Contrast-enhanced intraoperative ultrasonography for vascular imaging of hepatocellular carcinoma: Clinical and biological significance


  • Potential conflict of interest: Nothing to report.


Abnormal tumor vascularity is one of the typical features of hepatocellular carcinoma (HCC). In this study, the significance of contrast-enhanced intraoperative ultrasonography (CEIOUS) images of HCC vasculature was evaluated by clinicopathological and gene expression analyses. We enrolled 82 patients who underwent curative hepatic resection for HCC with CEIOUS. Clinicopathological and gene expression analyses were performed according to CEIOUS vasculature patterns. CEIOUS images of HCC vasculatures were classified as reticular HCC or thunderbolt HCC. Thunderbolt HCC was significantly correlated with higher alpha-fetoprotein levels, tumor size, histological differentiation, portal vein invasion, and tumor-node-metastasis stage, and these patients demonstrated a significantly poorer prognosis for both recurrence-free survival (P = 0.0193) and overall survival (P = 0.0362) compared with patients who had reticular HCC. Gene expression analysis revealed that a rereplication inhibitor geminin was significantly overexpressed in thunderbolt HCCs (P = 0.00326). In vitro knockdown of geminin gene reduced significantly the proliferation of human HCC cells. Immunohistochemical analysis confirmed overexpression of geminin protein in thunderbolt HCC (P < 0.0001). Multivariate analysis revealed geminin expression to be an independent factor in predicting poor survival in HCC patients (P = 0.0170). Conclusion: CEIOUS vascular patterns were distinctly identifiable by gene expression profiling associated with cellular proliferation of HCC and were significantly related to HCC progression and poor prognosis. These findings might be clinically useful as a determinant factor in the postoperative treatment of HCC. (HEPATOLOGY 2013)

Hepatocellular carcinoma (HCC) is the fifth most common malignancy and one of the most common causes of cancer-related death in the world.1, 2 Surgical resection is considered the primary curative therapy in the treatment of HCC.3-5 During hepatic resection, intraoperative ultrasonography (IOUS) of the liver is used as an aid for surgical navigation. IOUS provides crucial diagnostic and staging information to the surgeon during the procedure.6 Recently, contrast-enhanced ultrasonography techniques using microbubble agents have been developed.7 Among these agents, Sonazoid (gaseous perflubutane; Daiichi-Sankyo, Tokyo, Japan) is a unique ultrasound contrast agent that is accumulated in Kupffer cells.8-12 We reported recently that Kupffer imaging of Sonazoid with contrast-enhanced intraoperative ultrasonography (CEIOUS) is quite useful for detailed detection of tumors in real time during hepatic resection.13

Tumor angiogenesis is one of the critical features in determining overgrowth and metastatic potential.14, 15 In contrast to normal vessels, tumor vessels are tortuous, excessively branched, and short-circuited. In this manner, tumor vasculature is highly disorganized.14-18 HCC is a tumor that is typically known to exhibit angiogenesis.2, 3, 18-23 In particular, a dramatic alteration in arterial hypervascularity is observed in moderately and poorly differentiated HCC.18, 20, 21 Such hypervascularity can be observed using angiography and contrast-enhanced computed tomography,21, 24, 25 but it is quite difficult to analyze the detailed intratumoral vasculature in real time.

Recently, it has been reported that contrast-enhanced ultrasonography can be used to evaluate tumor vasculature similarly to what is seen with other contrast-enhanced radiological imaging techniques.22, 23, 34 In the present study, the HCC vasculature was analyzed in detail by CEIOUS with Sonazoid to identify the specific patterns associated with clinicopathological features. Additionally, genome-wide gene expression was assessed via DNA microarray analysis, which offers a systematic approach to acquire comprehensive information regarding gene transcription profiles.26 Such studies could lead not only to the identification of unique biomarkers but also to the development of a novel molecular-targeted therapy for HCC.27-30 The present study demonstrates the evidence indicating the biological and clinical significance of CEIOUS microflow imaging (MFI).


AFP, alpha-fetoprotein; CEIOUS, contrast-enhanced intraoperative ultrasonography; FC, fold change; HCC, hepatocellular carcinoma; IOUS, intraoperative ultrasonography; MFI, micro-flow imaging; PBS, phosphate-buffered saline; preRC, prereplication complex; siRNA, small interfering RNA; TNM, tumor-node-metastasis.

Patients and Methods

Patients and Samples.

