Comparison of printed glycan array, suspension array and ELISA in the detection of human anti-glycan antibodies
- First Online:
- Cite this article as:
- Pochechueva, T., Jacob, F., Goldstein, D.R. et al. Glycoconj J (2011) 28: 507. doi:10.1007/s10719-011-9349-y
- 849 Downloads
Anti-glycan antibodies represent a vast and yet insufficiently investigated subpopulation of naturally occurring and adaptive antibodies in humans. Recently, a variety of glycan-based microarrays emerged, allowing high-throughput profiling of a large repertoire of antibodies. As there are no direct approaches for comparison and evaluation of multi-glycan assays we compared three glycan-based immunoassays, namely printed glycan array (PGA), fluorescent microsphere-based suspension array (SA) and ELISA for their efficacy and selectivity in profiling anti-glycan antibodies in a cohort of 48 patients with and without ovarian cancer. The ABO blood group glycan antigens were selected as well recognized ligands for sensitivity and specificity assessments. As another ligand we selected P1, a member of the P blood group system recently identified by PGA as a potential ovarian cancer biomarker. All three glyco-immunoassays reflected the known ABO blood groups with high performance. In contrast, anti-P1 antibody binding profiles displayed much lower concordance. Whilst anti-P1 antibody levels between benign controls and ovarian cancer patients were significantly discriminated using PGA (p = 0.004), we got only similar results using SA (p = 0.03) but not for ELISA. Our findings demonstrate that whilst assays were largely positively correlated, each presents unique characteristic features and should be validated by an independent patient cohort rather than another array technique. The variety between methods presumably reflects the differences in glycan presentation and the antigen/antibody ratio, assay conditions and detection technique. This indicates that the glycan-antibody interaction of interest has to guide the assay selection.
KeywordsGlycan arrayCarbohydrateMultiplex assayOvarian cancer
Area under the curve
- Atri (GalNAcα1-3(Fucα1-2)Galβ-)
- Btri (Galα1-3(Fucα1-2)Galβ-)
Concordance correlation coefficient
Conjugate of glycan with poly[N-(2-hydroxyethyl)acrylamide]
End biotin labeled conjugate of glycan with poly[N-(2-hydroxyethyl)acrylamide]
Enzyme linked immunosorbent assay
Median fluorescent intensity
Median total signal intensities
Phosphate buffered saline
Printed glycan array
Receiver operating characteristics
Tumor associated carbohydrate antigens
Interactions of glycans, namely complex oligosaccharides attached to a protein or lipid mediate many important biological processes, such as pathogen recognition , malignant transformation , autoimmune disease , neurological disorders  and host-versus-graft rejection . Naturally occurring and adaptive anti-glycan antibodies are currently the object of both basic research [6, 7] and biomarker discoveries [8–12].
Ovarian cancer is the sixth most common cause of cancer-related death of women and the leading cause of death from gynaecological malignancies . Despite considerable efforts to improve early detection and advances in chemotherapy, intra-abdominal metastasis remains a major challenge in the clinical management of undifferentiated high FIGO (International Federation of Gynecology and Obstetrics) stage serous ovarian cancers with an overall five-year survival of 20%. This poor overall prognosis is due to multiple factors including a lack of early symptoms and specific screening methods, ineffective therapy for advanced or recurrent disease, and from limited understanding of the early-initiating events and stages of ovarian cancer development. Therefore, there is an urgent clinical need for a reliable, non-invasive and affordable serum-based screening test for ovarian cancers.
In a previous study we used the first custom-made high throughput printed glycan array (PGA, Cellexicon, San Diego, USA) and identified specific anti-glycan antibody patterns in blood of ovarian cancer patients compared to healthy controls . Whilst it is known that aberrant glycosylations are found in most cancers, only a limited number of human antibodies towards tumor associated carbohydrate antigens (TACA) have been evaluated for their potential significance in cancer biology. The printed glycan array [6, 14] consists of a library of over 200 glycans, including well known TACA. Linear modelling revealed 24 carbohydrate structures for which the amount of anti-glycan antibodies was significantly lower in the above mentioned non-mucinous ovarian borderline and cancer cohort as compared to the healthy patient cohort. The glycan structure with the most significant discriminatory ability was P1 (Galα1-4Galβ1-4GlcNAcβ; P < 0.001), a member of the P blood group system.
