Development and validation of nomogram including mutations in angiogenesis-related genes as risk factors for HCC patients treated with TACE.

INTRODUCTION
Hepatocellular carcinoma (HCC) is the fourth most common cancer and a leading cause of cancer-related deaths globally [1]. Liver resection and transplantation are the most effective surgical treatments for patients with early HCC [2]. However, most patients with HCC are diagnosed at intermediate or advanced stages [3]. Based on the Barcelona Clinic Liver Cancer (BCLC) staging system, transarterial chemoembolization (TACE) is recommended as the first-line therapy for intermediate-stage HCC [4,5]. TACE is also used after the recommended methods fail to achieve satisfactory results [6,7]. However, effective local therapy can be challenging in patients with HCC, and patients enter a state of TACE failure/refractoriness [5,8,9].
Despite TACE achieving an objective tumor response rate of over 60 %, approximately 40 % of patients do not respond to the therapy [10,11]. Patient responses of TACE exhibit significant diversity owing to the variability associated with intermediate-stage HCC and extensive application of TACE beyond authorized settings [12]. In patients with HCC, TACE failure may be related to tumor angiogenesis in residual disease [13]. Angiogenesis is a biological process that generates new vessels from the endothelium of existing vasculature for tissue growth, wound healing, and pregnancy [14], and can supply oxygen and nutrients to the whole body. However, angiogenesis could also nourish rapid growth and metastasis of tumor cells [14]. Folkman speculated that the early stages of tumor angiogenesis occur coextensively with tumor development. This hypothesis is supported by the observation that fast-developing tumors are substantially vascularized, whereas dormant tumors are not [15]. When the ratio of proangiogenic to antiangiogenic signals increases, endothelial cells proliferate and move to support pathological angiogenesis, making the tumor more aggressive [14].
HCC, a solid tumor rich in blood vessels, exhibits significant blood vessel growth and abnormalities, with tumor angiogenesis contributing to its growth, invasion, and metastasis. In the absence of angiogenesis, HCC tumors do not grow more than 1–2 mm in size [16]. Of the numerous angiogenesis pathways, the vascular endothelial growth factor/vascular endothelial growth factor receptor (VEGF/VEGFR) signaling pathway has been verified as the target of HCC precision medicine, and Sorafenib, by targeting this pathway to inhibit angiogenesis, achieves treatment for advanced HCC [17]. In addition, several other genes have been suggested to be biomarkers or modulators of angiogenesis. Certain genes, such as COMMD3 and BMP9, promote angiogenesis, which is essential for HCC invasion and metastasis [16,18]. In contrast, some angiogenesis-related genes, including EYA4, suppress angiogenesis and metastasis in HCC [19]. However, the relationship between the angiogenesis-related genes and TACE failure or refractoriness remains unclear.
Whole-exome sequencing (WES) enables the swift and effective identification of potential therapeutic targets, and patient genetic traits [[20], [21], [22], [23]]. In this study, we conducted a retrospective analysis using WES data and the clinical information of patients with HCC collected during our clinical work. We aimed to demonstrate the relationship between angiogenesis and the exome characteristics of patients who experience TACE failure or refractoriness and to construct a predictive model for HCC patients undergoing TACE.

Materials and methods

Patients
A total of 165 patients diagnosed with intermediate- and advanced-stage HCC and admitted to The First Affiliated Hospital, Sun Yat-sen University, Guangzhou, China, between January 2013 to December 2023 were recruited. All subjects were followed up via telephone for three years to obtain prognostic information. Following informed consent, we collected 5 mL peripheral blood samples and surgically resected tumor tissues for WES analysis. Moreover, we collected demographic characteristics and blood test results using an electronic medical record system.
Patients meeting the following criteria were enrolled in the study: (1) pathologically confirmed HCC and histological biopsy performed within two weeks after computed tomography (CT) examination; (2) performed multi-phase enhanced-contrast CT scan and classified as stage B or C according to the BCLC staging system [24]; (3) Child-Pugh class A or B; (4) Eastern Cooperative Oncology Group (ECOG) performance status (PS) of 0 or 1; and (5) underwent TACE as the first treatment. In addition, we evaluated the efficacy of TACE treatment based on the expert consensus issued by the Clinical Guidelines Committee of the Interventional Physicians Branch of the Chinese Medical Doctor Association.
This study was reviewed and approved by the Institutional Ethics Review Board and the Hospital National Cancer Center of The First Affiliated Hospital, Sun Yat-sen University (IRB No.[2018] 43).

