Quantitative Analysis of Serum IgG Galactosylation Assists Differential

Jul 15, 2013 - The united glycoform profile was able to enhance the differential diagnosis of ovarian cancer with a specificity of 61.1% at 90% sensit...
0 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/jpr

Quantitative Analysis of Serum IgG Galactosylation Assists Differential Diagnosis of Ovarian Cancer Yifan Qian,†,⊥ Yisheng Wang,#,‡,⊥ Xingwang Zhang,† Lei Zhou,† Zejian, Zhang,† Jiejie Xu, Yuanyuan Ruan, Shifang Ren,*,† Congjian Xu,*,#,‡,§,∥ and Jianxin Gu† †

Key Laboratory of Glycoconjugate Research Ministry of Public Health, Department of Biochemistry and Molecular Biology, School of Basic Medical Sciences, ‡Department of Obstetrics and Gynecology of Shanghai Medical School, and §Institute of Biomedical Sciences, Fudan University, Shanghai, P. R. China # Obstetrics and Gynecology Hospital, Fudan University, Shanghai, P. R. China ∥ Shanghai Key Laboratory of Female Reproductive Endocrine Related Diseases, Shanghai, P. R. China S Supporting Information *

ABSTRACT: CA-125, the most frequently used biomarker for ovarian cancer detection, cannot provide accurate diagnosis due to its poor specificity as it may also increase in many benign gynecological conditions. Thus, reducing the false-positive outcomes is urgently needed. Decrease in terminal galactosylated N-glycans of serum IgG has been found in various malignancies compared to healthy controls. Here, this alteration of IgG galactosylation was extended to be investigated between ovarian cancer and benign conditions with similar elevated CA-125 levels, in an attempt to effectively distinguish between false-positive subjects and ovarian cancer patients. In the study of 58 patients with elevated CA-125 levels (>35 U/ mL), the degree of IgG galactosylation was measured from the relative intensities of IgG digalactosyl (G2), monogalactosyl (G1), and agalactosylated (G0) N-glycans according to the formula G0/(G1 + G2·2). This ratio was found significantly higher in the malignant group than in the benign group (0.74 vs 0.34; p < 0.0001). ROC analysis demonstrated an improved specificity from 65.2% (by CA-125 test alone) to 84.6%, while maintaining sensitivity at 90% by incorporating quantitative analysis of IgG galactosylation in the current assay. The results suggest that combining quantitative alteration of IgG galactosylation with CA-125 may generate an overall more robust approach for differential diagnosis of ovarian cancer. KEYWORDS: ovarian cancer, differential diagnosis, false-positive rate, galactosylation, IgG

■

INTRODUCTION Ovarian cancer, known as a highly lethal gynecological malignancy, is the fifth leading cause of cancer deaths among women and accounts for more than 140,000 deaths annually in females worldwide.1,2 Cancer antigen 125 (CA-125) is currently the most robust FDA-approved serum biomarker for routine ovarian cancer detection.3 However, many benign conditions may also cause elevated CA-125, such as endometriosis, pelvic inflammatory disease, menses, and pregnancy,4 which lead to low specificity of CA-125. This insufficient specificity can cause a high rate of false-positive results and limits the suitability of CA-125 for a requisite diagnosis of ovarian cancer.5,6 It is a problem especially when CA-125 level are moderately elevated in the range of 30−500 U/mL as higher levels strongly correlate with malignancy. High false-positive results are risky because they may lead to invasive follow-up testing, inappropriate management strategy, financial burden, and anxiety.7 Of note, it has been reported that over two-thirds of the patients who undergo surgery for suspected ovarian neoplasm do not have cancer.8 The unnecessary © 2013 American Chemical Society

surgical treatments should be avoided as far as possible because after extensive surgery not only is long rehabilitation time inevitable, but the life quality of patients may be affected as well.9 Accordingly, there is an urgent need to develop effective methods to reduce false-positive results of CA-125 test and improve the accuracy of preoperative assessment. Glycosylation is one of the most common post-translational modifications, and deregulation of glycosylation has been reported associated with a wide range of diseases including human carcinomas for many years.10,11 Analysis of glycan alteration in cancer serums or tissues can lead to the development of improved diagnostic methods.12 Thus, the combination of current biomarkers with glyco-markers to provide a further improvement is proposed as an important future goal.13 Indeed, a large number of glycomics-based4,14−18 studies have focused on the improvement in diagnostic accuracy of ovarian cancer, while the effect of N-glycan Received: April 28, 2013 Published: July 15, 2013 4046

