Glycomics based biomarkers of the rate of aging :
development and applications of high-throughput N-glycan analysis
Ruhaak, L.R.
Citation
Ruhaak, L. R. (2011, March 24). Glycomics based biomarkers of the rate of aging : development and applications of high-throughput N-glycan analysis.
Retrieved from https://hdl.handle.net/1887/16559
Version: Corrected Publisher’s Version License:
Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden
Downloaded from: https://hdl.handle.net/1887/16559
Note: To cite this publication please use the final published version (if
7
Chapter 7.
High-throughput immuno-affinity capturing procedure for targeted glycan-based biomarker discovery:
application to AAT and IgA
L.Renee Ruhaak1, Carolien A.M. Koeleman1, Hae-Won Uh2, P.Eline Slagboom3,4, André M. Deelder1, Manfred Wuhrer1
1 Dept. of Parasitology, Biomolecular Mass Spectrometry Unit, Leiden University Medical Center, Leiden, The Netherlands
2 Dept. of Medical Statistics and Bioinformatics, section Medical Statistics, Leiden University Medical Center, Leiden,
The Netherlands
3 Dept. of Medical Statistics and Bioinformatics, section Molecular Epidemiology, Leiden University Medical Center, Leiden, The Netherlands
4 Netherlands Consortium of Healthy Aging
Manuscript in preparation
Abstract
Protein N-glycosylation patterns show large variation. Since this large variation may re- flect the physiological state of cells and organs in humans, protein glycosylation has been marked as a large pool of potential biomarkers. Larger scale studies are needed for the identification and validation of biomarkers, and the analytical techniques required for the conduction of such studies assessing protein N-glycosylation in larger sample sets have recently been developed. While some studies have been aimed at glycosyla- tion analysis of immunoglobulin G, such methods have up to now mainly been applied to complex mixtures of glycoproteins in biofluids (e.g. plasma).
This chapter describes the use of large-scale immuno-affinity capturing, combined with high-throughput N-glycan analysis for targeted glycan based biomarker discovery.
Alpha-1-antitrypsin and immunoglobulin A are enriched from human plasma in a bead- based procedure in a 96-well microtitration plate based platform. The intra-batch va- riation of the total analysis strategy, consisting of protein enrichment, glycan release, glycan labeling and glycan analysis using multiplexed CGE-LIF, was determined to be 15-20%. The strategy has been applied to 2395 plasma samples from the Leiden Lon- gevity Study.
AAT-glycosylation patterns are associated with calendar age and differ between females and males. Moreover, several AAT-glycans could be associated with physiological pa- rameters marking cardiovascular and metabolic diseases. Two non-fucosylated glycans (AAT-16 and AAT-21) were shown to be positively related to the incidence of myocardial infarction. While several parameters could be related to AAT-glycosylation features, the IgA glycosylation patterns seem to be almost unaffected. The only difference in IgA gly- cosylation patterns that could be observed was between females and males.
It was shown that a strategy consisting of large-scale immuno-affinity capturing of pro- teins from human plasma using a bead-based method, coupled with high-throughput N-glycan analysis using multiplexed CGE-LIF is a powerfull tool for the analysis of N- glycosylation patterns of specific glycoproteins in large studies. The strategy has suc- cessfully been applied to 2395 plasma samples from the LLS and it could be observed that AAT glycosylation patterns are regulated by all kinds of biological processes, while the glycosylation pattern of IgA was not affected by most parameters tested.
7
Introduction
N-Glycosylation, the enzymatic addition of oligosaccharides to asparagine residues in proteins, occurs on more than 50% of human proteins [1] and proper glycosylation is essential for the survival of most multicellular organisms. N-Glycans have impor- tant functions in several biological processes [1-6]. N-glycans are often branched structures, where monosaccharides may be linked in several different ways; protein glycosylation may be very diverse. The glycan signature, i.e. the total plasma N- glycome of an individual, is highly reproducible in a given physiological state [7;8], however, when the physiological state changes, e.g. due to aging or disease, the glycan pattern can change dramatically (e.g [7]). Due to the large variability of gly- cosylation and the reflection of several physiological states in glycosylation patterns, protein glycosylation patterns have been marked as a large field of potential biomar- kers [9].
To identify and validate biomarkers, large scale studies are needed, and the analyti- cal methods required for the evaluation of protein glycosylation patterns at the glycan level in larger sample sets have only recently been developed (e.g. [80;97;99;204]).
Up to now, these methods have only been applied for the evaluation of plasma N- glycosylation profiles. However, some constraints are associated with the interpreta- tion of the results from such analyses. As the profiles originate from the total protein pool present in plasma, changes in the N-glycosylation profile may be caused by alterations in protein concentration, or by changes in protein-specific glycosylation patterns. Moreover, assuming the latter cause, information regarding the site (and thus protein) of attachment is lost, and it is therefore not clear from which glyco- protein the glycan signature is changed. A third aspect is that glycans from high abundant proteins dominate the glycan pattern, and therefore, changes in the glyco- sylation signature of less abundant proteins will not be visible.
