Clinical Biochemistry 40 (2007) 1113 – 1121
Identification of a predictive biomarker for the beneficial effect of a Kampo (Japanese traditional) medicine keishibukuryogan in rheumatoid arthritis patients Kazuo Ogawa a , Tetsuko Kojima a , Chinami Matsumoto a , Satoshi Kamegai b , Takuya Oyama b , Yukari Shibagaki b , Hiroshi Muramoto b , Tetsuo Kawasaki b , Hiroshi Fujinaga c , Kozo Takahashi c , Hiroaki Hikiami d , Hirozo Goto d,g , Chizuru Kiga e,f , Keiichi Koizumi e , Hiroaki Sakurai e,g , Yutaka Shimada d,g , Masahiro Yamamoto a , Katsutoshi Terasawa h , Shuichi Takeda a , Ikuo Saiki e,g,⁎ a
Central Research Laboratories, Tsumura and Co., Ibaraki, Japan Bioinformatics Division, INTEC Web and Genome Informatics Corporation, Toyama, Japan c Department of Japanese Oriental Medicine, Toyama Prefectural Central Hospital, Toyama, Japan Department of Japanese Oriental (Kampo) Medicine, Graduate School of Medicine, University of Toyama, Toyama, Japan e Division of Pathogenic Biochemistry, Institute of Natural Medicine, University of Toyama, Toyama, Japan f Toyama New Industry Organization, Toyama, Japan g The 21st Century COE Program, University of Toyama, Toyama, Japan h Department of Japanese-Oriental (Kampo) Medicine, Graduate School of Medicine, Chiba University, Chiba, Japan b
d
Received 1 February 2007; received in revised form 24 May 2007; accepted 9 June 2007 Available online 3 July 2007 This study is dedicated to the memory of the late Mr. Mizushima (INTEC Web and Genome Informatics Corporation).
Abstract Objectives: Kampo (Japanese traditional herbal) medicines are now ethically used in Japan as pharmaceutical grade prescription drugs. However, there are distinct groups of responders and non-responders to Kampo medicines. We searched for biomarker candidates to discriminate responders from non-responders to keishibukuryogan (KBG); one of the most frequently used Kampo medicines. Design and methods: A combination of SELDI technology and a decision tree analysis with proprietary developed bioinformatics tools was applied to 41 (32 for tree construction and 9 for validation test) plasma samples obtained from rheumatoid arthritis (RA) patients. A candidate biomarker protein was identified using LC–MS/MS. Results: The constructed tree with measurable reliability contained only a single peak which was identified as haptoglobin alpha 1 chain (Hpα1). Conclusion: Hpα1 is a biomarker candidate for discriminating responders from non-responders to KBG treatment for RA. The present results may open the way to the establishment of “evidence-based” complementary and alternative medicine. © 2007 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved. Keywords: Kampo medicine; Keishibukuryogan; Responder; SELDI–TOF MS; Cross detector; Peak separability analysis; Haptoglobin
Introduction The use of complementary and alternative medicine (CAM) has grown dramatically in recent years. The widespread use of CAM prompted us to conduct rigorous research to verify ⁎ Corresponding author. Division of Pathogenic Biochemistry, University of Toyama, 2630 Sugitani, Toyama 930-0194, Japan. Fax: +81 76 434 5058. E-mail address:
[email protected] (I. Saiki).
