Glycosylation Signatures of Thyroglobulin in Papillary Thyroid Carcinoma

2.1. Sample Collection and Deparaffinization Process

Thyroid cancer and adjacent non‐tumoral healthy thyroid tissues were obtained from the pathology unit of the Karabük Education and Research Hospital. The Non‐Interventional Ethics Committee of the Karabük University Faculty of Medicine approved this study (2023/1236). Nineteen formalin‐fixed paraffin‐embedded (FFPE) tissue blocks of PTC and 15 FFPE tissue blocks of adjacent non‐tumoral healthy thyroid tissue were retrospectively collected for analysis. Adjacent control tissues were selected from macroscopically normal thyroid tissue located outside the tumor margins. To ensure accurate classification, an experienced pathologist microscopically evaluated all sections to confirm the complete absence of tumor cells in the control samples and to verify PTC morphology in the cancerous samples. Approximately 10 µm thick FFPE tissue sections were obtained from cancerous and adjacent control tissues and stored at room temperature until sample preparation. Table 1 presents the demographic information of the study population.

2.2. Deparaffinization of Tissues and Extraction of Proteins From Thyroid Tissues

This procedure was applied with minor modifications to a previously described method [23]. FFPE tissue sections of 10 µm thickness were placed into 1.5 mL sample tubes and washed twice with xylene for 5 min under gentle shaking and then cleared twice with absolute ethanol for 5 min. The cleared sections were then allowed to dry. The deparaffinized and cleared samples were combined and homogenized in 0.1 M Tris–HCl (pH 8.0) and 0.1 M dithiothreitol (DTT) buffer on ice. This procedure used a desktop Branson B‐12 sonicator (power output control: 1.5; 3 × 10–30 s). SDS was added to a final concentration of 4%, and the samples were incubated at 99°C for 60 min with gentle shaking, then cooled to room temperature. Samples were centrifuged at 2000 rcf for 20 min to remove insoluble material, and 150 µL of the supernatant from each sample was placed in separate Eppendorf tubes. Then, 600 µL methanol, 150 µL chloroform, and 450 µL deionized water were added to all samples. Samples were vortexed for 1–2 s and centrifuged at 14 000 rcf for 5 min. The supernatant was gently removed to avoid damaging the central pellet. Subsequently, 450 µL of methanol was added, vortexed for 1–2 s, and centrifuged at 14 000 rcf for 10 min. The remaining protein precipitate at the bottom was dried using a fast vacuum concentrator at 45°C and stored at −20°C. Dried proteins were solubilized with 75 µL of 8 M urea to ensure complete denaturation of the precipitated proteins, facilitating subsequent reduction and Lys‐C digestion. A spectrophotometer (BioTek Epoch, USA) was used to quantify the amount of protein. Each sample was prepared and analyzed separately.

2.3. In‐Solution Digestion of Extracted Proteins From Thyroid Cancer and Adjacent Control Tissues

The following procedures were applied using 10 µg of proteins extracted from the tissue, and 1,4‐DTT was added to a final concentration of 5 mM to reduce the proteins. This process was applied at 56°C for 40 min to break the disulfide bonds and prepare the proteins for further processing. After the reduction process was completed, iodoacetamide (IAA) was added to a final concentration of 15 mM for alkylation of the proteins, and the samples were kept in the dark for 20 min. Protein digestion was initiated following alkylation. Proteins were first incubated at 37°C for 4 h using the Lys‐C protease enzyme. The incubation was carried out at a ratio of 1:100 (enzyme/protein, w/w). Following digestion with Lys‐C, trypsin was added to complete protein digestion. Digestion with trypsin was completed by overnight incubation (approximately 16 h) at 37°C at a ratio of 1:30 (enzyme/protein, w/w). After proteolytic digestion, the glycopeptides were enriched using a ZIC‐HILIC sorbent (SPE adsorbents, Bulk, CHROMABOND HILIC, MACHEREY‐NAGEL, Germany). ZIC‐HILIC beads were suspended in 0.1% trifluoroacetic acid (TFA) (50 mg/mL) and loaded (100 µL/well) onto a filter plate. After washing the wells twice with 0.1% TFA, they were equilibrated twice with 90% ACN containing 0.1% TFA. The samples were diluted in 90% ACN/0.1% TFA and applied to the conditioned beads. The impurities were removed by washing four times with the same solution. Enriched glycopeptides were eluted using 0.1% TFA and collected by low‐speed centrifugation, dried using a SpeedVac, and stored at −20°C until analysis.

