Overexpression of AKR1B10 Predicts Poor Prognosis in Gastric Cancer Patients Undergoing Surgical Resection

Aldo–keto reductase family 1 member B10 (AKR1B10) is associated with several cancers, but the prognostic role in gastric cancer (GC) remains unclear. We enrolled 359 GC patients who underwent a gastrectomy with D2 lymph node dissection. AKR1B10 expression was scored using an immunoreactive scoring system based on immunohistochemistry. Adjuvant chemotherapy with S-1 or oxaliplatin plus capecitabine was administered to pathological stage II or III disease patients. There were 117 (32.6%) and 242 (67.4%) patients with AKR1B10 overexpression and low expression, respectively. Patients overexpressing AKR1B10 had worse 5-year disease-free survival (DFS) and overall survival (OS) rates than those with low expression of AKR1B10. Pathological T3–T4 stage, pathological stage III, lymph node ratio ≥25%, and AKR1B10 overexpression were independent prognostic factors for worse DFS and OS in univariate and multivariate analyses. For 162 stage II or III patients who received adjuvant chemotherapy after surgical resection and 59 patients with signet ring cell carcinoma histology, AKR1B10 overexpression was also associated with inferior DFS and OS. AKR1B10 was not associated with clinical survival in stage I GC patients. In conclusion, AKR1B10 overexpression may be an independent prognostic factor for worse survival in GC patients who underwent gastrectomy with D2 lymph node dissection.


Introduction
Gastric cancer (GC) is one of the most aggressive malignancies worldwide and is the 8th leading cause of cancer-associated mortality in Taiwan [1]. For operable diseases, radical gastrectomy using extended regional lymph node dissection is the gold standard treatment. Recently, adjuvant chemotherapy has been approved to improve disease progression and decrease tumor recurrence [2,3]. Despite the development of surgical techniques, chemotherapy, radiotherapy, and immunotherapy, the long-term survival rate still remains poor. This indicates that identifying novel mechanisms to improve clinical outcomes is crucial.
The aldo-keto reductase protein superfamily consists of over 190 members and is a family of enzymes that includes many related monomeric NADPH-dependent oxidoreductases [4]. These enzymes reduce carbonyl substrates such as sugar aldehydes, keto-steroids, tion of the American Joint Committee on Cancer (AJCC) staging system [27]; (iv) Eastern Cooperative Oncology Group (ECOG) 0-1; (v) adjuvant chemotherapy with S-1 or oxaliplatin/capecitabine was prescribed if indicated [2,3]; and (vi) distant metastasis or history of second primary malignancy. In total, 359 GC patients who met these inclusion criteria were included in our study.
Post-operative patients were followed up at the outpatient clinic every 2-4 weeks during the first year. Each patient underwent computed tomography (CT) of the abdomen every 3-6 months; a panendoscopy was performed every 6 months for at least one year. Adjuvant chemotherapy with S-1 or oxaliplatin/capecitabine was administered to patients with pathological stage II or III. The duration and dose of chemotherapy were based on previous clinical trials [2,3]. The definition of LN ratio is the total number of involved LNs divided by the dissected LNs [28].

Immunohistochemistry
The formalin-fixed paraffin-embedded surgical specimen of each patient was sectioned to be 4 µm thick. Next, the deparaffinization step was performed in a dry oven at 60 • C for one hour. Antigen retrieval with 0.01 mol/L citrate buffer (pH 6.0) in a hot water bath (95 • C) for 20 min, and peroxidase blocking using 0.3% hydrogen peroxide for five minutes was done. The specimen was then incubated with the AKR1B10 antibody (Abnova, H00057016-M01, 1:80, Walnut, CA, USA) for 30 min at room temperature. A healthy human colon was used as a positive control, and the negative control was performed by replacing the primary antibody with an isotype-matched irrelevant antibody. These slides were independently investigated by two pathologists (S.L. Wang and W.T. Huang) in a blinded manner. The immunohistochemical staining was performed according to a previously published method [29].
during the first year. Each patient underwent computed tomography (CT) of the a every 3-6 months; a panendoscopy was performed every 6 months for at least Adjuvant chemotherapy with S-1 or oxaliplatin/capecitabine was administered to with pathological stage II or III. The duration and dose of chemotherapy were previous clinical trials [2,3]. The definition of LN ratio is the total number of invo divided by the dissected LNs [28].

