned, normalized peptide ion currents were annotated within the alignment software by generating database search files (*.dta) and were annotated. The ion current signals from all charge states for each peptide were concatenated unique using a visual script within the software. The table of peptides and peptide intensities was exported in Excel *.csv format and grouped as individual genes (Table S5).
Metabolic profiling The nontargeted metabolic profiling platform employed for this analysis combined three independent platforms: ultrahigh-performance liquid chromatography/tandem mass spectrometry (UHPLC/MS/MS) optimized for basic species, UHPLC/MS/MS optimized for acidic species, and gas chromatography/mass spectrometry (GC/MS). Samples were processed as described previously (Evans et al. 2009a,b; Ohta et al. 2009). For each sample, equivalent amounts of liver tissue were used liquid handler (Hamilton LabStar, Salt Lake City, UT). Protein was precipitated with methanol containing four standards to determine extraction efficiency. The supernatant was split into equal aliquots for analysis on the three platforms. Aliquots, dried under nitrogen and vacuum desiccated, were subsequently either reconstituted in 50 lL of 0.1% formic acid in water (acidic conditions) or in 50 lL of 6.5 mmol/L ammonium bicarbonate in water, pH 8 (basic conditions) for the two UHPLC/MS/MS analyses or derivatized to a final volume of 50 lL for GC/MS analysis using equal parts bistrimethyl-silyl-trifluoroacetamide and solvent
2014 | Vol. 2 | Iss. 6 | e00074 Page 4
mixture acetonitrile:dichloromethane:cyclohexane (5:4:1) with 5% triethylamine at 60°C for 1 h. In addition, three types of controls were analyzed in concert with the experimental samples: aliquots served as technical replicates. A cocktail of standards spiked into every analyzed sample allowed instrument performance monitoring. Experimental samples and controls were randomized across one platform run day. For UHLC/MS/MS analysis, aliquots were separated using a Waters Acquity UPLC (Waters, Millford, MA) and analyzed using an LTQ mass spectrometer (Thermo Fisher Scientific, Inc.) which consisted of an electrospray ionization (ESI) source and linear ion-trap (LIT) mass analyzer. The MS instrument scanned 99–1000 m/z and alternated between MS and MS2 scans using dynamic exclusion with ~6 scans per second. Derivatized samples for GC/MS were separated on a 5% phenyldimethyl silicone column with helium as the carrier gas and a temperature ramp from 60°C to 340°C and then analyzed on a Thermo-Finnigan Trace DSQ MS (Thermo Fisher Scientific, Inc.) operated at unit mass resolving power with electron impact ionization and a 50–750 atomic mass unit scan range. Metabolites were identified by automated comparison of the ion features in the experimental samples to a reference library of chemical standard entries that included retention time, molecular weight (m/z), preferred adducts, and in-source fragments as well as associated MS spectra, and curated for quality control using software (DeHaven et al. 2010). For statistical analyses and data display purposes, any missing values were assumed to be below the limit of detection and these values were imputed with the compound minimum (minimum value imputation). Statistical analysis of log-transformed data was performed using “R” (http://cran.r-project.org/), a freely available, open-source software package. A Welch’s two-sample t-test was used to identify biochemicals that differed significantly between groups. P ≤ 0.05 were considered statistically significant and P < 0.10 were reported as trends. Multiple comparisons were accounted for by estimating the false discovery rate (FDR) using q-values (Storey and Tibshirani 2003). In our analysis, we classified certain fatty acids as longchain fatty acids. These included palmitate (16:0), stearate (18:0), stearidonate (18:4n3), linoleate (18:2n6), linolenate (a or c; (18:3n3 or 6), nonadecanoate (19:0), 10-nonadecenoate (19:1n9), dihomo-linolenate (20:3n3 or 6), eicosapentaenoate (EPA; 20:5n3), eicosenoate (20:1n9 or 11), dihomo-linoleate (20:2n6), mead acid (20:3n9), arachidonate (20:4n6), docosapentaenoate (n3 DPA; 22:5n3), docosapentaenoate (n6 DPA; 22:5n6), docosahexaenoate (DHA; 22:6n3), docosadienoate (22:2n6), docosatrienoate (22:3n3), and adrenate (22:4n6).
