P48 is a predictive marker for outcome of postoperative interferon-α treatment in patients with hepatitis B virus infection-related hepatocellular carcinoma

Authors


Abstract

BACKGROUND.

Postoperative interferon-α (IFN-α) therapy improved survival in patients with hepatitis B virus (HBV)-related hepatocellular carcinoma (HCC). The identification of predictive markers of outcome will help to select patients who are most likely to benefit from treatment.

METHODS.

An immunohistochemical study of P48 was performed on specimens that were collected from patients in a randomized trial who received postoperative IFN-α therapy (Group 1; n = 80 patients) and who did not receive postoperative IFN-α therapy (Group 2; n = 75 patients). Positive P48 expression was graded as ≥20% positive cells in 1 sample.

RESULTS.

Eighty-one patients were positive for P48, and 74 patients were negative for P48. The clinicopathologic data were comparable between patients with P48-negative and P48-positive staining. Disease-free survival (DFS) and overall survival (OS) in P48-positive patients were better than that in P48-negative patients in Group 1 (DFS, P = .036; OS, P = .014), however, DFS and OS did not differ between patients with positive and negative P48 in Group 2. OS in P48-positive patients from Group 1 was better than that in patients with P48-positive patients from Group 2 (OS, P = .001) but did not differ when P48 was negative. In Group 1, the risk factors for DFS were cirrhosis and P48 staining, and the risk factors for OS were tumor differentiation and P48 staining. Receiver operating curve analysis indicated that, in the first 2 years of DFS, combined cirrhosis and P48 had good predictive accuracy; and, in the first 4 years of OS, combined tumor differentiation and P48 had good predictive accuracy.

CONCLUSIONS.

P48 was useful as a predictive marker of outcome after postoperative IFN-α treatment in patients with HBV-related HCC. Cancer 2006. © 2006 American Cancer Society.

Hepatocellular carcinoma (HCC) is one of major health problems in China, and >50% of new diagnoses worldwide are found in China.1, 2 Surgical resection can provide potentially curative outcome, but >50% of patients will develop a recurrence in the first 5 years after resection3–6 that usually results in death.7 To date, no standard postoperative adjuvant therapy has been established for this patient population.8

Interferon α (IFN-α) effectively prevents recurrence in hepatitis C virus (HCV)-related HCC by suppressing the inflammation and fibrosis caused by HCV viraemia.9 The first report of successful prevention of postoperative recurrence in patients with hepatitis B virus (HBV)-related HCC was Oon's study published in 1992.10 Results from another randomized controlled study indicated that IFN-α treatment after medical ablation of HBV-related HCC also was effective in reducing the recurrence rate,11 although the authors of that report did not explain the underlying mechanisms. From 1999 to 2004, we conducted a randomized controlled study to evaluate the effectiveness of IFN-α in the prevention of recurrence after curative resection for HBV-related HCC.12 The results indicated that 18 months of IFN-α treatment delayed recurrence and improved overall survival (OS). In our trial, 33 of 118 patients (28%) developed recurrent disease within 1 year after resection when they received IFN-α12; thus, it is important to identify predictive markers of response to IFN treatment to avoid unnecessary treatment costs and side effects.

DNA-binding protein p48-ISGFγ is a key factor in the signal-transduction pathway of IFN-α, which is stimulated by IFN-α signal and then binds to the IFN-stimulated response element (ISRE) to induce the transcription of several IFN-inducible genes. Our previous studies attributed P48 deficiency to the resistance of HCC cells to IFN-α treatment.13 In the current study, we analyzed the relation between P48 expression in HCC and the outcome of patients receiving IFN-α treatment who were enrolled in a randomized controlled trial.