We enrolled patients who underwent curative hepatic resection for HCC at the Tokyo Medical and Dental University Hospital between August 2007 and March 2010. From a total of 167 patients with HCC, 135 patients underwent CEIOUS of the main tumor during hepatic resection. Among them, 82 patients were technically eligible for MFI analysis. The other 53 patients were technically ineligible for MFI mainly because of pretreatment with sorafenib, radiofrequency ablation, and transcatheter arterial chemoembolization. The baseline characteristics of the enrolled patients are summarized in Table 1. Written informed consent was obtained from the patients, and the institutional review board approved the study. The preoperative evaluations have been described elsewhere.31 Resected tissue containing no necrosis was divided into two specimens immediately after surgery: one was snap-frozen in liquid nitrogen and stored at −80°C for microarray analysis; the other was fixed in 10% formaldehyde solution and embedded in paraffin for histopathological analysis. According to The General Rules for the Clinical and Pathological Study of Primary Liver Cancer,32 histopathological analysis was performed. To confirm the expression patterns detected by the microarrays, the median follow-up period was 784 days (interquartile range, 497-1,015).

Table 1. Patient Characteristics
  1. HBV, hepatitis B virus, HCV, hepatitis C virus; PIVKA-II, protein induced by vitamin K absence or antagonist II; PT%, prothrombin time.

Age, years, mean ± SD (range)70 ± 9.0 (34-84)
Sex, no., male:female62:20
Viral infection, no., HBV:HCV:non-B/C13:46:23
Background liver pathology, no. 
 Chronic hepatitis or liver fibrosis33
 Liver cirrhosis43
Child-Pugh classification, no., A:B78:4
Albumin, mg/dL, mean ± SD4.1 ± 0.4
Total bilirubin, mg/dL, mean ± SD0.90 ± 0.40
PT%, mean ± SD82.2 ± 8.0
AFP, ng/mL, mean ± SE5,751 ± 4,832
PIVKA-II, mAU/mL, mean ± SE7,355 ± 3,575
Tumor size, cm, mean ± SD4.2 ± 3.4 (1-25)
Tumor number, no., solitary:multiple49:33
TNM stage, I:II:III:IV5:28:32:17

Analysis of MFI via CEIOUS.

The ultrasound system Xario-XG (Toshiba Medical) was used for all IOUS and CEIOUS procedures with a 7-MHz, T-shaped linear probe (PLT-705BTH; Toshiba Medical). CEIOUS was performed with pulse inversion harmonic imaging capability. The mechanical index was set at 0.15 in all CEIOUS procedures. The acoustic power was altered to keep the mechanical index at 0.15 because the depth of the focus point varied. During CEIOUS procedures, a real-time fundamental mode image was displayed simultaneously with a pulse inversion harmonic image side-by-side for reference. Thus, the target lesions were not missed even if they were difficult to recognize in the pulse inversion harmonic imaging. During the operation, the liver was mobilized off the diaphragm for improved sonographic visualization. IOUS was then performed in a systematic fashion in baseline fundamental mode scan to confirm the preoperative tumor staging. There was no lesion discovered in the preoperative staging found during IOUS. After IOUS, Sonazoid was injected at a dose of 0.5 mL/body in approximately 1 second through a catheter inserted in the antecubital vein, followed by a 10-mL normal saline flush. Following injection, a dynamic CEIOUS was performed with the focus depth beyond the main tumor. The main tumor was observed continuously for approximately 1 minute from the time of injection (vascular phase). The arterial phase was timed for 45 seconds after completion of the flash, after which the portal venous phase was timed from 45 to 70 seconds after injection. We performed MFI after observation of the portal venous phase. For MFI analysis, the combination of flash replenishment sequence and the maximum intensity holding sequence was expected to make it possible to visualize the tumor vasculature clearly with high special resolution and vascular continuity.23 The accumulation time for each MFI sequence was 10-15 seconds, depending on the perfusion of the target tissue. As the standardized procedure, we applied three-time flashes on the main tumor and confirmed the MFI pattern based on the dominant image during CEIOUS. At approximately 10-15 minutes after injection, ultrasonic observation was resumed using pulse inversion harmonic imaging in the systematic liver (Kupffer phase). Hypoechoic lesions were searched for in hyperechoic surrounding liver with accumulated microbubbles. The focus point was set at the bottom of the liver. In the case of new lesions detected, we performed defect reperfusion imaging with an additional injection of Sonazoid (0.5 mL/body). The observation in the Kupffer phase was not repeated after the second injection because all focal liver lesions were theoretically examined by a thorough scan in the first Kupffer phase. We removed all lesions diagnosed as HCC by each modality under conditions that ensured safety. The IOUS and CEIOUS scans and image analysis were performed in consensus reading by surgeons with 8, 10, 20, and 40 years of experience in liver surgery.

