Site-specific N-glycosylation alterations in serum proteins across subtypes of adolescent depressive disorder

Abstract

Background

Adolescent depressive disorder is a clinically heterogeneous condition with poorly understood molecular mechanisms. Protein N-glycosylation, a key post-translational modification involved in immune regulation and neuronal communication, has not been systematically investigated across depressive subtypes in youth. Understanding glycosylation alterations may reveal novel biochemical pathways underlying disease heterogeneity and pathophysiology.

Methods

We applied a site-specific glycoproteomic approach integrating liquid chromatography-tandem mass spectrometry (LC-MS/MS) and StrucGP to profile serum N-glycosylation in adolescents with major depressive disorder (MDD), MDD with non-suicidal self-injury (NSSI), and MDD with suicide attempts (SA), compared with healthy controls. Serum samples were pooled by group (10 individuals per pool), and glycan alterations were identified using predefined fold-change thresholds. This approach is classified as an exploratory research strategy, which does not support statistical inference.

Results

All depressive subtypes displayed shared alterations, characterized by reduced bi-antennary glycans and LacNAc/Lewis structures, alongside elevated tetra-antennary sialylation, particularly N6H7S4. Candidate subtype-associated patterns were also observed: increased tri-antennary sialylation in NSSI, and selective upregulation of N6H7S4 on ORM2 in SA. Differentially glycosylated proteins, including SERPING1 and A2M, were enriched in immune, complement, and coagulation pathways. Notably, protein-level changes were minimal in label-free quantification, suggesting glycan-specific dysregulation.

Conclusions

This exploratory study identifies candidate subtype-dependent alterations in N-glycosylation in adolescent depression, which may implicate glycan-mediated immune and neuroinflammatory mechanisms in its molecular heterogeneity. This study provides a preliminary, pooled-sample based catalog of candidate site-specific N-glycosylation alterations for adolescent depression, warranting validation in larger, individual-level cohorts.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12888-026-08185-9.

Introduction

Depressive disorder is a predominant contributor to global disability-adjusted life years, manifesting as persistent affective disturbances (low mood, emotional dysregulation), cognitive dysfunction, and functional impairment-accounting for > 700,000 annual suicide deaths worldwide [1, 2]. Its rising prevalence among adolescents (10–24 years) is of particular concern, given this critical neurodevelopmental period involving dynamic brain maturation and heightened environmental stress sensitivity [3, 4]. This population shows marked clinical heterogeneity, ranging from classic MDD to subtypes with non-suicidal self-injury (NSSI) or recurrent suicide attempts [5]. However, current neurobiological frameworks insufficiently explain the multidimensional molecular pathways underlying these divergent phenotypes [6].

Although monoamine dysregulation hypotheses have been proposed, these models remain insufficient to explain the pronounced clinical heterogeneity observed in adolescent depression [7, 8]. Notably, emerging biomarker research increasingly supports the existence of distinct molecular pathways underlying depressive subtypes. For example, peripheral biomarkers exhibit subtype-specific associations: circulating endocannabinoid levels inversely correlate with depression severity [9], while discrete immune signatures are linked to suicidality among adolescents with internalizing symptoms [10], underscoring the importance of immunological differentiation in the biological stratification of depressive presentations.

Consistent with these findings, growing evidence implicates N-linked glycosylation as a key regulator of neuroimmune signaling, synaptic plasticity, and neuronal excitability [11, 12]. Glycans modulate the folding, trafficking, and function of essential synaptic proteins—including serotonin, glutamate, and GABA receptors—as well as ion channels and neurotransmitter transporters [13, 14]. Mechanistically, sialylated glycans have been shown to regulate voltage-gated ion channel activity and neurotransmitter uptake, thereby influencing neuronal communication and stress responsiveness. Generally, these molecular effects from preclinical and in vitro studies suggest that glycosylation dynamics may play a critical role in regulating mood, which supports the hypothesis that glycosylation may act as a potential mediator involved in the pathophysiology of depressive disorders. However, this potential mediating role has not been empirically verified in clinical samples of adolescent depression.

As described previously, experimentally induced alterations in N-linked glycosylation have been shown to elicit depressive-like behaviors in preclinical models, underscoring the relevance of glycosylation pathways in the pathophysiology of mood disorders [15].

Although glycoproteins serve as established biomarkers/therapeutic targets in oncology and neurodegeneration [16, 17], subtype-specific N-glycosylation in adolescent depression remains uncharacterized. This gap impedes mechanistic insights and precision biomarker development. Glycosylation signatures are emerging as potential biomarkers for adult depression, with alterations in sialylation and glycan branching linked to neuroinflammatory pathways [18]. However, research on glycosylation in adolescent MDD remains limited, with no subtype-specific profiles reported to date. In other psychiatric disorders, bipolar disorder is associated with alterations in protein glycosylation, including changes in both core mannosylation and terminal sialylation [19–20], while schizophrenia exhibits dysregulated fucosylation and sialylation [21]. Studying adolescent depressive subtypes is clinically critical: this developmental period is characterized by unique neuroimmune dynamics, high treatment heterogeneity, and elevated suicide risk, making the identification of subtype-specific biomarkers essential for guiding early intervention and advancing precision care.

To address this, we employed a multidimensional glycoproteomics platform to systematically profile serum N-glycosylation across three clinically defined adolescent cohorts: major depressive disorder (MDD), MDD comorbid with non-suicidal self-injury (MDD-NSSI), and MDD with prior suicide attempts (MDD-SA). This exploratory study aims to: (1) screen candidate N-glycosylation patterns potentially associated with depression subtypes, and (2) propose hypotheses regarding subtype-specific glycoprotein glycosylation patterns, thereby providing resources for future biomarker-based research in the field of precision psychiatry. This exploratory study provides a preliminary glycoproteomic profile of adolescent depressive subtypes, laying a foundation for future validation of candidate glycosylation biomarkers in individual-level cohorts.

