A panel of protein kinase high expression is associated with postoperative recurrence in cholangiocarcinoma

Background Cancer recurrence is one of the most concerning clinical problems of cholangiocarcinoma (CCA) patients after treatment. However, an identification of predictive factor on Opisthorchis viverrini (OV)-associated CCA recurrence is not well elucidated. In the present study, we aimed to investigate the correlation of twelve targeted protein kinases with CCA recurrence. Methods Twelve protein kinases, epidermal growth factor receptor (EGFR), human epidermal growth factor receptor 2, 3, 4 (HER2, HER3, HER4), vascular endothelial growth factor receptor 3 (VEGFR3), vascular endothelial growth factor-C (VEGF-C), erythropoietin-producing hepatocellular carcinoma receptor type-A3 (EphA3), EphrinA1, phosphor-serine/threonine kinase 1 (p-Akt1), serine/threonine kinase 1 (Akt1), beta-catenin and protein Wnt5a (Wnt5a) were examined using immunohistochemistry. Pre-operative serum tumor markers, CA19–9 and CEA were also investigated. Results Among twelve protein kinases, EGFR, HER4, and EphA3 were associated with tumor recurrence status, recurrence-free survival (RFS) and overall survival (OS). Multivariate cox regression demonstrated that EGFR, HER4, EphA3 or the panel of high expression of these proteins was an independent prognostic factor for tumor recurrence. The combination of high expression of these proteins with a high level of CA19–9 could improve the predictive ability on tumor recurrence. Moreover, the patients were stratified more accurately when analyzed using the combination of high expression of these proteins with primary tumor (T) or lymph node metastasis (N) status. Conclusion EGFR, HER4, EphA3 or the panel of high expression of these proteins is an independent prognostic factor for post-operative CCA recurrence.


Background
Cholangiocarcinoma (CCA) is a malignant tumor of bile duct epithelium with very high incidence in Thailand, particularly in northeastern region, of which Opisthorchis viverrini (OV) infection is reported as the major risk factor of CCA development in this area [1]. CCA is usually asymptomatic in early stage and most patients are diagnosed with CCA when the disease becomes advanced, resulting in poor outcome [2]. Moreover, the recurrence after treatment is nowadays very important, because it is a significant problem for many patients with cancer and is involved in poor prognosis of patients [3]. In CCA, the high recurrence rate was reported in many studies [4,5]. A precious study reported that most CCA patients developed recurrence within 2 years after surgery and the percentage of recurrence accounted for 62.2% [5]. Recently, recurrence rate in mass-forming type of intrahepatic CCA patients was reported with the recurrence rate of 80%. 1-, 2-, and 3-year RFS rate were very low which were 16.2, 5.4, and 2.7%, respectively. However, the association between RFS and clinicopathological data was not significant [4]. Thus, the effective prognostic biomarkers are required to assess outcome of CCA patients as well as the probability of recurrence after treatment.
Nowadays, there are several markers reported as tumor behavior predictors. They can be used for disease management including progression and the relapse indicators of cancer. Serum tumor markers are the wellestablished markers for monitoring tumor and have been reported to predict tumor recurrence in many types of cancer [6,7]. However, molecular biomarkers are widely studied because it is not only used for the predicting of tumor progression or recurrence, but can also be employed as drug target for cancer treatment. Our group previously reported the alteration of protein kinase expression in CCA. We found that many protein kinases were upregulated in CCA tissue and cell lines, including receptor tyrosine kinase, the epidermal growth factor receptor (EGFR) family, vascular endothelial growth factor (VEGFR) receptor, erythropoietinproducing hepatocellular carcinoma (Eph) receptor, and also many down-steam kinases such as serine/threonine kinase or protein kinase B (Akt), and Wnt/beta-catenin signaling pathways [8]. The evaluation of EGFR expression was reported in CCA and associated with poor prognosis of CCA patients [9]. Furthermore, our group also reported that high expressions of VEGFR3, EphA3 and their ligands were correlated with CCA metastasis [10]. The role of protein kinase in PI3K/Akt signaling pathway was also studied in CCA. The results showed that high expression of protein in this pathway was mostly involved in the worse clinical outcome of CCA patients. Moreover, targeting of this pathway using NVP-BEZ235 could inhibit tumor growth and metastasis through reduced protein kinase activation [11]. The association of Wnt/beta-catenin signaling pathway with CCA progression was also reported. The result showed the alteration of Wnt proteins was associated with poor prognosis of CCA patients, and inhibition of betacatenin expression could inhibit CCA cell growth [12].
Large-scale multi-omics have also been employed in many studies in order to understand the carcinogenesis as well as the progression of disease. In 2015, a previous study reported the genomic alteration which characterized biliary tract cancer (BTC) patients. EGFR family genes including EGFR, ERBB2 (HER2), ERBB3 (HER3) were the most activating gene in gallbladder cancer while EPHA2 mutation was found frequently in intrahepatic CCA (iCCA) [13]. ERBB2 amplification was reported for 3.9-8.5% of CCAs. This was more frequent in fluke-associated CCA which account for 10.4% compared with 2.7% of fluke-negative CCA, resulting in the elevation of ERBB2 gene expression in fluke-associated CCA compared with fluke-negative cases. In addition, the upregulation of AKT1 and WNT5B was also reported [14]. Single-nucleotide variations (SNVs) and insertion-deletions (indels) were found in ERBB3 gene in BTCs (5%). This mutation was significantly enriched in extrahepatic CCA (eCCA) [15]. Recently, Nepal et al. reported that the mutation of ERBB4 gene was also found in intrahepatic CCA (iCCA). In addition, pathway dysregulation in each subgroup of patients was explored. They found that the patients who have KRAS mutation were enriched for immune-related pathways, ErbB and VEGF pathways. On the other hand, WNT pathway was enriched in patients with TP53 gene mutation [16].
Since protein kinases play an important role in CCA progression and are involved in poor prognosis of CCA patients. In the current study, we hypothesized that the alteration of these protein kinases including EGFR family, VEGFR3 and its ligand, Eph receptor and its ligand, Akt1 and its activated form, Wnt, and beta-catenin may be used as the predicting markers for post-operative CCA recurrence. Therefore, twelve protein kinases were examined using immunohistochemistry and analyzed against CCA recurrence status, recurrence location, recurrence-free survival (RFS) and overall survival (OS).

