Circulating cell-free DNA as a potential biomarker for minimal and mild endometriosis

Circulating cell-free DNA as a potential biomarker for minimal and mild endometriosis

RBMOnline - Vol 18 No 3. 2009 407-411 Reproductive BioMedicine Online; www.rbmonline.com/Article/3577 on web 27 January 2009 Article Circulating cell...

354KB Sizes 0 Downloads 42 Views

RBMOnline - Vol 18 No 3. 2009 407-411 Reproductive BioMedicine Online; www.rbmonline.com/Article/3577 on web 27 January 2009

Article Circulating cell-free DNA as a potential biomarker for minimal and mild endometriosis Professor Xiao Yan Zhong was born and studied medicine in China. Before she moved to Europe for her research activity, she was appointed as an Associate Professor and vicechief physician at University of Tongji. She is focusing on risk-free non-invasive diagnosis for prenatal medicine and gynaecological cancers using tissue-derived genetic materials in circulation under the leadership of Professor Wolfgang Holzgreve at University Medical Center, Freiburg, Germany. Dr Zhong was appointed Professor in Molecular Reproduction at the University of Basel and is the chief of Prenatal Medicine and Gynaecological Oncology.

Professor Xiao Yan Zhong Rebecca Zachariah1, Seraina Schmid2, Ramin Radpour1, Nicole Buerki3, Alex Xiu-Cheng Fan1, Sinuhe Hahn1, Wolfgang Holzgreve4, Xiao Yan Zhong1,5 1 Laboratory for Prenatal Medicine and Gynecologic Oncology, Women’s Hospital, Department of Biomedicine, University of Basel, Switzerland; 2Department of Obstetrics and Gynecology, Women’s Hospital, University of Basel, Switzerland; 3Department of Obstetrics and Gynecology, Kantonsspital, Liestal, Switzerland; 4 University Medical Center, Freiburg, Germany 5 Correspondence: Tel: +41 61 265 9224, 9595; Fax: +41 61 265 9399; e-mail: [email protected]

Abstract It has recently been reported that high concentrations of circulating cell-free (ccf) nucleic acids in plasma and serum could be used as biomarkers for non-invasive monitoring a wide variety of malignant and benign proliferations and inflammatory conditions. Endometriosis is one of the most common benign gynaecological proliferations with inflammatory activation in premenopausal women. Real-time multiplex polymerase chain reaction was used for synchronized quantification of the glyceraldehyde-3-phosphate dehydrogenase gene sequence in nuclear DNA (nDNA) and the ATP synthase-8 gene sequence in mitochondrial DNA (mtDNA). DNA was extracted from 500 µl serum and plasma of 19 cases with endometriosis to measure the total amount of ccf nDNA and ccf mtDNA. The concentration of ccf nDNA in plasma was significantly higher in the endometriosis group than in the control group (P = 0.046). The cut-off value selected by a receiver operating characteristic curve could provide a sensitivity of 70% and a specificity of 87% to discriminate between the minimal or mild cases and normal controls. The finding of significantly increased concentrations of ccf nDNA in plasma of patients with endometriosis suggests that ccf nDNA might be a potential biomarker for developing non-invasive diagnostic test in endometriosis. Keywords: cell-free DNA, circulating DNA, endometriosis, mtDNA, nDNA

Introduction Endometriosis, defined as the presence of endometrial tissue outside of the uterus, is one of the most common benign gynaecological proliferations in premenopausal women. It contains features similar to those found in malignancies, e.g. progressive growth, invasive growth, oestrogen-dependent growth, recurrence and a tendency to metastasize (Van Gorp et al., 2004; Flores et al., 2007). Patients with endometriosis often have dysfunctional uterine bleeding, infertility, chronic pelvic pain (CPP) or pelvic pain and dysmenorrhoea (D’Hooghe and Hummelshoj, 2006; Gao et al., 2006). The biology of endometriosis is unclear (Tomassetti et al., 2006; Flores et al., 2007). Endometriosis is classified in four stages based on the severity, location, amount, depth and size of growths. They are stage I