Materials and methods

Patients and specimens

This study enrolled a total of 40 adolescents participants based on the 17-item Hamilton Depression Rating Scale (HAMD-17), the Beck Scale for Suicide Ideation (BSSI), and the DSM-5 diagnostic criteria [22–24]. From the inpatients at Shandong Mental Health Center, China, we selected three patient groups meeting the DSM-5 criteria for major depressive disorder (MDD): a major depressive disorder group without NSSI or suicide attempts (MDD group, n = 10), a depression with non-suicidal self-injury group (NSSI group, n = 10), and a depression with a history of suicide attempts group (SA group, n = 10). All patients met the criteria of HAMD-17 score ≥ 24 (indicating severe depression) and BSSI score ≥ 6 (indicating clinically significant suicide risk). Participants with comorbid autoimmune disorders, recent infections, or other major neurological or psychiatric conditions were excluded. To minimize potential confounding, we collected detailed medication histories for all participants. All patients were treated in accordance with the clinical guidelines for adolescent major depressive disorder, with selective serotonin reuptake inhibitors (SSRIs) as the first-line treatment. Importantly, the distribution of psychotropic medication classes (e.g., sertraline, escitalopram) and dosages was comparable across the three patient subgroups (MDD, MDD-NSSI, and MDD-SA), thereby reducing the likelihood that the observed differences in glycosylation were solely attributable to heterogeneous pharmacotherapy. Concurrently, 10 healthy adolescents were recruited to serve as the healthy control group (HC group). All diagnoses were independently confirmed by two psychiatrists. Serum samples are collected from all participants for further research.After a 30-minute incubation at room temperature, serum was isolated by centrifugation (3000 g, 10 min) from blood collected in anticoagulant tubes. The supernatant was stored at -80 °C until analysis. This study was approved by the Ethics Committee of Shandong Mental Health Center (Approval No. 2024-RE-02), and written informed consent was obtained from all participants.

Protein extraction and digestion

Equal volumes of serum samples from each group were pooled, and protein concentration was determined using BCA reagent. 500 µg of protein was aliquoted on ice and mixed with 8 volumes of 9 M urea/1.125 M NH₄HCO₃ to adjust the final concentration to 8 M urea/1 M NH₄HCO₃ for denaturation. The mixture was sonicated at 60 Hz for 2 min and centrifuged at 15,000 g for 15 min to collect the supernatant. Dithiothreitol (DTT) was added to a final concentration of 5 mM, followed by incubation at 37 °C with shaking for 1 h. Subsequently, 15 mM iodoacetamide (IAM) was added for incubation at room temperature in the dark for 30 min, and additional DTT was added to a final concentration of 2.5 mM for 10 min of incubation at room temperature. Urea was diluted to 4 M with ultrapure water, and 5 µg of sequencing-grade trypsin (protein: enzyme ratio = 100:1) was added, followed by incubation at 37 °C with shaking at 160 rpm for 2–3 h. After further diluting urea to 1 M, additional trypsin was supplemented for overnight incubation at 37 °C. Post-digestion, 50% TFA was added to adjust the final concentration to 1% (pH < 2), and the sample was centrifuged at 13,000 g for 15 min. The supernatant was loaded onto an activated C18 column (1 cc, 50 mg) and eluted twice with 200 µL of 50% ACN/0.1% TFA, yielding 400 µL of total eluate; the concentration was measured using a DeNovix instrument. 20 µg of the peptide was dried in a Speed-Vac, redissolved in 0.1% FA for proteomic analysis, and the remainder was dried and stored at − 80 °C.

Enrichment of intact glycopeptides

The desalted peptides were dissolved in 80% ACN/1% TFA and centrifuged to collect supernatant. The Hydrophilic interaction chromatographic column (cotton-based HILIC) sorbent (5 mg) was washed by ddH₂O and conditioned with 80% ACN/1% TFA 10 times used for activation, respectively. Then peptide samples were loaded on column and washed by 80% ACN/1% TFA 10 times. Finally, glycopeptides on column were eluted by 0.1% TFA and dried by vacuum centrifuging.

LC − MS/MS analysis

We integrated liquid chromatography-tandem mass spectrometry (LC-MS/MS) with the StrucGP bioinformatics pipeline for site-specific N-glycosylation profiling, enabling accurate identification of glycan structures and their attachment sites [25]. LC-MS/MS analysis of glycopeptides from human serum were performed on an EASY-nano-LC 1000 system (Thermo Scientific) coupled with Orbitrap Fusion Lumos mass spectrometer (Thermo Scientific). Peptides separation was conducted on a 75 μm × 50 cm Acclaim PepMap100 C18 column (cat. no. 164570, Thermo Fisher Scientific, USA), complemented by a 75 μm × 2 cm pre-column (cat. no. 164946, Thermo Fisher Scientific, USA). The glycopeptides were dissolved in buffer A (buffer A, 0.1% FA in water; buffer B, 0.1% FA in 80% ACN). The flow rate was 200 nL/min and the separation gradient was 120 min (0–2 min, 3–7% B; 2–85 min, 7–35% B; 85–105 min, 35–68% B; 105–120 min, 100% B. The global peptides were dissolved in buffer (A) The flow rate was 200 nL/min and the separation gradient was 130 min (0–3 min, 3–7% B; 3–94 min, 7–35% B; 94–113 min, 35–68% B; 113–130 min, 100% (B) The parameters for glycoproteomics analysis were: (1) MS: scan range (m/z) = 375–1800; resolution = 120 000; AGC target = 400 000; included charge state = 2–5; (2) MS/MS: MS scan range (m/z) = 120–3000; isolation window (m/z) = 2; detector type = Orbitrap; resolution = 30 000; maximum injection time = 100 ms; collisional mode = HCD; collisional energy = 20% and 33%. The parameters for proteomics analysis were: (1) MS: scan range (m/z) = 350–1800; resolution = 60 000; AGC target = 400 000; included charge state = 2–7; (2) MS/MS: MS scan range (m/z) = 120–2000; isolation window (m/z) = 1.6; detector type = Orbitrap; resolution = 15 000; maximum injection time = 30 ms; collisional mode = HCD; collisional energy = 30%. Each pooled serum sample was analyzed by LC-MS/MS in three technical replicates.

