Concordance of human equilibrative nucleoside transporter‐1 expressions between murine (10D7G2) and rabbit (SP120) antibodies and association with clinical outcomes of adjuvant chemotherapy for pancreatic cancer: A collaborative study from the JASPAC 01 trial

Abstract Background Expression of human equilibrative nucleoside transporter‐1 (hENT1) is reported to predict survival of gemcitabine (GEM)‐treated patients. However, predictive values of immunohistochemical hENT1 expression may differ according to the antibodies, 10D7G2 and SP120. Aim We aimed to investigate the concordance of immunohistochemical hENT1 expression between the two antibodies and prognosis. Methods The subjects of this study were totally 332 whose formalin‐fixed paraffin‐embedded specimens and/or unstained sections were obtained. The individual H‐scores and four classifications according to the staining intensity were applied for the evaluation of hENT1 expression by 10D7G2 and SP120, respectively. Results The highest concordance rate (79.8%) was obtained when the cut‐off between high and low hENT1 expression using SP120 was set between moderate and strong. There were no correlations of hENT1 mRNA level with H‐score (p = .258). Although the hENT1 mRNA level was significantly different among four classifications using SP120 (p = .011), there was no linear relationship among them. Multivariate analyses showed that adjuvant GEM was a significant predictor of the patients with low hENT1 expression using either 10D7G2 (Hazard ratio [HR] 2.39, p = .001) or SP120 (HR 1.84, p < .001). In contrast, agent for adjuvant chemotherapy was not significant predictor for the patients with high hENT1 expression regardless of the kind of antibody. Conclusion The present study suggests that the two antibodies for evaluating hENT1 expression are equivalent depending on the cut‐off point and suggests that S‐1 is the first choice of adjuvant chemotherapy for pancreatic cancer with low hENT1 expression, whereas either S‐1 or GEM can be introduced for the pancreatic cancer with high hENT1 expression, no matter which antibody is used.

(p = .258). Although the hENT1 mRNA level was significantly different among four classifications using SP120 (p = .011), there was no linear relationship among them.
Multivariate analyses showed that adjuvant GEM was a significant predictor of the patients with low hENT1 expression using either 10D7G2 (Hazard ratio [HR] 2.39, p = .001) or SP120 (HR 1.84, p < .001). In contrast, agent for adjuvant chemotherapy was not significant predictor for the patients with high hENT1 expression regardless of the kind of antibody.

Conclusion:
The present study suggests that the two antibodies for evaluating hENT1 expression are equivalent depending on the cut-off point and suggests that S-1 is the first choice of adjuvant chemotherapy for pancreatic cancer with low hENT1 expression, whereas either S-1 or GEM can be introduced for the pancreatic cancer with high hENT1 expression, no matter which antibody is used.
10D7G2, gemcitabine, human equilibrative nucleoside transporter-1, pancreatic cancer, SP120 1 | INTRODUCTION Gemcitabine (GEM) is a key drug of pancreatic cancer (PC). [1][2][3][4] Expression of human equilibrative nucleoside transporter-1 (hENT1) has been reported to be related with sensitivity to GEM in several cancer types including PC. [5][6][7] High immunohistochemistry (IHC) expression of hENT1 in tumor tissue (hENT1 High ) is associated with better survival benefit from adjuvant GEM. [8][9][10][11][12][13][14][15] However, most of these results were based on the IHC using murine 10D7G2 monoclonal anti-hENT1 antibody (Ab), which is not commercially available. [8][9][10][11][12][13][14] Alternatively, the SP120 rabbit monoclonal anti-hENT1 Ab has been developed and used to evaluate hENT1 expression, and three studies did not find consistent association between IHC expression of hENT1 using the SP120 Ab and prognosis in PC patients treated with GEM. [15][16][17] Moreover, three studies comparing the 10D7G2 and SP120 Abs have been reported. 14,18,19 One study found that hENT1 expression, as evaluated by IHC using tissue microarray (TMA), matched between the two Abs in two thirds of cases, and concluded that both Abs could predict prognosis in patients who received GEM as adjuvant chemotherapy. 19 In contrast, the other study found that IHC hENT1 expression matched in half of cases, and concluded that only the 10D7G2 Ab was useful for predicting prognosis. 14,18 However, both studies used specimens collected over ≥10 years, 14,18,19 and during such a long study period their outcomes have been affected by changes in treatment strategies after introduction of adjuvant chemotherapy. 20 Previously, we reported on the association of IHC expression of hENT1 using SP120 Ab with overall survival in patients enrolled into the Japan Adjuvant Study Group of Pancreatic Cancer (JASPAC) 01 study which randomized 377 pancreatic cancer patients to receive either GEM or S-1 after curative resection. 20,21 Interestingly, in the S-1 arm, the median overall survival in patients with high hENT1 expression was significantly shorter than the patients with low hENT1 expression. 21 Thus, predictive values of hENT1 IHC expression may differ according to the Abs, 10D7G2 and SP120. Furthermore, there may be some patients for whom GEM would show equivalent or better efficacy than S-1, despite the results of the JASPAC 01 study showing significantly better prognosis of S-1 than GEM in overall.
In the present study, we evaluated the concordance of hENT1 expression between 10D7G2 and SP120 Abs, and explored their relation to hENT1 mRNA level and survival in subsets of patients enrolled into the JASPAC 01 study. 20,21 2 | MATERIALS AND METHODS