2.4. Mass Spectrometric Analyses

Peptides were analyzed using a Dionex UltiMate 3000 UHPLC system (Thermo Fisher Scientific, Waltham, MA, USA) coupled online to a trapped ion mobility spectrometry‐time‐of‐flight mass spectrometer (timsTOF; Bruker Daltonics, Bremen, Germany). Ionization was performed in the positive mode using a CaptiveSpray source equipped with a nanoBooster, which continuously delivered nitrogen gas enriched with acetonitrile to enhance the ionization efficiency, particularly for glycopeptides. Chromatographic separation was achieved using a two‐column setup consisting of an Acclaim PepMap 100 C₁₈ trap column (100 µm × 5 cm, 5 µm particle size; Thermo Scientific) and an Acclaim PepMap RSLC C₁₈ analytical column (75 µm × 25 cm, 2 µm particle size). Peptide samples were initially loaded onto the trap column at a flow rate of 15 µL/min using a loading buffer composed of dH₂O with 0.02% TFA. Analytical separation was conducted at a flow rate of 400 nL/min using a binary gradient: mobile phase A was water with 0.1% formic acid, and mobile phase B was acetonitrile with 0.1% formic acid. The gradient was applied as follows: 5%–30% B over 60 min, followed by 30%–90% B in 1 min, then held at 90% B for 5 min before re‐equilibration. The column temperature was maintained at 40°C throughout the analysis. Mass spectrometry was performed using a timsTOF in the positive ion mode. Full MS spectra were acquired over an m/z range of 100–2800 at a scanning rate of 1 Hz. For MS/MS analysis, a basic‐stepping collision energy was applied during CID fragmentation to optimize the fragmentation efficiency across glycopeptides of m values/z, as described previously [7]. Each individually prepared sample was injected into the nLC–MS/MS system as a technical replicate.

2.5. Data Analysis

This study investigated glycopeptides belonging to human thyroglobulin, as reported in the literature. We used our previous report to determine the mass spectrometric signals of N‐glycopeptides of thyroglobulin. In addition, MSFragger software was employed for data analysis using the N‐glyco HCD method defined in the software [24]. The results were transferred to Excel files. Moreover, a classification process was performed to identify glycopeptide signals in the obtained MS/MS spectra using Python‐based software developed in our laboratory (Figure S1). Glycopeptide signals were determined by scanning the diagnostic oxonium ion signals of glycopeptides using the software throughout the MS/MS spectrum. The list obtained from these analyses was classified according to peptide fragment mass. Glycopeptide spectra were manually confirmed using data analysis software (Bruker Daltonik, GmHB). In addition to these analyses, glycopeptide signals were determined manually by extracting oxonium ions such as 204.0867 [HexNAc1 + H]⁺, 366.1394 [Hex1HexNAc1 + H]⁺, 292.1027 [NeuAc1 + H]⁺, and 657.2348 [Hex1HexNAc1NeuAc1 + H]⁺. The areas of the signals of the detected glycopeptides were extracted manually using the data analysis software. The first three isotopes of the glycopeptide signal were used to calculate the relative abundance of each glycopeptide. After obtaining the area information of each glycopeptide, the relative abundance of each glycopeptide in terms of glycosylation sites was calculated using the total area normalization approach and used in the statistical analyses.