WES and data processing
All samples were stored at −80°C, and DNA was extracted using the Maxwell RSC Blood DNA Kit (Promega, USA). Then, the exome DNA was captured using the SureSelect Human All Exome Kit V5 (Agilent Technologies, Santa Clara, CA, USA). The exome shotgun library was sequenced on an Illumina NovaSeq-6000 sequencer (Illumina, San Diego, CA, USA) for double-ended WES with a length of 150 bp. Furthermore, the default parameters of CAVSAVR (Illumina, San Diego, CA, USA) were used for image analysis and base identification to filter out low-quality reads and obtain sequencing files with high-quality reads.
Tumor somatic mutation analysis of whole-exome sequencing data was performed using GATK software, and reads were aligned to the human hg38 reference genome. To identify mutations, we compared the sequencing data from a tumor sample with a paired blood sample from the same patient, filtering out germline mutations to get a list of somatic mutations, such as single nucleotide variation and short insertion and deletion somatic mutations.

TACE procedure
TACE was performed by two experienced interventional radiologists (W.F. and J.L.). Hepatic angiography was performed by placing a 5-F catheter (Terumo, Tokyo, Japan) or 2.7-F microcatheter (Renegade Hi-Flo Straight, Boston Scientific, Natick, MA, USA; Progreat, Terumo) as selectively as possible into the tumor-feeding arteries. Next, an emulsion of 5–20 mL of lipiodol (Lipiodol, Guerbet, Aulnay-Sous-Bois, France) or drug-loaded microspheres (DC-Bead [DCB], SciClone, Shanghai, China) and 20–80 mg of epirubicin (Pharmorubicin, Pfizer, New York, NY, USA) were injected. Subsequently, 350–560 μm absorbable gelatin sponge particles (Gelfoam, Hangzhou Pharmaceutical, Hangzhou, China) were administered into tumor-feeding vessels. The particle size of the DCB microspheres was selected according to the angiographic blood flow velocity. Following embolization, angiography of the feeding artery was performed to assess the degree of vascular occlusion. The use of the DCB microspheres, lipiodol, or absorbable gelatin sponge particles was terminated if reflux occurred.

Evaluation of TACE response
Under the modified response evaluation criteria in solid tumors (mRECIST) standard, according to the actual situation of the patient, we can evaluate the patient's objective response using computed tomography (CT) or magnetic resonance imaging (MRI) performed 1–3 months after the first TACE procedure [25]. Complete Response (CR) is defined as the total absence of arterial enhancement in all intrahepatic target lesions, indicating complete necrosis of treated tumors. Partial Response (PR) requires a ≥ 30 % reduction in the sum of diameters of viable (enhancing) tumor components compared to baseline measurements. Progressive Disease (PD) is characterized by either a ≥ 20 % increase in the sum of viable lesion diameters or the emergence of new enhancing intrahepatic lesions, while Stable Disease (SD) encompasses changes that neither meet PR nor PD thresholds. Patients achieving CR/PR are classified as remission, whereas SD/PD indicates non-remission.. In this trial, new intrahepatic lesions were not regarded as progressive disease as they are indicative of the natural tumor biology of HCC and do not imply treatment failure or progression to the next line of treatment [26]. Progression-free survival (PFS) was defined as the time from baseline to tumor progression, and overall survival (OS) was determined at the time of death or final follow-up.

Differential mutation gene identification and genetic risk scores (GRS) calculation
Chi-square tests were used to analyze the differences in gene mutations with a frequency greater than 5 % between the remission and non-remission groups, and the intersection with the angiogenesis pathway was performed to screen the differential genes further. Multifactor Cox regression was used to construct GRS for these differential genes. Moreover, all samples were classified into high- and low-risk groups based on the median GRS, and the prognoses of the remission and the non-remission groups using Kaplan–Meier analysis.