(minimal disease), stage II (mild disease), stage III (moderate disease) and stage VI (severe disease) (Kennedy et al., 2005). Laparoscopy is the standard technique for visual inspection of the pelvis, and for establishing a definitive diagnosis, as well as for simultaneous treatment (Kennedy et al., 2005). Additional tools are needed for non-invasive classifications in order to reduce the number of unnecessary laparoscopies without adversely affecting outcomes according to patients wishes, especially the wishes of patients with minimal or mild disease (D’Hooghe et al., 2006; Kennedy, 2006). Concentrations of CA125 could be significantly higher in women with moderate or severe endometriosis, and could be normal in women with minimal or mild disease (Riley et al., 2007). The lack of sensitivity and specificity regarding a cut-off value for this

© 2009 Published by Reproductive Healthcare Ltd, Duck End Farm, Dry Drayton, Cambridge CB23 8DB, UK

407

Article - Circulating cell-free DNA in endometriosis - R Zachariah et al. marker limits its diagnostic and clinical application (Rosa et al., 2007). Compared with laparoscopy, measurement of serum CA125 concentrations has no value as a diagnostic tool, and there is, therefore, no blood test available for the diagnosis of endometriosis (Panel and Renouvel, 2007).

ultrasound, were suggested for a further laparoscopy. Before the invasive procedure, blood samples were taken from the patients. Patients with co-diagnosis, such as chronic inflammatory disease and/or cancer and/or under any therapy were excluded from this study.

The discovery of circulating cell-free (ccf) DNA in circulation has opened up the possibilities of non-invasive diagnosis and monitoring of a wide variety of malignant diseases. Increased concentrations of ccf DNA have also been found in inflammatory conditions, such as systemic lupus erythematosus and rheumatoid arthritis (Galeazzi et al., 2003; Zhong et al., 2007a). Additionally, there have been several recent studies demonstrating the existence of another species of circulating nucleic acids, i.e. mitochondrial DNA (mtDNA). Both ccf nDNA and ccf mtDNA in circulation have been found to be elevated in trauma (Lo et al., 2000; Lam et al., 2004), suggesting that cell death is the source of ccf DNA, including the proportion of histone-protein bound molecules and unbound part molecules (Seefeld et al., 2008).

The control group was composed of women of reproductive age reporting for annual checkup in the study hospital who agreed to have blood sampling but had neither symptoms nor clinical signs for endometriosis or other pelvic pathologies, no diagnosed chronic inflammatory disease and/or cancers and/ or under any therapy. No laparoscopy was performed for the controls.

Finding specific and more sensitive biomarkers in endometriosis is critical, because endometriosis is usually diagnosed only in advanced stages, and there is a high rate of morbidity for this disease. Recently, Gajbhiye et al. (2008) have found antiendometrial antibodies in sera of endometriosis in common with human autoimmunity, suggesting that endometriosis may correlate with autoimmune inflammation, and the targets may serve as markers for developing non-invasive diagnostic tests. Apoptosis has also been observed in eutopic and ectopic endometria of patients with endometriosis (Harada et al., 2007). Based on the observations of autoimmune reaction and apoptotic events in endometriosis, it is proposed here that high concentrations of ccf DNA will be seen in endometriosis. The present study developed a simple and accurate multiplex realtime polymerase chain reaction (PCR) method for synchronized quantification of ccf nDNA and ccf mtDNA in plasma and serum of patients with endometriosis. ccf DNA in plasma is regarded as a natural release of the species, and the main part of ccf DNA in serum is considered as an unnatural release of the species during blood clotting procedures. The concentrations of serum ccf DNA also reflects the fragility of blood cells (Zhong et al., 2007b). The detection is based on a single-step real-time PCR, using FAM- and VIC-labelled probes for determining selected mtDNA and nDNA in the area of interest. Through the use of this new method, total ccf nDNA and ccf mtDNA in serum and plasma of patients with endometriosis was measured and quantified.

Materials and methods Patients A retrospective study was undertaken at the University Women’s Hospital, Department of BioMedicine, University of Basel. The study was approved by the local institutional review board. A total of 34 women were recruited for this study within 1 year. Written consent forms were collected from all individuals involved in this study.