It would thus be advantageous to have a method for fast, large scale glycan analysis of specific glycoproteins or groups of glycoproteins. Current methods for glycopro- tein enrichment may comprise lectins or (immuno-)affinity purification (e.g. [48;205]).
Except for affinity capturing of IgG [183;185], however, this has to our knowledge not been applied in large scale studies.
Large scale analysis of glycans from single glycoproteins or mixtures of a few gly- coproteins may be performed at the glycopeptide level or at the glycan level. Using glycopeptide analysis, site specific glycosylation profiles may be obtained. However, analysis at the glycan level may also be advantageous as such an approach is uni- versal, while analytical procedures at the glycopeptide level have to be optimized for each glycoprotein and thus require additional skills to adjust the procedure.
We here describe an approach for high-throughput immuno-affinity capturing in com- bination with N-glycan analysis for targeted glycan based biomarker discovery. To this end, both commercially available antibody coated beads (anti-immunoglobulin A, IgA), as well as beads that were coated in-house with VHH-antibodies (small, camelid-derived single domain antibody fragments, www.bac.nl , anti- alpha-1-anti- trypsin, AAT) were used. We have applied our previously published workflow for high-throughput N-glycan analysis at the glycan level using multiplexed CGE-LIF analysis.
The described high-throughput immuno-affinity capturing procedure for profiling of AAT- and IgA derived glycosylation patterns has subsequently been applied to 2395 samples of the Leiden Longevity Study. This is a cohort which consists of long-lived siblings together with their offspring and the partners thereof, and is aimed at the identification of markers for healthy aging [160].
7
Materials and Methods Materials
Dimethylsulfoxide (DMSO), Nonidet P-40 (NP-40), formic acid, triethylamine (TEA), aminopyrene-1,3,6-trisulfonic acid (APTS), sodium cyanoborohydride (NaBH3CN) and 2-picoline-borane, trifluoroacetic acid (TFA), and anti-human IgA−Agarose an- tibody produced in goat were obtained from Sigma-Aldrich (Zwijndrecht, The Ne- therlands). Sodium dodecyl sulfate (SDS) was bought from United States Biochemi- cals (Cleveland, OH). PNGase F was obtained from Roche Diagnostics (Mannheim, Germany). Biogel P-10 was obtained from Bio-Rad (Veenendaal, The Netherlands), while citric acid was from Merck (Darmstadt, Germany). Acetonitrile was purchased from Biosolve (Valkenswaard, The Netherlands). 0.45 μm GHP filter plates were ob- tained from Pall Corporation (Ann Arbor, MI) and 96-well V-bottom deep well plates from Westburg (Leusden, The Netherlands). PCR plates for measurement in the DNA-sequencer were obtained from Thermo Fischer Scientific via Westburg (Leus- den, The Netherlands). 96-well polypropylene filterplates containing a 10 μm PE frit were obtained from Orochem (Lombard, IL). Anti-human AAT-coated Sepharose beads were kindly provided by Dr. J. Stam, Utrecht University, The Netherlands.
Plasma was obtained from a healthy donor for the repeatability study
Participants of the Leiden Longevity Study
In the Leiden Longevity study, Caucasian families were recruited if at least two long- lived siblings were alive and fulfilled the age-criterion of 89 years or older for males and 91 year or older for females, representing less than 0.1 % of the Dutch popula- tion in 2001. In total, 960 long-lived proband siblings were included, 1710 offspring with a mean age of 59.4 and 761 partners with a mean age of 58.9 [160].
The study protocol was approved by the Leiden University Medical Centre ethical committee and an informed consent was signed by all participants prior to participa- tion in the study.
Phenotypic parameters
All serum measurements were performed with fully automated equipment. For insu- lin, the Immulite 2500 from DPC (Los Angeles, CA, USA) was applied. For glucose, total cholesterol, HDL-cholesterol (HDL-C) and triglycerides the Hitachi Modular or the Cobas Integra 800, both from Roche, Almere, the Netherlands were applied. For free triiodothyronine, the Modular E170 was used from Roche, Almere, the Nether- lands. LDL-cholesterol level (LDL-C) was calculated using the Friedewald formula (LDL-C = total cholesterol - HDL-C - (triglycerides/2.2); unit mmol/l) and set to mis- sing if plasma triglyceride concentration exceeded 4.52 mmol/l.
Immunoaffinity enrichment of AAT and IgA
Antibody coated beads were washed three times with 10 volumes of PBS. Six or twenty microliters of anti-AAT or anti-IgA beads, respectively, were applied per well to a 96-well filter plate. The volume was brought to 200 μL with PBS, and 10 μL of plasma were applied per well. The plate was sealed with tape and incubated on a shaker for 1 h. The beads were washed with 3 x 200 μL PBS by vacuum filtration.
