Human hepatocellular carcinomas with “Stemness”-related marker expression: keratin 19 expression and a poor prognosis


  • Potential conflict of interest: Nothing to report.

  • This work was supported by a Korea Science and Engineering Foundation grant (KOSEF) (20100008075), a National Research Foundation grant (R13-2002-054-03004-0) funded by the Korean government (MOST; to Y.N.P.), and grants from the SNUBH Research Fund (11-2009-008) and Korean Health Technology R&D Project, Ministry for Health, Welfare and Family Affairs, Republic of Korea (A101397; to H.K.).


There is a recently proposed subtype of hepatocellular carcinoma (HCC) that is histologically similar to usual HCC, but characterized by the expression of “stemness”-related markers. A large-scale study on two different cohorts of HCCs was performed to investigate the clinicopathologic features and epithelial-mesenchymal transition (EMT)-related protein expression status of this subtype of HCCs. The expression status of stemness-related (e.g., keratin 19 [K19], cluster of differentiation [CD]133, epithelial cell adhesion molecule [EpCAM], and c-kit) and EMT-related markers (e.g., snail, S100A4, urokinase plasminogen activator receptor [uPAR], ezrin, vimentin, E-cadherin, and matrix metalloproteinase [MMP]2) were examined using tissue microarrays from cohort 1 HCCs (n = 137). K19 protein expression in cohort 2 HCCs (n = 237) was correlated with the clinicopathologic parameters and messenger RNA (mRNA) levels of K19, uPAR, VIL2, Snail, Slug, and Twist. K19, EpCAM, c-kit, and CD133 positivity were observed in 18.2%, 35.0%, 34.3%, and 24.8%, respectively. K19 was most frequently expressed in combination with at least one other stemness-related marker (92.0%). K19-positive HCCs demonstrated more frequent major vessel invasion and increased tumor size, compared to K19-negative HCCs (P < 0.05). K19 was most significantly associated with EMT-related protein expression (e.g., vimentin, S100A4, uPAR, and ezrin) (P < 0.05) and a poor prognosis (overall survival: P = 0.018; disease-free survival: P = 0.007) in cohort 1. In cohort 2, HCCs with high K19 mRNA levels demonstrated higher mRNA levels of Snail, uPAR, and MMP2 (P < 0.05). K19-positive HCCs demonstrated more frequent microvascular invasion, fibrous stroma, and less tumor-capsule formation, compared to K19-negative HCCs (P < 0.05). K19 expression was a significant independent predictive factor of poor disease-free survival (P = 0.032). Conclusion: K19 was well correlated with clinicopathologic features of tumor aggressiveness, compared to other stemness-related proteins. K19-positive HCCs showed significantly increased EMT-related protein and mRNA expression, suggesting that they may acquire more invasive characteristics, compared to K19-negative HCCs through the up-regulation of EMT-associated genes. (HEPATOLOGY 2011;)

Cancer stem cells have the ability to self-renew, differentiate, and proliferate, have greater tumorigenicity and chemoresistance, and have been associated with a poor prognosis in several human malignancies.1, 2 They have also been identified in hepatocellular carcinoma (HCC): Among the HCCs with the conventional histomorphological features, there is a recently proposed subtype characterized by the expression of “stemness”-related markers, such as keratin 19 (K19), cluster of differentiation (CD)133, and epithelial cell adhesion molecule (EpCAM), and is associated with a poorer prognosis, compared to usual HCCs without these markers.3

The poor prognosis of stemness-related marker expressing HCCs may partly be attributed to the relationship between this subset of HCC with invasion and metastasis-related gene expression. Up-regulation of invasion and metastasis-related genes, such as VIL2 (encoding ezrin), PLAUR (uPAR: urokinase plasminogen activator receptor), and CD44, was demonstrated in HCCs with progenitor cell features (i.e., “hepatoblast”-like subtype) in a gene-expression profiling study,3 and up-regulated invasion and epithelial-mesenchymal transition (EMT)-related genes was demonstrated in CD133-expressing HCCs.4 Moreover, an association between high expression levels of stemness-related markers in HCCs and tumor angiogenesis was recently reported.5

EMT is a critical part of the tumor-cell invasion process and results in the loss of epithelial characteristics and the acquisition of mesenchymal features, which include the expression of vimentin, N-cadherin, S100A4, metalloproteinases, snail, and twist.6 Loss of normal epithelial intercellular contact is an important feature and is demonstrated by the loss of membranous E-cadherin expression in the tumor cells. The expression of E-cadherin is controlled by transcription factors, such as snail and twist, which bind to consensus E-box sequences in the E-cadherin gene promoter.7 In addition, E-cadherin may also undergo cleavage by matrix metalloproteinases (MMPs), resulting in the down-regulation of E-cadherin-mediated intercellular adhesion. MMPs have been shown to play important roles in tumor invasion, and MMP2 overexpression in HCC has been associated with a poor prognosis.8 VIL2 and its protein, ezrin, interact with E-cadherin and have been implicated in the invasiveness and metastasis of HCCs.9 Ezrin overexpression has been associated with K19 expression10 and poor prognosis.10, 11 An association between uPAR and EMT has also been demonstrated in breast cancer cells,12 and interactions between urokinase plasminogen activator (uPA) and uPAR have been observed in the invasion and metastasis of various tumors, including HCCs.13

