Integrated LiP-MS and quantitative proteomics reveal coordinated alterations in protein conformation and expression across tumor and peritumoral regions in hepatocellular carcinoma Authors: Zhen-dong Chen, Min-ying Yang, Xi-ni Wei, Wei Qin, Li Xiao, Yin-yun Qin, Rui-ling Su, Yi-wu Dang, Dan-dan Xiong Published: 2026 ABSTRACT Hepatocellular carcinoma (HCC) exhibits substantial molecular heterogeneity, yet protein-level alterations beyond abundance remain insufficiently characterized. Here, we integrated limited proteolysis mass spectrometry (Lip-MS) with 4D label-free quantitative proteomics to investigate conformational accessibility and protein abundance across tumor, peritumoral-near, and peritumoral-far tissues from HCC patients. Differential LiP peptides identified by both DDA and DIA corresponded to 725, 674, and 33 differentially conformed proteins in the Tumor vs. Peritumor-far, Tumor vs. Peritumor-near, and Peritumor-near vs. Peritumor-far comparisons, respectively. Quantitative proteomics identified 405, 365, and 4 differentially expressed proteins in the corresponding comparisons. Integrated analysis identified 488 and 469 conformation-specific altered proteins (CSAPs), which showed altered conformational accessibility without significant abundance changes, and 237 and 205 conformation-expression coupled proteins (CECPs) in the two tumor-involved comparisons. LiP peptide and protein abundance changes were positively correlated, with Spearman coefficients of 0.69–0.72, and more than 99% of CECPs showed concordant directions. Among them, 169 region-conserved CECPs (rcCECPs) were predominantly associated with metabolic and redox-related pathways. Protein-protein interaction analysis identified 30 hub rcCECPs. ACLY, ALDH18A1, GMPS, and DHX9 showed increased representative LiP peptide signals and protein abundance, elevated transcript expression in HCC, and associations with poorer overall survival. Peptide mapping further localized their differential LiP signals to specific sequence regions and annotated domains. Collectively, these findings provide an integrated view of regional conformational accessibility and protein abundance alterations in HCC and identify candidate proteins for further structural and functional investigation.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-62723-9. Introduction Hepatocellular carcinoma (HCC) is a leading cause of cancer-related death worldwide, and its disease burden continues to increase1,2. Its marked molecular heterogeneity, complex tumor biology, and limited options for early detection and effective treatment remain major challenges in clinical management. Although advances in immunotherapy and multi-targeted therapies have broadened treatment options, the overall prognosis of patients with HCC remains unsatisfactory3. A deeper understanding of the molecular alterations associated with hepatocarcinogenesis is therefore needed to facilitate the identification of potential biomarkers and therapeutic targets. Proteomics has become an important strategy for characterizing tumor biology by quantifying protein abundance and reconstructing tumor-associated regulatory networks4,5. However, conventional proteomic workflows generally rely on complete enzymatic digestion and primarily provide abundance-based information, with limited assessment of protein structural states6. Protein conformation is tightly linked to enzymatic catalysis, substrate recognition, and protein–protein interaction affinity, and its dysregulation contributes to cellular stress adaptation, signal transduction, and metabolic reprogramming in tumors7,8. Limited proteolysis mass spectrometry (LiP-MS) addresses this gap by detecting proteome-wide changes in peptide-level conformational accessibility through structure-sensitive proteolytic patterns, enabling the identification of proteins with altered folding, domain accessibility, or interaction states under pathological conditions9–11. Although LiP-MS has been applied to characterize protein structural states in various biological contexts, its integration with quantitative proteomic platforms such as 4D label-free quantitative proteomics remains largely unexplored. Leveraging the complementary strengths of these platforms enables the parallel interrogation of protein conformational accessibility and expression changes in complex tissues, thereby uncovering protein alterations that cannot be captured through abundance-based proteomics alone. In this study, we applied LiP-MS together with 4D label-free quantitative proteomics to paired tumor, peritumor-near, and peritumor-far tissues from HCC patients. By integrating peptide-level conformational accessibility with protein abundance data, we aimed to characterize regional protein alterations and identify candidate molecular features associated with hepatocarcinogenesis. Materials and methods Sample collection Three patients with primary HCC who underwent curative liver resection at the First Affiliated Hospital of Guangxi Medical University were included. All patients were histologically diagnosed with HCC and had not received any systemic anticancer therapy before surgery. The study was approved by the Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (No. 2025-S0002), and written informed consent was obtained from all participants prior to sample collection. All methods were performed in accordance with the relevant guidelines and regulations. During resection, three types of tissue samples were obtained from each patient: HCC tumor tissue (Tumor), near peritumoral tissue (Peritumor-Near; < 1 cm from the tumor margin), and far peritumoral tissue (Peritumor-Far; > 3 cm from the tumor margin). All samples were immediately snap-frozen in liquid nitrogen and stored at -80 °C for subsequent proteomic sequencing analysis. All tissue samples were collected and processed using the same procedure to reduce pre-analytical variation. Given that LiP-MS was performed on frozen tissue lysates, the resulting LiP-MS signals were interpreted as relative ex vivo proteinase K accessibility patterns under controlled processing conditions, rather than as direct measurements of protein conformations in living tissues. Lip-MS sample preparation Tissue samples were ground into fine powder in liquid nitrogen using a pre-cooled mortar. No repeated freeze-thaw cycles were applied during tissue lysis. The tissue powder was lysed in ice-cold lysis buffer containing 100 mM HEPES, 150 mM KCl, and 1 mM MgCl2, pH 7.5. After vortexing and mixing, the lysates were centrifuged at 20,000 × g for 10 min at 4 ℃, and the supernatants were collected as clarified protein lysates. Protein concentrations were determined using the BCA assay. For limited proteolysis, 100 µg of protein from each sample was incubated at room temperature for 10 min. Proteinase K was then added at an enzyme-to-substrate mass ratio of 1:100, and the reaction was carried out at room temperature for 5 min. The reaction was immediately terminated by heating at 98 °C for 3 min. Then sodium dodecyl sulfate (SDS, final concentration 4%) was added and heated again at 98 °C for 3 min. The samples were reduced with tris(2-carboxyethyl)phosphine (TCEP) at a final concentration of 40 mM and alkylated with iodoacetamide (IAA) at a final concentration of 100 mM. The samples were incubated in a boiling water bath for 5 min. Hydroxypropyl-β-cyclodextrin solution (250 mM) was then added to reduce SDS interference during digestion, followed by the addition of 100 µL of 100 mM ammonium bicarbonate solution. Trypsin was added at an enzyme-to-substrate mass ratio of 1:50, and digestion was performed overnight at 37 °C. After digestion, the reaction was terminated by adding trifluoroacetic acid to reduce the pH below 2. The resulting peptide mixtures were desalted using C18 cartridges (Waters, WAT023590) before nanoLC-MS/MS analysis. LC-MS/MS acquisition of LiP-MS samples LiP-MS peptide samples were separated using a nano-flow Easy-nLC system and subsequently analyzed on a Q Exactive HF-X mass spectrometer. Mobile phase A was 0.1% formic acid in water, and mobile phase B was 0.1% formic acid in 80% acetonitrile. Before sample loading, the C18 reversed-phase analytical column was equilibrated with 100% mobile phase A. Peptide mixtures were loaded onto the C18 analytical column (Thermo Fisher Scientific, Acclaim PepMap RSLC 50 μm × 15 cm, nanoViper, P/N 164943) by the autosampler and separated at a flow rate of 300 nL/min. A 120-min gradient was used as follows: 1%–4% buffer B from 0 to 5 min, 4%–28% buffer B from 5 to 101 min, 28%–38% buffer B from 101 to 109 min, 38%–100% buffer B from 109 to 114 min, and 100% buffer B from 114 to 120 min. For data-dependent acquisition (DDA), the mass spectrometer was operated in positive ion mode. MS data were acquired using a data-dependent top-10 method, in which the 10 most abundant precursor ions from each full MS scan were selected for higher-energy collisional dissociation (HCD) fragmentation. Full MS scans were acquired over an m/z range of 350–1800 at a resolution of 60,000, with an