Development and validation of a nomogram
The dataset was randomly divided into training and validation sets in a ratio of 7:3. In the training set, univariate Cox analysis was used for preliminary screening of angiogenesis-related genes, followed by further dimensional reduction using the least absolute shrinkage and selection operator (LASSO) regression model. Independent risk factors affecting the prognosis of HCC patients treated with TACE were screened using stepwise Cox regression, and a prognostic model and nomogram were established. The Receiver Operating Characteristic curve (ROC), calibration curve, and decision curve analysis (DCA) were constructed for the training and validation sets to evaluate the accuracy and stability of the prediction model.

Statistical analysis
Statistical methods such as differential mutation gene screening and model construction are described above. All statistical analyses were performed using R Project for Statistical Computing (R software, version 4.0.3; University of Auckland, New Zealand). When comparing clinical information, if the quantitative data conformed to a normal distribution, a two-sample independent t-test was used; otherwise, a nonparametric Mann-Whitney test was used. Qualitative data were expressed as percentages (%) and compared using the chi-square test or Fisher's exact probability method.
All statistical tests were two-tailed, and P < 0.05 indicated a significant difference.

Results

Demographic characteristics
A total of 165 patients were included in the study; their demographic characteristics are shown in Supplementary Table 1. These patients comprised 164 males and 1 female, with the majority being over 50 years old. A total of 162 patients had a history of chronic hepatitis B and 55 had cirrhosis. According to the BCLC staging system, 54 patients were classified as stage B and 111 were classified as stage C. More than half of the patients had a maximum lesion diameter <8.5 cm.

Overview of somatic genes mutations
Based on WES analysis, all 165 patients harbored genetic alterations (Supplementary Figure 1), with missense mutations (Supplementary Figure 1A) and single nucleotide polymorphisms (SNP, Supplementary Figure 1B) accounting for the vast majority. After processing the microarray data, patients with the largest mutation load had 3967 gene alterations (Supplementary Figure 1D). TP53 was the most frequently altered gene with a mutation rate of 56 % (Supplementary Figure 1F). Gene mutations in 10 classical carcinogenic signaling pathways known in TCGA were analyzed in 165 patients; 72.73 % of patients had at least one mutation in the Notch signaling pathway, which was the pathway with the highest frequency. Additionally, at least half of the patients harbored at least one mutation in the RTK-Ras, Wnt, or TP53 pathways (Supplementary Figure 2).
Further, we compared our WES data (WES_HCC) with the TMB of 33 cancer species in the TCGA database. As shown in Supplementary Figure 3, the TMB of patients in this study was higher than that of the HCC cohort (LIHC) in TCGA, and the mutation load was between those of bladder urothelial cancer (BLCA) and lung adenocarcinoma (LUAD).

Analysis of differentially mutated genes between the remission group and the non-remission group
To explore the relationship between gene mutations and TACE efficacy, we analyzed differences in remission status and gene mutations with frequencies greater than 5 % in the subjects.
As shown in Supplementary Table 2, differences in the 494 mutated genes between the two groups were identified. The top 10 genes with the greatest differences are shown in Fig. 1A. Among them, missense mutations in ZNF717 mainly occurred in patients in TACE remission, and mutations in other genes were mainly observed in the other group.

Mutations in angiogenesis-related genes were associated with poor prognosis in HCC patients treated with TACE
We collected 97 angiogenesis genes from GeneCards (https://pathcards.genecards.org/). These 494 genes intersected with angiogenesis genes to obtain five common genes (Fig. 1B). Furthermore, we performed multivariate Cox regression analysis for these five genes and identified three genes (FGFR4, MST1R, and RECK) to construct the GRS (Table 1; GRS=FGFR4×10.58 + MST1R×6.09 + RECK×2.24 (If the gene is mutated, the value is 1. Otherwise, the value is 0)).
Based on the median value of the GRS, the participants were classified into high-risk (n = 14) or low-risk (n = 151) groups. Survival curve analysis showed that the high-risk group GRS had a shorter median survival time and worse prognosis in HCC patients treated with TACE (Fig. 2, P < 0.001, HR=6.094, 95 % CI=3.116–11.920). This suggests that mutations in angiogenesis-related genes affect the efficacy of TACE.

