Identification of potential biomarkers in follicular thyroid carcinoma: bioinformatics and immunohistochemical analyses

Introduction

In recent decades, there has been a significant global rise in the incidence of differentiated thyroid carcinoma (DTC). According to the GLOBOCAN 2020 data on cancer incidence and mortality, thyroid cancer ranks ninth in terms of overall incidence and fifth among malignant tumors in women [1]. Follicular thyroid carcinoma (FTC), a subtype of DTC originating from thyroid follicular cells, represents a relatively less common cancer, constituting approximately 5–15 % of all thyroid malignancies. Despite its lower prevalence, FTC exhibits greater malignancy than papillary thyroid carcinoma (PTC) and a propensity for blood dissemination, elevating the risk of distant metastasis – most frequently observed in the bone and lungs [2], [3], [4]. The prognosis of FTC is predominantly influenced by the clinical stage of cancer, and early diagnosis and intervention typically result in a more favorable outcome. Surgical resection and other therapeutic modalities can effectively extend survival and control the disease in cases with limited local and distant metastases. However, prognosis tends to be poor for advanced or recurrent cases. Therefore, the precise identification of the genes implicated in FTC and the genetic exploration of FTC have emerged as key research areas. A growing body of evidence highlights the association between aberrant gene expression and the initiation and unfavorable prognosis of FTC [5, 6]. According to the TNM staging of thyroid cancer (8th Edition) by the American Joint Committee on Cancer [7], patients aged 55 years and older are classified as stage IVB in the presence of distant metastasis (M1), regardless of tumor size (T) and lymph node infiltration (N).

With the advantages of comprehensive data and large sample sizes, gene chips have become integral to biology and medicine. Recently, the application of bioinformatics and microarray techniques has gained prominence in investigating tumor occurrence and development attributed to gene inactivation or mutation [8]. However, these data, while abundant, lack robust experimental foundations. Consequently, the extraction of accurate and reliable gene information is emerging as a focal point of bioinformatics research. In this study, differentially expressed genes (DEGs) in FTC and normal thyroid tissue microarrays were analyzed through bioinformatics methods and subsequently validated via immunohistochemical (IHC) staining of clinical samples. Furthermore, the correlation between core DEGs and immune cell infiltration in tumor tissues was explored, aiming to identify key DEGs associated with the development and progression of FTC. The findings of this research are anticipated to provide a novel theoretical foundation for the targeted therapy of FTC and related conditions.

Materials and methods

Results

Identification of DEGs

Upon analysis of the GSE27155 dataset, a total of 298 DEGs were identified. Among these, 215 genes were upregulated, whereas 83 were downregulated. The upregulated and downregulated genes are shown in Figure 1 using cluster analysis.

Figure 1:

Heat map illustrating the DEGs between FTC and normal thyroid samples in the GSE27155 dataset. The normal thyroid tissue group is indicated by the grey module, whereas the FTC group is represented by the yellow module. Sample clustering is demonstrated in the upper branch tree, and differential gene clustering is showcased in the left branch tree. The chip sample number is annotated below, with the DEG names enumerated on the right. The color key extends from green to red, indicating the degree of expression enhancement. FTC, follicular thyroid carcinoma; DEG, differentially expressed gene.

GO and KEGG pathway enrichment analyses of DEGs

The DEGs were analyzed using GO enrichment analysis to evaluate their CC, MF, and BP. With regard to CC, the associated DEGs were notably enriched in components such as the extracellular matrix (ECM), extracellular exosome, membrane, and perinuclear region of the cytoplasm. In terms of MF, significant enrichments were observed in functions such as protein binding, peroxidase activity, enzyme binding, and ECM structural constituents. As for BP, DEGs were predominantly involved in processes such as immune response, ECM organization, cell adhesion, angiogenesis, cell migration, positive and negative regulation of cell proliferation, positive and negative regulation of the apoptotic process, positive regulation of ERK1/2 cascade, and cellular response to DNA damage stimulus. In KEGG signaling pathway analysis, the DEGs were primarily implicated in pathways including small cell lung cancer, focal adhesion, insulin signaling pathway, pathways in cancer, protein processing in the endoplasmic reticulum, phosphoinositide-3-kinase–protein kinase B (PI3K–Akt) signaling pathway, proteoglycans in cancer, mitogen-activated protein kinase signaling pathway, Rap1 signaling pathway, erythroblastic oncogene B signaling pathway, and ECM-receptor interaction. These results are illustrated in Figure 2.