ª 2014 The Authors. Pharmacology Research & Perspectives published by John Wiley & Sons Ltd, British Pharmacological Society and American Society for Pharmacology and Experimental Therapeutics.
P. T. Vedell et al.
Results Identification of expressed genes We initially examined the number of elements (genes, proteins) which were identified in rat liver samples (Fig. S1). There were many more genes identified in the transcriptome approach than in the proteomic approach, both in terms of detected expression and in terms of differential expression versus control. The more complete list of gene expression in control liver, TRG-treated liver, and UAB-30-treated liver have been deposited in GEO (GSE57638). A listing of the individual proteins and transcripts which were significantly modulated in these studies are shown in Tables S4 and S5.
Altered gene or protein expression by TRG and UAB30 We examined the effects on gene and protein expression of TRG and UAB30 using heat maps (Fig. 1). These showed there were many more elements (genes or proteins) whose expression was altered by TRG than by UAB30. Second, although the numbers of genes in overlap were few, there were some genes which show common response to the two agonists. Thus, TRG caused increased expression in about 25% of transcripts while decreasing expression of about 35% transcripts. However, in 35% of transcripts there was more overlap as UAB30 and TRG appeared to induce similar transcript alterations relative to control. In the proteomics study, roughly 50% of the altered protein expression profiles involved decreases in TRG-treated rats while 40% showed involved increases in TRG-treated rats. UAB-30 treated rats showed minimal changes for these proteins. Roughly 10–15% of proteins (the yellow-shaded branch of the dendrogram of Fig. 1B) showed similar changes for TRG and UAB30.
Pathway analysis of altered gene or protein expression employing GO terms GO terms are an ongoing effort to develop ontology for representing properties of gene products (Ashburner et al. 2000). A Fisher-like overpresentation test was applied to sets of genes or proteins with altered expression to see if any sets were overrepresented with genes of any GO terms or KEGG pathway maps. In defining these associations, we have used less stringent statistical criteria (P < 0.15 and absolute FC > 1.2) because of the relatively low statistical power to detect changes with a small number of replicates per group. This less stringent approach allowed one to observe multiple changes associated with the different RXR-agonist treatments in a single pathway even if
Effects of RXR on RNA and Protein Levels
some were of small magnitude. Lists of GO categories that were significantly altered are presented in Tables S1 (TRG) and S2 (UAB-30). Certain of the most overpresented pathways associated with TRG treatment, as determined separately in transcripts and proteins and then ranked based on the Fisher P-value are presented (Table 1A). We found that GO terms highly significantly overrepresented (P < 1016) at the transcript level were similarly highly significantly overrepresented at the protein level (Table 1A). Note however most of the listed GO terms are relatively large (hundreds of elements) and the arrays routinely detected >85% of these terms. In contrast, proteomics typically identified 1.2) most elements in these highly altered pathways were not changed. Thus, alterations in most GO term pathways are mediated by altered expression of a more limited number of key elements (genes/proteins). The complete list of transcripts and proteins altered in these studies are presented in Tables S4 and S5. There are fewer protein changes due primarily to fact we identified fewer proteins. However, if the same transcript and protein were altered, with similar magnitude of changes, there was a tendency for the protein changes to be much more significant statistically. There appeared to be lower variability in measuring proteins than transcripts.