After subsequent washing with 2 x 200 μL water, enriched glycoproteins were eluted with 100 μL of 100 mM formic acid into a V-bottom microtitration plate. Samples were dried by vacuum centrifugation. To allow reuse of the antibody coated beads, beads were washed in the filter plates using 3 x 100 μL of 100 mM formic acid, fol- lowed by equilibration using 3 x 200 μL PBS. 200 μL of PBS were added to the well prior to application of the following batch of plasma samples.
Glycan release, labeling and purification
N-glycans were released from the enriched glycoprotein samples as described pre- viously [97] with slight modifications. Shortly, proteins were denatured after addition of 2 μl 2% SDS by incubation at 60˚C for 10 min. Subsequently, 2 μl 2% NP-40 containing 0.2 mU of PNGase F was added to the samples. The samples were in- cubated over night at 37˚C for N-glycan release. Labeling of oligosaccharides was performed as published [87] with slight modifications: 2 μl of a freshly prepared solu- tion of label (APTS; 20 mM in 3.6 M citric acid) and 2 μl of freshly prepared reducing
7
agent solution (0.2M 2-picoline-borane in DMSO) were added, the plate was sealed using adhesive tape and after 5 min of shaking, the samples were incubated at 37
˚C for 16 hours.
HILIC-SPE was applied to purify labeled N-glycans. An amount of 100 μl of a 100 mg/mL suspension of Biogel P-10 in water/ethanol/acetonitrile (70:20:10, v/v) was applied to each well of a 0.45 μm GHP filter plate (Pall Corporation, Ann Arbor, MI).
Solvent was removed by application of vacuum using a vacuum manifold (Millipore, Billerica, MA). All wells were prewashed using 5 × 200 μl water, followed by equili- bration using 3 × 200 μl acetonitrile/water (80:20, v/v). The samples were loaded to the wells, and the plate was shaken for 5 min. The wells were subsequently washed using 5 × 200 μl acetonitrile/100 mM triethylamine (TEA) adjusted to pH 8.5 with ace- tic acid (80:20, v/v) , followed by 3 × 200 μl acetonitrile/water (80:20, v/v). Washing steps were performed by addition of solutions, incubation for 30 s, and removal of solvent by vacuum. Water (100 μl) was applied followed by 5 min incubation to allow swelling of the Biogel P-10 particles. Then N-glycans were eluted using 2 x 200 μl water. Samples were incubated for 5 min. prior to collection of eluates by vacuum in a 96-well V-bottom polypropylene deep well plate. The combined eluates were either analyzed immediately by CGE-LIF or stored at -20ºC until usage.
CGE-LIF analysis
Multiplexed CGE-LIF analysis was performed as previously described (Chapter 4).
Shortly, two μl of N-glycan eluate were added to 60 μl of DMSO in a PCR plate (Thermo Fischer Scientific via Westburg, Leusden, The Netherlands). Plates were sealed and turned several times for thorough mixing and subsequently centrifuged prior to analysis using an ABI-3730 DNA sequencer (Applied Biosystems). The in- jection voltage was set at 7.5 kV, while the running voltage was 10 kV. The system was equipped with a 48 channel array with capillaries of 50 cm, and the capillaries were filled with POP-7 buffer (Applied Biosystems). The 3730 running buffer was obtained from Applied Biosystems. Data was collected with a frequency of 10 Hz for 50 min using ABI Data Collection software v.2.0.
RP-ESI-IT-MS/MS of tryptic digest
Trypsin (Promega, Madison, WI) was dissolved in 50 mM ammonium hydrogen car- bonate buffer (Fluka) to a concentration of 0.02 μg/ml, and immediately 40 μL was added to the enriched glycoprotein samples. Samples were shaken for 10 min, and incubated overnight at 37°C.
5 μl of the tryptic digest were applied to a reverse-phase column (PepMap, 3 μm, 75 μm• 100 mm; Dionex /LC Packings, Amsterdam, The Netherlands) using an Ultimate 3000 nano-LC system (Dionex /LC Packings). The column was equilibrated at room temperature with eluent A (0.1% formic acid in water) at a flow rate of 200 nL min- 1. After injection of the sample, elution conditions were switched to 10% solvent B (95% acetonitrile, 0.1% formic acid), followed by a gradient to 60% B in 45 min and a subsequent isocratic elution of 10 min. The eluate was monitored by absorption at 215 nm. The LC column was coupled to an Esquire HCT-Ultra ESIion trap-MS (Bruker-Daltonics, Bremen, Germany) equipped with an online nanospray source operating in the positive-ion mode. For electrospray (1100-1250 V), electropolished, stainless steel LC-MS emitters (150 μmOD, 30 μmID) from Proxeon A/S (Odense, Denmark) were used. The solvent was evaporated at 175 ˚C employing a nitrogen stream of 7 L min-1. Ions from m/z 500 to m/z 1800 were registered in the MS mode.
