Introduction

Previous studies have elucidated the pivotal role of cartilage oligomeric matrix protein (COMP), also known as thrombospondin-5 (TSP-5), in various cancers. COMP's presence in multiple cell types and its contributions to maintaining extracellular matrix (ECM) integrity have been established. Notably, its functions encompass stabilizing ECM protein connections and enhancing tissue mechanical strength. COMP levels are elevated in a variety of musculoskeletal disorders1. High COMP levels have also been associated to breast cancer, hepatocellular carcinoma, prostate cancer, and colon cancer. In these cancers, COMP overexpression was associated with increased tumor growth, cancer metastasis, cancer recurrence, and overall shorter survival2,3,4,5. Against this backdrop, the prognostic significance of COMP in intrahepatic cholangiocarcinoma (iCCA) remains unexplored.

After hepatocellular carcinoma (HCC), cholangiocarcinoma (CCA) is the second most prevalent primary hepatic malignancy, accounting for around 15% of all primary liver tumors and 3% of all gastrointestinal cancers6,7. CCA is a diverse collection of malignant cancers that arise in various parts of the biliary tree. CCAs are divided into three types: intrahepatic (iCCA), perihilar (pCCA), and distal CCA (dCCA), each with its own etiologies, risk factors, prognosis, and clinical and therapies. Together, iCCA and pCCA account for more than 90% of all CCAs in the world8. CCA is a rare cancer, but its incidence (0.3–6 per 100,000 people per year) and mortality (1–6 per 100,000 people per year, globally, excluding specific regions with incidence > 6 per 100,000 people, such as South Korea, China, and Thailand) have been rising in recent decades, posing a global health problem9,10,11. Over the past few decades, the reported age-standardized incidence for iCCA has been steadily increasing in most locations worldwide, whereas the age-standardized incidence for dCCA has been decreasing12. Curative surgery remained the standard treatment for early CCA. Despite the improvement in CCA pathogenesis, diagnosis, and treatments over the last decade, patient prognosis has remained unchanged, with 5-year survival rates of 7–20 percent and tumor recurrence rates were as high as 48–56% after resection remaining unsatisfactory13,14,15,16,17. Several clinical markers, including as T stage, lymph node metastases, and histological grade, can be used to identify high-risk patients16. However, to establish further treatment plans after curative surgery, genomics-based prognostic biomarkers were warranted.

Given the rising incidence and mortality of iCCA globally, coupled with the challenges in improving patient prognosis, an investigation into COMP's role in iCCA assumes importance. Therefore, the primary aim of this study is to assess COMP expression levels in iCCA patients and determine its potential as a prognostic marker. In summary, this study investigates the expression of COMP in intrahepatic cholangiocarcinoma (iCCA) and its implications for prognostic evaluation. By shedding light on COMP's potential as a prognostic marker, our work seeks to contribute valuable insights into enhancing the management and treatment outcomes of iCCA patients.

Results

Upregulation of COMP gene links to extracellular matrix structure in the CCA transcriptome

For data mining, a published CCA transcriptome dataset (GSE26566) was used, which includes 104 patients who had radical surgery. We found 9 probes that covered 9 transcripts related to structural elements of the extracellular matrix (GO:0,005,201) (Table 1 and Fig. 1). When compared to neighboring liver tissue, COMP gene (ILMN 1,677,636) was elevated by up to 2.3107-fold log ratios (p < 0.0001) in CCA, as shown in Table 1. When compared to the normal intrahepatic bile duct, COMP was also enhanced in CCA, with 2.4308-fold log ratios (p < 0.0001). Thus, COMP was chosen for further analysis.

Table 1 Summary of the alterations of gene associated with extracellular matrix structural constituent (GO:0005201) in cholangiocarcinoma (GSE26566).
Full size table
Figure 1
figure 1

A published transcriptome dataset of intrahepatic cholangiocarcinoma (GSE26566) from GEO database showed gene expression associated with extracellular matrix structure constituent (GO:0005201). COMP is the most upregulated genes in cholangiocarcinoma compared to surrounding liver and normal biliary epithelium.

