Hu et al. J Transl Med (2021) 19:359 https://doi.org/10.1186/s12967-021-03038-3 Journal of Translational Medicine

RESEARCH Open Access Use of proteomics to identify mechanisms of hepatocellular carcinoma with the CYP2D6*10 polymorphism and identifcation of ANGPTL6 as a new diagnostic and prognostic biomarker

Guiming Hu1, Fei Gao1, Guanzhe Wang1, Yan Fang1, Yuanyuan Guo1, Jun Zhou1,2, Yuhan Gu1, Cunzhen Zhang1, Na Gao1, Qiang Wen1 and Hailing Qiao1* Abstract

Background: Although an association between the cytochrome P4502D6 (CYP2D6) *10 (100C>T) polymorphism and hepatocellular carcinoma (HCC) is known, the mechanism remains unclear. Here we aimed to explore mecha- nisms of CYP2D6*10 (100C>T) polymorphism conferring to HCC, and screen markers for HCC. Methods: Label-free global proteome profling with 34 normal livers and peritumor tissue from 61 HCC patients was performed, and angiopoietin-like protein-6 (ANGPTL6) was evaluated in 2 liver samples validation cohorts and 2 blood specimens validation cohorts. Results: We found a signifcantly decreased frequency of TT in HCC patients which reduced HCC susceptibility by 69.2% and was accompanied by lowered enzymatic activity for CYP2D6. Proteomic analysis revealed 1342 diferentially expressed proteins (DEPs) that were associated with HCC and 88 DEPs were identifed as 100 TT-related proteins, likely underlying the susceptibility to HCC. Twenty-two upregulated DEPs and 66 downregulated DEPs were mainly related to lipid metabolism and the extracellular matrix, respectively. High ANGPTL6 was associated with a higher risk to HCC and worse prognosis. ANGPTL6 was both an independent risk factor and an independent prognostic factor for HCC and exhibited strong poten- tial for predicting HCC occurrence, with comparable AUC values and higher sensitivity compared with alpha-fetoprotein. Conclusions: The TT genotype-associated decreased risk of HCC appears to be related to lowered CYP2D6 activ- ity and altered protein expression in the tumor microenvironment, and ANGPTL6 is a promising new diagnostic and prognostic biomarker for HCC. Our fndings reveal new mechanistic insights for polymorphisms related to HCC risk and provide avenues for screening for HCC. Keywords: Polymorphisms, CYP2D6, Hepatocellular carcinoma, ANGPTL6, Proteomics

Background Numerous studies have reported the important role of gene polymorphisms in the development of numerous *Correspondence: [email protected] diseases [1–5]. However, there is still a lack of under- 1 Institute of Clinical Pharmacology, Zhengzhou University, lying mechanisms for these polymorphisms. Indeed, Zhengzhou 450052, Henan, China Full list of author information is available at the end of the article

© The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat​ iveco​ mmons.​ org/​ licen​ ses/​ by/4.​ 0/​ . The Creative Commons Public Domain Dedication waiver (http://creat​ iveco​ ​ mmons.org/​ publi​ cdoma​ in/​ zero/1.​ 0/​ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Hu et al. J Transl Med (2021) 19:359 Page 2 of 13

phenotypic changes related to diseases caused by geno- reported to be most closely associated with signifcantly typic changes involve numerous proteins and complex altered metabolic activity and susceptibility to HCC, both processes, making it difcult to determine mechanisms. in our and other studies [24–26]. However, the underly- However, currently available mass-spectrometry-based ing mechanism by which this 100C>T mutation impacts proteomics, which had evolved as the preferred method hepatocarcinogenesis is not established. Tus, to address for the analysis of complex proteomes, provided unprec- this problem, we performed label-free global proteome edented opportunities to elucidate the mechanisms by profling with 34 normal livers and peritumors of 61 HCC which gene polymorphisms lead to cancer. patients to elucidate changes in the tumor microenviron- Hepatocellular carcinoma (HCC), which is character- ment (TME) related to the CYP2D6*10 polymorphism. ized by high morbidity and mortality rates, is the most In addition, ANGPTL6 as a potential diagnosis and prog- common type of liver cancer and accounts for 90% of nostic biomarker of HCC was evaluated in the above liver cases [6]. Hepatocarcinogenesis is well known to be tissues and validated in the GEO database, another set of associated with numerous genetic polymorphisms that liver samples, and serum samples of 23 normal healthy are involved in its complex and multi-factorial develop- controls and 42 HCC patients. ment and contribute to malignant transformation [7–10]. To date, no efective intervention methods for HCC are based on genetics. Terefore, discovery of new genetic Methods polymorphisms in HCC and exploration of their mech- Study design anisms might provide new approaches for identifying Te study included a training cohort and four independ- novel therapeutic targets and designing new or personal- ent validation cohorts. Te workfow of our study was ized treatment for HCC. summarized in Fig. 1. In the training cohort 105 healthy Cytochrome P450 2D6 (CYP2D6), is one of the most subjects and 102 HCC patients who had undergone investigated CYPs in relation to its highly complicated hepatic surgery were included. Human and rat liver tis- polymorphisms, with 144 single nucleotide polymor- sues, parafn specimens, human serum samples and phisms (SNPs) reported [11], resulting in remarkable dif- ferences in the activity of this enzyme [12, 13]. In recent years CYP2D6 has gained increasing attention on those polymorphisms that are related disease susceptibility, due to altered metabolism of its metabolically abundant substrates including approximately 25% of therapeutic drugs (about 223 drugs) and known and potential carcin- ogens such as nitrosamine, nicotine, and cotinine found in tobacco [14]. An increasing evidence suggest that CYP2D6 polymorphisms are related to susceptibility to numerous diseases, including autoimmune diseases [15], Alzheimer’s disease [16] and Parkinson’s disease [17]. It has been demonstrated that the frequency of CYP2D6- inactivating mutations is signifcantly lower among liver cancer patients than in healthy controls and cirrhotic subjects who did not develop liver cancer [18]. Moreo- ver, CYP2D6 poor metabolizer genotypes were reported signifcantly more frequent in healthy subjects and car- riers than in hepatitis/cirrhosis and HCC patients [19]. Although association between CYP2D6 gene polymor- phisms and disease risk had been investigated in several studies, it is still unclear how CYP2D6 polymorphisms Fig. 1 Workfow of the study. Clarifcation of mechanisms of confer susceptibility to HCC. HCC with CYP2D6*10 polymorphisms and screening of potential Our previous study systematically assessed changes biomarkers using label-free global proteome analysis. CYP2D6, cytochrome P4502D6. CC and TT refer to CYP2D6*10 CC or TT in 10 major CYP450 enzymes in HCC patients [20] and genotype patients. ANGPTL6, angiopoietin‐like 6. AUC, area under found that diferences in CYP activity between physi- the curve. BP, biological processes. GO, gene ontology. Con, control. ological and pathological states were associated with RNA-seq, RNA sequencing. HCC, hepatocellular carcinoma. ELISA, increased risk for hepatofbrosis [21] and hepatocar- enzyme-linked immunosorbent assay. WB, western blotting. IHC, cinogenesis [1, 22, 23]. CYP2D6*10 (100C>T) has been immunohistochemistry Hu et al. J Transl Med (2021) 19:359 Page 3 of 13