automatic gain control (AGC) target of 1 × 10⁶ and a maximum injection time of 50 ms. MS2 scans were acquired at a resolution of 15,000, with an AGC target of 2 × 10⁵, a minimum AGC target of 2 × 10³, a maximum injection time of 60 ms, an isolation window of 2 m/z, and a normalized collision energy of 28 eV. Ions with charge states between 2 and 5 were selected for fragmentation, and dynamic exclusion was set to 30 s. For data-independent acquisition (DIA), the MS scan range was set to 350–1500 m/z at a resolution of 60,000, with an AGC target of 3 × 10⁶ and a maximum injection time of 50 ms. The DIA acquisition was performed at a resolution of 15,000, with an AGC target of 2 × 10⁵, a maximum injection time of 45 ms, and a normalized collision energy of 28%. Variable isolation windows covering the m/z range of 350–1500 were used. A total of 42 DIA windows were applied, from 350 to 405 m/z for the first window to 1189–1500 m/z for the final window. The detailed DIA window settings are provided in Supplementary Table 1. DDA and DIA were used as complementary acquisition modes for LiP-MS peptide analysis. DDA acquisition provided high-quality MS2 spectra and was used to support peptide identification, whereas DIA acquisition was used to improve the coverage and reproducibility of LiP peptide detection across samples. The DDA and DIA datasets were processed separately and were not treated as technical replicates. In the downstream analysis, differential LiP peptides showing consistent changes in both acquisition modes were prioritized as a conservative high-confidence peptide set. 4D label-free quantitative proteomics Tissues samples were lysed in SDT buffer containing 4% SDS and 100 mM Tris-HCl, pH 7.6, with quartz sand and homogenized using an MP Fastprep-24 Automated Homogenizer (6.0 m/s, 30 s, twice), followed by ultrasonication, boiling for 10 min, and centrifugation at 14,000 × g for 15 min to obtain clarified supernatants. The supernatants were filtered through 0.22 μm filters. Protein concentrations were measured using the BCA assay, and 20 µg of protein from each sample was examined by 12% SDS-PAGE to assess protein integrity. For quantitative proteomics, 100 µg of protein from each sample was processed using the FASP protocol12. Briefly, samples were reduced with 100 mM dithiothreitol at 100 °C for 5 min, alkylated with 100 mM IAA in UA buffer in the dark for 30 min, and subjected to repeated centrifugation through 30-kDa ultrafiltration units to remove detergents and low-molecular-weight contaminants. The UA buffer contained 8 M urea and 150 mM Tris-HCl, pH 8.5. After buffer exchange with 50 mM NH₄HCO₃, proteins were digested with 40 µL trypsin buffer (4 µg trypsin in 40 µL 50 mM NH₄HCO₃) at 37 °C for 16–18 h. The resulting peptides were collected by centrifugation, desalted using C18 cartridges, lyophilized, reconstituted in 40 µL of 0.1% formic acid, and quantified by absorbance at 280 nm. Peptides were analyzed on a nanoElute coupled to a timsTOF Pro mass spectrometer (Bruker, Bremen, Germany) equipped with a CaptiveSpray source. Mobile phase A was 0.1% formic acid in water, and mobile phase B was 0.1% formic acid in acetonitrile. Separation was performed on an IonOpticks analytical column, 25 cm × 75 μm, packed with 1.6 μm C18 beads, at a flow rate of 300 nL/min. The column was maintained at 50 °C and equilibrated with four column volumes of 100% mobile phase A at 800 bar before sample loading. Data acquisition was conducted in DDA-PASEF mode on the timsTOF Pro mass spectrometer. The analysis time was 90 min, and the mass spectrometer was operated in positive ion mode. The mass range was 100–1700 m/z, and the ion mobility range was 0.75–1.4 V·s/cm². The ramp time was 100 ms, and the duty cycle was locked to 100%. The capillary voltage was 1500 V, the dry gas flow rate was 3 L/min, and the dry temperature was 180 °C. The PASEF settings were as follows: 10 MS/MS scans per cycle, total cycle time of 1.16 s, charge range of 0–5, active exclusion time of 0.5 min, scheduling target intensity of 10,000, intensity threshold of 2,500, and CID collision energy of 20–59 eV. Protein and peptide identification and quantification The raw MS data from the 4D label-free quantitative proteomics and LiP-MS DDA analyses were analyzed using MaxQuant software (v1.6.17.0; https://www.maxquant.org/) and searched against the Homo sapiens reference proteome from the UniProt database (http://www.uniprot.org, downloaded on 2024-05-15, containing 20434 canonical entries). For 4D label-free quantitative proteomics, Trypsin/P was specified as the cleavage enzyme, and up to two missed cleavages were allowed. The main search tolerance was set to 10 ppm, the first search tolerance was set to 20 ppm, and the MS/MS tolerance was set to 20 ppm. Carbamidomethylation of cysteine residues was specified as a fixed modification, whereas N-terminal protein acetylation and methionine oxidation were set as variable modifications for database searching. The target-reverse search strategy was used, and contaminants were included in the database search. The false discovery rate (FDR) was controlled at 0.01 at both peptide and protein levels. Protein abundance was quantified based on the normalized spectral protein intensity (LFQ intensity). For LiP-MS data, DDA data were processed using MaxQuant software, version 1.6.17.0 (https://www.maxquant.org/). To identify semi-tryptic peptides generated by limited Proteinase K digestion, the database search was performed using a semi-specific Trypsin/P strategy, allowing one peptide terminus to be non-tryptic. The same Homo sapiens UniProt reference proteome was used for database searching. DIA data from the LiP-MS experiment were analyzed using Spectronaut software (version 18.5; Biognosys AG; https://biognosys.com/software/spectronaut/), which was used for DIA peptide identification and intensity extraction. The data were searched against the same Homo sapiens UniProt reference proteome, and peptide-level quantitative values were exported for downstream differential analysis. Peptide and protein identifications from Lip-MS DDA and DIA analyses are summarized in Supplementary Table 2. For LiP-MS quantification, semi-tryptic peptides generated during limited proteolysis were used as readouts of Proteinase K accessibility. Fully tryptic peptides generated after trypsin digestion were quantified in parallel to help distinguish accessibility-related peptide changes from changes potentially driven by protein abundance. Global protein abundance changes were independently assessed using 4D label-free quantitative proteomics. Differential LiP accessibility and protein expression analysis For LiP-MS, peptides were first filtered by requiring non-missing intensity values in at least half of the samples within at least one comparison group before statistical testing. Groupwise differential analysis was then performed in R statistical computing environment (v4.3.1; https://www.r-project.org/) using the limma (Linear Models for Microarray Data) package (v3.58.1; https://bioconductor.org/packages/release/bioc/html/limma.html), which estimates empirical Bayes-moderated t-statistics and the corresponding raw and adjusted P values. Two types of differential peptide events were evaluated: (i) intensity-based differences (significantly differential peptides), defined as peptides with |log2 fold change| ≥ 0.585, corresponding to a fold change ≥ 1.5 or ≤ 0.67, and P < 0.05 between groups; and (ii) presence/absence differences (qualitative differential peptide events), defined as peptides that had non-missing values in at least half of the samples in one group while being completely undetected in the comparison group. DDA and DIA datasets were processed separately and were not treated as technical replicates. Differential results from the two acquisition modes were compared, and peptides showing consistent differential patterns in both DDA and DIA were prioritized as high-confidence differential LiP peptides. This intersection strategy was used as a conservative filtering approach. Differential LiP peptides passing these filters were mapped to their parent proteins; proteins with one or more such peptides were designated as differentially conformed proteins (DCPs), although the LiP-MS readout reflects differential Proteinase K accessibility rather than direct structural determination. For 4D label-free quantitative proteomics, differential protein expression analysis was also performed using limma. Proteins with |log2 fold change| ≥ 1, corresponding to a fold change ≥ 2 or ≤ 0.5, and P < 0.05 were defined as differentially expressed proteins (DEPs). Protein quantification and differential analysis results are providing in Supplementary Table 3. Functional annotation and biological analysis Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway annotation were performed using Blast2GO Command Line (v1.4.4; https://cli.docs.blast2go.com/older-versions/user-manual-1.4.4/) and KOALA (KEGG Orthology And Links Annotation, v3.0; https://www.kegg.jp/blastkoala/), respectively13,14. For