whether assumptions concerning the safety and efficacy of this treatment are valid [1]. Herbal medicines, which represent the most important therapies in CAM, are also an essential part of traditional medicine in almost any culture. A number of epidemiological and interventional studies using herbal products have produced positive results. However, trials for evaluating the quality of herbal medicines, especially the old trials, are generally of a low standard because most of them were performed by herbal medicine manufacturers with limited
0009-9120/$ - see front matter © 2007 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved. doi:10.1016/j.clinbiochem.2007.06.005
1114
K. Ogawa et al. / Clinical Biochemistry 40 (2007) 1113–1121
resources and expertise. A fundamental problem in all clinical research involving herbal medicine is whether different products, extracts or even different lots of the same extract have comparable, constant and acceptable efficacies. The unavailability of the high-quality, standardized herbal products is a major obstacle in obtaining a reliable assessment of the efficacy and safety of these products. Apparently the “same” herbal drugs are often made using different plant species or different parts of the same plant (e.g., leaf, root or flower) with a generally low level of quality control. This variation may be of even greater concern for botanicals in which the active ingredient is hardly known [2]. “Kampo” medicine is a Japanese traditional medicine that originated from ancient Chinese traditional medicine[3]. Among “herbal medicines”, Kampo medicine occupies a unique position because it is covered by public health insurance. Over 100 Kampo formulas have been approved as ethical drugs by the Ministry of Health, Labour and Welfare of Japan, and are clinically used for the treatment of a wide variety of diseases by physicians who are trained in Western medicine. These medicines are manufactured on a modern industrial scale, whereby the quality and quantity of the ingredients are standardized according to strict, scientific quality control practices. Many clinical trials using Kampo medicines – including more than 10 multicenter, placebo-controlled, double-blind studies – have suggested that these medicines possess significant beneficial effects in terms of modern medicine [4,5]. However, these trials have suggested that it is essential to distinguish the responder from the non-responder at the beginning of the therapy to achieve the expected therapeutic outcome. Certain Kampo medicines have been reported to produce a dramatic therapeutic effect in “responder” patients; however, it is difficult to confirm their efficacies in clinical trials because of the presence of a large number of associated “nonresponders” in the cohorts. Therefore, selective use of these medicines for the treatment of “responder” patients should increase their effectiveness and safety. A recent attention in modern medicine has been directed towards molecular-target therapies with higher efficacy and less toxicity. For example, a therapy with trastuzumab (herceptin) was used selectively and successfully for the treatment of metastatic breast cancer patients overexpressing human epithelial growth factor receptor type 2 (HER2) protein [6]. Another example is the diagnostic test of polymorphism of genes encoding the drug metabolizing enzyme, uridine diphosphate glucuronosyltransferase 1A1 (UGT1A1), which makes irinotecan therapy for colorectal and lung cancer patients safer and more efficient [7]. These approaches are achievable only when the drug is strictly defined and its metabolism, target molecules and mode of action are fully elucidated. A similar approach involving therapies using Kampo medicine are highly impracticable because Kampo formulas consist of multiple crude drugs (mostly of plant origin; some are of animal or mineral origin), and consequently contain a large number of chemical substances, and, above all, the active ingredients responsible for their therapeutic actions are poorly elucidated.
Recently, an alternative approach to tailor the treatment response is rapidly developing. Mass spectrometric-based protein analysis, such as surface-enhanced laser desorption/ ionization (SELDI) technology, is a novel high-throughput proteomic technique that has already been used for the discovery of disease-related biomarkers in biological fluids, including serum, plasma and urine. Results from a number of studies suggest that this technology can be used as a tool for clinical diagnosis when combined with n-dimensional analyses algorithms and bioinformatics tools, including genetic algorithms [8], decision trees [9–11] and neural network algorithms [12]. Analysis of high volumes of SELDI data using these powerful algorithm software packages can be a valuable aid to accurate diagnosis. For example, this technique has been successfully applied to discriminate neoplastic and nonneoplastic diseases of the ovary [8,9], renal cell carcinoma from other urologic diseases [10,12], and prostate cancer from both benign hyperplasia and healthy men [11]. The potential of the technology has been expanded to diagnose other diseases such as Alzheimer's diseases [13], renal allograft rejection [14] and non-alcoholic fatty liver diseases [15]. Undoubtedly, utilization of this technology is a promising approach to classify and/or to predict the outcome of any interventional therapy, including CAM. In the present study, we used SELDI–TOF MS to obtain proteomic profiles of plasma samples from rheumatoid arthritis (RA) patients treated with a Kampo medicine keishibukuryogan (KBG). KBG, one of the most frequently used Kampo medicines, is composed of five medical herbs (Cinnamomi cortex, Paeoniae radix, Moutan cortex, Hoelen and Persicae semen). Use of KBG is widely accepted in Japan as an effective alternative treatment for hypermenorrhea, dysmenorrhea [16,17] and climacteric symptoms [18,19], especially in cases where hormone replacement therapy were contraindicated. We have recently observed that KBG has been beneficial and improving in a proportion of RA patients[20]. Thus, we endeavored to search for a biomarker that could distinguish responder from non-responder to KBG therapy in RA patients. Using a combination of decision tree analysis and newly developed peak detection and selection methods (described in the Supplementary material), we found candidate biomarker peak(s) that may be used to improve the responder rate for KBG in RA patients. These results were obtained despite the limited number of subjects/samples in the present study. We have also identified the protein biomarker(s), which will allow additional validation using independent methods. Identification of the biomarker protein(s) is a first step towards understanding the mechanism of the heterogenic response to KBG therapy in patients. Materials and methods Patients and drugs This study was approved by the ethics committee of Toyama University Hospital (TUH) and Toyama Prefectural Central Hospital (TPCH), and all patients provided written informed consent.
K. Ogawa et al. / Clinical Biochemistry 40 (2007) 1113–1121
1115
age of these patients was 63.6 years (range 43–79) and 55.6% of these patients achieved ACR20 (Table 1).
