Microscaled Cell Surface Proteomics for Cryo-Preserved Cells and Tissue Samples.

To compare performance across input levels, we performed a titration series using 0.1 × 10 6 to 5 × 10 6 HEK293T cells with two mechanistically distinct surface enrichment strategies. The glycopeptide-based approaches labels surface glycoproteins via periodate oxidation of glycans followed by conjugation to amine-reactive biotin ( Supplemental Fig. S1 A ). After proteolytic digestion, and washing to remove non-glycosylated peptides, N-linked glycopeptides are selectively released by PNGase F. This method is abbreviated here as N-Glycopeptide Enrichment (NGE). Filtering of the data for peptides containing N-X-S/T motifs yields a list of formerly N-glycosylated proteins ( Fig. 1 A ) ( 5 ). Analyses were performed at the peptide level based on confidently localized deamidation sites in peptides containing the N-linked glycosylation sequence motif retained for downstream analysis ( Supplemental Fig. S1 B ). Some caveats with the approach include the need for the analysis system to be able to robustly and repeatedly detect single or small numbers of formerly glycosylated peptides derived from a protein, and that the ability to detect is dependent on the abundance and capture efficiency of underlying glycopeptides, that can vary depending on glycosylation occupancy and site accessibility. In contrast, the WGA-HRP strategy localizes horseradish peroxidase to cell surface glycoproteins containing NeuNAc (N-acetylneuraminic acid) and/or GlcNAc (N-acetylglucosamine), where it catalyzes oxidation of a biotin-phenol to label proximal tyrosine residues on nearby proteins ( Supplemental Fig. S1 C ) ( 10 , 20 , 21 ). Labeled proteins are then captured and digested directly on-bead, enabling recovery of peptides from both glycosylated and non-glycosylated surface proteins. In contrast to the NGE method, detection in the WGA-HRP method is done using non-glycosylated peptides ( Fig. 1 A ). A caveat of this method is that it produces a significant amount of intracellular background, complicating confident identification of true cell surface proteins ( 6 ). This is also the case for approaches using the Sulfo-NHS-biotin approach ( 10 ). Fig. 1 Validation and comparison of glycan oxidation and WGA-HRP labeling protocols across HEK293T cell input titration. A , Downstream workflows for cell surface proteomics. For glycan-labeled samples, cells are lysed, proteins digested, biotinylated peptides enriched, and N-glycopeptides released with peptide-N4-(N-acetyl-β-D-glucosaminyl) asparagine amidase F prior to mass spectrometry analysis. For WGA-horseradish peroxidase (HRP)–labeled samples, cells are lysed, biotinylated proteins enriched, digested on -bead, and resulting peptides are analyzed by mass spectrometry. All data were generated using an Exploris 480 and searched using Spectronaut version 19 (see Methods for details). B , Total number of surface protein identifications across input amounts ranging from 1 × 10⁵ to 5 × 10⁶ cells for each labeling strategy (N = 3). Control samples were generated using 1 × 10⁶ cells and represent labeling reactions performed in the absence of key reagents: sodium periodate (NaIO₄) for N-glycopeptide enrichment (NGE) and wheat germ agglutinin–horseradish peroxidase (WGA-HRP) for WGA-HRP labeling. C , Violin plots show the distribution of average cell surface protein (CSP) intensities. Dots represent median log₂ intensity per input. Lines connect medians across inputs. Numeric labels indicates the log₂ change in median intensity between adjacent inputs. ( i.e., the slope between points) D , Venn diagram showing overlap of identified surface proteins across all input amounts between glycan- and WGA-HRP–labeled samples, STRING network analysis of proteins uniquely identified by each method, annotated for functions in cell adhesion and ion transport with BH FDR <0.01. E . Distribution of predicted extracellular NXS/T glycosylation motifs per protein, showing that proteins unique to WGA-HRP have low numbers of extracellular potential glycosylation sites ( p -value <0.01, Wilcoxon rank-sum test). CSP, cell surface protein; HRP, horseradish peroxidase; NGE, N-glycopeptide enrichment; WGA-HRP: wheat germ agglutinin–horseradish peroxidase Validation and comparison of glycan oxidation and WGA-HRP labeling protocols across HEK293T cell input titration. A , Downstream workflows for cell surface proteomics. For glycan-labeled samples, cells are lysed, proteins digested, biotinylated peptides enriched, and N-glycopeptides released with peptide-N4-(N-acetyl-β-D-glucosaminyl) asparagine amidase F prior to mass spectrometry analysis. For WGA-horseradish peroxidase (HRP)–labeled samples, cells are lysed, biotinylated proteins enriched, digested on -bead, and resulting peptides are analyzed by mass spectrometry. All data were generated using an Exploris 480 and searched using Spectronaut version 19 (see Methods for details). B , Total number of surface protein identifications across input amounts ranging from 1 × 10⁵ to 5 × 10⁶ cells for each labeling strategy (N = 3). Control samples were generated using 1 × 10⁶ cells and represent labeling reactions performed in the absence of key reagents: sodium periodate (NaIO₄) for N-glycopeptide enrichment (NGE) and wheat germ agglutinin–horseradish peroxidase (WGA-HRP) for WGA-HRP labeling. C , Violin plots show the distribution of average cell surface protein (CSP) intensities. Dots represent median log₂ intensity per input. Lines connect medians across inputs. Numeric labels indicates the log₂ change in median intensity between adjacent inputs. ( i.e., the slope between points) D , Venn diagram showing overlap of identified surface proteins across all input amounts between glycan- and WGA-HRP–labeled samples, STRING network analysis of proteins uniquely identified by each method, annotated for functions in cell adhesion and ion transport with BH FDR <0.01. E . Distribution of predicted extracellular NXS/T glycosylation motifs per protein, showing that proteins unique to WGA-HRP have low numbers of extracellular potential glycosylation sites ( p -value <0.01, Wilcoxon rank-sum test). CSP, cell surface protein; HRP, horseradish peroxidase; NGE, N-glycopeptide enrichment; WGA-HRP: wheat germ agglutinin–horseradish peroxidase To benchmark performance in cell lines with a focus on specificity, sensitivity and reproducibility we compared both approaches across an input range of HEK293T cells and analyzed the resulting protein lists using the in silico human surfaceome database (SURFY, Supplemental Table S1 A ( 22 ), to define confident CSPs. This provided a consistent framework for evaluating coverage of proteins in the known CSP list, where specificity was defined here as the percentage of proteins identified as CSPs relative to the total number of proteins detected by that method ( 22 ). Both enrichment strategies recovered approximately 200 CSPs in the list of known CSPs at the lowest input (1 × 10 5 cells) and reached a maximum of ∼500 CSPs at 2.5 × 10 6 cells, with no appreciable gain observed at 5 × 10 6 ( Fig. 1 B , Supplemental Table S1 , B and C ). Consistent with these observations, intensity distributions plateaued at higher input levels for both methods, further indicating signal saturation beyond 2.5 × 10 6 cells ( Fig. 1 C ). At intermediate inputs (2.5 × 10 5 and 5 × 10 5 ), the NGE method consistently identified a slightly higher number of CSPs than WGA-HRP ( Fig. 1 B ). Cumulatively across all inputs, NGE identified 580 CSPs, while WGA-HRP recovered 511, with an overlap of 373 proteins—highlighting that each method captures a distinct subset of the surface proteome ( Fig. 1 D , Supplemental Table S1 D ). Gene ontology analysis showed that proteins uniquely identified by the NGE method were enriched for cell adhesion functions, whereas