When operated in the auto MS/MS mode, registering ions from m/z 140 to 2200, each MS scan was followed by the acquisition of MS/MS spectra of up to three of the most abundant ions in the MS spectrum. MS/MS spectra were searched against the H. Sapiens NCBI database using the Mascot search algorithm (Matrix Science Ltd., London, UK) allowing one missed cleavage site. Identification of at least two peptides was regarded significant for protein identification.
Data processing
Data files were converted to xml files using DataFileConverter, which is supplied by Applied Biosystems. Files were then loaded into Matlab (version 2007a) software (The Mathworks, Inc., Natick, MA), and after smoothing, the data were cropped to 17000 datapoints to reduce the data volume. Alignment using Correlation optimized
7
warping (COW) was performed as reported previously [204;206] using a represen- tative electropherogram as the reference file. Segment length and slack size were optimized according to [204]. After smoothing and background adjustment, the peak integrals were determined (22 for AAT and 37 for IgA). Peak integrals were normali- zed on the signal intensity of Peak 3 (AAT) or Peak 7 (IgA).
For the determination of the intra-batch variation 4 standard samples were analy- zed. Glycoproteins were enriched and N-glycans were released, labeled, purified, and subsequently analyzed using the DNA sequencer in parallel on one plate. The procedure was repeated on four consecutive days. Per day averages as well as standard deviations were calculated from the four samples. RSDs were determined and the average RSDs from the four consecutive days were calculated per peak.
To determine inter-batch repeatability, four standard samples were analyzed. N- glycans were released, labeled, purified, and subsequently analyzed using the DNA sequencer. The procedure was repeated on four consecutive days. For each sample position the averages and standard deviations over the four days were calculated.
RSDs and subsequently average RSDs over all four samples were calculated.
Statistics
The samples of 2395 participants were divided over 28 individual plates to analyze AAT- and IgA-derived N-glycosylation patterns. To correct for batch effects, the N- glycosylation values were regressed on the categorical variable batch memberships.
The standardized residuals of this model were used for further statistical analysis.
Since we have multiple participants from the same family, the sandwich estimator was used to obtain valid standard errors. P-values <0.05 were regarded statistically significant. First, linear regression was used to explore the relationships between each of the response variables and covariates - age and sex - adjusting for the fa- mily status (offspring or partner of the offspring). Then, linear regression was used to explore relationships between N-glycosylation features and covariates – BMI, levels of cholesterol, HDL-cholesterol, LDL-cholesterol, triglyceride and glucose as well as
insulin activity - adjusting for the family status (offspring or partner of the offspring) age, sex and the age x sex interaction.
To determine potential biomarkers for longevity, logistic regression was applied to investigate whether glycosylation feature (independent variable) was predictive in classifying the family status after adjustment for age, sex, and their interaction. In that respect, the response variable (the family status) is coded as 0 (= partner) and 1 (= offspring of long-lived sibling). Finally, linear regression was used to explore relationships between glycosylation features and disease status – levels of CRP and incidence of MI, CVA or diabetes- adjusting for age, sex and their interaction.
Analyses were performed using STATA 10 (StataCorp LP, College Station, Texas, USA) and R version 2.9.0 (R Development Core Team).
Results
Protein enrichment
Glycoproteins were enriched from human plasma using high-throughput immunoaf- finity capturing in a 96-well format. Antibody coated beads were applied to each of the wells of a filter plate, after which plasma was added to the wells. After incubation and thorough washing with PBS and water to limit unspecific binding, the enriched glycoprotein fraction was eluted using 100 mM formic acid. To check the purity of the enriched glycoprotein samples, an LC-MS/MS run was performed on the tryptic digests, followed by a MASCOT search. Identification of at least two peptide frag- ments was regarded necessary for protein identification. This analysis revealed that the AAT-enriched fraction was constituted primarily of AAT (mascot score >1000, peptide coverage 41%), while for IgA, three proteins were observed, all with a score above 100: IgA1 (127), IgG (125), and albumin (124), indicating that the enrichment of IgA was less specific.
7
Glycan peak annotation
We recently developed a platform for high-throughput glycan analysis using a mul- tiplexed CGE-LIF instrument (Chapter 4). Using this platform, 96 samples can be measured within 3 hours, which makes it highly suitable for large scale studies. A disadvantage of the use of this platform is that no direct peak annotation can be performed. Previously, we obtained annotations of 35 abundant plasma N-glycan structures by fractionation of plasma N-glycans using HILIC-HPLC, and subsequent annotation of the peaks in the fractions based on accurate masses obtained by MAL- DI-FTICR-MS, standardized migration times in the CGE-LIF and glycan standards.
