Establishment And Application Of Clinical Predicting Algorithm Models For Liver Cirrhosis,Liver Fibrosis And Primary Liver Cancer Based On Serum N-Glycan Fingerprint Profiling Technology

Glycosylation is among the most abundant and complex protein posttranslational modification and has been caught much attention in the diagnosis of liver diseases including liver cirrhosis and hepatocellular carcinoma(HCC)due to the special effect hepatocytes played in protein glycosylation.Although the study of glycosylation is complicated technically,lectin microarray,lectin blot,and other technologies were proved to be rapid and highly sensitive for N-glycan profiling in our previous study.In the present study,we used the optimize N-glycan Fingerprint(NGFP)profiling technology for the first time and applied the NGFP technology to liver cirrhosis(LC),liver fibrosis(LF)and primary liver cancer(PLC)serum samples.We have investigated the relationship between N-glycan structure abundances and the progression of LC/LF and PLC.These results should be helpful for the prediction of LC/LF and PLC and provide potential therapeutic targets and personalized medical treatment for LC/LF and PLC with the noninvasive and high-throughput diagnostic methods.Part 1: The establishment of N-glycan fingerprint technology to detect the Nglycan structure abundances of serum and conducting the reference intervals of N-glycan structure abundancesAs the NGFP profiling technology,we established in this study,to ensure the accuracy and reliability of the N-glycan test results,this part of the study refer to the relevant standards of Clinical and Laboratory Standards Institute(CLSI)and other related standards for the performance validation and the establishment of reference interval.Precision analysis results revealed that the N-glycan structure abundances displayed by NGFP profiling technology,except for NA3Fc(Peak10),all met the intrabatch precision <10% and inter-batch precision <15%.In addition,the interference test results showed that the interfering substances have no obvious interference to the detection system(CV<10%).The N-glycan structure abundances showed statistically significant in different genders and ages of healthy people(P <0.05).Therefore,we established stratified reference intervals for different ages(18-30 years,30-50 years,≥50 years)and different genders(male,female).Furthermore,we validate the reference interval,the results showed that most of 13 N-glycan structure abundances in the validation cohort were included in the established reference interval.These results revealed that the NGFP profiling technology and the reference interval established in this part of the study have certain clinical applicability,which established a foundation for the future clinical transformation and clinical application of the technology.Part 2: Serum N-glycan fingerprint predicts liver cirrhosis and fibrosis.Liver fibrosis(LF)and liver cirrhosis(LC)underlies hepatocellular carcinoma(HCC)in most patients and diminishes the postoperative prognosis as a risk factor for operative complications,postoperative liver dysfunction and tumor recurrence.In addition,early detection of liver fibrosis is an important indicator for therapeutic intervention so as to stop or even reverse the fibrosis.Therefore,proper assessment of liver fibrosis and liver cirrhosis is crucial to support therapeutic decisions,determine surveillance intervals,and predict clinical outcome.Based on NGFP profiling technology,a multicenter study was designed enrolling totally 1539 participants with LC/LF,chronic HBV infection(CHB),and healthy negative controls(NC).The results revealed that NGA2FB(Peak2)and NA3(Peak8)showed the opposite trend and the biggest difference among the three groups(P<0.001).Furthermore,we analyzed its diagnostic performance for the prediction of LC and compared with FIB-4.The area under the receiver operating characteristic curve(AUC)of log(P2/P8)was 0.888(95%CI: 0.864-0.911)in the training cohort,which showed better diagnostic power than FIB-4(AUC: 0.802,95% CI: 0.732-0.872)(P<0.001).In the validation cohort,log(P2/P8)and FIB-4 showed similar diagnosis performance for the differentiation of LC and nonLC,and the AUCs were all over 0.900.Furthermore,we found that log(P2/P8)was feasible for assessment of fibrosis progression/regression.The AUCs of log(P2/P8)were all over 0.750 for the prediction of significant liver fibrosis(S0-1 vs S2-4),severe liver fibrosis(S0-2 vs S3-4)and liver cirrhosis(S0-3 vs S4).In addition,we also found that log(P2/P8)had a good correlation with liver fibrosis inflammation grad,ChildPugh classification,and clinical liver function indicators.This study suggested that the detection of N-glycan structure abundances can effectively predict the presence of LC and the stages of LF,so as to give CHB patients timely and effective antiviral treatment,and could reduce the need for invasive liver biopsy and partially helpful in relieving the heavy clinical management burden of CHB in HBV infection endemic region including China.Part 3: Identification and assessment of N-glycomic based biomarkers for hepatocellular carcinoma with N-glycan fingerprint profiling technologyHepatocellular carcinoma(HCC)is the most frequent liver cancer and is associated to high morbidity and mortality rates.It presents poor prognosis,generally due to its late diagnostic.The early diagnosis of HCC is of great clinical desirable due to the improved prognosis of HCC if the patients could get surgical treatment early.Fortunately,we found that NA2(Peak5)and NA3Fb(Peak9)were decreased in nonHCC and increased in HCC,while NGA2F(Peak1),NGA2FB(Peak2),NG1A2F(Peak3/Peak4),NA2F(Peak6),NA2FB(Peak7)were increased in non-HCC and decreased in HCC.To make full use of the differences,we used the log ratio of these N-glycan structure abundances [log(P5/P6)、log(P9/P3)、log(P5/P4)、log(P9/P7)] to differentiate HCC from non-HCC.The AUCs showed that the diagnostic