Patient selection criteria and follow-up
OV-associated cholangiocarcinoma (CCA) patients who underwent surgery at Srinagarind Hospital, Khon Kaen University, Khon Kaen, Thailand between February, 2007 and December, 2016 were retrospectively studied. In order to avoid the effect of neoadjuvant on protein expression, the patients were excluded if they received either radiotherapy or chemotherapy before operation. Tissue samples were obtained from CCA patients and kept in the BioBank of the Cholangiocarcinoma Research Institute. The clinical information was assessed in all CCA patients including sex, age, tumor location, histology, size of primary tumor (T stage), lymph node metastasis status (N stage), distant metastasis status (M stage), and TNM staging. In addition, tumor makers (carbohydrate antigen 19-9; CA19-9 and carcinoembryonic antigen; CEA) were examined in pre-operative serum.
For the recurrence, first year after surgery, all CCA patients were followed-up every 3 months and every 6 months thereafter. Post-operative recurrence was defined in the patients who developed new tumor which confirmed by computed tomography (CT)/magnetic resonance imaging (MRI). The interval between the date of operation until the date of recurrence or until the last of follow-up was defined as recurrence-free survival (RFS) and the interval between the date of operation until the date of death or until the last of follow-up was defined as overall survival (OS). Early recurrence was defined if patients developed the new tumor within 1 year after surgery, while late recurrence was defined if patients developed the new tumor after 1 year. This study was approved by the Human Research Ethics Committees, Khon Kaen University, Thailand (HE611412).

