Supported by grants from the Japan Agency for Medical Research and Development (Grant Nos. JP17jm0210056, JP18ck0106425, JP19fk0210058, JP16im0302428, and JP19ae0101075, JP20fk0210049) and JSPS KAKENHI (JP18K19538, JP18H02792, and JP17H06329) and funded by AbbVie, Bristol‐Myers Squibb, Gilead Sciences, MSD, and Abbott Japan.
Potential conflict of interest: Dr. Koshikawa received grants from Abbott. Dr. Nakagawa is employed by, owns stock in, and received grants from Abbott. Dr. Yoshida is employed by, owns stock in, and received grants from Abbott. Dr. Toru Yoshimura is employed by, owns stock in, and received grants from Abbott.
[Correction added on July 21, 2021 after first online publication: Additional funding information was included to this article.]
Worldwide, HCC is the second‐most lethal cancer after pancreatic cancer, with a 5‐year survival of 18%.(1) Early detection of HCC prolongs patient survival, and several guidelines are available for HCC screening in high‐risk patients, including those with chronic hepatitis C (CHC).(2) Several reports have indicated the utility of molecular HCC subclasses, such as Hoshida’s S1‐3 subclasses(3) or Boyault’s G1‐6 subclasses,(4) classified by microarray data, to predict prognosis. However, alpha‐fetoprotein (AFP) is the only marker clinically available for detecting HCC with poor prognosis.(5)
HCV infection is a major global health problem that affects ~1% of the world’s population.(6) HCV infection results in chronic liver inflammation and fibrosis that ultimately lead to liver cirrhosis and HCC in the infected liver.(7) Oral direct‐acting antivirals (DAAs) targeting HCV nonstructural (NS)3/4A protease, the NS5A replication complex, or NS5B polymerase effectively eliminate HCV infection, and sustained virological response (SVR) rates of pangenotype DAAs exceed 95% with good tolerance.(8) DAA treatment effectively recovers liver function, improves overall survival, and reduces the risk of HCC development in CHC patients.(9‐12) However, HCC risk is not completely eliminated after DAA treatment.(9) HCC screening by imaging diagnostics, such as ultrasonography or CT, is particularly applied to patients with cirrhosis,(13) even after they achieve SVR. Globally, ~3 million persons underwent HCV treatment in 2015 and 2016, ~86% of whom received DAAs, suggesting that ~2.5 million persons achieved SVR in just 2 years.(14) Annual incidence of HCC is ~0.9 per 100 person‐years in CHC patients with SVR,(11) indicating that ~90,000 of 10 million CHC patients with SVR have HCC annually. Development of a marker to select CHC patients at high risk of HCC development after DAA treatment is required to detect HCC in the early stages, prevent excessive imaging screening, and reduce medical costs.
Although basement membrane proteins play a pivotal role in the maintenance of epithelial cells, their functional role as well as the expression and structural dynamics remain open for study in chronic hepatitis and HCC. Laminin 332 (Ln‐332) is a heterotrimeric basement membrane protein consisting of α3, β3, and γ2 laminin chains. We previously reported that laminin γ2 is strongly expressed as a monomeric form at the invasive front in tumor cells, but not in normal cells.(15) Recently, we developed a chemiluminescent immunoassay (CLIA) to detect the laminin γ2 monomer (LG2m), but not trimeric Ln‐332 γ2, in a highly sensitive and specific manner using a fully automated detection machine (Architect; Abbott Laboratories, Chicago, IL).(16,17) In that small‐scale cohort, LG2m was elevated in a subset of HCC patients, although their outcomes have remained unclear.
In this study, we evaluated the molecular characteristics of LG2m‐positive HCC using cell lines (n = 14), microarray data (n = 258), and sera from patients with chronic liver disease (CLD; n = 133) and patients with HCC (n = 142). We further tested the ability of serum LG2m levels to predict HCC metastasis (n = 128) or HCC development in CHC patients with SVR (n = 399) after DAA treatment.
Materials and Methods
Exploratory Studies
Study stages are depicted in Fig. 1. At the exploratory stage, we collected supernatants from 14 HCC cell lines or patient‐derived HCC cells to determine LG2m expression characteristics (Fig. 1; Supporting Table S1). We obtained HuH1, HuH7, Hep3B, PLC/PRL/5, JHH5, JHH7, HLE, HLF, SK‐Hep‐1, and JHH4 cell lines from the Cell Bank of the Japanese Collection of Research Bioresources (Osaka, Japan).(18,19) KH, MT, KM, and Kami41 are patient‐derived HCC cell lines developed in the Department of Gastroenterology, Kanazawa University Hospital (Ishikawa, Japan). Detailed information is available in Supporting Table S1. Cells were cultured in culture dishes at 37°C and 5% CO2 in DMEM or RPMI‐1640 medium supplemented with 10% fetal calf serum, 2 mM of l‐glutamine, and 1% penicillin‐streptomycin. After cultures reached confluency, they were cultured in serum‐free medium for 48 hours. Levels of AFP, des‐gamma‐carboxy prothrombin (DCP), and LG2m in culture supernatants were measured by CLIA using a fully automated detection machine (Architect; Abbott Laboratories). Cell‐surface expression levels of epithelial cell adhesion molecule (EpCAM) and CD90 were measured by fluorescence‐activated cell sorting as described.(20) RNA was extracted from these cells, and microarray analysis was performed using a U133 2.0 plus array (ThermoFisherScientific, Waltham, MA). Serum AFP, DCP, and LG2m were measured in 19 HCC patients whose tumor samples were analyzed with microarray (ThermoFisherScientific). Of the 19 HCC patients, 14 underwent surgery, and their serum AFP, DCP, and LG2m levels before and after surgical resection were measured. Additional array data from 239 HCCs were used to characterize the gene signature of the LG2m expression status (GSE14520). Sera from 24 healthy donors (HDs) were collected to determine the cut‐off value of LG2m. Sera from 133 CLD patients (HBV related, n = 40; HCV related, n = 70; non‐B/non‐C, n = 23) and 142 HCC patients (HBV related, n = 29; HCV related, n = 73; non‐B/non‐C, n = 40) were collected to evaluate the sensitivity and specificity of the HCC diagnosis by AFP, DCP, and LG2m. All patients and donors provided written informed consent. The serum LG2m level measured by CLIA was confirmed to be stable for at least 1 year when stored at −80°C. Although all serum samples were stored at −20°C, analysis of 70 CHC cohort samples revealed no correlation between serum LG2m levels and the duration of serum storage (data not shown).