In this study, we analyzed the expression of four of the most commonly used stemness-related markers (e.g., K19, CD133, EpCAM, and c-kit) in HCCs and correlated the results with the clinicopathologic characteristics. In addition, the biological features of human HCCs expressing stemness-related markers were analyzed with various EMT and invasion-associated proteins.


AFP, alpha-fetoprotein; ALT, alanine aminotransferase; AST, aspartate aminotransferase; CD, cluster of differentiation; cDNA, complementary DNA; EGFR, epidermal growth factor receptor; EMT, epithelial-mesenchymal transition; EpCAM, epithelial cell adhesion molecule; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; H&E, hematoxylin-eosin; K19, keratin 19; MMP, matrix metalloproteinase; mRNA, messenger RNA; miRNA, microRNA; PCR, polymerase chain reaction; SD, standard deviation; uPA, urokinase plasminogen activator; uPAR, urokinase plasminogen activator receptor.

Patients and Methods

Case Selection and Histopathologic Analysis.

This study was performed on two independent cohorts of patients with HCC. The HCCs included in this study were typical HCCs morphologically; cases that could be classified as combined hepatocellular-cholangiocarcinoma by hematoxylin-eosin (H&E) or mucin stains were excluded from both cohorts. This study was approved by the ethics committees of Seoul National University Bundang Hospital (Seoul, Korea) and Severance Hospital (Seoul, Korea). Cohort 1 consisted of 137 formalin-fixed, paraffin-embedded HCC specimens obtained from the archives of the Department of Pathology, Seoul National University Bundang Hospital, from 2003 to 2009. Patients were 15-87 years in age (range, 56.4 ± 12.2, mean ± standard deviation [SD]) and consisted of 108 males and 29 females. Mean follow-up time after surgery was 33.9 months (range, 0-91). Cohort 2 consisted of 237 paraffin-embedded human HCC specimens, surgically resected from January 2000 to December 2009 at Severance Hospital, Yonsei University Medical Center. Curative resection was performed for all patients. Major and minor resections were defined as resection of ≥3 segments and ≤2 segments, according to the Couinaud classification, respectively. The patient population consisted of 189 males and 48 females, and their ages ranged from 24 to 81 years (range, 55.0 ± 10.2, mean ± SD). All patients received no preoperative treatment, such as transarterial chemoembolization or radiation. All patients were checked for serum alpha-fetoprotein (AFP) and protein induced by vitamin K absence or antagonist II (PIVKA II) and underwent ultrasonography or dynamic computed tomography every 3-6 months after surgery. Median follow-up time after resection was 21.6 months (range, 1-109). Other important clinical data from each patient were obtained from a careful review of the medical records, including hepatitis B virus surface antigen status, serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), and albumin levels. One hundred-and-ninety cases were hepatitis B virus (HBV) related, and the remaining cases were hepatitis C virus (HCV) related (n = 13), alcohol related (n = 19), or of unknown etiology (n = 15). In addition, 43 snap-frozen human HCC specimens were available from cohort 2 for real-time polymerase chain reaction (PCR) analysis. The fresh frozen specimens were obtained from the Liver Cancer Specimen Bank (part of the National Research Bank Program, Korea Science and Engineering Foundation, Ministry of Science and Technology).

Histopathologic analysis was performed for both cohorts on whole tissue sections, and the variables recorded for each case included tumor size, differentiation according to Edmondson-Steiner grade, presence of multiple tumors, fibrous stroma, tumor capsules, microvascular and major vessel invasion, and background of cirrhosis. Tumor-capsule formation was defined as the presence of a fibrous capsule occupying >50% of the tumor circumference, and fibrous stroma was defined as the presence of fibrosis occupying 5%-30% of the tumor area. HCCs with fibrous stroma occupying >30% of the entire tumor area were excluded from this study to avoid confusion with “scirrhous HCC.”14, 15

Tissue Microarray Construction (Cohort 1).

Core tissue biopsies were taken from individual paraffin-embedded HCC donor blocks and arranged in recipient tissue-array blocks using a trephine apparatus (Superbiochips Laboratories, Seoul, Korea). At least 2 cores were sampled from each tumor, with the number of cores depending on the degree of heterogeneity present on histologic examination.