MATERIALS AND METHODS

Patients and Follow-Up

From January 1999 to December 2003, 232 patients who underwent curative resection by the same surgical team for pathologically confirmed HCC at the Liver Cancer Institute (Fudan University, Shanghai, China) were enrolled into a randomized trial of postoperative IFN-α treatment to prevent recurrence. All patients were randomized into an IFN-α treatment group (5 MU 3 times weekly until recurrence was diagnosed; otherwise, for 18 months) or a control group (without any antitumor treatment until recurrence was diagnosed). In our protocol for the follow-up of patients after hepatectomy for HCC, screening tests every 2 months include ultrasonography, serum α-fetoprotein (AFP) levels, and liver function tests. If a suspicious lesion or elevated AFP levels are found, then computed tomography (CT) scans or magnetic resonance images are used to confirm the diagnosis. Bone scans are obtained only from patients with bone pain. Chest X-rays or CT scans are obtained once each year if there is no other evidence of lung metastasis. All patients are followed every 2 months in the first year and at least every 3 or 4 months afterward with regular monitoring of recurrence by serum AFP level and ultrasonography or CT studies of the liver. New lesions in the liver, lung, or lymph nodes based on typical imaging findings on CT scans, with or without elevated serum AFP levels, are diagnosed as recurrent tumor. Postoperative recurrences are managed by using the same treatment protocol. The results of this trial have been reported elsewhere.12

Because some archived paraffin blocks were too small to analyze, 80 samples (Group 1) the IFN-α group and 75 samples (Group 2) from the control group were retrieved from the Department of Pathology at Zhongshan Hospital. All samples were from patients who followed the treatment protocol.

Tissue Microarray and Immunohistochemistry

After screening hematoxylin and eosin-stained slides for optimal tumor content, we constructed tissue microarray slides (Shanghai Biochip Company, Ltd., Shanghai, China). Two cores were taken from each formalin-fixed, paraffin-embedded HCC samples by using punch cores that measured 0.8 mm in greatest dimension from the center of tumor foci. Immunohistochemistry for P48 was performed by using the avidin-biotin complex method (ABC; Vector Laboratories, Burlingame, CA), including heat-induced antigen-retrieval procedures. Incubation with polyclonal antibodies against P48 (P48: C20; 1:100 dilution; Santa Cruz Biotechnology, Santa Cruz, CA) was carried out at 4°C for 18 hours.14 Negative controls were treated identically but with the primary antibody omitted.

Scoring of P48 Immunohistochemistry

P48 immunoreactivity was evaluated independently by 3 pathologists who were blinded to patient outcomes. The percentage of positive tumor cells was determined by each observer, and the average of 3 scores was calculated. We randomly selected 10 high-power fields (magnification, × 400; 100 cells per high-power field) and counted 1000 cells in each core. When the mean of percentage of P48-positive cells is close to 0% or 100%, the standard deviation (SD) is close to 0; and, when the mean is approximately 50%, the SD is approximately 5%. Thus, the SD does not increase with the mean. In this study, P48 expression was graded as follows: tumor with positive cytoplasmic staining in <20% of tumor cells was graded negative, and tumor with positive cytoplasmic staining in ≥20% of tumor cells was graded positive.

Statistical Analysis

Continuous variables were expressed as the means ± SD and were compared between groups by using the Student t test. Categorical variables were compared by using the chi-square test. Survival curves were computed with the Kaplan–Meier method and were compared between groups by using the log-rank test. To elucidate factors that may have contributed to improved survival results, an analysis was carried out to identify the prognostic factors for disease-free survival (DFS) and OS. Univariate analyses and multivariate analyses with the Cox proportional hazards model were performed on 13 factors that had potential prognostic significance, all of which were categorized as binary variables. These included 4 clinical factors (age <60 years >60 years, gender, hepatitis B e-antigen status, and serum AFP level ≤20 ng/mL or >20 ng/mL), and 9 pathologic factors (cirrhotic or noncirrhotic, tumor size ≤5 cm or >5 cm, single or multiple tumor nodules; well encapsulated or poorly encapsulation tumor; Edmondson Grade 1/2 or Grade 3/4; the presence or absence of microvascular invasion, pathologic TNM Stage I/II or Stage III or Cancer of the Liver Italian Program Stage 0 or 1/2,15 and P48 staining). The ROC analysis was used to assess the predictive accuracy of prognostic factors using the free software developed by Metz at the University of Chicago (available at URL: http://xray.bsd.uchicago.edu/krl/roc_soft.htm). All statistical analyses were completed by using statistical software (SPSS Inc., Chicago, IL). Statistical significance was defined as a P value < .05.