dx.doi.org/10.1021/pr4003992 | J. Proteome Res. 2013, 12, 4046−4055

Journal of Proteome Research

Article

matched benign gynecological condition patients (median age 45.0 years old, range 40−51), respectively. Notably, the ovarian cancer and benign gynecological condition patients were not only in the same age range to minimize the age-dependence of IgG galactosylation,19,26,27 but all had elevated CA-125 levels (>35 U/mL) as well. After the blood was allowed to clot for 30 min at ambient temperature, the serum layer was collected, centrifuged at 2000 × g for 10 min, aliquoted, and stored at −80 °C until analysis. The protocol for the present study was approved by the Institutional Review Board of Obstetrics and Gynecology Hospital of Fudan University and informed consents from all patients were acquired.

alteration as a complement to current biomarkers on reducing the false-positive rate has been rarely reported to date. In this study, we attempted to minimize the false-positive rate of ovarian cancer diagnosis by incorporating the quantitative alteration of IgG galactosylation as a supplement to CA-125 test for the first time. Aberrant glycosylation of IgG has been found to be associated with disease pathogenesis in many studies.19 For instance, a decline in galactosylated IgG glycoforms has been reported in tumor progression and metastasis in prostate,20 gastric,21,22 lung23,24 and ovarian cancer.14,16,17 In the case of ovarian cancer, Gercel-Talor et al. first depicted the presence of an aberrantly glycosylated IgG in the circulation of patients with ovarian cancer.17 Later, Saldova et al. demonstrated a trend toward decreasing levels of galactosylation in the serum IgG obtained from three ovarian cancer patients when compared to a pooled control sample, indicating an association between the alteration of IgG galacto-glycoforms in sera and ovarian cancer progression.16 Recently, as a validation, Alley et al. observed similar growth in the agalactosylation level of IgGderived glycans with a larger sample size of 19 ovarian cancer patients and 20 healthy controls.14 All of this research demonstrated quantitative alterations of IgG galactosylation between ovarian cancer patients and healthy controls. In this study, we first focused on the quantitative change of IgG galactosylation between patients with ovarian cancer and benign gynecological conditions when all had elevated CA-125 levels, in an attempt to investigate whether this glycosylation difference could be utilized to the assist routinely used CA-125 test. If the quantitative alteration in serum IgG galactosylation is statistically significant between two patient groups, the combination of the galacto-marker with the current proteinbiomarker, CA-125, will benefit in differentiating cancer from those benign conditions and thereby reduce the false-positive response caused by the utilization of serum CA-125 alone, promising a further improvements in the specificities of practical ovarian cancer detection. Accordingly, the aims of this study were to (i) identify whether there existed significant difference in IgG galactosylation alteration between two subject sets with ovarian cancer and benign conditions, all of which had elevated CA-125 levels, and (ii) determine whether this quantitative alteration in IgG galactosylation had the potential utility as an adjunct marker to CA-125 for differential diagnosis to accurately discriminate between benign conditions and ovarian cancer by statistical analysis. The methodology employed to address these questions included Protein A affinity purification of serum IgG, glycosidase degradation, solid-phase extraction (SPE) with porous graphic carbon (PGC), permethylation, and analysis by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) at glycan level. The main advantages of this procedure are simplicity, together with specificity, sensitivity, and accuracy, which are exactly in line with the most valued attributes required for potential clinical diagnostic methods.25

■

IgG Purification

The Protein A IgG Purification Kit (Thermo Fisher Scientific, Rockford, IL) was utilized in this study to isolate and purify IgG from blood serum sample. According to the manufacturer instructions, 100 μL of blood serum were first diluted to 5 mL using a proprietary Protein A IgG Binding Buffer to facilitate binding and obtain good IgG recovery from blood serum. To equilibrate the Protein A column, 5 mL of IgG Binding Buffer were applied and allowed to drain through the column. Then, 5 mL of diluted sample were applied and allowed to flow through the equilibrated column. Subsequently, 10 mL of Binding Buffer were added to the column to thoroughly wash away all unbound non-IgG protein components. Last, the bound IgG were eluted with 10 mL of the proprietary IgG Elution Buffer and separate 1 mL fractions of the elution were collected. To determine which fractions contained IgG, the absorbance of each fraction was measured at 280 nm by a bicinchoninic acid (BCA) test (Thermo Fisher Scientific, Rockford, IL). The purity of eluted IgG was further validated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) under reducing conditions and the fractions containing IgG were stored at −20 °C until the releasing of oligosaccharides. IgG N-Glycan Release and PGC Solid-Phase Extraction

