help button home button Endocrine Society JCEM JCEM Call for Nominations for EIC
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Journal of Clinical Endocrinology & Metabolism, doi:10.1210/jc.2005-1137
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
91/2/671    most recent
Author Manuscript (PDF)
Right arrow Submit a related Letter to the Editor
Right arrow Purchase Article
Right arrow View Shopping Cart
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow Request Copyright Permission
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Chung, L.
Right arrow Articles by Baxter, R. C.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Chung, L.
Right arrow Articles by Baxter, R. C.
Right arrowPubmed/NCBI databases
*Substance via MeSH
Related Collections
Right arrow Metabolism
Right arrow Neuroendocrinology and Pituitary
The Journal of Clinical Endocrinology & Metabolism Vol. 91, No. 2 671-677
Copyright © 2006 by The Endocrine Society

Novel Biomarkers of Human Growth Hormone Action from Serum Proteomic Profiling Using Protein Chip Mass Spectrometry

Liping Chung, David Clifford, Michael Buckley and Robert C. Baxter

Kolling Institute of Medical Research, University of Sydney, Royal North Shore Hospital (L.C., R.C.B.), New South Wales 2065, Australia; and CSIRO Mathematical and Information Sciences (D.C., M.B.), North Ryde, New South Wales 1670, Australia

Address all correspondence and requests for reprints to: Dr. Robert C. Baxter, Kolling Institute of Medical Research, Royal North Shore Hospital, St. Leonards, New South Wales 2065, Australia. E-mail: robaxter{at}med.usyd.edu.au.


   Abstract
Top
 Abstract
Introduction
 Subjects and Methods
 Results
 Discussion
 References
 
Context: The detection of exogenous human GH (hGH) administration in athletes poses unique analytical problems, because its short circulating half-life provides only a brief opportunity to detect the administered hormone above endogenous levels. Measurement of novel GH-regulated serum protein biomarkers might provide an indirect method to detect exogenous GH.

Objective: The objective of this study was to identify new serum biomarkers of GH administration using proteomic profiling.

Design: Sera from a previously reported, double-blind, placebo-controlled GH administration trial were analyzed by protein chip mass spectrometry.

Setting: The study was performed at clinical research centers.

Subjects: Sixty healthy subjects, aged 18–40 yr, who were not elite athletes, were studied.

Interventions: Placebo or recombinant hGH treatment (0.1 or 0.2 IU/kg·d; 20 subjects/group) was administered for 4 wk, followed by an 8-wk washout period.

Main Outcome Measures: Protein mass profiles were determined on immobilized Cu2+ chips on d 0 and 21 of GH administration, and multivariate analysis was used to classify subjects into GH and placebo administration groups.

Results: When assessed by cross-validation, the classification performance of classifiers based on multivariate analysis of several GH-regulated peaks performed no better than classifiers based on the single best peak. This peak, a prominent biomarker of 15.1 kDa, was purified and identified as hemoglobin {alpha}-chain. The time course of the GH response of this biomarker is similar to that of other GH-dependent markers, such as IGF-I.

Conclusion: This study demonstrates that protein mass profiling is an effective tool for the detection of GH administration and suggests that measurement of hemoglobin {alpha}-chain may have utility as a novel serum biomarker of GH action.


   Introduction
Top
 Abstract
 Introduction
Subjects and Methods
 Results
 Discussion
 References
 
DETECTING THE ILLICIT administration of human GH (hGH), which is believed to occur commonly in sports since the advent of recombinant hGH (1), is analytically challenging. Recombinant hGH is indistinguishable from the 22-kDa isoform of pituitary-derived hGH by current analytical methods. Exogenous hGH disappears rapidly from the circulation (2, 3, 4). The secretion of pituitary hGH is pulsatile and is regulated by factors such as sleep, nutrition, exercise, and emotional stress, resulting in serum concentrations that fluctuate widely throughout the day and frequently overlap with measurements obtained after exogenous GH administration (5, 6).

Currently, there are two promising approaches for detecting exogenous GH administration. The first uses immunoassays to measure the suppression of endogenous, pituitary-derived GH isoforms (e.g. 17 and 20 kDa) in response to exogenous hGH. The second involves the measurement of GH-responsive proteins in serum, including IGF-I, IGF-binding protein-3 (IGFBP-3), acid-labile subunit (ALS), C-terminal propeptide of type I procollagen, C-terminal cross-linked telopeptide of type I collagen, and procollagen type III N-terminal peptide (7, 8, 9, 10, 11).

The GH2000 project was an international collaborative, double-blind, placebo-controlled GH administration trial that aimed to develop a method for detecting GH doping by athletes. Dall et al. (8) reported the effects of 4 wk of supraphysiological GH administration to 99 healthy subjects on the IGF-I axis and bone markers. Most IGF-I circulates as a GH-dependent, 150-kDa ternary complex, which also contains the specific binding proteins IGFBP-3 or IGFBP-5 and ALS, all of which are GH regulated (12, 13). The increase in IGF-I was markedly higher than that in IGFBP-3 or ALS. Several bone and collagen markers also increased in response to GH (10, 14).