In recent years a number of glycan based microarray platforms, based on specific (covalent or non-covalent) immobilization of chemically conjugated glycans on a modified surface, have been developed for high-throughput investigations of the glycome . Glycan-based arrays vary in ligand presentation, glycan origin (isolated from natural sources or chemically synthesized), assay conditions, detection method, microspheres (suspension array) and immobilization on flat surfaces (printed glycan array, ELISA); all of which contribute to the affinity and selectivity of binding [16, 17]. There is a crucial need for standardization and comparison between various forms of glycan microarrays, especially taking into account their proven scientific potential for translational research and subsequent clinical applications. The increasing number of high-throughput glycan array studies requires the application of different bioinformatical analyses for validation and comparison of datasets. However, to date, only a few reports comparing glycan microarrays and conventional ELISA have been performed. Most of these studies only investigated them in terms of sample size, dynamic range and sensitivity of measurement (detection limits) [2, 18–20].
In this present study, which is in line with our previous development of a glycan-based suspension array, we compared three glycan-immunoassays for the detection of the most complicated glycan-binding protein class; antibodies in human plasma. These assays include (1) printed glycan array; (2) multiplex flow cytometric suspension assay; and (3) regular direct ELISA, the most established method. Printed Glycan Array (PGA) as one of the newest high-throughput microarray technologies [6, 14] allows the automated detection of an unlimited number of natural and synthesized glycans in one experimental setting. It is characterized by high sensitivity and a significant reduction of reagent consumption. The second approach, multiplex flow cytometric suspension assay (SA), incorporating fluorescent microspheres with distinct spectral addresses as a solid support for ligands, was recently developed for profiling of anti-glycan antibodies [9, 21]. The advantage of this assay is reflected by the flexibility and simultaneous detection of multiple ligands in one sample with minimal reagent consumption. The third method, conventional direct ELISA [7, 19, 22, 23], which allows assessment of a limited number of glycan-protein interactions, is well established and currently the most widely used method for this type of application. In both the PGA and ELISA glycans are positioned onto a flat surface; in PGA glycans are covalently attached to the surface of a glass slide in a monovalent form as ω-alkyl glycosides (Glyc-sp-NH2), whereas in ELISA they are physically adsorbed to polystyrene in multivalent form, as conjugates with a polyacrylamide or albumin carrier. In the case of suspension array (SA), glycans are coupled as end-biotinylated glycopolymers  via streptavidin-biotin reaction to pre-modified fluorescent beads. All three glycan immunoassays display various detection specificities resulting in specific advantages and limitations. ELISA is most relevant for the investigation of a limited panel of glycans, and would be suitable for a preliminary survey of new glycan-binding partners. Printed glycan array allows broad glycan library screening, and suspension assay has advantages for the rapid and flexible multiplex detection of up to several dozen samples. Therefore, all three glycan assays could be used to study different aspects of glycan-antibody interactions.
We performed a comparative analysis using three glycan-based immunoassays for a cohort of 48 patients with and without ovarian cancer. Two target anti-glycan antibodies were selected, anti-blood group A/B trisaccharides [6, 25–27] and our previously identified candidate P1 in order to investigate the expected anti-glycan antibody distribution in plasma for each assay.