One hundred microliter portions of IgG-containing fractions of Protein A column eluate were denatured in a boiling water bath for 5 min. The IgG N-glycans were then released by incubating with PNGase F (New England Biolabs, Inc., USA) for 12 h at 37 °C. The solution of PNGase F released oligosaccharides were subsequently purified by solid-phase extraction (SPE) using an Eppendorf dualfilter tip (10 μL) packed with porous graphic carbon (PGC) powder as described previously.28,29 Briefly, the PGC dualfilter tip was washed for 3 times with 100 μL of 0.1% (v/v) trifluoroacetic acid (TFA) in 80% acetonitrile (ACN)/ H2O (v/v) and followed by 0.1% (v/v) TFA in H2O. The solution of released oligosaccharides was applied to the PGC dualfilter tip repeatedly 3 times to allow complete N-glycan adsorption. Then, the dualfilter tip was washed 3 times with 100 μL of H2O to remove salts and buffer. The N-glycans derived from IgG were eluted with 100 μL of 0.05% (v/v) TFA in 25% ACN/H2O (v/v) and lyophilized for permethylation or direct MS analysis. IgG N-Glycan Permethylation

MATERIALS AND METHODS

According to the previously published protocol,30 freeze-dried N-glycans released from serum IgG were permethylated in a glass tube. Briefly, five pellets of NaOH were pulverized into a fine powder using a dry mortar as quickly as possible to minimize absorption of moisture from the atmosphere. Approximately 20 mg of NaOH powder was added to the

Serum Samples

A total of 58 untreated serum samples, diagnosed at the Obstetrics and Gynecology Hospital of Fudan University, China, were obtained preoperatively from 32 ovarian cancer patients (median age 50.5 years old, range 40−63) and 26 age4047

dx.doi.org/10.1021/pr4003992 | J. Proteome Res. 2013, 12, 4046−4055

Journal of Proteome Research

Article

Figure 1. Typical MALDI-QIT-TOF MS spectra of serum IgG N-glycan profiles acquired for (A) a benign gynecological condition patient and (B) an ovarian cancer patient. The representative MALDI MS spectra showed pronounced differences in the relative intensities of the three most significant N-glycans (G0, G1, and G2) in the highlighted areas.

sample with 500 μL of dimethyl sulfoxide (DMSO) and 100 μL of iodomethane. The reaction slurry was mixed vigorously and then placed in an ultrasonic bath for 30 min at room temperature. An additional 50 μL of iodomethane was added to the reaction mixture for further 30 min to ensure the permethylation. To quench the reaction, 1 mL of H2O was added and mixed vigorously. Subsequently, 600 μL of chloroform was added to allow the mixture, which was allowed to settle into two layers. The aqueous phase was discarded, and the lower chloroform phase was washed 8 times with 1 mL of H2O and dried down under a stream of nitrogen in a hood for MALDI MS analysis.

TOF has been described in detail elsewhere.31 The m/z range was monitored to span from 500 to 3500. Two laser shots were set to generate a profile, and 200 profiles were accumulated from different points of laser irradiation into one file for each sample spot. Tandem mass spectrometry (MS/MS) was utilized to validate the presence of the expected G2, G1, and G0 glycans, referring to the biantennary, core-fucosylated structures carrying two, one, and no galactose residues, respectively (data not shown). The GlycoWorkbench software was used for the annotation of MS spectra. Data Processing and Statistical Analysis

The MALDI MS spectra data were exported, extracted using Mascot Distiller, and then transferred to Microsoft Excel as text files. The following statistical analyses were performed with GraphPad Prism 5. Relative quantification of the three most significant IgG N-glycans (G0, G1, and G2) (Figure 1) using MALDI-TOF MS was based on the signal strength measured by monoisotopic peak height.32−35 Generally, permethyltated glycans were mostly observed as sodium adducts.36 If multiple cation adduct signals such as [M + Na]+ and [M + K]+ were present, those related adducts should be included in the quantification. Of note, triplicate MALDI MS analyses were performed for each sample, and each sample was spotted in triplicate. Thus, a total of nine spectra for each sample were averaged, and then subjected to calculation of the degree of galactosylation according to the formula G0/(G1 + G2·2). The ratios of G0/(G1 + G2·2) were then evaluated by performing t test, and p values of ≤0.05 were considered statistically significant. The data were further processed by receiveroperator characteristics (ROC) test to assess the specificity and sensitivity of the potential diagnostic variable, and resulting area-under-the-curve (AUC) values of ≥0.9 were considered “highly accurate”.