Wallace et al. (15) also studied the effects of exercise and supraphysiological GH administration on the GH/IGF-I axis and bone markers in 17 athletic adult males. Acute exercise increased all GH isoforms, with a peak at the end of acute exercise (9). Several bone and collagen markers also increased with exercise, whereas osteocalcin was unchanged. GH treatment resulted in an augmented response to exercise in markers of collagen turnover (10, 14). Age is an important predictor of serum levels of markers of GH, reflecting the fall in GH secretion with age (16).

Surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF MS) is a proteomic technique in which proteins are bound to proprietary protein chips with different types of adsorptive surfaces, e.g. hydrophobic, cation exchange, or anion exchange (17, 18, 19). SELDI-TOF MS can be used to analyze peptide and protein expression patterns in a variety of clinical and biological samples, and biomarker discovery can be achieved by comparing the protein profiles obtained from control and patient groups to elucidate differences in protein expression (20, 21, 22). Because SELDI analysis provides biochemical information about biomarkers (i.e. mass and adsorption conditions), more specialized protein biochemistry and mass spectrometry can then be used to identify unknown marker proteins for a particular condition. Thus, we aimed to develop a novel method for the detection of biomarkers of hGH administration using serum proteomic profiling. For this purpose, we obtained serum protein mass profiles by SELDI-TOF MS on samples from athletes participating in the double blind, placebo-controlled GH2000 study.


   Subjects and Methods
Top
 Abstract
 Introduction
 Subjects and Methods
Results
 Discussion
 References
 
Subjects

Serum samples were obtained from the previously reported GH2000 program (8), a double-blind, placebo-controlled trial of 49 women and 50 men, physically fit and aged 26 ± 1 yr, in which blood was collected weekly during a 4-wk treatment period (d 1–28), and the subjects were subsequently followed for an additional 8 wk (d 29–84). The treatments included placebo (n = 40), 0.1 IU/kg·d GH (low GH; n = 30), and 0.2 IU/kg·d GH (high GH; n = 29). Most samples from these three treatment groups at the various time points were available for the present study. All samples were stored at –80 C until use.

SELDI-TOF MS analysis

Preliminary studies were performed to optimize conditions for discriminating between sera from GH-treated and those from placebo-treated subjects. This involved a comparison of various protein chip surfaces [strong anion exchange, weak cation exchange, hydrophobic, and immobilized metal affinity (IMAC) loaded with Cu2+, Ni2+, or Zn2+ ions] and of various adsorption and washing conditions. These preliminary assays (data not shown) resulted in the selection of IMAC chips loaded with copper (Cu2+-IMAC) as the most effective chip array for this study.

Biomarker discovery was based on 120 serum samples, 20 subjects from each of the three treatment groups described above (placebo, low GH, and high GH), each sampled before treatment (d 0) and on d 21 of treatment. Samples were analyzed on Cu2+-IMAC chips. The array spots were preactivated with 100 mM CuSO4 for 10 min, then washed with deionized H2O. Serum samples (20 µl) were mixed with 30 µl denaturing buffer [8 M urea, 1% 3-[(3-cholamidopropyl) dimethylammonio]-1-propanesulfate, 10 mM sodium phosphate (pH 7.2), and 150 mM NaCl] and diluted an additional 1:20 in binding buffer [20 mM sodium phosphate (pH 7.2) and 0.5 M NaCl]. Diluted samples (5 µl, equivalent to 0.1 µl undiluted serum) were pipetted onto the spots. After incubation for 30 min, unbound proteins were removed by washing with binding buffer (three washes, 5 µl each). After two rinses with deionized H2O, 2x 1 µl matrix solution (20 mg/ml sinapinic acid, 50% acetonitrile, and 0.5% trifluoroacetic acid) was applied, and the spots were air dried for 10 min. The arrays were read in a ProteinChip reader (ProteinChip Biology System II, Ciphergen Biosystems, Inc., Fremont, CA).

TOF spectra of proteins were generated using an average of 80 laser shots. For data acquisition of low molecular weight proteins, the detection size range was 5,000–25,000 m/z (mass/charge ratio). Detector sensitivity was set at 8, and laser intensity was set at 210 arbitrary units. For the high molecular weight proteins, the detection range was 25,000–70,000 m/z, detector sensitivity was 8, and laser intensity was 220. The SELDI-TOF mass spectrometer was externally calibrated using the [M+H]+ ion peaks of bovine insulin (5,734.51 m/z), equine cytochrome c (12,361.96 m/z), equine apomyoglobin (16,952.27 m/z), bovine carbonic anhydrase (29,023.70 m/z), rabbit muscle aldolase (39,212.28 m/z), glucose-6-phosphate dehydrogenase (57,432.72 m/z), and bovine albumin (66,430.09 m/z; all standards from Sigma-Aldrich Corp., St. Louis, MO).

Reproducibility was determined using pooled human sera from 60 individuals (20 individuals/treatment group) as a quality control sample. Identical aliquots of this pool were applied to eight spots on each of four chips. Within- and between-chip precisions were assessed by ANOVA for five protein peaks of high intensity (mean intensity, 13.3) and five low-intensity peaks (mean intensity, 1.54). The mean within- and between-chip coefficients of variation were 17.3% and 18.1%, respectively, for low-intensity peaks, and 10.2% and 11.7% respectively, for high-intensity peaks.

Time course of GH action

Analysis of serum samples at various sampling times after d 21 was also performed for 10 subjects in each of the three treatment groups. Samples taken on d 0, 21, and 28 after the administration of placebo, low GH, or high GH and on d 30, 33, 42, and 84 of the washout period were analyzed by SELDI-TOF on Cu2+-IMAC chips as described above.