| Study population and design
This biomarker study was designed as a collaborative study of the JASPAC 01 study 21 after the completing its final analysis. The protocol of the present study was approved by the ethics committee of the Shizuoka Cancer Center (No. 27-22-27-1-5) and the institutional review board of each participating institution. From totally 332 of all 377 patients enrolled in the JASPAC 01 study at the 24 participating institutions, we collected the unstained sections for IHC using the SP120 Ab (n = 326: 86.5%) and for measuring mRNA level of hENT1, and/or formalin-fixed, paraffin-embedded (FFPE) specimen blocks for TMA using the 10D7G2 Ab (n = 114: 30.2%).
Distribution of patients whose samples were evaluable for IHC, TMA, and measurable for hENT1 mRNA level is shown in Figure 1.
The concordance between the two Abs was investigated in 89 patients for whom both IHC and TMA were evaluable (Figure 1

| TMA with the 10D7G2 Ab
The University of Liverpool UK prepared TMAs from the FFPE specimen blocks of the 114 patients and were blindly evaluated for hENT1 expression using the 10D7G2 Ab, as previously described. 11 The cutoff value of H-score between high hENT1 expression (10D7G2 High ) and low hENT1 expression (10D7G2 Low ) was determined by the minimum p value approach in survival analysis.

| IHC with the SP120 Ab
Expression of hENT1 IHC using the SP120 Ab was evaluated on unstained sections (anti-hENT1 rabbit monoclonal Ab SP120, Roche Tissue Diagnostics Co, Ltd, Basel, Switzerland; already diluted Ab) as previously described. 21

| Reverse transcription-polymerase chain reaction
Representative hematoxylin and eosin-stained sections (10 μm-thick) were reviewed by a pathologist who marked out cancer predominant areas, which were removed by macro-dissection. RNA was extracted from the removed tumor tissue with the RNeasy FFPE Kit (Qiagen, Chatsworth, CA), and cDNA was prepared using a High Capacity cDNA Reverse Transcription Kit with RNase Inhibitor (Life Technologies, Foster City, CA) according to the manufacturer's protocol.
Expression of the hENT1 mRNA was measured using a TaqMan realtime PCR (Life Technologies, Foster City, CA), as described previously. 22 Expression of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA was quantified in each sample and used to F I G U R E 1 Distribution of patients for three methods assessing hENT1 standardize the data. We measured the cycle threshold (Ct) value, which is inversely proportional to the amount of cDNA. Analysis was performed in triplicate for all samples. The expression level of each sample was shown as the value of each target divided by the GAPDH value.

| Statistical analysis
Continuous variables are presented as median and range, and were compared using the Mann-Whitney U test or Kruskal Wallis test, as appropriate. Categorical variables were compared using the chi-square test or Fisher's exact test, as appropriate. The cut-off value of hENT1 mRNA levels was explored using receiver operating characteristic curves for the overall survival and Youden's index. Overall survival rates were estimated using the Kaplan-Meier method and compared by the log-rank test. The Cox proportional hazards model was used for univariate and multivariate analyses, and treatment arms (GEM or S-1) and all factors found to be significant predictors of overall survival (p < .10) in univariate analysis were entered into multivariate analysis.
All statistical analyses were performed using the SPSS 24.0 software package (SPSS, Inc., Chicago, IL). p < .05 (two-tailed) was considered significant.