2.6. Statistical Analysis

The analyses were performed using the JASP software. Statistical analyses were conducted to identify significant differences in glycopeptide levels between the groups. The normality of all numerical variables was assessed using the Shapiro–Wilk test, considering variables with a p value greater than 0.05 as normally distributed. Missing data were excluded from the analyses. Independent samples t‐tests were applied for normally distributed variables, whereas the Mann–Whitney U test was used for non‐normally distributed variables. A significance threshold of p < 0.05 was applied to detect meaningful differences. The results from both statistical tests were combined to enhance the robustness of identifying significant changes and were saved in CSV format for further interpretation and visualization. Violin plots were generated using JASP. Receiver operating characteristic (ROC) curve analysis and area under the curve (AUC) calculations were performed using Origin 2024 Pro software to evaluate the diagnostic accuracy of the candidate site‐specific glycan variants. The mean values were calculated on the basis of the number of valid data points (n) for each group. The exact number of data points used for the statistical evaluation and mean calculation of each specific glycopeptide profile is detailed in Table S3.

3.1. Site‐Specific N‐Glycoproteomic Profiling of Thyroglobulin

In the glycoproteomic analysis of thyroid cancer tissues, a bottom‐up proteomic approach was performed using the following specific steps. First, thyroid cancer tissues were collected, and FFPE tissue samples were then deparaffinized. This process removed paraffin from the tissues, and protein extraction was performed thereafter. The extracted proteins were subjected to in‐solution proteolytic digestion using trypsin/Lys‐C for mass spectrometric analysis, and the resulting peptides were analyzed using an nLC–MS/MS instrument. Glycopeptide signals were identified through MSFragger analysis, cross‐referenced with our previous findings, and verified manually. The glycopeptides identified in these analyses are listed in Table S1 and Figure S2. The proteins identified in the thyroid glands in this study are also listed in Table S2. A total of 234 proteins associated with thyroid tissues were identified. Among these, thyroglobulin was the most abundant glycoprotein, as indicated by the highest number of identified peptides and spectral counts in the present study. In this study, nine of the 20 comprehensive N‐glycosylation sites, which were previously mapped in detail in our earlier study on isolated human thyroglobulin [7], were characterized in the analysis of tissues obtained from papillary thyroid cancer and adjacent healthy controls (Figure S3). This limited detection, compared to purified protein analysis, may be due to the challenges associated with working with unfractionated, complex tissue lysates. Although trypsin cleaves proteins based on specific amino acid sequences, some glycosylation sites are not optimally cleaved because of steric hindrance from the glycans, limiting the formation of the targeted glycopeptides. Furthermore, without an initial isolation step for thyroglobulin, the signals of its specific glycopeptides are heavily suppressed by the massive abundance of co‐eluting background peptides from the total tissue proteome [25]. Furthermore, in mass spectrometric analysis, the signals of glycopeptides can be suppressed by more abundant co‐eluting peptides [26]. Despite the use of ZIC‐HILIC enrichment in the current study, we did not observe an improvement in the glycopeptide detection rate. As a result, 39 different glycopeptides belonging to nine glycosylation sites were identified. The study identified 16 distinct N‐glycan compositions, with HexNAc(2)Hex(5) being the most prominent glycan composition observed (Figure S4). The analysis revealed that glycans with high mannose compositions, specifically HexNAc(2)Hex(4–9) glycans, were predominant, indicating that most glycosylation sites determined in this study were high‐man sites. This finding aligns with prior comprehensive site mapping of thyroglobulin, which reported that 6 out of 20 N‐glycosylation sites include exclusively high‐mannose glycans [7].