Information on antibodies used and antigen-retrieval conditions are summarized in Supporting Table 1. Immunohistochemical stains were performed as previously described.11 Brown membranous and/or cytoplasmic staining was counted as positive for K19, EpCAM, c-kit, CD133, vimentin, E-cadherin, MMP2, uPAR, and ezrin, and nuclear and/or cytoplasmic staining for snail and S100A4 was counted as positive. For all antibodies studied, except E-cadherin, the immunohistochemical stain results of cohort 1 were interpreted in a semiquantitative manner and given a score, from 0 to 3, as follows: 0: staining in <1% of tumor cells; 1: weak staining in ≥1%; 2: moderate staining in ≥1%; and 3: strong staining in ≥1% of tumor cells. Positive staining was defined as staining scores of 2 and 3, whereas 0 and 1 were regarded as negative. For E-cadherin, immunohistochemical scoring was performed as follows: 0: intact membranous positivity; 1: partial loss of membranous stain; and 2: complete loss of membranous E-cadherin expression. For cohort 2, K19 positivity was defined as membranous and/or cytoplasmic expression in ≥5% of tumor cells with moderate or strong intensity.

Total RNA Extraction, cDNA Synthesis, and Real-Time Quantitative Reverse-Transcriptase PCR (Cohort 2).

Total RNA was extracted from 43 snap-frozen human HCC tissues and subjected to complementary DNA (cDNA) synthesis and real-time quantitative reverse-transcriptase PCR, as described previously.4 Supporting Table 2 shows the primer and probe sequences (Roche Molecular Biochemicals, Indianapolis, IN) for K19, uPAR, VIL2, Snail, Slug, and Twist. The sequences of MMP1 and MMP2 primers and probes are under the copyright of Applied Biosystems (Carlsbad, CA).

Statistical Analysis.

Statistical analysis was performed using SAS software (version 9.1.3; SAS Institute, Inc., Cary, NC) and assessed using the t-test, Pearson's correlation test, or chi-square test, as deemed appropriate. Disease-free survival was calculated using the Kaplan-Meier method, and differences in survival rate were compared using the log-rank test. Significant variables from the univariable analysis were entered in the multivariable analysis, which was performed using the Cox-proportional hazards model with forward stepwise selection. Statistical significance and marginal significance were assumed when P < 0.05 and P < 0.1, respectively.


Stemness-Related Protein Expression in HCC and Clinicopathologic Correlations: Tissue Microarray Study Results (Cohort 1, n = 137).

K19, EpCAM, c-kit, and CD133 expression was seen in 25 of 137 (18.2%), 48 of 137 (35.0%), 47 of 137 (34.3%), and 34 of 137 (24.8%) cases, respectively (Fig. 1; Table 1). The expression status of the four stemness-related proteins in this study were positively correlated with each other: K19 versus EpCAM (P < 0.001), K19 versus CD133 (P = 0.040), EpCAM versus CD133 (P = 0.006), and c-kit versus CD133 (P = 0.006). K19 positivity was most frequently found in combination with at least one other stemness-related marker: the frequency of K19 expression alone in HCCs was 8.0% (2/25). On the other hand, the frequencies of CD133, EpCAM, and c-kit expression alone were higher than K19 (12.1% [4/33], 25.5% [12/47], and 39.1% [18/46], respectively). The expression of CD133, c-kit, and EpCAM was uniformly distributed in HCCs without any differences in staining pattern or intensity according to the histopathological features. K19 expression was either diffuse or patchy, and, occasionally, scattered K19-positive tumor cells were observed. K19 positivity could be seen in “hepatocyte-like” tumor cells constituting the majority of tumor cells (n = 14) and/or in smaller tumor cells located either at the periphery of the tumor-cell nests adjacent to the fibrous stroma or within the tumor cell nests (n = 11)—that is, K19 positivity was unpredictable, without predilection for a particular morphological type of tumor cell (Supporting Fig. 1).

Figure 1.

Stemness-related protein expression in HCCs. HCCs expressing stemness-related proteins show less frequent tumor capsules (A) and more frequent fibrous stroma between tumor-cell nests (B) [A, B; H&E, original magnification ×40 (A), ×200 (B)]. Immunohistochemical stain results for K19 (C), CD133 (D), and EpCAM (E) and Venn diagram showing the frequency of K19, EpCAM, and CD133 expression in cohort 1 HCCs (F).

Table 1. Immunohistochemical Stain Results and Clinicopathologic Features of 137 HCCs (Cohort 1)
 K19 CD133 EpCAM c-kit 
Clinicopathologic FeaturesPositiveNegativeP ValuePositiveNegativeP ValuePositiveNegativeP ValuePositiveNegativeP Value
  • *

    Edmondson-Steiner grades III or IV; n = 137, unless otherwise specified.