| Patient characteristics
The clinical and pathological characteristics of the 332 patients ( Figure 1(A)-(F)) in the present study was shown in Table 1. The median overall survival time after randomization was 2.16 years with GEM and 3.76 years with S-1 ( Figure S2(A)). The hazard ratio (HR) for mortality of S-1, compared with GEM, was 0.60 (95% confidence interval [CI] 0.46-0.78, p < .001). The median relapse-free survival time after randomization was 1.07 years with GEM and 1.88 years with S-1 ( Figure S2(B)).
3.2 | Concordance of hENT1 expression between TMA with 10D7G2 and IHC with SP120 H-score for 10D7G2-assessed hENT1 expression was 100 (0-186), and the cut-off value of H-score was determined to be 135 by the minimum p value approach in survival analysis. There were 16 patients (18%) whose H-scores were ≥135 (10D7G2 High ), and IHC expression of hENT1 using SP120 were strong in 17, moderate in 31, weak in 31, and absent in 10 patients (Table 2).
Correlation between IHC expression of hENT1 using SP120 and H-score (high/low) by TMA using 10D7G2 is shown in  categorization was obtained when the cut-off between high and low IHC expression using SP120 (SP120 Low and SP120 High ) was set between moderate and strong (concordance rate: 79.8%). When the cut-off between SP120 Low and SP120 High was set between weak and moderate, the concordance rate was 58.4%. The cut-off set between absent and weak showed a concordance rate as low as 32.6%.
3.3 | Relationship between hENT1 mRNA level and hENT1 expression with TMA by 10D7G2 (H-score) and with IHC by SP120 Relationship between hENT1 mRNA level and H-score was investigated in the 84 patients (Figure 1(C)). There were no significant associations between them (Pearson's correlation coefficient: 0.125; p = .258; Figure S3(A)) and between categorical 10D7G2 hENT1 expression (10D7G2 High /10D7G2 Low ) and hENT1 mRNA level (p = .350). Wallis test, p = .011; Figure S3(B)), there was no linear relationship between them.

| Overall survival by 10D7G2 hENT1 expression
Among the 95 patients whose H-score by TMA with 10D7G2 Ab were evaluable (Figure 1(A)-(C)), we compared the clinical and pathological characteristics between the patients with 10D7G2 High and 10D7G2 Low (Table S1). Although not significant, the proportions with lymph node metastasis and with low CA19-9 level were slightly higher in 10D7G2 High patients in both treatment groups.

| Overall survival by SP120 hENT1 expression
Among the 326 patients whose IHC with SP120 were evaluable ( Figure 1(A),(C)-(E)), the median survival times of patients treated with GEM who showed absent, weak, moderate and strong IHC with SP120 were 2.09, 2.19, 2.08 and 2.13 years, respectively ( Figure S4 (A)). There were no significant differences among the four groups according to IHC with SP120. Among patients treated with GEM, those with SP120 High showed an equivalent survival compared with those with SP120 Low (median 2.13 vs. 2.16 years, HR [SP120 High ] 0.78, 95% CI 0.42-1.45, p = .429, Figure 2(C)).
The median survival times of patients treated with S-1 who showed absent, weak, moderate and strong IHC with SP120 were: "not reached" due to 5-year overall survival rate >50%, 4.49, 2.81 and 2.56 years, respectively ( Figure S4(B)). As we previously reported, 20 there was a significant difference in overall survival between patients with SP120 Low and SP120 High when the cut-off was set between absent/weak and moderate/strong, and multivariate analysis showed that SP120 High (HR 1.61, 95% CI 1.06-2.45, p = .027) was one of significant predictors for the survival of the patients who received S-1 as adjuvant chemotherapy. However, when the cut-off was set between absent/weak/moderate and strong, which showed the highest concordance with TMA with 10D7G2 Ab ( showed that SP120 High whose cut-off set between moderate and strong was marginally not a significant prognostic factor of patients treated with S-1 (Table 4).

| hENT1 mRNA expression and overall survival
Among 310 patients whose hENT1 mRNA levels were available  Figure S5 (A)) and 0.531 for predicting 3-year overall survival ( Figure S5(B)).
Because these AUCs were lower than 0.7, it was difficult to determine the optimal cut-off values of the hENT1 mRNA levels for predicting survival.