;)
FIG. 1: Study design. Exploratory samples, test cohort samples, and validation cohort samples are depicted. Abbreviations: BCLC, Barcelona Clinic Liver Cancer; N/A, not applicable.
HCC Prognosis Studies
The relationship between the LG2m gene signature and prognosis was evaluated using 239 HCC microarray samples (GSE14520). All HCC patients enrolled in the test (n = 47, from February 2006 to December 2015) and validation (n = 81, from June 2006 to August 2015) cohorts for HCC prognosis analysis underwent surgery or radiofrequency ablation. Clinical background data are presented in Supporting Table S2, including tumor‐node‐metastasis stages. Overall survival and time to extrahepatic spread (EHS) were analyzed. All patients provided written informed consent.
HCC Prediction Studies
Sera collected from 70 CHC patients recruited from May 2004 to September 2013 served as a test cohort to evaluate subsequent HCC risk according to serum LG2m level (Fig. 1; Supporting Tables S3 and S4). These patients received interferon‐based treatment, and an SVR was achieved in 57% (40 of 70 patients). Potential predictors of clinical outcome assessed at entry were age, sex, alcohol use, obesity, diabetes mellitus, fibrosis scores (aspartate aminotransferase‐to‐platelet ratio index [APRI] and Fibrosis‐4 [FIB‐4] index), and biological variables (platelet count, albumin, aspartate aminotransferase [AST], and alanine aminotransferase [ALT]). Sera were sent to the Kanagawa Cancer Center Research Institute (Yokohama, Japan), and AFP, DCP, and LG2m levels were measured by the CLIA system. All patients provided written informed consent.
For a validation study, patients were recruited from 28 hospitals in three prefectures in the Hokuriku region of Japan (Fig. 1; Supporting Table S4; participating hospitals are listed in Supporting Table S5). A retrospective test cohort indicated that one third of CHC patients without an SVR showed elevated LG2m; HCC incidence at 3 years was 21.3% and 0% in the LG2m‐positive and ‐negative groups, respectively. Because DAA reduced HCC risk to approximately one third, we estimated a sample size of 387 patients (statistical power of 80%, α = 0.05, and LG2m‐positive/LG2m‐negative ratio = 1:2; HCC incidence at 3 years hypothesized to be 6% [LG2m positive] and 1% [LG2m negative], 44‐month recruitment period, and an additional 12‐month follow‐up period). All patients provided written informed consent and underwent HCC screening by imaging diagnostics within 4 weeks before initiation of DAA treatment. This study evaluated de novo HCC incidence, and patients who had a previous history of HCC were excluded because HCC recurrence might result from intrahepatic metastasis of previously treated HCC. The study organization is depicted in Supporting Fig. S1. The trial protocol was approved by Kanazawa University Hospital and the ethics committee of each hospital. Patients’ sera were collected at baseline and stored in a −20°C freezer, and times to SVR achievement were recorded. Sera were sent to the Kanagawa Cancer Center Research Institute, and AFP, DCP, and LG2m levels were measured by the CLIA system. Tumor marker data were sent to the Data Center of the Innovative Clinical Research Center at Kanazawa University (iCREK). The primary endpoint of this prospective study was HCC incidence. After SVR achievement by DAA treatment, all patients underwent HCC surveillance according to the Japan Society of Hepatology (JSH) algorithm.(21) Event information was centrally evaluated by two independent radiologists at the Image Center of Kanazawa University Hospital. HCC was diagnosed according to JSH guidelines. Certified event information was sent to the Data Center of iCREK for data management. Clinical trial registration University Hospital Medical Information IDs are 000016633 (supported by Bristol‐Myers Squibb), 000019801 (supported by Gilead Sciences), 000020911 (supported by AbbVie), and 000025575 (supported by MSD). Detailed protocols are available in the Supporting Information. All authors had access to the study data and reviewed and approved the final manuscript.
Statistical Analysis
All clinical data, serum tumor markers, and final outcome information of this multicenter, prospective study were recruited to, and managed at, the Data Center and the Biostatistical Analysis Center of iCREK. All analyses were performed in the intention‐to‐diagnose population. Time to HCC development was calculated as the time between SVR achievement and the last follow‐up visit or date of HCC development. Kaplan‐Meier curves of cumulative HCC incidence, survival, recurrence, or incidence of EHS were compared using log‐rank tests in GraphPad Prism software (ver. 8.2.0; GraphPad Software Inc., San Diego, CA). Cox proportional hazard models were used to identify risk factors for HCC, analyzed using SPSS software (ver. 23.0; IBM Japan, Ltd., Tokyo, Japan). Two‐sided P values of ≤0.05 were considered statistically significant. Correlation among AFP, DCP, and LG2m levels was evaluated using the Pearson correlation coefficient in GraphPad Prism. The area under the receiver operating characteristic (AUROC) was evaluated using GraphPad Prism. GeneSpring GX software (Agilent, Santa Clara, CA) was applied to identify differentially expressed genes between LG2m‐positive and ‐negative cell lines, using a t test. Gene set enrichment analysis (GSEA) was performed using microarray data on 14 HCC cell lines and GSEA software (ver. 4.0.3; http://www.gsea‐msigdb.org/).
Results
LG2m Reflects the CD90+ Metastatic Cancer Stem‐Cell Population and Is Expressed in a Distinct Molecular Subclass of HCC
Study stages are depicted in Fig. 1. We developed and optimized the CLIA system to determine serum LG2m levels. We previously proposed the classification of HCC cell lines into two groups based on expression of EpCAM and CD90 stem‐cell markers: AFP‐positive HCC cell lines with features of epithelial stem cells (EpCAM positive) or AFP‐negative HCC cell lines with features of mesenchymal stem cells (CD90 positive).(20) Importantly, presence of CD90‐positive cells was correlated with later development of distant organ metastasis.(20) We initially evaluated LG2m level in culture supernatants of HCC cell lines and found that AFP and DCP were detected mainly in EpCAM‐positive cell lines (Fig. 2A,B). In contrast, LG2m was abundantly produced in mesenchymal CD90‐positive HCC cell lines (Fig. 2C). These data suggest that LG2m could reflect the presence of CD90‐positive metastatic cancer cells and that LG2m measurement may fill in the diagnostic gaps for AFP‐ or DCP‐negative HCC.