Figure 2:

GO and KEGG pathway enrichment analyses of DEGs. (A) Cellular components; (B) molecular functions; (C) distribution of DEGs enriched in various biological processes in FTC; (D) KEGG signaling pathway. The dot sizes represent the count of the enriched DEGs, and the dot color represents the p-value. DEG, differentially expressed gene; GO, gene ontology; KEGG, Kyoto encyclopedia of genes and genomes; FTC, follicular thyroid carcinoma.

PPI construction and molecular module analysis

The PPI network was constructed, comprising 159 network nodes and 367 edges (Figure 3A). Four key modules (subnetworks) were identified from the PPI network (Figure 3B–E). Notably, Module A was enriched in ECM organization, participating in ECM receptor interaction, protein digestion and absorption, PI3K-Akt signaling, and focal adhesion pathways. Seven key network nodes (NID1, COL4A1, COL4A2, COL6A1, FBLN1, LUM, and DCN) were identified in Module A. Module B, located in the mitochondrial outer membrane, contributing to fatty acid biosynthesis, degradation, metabolism, and adipocyte signaling pathways. ACACB, ACSL1, ACSL3, and ACADVL are four vital network nodes that are involved in these processes. Module C engaged in adhesion, leukocyte migration, chemokine signaling, and other pathways. This module primarily consists of 14 key network nodes, including SOX4, CDC42, CCL2, KLF4, and CDH5. Module D was primarily enriched in BP involving the active regulation of cell migration and cell proliferation, participating in Rap1 signaling, EGFR tyrosine kinase inhibitor resistance, actin cytoskeleton regulation, and Ras signaling pathways. This process primarily involves eight key network nodes: TGFA, PRKCZ, MCAM, PRKCI, PIK3R3, FGFR2, PDGFRA, and CXCL12.

Figure 3:

DEGs encode the PPI network and four key modules. (A) The PPI network of DEGs. (B–E) Key modules 1, 2, 3, and 4, respectively. In these figures, circles represent genes and lines indicate the protein interactions among them. In particular, red circles denote upregulated genes, blue circles indicate downregulated genes, and the colors indicate fluctuations in gene expression. DEG, differentially expressed gene; PPI, protein–protein interaction.

Validation of core DEG expression using the TCGA database

The findings revealed a significant reduction in the expression of three key DEGs (DCN, GPC3, and PDGFRA) in FTC tissues compared with normal tissues (p<0.0001) (Figure 4).

Figure 4:

Expression of the three core genes: DCN, GPC3, and PDGFRA (****p<0.0001). DEG, differentially expressed gene; DCN, decorin; GPC3, glypican-3; PDGFRA, platelet-derived growth factor receptor alpha.

Correlation analysis between DCN and tumor pathological stage

The GEPIA database was utilized to analyze the tumor pathological stages in the three core DEGs, revealing a more significant correlation between DCN expression and thyroid cancer stage compared to PDGFRA (p<0.0001) (Figure 5). No significant difference was observed in thyroid cancer stage between GPC3 (p>0.05).

Figure 5:

Association between the three core DEGs expression and tumor stage in thyroid cancer.

Correlation analysis between DCN and immune infiltrating cells

Results from TIMER database analysis indicated a negative correlation (partial cor<0, p<0.01) between DCN expression and tumor purity in thyroid cancer. Moreover, DCN gene expression exhibited significant correlations (partial cor>0, p<0.01) with B cells, CD8⁺ T cells, CD4⁺ T cells, macrophages, neutrophils, and dendritic cells (Figure 6A). The “survival module” analysis of the TIMER database revealed a significant association between low levels of infiltrating CD8⁺ T cells in thyroid cancer and poor overall survival (p<0.05) (Figure 6B).

