Serum glycopattern and Maackia amurensis lectin-II binding glycoproteins in autism spectrum disorder

The pathophysiology of autistic spectrum disorder (ASD) is not fully understood and there are no diagnostic or predictive biomarkers. Glycosylation modified as many as 70% of all human proteins can sensitively reflect various pathological changes. However, little is known about the alterations of glycosylation and glycoproteins in ASD. In this study, serum glycopattern and the maackia amurensis lectin-II binding glycoproteins (MBGs) in 65 children with ASD and 65 age-matched typically developing (TD) children were compared by using lectin microarrays and lectin-magnetic particle conjugate-assisted LC-MS/MS analyses. Expression of Siaα2-3 Gal/GalNAc was significantly increased in pooled (fold change = 3.33, p < 0.001) and individual (p = 0.009) serum samples from ASD versus TD children. A total of 194 and 217 MGBs were identified from TD and ASD sera respectively, of which 74 proteins were specially identified or up-regulated in ASD. Bioinformatic analysis revealed abnormal complement cascade and aberrant regulation of response-to-stimulus that might be novel makers or markers for ASD. Moreover, increase of APOD α2-3 sialoglycosylation could sensitively and specifically distinguish ASD samples from TD samples (AUC is 0.88). In conclusion, alteration of MBGs expression and their sialoglycosylation may serve as potential biomarkers for diagnosis of ASD, and provide useful information for investigations into the pathogenesis of ASD.


Results
Alteration of Glycopattern in Sera from ASD versus TD. The layout of the lectin microarray, and the resulting glycopatterns of serum glycoproteins defined by the microarrays for the ASD and TD groups are shown in Fig. 2A,B. The original data were imported into EXPANDER 6.0 for hierarchical clustering analysis (Fig. 2C). The normalized fluorescent intensities (NFIs) and the sugar-binding specificities for each of the 37 lectins from the two groups are summarized in Table S1. As a result of differential analysis, five lectins showed significant differences between ASD and TD groups. MAL-II (Siaα 2-3 Gal/GalNAc) and MAL-I (Siaα 2-3Galβ -1,4GlcNAc and Galβ -1,4GlcNAc) showed the most significantly increased NFIs (fold change = 3.33 and 2.20, p < 0.001 and p = 0.030), and ACA and PNA (Galβ 1-3GalNAcα -Ser/Thr (T) and sialyl-T(ST)) showed also significantly increased NFIs (fold change = 1.65 and 1.90, p = 0.0014 and 0.046) in the ASD versus the TD group (Fig. 2D). However, STL (trimers and tetramers of the GlcNAc) showed significantly decreased NFIs (fold change = 0.54, p = 0.0057) in the ASD versus the TD group (Fig. 2D). To validate the different abundance of certain kinds of glycans between ASD and TD, the lectin blotting was performed with MAL-II and ACA in the pooled TD (n = 50, lane 1) and ASD (n = 50, lane 2) sera. The result of SDS-PAGE showed that sera proteins from ASD and TD were similar in their molecular weight. The lectin blotting analysis showed a total of nine apparent bands (b1-b9) and several minor bands belonging to different molecular weight ranging from 7 to 175 kDa (Fig. 2E). Similarly but differently, MAL-II and ACA showed stronger binding to glycoprotein bands (b2, b4, b6, and b9 for MAL-II, b3, b4, b6, b7, and b9 for ACA) in the ASD than the TD sera. In addition, the results of serum microarray revealed that expression of Siaα 2-3 Gal/GalNAc recognized by MAL-II was significantly increased (p = 0.0009) in individual serum samples from ASD versus TD, which were coincident with the results of the lectin microarrays (Fig. 2F).

Identification of MBGs.