Proteins identification and quantification

Label-free quantification from Proteome Discoverer2.3 (PD) was used for protein quantification. The search parameters for PD were set as follows: using the UniProt human proteome database (http://www.uniprot.org, accession number UP000005640); maximum missed cleavages for trypsin set to 2; mass tolerance ranges for precursor and fragment ions set to 10 ppm and 0.02 Da, respectively; fixed peptide modification of carbamidomethylation at cysteine residues (C, + 57.021464 Da); variable peptide modifications including methionine oxidation (M, + 15.9949 Da) and N-terminal acetylation (+ 42.010565 Da). This study employed a label-free quantitative approach, utilizing a parent ion peak area integration-based algorithm for peptide quantification analysis. Peptide identification required matching at least two peptide-spectrum matches (PSMs), with search results filtered using a 1% false discovery rate (FDR) as the screening threshold.

Intact glycopeptides identification and quantification

The raw data obtained from nano-LC-MS/MS was searched by StrucGP, referring to the UniProtKB human protein sequence database [26]. The protein enzymatic digestion was performed using trypsin with up to 2 missed cleavages allowed. Potential glycosylation-containing peptides were screened based on the N-X-S/T motif (where X represents any amino acid except proline). Carbamidomethylation (C, + 57.0215 Da) was set as a fixed modification, and oxidation (M, + 15.9949 Da) was designated as a variable modification. The mass tolerance was set to 10 ppm for MS1 and 20 ppm for MS2. For Y-ion determination, an optional mass shift of ± 1 Da or ± 2 Da was permitted in addition to the 20 ppm mass tolerance for MS2. Finally, both peptide and glycan components were required to achieve a false discovery rate (FDR) of < 1% for the identification of intact glycopeptides.

A label-free quantitative approach was employed to compare the abundance of intact N-glycopeptides among sample groups. Glycopeptide abundance was represented by peptide-spectrum matches (PSMs). A segmented threshold filtering strategy based on PSMs was adopted: a 2-fold threshold (Ratio > 2 or Ratio < 0.5) for glycopeptides with PSMs ≥ 20 in any pooled sample; a 4-fold threshold for PSMs between 10 and 20; and an 8-fold threshold for PSMs between 5 and 10. This is a heuristic filtering approach for discovery-phase screening and does not constitute statistical testing. For final differential comparisons, the PSMs used were normalized values calculated based on the mean total PSMs of intact glycopeptides in each group. For missing values (i.e., intact glycopeptides not identified in one group), the PSM value of the same glycopeptide in the other group included in the comparison was required to be no less than 5.

Gene ontology (GO) and reactome pathway analyses

Gene Ontology (GO) terms enrichment and Reactome pathway analyses were performed using the STRING platform (https://string-db.org) to characterize biological functions and pathways associated with differentially expressed glycoproteins in Adolescent depression patients plasma [27]. Statistically significant terms and pathways were identified using a dual threshold (gene count > 2, p-value < 0.05) adjusted for False Discovery Rate (FDR). GO highlighted overrepresented biological processes, molecular functions, and cellular components, while Reactome analysis revealed key pathways linked to depression progression. Results were visualized to prioritize functionally relevant networks.

Results

Baseline characteristics of participants

The baseline characteristics of the 40 included adolescents are shown in Table 1. Among the four groups (HC, MDD, MDD-NSSI, MDD-SA), no significant differences were found in age or gender distribution (P > 0.05). Within the MDD subgroups (MDD, MDD-NSSI, MDD-SA), there were no significant differences in illness duration or HAMD-17 scores (P > 0.05). Regarding BSSI scores, while no significant difference was observed between the MDD and MDD-NSSI subgroups (P > 0.05), both exhibited significantly lower BSSI scores compared to the MDD-SA subgroup (P < 0.05).

Table 1.

Demographic and clinical characteristics of study participants

Group	Sample size (n)	Sex (male/female)	Age (years, median)	HAMD-17 (x ± s)	BSSI (x ± s)
HC	10	5/5	15	5.4 ± 1.6	0 ± 0
MDD	10	5/5	15	27.3 ± 3.1	19.4 ± 5.3
MDD-NSSI	10	5/5	15	27.7 ± 3.3	23.6 ± 5.2
MDD-SA	10	5/5	15	28.5 ± 3.5	29.83 ± 4.6

Study design for quantitative analysis of intact N-glycopeptides in the serum of depressed patients

To investigate site-specific glycosylation differences of serum proteins between depressed patients and healthy controls, intact N-glycopeptides and proteomic analyses were performed on serum samples from healthy individuals and three subtypes of depressed patients (Fig. 1A). Specifically, 10 serum samples were collected from each of the healthy, depressed, suicidal depressed, and self-injurious depressed groups, with samples from each group pooled into one, resulting in four pooled serum samples.

Fig. 1.