GEO database (GSE142987) constituted the validation concentrations of substrate and 1 mM NADPH. Te cohorts. mixture was pre-incubated for 5 min at 37 °C. Opti- mal incubation time was 20 min for dextromethorphan Human liver samples O-demethylation. For biotransformation, 0.625– Normal liver samples were from hepatic haemangi- 960 μM dextromethorphan was added. Reactions were oma patients with normal liver function and histologic terminated by adding 20 μL ice-cold acetonitrile; sub- appearance. Peritumor liver specimens were obtained strate metabolite was determined by high performance about 2 cm distant from the tumor tissues from HCC liquid chromatography-fuorescence detection (HPLC- patients without tumor radiotherapy, chemotherapy FLD). Te values of the Michaelis constant ­(Km) and or targeted drug therapy before surgery. All liver speci- ­Vmax were determined by nonlinear regression analy- mens were stored in liquid nitrogen within 30 min of sis using GraphPad Prism 8.03 software. Te intrinsic resection and until use. Te data were recorded via an clearance ­(CLint) was calculated from the ratio of ­Vmax inpatient case record, including gender, age, laboratory to K­ m. tests such as aspartate amino transferase (AST), alanine aminotransferase (ALT), gamma-glutamyl transferase (GGT), alkaline phosphatase (ALP), platelets (PLT), Proteomics and data analysis alpha‐fetoprotein (AFP), bilirubin and so on. All 207 liver Proteomic analysis of 34 normal livers and peritumors tissue samples were collected at Te First Afliated Hos- of 61 HCC patients was performed by the State Key pital of Zhengzhou University, Henan Provincial People’s Laboratory of Proteomics, Beijing Proteome Research Hospital and Henan Provincial Tumor Hospital (Henan, Center, (Beijing, China). Briefy, proteins were pre- China); demographic characteristics are described in our pared by means of in-solution or in-gel digestion and previous research [24]. Parafn samples of human livers subjected to liquid chromatography tandem mass spec- included 6 normal livers, 6 peritumor tissues and tumor trometry analysis (Q-Exactive HF LC–MS/MS). Raw cores of HCC patients derived from the Second Afliated data from mass spectrometry were analyzed against Hospital of Zhengzhou University (Henan, China). the human UniProt protein sequence database (ver- sion 20140922, 20,193 sequences) with MaxQuant soft- Blood samples ware31 (version 1.5.3.8). Te false discovery rate (FDR) Serum samples were drawn at 6:00–8:00 AM after an for peptides was set to < 1%. Te maximum mass error overnight fast from a validation cohort that included 27 of the primary ion and product ion scans for the ini- healthy controls and 42 HCC patients from Te First tial mass spectrometry were set to 20 ppm and 0.5 Da, Afliated Hospital of Zhengzhou University. A single respectively. Te corrected error of the precursor venous blood sample (3–5 mL) was taken from each par- ion was set to 5 ppm. A 6947 × 95 protein expression ticipant under aseptic conditions. Serum was separated matrix was obtained from the 3.2.4 MaxQuant result fle. Intensity-based absolute protein quantifcation by centrifugation and stored at − 20 °C, avoiding repeated freeze–thaw cycles prior to analysis. (iBAQ) based on peak intensity was used to determine protein expression levels. Te R/Bioconductor package Genotypes of CYP2D6*10 (100C>T) limma v.3.24.15 34 was used to identify diferentially Genomic DNA was isolated from human liver tissues expressed proteins (DEPs) between normal liver tissues using a genomic DNA purifcation kit (QIAGEN Transla- and peritumor tissues or between CC and TT geno- tional Medicine Co., Ltd, China). Genotyping of CYP2D6 type groups. Proteins detected in < 50% of samples were 100C>T of all participants was determined by Polymer- excluded, and missing values were replaced with half ase Chain Reaction (PCR) sequencing according to our the minimal value for each protein. Further details of previous study [25]. the procedure for digesting proteins into peptides and proteomic analysis are described elsewhere [27]. Determination of enzyme activity of CYP2D6 Enzymic activity of CYP2D6 in human liver microtomes Functional enrichment analysis (HLMs) of 105 healthy subjects and 102 HCC patients was measured by metabolism of dextromethorphan Gene ontology (GO) analysis was performed to deter- (National Institute for the Food and Drug Control), mine the potential biological functions of ANGPTL6 a probe drug for CYP2D6. HLMs were prepared by and were carried out using Enrich GO function in the R hypothermal diferential centrifugation. Briefy, incu- package “clusterProfler”. GO analysis was according to bation mixtures contained HLMs (0.2 mg/mL pro- the threshold of P < 0.05 and q < 0.05. tein), 100 mM phosphate bufer (pH 7.4), diferent Hu et al. J Transl Med (2021) 19:359 Page 4 of 13

Validation of Angiopoietin‑like protein‑6 (ANGPTL6) and Graphpad Prism 8.3.0 software. Te data were in liver tissues summarized as mean ± SE. Diferences between two Expression of ANGPTL6 in livers from 6 normal sam- independent groups were tested with the Wilcoxon ples and 6 tumor-extrinsic tissues of HCC patients was Mann–Whitney test. Receiver operating characteris- determined by western blotting (WB) with primary anti- tics (ROC) curves were constructed to assess sensitivity, ANGPTL6 antibody (GTX118035, diluted at 1:1000). specifcity, and respective areas under the curves (AUCs). Further validation of ANGPTL6 in parafn samples A two-tailed P value < 0.05 was regarded as statistically including 6 human normal livers and 6 tumor-extrinsic signifcant. tissues of HCC patients was performed by immunohis- tochemistry (IHC) using anti-ANGPTL6 (GTX118035) Results antibody at 1:500 dilution. Antigen Retrieval: Trilogy™ Decreasing susceptibility to HCC is skewed in 100TT HCC (EDTA based, pH 8.0) bufer, 15 min. patients and associated with metabolic activity of CYP2D6 A Chi-square test revealed that the study population is Allograft transplantation model in Hardy–Weinberg equilibrium for the CYP2D6 gene Sprague Dawley (SD) rats (180–220 g), purchased from (P > 0.05), which suggest that the genotype frequency ZhengZhou University, were anesthetized with 300 mg/ distribution had a group representation (Additional kg chloral hydrate by intraperitoneal injection and fle 1: Table S1). Signifcantly decreased frequency of TT then placed in a supine position on the operating table. was discovered in HCC patients compared with healthy To expose the left lobe of the liver, a 1.5 cm longitudinal subjects (Fig. 2a, 16.7% vs 46.7%, p < 0.01), as evidenced incision was made below the xiphoid process. Walker256 by the frequency of CC, CT and TT (6.5%, 46.5% and 6 cells (1.5 × ­10 ) in 50 μL RPMI-1640 medium (GIBCO 16.7% respectively) in HCC patients and healthy subjects RPMI 1640, HCC model group, n = 7) or 50 μL serum- (31.4%, 21.9% and 46.7%, respectively) (Additional fle 2: free RPMI-1640 medium (Normal control group, n = 6) Table S2). were slowly injected into the left liver lobe using a micro- CYP2D6*10 polymorphisms were signifcantly asso- injector at a 30-degree angle. After allowing 3 min for the ciated with susceptibility to HCC, as the risk of HCC in cells to disperse into the liver tissue, the skin was sutured individuals carrying the TT genotype decreased by 69.2% and the surgical incision disinfected. Te rats were [(Odds Ratio (OR) = 0.308, 95% confdence interval (CI) kept in a Specifc Pathogen Free (SPF) warm incubator 0.147–0.644, P = 0.002)] when compared with CC geno- until they had recovered from the anesthesia and then type carriers (Additional fle 2: Table S2), suggesting returned to animal room. that the TT genotype is a protective factor for HCC. Te above fndings were confrmed by the overall survival Validation of ANGPTL6 in serum (OS) of HCC patients. A Kaplan–Meier survival curve Te serum level of ANGPTL6 in 23 healthy controls revealed that patients with the TT genotype had longer and 42 HCC patients was determined by ELISA. A OS than those with the CC genotype (median OS 702 vs human ANGPTL6 enzyme-linked immunosorbent assay 344 days, P = 0.365, Fig. 2b). Te survival diference did (ELISA) Kit (Catalog No. EK19979) was used to meas- not reach statistical signifcance, possibly due to the small ure ANGPTL6 levels according to the manufacturer’s number of cases. Tis conclusion was further substanti- instructions. Te analytical (linear) measurement range ated by the biochemical data obtained in HCC patients. was 0.312–20 ng/mL with sensitivity of 0.077 ng/mL. As shown in Table 1, platelets (PLT) and the albumin/ Te intra-assay coefcient of variation (CV) and inter- globulin ratio (A/G) were higher in patients with the assay CV of ANGPTL6 were less than 10% and 15% TT genotype (P = 0.022 and P = 0.020, respectively), respectively. In addition, quality controls were included while total bilirubin (TBIL), direct bilirubin (DBIL) and in all measurements with the results within the expected prothrombin time (PT) were lower when compared range. with individuals carrying the CC genotype (P = 0.009, P = 0.043, P = 0.043, respectively). Tere was no signif- Validation of ANGPTL6 in the GEO database cant diference in distribution of gender and age between Serum mRNA levels of ANGPTL6 in 30 healthy controls the two groups. Taken together, these results suggest that and 35 liver cancers were obtained from the GEO data- homozygous TT mutations may play an important role in base (GSE142987). the decreased susceptibility to HCC. To uncover the underlying mechanism, we next Statistical analysis assessed the effects of CYP2D6*10 polymorphisms Statistical analysis was performed using SPSS 23.0 soft- on enzyme activity. As measured by dextromethor- ware. Graphs were plotted by R software (Version 4.0.3) phan O-demethylation metabolism in human liver Hu et al. J Transl Med (2021) 19:359 Page 5 of 13

ab 50 ** ** Normal 1.0 CC f %)