Univariate analysis

All spectra were compiled, and qualified mass peaks (signal to noise ratio, >5) with m/z ratios between 3,000 and 70,000 were autodetected. Peak clusters were completed using second-pass peak selection (signal to noise ratio, >2; within 0.3% mass window), and estimated peaks were added. The data were analyzed with ProteinChip Biomarker Wizard software version 3.0.2 (Ciphergen). The peak intensities were normalized to the total ion current. To characterize protein peaks of potential interest, serum proteomic profiling of athletes from the three treatment groups were compared on d 0 and 21 of treatment.

Univariate statistical analysis was performed using StatView (SAS Institute, Inc., Cary, NC). Time courses were analyzed by repeated measures ANOVA, followed by Fisher’s protected least significant difference test. P < 0.05 was regarded as significant. All values are presented as the mean ± SE.

Multivariate analysis

This was performed on the raw data output from the ProteinChip reader. Registration of data was performed by selecting two spectral regions (m/z ranges, 7,500–7,800 and 8,800–9,100) containing peaks that were present in all spectra and isolated from other peaks. The peak selection algorithm described by Tibshirani et al. (23) was used to select peaks in these two regions for each of the 120 spectra. Each spectrum was transformed linearly to align these peaks to their average locations across spectra, and the intensity values were interpolated linearly onto a fixed set of 24,347 m/z values. In this way, the SELDI intensity data were reduced to a single 120 x 24,347 data matrix.

Baseline subtraction was performed on log-transformed intensities by the morphological opening technique (24). This is a nonlinear correction method used in image analysis that produces a background value at each location that is less than or equal to the data value, so corrected values are nonnegative. A moving window of 51 consecutive m/z values was used in this process.

As in the studies by Tibshirani et al. (23) and Wu et al. (25), a linear transformation was applied to the intensity values in each spectrum, so that the 10th and 90th percentiles were mapped to 0 and 1, respectively. Finally, a lower cutoff was applied, removing all data at m/z values below 3,000. This left a set of 19,830 m/z values and a matrix of 120 x 19,830 transformed intensities. Within this set of 120 transformed spectra, a total of 530 peak clusters were identified by the method described by Tibshirani et al. (23), each referred to by its centroid. Associated with each cluster is a range of m/z values and corresponding peak heights.

Binary classification of samples into GH and placebo groups was performed on these peak data using two techniques. The first was GeneRaVE (26), an algorithm that employs Bayesian stochastic variable selection to identify variables (in this case, peaks) that discriminate users of GH from nonusers. The second was a forward stepwise diagonal discriminant analysis method in which the number of peaks chosen can be controlled. Evaluation was based on a 10-fold cross-validation, i.e. repeatedly training on 90% of the data and using the remaining 10% as test data.

Purification and identification of protein biomarkers

Pooled sera were diluted 1:10 in 50 mM NH4 acetate (pH 6.0), centrifuged at 12,000 rpm for 10 min, loaded onto a 1-ml HiTrap CM Sepharose FF cartridge (Amersham Biosciences, Uppsala, Sweden), and eluted at room temperature with a linear gradient of 0–1 M NaCl in NH4 acetate buffer at 1 ml/min. Each 1-ml fraction was analyzed by SELDI-TOF MS on IMAC-Cu2+ chips. Proteins eluting from 0.15–0.20 M NaCl were pooled and loaded on a reverse phase HPLC C18 column (250 x 4.6 mm; 300 Å pore size; Phenomenex, Torrance, CA) and eluted with a 30-min linear gradient from 15–60% acetonitrile in 0.1% trifluoroacetic acid (TFA) at 1.5 ml/min. All peaks were collected, lyophilized, and analyzed on hydrophilic NP20 chips (Ciphergen). N-Terminal Edman sequencing was carried out using an Applied Biosystems 494 Procise Protein Sequencing System (Foster City, CA) at Australian Proteome Analysis Facility (North Ryde, Australia).

Validation of the protein chip determination of hemoglobin {alpha}-chain

To isolate hemoglobin {alpha}-chain (HbA1), human hemoglobin (Sigma-Aldrich Corp.), 1.7 ml of a 1 mg/ml solution in 0.1% TFA, was injected onto a 300-Å C18 column (Phenomenex) and eluted with a 15–60% acetonitrile gradient in 0.1% TFA, resulting in the separation of Hb subunits. Fractions containing HbA1 were lyophilized and reconstituted in PBS, then analyzed by SELDI-TOF MS on IMAC-Cu2+ protein chips (Ciphergen) at 5 µl/spot.

To confirm the identity of HbA1 in serum samples immunologically, epoxy-activated protein chips (RS100, Ciphergen) were precoupled with antiserum H80, a rabbit polyclonal IgG against HbA1 (catalogue no. sc-21005, Santa Cruz Biotechnology, Inc., Santa Cruz, CA) or nonimmune rabbit IgG. H80 or nonimmune IgG (2 µl, 0.2 mg/ml) was applied to protein chips and mixed with 2 µl 50 mM NaHCO3 buffer (pH 9.2). The arrays were shaken for 2 h at room temperature in a humid chamber. To block the remaining active sites, spots were washed with 50 µM BSA in PBS (pH 7.2) for 1 h. After two PBS washes, serum samples, diluted 2-fold in PBS containing 0.1% Triton X-100, were applied to the spots using an eight-well Bioprocessor (Ciphergen) (50 µl/spot). The chips were incubated for 2 h on a shaker to achieve optimal binding, and wells were washed with PBS; then with 50 mM Tris-HCl, 1 M urea, 0.1% 3-[(3-cholamidopropyl) dimethylammonio]-1-propanesulfate, and 0.5 M NaCl (pH 7.2), and then with 5 mM HEPES (pH 7.2) before air drying. Sinapinic acid (20 mg/ml in 50% acetonitrile and 0.5% TFA; 2x 1 µl) was added to each spot and air dried. Protein mass spectra were generated as described above.