  • Abbreviations: HCCs, hepatocellular carcinomas; HBs, hepatitis B surface; ALT, alanine aminotransferase; AST, aspartate aminotransferase; MMP2, matrix metalloproteinase 2; uPAR, urokinase plasminogen activator receptor.

Frequency (%)25 (18.2)112 (81.8) 34 (24.8)103 (75.2) 48 (35.0)89 (65.0) 47 (34.3)90 (65.7) 
Age (years, mean ± SD)55.4 ± 13.356.7 ± 12.00.64255.9 ± 12.256.6 ± 12.30.75553.2 ± 11.758.1 ± 12.20.02655.7 ± 12.456.8 ± 12.20.618
Gender (male:female)16:992:200.05824:984:200.33532:1576:140.04637:971:200.827
HBs antigen (%)16 (64.0)76 (67.9)0.68523 (69.7)69 (66.3)0.50436 (76.6)56 (62.2)0.05233 (71.7)59 (64.8)0.206
Serum ALT (IU/L, mean ± SD)35.6 ± 17.441.0 ± 30.10.39133.7 ± 15.342.0 ± 31.10.14433.6 ± 18.143.3 ± 31.80.05842.7 ± 21.838.7 ± 31.00.439
Serum AST (IU/L, mean ± SD)42.3 ± 15.142.8 ± 27.40.93037.3 ± 17.244.5 ± 27.50.08236.5 ± 14.145.9 ± 29.30.01345.8 ± 25.441.2 ± 25.60.319
Serum albumin (g/dL, mean ± SD)3.9 ± 0.44.1 ± 0.40.0434.0 ± 0.54.0 ± 0.40.6414.1 ± 0.44.0 ± 0.40.5884.0 ± 0.44.1 ± 0.40.435
Alpha-fetoprotein >1,000 IU/mL (%)10 (40.0)19 (17.0)0.02810 (30.3)19 (18.3)0.14718 (38.3)11 (12.2)0.0019 (19.6)20 (22.0)1.000
Cirrhosis (%)13 (52.0)54 (48.2)0.56816 (48.5)51 (49.0)0.69325 (53.2)42 (46.7)0.85621 (45.7)46 (50.5)0.581
Tumor size (≥5 cm) (%)13 (52.0)32 (28.6)0.0347 (21.2)38 (36.5)0.13716 (34.0)29 (32.2)0.85014 (30.4)31 (34.1)0.704
Microvascular invasion (%)12 (48.0)68 (60.7)0.26818 (54.5)62 (59.6)0.68627 (57.4)53 (58.9)1.00025 (54.3)55 (60.4)0.583
Major vascular invasion (%)7 (28.0)9 (8.0)0.0116 (18.2)10 (9.6)0.2155 (10.6)11 (12.2)1.0006 (13.0)10 (11.0)0.781
Multiple tumors (%)6 (24.0)25 (22.3)0.79810 (30.3)21 (20.2)0.23912 (25.5)19 (21.1)0.66811 (23.9)20 (22.0)0.831
Poor differentiation* (%)22 (88.0)75 (67.0)0.05024 (72.7)73 (70.2)0.83036 (76.6)61 (67.8)0.32633 (71.7)64 (70.3)1.000
Tumor capsule formation (%)3 (12.0)29 (25.9)0.2354 (12.1)28 (26.9)0.2626 (12.8)26 (28.9)0.0946 (13.0)26 (28.6)0.125
Fibrous stroma (%)11 (44.0)29 (25.9)0.08216 (48.5)24 (23.1)0.00822 (47.8)18 (20.0)0.00220 (43.5)20 (22.0)0.027
Vimentin (%)4 (16.0)5 (4.5)<0.0015 (15.2)7 (6.7)0.0085 (10.6)4 (4.4)0.1742 (4.3)7 (7.7)0.335
Snail (%) (n = 136)5 (20.0)12 (10.7)0.26116 (50.0)27 (26.0)0.01616 (34.8)27 (30.0)0.56620 (44.4)23 (25.3)0.031
S100A4 (%)9 (36.0)5 (4.5)<0.0015 (15.2)9 (8.7)0.3257 (14.9)7 (12.9)0.2373 (6.5)11 (12.1)0.383
E-cadherin (loss) (%) (n = 136)7 (28.0)17 (15.2)0.2807 (21.2)17 (16.3)0.0065 (10.9)19 (21.1)0.3338 (17.4)16 (17.8)0.732
MMP2 (%) (n = 133)13 (52.0)51 (45.5)0.47718 (56.3)46 (45.1)0.58521 (46.7)43 (48.3)0.19921 (45.7)43 (48.9)0.658
uPAR (%) (n = 133)12 (48.0)19 (17.0)0.00313 (40.6)18 (17.6)0.01513 (28.9)18 (20.2)0.2839 (19.6)22 (25.0)0.525
Ezrin (%) (n = 133)17 (68.0)32 (27.4)<0.00116 (50.0)33 (32.4)0.09224 (53.3)25 (28.1)0.00715 (32.6)34 (38.6)0.573