Immunohistochemistry
The formalin-fixed paraffin-embedded surgical specimen of each patient tioned to be 4 μm thick. Next, the deparaffinization step was performed in a dry 60 °C for one hour. Antigen retrieval with 0.01 mol/L citrate buffer (pH 6.0) in a h bath (95 °C) for 20 min, and peroxidase blocking using 0.3% hydrogen peroxid minutes was done. The specimen was then incubated with the AKR1B10 antib nova, H00057016-M01, 1:80, Walnut, CA, USA) for 30 min at room temperature. A human colon was used as a positive control, and the negative control was perfo replacing the primary antibody with an isotype-matched irrelevant antibody. The were independently investigated by two pathologists (S.L Wang and W.T Hu blinded manner. The immunohistochemical staining was performed according to ously published method [29]. The expression of AKR1B10 was scored based on the immunoreactive sc system. It provides a range of 0-12 as a product of multiplication between the intensity score (0: no staining, 1: weak, 2: moderate, and 3: strong) and positive portion score (0:0%, 1:1-10%, 2:11-50%, 3:51-80%, and 4:81-100%) [30]. Any s with a sum score > 6 was defined as overexpression of AKR1B10. The expression AKR1B10 are shown in Figure 1.

Statistical Analysis
Statistical analyses were conducted using SPSS software (version 26.0; International Business Machines Corp., New York, NY, USA) and R software version 3.3.0 (R foundation for Statistical Computing, Vienna, Austria). The differences in categorical variables between groups were examined using the chi-square test. DFS was defined as the duration from the date of GC diagnosis to death or last living contact. OS was calculated from surgery to tumor recurrence or death from any cause. Univariate analysis was performed using the Kaplan-Meier method, and the log-rank test was used to examine potential differences. All categorical variables (p-values < 0.1) in the univariate analyses were further entered into a multivariate Cox proportional hazards model to identify significant factors using forward stepwise selection independently. Statistical significance was defined as a twotailed p-value < 0.05.

Construction and Validation of Nomogram
A nomogram to predict rate of recurrence or death was constructed based on the significant factors of the multivariate analyses. An external independent validation cohort was used to validate the nomogram, and the predictive performance was evaluated by the concordance index (C-index) and calibration plot. The measurement of nomogram between performance and predicted outcomes was determined by C-index. The comparison between actual and nomogram-predicted outcome was performed by calibration plots using a 45-degree line as an optimal model.

Ethical Statement
The study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of Chang Gung Medical Foundation (201900004B0). Written informed consent was not required because of the retrospective design of this study.
However, for 127 patients with stage I disease, including 40 (31.5%) patients with overexpression of AKR1B10 and 87 (68.5%) with low expression of AKR1B10, the difference between the groups did not reach significance. This was despite inferior 5-year DFS (67.2% versus 81.2%, p = 0.39, Figure 3E) and 5-year OS rates (68.8% versus 86.1%, p = 0.33, Figure 3F), in addition to overexpression of AKR1B10 in comparison with those with low expression, in these patients.

Nomogram and Validation
The nomogram that integrated four independent risk factors to predict recurrence and death in the training cohort is shown in Figure 4. The C-index for the model prediction was 0.75 (95% CI: 0.70-0.80) for recurrence and 0.76 (95% CI: 0.71-0.81) for death in the training model. The calibration plot for the recurrence and death revealed an optimal agreement between the prediction by the nomogram and actual observation ( Figure 5A

