Heat-Map Visualization of Gas Chromatography-Mass Spectrometry Based Quantitative Signatures on Steroid Metabolism Ju-Yeon Moon,a,b Hyun-Jin Jung,a,b Myeong Hee Moon,b Bong Chul Chung,a and Man Ho Choia a Life Sciences Division, Korea Institute of Science and Technology, Seoul, Korea b Department of Chemistry, Yonsei University, Seoul, Korea Abnormalities in steroid hormones are responsible for the development and prevention of endocrine diseases. Due to their biochemical roles in endocrine system, the quantitative evaluation of steroid hormones is needed to elucidate altered expression of steroids. Gas chromatographic-mass spectrometric (GC-MS) profiling of 70 urinary steroids, containing 22 androgens, 18 estrogens, 15 corticoids, 13 progestins, and 2 sterols, were validated and its quantitative data were visualized using hierarchically clustered heat maps to allow “steroid signatures”. The devised method provided a good linearity (r2 Ͼ 0.994) with the exception of cholesterol (r2 ϭ 0.983). Precisions (% CV) and accuracies (% bias) ranged from 0.9% to 11.2% and from 92% to 119%, respectively, for most steroids tested. To evaluate metabolic changes, this method was applied to urine samples obtained from 59 patients with benign prostatic hyperplasia (BPH) versus 41 healthy male subjects. Altered concentrations of urinary steroids found and heat maps produced during this 70-compound study showed also differences between the ratios of steroid precursors and their metabolites (representing enzyme activity). Heat maps showed that oxidoreductases clustered (5␣-reductase, 3␣-HSD, 3␤-HSD, and 17␤-HSD, except for 20␣-HSD). These results support that data transformation is valid, since 5␣-reductase is a marker of BPH and 17␤-HSD is positively expressed in prostate cells. Multitargeted profiling analysis of steroids generated quantitative results that help to explain correlations between enzyme activities. The data transformation and visualization described may to be found in the integration with the mining biomarkers of hormone-dependent diseases. (J Am Soc Mass Spectrom 2009, 20, 1626–1637) © 2009 American Society for Mass Spectrometry ass spectrometry based metabolite profiling mone, two major types of enzymes are involved, cyto- reveals the metabolic states of biological sys- chrome P450 and steroid oxidoreductase. Abnormalities of Mtems and provides comprehensive insights by these enzymes often lead to hormonal imbalances that allowing comparisons between many metabolites si- have serious consequences, and which are responsible multaneously present in cells, tissues, or organisms [1, for the development of hormone-dependent diseases 2]. This technique promotes the establishment of rela- (see Supplementary Table 1, which can be found in the tionships between phenotypes and metabolisms by electronic version of this article). For example, concen- providing descriptions of the distributions of metabo- trations of corticoids and their metabolic ratios provide lites and their biological functions. In fact, quantitative diagnostic evidence of apparent mineralocorticoid ex- analyses of sets of metabolites in biochemical pathways cesses caused by 11␤-HSD deficiency [8] and congenital have been used for physiological monitoring, toxicolog- adrenal hyperplasia, which are caused by deficiencies ical evaluations, and clinical diagnosis [3–6]. of enzymes like hydroxylase (at C-11, 17, and 21) or Many naturally occurring steroids with similar 3␤-HSD [9]. In addition, enhanced androgen activity chemical structures could yield biological information generated by the conversion of testosterone to dihy- [7]. Endogenous steroids are divided into five groups, drotestosterone (DHT) by 5␣-reductase was utilized to namely, androgens, estrogens, corticoids, progestins, allow early therapeutic intervention in young men [10]. and sterols, which are generally synthesized from cho- Enzyme activity profiles can be used to describe the lesterol in the adrenal cortex, ovaries, and testes functional diversities of biological systems, which are (Scheme 1). In biosynthetic pathways of steroid hor- driven by genetic diversity. Although enzyme activities have been monitored by following reactions between targeted enzymes and substrate molecules, over- Address reprint requests to Dr. M. H. Choi, Life Sciences Division, Korea Institute of Science and Technology, 39-1 Hawolkok-dong, Seongbuk-ku, estimations by radioimmunoassay (RIA) and enzyme Seoul 136-791, Korea. E-mail: [email protected] immunoassay (EIA), because of cross reacting antibod- Published online May 5, 2009 © 2009 American Society for Mass Spectrometry. Published by Elsevier Inc. Received February 7, 2009 1044-0305/09/$32.00 Revised April 25, 2009 doi:10.1016/j.jasms.2009.04.020 Accepted April 28, 2009 J Am Soc Mass Spectrom 2009, 20, 1626–1637 