K19-positive HCCs demonstrated more frequent major vessel invasion (P = 0.011), increased tumor size (P = 0.034), poor differentiation (P = 0.050), and fibrous stroma (P = 0.082), compared to K19-negative HCCs. Proteins related to EMT and invasiveness were more frequently expressed in K19-positive HCCs, although statistical significance was found for only vimentin (P < 0.001), S100A4 (P < 0.001), uPAR (P = 0.003), and ezrin (P < 0.001) (Fig. 2; Supporting Fig. 2). Fibrous stroma was also more frequently observed in CD133, EpCAM, and c-kit-expressing HCCs (P = 0.008, P = 0.002, and P = 0.027, respectively), compared to HCCs negative for these markers; however, the other pathologic features were not significantly different, according to the expression status of these markers. CD133-positive HCCs were characterized by more frequent vimentin (P = 0.008), snail (P = 0.016), uPAR (P = 0.015), and ezrin (P = 0.092) positivity and loss of E-cadherin expression (P = 0.006), compared to CD133-negative HCCs. Ezrin expression was more common in EpCAM-positive HCCs, compared to EpCAM-negative HCCs (P = 0.007), and snail was more frequently expressed in c-kit-positive HCCs, compared to c-kit-negative HCCs (P = 0.031).

Figure 2.

A summary of the immunohistochemical stain results of cohort 1 (A). The cases are sorted from left to right in ascending order, based on the number of expressed stemness-related markers. The positive cases are indicated by black boxes. The numbers of expressed stemness-related markers for each case are summarized at the top of the figure. (B-F) EMT and invasion-associated protein expression in HCCs expressing stemness-related markers. Nuclear expression of snail (B), nuclear and cytoplasmic expression of S1004 (C), and cytoplasmic expression of MMP2 (D), uPAR (E), and Ezrin (F) are observed.

A univariable survival analysis, according to the expression status of the four stemness-related markers, demonstrated that K19 expression in HCC was associated with a poor overall (P = 0.018) and disease-free survival (P = 0.007) (Supporting Table 3). CD133-expressing HCCs showed decreased overall survival (P = 0.057), compared to CD133-negative HCCs, although marginally significant. EpCAM or c-kit expression status was not related to HCC prognosis in this study. In addition, expression of three or more stemness-related proteins in HCC was characterized by a decreased overall survival (P = 0.086); however, disease-free survival was not affected by this variable. Because the majority (n = 101) of cohort 1 HCCs were HBV-related, survival analysis was repeated separately for 101 HBV-related HCCs and 36 HBV-unrelated HCCs (11 HCV-related, 6 alcohol-related, and 19 of uncertain etiology) to see whether K19 would be still prognostically significant in non-B-viral HCCs. Disease-free survival was still significantly decreased in K19-positive, HBV-unrelated HCCs, compared to K19-negative cases (P = 0.005) (Supporting Fig. 3). Overall survival, however, was not significantly different between the two groups. As for the HBV-related HCCs, decreased overall survival (P = 0.002) and disease-free survival (P = 0.121) were noted in the K19-positive groups.

Clinicopathologic Analysis of 237 HCCs According to K19 Expression Status (Cohort 2, n = 237): Clinicopathological Features of HCC With K19 Expression.

Because K19 appeared to be the stemness-related marker that was most significantly associated with clinicopathologic features of tumor aggressiveness, the next cohort was divided into two groups, according to K19 protein expression status. Immunohistochemical stain for K19 protein was done using whole sections of representative paraffin blocks from each case, and K19 positivity was found in 68 (28.7%) cases. The pattern of K19 immunoreactivity was focal and heterogeneous. The smaller tumor cells at the periphery of the tumor-cell nests, or intermingled with the hepatocyte-like cells within the tumor-cell nests, expressed K19 in 16 cases, whereas the majority of the K19-expressing tumor cells were hepatocyte-like cells (n = 52).

HCCs were grouped according to K19 protein-expression status, and clinicopathological features were compared between the two groups (Table 2). The K19-positive group was composed of higher proportions of younger (P = 0.004) and female (P < 0.001) patients, compared to the K19-negative group. K19-positive HCCs were more frequently associated with high AFP levels (>1,000 IU/mL), compared to K19-negative HCCs (P < 0.001).

Table 2. Comparison of Clinicopathologic Features Between K19-Positive and K19-Negative HCCs*
Clinopathologic FeaturesK19-Positive Group (n = 68)K19-Negative Group (n = 169)P Value
  • *

    n = 237, cohort 2.

