High-Mobility Group Box 1 expression predicts survival of patients after resection of adenocarcinoma of the ampulla of Vater

Background Expression of High-Mobility Group Box 1 (HMGB1), a multifunctional protein involved in DNA function as well as cell proliferation, inflammation, and the immune response, has been reported to be prognostic in several types of malignancies. However, the prognostic value of HMGB1 in ampullary cancer has not been studied. Methods Patients with adenocarcinoma of the ampulla of Vater who underwent R0 resection with pancreaticoduodenectomy between 2001 and 2011 were included in the present multi-institutional study. The degree of HMGB1 expression was examined in each resected specimen by immunohistochemical staining. Results A total of 101 patients were enrolled of which, 79 patients were eligible. High expression of HMGB1 was observed in 31 (39%) patients. Blood loss, transfusion, tumor stage, nodal status, and HMGB1 expression were identified as predictors with univariate analysis. Multivariate analysis showed that transfusion, lymph-node metastasis, and high HMGB1 expression were independent predictors of poor overall survival. Subgroup analysis showed that high HMGB1 expression was predictive, especially in patients who did not receive adjuvant chemotherapy. Conclusions High HMGB1 expression is an independent predictor of poor prognosis in patients with adenocarcinoma of the ampulla of Vater not treated with adjuvant chemotherapy.


Introduction
Ampullary cancer is a rare biliary tract cancer that accounts for 0.2% of digestive tract cancers in the USA [1]. The Japanese Society of Hepato-Biliary-Pancreatic Surgery reported a total of 13,192 bile duct cancer cases, including 2161 (16.4%) ampullary cancer patients in 2016 [2]. Ampullary cancer patients are usually diagnosed at an early stage, with 30% of tumors localized in the mucosa or the sphincter of Oddi [2]. Therefore, more than half of the patients are treated with surgical resection, achieving relatively better long-term outcomes than other biliary tract cancers or pancreatic cancer [2,3]. However, considering that lymph-node metastasis occurs even at an early stage of ampullary cancer, and the 5-year survival for patients with lymph-node metastasis is less than 30%, improvement of patient outcome is still needed [2,4]. Discovery of new predictive markers can provide new therapeutic targets that could result in further improvements in longterm outcome.
Expression of High-Mobility Group Box 1 (HMGB1), a multifunctional protein involved in DNA function as well as other cell functions, has been shown to be a prognostic factor in many cancers, such as esophageal cancer, gastric cancer, colorectal cancer, pancreatic cancer, and hepatocellular carcinoma [5]. However, the relationship between HMGB1 overexpression and survival in ampullary cancer has not yet been elucidated. Therefore, the present study was performed to determine the prognostic impact of HMGB1 in adenocarcinoma of the

Inclusion criteria
Ampullary cancer was defined as a cancer derived from tissue surrounded by the Oddi muscle [6]. Patients with ampullary cancer treated with pancreaticoduodenectomy in YCOG were enrolled in the present study. Patients who met all of the following criteria were eligible for the present study: adenocarcinoma of the ampulla of Vater confirmed by histological examination; achievement of R0 resection by pancreaticoduodenectomy (PD), pylorus-preserving PD (PpPD), or subtotal stomach-preserving PD (SSPPD) between 2001 and 2011; and availability of tissue specimens. Pathologists in each institute reviewed the histological analysis. Clinicopathological data and paraffin-embedded resected tissue specimens or unstained slides from each participant were obtained and collected in the Department of Gastroenterological Surgery, Graduate School of Medicine, Yokohama City University.

Definition of lymph-node dissection for ampullary cancer
The level of lymph-node dissection for ampullary cancer was classified into the following three categories according to the General Rules for Surgical and Pathological Studies on Cancer of the Biliary Tract, 5th edition [6]: D1, dissection of retro-pancreatic lymph-nodes; D2, dissection of pre-pancreatic lymph-nodes, mesenteric lymph-nodes (proximal portion of the superior mesenteric artery, inferior pancreaticoduodenal artery, middle colic artery, and jejunal artery), and lymph-nodes along the distal bile duct in addition to D1 dissection; and D3, dissection of paraaortic lymph-nodes and lymph-nodes around the celiac trunk, common hepatic artery, and hepato-duodenal ligament along with D2 dissection.