GO annotation, protein sequences were aligned using National Center for Biotechnology Information Basic Local Alignment Search Tool Plus (NCBI BLAST+, v2.3.0+; https://www.ncbi.nlm.nih.gov/books/NBK279690/), and the top 10 hits with E-value ≤ 1 × 10⁻³ were retained for subsequent mapping and annotation. InterProScan (v5.30-69.0; https://www.ebi.ac.uk/interpro/about/interproscan/) was subsequently used to search for conserved motifs and supplement the functional annotations. The GO database version used in this study was go_20190701.obo. For KEGG annotation, protein sequences were assigned to KEGG Orthology groups, and pathway information was obtained based on KO annotation. The KEGG database version used in this study was KO_INFO_END.txt (2024-06-03). GO and KEGG enrichment analyses were conducted using a consistent statistical framework: Fisher’s exact test was applied to compare the distribution of GO terms or KEGG pathways between the target protein set and the overall background, thereby assessing the significance of functional over-representation. Protein structural domain annotation and domain enrichment analysis were carried out using the InterPro database (https://www.ebi.ac.uk/interpro/)15, which provides protein family classification, domain prediction, and identification of conserved sequence features. Domain enrichment was performed using Fisher’s exact test with the same background protein set used for GO and KEGG enrichment analyses. Subcellular localization analysis was performed using WoLF PSORT protein subcellular localization prediction server (https://wolfpsort.hgc.jp/)16. The software converts protein sequences into numerical localization descriptors based on sorting signals, amino acid composition, and functional motifs. Since WoLF PSORT supports only animal, plant, or fungal models, the biological origin of each protein had to be specified prior to analysis. Protein sequences in FASTA format were uploaded, the appropriate species model was selected, and the predicted subcellular localization for each target protein was obtained. Protein-protein interaction (PPI) network was constructed using the STRING (Search Tool for the Retrieval of Interacting Genes/Proteins) database (https://string-db.org) with a confidence score > 0.7. The network was visualized in Cytoscape (v3.10; https://cytoscape.org/), and hub proteins were identified based on network topology. Gene expression and survival analyses based on the TCGA-LIHC cohort were performed using the Xiantao Academic online platform (https://www.xiantaozi.com/) and Gene Expression Profiling Interactive Analysis 2 (GEPIA2; http://gepia2.cancer-pku.cn/), respectively. For survival analysis, patients were divided into high- and low-expression groups according to the median transcript level of each gene. Kaplan-Meier survival curves were generated, and statistical significance was defined as P < 0.05. Experimental design and statistical rationale This study was based on a small but carefully controlled sampling strategy. Three patients with primary HCC were enrolled, and for each patient, tumor, near-peritumoral, and far-peritumoral tissues were collected, yielding a total of nine biological samples. All specimens were processed under uniform conditions and immediately preserved at -80 °C to minimize pre-analytical variability. Each tissue sample was independently subjected to both LiP-MS and 4D label-free proteomic workflows, ensuring that LiP accessibility and expression-level measurements originated from the same set of biological materials. Because every sample represented an independent biological unit, no technical replicates were generated at the LC-MS acquisition stage. Each biological sample was analyzed once per acquisition mode. This design reflects the exploratory nature of the study, in which matched biological sampling across tumor, near-peritumoral, and far-peritumoral regions was prioritized. For LiP-MS, tissue samples were processed by liquid nitrogen grinding rather than repeated freeze–thaw lysis. Therefore, no repeated freeze-thaw cycles were applied during tissue lysis. Given that LiP-MS was performed on frozen tissue lysates, the resulting LiP-MS signals were interpreted as relative ex vivo Proteinase K accessibility patterns under controlled processing conditions, rather than as direct measurements of protein conformations in living tissues. To improve statistical robustness, several filtering and quality-control steps were applied. A FDR of 0.01 was applied to peptide and protein identifications. For LiP-MS, differential peptide identification required non-missing measurements in at least half of the samples within at least one group, followed by statistical analysis using the limma framework. DDA and DIA datasets were processed separately and were not treated as technical replicates. Peptides consistently detected as differential in both acquisition modes were prioritized for downstream interpretation. This intersection strategy was used as a conservative filtering approach, although it may exclude peptides detected in only one acquisition mode. For 4D label-free proteomics, protein quantification relied on MaxQuant LFQ algorithm, which normalizes intensity values to reduce run-to-run variation and facilitate comparison across samples. Group comparisons were performed using the limma statistical framework, which estimates empirical Bayes-moderated variances and is suitable for studies with limited sample numbers. Fold-change thresholds and significance criteria were selected to provide balanced stringency while retaining biological signal. Given the small sample size, the findings of this study should be considered exploratory and require validation in larger independent cohorts. Results Region-dependent differences in protein conformational accessibility in HCC tissues Peptide-level LiP-MS profiles across tissue regions LiP-MS profiling revealed marked regional differences in peptide-level signals among Tumor, Peritumor-near, and Peritumor-far tissues. Intensity-based differential peptides and qualitative presence/absence peptide events were evaluated separately in the DDA and DIA datasets. Among peptides quantified in both groups, the DDA dataset identified 1,097 significantly increased and 1,053 significantly decreased peptides in the Tumor vs. Peritumor-far comparison. The Tumor vs. Peritumor-near comparison contained 935 increased and 891 decreased peptides, whereas 97 increased and 83 decreased peptides were identified between Peritumor-near and Peritumor-far tissues (Fig. 1A–C). In the DIA dataset, 2,865 increased and 3,525 decreased peptides were identified in the Tumor vs. Peritumor-far comparison, 2,795 increased and 3,101 decreased peptides in the Tumor vs. Peritumor-near comparison, and 236 increased and 176 decreased peptides in the Peritumor-near vs. Peritumor-far comparison (Fig. 1D–F). For peptides exhibiting group-specific detection, the DDA dataset identified 1,130 peptides detected only in Tumor tissues and 456 detected only in Peritumor-far tissues in the Tumor vs. Peritumor-far comparison. In the Tumor vs. Peritumor-near comparison, 1,058 peptides were detected only in Tumor tissues and 450 only in Peritumor-near tissues. Between the two peritumoral regions, 206 peptides were detected only in Peritumor-near tissues and 192 only in Peritumor-far tissues (Fig. 1G). In the DIA dataset, 2,640 peptides were detected only in Tumor tissues and 445 only in Peritumor-far tissues in the Tumor vs. Peritumor-far comparison. The corresponding numbers were 2,274 and 476 in the Tumor vs. Peritumor-near comparison, and 382 and 182 in the Peritumor-near vs. Peritumor-far comparison, respectively (Fig. 1H). The intensity-based differential peptides and qualitative presence/absence peptide events were then combined to generate the overall differential peptide set for each comparison. In the DDA dataset, 3,736 differential peptides were identified in Tumor vs. Peritumor-far, 3,334 in Tumor vs. Peritumor-near, and 578 in Peritumor-near vs. Peritumor-far. The corresponding totals in the DIA dataset were 9,475, 8,646, and 976 peptides, respectively. Intersection analysis of the combined differential peptide sets from the DDA and DIA datasets identified 2,002 overlapping peptides in Tumor vs. Peritumor-far, 1,728 in Tumor vs. Peritumor-near, and 44 in Peritumor-near vs. Peritumor-far (Fig. 1I–K). These overlapping peptide sets corresponded to 725, 674, and 33 DCPs, respectively. Detailed peptide-level results, including acquisition modes, event types, tissue comparisons, peptide sequences, parent proteins, peptide positions, sample-level intensity values, fold changes, statistical values, and directions of change, are provided in Supplementary Table 4. Fig. 1LiP-MS peptide-level overview. (A–F) Intensity-based volcano plots for the Tumor vs. Peritumor-far, Tumor vs. Peritumor-near, and Peritumor-near vs. Peritumor-far comparisons using the DDA (A–C) and DIA (D–F) acquisition modes. Differential peptides were defined using |log2 fold change| > 0.585 and P < 0.05. (G, H) Peptide