  • †Edmondson-Steiner grades III or IV.

  • Abbreviations: K19, keratin 19; SD, standard deviation; ALT, alanine aminotransferase; AST, aspartate aminotransferase.

Age (years, mean ± SD)52.0 ± 10.256.2 ± 10.00.004
Gender (male:female)42:26147:22<0.001
HBs antigen (%)57 (83.8)133 (78.7)0.371
Serum ALT (IU/L, mean ± SD)36.4 ± 21.337.3 ± 22.50.777
Serum AST (IU/L, mean ± SD)46.0 ± 19.044.7 ± 34.60.780
Serum albumin (g/dL, mean ± SD)4.3 ± 0.54.4 ± 0.50.033
Serum alpha-fetoprotein >1,000 IU/mL (%)25 (36.8)18 (10.7)<0.001
Cirrhosis (%)41 (60.3)82 (48.5)0.101
Tumor size (cm, mean ± SD)3.92 ± 2.023.88 ± 2.580.911
Microvascular invasion (%)50 (73.5)96 (56.8)0.017
Major vascular invasion (%)6 (8.8)9 (5.3)0.317
Multiple tumors (%)12 (17.6)26 (15.4)0.688
Poor differentiation* (%)35 (51.5)92 (54.4)0.679
Tumor-capsule formation (%)11 (16.2)51 (30.2)0.027
Fibrous stroma (%)12 (17.6)12 (7.1)0.015
Minor/major resections33/3581/880.933

On pathologic examination, K19-positive HCCs showed more frequent microvascular invasion (P = 0.017) and fibrous stroma (P = 0.015) and less frequent fibrous capsule formation (P = 0.027). There was no significant difference in histologic differentiation between K19-positive and -negative HCCs.

Comparison of K19 mRNA Levels With Expression Levels of EMT-Associated Genes (Snail, Slug, and Twist) and Invasion-Associated Genes (uPAR, VIL2, MMP1, and MMP2).

mRNA expression levels of K19, three EMT-associated genes (Snail, Slug, and Twist), and four invasion-associated genes (uPAR, VIL2, MMP1, and MMP2) were investigated in 43 HCCs of cohort 2. The log-transformed K19 mRNA levels were significantly higher in K19 protein-positive HCCs than in K19 protein-negative HCCs (P = 0.048) (Fig. 3). Patients were classified into two groups (K19 mRNA high and K19 mRNA low), using the median of log-transformed K19 mRNA level as the cut-off value.

Figure 3.

Difference in K19 mRNA expression according to K19 protein expression status in human HCCs (A) and the mRNA levels of Snail (B), Twist (C), MMP2 (D), and uPAR (E) in HCCs, according to K19 mRNA expression status in cohort 2 HCCs (n = 43).

The log-transformed mRNA expression levels of EMT-associated genes were positively correlated with each other (P < 0.05), and positive correlations were observed between the log-transformed mRNA levels of invasion-associated genes (P < 0.05) (Supporting Table 4). Positive correlations were also observed between the mRNA levels of EMT-associated genes and invasion-associated genes (P < 0.05).

The K19 mRNA high HCCs showed significantly higher mRNA levels of Snail (P = 0.012), Twist (P = 0.069), uPAR (P = 0.040), and MMP2 (P = 0.040), whereas Slug, VIL2, and MMP1 mRNA levels were not significantly different between the two groups. Significant positive correlations were observed between the log-transformed mRNA levels of K19 and Twist (P = 0.012), uPAR (P = 0.005), and MMP2 (P = 0.009).

Prognostic Significance of K19.

Univariable analysis revealed that AST >50 IU/mL, K19 expression, tumor size >5 cm, multiple tumors, major vascular invasion, microvascular invasion, and AFP >1,000 IU/mL were adverse prognostic factors for disease-free survival after resection (Table 3; Fig. 4). Multivariable analysis indicated that K19 expression, tumor size >5 cm, multiple tumors, major vascular invasion, and AST>50 IU/mL were independent prognostic factors for disease-free survival after resection (Table 4).

Figure 4.

Survival curves demonstrating the decreased disease-free survival in HCCs with K19 expression (A), tumor size >5 cm (B), multiple tumors (C), gross vascular invasion (D), and high serum AST levels (E) in cohort 2 HCCs (n = 237).

Table 3. Univariable Analysis of Disease-Free Survival in Patients With HCC (Cohort 2)
VariableNo. of Patients (n = 237)1 Year of Disease-Free Survival (%)3 Years of Disease-Free Survival (%)P Value
  1. Abbreviations: HCC, hepatocellular carcinoma; HBs, hepatitis B surface; ALT, alanine aminotransferase; AST, aspartate aminotransferase; K19, keratin 19.