MALDI MS of N-Glycans

Prior to MALDI MS analysis, TOFMix (LaserBio Laboratories, France) containing an eight-peptide calibration standard was used for external calibration of MS. The dried N-glycans were resuspended in 50 μL of 50% (v/v) methanol in H2O. MALDI samples (1 μL) were deposited onto a standard MALDI plate and allowed to dry in air at ambient temperature. Then, 1 μL of matrix solution, 10 mg/mL 2,5-dihydroxybenzoic acid (DHB, Sigma-Aldrich, Germany) in 0.1% (v/v) TFA in 50% ACN/ H2O (v/v), was added onto the sample layer and allowed to dry under ambient conditions. Of note, 0.3 μL of ethanol was subsequently pipetted onto the dried spot to recrystallize the sample for a uniform layer of thin, fine crystals, which can help the homogeneity of the spot surface, making it easier to find a good signal. Each sample was spotted in triplicate. The samples were interrogated automatically in a “batch mode” by AXIMA Resonance MALDI-QIT-TOF MS (Shimadzu Corp. JP) equipped with a 337 nm nitrogen laser in reflector positive ionization mode. The laser power was set as low as 125 to minimize the “in-source decay” (ISD) with satisfied signal-tonoise ratios. The engineering design and the operation of QIT4048

dx.doi.org/10.1021/pr4003992 | J. Proteome Res. 2013, 12, 4046−4055

Journal of Proteome Research

Article

Table 1. Characteristics of Histological Subtypes amongst Patients (n = 58) with Benign Gynecological Diseases and Primary Invasive Ovarian/Tubal Cancer, Distribution of Age, and CA-125 Values histological subtype benign diseases endometriosis/adenomyosis ovarian abscess PID leiomyoma other ovarian/tubal cancer serous endometrioid mucinous clear cell fallopian tube cancer other

■

no. of patients

age median (25−75% percentile)

CA-125 (U/mL) median (25−75% percentile)

26 19 1 2 2 2 32 19 1 1 2 5 4

45.0 (42.0−48.0)

107.9 (79.2−182.8)

50.5 (46.3−54.8)

209.3 (136.2−600.0)

RESULTS

48.0) and 50.5 (25−75% percentile: 46.3−54.8) in patients with benign conditions and ovarian cancer, respectively (Table 1).

Selection Criteria of Patients

Serum from patients with ovarian cancer (n = 32) and benign gynecological conditions (n = 26) were collected preoperatively for IgG glycan analysis (Table 1). All patients were chosen primarily according to a CA-125 cutpoint of 35 U/mL, a commonly reported reference value that designates a clinically positive screening test, among which patients with moderately elevated CA-125 levels in the range of 30−500 U/mL are especially difficult to be diagnosed because higher levels strongly correlated with malignancy.18 The 32 patients with ovarian cancer were identified with serum CA-125 level between 38.8 and 2362 U/mL with a median value of 209.3 (25−75% percentile: 136.2−600.0), and the 26 patients with benign gynecological conditions between 44.6 and 2526 U/mL with a median value of 107.9 (25−75% percentile: 79.2−182.8) (Table 1). In addition, the Stage and Grade of the ovarian cancer patients are presented in Table 2. In this study,

IgG N-Glycan Profiling by MALDI MS

In this study, MALDI-QIT-TOF MS was employed for IgG Nglycan profling due to its speed, high sensitivity, and ease of use. The experimental procedure is shown in Chart 1. Specially, this Chart 1. Quantitative Procedure of Alterations in Serum IgG Galactosylation between Ovarian Cancer and Benign Gynecological Condition Patients

Table 2. Characteristics of Stage and Grade Distributions in Patients (n = 32) with Primary Invasive Ovarian/Tubal Cancer no. of patients Stage I II III IV not available

1 6 15 0 10

1 2 3 not available

4 1 8 19

study was designed to analyze three specific IgG glycoforms (G0, G1, and G2, Figure 1) rather than all released glycans (Supplementary Figure S1, Supporting Information) in order to simplify data processing.20 Subsequently, the relative intensities of IgG G0, G1, and G2 were integrated into one ratio index according to our algorithm for the change in IgG galactosylation to act as a candidate of adjunct glyco-marker to CA-125. We did not directly employ G0, G1, and G2 as three separate N-glycan indicators to evaluate their united abilities in supplement diagnosis of ovarian cancer because a combination of higher numbers of markers may sacrifice more coefficient of variation (CV) errors and thus generate less accurate and reliable diagnostic results due to the overfitting effect mentioned in a previous study.8 As shown in Figure 1, an obvious difference in the relative intensities of IgG G0, G1, and G2 in highlighted areas could be easily found by comparing the glycomic profiles of two patient groups visually. Thereby, the

