Hepatitis B virus (HBV) is a hepatotropic virus causing hepatitis, cirrhosis and hepatocellular carcinoma (HCC). The methylation status of the HBV DNA in its different forms can potentially provide insight into the pathogenesis of HBV-related liver diseases, including HCC, however this is unclear. The goal of this study is to obtain comprehensive DNA methylation profiles of the three putative CpG islands in the HBV DNA in infected livers, with respect to liver disease progression. The extent of methylation in these CpG islands was first assessed using bisulfite PCR sequencing with a small set of tissue samples, followed by analysis using both quantitative bisulfite-specific PCR and quantitative methylation-specific PCR assays in a larger sample size (n = 116). The level of HBV CpG island 3 methylation significantly correlated with hepatocarcinogenesis. We also obtained, for the first time, evidence of rare, non-CpG methylation in CpG island 2 of the HBV genome in infected liver. Comparing methylation of the HBV genome to three known HCC-associated host genes, APC, GSTP1, and RASSF1A, we did not identify a significant correlation between these two groups.

Hepatocellular carcinoma (HCC) has a 5-year survival rate of <10% because it is difficult to diagnose early. Mutations in the TP53 gene are associated with approximately 50% of human cancers. A hotspot mutation, a G:C to T:A transversion at codon 249 (249T), may be a potential DNA marker for HCC screening because of its exclusive presence in HCC and its detection in the circulation of some patients with HCC. A locked nucleic acid clamp-mediated PCR assay, followed by melting curve analysis (using the SimpleProbe), was developed to detect the TP53 249T mutation. In this assay, the locked nucleic acid clamp suppressed 10(7) copies of wild-type templates and permitted detection of 249T-mutated template, with a sensitivity of 0.1% (1:1000) of the mutant/wild-type ratio, assessed by a reconstituted standard within 2 hours. With an amplicon size of 41 bp, it detects target DNA sequences in short fragmented DNA templates. The detected mutations were validated by DNA sequencing analysis. We then tested DNA isolated from urine samples of patients with HCC for p53 mutations and identified positive TP53 mutations in 9 of 17 samples. The possibility of using this novel TP53 249T assay to develop a urine or blood test for HCC screening is discussed.

Hypermethylation of the glutathione S-transferase π 1 (GSTP1) gene promoter region has been reported to be a potential biomarker to distinguish hepatocellular carcinoma (HCC) from other liver diseases. However, reports regarding how specific a marker it is have ranged from 100% to 0%. We hypothesized that, to a large extent, the variation of specificity depends on the location of the CpG sites analyzed. To test this hypothesis, we compared the methylation status of the GSTP1 promoter region of the DNA isolated from HCC, cirrhosis, hepatitis, and normal liver tissues by bisulfite–PCR sequencing. We found that the 5′ region of the position −48 nt from the transcription start site of the GSTP1 gene is selectively methylated in HCC, whereas the 3′ region is methylated in all liver tissues examined, including normal liver and the HCC tissue. Interestingly, when DNA derived from fetal liver and 11 nonhepatic normal tissue was also examined by bisulfite-PCR sequencing, we found that methylation of the 3′ region of the promoter appeared to be liver-specific. A methylation-specific PCR assay targeting the 5′ region of the promoter was developed and used to quantify the methylated GSTP1 gene in various diseased liver tissues including HCC. When we used an assay targeting the 3′ region, we found that the methylation of the 5′-end of the GSTP1 promoter was significantly more specific than that of the 3′-end (97.1% vs. 60%, p<0.0001 by Fisher's exact test) for distinguishing HCC (n = 120) from hepatitis (n = 35) and cirrhosis (n = 35). Encouragingly, 33.8% of the AFP-negative HCC contained the methylated GSTP1 gene. This study clearly demonstrates the importance of the location of CpG site methylation for HCC specificity and how liver-specific DNA methylation should be considered when an epigenetic DNA marker is studied for detection of HCC.