DNA Microarray Analysis.

For the gene expression analysis, at least three sections of the largest nodule were used from the largest cross-section of the main tumor. Total RNA was extracted from the HCC specimens with an RNeasy kit (Qiagen, Hilden, Germany). The integrity of the RNA obtained was assessed with an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). Among the HCC tumors limited to ≤4 cm in diameter to exclude the bias of tumor size, 27 samples (11 reticular HCC, 16 thunderbolt HCC) were available for analysis of gene expression. In the 27 samples, the mean size of the tumor was 2.9 ± 0.2 cm. Contaminating DNA was removed by digestion with RNase-free DNase (Qiagen), and with 2 μg of total RNA, complementary RNA was prepared with a one-cycle target labeling and control reagents kit (Affeymetrix, Santa Clara, CA). The hybridization and signal detection of the Human Genome U133 (HG-U133) Plus 2.0 arrays (Affymetrix) were performed in accordance with the manufacturer's instructions. A total of 37,743 microarray data sets were normalized by the robust multiarray average method (R2.4.1 statistical software together with the BioConductor package), essentially as described in our previous report.31 The estimated gene expression levels were log 2 transformed, and the fold change (FC) values were calculated using ratios of genometric means of gene expression levels between two MFI patterns. A Wilcoxon rank sum test was performed to estimate the significance levels of the differences in gene expression between the two groups. For each statistical test, the obtained P values from the multiple hypothetical testing were adjusted by a false discovery rate, and probe sets with a false discovery rate <0.47 were considered for further analysis. Hierarchical clustering with the selected probe sets was performed using a plete linkage method. For visualization, the expression levels were standardized as z scores (mean = 0, variance = 1) for each probe set.

Cell Culture.

The human HCC cell lines SK-Hep1, Hep3B, and PLC/PRF/5 were obtained from American Type Culture Collection (Manassas, VA). The other human HCC cell lines Huh1, Huh7, HLE, HLF, and HepG2 were obtained from the Human Science Research Resources Bank (Osaka, Japan). The culture media used were Roswell Park Memorial Institute 1640 medium (SK-Hep1, Hep3B, Huh7, and HepG2) and Dulbecco's modified Eagle medium (PLC/PRF/5, Huh1, HLE, and HLF) supplemented with 5% fetal bovine serum for the HLF cells or 10% fetal bovine serum for the remaining cell lines. All media were supplemented with 100 U/mL of penicillin and 100 μg/mL of streptomycin. All cell lines were cultivated in a humidified incubator at 37°C in 5 % carbon dioxide and were collected with 0.25% trypsin-0.03% ethylene diamine tetraacetic acid.

Western Blotting and Immunocytochemistry.

Geminin protein expression in the cell lines was detected via western blotting analysis. The total protein was extracted from each cell line as described.33 The protein levels of geminin and α-tubulin (control) were detected via standard western blot analysis by using 8-15% sodium dodecyl sulfate–polyacrylamide gel electrophoresis. The blots were incubated overnight at 4°C with the primary antibody, anti-human geminin (1:200; Santa Cruz Biotechnology, Santa Cruz, CA; catalog #sc-13015), and then at room temperature for 1 hour with anti–α-tubulin (1:5,000; Sigma-Aldrich, St.Louis, MO; catalog #T9026). The appropriate secondary antibodies were added for 2 hours, and the protein expression was visualized with enhanced chemiluminescence by the ECL western blot testing detection system (GE Healthcare, Buckinghamshire, UK). The immunocytochemical analysis was performed with cultured cells on glass slides coated with saline. The cells were fixed in phosphate-buffered saline (PBS)-10% trichloroacetic acid for 15 minutes, permeabilized in PBS-0.2% Triton X-100 for 5 minutes, and then blocked in PBS-3% bovine serum albumin for the immunocytochemical detection of geminin. The primary antibody (Santa Cruz Biotechnology) was used at 1:50 dilution, and γ-tubulin antibody (Sigma-Aldrich; catalog #T6557) was used at 1:1,000. The secondary antibody for geminin was the Alexa Fluor 568 fragment of a donkey anti-rabbit immunoglobulin G (H+L) and antibody for γ-tubulin was the Alexa Fluor 488 fragment of a donkey anti-mouse immunoglobulin G (H+L) (Invitrogen, Carsbad, CA). The DNA was counterstained with 4′,6-dia-midino-2-phenylindeole. Image acquisition was performed on a confocal microscope (Axio Observer.ZL, Carl Zeiss Microimaging GmbH, Germany).