Comprehensive N-glycoproteomic analysis of serum from healthy and depressed patients. (A) Workflow of glycoproteomic analysis of serum from healthy and depressed patients. (B) Venn diagram of intact N-glycopeptides identified in the four groups of serum samples. (C) The number of intact glycopeptides with different glycan types identified from the four groups of serum

Given this pooled design, the study adopts an exploratory approach focused on fold-change-based comparisons rather than statistical inference. Prior to LC-MS/MS analysis, total proteins were extracted from the serum, digested with trypsin into peptides, and glycopeptides were enriched using a HILIC-based separation system enrichment method. Glycopeptide identification was carried out using StrucGP software, and peptides from the proteomic analysis were also subjected to mass spectrometry [28]. Relative quantification of proteins and intact N-glycopeptides was achieved by label-free quantification approach [29]. Subsequently, an integrated analysis of glycoproteomic and proteomic data was conducted to explore potential alterations in site-specific glycosylation on serum proteins among the different depressed patient groups (Fig. 1A). Using the above approach, 1432, 1164, 1217, and 1161 intact N-glycopeptides were identified from the serum of healthy individuals (HC, IGP1), self-injurious depressed patients (MDD-NSSI, IGP2), suicidal depressed patients (MDD-SA, IGP3), and depressed patients (MDD, IGP4), respectively (Fig. 1B). In total, 2044 intact N-glycopeptides were identified across the four serum samples, with 695 intact N-glycopeptides commonly detected in all groups (Fig. 1B). These intact N-glycopeptides encompassed 217 N-glycans distributed on 271 glycoproteins. Among these 217 N-glycans, nearly 70% (144) were complex glycans, followed by hybrid (57) and oligo-mannoses (12) glycans (Fig. 1C) [30].

Overview of site-specific N-glycosylation analysis in the serum of patients with depression

This study systematically compared the site-specific glycosylation profiles in serum between healthy adolescents and patients with depression through heatmap analysis (Fig. 2A). The analysis incorporated all identified intact glycopeptides from IGP1, IGP2, IGP3, and IGP4 serum samples, along with their peptide spectrum matches (PSMs). Mass spectrometry analysis identified 271 glycoproteins corresponding to 2,044 intact N-glycopeptides in serum samples, which collectively contained 217 distinct N-glycan structures and 470 specific glycosylation sites. Notably, individual glycosylation sites could be modified by up to 46 different N-glycan structures. Glycan composition analysis revealed that 70.9% of glycosylation sites were exclusively modified by complex-type glycans, while 7.7% of sites exhibited co-occurrence of complex and hybrid-type glycans (Fig. 2A).

Fig. 2.

Overview of the analysis of all intact glycopeptides identified from serum of healthy and depressed patients. (A) Heatmap of identified intact glycopeptides in serum from healthy and depressed patients. The peptide-spectrum-matches (PSMs) of the intact glycopeptides, comprising different glycans compositions (upper) and glycosites (left) were shown in the heatmap. The numbers of glycosites modified by each glycan and glycans at each glycosite were summarized at the bottom and right part of the figure, respectively. N: HexNAc; H: Hex; F: Fucose; S: Neu5Ac. (B) Intact glycopeptide PSMs of different glycan subtypes from serum of healthy and depressed patients. (C) The top ten glycan structures detected in serum based on the number of modified glycosites

Further characterization demonstrated that complex-type glycans predominated among the 2,044 N-glycopeptides (82.2%, 1,680), followed by hybrid-type (11.2%, 228) and high-mannose-type glycans (6.2%, 127) (Fig. 2B). Remarkably, over 85% of complex-type glycan-modified glycopeptides carried 1–4 sialic acid residues, with 46% of these glycopeptides displaying fucosylation modifications (Fig. 2B). Structural diversity analysis revealed that individual N-glycans could modify up to 291 distinct glycosylation sites (Fig. 2A). The top 10 most prevalent glycans modifying multiple sites primarily consisted of biantennary (N4H5S2, N4H5S1, N4H5F1S2, N4H5F1S1, N4H5, N5H6F1S2) and triantennary structures (N5H6S3, N5H6F1S3, N5H6S2, N5H6F1S2) (Fig. 2C). Notably, all dominant glycans contained sialic acid residues, with 50% exhibiting fucose modifications, suggesting that abnormal regulation of sialylation and fucosylation levels may critically contribute to depression pathogenesis [31].

Quantitative analysis of site-specific N-glycosylation in serum of MDD-NSSI patients (IGP2)

This study employed a label-free quantitative approach to compare the abundance of intact N-glycopeptides in serum samples from IGP2 and IGP1 patients. To enhance the reliability of quantitative results, we established a segmented threshold criterion based on PSMs: glycopeptides with higher PSMs exhibited superior identification stability and more credible differential ratios, whereas those with lower PSMs demonstrated greater quantitative variability, necessitating stricter filtering thresholds. Using the fold-change-based thresholding strategy, 70 differentially abundant glycopeptides derived from 27 proteins were identified in MDD-NSSI patients (IGP2) versus healthy controls (IGP1) serum (Fig. 3A), with 49 downregulated and 21 upregulated. Notably, label-free quantification using Proteome Discoverer 2.3 (PD) revealed minimal changes in protein expression levels corresponding to these glycopeptides (defined as fold-change < 1.5 between groups) (Supplementary Fig. 1A), suggesting that the observed differences in the MDD-NSSI group serum primarily originated from glycan structural modifications rather than alterations in protein abundance. Functional annotation analysis (Supplementary Fig. 1B) demonstrated that these glycoproteins were predominantly involved in immune responses (defense response, complement activation), biological regulation (negative regulation of coagulation and fibrinolysis), and stress responses, while also modulating enzymatic activity, endopeptidase inhibition, and heparin/complement binding at the molecular level. Reactome pathway analysis (Supplementary Fig. 1C) further revealed significant enrichment of differential glycoproteins in critical pathways, including platelet degranulation, complement cascade regulation, post-translational phosphorylation, intrinsic pathways of fibrin clot formation, and C3/C5 activation.