FIG. 2: Characteristics of LG2m expression in cell lines and HCC tissues. (A) AFP concentration detected in culture supernatants of HCC cell lines and patient‐derived HCC cells. EpCAM positivity and CD90 positivity are indicated in blue and yellow, respectively. HuH1, HuH7, Hep3B, PLC/PRL/5, JHH5, JHH7, HLE, HLF, SK‐Hep‐1, and JHH4 cell lines were obtained from the Cell Bank of the Japanese Cancer Research Resources Bank. KH, MT, KM, and Kami41 are patient‐derived HCC cells developed in the Department of Gastroenterology, Kanazawa University Hospital. (B) DCP concentration determined in culture supernatants in HCC cell lines and patient‐derived HCC cells. (C) LG2m concentration determined in culture supernatants in HCC cell lines and patient‐derived HCC cells. (D) Expression pattern of the LG2m gene signature in CD90‐positive (orange box), EpCAM‐positive (blue box), LG2m‐positive (red box), and LG2m‐negative (green box) cells after hierarchical clustering of genes and samples, shown as a heatmap image. Red indicates a high expression level; green indicates a low expression level. LG2m was defined as positive when the LG2m level in the culture supernatant exceeded 60 pg/mL. EpCAM and CD90 expression was determined by flow cytometry. (E) Hierarchical clustering of 19 HCC tissues by LG2m gene signature. Five and 14 HCCs were regarded as LG2m positive or LG2m negative, respectively, based on the serum LG2m level (cutoff, 30 pg/mL). (F) Activated pathways identified in Cluster A (orange bar) and B (blue bar). (G) GSEA of microarray data of 14 HCC cell lines. Genes related to Hoshida’s S1 subclass (NES, 2.08; nominal P value, <0.0001) and Boyault’s G3 subclass (NES, 1.50; nominal P value, <0.0001) were suggested to be activated in LG2m‐positive HCC cell lines. (H) Serum tumor marker values (AFP, DCP, and LG2m) before and after surgery in 14 HCC patients whose tissues were analyzed by microarray. (I) Kaplan‐Meier curves of HCC recurrence in LG2m‐positive (n = 7) and LG2m‐negative (n = 7) HCCs after surgery, based on serum LG2m level (cutoff, 30 pg/mL). Abbreviation: NES, normalized enrichment score.
Using microarray data, we identified 155 and 238 genes up‐regulated (Cluster A) or down‐regulated (Cluster B) >2‐fold in LG2m‐positive cell lines compared with LG2m‐negative lines (Supporting Table S6). This gene signature (LG2m gene signature) was able to separately classify LG2m‐positive and ‐negative HCC cell lines by hierarchical clustering (Fig. 2D). The signature could further classify HCC tissues with elevated serum LG2m (Fig. 2E), suggesting a correlation between the LG2m gene signature in the tumor and its serum LG2m level. Pathway analysis indicated that the top four pathways activated in LG2m‐positive HCC were related to cell motility, movement, locomotion, and migration (Cluster A; Fig. 2F).
We performed GSEA to identify molecular subclasses corresponding to LG2m‐positive HCC. By testing default gene sets described in Hoshida’s(3) and Boyault’s(4) subclasses, widely used for microarray‐based HCC classification, we determined that LG2m‐positive HCC cell lines corresponded to the Hoshida’s S1 and Boyault’s G3 subclasses (Fig. 2G), both associated with poor prognosis and early recurrence.(22) Importantly, these subclasses are still open for clinical diagnosis because of the absence of serum tumor markers.
We measured the serum values of AFP and DCP in 14 of the 19 HCC patients (analyzed by microarray as described above) who underwent surgery. Seven and nine HCCs were regarded as AFP positive (10 ng/mL) and DCP positive (40 mAU/mL) before surgery, respectively, and all AFP‐positive and 8 of the 9 DCP‐positive HCC patients showed reduction or normalization of serum AFP and DCP levels after surgery (Fig. 2H). In contrast, although five HCCs were regarded as LG2m positive (30 pg/mL), only two of them showed decreased serum LG2m after surgery. Furthermore, we found that 2 LG2m‐negative HCC patients became LG2m positive after surgery (Fig. 2H).
Because LG2m status may correlate with the molecular phenotype closely associated with cell migration, metastasis, and poor prognosis, we hypothesized that LG2m‐positive cells might remain and proliferate after surgery, which might be related to the high postoperative recurrence rate in these patients. We evaluated the outcomes of these 14 HCC patients, and LG2m‐positive HCC evaluated after surgery showed significantly higher incidence of recurrence compared with LG2m‐negative HCC (P = 0.023; Fig. 2I). AFP and DCP status after surgery were not related to recurrence (Supporting Fig. S2). Taken together, these data suggest that LG2m is a serum tumor marker distinct from AFP and DCP, activated in Hoshida’s S1 and Boyault’s S3 HCC subclasses with poor survival outcome, and correlates with features of metastatic CD90‐positive cancer stem cells.
Serum LG2m Is Elevated in a Subset of HCC and CLD Patients
For clinical diagnostic application, we tested serum levels of LG2m in 24 HDs, 133 CLD patients, and 142 HCC patients (Fig. 1). The cut‐off value of 30 pg/mL was set according to the data from the HD samples (Fig. 3A). Serum LG2m was elevated in HCC samples compared with CLD and HD samples, although its elevation was detected in a subset of CLD samples. Approximately 86% of HCC samples showed elevation of either AFP, DCP, or LG2m (Fig. 3B).
FIG. 3: LG2m test for the diagnosis of HCC. (A) Dot plots of serum LG2m in HDs, CLD patients, and HCC patients. A cut‐off value of 30 pg/mL was set according to the HD data. (B) Venn diagram of AFP (cut‐off value of 20 ng/mL), DCP (40 mAU/mL), and LG2m (30 pg/mL) positivity in HCC. The number is depicted of HCC cases (n = 142) positive for AFP, DCP, and/or LG2m. (C) ROC curve for AFP in the diagnosis of HCC. The AUROC was 0.77. (D) ROC curve for DCP in the diagnosis of HCC. The AUROC was 0.73. (E) ROC curve for LG2m in the diagnosis of HCC. The AUROC was 0.74. (F) Scatter plots of AFP and DCP in HCC. Correlation was evaluated by the Pearson correlation coefficient. (G) Scatter plots of LG2m and AFP in HCC. Correlation was evaluated by the Pearson correlation coefficient. (H) Scatter plots of LG2m and DCP in HCC. Correlation was evaluated by the Pearson correlation coefficient. Abbreviation: NBNC, non‐B/non‐C.