RESULTS

Clinicopathologic Data

Positive P48 expression was observed in 81 patients, and negative P48 expression was observed in 74 patients (Fig. 1A). P48 was stained mainly in the cytoplasm of HCC and occasionally in the nucleus of HCC and bile duct (Fig. 1B–D). There was no statistically significant difference in clinicopathologic data between patients with positive and negative P48 expression except for gender (Table 1). There was no statistically significant difference in Group 1 and Group 2 between patients with positive and negative P48 expression (data not shown). In Group 1, 27 patients died, including 21 patients (77.8%) who died of tumor recurrence, 1 patient who died of liver failure, 4 patients who died of cerebral hemorrhage, and 1 patient who died of myocardial infarction.

Figure 1.

These photomicrographs illustrate the immunohistochemistry of P48. (A) Overview of the tissue array. (B) Positive staining for P48 is observed in the cytoplasm of hepatocellular carcinoma cells. (C) Negative staining for P48 is shown. (D) Positive staining for P48 is shown. Original magnification × 400 (B); × 40 (C, D).

Table 1. Comparison of Clinicopathologic Profiles between Negative and Positive P48 Expression in all Patients
Clinicopathologic dataNo. of patients (%)OR95% CIP
P48-Negative (n = 74)P48-Positive (n = 81)
  • OR indicates odds ratio; 95% CI, 95% confidence interval; HBeAg: hepatitis B e-antigen; AFP, α-fetoprotein; pTNM, pathologic tumor, lymph node, metastasis classification; CLIP, Cancer of the Liver Italian Program.

  • *

    Chi-square test.

Age, y*
 ≤6060 (81)65 (80)1.10. 5–2.3.896
 >6014 (19)16 (20)
Gender*
 Male62 (84)76 (94)0.30.1–1.0.046
 Female12 (16)5 (6)
HbeAg*
 Negative62 (84)64 (79)1.40.6–3.1.447
 Positive12 (16)17 (21)
Cirrhosis*
 Absent8 (11)13 (16)0.60.2–1.6.341
 Present66 (89)68 (84)
AFP*
 ≤20 ng/mL25 (34)35 (43)0.70.3–1.3.229
 >20 ng/mL49 (66)46 (57)
Tumor size*
 ≤5 cm47 (64)58 (72)0.70.4–1.4.282
 >5 cm27 (36)23 (28)
No. of tumor nodules*
 Single64 (86)69 (85)1.10.4–2.8.817
 Multiple10 (14)12 (15)
Tumor capsule*
 Well differentiated34 (46)44 (54)1.40.7–2.6.298
 Poorly differentiated40 (54)37 (46)
Microvascular invasion*
 Negative55 (74)68 (84)0.60.3–1.2.139
 Positive19 (26)13 (16)
Edmondson grade*
 Low (Grade 1/2)47 (64)61 (75)1.80.9–3.5.111
 High (Grade3/4)27 (36)20 (25)
pTNM stage*
 Low (Stage I/II)53 (72)66 (81)0.60.3–1.2.146
 High (Stage III)21 (28)15 (19)
CLIP
 Low (Stage)38 (51)51 (63)0.10.3–1.2.621
 High (Stage 1/2)36 (49)30 (37)

Survival

In Group 1, DFS was significantly better in patients who had positive P48 expression compared with patients who had negative P48 expression (hazards ratio, 1.8; 95% confidence interval [95% CI], 1.0–3.2 [P = .036]). The cumulative 1-year, 3-year, and 5-year DFS rates were 84%, 55%, and 36%, respectively, in patients who had positive P48 expression and 53%, 34%, and 29%, respectively, in patients who had negative P48 expression. In Group 2, however, there was no significant difference in DFS between patients with positive and negative staining (hazards ratio, 1.4; 95% CI, 0.8–2.6 [P = .208]).