Construction and validation of the nomogram
The univariate Cox-LASSO-stepwise Cox regression results demonstrated that alpha-fetoprotein (AFP), alanine aminotransferase (ALT), aspartate aminotransferase (AST), platelet-to-lymphocyte ratio (PLR), PD-L1 immune drug treatment, and GRS were independent factors affecting the prognosis of patients with HCC treated with TACE (Fig. 3, Supplementary Table 3, and Table 2). Based on these independent prognostic factors, we constructed a prognostic model and nomogram (Fig. 4).
The receiver operating characteristic (ROC) curve, calibration curve, and DCA were used to evaluate the effectiveness of the model. At three, five, and seven months after TACE treatment, the AUC values in the training and validation sets were 0.91, 0.96, 0.97, and 0.84, 0.92, and 0.87, respectively (Fig. 5A). The predicted probabilities of the model were almost identical to the actual probabilities in the two datasets (Fig. 5B). Meanwhile, when the threshold probability ranged from 0.3 to 0.9, positive clinical benefits were observed in both the training and validation sets (Fig. 5C–D). Together, these results show that the nomogram based on the GRS of angiogenesis-related genes can be used as a quantitative analytical tool to predict the prognosis of patients with HCC treated with TACE.

Discussion
In this study, WES analysis was performed on 165 patients with HCC treated with TACE, and three angiogenesis-related gene mutations were found to be associated with patient prognosis. Furthermore, through univariate Cox-LASSO-stepwise Cox regression analysis, we identified that AFP, ALT, AST, PLR, PD-L1 immune drug treatment, and GRS served as independent prognostic factors and developed a nomogram for prognostic prediction within this patient population. Moreover, the nomogram demonstrated excellent performance in terms of discrimination, calibration, and clinical benefits during internal validation.
Predicting treatment outcomes in patients receiving TACE remains a significant clinical challenge. Although TACE is the standard treatment for intermediate-stage HCC, patient responses vary considerably, highlighting the need for reliable prognostic tools. The development, progression, and metastasis of solid tumors are closely related to angiogenesis (vessel-dependent). Many factors affect the efficacy of TACE in the treatment of HCC, and tumor angiogenesis is an essential factor. Kenji et al. showed that TACE is effective for HCCs with a rich blood supply and less effective for HCCs with a poor blood supply [13]. High expression of pro-angiogenesis genes is associated with obvious vascular proliferation and vascular abnormalities, including sinusoidal capillarization and arterialization, which promote metastasis in TACE patients. For example, fibroblast growth factor receptor 4 (FGFR4), interacts with primary fibroblast growth factor (bFGF) to activate essential signaling pathways, including the PLC-γ/PI3/Ca2+signaling pathway, leading to PKC activation and increased angiogenesis [27]. Moreover, investigation of recepteur d’origine nantais (RON, MST1R) activity in epithelial cancer cell lines revealed its roles in cell proliferation, survival, migration, and epithelial-mesenchymal transition. Loss of RON function has been demonstrated to affect tumor growth, angiogenesis, and metastasis in a mouse model of cancer [28]. In contrast, the upregulated expression of anti-angiogenic genes slows the progression or recurrence of HCC in patients treated with TACE. Previous studies have demonstrated the importance of angiogenesis in HCC progression; however, few have translated this knowledge into practical clinical tools. Therefore, it is essential to provide prognostic predictions for patients with HCC treated with TACE by integrating angiogenesis-related genes as effective and reliable biomarkers.
Our investigation was prompted by preliminary research showing correlations between specific genetic variations and the effectiveness of TACE. Building on this foundation, we systematically analyzed angiogenesis-related gene mutations to develop a more precise prognostic framework. This approach is particularly warranted given the direct relationship between the tumor vasculature and the mechanism of action of TACE. We identified mutations in the angiogenesis-related genes FGFR4, MST1R, and RECK, associated with a worse prognosis in patients with HCC treated with TACE, thus providing scientific references for TACE therapy. We speculate that this may be related to decreased gene expression caused by the mutations. For example, reversion-inducing cysteine-rich protein with Kazal motifs (RECK) is a cell membrane-anchored glycoprotein that acts as a metastasis suppressor against the activated Ras oncogene [29]. It is widely expressed in normal tissues. However, its expression was lower in HCC tissues than in the surrounding non-malignant tissues. In animal models, RECK expression inhibited tumor invasion, metastasis, and angiogenesis. Higher RECK expression in HCC tissues is associated with better survival. Thus, upregulation of RECK may be a promising approach for treating malignant tumors [29]. Although we did not observe differential results in the TCGA data (Supplementary Figure 4), this speculation is still worthy of attention, considering the reasons for the small sample size of the mutations. Furthermore, the GRS has proven to be an independent prognostic factor for clinical outcomes. The prognostic predictive ability of the nomogram, including the GRS, showed relatively robust diagnostic value (Fig. 5A). These results indicate that the GRS model can be utilized as a prognostic biomarker for patients with HCC treated with TACE.