The patients were selected as described below. Patients who displayed symptoms that fulfilled the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) 1987 revised criteria for the classification of RA were included in this study as eligible subjects [21]. Patients received concomitant drugs, disease-modifying anti-rheumatic drugs (DMARD), nonsteroidal anti-inflammatory drugs (NSAID), steroid or other Kampo medicines for RA. Some of these patients were undergoing treatment for concomitant diseases such as diabetes, hyperlipemia and hyperpiesia. In such cases, treatment was continued without changing the drugs and dosages 3 months before and during the period of this study. Patients were followed for 12 weeks at outpatient clinic. All patients were treated with KBG. Patients achieving the ACR 20% improvement criteria (ACR20) at 12 weeks were defined as “the responders to KBG treatment”. In the present study, two different dosage forms of KBG were used. At the TUH, 19 patients were treated with 4 g of “hospital-made” KBG three times daily for 12 weeks. The “hospital-made” KBG consisted of 5 medical plants, registered in Pharmacopoeia of Japan. Each KBG tablet weighs 2 g and contains 0.2 g of Cinnamomi cortex (Cinnamomum cassia Blume), 0.2 g of Paeoniae radix (Paeonia lacitiflora Pallas), 0.2 g of Persicae semen (Prunus persica Batsch), 0.2 g of Hoelen (Poria cocos Wolf), and 0.2 g of Moutan radix (Paeonia frutricosa Andrews), and 1 g of honey as an excipient. At the TPCH, twenty-two patients were treated with 2.5 g of KBG (TJ25; obtained from Tsumura and Co., Tokyo, Japan) three times daily for 12 weeks. TJ-25 is manufactured under a standardized process that was developed using a spray-dried powder of KBG. The contents of TJ-25 are equivalent to those of the “hospitalmade” KBG except for honey, which is replaced by lactose. The Ministry of Health, Labour and Welfare of Japan approves both dosage forms as ethical “KBG” drugs. Plasma samples were obtained before the treatment and stored at − 80 °C until SELDI analysis. Among these samples, the plasma samples of 32 patients collected between September 2002 and January 2004 were submitted for Discovery Set; the mean age of these patients was 59.4 years (range 38–79) and 37.5% of these patients achieved ACR20. Samples collected from 9 patients between November 2003 and January 2005 were submitted for the blind test set (Validation Set); the mean
SELDI protein profiling Each plasma sample was analyzed on four different array surfaces: anion exchange (Q10), cation exchange (CM10), metal binding (IMAC30–Cu) and hydrophobic (H50). Sinapinic acid (SPA) or α-cyano-4-hydroxycinnamic acid (CHCA) was used as the energy absorbing molecule (EAM). These materials were from Ciphergen Biosystems, Inc. (Fremont, CA, USA). In total, each plasma sample was subjected to 16 optimal binding and analysis conditions using a combination of denatured/non-denatured, chip surface and EAM (different SPA and CHCA mass range: SPA-Low, m/z 3000–10,000; SPA-High, m/z 10,000–30,000; CHCA, m/z 3000–10,000; m/z: mass to charge ratio). The arrays were analyzed using a Protein Biology System IIc Reader (PBS IIc; Ciphergen Biosystems, Inc.) and all spectra were collected using the ProteinChip Biomarker software version 3.1 (Ciphergen Biosystems, Inc.). After baseline subtraction on the software, each spectrum was exported to a CSV file. The exported spectrum data were analyzed for peak detection/alignment using a proprietary-developed peak detection procedure (Cross Detector). Obtained peaks were normalized to the total ion current of m/z starting from 3000 for SPALow and CHCA or from 10,000 for SPA-High. Details of the binding/analysis conditions and the “Cross Detector” method are described in the Supplementary material. Study design The predictive decision tree model for response to KBG was constructed by R 2.0.1 based on a proprietary-developed procedure named the Peak Separability Analysis (PSA) model. The analysis used a Discovery Set consisting of 32 samples (16 samples from TUH and 16 samples from TPCH). The constructed tree was applied to the blind test set (Validation Set) consisting of 9 samples (3 samples from TUH and 6 samples from TPCH) and assessed for accuracy, sensitivity, and specificity. Details of the PSA model are described in the Supplementary material.
Table 1 Details of the patients Data set
For Discovery Set
Institution
TUH a TPCH b
For Validation Set c a b c
TUH TPCH
Patients
Age
Percentage of patients achieving ACR20 (responders/patients)
Sex
n
Mean
Range
Male Female Male Female Male Female
1 15 3 13 1 8
58.3
38–75
31.3% (5/16)
60.4
46–79
43.8% (7/16)
63.6
43–79
55.6% (5/9)
TUH: Toyama University Hospital. TPCH: Toyama Prefectural Central Hospital. Validation Set: 3 patients were from TUH and 6 patients were from TPCH.