We evaluated the receiver operating characteristic (ROC) curves of tumor markers using serum data from 133 CLD and 142 HCC samples (Fig. 3C‐E). The AUROC was similar for AFP (0.77), DCP (0.73), and LG2m (0.74). When cut‐off values of AFP, DCP, and LG2m were set at 20 ng/mL (AFP high), 80 mAU/mL (DCP high), and 60 pg/mL (LG2m high; twice the upper normal range of each marker), the sensitivity and specificity for HCC diagnosis were 38.3% and 89.9% (AFP), 21.0% and 97.7% (DCP), and 21.0% and 96.9% (LG2m), respectively. Accordingly, we used these cut‐off values in two independent HCC prognosis cohorts (Fig. 1) to characterize the natures of each marker‐high HCC with high specificity. In contrast, when cut‐off values of AFP, DCP, and LG2m were set at 10 ng/mL (AFP positive), 40 mAU/mL (DCP positive), and 30 pg/mL (LG2m‐positive; upper normal range of each marker), the sensitivity and specificity for HCC diagnosis were 76.5% and 59.7% (AFP), 37.0% and 94.6% (DCP), and 62.9% and 70.5% (LG2m), respectively. We used these cut‐off values in two independent HCC prediction cohorts (Fig. 1) to increase the sensitivity of each marker for HCC diagnosis.
Evaluation of the serum levels of AFP, DCP, and LG2m in 142 HCC patients revealed a small positive correlation between AFP and DCP (r = 0.43; P < 0.0001; Fig. 3F), but not between AFP and LG2m (r = −0.06; P = 0.47) or DCP and LG2m (r = −0.15; P = 0.08; Fig. 3G,H). These data indicated that LG2m is a distinct tumor marker elevated in a subset of CLD and HCC patients.
High Serum LG2m Correlates With Later EHSs of HCC With Poor Prognosis
We next evaluated the prognostic value of LG2m measurement in HCC patients. Because the prognostic information on 142 HCC patients at the exploratory stage was incomplete, we first used the LG2m gene signature, which closely correlated serum LG2m status, to evaluate its prognostic value in 239 HCC patients. We classified the 239 HCC patients analyzed by microarray (Fig. 1) as LG2m gene signature positive (n = 104) and LG2m gene signature negative (n = 135; Fig. 4A). We also classified these HCCs as AFP positive (≥20 ng/mL; n = 166) and AFP negative (n = 73) and assessed overall survival. We found that both AFP and LG2m elevation was correlated with poor overall survival with borderline significance (HR, 1.4; P = 0.059 by log‐rank test) or statistical significance (HR, 1.7; P = 0.008), respectively (Fig. 4B,C).
FIG. 4: Survival outcomes of HCCs based on LG2m expression. (A) Hierarchical clustering of 239 HCC tissues by LG2m gene signature. AFP‐positive (orange box), AFP‐negative (blue box), LG2m‐positive (red box), and LG2m‐negative (green box). LG2m was defined as positive if HCC was clustered in the right branch with activation of Cluster A genes. AFP was defined as positive based on the serum AFP level (cutoff, 20 ng/mL). (B) Kaplan‐Meier curves of HCC patients (overall survival) according to serum AFP (cutoff, 20 ng/mL; AFP high, n = 166; AFP low, n = 73). (C) Kaplan‐Meier curves of HCC patients (overall survival) according to LG2m gene signature (LG2m+, n = 104; LG2m–, n = 135). (D) Kaplan‐Meier curves of 47 HCC patients (overall survival) according to serum AFP (cutoff, 20 ng/mL; AFP‐high, n = 19; AFP‐low, n = 28). (E) Kaplan‐Meier curves of 47 HCC patients (overall survival) according to serum DCP (cutoff, 80 mAU/mL; DCP‐high, n = 6; DCP‐low, n = 41). (F) Kaplan‐Meier curves of 47 HCC patients (overall survival) according to serum LG2m (cutoff, 60 pg/mL; LG2m‐high, n = 12; LG2m‐low, n = 35). (G) Kaplan‐Meier curves of the cumulative incidence of the EHS of 47 HCC patients according to serum AFP (cutoff, 20 ng/mL; AFP‐high, n = 19; AFP‐low, n = 28). (H) Kaplan‐Meier curves of the cumulative incidence of the EHS of 47 HCC patients according to serum DCP (cutoff, 80 mAU/mL; DCP‐high, n = 6; DCP‐low, n = 41). (I) Kaplan‐Meier curves of the cumulative incidence of the EHS of 47 HCC patients according to serum LG2m (cutoff, 60 pg/mL; LG2m‐high, n = 12; LG2m‐low, n = 35). (J) Kaplan‐Meier curves of 81 HCC patients (overall survival) according to serum AFP (cutoff, 20 ng/mL; AFP‐high, n = 31; AFP‐low, n = 50). (K) Kaplan‐Meier curves of 81 HCC patients (overall survival) according to serum DCP (cutoff, 80 mAU/mL; DCP‐high, n = 16; DCP‐low, n = 65). (L) Kaplan‐Meier curves of HCC patients (overall survival) according to serum LG2m (cutoff, 60 pg/mL; LG2m‐high, n = 17; LG2m‐low, n = 64). (M) Kaplan‐Meier curves of the cumulative incidence of the EHS of 81 HCC patients according to serum AFP (cutoff, 20 ng/mL; AFP‐high, n = 31; AFP‐low, n = 50). (N) Kaplan‐Meier curves of the cumulative incidence of the EHS of 81 HCC patients according to serum DCP (cutoff, 80 mAU/mL; DCP‐high, n = 16; DCP‐low, n = 65). (O) Kaplan‐Meier curves of the cumulative incidence of the EHS of 81 HCC patients according to serum LG2m (cutoff, 60 pg/mL; LG2m‐high, n = 17; LG2m‐low, n = 64). (D‐I) Test cohort results. (J‐O) Validation cohort results. Number of patients at risk is shown at the bottom.
