For this study, we obtained 162 immunoblots for the expression of Notch1, Snail, and p53 proteins and 110 sequences for p53 mutations in HCCs. All experiments were performed as described in the previous study.15 Briefly, rabbit monoclonal anti-Notch1 (diluted 1:100, Cell Signaling, Danvers, MA), rabbit polyclonal anti-Snail (diluted 1:100, Abcam, Cambridge, MA), and mouse monoclonal anti-p53 (diluted 1:100, Santa Cruz Biotechnology) antibodies were used to detect each protein in paraffin sections, and automated sequencing was performed to detect p53 mutations in frozen tissues (Appendix Fig. 1). Of the 162 immunoblots obtained, 147 were from patients with HCC (33 grade I, 40 grade II, and 74 grade III); their corresponding non-HCCs and 15 dysplastic nodules (DN) were used as controls. All tissue specimens, along with clinicopathological information, were obtained from Severance Hospital, Yonsei University College of Medicine (Seoul, Korea). The analyses focused on the effects of differentiation grade III HCC on postoperative prognosis, for which 47 cases were chosen on the basis of well-defined clinical records (mean age, 53±13 years; range, 28-76 years; 38 men and 9 women). Together with biochemical liver function, HBV viral markers, and serum AFP and protein induced by vitamin K absence (PIVKA) tests, all patients underwent a liver spiral computed tomography (CT) scan every 3-6 months to detect recurrent HCC after surgery. All patients had HBV infection and reserved liver function (Child-Pugh A). The reason for this selection criterion was that the differentiation grade4,5 and poor liver function7 themselves influenced the postoperative prognosis, and etiological factors could influence molecular expression patterns.16,17 For the evaluation, we kept a routine histological examination data file on differentiation grades, tumor size, tumor capsule formation (encapsulation), capsular infiltration, microvascular invasion, portal vein invasion, and satellite nodules. Differentiation was graded according to Edmondson and Steiner's criteria.18 Informed consent was obtained from each patient, and tissue collection was approved by the Institutional Review Board.

Statistical analysis was performed using R statistics software (Ver.2.12.2.; The R Foundation for Statistical Computing, Vienna, Austria), which is a free software used under the terms of the Free Software Foundation's GNU General Public License in source code form.19 The relationship between the expression of several proteins and the HCC grade was analyzed by calculating Spearman's correlation and by Fisher's exact test for count. Logistic regression analysis and a chi-square test were done to analyze the significance of the relationships among the expression patterns of the molecules. P-values of less than 0.05 were considered statistically significant.

Because HCCs are classified on the basis of cellular morphology, differentiation grades do not directly depict the malignant nature of HCC. Therefore, in comparison with grades I and II HCCs, we characterized grade III HCC in view of their relationship with other pathological features that directly indicated postoperative prognosis, such as tumor size, encapsulation status, capsular infiltration, microvascular invasion, portal vein thrombosis, and satellite nodules. As summarized in Table 1 and Figure 1, the correlation between differentiation grades and other pathological features suggests that grade III HCC was more malignant than grade I and/or II HCCs.

Of the 3 core molecules Notch1, Snail, and p53 proteins, which are components of the NSP pattern, the expression of Snail and p53 proteins were more significantly upregulated in grade III HCC than in grade I or II HCCs (P<0.01 and P<0.001, respectively), whereas Notch1 expression was not (P>0.1) (Table 1). Of the 8 NSP patterns, the proportional rates of NSP-IA1 (Notch1↑/Snail↑/p53↑), NSP-IIA1 (Notch1↔/Snail↑/p53↑), and NSP-IIB (Notch1↔/Snail↔/p53↑) patterns increased according to the differentiation grade; the NSP-IA2 (Notch1↑/Snail↑/p53↔) pattern was stationary; and the NSP-III pattern decreased (Appendix Table 1). This observation strongly suggests that the co-expression of these 3 core molecules may be regulated differently according to the differentiation grade of HCC.

In our previous study, we demonstrated that both the expression level and the mutation of p53 protein were key factors regulating the nature of co-expression of the 3 core molecules selected for this study.15 Although there was a positive correlation between differentiation grade and the total incidence of p53 upregulation by immunoblot analysis, the incidence of upregulation of wild-type p53 was not altered in the various differentiation grades, whereas the majority of p53 mutations occurred in grade III HCC (Table 1 and Appendix Table 2).

In this study cohort with grade III HCC, the p53 upregulation rate was 36.3% (16/47), but the majority of the p53 mutations (75%, 12/16) were found in the NSP-II category. The distribution of p53 mutation rates differed significantly according to the NSP pattern: 71.4% (5/7) occurred in NSP-IIA1 and 63.6% (7/11) in NSP-IIB, whereas 20% (3/15) occurred in NSP-IA1 (P<0.001) (Table 2). The p53 mutation rate was different between the NSP-I and NSP-II categories.