Fig. 3.

Quantification of site-specific glycosylation in the serum of MDD-NSSI patients. (A) Volcano plot of the PSMs vs. the log2 (IGP2/IGP1). The up- and down-regulated glycopeptides were highlighted in red and green, respectively.The plot does not include p-values, as the analysis is based on fold-change filtering without statistical inference. (B) Glycan types of differentially expressed glycopeptides. (C) Distribution of core structures of differentially expressed glycopeptides. (D) Distribution of branch structures of differentially expressed glycopeptides. “⌀“: lack of one branch. (E) Glycopeptide count of differentially expressed site-specific glycans in the serum of patients with self-injurious depression. The count was calculated on the basis of the number of glycosites modified by site-specific glycans

From 70 identified differential glycopeptides, 27 site-specific N-glycan structures were characterized. These N-glycans predominantly comprised high-mannose, bi-/tri-, and tetra-antennary structures, core-fucosylated bi-/tri-antennary glycans, and sialylated bi-/tri-, and tetra-antennary glycans. Among them, 80% were complex-type glycans, followed by hybrid-type (10%) and high-mannose-type (10%) glycans (Fig. 3B). These site-specific N-glycans derived from differential glycopeptides consisted of two core structures and seven branch structures. For core structures, 90% were standard core structures (HexNAc2Hex3), while the remainder were fucosylated cores (Fig. 3C). Among the seven branch structures, HexNAc + Hex+Neu5Ac, LacNAc (annotated as HexNAc + Hex), and Hex modified 79%, 44%, and 17% of the differential glycopeptides, respectively. Additionally, sialylated Lewisx/a, HexNAc, and Lewisx/a modified 13%, 3%, and 1% of the differential glycopeptides, respectively (Fig. 3D). Based on these differentially expressed intact glycopeptides in the MDD-NSSI group, Fig. 3E statistically summarizes the frequency of site-specific glycan modifications across the 27 glycosylation sites, with the Y-axis representing glycopeptide counts. Key findings include: (1) Bi-antennary sialylated glycans (N4H5S2 and N4H5S1) showed a consistent decrease in downregulated glycopeptides; (2) Tri- and tetra-antennary sialylated glycans (N5H6S3, N5H6S2, N6H7S4, and N6H7S3) were predominantly associated with upregulated glycopeptides; and (3) Sialylated Lewisx/a and Lewisx/a glycans almost exclusively originated from downregulated glycopeptides. In the MDD-NSSI group, the observed fold-change pattern in pooled sera showed a reduction of bi-antennary glycans and Lewis structure-containing glycans, coupled with an increase in tri/tetra-antennary sialylated glycans at specific glycosylation sites.

Quantitative analysis of site-specific N-glycosylation in serum from patients with MDD-SA patients (IGP3)

Using the same quantitative approach described above, a total of 34 differential glycopeptides derived from 19 proteins were identified in the serum of MDD-SA patients (IGP3) versus healthy controls (IGP1), with 11 upregulated and 23 downregulated glycopeptides (Fig. 4A). Further analysis of the protein expression levels corresponding to these differential glycopeptides revealed minimal variations at the protein level, suggesting that the observed glycopeptide differences in the MDD-SA group predominantly arise from structural alterations in the glycan chains of modified glycoproteins (Supplementary Fig. 2A). To elucidate the associated biological events, Gene Ontology (GO) and pathway enrichment analyses were performed on the 34 differential glycopeptides at the protein level. The results demonstrated that these glycoproteins are primarily involved in biological regulation, response to stimuli, stress response, defense response, immune system processes, inflammatory response, complement activation, and coagulation. At the molecular functional level, they predominantly exhibited endopeptidase inhibitor activity (Supplementary Fig. 2B). Reactome pathway analysis further revealed that these glycoproteins are significantly enriched in pathways including platelet degranulation, innate immune system, vesicle-mediated transport, regulation of the complement cascade, post-translational protein phosphorylation, intrinsic pathway of fibrin clot formation, and plasma lipoprotein assembly (Supplementary Fig. 1C).

Fig. 4.

Quantification of site-specific glycosylation in the serum of MDD-SA patients. (A) Volcano plot of the PSMs vs. the log2 (IGP3/IGP1). The up- and down-regulated glycopeptides were highlighted in red and green, respectively.The plot does not include p-values, as the analysis is based on fold-change filtering without statistical inference. (B) Glycan types of differentially expressed glycopeptides. (C) Distribution of core structures of differentially expressed glycopeptides. (D) Distribution of branch structures of differentially expressed glycopeptides. “⌀“: lack of one branch. (E) Glycopeptide count of differentially expressed site-specific glycans in the serum of patients with suicidal depression. The count was calculated on the basis of the number of glycosites modified by site-specific glycans

A total of 15 site-specific N-glycan structures were identified from 34 differential glycopeptides obtained through selection.These N-glycans primarily consisted of high-mannose, bi-, tri-, and tetra-antennary glycans, as well as terminally sialylated bi-, tri-, and tetra-antennary N-glycans. Among the N-glycans, 88% were complex-type glycans, followed by hybrid-type (6%) and high-mannose-type (6%) glycans (Fig. 4B). These site-specific N-glycans from differential glycopeptides comprised two core structures and six branch structures. For core structures, 94% were common core structures (HexNAc2Hex3), while the remainder were fucosylated core structures (Fig. 4C). Among the six distinct branch structures, HexNAc + Hex+Neu5Ac, LacNAc, and sialylated Lewisx/a modified 85%, 44%, and 18% of the differential glycopeptides, respectively. Hex and Lewisx/a modified 9% and 6% of the differential glycopeptides, respectively (Fig. 4D).For these intact glycopeptides differentially expressed in the MDD-SA group, Fig. 4E summarizes the frequency of these 15 site-specific glycan modifications at glycosylation sites. The results show that bi-antennary sialylated glycans (N4H5S2 and N4H5S1) modified the highest number of both upregulated and downregulated glycopeptides, while tetra-antennary sialylated glycans (N6H7S4 and N6H7S3) were exclusively present on upregulated glycopeptides. Additionally, glycans bearing sialylated Lewisx/a, Lewisx/a, and LacNAc were almost entirely derived from downregulated glycopeptides. Thus, in the MDD-SA group, a reduction of glycans with Lewis and LacNAc structures at specific glycosylation sites, coupled with an increase in tetra-antennary sialylated glycans, was observed.