408

In the case group, patients who were suspected of a diagnosis of endometriosis with chronic pelvic pain, as well as having a positive clinical examination and positive detection on

Sample preparation and DNA extraction The 7.5 ml of peripheral blood samples for coagulant serum and 7.5 ml of peripheral blood samples for EDTA plasma were sent to the laboratory for Prenatal Medicine and Gynaecologic Oncology, University of Basel. The blood samples were processed immediately (within 6 hours) by centrifugation at 1600 g for 10 min. Plasma and serum layers were transferred to new Eppendorf tubes, centrifuged once more for 10 min and immediately stored at –80°C (Zhong et al., 2007a). DNA was extracted from 500 µl of serum and plasma from 34 cases using TM the method with the automated MagNA Pure LC Instrument and MagNA Pure LC DNA Isolation Kit (large volume) (Roche Applied Science, Switzerland), and eluted into 100 µl elution buffer. The technical details of the DNA extraction are described elsewhere (Huang et al., 2008). Samples which were thawed once, not prepared according to the standard protocol in the study laboratory, or had preparation postponed, were excluded from the study. DNA samples processed using other commercial kits were not used in this study.

Real-time PCR Multiplex real-time PCR was performed using ABI Prism 7000 Sequence Detector (Applied Biosystems, ABI, USA). As a single-step real-time PCR, FAM- and VIC-labelled probes were used for determining selected mtDNA and nDNA area of interest. For analysing ccf nDNA, the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene sequence was amplified with forward 5′-CCCCACACACATGCACTTACG-3′ and reverse 5′-CCTAGTCCCAGGGCTTTGATT-3′ primers and with 5′-MGB-GTGAACGTGGATGAAGTTGG–VIC-3′ as a probe. For testing ccf mtDNA, a sequence of the MTATP-8 gene starting at locus 8446 was amplified with forward 5′-AATATTAAACACAAACTACCACCTACC-3′ and reverse 5′-TGGTTCTCAGGGTTTGTTATA-3′ primers and with 5′-6-FAM-CCTCACCAAAGCCCATA-MGB-3′ as a probe. Multiplex real-time amplification reaction was set up as a volume of 25 µl containing 5 µl of DNA, 12.5 µl of TaqMan® Universal PCR Master Mix, 0.3 µmol/l of each GAPDH primer and 0.1 µmol/l of GAPDH probe. The optimal concentrations of the mitochondrial primers and probe for duplex real-time PCR were 0.6 µmol/l for each primer and 0.4 µmol/l for the probe. The real-time quantitative PCR were performed using a 2 min incubation at 50°C, followed by an initial denaturation step at 95°C for 10 min and 40 cycles of 1 min at 60°C and 15 s at 95°C. The positive reaction was detected by accumulation of a RBMOnline®

Article -Circulating cell-free DNA in endometriosis - R Zachariah et al. fluorescent signal. The cycles required for the fluorescent signal to cross the threshold were defined as cycle threshold (CT). CT values of GAPDH assay and MTATP-8 assay can be simply used for the quantitative comparison between the two study groups. The CT values can also be converted into quantities according to standard curves generated by dilution of either high-performance liquid chromatography-purified single-stranded synthetic DNA oligonucleotides (Microsynth, Switzerland) specifying a 79-base pair mtDNA amplicon and a 97-base pair GAPDH amplicon, or human genomic DNA with a known concentration. For this study, a serial five-fold dilution with six concentration points of a genomic DNA sample with a known concentration measured by NanoDrop® ND-1000 Spectrophotometer was built for generating the standard curves and analysed in parallel with each assay (Zhong et al., 2007b). The efficiency of the calibration curve remained stable for more than 95% for all PCR runs. The results of ccf nDNA in plasma or serum were expressed as genome equivalent/ml. Amplification data were analysed and stored by the Sequence Detection System Software (v1.2.3; Applied Biosystems). The concentration of ccf mtDNA was obtained by calculating the ΔCT of mtDNA and nDNA (CTGAPDH – CTMTATP) in the same well, and used as exponent of two (2ΔCT).

Statistical analysis The data were analysed using Statistics Package for Social Sciences (SPSS) software (Statistical Software Package for Windows v. 15.0, SPSS, USA). The Shapiro–Wilk test showed that the data set was not normally distributed. Therefore, nonparametric tests were chosen for the analysis. The Mann– Whitney U-test was used to compare the concentrations of ccf nDNA and ccf mtDNA in patient and control groups. The Spearman rank test was applied to analyse the relationship between concentrations of ccf nDNA and quantities of ccf mtDNA. A receiver operating characteristic (ROC) curve was used to determine a cut-off point of ccf nDNA concentration for distinguishing between affected and non-affected individuals.