Next, we investigated the clinical outcomes related to AFP, DCP, and LG2m elevations in 47 (test cohort) and 81 (validation cohort) HCC patients (Fig. 1). The clinical backgrounds of these patients are shown in Supporting Table S2. We used the cut‐off values of 20 ng/mL (AFP), 80 mAU/mL (DCP), and 60 pg/mL (LG2m) to distinguish the characteristics of each marker‐high HCC. In the test cohort, Kaplan‐Meier survival analysis indicated that AFP‐ or DCP‐high HCC were not associated with survival (Fig. 4D,E). Nevertheless, LG2m‐high HCC was significantly correlated with poor overall survival (HR, 7.9; P = 0.006; Fig. 4F). We further evaluated the effect of elevated LG2m on subsequent EHS. Although AFP‐high HCC and DCP‐high HCC were not correlated with EHS, LG2m‐high HCC clearly showed a high incidence of the later development of EHS (HR, 21.5; P < 0.0001; Fig. 4G‐I). We validated the clinical outcome of LG2m elevation using an independent cohort, and Kaplan‐Meier survival analysis again indicated that LG2m‐high HCC was significantly correlated with poor overall survival (HR, 4.2; P = 0.0026; Fig. 4J‐L). Furthermore, although AFP‐ and DCP‐high HCC were not correlated with EHS, LG2m‐high HCC again showed a high incidence of the later development of EHS (HR, 8.6; P = 0.0002; Fig. 4M‐O).
We performed univariate Cox regression analysis to evaluate whether other clinicopathological factors were correlated with the clinical outcome of HCC (Supporting Tables S7‐S10). In the test cohort, LG2m elevation was the only factor correlated with poor prognosis and later EHS (Supporting Tables S7 and S8). In the validation cohort, LG2m and AFP were associated with poor survival (Supporting Table S9). LG2m was the only factor correlated with later EHS in the validation cohort (Supporting Table S10). All of these data suggest that LG2m is a marker of metastatic HCC.
Serum LG2m Elevation Predicts the Later Development of HCC in CHC Patients
The relatively low specificity of AFP and LG2m for HCC diagnosis (when the cut‐off values of AFP and LG2m were set at 10 ng/mL and 30 pg/mL, respectively) could be attributed to the slight elevation of these markers in CLD patients (Fig. 5A‐C). Although elevated serum AFP is a risk factor for HCC, the clinical outcome of elevated LG2m in CLD patients was unclear. Because CHC patients underwent HCC surveillance according to the guidelines, we evaluated serum LG2m levels in 70 CHC patients to predict HCC risk. Characteristics of the 70 patients at enrollment are shown in Supporting Table S3. Elevated LG2m was observed in 31% (22 of 70) of these CHC patients without HCC at the time of serum collection (Supporting Table S3). Elevated LG2m was associated with old age, elevated AST, thrombocytopenia, hypoalbuminemia, elevated fibrosis scores (APRI and FIB‐4 index), and high serum AFP (Supporting Table S4), suggesting that LG2m was elevated in the high‐risk group, leading to the development of HCC. We retrospectively evaluated the consequence of elevated LG2m in these CHC patients as a test cohort (median follow‐up period, 49.2 months; 25th‐75th percentile, 19.2‐90.0 months). In this cohort, non‐SVR achievement, thrombocytopenia, hypoalbuminemia, elevated fibrosis scores (APRI and FIB‐4 index), and elevated AFP were associated with an increased risk of HCC (Fig. 5D‐I). Elevated LG2m was a strong risk factor for HCC development (HR, undefined by log‐rank test; P < 0.001; Fig. 5J). Median time to HCC progression in these patients was 94 months, and no patients developed HCC without elevated LG2m.
FIG. 5: Cumulative HCC incidence in the test cohort. (A) Dot plots of serum AFP in CLD (n = 133) and HCC (n = 144) patients. (B) Dot plots of serum DCP in CLD (n = 133) and HCC (n = 144) patients. (C) Dot plots of serum LG2m in CLD (n = 133) and HCC (n = 144) patients. (D) Cumulative incidence of HCC according to SVR status. (E) Cumulative incidence of HCC according to platelet count (cutoff, 105/mm3). (F) Cumulative incidence of HCC according to APRI (fibrosis indicator; cutoff, 1). (G) Cumulative incidence of HCC according to FIB‐4 index (cutoff, 3.75). (H) Cumulative incidence of HCC according to serum albumin (cutoff, 4g/dL). (I) Cumulative incidence of HCC according to serum AFP (cutoff, 10 ng/mL). (J) Cumulative incidence of HCC according to serum LG2m (cutoff, 30 pg/mL). HR was not determined because the event was not detected in LG2m‐negative patients. Number of patients at risk is shown at the bottom.
Based on the retrospective data, a multicenter, prospective, clinical trial was performed. From October 2014 to May 2018, 399 patients who achieved SVR by DAA treatment provided consent for LG2m testing (intention‐to‐diagnose). One patient dropped out after achieving SVR, and the final outcome was fixed on July 30, 2019. Median follow‐up time of these patients was 30 months. Characteristics of the 399 patients at enrollment are shown in Supporting Table S3; 77% of the patients were aged ≥60 years, and two thirds were female. Diabetes was diagnosed in 21% of patients, and 23% were obese (body mass index, ≥25). Mean values of aminotransferases, albumin, and platelet count were within normal ranges at the time of SVR achievement. Approximately 93% of patients did not drink alcohol daily (>20 g per day). When compared with baseline data obtained before DAA treatment, AST (P < 0.001) and ALT (P < 0.001) levels were decreased, whereas albumin (P < 0.001) and platelet count (P < 0.001) were increased at the time of enrollment (SVR achievement), using matched pair samples (Supporting Fig. S3). We also observed a decrease in AFP (P < 0.001) and LG2m (P < 0.001), but not in DCP (P = 0.23), at the time of SVR. These data indicate that DAA treatment suppressed liver inflammation, improved liver function, and reduced production of the tumor markers, AFP and LG2m, when SVR was achieved.