Immunohistochemical staining
Fresh tumor samples were fixed in formalin and embedded in paraffin before sectioning and staining. Tissue sections (4-μm-thick) from the center of the tumors were deparaffinized in xylene and rehydrated in an ethanol series. Antigen retrieval was achieved by autoclaving in a citrate buffer (pH 6.0). Intrinsic peroxidase activity was quenched in a hydrogen peroxide phosphate-buffered saline for 30 min at room temperature. Then, bovine serum albumin (10%) was added to block non-specific antibody binding. After blocking non-specific antibody binding, sections were incubated with the primary antibody, rabbit polyclonal anti-HMGB1 antibody (ab18256, Abcam, Cambridge, UK), at 4°C overnight. The bound primary antibody was detected with an anti-rabbit secondary antibody. The labeled antigens were visualized with DAB. Sections were counterstained with hematoxylin.

HMGB1 expression analysis
The expression level of HMGB1 was evaluated by nuclear-staining intensity according to previously reported methods [7]. Four categories of staining were used: 1, no stain, 2-4, increasing intensity of brown stain (Fig. 1). Four categories of positivity were used, each with an increase of 25%. Positivity scores were multiplied to obtain a histochemical score. High expression was defined as a score of greater than or equal to 8, and low expression was defined as no more than 6.

Statistical analysis
Statistical analyses were with SPSS statistics version 21.0 (IBM, New York City, NY, USA). The chi-squared test was used for statistical significance of categorical variables. The Mann-Whitney U test was used for continuous variables. Kaplan-Meier analysis was used for survival and compared by the log-rank test. The Cox proportional hazards model was used for multivariate analysis on the outcome. Significant variables on univariate analysis were included in the multivariate analysis. A P value < 0.05 was considered significant.

Patient characteristics
A total of 101 patients were enrolled, of which 79 patients were eligible for the present study using the criteria described above under the "Patients and methods" section. The patients' characteristics are summarized in Table 1. High expression of HMGB1 was confirmed in 31 (39%) patients. Forty-five patients underwent conventional PD, and 31 underwent PpPD, while 3 patients were treated with SSPPD. Lymph-node dissection greater than D2 was performed on 55 (70%) patients. Median operation time and blood loss were 466 min and 850 ml, respectively; 24% of patients were transfused. Morbidity of greater than grade IIIa, according to the Clavien-Dindo classification, occurred in 35 (44%) patients [8]. Of these cases, 20 had a grade B or C postoperative pancreatic fistula, defined by the International Study Group of Pancreatic Fistula classification [9]. According to the 8th Edition of the American Joint Committee on Cancer Staging Manual, there were 20 patients in stage IA, 18 in stage IB, 15 in stage IIA, 22 in stage IIIA, and 4 in stage IIIB [10]. Gemcitabine monotherapy was administered to 19 patients, and S-1 monotherapy was given to 7 patients. Three patients were treated with UFT, and intra-arterial chemotherapy using a 5-FU plus cisplatin regimen was administered to 2 patients, all as adjuvant chemotherapy. Most patients with lymph-node metastasis were given adjuvant chemotherapy (Table 2). No patients received radiation therapy.

Relationship between HMGB1 expression and other clinicopathological variables
Relationships between HMGB1 expression and other clinicopathological variables are shown in Table 3. Patients with high-expression HMGB1 underwent PD more frequently and showed a larger amount of blood loss than patients with low-expression. None of the tumor factors, including tumor diameter, tumor ulceration, histological type of adenocarcinoma, T state, and N stage, significantly correlated with HMGB1 expression level.

Prognostic factors for patients with adenocarcinoma of the ampulla of Vater treated with R0 resection
The 5-year overall survival rate and median survival time in the whole cohort were 58.1% and 113 months, respectively. Univariate analysis was performed to identify factors affecting overall survival (Table 4). Blood loss (≥ 800 ml), intraoperative transfusion, advanced T stage (T2, T3a), lymph-node metastasis, and HMGB1 high expression were significantly associated with shorter overall survival. Multivariate analysis showed that intraoperative transfusion (hazard ratio [HR] 3.13; 95% confidence interval [CI] 1.26-7.78; P = 0.014), lymph-node metastasis (HR 7.43; 95% CI 2.83-19.54; P < 0.001), and HMGB1 high expression (HR 3.54; 95% CI 1.37-9.16; P = 0.009) were predictors of poor overall survival (Table 4). Kaplan-Meier analysis showed that the 5-year overall survival was significantly less in patients with high HMGB1 expression than a b c d HMGB1 expression was significantly prognostic only in patients who did not receive adjuvant chemotherapy High expression of HMGB1 was a good predictor of survival for patients without adjuvant chemotherapy (5-year overall survival with high and low HMGB1 expression: 36.8% and 71.8%, respectively; P = 0.019; Fig. 3a), but not for patients who received adjuvant chemotherapy (5-year overall survival in high and low HMGB1 expression: 32.8% and 57.0%, respectively; P = 0.63; Fig. 3b).