Results A total of 34 paired serum and plasma samples from 19 patients and 15 controls with the mean age of 36 years (range 21–50) were analysed by synchronized multiplex real-time PCR. All diagnoses for endometriosis were histopathologically

confirmed. The stage of the disease was obtained from all 19 cases (stage I = 5 cases, stage II = 8 cases, stage III = 0 and stage IV = 6 cases). ccf nDNA and ccf mtDNA were analysed in plasma and serum samples from those individuals. The concentration of ccf mtDNA was significantly higher than that of the ccf nDNA (Mann–Whitney U-test, P < 0.001 for serum and plasma samples) (Table 1). No correlation between ccf nDNA and ccf mtDNA was observed using the Spearman rank test. The ccf nDNA and ccf mtDNA concentrations in serum samples were significantly higher than those in plasma samples in the two study groups. Also no correlation was found between the concentrations of ccf DNA in plasma and the concentrations of ccf DNA in serum. The concentrations of ccf nDNA and ccf mtDNA were compared between the control group and the endometriosis group (Table 1). ccf nDNA concentration was significantly increased in the endometriosis group compared with that in the control group (660 genome equivalent/ml versus 285 genome equivalent/ml, P = 0.046. When the outliers from the two groups were excluded, P = 0.039 (Figure 1). The ccf nDNA concentration in the minimal or mild endometriosis subgroup were compared with that in the healthy controls (13 versus 15). A significantly increased concentration was found in the former group (Mann–Whitney U-test, P = 0.018). Using a ROC curve, the concentration of the ccf nDNA for discriminating minimal or mild endometriosis and healthy individuals was evaluated. Figure 2 shows graphics with the sensitivity plotted on the y coordinate versus 1 – specificity or the false positive rate plotted on the x coordinate. The area under the curve for the endometriosis group was 0.76 (95% confidence interval, 0.58– 0.95). An optimal cut-off value of a mean of 416 ccf nDNA genome equivalent/ml selected by the ROC provided a sensitivity of 70% and a specificity of 87% to discriminate between the minimal or mild cases and normal controls. Since the specificity in this disease is less important in daily clinical practice (D’Hooghe et al., 2006), alternative cut-off values for providing better sensitivity between 85–77% and less specificity between 33–47% were available. Based on the limited sample size for stage III and IV patients (six cases in total), similar analysis for this subgroup was not performed.

Table 1. Concentrations of circulating cell-free nuclear DNA and ccf mitochondrial DNA in serum and plasma. Sample type

ccf nDNA Mean Range

ccf mtDNA Mean Range

Plasma Serum

660 285 88,575 74,370

454,800 240,245 3,072,290 3,471,590

Endometriosis Control Endometriosis Control

140–2110 90–535 430–357,235 240–352,880

58,695–1,911,595 21,580–559,090 465,055–9,648,285 971,350–12,192,990

Values are genome equivalent/ml. ccf = circulating cell-free; mtDNA – mitochondrial DNA; nDNA = nuclear DNA. RBMOnline®

409

Article - Circulating cell-free DNA in endometriosis - R Zachariah et al.

a

b 2500.00

2000000.00

ccf mtDNA (GE/ml)

ccf nDNA (GE/ml)

2000.00 1500.00 1000.00 500.00 0.00

*

1500000.00

1000000.00

500000.00

0.00 Control

Endometriosis Groups

Control

Endometriosis Groups

Figure 1. Concentrations of plasma circulating cell-free nuclear DNA (a) and ccf mitochondrial DNA (b) in the two study groups. ccf = circulating cell-free; GE = genome equivalent; mtDNA – mitochondrial DNA; nDNA = nuclear DNA.

Currently, the possibility of circulating ccf DNA serving as a reliable apoptotic and necrotic marker for the detection of pathological processes has been raised. Elevated concentrations of ccf DNA in cancers and in inflammatory diseases have been suggested for developing non-invasive diagnosis (Zhong et al., 2001, 2007a). Since endometriosis is related to a chronic inflammatory reaction, this study took the opportunity to analyse the concentrations of ccf DNA in the patients and to investigate whether the circulating species could be potential biomarkers for developing a non-invasive diagnostic test for the disease.