   Results
Top
 Abstract
 Introduction
 Subjects and Methods
 Results
Discussion
 References
 
Serum proteomic profiling by SELDI-TOF MS

Figure 1Go shows SELDI-TOF mass spectra in the m/z range 3,000–25,000 obtained using the Cu2+-IMAC chip after baseline correction and normalization using Ciphergen software. Figure 1AGo shows a spectrum from a subject on d 0, with the corresponding spectrum on d 21 of GH treatment in Fig. 1BGo. Several peaks show increases in intensity in response to GH. Figure 1Go, C–H, shows expanded spectra in the 14,500–16,500 m/z range for three subjects, sampled on d 0 (C, E, and G) and d 21 (D, F, and H) of GH treatment. A prominent peak at approximately m/z 15,120 that increased on GH treatment is indicated, with a possibly related smaller peak at approximately m/z 15,330. Similar spectra were obtained for 60 subjects in three treatment groups (placebo, low-dose GH, and high-dose GH) on d 0 and 21 of treatment. By univariate analysis, 241 peaks common to all 120 profiles were identified. Among these 241 peaks, 17 demonstrated statistically significant differences between intensity values on d 21 compared with d 0 of treatment and dose-dependent responsiveness to hGH with no response to placebo treatment. Figure 2Go illustrates two such peaks: a peak at 15,120 m/z, up-regulated by GH, and a peak at 6,850 m/z, down-regulated by GH treatment, both significant by ANOVA. Among other GH-dependent peaks, the peak at 15,120 also gave a prominent, doubly-charged species at approximately 7,560 and appeared to have a related smaller peak at approximately 15,330 m/z (shown in Fig. 1Go); another GH-regulated peak of m/z 7,651 probably represents IGF-I (not shown). These data indicate that the Cu2+-IMAC chips bind a number of GH-regulated proteins that may serve as biomarkers of GH action.


Figure 1
View larger version (25K):
[in this window]
[in a new window]
  		FIG. 1. Serum proteomic profiling on Cu2+-IMAC chips. A and B, Normalized SELDI-TOF mass spectra in the m/z range 3,000–25,000 showing the response of serum proteins in a single subject sampled on d 0 (A) and d 21 (B) of GH treatment. C–H, Normalized spectra expanded in the m/z range 14,500–16,500 for three subjects sampled on d 0 (C, E, and G) and d 21 (D, F, and H), respectively, of GH treatment. The arrows indicate a putative biomarker at m/z approximately 15,120 and a possibly related smaller peak at approximately 15,330.

 

Figure 2
View larger version (16K):
[in this window]
[in a new window]
  		FIG. 2. Examples of SELDI-TOF MS peaks showing dose-dependent GH regulation by univariate analysis. Mean data are shown for 20 subjects in each group, sampled before ({square}) or after ({blacksquare}) 21-d treatment with placebo, low-dose GH, or high-dose GH. A, Up-regulated marker at m/z 15,120 (P = 0.001 for GH effect). B, Down-regulated marker at m/z 6,850 (P = 0.003 for GH effect). Values are shown as the mean ± SE.

 
Relationship between GH treatment and time

The time course of change over 28 d of GH treatment and the following 8-wk washout period is shown in Fig. 3Go for the 15,120 m/z peak. This putative biomarker reached peak levels after 21 d of GH treatment at both low and high doses, fell slightly on d 28 of treatment, and then declined rapidly after the cessation of treatment, returning to baseline within about 1 wk. The response to GH was dose dependent by repeated measures ANOVA, with the high-dose GH group significantly different from the low-dose GH group (P = 0.035) and also significantly different from the placebo group (P = 0.0003).


Figure 3
View larger version (19K):
[in this window]
[in a new window]
  		FIG. 3. Time course of changes in the SELDI-TOF MS peak at m/z 15,120. Time course of change in peak intensity of this biomarker over 28 d of GH treatment and the following 8-wk washout period. Data are the mean ± SE for eight, eight, and nine subjects in placebo, low-dose GH, and high-dose GH groups, respectively.

 
Multivariate analysis and sample classification

Of a total of 530 peak clusters identified in this analysis, the GeneRaVE algorithm selected eight peak clusters with centroids at 3,264, 3,652, 4,141, 6,376, 7,759, 9,056, 13,127, and 15,120 m/z. Figure 4AGo compares receiver operator characteristic curves for the classification based on a linear combination of these peaks with that obtained using the best single classifier, i.e. the 15,120 m/z peak. In this binary classification (GH vs. no GH), GeneRaVE clearly classifies subjects more effectively than the single marker, with approximately 95% of GH users identified with no false positives. Under cross-validation, the performance of GeneRaVE fell, so that analysis based on the single 15,120 m/z peak actually performed slightly better than the multivariate analysis (Fig. 4BGo).