presence/absence count plots for the Tumor vs. Peritumor-far, Tumor vs. Peritumor-near, and Peritumor-near vs. Peritumor-far comparisons using the DDA (G) and DIA (H) acquisition modes. (I–K) Venn diagrams showing the overlap between the combined differential peptide sets identified by DDA and DIA in the Tumor vs. Peritumor-far (I), Tumor vs. Peritumor-near (J), and Peritumor-near vs. Peritumor-far (K) comparisons. The combined differential peptide sets included both intensity-based differential peptides and qualitative presence/absence peptide events. The Venn diagrams are schematic and not drawn to scale; the numerical labels indicate the exact numbers of elements in each set and intersection. LiP-MS peptide-level overview. (A–F) Intensity-based volcano plots for the Tumor vs. Peritumor-far, Tumor vs. Peritumor-near, and Peritumor-near vs. Peritumor-far comparisons using the DDA (A–C) and DIA (D–F) acquisition modes. Differential peptides were defined using |log2 fold change| > 0.585 and P < 0.05. (G, H) Peptide presence/absence count plots for the Tumor vs. Peritumor-far, Tumor vs. Peritumor-near, and Peritumor-near vs. Peritumor-far comparisons using the DDA (G) and DIA (H) acquisition modes. (I–K) Venn diagrams showing the overlap between the combined differential peptide sets identified by DDA and DIA in the Tumor vs. Peritumor-far (I), Tumor vs. Peritumor-near (J), and Peritumor-near vs. Peritumor-far (K) comparisons. The combined differential peptide sets included both intensity-based differential peptides and qualitative presence/absence peptide events. The Venn diagrams are schematic and not drawn to scale; the numerical labels indicate the exact numbers of elements in each set and intersection. Functional characteristics of DCPs Functional enrichment analysis showed that DCPs from both tumor-involved comparisons were predominantly associated with metabolic processes. In the Tumor vs. Peritumor-far comparison, enriched pathways included fatty acid degradation, carbon metabolism, glycolysis/gluconeogenesis, amino-acid-related pathways, retinol metabolism, beta-alanine metabolism, pyruvate metabolism, and cytochrome P450-related drug/xenobiotic metabolism (Fig. 2A, B). A broadly overlapping metabolic pattern was observed in the Tumor vs. Peritumor-near comparison, with additional enrichment in spliceosome components and 2-oxocarboxylic acid metabolism (Fig. 2C, D). Although far fewer DCPs were identified between Peritumor-near and Peritumor-far tissues, these proteins were associated with several metabolic pathways, including beta-alanine, lysine, histidine, branched-chain amino acid, and alanine-aspartate-glutamate metabolism, together with pyruvate and fatty-acid pathways. Peroxisome-related and tight-junction pathways were also represented (Fig. 2E, F). Fig. 2Functional enrichment analyses of differentially conformed proteins (DCPs) in hepatocellular carcinoma tissues. GO and KEGG enrichment analyses of DCPs in the Tumor vs. Peritumor-far (A, B), Tumor vs. Peritumor-near (C, D), and Peritumor-near vs. Peritumor-far (E, F) comparisons. The GO enrichment plots show the top 5 significantly enriched terms in the biological process (BP), cellular component (CC), and molecular function (MF) categories. The KEGG enrichment plots show the top 15 pathways identified in each comparison. Functional enrichment analyses of differentially conformed proteins (DCPs) in hepatocellular carcinoma tissues. GO and KEGG enrichment analyses of DCPs in the Tumor vs. Peritumor-far (A, B), Tumor vs. Peritumor-near (C, D), and Peritumor-near vs. Peritumor-far (E, F) comparisons. The GO enrichment plots show the top 5 significantly enriched terms in the biological process (BP), cellular component (CC), and molecular function (MF) categories. The KEGG enrichment plots show the top 15 pathways identified in each comparison. Domain characteristics and subcellular localization of DCPs Domain enrichment analysis revealed that DCPs from both the Tumor vs. Peritumor-far and Tumor vs. Peritumor-near comparisons were mainly associated with NAD(P)-binding, alcohol/aldehyde dehydrogenase, and pyridoxal phosphate-dependent transferase domains, which are commonly found in metabolic and redox-related enzymes (Fig. 3A, B). In the Peritumor-near vs. Peritumor-far comparison, the enriched domains included S-adenosylmethionine synthetase, glutamine transaminase, and transglutaminase family domains (Fig. 3C). Subcellular localization analysis showed that DCPs across all three comparisons were predominantly assigned to the cytosol, followed by the nucleus, mitochondria, and extracellular compartment (Fig. 3D). The two tumor-related comparisons exhibited broadly similar localization patterns. In contrast, the Peritumor-near vs. Peritumor-far comparison showed a relatively higher mitochondrial proportion and a lower extracellular proportion; however, these proportional differences should be interpreted cautiously because only 33 DCPs were identified in this comparison. Fig. 3Domain enrichment and subcellular localization of differentially conformed proteins (DCPs) in hepatocellular carcinoma tissues. (A–C) Domain enrichment analyses of DCPs identified in the Tumor vs. Peritumor-far (A), Tumor vs. Peritumor-near (B), and Peritumor-near vs. Peritumor-far (C) comparisons. The top 15 most significantly enriched domains are presented for each comparison. (D) Subcellular localization profiles of DCPs identified in the Tumor vs. Peritumor-far, Tumor vs. Peritumor-near, and Peritumor-near vs. Peritumor-far comparisons. Domain enrichment and subcellular localization of differentially conformed proteins (DCPs) in hepatocellular carcinoma tissues. (A–C) Domain enrichment analyses of DCPs identified in the Tumor vs. Peritumor-far (A), Tumor vs. Peritumor-near (B), and Peritumor-near vs. Peritumor-far (C) comparisons. The top 15 most significantly enriched domains are presented for each comparison. (D) Subcellular localization profiles of DCPs identified in the Tumor vs. Peritumor-far, Tumor vs. Peritumor-near, and Peritumor-near vs. Peritumor-far comparisons. Region-dependent patterns of protein expression differences in HCC tissues Regional distribution and quantitative characteristics of DEPs Using 4D label-free quantitative proteomics, we compared protein abundance profiles among Tumor, Peritumor-near, and Peritumor-far tissues. A total of 405 DEPs were identified in the Tumor vs. Peritumor-far comparison (Fig. 4A), and 365 DEPs were identified in the Tumor vs. Peritumor-near comparison (Fig. 4B). In contrast, only four DEPs were detected between Peritumor-near and Peritumor-far tissues (Fig. 4C), indicating broadly similar global protein abundance profiles between the two peritumoral regions under the present analytical conditions. Hierarchical clustering analysis further illustrated these patterns. Tumor and non-tumor samples formed clearly separated clusters in both the Tumor vs. Peritumor-far and Tumor vs. Peritumor-near comparisons (Fig. 4D, E). The heat map of the four DEPs identified between Peritumor-near and Peritumor-far tissues also showed differences in their abundance patterns between the two regions (Fig. 4F). However, because this clustering was based on only four proteins, it should not be interpreted as evidence of broad proteome-wide separation. Taken together, the DEP numbers and clustering profiles indicate pronounced protein abundance differences between tumor and non-tumor tissues, whereas the two peritumoral regions exhibit only limited differences at the global protein abundance level. Fig. 4Differential protein expression across Tumor, Peritumor-near, and Peritumor-far regions. (A–C) Volcano plots showing differentially expressed proteins (DEPs) identified in the Tumor vs. Peritumor-far (A), Tumor vs. Peritumor-near (B), and Peritumor-near vs. Peritumor-far (C) comparisons. DEPs were defined by |log2 fold change| > 1 and P < 0.05. (D–F) Hierarchical clustering heatmaps of DEPs identified in the corresponding Tumor vs. Peritumor-far (D), Tumor vs. Peritumor-near (E), and Peritumor-near vs. Peritumor-far (F) comparisons. Differential protein expression across Tumor, Peritumor-near, and Peritumor-far regions. (A–C) Volcano plots showing differentially expressed proteins (DEPs) identified in the Tumor vs. Peritumor-far (A), Tumor vs. Peritumor-near (B), and Peritumor-near vs. Peritumor-far (C) comparisons. DEPs were defined by |log2 fold change| > 1 and P < 0.05. (D–F) Hierarchical clustering heatmaps of DEPs identified in the corresponding Tumor vs. Peritumor-far (D), Tumor vs. Peritumor-near (E), and Peritumor-near vs. Peritumor-far (F) comparisons. Functional characteristics of DEPs In the Tumor vs. Peritumor-far comparison, enriched GO terms were dominated by branched-chain amino acid catabolism (isoleucine, leucine, and valine), tryptophan and benzoate metabolic processes, and mitochondrial or peroxisomal matrix components. Enzyme-related categories, including NAD binding, retinol/aldehyde dehydrogenase activity, and ethanol binding, were also significantly enriched (Fig. 