In addition, previous studies have shown that pro-angiogenic genes, such as VEGF, could lead to progression and promote neovascularization in patients with HCC treated with TACE [17], although this result was not observed in this study. Drugs that target these pathways have been studied and used in clinical settings [30]. Nevertheless, the relationship between these genes and OS in HCC remains unclear because of the intricate mechanisms of most other angiogenesis-related genes and their combined effects. In this study, the expression levels of 97 angiogenesis-related genes in cancer and adjacent tissues of HCC and their relationship with OS were observed through the TCGA database, which showed that 36 genes (37.11 %) had a gap in expression between the two groups (Supplementary Table4), and 24 genes (24.74 %) were correlated with OS probability through univariate Cox regression analysis (Supplementary Table5). These results suggest that angiogenesis-related genes play a potential role in the tumorigenesis and progression of HCC, providing the possibility of establishing a potential prognostic signature with multiple angiogenesis-related genes.
There are currently few biomarker data available to support decision-making and direct TACE therapy, such as Platelet-to-Lymphocyte Ratio (PLR) [6] and α-fetoprotein (AFP) level [31]. The clinical data of patients, such as results from preoperative laboratory testing and the decision to employ immune medications, were included in our risk assessment algorithm. The treatment for patients with HCC must be determined by assessing liver function. AST, ALB, and PT are serum biomarkers that reveal the liver status of patients with HCC [32]. Inflammation-based scores, such as the PLR, depend on circulating lymphocyte and platelet counts. Platelets facilitate tumor progression by supporting cancer stem cells, inducing angiogenesis, sustaining cell proliferation, and evading immune surveillance. Additionally, a poor cell-mediated immune response to cancer may be associated with relative lymphocytopenia. Tumor growth and immune response escape are exacerbated by these conditions. Therefore, high inflammation-based scores may indicate an HCC-promoting state, which is associated with poor prognosis following TACE [33]. Increasing evidence has demonstrated that response to TACE is a surrogate marker of aggressive tumor biology in patients with early- and intermediate-stage HCC. Conversely, inflammation-based scores have been proposed as prognostic factors for predicting the response to TACE, disease progression, and survival in patients with HCC. They are also biomarkers of the interaction between the tumor stromal microenvironment and immune responses [33]. Xue et al. showed that PLR can be utilized to predict poor survival in HCC patients who underwent TACE, supporting the predictive function of PLR in these patients [34].
We selected these indices for analysis and confirmed by univariate Cox-Lasso stepwise Cox regression analysis that AFP, ALT, AST, PLR, and GRS can be used as independent prognostic factors for HCC patients treated with TACE. It also confirmed existing biomarker evidence and identified the role of liver function-related indicators in predicting patient prognosis. Nonetheless, owing to the heterogeneity of patients, survival rates can differ significantly even among HCC patients with comparable clinicopathological features. Therefore, additional and more precise predictors are required to distinguish patients who can benefit from TACE from those with TACE failure or refractoriness.
Nomograms can provide fast and intuitive tools for the clinical application of predictive models [35]. Currently, some tools have been used to predict the prognosis of HCC, but their performance requires improvement [14,17,22,[36], [37], [38]]. In this study, we developed a nomogram for the prognostic prediction of patients with HCC treated with TACE based on mutations in angiogenesis-related genes, which were evaluated using the ROC curve, calibration curve, and DCA. Moreover, the nomogram demonstrated excellent predictive accuracy in both the training and validation sets, as well as calibration and clinical benefits, which were significantly better than the angiogenesis-related scoring model constructed by Tang et al. (AUCmax=0.79) [14] and the prognostic model constructed by Zhang et al. (AUCmax=0.816) [37,39]. Furthermore, this represents a significant advance over conventional prognostic methods that rely solely on clinical and imaging parameters [40]. These findings have significant clinical implications. Our model provides clinicians with a practical tool for stratifying patients based on their likelihood of responding to TACE, thereby potentially enabling personalized treatment strategies. WES data from hundreds of patients confirmed the reliability and generalizability of the model, addressing a critical gap in current clinical practice. Furthermore, this study established a foundation for future research on targeted therapies that could potentially overcome TACE resistance in high-risk patients. Identification of specific angiogenic pathways associated with poor outcomes may guide the development of complementary therapeutic approaches. These findings contribute to the growing field of precision medicine for HCC treatment, offering a more nuanced approach to patient selection for TACE. By incorporating genetic information into clinical decision making, we can potentially improve treatment outcomes and resource utilization in HCC management.
This study has several shortcomings that need to be addressed in future work. First, only data from one research facility were included in this analysis; further prospective clinical data from several centers are required for validation. Second, there was an inherent flaw in assuming a single phenotype to create a prognostic signature because this kind of analysis may rule out some additional essential prognostic factors in the occurrence and development of HCC. The pathophysiology of HCC involves a complex interplay of multiple genetic and environmental factors, the interplay of which is the origin of the early steps of hepatocyte malignant transformation and HCC development [41]. Thus, further experimental studies are required to determine the underlying molecular mechanism of HCC tumorigenesis and the roles of angiogenesis-related genes.
In summary, we identified mutations in three angiogenesis-related genes associated with poor outcomes in patients with HCC treated with TACE and established GRS as a novel prognostic signature. We constructed and validated an excellent prognostic model and nomogram based on the GRS of the genes mentioned above, which can provide clinicians with more decision-making and reference suggestions regarding the prognosis of HCC patients treated with TACE.

