Rs9679162 genotype predicts prognosis of real-world advanced hepatocellular carcinoma treated by sorafenib

2.1 Patients and samples

This retrospective cohort study was conducted under the approval of the Institutional Review Board of Chang Gung Memorial Hospital, Taiwan (approval numbers: 201506308A3 and 201900261B0). From January 2010 to December 2020, a total of 350 patients with advanced HCC at BCLC stage C who received targeted therapy were retrospectively included. Of these, 298 were treated with sorafenib, and the remaining 52 were treated with lenvatinib at the Linkou and Keelung branches of Chang Gung Memorial Hospital. Their blood samples were retrieved for genomic DNA isolation and their clinical parameters were reviewed, including gender, age, HBV surface antigen (HBsAg), anti-HCV antibody, alcoholism, liver cirrhosis status, presence of ascites, Child-Pugh classification, Eastern Cooperative Oncology Group (ECOG) performance status, initial metastasis, main portal vein thrombosis, previous anticancer treatments (TACE, radiotherapy, or chemotherapy), additional anticancer therapies following TKI failure, total dosage of TKIs, largest tumor size (in diameter), date of response, date of tumor progression, and date of last follow-up or HCC related death. Biochemistry and hemogram analysis included alpha-fetoprotein, albumin, bilirubin, aspartate transaminase (AST), alanine transaminase (ALT), creatinine, prothrombin time, and platelet count. In addition, the APRI (aspartate aminotransferase-to-platelet ratio index) [23], ALBI (Albumin-Bilirubin) grade [24], FIB-4 (Fibrosis-4) score [25], and PALBI (Platelet-Albumin-Bilirubin) grade were calculated according to the published formulas.

HCC was diagnosed by dynamic computed tomography or dynamic magnetic resonance imaging, plus angiography, when the size of HCC in a cirrhotic liver exceeded 2 cm and typical HCC characteristics were present. In the absence of such typical features, liver biopsy or aspiration cytology was required for diagnosis. Response to TKIs treatment was assessed using alpha-fetoprotein (AFP) and modified response evaluation criteria in solid tumors (mRECIST), as previously described [26, 27, 28]. A 20% reduction in AFP was also considered to achieve a therapeutic response [27].

2.2 GALNT14 genotyping

Genotyping of the SNP GALNT14-rs9679162 was performed as previously described [16]. Briefly, genomic DNA isolated from patient blood samples was amplified using the QIAamp DNA mini kit (QIAGEN, Cat: 51306) to obtain 172-bp DNA containing the GALNT14-rs9679162 by polymerase chain reaction (PCR). Two primers were used: 5’-TCACGAGGCCAACATTCTAG-3’ and 5’-TTAGATTC-TGCATGGCTCAC-3’. PCR was performed using Hi-Fi Tag polymerase (BIOTOOLS, Cat: KTT-BB05) according to the protocol provided by the manufacturer. SNP genotypes were determined by direct sequencing.

2.3 GALNT14 promoter luciferase assay

The primers, forward: 5’-ATCGGGTACCTCCAACTTTGGACCTCTTTT-3’ and reverse: 5’-ATCGAAGC TTCCAGACCATCGTGGGTTG-3’, were used to amplify a 507-bp DNA fragment containing the GALNT14 promoter. The underlined sequence is recognized by KpnI in the forward primer and HindIII in the reverse primer, respectively. The PCR product containing GALNT14 promoter region was digested with the aforementioned restriction enzymes and inserted into pGL3. After verification of the sequence, the plasmids were transfected into hepatoma cells for the luciferase assay, with or without co-expression of wild type HBx. The procedure was performed as previously described [29].

2.4 Immunohistochemical staining (IHC) and western blot analysis

IHC and western blot analysis were performed as described in previous studies [9, 30]. The GALNT14 rabbit polyclonal antibody (Proteintech, Cat: 16939-1-AP) was used for IHC in 1:200 and for western blot analysis in 1:2000 dilutions.