During the follow‐up period, 10 of the 399 patients developed HCC. Clinical characteristics of these 10 patients are available in Supporting Table S11. HCC incidence at 12, 24, and 36 months was 1.1%, 1.7%, and 4.6%, respectively (Fig. 6A). In total, 122 patients were regarded as LG2m positive (≥30 pg/mL). Kaplan‐Meier survival analysis indicated that elevated LG2m at SVR was associated with significantly increased risk of HCC (HR, 19.6; 95% CI, 5.2‐74.6; P < 0.001; Fig. 6B). HCC incidence in LG2m‐negative patients remained very low (0.4%) for 3 years. When the cut‐off value of AFP was 10 ng/mL, as used in the test cohort, only 9 patients were regarded as AFP positive. Therefore, we set the cut‐off value of AFP at 5 ng/mL (n = 74) for HCC risk prediction, and found that elevated AFP was associated with an increased risk of HCC (HR, 8.8; 95% CI, 1.9‐40.5; P < 0.001), although the HR was lower than that of LG2m (Fig. 6C). HCC developed in some AFP‐negative patients, and its incidence gradually increased over time (0% at 1 year, 0.8% at 2 years, and 2.8% at 3 years). Elevated DCP was observed in 17 patients based on a cut‐off value of 40 mAU/mL, and 6 were prescribed warfarin. Elevated DCP was not associated with HCC risk.
FIG. 6: Cumulative HCC incidence in the validation cohort. (A) Kaplan‐Meier curve of HCC incidence in all enrolled patients. (B) Kaplan‐Meier curves of cumulative HCC incidence in serum LG2m‐positive (red) and LG2m‐negative (blue) CHC patients. (C) Kaplan‐Meier curves of cumulative HCC incidence in serum AFP‐positive (red) and AFP‐negative (blue) CHC patients. (D) Kaplan‐Meier curves of cumulative HCC incidence in either serum AFP‐ or LG2m‐positive (red) and double‐negative (blue) CHC patients. Number of patients at risk is shown at the bottom.
Because we found one event in the LG2m‐negative group and four events in the AFP‐negative group during the follow‐up period in this cohort, we further tested the value of combined serum LG2m and AFP levels for HCC risk prediction. In this setting, either LG2m or AFP was elevated in 153 CHC patients, and HCC development was not observed in 246 LG2m‐ and AFP‐negative patients (Fig. 6D).
We performed univariate Cox regression analysis to identify risk factors for HCC in these patients and found that thrombocytopenia, elevated APRI, FIB‐4 index, AFP, or LG2m were associated with a significantly increased risk of HCC (Table 1). Multivariate Cox regression analysis indicated that the significant independent risk factors for HCC were AFP (HR, 4.5; 95% CI, 1.1‐18.1; P = 0.034) and LG2m (HR, 12.2; 95% CI, 1.5‐101.8; P = 0.02).
TABLE 1 -
Univariate and Multivariate Cox Regression Analyses of HCC Risk
| Variables |
Univariate |
Multivariate |
| HR (95% CI) |
P Value |
HR (95% CI) |
P Value |
| Age (years; ≥60/<60) |
N.C. |
|
— |
|
| Sex (male/female) |
0.82 (0.21‐3.20) |
0.770 |
— |
|
| AST (≥40/<40 IU/mL) |
2.30 (0.29‐18.30) |
0.430 |
— |
|
| ALT (≥35/<35 IU/mL) |
1.57 (0.2‐12.4) |
0.670 |
— |
|
| Platelet count (≥10/<10 × 104/mm3) |
0.16 (0.045‐0.540) |
0.004 |
N.E. |
|
| Albumin (≥4.0, <4.0 g/dL) |
0.82 (0.17‐3.90) |
0.800 |
— |
|
| Diabetes (yes/no) |
0.90 (0.19‐4.20) |
0.890 |
— |
|
| Alcohol use (<20/≥20 g per day) |
N.C. |
|
— |
|
| Body mass index (kg/m2; ≥25/<25) |
0.85 (0.18‐4.00) |
0.840 |
— |
|
| APRI (≥1/<1) |
7.8 (2.0‐30.5) |
0.003 |
N.E. |
|
| FIB‐4 index (≥3.75/<3.75) |
9.3 (2.4‐36.0) |
0.001 |
N.E. |
|
| AFP (≥5/<5 ng/mL) |
8.9 (2.3‐34.6) |
0.002 |
4.5 (1.1‐18.1) |
0.034 |
| DCP (≥40/<40 mAU/mL) |
N.C. |
|
— |
|
| LG2m (≥30/<30 pg/mL) |
19.7 (2.5‐155.1) |
0.005 |
12.2 (1.5‐101.8) |
0.020 |
Multivariate analysis was performed by the forward selection method (likelihood ratio) using all variables identified as HCC risk factors by univariate analysis with statistical significance (P < 0.05).
Abbreviations: N.C., not calculable; N.E., not eligible.
We further performed a subanalysis of HCC risk in this cohort based on liver fibrosis status, to evaluate the ability of LG2m measurement to predict HCC risk according to liver fibrosis/nonfibrosis status. We used platelet counts (fibrosis: platelet count, <15 × 104/mm3; cirrhosis: platelet count, <10 × 104/mm3) and the FIB‐4 index (fibrosis: FIB‐4, >2.67; cirrhosis: FIB‐4, >3.75) to stratify CHC patients into cirrhosis/noncirrhosis or fibrosis/nonfibrosis groups. Accordingly, 47 and 145 (defined by platelet counts) or 75 and 144 (defined by the FIB‐4 index) CHC patients were regarded as having liver cirrhosis and fibrosis, respectively (Supporting Fig. S4). In the liver cirrhosis group, LG2m elevation was associated with high risk of HCC development, although the finding did not reach statistical significance, most likely attributable to the small sample size of liver cirrhosis patients (Supporting Fig. S4A,E). Indeed, 12 LG2m normal patients with platelet counts <10 × 104/mm3 did not develop HCC during the follow‐up period (Supporting Fig. S4E). LG2m elevation was significantly associated with high risk of HCC in patients without cirrhosis (Supporting Fig. S4B,F). We further evaluated the value of LG2m measurement in liver fibrosis/nonfibrosis patients, based on the FIB‐4 index (Supporting Fig. S4C,D) and platelet counts (Supporting Fig. S4G,H). In this setting, LG2m elevation was significantly associated with increased risk of HCC development in both liver fibrosis and nonfibrosis CHC patients. Taken together, these data demonstrated that measurement of serum LG2m may provide useful information for evaluating HCC risk in CHC patients irrespective of their liver fibrosis status.