)( TT no 40 HCC al 0.8 >T Median survival rviv

30 su 0.6 344 vs 702 0C istributio 10 yd 0( 20 0.4 Fraction 0.2 10 P2D6*1 Frequenc

CY 0.0 0 0500 1000 1500 2000 CC CT TT Survival days after surgery cd 12 12 ** * CC ) ) TT 9 9 *** min/mg min/mg l/ l/ µ µ ( (

6 nt 6 nt Li Li C C 3 3 P2D6 P2D6 CY CY 0 0 Normal HCC Normal HCC Fig. 2 CYP2D6*10 genotype frequency and enzyme activity in healthy subjects and HCC patients. a The frequency distribution of CYP2D6*10 genotypes in HCC patients and a normal population. **P < 0.01 vs heathy subjects. b Kaplan–Meier curve for overall survival (OS) of HCC patients with the CC and TT genotype, P 0.365. c CYP2D6 activity of in peritumor liver tissues of HCC patients and normal liver tissues. **P < 0.01 vs log-rank = heathy subjects. d CYP2D6 activity in subjects with the TT and CC genotype in healthy subjects and HCC patients. *P < 0.05, ***P < 0.001, TT group vs CC group. HCC, hepatocellular carcinoma; Healthy means healthy subjects. Data were expressed as mean SEM, P value was calculated by Mann– ± Whitney U test

Proteomic mechanisms of CYP2D6*10 polymorphisms microsomes, the intrinsic clearance (Cl­ int) of CYP2D6 increased significantly in HCC patients compared that confer susceptibility to HCC with normal controls (Median: 6.86 μL/min/mg vs To investigate the mechanism by which the CYP2D6*10 3.49 μL/min/mg, P < 0.01, Fig. 2c). What’s more, lower mutation reduces HCC susceptibility, we performed pro- metabolic activity was observed among TT genotype teomic analysis of liver tissues from 34 normal subjects subjects when compared with CC genotype individu- and 61 HCC patients to identify HCC related proteins. A als both in healthy samples (Median: 1.775 μL/min/ total of 1342 DEPs were identifed. We then focused on mg vs 5.079 μL/min/mg, P < 0.001) and in the HCC 18 CC genotypes and 10 TT genotypes derived from the group (Median: 1.792 μL/min/mg vs 6.228 μL/min/ above HCC patients to identify DEPs by using a flter cri- mg, P = 0.045, Fig. 2d). Taken together, these results terion of |log2-fold change (FC)|≥ 0.20 and P value < 0.05. suggest that a significantly reduced frequency of TT As shown in the volcano plot (Fig. 3a) and heat-map genotype was associated with decreased HCC suscep- (Fig. 3b), 88 statistically signifcant DEPs were identifed, tibility, at least partially, probably by decreasing the including 22 upregulated and 66 downregulated proteins metabolic activity of CYP2D6. (Additional fle 3: Table S3). Te gene ontology (GO) term enrichment biological process analysis revealed that upregulated DEPs were sig- nifcantly enriched in lipid transport, lipid localization, Hu et al. J Transl Med (2021) 19:359 Page 6 of 13

Table 1 The basic clinical characteristics of HCC patients with CC Subsequently, we investigated the diferential expres- (n 35) and TT (n 16) genotype = = sion of these CYP2D6*10-related DEPs in normal sub- Variables CC n (%) TT n (%) P jects and HCC patients. Surprisingly, an opposite trend further confrmed that the peritumoral microenviron- Age (year) 0.091 ment of these two genotypes was diferent. As shown in 50 13 (37.14) 10 (62.50) ≤ Fig. 3e, most of the up-regulated proteins (68.2%, 15 out > 50 22 (62.86) 6 (37.50) of 22) in the TT genotype were lowly expressed, while Gender 0.370 the vast majority of downregulated proteins (97.0%, 64 Man 31 (88.57) 13 (81.25) out of 66) in this genotype were highly expressed in HCC Female 4 (11.43) 3 (18.75) (Additional fle 4: Table S4). Te signifcantly upregulated PLT ( ­109/L) 0.022 × top nine DEPs—EFEMP1, CYP3A7, LOXL1, EFEMP2, < 180 23 (65.71) 5 (37.50) MFI2, CHST4, CCL21, NPNT, LTBP4 and ANGPTL6, 180 12 (34.29) 11 (62.50) ≥ and downregulated top fve DEPs—MIF2, ACE, SLC9B2, ALT (U/L) 0.360 GRAMD4 and ACACB in the peritumoral microenviron- 40 23 (65.71) 8 (50.00 ≤ ment of HCC are shown in Fig. 3f and g. Tese results > 40 12 (34.29) 8 (50.00) clearly demonstrate that patients with the TT genotype AST (U/L) 0.179 had lower risk of HCC that might result, at least in part, 40 18 (51.43) 5 (31.25) ≤ from an altered tumor-extrinsic microenvironment, per- > 40 17 (48.57) 11 (68.75) haps by upregulating proteins related to lipid metabolism Albumin/globulin 0.020 and downregulating proteins related to the extracellular 1.5 21 (60.00) 4 (33.33) ≤ matrix. > 1.5 14 (40.00) 12 (66.67) ALB (g/L) 0.355 Screening biomarkers for HCC based on proteomics 70 18 (51.43) 6 (37.50) ≤ A receiver operating characteristic (ROC) curve on the > 70 17 (48.57) 10 (62.50) 100C >T-related DEPs revealed that 30 of the 88 proteins TBIL (μmol/L) 0.009 showed high ability to distinguish HCC patients from 17 20 (57.14) 15 (93.75) ≤ healthy subjects with a high AUC (> 0.70) (Additional > 17 15 (42.86) 1 (6.67) fle 5: Table S5). Te top fve proteins with the AUC > 0.80 DBIL (μmol/L) 0.043 were ANGPTL6 (AUC 0.900, sensitivity 91.8%, specifcity 7 18 (51.43) 13 (81.25) ≤ 81.1%), LTBP1 (AUC 0.830, sensitivity 77.0%, specifcity > 7 17 (48.57) 3 (18.75) 88.2%), EFEMP2 (AUC 0.814, sensitivity 82.0%, specifcity ALP (U/L) 0.242 73.5%), EFEMP1 (AUC 0.813, sensitivity 78.7%, specifc- 130 27 (77.14) 15 (93.75) ≤ ity 79.4%) and KRT7 (AUC 0.804, sensitivity 77.0%, speci- > 130 8 (22.86) 1 (6.67) fcity 79.4%). Te largest AUC was for ANGPTL6, which AFP (ng/mL) 0.246 was comparable to serum AFP with an AUC of 0.900 300 17 (48.57) 5 (31.25) ≤ (Fig. 4a and b). Interestingly, the detection rate of AFP in > 300 18 (51.43) 11 (68.75) normal liver tissues and peritumoral tissues of HCC was PT (s) 0.043 2.94% (1/34) and 18.03% (11/61), respectively, which was 14 26 (74.29) 16 (100.00) ≤ much lower than those of the top fve proteins. What’s > 14 9 (25.71) 0 (0) more, the combination of ANGPTL6 and AFP increased P values were calculated from two‐side chi‐square test the diagnostic accuracy to a great extent when compared AFP: alpha-fetoprotein; ALB: albumin; ALT: alanine amino transferase; ALP: with ANGPTL6 or AFP alone, yielding an AUC of 0.963 alkaline phosphatase; AST: Aspartate amino transferase; DBIL: direct bilirubin; PLT: platelet; PT: prothrombin time; TBIL: total bilirubin and a sensitivity of 90.2% for HCC (Fig. 4c). Collectively, the above results show that in the parallel study the top fve proteins, particularly ANGPTL6, exhibited a signif- cantly high diagnostic performance whether alone or in sterol metabolic process, lipid storage and cholesterol combination with AFP in discriminating HCC from the biosynthetic processes (Fig. 3c). Downregulated DEPs healthy controls. were mostly enriched in extracellular matrix organiza- Among the 88 newly identifed CYP2D6*10 related tion, extracellular structure organization, and cell sub- − DEPs, we focused on ANGPTL6, a secretory protein, due strate adhesion categories (Fig. 3d). Patients with the to its high expression in peritumoral tissues and high- TT genotype had a lower risk of HCC which might be est AUC for predicting HCC. In line with the proteomic related to an altered tumor-extrinsic microenvironment. results, ANGPTL6 was higher in peritumor specimens Hu et al. J Transl Med (2021) 19:359 Page 7 of 13