Immunohistochemical staining (IHC)
A CCA tissue microarray (TMA) was prepared from two independent puncture from each patient and cut into 4 μm for each section. The expression of protein was investigated using IHC. Briefly, the sections were deparaffinized with xylene and rehydrated with stepwise of 100, 90, 80 and 70% ethanol, respectively. Microwave cooking was used for antigen retrieval for 10 mins. Then tissue sections were incubated with 0.3% hydrogen peroxide followed by 10% skim milk for 30 mins of each in order to inhibited endogenous hydrogen peroxide activity, and nonspecific binding. After washing the sections were incubated with primary antibodies at room temperature for 1 h followed by 4°C overnight. The excess antibodies were washed for 3 times using phosphate buffered saline (PBS) with 0.1% tween20 followed by PBS for 5 mins of each. The sections were then incubated with HRP-conjugated secondary antibodies for 1 h, and the excess antibodies were also washed using PBS with 0.1% tween20 followed by PBS for 5 mins of each. A 3, 3'diaminobenzidine tetrahydrochloride (DAB) substrate kit (Vector Laboratories, Inc., CA) was used to develop the signal. The tissues were then counterstained using hematoxylin for 2 mins. After washing, the tissue sections were dehydrated with stepwise of 70, 80, 90, 100% ethanol and xylene, respectively. Tissue sections were mounted with permount, and finally observed under light microscopy.

Immunohistochemical (IHC) scoring
The expression of each protein was scored based on intensity and frequency which is the proportion of positive cells stanning. The intensity of protein expression was classified into four levels including 0 = negative, 1 = weak, 2 = moderate, and 3 = strong stanning. The proportion of positive cells stanning was semi-qualitatively, and classified into negative = 0%, 1 = 1-25%, 2 = 26-50%, and 3 = more than 50% positive stanning. The grading score was calculated by multiplying between intensity and frequency, and the minimum score was 0 while the maximum score was 9. The grading score of each patient was calculated from the average value of two independent punctures. Finally, the median value was calculated from all cases and used as cut-off point. The patients having a grading score lower, equal to or higher than the median was classified as the low or high expression group, respectively. For the proteins which have a median equal to zero, the patients have a grading score equal to zero, being classified as the negative group, while those with a grading score above zero are classified as the positive group.

Statistical analysis
Statistical Package for the Social Science; SPSS software v.25 was used to analyze data in this study. Chi-square test was used to analyze the correlation between protein kinase expression with recurrence status and clinicopathological characteristics of CCA patients. The difference in IHC score and tumor marker levels on recurrence and recurrence location was analyzed using the Kruskal-Wallis test and Mann-Whitney U-test. Kaplan-Meier (log-rank) analysis was used to analyze RFS and OS. The predictive ability of protein kinases on RFS and OS was analyzed by Cox proportional hazards regression. Statistical significance was considered if p-value less than 0.05.   EGFR Epidermal growth factor receptor, HER Human epidermal growth factor receptor, VEGFR3 Vascular endothelial growth factor receptor 3, VEGF-C Vascular endothelial growth factor-C, EphA3 Erythropoietin-producing hepatocellular carcinoma receptor type-A3, p-Akt1: Phosphor-serine/threonine kinase 1, Akt1 Serine/threonine kinase 1, Wnt5a Protein Wnt5a, NA Not applicable

Patients characteristics
A total of 190 CCA patients (35% female and 65% male) were recruited in the current study. The median of age was 61 years (rang between 39 and 82). 55% of patients were classified as intrahepatic CCA cases while 45% were extrahepatic CCA cases. 43% of patients were characterized as papillary type and 57% were other types. Size of primary tumor (T) was also classified and 57% of patients were T stage I and II, whereas 43% were T stage III and IV. From 190 patients, lymph node (N) and distant (M) metastasis were shown in 55 and 6% of patients, respectively. TNM staging was also characterized according to size of primary tumor, lymph node and distant metastasis status. In this study, 40% of patients were stage I and II and 60% were stage III and IV and recurrence after surgery was also detected in 31% ( Fig. 1) (Table S1). Among patients with recurrence, 53% were classified as early recurrence while 47% were late recurrence. The median follow-up was 16, 28, and 13 months for no recurrence, late recurrence and early recurrence groups, respectively.