Effects of RXR on RNA and Protein Levels
In contrast, when pathways not highly modulated by either RNA or protein expression are examined (i.e., P > 1010), these pathways were often not identified by the complementary method (Tables 1B and 2). In general, limited numbers of modulated transcripts for a given pathway yielded few altered proteins. For UAB-30 (Table 2), although most pathways were not strongly altered, transcripts and protein analysis reflected GO terms cellular respiration; xenobiotic metabolism. These pathway results can be used for two further comparisons of genomic and proteomic results. First, the correlation between the magnitude of transcript and protein changes (Fig. 2). Examining fatty acid oxidation (Fig. 2A) the GO term for this process has over 330 elements, >90 exhibited differential transcript expression, while roughly 30 exhibited differential protein expression. Comparing altered elements that overlapped, one finds a strong correlation between the magnitudes of the transcript/protein changes. This correlation was illustrated for three additional GO biological processes (Fig. 2B–D), and similarly showed a strong correlation. In order to present the overlap in an easier visual presentation, we looked at GeneGO (KEGG) pathways maps (Figs. 3 and 4). Since these are smaller pathways one can present specific genes or proteins. Because these pathways have fewer elements than typical GO terms, there is less statistical power which may explain the tendency toward lower statistical significance for KEGG pathways relative to GO terms. Interestingly, a relatively high percentage of the genes and proteins modulated in the butanoate and carboxylic acid terms (Fig. 3A and B) and the oxidative phosphorylation (Fig. 4B) overlapped one another. This is in contrast to our overall degree of overlap of proteins and genes (Fig. S1). In these cases, however, we had chosen GeneGo pathway maps which were highly significant for changes at both the transcript and protein levels. For retinol metabolism (Fig. 4A) where there was limited statistical significance at the transcript level, and none for protein, we find minimal overlap. Metabolomic changes can serve as a surrogate for important biochemical and physiological processes (Fig. 5A and B). Metabolomics demonstrated perturbations in fatty acid and bile acid metabolism by TRG, while UAB30 exhibited limited effects. These changes in fatty acid and bile acid metabolism are consistent with the differential gene and protein expression related to these pathways that we observed by GO categories (Table 1A). Formation of malo-
Figure 3. Differential expression due to targretin treatment within GeneGo pathway maps. The GeneGo pathway maps for butanoate metabolism (A) and mitochondrial unsaturated b fatty acid oxidation metabolism (B) are shown. The thermometer-like shapes indicate relative fold changes (red: positive, blue: negative) for targretin relative to controls for proteomics (index 1) and transcriptome (index 2) where the differential expression criteria are met.
ª 2014 The Authors. Pharmacology Research & Perspectives published by John Wiley & Sons Ltd, British Pharmacological Society and American Society for Pharmacology and Experimental Therapeutics.
2014 | Vol. 2 | Iss. 6 | e00074 Page 11
Effects of RXR on RNA and Protein Levels
P. T. Vedell et al.
(A)
(B)
2014 | Vol. 2 | Iss. 6 | e00074 Page 12
ª 2014 The Authors. Pharmacology Research & Perspectives published by John Wiley & Sons Ltd, British Pharmacological Society and American Society for Pharmacology and Experimental Therapeutics.
P. T. Vedell et al.
Effects of RXR on RNA and Protein Levels
nyl-CoA catalyzed by acetyl-CoA carboxylase in the cytosol is the committal and rate-limiting step for fatty acid biosynthesis. Malonylcarnitine, a metabolite whose levels reflect that of malonyl-CoA was increased in liver from TRG-treated rats reflecting increased fatty acid biosynthesis. This observation coupled with significantly decreased levels of long-chain fatty acids parallels increased triglycerides in TRG-treated rodents and humans. These alterations reflect increases in expression of various genes involved in fatty acid synthesis; for example, Scd1 which we previously observed in the livers of TRG-treated rats (Vedell et al. 2013). Malonyl-CoA can also inhibit the activity of carnitine acyltransferase I, which is essential for moveMalonylcarnitine - L
Components of triglycerides
3.5
Glycerol 3-phosphate (G3P) - L
1.4
3
1.2
2.5
1
2
0.8
1.5
0.6
• Targretin increased triglyceride levels
0.4
1
0.2
0.5
0
0 Control
Targretin
UAB-30
Eicosenoate (20:1n9 or 11) - L
1.4
ment of long-chain fatty acids to the mitochondrial interior for b-oxidation. In the mitochondria, acylcarnitines decrease the b-oxidation free fatty acids. Metabolomics showed that tissue from TRG-treated animals had significantly increased levels of the acyl-carnitines, palmitoylcarnitine, stearoylcarnitine, and oleoylcarnitine and decreased levels of two end products of BHBA. Hepatic levels of acylcarnitine and fatty acid b-oxidation products were not altered following UAB30 treatment. These data suggest that TRG, but not UAB30, treatment activates fatty acid synthesis, which inhibits fatty acid b-oxidation and provides a physiological context for their differential effects on the levels of triglycerides (Grubbs et al. 2006).