Type Type HMHA1 3 CC a b LSP1 c LMNB1 TT lipid transport ITGA3 2 3 THSD4 CHST4 GZMK GPC4 1 lipid localization qvalue LRRC32 LOXL1 NPNT 0 OLFML1 protein processing 0.015 OLFML3 COL12A1 AEBP1 −1 ANGPTL6 0.020 PRELP protein maturation IGHA1 −2 MEGF6 ACAP1 0.025 DBN1 −3 regulation of plasminogen activation GPC6 S100A10 ANXA2;ANXA2P2 THY1 plasminogen activation EFEMP1 Count SGCD 2 TACSTD2 LTBP2 2 VIM positive regulation of protein processing positive EMILIN1 3 alue) FLNA KRT7 VCAN regulation of protein maturation peptide 4 LAMB1 SLC12A2 SUSD2 5 MRC2 hormone processing EFEMP2 ALPK3 FBLN5 LTBP4 LTBP1 signaling receptor ligand precursor processing HSPG2 NID1 LAMA5 AGRN 0.10 0.15 0.20 FBN1 VWA1 GeneRatio 1 PLVAP ATP8B1 −log10 (q−v HMCN1 d TRAF3IP3 CYP3A7 ANKRD44 AQP1 extracellular matrix organization HLA−DRA PFKP SAMHD1 ANGPTL2 PTPRCAP extracellular structure organization qvalue PFKFB2 CHST14 CCL21 MAMDC2 cell−substrate adhesion 2.5e−06 ERO1LB CYP4V2 PLIN4 5.0e−06 AGFG1 GLDC regulation of cell−substrate adhesion ACE 7.5e−06 SLC9B2 0 GRAMD4 LLGL2 positive regulation of cell−substrate adhesion 1.0e−05 HPN DHRS1 ACACB −6 −3 03 LSS glycosaminoglycan biosynthetic process CHN2 Count NDUFA10 ESYT1 log2 (fold change) ITGB3 aminoglycan biosynthetic process 4 SLC35F6 ATXN2L 8 ATP11C PEX5 EXOC3 glycosaminoglycan catabolic process 12 MFI2 CC1 CC2 CC3 CC4 CC5 CC6 CC7 CC8 CC9 CC1 0 CC1 1 CC1 2 CC1 3 CC1 4 CC1 5 CC1 6 CC1 7 CC1 8 TT1 TT2 TT3 TT4 TT5 TT6 TT7 TT8 TT9 TT1 0 aminoglycan catabolic process elastic 16

fiber assembly g 0.10 0.15 0.20 0.25 Type 4 Type GeneRatio e CHST14 Normal f ANGPTL2 CCL21 Tumor NPNT OLFML3 2 OLFML1 COL12A1 PRELP 25 25 Healthy AEBP1 *** ** ITGA3 VCAN 0 *** VWA1 HCC HSPG2 AGRN *** LAMA5 NID1 −2 *** EFEMP2 *** * SUSD2 FBN1 *** *** *** EMILIN1 *** FBLN5 LTBP1 20 20 −4 *** EFEMP1 LTBP4 *** CHST4 ANGPTL6 LOXL1 AQP1 GPC4 THSD4 MEGF6 MAMDC2 MRC2 SGCD THY1 15 bundance 15 HLA−DRA GPC6 ANXA2;ANXA2P2 S100A10 SAMHD1 abundance PFKP

LTBP2 na VIM LAMB1 FLNA IGHA1 ALPK3 TRAF3IP3 10 10 ESYT1 ACAP1 ANKRD44 LSP1 DBN1 LRRC32 ATP8B1

CYP3A7 protei HMCN1

PTPRCAP protein

PLVAP 2 TACSTD2 KRT7 2 5 SLC12A2 5 NDUFA10 ERO1LB HMHA1 GZMK EXOC3

SLC35F6 Lo g PLIN4 AGFG1 ATXN2L Lo g LMNB1 GLDC MFI2 ITGB3 0 0 ACE PFKFB2 CHN2 LLGL2 DHRS1 ATP11C ACACB 1 6 2 2 LSS E B GRAMD4 L T4 NT P4 CYP4V2 FI D4 SLC9B2 3A7 X S B PEX5 HPN P AC N1 N2 N3 N4 N5 N6 N7 N8 N9 N1 N1 N1 N1 N1 N1 N1 N1 N1 N1 N2 N2 N2 N2 N2 N2 N2 N2 N2 N2 N3 N3 N3 N3 N3 T1 T2 T3 T4 T5 T6 T7 T8 T9 T1 T1 T1 T1 T1 T1 T1 T1 T1 T1 T2 T2 T2 T2 T2 T2 T2 T2 T2 T2 T3 T3 T3 T3 T3 T3 T3 T3 T3 T3 T4 T4 T4 T4 T4 T4 T4 T4 T4 T4 T5 T5 T5 T5 T5 T5 T5 T5 T5 T5 T6 T6 EMP1 EMP2 NP M 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 LO CH LT GPTL AM EF CY EF LC9B R ACAC AN S G Fig. 3 Proteomic mechanisms of CYP2D6*10 polymorphisms on susceptibility to HCC. a Volcano plot showing 22 upregulated and 66 downregulated proteins. b Total of 88 DEPs between CC and TT groups of HCC patients shown in heat map. GO enrichment analysis shows up-regulated (c) and down-regulated (d) proteins in the TT group. e Heat-map showing 88 statistically signifcant DEPs between normal individuals (n 34) and HCC patients (n 61). f Expression profles of the top nine signifcantly up-regulated DEPs in peritumoral tissues of HCC. g Expression = = profles of the top fve signifcantly down-regulated DEPs in peritumoral tissues. Data were shown with mean SEM, *P < 0.05, **P < 0.01, ± ***P < 0.001 (M-U-test). DEPs, diferentially expressed proteins; HCC, hepatocellular carcinoma