In Group 1, OS was significantly better in patients who had positive P48 expression compared with patients who had negative P48 expression (hazards ratio, 2.7; 95% CI, 1.2–6.2 [P = .014]). Again, in Group 2, there was no significant difference in OS between patients who had positive and negative P48 expression (hazards ratio, 1.2; 95% CI, 0.6–2.3 [P = .545]).

When we analyzed the patients with positive P48 expression, the results indicated that the OS of patients in Group 1 was better compared with the OS in Group 2 (hazards ratio, 3.9; 95% CI, 1.7–9.2 [P = .001]). The mean OS was 57.3 months (95% CI, 53.0–61.5 months) in Group 1 and 41.5 months (95% CI, 34.0–49.0 months) in Group 2. The cumulative 1-year, 3-year, and 5-year OS rates were 100%, 93%, and 69%, respectively, in Group 1 and 84%, 63%, and 46%, respectively, in Group 2. However, there was no significant difference between the groups for the patients who had negative P48 expression (hazards ratio, 1.5; 95% CI, 0.8–2.9 [P = .246]).

Prognostic Factors for DFS in Group 1

Univariate analysis of prognostic factors for DFS indicated that 2 of 13 evaluated factors had significant prognostic influence (Table 2). After multivariate analysis, only cirrhosis (present vs. absent: risk ratio [RR], 4.9; 95% CI, 1.4–16.9 [P = .012]) and the pattern of P48 staining (negative vs. positive: RR, 2.0; 95% CI, 1.1–3.8 [P = .028]) were significant prognostic factors for DFS (Table 3). Receiver operating curve (ROC) analysis indicated that these 2 prognostic factors had good predictive accuracy for 1-year DFS (area under the concentration-time curve (AUC) = 0.649; 95% CI, 0.5–0.8 [P < .01]) and acceptable accuracy in the first 2 years of DFS (AUC = 0.604; 95% CI, 0.5–0.7 [P = .05]) (Fig. 2).

Figure 2.

Receiver operating characteristic analysis showed that P48 staining combined with cirrhosis (marker A) had good predictive value for disease-free survival (DFS) in the first 2 postoperative years. The area under the concentration time curve (AUC) was 0.649 (P < .01) at 1 year and 0.604 (P = .05) at the first and second postoperative year, but the AUC was poor in later years (P < .05). T = 1, DFS for the 1st postoperative year; T = 2, DFS for the 2nd postoperative year; T = 3, DFS for the 3rd postoperative year; T = 4, DFS for the 4th postoperative year; T = 5, DFS for the 5th postoperative year. TPF indicates true-positive fraction; FPF, false-positive fraction. a: P48 staining; b: cirrhosis; AZ: AUC.

Table 2. Univariate Analysis of Significant Prognostic Factors for Disease-Free Survival in the Interferon-α Treatment Group
VariableNo. of patientsMean survival, monthsP
  1. HbeAg indicates hepatitis B e-antigen; AFP, α-fetoprotein; pTNM, pathologic tumor, lymph node, metastasis classification; CLIP, Cancer of the Liver Italian Program.

Age, y
 ≤606223.4.620
 >601823.2
Gender
 Male7524.2.785
 Female523.2
HbeAg
 Negative6524.2.186
 Positive1514.1
Cirrhosis
 Absent1353.5.000
 Present6728
AFP
 ≤20 ng/mL3724.2.775
 >20 ng/mL4323
Tumor size
 ≤5 cm5724.2.917
 >5 cm2319.8
No. of tumor nodules
 Single6832.1.098
 Multiple1211.6
Tumor capsule
 Good4424.2.480
 Poor3623.4
Microvessel invasion
 Absent6723.4.608
 Present1321.5
P48 staining
 Negative3726.6.011
 Positive4338.3
Edmondson grade
 Low (Grade 1/2)6135.6.025
 High (Grade 3/4)1924.0
pTNM
 Low (Stage I/II)6632.1.709
 High (Stage III)1414.1
CLIP
 Low (Stage 0)5223.4.618
 High (Stage 1/2)2823.0
Table 3. Multivariate Analysis of Significant Prognostic Factors for Disease-Free Survival in the Interferon-α Treatment Group
VariableRR95% CIP
  1. RR indicates risk ratio; 95% CI, 95% confidence interval.