efficacies of AFP(AUC: 0.820,95%CI: 0.800-0.840)was better than the N-glycan markers(AUC:over 0.700).Therefore,we further combined with clinical liver function indicators to construct a diagnosis model for the prediction of the presence of HCC(Pre HCC =0.054 × Age-0.020 × TBIL-0.106 × TP + 0.180 × ALB + 4.879 ×log(P5/P6)+ 2.129 × log(P9/P3)-1.369 × A/G-0.357).In the training and validation cohort,the AUCs of the diagnostic model Pre HCC were all over 0.810,which showed similar diagnostic power of AFP.Especially,Pre HCC also showed an excellent diagnostic efficiency in identifying HCC with negative AFP and non-HCC in the training and validation cohort(AUCs :0.800).In addition,a follow-up study was conducted on LC patients.Among them,11 LC patients who were diagnosed with HCC during the follow-up study were included.Serum samples at the time of HCC diagnosis and at weeks-24 and-48(24 and 48 weeks before the diagnosis of HCC)and at the corresponding time points from match controls were tested.At HCC-24 w diagnosis,the area under the curves(AUCs)for Pre HCC and AFP were 0.730(95%CI: 0.532-0.928)and 0.824(95%CI: 0.645-1.000).The sensitivity of both Pre HCC and AFP were all 72.72%(8/11).At HCC-48 w diagnosis,The Pre HCC showed a better diagnostic performance than AFP.The AUCs of Pre HCC and AFP were 0.789(95%CI: 0.628-0.950)and 0.599(95%CI: 0.355-0.843).The sensitivity of Pre HCC and AFP were80.00%(8/10)and 36.36%(4/11).In conclusion,by using NGFP profiling technology in patients with HCC and non-HCC,we found that the combination of NGFP and clinical biochemical indicators(Pre HCC)could be used as a supplementary test to AFP.Part 4: Based on N-glycan fingerprint and clinical laboratory indicators to predict the vascular invasion in hepatocellular carcinomaAlthough liver resection and liver transplantation could offer a promising for selected patients,the high postoperative recurrence rate has impaired long-time survival.Among various risk factors,vascular invasion,including microvascular invasion(MVI)and macrovascular invasion,has been proven to be an independent risk factor for predicting high recurrence and poor survival rate.An accurate preoperative estimation of the presence of vascular invasion can help surgeons choose appropriate surgical procedures for patients.This part of the study enrolled 1839 consecutive patients who had undergone liver resection for histologically confirmed HCC from 2015 to 2018.Eligible patients who underwent surgery between January 2015 and June 2016 were included into the training cohort(n=961)for the development of the diagnostic model,and those who underwent surgery between June 2016 and December 2018 were entered into the validation cohort(n=878).In the training cohort,we found that NG1A2F(Peak4),NA3Fb(Peak9)and NA3Fc(Peak10)were the independent risk factors of vascular invasion(VI)of HCC patients.Based on the VI related N-glycan structure and clinical laboratory indexes,we constructed the vascular invasion predict model named VI-G+ by logistic regression analysis(VI-G+=-0.015 × Age + 0.139 × Peak9 +1.041 × Peak10-0.031 × ALB + 0.158 × log AFP + 0.338 × log PIVKA-II +0.619 × AFP-L3 + 0.577).The AUC of VI-G+ was 0.739(95%CI: 0.708-0.711)in the training cohort for the prediction the presence of VI in HCC patients.Especially,when identifying M2 from M0 and PVTT from M0-M2,the AUCs of VI-G+ were all over 0.700.Kaplan-Meier curve and COX regression analysis further confirmed that VI-G+ was an independent risk factor of the overall survival rate of postoperative HCC patients.These results suggested that the use of the predict model(VI-G+)in estimating the risk of a patient harboring VI to direct clinical treatment is a new concept and has broad application prospects.Part 5: Based on N-glycan and clinical laboratory parameters to discriminate intrahepatic cholangiocarcinoma from benign liver disease and hepatocellular carcinomaPrimary liver cancer can be roughly divided into three main subtypes:hepatocellular carcinoma(HCC),intrahepatic cholangiocarcinoma(ICC)and mixed hepatocellular cholangiocarcinoma according to different cell origin.ICC ranks the second most prevalent primary hepatic malignancy,accounting for 10-15% of PLC,and bears a 5-year survival of only about 5%.As treatment modalities and clinical outcomes of ICC and HCC differ significantly,confidently discriminating between ICC and HCC before making a medical decision has attracted more emphasis.Therefore,in the study,we constructed two algorithm models for the early detection of ICC and discrimination ICC from HCC,respectively.Between 2014 and 2018,244 patients with ICC,270 patients with HCC and 59 patients with benign liver disease were enrolled to construct diagnostic models for early diagnosis of ICC and identification ICC from HCC,respectively.When differentiating ICC and non-ICC,we found that NA2(Peak5)was the independent risk factor of ICC.Furthermore,we construct a diagnostic model for the early diagnosis of ICC based on N-glycan marker and clinical laboratory test(ICC-G+ =-1.382 × Sex + 0.432 × Peak5 + 0.082 × CEA + 0.005 × CA199-15.635(Male = 1,Female = 2)).ICC-G+ showed an excellent diagnostic efficiency for the early detection of ICC in the training cohort,the AUC of ICC-G+ was 0.911(95%CI: 0.844-0.978).In the validation cohort,ICC-G+ displayed an AUC of 0.800(95%CI: 0.677-0.907)for the estimation of ICC risk.For the discrimination of ICC and HCC,we also developed an algorithm model named IHCC-G+(IHCC-G+ =0.020 × AFP-0.057 × CA199 + 2.553 × Peak2 + 18.448 × Peak10-7.190).The IHCC-G+ also showed a good performance for identification ICC and HCC,the AUCs of IHCC-G+ were all over 0.800 in the training and validation cohort.These results indicated that the serum N-glycan is an innovative approach to screening patients with ICC form patients with BLD and HCC.The usefulness of N-glycan markers for screening,follow-up,and management of patients with ICC should be evaluated further.