those unique to WGA-HRP were enriched for ion transport ( Fig. 1 D ). Domain-based analysis using topological prediction revealed that many of the proteins uniquely identified by the WGA-HRP method had few N-X-S/T glycosylation motifs in predicted extracellular regions ( Fig. 1 E ) highlighting its ability to identify proteins that may not be captured using the NGE method ( 23 ). Among these, a subset of ion transporters exhibited structural features, such as short extracellular loops with limited glycosylation, that likely explain their absence from the NGE dataset. A representative topological map of one such protein is shown in Supplemental Fig. S1 D ( 24 ). Follow-up experiments described below provide further insight into the enrichment of cell adhesion proteins by the NGE approach. Reproducibility was high for both approaches, with WGA-HRP exhibiting slightly lower CSP coefficients of variation (CVs) across inputs (median CV < 20%). NGE CVs were higher at the lowest inputs but dropped below 20% at 5 × 10 5 cells and above ( Supplemental Fig. S1 E , Supplemental Table S1 , E and F ). To assess consistency across the input range, we calculated the correlation of CSP quantification between 1 × 10 5 and 5 × 10 6 cells for each method. NGE showed higher quantitative agreement across inputs (Pearson r = 0.87), compared to WGA-HRP (r = 0.65) ( Supplemental Fig. S1 F ). This likely reflects greater confidence in low-input NGE measurements, supported by oxidation-minus controls in which CSP identifications dropped to near zero ( Fig. 1 A ). In contrast, WGA-HRP identifications in the absence of WGA-HRP remained elevated (∼80 CSPs), consistent with background signal independent of WGA-mediated binding. ( Fig. 1 A ). Correlation coefficients between the two enrichment methods at each input level ranged from 0.4 to 0.6, which is unsurprising given the distinct quantification strategies employed: protein-level quantification derived from captured glycopeptides in NGE experiments, compared to the use of all identifiable peptides from WGA-HRP–captured proteins. ( Supplemental Fig. S1 G ). Lastly, the proportion of identified proteins classified using SURFY as true CSPs was markedly higher for NGE (∼50%) compared to WGA-HRP (<10%), consistent with the elevated background associated with protein capture–based enrichment strategies ( Supplemental Fig. S1 H ). Both strategies were deployed to evaluate both the robustness of low-input workflows across different cell types and their compatibility with cryopreserved material. We used the suspension myeloma cell line KMS-12-BM and the adherent lung adenocarcinoma cell line A549 under fresh and cryopreserved conditions. KMS-12-BM samples were processed at 1 × 10 6 (low) and 15 × 10 6 (high) cell inputs for fresh conditions, and at ∼1 × 10 6 input from cryopreserved material. Approximately 1.3 × 10 6 cells were frozen using a DMSO-based freezing protocol and stored in liquid nitrogen (see Methods) ( Fig. 2 A ). After rapid thawing in warm media and an additional PBS wash, recovered cell numbers based on manual counting were comparable to the fresh low-input condition ( Supplemental Fig. S2 A ). Fig. 2 Surface proteomic profiling of cryopreserved and fresh cell lines across input levels. A , Experimental design for KMS-12-BM cryopreservation and surface labeling. Cells were frozen in 10% dimethyl sulfoxide/fetal bovine serum, stored in liquid nitrogen, and rapidly thawed for enrichment and analysis. B , Bar plot showing CSP overlap between NGE and WGA-HRP methods at high and low inputs in KMS-12-BM cells. Partial overlap indicates method-specific subsets. C , Rank plot of RNA expression (log₂ TPM) for surface proteins uniquely identified by NGE versus WGA-HRP in KMS-12-BM. NGE proteins are enriched at lower transcript levels. D , Log₂ (fresh/frozen) differences are plotted against the average protein intensity across both methods. Solid lines indicate the mean difference (NGE: −0.051, orange ; WGA-HRP: 0.171, blue ); dashed lines represent ±1.96×SD (NGE: ±0.431; WGA-HRP: ±0.462), capturing the expected spread of variation for each method. E , Experimental design comparing A549 surface proteomes across three conditions: fresh adherent, freshly detached, and cryopreserved post-dissociation. Low input samples used 6-well format; high input fresh adherent cells used 100 mm dishes. F , Bar plot showing CSP overlap between NGE and WGA-HRP at high and low fresh inputs in A549. G , Scatter plots and histograms showing consistent surface protein quantification across A549 conditions (plate-based labeling and versene detachment—fresh and frozen) for the NGE method. CSP, cell surface protein; HRP, horseradish peroxidase; NGE, N-glycopeptide enrichment; WGA-HRP: wheat germ agglutinin–horseradish peroxidase Surface proteomic profiling of cryopreserved and fresh cell lines across input levels. A , Experimental design for KMS-12-BM cryopreservation and surface labeling. Cells were frozen in 10% dimethyl sulfoxide/fetal bovine serum, stored in liquid nitrogen, and rapidly thawed for enrichment and analysis. B , Bar plot showing CSP overlap between NGE and WGA-HRP methods at high and low inputs in KMS-12-BM cells. Partial overlap indicates method-specific subsets. C , Rank plot of RNA expression (log₂ TPM) for surface proteins uniquely identified by NGE versus WGA-HRP in KMS-12-BM. NGE proteins are enriched at lower transcript levels. D , Log₂ (fresh/frozen) differences are plotted against the average protein intensity across both methods. Solid lines indicate the mean difference (NGE: −0.051, orange ; WGA-HRP: 0.171, blue ); dashed lines represent ±1.96×SD (NGE: ±0.431; WGA-HRP: ±0.462), capturing the expected spread of variation for each method. E , Experimental design comparing A549 surface proteomes across three conditions: fresh adherent, freshly detached, and cryopreserved post-dissociation. Low input samples used 6-well format; high input fresh adherent cells used 100 mm dishes. F , Bar plot showing CSP overlap between NGE and WGA-HRP at high and low fresh inputs in A549. G , Scatter plots and histograms showing consistent surface protein quantification across A549 conditions (plate-based labeling and versene detachment—fresh and frozen) for the NGE method. CSP, cell surface protein; HRP, horseradish peroxidase; NGE, N-glycopeptide enrichment; WGA-HRP: wheat germ agglutinin–horseradish peroxidase Across the fresh input range, we identified approximately 400 to 440 confident surface proteins per protocol ( Fig. 2 B , Supplemental Table S2 , A and B ). Intra-method overlap of identified CSPs between low and high input levels was high indicating consistent detection across input levels. However, overlap between methods was modest (∼260 CSPs shared), reinforcing the observation that each strategy captures a partially distinct subset of surface proteins ( Fig. 2 B , Supplemental Table S2 C ). As seen in HEK293T cells, proteins uniquely identified by WGA-HRP were enriched for ion transport functions, while NGE-specific proteins were dominated by cell adhesion functions with these observed enrichments based at least in part by the presence/absence and abundance of accessible N-glycosylation sites ( Supplemental Fig. S2 B ). While the dynamic range of signal intensities appeared similar between methods ( Supplemental Fig. S2 C ), integration with KMS-12-BM RNA-seq data revealed that NGE uniquely captured surface proteins with lower transcript abundance ( Fig. 2 C ). These proteins were primarily associated with cell adhesion and likely detected due to the large, heavily glycosylated extracellular domains characteristic of this functional class. A representative topological map of one such cell adhesion protein is shown in Supplemental Fig. S2 D . Comparison of low-input fresh versus cryopreserved KMS-12-BM samples showed near-complete overlap in