than normal livers both in human liver samples and in Kaplan–Meier survival analysis indicated that higher a rat HCC model (P < 0.01, respectively, Fig. 5a). Mean- ANGPTL6 in patients with HCC was associated with while, parafn specimen analysis further confrmed worse overall survival (OS), suggesting an important role these results and demonstrated almost no expression of of ANGPTL6 in HCC progression (Fig. 6c). Moreover, ANGPTL6 in tumor cores (Fig. 5b). compared with the low expression group, patients with higher ANGPTL6 had lower PTL (P = 0.048), higher High ANGPTL6 is a risk factor for HCC and associated alanine aminotransferase (ALT, P = 0.030) and aspar- with poor prognosis tate aminotransferase (AST, P = 0.032 Fig. 6d). Interest- To further elucidate the diagnostic potential of ingly, older patients had a signifcantly higher level of ANGPTL6, univariate and multivariate logistic regres- ANGPTL6 than young patients (Additional fle 6: Fig- sion analysis was performed on the training set. As ure S1A). Serum levels of gamma-glutamyl transpepti- shown in the forest maps (Fig. 6a), the univariate logis- dase (GGT) and PT showed an increasing trend in the tic regression analysis revealed that a high level of high ANGPTL6 groups (Additional fle 6: Figure S1B ANGPTL6 was associated with high risk for HCC [Odds and S1C). What’s more, ANGPTL6 seemed to be higher ratio (OR): 3.417, 95% confdence interval (CI) 2.124– in CC genotype patients than TT group (Additional 5.498; P < 0.001]. Multivariate logistic analysis showed fle 6: Figure S1D). Univariate analysis revealed that high that ANGPTL6 was an independent risk factor for HCC expression of ANGPTL6 showed a signifcant correlation (OR: 2.327, 95%CI 1.191–4.547, P = 0.013, Fig. 6b). Col- with poor OS [hazard ratio (HR): 1.781, 95% confdence lectively, these data strongly suggest that up-regulation of interval (CI) 1.009–3.144; P = 0.047, Fig. 6e]. Multivari- ANGPTL6 in peritumoral tissues of HCC might contrib- ate Cox analysis showed that ANGPTL6 was an inde- ute to the pathogenesis of HCC. pendent prognostic factor for HCC (OR: 2.336, 95%CI Hu et al. J Transl Med (2021) 19:359 Page 8 of 13

a ANGPTL6 EFEMP2 EFEMP1 LTBP1 100 100 100 100 80 80 80 80

60 60 60 60 nsitivity% nsitivity%

nsitivity% 40 nsitivity% 40 40 AUC = 0.900 40 AUC = 0.830 AUC = 0.814 AUC = 0.813 Se Se Se P < 0.0001 Se P < 0.0001 P < 0.0001 P < 0.0001 20 20 20 20

0 0 0 0 020406080100 020406080100 020406080100 020406080100 100% - Specificity% 100% - Specificity% 100% - Specificity% 100% - Specificity% b c Serum AFP Combination AFP and ANGPTL6 KRT7 100 100 100 80 80 80 60 60 60 nsitivity% 40 nsitivity% nsitivity% AUC = 0.804 AUC = 0.902 40 40 AUC = 0.963 Se Se

Se P < 0.0001 P < 0.0001 P < 0.001 20 20 20

0 0 0 020406080100 020406080100 020406080100 100% - Specificity% 100% - Specificity% 100% - Specificity% Fig. 4 Diagnostic performance of the top fve DEPs with AUC > 0.80 and serum AFP. a Receiver operating characteristic (ROC) curve shows the predictive power of the top 5 DEPs with AUC > 0.80. b ROC curve of serum AFP for HCC patients versus healthy subjects. c ROC curve of combination AFP and ANGPTL6. HCC, hepatocellular carcinoma, DEPs, diferentially expressed proteins

serum ANGPTL6 (AUC 0.826, sensitivity 92.6%, specifc- 1.186–4.600, P = 0.014, Fig. 6f). Taken together, these data strongly suggest that high expression of ANGPTL6 ity 69.6%) had a comparable diagnostic value with AFP is an independent prognostic indicator that correlates (AUC 0.807, sensitivity 72.7%, specifcity 81.2%, Fig. 7c). with poor prognosis. Te combination of ANGPTL6 with AFP provided bet- ter diagnostic performance of preclinical HCC before clinical diagnosis (AUC 0.852, sensitivity 72.7%, speci- Validation ANGPTL6 as diagnostic marker in serum fcity 87.5%, Fig. 7c). Similar results were corroborated Serum ANGPTL6 and AFP from another validation by the validation set obtained from the GEO database cohort with 27 normal healthy volunteers and 42 HCC (GSE142987, AUC, 0.688 vs 0.708, Fig. 7d). Tese results patients was determined to explore whether the serum conclusively demonstrate that serum ANGPTL6 is an level of ANGPTL6 can be used as an indicator for the indicator for the diagnosis of HCC. diagnosis of HCC. Tere was a signifcant increase in serum ANGPTL6 and AFP levels in patients with HCC Discussion before surgery compared with healthy controls (Fig. 7a). In the present study, we demonstrated that CYP2D6 Similar results were corroborated by the validation set polymorphisms alter the susceptibility to HCC by obtained from the GEO database (GSE142987, Fig. 7b) decreasing the enzyme’s activity. Furthermore, based in 30 healthy control and 35 liver cancers for mRNA on proteomic analysis, we demonstrated a signifcant expression levels. Surprisingly, the mean concentration diference in the TME between CC- and TT-genotype of plasma ANGPTL6 before surgery was 3.51 (SD 0.83) patients that decreases the susceptibility to HCC in TT ng/mL, and values dropped to 2.02 (SD 0.88) ng/mL carriers, at least partly, by upregulating proteins related (P < 0.001) after surgery; however, AFP was still high in to lipid metabolism and downregulating proteins patients seven days after undergoing hepatic surgery or related to the extracellular matrix. Our current fndings receiving interventional therapy (Fig. 7a), suggesting that also reveal ANGPTL6 as a promising diagnostic and ANGPTL6 can monitor the curative efect dynamically. prognostic biomarker for HCC. We further assessed the accuracy of serum ANGPTL6 levels in the diagnosis of HCC. ROC curves showed that Hu et al. J Transl Med (2021) 19:359 Page 9 of 13

a Normal HCC model 1.5 ** ) evel

nl 1.0 ei ot pr

ve 0.5 ti NGPTL6/GAPDH la (A Re

0.0 Normal HCC model

Normal HCC model 1.5 ** ANGPTL6 52 ) ev el DH nl 1.0

GAPDH 37 AP /G Control HCC patients protei

PTL6 0.5

ANGPTL6 52 NG Relative (A

GAPDH 37 0.0 ControlHCC patients b Normal Adjacent Tumor 0.3 *** ** ### y it

dens 0.2 al ic pt

eo 0.1 Averag 0.0 Nomal AdjacentTumor Fig. 5 Verifcation of ANGPTL6 in animals and human samples. a General picture and western blot analysis of ANGPTL6 in Sprague Dawley rat model and human samples. b Representative images of H&E staining (top) and immunohistochemistry staining for ANGPTL6 (bottom), from human normal liver tissue, HCC tumor, and its peritumoral tissues. Scale bar, 100 μm. Data were expressed as mean with SEM. Statistical analysis was performed by Mann–Whitney test. **P < 0.01 vs Normal or Control, ###P < 0.001 vs adjacent. HCC, hepatocellular carcinoma

Relationships between genetic polymorphisms and In this study we found a signifcantly decreased fre- cancer risk have been widely studied [28, 29]. However, quency of 100TT-related genotypes with decreased most studies focused on polymorphisms and tumor susceptibility to HCC, consistent with previous stud- susceptibility on the basis of epidemiologic studies, ies that reported that CYP2D6 poor metabolizer (PM) with the underlying mechanisms undiscovered. Recent genotypes or inactivating alleles were signifcantly advances in high-throughput proteomic technology more frequent in healthy subjects and carriers than in provide expanded approaches into proteomic analy- patients with hepatitis/cirrhosis and HCC [18, 19]. We ses with breadth and throughput not previously possi- speculated that the decrease in HCC susceptibility in ble [30, 31]. Here, proteomic analysis was carried out TT genotypes may be partly explained by decreased to uncover the mechanisms of susceptibility to HCC CYP2D6 activity, due to its important role in activation underlying a CYP2D6 polymorphism. Our fndings of procarcinogens such as nitrosamine, nicotine, coti- provide a unique insight into possible early interven- nine [14, 32]. tion for HCC, especially in patients with cirrhosis. Subsequently, based on proteomics, we identifed a total of 1342 DEPs related to HCC, among which, 88 Hu et al. J Transl Med (2021) 19:359 Page 10 of 13

a Hazard Ratio Plot b Hazard Ratio Plot Variable Crude OR (95%CI) P Value Variable Adjust OR (95%CI) P Value