Material and methods
Clinicopathological characteristics. Patient numbers and percentage (in brackets)
Patient cohort (n = 48)
Stage I / II
Stage III / IV
NUNC MaxiSorp 96-well immunoplates (Thermo Fisher Scientific, Roskilde, Denmark) were coated with Glyc-PAA (Lectinity Holdings, Moscow, Russia), 10 μg/mL, 100 μl per well in carbonate buffer (50 mM Na2CO3/ NaHCO3, pH 9.6) for 12 h at 4°C. Carbohydrate-free PAA was applied as negative control. Plates were blocked with 1% (w/v) BSA (Sigma-Aldrich Chemie GmbH, Buchs, Switzerland) in PBS for 40 min at 37°C and washed four times with PBS containing 0.5% (v/v) Tween-20 after incubation. Plasma samples were diluted 1:1000 in incubation buffer (PBS, 0.3% (w/v) BSA, 0.02% (v/v) Tween 20), added to plates in duplicate and incubated for 60 min at 37°. Between each of the following steps plates were washed four times with PBS containing 0.5% (v/v) Tween-20: incubation with 100 μl per well of goat anti-human Ig (IgA + IgG + IgM) conjugated to long chain biotin for 60 min at 37° (Pierce, Rockford, IL, USA, 0.16 μg/mL in incubation buffer); streptavidin horse raddish peroxidase conjugate for 60 min at 37°C (Southern Biotechnology Associates, Inc., Birmingham, AL, USA, 0.083 μg/mL in incubation buffer); and chromogen substrate 3,3′,5,5′-tetramethylbenzidine (Sigma-Aldrich Chemie GmbH, Buchs, Switzerland) for 5 min at RT. The peroxidase reaction was stopped by addition of equal volumes of 1M H2SO4. Absorbance was measured at 450 nm using a TECAN plate reader (Tecan Spectrafluor Plus, Tecan Trading AG, Männedorf, Switzerland).
Suspension array (SA)
The Bio-Plex Suspension Array (Bio-Rad Laboratories, Hercules, CA, USA) is a multiplex analysis system that permits the simultaneous analysis of up to 100 different biomolecules in a single microplate well. The constituents of each well are drawn up into the flow-based Bio-Plex array reader, which quantifies each specific reaction based on its bead color using fluorescently labeled reporter molecules specific for each target protein followed by Bio-Plex Manager software data analysis. A 96-well Multiscreen HTS filter plate (Millipore Corp., Billerica, MA, USA) was soaked in 100 μl of antibody diluent for 5 min (PBS-0.05 M Tris, pH 7.2, 0.25% BSA, Sigma-Aldrich Chemie GmbH, Buchs, Switzerland). Antibody diluent incorporating 2000 beads/well (50 μl/well) was added. The plate was washed three times with 100 μl of washing buffer (PBS-0.05 M Tris, pH 7.2) using a vacuum manifold (Bio-Rad, Munich, Germany). Human samples were added in duplicates to wells (in antibody diluent 1:40 (50 μl/well)) and agitated at 1,100 rpm for 30 s on a microplate shaker before incubation on a shaker (200–300 rpm) for 1 h at RT in the dark. After incubation, the plate was washed three times using washing buffer. Secondary antibodies (R-phycoerythrin conjugated goat anti-human Ig (IgM + IgG + IgA, H + L; Southern Biotechnology Associates Inc., Birmingham, AL, USA, 25 ng/well) were added and incubated for 30 min on the plate shaker in the dark. The plate was washed three times with washing buffer, beads were then resuspended and shaken for 30 s at 1,100 rpm in 100 μl of washing buffer before being analyzed on the Bio-Plex array reader. Data were acquired in real time analyzing 100 beads by their median fluorescence intensitiy (MFI) using a computer software package (Bio-Plex Manager 4.1; Bio-Rad Laboratories, Hercules, CA, USA).