3.2. Differential Distribution of Glycan Compositions in PTC vs. Control Tissues

Quantitative comparisons of site‐specific glycan composition abundances revealed clear differences between PTC tissues and adjacent normal tissues. The descriptive statistical data for the N‐glycosylation sites identified in this study are presented in Table S3. Table 2 presents the glycopeptides that were significantly altered between the control and tumor groups on the basis of their relative amounts. The table presents the results of independent sampling tests for glycan compositions at each glycosylation site. The Student and Mann–Whitney tests were applied to determine the differences between the two groups, and p values were determined. When thyroid cancer‐associated glycan composition changes were evaluated, some glycan compositions were significantly different in tumor tissue compared to the control group. Asn484‐HexNAc(4)Hex(5)Fuc(1)NeuAc(1) glycan composition was found to be statistically significant (p = 0.012) and decreased in tumor tissue with a fold ratio of 0.62. For the Asn1365‐HexNAc(3)Hex(5) glycan composition, p = 0.037, this glycan composition increased in tumor tissues with a fold change of 1.57. Similarly, the Asn1365‐HexNAc(3)Hex(6) glycan composition was significant with a p value of 0.007, and its level in the tumor increased (fold‐change ratio = 1.68). Asn1365‐HexNAc(3)Hex(5)NeuAc(1) was significant with p = 0.020 and showed a decrease in tumors (fold ratio = 0.45). The Asn2582‐HexNAc(2)Hex(5) glycan composition was significantly different (p = 0.033), indicating a decrease in tumor tissue (fold ratio = 0.37), whereas the p = 0.030 value for the Asn2582‐HexNAc(2)Hex(8) glycan composition revealed an increase in tumor tissue (fold ratio = 1.50). In addition, Asn2582‐HexNAc(2)Hex(9) was significantly increased in tumor tissues (fold change = 1.32).

We observed a notable decrease in several complex‐type glycan compositions containing sialic acid in the tumor group compared to the controls. For example, the average relative abundance of the mono‐sialylated Asn1365‐HexNAc(3)Hex(5)NeuAc(1) glycan in the control group was 30.79, which dropped to 18.27 in the tumor group. Similarly, the Asn1365‐HexNAc(3)Hex(6)NeuAc(1) glycan levels averaged 29.69 in controls and 22.32 in tumors. In addition, Asn484‐HexNAc(4)Hex(5)NeuAc(1) levels were found to be lower in tumors than in controls. All three significantly altered sialylated glycans showed a reduction in the tumor samples, indicating a decrease in sialic acid–terminated structures during the malignant transformation of the thyroid. Consistent with our results, earlier biochemical analyses found that tumor‐derived Tg carries significantly less sialic acid than normal thyroid tissue [27]. A previous study suggested that diminished sialylation may be a distinguishing feature of malignant thyroid lesion progression [28]. However, other studies have reported increased sialylation in thyroid diseases, especially upregulation of α2,3‐linked sialic acids [29].

One specific glycopeptide highlight is the N‐glycan profile at Asn484. The dominant glycan composition at this site in normal tissue was a fucosylated, monosialylated complex glycan Asn484‐HexNAc(4)Hex(5)Fuc(1)NeuAc(1), which was significantly reduced in the tumor. In the control group, this glycan composition was predominant (tumor/control fold change: 0.62), whereas in PTC tissue, its level was much lower. Interestingly, a prior global N‐glycome profiling of PTC tissue observed the opposite trend [15]. In that MALDI‐TOF‐MS‐based glycomic study, the same fucosylated monosialylated glycan (composition H4N5F1E1 in their notation) was found to be upregulated in PTC tumor tissue relative to normal tissue. This discrepancy between our site‐specific findings and bulk tissue glycomics highlights the importance of a site‐resolved analysis. Although our analysis provides detailed insights into nine specific N‐glycosylation sites, the overarching trends observed in previous whole‐tissue studies may be largely driven by the remaining uncharacterized sites, highlighting the complementary nature of site‐specific and global glycan profiling. Thus, site‐specific N‐glycosylation analysis provides a finer lens for detecting changes that may be diluted or even counterintuitive in total glycan measurements.