Age 1.054 (1.010−1.100) 0.015 Age 1.089 (0.998−1.303) 0.057

Gender Gender

Male 1 Male 1

Female 0.134 (0.049−0.366) < 0.001 Female 0.101 (0.010−1.054) 0.055

AFP 1.826 (1.204−2.768) 0.005 AFP 1.785 (0.975−3.269) 0.061

ANGPTL6 3.417 (2.124−5.498) < 0.001 ANGPTL6 2.327 (1.191−4.547) 0.013

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 c

d

Hazard Ratio Plot Hazard Ratio Plot efVariable HR (95%CI) P Value Variable HR (95%CI) P Value Age (years) Age (years) ≤ 55 1 ≤ 55 1 > 55 1.215 (0.568-2.598) 0.616 > 55 1.026 (0.533-1.976) 0.938 Gender Gender Male 1 Male 1 0.964 (0.547-1.701) 0.900 Female Female 1.846 (0.664-5.136) 0.240 AFP (ng/ml) AFP (ng/ml) ≤ 300 1 ≤ 300 1 > 300 1.666 (0.910-3.050) 0.098 > 300 1.928 (1.029-3.612) 0.040 ANGPTL6 (ng/ml) ANGPTL6 (ng/ml) ≤ 1 24 ≤ 24 1 > 24 1.781 (1.009-3.144) 0.047 > 24 2.336 (1.186-4.600) 0.014

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 00.5 11.5 22.5 33.5 44.5 55.5 Fig. 6 Clinical signifcance of ANGPTL6. Forrest plot of the univariate (a) and multivariate (b) logistic regression analysis. c High expression of ANGPTL6 conferring to a poor prognosis. d ANGPTL6 correlated negatively with PLT and positively with ALT and AST. Univariate (e) and (f) multivariate Cox regression analysis for OS. AFP, alpha-fetoprotein. PLT, platelet. ALT, alanine aminotransferase. AST, aspartate aminotransferase. OS, overall survival Hu et al. J Transl Med (2021) 19:359 Page 11 of 13

GSE142987 P < 0.001 P < 0.001 GSE142987 ab4 P < 0.05 P = 0.008 ) 4000 10 5 P = 0.002 ml

g/ 3000 )

(n 3 8 4 ml 2000 FP g/ ANGPTL6

TL6 1000 (n mA

P 6 3 2 m FP NG

15 se ru

se ru 4 2 mA aA of of 1 ru 10 KM Se asm 2 1

Pl 5 FP FPKM 0 0 0 0 n n HealthyLiver cancer Healthy Liver cancer al o o l n n i a o io rm rat rm at e o r No perati p N e o o perati p - -o to e st e s r o r o P P P P

c Human serum ANGPTL6 Human serum AFP ANGPTL6 + AFP 100 100 100

80 80 80 ity% ity% ity% 60 60 60 iv iv iv it it it 40 AUC = 0.826 40 AUC = 0.807 40 AUC = 0.852 Sens Sens Sens P < 0.001 P = 0.008 P = 0.002 20 Sensitivity: 92.6% 20 Sensitivity: 72.7% 20 Sensitivity: 72.7% Specificity: 69.6% Specificity: 81.2% Specificity: 87.5% 0 0 0 020406080100 020406080 100 020406080100 100% - Specificity% 100% - Specificity% 100% - Specificity% d GSE142987 GSE142987 100 GSE142987 100 (Serum ANGPTL6) 100 (ANGPTL6 + AFP) (Serum AFP) 80 80 80 ity%

ity% 60 ity% 60 60 iv iv it iv it it AUC = 0.708 40 40 40 AUC = 0.703 AUC = 0.688 P = 0.004 Sens Sens

Sens P = 0.005 P = 0.010 Sensitivity: 54.3% Sensitivity: 45.7% 20 Sensitivity: 51.4% 20 Specificity: 83.3% 20 Specificity: 90.0% Specificity: 96.7% 0 0 0 020406080100 020406080 100 020406080100 100% - Specificity% 100% - Specificity% 100% - Specificity% Fig. 7 Identifcation of serum ANGPTL6 as a sensitive biomarker for HCC. ANGPTL6 and AFP protein level (a) and mRNA level (b) in plasma or serum. ROC curve of serum ANGPTL6, AFP alone or combination of ANGPTL6 and AFP for diagnosis of HCC in our validation set (c) and GEO database (d)

DEPs constitute a diferent TME between CC and TT markers for screening HCC at early stage. Alpha feto- and susceptibility to HCC. Tese fndings could be protein (AFP) is the only currently available blood test explained by upregulated and downregulated DEPs that commonly regarded as a characteristic biomarker for were mainly involved in lipid metabolism and the extra- screening, detection and surveillance of HCC [35]. cellular matrix, respectively, two biologic processes Unfortunately, the clinical utility is limited because of closely related with HCC development [33, 34]. Moreo- low sensitivity. Terefore, it is urgent to fnd novel efec- ver, most of downregulated DEPs, which are involved in tive predictive biomarkers with early diagnostic value extracellular matrix organization, extracellular structure for HCC. ANGPTL-6, also called angiopoietin-related organization, and cell − substrate adhesion were upregu- growth factor protein 5 (ARP5), initially known as a cir- lated in HCC. Tese fndings together further validated culating proangiogenic factor mainly secreted from the that the TT variant was a protective factor in the occur- liver [36–38], represents an attractive therapeutic target rence of HCC and confrmed that an altered TME played for anti-tumor therapy [39]. In the current study, ANG- a key role in polymorphisms related to susceptibility to PLT6 is dramatically increased both in peritumor tissues HCC. and serum and enriched in the extracellular matrix of One of the main reasons for the high mortality rate of HCC, suggesting an important role in the initiation and HCC is the lack of sensitive and reliable early diagnostic development of HCC. Upregulated ANGPTL6 expression Hu et al. J Transl Med (2021) 19:359 Page 12 of 13

resulted in the proliferation of undiferentiated glioblas- Additional fle 4: Table S4. Compared with healthy subjects, the 17 toma cells [40], and knockdown of ANGPLT6 in vivo upregulated and 71 downregulated DEPs in HCC groups based on prot- signifcantly inhibited AFP-producing gastric cancer by eomics analysis. regulating endothelial cell migration and tube formation Additional fle 5: Table S5. DEPs with the area under the ROC curve (AUC) > 7.0. [39]. Te diagnostic accuracy of ANGPTL6 was com- Additional fle 6: Figure S1. ANGPTL6 has clinical signifcance. a High parable to AFP; the AUC values for the two biomarkers ANGPTL6 expression was more common in the elderly group. Serum were 0.900 and 0.902, respectively, while the sensitivity levels of GGT (b) and PT (c) show an increasing trend in the high ANGPTL6 of ANGPTL6 was markedly superior to AFP (91.8% vs groups. d ANGPTL6 in TT and CC group. HCC, hepatocellular carcinoma. GGT, gamma-glutamyl transpeptidase. PT, prothrombin time. 73.8%). Regrettably, the number of serum samples in our study was too small to allow stratifed analyses. In par- ticular, some AFP data were missing, possibly leading to Acknowledgements We thank the department of liver and gallbladder surgery, Pathology and bias in the results, such as combination of ANGTPL6 Laboratory of The First and Second Afliated Hospital of Zhengzhou University, and AFP not showing signifcant advantages over a single Henan Provincial People’s Hospital and Henan Provincial Tumor Hospital biomarker. A larger sample size is needed to confrm the (Henan, China). clinical application value of ANGPTL6. Authors’ contributions Hematological results showed high ANGPTL6 lev- GMH: Conceptualization, methodology, data curation, writing—original draft, els were associated with lower PTL and higher ALT and writing—review & editing; YF: Investigation, methodology, resources; YYG: Data curation, resources; FG: Resources, investigation, data curation; NG: Methodol- AST levels, and conferred a shorter OS. Univariate and ogy, data curation, resources; GZW: Resources, investigation, resources; YHG: multivariate Cox proportional hazards regression analy- Investigation, methodology, data curation; CZZ: Resources, supervision, valida- sis revealed that high expression of ANGPTL6 was an tion; QW and JZ: Validation, writing—original draft; HLQ: Project administration, validation, writing—original draft, writing—review & editing. All authors read independent prognostic indicator that correlated with and approved the fnal manuscript. poor prognosis. As a consequence, our current fndings highlight ANGPTL6 as a promising diagnostic and prog- Funding This work was supported by National Natural Science Foundation of China nostic biomarker for HCC. (NSFC) (Nos. 81473279, 81673507, 81872931 and 82073930); Zhengzhou Major Scientifc and Technological Innovation Projects (No. 2020CXZX0076).