CSP identifications for both enrichment methods, with Pearson correlation coefficients ≥0.95 for both ( Supplemental Fig. S2 E , Supplemental Table S2 , D and E ). To assess the presence of any systematic intensity shift, we performed a Bland-Altman analysis, which confirmed strong agreement, with the vast majority of proteins falling within ±2 standard deviations of the average log 2 fold change (frozen versus fresh). Minor shifts in average fold change were observed (−0.051 for NGE, 0.171 for WGA-HRP), reflecting minimal changes in surface signal post-thaw ( Fig. 2 D ). We then extended our analysis to the adherent A549 cell line, which presented the additional challenge of detachment prior to freezing. Cells were labeled in culture dishes at two input scales: 6-well plates (low input) and 10 cm dishes (high input). For the low-input condition, we used the NGE method to compare fresh cells labeled directly on-plate with freshly detached cells using a cell scraper, versene, or trypLE, all of which have been shown to better preserve surface protein integrity compared to classic trypsinization ( Fig. 2 E ). Similar to the KMS-12-BM findings, strong overlap was observed between high- and low-input samples within each method, and moderate overlap was observed between the different methods ( Fig. 2 F , Supplemental Table S2 , F and G ). Additionally, RNAseq data from A549 cells validated the trend of NGE’s unique ability to capture a low abundance set of surface proteins ( Supplemental Fig. S3 A ). In comparing plate-labeled and detached-labeled cells, protein yields following lysis showed that trypLE recovered the largest number of cells (avg. 385 μg), versene and scraping recovered the lowest (avg. 296 μg, 276 μg, respectively), and plate-based labeling recovered an intermediate amount (avg. 328 μg) ( Supplemental Fig. S3 B ). Despite recovering the least total protein, versene-treated and scraped cells recovered the largest number of surface proteins and NXS/T peptides ( Supplemental Fig. S3 , C and D , Supplemental Table S2 H ). While the total number of surface proteins was similar and largely overlapping across methods, differences in NXS/T peptide identification were more pronounced. For example, in versene- and trypLE-treated cells, CSP identifications were 566 and 549, respectively, and NXS/T peptides were 3461 and 2915, respectively. The peptides missing in the trypLE condition came from proteins primarily associated with cell adhesion, cell motility, and neurogenesis, suggesting that minor surface shaving occurs during detachment ( Supplemental Fig. S3 E ). Despite these differences, correlation coefficients between all techniques exceeded 0.9, indicating that all are viable options for surface protein detachment while retaining quantitative integrity ( Supplemental Fig. S3 F ). However, for adherent cell lines, gentle scraping and versene treatment are preferable to trypLE to minimize shaving-related losses. Lastly, we compared frozen A549 cells treated with versene and found a strong correlation (r > 0.9) with both fresh versene-detached and plate-labeled cells ( Fig. 2 G , Supplemental Table S2 H ). We next evaluated the ability of these surface proteomic workflows to capture biologically meaningful, stimulus-induced changes in surface protein abundance. A549 cells stimulated with EGF, a canonical model of receptor internalization ( Fig. 3 A ) ( 25 , 26 ) was used as a model system to capture dynamic changes and evaluate quantitative changes. This experiment also allowed us to leverage the peptide-level nature of the NGE method for matched surface and global deep-scale proteome analysis. Peptides for global proteome analysis were taken prior to streptavidin enrichment of biotinylated glycopeptides, as non-specific losses during enrichment preclude reliable quantification of flow-through peptides. Although biotinylated glycopeptides themselves are not detectable in the global proteome data, non-glycosylated peptides from surface proteins or intracellular pools of these proteins can still be quantified. In contrast, WGA-HRP captures intact proteins rather than peptides, making it incompatible with paired surface and total proteome analysis from the same input. Following EGF stimulation for 10 min or 3 h in 6-well culture plates, both protocols yielded consistent coverage of A549 cell surface proteins (∼530–560 CSPs) with median CSP CVs ∼20% across biological replicates (N = 3). ( Supplemental Fig. S4 , A and B , Supplemental Table S3 , A and B ). Surface epidermal growth factor receptor (EGFR) levels, as measured by the NGE workflow, were significantly reduced at both time points, consistent with rapid internalization ( Fig. 3 B ). In contrast, global proteome EGFR abundance remained essentially unchanged at 10 min, reflecting receptor internalization without degradation, and showed a pronounced decrease by 3 h, consistent with lysosomal degradation ( Fig. 3 C , Supplemental Table S3 C ). Fig. 3 Contrasting surfaceome and proteome dynamics of receptor trafficking in response to EGF stimulation. A , Schematic of epidermal growth factor (EGF)-induced epidermal growth factor receptor (EGFR) internalization and lysosomal degradation, with or without inhibition of lysosomal function by chloroquine. B , Rank plot of NGE–based surfaceome data from control, 10 min. EGF, and 3 h. EGF conditions, highlighting EGFR and high-abundance stable surface proteins. C , Total EGFR protein levels from N-glyco labeling of A549 cells across four conditions: untreated control, 10-min EGF, 3-h EGF, and chloroquine pretreatment followed by 3-h EGF stimulation. EGFR abundance is quantified from surfaceome data (N = 3). Significance determined by unpaired two -tailed t-tests: p > 0.05 (NS), p < 0.0001 (∗∗∗∗). D . Surface EGFR levels from the same conditions and samples, measured by whole-proteome analysis (N = 3). Significance determined by unpaired two -tailed t-tests: p < 0.05 (∗), p < 0.01 (∗∗), p < 0.0001 (∗∗∗∗). E , Volcano plot comparing surfaceome profiles at 3-h EGF versus chloroquine + 3-h EGF, showing reduced surface EGFR and downregulation of proteins involved in EGFR recycling in chloroquine-pretreated samples (log₂-fold change > |1|; p < 0.05 by t test). F , Comparison of surfaceome and proteome log₂ fold changes for overlapping proteins for one time point (3 h. versus control), highlighting that significant regulation is primarily detected in the surfaceome data (log₂-fold change > |1| for proteome and surfaceome data). EGF, epidermal growth factor; EGFR, epidermal growth factor receptor; NGE, N-glycopeptide enrichment. Contrasting surfaceome and proteome dynamics of receptor trafficking in response to EGF stimulation. A , Schematic of epidermal growth factor (EGF)-induced epidermal growth factor receptor (EGFR) internalization and lysosomal degradation, with or without inhibition of lysosomal function by chloroquine. B , Rank plot of NGE–based surfaceome data from control, 10 min. EGF, and 3 h. EGF conditions, highlighting EGFR and high-abundance stable surface proteins. C , Total EGFR protein levels from N-glyco labeling of A549 cells across four conditions: untreated control, 10-min EGF, 3-h EGF, and chloroquine pretreatment followed by 3-h EGF stimulation. EGFR abundance is quantified from surfaceome data (N = 3). Significance determined by unpaired two -tailed t-tests: p > 0.05 (NS), p < 0.0001 (∗∗∗∗). D . Surface EGFR levels from the same conditions and samples, measured by whole-proteome analysis (N = 3). Significance determined by unpaired two -tailed t-tests: p < 0.05 (∗), p < 0.01 (∗∗), p < 0.0001 (∗∗∗∗). E , Volcano plot comparing surfaceome profiles at 3-h EGF versus chloroquine + 3-h EGF, showing reduced surface EGFR and downregulation of proteins involved in EGFR recycling in chloroquine-pretreated samples (log₂-fold change > |1|; p < 0.05 by t test). F , Comparison of surfaceome and proteome log₂ fold changes for overlapping proteins for one time point (3 h. versus control), highlighting that significant regulation is primarily detected in the surfaceome data (log₂-fold change > |1| for proteome and surfaceome data). EGF, epidermal growth factor; EGFR, epidermal growth factor receptor; NGE, N-glycopeptide enrichment. To test whether receptor internalization and degradation could be blocked, we pretreated cells with chloroquine at levels shown to inhibit autophagy prior to 3-h EGF stimulation ( 27 ). Chloroquine led to increased total EGFR levels, consistent with impaired lysosome mediated-degradation, but resulted in even greater depletion of surface EGFR, suggesting continued internalization without efficient recycling ( Fig. 3 D , Supplemental Table S3 D ). When comparing surface profiles between 3-h EGF-stimulated samples with and without chloroquine pretreatment, we observed a significant reduction in SORL1 alongside EGFR ( Fig. 3 E ). SORL1 is known to mediate ERBB2 recycling and was recently implicated in EGFR trafficking ( 28 , 29 ). These findings suggest that lysosomal disruption may interfere with SORL1 function, leading to impaired recycling and enhanced surface depletion of EGFR. While prior studies have shown that SORL1 silencing can impair lysosomal activity, our data reveal the inverse relationship—highlighting how lysosomal dysfunction can potentially reveal SORL1’s role in trafficking regulation ( 29 ). To further assess whether the CSP changes observed could be obtained just using whole proteome analysis, we compared the log fold-change of CSPs identified in both the surface and total proteomes following 3-h EGF stimulation. EGFR was the only protein to show consistent downregulation in both the surface and total proteomes, reflecting its well-characterized internalization and degradation dynamics. In contrast, the majority of surface-regulated proteins showed little to no change at the total proteome level, while significant fold changes were observed exclusively in the surface-enriched fraction ( Fig. 3 F ). These findings highlight the limitations of global proteomics in capturing spatial and dynamic remodeling and underscore the value of surface-specific enrichment for detecting dynamic regulation. Using WGA-HRP enrichment, we similarly observed EGFR surface depletion following 10-min and 3-h EGF stimulation ( Supplemental Fig. S4 C , Supplemental Table S3 E ). However, the chloroquine-treated condition showed only a modest, non-significant decrease in EGFR alongside a significant reduction in SORL1 ( Supplemental Fig. S4 , D and E ). This may reflect labeling and enrichment of EGFR by WGA-HRP in the intracellular pool or compression of the actual fold change due to the significantly higher background protein levels in the WGA-HRP method, both of which are potential limitations in distinguishing surface-localized versus internal protein pools. To enable surface proteomic profiling of clinical lesions with limited input material, we optimized a dissociation and cryopreservation workflow using healthy human endometrial tissue as a model system. We sought to establish a protocol that maximizes post-thaw viability and surface labeling compatibility while working within input constraints typically encountered in this clinical context. Using endometrial tissue from a single donor, initially, we compared three dissociation protocols (see Methods) (1): gentle enzymatic digestion with TrypLE and DNase I for 20 min at 37 °C (T-20) (2), same enzymatic digestion for 60 min (T-60), and (3) mechanical dissociation using the gentle MACS Dissociator ( Fig. 4 A ). Prior experiments performed in our group suggested that TrypLE has superior ability to preserve cell surface epitopes, critical for downstream proteomic applications for complex solid tissue. Unlike traditional enzyme blends such as Liberase, TrypLE also offers higher purity and specificity, minimizing degradation of surface markers during dissociation. Tissue was manually divided into 150 to 200 mg wet weight aliquots and processed in replicate to generate technical pairs. Cell yields ranged from 20,000 to 30,000 cells per mg of tissue across all conditions, resulting in total recoveries sufficient to generate cryopreserved aliquots containing >1.5 × 10 6 cells ( Supplemental Fig. S5 A ). Fig. 4 Optimization and evaluation of a low-input surface proteomics workflow using dissociated endometrial tissue. A , Schematic of the dissociation and cryopreservation workflow. Fresh endometrial tissue was subjected to either enzymatic digestion with TrypLE/deoxyribonuclease I for 20 min 20-Minute TrypLE digestion protocol (T-20) or 60 min (60-Minute TrypLE Digestion Protocol, or mechanical dissociation using the gentle magnetic-activated cell sorting system. Resulting single-cell suspensions were cryopreserved and later used for surface proteomic analysis. B , Overlap of confident CSP identifications among the three dissociation protocols. The majority of proteins identified in the gentle magnetic-activated cell sorting system and 60-Minute TrypLE Digestion Protocol conditions overlapped with those captured using T-20. C , CSP counts, NXS/T motif-containing peptide counts, and total protein input for each titration sample using the T-20 method. Inputs of 0.5 × 10⁶, 1 × 10⁶, and 2 × 10⁶ cells were analyzed for three donors, with one donor profiled in triplicate. CSP yields remained robust even at the lowest input levels. D , UpSet plot showing the number of CSPs shared and uniquely identified across donor–input combinations. High overlap in CSP identifications was observed across donors and input levels. E , Principal component analysis of CSP intensities after experiment-wide median intensity normalization reveals clear donor-specific clustering, demonstrating reproducibility and biological distinction across samples. CSP, cell surface protein; T-20, 20-Minute TrypLE digestion protocol. Optimization and evaluation of a low-input surface proteomics workflow using dissociated endometrial tissue. A , Schematic of the dissociation and cryopreservation workflow. Fresh endometrial tissue was subjected to either enzymatic digestion with TrypLE/deoxyribonuclease I for 20 min 20-Minute TrypLE digestion protocol (T-20) or 60 min (60-Minute TrypLE Digestion Protocol, or mechanical dissociation using the gentle magnetic-activated cell sorting system. Resulting single-cell suspensions were cryopreserved and later used for surface proteomic analysis. B , Overlap of confident CSP identifications among the three dissociation protocols. The majority of proteins identified in the gentle magnetic-activated cell sorting system and 60-Minute TrypLE Digestion Protocol conditions overlapped with those captured using T-20. C , CSP counts, NXS/T motif-containing peptide counts, and total protein input for each titration sample using the T-20 method. Inputs of 0.5 × 10⁶, 1 × 10⁶, and 2 × 10⁶ cells were analyzed for three donors, with one donor profiled in triplicate. CSP yields remained robust even at the lowest input levels. D , UpSet plot showing the number of CSPs shared and uniquely identified across donor–input combinations. High overlap in CSP identifications was observed across donors and input levels. E , Principal component analysis of CSP intensities after experiment-wide median intensity normalization reveals clear donor-specific clustering, demonstrating reproducibility and biological distinction across samples. CSP, cell surface protein; T-20, 20-Minute TrypLE digestion protocol. After thawing, cell viability remained high for both enzymatic conditions, with ∼70% of input cells recovered, whereas the gentle MACS protocol yielded substantially lower post-thaw recovery (∼40%) ( Supplemental Fig. S5 B ). This reduced viability may reflect accumulated mechanical stress