Gene Silencing.

The knockdown of geminin was performed by using small interfering RNA (siRNA) (Invitrogen; catalog #1299003) and negative control siRNA duplexes (Invitrogen; catalog #12935112). HLF, SK-Hep1, and Hep3B cells were seeded at a density of 1.0 × 105 cells into 6-well plates in 2,000 μL of culture medium with 5% fetal bovine serum for the HLF cells or 10% fetal bovine serum for Hep3B and SK-Hep1 (for western blot testing, cell proliferation analysis, cell cycle analysis). Thereafter, transfection with the siRNA was performed by using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. After transfection, the cells were incubated for 96 hours at 37°C in a 5% carbon dioxide atmosphere. At the 0-, 24-, 48-, 72-, and 96-hour time points after siRNA transfection, cells were detached from each plate. The number of viable cells was counted by an automatic cell counting machine (CYTORECON; GE Healthcare) according to the manufacturer's instructions. The number of nonviable cells was assessed using CYTORECON and trypan blue dye exclusion. These experiments were independently evaluated in triplicate for cell proliferation analysis. And fluorescence-activated cell sorting for cell cycle analysis was done on cells that were collected by trypsinization on each point after siRNA transfection cells, fixed with 70% ethanol overnight at 4°C. Cells were rehydrated in PBS and then resuspended in PBS containing 100 μg/mL RNase (Sigma) and 10 μg/mL propidium iodide. Cellular DNA content was analyzed with a FACSCaliber flow cytometer (Becton Dickinson Biosciences, San Joes, CA) using Cellquest software.

Immunohistochemical Analysis.

Immunohistochemical analysis was performed on HCC tissue samples. The primary antibodies were used at the following concentrations diluted in PBS containing 1% bovine serum albumin: geminin (1:500; Santa Cruz Biotechnology), Ki67 (1:100; Abcam, Cambridge, UK; catalog #ab833), EpCAM (1:3,000; AbD Serotec, Oxford, UK; catalog #MCA1870G), CK19 (1:100; Dako, Glostrup, Denmark; catalog #M088801), and c-KIT (Ventana XT System; Ventana, Tucson, AZ; catalog 790-2951). The tissue sections were stained by an automated immunostainer (Ventana XT System) using heat-induced epitope retrieval and a standard DAB detection kit (Ventana). The immunostaining was evaluated quantitatively by counting at least 500 cells in three different random fields (magnification ×100) under a light microscope by three independent investigators (Sato, Tanaka, and Arii). The mean value was calculated for the final result of each case.

Statistical Analysis.

Statistical comparisons of clinicopathological characteristics for significance were performed using a χ2 test or Fisher's exact test with a single degree of freedom, and a Student t test was used to analyze the differences between continuous values. Overall survival and recurrence rates were determined by the Kaplan-Meier method, and for comparisons, log-rank tests were used. P values less then 0.05 were considered to have statistical significance. To investigate those factors that predicted overall survival, multivariate analyses were performed using Cox proportional hazard models and logistic regression models. All statistical analyses were performed using SPSS version 17.0 (SPSS, Chicago, IL).


Classification of MFI Patterns and Postoperative Outcomes of Patients with HCC.

A total of 82 patients who underwent CEIOUS examination during hepatic resection of HCC were analyzed. According to MFI data, tumor vasculatures were classified as two characteristic patterns: a thin, ramified pattern (reticular HCC) and a thick, linear pattern (thunderbolt HCC), although such a difference could not be detected preoperatively using other imaging modalities (Supporting Table 1). In the 82 cases, 26 were classified as reticular HCC, and the remaining 56 cases were classified as thunderbolt HCC (Fig. 1a,b). The clinicopathological significance of the MFI patterns was then evaluated. Univariate analysis revealed that thunderbolt HCC correlated significantly with high levels of alpha-fetoprotein (AFP) (P = 0.0003), tumor size (P = 0.0048), histological dedifferentiation (P = 0.0004), infiltrative growth type (P = 0.0215), portal vein invasion (P < 0.0001), and tumor-node-metastasis (TNM) stage (P < 0.0001) (Table 2).

Figure 1.