5A). Consistent with these findings, KEGG analysis highlighted metabolic reprogramming pathways such as carbon metabolism, fatty acid degradation, glycolysis/gluconeogenesis, amino acid biosynthesis and degradation, tryptophan metabolism, and cytochrome P450-mediated xenobiotic and drug metabolism (Fig. 5B). In the Tumor vs. Peritumor-near comparison, GO enrichment similarly emphasized branched-chain amino acid catabolism, tryptophan and retinol metabolic processes, and mitochondrial/peroxisomal compartments, along with complexes such as the prefoldin complex and centrosome (Fig. 5C). KEGG pathways remained predominantly metabolism-centered, including fatty acid degradation, β-alanine metabolism, valine/leucine/isoleucine degradation, biosynthesis of amino acids, and multiple xenobiotic-metabolism and drug-metabolism pathways. Additional enrichment was observed in spliceosome-related processes and chemical carcinogenesis-DNA adduct pathways (Fig. 5D). GO and KEGG enrichment analyses were not performed for the Peritumor-near vs. Peritumor-far comparison because only four DEPs were identified. Fig. 5GO and KEGG enrichment analyses of differentially expressed proteins (DEPs) in hepatocellular carcinoma tissues. GO and KEGG enrichment analyses of DEPs identified in the Tumor vs. Peritumor-far (A, B) and Tumor vs. Peritumor-near (C, D) comparisons. The GO enrichment plots show the top 5 significantly enriched terms in the biological process (BP), cellular component (CC), and molecular function (MF) categories. The KEGG enrichment plots show the top 15 enriched pathways in each comparison. GO and KEGG enrichment analyses of differentially expressed proteins (DEPs) in hepatocellular carcinoma tissues. GO and KEGG enrichment analyses of DEPs identified in the Tumor vs. Peritumor-far (A, B) and Tumor vs. Peritumor-near (C, D) comparisons. The GO enrichment plots show the top 5 significantly enriched terms in the biological process (BP), cellular component (CC), and molecular function (MF) categories. The KEGG enrichment plots show the top 15 enriched pathways in each comparison. Domain characteristics and subcellular localization of DEPs Domain enrichment analysis revealed that DEPs in both tumor-involved comparisons were enriched in NAD(P)-binding and dehydrogenase-related domains (Fig. 6A, B). For the Peritumor-near vs. Peritumor-far comparison, the four DEPs showed enrichment for domains such as Tetraspanin, DNA-binding, Vimentin, and SH3 (Fig. 6C). Subcellular localization analysis showed that DEPs in the two tumor-involved comparisons were mainly distributed in the cytosol, followed by the nucleus, mitochondria, and extracellular compartment (Fig. 6D). Notably, the Peritumor-near vs. Peritumor-far group showed a relatively higher proportion of mitochondrial and extracellular annotations, although the absolute number of DEPs was small. Overall, the DEP results revealed pronounced metabolism-related protein abundance changes in tumor tissues and provided expression-level information complementary to the differential conformational accessibility detected by LiP-MS. Fig. 6Domain enrichment and subcellular localization analyses of differentially expressed proteins (DEPs) in hepatocellular carcinoma tissues. (A–C) Protein domain enrichment analyses of DEPs in the Tumor vs. Peritumor-far (A), Tumor vs. Peritumor-near (B), and Peritumor-near vs. Peritumor-far (C) comparisons, with the top 15 significantly enriched domains shown for each comparison. (D) Subcellular localization profiles of DEPs identified in the Tumor vs. Peritumor-far, Tumor vs. Peritumor-near, and Peritumor-near vs. Peritumor-far comparisons. Domain enrichment and subcellular localization analyses of differentially expressed proteins (DEPs) in hepatocellular carcinoma tissues. (A–C) Protein domain enrichment analyses of DEPs in the Tumor vs. Peritumor-far (A), Tumor vs. Peritumor-near (B), and Peritumor-near vs. Peritumor-far (C) comparisons, with the top 15 significantly enriched domains shown for each comparison. (D) Subcellular localization profiles of DEPs identified in the Tumor vs. Peritumor-far, Tumor vs. Peritumor-near, and Peritumor-near vs. Peritumor-far comparisons. Combined profiling of conformational accessibility and protein abundance alterations in HCC Relationship between LiP-MS peptide changes and protein abundance changes To evaluate the relationship between peptide-level LiP-MS changes and global protein abundance alterations, LiP peptide log2 fold changes were compared with the corresponding 4D label-free protein log2 fold changes in the two tumor-involved comparisons. In the Tumor vs. Peritumor-far comparison, significant positive correlations were observed in both the DDA dataset (Spearman R = 0.69, P < 2.2 × 10–16) and the DIA dataset (Spearman R = 0.72, P < 2.2 × 10–16) (Fig. 7A). Similarly, in the Tumor vs. Peritumor-near comparison, LiP-MS peptide changes were positively correlated with protein abundance changes in both the DDA dataset (Spearman R = 0.69, P < 2.2 × 10⁻16) and the DIA dataset (Spearman R = 0.72, P < 2.2 × 10⁻16) (Fig. 7B). These findings indicate a consistent association between peptide-level LiP-MS signals and global protein abundance changes in both tumor-involved comparisons. However, the correlations were not complete, suggesting that some peptide-level differences may not be fully explained by protein abundance changes alone. To further distinguish these patterns, DCP and DEP datasets were integrated in the subsequent analyses. Conformation-specific altered proteins Proteins containing differential LiP peptides but lacking significant changes in global protein abundance were classified as conformation-specific altered proteins (CSAPs). This analysis identified 488 proteins in the Tumor vs. Peritumor-far comparison and 469 proteins in the Tumor vs. Peritumor-near comparison (Fig. 7C, D). Because only four DEPs were detected between Peritumor-near and Peritumor-far tissues, CSAPs were not evaluated for this direct comparison. GO enrichment analysis showed that CSAPs in the Tumor vs. Peritumor-far comparison were associated with processes such as amino acid metabolism, cytoplasmic translation, and small-molecule or organic acid catabolism, together with compartment terms including vesicle-lumen structures and the mitochondrial matrix (Fig. 7E). KEGG pathways were enriched in carbon and amino acid metabolism, glutathione and drug-related metabolism, proteasome components, and nucleocytoplasmic transport (Fig. 7F). For the Tumor vs. Peritumor-near comparison, GO enrichment included small-molecule catabolism, cellular amino acid metabolism, RNA splicing, and mRNA splicing via the spliceosome, together with cellular-component annotations for vesicle lumen and the mitochondrial matrix (Fig. 7G). KEGG pathways highlighted carbon metabolism, glycolysis/gluconeogenesis, amino acid biosynthesis, spliceosome-related mechanisms, and glutathione metabolism (Fig. 7H). In summary, CSAPs were distributed across metabolic, translational, and RNA-processing pathways in both comparisons. Although their global protein abundance remained statistically unchanged, these proteins contained differential LiP peptides, providing structural accessibility-related information that was not captured by abundance-based proteomics alone. Fig. 7Identification and functional enrichment analyses of conformation-specific altered proteins (CSAPs). (A, B) Correlation between LiP-MS peptide log2 fold changes and 4D label-free protein abundance log₂ fold changes in the Tumor vs. Peritumor-far (A) and Tumor vs. Peritumor-near (B) comparisons. Spearman’s rank correlation coefficients and corresponding P values are shown, with results from the DDA and DIA datasets are shown separately. (C, D) Venn diagrams showing the overlap between differentially conformed proteins (DCPs) and differentially expressed proteins (DEPs) in the Tumor vs. Peritumor-far (C) and Tumor vs. Peritumor-near (D) comparisons. Proteins containing differential LiP peptides without corresponding significant changes in protein abundance were classified as CSAPs. The Venn diagrams are schematic and not drawn to scale; the numerical labels indicate the exact numbers of elements in each set and intersection. (E–H) GO and KEGG enrichment analyses of CSAPs identified in the Tumor vs. Peritumor-far (E, F), and Tumor vs. Peritumor-near (G, H) comparisons. GO terms are categorized into biological process (BP), cellular component (CC), and molecular function (MF), with the top five significantly enriched terms presented for each category. KEGG plots display the top 15 enriched pathways. Identification and functional enrichment analyses of conformation-specific altered proteins (CSAPs). (A, B) Correlation between LiP-MS peptide log2 fold changes and 4D label-free protein abundance log₂ fold changes in the Tumor vs. Peritumor-far (A) and Tumor vs. Peritumor-near (B) comparisons. Spearman’s rank correlation coefficients and corresponding P values are shown, with results from the DDA and DIA datasets are shown separately. (C, D) Venn diagrams showing the