Full size image

Clinicopathological features of the iCCA cohorts

Our study includes total 182 iCCA patients who had undergone radical surgery (Table 2). Males account for a slightly larger proportion (59.3%). Most patients (59%) were younger than 65 years old. 72 patients (40%) were chronic hepatitis B carrier, 29 (16%) patients were chronic hepatitis C carrier. Intrahepatic lithiasis was identified in 80 patients (44%). R0 resection was achieved in 163 patients (90%). 87 patients (48%) had T1 lesions, 61 patients (34%) had T2 lesions, while 34 patients (18%) had T3 lesions. Regarding histological findings, 105 patients (58%) had large duct type cholangiocarcinoma, while 77 patients (42%) had small duct type cholangiocarcinoma. 61 patients (34%) had well differentiated tumor, 66 patients (36%) had moderate differentiated tumor, and 55 patients (30%) had poorly differentiated tumor.

Table 2 Correlations between COMP expression and other important clinicopathological parameters in primary localized iCCA analyzed by Pearson's chi-square test.
Full size table

Correlations between COMP expression and pathological features in iCCA

The association between COMP expression and various clinicopathological features was assessed using Pearson’s chi-square test. As determined using immunohistochemistry, poorly differentiated and higher T stage cholangiocarcinoma had greater COMP immunoreactivity than well differentiated and lower T stage cholangiocarcinoma (Fig. 2). Table 2 summarized the relationships between COMP expression levels and clinicopathological parameters in iCCA cases. COMP overexpression was associated with R1 resection (p = 0.008), advanced T stage (p < 0.001), large duct type (p = 0.004) and poor differentiated histology (p = 0.002).

Figure 2
figure 2

Immunohistochemistry staining showed lower COMP expression in well-differentiated and pT1 (A, B, C, D) iCCA compared to poorly-differentiated and pT3 stage (E, F, G, H) iCCA. (Magnification: A, C, E, G, 200X; B, D, F, H, 400X).

Full size image

Prognostic significance of COMP expression

The predictive effects of COMP expression on cancer metastasis and patient survival in iCCA were assessed using univariate and multivariate analyses. (Table 3). In univariate analysis, female gender, R1 resection, advanced T stage and COMP overexpression were significantly associated with worse disease-specific survival (DSS) and overall survival (OS) (Table 3). To further evaluate the recurrence pattern, R1 resection, advanced T stage, large duct type CCA, well differentiated CCA and high COMP expression were revealed to be significant predictive factors for shorter local recurrence-free survival (LRFS). While R1 resection, advanced T stage and high COMP expression were significantly linked to worse metastases-free survival (MFS) (Table 4). Multivariate Cox Regression analysis revealed that R1 resection, advanced T stage and high COMP expression were significant indicators for worse DSS and OS (Table 3). Only R1 section and high COMP expression were associated with shorter LRFS, while R1 resection, advanced T stage and high COMP expression were significantly associated with poor MFS (Table 4). COMP overexpression was found to be substantially linked with poor DFS (hazard ratio [HR], 3.651; 95% confidence interval [CI], 1.675–7.958; p = 0.001), poor OS (HR, 1.827; 95% CI, 1.087–3.070; p = 0.023), poor LRFS (HR, 4.077; 95% CI, 2.350–7.072; p < 0.001), and poor MFS (HR, 3.718; 95% CI, 2.026–6.823; p < 0.001) in multivariate analysis (Tables 3 and 4).

Table 3 Univariate log-rank and multivariate analyses for overall and disease-specific survivals in primary localized iCCA.
Full size table
Table 4 Univariate log-rank and multivariate analyses for local recurrence-free and metastasis-free survivals in primary localized iCCA.
Full size table

Survival analysis in iCCA

Kaplan–Meier analysis showed COMP overexpression was correlated with poorer overall survival (Fig. 3A; p = 0.0001), DSS (Fig. 3B; p < 0.0001), LRFS (Fig. 3C; p < 0.0001) and MFS (Fig. 3D; p < 0.0001) in iCCA patients.

Figure 3
figure 3

Kaplan–Meier analysis showed COMP overexpression associated with (A) worse overall survival (B) worse disease-specific survival (C) worse local recurrence free survival (D) worse metastasis free survival in iCCA patient.