Because of the significant increase of Siaα 2-3 Gal/GalNAc expression (α 2-3 linked sialoglycosylation) on glycoproteins in sera from ASD versus TD groups, two new questions were then raised: firstly, whether the Siaα 2-3 Gal/GalNAc or the α 2-3 linked sialoglycosylated proteins were increased, and secondly, what kinds of proteins were α 2-3 linked sialoglycosylated and what are their potential biological functions in ASD sera. To solve these problems, MMPCs were applied to isolate MBGs from the pooled sera in two groups. The isolated protein fractions were analyzed by SDS-PAGE (Supplementary Figure S1). The eluted protein factions were slightly darker for ASD (green frame) than TD sera (blue frame). The peptide mixtures were identified in triplicate to reduce variances for individual proteins. A total of 1081 (corresponding 194 glycoproteins) and 1248 (corresponding 217 glycoproteins) unique peptides were identified from TD and ASD sera, respectively (Table S2). Notably, 983 (75.0%) of the peptides, corresponding 168 (69.1%) proteins, were common to both sera, whereas 26 proteins and 49 proteins were specially identified in TD and ASD sera respectively (Fig. 3A,B). By mapping to the UniProtKB/Swiss-Prot database, 213 of the identified proteins were known proteins, of which 146 proteins (68.5%) were known N-glycoproteins (N Y ) and 45 proteins (21.1%) were O-glycoproteins (O Y ) (Table S2 and Fig. 3C). Other 90 proteins including 30 unknown proteins (cannot search in UniProt database) and 60 "unproven" proteins (without glycosylation information by UniProt database) were novel identified glycoproteins in the study (Table S2). According to two online glycosylation site prediction servers (NetNGlyc 1.0 32 and NetOGlyc 4.0 33 ), 35 of the 90 proteins were predicted to have potential N-glycosylation sites (N P ), and 79 were predicted to have potential O-glycosylation sites (O P ). In addition, 106 known N-glycoproteins were also predicted to have potential O-glycosylation sites (O P ) (Table S2 and Fig. 3C). There were still seven non-glycoproteins (Table S2). A spectral index (SI) based on the spectra and peptide counts was calculated to compare protein expression level between TD and ASD sera. Totally, 25 proteins (e.g., apolipoprotein D [APOD] and complement component C8 [C8B]) were up-regulated (ratio ≥ 1.5), while 23 proteins (e.g., complement C1q subcomponent subunit A [C1QA] and Neuropilin-1 [NRP1]) were down-regulated (ratio ≤ 0.67) in ASD relative to TD sera (Table S2 and Supplementary Figure S1). Comparing the molecular weights (MW) and isoelectric points (pI) of the MBGs, the majority had MWs lower than 200 kDa and pI values lower than 7 in both TD and ASD sera (Supplementary Figure S1).

Gene Ontology Analysis of the MBGs.
To investigate the major biological functions of the identified MBGs, Blast2GO was applied to analyze them for functional enrichment according to three grouping classifications: cellular components, biological processes, and molecular functions. Of the 243 identified MBGs, 216 gene ontology (GO) annotations were available (Supplementary Figure S1). In the cellular component group, 150 proteins (61.2%) were extracellular region proteins, and 90 proteins (36.7%) were membrane proteins. In biological processes, 152 proteins (62.0%) were involved in single-organism processes and 136 proteins (55.5%) were involved in response-to-stimulus processes. In terms of molecular function, proteins with binding ability formed the largest group (146, 59.6%), and other smaller groups identified included enzyme regulatory activity (38,15.5%) and catalytic activity (38,15.5%). To analyze potential differences in the GO annotations between ASD and TD serum samples, GO enrichment analysis was performed. ASD serum samples were enriched versus TD serum samples in annotations including: fibrinogen complex, lyase, transposase, chromosome segregation, and maintenance of location in cell. On the other hand, ASD serum samples were depleted versus TD serum samples in annotations including: cell leading edge, nucleoside-triphosphatase regulator, peptide binding, and developmental growth (Supplementary Figure S1).