Control
1.4
1.2
1.2
1
1
0.8
0.8
0.6
0.6
Targretin
UAB-30
• Observed elevated fatty acid synthesis, which inhibited β-oxidation and caused the build-up of acyl-carnitine intermediates
Dihomo-linoleate (20:2n6) - L
triacylglycerides membrane phospholipids
0.4
0.4
0.2
0.2
free fatty acid
0
0 Control
Targretin
UAB-30
Control
Targretin
fatty acid s ynthesis
UAB-30
Acyl-carni nes
malonyl-CoA Stearoylcarnitine - L
3 2.5 2 1.5 1 0.5 0 Control
cytosol mitochondria
UAB-30
Control
Targretin
UAB-30
3-hydroxybutyrate (BHBA) - L
citrate
acetyl-CoA
TCA cycle
pyruvate glycolysis
0.8
0.5
0.4
0
0 Control
Targretin
UAB-30
Cholesterol - L
Control
1.2
7
1
6
Targretin
UAB-30
7-alpha-hydroxycholesterol - L
4
0.6
3
0.4
7α-hydroxycholesterol glycocholic acid glycochenodeoxycholic acid
2
0.2
1 0
0 Control
Targretin
UAB-30
Taurocholate - L
Control
UAB-30
Glycocholate - L
6
1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0
Targretin
urinary excretion
taurocholic acid taurochenodeoxycholic acid tauromuricholic acid
glycodeoxycholic acid glycine glyco-lithocholic acid taurine
4 3
tauro-ursodeoxycholic acid
2
taurodeoxycholic acid tauro-lithocholic acid
1 0 Targretin
UAB-30
Glycochenodeoxycholate - L
2.5
Control
2.5
Targretin
UAB-30
2
1 0.5
0.5
0
0 Control
Targretin
UAB-30
α-muricholic acid β-muricholic acid
bacterial metabolism/deconjugation ursodeoxycholic acid deoxycholic acid lithocholic acid
ursodeoxycholic acid deoxycholic acid
enterohepatic recirculation
lithocholic acid cholic acid chenodeoxycholic acid α-muricholic acid β-muricholic acid
fecal excretion
compounds with gut microbiome metabolic origin or contribution
• Targretin changed the composition and amounts of bile acids in liver, possibly through the RXR-FXR heterodimer.
1.5
1
cholic acid chenodeoxycholic acid
intestinal microbiota metabolism
Taurochenodeoxycholate - L
2
1.5
glycine taurine
glyco-ursodeoxycholic acid
5
Control
cholesterol
host liver metabolism
systemic circulation
5
0.8
3
mitochondria
1.2
1
acetyl-CoA
acyl-CoA fatty acid βoxidation
cytosol
1.6 1.5
acyl-CoA acyl-carnitine
Acetylcarnitine - L
2
Products of β-oxida on
Targretin
Oleoylcarnitine - L
8 7 6 5 4 3 2 1 0
Control
Targretin
UAB-30
Figure 5. Effects of RXRs. Effects of RXRs on synthesis of fatty acids and b-oxidation (A) and cholesterol and bile acid synthesis (B).
Figure 4. Differential expression due to UAB30 treatment within GeneGo pathway maps. The GeneGo pathway maps for retinol metabolism (A) and oxidative phosphorylation (B) are shown. The thermometer-like shapes indicate relative fold changes (red: positive, blue: negative) for targretin relative to controls for proteomics (index 1) and transcriptome (index 2) where the differential expression criteria are met.