This previous annotation of plasma N-glycans, together with literature, was now used to annotate the observed glycans in the AAT- and IgA- enriched glycan samples. An overview of the annotation is given in Table 7 1 (AAT) and Table 7 2 (IgA). An over- lay of ten annotated electropherograms is depicted in Figure 7 1. Glycan structures are schematically represented and are assigned based on migration positions, Table 7 1 and Table 7 2, and literature.
Fraction no. Composition Spiking Literature
AAT-2 Hex6HexNAc5Sia3 X [207;208]
AAT-3 Hex5HexNAc4Sia2 X [207;208]
AAT-4 Hex5HexNAc5Sia2 X
AAT-5 Hex5HexNAc4Fuc1Sia2 X [207;208]
AAT-6 Hex5HexNAc5Fuc1Sia2 X
AAT-7 Hex7HexNAc6Fuc1Sia3 X [207]
AAT-8 Hex6HexNAc5Sia2 X [207]
AAT-9 Hex6HexNAc5Sia2 X [207]
AAT-10 Hex6HexNAc5Fuc1Sia2 X [207]
AAT-11 Hex6HexNAc5Fuc1Sia2 X [207]
AAT-12 Hex5HexNAc4Sia1 X [207;208]
AAT-13 Hex5HexNAc4Sia1 X [207;208]
AAT-14 Hex5HexNAc4Fuc1Sia1 X [207]
AAT-15 Hex5HexNAc5Fuc1Sia1 X [207]
AAT-16 Hex6HexNAc5Sia1 X
AAT-17 Hex3HexNAc4Fuc1 X
AAT-20 Hex4HexNAc4 X
AAT-21 Hex5HexNAc4 X
AAT-22 Hex5HexNAc4Fuc1 X
Table 7 1. Peak annotation of AAT-enriched glycans. Glycan compositions were annotated based on spiking experiments with human plasma and literature. Compositions are given in terms of Hex: hexose, HexNAc: N-acetyl-hexosamine, Fuc: fucose, Sia: sialic acid.
Fraction no. Composition Spiking Literature
IgA-1 Hex5HexNAc4Sia2 X [209;210]
IgA-4 Hex5HexNAc5Sia2 X
IgA-5 Hex5HexNAc4Fuc1Sia2 X [209;210]
IgA-6 Hex5HexNAc5Fuc1Sia2 X [209;210]
IgA-7 Hex5HexNAc4Sia1 X [209;210]
IgA-8 Hex5HexNAc4Sia1 X [209;210]
IgA-9 Hex5HexNAc4Fuc1Sia1 X [209;210]
IgA-10 Hex5HexNAc5Fuc1Sia1 X [209;210]
IgA-11 Hex4HexNAc5Fuc1 X
IgA-14 Hex5HexNAc4Fuc1 X
IgA-15 Hex5HexNAc5Fuc1 X
Table 7 2. Peak annotation of IgA-enriched glycans. Glycan compositions were annotated based on spiking experiments with human plasma and literature. Compositions are given in terms of Hex: hexose, HexNAc: N-acetyl-hexosamine, Fuc: fucose, Sia: sialic acid.
7
Figure 7 1. Representative pool of electropherograms from enriched glycans. Electropherograms from the AAT-enriched pool are depiced in (A), while electropherograms from the IgA-enriched pool are de- piced in (B). Compositions and structural schemes are given in terms of N-acetylglucosamine (square), mannose (dark circle), galactose (light circle), sialic acid (diamond) and fucose (triangle).
Repeatability of the method
To assess the repeatability of the procedure for large scale protein enrichment and subsequent glycan analysis, the intra-batch as well as the inter-batch variation were calculated for both procedures (AAT and IgA). The intra-batch relative standard de- viation (RSD) for AAT, averaged over the 21 peaks was found to be 15.8 %. The intra-batch relative standard deviation (RSD) for IgA, averaged over 14 peaks was found to be 20.2 %. The inter-batch standard deviation was determined by compa- ring the results from the four experiments, and is depicted in Figure 7 2. The average inter-batch RSD for the 21 peaks from AAT was found to be 33.8 %. The average inter-batch RSD for the 14 peaks from IgA was found to be 32.1 %.
Figure 7 2. Repeatability of the enrichment procedure with subsequent analysis. Relative intensities toge- ther with their inter-batch standard deviations are depicted for AAT (A) and IgA (B).
7
Relation of protein glycosylation with calendar age and sex
Recent studies in human plasma have revealed that calendar age and sex influ- ence the plasma N-glycosylation pattern. As AAT and IgA are enriched from human plasma, it may be speculated that glycosylation of the enriched glycoproteins is also altered with calendar age, and influenced by sex.