KEGG Pathway and Protein Interaction Network Analysis.
In total, 184 of 243 identified MBGs were annotated in DAVID Bioinformatics Resources (version 6.7). These MBGs were mapped to 6 KEGG pathways with thresholds of count ≥ 5 and a P-value < 0.05 versus the background signal of the human genome; the identified KEGG pathways included complement and coagulation cascades, systemic lupus erythematosus, ECM-receptor interaction, and others (Table S3). A total of 38 MBGs were involved in complement and coagulation cascades, of which most MBGs (e.g., carboxypeptidase B2 [CPB2], kininogen-1 [KNG1], and complement C5 [C5]) were up-regulated except that vitamin K-dependent protein S (PROS1), alpha-2-antiplasmin (SERPINF2), and complement factor H (CFH) were down-regulated in ASD sera compared to TD sera (p = 3.79E-54) (Fig. 3D). In addition, 217 matched MBGs were queried against the STRING Homo sapiens database to determine their functional relevance. Through enrichment analysis of biological processes, 18 versus 5 of the 49 proteins responsible for positive regulation of response to stimulus (p = 1.53E-18) exhibited decreased versus increased expression (Fig. 3E), meanwhile 11 versus 4 of the 39 proteins responsible for negative regulation of response to stimulus processes (p = 2.76E-15) showed increased versus decreased expression ( Fig. 3F) in ASD sera.

Sequence Motif Preference of MBGs.
Typically, N-glycosylation occurs at N-X-S/T motifs (where X cannot be proline) in mammals. Our data set provided a good basis to test the generality of this motif and to identify Images of Cy3-labeled sera proteins from TD and ASD bound to the lectin microarrays. Fluorescent images were scanned with a 70% photomultiplier tube and 100% laser power settings in a Genepix 40 00B confocal scanner. A portion of the slide with three replicate lectin arrays is shown. The lectins exhibited significant differences, marked with white frames. (C) Hierarchical clustering analysis of NFIs for the 37 lectins from TD-1~5 and ASD-1~5. The samples are listed in columns, and lectins are listed in rows. The color and intensity of each square indicates expression levels relative to the other data in the row. Red, high; green, low; black, medium. Yellow and blue frame marked higher and lower binding intensity of lectin in ASD vs. TD sera. (D) Differential analysis of NFIs for five lectins from TD-1~5 and ASD-1~5. NFIs of each lectin for the TD and ASD groups were compared according to the following criteria: fold change ≥ 1.5 or ≤ 0.67 indicated up-regulation or downregulation. Differences for each lectin between TD and ASD groups were further tested by Paired student's t-test using SPSS Statistics 19 (p < 0.05). (E) Binding patterns of glycoproteins in pooled sera from TD (n = 50, lane 1) and ASD (n = 50, lane 2) samples for MAL-II (middle) and ACA (right). Gels were stained directly with alkaline silver as control (left). The average gray value for each lane was from Image pro-Plus 6.0 analysis and compared between two groups (*P < 0.05, **P < 0.01, and ***P ≤ 0.001). (F) Validation of Siaα 2-3 Gal/GalNAc expression in individual serum samples using serum micorarrays. Fluorescent images were scanned with the 50% photomultiplier tube and 100% laser power settings using a LuxScan 10 K Microarray Scanner. Scatter plot analysis of the original data achieved from the serum microarrays. Statistical significance of differences between groups was indicated by the p-value.
further consensus sequences. Notably, 12 specific nonredundant consensus sequences with a high motif score and fold increase > 30 were identified (Supplementary Figure S1). The position-specific amino acid frequencies of the surrounding asparagine residues (13 amino acids to both termini) were compared, and the motif [AVH] [KR]xNxxNxSxxxY (where "x" denotes any residue, [AVH] and [KR] represent several amino acid residues that might appear in the position, and the bullet point denotes a possible glycosite) was identified as a possible N-glycosylation motif around asparagine (Fig. 3G). Interestingly, xxxxxxQSDxxYK and xxxxxxHGSxSGx motifs were significantly overrepresented (fold increase = 81.62 and 67.07) in the MBG data (Fig. 3G), which might represent O-linked glycosylation motifs around serine residues for the α 2-3-linked sialylated glycopeptide domain. However, to further confirm the O-glycosites in MBGs, it still needs much more in-depth studies.