The end-biotinylated glycopolymers for coupling to fluorescent microspheres, Glyc-PAA-biot1, were produced in-house (Laboratory of Carbohydrates, Shemyakin-Ovchinnikov Institute of Bioorganic Chemistry, Russian Academy of Sciences, Moscow, Russian Federation). Biotinylated glycopolymers were coupled to fluorescent carboxylated beads of 5.5 μm diameter with distinct spectral “addresses” (Bio-Rad Laboratories Inc., Hercules, CA, USA). Each set of beads was embedded with a precise ratio of red and infrared fluorescent dyes allowing its identification by measuring the intensities of the two classifier fluorochromes. The stock vial of microspheres (1.25 × 107 microspheres/mL) was vortexed for 30 s and sonicated for 30 s in a water bath prior its use. Bead suspension (100 μl; 1.25 × 106 microspheres, 0.2 nmol –COOH groups in total, according to the supplier’s information) was centrifuged for 4 min, 14,000 g at RT. The pellet was resuspended in bead wash buffer (100 μl; Bio-Plex amine coupling kit, Bio-Rad Laboratories Inc., Hercules, CA, USA) by vortexing and sonication, and washed by centrifugation as described above. After gentle removal of supernatant, the pellet was resuspended in 80 μl of bead activation buffer (Bio-Plex amine coupling kit, Bio-Rad Laboratories Inc., Hercules, CA, USA), vortexed and sonicated. Sulfo-N-hydroxysuccinimide sodium salt (S-NHS) and 1-ethyl-3-[3,3-dimethylaminopropyl]carbodiimide hydrochloride (EDC; Pierce Biotechnology, Rockford, IL, USA, both 50 mg/mL in activation buffer) were prepared immediately prior to use, and 10 μl of each solution was added to the bead suspension, followed by vortexing for 30 s. Beads were incubated on a vertical rotor in the dark for 20 min at RT. The activated beads were centrifuged and supernatant removed. The pellet was resuspended in 150 μl biotin-solution (0.1 M NaHCO3, pH 8.3, containing 1 μg (≈ 2 nmol) of biotin-NH(CH2)6NH2. Lectinity Holdings, Moscow, Russia) and incubated on a vertical rotor with medium speed for 2 h at RT in the dark. Obtained biotinylated beads were pelleted by centrifugation and resuspended in 150 μl of 50 mM ethanolamine (Sigma-Aldrich Chemie GmbH, Buchs, Switzerland) in 0.1 M NaHCO3, pH 9.0 to quench unbound activated groups. Beads, protected from light, were agitated on a rotator for 30 min at RT and centrifuged. The pellet was washed twice with 500 μl PBS, pH 7.4 and resuspended in streptavidin-solution (400 pmol streptavidin in 150 μl PBS; Bio-Rad Laboratories Inc., Hercules, CA, USA). Suspended bead were vortexed and agitated on a rotator in the dark for 2 h at RT or 12 h at 4°C. Beads were washed twice with 500 μl PBS followed by centrifugation. Glyc-PAA-biot1 solutions in water (20 pmol end-biotin labeled Glyc-PAA, ∼30 kDa, Glyc contents - 20% mol) , were added with 1.25 × 106 streptavidin-coated beads in 150 μl PBS to the reaction tubes. The mixture was protected from light and agitated on a rotator for 6 h at RT. Modified microspheres were centrifuged, supernatant removed and beads washed twice with 500 μl of bead storage buffer (Bio-Rad Laboratories Inc., Hercules, CA, USA). Beads were centrifuged and resuspended in 100 μl of bead storage buffer and concentration determined using a hemocytometer (Roth AG, Karlsruhe, Germany) before storing at 4°C, protected from light.
