The collection and use of the whole saliva for research in this study were approved by the Ethical Committee of Northwest University (Xi’an, China) and the Second Affiliated Hospital of Zhengzhou University (Zhengzhou, China). Written informed consent was received from participants. And this study was conducted by the ethical guidelines of the Declaration of Helsinki. After a standardized endoscopic procedure and histopathological evaluation, the individuals diagnosed with primary ESCC (n =16) were enrolled in this study. The age- and sex-matched HVs (n = 25) were recruited from the health checkup center in the same hospital where they underwent a medical examination. All the participants had no significant differences between the two groups regarding demographic, socioeconomic, and lifestyle characteristics, and those who received preoperative radiotherapy, chemotherapy, chemoradiotherapy, or antibiotic therapy were excluded from this study. The clinical characteristics of HVs and ESCC patients were summarized in Table 1.

The collection protocol has been described in previous literature (Liu et al., 2018; Shu et al., 2018). The collected whole saliva was centrifuged at 10,000 g at 4°C for 15 min to remove insoluble components. And the cocktail of protease inhibitor (Sigma-Aldrich, United States) was added to the collected supernatant according to the manufacturer’s recommendations and lyophilized at −80°C until use.

The lectin microarrays were produced using 37 lectins with different binding preferences covering N- and O-linked glycans (Shu et al., 2017; Yu et al., 2020b). The Cy3-labeled glycoproteins were incubated on the lectin microarray with gentle rotation in the dark (37°C, 3 h). The microarrays were washed with PBST and PBS, respectively, centrifuged dry, and scanned immediately using a Genepix 4000B confocal scanner (Axon Instruments, United States). The generated images were analyzed by Gene pix software (version 6.0, Axon Instruments Inc., Sunnyvale, CA). The unsupervised average hierarchical cluster analysis (HCA) and principal component analysis (PCA) were performed by Expander 6.0 (http://acgt.cs.tau.ac.il/expander/) and Multivariate Statistical Package (United Kingdom), respectively. And the p value was derived from the nonparametric Mann–Whitney test using the GraphPad Prism software (Version 7.0, GraphPad Software, Inc., San Diego, CA).

The expression levels of glycan structures were analyzed by lectin blotting as described previously (Zhang et al., 2019; Yu et al., 2020b). The pooled salivary protein was separated by 10% SDS-PAGE and transferred onto PVDF membrane (0.22 mm Millipore, Bedford, MA, United States). The PVDF membrane was blocked by the Carbo-Free Blocking Solution (Vector Labs, Burlingame, CA) and incubated with Cy5-labeled DSA, ECA, LCA, PSA, and UEA-I, respectively. The membrane was scanned by the STORM FluorImager (Molecular Dynamics, Sunnyvale, CA, United States) and measured using the ImageJ software (NIH).

The epoxy-coated magnetic particles (2 mg) were rinsed with ethanol and coupling buffer (5 mM NaB₄O₇, 180 mM H₃BO₄, 150 mM Na⁺, pH 7.4), respectively, and reacted with DSA solution (DSA dissolve in coupling buffer) according to the protocol (Zhang et al., 2019; Yang et al., 2020b). Then, conjugates were washed with a coupling buffer to remove the unbound lectins.

The GlcNAc and Galβ1-4GlcNAc-containing glycoproteins were isolated from ESCC patients and HVs using DSA-magnetic particle conjugates as described previously (Zhang et al., 2019; Yang et al., 2020b). Briefly, the pooled salivary protein was diluted with the binding buffer (100 Mm Tris-HCl; 150 mM NaCl; 1 mM CaCl₂, MgCl₂, and MnCl₂; pH 7.4) and incubated with the DSA conjugates (room temperature, 3 h). After 3 h gentle shaking, the conjugates were washed using washing buffer (0.1% Tween-20 in binding buffer, pH 7.2) to remove unbound proteins, and the glycoproteins bound to the conjugates were eluted with an eluting buffer (8 M urea, 40 mM NH₄HCO₃).

The isolation and purification of N-glycans were performed based on previously described methods (Qin et al., 2017; Yang et al., 2020b). The glycoproteins were concentrated and desalted by Amicon Ultra-0.5 3 KDa ultrafiltration units (Millipore, United States) and then denatured by the addition of 8 M urea, 10 mM DTT, and 10 mM IAM. The denatured glycoproteins were exchanged to 40 mM NH₄HCO₃ buffer and incubated with trypsin (37°C, overnight). The trypsin in the mixture was inactivated by heating (80°C, 5 min) and then the PNGase F (New England Biolabs, Beverly, MA) was added to release the N-glycans (37°C, overnight). Subsequently, the digest was subjected to HyperSep Hypercarb SPE cartridges (25 mg, 1 mL; Thermo Scientific) to remove peptides. The purified N-glycans were collected and lyophilized.

The N-glycans were characterized by matrix-assisted laser desorption ionization time-of-flight/time-of-flight mass spectroscopy (MALDI-TOF/TOF-MS, UltrafleXtreme, Bruker Daltonics; Bremen, Germany) as described previously (Qin et al., 2017; Zhang et al., 2019; Yang et al., 2020b). Glycans were resuspended and spotted onto an MTP AnchorChip sample target. Then, 2 μL of 10 mg/mL 2,5-dihydroxybenzoic acid (DHB) with 1 mM NaCl in 50% (v/v) methanol solution was spotted to recrystallize the N-glycans and vacuum dried for analysis. Peptide calibration standards (250 calibration points; Bruker) were used as mass calibration. Positive ion reflection mode was performed, and a mass range of 1,000–4,000 Da was analyzed. Representative MS spectra of N-glycan mass peaks with signal-to-noise ratio above three were generated and annotated using FlexAnalysis and GlycoWorkbench software.