Expression and Sialoglycosylation of MBGs in Individual Serum Samples. A western blot was
performed to verify the expression of C8B, serotransferrin (TF), C1QA, and APOD in individual serum samples. As a result, expression of C8B and TF were increased and expression of C1Q was decreased in four tested ASD samples compared to four TD samples, which were consistent with results of MS (Fig. 4A). However, expression of APOD was not significantly different between TD and ASD samples (Fig. 4A).
LGAMs were designed to detect α 2-3 linked sialoglycosylation of C8B, TF, C1QA, and APOD in 15 TD and 15 ASD individual serum samples (Fig. 4B). As a result, no significant differences for C8B, TF, and C1QA sialoglycosylation were detected between two groups. However, α 2-3 siologlycosylation of APOD was significantly increased in ASD samples relative to TD samples (p = 0.004) (Fig. 4C). ROC curve analysis revealed that serum levels of α 2-3 sialoglycosylated APOD resulted in an AUC of 0.88, with a specificity of 86.7% and a sensitivity of 80.6% for differentiating ASD from TD) (Fig. 4D).

Discussion
Lectins are carbohydrate-binding proteins that are neither antibodies nor enzymes, which have a wide range of glycan-binding specificities. These characteristics make them suitable for the characterization of a glycome of cell, tissue, serum, saliva, and so on. In this study, expression of Siaα 2-3 Gal/GalNAc (MAL-II) showed the most significant increase (fold change = 3.3, p < 0.001) in sera from ASD versus TD (Fig. 2D), and therefore, the intriguing MBGs were further captured by MMPCs from ASD and TD sera and identified by LC-MS/MS. MAL-II is a known glycoprotein, so it was possible that a few glycan-recognition proteins bound to glycans on MAL-II be pulled down together using the MMPCs. This factor, together with nonspecific protein adsorption, resulted in the identification of "non-glycoproteins" by MS (Table S2). To eliminate the impact of these two problems as effectively as possible, an optimum binding buffer conducive to high affinity interactions between the carbohydrate-binding domain of MAL-II and α 2-3 linked sialic acids on MBGs was used, together with appropriate denaturing by washing with 0.1% (v/v) Tween 20 several times.
There was a notable phenomenon that the binding patterns of MAL-II and ACA to the glycoproteins in sera of ASD and TD were extremely similar according to lectin blotting (Fig. 2E). Now that MAL-II is proven to bind specifically to α 2-3 sialic acid on T antigen 34 , it could be speculated that increased NFIs of MAL-II bound to ASD sera vs. TD sera was actually resulted from higher expression of α 2-3 sialosyl-T antigens on MBGs. In the subsequent protein identification and characterization, 46 of 243 MBGs were known O-linked glycoproteins, and 145 MBGs were potential O-glycosylated proteins. Sequence motif preference analysis also indirectly indicated that xxxxxxHGSxSGx and xxxxxxQSDxxYK motifs surrounding serine residues significantly overrepresented in MBGs might be potential O-linked glycosylation motifs (Fig. 3G). A previous study found that no changes in the plasma N-glycome were associated with ASD using hydrophilic interaction high performance liquid chromatography 35 , which was complemented in this study. However, although these data provided clues for predicting glycan structures of MBGs, the fact still needed to be experimentally proved by possibly employing advanced glycomic techniques, such as matrix-assisted laser desorption ionization time-of-flight mass spectrometry 36 .