Discussion
HCC is a heterogeneous disease in terms of cellular morphology, behavior, and clinical outcomes.(23) Although AFP and DCP are serum markers used to detect HCC, they can reflect the limited population of heterogeneous HCC cells. Because AFP is an albumin family protein and DCP is an abnormally matured prothrombin, these markers also generally reflect the hepatocytic nature of tumor cells and cannot reflect the status of metastatic cells that have undergone epithelial‐mesenchymal transition (EMT) and have therefore lost their epithelial cell features.
In this study, we used a cell‐culture system and clinical HCC samples to show that LG2m is a marker for the diagnosis of metastatic HCC with mesenchymal cell features and poor prognosis when a cut‐off value of 60 pg/mL is applied. LG2m elevation potentially reflects the molecular HCC subclasses proposed by Hoshida (S1) and Boyault (G3), and its clinical application could foster the diagnosis of molecular HCC subclasses without the need for microarray analysis (Fig. 7). LG2m is a biomarker for detecting metastatic HCC that will develop EHS, and its measurement has had a profound impact on HCC diagnostic and treatment strategies. However, the diagnostic ability of serum LG2m in metastatic HCC was based on two small‐scale retrospective cohorts, and large‐scale validation is required. Furthermore, we found strong evidence that elevated serum LG2m, at a cut‐off value of 30 pg/mL, is a strong risk factor for HCC in CHC patients even after DAA treatment. Thus, using two different cut‐off values, LG2m measurement can serve as both a diagnostic and predictive marker for HCC. Because the number of CHC patients with SVR after DAA treatment has markedly increased worldwide, patients with elevated LG2m may benefit from intensive HCC surveillance for early diagnosis. LG2m levels can be measured using the automated CLIA system, and our findings can be applied at a low cost worldwide.
FIG. 7: Molecular HCC subclass defined by LG2m elevation. Serum LG2m elevation is correlated with poor survival and metastasis. LG2m‐positive HCC corresponds to Hoshida’s S1 and Boyault’s G3 subclasses, which exhibit activation of EMT signaling (e.g., TGF‐β) and expression of the metastatic cancer stem‐cell marker, CD90 (adapted from Hoshida et al.
( 3 )). Abbreviations: Akt, protein kinase B; CTNNB1, catenin beta 1; E2F1, E2F transcription factor 1.
Our data indicate that both serum AFP and LG2m levels decreased with DAA treatment, unlike DCP. AFP is produced in normal hepatic progenitors accompanied by impaired hepatocyte regeneration by HCV infection.(24) Because the emergence of hepatic progenitors is significantly correlated with a high risk of HCC,(23) elevated LG2m might also reflect the status of the normal hepatic progenitor cell population in the liver. In contrast, elevated DCP was mainly observed in CHC patients who were prescribed warfarin as anticoagulation therapy; 96% of these patients did not show elevated DCP in this study, indicating that it is not useful for assessing the risk of HCC in CHC patients.
Although LG2m is detected in the tissues and sera of CHC and HCC patients, it is still unclear how LG2m production is regulated. AFP and DCP are detected in immature hepatic progenitors or in HCC with hepatic epithelial cell features.(25) In contrast, no serum markers are thus far available to detect HCC with mesenchymal cell features.(26) As clarified by this study, LG2m is a tumor marker detected in mesenchymal CD90‐positive HCC cell lines and the Hoshida’s S1 and Boyault’s G3 subclasses. Because we previously demonstrated that the presence of CD90‐positive cells in HCC is characterized by the activation of EMT signaling with frequent later distant organ metastasis,(20,26) it is plausible that LG2m might be elevated as a consequence of EMT signaling activated by Wnt and TGF‐β. Consistent with this idea, we found a high frequency of EHS in LG2m‐high HCC. Laminin subunit gamma‐2, which encodes laminin γ2, is up‐regulated by EMT in several cancers.(27‐29) Furthermore, another group reported that Ln‐332 γ2 plays a crucial role in maintaining and supporting cell stemness as part of the specialized cancer stem‐cell niche.(30) Increased HCC risk represented by slightly elevated LG2m in CLD patients might be associated with its reflection of activated EMT signaling in the liver because the EMT is a hallmark of cancer progression.(31) Given that AFP and LG2m expression are unique and distinct, combined serum AFP and LG2m measurements may mechanistically be an ideal complementary method for risk prediction and HCC screening.
In conclusion, our data demonstrated that LG2m is a tumor marker for predicting the risk of HCC in CHC patients as well as diagnosing metastatic HCC with frequent later EHS. The utility of LG2m measurement for HCC diagnosis is currently under evaluation in other ethnicities, and future studies involving large‐scale, multicenter, prospective cohorts are needed to further evaluate the nature of LG2m‐high HCC. A multicenter, prospective, observational study (UMIN000039467) is now in the process of recruiting ~6,000 CLD patients with various etiologies, including HBV infection, alcohol, diabetes, and NAFLDs, and may unveil the utility of serum LG2m measurement in the prediction of HCC risk in the future.
Author Contributions
T.Y., N.K., M.S., and S.K. designed the study. T.Y., T.S., T.T., R.H., K.N., N.I., K.K., K.A., Y.S., T.Y., E.M., M.H., and Y.N. recruited participants. N.K., M.N., E.Y., and T.Y. measured LG2m. A.K. and S.K. performed image reviews. S.T., Y.I., K.Y., and T.M. managed the clinical trial data and performed the statistical analysis. T.Y. prepared the manuscript and created the figures.
Acknowledgment
We thank Mss. Masayo Baba, Masayo Matsumoto, Mai Nakayama, Nami Nishiyama, and Mikiko Nakamura for excellent technical assistance.
References
1. Villanueva A. Hepatocellular carcinoma. N Engl J Med 2019;380:1450‐1462.
2. Della Corte C, Triolo M, Iavarone M, Sangiovanni A. Early diagnosis of liver cancer: an appraisal of international recommendations and future perspectives. Liver Int 2016;36:166‐176.
3. Hoshida Y, Nijman SMB, Kobayashi M, Chan JA, Brunet JP, Chiang DY, et al. Integrative transcriptome analysis reveals common molecular subclasses of human hepatocellular carcinoma. Cancer Res 2009;69:7385‐7392.
4. Boyault S, Rickman DS, de Reyniès A, Balabaud C, Rebouissou S, Jeannot E, et al. Transcriptome classification of HCC is related to gene alterations and to new therapeutic targets. Hepatology 2007;45:42‐52.