Grade

detection of the degree of IgG galactosylation, calculated from the relative intensity ratios of IgG agalactosylated (G0), monogalactosyl (G1), and digalactosyl (G2) N-glycans (Figure 1), was applied for aiding differential diagnosis of ovarian cancer when all patients were with elevated CA-125 levels. In addition, these two patients groups were also age-matched in our work in order to minimize the contribution of age to the IgG galactosylation discrepancy between two patient groups.37 The median patient age was 45.0 (25−75% percentile: 42.0− 4049

dx.doi.org/10.1021/pr4003992 | J. Proteome Res. 2013, 12, 4046−4055

Journal of Proteome Research

Article

gynecological conditions compared with those with ovarian cancer (p < 0.0001, Figure 2). Obviously, these ratios were more than doubled in the malignant group when compared to the benign group. Furthermore, two other formulas, G0/(G1 + G2) and G0/(G1·0.5 + G2), were evaluated as well, which produced almost the same results as the G0/(G1 + G2·2) ratios (Supplementary Figure S2 and Table S1, Supporting Information). However, since the relative intensities of G1 and G2 were not normalized in the formula G0/(G1 + G2) and the individual variations (patient-to-patient) in ratios of G0/ (G1·0.5+G2) were a little bit larger, we finally identified G0/ (G1 + G2·2) as the most reliable and valid algorithm for the degree of IgG galactosylation. In addition, it has long been reported that IgG galactosylation decrease with increasing age,10,42 which indicates that change in IgG galactosylation is age-dependent. Considering that the malignant patients (median age: 50.5, 40− 63) covered a wider age range than the benign controls (median age: 45.0, 40−51), whether the partially matched age would contribute to the significant difference in IgG galactosylation between two patient groups needed evaluation. Thus, 14 ovarian cancer patients of age 52−63 were excluded temporarily, and the corresponding statistic analyses were performed separately between 26 benign controls and 18 malignant patients all of age 40−51. Noticeably, a significant difference (0.34 vs 0.62; p < 0.0001) was still observed between benign controls and their malignant counterparts in the same age range from 40 to 51 (Supplementary Figure S3, Supporting Information). Moreover, the G0/(G1 + G2·2) ratios of ovarian cancer patients were consistently higher than the corresponding ratios of benign controls of the same age (Supplementary Figure S4, Supporting Information), which indicated that the influence of age-dependence of IgG galactosylation was minimal and would not partly contribute to the considerable discrepancy of the G0/(G1 + G2·2) ratios between two patient groups.

degree of IgG galactosylation was further determined on the basis of normalized intensities of these three glycoforms according to the formula G0/(G1 + G2·2). In the differential measurements of G0, G1, and G2 data for two patient groups, it was essential to achieve a high degree of methodological precision, so as to (i) ensure that the individual variations (patient-to-patient) were not overshadowed by the measurement errors and (ii) fulfill the demand for intraday repeatability for the entire profiling procedure. Considering that the precision of relative quantitation based on MALDI MS might be affected by the matrices selection and adjustment of the MS settings,33,38,39 the “classical” DHB with ethanol recrystallization was chosen as the more versatile matrix, and all samples were interrogated automatically in a “batch mode” by MALDI MS as sodium was not uniformly distributed on DHB spots,40,41 which might influence the laser shot location if the spectra were acquired manually and thereby affect the accuracy of relative quantitation. Fortunately, our quantitative reproducibility yielded acceptable coefficients of variation below 5% (4.6% ± 2.8%) on running repeated biological samples, making it possible to dependably evaluate N-glycan profiles from sera of two patient groups. Statistical Analysis of IgG Galactosylation Change

To identify whether there existed significant difference in IgG galactosylation between two patient groups with ovarian cancer and benign gynecological conditions, we carried out an extensive study into the ratios calculated on the formula G0/ (G1 + G2·2), and the corresponding G0, G1, and G2 relative intensity data sets were obtained in total 58 serum samples from two patient groups (Figure 2). Based on our results, the

Diagnosis Evaluation of IgG Galactosylation Analysis

As shown in Figure 3, ROC analysis was further conducted for the G0/(G1 + G2·2) ratios of IgG galactosylation in order to

Figure 2. Comparison of the G0/(G1 + G2·2) ratios for the change of IgG galactosylation in serum samples obtained from benign gynecological condition controls (n = 26) and ovarian cancer patients (n = 32) by t test. Error bars represent SEM of each data set. The summary for p value of