1116
K. Ogawa et al. / Clinical Biochemistry 40 (2007) 1113–1121
Purification and identification of biomarker candidate Ion exchange fractionation was undertaken on a Q HyperD F Spin Column (Ciphergen Biosystems, Inc.) pre-equilibrated with glycine buffer (100 mM NaCl, 50 mM glycine, pH 10). Plasma was diluted at a ratio of 1:1.5 in urea denaturing buffer (7 M urea, 2 M thiourea, 1% dithiothreitol, 4% CHAPS and 2% ampholyte) and incubated for 20 min at 4 °C on a rotator. Samples were then diluted 1:9 in glycine buffer and applied to the column. Bound proteins were eluted using the glycine buffer with a stepwise increase in NaCl (100, 150, and 200 mM). Eluted fractions were pooled and concentrated by cut-off membrane fractionation (YM3; Millipore, Billerica, MA, USA). SELDI–TOF MS was used to monitor each fraction for ions of interest. Biomarker candidates were identified using high resolution two-dimensional gel electrophoresis (2-DE) coupled with mass spectrometry analysis (performed by APRO Life Science Institute, Inc., Tokushima, Japan). Separation in the first dimension was carried out using an immobilized pH gradient (IPG) ReadyStrip gel (7 cm, pH 3–10NL; Bio-Rad, Hercules, CA, USA). The IPG strips were equilibrated for 15 min in equilibration buffer A (6 M urea, 30% glycerol, 2% sodium dodecyl sulfate, 0.005% bromophenol blue, 50 mM Tris–HCl, pH 8.5, and 1% dithiothreitol) followed by another 15 min in equilibration buffer B (Equilibration Buffer A with 4.5% iodoacetamide instead of dithiothreitol). Equilibrated IPG strips were transferred onto precast 16.5% gels, and separation in the second dimension was performed by Tris–Tricine sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS– PAGE). Gels were stained with SYPRO Ruby (Invitrogen,
Carlsbad, CA, USA) and scanned with ProFINDER 2D (PerkinElmer, Wellesley, MA, USA). Gel pieces containing the proteins of interest were excised and subjected to in-gel trypsin digestion followed by liquid chromatography–tandem mass spectrometry (LC–MS/MS) analysis. MS/MS spectra were submitted to the database mining tool Mascot (Matrix Science, London, UK) for identification. Results Determination of predictive tree model and blind test In order to obtain more accurate and exhaustive peak profiles, we used a newly developed proprietary peak detection algorithm “Cross Detector” for alignment of the peaks in SELDI profiles (see Supplementary material). Cross validation among the datasets of different hospitals (TUH and TPCH) and of different lots of measurements were conducted to evaluate the performance of the PSA method compared to the conventional peak selection method based on the significant difference of peak intensities. Decision trees based on the PSA method resulted in higher and more reproducible prediction performances compared to the peak intensity method (see Supplementary material). In the main test, the decision tree based on the PSA was constructed using the Discovery Set consisting of 32 SELDI profiles. The tree that most correctly discriminated responder from non-responder to KBG in these sets used only one splitter as m/z 9200 (Figs. 1A, B), and was obtained with an accuracy of 68.8%, sensitivity of 75.0%, and specificity of 65.0% (Fig. 1C). The constructed tree was applied to 9 blind test samples (Validation Set). The predictions made by the tree
Fig. 1. Classification of “responder” and “non-responder” to KBG in the Discovery Set. (A) Diagram of the best decision tree algorithm based on PSA method (red: responder, blue: non-responder). (B) Representative spectra from the m/z 9200-positive (upper panel) and m/z 9200-negative samples (lower panel). (C) Results of classification of the Discovery Set and Validation Set using the best decision tree (A).