Figure 1 shows the violin plots of the differential glycosylation levels of glycopeptides between the tumor and control groups. For example, Asn1365‐HexNAc(3)Hex(5), Asn1365‐HexNAc(3)Hex(6), Asn2582‐HexNAc(2)Hex(8), and Asn2582‐HexNAc(2)Hex(9) had substantially increased mean glycosylation levels in the tumor group but showed a more significant variance, indicating their cancer‐specific glycosylation heterogeneity. In contrast, Asn1365‐HexNAc(3)Hex(6)NeuAc(1) had less variance but consistently high means, indicating that they were more homogeneously altered as a function of tumor samples. In addition, glycopeptides, including Asn2582‐HexNAc(2)Hex(5), Asn2582‐HexNAc(2)Hex(8), and Asn2582‐HexNAc(2)Hex(9), displayed broad variance in the samples taken from the tumor with increasing variability, suggesting a link to tumor heterogeneity or microenvironmental differences. Among the identified glycopeptides, Asn1365‐HexNAc(3)Hex(6) exhibited the highest fold change, with a fold ratio value of 1.68 (Table 2).

The ROC curves in Figure 2 show the diagnostic performances of the glycosylation profiles. Asn1365‐HexNAc(3)Hex(6) exhibited the highest diagnostic potential, with an AUC of 0.81, indicating good discrimination between the two groups. Similarly, Asn1365‐HexNAc(3)Hex(5) and Asn2582‐HexNAc(2)Hex(8) were determined to have moderate diagnostic abilities, with AUC values of 0.76 and 0.72, respectively. Although Asn2582‐HexNAc(2)Hex(9) showed an AUC of 0.72, its performance suggests that it could be complementary when combined with other profiles. The ROC analysis results suggest that these glycosylation profiles, especially Asn1365‐HexNAc(3)Hex(6), could be candidates for discriminating between the tumor and control groups.

The distribution of the four glycopeptide variants [Asn1365‐HexNAc(3)Hex(6), Asn1365‐HexNAc(3)Hex(5), Asn2582‐HexNAc(2)Hex(8), and Asn2582‐HexNAc(2)Hex(9)], identified based on their ROC scores, is presented in Figure 3. The t‐SNE dimensionality reduction (Figure 3A) indicates a general tendency for the control and tumor samples to form observable groups in a two‐dimensional space based on these four candidates. In contrast, the heat map generated following the k‐means merge step (Figure 3B) reflects the overall distribution of the samples based on the profiles of the four selected key glycopeptide variants. An examination of the hierarchical clustering dendrogram revealed that although separation between the tumor and control groups was observable in certain regions, strict boundaries were not fully formed, resulting in a partial overlap where samples from both groups clustered together within some main branches. The color distribution in the heat map illustrates the changes in the relative glycopeptide abundance. Overall, the glycan forms attached to the Asn2582 site generally increased, particularly in clusters predominantly composed of tumor samples, whereas the levels of the Asn1365 variants displayed a more variable profile of increases and decreases across the samples and clusters.

The capability of the four selected key glycopeptide variants to differentiate between the tumor and control groups was evaluated using various supervised machine learning algorithms, with the resulting performance metrics summarized in Table S4. Among the tested models, the highest AUC values were obtained by the neural network (AUC = 0.828) and support vector machines (SVM, AUC = 0.818). Conversely, the random forest algorithm presented a relatively more balanced profile across other metrics, such as classification accuracy (CA = 0.765) and F1‐score (F1 = 0.762). The evaluation results for the neural network model are shown in Figure 4. The ROC curve of the model demonstrated an overall AUC value of 0.83 (Figure 4A). To determine the contribution of each glycopeptide to the predictive power of the model, a feature importance analysis based on the decrease in AUC was conducted (Figure 4B). The results indicated that Asn1365‐HexNAc(3)Hex(6) and Asn1365‐HexNAc(3)Hex(5) provided the highest contributions to the model performance, whereas the variants at the Asn2582 site exhibited a relatively lower impact. Additionally, the distribution plot illustrating the impact of feature values on the model output (Figure 4C) shows that across all four glycopeptides, higher relative abundance values (red dots) generally exert a positive impact on the model prediction output, whereas lower values (blue dots) exert a negative impact.