Age (years)   0.152
Gender   0.624
HBs antigen   0.142
Serum albumin (g/dL)   0.256
Serum ALT (IU/L)   0.528
Serum AST (IU/L)   0.003
Cirrhosis   0.377
Extent of resection   0.707
K19 expression   0.008
Tumor size    
 ≤5 cm19178.860.60.002
 >5 cm4663.236.3 
Multiple tumors   <0.001
Major vascular invasion   0.005
Microvascular invasion   0.033
Edmondson-Steiner grade   0.270
Serum alpha-fetoprotein (IU/mL)   0.015
Table 4. Independent Prognostic Factors for Disease-Free Survival by Multivariable Analysis (Cohort 2)
VariableCoefficientStandard ErrorP ValueRelative Risk (95% CI)
  1. Abbreviations: 95% CI, 95% confidence; K19, keratin 19; AST, aspartate aminotransferase.

K19 expression0.5220.2430.0321.685 (1.046-2.716)
Multiple tumors0.8830.2780.0022.417 (1.401-4.172)
Major vascular invasion0.7980.3650.0292.221 (1.085-4.545)
AST >50 IU/L0.7860.3130.0122.194 (1.187-4.054)
Tumor size >5 cm0.6720.2630.0101.958 (1.170-3.277)

When survival analysis was performed separately for HBV-related (n = 190) and HBV-unrelated HCCs (n = 47), K19 expression was associated with decreased disease-free survival in both etiologic groups, but statistical significance was demonstrated only in the HBV-related HCCs (P = 0.011) (Supporting Fig. 3). No significant differences in overall survival were observed for both groups.


HCC with stemness-related marker expression is a recently proposed subtype of HCC in which a fraction of tumor cells (>5%) expresses stem/progenitor cell markers, but is not otherwise recognizable by routine hematoxylin-eosin (H&E) stain.16 Because this subgroup of HCCs has been reported to show more aggressive behavior, compared to conventional HCCs without stemness-related marker expression, it is important that a suitable marker is developed to facilitate its diagnosis. This subtype is different from combined hepatocellular-cholangiocarcinomas—which include the recently described combined hepatocellular-cholangiocarcinoma with stem cell features (i.e. typical, intermediate cell, and cholangiolocellular subtypes)—because these tumors can be recognized on H&E stain and confirmed by immunohistochemical stains for stemness markers.17

Several markers, including K19, CD133, c-kit, and EpCAM, which are expressed in liver stem/progenitor cells, have been suggested as markers for cancer stem cells in HCC, and although there is an increasing number of candidate stemness-related markers for HCCs, it is still uncertain which of these markers is the best one for representing the stemness of HCCs and how the expression of these various candidate markers are related to each other. We conducted an immunohistochemical analysis of four stemness-related proteins (e.g., K19, CD133, c-kit, and EpCAM), using a tissue microarray from the first cohort of patients, and compared the various clinicopathologic features according to the expression status of each of these markers.

EpCAM and c-kit expression were seen in approximately one-third of the cohort 1 HCCs, and these markers were less frequently expressed in combination with other stemness-related markers, compared to K19 and CD133. Different studies have reported a wide range of frequencies of c-kit expression in HCCs (2.3%-80%),18, 19 and EpCAM expression has been reported in up to 47% of HCCs.20 Although fibrous stroma was more frequently observed in HCCs with c-kit, CD133, and EpCAM expression, the other clinicopathologic features, including prognosis, did not significantly differ according to the expression status of these three markers. CD133-positive HCCs were characterized by increased expression of EMT-related proteins and loss of E-cadherin expression. In contrast, there were no significant relationships between c-kit and EpCAM expression and the expression status of EMT-associated markers, except for Snail and Ezrin, respectively.

In contrast, the expression of K19 in HCC (18.2%) was more frequently associated with clinicopathologic features, including larger tumor size, more frequent vascular invasion, and poor differentiation. Fibrous stroma and lack of tumor capsules (i.e., infiltrative growth) were also more frequently observed, although not statistically significantly. EMT-related proteins, such as vimentin, S100A4, uPAR, and ezrin, were significantly expressed in K19-positive HCCs, and these tumors showed significantly decreased overall and disease-free survival. Therefore, K19 was more closely related to aggressive behavior and EMT in HCCs, compared to CD133, c-kit, and EpCAM in this group of patients. Because more than 90% of K19-positive HCCs in cohort 1 expressed at least one other stemness-related marker, and K19-expressing HCCs were significantly associated with poor overall and disease-free survivals, we proceeded to the second cohort of patients, from a different institution, to examine the characteristics of K19-expressing HCCs in more detail. The second cohort contained a larger number of patients, and fresh tissue was available for further examination of the relationship between K19 and EMT-related marker expression at the mRNA level.