Figure 4
View larger version (16K):
[in this window]
[in a new window]
  		FIG. 4. Receiver operator characteristic (ROC) curves plotting the true positive rate (sensitivity) against the false positive rate (1 – specificity) in the binary classification of subjects into GH or no GH groups using the GeneRaVE algorithm based on eight peaks and the single best-performing peak at 15,120 kDa. A, Comparison without cross-validation: 1, GeneRaVE; 2, best peak. B, Comparison after cross-validation: 3, GeneRaVE; 4, best peak.

 
Application of a forward stepwise diagonal discriminant analysis showed that the inclusion of more than two peaks led to overfitting of the classifier. Overfitting occurs when peak clusters are included that improve the performance of the training data at a cost to the performance on test data. The analysis described assumes no difference between d 0 samples (before treatment commenced) and d 21 placebo samples, i.e. no placebo effect. In case a placebo effect does exist, a similar analysis of only the d 21 data was carried out, and similar results were obtained (data not shown).

Characterization, identification, and validation of the 15.12-kDa putative marker

The protein corresponding to the peak of 15.12 kDa was purified from pooled serum samples by cation exchange chromatography, followed by reverse phase HPLC (Fig. 5AGo), monitoring fractions by SELDI-TOF MS on NP20 chips. A major species of apparent mass 15.12 kDa was detected in fraction 31 (Fig. 5BGo). N-Terminal Edman sequencing yielded VLSPADKTNVKA, which corresponds to HbA1, a protein of 15,126 molecular weight.


Figure 5
View larger version (19K):
[in this window]
[in a new window]
  		FIG. 5. Purification and validation of the 15.12-kDa putative GH biomarker. A, C18 reverse phase liquid chromatography profile at 280 nm, indicating the elution time of the candidate biomarker (arrow). The gradient was 15–60% acetonitrile in 0.1% trifluoroacetic acid. B, Fractions were monitored by SELDI-TOF MS using NP20 protein chips. The arrow indicates the peak representing the candidate biomarker in the protein mass profile of the chromatography fraction indicated in A. C, Dose-response curve showing that the intensity of the 15,126 m/z peak by SELDI-TOF MS is directly related to the concentration of HbA1. D, Immunological confirmation that the 15,126 m/z peak represents HbA1. Proteins were immunoadsorbed out of serum samples onto activated RS100 protein chips to which rabbit antiserum H80 against HbA1 had been covalently coupled. Nonimmune rabbit IgG was used as a control. Mass spectra were then obtained by SELDI-TOF MS. Data from serum samples from a single subject on d 0 and 21 after hGH administration, as indicated, show that the protein of approximately 15.1 kDa was selectively bound to the HbA1 antibody and not to the control IgG. Note that a smaller peak at 15.3 kDa, also shown in Fig. 1Go, was extracted by the HbA1 antiserum and is therefore immunologically related to HbA1.

 
Figure 5CGo illustrates the linear relationship between the HbA1 concentration and peak intensity by SELDI-TOF MS, with 1 intensity unit approximately equivalent to 5 ng HbA1. Thus, in Fig. 2Go, where high-dose hGH treatment increased the mean peak intensity of the 15.12-kDa peak from approximately 1 to 4 intensity units (for samples equivalent to 0.1 µl serum), this is equivalent to an increase from 50 to 200 µg HbA1/ml.

Immunological confirmation that the 15.12-kDa peak represented HbA1 was obtained by reacting serum samples with activated protein chips to which a specific HbA1 antiserum had been coupled. Sera from two subjects, on d 0 or 21 of hGH administration, were used in this study. Figure 5DGo illustrates the results from one subject. A protein of 15.12 kDa was adsorbed from these sera onto antibody-coupled chips, but not onto control nonimmune IgG-coupled chips, and was detected by SELDI-TOF MS. A minor peak of approximately 15.33 kDa was also bound by the HbA1 antiserum, indicating that this species is immunologically related to HbA1 (possibly a glycation adduct). Identical results were obtained with sera from the second subject.


   Discussion
Top
 Abstract
 Introduction
 Subjects and Methods
 Results
 Discussion
References
 
In this study we applied the proteomic technique of SELDI-TOF MS to discover novel biomarkers of human GH in serum samples obtained from athletes under controlled conditions. Proteomic profiling using the SELDI-TOF system is a high-throughput, technically simple, monitoring platform for rapid characterization of serum protein mass profiles. This technique has been used to detect various cancers (20, 21, 27) and HIV infection (22), but has not previously been reported in either endocrine or sports drug testing. Using Cu2+-IMAC chips, we initially identified 241 peaks common to all 120 profiles, using the manufacturer’s normalization, baseline subtraction, and peak identification software. Univariate analysis identified 17 of these 241 peaks (i.e. ~7% of the peaks) significantly different after GH treatment compared with placebo treatment. Although this suggests the possible utility of these peaks as single biomarkers of GH action, there is a multiple comparison problem, because almost as many peaks would be found to be statistically significant at the 5% level purely by chance.