Cirrhosis (present)4.91.4–16.9.012
P48 staining (negative)2.01.1–3.8.028

Prognostic Factors for OS in Group 1

Univariate analysis of prognostic factors for OS indicated that 3 of 12 evaluated factors had significant prognostic influence (Table 4). After multivariate analysis, only Edmondson stage (high grade [Grade 3 or 4] vs. low grade [Grade 1 or 2]; RR, 5.6; 95% CI, 2.0–15.7 [P = .001]) and pattern of P48 staining (negative vs. positive: RR, 3.8; 95%CI, 1.4–10.3 [P = .008]) were significant prognostic factors for OS (Table 5). ROC analysis showed that these 2 prognostic factors had impressive predictive accuracy for OS in the first 4 years: For 1-year, 2-year, 3-year, and 4-year survival, the AUCs were 0.883 (P < .01), 0.737 (P < .01), 0.646 (P = .01), and 0.634 (P = .03), respectively (Fig. 3).

Figure 3.

Receiver operating curve analysis showed that P48 staining combined with Edmondson grade (marker B) had impressive prediction values for overall survival (OS) for the first 4 postoperative years. For 1-year, 2-year, 3-year, and 4-year survival, the areas under the concentration time curve (AUCs) were 0.883 (P < .01), 0.737 (P < .01), 0.646 (P = .01), and 0.634 (P = .03), respectively, but the AUC was poor in the 5th year. T = 1, OS for the 1st postoperative year; T = 2, disease-free survival (DFS) for the 2nd postoperative year; T = 3, DFS for the 3rd postoperative year; T = 4, DFS for the 4th postoperative year; T = 5, DFS for the 5th postoperative year. TPF indicates true-positive fraction; FPF, false-positive fraction. a: P48 staining; b: Edmonson grade; AZ: AUC.

Table 4. Univariate Analysis of Significant Prognostic Factors for Overall Survival in the Interferon-α Treatment Group
VariableNo. of patientsMean survival, monthsP
  1. HbeAg indicates hepatitis B e-antigen; AFP, α-fetoprotein; pTNM, pathologic tumor, lymph node, metastasis classification; CLIP, Cancer of the Liver Italian Program.

Age, y
 ≤606249.3.408
 >601856.4
Gender
 Male7550.7.817
 Female538.5
HBeAg
 Negative6553.6.009
 Positive1535.6
Cirrhosis
 Absent1364.3.032
 Present6747.9
AFP
 ≤20 ng/mL3753.1.373
 >20 ng/mL4348.5
Tumor size
 ≤5 cm5753.2.154
 >5 cm2344.5
No. of tumor nodules
 Single6851.3.812
 Multiple1247.1
Tumor capsule
 Good4452.1.686
 Poor3649.4
Microvessel invasion
 Absent6751.6.642
 Present1346.6
P48 staining
 Negative3743.2.014
 Positive43)57.3
Edmonson grade
 Low (Grade 1/2)6154.8.003
 High (Grade 3/4)1937.8
pTNM
 Low (Stage I/II)6652.390
 High (Stage III)1444.5
CLIP
 Low (Stage 0)5254.8.051
 High (Stage1/2)2843.2
Table 5. Multivariate Analysis of Significant Prognostic Factors for Overall Survival in the Interferon-α Treatment Group
VariableRR95% CIP
  1. RR indicates risk ratio; 95% CI, 95% confidence interval.

