Gene Expression Profiling Reveals Potential Biomarkers Of Human Hepatocellular Carcinoma Including Those With Normal Alpha-Fetoprotein Level And Smaller Tumors

Hepatocellular carcinoma (HCC), a common cancer worldwide, it has been the third cancer killer in the world and the second cancer killer in China since the 1990s. Globally, the five year survival rate of HCC is less than 5% and approximately 598,000 HCC patients die each year. The high mortality is mainly attributed to the inability to diagnose HCC patients at an early stage. Currently, alpha-fetoprotein (AFP) is the only diagnostic serum marker that universally accepted for HCC patients. However, elevated serum AFP is only observed in about 60-70% of HCC patients and to a lesser extent (33-65%) in patients with smaller HCCs. Moreover, nonspecific elevation of serum AFP has been found in 15-58% of patients with chronic hepatitis and 11-47% patients with liver cirrhosis. Therefore, it is necessary to identify new serological HCC biomarkers that have a sufficient sensitivity and specificity for the diagnosis of HCC patients, especially in AFP negative (≤20 ng/ml) and/or smaller HCC cases.In this study, we utilized the microarray technique to determine the expression profiles of 218 HCC specimens from patients with either high (AFP≥300 ng/ml) or low (AFP<300 ng/ml) serum AFP. We get 46 potential candidates with average expression ratio ≥2-fold in AFP-negative cases, then we subjectively selected 5 genes as potential diagnostic markers based on the following considerations: 1) their expression levels were relatively high in most of the low AFP tumor samples as compared to the matched non-cancerous liver tissues; 2) their encoded proteins are either novel or secretory and thus, could potentially be detected in serum or plasma samples. The 5 genes: GPC3, PEG 10, MDK, SERPIN11, and QP-C, are all overexpressed in HCCs including those with low serum AFP. For example, 92% of low AFP HCC cases had a ≥2-fold expression in the MDK gene while 69%-81% of low AFP HCC cases had a ≥2-fold expression in the GPC3 gene.Using quantitative real-time polymerase-chain-reaction analyses, we validated the expression of these 5 genes in an additional 50 AFP-negative (≤20 ng/ml serum AFP) and 8 high AFP (≥300 ng/ml serum AFP) HCC specimens and 36 cirrhoticnoncancerous hepatic specimens. A significant increase in the expression of these 5 genes could be detected in most of the HCC samples including those with negative serum AFP and small tumor size. We found that the relative mean levels (fold increase as compared to a disease-free normal liver pool) of GPC3, PEG 10, MDK, SERPINI1 and QP-C were 26.9, 23.8, 40.2, 6.2, 4.8 in tumors, and 0.2, 0.2, 1.4, 1.1, 0.8 in noncancerous cirrhotic tissues, respectively. The difference between cancer and noncancerous cirrhotic tissues is statistically significant (p<0.0001; nonparametric Mann Whitney tests in all 5 genes).Using PAM (Prediction Analysis of Microarray) analysis with 10-fold crossvalidation resulted in 100% (36/36) correct classification of non-tumor cirrhotic samples, 84.5% (49/58) correct classification of HCC samples, 82% (41/50) correct classification of AFP negative HCC samples and 87.5% (35/40) correct classification of small tumor size HCC samples. Similar results were obtained when these samples were randomly partitioned into 50% training and 50% testing cases, a process that was repeated 5 times. We demonstrated that these 5 genes can accurately classify non-cancerous cirrhotic tissues and HCC tissues.We also performed sample prediction by PAM based on markers in various combinations, it appears that markers in combination provided a much better sensitivity and specificity than that of any of the single marker alone. For example, when use these 5 markers in combination for PAM analysis, the sentistivity is 84.5% and specificity is 100%; when just use GPC3 alone, the sentistivity is 100% while specificity is 17%. And we found that GPC3, PEG10 and MDK provided the most weight to discriminate HCC from non-tumor liver specimens.GPC3, MDK and SERPINI1 encode known serum proteins. Because of the availability of a commercial ELISA-based assay for midkine, we tested the potential utility of this marker using serum samples. We determined the level of serum midkine in an additional 64 HCC patients (cases) and 26 healthy blood donors (controls) using this commercial kit. The medians (25th and 75th percentiles) of serum midkine were 442.8 (288, 666.5) pg/ml in the cases and 11.65 (3.83, 42.3) pg/ml in the controls. A statistically significant difference (p<0.0001; two-tailed) was observed between cases and controls, as determined by unpaired t test with Welch's correction. Consistently a significant increase in serum midkine, encoded by MDK, was associated with HCC patients, including those with negative AFP [the medians (25th and 75th percentiles)were 445.6 (381.9, 673.4) pg/ml] and smaller tumors [the medians (25th and 75th percentiles) were 392.1 (250.9, 549.3) pg/ml].When using the median tumor size (5cm in diameter) as the cutoff, the medians (25th and 75th percentiles) of serum AFP were 1634 (21.85, 41910) ng/ml in the larger tumor group and 74.3 (11, 477) ng/ml in the smaller tumor group, we found that larger tumors (≥5cm) had significantly higher serum AFP than that of smaller tumors (<5cm) (p=0.02, nonparametric Mann Whitney tests). In contrast, the difference in serum midkine between small and large tumors was insignificant (p>0.05, nonparametric Mann Whitney tests), the medians (25th and 75th percentiles) of serum midkine were 553.1 (373.3, 682.8) pg/ml in the larger tumor group and 392.1 (250.9, 549.3) pg/ml in the smaller tumor group. These results suggest that serum AFP level may reflect the degree of a tumor burden while serum midkine is independent of the tumor size, and thus it is feasible to diagnose AFP-negative and small HCC with serum samples.We suggest that a diagnostic signature approach using a combined weight of these 5 biomarkers rather than a single marker, may enhance the identification accuracy of HCC patients, including those with negative serum AFP and smaller sized tumors.We demonstrate that these 5 genes provide sufficient weights to separate HCC specimens from cirrhotic and non-cancerous liver specimens. A significant elevation of their expressions can be found in most of the HCC specimens, including cases with negative serum AFP and small tumor size. These results warrant further efforts in the development of serum-based detection assays for GPC3 and SERPINI1, and possibly PEG10 and QP-C, and an algorithm by combining the readings of 5 markers as a potential early indicator for HCC, suggesting that the products of these 5 genes may serve as biomarkers, in concert with AFP to aid HCC diagnosis.