Classification of CEIOUS MFI patterns with HCC. Using the MFI data, the HCC vasculatures were classified into two characteristic types; (a) reticular HCC and (b) thunderbolt HCC. While reticular HCC was gradually enhanced to the entire tumor with thin and ramified vessels (B-D), thunderbolt HCC was enhanced in a part of tumor with thick and linear vessels (F-H). The fundamental mode of ultrasonography is shown (A, E). (c) Recurrence-free survival and (d) overall survival of patients with HCC according to the MFI patterns. A log-rank test demonstrated statistically significant differences in recurrence-free and overall survival rates (P = 0.0193 and P = 0.0362, respectively). (e) Recurrence-free survival and (f) overall survival curves of the patients with HCC ≤4 cm. A log-rank test demonstrated statistically significant differences in the recurrence-free survival rates (P = 0.0434).

Table 2. Clinicopathological Findings in Patients with HCC in Relation to MFI Patterns of CEIOUS
Clinicopathological FactorMFI PatternP
Reticular HCC (n = 26)Thunderbolt HCC (n = 56)
  1. Abbreviations: HBV, hepatitis B virus; HCV, hepatitis C virus; PIVKA-II, protein induced by vitamin K absence or antagosists II.

Age, years, mean ± SD69.5 ± 8.065.8 ± 9.30.0868
Sex, no., male:female20:642:140.8503
Viral infection, no., HBV:HCV:non-B/C4:12:109:34:130.3416
Background liver pathology, no.  0.5341
 Chronic hepatitis or liver fibrosis1122 
 Liver cirrhosis1231 
Child-Pugh classification, no.  0.7676
Albumin, mg/dL, mean ± SD4.1 ± 0.44.0 ± 0.50.6408
Total bilirubin, mg/dL, mean ± SD0.82 ± 0.400.94 ± 0.400.2060
PT%, mean ± SD83.3 ± 6.281.7 ± 8.60.3899
AFP, no., ≥20 ng/mL  0.0003
PIVKA-II, no., ≥40 mAU/mL  0.6779
Tumor size, cm, mean ± SD2.6 ± 0.94.9 ± 3.90.0048
Tumor number, no., solitary:multiple17:932:240.4788
Histological differentiation, no.  0.0004
Tumor morphology, no.  0.1981
 Simple nodular type1324 
 Simple nodular type with  extranodular growth913 
 Confluent multinodular type419 
Tumor growth type, no.  0.0215
 Expansive growth2646 
 Infiltrative growth010 
Portal vein invasion, no.  <0.0001
Hepatic vein invasion, no.  0.1680
Bile duct invasion, no.  0.1623
Capsule formation, no.  0.7087
Cancerous infiltration of the capsule, no.  0.7632
Formation of fibrous septum within the tumor, no.  0.3594
TNM stage, no., I/II:III/IV20:613:43<0.0001

The recurrence-free survival and overall survival rates were then compared between the two groups. The thunderbolt HCC group demonstrated a significantly poorer prognosis than the reticular HCC group for both recurrence-free survival (P = 0.0193) and overall survival (P = 0.0362) (Fig. 1c,d). In order to exclude the bias of tumor size, we analyzed the significance of MFI patterns by limiting the tumor size to ≤4 cm in diameter. As a result, patients with thunderbolt HCC demonstrated a poorer recurrence-free survival than those with reticular HCC (Fig. 1e,f). As shown in Table 3, thunderbolt HCC also correlated significantly with high levels of AFP, histological dedifferentiation, portal vein invasion, and TNM stage, even if the tumor size was ≤4 cm.

Table 3. Clinicopathological Findings in Patients with HCC in Relation to MFI Patterns of CEIOUS (Tumor Size Limited to ≤4 cm)
Clinicopathological FactorMFI PatternP
Reticular HCC (n = 26)Thunderbolt HCC (n = 32)
  1. Abbreviations: HBV, hepatitis B virus; HCV, hepatitis C virus; PIVKA-II, protein induced by vitamin K absence or antagosists II.