during dissociation, which can compromise membrane integrity or increase sensitivity to freeze–thaw damage. Surface proteomic analysis on the cryopreserved replicates yielded an average of approximately 400 confident surface protein identifications ( Supplemental Fig. S5 C , Supplemental Table S4 A ). CSP profiles were highly consistent across T-20 and T-60 replicates, indicating minimal impact of extended incubation time, and nearly all proteins identified in the gentle MACS condition overlapped with those captured by the enzymatic protocols ( Fig. 4 B ). Based on comparable proteomic depth, high post-thaw viability, and reduced incubation time, the T20 protocol was selected for subsequent use. To evaluate the performance of our optimized workflow under clinically relevant constraints, we selected the T-20 protocol for a titration series using cryopreserved cells from three independent endometrial donors. Cell inputs of 0.5 × 10 6 , 1 × 10 6 , and 2 × 10 6 were chosen to reflect the range of yields that might be obtained from dissociated endometriotic lesions. Each donor contributed a full titration series; for one donor, triplicate input sets were generated from independent pieces of tissue. Tissue input weights for dissociation ranged from 158 to 170 mg, yielding higher cell recoveries than in the optimization experiments, at approximately 60,000 to 90,000 cells per mg of tissue ( Supplemental Fig. S5 D ). Across all conditions, surface proteomic analysis yielded high CSP identification rates (typically >400 CSPs per sample), with only modest reductions observed at the lowest input level ( Supplemental Fig. S4 C , Supplemental Table S4 B ). Protein identifications demonstrated strong overlap between input amounts, indicating consistent enrichment performance even with limited material ( Supplemental Fig. S4 D ). Principal component analysis further showed clear donor-specific separation, emphasizing both the reproducibility of the method and its sensitivity to underlying biological variation ( Fig. 4 E ). These results establish the feasibility of robust surfaceome profiling from limited clinical material.

This work addresses two key gaps in the field: first, by providing a direct, head-to-head comparison of two mechanistically distinct, low-input-compatible surface enrichment strategies under standardized conditions; and second, by demonstrating that these optimized workflows can be successfully applied to cryopreserved cell lines and single-cell suspensions derived from solid tissues. Together, these advances broaden the applicability of surface proteomics to clinically relevant, low-input samples and support its use in translational research settings. Despite growing interest in surface proteomics for translational applications, there remains a lack of standardized frameworks for low-input or cryopreserved samples. By systematically evaluating two distinct enrichment strategies across input levels, preservation states, and dissociation protocols, this work provides practical guidance for implementing surfaceome profiling in resource-limited or clinically derived contexts. While various protein capture methods employing different chemistries have been explored individually ( 10 ), this study offers the first direct comparison of two mechanistically distinct surface enrichment approaches—from labeling strategy through downstream enrichment. Both strategies use commercially available reagents and are straightforward to implement, enabling high-quality surfaceome coverage from inputs below 1 × 10 6 cells—cell numbers that previously would have been considered insufficient for deep profiling. Additionally, all analyses were performed on a single Exploris 480 instrument, using manual implementations of the protocols. While others have explored automation to improve throughput, our results demonstrate that robust low-input surface profiling can be achieved with accessible instrumentation and hands-on workflows ( 9 , 17 ). We anticipate that this comparison will help researchers select appropriate methods based on sample characteristics, experimental goals, and analytical constraints. In our optimizations and comparative analyses, we leveraged SURFY as a uniform way to assess the degree of cell surface protein coverage and selectivity of the method for CSPs versus background proteins. While curated databases such as SURFY provide the ability to detect novel, previously unknown true cell surface localized proteins, including proteins with predicted transmembrane domains, signal peptides, or surface annotations from databases like UniProt—will likely yield broader coverage, and may improve the ability to detect novel CSPs. We selected the NGE approach for tissue-derived applications based on several key advantages. It demonstrated high confidence in low-input surface protein quantification, with consistent CSP recovery and strong correlation across input levels. Its reliance on glycan-directed labeling and selective release of N-linked glycopeptides resulted in lower background and reduced need for aggressive post-hoc filtering, enabling more complete analysis of all identified proteins. Additionally, the method showed enhanced dynamic range and successfully captured biologically meaningful changes in surface protein abundance, as seen in our EGF stimulation model. These features make the NGE workflow particularly well suited for complex or limited-input samples such as cryopreserved dissociated tissue. Finally, while we selected the 20-min TrypLE/DNase I dissociation protocol based on performance and handling efficiency, we acknowledge that optimal dissociation strategies may differ across tissue types and should be empirically evaluated based on cell type and structural composition. The merits of the WGA-HRP approach should not be overlooked. Protein-level capture may offer improved quantification within a sample due to broader sequence coverage, particularly in cases where differential glycosylation influences the yield and apparent abundance of target proteins, as we observed for certain cell adhesion molecules. Moreover, differences in instrumentation or refinements to labeling workflows may mitigate some of the limitations we encountered related to background and sensitivity. However, protein-level enrichment produces much larger protein lists that are dominated by secreted and intracellular background proteins, complicating identifcation of true cell surface proteins, which is a general challenge with protein capture-based strategies. A recent study employing protein-level capture of glycoproteins using labeling chemistries analogous to those used in our NGE strategy similarly reported background signal as a concern ( 6 ). Thus, for applications involving extremely limited material ( e.g., <5 × 10 5 cells), the NGE approach may offer greater reliability for quantitative surface profiling. Additionally, alternative lectin–HRP conjugates recognizing other glycan classes, such as ConA for mannose-rich structures, could provide complementary or improved recovery of surface proteins. Future studies systematically comparing multiple lectin-based enrichment reagents will help define optimal lectin selection for different biological systems. Looking ahead, this work expands the applicability of surface proteomics to clinically relevant materials that have traditionally been inaccessible to enrichment-based workflows. In particular, it enables investigation of the surface proteome in contexts such as therapeutic response, receptor trafficking, and immune evasion—where cell surface remodeling can have functional and clinical significance. Ongoing refinement and validation of flexible, low-input-compatible workflows will be critical for realizing the full translational potential of surface proteomics, and the approaches described here contribute toward that goal.