Of the pathological features associated with differentiation grades, tumor size, microvascular invasion, portal vein invasion, and satellite nodules did not differ according to the NSP pattern (Appendix Table 3 and 4). Interestingly, the NSP-IIB pattern (Notch1↔/Snail↔/p53↑) showed a drastic increase in the incidence of fibrous capsule formation in grade III HCC than in grade II HCC (93.8% (15/16) vs. 12.5% (1/8), respectively) (Appendix Table 5). Furthermore, both NSP-IA1 (Notch1↑/Snail↑/p53↑) and NSP-IIB patterns showed significant trends of capsular infiltration, compared to the NSP-IA2 (Notch1↑/Snail↑/p53↔) pattern with incidences of 50% (8/18), 53.3% (8/15), and 0% (0/6), respectively (P=0.002 and P=0.004, respectively) (Appendix Table 5).

On the basis of the data of pathological and molecular characterizations related to grade III HCC described above, we focused our analysis on how grade III HCC might affect postoperative recurrence. Of the 47 grade III HCC patients, 19 (41.7%) relapsed 8.6±5.9 months after surgical resection (range of DFI, 1-20.5 months). There were no significant differences in clinical variables, including age, sex, and preoperative serum AFP and PIVKA-II levels between the recurrent and non-recurrent groups (Table 3). In this cohort, only portal vein invasion was significantly correlated with recurrence (P=0.003): 42.1% (8/19) in the recurrence and 7.1% (2/28) in the non-recurrence groups (Table 3). These data suggest that, when limited to grade III HCC, portal vein invasion is the most important pathological feature for predicting the risk of postoperative recurrence.

The individual expression status of Notch1, Snail, and p53 proteins was not associated with recurrence. Moreover, p53 mutation was not a significant prognostic factor for postoperative recurrence when the analysis was restricted to the grade III HCC group (Table 3). In addition, the recurrence rates of the NSP-IA1 (Notch1↑/Snail↑/p53↑) with p53 wild dominant and NSP-IIB (Notch1↔/Snail↔/p53↑) with p53 mutant dominant were 53.3% (8/15) and 18.2% (2/11), respectively (P=0.1092) (Table 2).

In HCCs with the NSP-I pattern, the mean tumor size was not significantly different between the recurrence and non-recurrence groups (smaller than 5 cm in both groups). However, in HCCs with the NSP-IIA1 and NSP-IIB patterns, the mean tumor size was significantly larger in the recurrence group than in the non-recurrence group (Table 2). Other pathological prognostic features were not significantly different between the recurrence and non-recurrence groups according to the NSP patterns.

Twelve of 19 cases (63.2%) recurred within the liver and 7 (36.8%) recurred at extrahepatic sites, including 6 lung and 1 mediastinal lymph node metastases (Appendix Table 1). Two of 6 cases of pulmonary recurrence were accompanied by metastasis in other sites: 1 in the brain and 1 in an abdominal lymph node. There were no significant differences between the intrahepatic and extrahepatic recurrence groups in clinical parameters, including age, sex, and DFI (Table 4).

Among the pathological prognostic features, only tumor size was significantly associated with relapse sites (P=0.002): 40.5±17.3 cm in cases with intrahepatic recurrence and 70.0±10.8 cm in cases with extrahepatic recurrence (Table 4). The presence of microvascular tumor emboli, portal vein invasion, non-capsulation, or satellite nodules were not significant (Table 4). These data suggest that, of all clinicopathological parameters, only tumor size is important as an indicator for extrahepatic recurrence of grade III HCC and that the maximum cut-off size would be a diameter of 5 cm.

While the Notch1 and Snail proteins were not associated with the recurrence itself, the upregulation of Notch1 and Snail proteins were individually associated with the relapse sites (P=0.035 and P=0.017, respectively) (Table 4). In NSP pattern analysis, intrahepatic recurrence tended to be clustered in the NSP-IA1 pattern, while scattered in the non-NSP-IA1 patterns, although there was no statistical significance between both (87.5% [7/8] vs. 45.5% 5/11), respectively, P=0.1473). Extrahepatic recurrence rate of the NSP-IA1 and non-NSP-IA1 patterns was 12.5% (1/8) and 54.5% (6/11), respectively (P=0.1473) (Table 2). There was no deviation of the NSP patterns in cases with extrahepatic recurrence.

Although the NSP-IA1 pattern and tumor size showed potential as significant markers for indicating postoperative recurrence in grade III HCC, the Kaplan-Meier curve analysis showed no significant difference in the recurrence time based on NSP patterns (NSP-IA1 vs. non-IA1) and tumor size (cut-off diameter, 5 cm), as well as the relapse sites.