ª 2014 The Authors. Pharmacology Research & Perspectives published by John Wiley & Sons Ltd, British Pharmacological Society and American Society for Pharmacology and Experimental Therapeutics.
2014 | Vol. 2 | Iss. 6 | e00074 Page 13
P. T. Vedell et al.
Effects of RXR on RNA and Protein Levels
Consistent with gene expression and proteomic analysis, metabolic profiling showed altered hepatic bile acid metabolism in TRG-treated and to a lesser degree in UAB30-treated rats (Fig. 5B). Liver samples from TRGtreated animals had significantly decreased cholesterol and increased 7-a-hydroxycholesterol, reflecting altered bile acid synthesis that produces this compound. Interestingly, these animals had significantly decreased glycocholate, taurocholate, and taurodeoxycholate along with significantly increased levels of taurochenodeoxycholate, glycochenodeoxycholate, b-muricholate, tauro-b-muricholate, and tauroursodeoxycholate; in agreement with gene expression and proteomic data showing interference of TRG on bile acid and cholesterol metabolism. Many or most of these changes were probably secondary to altered expression of genes related to the LXR and FXR nuclear receptors previously demonstrated (Vedell et al. 2013). Altered gene expression in the LXR heterodimer was correlated with the triglycerides effects of multiple RXRs (Lalloyer et al. 2009; Vedell et al. 2013). Furthermore, we have shown that alteration in genes associated with the Ah receptor allowed us to predict altered carcinogen metabolism (Grubbs et al. 1995; Vedell et al. 2013). In fact, we observed alterations in the same pathway (Cyp1a1, Nqo1 [quinone reductase], and Ugt1a6) at the protein level (data not shown). Thus, knowledge of gene or protein expression data alone (particularly in liver) may allow one to predict many physiologic changes in the target tissue and serum as well. TRG and UAB-30 yielded strong concordance between measurements of transcript level alterations and protein level alterations at the pathway levels and in terms of relative magnitude of change. Furthermore, the transcript/ protein alterations correlated with/predicted many of the changes observed by metabolomics.
Acknowledgements The authors thank Petra Erdmann-Gilmore and Alan Davis for expert proteomics technical assistance and LC-MS analysis. The proteomics data processing and analysis was performed by James Malone at the WU Proteomics Core Laboratory. The PCL laboratory is supported, in part, with grants P41RR000954 and UL1 RR024992 from the National Center for Research Resources, a component of the National Institutes of Health (NIH) and NIH Roadmap for Medical Research. Additional funding for this work was provided by The National Cancer Institute (Breast SPORE CA089019) and contract HHSN261200433001C.
Disclosures None declared.
2014 | Vol. 2 | Iss. 6 | e00074 Page 14
References Andley UP, Malone JP, Hamilton PD, Ravi N, Townsend RR (2013). Comparative proteomic analysis identifies age-dependent increases in the abundance of specific proteins after deletion of the small heat shock proteins aA- and aB-crystallin. Biochemistry 52: 2933–2948. Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, et al. (2000). Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet 25: 25–29. Atigadda VR, Vines KK, Grubbs CJ, Hill DL, Beenken SL, Bland KI, et al. (2003). Conformationally defined retinoic acid analogues. 