Figure 6:

Correlation analysis between DCN expression and infiltrating immune cells in thyroid cancer. (A) The relationship between DCN expression and proportions of infiltrating immune cells in THCA. (B) Cumulative survival of infiltrating immune cells in THCA. THCA, thyroid cancer; DCN, decorin.

Correlation analysis between DCN and hotspot signature genes in thyroid cancer

Using TIMER database analysis, DCN was found to be positively correlated with NRAS and KRAS genes with statistical significance (partial. cor>0, p<0.05). Furthermore, DCN exhibited a negative correlation with HRAS and PAX8 genes with statistical significance (partial. cor<0, p<0.05). However, no significant correlations were observed between DCN and BRAF and PPARG or PPAR-γ genes (p>0.05) (Figure 7). These findings provide valuable insights into the relationship between DCN and characteristic hotspot genes in thyroid cancer.

Figure 7:

Correlation analysis between DCN and hotspot signature genes in thyroid cancer.

IHC validation

IHC staining experiments were conducted to select the most differentially expressed DCN among the core DEGs. The staining score of DCN in tumor tissues (3.071 ± 2.493) was significantly lower (p<0.0001) than that in adjacent normal tissues (8.643 ± 2.094), confirming the underexpression of DCN in FTC (Figures 8 and 9). The immunohistochemical images of 28 patients with follicular thyroid carcinoma (FTC) and their corresponding adjacent non-cancerous tissues can be found in the Supplementary Materials.

Figure 8:

Immunohistochemical images of DCN expression in FTC tissues and paired adjacent noncancerous tissues (×200). Scale bar: 100 μm. (A) Negative DCN staining in the FTC tissue. (B) Positive DCN staining in the adjacent noncancerous tissues. FTC, follicular thyroid carcinoma; DCN, decorin.

Figure 9:

Immunohistochemical scoring plots of FTC (28 samples) and adjacent noncancerous tissues (28 samples) (****p<0.0001).

Discussion

The incidence of thyroid carcinoma, particularly DTC, has seen a notable rise in recent years. While surgical intervention has generally proven effective, cases of advanced FTC or recurrence may lead to poor prognoses [12]. In this study, the integration of data mining from FTC and normal thyroid microarrays, utilizing bioinformatics and high-throughput sequencing, identified three significant core DEGs: DCN, GPC3, and PDGFRA, all of which were underexpressed in FTC tumor tissues. Among these genes, DCN, identified as a leucine-rich low-protein proteoglycan primarily found in the mammalian connective tissue within the ECM [13], has multifaceted biological activities. These include the inhibition of collagen fiber formation and the regulation of cell proliferation, migration, and adhesion [13]. While previous studies, such as Arnaldi et al. [14], reported a high percentage of DCN underexpression in various thyroid conditions, its specific role in FTC has not been extensively explored. Therefore, investigating the protein expression profile and developmental roles of DCN in FTC tissues becomes crucial.

In FTC, DCN genes were found to be primarily involved in BPs such as ECM composition, cell aging, negative regulation of protein kinase activity, positive regulation of phosphatidylinositol 3-kinase signaling, and negative regulation of angiogenesis and primarily enriched in proteoglycan signaling pathways in cancer (Table 1). These processes and pathways are closely linked to the development and progression of various malignancies, suggesting that DCN regulates FTC progression by influencing these pathways. Notably, DCN expression is considered crucial for tumor initiation and progression, possibly acting as a tumor suppressor [15]. Goldoni et al. [16] revealed that DCN can effectively inhibit the growth and metastasis of breast cancer cells, proposing it as a drug candidate for targeted therapy. Horvath et al. [17] reported that hepatocarcinoma tissues exhibited low DCN expression, suggesting that DCN inhibits liver cancer growth via cell proliferation inhibition, cell cycle arrest, and increased caspase-3 enzyme activity. Bi et al. [18] reported low DCN expression in colorectal cancer tissues, and increased DCN expression was found to inhibit the growth and metastasis of colorectal cancer cells. These findings align with our observations, highlighting the importance of DCN in patients with FTC.