Availability of data materials Conclusions All the data generated and analyzed during this study are included in the In summary, our fndings support the mechanism by manuscript and the additional materials. which CYP2D6*10 (100C>T) polymorphisms confer sus- ceptibility to HCC by altering CYP2D6 activity and TME, Declarations which contributes to hepatocarcinogenesis. We further Ethics approval and consent to participate highlight ANGPTL6 as a promising diagnostic and prog- This study was carried out abiding by the principles established by the Declara- nostic biomarker for HCC. Our fndings open new mech- tion of Helsinki and approved by the Ethical Committee of the Zhengzhou University (Henan, China). All participants were provided with an informed anistic insights for polymorphisms related HCC risk and consent that they signed prior to the experiments. provide avenues for clinical screening for HCC. Consent for publication Not applicable. Abbreviations AFP: Alpha-fetoprotein; A/G: Albumin/globulin ratio; ALB: Albumin; ALT: Alanine Competing interests aminotransferase; ANGPTL6: Angiopoietin-like protein-6; AST: Aspartate ami- The authors declare no potential conficts of interest. notransferase; AUC:​ Area under the curve; CI: Confdence interval; Clint: Intrinsic clearance; CYP2D6: Cytochrome P4502D6; DBIL: Direct bilirubin; DEPs: Diferen- Author details tially expressed proteins; HCC: Hepatocellular carcinoma; HR: Hazard ratio; OR: 1 Institute of Clinical Pharmacology, Zhengzhou University, Zhengzhou 450052, Odds ratio; OS: Overall survival; PTL: Platelets; SD: Sprague Dawley; SNPs: Single Henan, China. 2 Afliated People’s Hospital of Zhengzhou University, Zheng- nucleotide polymorphisms; TBIL: Total bilirubin; TME: Tumor microenvironment; zhou, China. ROC: Receiver operator characteristics curve. Received: 7 July 2021 Accepted: 10 August 2021 Supplementary Information The online version contains supplementary material available at https://doi.​ ​ org/10.​ 1186/​ s12967-​ 021-​ 03038-3​ . References 1. Khera AV, Chafn M, Aragam KG, Haas ME, Roselli C, Choi SH, et al. Additional fle 1: Table S1. Hardy–Weinberg Equilibrium (HWE). Genome-wide polygenic scores for common diseases identify Additional fle 2: Table S2. The genotype frequency of CYP2D6*10 (100 individuals with risk equivalent to monogenic mutations. Nat Genet. C>T) in healthy volunteers (n 105) and HCC patients (n 96). 2018;50:1219–24. https://doi.​ org/​ 10.​ 1038/​ s41588-​ 018-​ 0183-z​ . = = 2. Whibley C, Pharoah PD, Hollstein M. p53 polymorphisms: cancer implica- Additional fle 3: Table S3. Compared with CC groups, the 22 upregulated tions. Nat Rev Cancer. 2009;9:95–107. https://doi.​ org/​ 10.​ 1038/​ nrc25​ 84​ . and 66 downregulated DEPs in TT groups based on proteomics analysis. Hu et al. J Transl Med (2021) 19:359 Page 13 of 13