All experimental design details are provided in the subsequent sections and the number of replicates is indicated in the result section and figure legends when appropriate, along with p -values and statistical tests. HEK293T and A549 cells were obtained from internal Broad Institute repositories, and KMS-12-BM cells were provided by the Ghobrial laboratory at Dana-Farber Cancer Institute. KMS-12-BM cells were cultured in RPMI-1640 medium (Gibco) supplemented with 10% fetal bovine serum (FBS), while HEK293T and A549 cells were maintained in DMEM (Gibco) with 10% FBS. All cells were grown at 37 °C in a humidified incubator with 5% CO 2 . Cryopreserved single-cell suspensions from dissociated endometrial tissue were provided by the O’Neill Laboratory and collaborators at the University of Pennsylvania School of Medicine. Tissue dissociation and cryopreservation were performed on-site prior to shipment. Patients were consented for use. All human studies reported herein abide by the Declaration of Helsinki principles. Participant consenting was performed by the primary investigator or members of the study team and the informed consent form IRB-approved. A verbal explanation was provided in terms suited to the participant’s comprehension of the purposes, procedures, and potential risks of the study and of their rights as research participants, and ample time given to review the written consent form and ask questions prior to signing. Participants were informed that participation was voluntary and could be withdrawn at any time. Immediately upon collection, all biospecimens were assigned a unique, arbitrary alphanumeric identifier that is used for anonymization. The alphanumeric identifier is linked to PHI in a manner that the PHI cannot be ascertained by the researchers and adheres to HIPAA regulations. The biospecimens and data are managed through a secure access-controlled LIMS system (LabVantage). Clinical patient-level information is stored in a separate access-controlled database (RedCap), which contains no PHI and no patient identifiers. Only authorized personnel on this IRB with a "need-to-know" have access to the data, and the data is backed up on a secure server. The IRB protocol number is 828613 For stimulation experiments, A549 cells were treated with 100 ng/ml epidermal growth factor (EGF; PeproTech) in DMEM containing 2% FBS. For chloroquine pretreatment conditions, cells were first incubated for 2 h in DMEM with 2% FBS containing 10 μM chloroquine (Selleck Chemicals), after which EGF was added directly to the same medium and cells were stimulated for an additional 3 h. For cryopreservation, cells were resuspended in 10% DMSO in FBS, slowly cooled in an ethanol-based freezing container at −80 °C overnight, and transferred to liquid nitrogen the following day. After an additional overnight storage in liquid nitrogen, cells were thawed and processed. For suspension cells (KMS-12-BM), fresh and freshly thawed cells were labeled and processed in parallel. For adherent A549 cells, fresh cells were labeled immediately, while the dissociated cells were cryopreserved using the same protocol. Thawed A549 cells were then labeled and both fresh and frozen samples were processed together through all downstream steps. Endometrial biopsies were obtained from the study participants using Pipelle Endometrial Biopsy Curette (3.1 mm Tip, Cooper Surgical). Biopsies were obtained in the luteal phase of the menstrual cycle in ovulatory women using the date of the first day of their last menstrual period to establish the cycle day of biopsy. Tissue was placed in ice-cold calcium and magnesium free Dulbecco's (1X PBS), rinsed twice to remove excess mucus and blood and placed on ice for transport back to the laboratory. Tissue was weighed and placed into a Petri dish and tissue was finely minced with two razor blades for 30 s. Using a large-bore transfer pipette, tissue was collected and placed in ice-cold digestion media comprised of TrypLe Express (Gibco) with deoxyribonuclease I (DNase I) (Roche, #10104159001, 30ug/ml stock concentration, 30 ng/ml working concentration). For the 20-Minute TrypLE Digestion Protocol (T-20) protocol, tissue was incubated in digestion media for 20 min at 37 °C in a shaking incubator (75 rpm), and a wide-bore transfer pipet was used to mix the tissue every 5 to 7 min during the incubation. For the 60-Minute TrypLE Digestion Protocol (T-60) protocol, tissue was incubated in digestion media for 60 min at 37 °C in a rotating incubator (40 rpm), and a wide-bore transfer pipet was used to mix the tissue every 15 min during the incubation. Cell suspensions were then applied to a 70-μm cell strainer (Corning, #431750) over a 50 ml conical tube and the suspension was pushed through the filter using the plunger end of a syringe. Digestion media was diluted with an ice-cold magnetic-activated cell sorting (MACS) buffer (PBS (without calcium, without magnesium), 2 mM EDTA, 2% fetal bovine serum) by pouring MACS buffer through the filter. The sample was pelleted at 300 g for 5 min at 4 °C and the supernatant was discarded. The cell pellet was resuspended in 3 to 5 ml of ACK Lysing buffer (Gibco, #A1049201) and incubated at room temperature for 5 min. Ice-cold 1X PBS was added to the cell suspension to stop the lysis reaction, and the suspension was spun at 300 g for 5 min at 4 °C. The supernatant was discarded and the cell pellet was resuspended in 1X PBS +0.04% BSA (Sigma, #B2518-100MG). Cell viability was determined by Trypan Blue staining and hand counting with a hemocytometer. Cells were cryopreserved in 90% FBS with 10% DMSO and stored in liquid nitrogen until ready for analysis. Tissue was dissociated according to protocol https://doi.org/10.17504/protocols.io.bvy8n7zw . The final cell pellet was resuspended in 1X PBS +0.04% BSA (Sigma, #B2518-100MG). Cell viability was determined by Trypan Blue staining and hand counting with a hemocytometer. Cells were cryopreserved in 90% FBS with 10% DMSO and stored in liquid nitrogen until ready for analysis. For tube-based labeling, cells were transferred to either 0.5 ml or 1.5 ml Protein LoBind tubes (Eppendorf) and washed twice with warm PBS. For adherent cell lines, PBS containing calcium and magnesium was used. All subsequent steps were performed on ice unless otherwise noted. Cells were incubated with 1.6 mM sodium metaperiodate (Thermo Fisher Scientific) in PBS (pH 6.5) for 15 min at 4 °C, followed by a single wash with 1% FBS in PBS (pH 6.5). Surface aldehydes were then labeled with 500 μM alkoxyamine–PEG4–biotin (Thermo Fisher Scientific) supplemented with 5 mM 5-methoxyanthranilic acid (Thermo Fisher Scientific) in 1% FBS/PBS (pH 6.5) for 30 min at 4 °C. Cells were washed once more with 1% FBS/PBS (pH 6.5) and twice with PBS (pH 6.5). For plate-based labeling (6-well plates and 10 cm dishes), the same protocol was applied, with reagent volumes scaled to fully cover the adherent monolayer and all steps performed using manual aspiration. After the final PBS wash, cells were gently scraped into tubes and processed identically to suspension samples. All centrifugation