1.0

Sensitivity

0.8

0.6

0.4

0.2

0.0

0.0

0.2

0.4

0.6

0.8

1.0

1-Specificity

Figure 2. A receiver-operating characteristic curve for selecting the cut-off point of circulating cell-free nuclear DNA concentration to distinguish between minimal or mild cases and unaffected controls.

No significant differences in the serum ccf DNA (nDNA and mtDNA) and plasma ccf mtDNA concentrations between the study groups were found.

Discussion

410

The need for a non-invasive diagnostic test for minimal to mild endometriosis with a high sensitivity before invasive laparoscopic investigation and treatments was discussed recently (D’Hooghe et al., 2006). A non-invasive diagnostic test could be developed in circulation or menstrual fluid based on the pathology of endometriosis, such as immune function, cell-death event and, or inflammatory reactions in the patients with endometriosis. Gajbhiye et al. (2008) found serum immunoglobulin G (IgG) and IgM anti-endometrial antibodies presenting in almost 60% of patients with endometriosis, suggesting that the identification would help in developing a non-invasive diagnostic test.

Significantly increased concentrations of ccf nDNA were found in plasma from patients with endometriosis compared with the age-matched controls. It corresponds with the results of different inflammatory diseases presenting elevated ccf nDNA in plasma, which reflects the increased frequency of cell-death events in the conditions (Galeazzi et al., 2003; Zhong et al., 2007a). The clinical relevance of using the plasma ccf nDNA as a marker to distinguish between the minimal or mild disease and healthy controls was evaluated by ROC analysis. A sensitivity of 70% and a specificity of 87% using the ccf DNA assay can be achieved to discriminate between the affected cases and individuals. Since a high sensitivity is more important to reduce false negative cases for endometriosis, more alternative cutoff points can also be obtained using the ROC to improve the sensitivity. This study also analysed ccf serum DNA in the two study groups. Significantly increased cell-free serum DNA were found in the both patient and control groups. However, the concentrations were not elevated in endometriosis. This finding is reinforced by similar research that also found that the amounts of ccf DNA in serum could be several fold higher than those in plasma (Galeazzi et al., 2003; Zhong et al., 2007a,b). Umetani et al. (2006) observed the unequal distribution of DNA during blood sampling. During the clotting process, cell lyses and destruction could contribute to the increased release of ccf DNA in serum. Therefore, plasma ccf DNA seems to be more useful as a biological marker in apoptotic and necrotic cell-death events. Since human cells contain both nDNA and mtDNA, ccf mtDNA was also analysed. No significant difference could be found in RBMOnline®

Article -Circulating cell-free DNA in endometriosis - R Zachariah et al. the ccf mtDNA concentrations in the serum and plasma samples between the two study groups, suggesting that mtDNA might not be a sensitive marker in endometriosis. Several studies also support this finding: while high concentrations of ccf nDNA were observed in many chronic inflammatory conditions (Galeazzi et al., 2003; Zhong et al., 2007a) and acute trauma (Lo et al., 2000), high concentrations of ccf mtDNA were only reported in acute trauma (Lam et al., 2004). This study also could not find any association between ccf nDNA and ccf mtDNA concentrations in circulation. Therefore, it is assumed that mtDNA might have a different releasing mechanism and metabolism in the case of chronic inflammatory disease. In conclusion, these data showed that the concentration of ccf nDNA is significantly elevated in individuals with endometriosis as compared with a healthy control group. Optimal cut-off points for achieving appropriate sensitivity and specificity can be selected to distinguish minimal or mild disease and healthy controls according to the aims of gynaecologists and the wishes of patients. This study suggested that ccf plasma nDNA might be a potential biomarker to develop non-invasive test for endometriosis. However, further research, with a larger sample size, into better understanding the biology and clinical relevance of increased ccf nDNA in the condition is warranted.

Acknowledgements The authors thank Vivian Kiefer for her excellent assistance and support, and Nicole Chiodetti for her help. They are also grateful to Mrs Regan Geissmann for proofreading the text. This work was supported in part by Swiss National Science Foundation, Swiss Cancer League, Krebsliga Beider Basel and Dr Hans Altschueler Stiftung.