K. Ogawa et al. / Clinical Biochemistry 40 (2007) 1113–1121
1117
Table 2 Comparison of the responder rate
Protein purification and identification
Cohort
In order to further characterize the candidate marker, fractions eluted from the anion exchange spin column were applied to SELDI protein chip arrays. We identified the m/z 9200 and m/z 15,970 peaks using this method. The two candidate marker proteins were identified using a combination of 2-DE, enzymatic digestion and LC–MS/MS. Four major spots with different pI values (spots a, b, c and d boxed in Fig. 3A) that were absent on the 2-DE gel from m/z 9200-negative plasma (Fig. 3B) were extracted from the gel indicated on Fig. 3A and analyzed by SELDI–TOF MS. All four samples produced SELDI peaks identical to the m/z 9200 peak (Fig. 4A). Each spot was subjected to in-gel trypsin digestion followed by LC–MS/MS. The MS/MS spectra were then submitted to Mascot for identification. Spots a, b and c were identified as haptoglobin alpha 1 chain (Hpα1). Spot d, which was detected at a more basic pI location compared to the others, was identified as Hpα1 with an additional arginine at the C-terminus (Table 3). However, a further four major spots with different pI values (spots e, f, g and h boxed in Fig. 3A) that were absent on the 2-DE gel from m/z 15,970-negative plasma (Fig. 3C) produced SELDI peaks identical to the m/z 15,970 peak (Fig. 4B). Spots e and f were identified as haptoglobin alpha 2 chain (Hpα2) by the same methods described above. Spots g and h, which were detected at more basic pI location compared to spots e and f, were identified as Hpα2 with an additional arginine at the C-terminus (Table 4). The mass values of these extracted spots were greater than 9200 or 15,970, which can be attributed to the carbamidomethylation of cysteine residues caused in the process of 2-DE (i.e., 57 Da difference per cysteine residue).
Number Number Responder rate of patients of responders All m/z 9200 m/z 9200 (+) (−)
All 41 For Discovery Set TUH 16 TPCH 16 For Validation Set TUH + TPCH 9
17
41.5% 60.0%
23.8%
5 7
31.3% 57.1% 43.8% 55.6%
11.1% 28.6%
5
55.6% 75.0%
40.0%
The “responder rate” is defined for each cohort as follows: column of “All” is percentages of responders in all patients; column of “m/z 9200 (+)” is % of responders in m/z 9200-positive patients; column of “m/z 9200 (−)” is % of responders in m/z 9200-negative patients.
scored an accuracy of 66.7%, sensitivity of 60.0%, and specificity of 75.0% (Fig. 1C). Supplementary analysis Table 2 shows the relationship between the presence/absence of m/z 9200 peak and responder rate to KBG in three cohorts: two for the Discovery Set (16 patients from TUH and 16 patients from TPCH) and one for the Validation Set (3 patients from TUH and 6 patients from TPCH). In these three sets, the responder rates were between 55.6% and 75.0% for the m/z 9200-positive patients (total rate for the positive group was 60.0%) and between 11.1% and 40.0% for the m/z 9200negative patients (total rate for the negative group was 23.8%). Additionally, to find other candidate markers, a series of the PSA method was re-applied only to the m/z 9200-positive profiles. The number of m/z 9200-positive patients was 20. The new tree that most correctly discriminated responder from nonresponder to KBG used only one splitter as m/z 15,970 (Figs. 2A, B). The responder rate in the presence of m/z 15,970 peak, 73.3%, was higher than the absence of the peak, 20.0%. However, a decision tree was not built using m/z 9200-negative profiles. All m/z 9200-negative patients have been found to be m/z 15,970 positive. These results suggest that the presence of the m/z 9200 and 15,970 peaks in a plasma sample indicates higher responsiveness to KBG treatment.
Discussion In this study, we searched for biomarker candidates in the plasma of RA patients in order to predict response to KBG, one of the most frequently used Kampo medicines. Using a combination of SELDI technology with two new analysis procedures, Cross Detector and PSA described in the Supplementary material, we built a decision tree that discriminated
Fig. 2. Classification of “responder” and “non-responder” to KBG in the all m/z 9200-positive patients. (A) Diagram of the best decision tree algorithm based on PSA method (red: responder, blue: non-responder). (B) Representative spectra from the m/z 15,970-positive (upper panel) and m/z 15,970-negative samples (lower panel).
1118
K. Ogawa et al. / Clinical Biochemistry 40 (2007) 1113–1121
Fig. 3. 2-DE protein patterns from three types of RA patient's plasma: (A) m/z 9200 and m/z 15,970 both positive; (B) m/z 9200-negative; (C) m/z 15,970-negative. Area that we have hoped to compare was zoomed (below). Four major spots on gel A that are absent on gel B are marked by arrows a, b, c and d. Another four major spots on gel A that are absent on gel C are marked by arrows e, f, g and h. The fractions eluted from the anion exchange spin column from each plasma sample were separated by 2-DE and stained with SYPRO Ruby stain.