Printed glycan array (PGA)
Printed glycan array slide fabrication and high-throughput profiling was performed as previously described [6, 11, 14]. Briefly, monomeric glycans as ω-aminopropyl glycosides of 95–98% purity (Lectinity Holdings, Moscow, Russia) were diluted in 300 mM phosphate buffer pH 8.5, containing 0.005% Tween 20 and printed by robotic pin deposition on N-hydroxysuccinimide activated glass slides (Nexterion Slide H, Schott, Jena, Germany). Glycans were printed at a 50 μM concentration in eight replicates. Free N-hydroxysuccinimide activated groups were blocked with 50 mM ethanolamine in 50 mM borate buffer at a final pH of 9.2. Slides were then rinsed with deionized water, dried and stored at room temperature in a desiccator. Each plasma sample was diluted 1:15 with PBS containing 0.1% v/v Tween20 and 3% w/v BSA, thoroughly vortexed for 15 s and incubated at 37°C for 15 min to dissolve potential lipid aggregates. Samples were transferred to the array slides and gently rocked in a sealed humidified incubator for 2 h at 37°C. Unbound sample components were washed with a series of 0.1% and 0.001% Tween-20 in PBS. Antibody solution and goat anti-human IgA + IgG + IgM conjugated to long chain biotin (Pierce Biotechnology, Rockford, IL, USA, 1:100 (20 μg/mL) in PBS containing 0.1% Tween-20 and 3% BSA) was added. Slides were incubated at RT in a humidified chamber for 45 min and then washed. Bound antibodies were visualized by incubating slides with fluorescent dye streptavidin solution (Alexa Fluor555, Molecular Probes, Invitrogen, Carlsbad, CA, USA, 1:1000 in PBS/0.1% Tween-20) at RT for 30 min. Fluorescence signals corresponding to glycan-bound antibodies were measured and quantified using ImaGene analysis software version 6.1 and 7.5 (BioDiscovery, El Segundo, USA). Signals were measured as total signal intensity (medTSI) per glycan and were expressed as median across eight intra-array replicates. In a prior discovery approach we screened plasma of non-mucinous ovarian cancer patients in comparison to healthy controls and revealed anti-P1 antibodies to be the most specific differentiating anti-glycan antibody. Apart from the standard ABO blood group system we have therefore used anti-P1 in this comparative analysis as an example of a new biomarker identified by PGA only.
Controls and validation procedures
Due to the known high expression levels in healthy controls α-rhamnose was used as a “positive biological control” in the PGA experiments . Aminoglucitol, an opened reduced form of D-glucose which is not present as a structural component of regular glycosylation, has shown to be negative for anti-glycan binding at similar values to the technical background [6, 11]. We therefore used it as a negative biological control in both PGA and SA experiments, where MFI signals for aminoglucitol binding did not exceed 200 and were close to the technical cut-off in all individual samples (data not shown). A glycan-free polymer (PAA) was used in ELISA as a control for non-specific binding and displayed lower signals than the lowest signals for Glyc-PAA coating in all cases (data not shown). Human polyclonal antibodies against A/B blood group antigens were used to test the coupling effectiveness in SA. Anti-glycan antibodies were purified from pooled human plasma of the same blood group using affinity chromatography as previously described [7, 21]. Polyclonal anti-glycan antibodies bound to cognate glycans in a dose-dependent manner displayed no or very low cross-reactivity to other glycans in both SA and ELISA (data not shown).
Anti-A/B antibody screening
The well known ABO blood group carbohydrate antigens were used as a model for the comparison of our three glycan immunoassays. ABO-specific antibodies (isohaemagglutinins) belong to a group of naturally occurring antibodies  which can be detected by the haemagglutination test, ELISA [22, 23] and flow cytometry . Blood group information was available for 31 patients, of whom 13 patients had blood group A (41.9%), 6 patients blood group B (19.3%), 2 patients blood group AB (6.4%) and 10 patients blood group O (32.2%). Antibody profiles for Atri (GalNAcα1-3 (Fucα1-2)Galβ) and Btri (Galα1-3(Fucα1-2)Galβ) blood group antigens were generated using SA, ELISA and PGA. Raw data were acquired for all technologies in different units; values varied from 74 to 13595 (MFI, SA), from 0.152 to 2.305 (OD, 450 nm, ELISA) and from 0 to 51.28*105 (medTSI, PGA). Anti-A/B antibody levels were calculated as distribution of the median and additional median absolute deviation.