Age, years, mean ± SD69.5 ± 8.067.5 ± 7.90.3548
Sex, no., male:female20:622:100.4886
Viral infection, no., HBV:HCV:non-B/C4:12:103:21:80.3291
Background liver pathology, no.  0.2809
 Chronic hepatitis or liver fibrosis1112 
 Liver cirrhosis1219 
Child-Pugh classification, no.  0.4086
Albumin, mg/dL, mean ± SD4.1 ± 0.44.1 ± 0.50.8200
Total bilirubin, mg/dL, mean ± SD0.82 ± 0.400.97 ± 0.420.1853
PT%, mean ± SD83.3 ± 6.281.0 ± 8.40.2383
AFP, no., ≥20 ng/mL  0.0009
PIVKA-II, no., ≥40 mAU/mL  0.5592
Tumor size, cm, mean ± SD2.6 ± 0.92.9 ± 0.80.3004
Tumor number, no., solitary:multiple17:921:110.9847
Histological differentiation, no.  0.0014
Tumor morphology, no.  0.5256
 Simple nodular type1319 
 Simple nodular type with  extranodular growth95 
 Confluent multinodular type48 
Tumor growth type, no.  0.2217
 Expansive growth2628 
 Infiltrative growth04 
Portal vein invasion, no.  <0.0001
Hepatic vein invasion, no.  0.2245
Bile duct invasion, no.  0.3632
Capsule formation, no.  0.6249
Cancerous infiltration of  the capsule, no.  0.8128
Formation of fibrous septum  within the tumor, no.  0.4417
TNM stage, no., I/II:III/IV20:610:220.0005

Genome-wide Gene Expression Analysis Correlated to the MFI Patterns of Human HCC.

Gene expression was analyzed in 27 samples of HCC, including 11 reticular HCCs and 16 thunderbolt HCCs. The gene expression changes in 72 probe sets were evaluated by microarray analysis in 27 HCCs (FC >2, P < 0.01). As shown in Fig. 2a, the hierarchical clustering clearly divided the samples, and 14 genes were up-regulated in thunderbolt HCC. As displayed in Fig. 2b, in order of the FC and P values, a rereplication inhibitor geminin was significantly overexpressed in thunderbolt HCCs.

Figure 2.

Molecular and biological analysis in relation to the MFI patterns of HCC. (a) Gene expression profiling and MFI patterns. The hierarchical clustering of the gene expression (FC >2 and P < 0.01) in reticular HCC (blue bars) and thunderbolt HCC (orange bars). Dendrograms show the classification determined by the hierarchical clustering analysis. The red and green areas indicate relative overexpression and underexpression, respectively. (b) The differentially expressed genes in thunderbolt HCC by ranking of FC and P value (FC > 2, P < 0.01). The profiling identified geminin as one of the dominant molecules significantly overexpressed in cancer tissue of thunderbolt HCC. (c) Immunocytochemical analysis of geminin in human HCC cell lines (original magnification ×200). The predominant expression was recognized in the nucleus. DAPI, 4′,6-diamidino-2-phenylindeole. (d) The silencing effects of the geminin gene. Western blot analysis confirmed that the geminin expression was markedly suppressed by the specific siRNA, compared with control siRNA (siNegative) in HLF, SK-Hep1, and Hep3B human HCC cells. (e) Growth curve analysis up after siRNA transfection. The silenced HLF, SK-Hep1 and Hep3B cells showed a significant reduction in cell proliferation. *P < 0.01 versus control. The results are presented as the mean ± SD from triplicate experiments. (f) Cell cycle analyses. The percentage of <2N DNA content cells was increased in all cell lines after transfection of geminin siRNA.

Significance of Geminin Expression in Human HCC Cells.

The expression of the geminin protein was then analyzed in human HCC cell lines. The protein expression was recognized by western blotting in all of eight cell lines examined. Among them, we selected the three cell lines according to the expression level of geminin (HLF, high; SK-Hep1, moderate; Hep3B, low). Immunocytochemical analysis in these cell lines showed the potent expression of geminin protein, mainly in the nucleus (Fig. 2c). Next, the effects of the specific siRNAs against geminin were assessed in the HCC cell lines (HLF, SK-Hep1, and Hep3B). Western blot analysis certified that the expression of geminin was markedly silenced by the specific siRNA, but not by the control siRNA (Fig. 2d). As shown in the cell growth assays of HCC (Fig. 2e), geminin siRNA significantly depressed cellular proliferation 96 hours after transfection, compared with cells treated with control siRNA (P < 0.01). In addition, cell cycle analysis demonstrated the accumulation of <2N DNA content in all of the cell lines after transfection of geminin siRNA (Fig. 2f). These results suggest that geminin has an essential role in the cell cycling of human HCC.

MFI Pattern of HCC Tumors and the Expression of Geminin Protein.

Immunohistochemical validation of geminin protein was then performed using 82 tissue samples of HCC. As a shown in Fig. 3, the geminin protein stained mainly in the nucleus was clearly detected in cancer cells of HCC, but not in noncancerous liver tissue. The frequency of geminin-positive cancer cells was 9.3% in thunderbolt HCCs, but only 3.1% in reticular HCCs, indicating the statistical significance between the two types (P < 0.0001). The statistical significance was also observed even if the tumor size was limited (Supporting Table 3). The geminin expression and MFI patterns were not associated with cholangiocellular differentiation as well as progenitor cell markers (Supporting Fig. 1). According to the multivariate analysis for the MFI patterns of HCC using logistic regression mode, the expression of geminin protein (P = 0.0180), tumor size (P = 0.0222), and portal vein invasion (P = 0.0174) were statistically independent factors of MFI thunderbolt HCC (Table 4).