The correlation of protein kinases with post-operative recurrence and clinicopathological characteristics
In the present study, 12 protein kinases including EGFR, HER2, HER3, HER4, VEGFR3, VEGF-C, EphA3, EphrinA1, p-Akt1, Akt1, beta-catenin and Wnt5a were examined in CCA tissues obtained from190 cases using IHC. The expression of each protein was defined as high  = 19). b, the levels of CA19-9 and CEA in different recurrence location, locoregional (n = 11) and distant recurrence/ combination between locoregional recurrence with distant recurrence (n = 27). p-value less than 0.05 was considered as statistical significance and low expression or positive and negative. The expression in individual patients was showed in Fig. 1 (Fig. 2). The expressions of all proteins were analyzed with post-operative recurrence including early and late recurrence in order to identify proteins that can be used for the prediction of tumor recurrence. In addition, the expression of beta-catenin was examined in the different cellular compartments, cytoplasm, membrane and nucleus. Positive expression of beta-catenin in cytoplasm, membrane and nucleus were 17, 8 and 2%, respectively. Among 12 protein kinases, the expression of EGFR, HER4, and EphA3 was significantly associated with early recurrence (p = 0.038: p = 0.033: p = 0.008; Table 2), while HER2 and p-Akt1 were significantly correlated with late recurrence (p = 0.035: p = 0.029; Table 2). In contrast, there was no correlation between HER3, VEGFR3, VEGF-C, EphrinA1, Akt1, beta-catenin, Wnt5a and post-operative recurrence ( Table 2). The IHC scores of EGFR, HER2, HER4, EphA3 and p-Akt1 were also compared between patients with and without recurrence. The IHC score of EGFR was significantly different between patients with early recurrence compared with late or without recurrence (p = 0.029: p = 0.024; Fig. 3a). The IHC scores of HER2 and p-Akt1  were significantly higher in patients with late recurrence compared with no-recurrence (p = 0.002: p = 0.013; Fig.  3a), while IHC scores of HER4 and EphA3 were significantly higher in patients with early recurrence compared with no-recurrence (p = 0.003: p = 0.004; Fig. 3a). On the contrary, there was no difference between IHC scores of HER3, VEGFR3, VEGF-C, Ehprin-A1, Akt1, beta-catenin and Wnt5a (Fig. S1 and S2). The IHC scores of these proteins were also analyzed with recurrent location. The expressing level of p-Akt1 was significantly higher in the patients with distant recurrence/combination between locoregional recurrence with distant recurrence compared with locoregional recurrence (p = 0.004; Fig. 3b), while there was no statistical difference in EGFR, HER2, HER4 and EphA3 (Fig. 3b). The expression levels of EGFR, HER2, HER4, EphA3 and p-Akt1 were also analyzed with clinicopathological characteristics. Our finding only showed the sigfinicant correlation between expression of HER4 and lymph node metastasis (p = 0.045; Table 3).   TNM Size of primary tumor-node metastasis-distant metastasis, EGFR Epidermal growth factor receptor, HER Human epidermal growth factor receptor, EphA3 Erythropoietin-producing hepatocellular carcinoma receptor type-A3, Protein panel a : the expression of EGFR, HER4 and EphA, Others b : 0-1 marker high, Others c : three groups of patients (0-1 marker high and CA19-9 low, 0-1 marker high and CA19-9 high or 2-3 markers high and CA19-9 low) Fig. 8 Kaplan-Meier analysis for RFS and OS according to the combined of three protein kinase expression (EGFR, HER4 and EphA3). Upper and lower panels demonstrated the prognostic significantly of the combined of three protein kinase expression or the combined of three protein kinase expression with CA19-9 level on RFS and OS, respectively. 0-1 marker high represented the patients with all markers low or one marker high, 2-3 markers high represented the patients with at least two markers high, others represented three groups of patients (0-1 marker high and CA19-9 low, 0-1 marker high and CA19-9 high or 2-3 markers high and CA19-9 low). p-value less than 0.05 was considered as statistical significance