steps were performed at 650 rpm for 2.5 min, and resulting pellets were frozen at −80 °C until lysis. Samples were processed similar to previously published methods ( 9 ), with minor modifications. Cell pellets were lysed in 8 M urea, 50 mM ammonium bicarbonate (ABC), 2% SDS, and 1 mM magnesium chloride (MgCl 2 ), and mixed at 1000 rpm for 10 min at room temperature using an Eppendorf ThermoMixer. Lysates were treated with 1 μl Benzonase (Millipore Sigma) and mixed again at 1000 rpm for 20 min. Following nuclease treatment, samples were clarified by centrifugation at 15,000 g for 12 min. Protein concentrations were determined using the Pierce bicinchoninic acid (BCA) assay (Thermo Fisher Scientific). Disulfide bonds were reduced with 5 mM tris(2-carboxyethyl)phosphine (tris(2-carboxyethyl)phosphine, Thermo Fisher Scientific) and alkylated with 10 mM iodoacetamide (Thermo Fisher Scientific) for 30 min at room temperature. Samples were acidified with 25% phosphoric acid to a final concentration of 2.5% and processed using the S-Trap (Protifi) protocol. Briefly, six volumes of 50 mM triethylammonium bicarbonate (TEAB, pH 7.2) in methanol were added relative to the total lysate volume, including reducing, alkylating, and acidifying reagents, and prior to S-Trap loading. Samples were washed three times with 50 mM TEAB (pH 7.2) in methanol and digested overnight at 37 °C using sequencing-grade trypsin and Lys-C (Promega) at a 1:50 enzyme-to-protein ratio in 50 mM TEAB. Peptides were eluted from the S-Trap columns using sequential washes of 50 mM TEAB, 0.2% formic acid (FA), and 50% acetonitrile (ACN) with 0.2% FA. Eluted peptides were quantified using the Pierce Fluorometric Peptide Assay (Thermo Fisher Scientific). For experiments involving paired global proteome analysis, a specified amount of peptide was removed prior to enrichment and desalted separately. The remaining peptides were dried using a vacuum concentrator and resuspended in 100 μl PBS in 0.5 ml Protein LoBind tubes for surface glycopeptide capture. For glycopeptide capture, streptavidin magnetic beads (Genscript) were added at a ratio of 10 μl per 100 μg of peptide (minimum 5 μl, maximum 50 μl), and samples were incubated for 3 h at room temperature with mixing at 1100 rpm. Beads were washed sequentially for 2 min each with: 1 × PBS, 2 × 5 M NaCl, 2 × StimLys buffer (100 mM NaCl, 100 mM glycerol, 50 mM Tris, 1% Triton X-100), 2 × 50% ACN, 2 × 100 mM sodium carbonate (Na 2 CO 3 , pH 11), and 3 × 50 mM ABC). Beads were then resuspended in 50 μl of 50 mM ABC and transferred to 1.5 ml or 0.5 ml Protein LoBind tubes. Rapid peptide-N4-(N-acetyl-β-D-glucosaminyl) asparagine amidase F (PNGase F) (New England Biolabs), diluted 1:10 in 50 mM ABC, was added at 2 μl per sample, and deglycosylation was carried out overnight at 37 °C with mixing at 1350 rpm for 1.5 ml tubes or 1100 rpm for 0.5 ml tubes. The next day, the 50 μl eluate was transferred to styrenedivinylbenzene-reverse phase sulfonate (SDB-RPS) stage tips (Empore) for desalting. Beads were then washed with 50 μl of 50 mM ABC in 20% ACN, and the wash was also transferred to the stage tip, bringing the total load volume to 100 μl. Samples were acidified with 15 μl of 10% trifluoroacetic acid (TFA), bringing the final volume to 115 μl, and centrifuged to bind peptides to the resin. Peptides were washed with 125 μl of 1% TFA in isopropanol (IPA), followed by 125 μl of 0.2% TFA. Peptides were eluted with 80% ACN and 5% ammonium hydroxide (NH 4 OH), collected into LC-MS vials preloaded with 4.5 μl of 0.2% n-dodecyl-β-D-maltoside (DDM), dried, and reconstituted in 9 μl of 3% ACN/1% FA. A 4 μl injection was used for LC-MS analysis. Peptide aliquots taken for global proteome analysis were adjusted to 100 μl with 1% FA, and an additional volume of 10% FA was added to bring the final FA concentration to 1% prior to desalting. styrenedivinylbenzene-reverse phase sulfonatestage tips (Empore) were conditioned with 100% ACN, and samples were loaded directly onto the tips. Bound peptides were washed twice with 100 μl of 1% FA in IPA, followed by two washes with 100 μl of 1% FA. Peptides were eluted with 80 μl of 80% ACN containing 5% NH 4 OH, dried by vacuum centrifugation, and reconstituted to a final concentration of 1 μg/μl in 3% ACN with 1% FA. A 1 μl injection (1 μg total) was used for LC-MS/MS analysis. For tube-based labeling, cells were transferred to either 0.5 ml or 1.5 ml Protein LoBind tubes (Eppendorf) and washed three times with 300 μl of PBS (pH 6.5) at room temperature. The following volumes correspond to the 0.5 ml format; volumes were scaled proportionally for 1.5 ml tubes and plate-based formats. Cells were incubated with 200 μl of 0.5 μM WGA-HRP (Vector Laboratories) in PBS (pH 6.5) supplemented with 1 mM calcium chloride (CaCl 2 ) for 5 min on ice. Biotinyl tyramide (Sigma-Aldrich) was added to a final concentration of 0.5 mM by spiking in 2 μl of 50 mM stock in DMSO, followed by thorough mixing. Hydrogen peroxide (30%, Thermo Fisher Scientific) was added to a final concentration of 0.003% (2 μl of 0.3%), and samples were incubated at 37 °C for 2 min to initiate labeling. Labeling was quenched by adding 200 uL of cold quench buffer containing 2 mM sodium pyruvate (Thermo Fisher Scientific), 20 mM sodium ascorbate (Thermo Fisher Scientific), and 10 mM Trolox (Thomas Scientific) in PBS (pH 6.5), followed by two washes with 1X cold quench buffer (1 mM sodium pyruvate, 10 mM sodium ascorbate, and 5 mM Trolox in PBS (pH 6.5)), and one wash with ice-cold PBS (pH 6.5). For plate-based labeling (6-well plates and 10 cm dishes), the same protocol was applied, with reagent volumes scaled to fully cover the adherent monolayer and all steps performed using manual aspiration. After the final PBS wash, cells were gently scraped into tubes and processed identically to suspension samples. All centrifugation steps were performed at 650 g for 2.5 min, and resulting pellets were frozen at −80 °C until lysis. Cells were lysed in 1× radioimmunoprecipitation assay buffer (Thermo Fisher Scientific) supplemented with 1% Triton X-100 and 1× Halt Protease Inhibitor Cocktail (100×, Thermo Fisher Scientific). Samples were incubated at 4 °C for 30 min with gentle mixing at 500 rpm on a thermomixer. Lysates were clarified by centrifugation at 15,000 g for 12 min at 4 °C, and protein concentration was determined using the Pierce BCA Protein Assay. For enrichment, 100 μl of High Capacity NeutrAvidin resin (Thermo Fisher Scientific) was added to each sample and incubated at 4 °C for 1 h with mixing at 1300 rpm. Samples were transferred to Bio-Rad Bio-Spin columns (Bio-Rad) and washed sequentially with 3 ml each of radioimmunoprecipitation assay buffer, 1 M NaCl in PBS, 50% ACN, 3M Urea in 50 mM TEAB, 6M Urea in 50 mM TEAB, and 50 mM TEAB. After the final wash, beads were transferred to 1.5 ml Protein LoBind tubes with 200 μl of 50 mM TEAB. Disulfide bonds were reduced with 5 mM tris(2-carboxyethyl)phosphine and alkylated with 10 mM Iodoacetamide for 30 min at room temperature. Reagents were removed, and resin was resuspended in 100 μl of 