AAT-glycosylation profiles could be obtained from 2271 individuals in the LLS, while IgA profiles were obtained from 2382 individuals. The participants in the LLS cover a large age-range (30-80 years of age) and after batch-correction, the effects of age and sex were evaluated in both datasets. The results are depicted in Table 7 3 and Table 7 4. Several effects can be observed: non-fucosylated triantennary structures in AAT (AAT-2, AAT-8, AAT-9 and AAT-16) are all negatively correlated to calendar age. Moreover, levels of most IgA derived glycan features are less abundant in ma- les than in females. Previous studies have reported that calendar age-associated changes in plasma glycosylation are usually more profound in females than in ma- les. Therefore, we investigated whether the interaction between calendar age and sex influences AAT- and/or IgA- glycosylation patterns, but no significant association with AAT- or IgA glycosylation features was recorded (data not shown).
Age Sexa
Coef P Coef P
AAT_1 + 0.000 - 0.000
AAT_2 - 0.000 - 0.000
AAT_3
AAT_4 - 0.000 - 0.000
AAT_5 + 0.004 - 0.000
AAT_6 + 0.058 + 0.000
AAT_7 + 0.172 - 0.000
AAT_8 - 0.002 - 0.000
AAT_9 - 0.000 - 0.000
AAT_10 + 0.056 + 0.000
AAT_11 + 0.000 + 0.000
AAT_12 - 0.000 - 0.129
AAT_13 - 0.000 - 0.000
AAT_14 - 0.104 + 0.854
AAT_15 - 0.000 - 0.000
AAT_16 - 0.002 - 0.000
AAT_17 + 0.004 - 0.000
AAT_18 - 0.000 - 0.000
AAT_19 + 0.032 + 0.000
AAT_20 + 0.000 + 0.000
AAT_21 - 0.017 - 0.075
AAT_22 - 0.000 - 0.682
Table 7 3. Calendar age and sex influences AAT glycosylation. The direction of the regression coefficient is depicted together with its P-value. Significant results are highlighted in bold (P ≤ 0.002, after Bonferroni correction). No results were obtained for AAT_3, as values were normalized to this glycan. a Female = 0, Male = 1
Age Sexa
Coef P Coef P
IgA_1 - 0.194 - 0.000
IgA _2 + 0.241 - 0.000
IgA _3 + 0.028 + 0.040
IgA _4 - 0.254 - 0.000
IgA _6 + 0.028 - 0.000
IgA _7 + 0.021 - 0.002
IgA _8 + 0.331 - 0.000
IgA _9 + 0.101 - 0.424
IgA _10 + 0.073 - 0.000
IgA _11 + 0.000 - 0.001
IgA _12 + 0.000 - 0.000
IgA _13 + 0.518 - 0.000
IgA _14 - 0.362 + 0.982
IgA _15 + 0.044 - 0.000
Table 7-4. Calendar age and sex influences IgA glycosylation. The direction of
the regression coefficient is depicted together with its P-value. Significant results are highlighted in bold (P ≤ 0.002, after Bonferroni correction). No results were obtained for IgA_5, as values were normalized to this glycan. a Female = 0, Male = 1
7
Protein glycosylation is affected by physiological parameters
In two recent studies, plasma glycosylation patterns were associated with altered lipid status and changes in BMI and glucose homeostasis [24]. Therefore, we eva- luated whether AAT- and/or IgA-glycosylation features were associated with BMI, plasma levels of cholesterol, HDL-cholesterol, LDL-cholesterol, triglycerides and glucose and insulin activity. While (almost) no associations could be observed for IgA glycosylation features (data not shown), several association could be reported for AAT-glycosylation features (see Table 7 5). Levels of HDL-cholesterol, BMI, insu- lin activity, and mainly levels of triglycerides are associated with several AAT-derived glycan features. This immediately raises the question whether the changes in AAT- glycosylation features may be associated with health or disease states.