The correlation of tumor maker level with post-operative recurrence
Since tumor markers were also used to monitor patients after treatment. Therefore, in the present study, CA19-9 and CEA levels were analyzed with tumor recurrence. The result revealed that the level of CA19-9 was significantly higher in early recurrence compared with no-recurrence (p = 0.017) (Fig. 4a), whereas there was no difference between CEA level in patients with and without recurrence (Fig. 4a). In addition, the levels of CA19-9 and CEA were also analyzed with recurrence location. All markers were likely to increase in distant recurrence/combination between locoregional recurrence with distant recurrence, compared with locoregional recurrence. However, there was no such statistically significant correlation in this study (Fig. 4b).
The above results demonstrate that the expression of EGFR, HER2, HER4, EphA3 and p-Akt1 was significantly associated with post-operative recurrence. Thus, the  Fig. 6) and OS (p = 0.016: p = 0.025: 0.018; Fig. 6), compared with those patients with low expression. However, there was no significance found in HER2 and p-Akt1 (Fig. 6). Because the expressing levels of EGFR, HER4 and EphA3 were highly correlated with each other, their expressing levels were also associated with patient prognosis. Therefore, the combination of these proteins was also analyzed with patient prognosis. High expression of the protein pairs, EGFR and HER4, EGFR and EphA3, and HER4 and EphA3 was significantly associated with shorter RFS (p = 0.001: p = 0.008: p = 4.0 × 10 − 4 ; Fig. 7). High expression of EGFR and HER4, HER4 and EphA3 was also associated with a shorter OS (p = 0.043: p = 0.002; Fig. 7). In addition, patients who had high expression of two and three proteins were significantly associated with shorter RFS (p = 3.5 × 10 − 4 ; Fig. 8) and OS (p = 0.012; Fig. 8). The level of tumor marker CA19-9 was also correlated with tumor relapse. Thus, the prognostic efficiency of the combination of protein kinases expression and tumor marker level was also explored. It was significantly associated Erythropoietin-producing hepatocellular carcinoma receptor type-A3, Protein panel a : the expression of EGFR, HER4 and EphA3, Protein panel a : the expression of EGFR, HER4 and EphA, Others b : 0-1 marker high, Others c : three groups of patients (0-1 marker high and CA19-9 low, 0-1 marker high and CA19-9 high or 2-3 markers high and CA19-9 low) with shorter RFS (p = 1.5 × 10 − 4 ; Fig. 8) and OS (p = 0.008; Fig. 8).

Independent prognostic value of EGFR, HER4 and EphA3
In order to investigate whether EGFR, HER4 and EphA3 could be used as prognostic factors, independent of clinicopathological characteristics, Cox regression analysis was used. The univariate result for factors predicting the RFS and OS is shown in Table 5. Multivariate Cox regression for RFS and OS was analyzed using the different models that are summarized in Table 6 and Table 7.
The result demonstrated that EGFR, HER4 and EphA3 were the independent prognostic factors for RFS (HR: 1.542; p = 0.006, HR:1.388; p = 0.042, HR: 1.469; p = 0.001; Table 6). EGFR and EphA4 were also independent prognostic factors for OS (HR: 1.450; p = 0.019, HR: 1.372; p = 0.040; Table 7). The combination of high expression of two and three markers or the high expression of two and three markers with high level of CA19-9 could be used to improve the predictive ability for RFS (HR: 1.528; p = 0008, HR: 2.080; p = 0.004; Table 6). Moreover, the patients were stratified more accurately when analyzed using the combintion of protein kinase expression and primary tumor (T) or lymph node metastasis (N) status. The patients with high stage T and high expression of two and three markers or high expression of two and three markers with high level of CA19-9 have shorter RFS, compared with other groups (p = 2.1 × 10 − 9 : p = 6.9 × 10 − 9 ; Fig. 9). Similarly, patients with lymph node metastasis and high expression of two and three markers or high expression of two and three markers with a high level of CA19-9 have shorter RFS compared with other groups (p = 9.0 × 10 − 7 : p = 3.8 × 10 − 5 ; Fig. 9). Fig. 9 Kaplan-Meier analysis for RFS according to the combination of three protein kinase expression (EGFR, HER4 and EphA3) and clinicopathological features. 0-1 marker high represented the patients with all markers low or one marker high, 2-3 markers high represented the patients with at least two markers high, others represented three groups of patients (0-1 marker high and CA19-9 low, 0-1 marker high and CA19-9 high or 2-3 markers high and CA19-9 low). T represented the primary tumor stage, N represented the lymph node metastasis status (N0: no lymph node metastasis, N1: lymph node metastasis). p-value less than 0.05 was considered as statistical significance may be used as a guideline for clinical intervention in order to improve patient survival.

Conclusion
Our results demonstrate that the elevated expression of EGFR, HER4, and EphA3 is correlated with OVassociated CCA recurrence. Moreover, the panel of high expression of EGFR, HER4, and EphA3 can be used as a prognostic factor for CCA recurrence, especially when combined with CA19-9 or clinicopathological features, primary tumor (T) or lymph node metastasis (N) status.