Conclusion
This study identified mutations in three angiogenesis-related genes (FGFR4, MST1R, and RECK) associated with poor outcomes in HCC patients treated with TACE, and developed a novel prognostic model integrating GRS with clinical parameters. The resulting nomogram demonstrated excellent predictive accuracy in internal validation, offering clinicians a practical tool for patient stratification and treatment planning. While multi-center validation is needed, this integrated approach represents a significant advancement toward personalized medicine in HCC management.

Funding

This study was supported by the Clinical characteristic technology projects in Guangzhou City [grant number: C3230118], the National Natural Science Foundation of China [grant number:82172036,82372059], Undergraduate teaching quality and teaching reform cultivation project of the First Affiliated Hospital of Sun Yat-sen University [grant number:2023P12220011–230106], and the Undergraduate teaching quality and teaching reform cultivation project of the First Affiliated Hospital of Sun Yat-sen University in [grant number:2023 P12220011–230134]

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.

CRediT authorship contribution statement
Wanqing Shen: Writing – original draft, Software, Investigation, Formal analysis. Zhi Li: Methodology, Conceptualization. Jinyi Huang: Writing – original draft, Data curation. Guixiong Zhang: Visualization. Yiyang Tang: Validation. Wenzhe Fan: Writing – review & editing, Visualization, Software. Jiaping Li: Writing – review & editing, Supervision, Resources, Funding acquisition.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.