4.1. Biological Implications of Aberrant Thyroglobulin Glycosylation in PTC

Studies have revealed that thyroid cancer tissues significantly increase α1–6 fucosylation (core fucosylation) via the Fut8 enzyme in papillary thyroid cancer [30]. Fut8 expression is elevated during the early stages of carcinogenesis but declines in the advanced stages [16]. Figure S5 shows the control tumor comparisons of fucose‐containing glycan compositions. This graph compares the mean glycan composition values between the control and tumor groups. For each glycan composition, the mean values in the tumor group were analyzed to determine whether they increased or decreased compared to those in the control group. In six of the glycan compositions examined, the mean values were higher in the tumor group than in the control group. It can be said that these glycan compositions are found at higher levels in tumor samples, so there is an increasing trend toward these glycan compositions in the tumor group. In contrast, the mean values of the remaining five glycan compositions decreased in the tumor group compared to the control group. These structures were found at lower levels in the tumor group, which can be interpreted as indicating that these glycan compositions are less common in tumor samples. According to the data shown in Figure S5, although there was an increase in some fucose‐containing glycan compositions in the tumor group, other structures tended to decrease. However, there was a relative decrease in fucosylation in the tumor group compared to the control group; however, when this was evaluated with standard errors, the change was insignificant.

The altered N‐glycan patterns observed in PTC thyroglobulin align with the known changes in the expression of glycosylation enzymes in thyroid cancer. Notably, the N‐acetylglucosaminyltransferase V enzyme (GnT‐V, encoded by MGAT5), which creates β1,6‐linked GlcNAc branching on N‐glycans, has been reported to be highly expressed in PTC tissues [31]. Increased GnT‐V activity is expected to drive more extensive branching (complex‐type elaboration) of N‐glycans. Indeed, a recent study found that PTC specimens exhibit higher levels of N‐glycan branching compared to normal thyroid, consistent with up‐regulated glycosyltransferases in the tumor [32]. These enzyme‐driven modifications in PTC would lead to fewer high‐mannose glycans and more complex multi‐antennary glycans with core fucosylation. In our data, when considering all identified high‐mannose‐type glycans collectively (Figure S6), we found a general decrease in smaller high‐mannose species (HexNAc(2)Hex(4), ‐(5), and ‐(6)) in the tumor compared to the control for several sites. This indicates that those less processed glycans are present at lower levels in cancer. Conversely, we observed an increase in the larger high‐mannose structures, HexNAc(2)Hex(8) and HexNAc(2)Hex(9), in the tumor group compared to the control group, and these increases were statistically significant. This somewhat counterintuitive pattern (loss of Man_4–6 but gain of Man_8–9) may reflect site‐dependent differences in processing efficacy or altered glycan trimming in tumor thyroid tissue. It is possible that at certain glycosylation sites, glycan processing stands at larger high‐mannose structures in the tumor, whereas other sites that would normally carry smaller high‐mannose forms in healthy tissue become further processed or are differentially regulated in PTC.

4.2. Clinical Relevance and Biomarker Potential of Site‐Specific Glycan Traits

Glycosylation‐based biomarkers are emerging as important tools for the diagnosis and prognosis of thyroid cancer. Our identification of a tumor‐enriched Tg glycan composition adds to this growing body of evidence. For example, recent plasma glycomic studies have defined panels of N‐glycan traits that distinguish malignant from benign thyroid nodules [33]. Zhang et al. demonstrated that differences in glycan complexity, galactosylation, fucosylation, and sialylation in patient plasma could classify thyroid cancer versus benign nodules. Interestingly, N‐glycosylation changes in IgG have also shown diagnostic value: a specific decrease in IgG bisected N‐glycans (trait “BN”) was able to discriminate PTC from non‐cancer with ∼92% accuracy in a training cohort [19]. Beyond the initial diagnosis, glycan alterations have been linked to disease recurrence. In one study, three structurally related serum N‐glycans (core‐fucosylated biantennary glycans differing in galactose content) have been found at altered levels in patients with recurrent thyroid cancer, and the ratio of agalactosylated to mono‐galactosylated species (G0F/G1F) predicted recurrence with an AUC of 0.82 [13]. Likewise, increases in specific linkage‐specific sialylation patterns have been observed in PTC patients who later developed recurrence [34], suggesting sialylated glyco‐epitopes can signal aggressive disease. Taken together, these studies illustrate the considerable potential of glycosylation changes as biomarkers for thyroid cancer. Our finding of a site‐specific glycan composition (HexNAc(3)Hex(6) at Asn1365 that is significantly upregulated in PTC may further support this paradigm.