In the combination of three different immuno-assays one of the major problems is the absence of a ‘gold’ standard method, which means there is no experimental approach available, which can be taken as a norm of measuring anti-glycan antibodies. Neither has an internal standard nor a background threshold been established , and the data values achieved in each method are independent from each other, which make the direct comparison impossible: (A) ELISA values reflect the oxidized product of the chromogen TMB (3, 3′, 5, 5′-tetramethylbenzidine) and is a substrate conversion based on an enzymatic reaction; (B) printed glycan array (PGA) and suspension array (SA) (R-phycoerythrin) results are achieved due to fluorescence measurements. These differences lead to data values, which require data mining before method comparison. When dealing with parameters of different units and scales it is very important to standardize or normalize data in order to allow for a reliable comparison. Here we used standardization of data as it transforms datasets into a mean zero and a unit variance and keeps ranges similar and variables different. This is also the most commonly used method for normalization and comparison of methods incorporating differing datasets. Whilst PGA data were pre-processed , SA raw data were log-transformed to improve interpretability and visualization. To solve the problem of different data values, all data sets were standardized as follows: z = (xi – xmean)/xsd. Standardized data (z) were generated for each vector data set (xi) by subtraction by their mean (xmean) and division of their standard deviation (xsd), and called ‘standardized antibody measurements’ (SAM). Combined graphical and statistical interpretations of method-comparison studies were performed and included scatter plots combined with correlation and regression analysis . Data analysis, including calculation of mean, median, standard deviation and coefficient of variation was performed using the open source statistical programming language R (http://CRAN.R-project.org/, version 2.8.1). Statistically significant differences were proved for each method by the Wilcoxon rank sum test or student t test. Concordance correlation coefficients (rCCC) , which evaluate the degree to which pairs of observations fall on the 45° line through the origin, were calculated and compared within all independent methods (R package epiR). Direct comparison of two methods was performed using parametric linear regression [32, 33]. Non-parametric testing of median signals among all known blood groups was measured using the Kruskal-Wallis rank sum test. For comparisons of high versus low anti-glycan antibody levels in correlation with known blood groups, sensitivity, specificity and area under the curve (AUC) were calculated for each method (R package ROCR ). The best cut-off between observed false negative and false positive values was described as the “precision-recall break-even point” , the point at which precision equals recall and predictions are made due to the prevalence within the data given. All p-values < 0.05 were taken as significant.
Detection of anti-A and anti-B antibodies
Using PGA, we observed higher antibody binding levels for Btri than for Atri with higher individual variability of the latter, consistent with the literature . The same pattern was observed in ELISA, followed by lower individual variability of both anti-glycan antibodies, indicated by median absolute deviation (data not shown). In contrast to PGA and ELISA, data using SA revealed no significant difference between general levels of anti-A and anti-B antibody binding, accompanied by higher individual variability in anti-A antibody binding.
Discrimination between ABO blood groups
Expression of ovarian cancer biomarker P1 trisaccharide
In the current study, which is in continuity of two previous publications of ours using both a newly described PGA and SA [11, 21], we investigated the efficacy and selectivity of three custom-developed immunoassays to profile anti-glycan antibodies in individual human plasma samples. All three methods adequately reflected the known serological isohaemagglutinin distribution within ABO blood groups, with the best being SA and PGA. The correlation between all three assays was more distinctive for the detection of anti A-antibodies than for anti-B antibodies. Our results indicate that anti-A/B blood group antibodies positively correlated with each other and that all glycan immunoassays could detect these with statistical significance. They demonstrated method-independent tendencies due to their high affinity, which reflects their strong biological importance even in a small patient cohort .
In contrast, anti-P1 correlations across the three assays were considerably less pronounced although it was primarily detected by PGA with the highest discriminative power in a discovery approach in non-mucinous ovarian cancers. Nevertheless, with equal control to cancer patient numbers we did improve the correlation of this cancer-specific marker for the comparison of PGA and SA.
This work was supported by Oncosuisse (02115-08-2007 to V.H.S.); Swiss National Foundation (320000-12543 to V.H.S.; PBZHP3-133289 to F.J.); European Science Foundation (V.H.S.); Cancer Institute NSW (09/CRF/2-02 to V.H.S.); Mary Elizabeth Courier Research Scholarship, RANZCOG (to V.H.S.); William Maxwell Trust (to V.H.S.); Cellexicon, Inc. La Jolla, CA, USA (to M.E.H.); National Cancer Institute, USA, grant 1U01CA128526 (to M.E.H); RAS Presidium Program ‘Molecular and Cell Biology’ (to N.V.B.) and State grant #2011-1.2-512-013-008 (to N.V.B.).
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.