overlap between differentially conformed proteins (DCPs) and differentially expressed proteins (DEPs) in the Tumor vs. Peritumor-far (C) and Tumor vs. Peritumor-near (D) comparisons. Proteins containing differential LiP peptides without corresponding significant changes in protein abundance were classified as CSAPs. The Venn diagrams are schematic and not drawn to scale; the numerical labels indicate the exact numbers of elements in each set and intersection. (E–H) GO and KEGG enrichment analyses of CSAPs identified in the Tumor vs. Peritumor-far (E, F), and Tumor vs. Peritumor-near (G, H) comparisons. GO terms are categorized into biological process (BP), cellular component (CC), and molecular function (MF), with the top five significantly enriched terms presented for each category. KEGG plots display the top 15 enriched pathways. Conformation–expression coupled proteins The proteins located in the overlapping regions of the DCP and DEP sets were defined as conformation-expression coupled proteins (CECPs), as they contained differential LiP peptides and also exhibited significant changes in global protein abundance. A total of 237 CECPs were identified in the Tumor vs. Peritumor-far comparison, and 205 in the Tumor vs. Peritumor-near comparison (Fig. 7C, D). In the Tumor vs. Peritumor-far comparison, GO enrichment indicated that CECPs were mainly involved in small-molecule and amino acid catabolic processes, together with mitochondrial and peroxisomal annotations, and catalytic functions dominated by oxidoreductase and lyase activities (Fig. 8A). KEGG pathways were largely centered on amino acid biosynthesis and degradation, carbon metabolic routes including glycolysis/gluconeogenesis and pyruvate metabolism, as well as cofactor biosynthesis and peroxisome-associated pathways (Fig. 8B). In the Tumor vs. Peritumor-near comparison, CECPs exhibited a highly similar GO pattern, with enrichment in amino acid and organic acid catabolism, mitochondrial/peroxisomal compartments, and NAD-dependent oxidoreductase activities (Fig. 8C). KEGG analysis also reflected comparable metabolic themes, covering amino acid metabolism, fatty-acid degradation, glycolytic and one-carbon pathways, together with additional terms such as cytochrome-P450-related drug metabolism and retinol metabolism (Fig. 8D). Molecular functional categorization further showed that CECPs were predominantly metabolic enzymes. In the Tumor vs. Peritumor-far comparison, the 237 CECPs included 125 metabolic enzymes (MEs), 37 RNA-binding proteins (RBPs), 5 transcription factors (TFs), 12 kinases/signaling proteins (KSPs), 8 transmembrane/channel proteins (TCPs), 14 structural proteins (SPs), and 36 other functional proteins (OFPs). In the Tumor vs. Peritumor-near comparison, the 205 CECPs comprised 114 MEs, 27 RBPs, 3 TFs, 7 KSPs, 11 TCPs, 12 SPs, and 31 OFPs (Fig. 8E, F and Supplementary Table 5). Collectively, CECPs from both comparisons were predominantly associated with metabolism- and redox-related pathways, highlighting proteins that exhibited coordinated changes in conformational accessibility and global protein abundance. Fig. 8Functional enrichment and classification analysis of conformation-expression coupled proteins (CECPs). (A–D) GO and KEGG enrichment analyses of CECPs identified in the Tumor vs. Peritumor-far (A, B) and Tumor vs. Peritumor-near (C, D) comparisons. GO analysis includes three categories: biological process (BP), cellular component (CC), and molecular function (MF), with the top five most significantly enriched terms shown for each category. KEGG plots display the top 15 enriched pathways in each comparison. (E–F) Functional categorization of CECPs identified in the Tumor vs. Peritumor-far (E) and Tumor vs. Peritumor-near (F) comparisons. MEs, metabolic enzymes; RBPs, RNA-binding proteins; TFs, transcription factors; KSPs, kinases/signaling proteins; TCPs, transmembrane/channel proteins; SPs, structural proteins; OFPs, other functional proteins. Functional enrichment and classification analysis of conformation-expression coupled proteins (CECPs). (A–D) GO and KEGG enrichment analyses of CECPs identified in the Tumor vs. Peritumor-far (A, B) and Tumor vs. Peritumor-near (C, D) comparisons. GO analysis includes three categories: biological process (BP), cellular component (CC), and molecular function (MF), with the top five most significantly enriched terms shown for each category. KEGG plots display the top 15 enriched pathways in each comparison. (E–F) Functional categorization of CECPs identified in the Tumor vs. Peritumor-far (E) and Tumor vs. Peritumor-near (F) comparisons. MEs, metabolic enzymes; RBPs, RNA-binding proteins; TFs, transcription factors; KSPs, kinases/signaling proteins; TCPs, transmembrane/channel proteins; SPs, structural proteins; OFPs, other functional proteins. Region-conserved and region-specific CECPs To identify CECPs shared between the two tumor-involved comparisons, the intersection of the CECPs sets was defined as region-conserved CECPs (rcCECPs), whereas non-overlapping proteins were classified as region-specific CECPs (rsCECPs) (Fig. 9A). This analysis yielded 169 rcCECPs, 36 rsCECPs unique to the Tumor vs. Peritumor-near comparison, and 68 rsCECPs unique to the Tumor vs. Peritumor-far comparison. GO enrichment of rcCECPs showed a strong focus on small-molecule and amino/organic acid catabolism, peroxisomal and mitochondrial compartments, and redox-related molecular functions such as NAD-dependent oxidoreductase and lyase activities (Fig. 9B). KEGG pathways were mainly enriched in fatty-acid and branched-chain amino acid degradation, pyruvate and carbon metabolism, one-carbon and folate-related processes, and cofactor biosynthesis (Fig. 9C, D), reflecting conserved metabolic remodeling across the sampled tumor regions. The 36 rsCECPs unique to the Tumor vs. Peritumor-near comparison were enriched in amino acid and organic acid biosynthesis, xenobiotic metabolism, mitochondrial and granule luminal structures, and ligand-binding activities including glutathione and oligopeptide binding (Fig. 10A). KEGG terms highlighted cytochrome-P450-mediated drug metabolism, amino acid and glutathione metabolism, and pathways linked to chemical carcinogenesis-DNA adducts (Fig. 10B). The 68 rsCECPs unique to the Tumor vs. Peritumor-far comparison showed a different pattern, characterized by RNA-splicing processes, spliceosomal complexes, RNA-binding activities, and antioxidant functions (Fig. 10C). KEGG enrichment included spliceosome, glycolysis/gluconeogenesis, amino acid biosynthesis, carbon and glutathione metabolism, and PPAR signaling pathway (Fig. 10D). Overall, rcCECPs reflect conserved metabolic alterations shared between the two tumor-involved comparisons, whereas rsCECPs showed distinct functional patterns depending on the peritumoral reference region. The Tumor vs. Peritumor-near-specific proteins were mainly associated with xenobiotic and amino acid metabolism, whereas the Tumor vs. Peritumor-far-specific proteins were more closely related to RNA processing and antioxidant activity. Fig. 9Functional enrichment analysis of region-conserved conformation-expression coupled proteins (rcCECPs). (A) Venn diagram showing the overlap between CECPs identified in the Tumor vs. Peritumor-far and Tumor vs. Peritumor-near comparisons. The overlapping proteins were classified as rcCECPs. The Venn diagrams are schematic and not drawn to scale; the numerical labels indicate the exact numbers of elements in each set and intersection. (B, C) GO and KEGG enrichment analyses of rcCECPs. GO terms are categorized into biological process (BP), cellular component (CC), and molecular function (MF), with the top five most significantly enriched terms displayed for each category. KEGG analysis presents the top 15 most enriched pathways. (D) Emapplot showing the functional relationships among the top 15 enriched KEGG pathways. Functional enrichment analysis of region-conserved conformation-expression coupled proteins (rcCECPs). (A) Venn diagram showing the overlap between CECPs identified in the Tumor vs. Peritumor-far and Tumor vs. Peritumor-near comparisons. The overlapping proteins were classified as rcCECPs. The Venn diagrams are schematic and not drawn to scale; the numerical labels indicate the exact numbers of elements in each set and intersection. (B, C) GO and KEGG enrichment analyses of rcCECPs. GO terms are categorized into biological process (BP), cellular component (CC), and molecular function (MF), with the top five most significantly enriched terms displayed for each category. KEGG analysis presents the top 15 most enriched pathways. (D) Emapplot showing the functional relationships among the top 15 enriched KEGG pathways. Fig. 10Functional enrichment analysis of region-specific conformation-expression coupled proteins (rsCECPs). GO and KEGG enrichment analyses of rsCECPs identified in the Tumor vs. Peritumor-near (A, B) and Tumor vs. Peritumor-far (C, D) comparisons. GO categories include Biological Process (BP), Cellular Component (CC), and Molecular Function (MF), with