References D’Hooghe T, Hummelshoj L 2006 Multi-disciplinary centres, networks of excellence for endometriosis management and research: a proposal. Human Reproduction 21, 2743–2748. D’Hooghe TM, Mihalyi AM, Simsa P et al. 2006 Why we need a noninvasive diagnostic test for minimal to mild endometriosis with a high sensitivity. Gynecologic and Obstetric Investigation 62, 136–138. Flores I, Rivera E, Ruiz LA et al. 2007 Molecular profiling of experimental endometriosis identified gene expression patterns in common with human disease. Fertility and Sterility 87, 1180– 1199. Gajbhiye R, Suryawanshi A, Khan S et al. 2008 Multiple endometrial antigens are targeted in autoimmune endometriosis Reproductive BioMedicine Online 16, 817–824. Galeazzi M, Morozzi G, Piccini M et al. 2003 Dosage and characterization of circulating DNA: present usage and possible applications in systemic autoimmune disorders. Autoimmunity Reviews 2, 50–55. Gao X, Outley J, Botteman M et al. 2006 Economic burden of endometriosis. Fertility and Sterility 86, 1561–1572. Harada T, Taniguchi F, Izawa M et al. 2007 Apoptosis and endometriosis. Frontiers in Bioscience 12, 3140–3151. Huang DJ, Mergenthaler-Gatfield S, Hahn S et al. 2008 Isolation of cell-free DNA from maternal plasma using manual and automated systems. Methods in Molecular Biology 444, 203–208. Kennedy S 2006 Should a diagnosis of endometriosis be sought in all symptomatic women? Fertility and Sterility 86, 1312–1313, discussion 1317. Kennedy S Bergqvist A, Chapron C et al. 2005 ESHRE guideline for the diagnosis and treatment of endometriosis. Human RBMOnline®

Reproduction 20, 2698–2704. Lam NY, Rainer TH, Chiu RW et al. 2004 Plasma mitochondrial DNA concentrations after trauma. Clinical Chemistry 50, 213–216. Lo YM, Rainer TH, Chan LY et al. 2000 Plasma DNA as a prognostic marker in trauma patients. Clinical Chemistry 46, 319–323. Panel P, Renouvel F 2007 [Management of endometriosis: clinical and biological assessment]. Journal de Gynécologie, Obstétrique et Biologie de la Reproduction 36, 119–128. Riley CF, Moen MH, Videm V 2007 Inflammatory markers in endometriosis: reduced peritoneal neutrophil response in minimal endometriosis. Acta obstetricia et gynecologica Scandinavica 86, 877–881. Rosa ESAC, Rosa ESJC, Ferriani RA 2007 Serum CA-125 in the diagnosis of endometriosis. International Journal of Obstetrics and Gynecology 96, 206–207. Seefeld M, Tarhouny S, Fan AX et al. 2008 Parallel assessment of circulatory cell-free DNA by PCR and nucleosomes by ELISA in breast tumors. International Journal of Biological Markers 23, 69–73. Tomassetti C, Meuleman C, Pexsters A et al. 2006 Endometriosis, recurrent miscarriage and implantation failure: is there an immunological link? Reproductive BioMedicine Online 13, 58–64. Umetani N, Hiramatsu S, Hoon DS 2006 Higher amount of free circulating DNA in serum than in plasma is not mainly caused by contaminated extraneous DNA during separation. Annals of the New York Academy of Sciences 1075, 299–307. Van Gorp T, Amant F, Neven P et al. 2004 Endometriosis and the development of malignant tumours of the pelvis. A review of literature. Best Practice and Research. Clinical Obstetrics and Gynaecology 18, 349–371. Zhong XY, von Muhlenen I, Li Y et al. 2007a Increased concentrations of antibody-bound circulatory cell-free DNA in rheumatoid arthritis. Clinical Chemistry 53, 1609–1614. Zhong XY, Hahn S, Kiefer V, Holzgreve W 2007b Is the quantity of circulatory cell-free DNA in human plasma and serum samples associated with gender, age and frequency of blood donations? Annals of Hematology 86, 139–143. Zhong XY, Holzgreve W, Hahn S 2001 Risk free simultaneous prenatal identification of fetal Rhesus D status and sex by multiplex real-time PCR using cell-free fetal DNA in maternal plasma. Swiss Medical Weekly 10, 70–74.

Declaration: The authors report no financial or commercial conflicts of interest. Received 31 March 2008; refereed 30 July 2008; accepted 29 October 2008.

411