Edmondson grade (high)5.62.0–15.7.001
P48 staining (negative)3.81.4–10.3.008

DISCUSSION

It has been reported that a gene expression profile and IFN-α/type 2 IFN receptor (IFNAR2) may predict the response of patients with HCC to the combination treatment of IFN-α and 5-fluoroucil.16, 17 A decrease in P48 expression has been associated with a poor response to IFN-α in patients with melanoma and transitional cell carcinoma.18, 19 Other molecular markers, such as suppressor of cytokine signaling 1 (SOCS-1) and SOCS-3, reportedly act as negative feedback regulators of Janus kinase (Jak)/signal transducer and activator of transcription (STAT) and determine sensitivity to IFN.20, 21 However, to date, there have been no reports on predictive markers for outcome after postoperative IFN-α treatment in patients with HBV-related HCC.

A wealth of information is available on the molecular processes that underlay IFN-α-induced signaling. Binding of IFN-α to Type I receptor brings the receptor-associated tyrosine kinases Tyk2 and Jak1 in close proximity, which allows the kinases to phosphorylate, leading to their activation and the generation of receptor docking sites for the latent cytoplasmic protein STAT.22, 23 Tyrosine-phosphorylated STAT1 and STAT2 also associate as heterodimers to form the IFN-stimulated gene factor 3 (ISGF3α) and translocate to the nucleus, where they associate with the DNA-binding protein ISGF3γ (P48) to form the ISGF3 and then bind to the IFN-stimulated response element (ISRE) to induce transcription of several IFN-inducible genes.24 In our previous study, we demonstrated that the transduction of P48 in resistant HCC cells can restore sensitivity to IFN-α treatment, which further confirms the importance of Jak/STAT signaling in the antiproliferation effects of IFN-α in HCC cells,13 a finding that also has been supported by others.25 Therefore, we believe that P48 expression in tumor cells is related to the effect of IFN-α.

In the current study, we observed that DFS and OS were similar in patients with positive and negative P48 expression who did not receive postoperative IFN-α; however, DFS and OS differed significantly between patients with positive and negative P48 expression who did receive postoperative IFN-α. Furthermore, OS was better in patients with positive P48 expression who received IFN-α than in patients with positive P48 expression who did not receive IFN-α. These data clearly demonstrate that P48 expression was associated with outcome after postoperative IFN-α treatment in our patients. The results also indicated that, in patients who had poorly differentiated HCC and/or the presence of cirrhosis, the outcomes may be less optimal. We noted that the predictive accuracy of P48 expression for DFS and OS diminished over time. A possible explanation for this may be that IFN-α treatment was given for 18 months to patients in the treatment group, and P48 expression was used to predict the effect of IFN-α treatment; thus, the predictive accuracy of P48 expression was good in the first several years and diminished after IFN-α treatment was stopped. Similar to what we observed in our clinical trial,12 18 months of IFN-α treatment delayed tumor recurrence but did not reduce the incidence of recurrence during 5 years of observation, because a number of patients developed recurrent tumors after IFN-α treatment was stopped; however, OS within 5 years was improved by IFN-α treatment because recurrences were delayed in the treatment group.

Although most patients who had tumors with positive staining for P48 were long-term survivors, approximately 16% of them were diagnosed with recurrent disease within 1 year. An explanation for this may be that the expression of P48 is not sufficient for antiproliferation in Jak/STAT signaling. In a previous study, it was observed that a lack of stat1 expression also lead to resistance to the growth-inhibitory effect of IFN-α in CTCL cells, a human melanoma and a breast cancer cell line.18, 26–29 In addition to antiproliferation, IFN-α has other antitumor mechanisms, such as immunomodulation and antiangiogenesis.30 In our previous study, we demonstrated that the antiangiogenesis mechanism involved another signal-transduction pathway despite a deficiency of P48 in HCC cells.31

In conclusion, the results of the current study indicate that P48 assessed immunohistochemically in surgical samples from patients with HCC may be a useful marker in for prediction of outcome after postoperative IFN-α therapy. Postoperative IFN-α treatment should be given based on tumor P48 expression status, the presence of cirrhosis, and Edmondson grade to provide maximal benefit and avoid the toxicities of IFN-α.

Ancillary