responder from non-responder, with a certain degree of accuracy, to KBG in plasma samples derived from RA patients. By adopting this approach, we found a biomarker (m/z 9200) whose presence in plasma samples may indicate a higher responsiveness to KBG treatment. We identified this biomarker to be Hpα1 by using 2-DE and LC–MS/MS. KBG is one of the most important Kampo medicines for improving “Oketsu”. The “Oketsu” symptom, a pathological concept diagnosed by Kampo medicine [22], is thought to
correlate with hemorheological abnormalities such as elevation of blood viscosity, acceleration of erythrocyte aggregability and a deterioration of erythrocyte deformability [23–25]. Recent studies have reported that KBG improves peripheral blood flow in post-menopausal women with hot flashes [19], and increases deformability of red blood cell in patients with multiple lacunar infarctions [26]. In addition, Sekiya et al. [27,28] have reported that the antioxidative effect of KBG prevented the progression of atherosclerosis in cholesterol-fed rabbits and inhibited free
Fig. 4. Re-analysis by SELDI–TOF MS of four major spots with different pI values (spots a, b, c and d boxed in Fig. 3A) that are absent on the 2-DE gel from m/z 9200-negative plasma (Fig. 3B) and SELDI–TOF MS of four other major spots with different pI values (spots e, f, g and h boxed in Fig. 3A) that are absent on the 2DE gel from m/z 15,970-negative plasma (Fig. 3C). The gel strips containing each spot were excised and each protein was extracted and further subjected to SELDI– TOF MS analysis. The mass values of these extracted spots were larger than 9200 or 15,970, which can be attributed to the carbamidomethylation of cysteine residues occurred in the process of 2-DE (see Tables 3 and 4).
K. Ogawa et al. / Clinical Biochemistry 40 (2007) 1113–1121
1119
Table 3 Peptide sequences of m/z 9200 determined from the LC–MS/MS analysis
A: amino acid sequence of human Hpα1 (Swiss-Prot entry P00737) and the matched sequences printed in red and blue. B: peptide sequences determined from the LC–MS/MS analysis. Samples were carbamidomethylated before the second dimension of electrophoresis by SDS–PAGE. Blue sequence, NPANPVQR, was determined only from spot d, which was the spot found on the most basic side of the four spots (see Fig. 3). (For interpretation of the references to colour in this table legend, the reader is referred to the web version of this article.)
radical-induced lysis of rat red blood cells. These studies strongly suggest that the effects of KBG are well associated with the hemolytic–antihemolytic status of the patients. In light of the above-described pharmacological properties of KBG, it is of primary interest that a possible marker for a responder to KBG therapy was identified as Hpα1. Hp, one of the acute-phase proteins induced by infection, tissue injury, and malignancy, is an α2-sialoglycoprotein with hemoglobin (Hb)binding capacity and consists of two different polypeptides (αand β-chain). The β-chain (40 kDa) is heavier than the α-chain and is identical in all Hp types. Consequently, Hp polymorphism arises from the variant α-chains (Hpα1 and Hpα2) to form three major phenotypes, Hp 1-1 (allelic genes are both Hpα1), Hp 2-1 (allelic genes consist of Hpα1 and Hpα2), and Hp 2-2 (allelic genes are both Hpα2). Hp binds to Hb to prevent both iron loss and kidney damage during hemolysis. In addition, Hp is known to act as an antioxidant, inhibit prostaglandin synthesis and possesses angiogenic and bacteriostatic activities [29]. KBG is thought to exert its beneficial effect by directly enhancing the heme-scavenging activity of the Hb–Hp complex. Alternatively, Hp might amplify the modulating effect of KBG on the hemolytic–antihemolytic status by binding to Hb or by scavenging oxidative radicals. Differences detected in the peak intensity of Hpα1 among patients in the present study may reflect the existence or nonexistence of the Hpα1 gene, rather than just differing amounts of Hpα1 protein. Peak intensities of m/z 9200 in the samples judged as “m/z 9200 absent” by PSA were extraordinary weak, suggesting a complete absence of Hpα1 protein. In this case, a difference in the response to KBG treatment may have originated not by altered expression levels of Hpα1 proteins (induced by,
for example, inflammation or injury as an acute-phase protein) but by a difference of Hp genotype. A preliminary follow-up study further suggested a possible relationship between the improved efficacy of KBG and Hp genotype. When a series of the PSA process, from the peak selection to the decision tree analysis, was re-applied only to the m/z 9200-positive profiles, we found a second biomarker candidate, m/z 15,970 polypeptide (Fig. 2). Surprisingly, 2-DE and LC–MS/MS analysis revealed that the m/z 15,970 polypeptide was Hpα2 (Table 4). Thus the combination of presence/absence of m/z 9200 and/or m/z 15,970 directly represents the genetic difference of Hp phenotype. The present study suggests that the Hp genotype can be used to differentiate between KBG responder and KBG non-responder patients (i.e., Hp 1-1 and Hp 2-2 phenotypes represent “nonresponders” whereas the Hp 2-1 phenotype represents “responders”). The difference in responder rates among patients with different Hp genotypes have also been noted for vitamin E and vitamin C in double-blind placebo-controlled randomized clinical trials [30]. Because the sample size in the present study was limited, extensive verification studies, including large-scale clinical trials, are necessary to evaluate the usefulness of this hypothesis. Although Kampo medicines are believed to elicit a dramatic therapeutic effect on “responder” patients, it is difficult to confirm their efficacies in clinical trials in cohorts that include a large number of “non-responders”. In the present study, 41.5% of the patients achieved ACR20 (Table 2). The outcome may not be “dramatic” in itself, however, it must be noted that all the patients in the present study had received various conventional, sometimes extensive, therapies based on modern Western medicines for a considerable time before KBG treatment.