Quantitative analysis of site-specific N-glycosylation in serum from patients with MDD patients (IGP4)

Using the same quantitative methodology described above, 40 differential glycopeptides derived from 19 proteins were identified in serum samples from MDD patients (IGP4) versus healthy controls (IGP1) cohorts, including 11 upregulated and 29 downregulated glycopeptides (Fig. 5A). Further analysis of protein expression levels corresponding to these differential glycopeptides using label-free quantification (PD 2.3) revealed minimal changes at the protein level (fold-change < 1.5) for most glycoproteins, with the exception of FGB and FGG proteins, which showed modest protein-level differences (fold-change > 1.5). This suggests that for the majority of glycopeptides, the observed variations in the MDD group were primarily attributed to structural alterations in glycan chains, although protein abundance changes for FGB and FGG may confound interpretation of glycan-level differences for these specific proteins (Supplementary Fig. 3A). To elucidate the associated biological events, Gene Ontology (GO) and Reactome pathway analyses were performed on the 40 differential glycopeptides at the protein level. The results demonstrated that these glycoproteins predominantly participate in biological processes such as response to stimulus, defense response, regulation of biological quality and catalytic activity, immune system processes, regulation of protein metabolic processes, positive regulation of transport, acute-phase response, complement activation, and coagulation. Molecular functions were primarily linked to signaling receptor binding and endopeptidase inhibitor activity (Supplementary Fig. 3B). Reactome pathway analysis further revealed that these glycoproteins are predominantly involved in platelet degranulation, innate immune system, platelet activation and aggregation, coagulation cascade, regulation of complement cascade, post-translational protein phosphorylation, neutrophil degranulation, intrinsic pathway of fibrin clot formation, and integrin cell surface interactions (Supplementary Fig. 1C).

Fig. 5.

Quantification of site-specific glycosylation in the serum of MDD patients. (A) Volcano plot of the PSMs vs. the log2 (IGP4/IGP1). The up- and down-regulated glycopeptides were highlighted in red and green, respectively.The plot does not include p-values, as the analysis is based on fold-change filtering without statistical inference. (B) Glycan types of differentially expressed glycopeptides. (C) Distribution of core structures of differentially expressed glycopeptides. (D) Distribution of branch structures of differentially expressed glycopeptides. “⌀“: lack of one branch. (E) Glycopeptide count of differentially expressed site-specific glycans in the serum of patients with depression. The count was calculated on the basis of the number of glycosites modified by site-specific glycans

Among the 40 differential glycopeptides identified, 16 site-specific N-glycans structures were characterized. These N-glycans primarily consisted of high-mannose, bi-, tri-, and tetra-antennary glycans, as well as sialylated bi-, tri-, and tetra-antennary N-glycans. Among these structures, 83% were complex-type glycans, followed by hybrid-type (10%) and high-mannose-type glycans (8%) (Fig. 5B). The site-specific N-glycans from differential glycopeptides were composed of two core structures and seven branch structures. For core structures, 90% exhibited the conventional core structure (HexNAc₂Hex₃), while the remaining displayed a fucosylated core structure (Fig. 5C). Among the seven distinct branching structures, HexNAc + Hex+Neu5Ac, LacNAc, and sialylated Lewisx/a modifications were observed in 78%, 38%, and 13% of differential glycopeptides, respectively. Additionally, Hex, Lewisx/a, and HexNAc modifications were present in 13%, 3%, and 3% of the differential glycopeptides, respectively (Fig. 5D).

For the intact glycopeptides differentially expressed in the MDD group, excluding those derived from FGB and FGG proteins, the frequency of site-specific glycosylation at the 16 identified sites was quantified (Fig. 5E). Notably, the sialylated bi- and tri-antennary glycans (N4H5S2 and N5H6S3) were predominantly associated with downregulated glycopeptides, whereas the tetra-antennary sialylated glycan (N6H7S4) was exclusively present in upregulated glycopeptides. Furthermore, glycans carrying sialylated Lewisx/a, Lewisx/a, and LacNAc motifs were almost entirely derived from downregulated glycopeptides. Across all depressive disorder subtypes (MDD, MDD-NSSI, and MDD-SA), in pooled-sample comparisons, a consistent reduction in glycans bearing Lewis and LacNAc structures, along with an increase in tetra-antennary sialylated glycans, was observed.