Hypermethylation of the promoter of the tumor suppressor gene, adenomatous polyposis coli (APC), occurs in various malignancies, including hepatocellular carcinoma (HCC). However, reports on the specificity of the methylation of the APC gene for HCC have varied. To gain insight into how these variations occur, bisulfite PCR sequencing was performed to analyze the methylation status of both sense and antisense strands of the APC gene in samples of HCC tissue, matched adjacent non-HCC liver tissue, hepatitis, cirrhosis, and normal liver tissues. DNA derived from fetal liver and 12 nonhepatic normal tissue was also examined. These experiments revealed liver-specific, antisense strand-biased CpG methylation of the APC gene and suggested that, although methylation of the antisense strand of the APC gene exists in normal liver and other non-HCC disease liver tissue, methylation of the sense strand of the APC gene occurs predominantly in HCC. To determine the effect of the DNA strand on the specificity of the methylated APC gene as a biomarker for HCC detection, quantitative methylation-specific PCR assays for sense and antisense strand DNA were developed and performed on DNA isolated from HCC (n = 58), matched adjacent non-HCC (n = 58), cirrhosis (n = 41), and hepatitis (n = 39). Receiver operating characteristic curves were constructed. With the cutoff value set at the limit of detection, the specificity of sense and antisense strand methylation was 84% and 43%, respectively, and sensitivity was 67.2% and 72.4%, respectively. This result demonstrated that the identity of the methylated DNA strand impacted the specificity of APC for HCC detection. Interestingly, methylation of the sense strand of APC occurred in 40% of HCCs from patients with serum AFP levels less than 20 ng/mL, suggesting a potential role for APC as a biomarker to complement AFP in HCC screening.

Aim Aberrant methylation of the promoter, P2, and the first exon, E1, regions of the tumor suppressor gene RASSF1A, have been associated with hepatocellular carcinoma (HCC), albeit with poor specificity. This study analyzed the methylation profiles of P1, P2 and E1 regions of the gene to identify the region of which methylation most specifically corresponds to HCC and to evaluate the potential of this methylated region as a biomarker in urine for HCC screening. Methods Bisulfite DNA sequencing and quantitative methylation-specific polymerase chain reaction assays were performed to compare methylation of the 56 CpG sites in regions P1, P2 and E1 in DNA isolated from normal, hepatitic, cirrhotic, adjacent non-HCC, and HCC liver tissue and urine samples for the characterization of hypermethylation of the RASSF1A gene as a biomarker for HCC screening. Results In tissue, comparing HCC (n = 120) with cirrhosis and hepatitis together (n = 70), methylation of P1 had an area under the receiver operating characteristics curve (AUROC) of 0.90, whereas methylation of E1 and P2 had AUROC of 0.84 and 0.72, respectively. At 90% sensitivity, specificity for P1 methylation was 72.9% versus 38.6% for E1 and 27.1% for P2. Methylated P1 DNA was detected in urine in association with cirrhosis and HCC. It had a sensitivity of 81.8% for α-fetoprotein negative HCC. Conclusion Among the three regions analyzed, methylation of P1 is the most specific for HCC and holds great promise as a DNA marker in urine for screening of cirrhosis and HCC.

In this paper, we study differentiable neural architecture search (NAS) methods for natural language processing. In particular, we improve differentiable architecture search by removing the softmax-local constraint. Also, we apply differentiable NAS to named entity recognition (NER). It is the first time that differentiable NAS methods are adopted in NLP tasks other than language modeling. On both the PTB language modeling and CoNLL-2003 English NER data, our method outperforms strong baselines. It achieves a new state-ofthe-art on the NER task.

No abstract

Compared with AI and WI, the superior ADR of WE offsets the drawback of a significantly longer insertion time. For quality improvement focused on increasing adenoma detection, WE is preferred over WI. The hypothesis that WE could lower the risk of interval colorectal cancers and related deaths should be tested.