1120
K. Ogawa et al. / Clinical Biochemistry 40 (2007) 1113–1121
Table 4 Peptide sequences of m/z 15,970 determined from the LC–MS/MS analysis
A: amino acid sequence of human Hpα2 (Swiss-Prot entry P00738) and the matched sequences printed in red and blue. B: peptide sequences determined from the LC–MS/MS analysis. Samples were carbamidomethylated before the second dimension of electrophoresis by SDS–PAGE. Blue sequence, NPANPVQR, was determined from spots g and h, which were the spots found on the basic side of the four spots (see Fig. 3). (For interpretation of the references to colour in this table legend, the reader is referred to the web version of this article.)
Furthermore, 6 patients achieved ACR50 (5 of them were identified as Hp 2-1 phenotype). Although not dramatic, we believe that the add-on improvement elicited by KBG treatment is at least noteworthy. However, the necessity to select patients as potential responders is obviously required to improve the efficacy of KBG therapy for RA. We found that the responder rates of the m/z 9200 (−) patients were lower than those of the m/z 9200 (+) patients in all three cohorts tested (Table 3). However, the sensitivity and specificity of the biomarker candidates found in this study is not high enough to use as a reliable predictive marker in clinically diagnosing responder from non-responder for KBG. Nevertheless, using the criteria that the m/z 9200 positives (or Hp 2-1 phenotypes) are “possible KBG responders”, we will be able to pre-screen subjects in our upcoming clinical trial. This approach will allow us to obtain consistent trial results with higher responder rates and may result in producing uniform efficacy for KBG. In conclusion, the present study provides a promising approach to finding biomarkers for predicting responders to
various Kampo and traditional medicine-based therapies. Moreover, the results obtained might lead to the development of “tailor-made” therapeutic modalities of Kampo medicine and CAM, which provides more consistent and reliable therapeutic outcomes and opens the way to the integration of traditional medicine/CAM and modern medicine. Acknowledgments This study was supported by a grant for CLUSTER (Cooperative Link of Unique Science and Technology for Economy Revitalization) from the Ministry of Education, Culture, Sports, Science and Technology, Japan. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.clinbiochem.2007.06. 005.
K. Ogawa et al. / Clinical Biochemistry 40 (2007) 1113–1121
References [1] Miller FG, Emanuel EJ, Rosenstein DL, Straus SE. Ethical issues concerning research in complementary and alternative medicine. JAMA 2004;291:599–604. [2] Shekelle PG, Morton SC, Suttorp MJ, Buscemi N, Friesen C. Challenges in systematic reviews of complementary and alternative medicine topics. Ann Intern Med 2005;142:1042–7. [3] Yamaura T. Preface. In: Hosoya E, Yamada Y, editors. Recent advance in the pharmacology of Kampo (Japanese herbal) medicines, International Congress Series 854. Amsterdam, MD: Excerpta Medica; 1988. p. 3. [4] Terasawa K, Shimada Y, Kita T, et al. Chotosan in the treatment of vascular dementia: a double-blind, placebo-controlled study. Phytomedicine 1997;4:15–22. [5] Miyamoto T, Inoue H, Kitamura S, et al. Effect of TUMURA Sho-seiryuto (TJ-19) on bronchitis in a double-blind placebo-controlled study. J Clin Ther Med 2001;17:1189–214. [6] Slamon DJ, Leyland-Jones B, Shak S, et al. Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer that overexpresses HER2. N Engl J Med 2001;344:783–92. [7] Ando M, Hasegawa Y, Ando Y. Pharmacogenetics of irinotecan: a promoter polymorphism of UGT1A1 gene and severe adverse reactions to irinotecan. Invest New Drugs 2005;23:539–45. [8] Petricoin EF, Ardekani AM, Hitt BA, et al. Use of proteomic patterns in serum to identify ovarian cancer. Lancet 2002;359:572–7. [9] Vlahou A, Schorge JO, Gregory BW, Coleman RL. Diagnosis of ovarian cancer using decision tree classification of mass spectral data. J Biomed Biotechnol 2003;2003:308–14. [10] Won Y, Song HJ, Kang TW, Kim JJ, Han BD, Lee SW. Pattern analysis of serum proteome distinguishes renal cell carcinoma from other urologic diseases and healthy persons. Proteomics 2003;3:2310–6. [11] Adam BL, Qu Y, Davis JW, et al. Serum protein fingerprinting coupled with a pattern-matching algorithm distinguishes prostate cancer from benign prostate hyperplasia and healthy men. Cancer Res 2002;62: 3609–14. [12] Rogers MA, Clarke P, Noble J, et al. Proteomic profiling of urinary proteins in renal cancer by surface enhanced laser desorption ionization and neural-network analysis: identification of key issues affecting potential clinical utility. Cancer Res 2003;63:6971–83. [13] Goldstein LE, Muffat JA, Cherny RA, et al. Cytosolic beta-amyloid deposition and supranuclear cataracts in lenses from people with Alzheimer's disease. Lancet 2003;361:1258–65. [14] Tomosugi N. Discovery of disease biomarkers by ProteinChip system; clinical proteomics as noninvasive diagnostic tool. Rinsho Byori 2004; 52:973–9. [15] Younossi ZM, Baranova A, Ziegler K, et al. A genomic and proteomic study of the spectrum of nonalcoholic fatty liver disease. Hepatology 2005;42:665–74.