5. Yamashita T, Kitao A, Matsui O, Hayashi T, Nio K, Kondo M, et al. Gd‐EOB‐DTPA‐enhanced magnetic resonance imaging and alpha‐fetoprotein predict prognosis of early‐stage hepatocellular carcinoma. Hepatology 2014;60:1674‐1685.
6. Thrift AP, El‐Serag HB, Kanwal F. Global epidemiology and burden of HCV infection and HCV‐related disease. Nat Rev Gastroenterol Hepatol 2017;14:122‐132.
7. Westbrook RH, Dusheiko G. Natural history of hepatitis C. J Hepatol 2014;61(1 Suppl.):S58‐S68.
8. Asselah T, Marcellin P, Schinazi RF. Treatment of hepatitis C virus infection with direct‐acting antiviral agents: 100% cure? Liver Int 2018;38(Suppl. 1):7‐13.
9. Carrat F, Fontaine H, Dorival C, Simony M, Diallo A, Hezode C, et al. Clinical outcomes in patients with chronic hepatitis C after direct‐acting antiviral treatment: a prospective cohort study. Lancet 2019;393:1453‐1464.
10. Ioannou GN, Feld JJ. What are the benefits of a sustained virologic response to direct‐acting antiviral therapy for hepatitis C virus infection? Gastroenterology 2019;156:446‐460.e2.
11. Kanwal F, Kramer J, Asch SM, Chayanupatkul M, Cao Y, El‐Serag HB. Risk of hepatocellular cancer in HCV patients treated with direct‐acting antiviral agents. Gastroenterology 2017;153:996‐1005.e1.
12. Roche B, Coilly A, Duclos‐Vallee JC, Samuel D. The impact of treatment of hepatitis C with DAAs on the occurrence of HCC. Liver Int 2018;38(Suppl. 1):139‐145.
13. Marrero JA, Kulik LM, Sirlin CB, Zhu AX, Finn RS, Abecassis MM, et al. Diagnosis, staging, and management of hepatocellular carcinoma: 2018 practice guidance by the American Association for the Study of Liver Diseases. Hepatology 2018;68:723‐750.
14. WHO . Progress report on access to hepatitis C treatment: focus on overcoming barriers in low‐ and middle‐income countries. 2018.
apps.who.int/iris/handle/10665/260445. Accessed May 22, 2020.
15. Koshikawa N, Moriyama K, Takamura H, Mizushima H, Nagashima Y, Yanoma S, et al. Overexpression of laminin gamma2 chain monomer in invading gastric carcinoma cells. Cancer Res 1999;59:5596‐5601.
16. Kiyokawa H, Yasuda H, Oikawa R, Okuse C, Matsumoto N, Ikeda H, et al. Serum monomeric laminin‐gamma2 as a novel biomarker for hepatocellular carcinoma. Cancer Sci 2017;108:1432‐1439.
17. Koshikawa N, Minegishi T, Nabeshima K, Seiki M. Development of a new tracking tool for the human monomeric laminin‐gamma 2 chain in vitro and in vivo. Cancer Res 2008;68:530‐536.
18. Hasumura S, Sujino H, Nagamori S, Kameda H. Establishment and characterization of a human hepatocellular carcinoma cell line JHH‐4. Hum Cell 1988;1:98‐100.
19. Homma S, Nagamori S, Fujise K, Hasumura S, Sujino H, Matsuura T, et al. Establishment and characterization of a human hepatocellular carcinoma cell line JHH‐7 producing alpha‐fetoprotein and carcinoembryonic antigen—changes in secretion of AFP and CEA from JHH‐7 cells after heat treatment. [in Japanese] Hum Cell 1990;3:152‐157.
20. Yamashita T, Honda M, Nakamoto Y, Baba M, Nio K, Hara Y, et al. Discrete nature of EpCAM(+) and CD90(+) cancer stem cells in human hepatocellular carcinoma. Hepatology 2013;57:1484‐1497.
21. Kokudo N, Hasegawa K, Akahane M, Igaki H, Izumi N, Ichida T, et al. Evidence‐based clinical practice guidelines for hepatocellular carcinoma. The Japan Society of Hepatology 2013 update (3rd JSH‐HCC Guidelines). Hepatol Res 2015;45:e12464.
22. Villanueva A, Hoshida Y, Battiston C, Tovar V, Sia D, Alsinet C, et al. Combining clinical, pathology, and gene expression data to predict recurrence of hepatocellular carcinoma. Gastroenterology 2011;140:1501‐1512.e2.
23. Yamashita T, Wang XW. Cancer stem cells in the development of liver cancer. J Clin Invest 2013;123:1911‐1918.
24. Yamashita T, Forgues M, Wang W, Kim JW, Ye Q, Jia H, et al. EpCAM and alpha‐fetoprotein expression defines novel prognostic subtypes of hepatocellular carcinoma. Cancer Res 2008;68:1451‐1461.
25. Yamashita T, Ji J, Budhu A, Forgues M, Yang W, Wang H, et al. EpCAM‐positive hepatocellular carcinoma cells are tumor‐initiating cells with stem/progenitor cell features. Gastroenterology 2009;136:1012‐1024.
26. Yamashita T, Kaneko S. Orchestration of hepatocellular carcinoma development by diverse liver cancer stem cells. J Gastroenterol 2014;49:1105‐1110.
27. Drake JM, Barnes JM, Madsen JM, Domann FE, Stipp CS, Henry MD. ZEB1 coordinately regulates laminin‐332 and {beta}4 integrin expression altering the invasive phenotype of prostate cancer cells. J Biol Chem 2010;285:33940‐33948.
28. Moon YW, Rao G, Kim JJ, Shim HS, Park KS, An SS, et al. LAMC2 enhances the metastatic potential of lung adenocarcinoma. Cell Death Differ 2015;22:1341‐1352.
29. Pei YF, Liu J, Cheng J, Wu WD, Liu XQ. Silencing of LAMC2 reverses epithelial‐mesenchymal transition and inhibits angiogenesis in cholangiocarcinoma via inactivation of the epidermal growth factor receptor signaling pathway. Am J Pathol 2019;189:1637‐1653.
30. Govaere O, Wouters J, Petz M, Vandewynckel YP, Van den Eynde K, Van den broeck A, et al. Laminin‐332 sustains chemoresistance and quiescence as part of the human hepatic cancer stem cell niche. J Hepatol 2016;64:609‐617.
31. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell 2011;144:646‐674.