| Treatment selection according to hENT1 expression
Significance of hENT1 expression with 10D7G2 for selecting adjuvant chemotherapy with GEM or S-1 was investigated in the 95 patients ( Figure 1(A)-(C)), and that with SP120, with the cut-off set between moderate and strong (the highest concordance rate between 10D7G2 Ab and SP120 Ab), was done in 326 patients (Figure 1(A)-(E)).
Similarly to the overall population in this study ( Figure S2(A)), the overall survival of the S-1-treated groups was significantly better than that of the GEM-treated group both in 10D7G2 Low (n = 79, Figure S6 (A)) and SP120 Low (n = 295, Figure S6(C)) subgroups, and multivariate analyses showed that adjuvant chemotherapy with GEM was a significant predictor for poor survival of the patients in both subgroups (Table 5).
In contrast, the overall survival between the GEM-and S-1-treated groups was not significantly different either in 10D7G2 High (n = 16, Figure S6(B)) and SP120 High (n = 31, Figure S6(D)) subgroups.
Moreover, multivariate analysis showed that agents for adjuvant chemotherapy, GEM or S-1, was not a significant predictor for the prognosis of the patients with high hENT1 expression regardless of the used Ab, either 10D7G2 (n = 16) or SP120 (n = 31) ( Table 5).

| DISCUSSION
The patients' clinical and pathological factors in the present study was not much different from those in the JASPAC 01 study, 20 and the all subjects of this study recapitulated the clinical outcomes in JASPAC 01 study. Moreover, the freshness of the specimens in this study is much better than those of the earlier three studies, 14,18,19 as the present study used specimens from a randomized controlled trial performed during short period (3 years). From these facts, it is considered that the present study have a certain degree of quality.
The present study showed no correlations between the hENT1 mRNA level and hENT1 expressions using either 10D7G2 or SP120 Ab and suggests that the concordance rate between the two Abs for evaluating hENT1 expression depends on the cut-off point set among four groups according to IHC staining with SP120 Ab. However, there were several problems in the studies focusing on this issue. First, the platform between the two Abs, IHC and TMA, was different. Second, even among studies that used 10D7G2 Ab, some studies evaluated hENT1 expression with H-scores, 11,18,23 and other studies categorized samples into three or four groups according to tumor staining intensity. 8,10,13,14,19,23 The two Abs should be compared based on the same assessment methods.
H-scores are obtained as continuous values, and cut-off values for high and low expression, to predict sensitivity to GEM, were different among studies. In earlier studies which used the median Hscore as their cut-off value, median H-scores (48 11 and 90 18 ) varied even using same Ab (10D7G2). The median H-score of the present study was 100, which was higher than that of the earlier studies, 11,18 and the cut-off value of H-scores (135)  On the other hands, almost all studies using SP120 Ab [14][15][16][17]19,23 evaluated hENT1 expression by classifying into 2-4 groups according to tumor staining intensity except one study. 18 The positive rates of hENT1 expression by SP120 Ab varied from 21% to 72%, similarly to those by 10D7G2 Ab. [14][15][16][17][18][19]23 Compared to earlier studies, the positive rate of hENT1 expression with SP120 Ab in the present study was extremely low (11%) when the cut-off between SP120 Low and SP120 High was set between moderate and strong, which showed the highest concordance with TMA with 10D7G2 Ab. However, with cutoff set between weak and moderate, the positive rate will be compatible to those of previous studies. Therefore, the utility of hENT1 expression in predicting survival will differ depending on the cut-off value regardless the platform and Abs for its evaluation.
We used the different measurement platforms for hENT1 expression with two Abs (TMA for 10D7G2 and IHC for SP120) in the present study. Expression of hENT1 as assessed by both 10D7G2 and SP120 Abs was matched in 79.8% when the cut-off was set between moderate and strong, leading to a low positive rate. The concordance rate was higher than those of earlier studies (59.7%, 69.1% and 50.7%). 14,18,19 However, the concordance rate in this study decreased when the cut-off was set between weak and moderate, leading to a higher positive rate. Many of studies that examined specimens from large-scale clinical trials used TMAs with small cores from the FFPE specimens. [8][9][10][11][14][15][16][17][18][19]