3. Schmiedel BJ, Singh D, Madrigal A, Valdovino-Gonzalez AG, White BM, diethylnitrosamine. J Pharmacol Exp Ther. 2018;365:398–407. https://​ Zapardiel-Gonzalo J, et al. Impact of genetic polymorphisms on human doi.org/​ 10.​ 1124/​ jpet.​ 117.​ 245555​ . immune cell gene expression. Cell. 2018;175:1701-15.e16. https://doi.​ ​ 24. Zhou J, Wen Q, Li SF, Zhang YF, Gao N, Tian X, et al. Signifcant change of org/10.​ 1016/j.​ cell.​ 2018.​ 10.​ 022​ . cytochrome P450s activities in patients with hepatocellular carcinoma. 4. Aleksandrova E, Vlaykova T, Ananiev J, Gulubova M. Association of Oncotarget. 2016;7:50612–23. https://doi.​ org/​ 10.​ 18632/​ oncot​ arget.​ ​ IL-12Bpro polymorphism with tumor-infltrating dendritic cells in colo- 9437. rectal cancer. J BUON. 2017;22:888–93. 25. Gao J, Tian X, Zhou J, Cui MZ, Zhang HF, Gao N, et al. From genotype to 5. Dopazo J, Amadoz A, Bleda M, Garcia-Alonso L, Alemán A, García-García phenotype: cytochrome P450 2D6-mediated drug clearance in humans. F, et al. 267 Spanish exomes reveal population-specifc diferences Mol Pharm. 2017;14:649–57. https://doi.​ org/​ 10.​ 1021/​ acs.​ molph​ armac​ ​ in disease-related genetic variation. Mol Biol Evol. 2016;33:1205–18. eut.6b009​ 20​ . https://doi.​ org/​ 10.​ 1093/​ molbev/​ msw005​ . 26. Mochizuki J, Murakami S, Sanjo A, Takagi I, Akizuki S, Ohnishi A. Genetic 6. Laursen L. A preventable cancer. Nature. 2014;516:S2-3. https://doi.​ org/​ ​ polymorphisms of cytochrome P450 in patients with hepatitis C 10.1038/​ 516S2a​ . virus-associated hepatocellular carcinoma. J Gastroenterol Hepatol. 7. Nahon P, Zucman-Rossi J. Single nucleotide polymorphisms and risk 2005;20:1191–7. https://doi.​ org/​ 10.​ 1111/j.​ 1440-​ 1746.​ 2005.​ 03808.x​ . of hepatocellular carcinoma in cirrhosis. J Hepatol. 2012;57:663–74. 27. Jiang Y, Sun A, Zhao Y, Ying W, Sun H, Yang X, et al. Proteomics identi- https://doi.​ org/​ 10.​ 1016/j.​ jhep.​ 2012.​ 02.​ 035​ . fes new therapeutic targets of early-stage hepatocellular carcinoma. 8. Thomas DL, Thio CL, Martin MP, Qi Y, Ge D et al. Genetic variation Nature. 2019;567:257–61. https://doi.​ org/​ 10.​ 1038/​ s41586-​ 019-​ 0987-8​ . in IL28B and spontaneous clearance of hepatitis C virus. Nature. 28. Turajlic S, Sottoriva A, Graham T, Swanton C. Resolving genetic hetero- 2009;461:798–801. doi: https://doi.​ org/​ 10.​ 1038/​ natur​ e08463​ geneity in cancer. Nat Rev Genet. 2019;20:404–16. https://doi.​ org/​ 10.​ ​ 9. Bressac B, Kew M, Wands J, Ozturk M. Selective G to T mutations of 1038/s41576-​ 019-​ 0114-6​ . p53 gene in hepatocellular carcinoma from southern Africa. Nature. 29. Bao Y, Prescott J, Yuan C, Zhang M, Kraft P, Babic A, et al. Leucocyte 1991;350:429–31. https://doi.​ org/​ 10.​ 1038/​ 35042​ 9a0​ . telomere length, genetic variants at the TERT gene region and risk 10. Niu ZS, Niu XJ, Wang WH. Genetic alterations in hepatocellular carci- of pancreatic cancer. Gut. 2017;66:1116–22. https://doi.​ org/​ 10.​ 1136/​ ​ noma: an update. World J Gastroenterol. 2016;22:9069–95. https://doi.​ ​ gutjnl-2016-​ 312510​ . org/10.​ 3748/​ wjg.​ v22.​ i41.​ 9069​ . 30. Zhang Z, Wu S, Stenoien DL, Paša-Tolić L. High-throughput proteomics. 11. Preissner SC, Hofmann MF, Preissner R, Dunkel M, Gewiess A, Preissner Annu Rev Anal Chem. 2014;7:427–54. https://doi.​ org/​ 10.​ 1146/​ annur​ ​ S. Polymorphic cytochrome P450 enzymes (CYPs) and their role in per- ev-anchem-​ 071213-​ 020216​ . sonalized therapy. PLoS ONE. 2013;8: e82562. https://doi.​ org/​ 10.​ 1371/​ ​ 31. Karczewski KJ, Snyder MP. Integrative omics for health and disease. Nat journal.​ pone.​ 00825​ 62​ . Rev Genet. 2018;19:299–310. https://doi.​ org/​ 10.​ 1038/​ nrg.​ 2018.4​ . 12. Neafsey P, Ginsberg G, Hattis D, Sonawane B. Genetic polymorphism 32. Nebert DW, McKinnon RA, Puga A. Human drug-metabolizing enzyme in cytochrome P450 2D6 (CYP2D6): Population distribution of CYP2D6 polymorphisms: efects on risk of toxicity and cancer. DNA Cell Biol. activity. J Toxicol Environ Health B Crit Rev. 2009;12:334–61. https://doi.​ ​ 1996;15:273–80. https://doi.​ org/​ 10.​ 1089/​ dna.​ 1996.​ 15.​ 273​ . org/10.​ 1080/​ 10937​ 40090​ 31583​ 42​ . 33. Ma C, Kesarwala AH, Eggert T, Medina-Echeverz J, Kleiner DE, Jin P, 13. Zanger UM, Schwab M. Cytochrome P450 enzymes in drug metabolism: et al. NAFLD causes selective CD4( ) T lymphocyte loss and promotes regulation of gene expression, enzyme activities, and impact of genetic hepatocarcinogenesis. Nature. 2016;531:253–7.+ https://doi.​ org/​ 10.​ 1038/​ ​ variation. Pharmacol Ther. 2013;138:103–41. https://doi.​ org/​ 10.​ 1016/j.​ ​ nature16969​ . pharmthera.​ 2012.​ 12.​ 007​ . 34. Afo S, Yu LX, Schwabe RF. The role of cancer-associated fbroblasts and 14. Zhou SF. Polymorphism of human cytochrome P450 2D6 and its clinical fbrosis in liver cancer. Annu Rev Pathol. 2017;12:153–86. https://doi.​ ​ signifcance: Part I. Clin Pharmacokinet. 2009;48:689–723. https://doi.​ ​ org/10.​ 1146/​ annur​ ev-​ pathol-​ 052016-​ 100322​ . org/10.​ 2165/​ 11318​ 030-​ 00000​ 0000-​ 00000​ . 35. Chen T, Dai X, Dai J, Ding C, Zhang Z, Lin Z, et al. AFP promotes HCC 15. Lee YH, Bae SC. Association between Functional CYP2D6 polymor- progression by suppressing the HuR-mediated Fas/FADD apop- phisms and susceptibility to autoimmune diseases: a meta-analysis. totic pathway. Cell Death Dis. 2020;11:822. https://doi.​ org/​ 10.​ 1038/​ ​ Immunol Invest. 2017;46:109–22. https://doi.​ org/​ 10.​ 1080/​ 08820​ 139.​ ​ s41419-020-​ 03030-7​ . 2016.12268​ 98​ . 36. Kim I, Kim HG, Kim H, Kim HH, Park SK, Uhm CS, et al. Hepatic expres- 16. Lu Y, Qin X, Li S, Zhang X, He Y, Peng Q, et al. Quantitative assessment of sion, synthesis and secretion of a novel fbrinogen/angiopoietin-related CYP2D6 polymorphisms and risk of Alzheimer’s disease: a meta-analysis. protein that prevents endothelial-cell apoptosis. Biochem J. 2000;346(Pt J Neurol Sci. 2014;343:15–22. https://doi.​ org/​ 10.​ 1016/j.​ jns.​ 2014.​ 05.​ 033​ . 3):603–10. 17. Lu Y, Peng Q, Zeng Z, Wang J, Deng Y, Xie L, et al. CYP2D6 phenotypes 37. Oike Y, Ito Y, Maekawa H, Morisada T, Kubota Y, Akao M, et al. Angi- and Parkinson’s disease risk: a meta-analysis. J Neurol Sci. 2014;336:161– opoietin-related growth factor (AGF) promotes angiogenesis. Blood. 8. https://doi.​ org/​ 10.​ 1016/j.​ jns.​ 2013.​ 10.​ 030​ . 2004;103:3760–5. https://doi.​ org/​ 10.​ 1182/​ blood-​ 2003-​ 04-​ 1272​ . 18. Agúndez JA, Ledesma MC, Benítez J, Ladero JM, Rodríguez-Lescure 38. Urano T, Ito Y, Akao M, Sawa T, Miyata K, Tabata M, et al. Angiopoietin- A, Díaz-Rubio E, et al. CYP2D6 genes and risk of liver cancer. Lancet. related growth factor enhances blood fow via activation of the ERK1/2- 1995;345:830–1. https://doi.​ org/​ 10.​ 1016/​ s0140-​ 6736(95)​ 92965-7​ . eNOS-NO pathway in a mouse hind-limb ischemia model. Arterioscler 19. Silvestri L, Sonzogni L, De Silvestri A, Gritti C, Foti L, Zavaglia C, et al. CYP Thromb Vasc Biol. 2008;28:827–34. https://doi.​ org/​ 10.​ 1161/​ ATVBA​ HA.​ ​ enzyme polymorphisms and susceptibility to HCV-related chronic liver 107.149674​ . disease and liver cancer. Int J Cancer. 2003;104:310–7. https://doi.​ org/​ ​ 39. Chen E, Tang C, Peng K, Cheng X, Wei Y, Liu T. ANGPTL6-mediated angio- 10.1002/​ ijc.​ 10937​ . genesis promotes alpha fetoprotein-producing gastric cancer progres- 20. Zhang HF, Wang HH, Gao N, Wei JY, Tian X, Zhao Y, et al. Physiologi- sion. Pathol Res Pract. 2019;215: 152454. https://doi.​ org/​ 10.​ 1016/j.​ prp.​ ​ cal content and intrinsic activities of 10 cytochrome P450 isoforms in 2019.152454​ . human normal liver microsomes. J Pharmacol Exp Ther. 2016;358:83–93. 40. Cui JG, Zhao Y, Sethi P, Li YY, Mahta A, Culicchia F, et al. Micro-RNA-128 https://doi.​ org/​ 10.​ 1124/​ jpet.​ 116.​ 233635​ . (miRNA-128) down-regulation in glioblastoma targets ARP5 (ANGPTL6), 21. Thomson S. A system for quantitative studies on interactions between Bmi-1 and E2F–3a, key regulators of brain cell proliferation. J Neuroon- Friend leukemia virus and hemopoietic cells. Proc Soc Exp Biol Med. col. 2010;98:297–304. https://doi.​ org/​ 10.​ 1007/​ s11060-​ 009-​ 0077-0​ . 1969;130:227–32. https://doi.​ org/​ 10.​ 3181/​ 00379​ 727-​ 130-​ 33527​ . 22. Gao J, Wang Z, Wang GJ, Gao N, Li J, Zhang YF, et al. From hepatofbrosis to hepatocarcinogenesis: Higher cytochrome P450 2E1 activity is a Publisher’s Note potential risk factor. Mol Carcinog. 2018;57:1371–82. https://doi.​ org/​ 10.​ ​ Springer Nature remains neutral with regard to jurisdictional claims in pub- 1002/mc.​ 22851​ . lished maps and institutional afliations. 23. Gao J, Wang Z, Wang GJ, Zhang HX, Gao N, Wang J, et al. Higher CYP2E1 activity correlates with hepatocarcinogenesis induced by