For multivariate analysis, we started with raw mass spectral data, applying published or in-house algorithms for registration, baseline correction, and peak identification. A total of 530 peaks (or peak clusters) were identified using this approach. The GeneRaVE algorithm has previously been used successfully for gene expression profiling analysis (28, 29), but not for the analysis of protein mass profiling data. GeneRaVE selected eight of the 530 peaks for its optimal multivariate classification of GH-treated vs. placebo-treated subjects. In cross-validation studies, a single protein peak of 15.12 kDa proved to be the most effective classifier of subjects administered GH. In this example, the classifier found using GeneRaVE is overfitted. Stepwise methods allow one to control the number of peaks that contribute to the classifier and are a means of avoiding overfitting. In this study we found that using more than two peaks led to overfitting, and that classifiers with two peaks perform no better than those with the single peak at 15,120 m/z.

An extensive body of evidence supports the dependence of serum IGF-I on GH status (30, 31, 32, 33, 34). Previous immunoassay analyses of the GH2000 sample set used in this study found that IGF-I was the most sensitive marker of GH exposure (8). In response to 28 d of GH administration in that study, the IGF-I response showed dose-dependent changes remarkably similar to those seen in our study for the 15.12-kDa protein, measured by SELDI-TOF MS peak intensity (8). This suggests that this protein may have similar potential diagnostic utility as an index of GH status.

The 15.12-kDa marker protein was characterized after chromatographic purification, by N-terminal Edman sequencing. Human HbA1 (15.126 kDa) was identified as a previously unrecognized biomarker of GH action, and its identity was confirmed by immunoabsorption from serum using an HbA1 antiserum coupled to epoxy-activated protein chips. GH and IGF-I have been proposed as significant determinants of the blood Hb concentration in elderly subjects, and the total Hb concentration has been positively correlated with concentrations of IGF-I and IGFBP-3 (35, 36). Multiple lines of evidence have also implicated GH-IGF axis proteins as stimulatory factors for erythropoieisis, resulting in increases in oxygen transport and Hb levels (36, 37). These studies all relate to heterotetrameric Hb in erythrocytes. In contrast, our profiling study measured HbA1 in serum. Because the reported effects of GH or IGF-I on total red blood cell Hb are relatively mild, the very significant changes in serum HbA1 observed in our study in response to GH may reflect increases in the free HbA1 chain rather than tetrameric hemoglobin. Interestingly, our analysis of HbA1 by SELDI-TOF MS in the sera of 10 subjects treated with erythropoietin under controlled conditions also showed a trend toward increased levels of the 15.12-kDa peak (HbA1), although the change was not statistically significant (data not shown).

There is considerable literature on the effects of sport on blood Hb levels (38), but there is a dearth of information on factors influencing serum levels of HbA1, which are less than 0.1% of blood Hb levels (~150 g/liter). It therefore remains to be tested how factors such as fitness, sport type, training, and injury might affect the utility of HbA1 measurement in detecting GH administration. If its utility is ultimately confirmed, SELDI-TOF MS will not necessarily be the most precise or the most sensitive method for its routine measurement. There are currently no published quantitative data for this analyte in serum (because Hb variants and subunits are generally measured in whole blood), so sensitivity comparisons are impossible, but the approximate limit of detectability by SELDI-TOF MS under the conditions described in this report would be 2.5 µg/ml (for a MS peak intensity of 0.05 units). Sample pretreatment, varying the sample size, or changing the conditions of the MS analysis might all improve the sensitivity. For other analytes, such as IGF-I (detected as a 7651 m/z peak in the univariate analysis), it is impossible to compare SELDI-TOF MS sensitivity to that of immunoassay, because the chip-binding conditions used were not optimized for IGF-I.

This study illustrates the novel use of serum proteomic profiling by SELDI-TOF MS to discover biomarkers of GH action. If fully validated, such biomarkers may be of value, not only in the detection of GH abuse by athletes, but in routine endocrine practice, perhaps as an adjunct to IGF-I measurement in GH deficiency, acromegaly, and other disorders. Although multivariate analysis techniques offer the potential for multiple protein peaks to be combined in a powerful classification algorithm, the considerable discriminatory potential of HbA1 alone in classifying subjects as GH or placebo treated, suggests that after additional validation, serum levels of this protein may provide a valuable new tool for the assessment of GH action.


   Acknowledgments
 
We thank the international collaborative GH2000 consortium, jointly funded by the European Union and the International Olympic Committee, for access to stored serum samples. Chris Howe, Australian Government Analytical Laboratories, National Measurement Institute, Sydney, generously provided sera from 10 subjects treated with erythropoietin.


   Footnotes
 
This work was supported by the Australian Government Anti-Doping Program, Department of Communications, Information Technology and the Arts.

First Published Online November 22, 2005

Abbreviations: ALS, Acid-labile subunit; HbA1, hemoglobin {alpha}-chain; hGH, human GH; IGFBP, IGF-binding protein; IMAC, immobilized metal affinity; SELDI-TOF MS, surface-enhanced laser desorption/ionization time of flight mass spectrometry; TFA, trifluoroacetic acid.

Received May 20, 2005.

Accepted November 7, 2005.