A total of 16 ESCC and 25 HV samples were detected by lectin microarrays independently to investigate the altered salivary protein glycopatterns associated with ESCC patients. The layout of the lectin microarrays and typical representative images were listed in Figure 1A,B. HCA and PCA were executed to evaluate the glycan expression profiles of ESCC and HV cases and to provide a graphical representation of relationships among the subjects and diseases. The normalized fluorescent intensities (NFIs) for each lectin were distributed in the heat map by unsupervised clustering method to achieve the hierarchical relationship of the samples based on similarities in their glycan expression pattern. As shown in Figure 1C, 16 ESCC cases were classified into one category, and 25 HV cases were classified into another class, indicating that the salivary glycopatterns identified by these lectins were different between HVs and ESCC patients. A similar result was obtained by PCA as well; the subjects clustered separately by principal components 1 and 2 to form HV and ESCC clusters with different colors and symbols in Figure 1D, representative of the salivary glycopatterns differences for the two groups in a manner.

To further investigate the alteration of salivary protein glycopatterns in ESCC patients, the Mann–Whitney test was used to compare the variance between ESCC and HV groups. In total, 13 lectins exhibited significantly altered NFIs in ESCC cases. As shown in Figure 2 and Table 2, the Galβ1-3/4GlcNAc binder LEL and MAL-I, Galα1-3GalNAc binder BPL, T antigen and sialyl-T antigen binder Jacalin, Siaα2-3/6Gal binder MAL-II/SNA, GlcNAc and Galβ1-4GlcNAc binder DSA, and GSL-II showed increased NFIs in ESCC salivary glycoproteins against HV cases. On the contrary, the high-mannose binder of ConA and GNA, Fucα-1,6GlcNAc binder LCA and PSA, and Fucα1-2Galβ1-4Glc (NAc) binder UEA-I showed decreased NFIs in ESCC salivary glycoproteins compared with HV cases. As a result, ESCC patients showed higher levels of GalNAc and Gal expression profile, sialic acid expression profile, GlcNAc expression profile, and lower levels of mannose expression profile and fucose expression profile.

SDS-PAGE and lectin blotting analysis were performed with silver staining and Cy5-labeled lectin staining to confirm the different abundances of glycopatterns in pooled saliva from ESCC and HV cases. The results of SDS-PAGE demonstrated that the distribution and abundance of salivary protein bands were similar in ESCC and HV cases (Figure 3A). The result of the lectin blotting analysis showed that the obviously different bands range from 30 to 100 kDa (Figure 3B). The GlcNAc and Galβ1-4GlcNAc binder DSA showed a distinctly increased binding to four apparent bands with molecular weights of approximately 90 kDa (b1), 55 kDa (b2), 60 kDa (b3), and 25 kDa (b4) in the ESCC patients compared with HVs, and Galβ-1,3/4GlcNAc binder ECA staining showed stronger binding intensity to two apparent bands (b3 and b4) in the ESCC patients. On the contrary, Fucα-1,6GlcNAc binder LCA and PSA and Fucα1-2Galβ1-4Glc (NAc) binder UEA-I showed weaker binding to b1, b2, b3, or b4 in ESCC than in HV (Figure 3B,C). The relative binding intensity of these randomly selected lectins to pooled saliva samples was almost coincident with the results from the lectin microarrays.

To obtain the GlcNAc, Galβ1-4GlcNAc N-glycan structures of glycoproteins in saliva from ESCC and HV cases. The glycoproteins were isolated using the DSA-magnetic particle conjugates, then N- glycans were released by PNGase F, purified by HyperSep Hypercarb SPE cartridges, and identified by MALDI-TOF/TOF-MS, respectively. A total of 48 and 56 N-glycan peaks from the pooled saliva samples of ESCC and HV were identified and annotated with proposed structures in Figure 4A,B. Of these, there were 44 and 52 GlcNAc or Galβ1-4GlcNAc containing N-glycans that could be recognized by DSA to be identified in HV and ESCC, respectively, and their proposed structures were listed in Table 3. It was noticeable that there was an overlap of 31 N-glycans (e.g., m/z 1565.491, 1593.565, and 1625.605) presented in both HV and ESCC cases, while 13 N-glycans (e.g., m/z 1622.555, 1926.732, and 2084.740) were observed only in HV samples and 21 N-glycans (e.g., m/z 1501.529, 1792.624, and 2018.639) detected only in ESCC samples. The increased tendency of GlcNAc or Galβ1-4GlcNAc-containing N-glycans in ESCC samples compared with HVs was presented in the numbers and relative abundance levels. The proportion of the N-glycans with the GlcNAc or Galβ1-4GlcNAc moieties was increased in ESCC (79.04%) compared with HV (63.20%), which was consistent with the results of lectin microarrays. The MS/MS analysis was further performed to determine the exact glycan structures. For example, the MS/MS spectra of the precursor ions m/z 1625.605, 1656.491, and 2830.999 were illustrated in Figure 5.