the top five significantly enriched terms displayed for each category. KEGG analysis presents the top 15 most enriched pathways. Functional enrichment analysis of region-specific conformation-expression coupled proteins (rsCECPs). GO and KEGG enrichment analyses of rsCECPs identified in the Tumor vs. Peritumor-near (A, B) and Tumor vs. Peritumor-far (C, D) comparisons. GO categories include Biological Process (BP), Cellular Component (CC), and Molecular Function (MF), with the top five significantly enriched terms displayed for each category. KEGG analysis presents the top 15 most enriched pathways. Molecular characteristics and clinical significance of hub CECPs Directional concordance between conformational accessibility and protein abundance changes Directional concordance analysis showed that all the 237 CECPs identified in the Tumor vs. Peritumor-far comparison (100%) exhibited the same direction of change between the representative LiP peptide signal and protein abundance. Similarly, 204 of the 205 CECPs in the Tumor vs. Peritumor-near comparison (99.5%) showed concordant directional changes. Notably, all the 169 rcCECPs displayed concordant changes in representative LiP peptide signals and protein abundance across both tumor-involved comparisons (Supplementary Table 6). Together with the correlation analyses shown in Fig. 7A–B, these findings indicate coordinated changes in peptide-level conformational accessibility and protein abundance among CECPs, although the LiP-MS signals should not be interpreted as direct evidence of uniform conformational changes across the entire protein. Identification and molecular characterization of hub rcCECPs To identify highly connected proteins within the region-conserved CECP set, we constructed a PPI network using the 169 rcCECPs and selected the top 30 proteins with the highest degree of connectivity as hub rcCECPs (Fig. 11A). Domain enrichment analysis showed that these hub proteins were predominantly associated with metabolic and enzyme-related domains, including aldehyde dehydrogenase and thiolase families, as well as NAD(P)-binding and pyridoxal-phosphate-dependent transferase domains (Fig. 11B). These findings indicate that the hub rcCECPs were mainly composed of proteins with enzymatic and metabolic functions. Fig. 11Protein-protein interaction (PPI) network and domain enrichment analysis of region-conserved conformation-expression coupled proteins (rcCECPs). (A) PPI network constructed using the 169 rcCECPs, with the top 30 proteins ranked by degree identified as hub rcCECPs. (B) Domain enrichment analysis of the 30 hub rcCECPs, showing the top 20 significantly enriched protein domains. Protein-protein interaction (PPI) network and domain enrichment analysis of region-conserved conformation-expression coupled proteins (rcCECPs). (A) PPI network constructed using the 169 rcCECPs, with the top 30 proteins ranked by degree identified as hub rcCECPs. (B) Domain enrichment analysis of the 30 hub rcCECPs, showing the top 20 significantly enriched protein domains. Peptide-level mapping and clinical relevance of selected hub rcCECPs We next assessed the mRNA expression profiles of the 30 hub rcCECPs to explore their potential clinical relevance. Among them, ACLY, ALDH18A1, GMPS, and DHX9, which showed increased representative LiP peptide signals and higher protein abundance in both tumor-involved comparisons, also exhibited significantly elevated transcript levels in HCC tissues compared with non-tumor liver tissues in the TCGA-GTEx dataset (P < 0.001; Fig. 12A–D). These four proteins were therefore selected for further peptide-level mapping and prognostic analysis. Differential LiP peptides from ACLY, ALDH18A1, GMPS, and DHX9 were mapped to their corresponding canonical protein sequences together with the annotated InterPro domains (Fig. 12E, F). The DIA dataset generally contained more mapped differential peptides than the DDA dataset. In ACLY and DHX9, differential peptides were distributed across several regions of the protein sequences, whereas the mapped peptides in GMPS and ALDH18A1 were concentrated within a more limited sequence range. The peptides were located both within annotated domains and in adjacent inter-domain regions. Most mapped peptides showed positive log₂ fold changes in the two tumor-involved comparisons, although a small number of peptides, particularly in ALDH18A1, showed negative changes. Overall, the mapping results show that the LiP-MS differences were localized to specific peptide regions and were not uniformly distributed across the entire protein sequence. Kaplan-Meier analyses further showed that higher expression of each gene was associated with shorter overall survival: ACLY, HR = 1.6 and Logrank P = 0.011; ALDH18A1, HR = 1.4 and Logrank P = 0.042; GMPS, HR = 1.7 and Logrank P = 0.0041; DHX9, HR = 1.6 and Logrank P = 0.008 (Fig. 12G–J). These results indicate that the four selected hub rcCECPs showed consistent increases at the LiP peptide, protein abundance, and transcript levels and were associated with unfavorable survival outcomes in HCC. Among the remaining 26 hub rcCECPs showing decreased representative LiP peptide signals and lower protein abundance, 20 also showed significantly reduced transcript levels in TCGA-GTEx dataset, whereas five displayed no significant transcript difference, and one showed increased transcript expression (Supplementary Table 7). Thus, transcript-level changes were consistent with the protein abundance patterns for most hub rcCECPs, although several discordant cases were also observed. These discrepancies may reflect post-transcriptional regulation, altered protein stability, post-translational modifications, or changes in conformational accessibility that are not captured by transcript abundance alone. Fig. 12Expression, peptide-level mapping, and prognostic value of the selected hub region-conserved conformation-expression coupled proteins (rcCECPs) in hepatocellular carcinoma (HCC). (A–D) Differential mRNA expression of ACLY, ALDH18A1, GMPS, and DHX9 between HCC and non-tumor liver tissues based on the TCGA-GTEx dataset (normal, n = 160; tumor, n = 371). (E, F) Mapping of differential LiP peptides from ACLY, ALDH18A1, GMPS, and DHX9 to their canonical protein sequences and annotated InterPro domains in the Tumor vs. Peritumor-far and Tumor vs. Peritumor-near comparisons. (G–J) Results from the DDA and DIA datasets are shown in E and F, respectively. Kaplan–Meier overall survival analyses of ACLY, ALDH18A1, GMPS, and DHX9 in patients with HCC. *** P < 0.001. Expression, peptide-level mapping, and prognostic value of the selected hub region-conserved conformation-expression coupled proteins (rcCECPs) in hepatocellular carcinoma (HCC). (A–D) Differential mRNA expression of ACLY, ALDH18A1, GMPS, and DHX9 between HCC and non-tumor liver tissues based on the TCGA-GTEx dataset (normal, n = 160; tumor, n = 371). (E, F) Mapping of differential LiP peptides from ACLY, ALDH18A1, GMPS, and DHX9 to their canonical protein sequences and annotated InterPro domains in the Tumor vs. Peritumor-far and Tumor vs. Peritumor-near comparisons. (G–J) Results from the DDA and DIA datasets are shown in E and F, respectively. Kaplan–Meier overall survival analyses of ACLY, ALDH18A1, GMPS, and DHX9 in patients with HCC. *** P < 0.001. Discussion Advances in mass-spectrometry-based proteomics have made it possible to examine protein features beyond abundance alone. Building on this trend, the present work combined LiP-MS with 4D label-free quantitative proteomics to explore conformational accessibility and expression changes across different regions of HCC tissues. Rather than focusing solely on expression shifts, this approach provided complementary information on peptide-level conformational accessibility and global protein abundance, adding another layer to the characterization of regional molecular variability in HCC. Across tumor and peritumoral regions, tumor tissues showed substantially more alterations in conformational accessibility, whereas the two peritumoral regions were comparatively similar. This pattern is consistent with recent spatial transcriptomic and metabolomic studies showing marked molecular differences between liver tumors and surrounding liver tissues17,18. As only three predefined tissue regions were examined, further spatially resolved analyses are needed to determine whether these changes follow a continuous gradient. Most DCPs were enriched for domains typical of metabolic enzymes, including NAD(P)-binding, aldehyde/alcohol dehydrogenase, and pyridoxal-phosphate-dependent transferase, which mirrors the metabolic emphasis seen throughout our pathway analyses. Because the enrichment analyses were conducted at the parent-protein level, whether the differential peptides are preferentially located within catalytic, substrate-binding, or coenzyme-recognition regions requires further site-specific structural analysis. These findings expand the current understanding of metabolic dysregulation in HCC by adding a conformational accessibility dimension