BMI Cholesterol HDL-choles-
terol LDL-choles-
terol Triglycerides Glucose Insulin
Coef P Coef P Coef P Coef P Coef P Coef P Coef P
AAT_1 + 0.000 + 0.459 - 0.000 - 0.779 + 0.000 + 0.000 + 0.000
AAT_2 + 0.511 + 0.000 + 0.755 + 0.037 + 0.000 - 0.313 - 0.151
AAT_3
AAT_4 - 0.000 + 0.056 + 0.000 + 0.132 - 0.000 - 0.000 - 0.000
AAT_5 + 0.000 + 0.347 - 0.000 + 0.344 + 0.000 + 0.002 + 0.000
AAT_6 - 0.002 - 0.021 + 0.005 - 0.370 - 0.000 - 0.703 - 0.000
AAT_7 + 0.000 + 0.059 - 0.000 + 0.216 + 0.000 + 0.035 + 0.000
AAT_8 + 0.005 + 0.007 - 0.016 + 0.235 + 0.000 - 0.821 + 0.063
AAT_9 + 0.886 + 0.001 + 0.880 + 0.033 + 0.000 - 0.023 - 0.092
AAT_10 + 0.862 - 0.745 - 0.820 - 0.791 - 0.768 + 0.115 + 0.493
AAT_11 + 0.877 - 0.231 + 0.949 - 0.503 - 0.002 + 0.057 + 0.298
AAT_12 - 0.098 + 0.942 + 0.707 + 0.720 + 0.945 - 0.202 - 0.706
AAT_13 - 0.000 + 0.190 + 0.000 + 0.158 - 0.052 - 0.000 - 0.000
AAT_14 + 0.000 - 0.837 - 0.000 + 0.767 + 0.001 + 0.018 + 0.000
AAT_15 + 0.959 + 0.773 - 0.336 + 0.479 + 0.511 + 0.955 - 0.067
AAT_16 + 0.127 + 0.013 - 0.086 + 0.074 + 0.000 - 0.635 - 0.840
AAT_17 + 0.089 + 0.113 - 0.022 + 0.211 + 0.000 - 0.453 + 0.950
AAT_18 - 0.350 + 0.010 + 0.549 + 0.150 + 0.000 + 0.874 - 0.142
AAT_19 - 0.827 - 0.287 + 0.820 - 0.753 - 0.002 + 0.780 - 0.525
AAT_20 + 0.149 - 0.530 - 0.102 + 0.788 - 0.511 + 0.301 + 0.631
AAT_21 - 0.359 + 0.529 - 0.962 + 0.269 + 0.526 - 0.318 - 0.664
AAT_22 + 0.059 + 0.641 - 0.054 + 0.548 + 0.006 + 0.949 - 0.883
Table 7-5. AAT-glycosylation is associated with several biological parameters. The direction of the re- gression coefficient is depicted together with its P-value. Significant results are highlighted in bold (P ≤ 0.002, after Bonferroni correction).
Relation of protein glycosylation with familial longevity
The Leiden Longevity study was originally designed to study biological mechanis- ms that influence human life-span. By comparing the offspring of long-lived siblings with their age-matched partners, parameters that associate with familial longevity could be reported (e.g. [161;162;164;185]). Here we evaluated whether AAT and/
or IgA glycosylation patterns are associated with familial longevity. Using a logistic regression model, none of the glycosylation features showed a significant difference between the group of offspring and the group of partners (data not shown).
Altered protein glycosylation is associated with disease
The glycosylation features could not be associated with familial longevity, but sever- al associations were found between AAT glycosylation patterns and several physio- logical parameters (Table 7 5). This raises the question whether AAT-glycosylation features may reflect other disease states. High levels of plasma C-reactive protein (CRP) indicate an acute inflammatory response, and may be regarded as a marker for the presence of inflammatory diseases [193]. Moreover, plasma N-glycosylation was previously reported to be correlated to CRP levels [33]. Therefore, we assessed the relation between CRP levels and AAT-glycosylation features. The results are depicted in Table 7 6. A large amount of glycosylation features were found to be associated with CRP levels with high significance, indicating that AAT-glycosylation patterns are indeed altered in inflammatory diseased individuals.
7
CRP
Coef P
AAT_1 + 0.000
AAT_2 - 0.000
AAT_4 - 0.000
AAT_5 + 0.000
AAT_6 + 0.070
AAT_7 + 0.000
AAT_8 - 0.471
AAT_9 - 0.000
AAT_10 + 0.000
AAT_11 + 0.000
AAT_12 - 0.005
AAT_13 - 0.000
AAT_14 + 0.000
AAT_15 - 0.077
AAT_16 - 0.011
AAT_17 - 0.068
AAT_18 - 0.000
AAT_19 + 0.000
AAT_20 + 0.000
AAT_21 - 0.015
AAT_22 - 0.007
Table 7-6. Relation of AAT-glycosylation features with the occurrence of inflammatory disease.The direc- tion of the regression coefficient is depicted together with its P-value. Significant results are highlighted in bold (P ≤ 0.002, after Bonferroni correction).
As biophysical parameters indicating cardiovascular and metabolic disease could be associated to several AAT-glycosylation features (Table 7 5), an attempt was made to evaluate whether AAT-gly- cosylation may reflect such diseases. Even though the incidence of myocardial infarction (MI, 2,9%), cerebrovascular accident (CVA, 2.9%) and diabetes (5.1%) in the studied population is rather low, significant positive associations could be observed for glycan features AAT-16 and AAT-21 with the occurrence of myocardial infarction (Table 7 7), indicating that these glycan features might
indeed reflect this disease.
myocardial infarction
coef P
AAT-16 + 0.002
AAT-21 + 0.000
Table 7 7. Relation of AAT-glycosylation with the incidence of myocardial infarction. The direction of the regressio coefficient is depicted together with its P-value. Only significant results are depicted (P ≤ 0.002, after Bonferroni correction).