Discussion
Gastric cancer is an aggressive malignancy; radical gastrectomy with extended regional lymph node dissection is the gold standard treatment for operable diseases. In addition to the well-known risk factors, such as advanced tumor stage or higher LN ratio, several predictive biomarkers have been identified in previous studies [25,31]. AKR1B10 has been reported to be associated with cancer development, progression, and treatment through several mechanisms, such as detoxification of RCS or regulation of lipid synthesis [4]. AKR1B10 is expressed in many organs and it has also been studied in a variety of cancers, such as in liver, lung, gynecological cancer, and breast [5,[8][9][10][11][12][13][14]. Moreover, AKR1B10 positively correlated with tumor size, lymph node metastasis, and clinical outcome in breast cancer. Silencing of AKR1B10 also inhibited tumor proliferation in cell lines and animal studies. In addition, the serum AKR1B10 was higher in patients with breast cancer than in the healthy cohort. In our study, poor clinical outcomes were noted in patients with advanced pathological T or N status, tumor stage, total gastrectomy, PNI, LVI, and a higher LN ratio. In addition, 117 (32.6%) patients were found to overexpress AKR1B10; univariate and multivariate analyses showed that this overexpression was an independent prognostic factor for worse DFS and OS in GC patients.
However, the role of AKR1B10 remains unclear. Growing evidence has shown that AKR1B10 is a tumor suppressor [23,24]. Shao et al. reported that AKR1B10 expression was significantly related to smaller tumor size, lower depth of invasion, negative lymph node metastasis, negative venous invasion, and advanced tumor stage [23]. In addition, AKR1B10 knockdown increased tumor cell proliferation, whereas its overexpression reduced tumor growth by enhancing EMT in an in vitro study. Depletion of AKR1B10 could