QUANTITATIVE STEROID SIGNATURES BY GC-MS 1627 OH Progestins, Sterols OH Androgens HO H HO βββαβααααα-diol 24S-OH-Chol 17α-HSD O O OH O 3α-HSD 17β-HSD HO O OH HO β 5 -reductase O HO HO Chol H H H HO βββαβαααββββ 16ααα-OH-DHEA 5βββ−β−−−dione Etio -diol O O 20,22-desmolase H O 16α-hydroxylase O O 20ααα-OH-prog ααα O 5 -DHP O OH β OH α 3 -HSD 20α-HSD 17α-hydroxylase 17,20-lyase 17β-HSD 5 -reductase HO O 5α-reductase O HO H HO HO HO HO H 20α-HSD 3α-HSD H 5ααα−α−−−dione Epi-An Preg 17ααα-OH-Preg DHEA A-diol 17β-HSD β α 3β-HSD 3β-HSD 3 -HSD 3β-HSD 3 -HSD HO HO O O O O OH OH OH O H H α P-diol P-one HO 11β-hydroxylase 17α-hydroxylase OH 17,20-lyase 17β-HSD 5α-reductase 3 -HSD 17β-HSD O O O HO HO H H H O O O ααααααββββ 11βββ-OH-Prog Prog 17ααα-OH-Prog A-dione T DHT -diol An 17α-HSD 3β-HSD OH OH 21-hydroxylase 21-hydroxylase aromatase O O O aromatase O HO OH OH OH H OH Epi-T αααβαβββββ-diol 5β-reductase β HO 3 -HSD O O O OH H 17β-HSD THDOC 11-DeoxyB 11-DeoxyF 11β-hydroxylase 11β-hydroxylase 11β-hydroxylase HO HO O O O O E1 17βββ-E2 OH HO HO OH HO HO OH OH OH 2-hydroxylase 4-hydroxylase 16-hydroxylase O O O OH OH 5α-reductase β β O O O 21 -hydroxylaseO OH 17 -HSD OH O H allo-DHB B Cortisol (F) 21-DeoxyF HO β α HO HO HO HO HO 11β-HSD 5 - / 5 -reductase 11β-HSD 3β-HSD 2-OH-E1 OH 4-OH-E1 16ααα-OH-E1 Estriol 16-keto-E2 O O O OH OH OH OH methyltransferase methyltransferase OH O HO OH O OH OH OH OH O OH O HO H CO HO HO O HO O 3 H 17-Epi-E3 16-Epi-E3 11-DehydroB THF / allo-THF, allo-DHF Cortisone (E) HO HO HO HO OH OCH 5β- / 5α-reductase 2-OH-E2 2-MeO-E1 4-OH-E2 3 4-MeO-E1 3β-HSD O OH methyltransferase OH methyltransferase OH O OH H3CO HO HO Corticoids H HO THE 2-MeO-E2 OCH3 4-MeO-E2 Estrogens Scheme 1. General scheme for steroid metabolism in man. See Supplementary Table 1 for the full names of steroid hormones. ies, limit the applicability of these assays, and further- groups have used GC-MS based steroid analysis for more, only single enzymes can be estimated at a time mining biomarkers, technical improvements that more [11–14]. In contrast to conventional enzyme assays, gas effectively allow the visualization of steroid metabolites chromatography-mass spectrometry (GC-MS)-based are required [26]. However, no studies have reported techniques have better quantitative reproducibility [15, MS-based steroid signatures generated by HCA to date. 16]. For these reasons, GC-MS profiling has been widely Here, we introduce the GC-MS quantitative profiling used for steroid analysis [17–20], and offers the basis for of 70 urinary steroids, including 22 androgens, 18 techniques that can be applied to large-scale clinical estrogens, 15 corticoids, 13 progestins, and 2 sterols, studies [21]. generated by HCA to evaluate metabolic changes and Clinical significances are generally expressed in ta- enzyme reactions in steroid analysis. The aim of this bles or bar graphs that show changes in analytes across study was to validate a GC-MS profiling method and a groups of interest. For studies involving few com- method that allows the quantitative visualization of pounds, these visualizations are enough to differentiate multiple urinary steroids. To visualize quantitative re- classes showing metabolic differences. However, quan- sults, a microarray map (a type of heat map) was titative datasets of multiple compounds are much more designed to present the urine sample results of 59 difficult to represent visually. Statistical clustering of- fers one such approach, and has been utilized to sup- patients with benign prostatic hyperplasia (BPH) and 41 port genomic and proteomic studies [22–24]. In a simi- healthy male subjects; BPH was chosen because steroid lar fashion, quantitative results obtained by metabolite metabolism is known to play a role in the progress of profiling can be directly compared between samples prostate diseases [27–29]. This study focused on illus- and utilized as metabolic biomarkers. In recent, MS- trating the usefulness of steroid signatures for explain- based quantitative data generated by hierarchical clus- ing both the concentrations of individual steroids and tering analysis (HCA) has been subjected to pattern the activities of enzymes correlated global steroidogen- analysis for metabolite profiling [2, 25]. The concept of esis in BPH, and suggest that enzyme activity profiling “metabolite signature” is a result of this process, and may be a useful diagnostic tool and provide a means of such signatures are useful for measuring and visualiz- identifying mining biomarkers in hormone-dependent ing the relative analyte concentrations. Although many diseases. 1628 MOON ET AL. J Am Soc Mass Spectrom 2009, 20, 1626–1637 Experimental tant biopsy and had either a suspicious finding by digitorectal examination or an elevated level of serum Chemicals and Materials prostate-specific antigen (PSA).