Full size image

COMP gene function prediction

To recognize the potential functions of COMP in IHCC, we downloaded the top two hundred differentially expressed transcripts showing positive connections (Supplementary Table 1) or negative connections (Supplementary Table 2) with COMP from the cholangiocarcinoma dataset (TCGA, n = 51). Later, employing the Gene Ontology (GO) classification system, these transcripts were utilized to prognosticate the functions of COMP. The results revealed that, in respect to biological processes (Fig. 4A), the most significant term correlated with COMP upregulation was positive regulation of transcytosis (GO: 1,904,300, fold enrichment: > 100), and the platelet-activating factor receptor (PTAFR) gene was identified. In view of molecular functions (Fig. 4B), the most remarkable term correlated with COMP upregulation was platelet-derived growth factor binding (GO: 0,048,407, fold enrichment: 55.32) that includes the platelet-derived growth factor subunit B (PDGFB), platelet-derived growth factor receptor beta (PDGFRB), collagen type I alpha 1 chain (COL1A1), COL1A2, COL3A1, and COL5A1 genes. Furthermore, with regard to cellular components (Fig. 4C), the most important terms correlated with COMP upregulation were short-chain collagen trimer (GO: 0,005,598, fold enrichment: > 100), collagen type I trimer (GO: 0,005,584, fold enrichment: > 100), and collagen sheet (GO: 0,098,646, fold enrichment: > 100).

Figure 4
figure 4

The outstanding GO terms enriched in COMP upregulation. Employing the GO classification system based on (A) biological processes, (B) molecular functions, and (C) cellular components, the top 20 GO terms presenting positive correlations with COMP were displayed and ordered by fold enrichment.

Full size image

Discussion

In contrast to HCCs, mass-forming iCCAs are characterized by a significant desmoplastic and hypovascularized tumor stroma, which is frequently the predominant histological characteristic of the tumor18. Several research have studied not only the morphological differences, but also the molecular distinctions between iCCA, pCCA, and dCCA19,20,21. Nakamura et al. demonstrated a variation in tumor anatomical location, indicating IDH, EPHA2 and BAP1 mutations and FGFR2 fusions in iCCA, whereas extrahepatic tumors preferentially show PRKACA and PRKACB fusions as well as mutations in ELF3 and ARID1B. Based on these findings, cholangiocarcinoma in different anatomical locations should be treated and examined differently20. Furthermore, cholangiocarcinoma's highly desmoplastic nature, substantial support from a rich tumor microenvironment, and significant genetic heterogeneity all contribute to its treatment resistance22.

COMP is a large pentameric cartilage protein and is found in articular cartilage, ligaments, and tendons. The protein can also be found in the skin, breast tissues, and liver tissues3,23,24. Although the function of COMP glycoprotein in a number of connective tissue illnesses has been well documented25, its function in carcinogenesis is still unknown. Evidence from recent studies has shown that COMP overexpression is associated with a poor prognosis and metastases in breast, prostate, thyroid, colon, and hepatocellular carcinoma2,3,5,26,27,28. In our previous study, we found COMP overexpression associated with poor survival and tumor invasiveness in urothelial carcinoma29.

As the extracellular matrix degrades with continued liver damage and fibrotic scar tissue builds up, the liver undergoes remodeling, which eventually results in fibrosis, cirrhosis, and ultimately hepatocellular carcinoma (HCC). Norman et el. demonstrated serum COMP could be an early marker of fibrosis, and that increased serum COMP levels could reflect the degree of cartilage breakdown during liver destruction and re- modeling, and also associated with HCC development2. Liver cirrhosis and viral hepatitis C and B have been recognized as risk factors for cholangiocarcinoma, especially intrahepatic disease30,31,32,33. The following associations were found in a meta-analysis34 of several case–control studies on the risk factors for intrahepatic cholangiocarcinoma: cirrhosis had a combined odds ratio (OR) of 22·92 (95% CI 18·24–28·79), hepatitis C of 4·84 (2·41–9·71), and hepatitis B of 5·10 (2·91–8·95). Based on this, we think that the COMP may also affect the prognosis of iCAA due to the shared etiology of HCC and iCCA. Moreover, cholangiocarcinoma is usually characterized by a prominent desmoplastic and hypovascularized stroma. There is now increasing evidence to suggest that the desmoplastic reaction, marked by a dramatic accumulation of α-smooth muscle actin positive cancer-associated fibroblasts (α-SMA + CAFs) with increased production of extracellular matrix proteins18. Based on these, we use the transcriptome dataset to assess the genes related to extracellular matrix structure constituent (GO:0005201). When compared to neighboring liver tissue, COMP gene was significantly elevated in CCA among these genes. Thus, COMP was chosen as a candidate for further survival analysis. To our knowledge, this is the first study to demonstrate COMP overexpression associated with poor survival in iCCA.