   References
Top
 Abstract
 Introduction
 Subjects and Methods
 Results
 Discussion
 References
 

  1. Macintyre JG 1987 Growth hormone and athletes. Sports Med 4:129–142[Medline]
  2. Parker ML, Utiger RD, Daughaday WH 1962 Studies on human growth hormone. II. The physiological disposition and metabolic fate of human growth hormone in man. J Clin Invest 41:262–268
  3. Owens D, Srivastava MC, Tompkins CV, Nabarro JD, Sonksen PH 1973 Studies on the metabolic clearance rate, apparent distribution space and plasma half-disappearance time of unlabelled human growth hormone in normal subjects and in patients with liver disease, renal disease, thyroid disease and diabetes mellitus. Eur J Clin Invest 3:284–294[Medline]
  4. Jorgensen JO, Pedersen SA, Thuesen L, Jorgensen J, Ingemann-Hansen T, Skakkebaek NE, Christiansen JS 1989 Beneficial effects of growth hormone treatment in GH-deficient adults. Lancet 1:1221–1225[Medline]
  5. Giustina A, Veldhuis JD 1998 Pathophysiology of the neuroregulation of growth hormone secretion in experimental animals and the human. Endocr Rev 19:717–797[Abstract/Free Full Text]
  6. Healy ML, Russell-Jones D 1997 Growth hormone and sport: abuse, potential benefits, and difficulties in detection. Br J Sports Med 31:267–268[Medline]
  7. Sonksen PH 2001 Insulin, growth hormone and sport. J Endocrinol 170:13–25[Abstract]
  8. Dall R, Longobardi S, Ehrnborg C, Keay N, Rosen T, Jorgensen JO, Cuneo RC, Boroujerdi MA, Cittadini A, Napoli R, Christiansen JS, Bengtsson BA, Sacca L, Baxter RC, Basset EE, Sonksen PH 2000 The effect of four wk of supraphysiological growth hormone administration on the insulin-like growth factor axis in women and men. GH-2000 Study Group. J Clin Endocrinol Metab 85:4193–4200[Abstract/Free Full Text]
  9. Wallace JD, Cuneo RC, Baxter R, Orskov H, Keay N, Pentecost C, Dall R, Rosen T, Jorgensen JO, Cittadini A, Longobardi S, Sacca L, Christiansen JS, Bengtsson BA, Sonksen PH 1999 Responses of the growth hormone (GH) and insulin-like growth factor axis to exercise, GH administration, and GH withdrawal in trained adult males: a potential test for GH abuse in sport. J Clin Endocrinol Metab 84:3591–3601[Abstract/Free Full Text]
  10. Longobardi S, Keay N, Ehrnborg C, Cittadini A, Rosen T, Dall R, Boroujerdi MA, Bassett EE, Healy ML, Pentecost C, Wallace JD, Powrie J, Jorgensen JO, Sacca L 2000 Growth hormone (GH) effects on bone and collagen turnover in healthy adults and its potential as a marker of GH abuse in sports: a double blind, placebo-controlled study. The GH-2000 Study Group. J Clin Endocrinol Metab 85:1505–1512[Abstract/Free Full Text]
  11. Wu Z, Bidlingmaier M, Dall R, Strasburger CJ 1999 Detection of doping with human growth hormone. Lancet 353:895
  12. Baxter RC, Martin JL 1986 Radioimmunoassay of growth hormone-dependent insulinlike growth factor binding protein in human plasma. J Clin Invest 78:1504–1512
  13. Baxter RC 1990 Circulating levels and molecular distribution of the acid-labile ({alpha}) subunit of the high molecular weight insulin-like growth factor-binding protein complex. J Clin Endocrinol Metab 70:1347–1353[Abstract]
  14. Ehrnborg C, Lange KH, Dall R, Christiansen JS, Lundberg PA, Baxter RC, Boroujerdi MA, Bengtsson BA, Healey ML, Pentecost C, Longobardi S, Napoli R, Rosen T 2003 The growth hormone/insulin-like growth factor-I axis hormones and bone markers in elite athletes in response to a maximum exercise test. J Clin Endocrinol Metab 88:394–401[Abstract/Free Full Text]
  15. Wallace JD, Cuneo RC, Bidlingmaier M, Lundberg PA, Carlsson L, Boguszewski CL, Hay J, Healy ML, Napoli R, Dall R, Rosen T, Strasburger CJ 2001 The response of molecular isoforms of growth hormone to acute exercise in trained adult males. J Clin Endocrinol Metab 86:200–206[Abstract/Free Full Text]
  16. Healy ML, Dall R, Gibney J, Bassett E, Ehrnborg C, Pentecost C, Rosen T, Cittadini A, Baxter RC, Sonksen PH 2005 Towards the development of a test for growth hormone abuse: a study of extreme physiological ranges of growth hormone dependent markers in 813 elite athletes in the post-competition setting. J Clin Endocrinol Metab 90:641–649[Abstract/Free Full Text]
  17. Yip TT, Lomas L 2002 SELDI ProteinChip array in oncoproteomic research. Technol Cancer Res Treat 1:273–280[Medline]
  18. Issaq HJ 2003 Application of separation technologies to proteomics research. Adv Protein Chem 65:249–269[Medline]
  19. Tang N, Tornatore P, Weinberger SR 2004 Current developments in SELDI affinity technology. Mass Spectrom Rev 23:34–44[CrossRef][Medline]