As the understanding of tumor biology and immunology deepens, researchers increasingly acknowledge the close relationship between tumorigenesis and the tumor microenvironment (TME). The TME, consisting of the ECM, cancer-associated fibroblasts, vascular epithelial cells, and infiltrating immune cells, plays a critical role in cancer development [19]. Immune cell infiltration within solid tumors is recognized as a key factor in TME-associated carcinogenesis [20]. Morana et al. [21] reported that changes in the TME affect apoptosis in tumor cells, promoting tumor cell occurrence and development. In this study, a positive correlation was found between the expression of DCN and the infiltrating abundance of tumor-infiltrating lymphocytes (TILs), including B cells, CD8⁺ T cells, CD4⁺ T cells, macrophages, neutrophils, and dendritic cells, in cancer tissues from patients with thyroid cancer. Thus, low DCN expression in thyroid cancer tissues is associated with reduced TIL infiltration, potentially altering the TME and promoting tumorigenesis and progression. Moreover, DCN may affect the survival time of patients with thyroid cancer via the regulation of CD8⁺ T cell infiltration, indicating its potential role as an immune regulator in thyroid cancer.

FTC poses challenges in clinical diagnosis as accurate identification via fine needle aspiration biopsy or intraoperative frozen section pathology is elusive. Confirmation typically relies on postoperative paraffin section pathology. Consequently, the implementation of molecular diagnostic technologies before clinical intervention has become a central focus in FTC research. Currently, genes such as BRAF, RAS, and PAX8-PPARG, recognized as characteristic hotspot genes for thyroid cancer, are widely employed in clinical molecular testing to differentiate between benign and malignant thyroid nodules. BRAF, extensively studied in thyroid fine needle aspiration cytopathology, is frequently mutated in PTC (in up to 80 % of cases) but less frequently mutated in FTC [22]. This study corroborates previous findings by revealing no correlation between DCN and BRAF, consistent with the results reported by Mayson et al. [22]. The RAS family genes are a class of proto-oncogenes that are very conservative during biological evolution, including KRAS, HRAS, and NRAS. They are closely related in function and structure, and both encode G proteins [23]. The mutation rates of RAS are second only to BRAF and are most common in FTC, followed by follicular type PTC, and less frequently in classical type PTC [24]. This study reveals a positive correlation between the DCN gene and NRAS and KRAS genes, as well as a negative correlation with HRA, indicating the involvement of DCN in the BP of RAS mutation in FTC. Rearrangements of the PPARG gene, particularly PAX8-PPARG, are common in follicular lesions, including FTC [25]. This study reveals an inverse correlation between DCN and PAX8 but no correlation with PPARG, suggesting a correlation between DCN expression and PAX8-PPARG rearrangements.

We retrospectively validated DCN expression in FTC tumor tissues and paracarcinoma tissues using IHC analysis, confirming its low expression in cancer tissues, consistent with DCN mRNA expression findings from Arnaldi et al. [14]. This validation enhances the reliability of the low expression data of DCN in FTC obtained through mining. Furthermore, they verified the authenticity and accuracy of the bioinformatics analysis results, highlighting the significant role of the DCN gene in FTC progression. A significant association between DCN and thyroid cancer stage was identified, suggesting its potential as a tumor suppressor gene in FTC pathogenesis. These findings align with existing literature reports [14], [15], [16], [17], [18].

Altogether, the downregulation of DCN expression may be associated with the development, progression, and immune infiltration of FTC. The study findings establish a theoretical basis for understanding the molecular mechanisms underlying FTC pathogenesis and the translational application of immunotherapy. Additionally, the DCN gene emerges as a potential biomarker and therapeutic target for FTC in clinical practice. Nevertheless, the study acknowledges limitations in not elucidating the specific mechanism and related signaling pathways of DCN in FTC, calling for further in-depth investigation.