The Notch, WNT, and transforming growth factor (TGF) signaling pathways, among others, are highly active in iCCA. It is known that the Notch pathway has a role in biliary repair, growth, fibrosis, and stem cell niche maintenance. NOTCH3 overexpression was linked to the development and progression of iCCA, promoting cell survival through PI3K-AKT signaling12,35,36. Interestingly, Konstantinos et al. demonstrated that COMP regulates the cancer stem cell population through increasing the interaction between Notch3 and Jagged1, leading to increased activation of Notch3 signaling37.

In the other way, the epithelial to mesenchymal transition (EMT) is a critical biological step in the migration and invasion of malignant tumor cells38. It gives epithelial cancer cells the ability to develop mesenchymal characteristics with invasive capabilities, which promote colonization of metastatic sites39. The TGF-dependent pathway, whose signature has been discovered in iCCA stroma, is the prototypical inducer of EMT. In CCA, TGF either directly induces EMT or collaborates with other key EMT inducer pathways like EGFR40,41,42. Notably, transcription factors (EMT-TFs) that control the expression of epithelial and mesenchymal genes, including as the SNAIL, ZEB, and TWIST families, orchestrate EMT. Regardless of anatomical location, CCAs express EMT-TFs, which are linked to a poor prognosis12,43,44. Interestingly, COMP can stimulate tumor EMT, although the mechanism is unknown27,45. COMP glycoprotein has been demonstrated to be co-expressed with many EMT genes, and a clear association between high COMP expression and poor colon cancer survival has also been reported45,46. The evidences to support the link of NOTCH3 pathway, EMT pathway between COMP and iCCA are ambiguous initially but gradually apparent. However, the exact mechanism by which COMP acts as a poor prognostic factor in iCCA remains unclear and need further exploration.

Recently, microRNAs (miRs) and extracellular vesicles (EVs) have gained attention as promising non-invasive biomarkers for diagnosis and prognosis in cholangiocarcinoma47. MiRs can be secreted extracellularly or encapsulated in EVs, thereby taking part in intercellular communication48. Transcytosis, a type of transcellular mechanism for crossing of EVs through the interior of a cell, has also been implicated in various solid tumors49. Interestingly, the results of our bioinformatic analysis revealed that, in terms of biological processes, the most significant term correlated with COMP upregulation was positive regulation of transcytosis (Fig. 4A) that contains the PTAFR gene. Activation of the platelet-activating factor (PAF)/PTAFR pathways has also been suggested to cause hepatocellular carcinoma cell migration and invasion50. However, whether COMP may promote IHCC progression through transcytosis and PTAFR needs further investigation. Moreover, in terms of molecular functions, the most remarkable term correlated with COMP upregulation was platelet-derived growth factor binding (Fig. 4B) that includes the PDGFB, PDGFRB, COL1A1, COL1A2, COL3A1, and COL5A1 genes. The critical role of the PDGFB/PDGFRB axis in angiogenesis has been demonstrated in tumors51. PDGFD can also bind PDGFRB and activate cancer-associated fibroblasts (CAFs) that play crucial roles in modulating cholangiocarcinoma development52. The COL1A1, COL1A2, COL3A1, and COL5A1 genes are also identified as CAF signature in cholangiocarcinoma53,54. Accordingly, the associations among the expression levels of COMP, platelet-derived growth factors, and CAF-derived collagen in IHCC development are quite interesting and deserve further examination.

Our research has certain limitations. Firstly, it is a retrospective study conducted at a single institution and lack of experimental validation. Secondly, the exact molecular mechanism underlying disease progression and adverse outcomes in COMP-overexpressing IHCC remains unclear. Thirdly, there is currently no standardized immunostaining and scoring scheme for assessing COMP expression. Due to the lack of agreed staining standards, it is difficult to reach a consensus in this type of research. Fourthly, The GEO transcriptome dataset we use is from cholangiocarcinoma patients rather than iCCA patients, which may also cause some limitations in this study. Lastly, to validate our findings, prospective multicenter studies are required.