  20. Petricoin III EF, Ornstein DK, Paweletz CP, Ardekani A, Hackett PS, Hitt BA, Velassco A, Trucco C, Wiegand L, Wood K, Simone CB, Levine PJ, Linehan WM, Emmert-Buck MR, Steinberg SM, Kohn EC, Liotta LA 2002 Serum proteomic patterns for detection of prostate cancer. J Natl Cancer Inst 94:1576–1578[Abstract/Free Full Text]
  21. Kozak KR, Amneus MW, Pusey SM, Su F, Luong MN, Luong SA, Reddy ST, Farias-Eisner R 2003 Identification of biomarkers for ovarian cancer using strong anion-exchange ProteinChips: potential use in diagnosis and prognosis. Proc Natl Acad Sci USA 100:12343–12348[Abstract/Free Full Text]
  22. Zhang L, Yu W, He T, Yu J, Caffrey RE, Dalmasso EA, Fu S, Pham T, Mei J, Ho JJ, Zhang W, Lopez P, Ho DD 2002 Contribution of human {alpha}-defensin 1, 2, and 3 to the anti-HIV-1 activity of CD8 antiviral factor. Science 298:995–1000[Abstract/Free Full Text]
  23. Tibshirani R, Hastie T, Narasimhan B, Soltys S, Shi G, Koong A, Le QT 2004 Sample classification from protein mass spectrometry, by ‘peak probability contrasts.’ Bioinformatics 20:3034–3044[Abstract/Free Full Text]
  24. Soille P 1999 Morphological image analysis: principles and applications. New York: Springer
  25. Wu B, Abbott T, Fishman D, McMurray W, Mor G, Stone K, Ward D, Williams K, Zhao H 2003 Comparison of statistical methods for classification of ovarian cancer using mass spectrometry data. Bioinformatics 19:1636–1643[Abstract/Free Full Text]
  26. Kiiveri H 2003 A Bayesian approach to variable selection when the number of variables is very large. In: Goldstein DR, ed. Science and statistics: a festschrift for Terry Speed. Beachwood, OH: Institute of Mathematical Statistics; 127–143
  27. Ebert MP, Meuer J, Wiemer JC, Schulz HU, Reymond MA, Traugott U, Malfertheiner P, Rocken C 2004 Identification of gastric cancer patients by serum protein profiling. J Proteome Res 3:1261–1266[Medline]
  28. Haven CJ, Howell VM, Eilers PH, Dunne R, Takahashi M, van Puijenbroek M, Furge K, Kievit J, Tan MH, Fleuren GJ, Robinson BG, Delbridge LW, Philips J, Nelson AE, Krause U, Dralle H, Hoang-Vu C, Gimm O, Morreau H, Marsh DJ, Teh BT 2004 Gene expression of parathyroid tumors: molecular subclassification and identification of the potential malignant phenotype. Cancer Res 64:7405–7411[Abstract/Free Full Text]
  29. Ponnampalam AP, Weston GC, Trajstman AC, Susil B, Rogers PA 2004 Molecular classification of human endometrial cycle stages by transcriptional profiling. Mol Hum Reprod 10:879–893[Abstract/Free Full Text]
  30. D’Ercole AJ 1996 Insulin-like growth factors and their receptors in growth. Endocrinol Metab Clin North Am 25:573–590[CrossRef][Medline]
  31. Daughaday WH 2000 Growth hormone axis overview–somatomedin hypothesis. Pediatr Nephrol 14:537–540[CrossRef][Medline]
  32. Le Roith D, Scavo L, Butler A 2001 What is the role of circulating IGF-I? Trends Endocrinol Metab 12:48–52[CrossRef][Medline]
  33. Isaksson O 2004 GH, IGF-I, and growth. J Pediatr Endocrinol Metab 17(Suppl 4):1321–1326
  34. Jorgensen JO, Vahl N, Hansen TB, Skjaerbaek C, Fisker S, Orskov H, Hagen C, Christiansen JS 1998 Determinants of serum insulin-like growth factor I in growth hormone deficient adults as compared to healthy subjects. Clin Endocrinol (Oxf) 48:479–486[CrossRef][Medline]
  35. Nilsson-Ehle H, Bengtsson BA, Lindstedt G, Mellstrom D 2005 Insulin-like growth factor-1 is a predictor of blood haemoglobin concentration in 70-yr-old subjects. Eur J Haematol 74:111–116[Medline]
  36. Sohmiya M, Kato Y 2005 Human growth hormone and insulin-like growth factor-I inhibit erythropoietin secretion from the kidneys of adult rats. J Endocrinol 184:199–207[Abstract/Free Full Text]
  37. Vihervuori E, Virtanen M, Koistinen H, Koistinen R, Seppala M, Siimes MA 1996 Hemoglobin level is linked to growth hormone-dependent proteins in short children. Blood 87:2075–2081[Abstract/Free Full Text]
  38. Shaskey DJ, Green GA 2000 Sports haematology. Sports Med 29:27–38[CrossRef][Medline]




This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
91/2/671    most recent
Author Manuscript (PDF)
Right arrow Submit a related Letter to the Editor
Right arrow Purchase Article
Right arrow View Shopping Cart
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow Request Copyright Permission
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Chung, L.
Right arrow Articles by Baxter, R. C.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Chung, L.
Right arrow Articles by Baxter, R. C.
Right arrowPubmed/NCBI databases
*Substance via MeSH
Related Collections
Right arrow Metabolism
Right arrow Neuroendocrinology and Pituitary

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Endocrinology Endocrine Reviews J. Clin. End. & Metab.
Molecular Endocrinology Recent Prog. Horm. Res. All Endocrine Journals