1121
[16] Sakamoto S. In: Hosoya E, Yamaura Y, editors. Recent advances in the pharmacology of Kampo (Japanese herbal) medicines. Amsterdam, MD: Excepta Medica; 1998. p. 170–6. [17] Mori T, Sakamoto S, Singtripop T, et al. Suppression of spontaneous development of uterine adenomyosis by a Chinese herbal medicine, keishibukuryo-gan, in mice. Planta Med 1993;59:308–11. [18] Mochimaru F, Toyama M, Kanakura Y, Inde S. Objective indicator for the assessment of postmenopausal hot flashes. Nippon Sanka Fujinka Gakkai Zasshi 1984;36:643–5. [19] Chen JT, Shiraki M. Menopausal hot flash and calcitonin gene-related peptide; effect of Keishi-bukuryo-gan, a Kampo medicine, related to plasma calcitonin gene-related peptide level. Maturitas 2003;45: 199–204. [20] Nozaki K, Hikiami H, Goto H, Nakagawa T, Shibahara N, Shimada Y. Keishibukuryogan (gui-zhi-fu-ling-wan), a Kampo formula, decreases disease activity and soluble vascular adhesion molecule-1 in patients with rheumatoid arthritis. Evid Based Complement Alternat Med 2006;3: 359–64. [21] Arnett FC, Edworthy SM, Bloch DA, et al. The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum 1988;31:315–24. [22] Terasawa K, Shinoda H, Imadaya A, Tosa H, Bandoh M, Sato N. The presentation of diagnostic criteria of “Yu-xie” (stagnated blood) conformation. Int J Orient Med 1989;14:194–213. [23] Terasawa K, Toriizuka K, Tosa H, Ueno M, Hayashi T, Shimizu M. Rheological studies on “Oketsu” syndrome: I. The blood viscosity and diagnostic criteria. J Med Pharm Soc WAKAN-YAKU 1986;3:98–104. [24] Hikiami H, Kohta K, Sekiya N, Shimada Y, Itoh T, Tosa H. Erythrocyte deformability in “Oketsu” syndrome and its relation to erythrocyte viscoelasticity. J Trad Med 1996;13:156–64. [25] Kohta K, Hiyama Y, Terasawa K, Hamazaki T, Itoh T, Tosa H. Hemorheological studies of “Oketsu” syndrome – erythrocyte aggregation in “Oketsu” syndrome. J Med Pharm Soc WAKAN-YAKU 1992; 9:221–8. [26] Hikiami H, Goto H, Sekiya N, et al. Comparative efficacy of Keishibukuryo-gan and pentoxifylline on RBC deformability in patients with “oketsu” syndrome. Phytomedicine 2003;10:459–66. [27] Sekiya N, Tanaka N, Itoh T, Shimada Y, Goto H, Terasawa K. Keishibukuryo-gan prevents the progression of atherosclerosis in cholesterol-fed rabbit. Phytother Res 1999;13:192–6. [28] Sekiya N, Goto H, Shimada Y, Terasawa K. Inhibitory effects of Keishibukuryo-gan on free radical induced lysis of rat red blood cells. Phytother Res 2002;16:373–6. [29] Langlois MR, Delanghe JR. Biological and clinical significance of haptoglobin polymorphism in humans. Clin Chem 1996;42:1589–600. [30] Levy AP, Friedenberg P, Lotan R, et al. The effect of vitamin therapy on the progression of coronary artery atherosclerosis varies by haptoglobin type in postmenopausal women. Diabetes Care 2004;27:925–30.