2.5 Cell culture and cell viability assay

In this study, three HCC-derived cell lines were used, including J7 [Research Resource Identification (RRID): CVCL_4Z69], Huh7 (RRID: CVCL_0336) and HepG2 (RRID: CVCL_0027). All three cell lines were gifts from Dr. Kwang-Huei Lin, Chang Gung University, Taoyuan, Taiwan. J7 and Huh7 were cultured in Dulbecco’s Modified Eagle Medium (DMEM), while HepG2 was maintained in minimal essential medium (MEM) under standard culture conditions with 5% CO 2 in a humidified incubator at 37 ∘ C. All the cell lines were routinely examined for mycoplasma contamination. The Alarmar Blue-based cell viability assay (Invitrogen, DAL1100) was performed as previously described [31].

2.6 Statistical analysis

Parametric data are presented as the mean ± standard deviation and compared using a two-sample t -test. Dichotomized data are presented as numbers and percentages (%) and compared using the chi-square test or Fisher’s exact tests. Univariate and following multivariate Cox proportional hazard models were performed to estimate overall survival for clinical factors and genotypic variables. In this study, the significant factors identified from the univariate analysis could be included in the multivariate Cox proportional hazards. The Kaplan-Meier method was used to estimate the probability of survival between the different groups, and the log-rank test was used to compare the survival rate. All tests were two-tailed, and 0.05 was considered statistically significant. All statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS) Version 20 (SPSS, Chicago, IL, USA).

3.1 Baseline demographic characteristics of HCC patients treated with sorafenib

A total of 298 patients treated with sorafenib were included in this study. Their baseline clinical and genotypic characteristics are listed in Table 1. Among them, 52 (17.4%) were GALNT14-rs9679162 “GG” genotype, and 246 (82.6%) were GALNT14 “non-GG” (“GT” + “TT”) genotype. The mean age was 63.4 ± 12.1 years; 234 (78.5%) were male; 256 (85.9%) had histological or clinical evidence of liver cirrhosis; 267 (89.6%) had previously received anticancer treatments, including TACE, chemotherapy and radiotherapy. Higher rates of initial metastases (126, 42.3%) and portal vein thrombosis (149, 50%) were noted due to the inclusion of patients with advanced HCC. The etiological analysis showed that 108 (36.2%) were anti-HCV positive, 181 (60.7%) were HBsAg positive and 94 (31.5%) were associated with alcohol abuse. It was not uncommon for patients to have more than one etiology.

OPEN IN VIEWER

Table 1 also shows the results of a detailed comparison of all clinical characteristics and scores between patients with advanced HCC with GALNT14-rs9679162 “GG” and “non-GG” (“GT” + “TT”) genotypes. Patients with the “GG” genotype had lower rates of alcohol abuse (19.2% versus 34.1%; P = 0.035) and lower ECOG performance status ( P = 0.034). No significant difference was observed in the other variables and clinical scores. Furthermore, no statistical difference was found between the patients of the two genotypes after correction of the multiple comparisons by Bonferroni method.

3.3 The independent prognostic predictors before treatment in all patients receiving sorafenib

Kaplan-Meier analysis was used to investigate the prognostic significance of the GALNT14-rs9679162 genotype and found that the “GG” genotype was not associated with OS ( P = 0.218), progression-free survival (PFS) ( P = 0.893), and time-to-response (TTR) ( P = 0.272) (Figs 1A, S1A, and S2A) when all patients were included. Univariate analysis and multivariate Cox proportional hazard analysis were performed to understand the relationship between clinical factors and OS after sorafenib treatment (Table 2). Univariate analysis revealed that HBsAg-positive, ascites, initial metastasis, lower albumin, shorter prothrombin time, higher bilirubin, higher AST, higher ALBI score, higher ALBI grade, higher Child-Pugh score, higher PALBI score, higher PALBI grade, higher APRI grade, and a higher FIB-4 score were significantly associated with shorter OS. After adjusting for confounding factors, multivariate analysis revealed positivity for HBsAg ( P = 0.002), ascites ( 0.001), initial metastasis ( P = 0.010), lower albumin level ( P = 0.010), shorter prothrombin time ( P = 0.002), higher AST ( P = 0.010), and higher PALBI grade ( P = 0.014) remained as independent predictors of shorter OS. Notably, alcoholism ( P = 0.017), higher ECOG grade ( P = 0.025), initial metastasis ( P = 0.001), higher AFP ( P = 0.008), and higher prothrombin time ( P = 0.017) were markedly correlated to shorter PFS, while only alcoholism ( P = 0.006), initial metastasis ( 0.001), higher AFP ( P = 0.036) and higher prothrombin time ( P = 0.027) were an independent predictor for PFS in multivariate analysis (Table S1). Absence of cirrhosis ( P = 0.016), and lower neutrophil/platelet ratio ( P = 0.036) were associated with shorter TTR. However, no factor was an independent predictor of TTR in the multivariate analysis (Table S2). These findings suggest that these baseline characteristics identified from multivariate analysis were independent prognostic predictors for all patients treated with sorafenib.