Differential intact N-glycopeptides with similar trends in serum from MDD-NSSI, MDD-SA, and MDD patients

A systematic analysis of differentially expressed glycopeptides identified in the sera of MDD-NSSI (IGP2), MDD-SA (IGP3), and MDD (IGP4) patients revealed a consistent trend in glycopeptide alterations. Specifically, one glycopeptide was commonly upregulated (Fig. 6A), while 11 glycopeptides were consistently downregulated (Fig. 6B). These 12 differentially expressed N-glycopeptides were further illustrated using a heatmap (Fig. 6C). The identified glycopeptides consisted of 10 distinct N-glycan structures and were derived from 8 glycoproteins (ORM2, HP, TF, SERPING1, SERPINA3, KNG1, CFB, and A2M) at 9 glycosylation sites. Based on their biological functions, they can be categorized into four groups: (1) Immune regulation—ORM2 inhibits inflammation and modulates immune cell function, while CFB enhances innate immunity by activating the alternative complement pathway [32, 33]; (2) Coagulation and fibrinolysis regulation—SERPING1 regulates the coagulation cascade by inhibiting coagulation factors and complement, KNG1 modulates vascular permeability and coagulation through bradykinin, and A2M influences the coagulation-fibrinolysis balance by clearing thrombin/plasmin. Additionally, TF, apart from its role in iron transport, is involved in platelet activation and coagulation [34–38]; (3) Inflammation and oxidative stress regulation—HP binds free hemoglobin to reduce oxidative damage, while SERPINA3 inhibits neutrophil elastase, mitigating inflammatory injury [39, 40]. Functional analysis suggests that these glycoproteins play crucial roles in immune modulation, platelet function regulation, and coagulation-fibrinolysis processes. Furthermore, by influencing the serotonin system, coagulation mechanisms, and inflammatory status, they may be indirectly involved in the pathogenesis of depression, consistent with previous studies.

Fig. 6.

Differential intact N-glycopeptides with the same changing trend in serum of MDD-NSSI (IGP2), MDD-SA (IGP3), and MDD (IGP4) patients. (A) Venn diagram of up-regulated intact N-glycopeptides in serum for IGP2, IGP3, and IGP4. (B) Venn diagram of down-regulated intact N-glycopeptides in serum for IGP2, IGP3, and IGP4. (C) Heatmaps of 12 differentially expressed intact glycopeptides of eight glycoproteins (ORM2, HP, TF, SERPING1, SERPINA3, KNG1, CFB, A2M) and their quantitative information

It is important to note that these 12 glycopeptides represent only a small fraction of all identified glycopeptides from the 8 glycoproteins. Interestingly, the only upregulated glycan was the tetra-antennary sialylated glycan (N6H7S4) on ORM2, whereas 73% of the downregulated glycans carried sialylation, predominantly in bi- and tri-antennary structures. The glycosylation levels of SERPING1, SERPINA3, KNG1, and CFB were all decreased, with the downregulation of Lewisx/a structures on SERPINA3 and KNG1 potentially affecting their roles in inflammation, immunity, and neurological functions [41]. These findings align with previous literature, suggesting an association between these glycoprotein modifications and depression. This adds to the growing evidence of their role in disease pathophysiology and highlights their promise for future translational research.

Discussion

This exploratory study provides preliminary evidence for subtype-specific N-glycosylation alterations in adolescent depression. Through integrated glycoproteomic profiling, we identified site-specific glycan changes associated with distinct pathogenic mechanisms. In non-suicidal self-injury depression, upregulated glycopeptides were predominantly modified by tri- and tetra-antennary sialylated glycans. In contrast, suicidal depression was characterized by the exclusive upregulation of the tetra-antennary sialylated glycan N6H7S4 on ORM2. Notably, N6H7S4 has also attracted attention in cancer research, where its increased expression is known to disrupt cell adhesion, thereby facilitating metastasis and immune evasion [42]. Although the functional implications of N6H7S4 upregulation in depression remain speculative, it is plausible that such glycosylation changes could influence neuroinflammatory processes or neural signaling related to emotional regulation, a hypothesis that warrants further investigation. These glycosylation differences might reflect subtle pathophysiological distinctions among different depression subtypes, suggesting a potential avenue for refining classification systems, though further validation is required.

Distinct glycan alterations were observed across all depressive disorder subtypes, characterized by a consistent trend toward upregulation of complex-type glycans-particularly tetra-antennary sialylated N6H7S4 and downregulation of bi-antennary glycans and Lewis/LacNAc structures. While these changes were partially shared among different depression subtypes (12 commonly altered glycopeptides), specific glycosylation patterns exhibited subtype specificity, supporting the concept of glycosylation as a potential molecular signature for depression subtyping. Furthermore, complex-type glycans, particularly N6H7S4, may possess functional significance in depression. Highly sialylated glycans are thought to contribute to chronic inflammation in both the peripheral and central nervous systems and have been linked to complement system activation, cytokine release, and glycosylation modifications [43, 44]. These preliminary glycan alterations may represent candidate targets for future biomarker exploration.

Significant differences in serum glycosylation profiles were observed among patients with non-suicidal self-injury depression, suicidal depression, and major depressive disorder. The reduction of bi-antennary glycans and Lewis/LacNAc structures could potentially disrupt processes cell adhesion, complement activation, and coagulation pathways, which may be relevant to pathophysiology of depression [45]. In contrast, highly sialylated complex-type glycans are implicated in inflammation, immune regulation, and neuronal signaling [46, 47]. Proteins such as ORM2, SERPING1, and A2M, along with their glycosylation modifications, are increasingly recognized as relevant to depression. Functional annotation of these differential glycoproteins revealed their involvement in key biological processes, including immune system regulation, inflammatory responses, complement activation, and coagulation cascades. This integrated view is consistent with the hypothesis that depression involves a state of chronic immune dysregulation [48–51]. Previous studies have indicated that chronic inflammation might influence depression through mechanisms such as neurotransmitter metabolism, neuroplasticity, and stress responses [52, 53]. Collectively, these findings lend preliminary support to the inflammation hypothesis of depression and suggest that N-glycosylation modifications could be a pivotal mechanism mediating the interplay between immune dysregulation and depressive pathology, potentially contributing to heterogeneity as seen in peripheral biomarker studies [54, 55].