Although serum thyroglobulin (Tg) levels are widely used in the clinical diagnosis and monitoring of PTC, their specificity can be compromised in the presence of benign conditions such as thyroiditis or goiter, particularly in patients with residual thyroid tissue following surgery. Although useful for monitoring, this physiological feature may limit its diagnostic accuracy when used standalone. In contrast, the molecular‐level resolution of site‐specific glycosylation profiling in tissue samples may capture cancer‐associated alterations that are not evident in total serum Tg levels. Although a direct comparison with serum Tg measurements is beyond the scope of this study, our findings suggest that glyco‐biomarkers may serve as valuable adjuncts to current diagnostic tools. Future studies integrating glycoproteomic data with serum‐based assays may enhance diagnostic precision and aid personalized clinical decision‐making in thyroid cancer management. Although the extensive glycosylation of thyroglobulin offers profound biological insights, its inherent complexity presents distinct analytical challenges for site‐specific characterization in clinical diagnostics. Therefore, future biomarker studies may benefit from focusing on glycoproteins with fewer, well‐characterized glycosites or a select group of thyroglobulin sites with the most consistent and diagnostically important glycan changes.

4.3. Study Limitations and Technical Perspectives

One limitation of the current study is the number of identified glycopeptides and glycosylation sites, which may appear to be limited given the complexity of the thyroid tissue proteome. In this study, nLC–MS/MS analysis was conducted at a relatively slow scan speed (1 Hz), which improved the quality of the MS/MS spectra for glycopeptides but reduced the total number of acquired spectra per run. Another contributing factor is the high abundance of thyroglobulin (Tg) in thyroid tissues, which may suppress the ionization and detection of less abundant proteins. Tg is the most abundant glycoprotein in the thyroid gland, often dominating the protein content and masking signals from lower abundance components [35]. In addition, the complex nature of FFPE tissue samples and the absence of fractionation steps further constrain the detection of low‐abundance or poorly ionized glycopeptides. Consequently, not all potential glycosylation sites of Tg could be reliably identified, and the full spectrum of Tg glycan compositions may not have been comprehensively captured in this study. Future studies would benefit from targeted fractionation strategies, such as immunoaffinity purification or immunoprecipitation, to specifically isolate thyroglobulin prior to digestion. Such approaches would be highly viable for drastically reducing background proteome complexity, minimizing ion suppression, and ultimately achieving a much deeper and more comprehensive coverage of Tg's N‐glycosylation sites.

In summary, the N‐glycosylation changes in thyroglobulin in PTC versus normal tissue reveal meaningful differences at both the site‐specific and global levels. Papillary thyroid tumor tissues tend to exhibit aberrant glycosylation, characterized by a prevalence of high‐mannose glycans, loss of specific terminal sialic acid‐containing structures, and emergence of highly branched modified glycans. By providing a detailed map of site‐specific glycan composition alterations in a key thyroid protein, our results may offer new insights into the dysregulation of glycan processing in PTC. This study lays the foundation for future research to validate glycopeptide markers in larger cohorts and explore the functional consequences of these glycosylation changes on thyroid cancer biology and progression. Integrating glycoproteomic markers with existing diagnostic modalities could improve the accuracy of thyroid cancer diagnosis and disease recurrence monitoring.