Discussion
We here present the first study on large scale protein enrichment using antibody affi- nity capturing with subsequent high-throughput N-glycan analysis. Such procedures could be performed using commercially available anti-IgA antibody coated beads (enrichment of IgA), but the use of in-house coated beads, where small VHH anti- bodies were immobilized on NHS activated Sepharose beads (enrichment of AAT), worked similarly well.
In the present study we have chosen to use a bead-based enrichment strategy, which is suitable for medium to high abundant glycoproteins; however, to allow en- richment of lower abundant glycoproteins from complex mixtures such as human plasma other platforms may have to be considered. Such platforms will need to further reduce aspecific binding by minimizing contact surfaces, in order to allow ef- ficient affinity-enrichment.
Coupling of proteins to beads is relatively easy, though larger amounts of antibo- dies (approximately 3 mg/ml of beads) are needed to allow efficient coupling and to limit non-specific binding of proteins to the beads. The selection of antibodies is important, and has turned out to be one of the limiting factors in this study. While lots of antibodies are commercially available, many of them are very expensive, or not suitable for immunopurification. The VHH-antibody against AAT that was used in this study had previously been tested for immunopurification, and was kindly provided
7
to us. While the anti-AAT antibody was shown to be highly specific, it was observed that the anti-IgA antibody is less specific. This resulted in a simple mixture of pro- teins, which hampers the interpretation of the data obtained in the study.
The intra-batch repeatability of the strategy described in this study is 15-20%. This value includes the variation due to the glycan release, labeling, purification and mea- surement of which the intra-batch variation was shown to be 6.2% (Chapter 4). As the inter-batch variation is even larger (around 33%), batch-corrections are needed to compensate for inter-batch variations when analyzing large sample sets.
Large scale enrichment of AAT and IgA from plasma samples from participants of the LLS using antibody affinity capturing followed by subsequent high-throughput N-glycan analysis, revealed that AAT-glycosylation patterns are associated with ca- lendar age and differ between females and males (Table 7 3). Moreover, several AAT-glycans could be associated with physiological parameters marking cardiovas- cular and metabolic diseases. Two non-fucosylated glycans (AAT-16 and AAT-21) could be shown to be positively related to the incidence of myocardial infarction (Table 7 7).
While several parameters could be related to AAT-glycosylation features, the IgA glycosylation pattern seems to be almost unaffected. The only difference in glyco- sylation patterns that could be observed was between females and males. As the IgA protein fractions were not free of contaminating proteins, and IgG was one of the proteins observed by LC-MS/MS of the enriched protein samples, it may be specu- lated that some IgG glycans are present in the IgA enriched glycosylation pattern.
Glycans that could originate from IgG are indeed observed (IgA-13, IgA-14 and IgA- 15), but these glycans could not be related to calendar age, while IgG glycosylation was previously shown to be highly associated with calendar age [40-42;185]. This argues against large influences of IgG glycans in the IgA enriched-glycosylation pat- terns that were observed in this study.
The group of Prof. Y. Wada recently published a study on O-glycosylation of IgA,
and compared O-glycan profiles from IgA1 from 26 patients with Rheumatoid arthritis (RA) with profiles from nine healty individuals. Decreased levels of GalNAc were observed in patients with RA, and a positive correlation was observed between galactosylation of IgG glycans and N-acetyl-galactosylation of IgA O-glycans. N- glycosylation patterns of IgA were not studied, but in a previous study, no significant difference in the degree of galactosylation of IgA N-glycans was observed in patients with RA [211]. This indicates that IgG N-glycosylation is differently regulated than IgA N- glycosylation. No age-related alterations in the IgA glycosylation profile were observed, while we did observe associations between IgG N-glycosylation patterns and calendar age in the LLS cohort (Chapter 6), thus confirming the hypothesis that the regulation of IgG N-glycosylation is different compared to the regulation of IgA N-glycosylation.
In this Chapter, it was shown that a strategy consisting of large-scale immuno-affinity capturing of proteins from human plasma using a bead-based method, coupled with high-throughput N-glycan analysis using multiplexed CGE-LIF is a powerfull tool for the analysis of N-glycosylation patterns of specific glycoproteins in large studies.
The strategy has successfully been applied to approximately 2400 plasma samples from the LLS and it could be observed that AAT glycosylation patterns are regulated by all kinds of biological precesses, while the glycosylation pattern of IgA was not affected by most parameters tested.
Acknowledgements
We would like to thank Dr. J. Stam from Utrecht University for providing us with anti- AAT-coated beads and ms. E. Steenvoorden for technical assistance. This work was supported by IOP Grant IGE05007.