Recent interest in profiling the glycome stems from the potential of glycans as disease markers 37,38 . With glycans as disease markers there are several intrinsic advantages compared to other biomolecules, specifically proteins: (1) glycan biosynthesis is more significantly affected by disease states than protein production. (2) Aberrant glycosylation can potentially affect nearly every glycoprotein produced in the diseased cell. (3) Given the current technology, it is far simpler to quantitate oligosaccharide expression than protein expression 39 . Analysis of glycan on protein involves several levels of complexity, which includes simple compositional profile, glycan structure, protein-specific glycosylation, and the site-specific glycosylation (with increasing complexity) 38 . In this study, the glycopattern was detected firstly by using lectin microarrays, and then the glycan associated proteins were further isolated and identified based upon LC-MS/MS. It is obvious that even the same glycan on various glycoproteins may play different roles that depend on functions of glycoproteins themselves in disease 40 . Therefore, this study revealed not just the altered glycans but also the glycan associated proteins so as to demonstrate the capacity of altered glycosylation of protein as biomarkers for ASD diagnosis and provide further information for investigations into the mechanisms of ASD. Previous studies showed that human serum N-glycan profiles are age and sex dependent 41 . In this study, to minimize the effects of inter-and intra-patient variations, 65 children with ASD and 65 age-matched TD children with similar sex ratio were enrolled. Then, 50 TD and 50 ASD serum samples were pooled respectively for lectin microarray and LC-MS/MS identification, and other 15 TD and 15 ASD samples maintained individually were used for LGAMs detection.
Many researchers have repeatedly described immune dysfunction in ASD, symptoms of which include neuroinflammation, the presence of autoantibodies, increased T cell responses, and enhanced innate NK cell and monocyte immune responses 42 . The complement system is a part of the immune system that helps or complements the ability of antibodies and phagocytic cells to clear pathogens from an organism 43 . In recent years, studies have shown that complement cascade, a major effecter arm of the innate immune system, is almost certainly involved in synaptic remodeling by tagging destined neurons and synapses for destruction 44 . In addition, developing astrocytes release signals that induce the expression of complement components in the central nervous system (CNS). In the mature brain, early synapse loss is a hallmark of several neurodegenerative diseases. Complement proteins are profoundly upregulated in many CNS diseases prior to signs of neuron loss 45 . Therefore, the abnormal complement cascade displayed in this study might be a pivotal manifestation mode of immune dysfunction in ASD (Fig. 3D). Besides, this study found that almost all MBGs (e.g., NRP1 and CFHR5) responsible for positive regulation of response-to-stimulus processes were down-regulated, and most MBGs (e.g., FGB and APOD) responsible for negative regulation of response-to-stimulus processes were up-regulated in the ASD sera (Fig. 3E,F), which might be one important maker or marker for ASD, and provide useful information for further in-depth investigations of the pathogenesis and treatment of ASD.
Typically, sialic acid is found as a component of the oligosaccharide chains of mucins, glycoproteins, and glycolipids occupying terminal, nonreducing positions of N-or O-glycans. Sialic acid levels in serum are associated with liver diseases 46 , rheumatic diseases 47 , and type-2 diabetes 48 . In this study, western blot analysis validated the alteration of C8B, TF, and C1QA expression, but not APOD expression, in individual ASD and TD serum samples (Fig. 4A).
LGAMs revealed significantly increased expression of α 2-3 sialoglycosylation of APOD in individual ASD serum samples, which greatly explained the no difference in expression of APOD protein between ASD and TD sera, and emphasized that both MBGs and their α 2-3 sialoglycosylation were associated to ASD. ROC curve analysis noted that sialoglycosylated APOD could sensitively and specifically distinguish ASD from TD children as candidate biomarkers (AUC = 0.88), and indicated the importance and necessary of studying the alteration of glycoproteins glycosylation in sera for diagnose of ASD.