OPEN IN VIEWER

Although in the present study, when all patients were included, the GALNT14-rs9679162 genotype was not associated with prognosis in patients receiving sorafenib, our previous study showed that in treatment-naïve BCLC stage C HCC patients receiving sorafenib as first-line therapy, the GALNT14-rs9679162 “GG” genotype was significantly associated with shorter OS in specific patient subgroups [22]. To understand whether this association stands valid in real-world BCLC stage C patients who received another anticancer treatment before sorafenib, we divided the patients into different subgroups and asked whether GALNT14-rs9679162 “GG” in these subgroups was associated with OS using Cox proportional hazard analysis (Fig. 2). We found in certain subgroups of patients (anti-HCV-positive, HBsAg-negative, tumor size ⩽ 5.4 cm, creatinine ⩽ 0.9 mg/dL, ALT $>$ 51 U/L, AST ⩽ ALT, platelet ⩽ 161 × 1000/ μ L, and APRI grade $>$ 1), the “GG” genotype was significantly associated with shorter OS (Fig. 2). Subsequently, Kaplan-Meier plots with log-rank analysis were used to further confirm these data. The “GG” genotype effectively predicted unfavorable OS after sorafenib therapy in the following subgroups: anti-HCV-positive (Fig. 1B, P = 0.004), HBsAg-negative (Fig. 1C, P = 0.001), tumor size ⩽ 5.4 cm (Fig. 1D, P = 0.026), Creatinine ⩽ 0.9 mg/dL (Fig. 1E, P = 0.049), ALT $>$ 51 U/L (Fig. 1F, P = 0.010), AST ⩽ ALT (Fig. 1G, P = 0.004), platelet ⩽ 161 × 1000/ μ L (Fig. 1H, P = 0.035), or APRI grade $>$ 1 (Fig. 1I, P = 0.017). However, the “GG” genotype could not predict PFS (Fig. S1B–I) and TTR (Fig. S2B–I) after sorafenib treatment in these subgroups.

3.5 Expansion of the patient subgroups eligible for survival prediction by GALNT14-rs9679162 genotype

In Fig. 2, the GALNT14-rs9679162 “GG” genotype could only predict unfavorable response to sorafenib in certain specific subgroups, including the red lines subgroups: HBsAg-negative and anti-HCV positive (Fig. 2, red lines), and the green line subgroups: smaller tumors, hepatitis, and cirrhosis (Fig. 2, green lines). Accordingly, we examined a group of patients without the following clinical conditions: HBsAg positivity, APRI grade ⩽ 1 and tumor size $>$ 5.4 cm. In other words, patients with HBsAg-positive, non-cirrhotic and larger tumors were excluded. After exclusion of these patients, 244 patients (81.9%) remained. In this large subgroup, the “GG” genotype was significantly associated with shorter OS ( P = 0.016), but not with PFS ( P = 0.461) and TTR ( P = 0.266) (Fig. 3A). Similarly, we were able to select patients without a combination of: HBsAg positivity, platelet $>$ 161 × 1000/ μ L and tumor size $>$ 5.4 cm ( n = 257, 86.3%). In this subgroup, the “GG” genotype was also significantly associated with shorter OS ( P = 0.037), but not with PFS ( P = 0.624) and TTR ( P = 0.474) (Fig. 3B). Further exclusion of patients meeting all four factors: HBsAg-positive, APRI grade ⩽ 1, platelet $>$ 161 × 1000/ μ L, and tumor size $>$ 5.4 cm resulted in a larger subgroup of 263 patients (88.3%). In this subgroup, the “GG” genotype was significantly associated with shorter OS ( P = 0.049), but not with PFS ( P = 0.583) and TTR ( P = 0.492) (Fig. 3C).