Discussion
The only known treatment to achieve cure for ampullary cancer is surgical resection. Since 4.6 to 9% of T1 ampullary cancer cases have lymph-node metastasis, the standard treatment method in Japan is pancreaticoduodenectomy with lymph-node dissection [2,4,11]. It has been shown that T stage, lymph-node metastasis, residual cancer cells, and intraoperative transfusion are predictors of poor prognosis in ampullary cancer [12][13][14][15][16]. Moreover, the pancreatobiliary type on histological subtyping, which was first reported by Kimura et al., has also been demonstrated to be associated with survival [14,17].
Fluorouracil-based adjuvant chemoradiotherapy may prolong survival of ampullary cancer patients following pancreaticoduodenectomy, but the evidence is limited [18]. HMGB1, a non-histone DNA-binding protein first reported in the calf thymus in 1973, is present in most mammalian cells [19,20]. Wang et al. showed that HMGB1 was involved in inflammatory responses [21]. Intranuclear HMGB1 participates in several DNA functions, including DNA chaperone, DNA repair, transcription, maintenance of telomeres, and genome stability. Moreover, previous studies reported that HMGB1 enhances the functions of key immune suppressor cells such as myeloid-derived suppressor cells and regulatory T cells, inhibiting anti-tumor immune responses [22,23]. Furthermore, HMGB1 expression was inversely associated with CD45RO + T cell infiltration in colon cancer [24]. Thus, HMGB1 affects anti-tumor immune responses, but further investigations are necessary to better understand the role of HMGB1 in anti-tumor immunity.
Importantly, recent studies have found that HMGB1 expression is associated with tumorigenesis and tumor progression [25,26]. HMGB1 appears to play a particular role in apoptosis inhibition, cell proliferation, angiogenesis, invasion, and metastasis in many malignancies [26]. Wu et al. reviewed the impact of HMGB1 overexpression on patient survival in several cancers [5]. Overexpression of HMGB1 was significantly correlated with poorer overall survival in esophageal cancer, gastric cancer, colorectal cancer, pancreatic cancer, hepatocellular carcinoma, nasopharyngeal cancer, head and neck cancer, bladder cancer, cervical cancer, and pleural mesothelioma. Moreover, there are some reports of HMGB1 in biliary tract cancers. RAGE, a receptor binding to HMGB1, was reportedly associated with invasion capacity in a study using bile duct-cancer cell lines [27]. Xu et al. investigated HMGB1 expression in intrahepatic cholangiocarcinoma, concluding that HMGB1 overexpression was significantly correlated with worse survival [28]. The level of HMGB1 expression was reported to be associated with lymph-node metastasis and advanced TNM stage in esophageal cancer, gastric cancer, colorectal cancer, head and neck cancer, and cervical cancer [29][30][31][32][33]. Moreover, HMGB1 expression was positively correlated with lymphatic-vessel density in esophageal cancer and intrahepatic cholangiocarcinoma [28,29]. These results suggest possible reasons why HMGB1 overexpression may contribute to poorer patient survival. The present study is the first to examine the association of HMGB1 expression and ampullary cancer. Our study demonstrated that high HMGB1 expression was significantly associated with shorter survival for patients with adenocarcinoma of the ampulla of Vater following R0 resection. HMGB1 was predictive for patients only without adjuvant chemotherapy. Adjuvant chemotherapy for patients with high HMGB1 expression, as well as patients with lymph-node metastasis, may be effective for some patients. Recently, there have been reports on novel HMGB1-targeted therapy in preclinical models. In a mouse model, anti-HMGB1 antibody therapy was reported to reduce colorectal cancer liver metastasis  [34]. Similarly, a study using mouse models showed that HMGB1-RAGE-signaling blockade suppressed tumor growth in which both RAGE and HMGB1 were expressed [35]. Based on these results, HMGB1-targeted therapy for ampullary cancer with HMGB1 overexpression is promising. Our results suggest a general phenomenon that has to be further studied.
There are several limitations in this study. First, the total number of patients was not high, despite it being a multi-institutional study. Second, the patients underwent different adjuvant chemotherapy regimens. Finally, the detailed molecular biological mechanisms of HMGB1 examined in the present study were not studied. Even after taking these limitations into consideration, however, the results of the present study suggests a new biomarker for survival, as well as a potential treatment target for adenocarcinoma of the ampulla of Vater. We have not studied the expression of HMGB1 as a function of the histologic subtype and plan to do in future studies.

Conclusions
The present study suggests that HMGB1 is a prognostic factor and potential biomarker for survival in patients not treated with adjuvant chemotherapy with adenocarcinoma of the ampulla of Vater not treated with adjuvant chemotherapy.