In conclusion, expression of α 2-3 sialosyl-T antigens was significantly increased in sera of ASD versus TD. A total of 194 and 217 MBGs were identified from TD and ASD sera respectively, of which 74 proteins were specially identified or up-regulated in ASD sera. Bioinformatic analysis revealed that abnormal complement cascade and aberrant cellular regulation of response-to-stimulus might be novel makers or markers for ASD, which provide novel information for further in-depth investigations into the pathogenesis of ASD. More importantly, LGAMs revealed significantly increased expression of α 2-3 sialoglycosylation of APOD in individual ASD serum samples, which might serve as potential biomarkers for diagnosis of ASD. Subjects. Sixty-five children with ASD and 65 age-matched TD children between 2.5 and 6 years of age were enrolled. Children in ASD group were recruited from Xi'an Children's Hospital, the First Affiliated Hospital of Xi'an Jiaotong University, and the Second Affiliated Hospital of Xi'an Jiaotong, Xi'an, China. All children with ASD were examined by clinical experts on autism. A developmental behavioral pediatrician and a pediatric neurologist or psychiatrist examined all the children. All consultants agreed on the diagnosis of ASD according to DSM-V criteria 49 . Subjects with tuberous sclerosis complex, Rett syndrome, Prader Willi syndrome, Angelman syndrome, or Fragile X syndrome were excluded. All participants were screened via a parental interview for current and past physical illness. The control group consisted of healthy TD children recruited from the same area to minimize the influence of different environments. Children in both the ASD and TD groups who had any type of infection or disease less than 2 weeks before the time of examination were excluded. Intelligence quotient was measured using the Gesell Development Schedule. ASD was evaluated with the autism diagnostic observation schedule (Table 1 and Table S4).
Approval for this research was obtained from the Ethics Committee and the Human Research Review Committee of Xi'an Jiaotong University (Xi'an, China). All parents of the participants enrolled in the study provided written informed consent. All experiments were carried out in accordance with the approved guidelines.
Sample Collection and Preparation. All blood samples were collected by a pediatric nurse and venous blood was collected. The blood was allowed to clot at room temperature for 25 min. The clot was then removed by centrifuging at 1, 500 g for 10 minutes in a refrigerated centrifuge. The resulting supernatant is immediately transferred to a clean polypropylene tube added with EDTA-free inhibitor cocktail (Halt protease inhibitor; Thermo Scientific Pierce Protein Research Products, Rockford, IL, USA) at a concentration of 10 μ L/mL serum. The produced serum was aliquoted into small portions and immediately frozen on dry ice and stored at − 80 °C. To normalize the differences between subjects and to tolerate individual variation, 50 μ L of 50 serum samples from TD and ASD groups were pooled respectively for lectin microarray and LC-MS/MS detection. The other 15 samples from each group were maintained individually for further validation.
Lectin Microarray and Data Analysis. The lectin microarray was produced and incubated with Cy3 fluorescent dye (GE Healthcare) labelled serum proteins according to our previous protocol [50][51][52][53] that are described in the Supplementary Materials and Methods in detail. Fifty TD and 50 ASD serum samples were used for lectin microarray detection. Twenty microliter (20 μ L) from each sample and 10 samples in a pool were prepared to form TD-1~5 and ASD-1~5 subgroups. The acquired images were analyzed at 532 nm for Cy3 detection using Genepix 3.0 software. The averaged background was subtracted, and values less than the average background ± 2 standard deviations (SD) were removed from each data point. The median of the effective data point for each lectin was globally normalized to the sum of the median of all effective data points for each lectin in a block. Each sample was observed consistently with three repeated slides, and the normalized median of each lectin from 9 repeated blocks was averaged and the SD determined. Normalized data for the TD and ASD groups were compared according to the following criteria: fold change ≥ 1.5 or ≤ 0.67 indicated up-regulation or down-regulation. Differences between the two arbitrary data sets were tested by Paired student's t-test using SPSS Statistics 19. The original data were further analyzed with Expander 6.0 (http://acgt.cs.tau.ac.il/expander/) to perform a hierarchical clustering analysis.