Figure 3.

Immunohistochemical analysis of expression of geminin protein in cancer tissues (C) and noncancerous tissues (NC) of human HCC. Compared with the cancer tissue of reticular HCC (a), the expression of geminin protein was increased in the cancer tissue of thunderbolt HCC (b). In noncancerous tissue of HCC, the expression of geminin was not detected. (c) Mean percentages of geminin-positive cells in cancer tissues (upper) and the case numbers of each type HCC (lower).

Table 4. Multivariate Analysis for Independent Predictors of MFI Patterns of HCC
Clinicopathological FactorHR (95% CI)P
  1. CI, confidence interval; Eg, expansive growth; HR, hazard ratio; Ig, infiltrative growth; NA, not applicable.

AFP, ≥20 ng/mL3.276 (0.464-23.145)0.2342
Tumor size, cm3.704 (1.206-11.378)0.0222
Histological differentiation  
 Well/moderate:poor6.082 (0.669-55.320)0.1090
Tumor growth type  
Portal vein invasion  
 Negative:Positive43.968 (1.945-993.935)0.0174
TNM stage (I/II:III/IV)4.341 (0.287-65.582)0.2892
Geminin1.289 (1.044-1.590)0.0180

Postoperative Outcomes of HCC Patients and the Expression of Geminin Protein.

The clinicopathological significance of geminin expression was then evaluated in the 82 patients enrolled in this study (Table 5). The univariate analysis for overall survival of HCC revealed that the expression of geminin protein (P = 0.0003), serum level of albumin (P = 0.0187) and AFP (P = 0.0071), tumor number (P = 0.0242), histological dedifferentiation (P = 0.0140), infiltrative growth type (P = 0.0093), hepatic vein invasion (P = 0.0030), and TNM stage (P = 0.0074) were significantly for overall survival of HCC. The multivariate analysis revealed that the expression of geminin protein (P = 0.0170) and hepatic vein invasion (P = 0.0113) were statistically independent factors of overall survival.

Table 5. Univariate and Multivariate Analysis of Overall Survival
Clinicopathological FactorUnivariate AnalysisMultivariate Analysis
HR (95% CI)PHR (95% CI)P
  1. CI, confidence interval; CM, confluent multinodular type; Eg, expansive growth; HBV, hepatitis B virus; HCV, hepatitis C virus; HR, hazard ratio; Ig, infiltrative growth; PIVKA-II, protein induced by vitamin K absence or antagosists II; SN, simple nodular type; SNEG, simple nodular type with extranodular growth.

Age, years1.026 (0.972-1.083)0.3745  
Sex, male versus female1.333 (0.441-4.034)0.6107  
Viral infection 0.8216  
 HBV1.814 (0.408-8.076)   
 HCV0.456 (0.168-1.297)   
 Non-B/C1.703 (0.556-5.124)   
Background liver pathology 0.5585  
 Normal0.635 (0.145-2.780)   
 Chronic hepatitis or liver fibrosis0.949 (0.371-2.430)   
 Liver cirrhosis1.218 (0.488-3.040)   
Child-Pugh classification, A versus B2.246 (0.516-9.784)0.2810  
Albumin0.355 (0.150-0.842)0.01870.619 (0.213-1.804)0.3797
PT%0.969 (0.921-1.019)0.2205  
Total bilirubin0.894 (0.293-2.728)0.8446  
AFP, ng/mL, ≥20 versus <204.562 (1.511-13.771)0.00711.780 (0.484-6.555)0.3856
PIVKA-II, mAU/mL, ≥40 versus < 401.856 (0.668-5.160)0.2357  
Tumor size, cm1.035 (0.978-1.157)0.1510  
Tumor number, solitary versus multiple2.930 (1.151-7.462)0.02422.254 (0.607-8.371)0.2247
Histological differentiation, well/moderate versus poor3.327 (1.276-8.677)0.01401.137 (0.303-4.259)0.8492
Tumor morphology, SN/SNEG versus CM2.178 (0.895-5.299)0.0863  
Tumor growth type, Eg versus Ig3.636 (1.374-9.624)0.00932.588 (0.664-10.088)0.1706
Portal vein invasion2.183 (0.869-5.481)0.0965  
Hepatic vein invasion4.140 (1.621-10.575)0.00305.039 (1.441-17.627)0.0113
Bile duct invasion2.103 (0.273-16.217)0.4757  
Capsule formation1.571 (0.168-3.996)0.3427  
Cancerous infiltration of the capsule1.455 (0.584-3.624)0.4205  
Formation of fibrous septum within the tumor1.251 (0.415-3.778)0.6907  
TNM stage, I/II versus III/IV3.392 (1.129-10.186)0.02951.332 (0.255-6.965)0.7339
Geminin1.120 (1.053-1.192)0.00031.098 (1.017-1.185)0.0170


