The localization and interaction of proteins often underpin their biological functions, and are essential to understanding what proteins do within cells and tissues. Spatial proteomics technologies allow high-resolution, high-depth analysis of proteins while preserving the structural integrity of cells and tissues, thereby providing a better understanding of biological processes such as intracellular signal transduction, intercellular communication, and the interaction between cells and the extracellular matrix. Spatial proteomics was named Method of the Year 2024 by Nature Methods, underscoring its revolutionary role in revealing biological complexity and advancing precision medicine and cancer research.
In March 2025, a team led by Academician He Fuchu from the Guangdong Institute of Intelligence Science and Technology and the Huayan Large-Scale Facility Project, together with the research group of Liu Chao at Beihang University and Dr. Zhang Honghua at Sun Yat-sen Memorial Hospital of Sun Yat-sen University, published an article in Analytical Chemistry entitled "High-Efficiency Cell-Type Proteomics Strategy Deciphers Cholangiocarcinoma Fibrosis-Associated PathologicalHeterogeneity." The paper describes a highly efficient cell-type-resolved spatial proteomics technology that combines the pathological image analysis algorithm π-CellSeg-CCA, laser microdissection (LMD), SISPROT sample preparation and high-sensitivity mass spectrometry, and was successfully applied to decipher the heterogeneity of fibrosis-associated pathology in cholangiocarcinoma. The π-CellSeg-CCA algorithm can automatically outline clusters of cholangiocarcinoma cells and normal bile duct cells in pathological images of cholangiocarcinoma sections, with a recognition accuracy of 90%, and the outlines are imported into a laser microdissection instrument for efficient cutting. Using this method, the team identified more than 8,000 proteins from a 1 mm² section sample, including marker proteins specifically expressed in cholangiocarcinoma; they also discovered MUC16, a marker protein specifically upregulated in fibrotic regions of cholangiocarcinoma, revealing its association with poor patient prognosis and disease progression.
Article Summary
Main Content
This study established a highly efficient cell-type-resolved spatial proteomics technology that integrates the pathological image analysis algorithm π-CellSeg-CCA, LMD, SISPROT sample preparation and high-sensitivity mass spectrometry detection. After cholangiocarcinoma sections were stained with H&E and scanned, the π-CellSeg-CCA algorithm automatically outlined the cholangiocarcinoma and normal bile duct cell clusters in the images; the cluster outlines were then imported into a laser microdissection instrument for efficient cutting. The collected samples were processed by SISPROT and analyzed on a timsTOF Pro high-sensitivity mass spectrometer (Figure 1).
Figure 1. The highly efficient cell-type-resolved spatial proteomics technology
To outline cholangiocarcinoma/normal bile duct cell clusters efficiently in H&E-stained images of cholangiocarcinoma sections, the team developed the pathological image analysis algorithm π-CellSeg-CCA. Built on the machine learning algorithms Mask R-CNN and ResNet-18, π-CellSeg-CCA can segment and classify cholangiocarcinoma/normal bile duct cell clusters and feed the cluster outlines into a laser microdissection instrument for efficient cutting and collection. The segmentation function of π-CellSeg-CCA is implemented by the Mask R-CNN-DCT model, which is based on the instance segmentation model Mask R-CNN with a discrete cosine transform (DCT) module added to its segmentation branch; this represents high-resolution masks as vectors without increasing model complexity, and the higher-quality masks improve the model's segmentation performance on cell clusters. As shown in Figure 2A, the improved Mask R-CNN-DCT achieved better segmentation and higher accuracy on cholangiocarcinoma cell clusters. The benign/malignant classification function of π-CellSeg-CCA is implemented by the ResNet-18 model, reaching 90% recognition accuracy. Figure 2B shows examples of images identified as cholangiocarcinoma and normal bile duct; the vast majority of images identified as cholangiocarcinoma reached a confidence level of 99%.
Figure 2. Performance of π-CellSeg-CCA in the segmentation (A) and classification (B) of cholangiocarcinoma/normal bile duct cell clusters; the percentages in panel B indicate the probability of cholangiocarcinoma
Using the highly efficient cell-type-resolved spatial proteomics technology, the team investigated the heterogeneity of fibrosis-associated pathology in cholangiocarcinoma. As shown in Figure 3A, a total of 6 normal bile duct samples (Normal BD), 5 non-fibrotic cholangiocarcinoma samples (NF CCA) and 7 fibrotic-region cholangiocarcinoma samples (Fibro CCA) were analyzed, yielding the following main findings:
1. An average of 8,000 proteins were identified from each 1 mm² section sample, and the Pearson correlation coefficient between samples of the same type exceeded 0.85, demonstrating the stability and analytical performance of the method (Figures 3B and 3C).
2. Comparative analysis of the two cholangiocarcinoma tissue types against normal bile duct tissue identified three well-established biomarkers highly expressed in cholangiocarcinoma tissue—S100P, MUC1 and MUC5AC—confirming the accuracy of the method (Figure 3D).
3. Comparative analysis of fibrotic-region cholangiocarcinoma against non-fibrotic cholangiocarcinoma identified MUC16 and S100A2 as upregulated in fibrotic-region cholangiocarcinoma, which were taken forward for further validation (Figure 3E).
Figure 3. Cell-type proteomic analysis of cholangiocarcinoma heterogeneity
Of the two proteins found to be upregulated in fibrotic-region cholangiocarcinoma, MUC16 and S100A2, MUC16 was selected for further study given its membrane protein localization and broader coverage in the literature, and the team investigated its relationship with fibrotic progression in cholangiocarcinoma, reaching the following main conclusions:
1. Multiplex immunofluorescence staining showed that MUC16 is specifically and highly expressed in fibrotic-region cholangiocarcinoma, and its expression level correlates positively with the degree of fibrosis (Figures 4A and 4B).
2. Statistical analysis combined with clinical data showed that patients with high MUC16 expression presented a higher T stage and more severe lymph node metastasis (LNM) and distant metastasis (DM) (Figures 4C and 4D).
Figure 4. Validation of the association between MUC16 and fibrosis in cholangiocarcinoma
In summary, this study developed a highly efficient cell-type-resolved spatial proteomics technology and successfully applied it to decipher the heterogeneity of fibrosis-associated pathology in cholangiocarcinoma. The π-CellSeg-CCA algorithm developed here markedly improves the efficiency of spatial proteomics analysis and has good scalability: with adjustment and training it can be applied to other cancer types, and it has promising application prospects.