Discussion
Gastric cancer is an aggressive malignancy; radical gastrectomy with extended regional lymph node dissection is the gold standard treatment for operable diseases. In addition to the well-known risk factors, such as advanced tumor stage or higher LN ratio, several predictive biomarkers have been identified in previous studies [25,31]. AKR1B10 has been reported to be associated with cancer development, progression, and treatment through several mechanisms, such as detoxification of RCS or regulation of lipid synthesis [4]. AKR1B10 is expressed in many organs and it has also been studied in a variety of cancers, such as in liver, lung, gynecological cancer, and breast [5,[8][9][10][11][12][13][14]. Moreover, AKR1B10 positively correlated with tumor size, lymph node metastasis, and clinical outcome in breast cancer. Silencing of AKR1B10 also inhibited tumor proliferation in cell lines and animal studies. In addition, the serum AKR1B10 was higher in patients with breast cancer than in the healthy cohort. In our study, poor clinical outcomes were noted in patients with advanced pathological T or N status, tumor stage, total gastrectomy, PNI, LVI, and a higher LN ratio. In addition, 117 (32.6%) patients were found to overexpress AKR1B10; univariate and multivariate analyses showed that this overexpression was an independent prognostic factor for worse DFS and OS in GC patients.
However, the role of AKR1B10 remains unclear. Growing evidence has shown that AKR1B10 is a tumor suppressor [23,24]. Shao et al. reported that AKR1B10 expression was significantly related to smaller tumor size, lower depth of invasion, negative lymph node metastasis, negative venous invasion, and advanced tumor stage [23]. In addition, AKR1B10 knockdown increased tumor cell proliferation, whereas its overexpression reduced tumor growth by enhancing EMT in an in vitro study. Depletion of AKR1B10 could facilitate tumor cell proliferation, resulting in tumor size progression and body weight gain in an in vivo xenograft model [23]. Another GC study demonstrated that the expression of AKR1B10 mRNA was higher in normal tissues than in GC tissues [24]. The OS was significantly higher in patients with positive AKR1B10 expression than in those with negative AKR1B10 expression. Its expression was an independent prognostic factor of better survival in Cox regression analysis [24]. In contrast, GC patients with AKR1B10 overexpression had advanced lymph node metastasis, fewer tumor regression, and worse overall survival (OS) compared to those with negative expression of AKR1B10 [25]. Its expression was also correlated with better tumor regression for patients receiving neoadjuvant chemotherapy. A meta-analysis further revealed that high expression of AKR1B10 was not found related to DFS and OS [26]. In our study, we enrolled a large sample size and tried to elucidate the clinical impact of AKR1B10 in GC patients receiving surgical resection. We found that overexpression of AKR1B10 was an independent prognostic factor of worse DFS and OS in the univariate and multivariate analyses. This indicated that AKR1B10 might be regarded to possess an oncogenic role in GC patients. In addition, the prognostic role of AKR1B10 was also validated in TCGA database. In this gastric adenocarcinoma cohort from TCGA database, there were no significant differences of OS between overexpression and low expression of AKR1B10 in whole population; however, in the subgroup analysis, the survival outcome in difference races is a little different, better OS was mentioned in Asian patients with low expression of AKR1B10 compared to Asian patients with overexpression of AKR1B10. On the other hand, the expression of AKR1B10 increased with a trend from early stage to advanced stage, and a similar finding was also mentioned in the status of lymph node metastasis [32].
Adjuvant chemotherapy has been proven to be a standard treatment for stage II or stage III GC patients who underwent surgical resection [33,34]. A previous study showed that AKR1B10 expression could predict the response to neoadjuvant chemotherapy in GC patients (25). Better tumor regression was found to be higher in patients with negative AKR1B10 expression than in those with positive AKR1B10 expression (60% versus 40%, p = 0.033). However, the role of AKR1B10 in adjuvant therapies remains unclear. In our study, we investigated the role of AKR1B10 in GC patients who received adjuvant chemotherapy after surgery. In 162 patients who received S-1 or oxaliplatin plus capecitabine as adjuvant chemotherapy, overexpression of AKR1B10 was strongly associated with worse 5-year DFS (30.4% vs. 46.4%, p = 0.001) and OS (34.6% vs. 63.5%, p < 0.001) rates. This strongly suggests that AKR1B10 may be regarded as a poor prognostic factor in GC patients undergoing adjuvant chemotherapy.
SRC carcinoma is a rare subtype of gastric adenocarcinoma in which cells present a large vacuole in histology. It is different from conventional GC to some degree; for example, its higher prevalence in younger age, location in the upper third stomach, and presence of higher number of Borrmann type IV tumors. Growing evidence has demonstrated that SRC carcinoma is associated with poor survival outcome [35][36][37]. There were 59 (16.4%) patients with SRC carcinoma in our cohort, and AKR1B10 expression was still related to clinical outcomes. Patients overexpressing AKR1B10 were found to have worse 5-year DFS (36.5% vs. 64.0%, p = 0.011) and OS (50.0% vs. 71.4%, p = 0.007) rates than those with low AKR1B10 expression. However, the prognostic value of AKR1B10 in early-stage GC remains unclear. We enrolled 127 stage I GC patients, including 40 with overexpression of AKR1B10 and 87 with low AKR1B10 expression. Patients overexpressing AKR1B10 had relatively inferior 5-year DFS (67.2% versus 81.2%) and OS (68.8% versus 86.1%) rates than those with low expression of AKR1B10. However, the difference between the two groups did not differ significantly, indicating that the role of AKR1B10 may not be significant in stage I GC.
Our study had some limitations. First, because of the retrospective design and the fact that all patients were treated at a single institution which may have led to a selection bias. For example, GC patients without surgical specimens for analysis were not enrolled in our study. Second, there was a higher percentage of total gastrectomy in patients overexpressing the AKR1B10 group compared to those with a lower expression (20.5% versus 8.3%). Although total gastrectomy was a prognostic factor for worse DFS and OS in univariate analysis, it was not significant in the multivariate analysis. This indicates that AKR1B10-related poor prognosis may not be completely associated with this reason but owing to its oncogenic mechanisms. However, to the best of our knowledge, the current study constitutes the largest analysis that explores the prognostic value of AKR1B10 in GC patients who underwent gastrectomy with D2 lymph node dissection. It may provide more evidence for further basic research.

Conclusions
The results of our study confirm that overexpression of AKR1B10 may be regarded as an independent prognostic factor for GC patients with worse survival and who underwent gastrectomy with D2 lymph node dissection.