Interestingly, the clinicopathologic characteristics of K19-expressing HCCs were similar in both cohorts, although we found that K19 positivity increased from 18.2% in cohort 1 to 28.7% in cohort 2. Because K19-positive tumor cells were not diffusely present, the expression frequencies of K19 in cohort 1, using 2-mm core microarrays, may have been somewhat underestimated, despite the fact that we used a lower cut-off value of 1% for the tissue microarray cases, compared to the 5% cut-off value for the whole tissue sections of cohort 2. The histologic features, such as the presence of fibrous stroma and the lack of tumor capsules, were recognized in K19-positive HCCs of both cohorts, and vascular invasion and high serum AFP levels were also common features. In addition, although statistical significance was not reached, these tumors were more frequent in younger and female patients, compared to K19-negative HCCs, were larger in size, and were more frequently multiple. The immunostaining patterns of K19 were variable, in contrast to CD133, c-kit, and EpCAM; K19 expression was observed in both tumor cells with typical hepatocyte-like features and in the slightly smaller cells, which were located at the periphery or within the cell nests. However, the latter group of K19-expressing cells could not be readily identified by H&E stain, and these tumors could not be classified as the recently described combined hepatocellular-cholangiocarcinoma with stem cell features. K19 positivity in HCCs was associated with a decreased disease-free survival in the second cohort after both univariable and multivariable analyses, therefore showing that K19 is a significant independent prognostic factor, which is in line with previous studies regarding the prognostic significance of K19 expression in HCCs.3, 21

The molecular features that explain the aggressive behavior of HCCs with high K19 expression are still unclear, and, to our knowledge, this is the first study that compares the differences between K19-high and K19-low HCCs with regard to the expression of EMT and invasion-related molecules. The mRNA levels of K19 were well correlated with K19 protein expression detectable by immunohistochemistry, and HCCs with high K19 mRNA levels were significantly associated with up-regulated EMT and invasion-associated genes (e.g., snail, twist, uPAR, and MMP2). In addition, K19 protein expression was significantly related with vimentin, S100A4, uPAR, and ezrin expression, and Snail and MMP2 expression and loss of E-cadherin were also more frequent in K19-expressing HCCs, although not statistically significantly.

Fibrous stroma was more frequently observed in HCCs expressing any of the four stemness markers, and fibrous capsules were less common in these tumors. The presence of fibrous stroma and infiltrative growth may be characteristic histologic features of HCCs expressing stem/progenitor cell markers. Interestingly, a previous study demonstrated more frequent expression of stemness markers in “scirrhous” HCCs, compared to ordinary HCCs, and it was suggested that these stemness marker-expressing tumor cells may be involved in the fibrogenesis of these tumors.14, 22 It may, indeed, be possible that these tumor cells have the potential to produce fibrous stroma through the up-regulation of EMT-related markers.

There is increasing evidence suggesting K19 as a progenitor cell marker. An extensive gene-expression profiling study demonstrated the presence of a “hepatoblast subtype” of human HCC, which was characterized by K19 expression, and was suggested to arise from hepatic progenitor cells,5 and, recently, a progenitor-derived HCC model was established in the rat, in which the K19-positive gene signature was well correlated with the former group of human HCCs.23 Furthermore, more than 15% of the genes in the K19 gene signature overlapped with the genes listed in the human embryonic stem cell-like module.23, 24 It is still uncertain whether the expression of K19 proteins in HCCs implies that these HCCs actually do carry stemness functions, as in the K19-positive ductular reactions of the regenerating liver, or whether K19 expression in HCC is merely an epiphenomenon of poor differentiation. Indeed, some K19-positive HCCs were poorly differentiated tumors, where it is possible that increasing genomic instability may have resulted in the expression of K19. Interestingly, a very recent study demonstrated that K19 gene activation may result in the expression of microRNA (miRNA)-492, and that this miRNA—and not the K19 protein itself—may be responsible for the aggressive behavior of hepatoblastomas.25

K19 expression in HCCs may also have therapeutic implications; an association between the epidermal growth factor-epidermal growth factor receptor (EGFR) pathway and K19 expression in HCCs has recently been demonstrated, suggesting a possible role for therapeutic agents targeted against EGFR, such as Gefitinib and Erlotinib, in the treatment of this aggressive subset of HCCs.26

In conclusion, K19 positivity in HCC—which is easily detected by immunohistochemistry, and is reliable and reproducible—was well correlated with the clinicopathologic features of tumor aggressiveness and a poor prognosis, compared to other stemness-related proteins. K19 positivity in HCC was associated with increased expression of EMT and invasion-related proteins, both at the protein and mRNA level, and these results suggest that this subset of HCCs may acquire more invasive characteristics, compared to K19-negative HCCs, through the up-regulation of EMT and invasion-associated genes.