50 mM TEAB containing 1 μg each of trypsin and Lys-C (Promega). Digestion was carried out overnight at room temperature with mixing at 1300 rpm. The following day, resin was separated using Pierce Snap Cap Spin Columns (Thermo Fisher Scientific), and the peptide-containing flow-through was collected. Peptides were quantified using the Pierce Fluorometric Peptide Assay and acidified to 1% FA using 10% FA. Peptides were desalted using in-house C18 stage tips prepared with two stacked punches of Empore C18 material (3M). Tips were conditioned sequentially with 100% methanol, 50% ACN containing 0.1% FA, and twice with 0.1% FA. Acidified peptide samples were loaded directly onto the stage tips, followed by two washes with 0.1% FA. Peptides were eluted with 80 μl of 50% ACN containing 0.1% FA and dried by vacuum centrifugation. Final resuspension volumes were adjusted based on peptide yields from the Pierce Fluorometric Peptide Assay to achieve a target concentration of 100 ng/μl in 3% ACN with 1% FA. A total of 2 μl (200 ng) was injected for LC-MS/MS analysis. For HEK293T titration experiments, enrichment was performed on defined cell numbers without protein normalization. For both KMS-12 and A549 samples, input was guided by cell counts or equivalent seeding conditions, and protein amounts were normalized based on BCA measurements to account for minor differences in recovery. This normalization approach was applied consistently across both NGE and WGA-HRP workflows. LC-MS/MS analysis was conducted using a Vanquish Neo UHPLC system (Thermo Fisher Scientific) coupled to an Orbitrap Exploris 480 mass spectrometer (Thermo Fisher Scientific) operating in data-independent acquisition (DIA) mode. Peptides were separated on a 25 cm New Objective PicoFrit column (75 μm inner diameter, 10 μm tip ID) packed in-house with 1.5 μm C18 particles (ReproSil-Pur C18-AQ, Dr Maisch) at a flow rate of 200 nl/min. Surface-enriched samples from both the N-glyco and WGA-HRP workflows were analyzed using an 85-min active gradient: 1% to 4% buffer B in 1 min, 4% to 5% in 1 min, 5% to 24% over 58 min, 24% to 28% over 7.5 min, 28% to 36% over 7.5 min, and 36% to 72% over the final 10 min. MS1 scans were acquired from m/z 350 to 1800 at a resolution of 120,000 with an AGC target of 3E6 and a maximum injection time of 20 ms. MS2 spectra were acquired across m/z 350 to 1450 using 20 variable windows, at a resolution of 30,000, with an AGC target of 1E6, maximum injection time of 54 ms, and normalized collision energy of 27% with a default charge state of +3 ( Supplemental Table S1 G ). For global proteome analysis, a 94-min active gradient was used: 1.6% to 4.8% buffer B in 1 min, 4.8% to 24% over 84 min, and 24% to 48% over the final 9 min. MS1 spectra were acquired over m/z 350 to 1800 with a resolution of 120,000 at 200 m/z, an AGC target of 3E6, and a maximum injection time of 20 ms. DIA was performed using 30 variable windows covering m/z 400 to 850. MS2 scans were acquired with an AGC target of 5E5, maximum injection time of 54 ms, and NCE set to 27% with a default charge state of +3 ( Supplemental Table S1 H ). All DIA data were analyzed using Spectronaut (version 19, Biognosys AG) in directDIA mode. Spectra were searched against the UniProt Swiss-Prot human reference proteome (20,421 entries, downloaded June 2025) using Trypsin/P specificity, allowing up to 2 missed cleavages. Carbamidomethylation of cysteine was set as a fixed modification. Oxidation of methionine and protein N-terminal acetylation were included as variable modifications for all datasets. For N-glycoprotein-enriched samples, asparagine deamidation was added as a variable modification to detect PNGase F-cleaved glycosylation sites, and site localization filtering was applied with a minimum probability cutoff of 0.75. Spectronaut’s modification filter was then used to retain only peptides containing deamidation, and these filtered peptides were used for protein quantification. “Exclude De-amidated Peptides” was also turned off in the calibration DIA analysis settings. Only proteins containing at least one peptide mapping to an N-X-S/T consensus motif, where X was any amino acid except Proline, were retained for downstream analysis. Peptides ending at an N-R/K motif were mapped back to the FASTA database used in the search, and the corresponding protein sequences were examined to determine whether the residue following the cleavage site was an S or T—consistent with the N-linked glycosylation consensus motif (N-X-S/T). Multiple instances of each peptide were reviewed to confirm correct mapping and potential motif identification. Protein-level quantification was performed using Spectronaut’s default protein inference and abundance estimation algorithms. For WGA-HRP-enriched and global proteome datasets, a standard search workflow was used without the deamidation modification. Protein-level filtering was applied to remove experiment-wide single-peptide identifications. For all Spectronaut analyses, protein (experiment-level) and precursor identifications were filtered at 1% false discovery rate and protein (run-level) identifications were filtered at 5% false discovery rate using Spectronaut’s q-value score. Additionally, precursor and fragment ion tolerance parameters were set to “Dynamic” with a correction factor of 1. For cell surface protein (CSP) ID figures (HEK titration, KMS12 and A549 high/low input, endometrial titration, and endometrial method evaluation), the method evaluation option in Spectronaut was enabled to eliminate match-between-runs (MBR) effects on identification rates and obtain more accurate identifications reflective of each input amount or dissociation technique. All figures referencing quantitative changes or comparisons were generated without method evaluation enabled, per Spectronaut’s recommendations for accurate quantification. Protein intensity values were log 2 -transformed and normalized using sample-wise median centering in R (Protigy pipeline), to account for global differences in sample loading and MS signal. For the HEK titration and KMS12 fresh/frozen experiments, sample-wise median normalization was performed within each group, not experiment-wide. For comparative analyses, fold changes between conditions were calculated based on log 2 -transformed protein intensities. For differential abundance analysis, two-sample t -tests were performed, and results were visualized as volcano plots displaying log 2 fold-change versus –log 10 p -value (nominal p -values reported). All statistical analyses were performed in R or Python, and figures were generated using ggplot2 or matplotlib as appropriate.

S. A. C. is a member of the scientific advisory boards of Kymera, PTM BioLabs, MOBILion, PrognomIQ. M. A. G. is on the scientific advisory board of PrognomIQ. The other authors declare no competing interests.

The original mass spectra, protein sequence database and associated parameter files have been deposited in the public proteomics repository MassIVE ( http://massive.ucsd.edu ) and are accessible at MSV000098694.

This article contains supplemental data .