In conclusion, COMP overexpression was associated with worsening clinical-pathological characteristics. COMP is also related with poorer survival in iCCA, supporting its function as a biomarker for iCCA prognosis. This is the first study to our knowledge that clarifies the role of COMP in iCCA. To completely comprehend the mechanism and apply these findings to clinical practice, additional research is required.

Methods

Data mining of the gene expression omnibus (GEO) dataset

A transcriptome dataset (GSE26566) containing 104 iCCA patients who received curative surgery was obtained from the NCBI Gene Expression Omnibus (GEO) database. Without pre-selection, all probe sets were used. The raw data was then transferred into the Nexus Expression 3 software, which was used to calculate gene expression levels. Comparative analyses were conducted to detect the significantly differently expressed genes connected to extracellular matrix structural constituents (GO:0,005,201) by comparing CCA part vs. normal surrounding liver and CCA part vs. normal intrahepatic duct. After analysis, differentially regulated genes in CCA part vs. normal surrounding liver and CCA part vs. normal intrahepatic duct were discovered (P < 0.0001 and log ratio > 2) (Fig. 1 and Table 1).

Study population

Between 1990 and 2010, 182 patients with iCCA who received curative surgery were enrolled at Chi Mei Medical Center. The presence of lymph node involvement or distant metastasis was ruled out to guarantee curability. Only individuals with T1-3N0M0 disease were included. Two pathologists investigated tumor samples to rule out the possibility of other malignancies arising from the biliary system. In this study, we used anonymous patient sample information from biobank as approved by IRB. As a rule, inform consent has been sign by every patient before their sample/information collected into biobank. The study had been approved by the Institutional Review Board (IRB) of Chi Mei Medical Center with the approval number of 09,912,003. All research was performed in accordance with relevant guidelines/regulations. We gathered patients' retrospective demographic and clinical data, including pathological characteristics, oncological survival follow-up, and cause of mortality. Patients with acute blood diseases, abnormal bone marrow, concurrent cancer, or insufficient clinical data were excluded from the research. The tumor stage was assessed using the eighth edition of American Joint Committee on Cancer (AJCC) Tumor, Node, Metastasis (TNM) system developed in 2017. This study followed the REMARK (Reporting Recommendations for Tumor Marker Prognostic Studies) guidelines.

Immunohistochemistry and scoring

Specimens were prepared according to protocol. The primary COMP antibody was incubated for 1 h on the sections (Clone: EPR22857-38, Abcam). Then the primary antibody was diluted 1:200 in TBS with 1% BSA. Antibodies were then detected using a DAKO ChemMate EnVision Kit (K5001, Carpinteria, CA, USA). Cell blocks from cell lines known to express COMP were used as positive controls. Sections were processed without the primary anti-COMP antibody as negative controls. Two pathologists used the following equation to calculate the H-score to estimate COMP immunoreactivity: H-score = SPi (i + 1), where Pi is the percentage of stained tumor cells in various intensities ranging from 0 to 100%, and I is the degree of staining (0 to 3 +). If there were any scoring disagreements, the two pathologists assessed the slides at the same time and agreed on an H-score. The immunostaining was categorized into low and high expression levels based on the median H-score as previously mentioned55.

Functional gene annotation

To recognize the potential functions of COMP in intrahepatic cholangiocarcinoma, we appraised the relations between the mRNA expression levels of COMP and its coexpressed genes from the cholangiocarcinoma dataset (Firehose Legacy, n = 51) in The Cancer Genome Atlas (TCGA) database. Later, the top 200 differentially expressed transcripts showing positive connections or negative connections with COMP were downloaded. These transcripts were then annotated by the Gene Ontology (GO) classification system and ranked by fold enrichment.

Statistical analysis

The connection between COMP expression and other clinicopathological characteristics was assessed using Pearson's chi-square test. We measured two outcomes: metastasis-free survival (MFS) and disease-specific survival (DSS). Using univariate and multivariate analysis, relevant COMP expression and clinicopathological variables were discovered as predictors of DSS (measured from curative surgery to the time of cancer mortality) and MFS (measured from curative surgery to the first metastasis). The survival curves were created using the Kaplan–Meier method and a log-rank test. All relevant parameters from the univariate analysis were integrated into the multivariate Cox proportional hazards model to find the independent variables. IBM's SPSS Statistics V.17.0 software (Armonk, NY, USA) was utilized for statistical analysis. The statistical significance criterion was chosen at P < 0.05.