5. Large-scale synthesis and mammary cancer chemopreventive activity for (2E, 4E, 6Z, 8E)-8-30 , 40 -dihydro-10 (20 H)-naphthalen-10 -ylidene)-3,7-dimethyl-2, 4, 6-octatrienoic acid (9cUAB30). J Med Chem 46: 3766–3769. Bastien J, Rouchette-Egly C (2004). Nuclear retinoid receptors and the transcription of retinoid target genes. Gene 328: 1–16. Boerma LJ, Xia G, Qui C, Cox BD, Chalmers MJ, Smith CD, et al. (2014). Defining the communication between agonist and coactivator binding in the retinoid X receptor ligand binding domain. J Bio Chem 289: 814–826. Carlson M. (2012). GO.db: a set of annotation maps describing the entire Gene Ontology. R package version 2.8.0. R Foundation for Statistical Computing, Vienna, Austria. Chambon P (1996). A decade of molecular biology of retinoic acid receptors. FASEB J 10: 940–954. Cui X, Hwang JT, Qiu J, Blades NJ, Church GA (2005). Improved statistical tests for differential gene expression by shrinking variance components estimates. Biostatistics 6: 59–75. DeHaven CD, Evans AM, Dai H, Lawton KA (2010). Organization of GC/MS and LC/MS metabolomics data into chemical libraries. J Cheminform 2: 9. Evans AM, DeHaven CD, Barrett T, Mitchell M, Milgram E (2009a). Integrated, nontargeted ultrahigh performance liquid chromatography/electrospray ionization tandem mass spectrometry platform for the identification and relative quantification of the small-molecule complement of biological systems. Anal Chem 81: 6656–6667. Evans MJ, Mahaney PE, Borges-Marcucci L, Lai K, Wang S, Krueger JA, et al. (2009b). A synthetic farnesoid X receptor (FXR) agonist promotes cholesterol lowering in models of dyslipidemia. Am J Physiol Gastrointest Liver Physiol 296: G543–G552. Farol LT, Hymes KB (2004). Bexarotene: a clinical review. Expert Rev Anticancer Ther 4: 180–188. Gniadecki R, Assaf C, Bagot M, Dummer R, Duvic M, Knobler R, et al. (2007). The optimal use of bexarotene in cutaneous T-cell lymphoma. Br J Dermatol 157: 433–440.
ª 2014 The Authors. Pharmacology Research & Perspectives published by John Wiley & Sons Ltd, British Pharmacological Society and American Society for Pharmacology and Experimental Therapeutics.
P. T. Vedell et al.
Gorman GS, Coward L, Kerstner-Wood C, Cork L, Kapetanovic IM, Brouillette WJ, et al. (2007). In vitro metabolic characterization, phenotyping, and kinetic studies of 9cUAB30, a retinoid X receptor-specific retinoid. Drug Metab Dispos 35: 1157–1164. Grubbs CJ, Steele VE, Casebolt T, Juliana MM, Eto I, Whitaker LM, et al. (1995). Chemoprevention of chemically-induced mammary carcinogenesis by indole-3-carbinol. Anticancer Res 15: 709–716. Grubbs CJ, Lubet RA, Attigadda VR, Christov K, Deshpande AM, Tirmal V, et al. (2006). Efficacy of new retinoids in the prevention of mammary cancers and correlations with short-term biomarkers. Carcinogenesis 27: 1232–1239. Heyman RA, Mangelsdorf DJ, Dyck JA, Stein RB, Eichele G, Evans RM, et al. (1992). 9-cis retinoic acid is a high affinity ligand for the retinoid X receptor. Cell 68: 397–406. Karpievitch Y, Stanley J, Taverner T, Huang J, Adkins JN, Ansong C, et al. (2009). A statistical framework for protein quantitation in bottom-up MS-based proteomics. Bioinformatics 25: 2028–2034. Keller A, Nesvizhskii AI, Kolker E, Aebersold R (2002). Empirical statistical model to estimate the accuracy of peptide identifications made by MS/MS and database search. Anal Chem 74: 5383–5392. Kolesar JM, Hoel R, Pomplun M, Havighurst T, Stublaski J, Wollmer B, et al. (2010). A pilot, first-in-human, pharmacokinetic study of 9cUAB30 in healthy volunteers. Cancer Prev Res 3: 1565–1570.