OPEN IN VIEWER

Collectively, in HBsAg-negative or anti-HCV-positive patients, the “GG” genotype effectively predicted shorter OS, but not PFS and TTR, in sorafenib-treated patients. However, if we wanted to apply this marker to a larger proportion of patients, patients with both HBsAg-positive, larger tumors, and non-cirrhosis (or higher platelet counts) should be excluded.

In the aforementioned findings, it was found that the GALNT14-rs9679162 genotype was unable to predict the prognosis of HBV-related HCC. A previous study also demonstrated that in patients with BCLC stage B HCC, GALNT14 levels were higher in cancerous liver tissue than in non-cancerous parts [9]. Therefore, here we examined 85 HCC tissues obtained from patients with BCLC stage A. GALNT14 expression was stained by IHC and scored according to a computer algorithm (Fig. 4A). Higher levels of GALNT14 were found in cancerous tissue, which was consistent with what has been observed in mid-stage HCC [9]. By performing a cell-based sorafenib susceptibility assay, we found that overexpression of GALNT14 significantly suppressed drug sensitivity to sorafenib in two HCC cell lines, J7 and Huh7 (Fig. 4B and C). Therefore, overexpression of GALNT14 may play an important role in controlling susceptibility of HCC cells to sorafenib.

OPEN IN VIEWER

Intriguingly, when 85 samples were stratified into HBV ( n = 61) and non-HBV ( n = 24) groups, elevated expression of GALNT14 at tumor parts was only found in non-HBV HCCs (Fig. 4D and E). In contrast, in HBV-related HCC tissues, the levels of GALNT14 in tumor tissues were highly diverse. Some GALNT14 levels decreased (e.g., patient-1, Fig. 4D), some increased (e.g., patient-2), and some remained unchanged (e.g., patient-3), when compared to the nontumor part.

To understand why the high expression of GALNT14 in cancerous liver tissue is absent in HBV-related HCC, a GALNT14 promoter fragment was inserted upstream of the luciferase gene as a reporter plasmid. This plasmid was co-transfected with mock or HBx expression plasmid into HCC cells. HBx is known as a promiscuous transactivator and plays crucial roles in hepatocarcinogenesis. Surprisingly, the presence of HBx was found to markedly inhibit the activity of the GALNT14 promoter in three independent HCC cell lines, Huh7, HepG2 and J7 (Fig. 4F), indicating that HBx serves a trans-repressor of the GALNT14 promoter, but not transactivator. Due to different levels of HBx in liver tissue between different individuals, diverse levels of GALNT14 were observed, as shown in Fig. 4D.

In the results above, the GALNT14-rs9679162 “GG” genotype was found to predict unfavorable outcomes in specific subgroups of HCC patients treated with sorafenib, such as the HBsAg-negative group. However, it is unclear whether these rules also apply to HCC patients treated with lenvatinib. To answer this question, another cohort of 52 patients with advanced HCC treated with lenvatinib was included in this study. Their baseline clinical and genotypic characteristics are listed in Table 3. Among them, 14 (26.9%) were GALNT14-rs9679162 “GG” genotype, and 38 (73.1%) were GALNT14 “non-GG” (“GT” + “TT”) genotype. The mean age was 65.4 ± 10.3 years; 40 (76.9%) were male; 44 (84.6%) had histological or clinical evidence of liver cirrhosis; 43 (82.7%) had previously received other anticancer treatments, including TACE, chemotherapy, and radiotherapy. Due to the inclusion of patients with advanced HCC, the rates of initial metastases (18, 34.6%) and portal vein thrombosis (33, 63.5%) were higher. Etiological analysis showed that 19 (36.5%) were anti-HCV-positive, 34 (65.4%) were HBsAg-positive, and 7 (15.3%) were alcohol-related. Patients with the “GG” genotype were associated with higher APRI ( P = 0.030) and FIB-4 ( P = 0.038) grades. Notably, the difference in objective response rate (ORR) was borderline significance between patients with “GG” and “non-GG” genotypes, with “GG” genotype associated with a higher ORR (53.1% vs 23.7%, P = 0.051). No significant difference was observed in other variables and clinical scores.