to previously reported transcriptional and proteomic changes19. Subcellular localization analysis further indicated that most DCPs were cytosolic, with substantial representation in the mitochondria and nucleus. This distribution aligns with the metabolic and RNA-processing pathways enriched in our data. Differential protein expression analyses revealed that DEPs of both tumor-related comparisons were enriched in amino acid biosynthesis and metabolism, carbon metabolism, fatty acid degradation, glycolysis/gluconeogenesis, and tryptophan metabolism, which are metabolic signatures consistent with known metabolic reprogramming in HCC20,21. In contrast, the two peritumoral regions exhibited almost identical protein abundance profiles, despite detectable differences in peptide-level conformational accessibility. This discrepancy suggests that LiP-MS may capture regional protein alterations that are not accompanied by statistically significant changes in total protein abundance. However, whether these accessibility changes arise before, alongside, or independently of abundance alterations still remains to be clarified. Local peptide accessibility may also be influenced by ligand binding, protein complex formation, post-translational modifications, or other conformational changes10,22. By integrating conformational accessibility and quantitative proteomics, we identified CSAPs, defined as proteins containing differential LiP peptides without significant abundance changes, and CECPs, which showed both altered conformational accessibility and protein abundance. CSAPs were broadly enriched in amino acid metabolism, carbon metabolism, glycolysis/gluconeogenesis, glutathione metabolism, and RNA-processing pathways. These pathways were also prominent among CECPs, highlighting conserved metabolic and redox-related alterations in tumor tissues. The large number of CSAPs suggests that conformational accessibility provides protein-level information not captured by abundance measurements alone. Such differences may arise from post-translational modifications, ligand binding, protein interactions, or chaperone-mediated folding, all of which can alter the exposure of specific protein regions23–25. Further structural and biochemical studies are warranted to elucidate the molecular basis and functional significance of these conformational changes in CSAPs, which occur without significant alterations in protein abundance. Correlation analyses further showed positive associations between representative LiP peptide changes and corresponding protein abundance changes, and most CECPs exhibited concordant directions between the two measurements. These findings indicate that alterations in conformational accessibility frequently accompanied changes in protein abundance. CECPs therefore represent proteins with coordinated conformational accessibility and abundance changes, although the relative contribution of each component requires further experimental dissection. Similar relationships between protein structural state, stability, and abundance have been described in other tumor contexts26,27. The comparisons between Tumor vs. Peritumor-near and Tumor vs. Peritumor-far further allowed CECPs to be divided into region-conserved CECPs (rcCECPs) and region-specific CECPs (rsCECPs). The rcCECPs were mainly enriched in core metabolic pathways, including branched-chain amino acid degradation, fatty-acid degradation, pyruvate/carbon metabolism, and one-carbon metabolism, indicating shared metabolic alterations across the two tumor-involved comparisons. In contrast, the rsCECPs showed different functional patterns. Those specific to the Tumor vs. Peritumor-near comparison were associated with xenobiotic, drug, glutathione, and amino acid metabolism, whereas those specific to the Tumor vs. Peritumor-far comparison were linked to RNA splicing, antioxidant activity, glycolysis/gluconeogenesis, and PPAR signaling. These comparison-specific patterns may reflect regional molecular heterogeneity surrounding HCC. Validation in larger cohorts and more finely resolved spatial samples will be needed to determine whether they represent reproducible biological differences between peritumoral regions. These patterns are consistent with recent spatial studies revealing molecular heterogeneity within and surrounding HCC tissues28,29. To further characterize the molecular relevance of hub rcCECPs, a PPI network was constructed and the top 30 highly connected proteins were examined. Domain enrichment analysis showed that these proteins were predominantly associated with metabolic enzyme-related domains, in agreement with the broader metabolism-centered patterns observed throughout the study. Transcript-level assessment further identified ACLY, ALDH18A1, GMPS, and DHX9 as proteins showing increased representative LiP peptide signals, increased protein abundance, elevated mRNA expression, and associations with poorer overall survival. Specifically, ACLY serves as a key enzyme linking carbon metabolism to lipid biosynthesis and histone acetylation30,31, and is widely implicated in cancer-associated metabolic reprogramming32. ALDH18A1 catalyzes a rate-limiting step in proline and arginine biosynthesis33, and has been associated with redox homeostasis and metabolic plasticity in malignant cells34,35. GMPS governs the final step of de novo GMP synthesis and supports nucleotide supply required for rapid tumor cell proliferation36. DHX9 functions as a multifunctional RNA helicase involved in transcriptional regulation and RNA processing and has been linked to oncogenic stress responses and genome stability in cancer37,38. These known functions provide biological context for their identification, while functional experiments are required to establish whether the observed LiP-MS changes affect their enzymatic or regulatory activities. Peptide-level mapping showed that differential LiP peptides from ACLY, ALDH18A1, GMPS, and DHX9 were distributed within both annotated domains and intervening regions. The mapped peptides were localized to specific sequence regions rather than uniformly distributed across the complete protein sequences, and DIA generally provided broader peptide coverage than DDA. These findings support localized differences in conformational accessibility within the selected proteins. Higher-resolution structural mapping and activity assays will be needed to determine whether these regions correspond to specific three-dimensional rearrangements or functional sites. Among the remaining hub rcCECPs showing decreased representative LiP peptide signals and lower protein abundance, most also showed reduced transcript expression, whereas several displayed discordant transcript-protein patterns. These discrepancies may reflect post-transcriptional regulation, altered protein turnover, or changes in conformational accessibility that are not captured by transcript abundance alone. Similar transcriptome-proteome divergences have been reported in proteogenomic studies of HCC5,39, highlighting that protein-level alterations cannot always be inferred directly from transcriptomic data. Several limitations should be acknowledged. First, the small patient cohort limits the statistical power and generalizability of the findings, although the matched multi-region design reduced inter-individual variation. Second, LiP-MS was performed using frozen tissue lysates, and sample processing may influence local peptide accessibility; therefore, the observed signals represent relative ex vivo conformational accessibility patterns rather than direct measurements in living tissues. Third, LiP-MS reflects susceptibility to limited proteolysis and may also be influenced by protein abundance, ligand binding, protein interactions, and post-translational modifications. Fourth, using the overlap between DDA and DIA increased analytical stringency but may have excluded valid peptides detected by only one acquisition mode. Finally, the TCGA-GTEx analyses provided transcript-level and prognostic support but did not independently validate the protein-level or functional changes. Larger cohorts and complementary structural and functional studies are needed to confirm these findings. Conclusion In summary, the integration of LiP-MS and 4D label-free quantitative proteomics revealed regional alterations in protein conformational accessibility and abundance in HCC tissues. Metabolism- and redox-related proteins were prominently represented among CSAPs, CECPs, and rcCECPs, while ACLY, ALDH18A1, GMPS, and DHX9 emerged as hub candidates with coordinated peptide-level, protein, transcript, and prognostic patterns. These findings provide complementary protein-level insights into HCC heterogeneity and a basis for further structural and functional investigation. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2 Supplementary Material 3 Supplementary Material 3 Supplementary Material 4 Supplementary Material 4 Supplementary Material 5 Supplementary Material 5 Supplementary Material 6 Supplementary Material 6 Supplementary Material 7 Supplementary Material 7