Tumor angiogenesis is one of the special features of HCC progression, and the detection of abnormal vasculature in a focal liver lesion aids in the identification of HCC.14, 15, 18-21 Previous studies have reported that contrast-enhanced ultrasonography using microbubble-based contrast agents is useful for visualizing the vasculature of HCC.22, 23 In particular, the difference in vasculature of HCC and surrounding liver could be evaluated in detail using MFI with contrast agents.23, 34 Although some studies reported that CEIOUS was useful for detection of focal HCC lesions during hepatic resection, the significance of CEIOUS findings of MFI has not yet been clarified.35, 36

In this study, we investigated the MFI with Sonazoid CEIOUS in hepatic resection for HCC. The MFI pattern was classified as reticular HCC and thunderbolt HCC, and as shown in Table 2 and Fig. 1c,d, thunderbolt HCC was significantly associated with the advanced progression and poor prognosis. We also revealed that the thunderbolt pattern was useful to predict the microinvasion into portal veins, which were undetectable by preoperative imaging modalities (Table 2). A subgroup analysis limiting the tumor size to ≤4 cm also demonstrated advanced progression and poorer prognosis in thunderbolt HCC compared with reticular HCC (Fig. 1e,f and Table 3). Additionally, we found the clinicopathological differences between each MFI pattern even when the tumor size was limited to <3 cm or <5 cm (Supporting Table 2).

To clarify the molecular and biological features associated with the MFI patterns, the gene expression profiling was further evaluated in human HCC samples (Fig. 2a). The gene expression analysis identified geminin as one of the dominant molecules significantly overexpressed in thunderbolt HCC tissue (Fig. 2b). Geminin is known as a DNA rereplication inhibitor that is expressed in the G2, S, and M phases but not in the G0 and G1 phases.37, 38 DNA replication begins with assembly of prereplication complexes (preRCs) at multiple sites throughout the genome as cells exit metaphase.39 preRCs are then assembled into preinitiation complexes that are subsequently activated by protein kinases to begin DNA synthesis (S phase). Once S phase begins, further assembly of preRCs is prevented by phosphorylation, ubiquitination, and degradation of preRC proteins ORC1, CDC6, and CDT1, and by geminin, a specific protein inhibitor of CDT1 activity unique to metazoan.40, 41

Previous studies have reported that geminin is specifically expressed in proliferating cells, including lymphocytes, crypt epithelial cells, and sperm cells, but not in nonproliferating epithelial cells, including normal neurocytes, muscle cells, and hepatocytes.42, 43 As for malignant neoplasms, the overexpression of geminin was reported in advanced human cancers such as breast, lung, renal, and colorectal carcinomas.41, 44-48 According to our additional analysis, Ki67 expression was significantly associated with geminin expression (P = 0.0112), supporting the proliferative role of geminin in HCC (Supporting Fig. 2). Indeed, in vitro knockdown of geminin significantly suppressed the cell proliferation of human HCC cell lines (Fig. 2d-f).

Our immunohistochemical analysis on clinical samples clarified that the frequency of geminin-positive cancer cells correlated significantly with thunderbolt HCCs (Fig. 3 and Supporting Table 3). In addition, the clinical significance of geminin was noted in the overall survival (Table 5). Quaglia et al.49 reported that the expression level of geminin was increased from regenerative and dysplastic nodules to HCC nodules, indicating its potential association with hepatocarcinogenesis. The molecular and biological role of geminin in HCC progression should be studied further.

In conclusion, the intraoperative MFI patterns were independent predictors of HCC progression, resulting in poor prognosis of the patients. Such vascular patterns were distinctly identified by gene expression profiling, such as geminin, which might play roles in HCC cell proliferation. Application of CEIOUS might be useful in determining the postoperative treatment of HCC. Additional studies should clarify the clinical impacts of CEIOUS MFI patterns.