Effects of RXR on RNA and Protein Levels
Neubert H, Bonnert TP, Rumpel K, Hunt BT, Henle ES, James IT (2008). Label-free detection of differential protein expression by LC-MALDI mass spectrometry. J Proteome Res 7: 2270–2279. Ohta T, Masutomi N, Tsutsui N, Sakairi T, Mitchell M, Milburn MV, et al. (2009). Untargeted metabolomic profiling as an evaluative tool of fenofibrate-induced toxicology in Fischer 344 male rats. Toxicol Pathol 37: 521–535. Perkins DN, Pappin DJ, Creasy DM, Cottrell JS (1999). Probability-based protein identification by searching sequence databases using mass spectrometry data. Electrophoresis 20: 3551–3567. Rigas JR, Dragnev KH (2005). Emerging role of rexinoids in non-small cell lung cancer: focus on bexarotene. Oncologist 10: 22–33. Saldanha AJ (2004). Java Treeview–extensible visualization of microarray data. Bioinformatics 20: 3246–3248. Storey JD, Tibshirani R (2003). Statistical significance for genomewide studies. Proc Natl Acad Sci USA 100: 9440–9445. Sussman F, de Lera AR (2005). Ligand recognition by RAR and RXR receptors: binding and selectivity. J Med Chem 48: 6212–6219. Szanto A, Narkar V, Shen Q, Uray IP, Davies PJ, Nagy L (2004). Retinoid X receptors: X-ploring their (patho) physiological functions. Cell Death Differ 11: S126–S143. ThomsonReuters. (2010). MetaCore: Pathway Analysis and Data Mining for Gene Expression. In Thomson Reuters. Available at http://thomsonreuters.com/content/science/pdf/ls/ metacore-cfs-en.pdf (accessed August 17, 2014).
Lalloyer F, Pedersen TA, Gross B, Lestavel S, Yous S, Vallez E, et al. (2009). Rexinoid bexarotene modulates triglyceride but not cholesterol metabolism via gene-specific permissivity of the RXR/LXR heterodimer in the liver. Arterioscler Thromb Vasc Biol 29: 1488–1495.
Vedell P, Lu Y, Grubbs C, Yin Y, Jiang H, Bland K, et al. (2013). Effects on gene expression in rat liver after administration of RXR agonists: UAB30, 4-methyl-UAB30, and targretin (Bexarotene). Mol Pharmacol 83:698–708.
Langfelder P, Horvath S (2008). WGCNA: a R package for weighted correlation network analysis. BMC Bioinformatics 9: 559.
Viswakarma N, Jia Y, Bai L, Vluggens A, Borensztajn J, XU J, et al. (2010). Coactivators in PPAR-Regulated Gene Expression. PPAR Res 2010: pii.250126.
Lansigan F, Foss FM (2010). Current and emerging treatment strategies for cutaneous T cell lymphoma. Drugs 70: 273–286. Levin AA, Sturzenbecker LJ, Kazmer S, Bosakowski T, Huselton C, Allenby G, et al. (1992). 9-cis retinoic acid stereoisomer binds and activates the nuclear receptor RXRa. Nature 355: 359–361. Lucas A, Gautier L (2005). ctc: cluster and tree conversion. R package version 1.38.0. R Foundation for Statistical Computing, Vienna, Austria. Muccio DD, Brouillette WJ, Breitman TR, Taimi M, Emanuel PD, Zhang X, et al. (1998). Conformationally defined retinoic acid analogues. 4. Potential new agents for acute promyelocytic and juvenile myelomonocytic leukemias. J Med Chem 41: 1679–1687.
Wang Y, Yao R, Maciag A, Grubbs CJ, Lubet RA, You M (2006). Organ-specific expression profiles of rat mammary gland, liver, and lung tissues treated with targretin, 9-cis retinoic acid, and 4-hydroxyphenylretinamide. Mol Cancer Ther 5: 1060–1072. Wolf G, Phil D (2006). Is 9-cis retinoic acid the endogenous ligand for the retinoic acid-X receptor? Nutr Rev 64: 532–538. Zybailov B, Coleman MK, Florens L, Washburn MP (2005). Correlation of relative abundance ratios derived from peptide ion chromatograms and spectrum counting for quantitative proteomic analysis using stable isotope labeling. Anal Chem 77: 6218–6224.
ª 2014 The Authors. Pharmacology Research & Perspectives published by John Wiley & Sons Ltd, British Pharmacological Society and American Society for Pharmacology and Experimental Therapeutics.