Furthermore, our findings resonate with existing literature on glycosylation in adult depression and other psychiatric disorders. For instance, the increased sialylation and decreased Lewis structures observed in our adolescent cohorts parallel some reports in adult MDD [56], suggesting a common inflammation-related glycan signature across ages. Similarly, alterations in complex-type N-glycans have been noted in schizophrenia, hinting at shared neuroimmune mechanisms across diagnostic boundaries. The specific glycan alterations we observed—particularly the reduction of bi-antennary glycans and Lewis/LacNAc structures—have been previously implicated in neuroinflammatory processes and behavioral symptoms in neurological disorders [32, 33], supporting their potential relevance to depression pathophysiology [34]. This convergence underscores the potential of glycosylation as a transdiagnostic marker of neuroinflammation. The clinical implications of these findings, while preliminary, are also noteworthy. If validated, these subtype-specific glycosylation patterns could, in the future, aid in risk stratification (e.g., the upregulation of N6H7S4 on ORM2 as a potential marker for suicide risk), provide objective molecular information to support differential diagnosis, and potentially guide personalized treatment strategies by identifying patients who might benefit from immunomodulatory interventions. Collectively, these translational possibilities highlight the relevance of glycosylation-informed approaches for advancing precision psychiatry.

As an exploratory study utilizing pooled samples, this work provides an initial characterization of site-specific serum N-glycosylation profiles in adolescent depression subtypes. Sample pooling was intentionally adopted at this discovery stage to enhance the depth of glycopeptide detection and mitigate the impact of inter-individual biological and technical variability. Importantly, this strategy also reduces the risk of false-positive discoveries driven by single outlier samples, as the signal from any individual is averaged with those of the other nine samples within the same pool. Consequently, a substantial fold-change can only be observed when multiple samples within the group exhibit a consistent trend, thereby prioritizing candidate alterations that are more likely to reflect group-level trends rather than individual anomalies. We fully acknowledge that this approach inherently limits statistical inference and a direct assessment of inter-individual heterogeneity. To address this limitation, we have outlined future directions emphasizing validation at the individual-sample level, which will enable statistical assessment and replication across independent cohorts.

Additionally, while emerging technologies like AI-driven glycan sequencing hold promise for psychiatric biomarker development, robust validation of these signatures remains paramount. Glycomics offers a novel framework for deconstructing depression’s heterogeneity, potentially informing precision medicine approaches through glycosylation-focused biomarkers.

Conclusion

This exploratory study identifies distinct serum N-glycosylation patterns across subtypes of adolescent depressive disorder using pooled-sample glycoproteomics. We observed a consistent decrease in bi-antennary glycans and Lewis/LacNAc structures, alongside increased tetra-antennary sialylation (particularly N6H7S4 on ORM2) in suicidal depression, whereas tri-antennary sialylation characterized non-suicidal self-injury. Integrated analysis indicates these alterations may link to dysregulated immune, complement, and coagulation pathways, independent of protein abundance—supporting glycan-specific mechanisms underlying depression heterogeneity. While our approach revealed potential subtype-specific signatures, the absence of individual-level data and statistical validation limits the interpretability of these findings. Future work should prioritize validation in individual serum samples using orthogonal glycomic or glycoproteomic methods to confirm the observed structures and quantitative differences. Moreover, expanding this framework to larger, longitudinal cohorts will enable assessment of temporal dynamics and causal relationships between glycan alterations, inflammatory activity, and clinical outcomes.

Collectively, these directions will establish a mechanistic foundation for glycan-based biomarkers in adolescent depression and advance the understanding of glycan-immune modulation in psychiatric disorders.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Abbreviations

LC-MS/MS

Liquid chromatography-tandem mass spectrometry

HC

Healthy individuals

MDD

Major depressive disorder

MDD-NSSI

MDD with non-suicidal self-injury

MDD-SA

MDD with suicide attempts

HAMD-17

17-item Hamilton Depression Rating Scale

BSSI

Beck Scale for Suicide Ideation

GO

Gene Ontology

PSMs

Peptide spectrum matches

A2M

Alpha-2-Macroglobulin

CFB

Complement Factor B

HP

Haptoglobin

KNG1

Kininogen-1

ORM2

Orosomucoid 2

SERPINA3

Serpin Family A Member 3

SERPING1

Serpin Family G Member 1

TF

Transferrin

Author contributions

X.H. and C.L. conceived and designed the study. X.H., X.L., Y.W., Y.Z. performed the experiments. X.H., X.L., Y.Z. contributed the materials. Y.W., K.Z., C.L., X.L., Y.Z. analysed the data. F.Z. and J.D. prepared the figures. X.H. wrote the manuscript. X.H. supervised the study. X.H. acquired funding. All authors reviewed and approved the manuscript.

Funding

This work was supported by Shandong Provincial Natural Science Foundation Youth Program (ZR2021QH283), Shandong Provincial Medical and Health Technology Development Plan (No.202201060948) and Shandong Provincial Traditional Chinese Medicine Science and Technology Project (Z-2023061).

Data availability

The mass spectrometry data have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository with the dataset identifier IPX0012470000.

Declarations

Ethics approval and consent to participate

All procedures involving human participants in this study, including the collection and use of adolescent serum samples, were conducted in accordance with the ethical standards of the Declaration of Helsinki. The study protocol was reviewed and approved by the Ethics Committee of the Shandong Mental Health Center (Approval No. 2024-RE-02). Prior to the commencement of the study, written informed consent was obtained from all individual participants and their legal guardians after a detailed explanation of the study’s purpose, procedures, potential risks, and benefits.

Consent for publication

Not applicable.

Supplementary information

Details of the experimental procedures, data processing, supplementary Figures S1-S3, list of all next generation sequencing files using in this study and instructions for access (Tables S1-S3).

Competing interests

The authors declare no competing interests.