Serum Microarray and Data Analysis.
A serum microarray was produced by using 30 individual serum samples from 15 TD and 15 ASD children each. The Cy3-labeled MAL-II was applied to detect the specific sugar structure in the minimal amount of serum samples that immobilized on the slides according to the fabrication protocol of saliva microarray 51 with some modifications. Detailed information is provided in the Supplementary Materials and Methods.
Isolation and Digestion of MBGs. MAL-II-magnetic particle conjugates (MMPCs) were prepared as described 54,55 . Two milligrams (~30 μ L, measured with Bradford reagent) of protein from pooled TD and ASD sera were incubated with the MMPCs 54,55 . The obtained glycoproteins (about 150 μ g) were digested by trypsin and PNGase F as described previously [54][55][56] . Detailed information is provided in the Supplementary Materials and Methods.

Label-Free Relative Quantification by Spectral Index Calculation. After peptide identification, an
algorithm similar to the ProteinExtractor in ProteinScape, which uses a given minimal peptide score (minPep-Score) and minimal peptide count per protein (minNrPeps), was applied as described 57 . Among the listed proteins, every peptide spectrum match (PSM) was extracted. A spectral index (SI) based on spectral and peptide counts was calculated as described previously 58 . The raw spectral counts for identified proteins were normalized using the following formulas (Formula 1 and Formula 2): where C i is the total spectral count of run i; and C is the averaged total spectral count of all the runs under comparison; N i and R i are the normalized and raw spectral counts of a protein in run i, respectively. The SI, C/C i , was used to normalize the total spectral count of each run to reduce run-to-run variability.
Data Mining and Bioinformatics. Data analysis and professional softwares used in this study are described in the Supplementary Materials and Methods in detail.
Lectin/Glyco-Antibody Microarrays and Data Analysis. The lectin/glyco-antibody microarrays (LGAMs) analysis was designed as described 55,59 previously with some modifications. Briefly, rabbit polyclonal antibodies for human C8B, TF, C1QA, and APOD were spotted onto the homemade epoxysilane-coated slides with Stealth microspotting pins (SMP-10B) (TeleChem; Atlanta, GA) using a Capital Smart Arrayer (CapitalBio; Beijing, China). Dilution buffer and BSA were negative controls. Each antibody was printed in quintuplicate per block with triplicate blocks on one slide. Slides were immobilized in a humidity-controlled incubator at 50% humidity overnight. To prevent subsequent interference from glycans on antibodies, the printed slides were oxidized with 200 mM NaIO 4 solution at room temperature (18-22 °C) for 30 min in the dark to remove all glycans. Slides were then immersed in 1 mM 4-hydroxybenzhydrazide in dimethylformamide at room temperature for 2 h to derivatize the carbonyl groups. After blocking with 1× Carbo-Free TM Blocking Solution (diluted with PBST) for 1 h, 20 μ L of serum sample (diluted 1:10 in Carbo-Free TM Blocking Solution with 1% BSA) was applied to the antibody microarrays and rotated in a humidified chamber at 4 °C overnight. The slide was then rinsed with PBST and PBS to remove unbound proteins, and incubated with Cy3 fluorescent dye labeled MAL-II solution and rotated at 37 °C for 2 h. After a final wash, the slides were dried and scanned with a Genepix 4000B microarray scanner (Axon Instruments, CA, USA). Genepix Pro 3.0 was used to extract the spot data. The average background was subtracted, and values less than the average background ± 2 SDs were removed from each data point. The median of the effective data points of each antibody for one sample was calculated. Differences between medians of the TD and ASD groups (n = 15) for each antibody were tested by Paired t-test using SPSS Statistics 19. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the potential use of sialoglycosylation of MBGs as biomarkers of ASD.