Proteomic Analysis of Tissue Proteins Related to Lateral Lymph Node Metastasis in Papillary Thyroid Microcarcinoma

Abstract

Patients with lateral lymph node metastasis (LLNM) may experience higher locoregional recurrence rates and poorer prognoses compared to those without LLNM, highlighting the need for effective preoperative stratification to reliably assess risk LLNM. In this study, we collected PTMC samples from Peking Union Medical College Hospital and employed data-independent acquisition mass spectrometry proteomics technique to identify protein profiles in PTMC tissues with and without LLNM. Pseudo temporal analysis and single sample gene set enrichment analysis were conducted in combination with The Cancer Genome Atlas Thyroid Carcinoma for functional coordination analysis and the construction of a prediction model based on random forest. Non-negative matrix factorization (NMF) clustering was utilized to classify molecular subtypes of PTMC. Our findings revealed that the differential activation of pathways such as MAPK and PI3K was critical in enhancing the lateral lymph node metastatic potential of PTMC. We successfully screened biomarkers via machine learning and public databases, creating an effective prediction model for metastasis. Additionally, we explored the mechanism of metastasis-associated PTMC subtypes via NMF clustering. These insights into LLNM mechanisms in PTMC may contribute to future biomarker screening and the identification of therapeutic targets.

Keywords: biomarker; lateral lymph node metastasis; mass spectrometry; papillary thyroid microcarcinoma; proteomics.

Figure 1

General features of PTMC proteomics. (A) Principal component analysis of metastatic vs nonmetastatic groups. (B) Pooled analysis of cancer tissues in metastatic vs nonmetastatic groups. (C) Pooled analysis of cancer tissues in metastatic groups vs metastatic cancer tissues in lymph nodes. (D) Difference volcano plots of LLNM-tumor vs LLNM-normal groups. (E) Difference volcano plots of NLNM-tumor vs NLNM-normal groups. (F) Difference volcano plots of LLNM-mln vs LLNM-nmln groups. (G) Comparison heatmap of classic tumor pathway ssGSEA scores among NLNM-tumor, LLNM-tumor, and LLNM-mln groups. (H) Differences in immune infiltration were calculated based on ssGSEA.

Figure 2

Proteomics pseudotemporal analysis and pathway protein expression differences. (A) Clustering of protein expression trends based on the Mfuzz package, with better clustering at K = 6. (B) GSEA of Cluster 1. (C) Expression trends of MAPK pathway genes in Cluster 1. (D) Expression trends of PI3K pathway genes in Cluster 1.

Figure 3

Pathway expression differences in metastasis and nonmetastasis groups. (A) Multiple MAPK-related pathways were activated in the LLNM-tumor group relative to the NLNM-tumor group. (B) MAPK and PI3K pathways were activated in the LLNM-mln group relative to the LLNM-tumor group. (C) Expression of MAPK subfamily hub proteins in the differential proteins. (D) Expression of MAPK pathway hub proteins in the differential proteins uniquely present in the LLNM-tumor vs LLNM-normal relative to the NLNM-tumor vs NLNM-normal. Pathway hub proteins. (E) The expression of PI3K pathway hub proteins is uniquely present in LLNM-mln vs LLNM-nmln differential proteins relative to LLNM-tumor vs LLNM-normal. *P < 0.05, **P < 0.01, ***P < 0.001.

Figure 4

Random forest algorithm-based screening of metastasis-associated genes and constructing of predictive models. (A) Metastasis-associated proteins in the N0–N1a group, with importance greater than 0.002 as the threshold. (B) The five-gene model with high predictive efficacy in the N0–N1a group training set. (C) ROC of the model in the N0–N1a group validation set. (D) N0–N1b group of metastasis-associated proteins with importance greater than 0.002 as the threshold. (E) The five-gene model with high prediction efficacy in the training set of the N0–N1b group. (F) ROC of this model in the validation set of the N0–N1b group. (G) Expression differences of the nine metastasis-associated genes in N0, N1a, and N1b.

Figure 5

New PTMC subtypes and mechanism exploration based on metastasis-related genes. (A,B) Correlation analysis of metastasis-related genes in the two 5-gene models. (C) NMF clustering of TCGA-THCA samples based on the nine metastasis-related genes, and the clustering was the most stable at K = 3. (D) Proportion of different LN metastasis states in the new PTMC subtypes. (E–G) GO analysis of the differential expression genes among the three subtypes. (H) GSEA of differentially expressed genes among the three subtypes. (I) The 30 pathways with the greatest differences among the three subtypes. (J) Significant differences in immune infiltration among the three subtypes.