Table S3 also shows the comparison between the cohorts treated with sorafenib and lenvatinib, respectively. Among them, number of alcoholic patients ( P = 0.013, 31.5% vs 15.3%), ECOG performance status ( 0.001), total TKI dose ( 0.001), ORR ( P = 0.003, 14.8% vs 32.7%), PALBI score ( P = 0.024) and FIB grade ( 0.001) were significantly different in the two cohorts.

To understand the predictive value of the GALNT14-rs9679162 “GG” genotype in HCC patients receiving lenvatinib, the same analysis strategy as in sorafenib-treated patients was used. As with sorafenib-treated patients, the “GG” genotype was not significantly associated with OS ( P = 0.791) when all patients were included (Fig. S3A). Univariate followed by multivariate Cox proportional hazard analysis was performed to examine the relationship between clinical factors and OS after receiving lenvatinib (Table 4). In univariate analysis, higher bilirubin, higher AST, higher Child-Pugh grade, higher PALBI score, higher PALBI grade, higher APRI score, higher APRI grade, and higher FIB-4 score were significantly associated with shorter OS. In multivariate analysis, higher AST ( P = 0.023), higher Child-Pugh grade ( P = 0.010), and higher APRI score ( P = 0.005) remained as independent predictors of shorter OS. Notably, in patients with advanced HCC treated with lenvatinib, only a higher APRI score ( P = 0.019) and the “GG” genotype ( P = 0.022) were associated with shorter PFS and TTR (Tables S4 and 5).

OPEN IN VIEWER

Subsequently, to examine the predictive value of GALNT14-rs9679162 “GG” in any subgroup, Cox proportional hazard analysis and Kaplan-Meyer analysis were performed. However, no significant association was found in any subgroup using Cox regression analysis (Fig. S4) and Kaplan-Meyer analysis (Fig. S3B–I). Moreover, except for the AST ⩽ ALT subgroup ( P = 0.024), there was no significant difference in PFS between the “GG” and “non-GG” genotypes in the other subgroups (Fig. S5). Additionally, according to subgroups defined among sorafenib-treated patients, those with HBsAg-positive, non-cirrhosis, and larger tumors ( n = 37, 71.2%; Fig. S6A); with HBsAg-positive, higher platelet counts, and large tumors ( n = 40, 76.9%; Fig. S6B); or with four criteria ( n = 40, 76.9%; Fig. S6C) were excluded from the comparison. However, OS and PFS were not significantly different between patients with GALNT14-rs9679162 “GG” and “non-GG” genotypes. Interestingly, all of these subgroups were significantly different between patients with “GG” and “non-GG” genotypes in TTR (Fig. S6A–C, P = 0.035, 0.018, and 0.018, respectively).

Some patients treated with lenvatinib had partial responses, assessed by AFP and mRECIST [26, 27, 28], we then estimated the TTR in this group of patients. Compared with “non-GG” genotype patients, “GG” genotype patients had a shorter (favorable) TTR (Fig. 5A, P = 0.013). This was also true compared to those with the “TT” genotype (Fig. 5B, P = 0.007, “GG” vs “TT”).

In summary, in HBsAg-negative or anti-HCV-positive patients, the “GG” genotype effectively predicted shorter OS in patients with advanced HCC treated with sorafenib but not lenvatinib. Conversely, in patients treated with lenvatinib, the “GG” genotype was associated with a favorable TTR. Therefore, it is recommended to choose lenvatinib over sorafenib when using targeted agents to treat advanced HCC in “GG” genotype patients.