2014 | Vol. 2 | Iss. 6 | e00074 Page 15
P. T. Vedell et al.
Effects of RXR on RNA and Protein Levels
Supporting Information Additional Supporting Information may be found in the online version of this article: Figure S1. The Venn diagram shows the relationships for gene frequencies between the detected gene expression at the protein level, the detected gene expression at the transcript level. The detection of transcript expression is defined by the E/xps package CABG (detected above background correction) P-value being less than 0.05 for at least one exon of one individual. Table S1. Differential expression attributed to targretin treatment by GO categories. Table S2. Differential expression attributed to UAB30 treatment by GO category. Table S3. Mass spectrometry and database search results. Table S4. The Fs statistic fold change and significance Pvalues are provided for the targretin versus control contrast and also for the UAB-30 versus control contrast for protein expression. In this table, genes are indicated by the “Gene Symbol” column, the “Fold Change TRG vs. CTL” and “P-Value TRG vs. CTL” indicate the fold change and Fs test significance P-value for TRG vs. Control, the “Fold Change UAB30 vs. CTL” and “P-Value UAB30 vs. CTL” indicate the fold change and Fs test significance P-value for UAB30 vs. Control. The “TRG Diff Expr Indicator” and “UAB30 Diff Expr Indicator” columns indicate genes for which the differential expression criteria are satisfied for the two different contrasts. Only genes that satisfied at least one of these. Table S5. The Fs statistic fold change and significance Pvalues are provided for the targretin vs. control contrast and also for the UAB30 vs. control contrast for transcriptional expression. In this table, genes are indicated by the “Gene Symbol” column, the “TRG Fold Change @ min P-value” and “TRG min. P-value” indicate the exon with the smallest P-value for the gene and the fold change in that exon. The “TRG mean Fold Change”: and “TRG mean P-value” indicated average folder change and Pvalue across the gene. For UAB30, there are columns analogous to the four aforementioned TRG columns. The “TRG Diff Expr Indicator” and “UAB30 Diff Expr Indicator” columns indicate genes for which the differential
2014 | Vol. 2 | Iss. 6 | e00074 Page 16
expression criteria are satisfied for the two different contrasts. This table only included genes for which expression was detected above background in at least one exon and for which differential expression is detected for at least one treatment. Table S6. The specific proteins examined in this table are related to the KEGG terms in Figures 3 and 4. The Fs statistic fold change and significance P-values are provided for the targretin vs. control contrast and also for the UAB30 vs. control contrast for transcriptional expression. In this table, proteins are indicated by the combination of the Kegg pathway and the specific gene symbol. The “TRG mean Fold Change” and “TRG mean P-value” indicated average folder change and P-value across the protein. For UAB30, there are columns analogous to the four aforementioned TRG columns. The “TRG Diff Expr Indicator” and “UAB30 Diff Expr Indicator” columns indicate genes for which the differential expression criteria are satisfied for the two different contrasts. This table only included genes for which expression was detected above background in at least one exon and for which differential expression is detected for at least one treatment. Table S7. The specific transcripts examined in this Table are related to the KEGG terms in Figures 3 and 4. The Fs statistic fold change and significance P-values are provided for the targretin vs. control contrast and also for the UAB30 vs. control contrast for transcriptional expression. In this table, genes are indicated by the combination of KEGG pathway and the specific gene symbol the “TRG Fold Change @ min P-value” and “TRG min. P-value” indicate the exon with the smallest P-value for the gene and the fold change in that exon. The “TRG mean Fold Change” and “TRG mean P-value” indicated average folder change and P-value across the gene. For UAB30, there are columns analogous to the four aforementioned TRG columns. The “TRG Diff Expr Indicator” and “UAB30 Diff Expr Indicator” columns indicate genes for which the differential expression criteria are satisfied for the two different contrasts. This table only included genes for which expression was detected above background in at least one exon and for which differential expression is detected for at least one treatment.
ª 2014 The Authors. Pharmacology Research & Perspectives published by John Wiley & Sons Ltd, British Pharmacological Society and American Society for Pharmacology and Experimental Therapeutics.
