Homocysteine: A novel prognostic biomarker in liver transplantation for alpha-fetoprotein- negative hepatocellular carcinoma

Abstract

BACKGROUND:

Precise recipient selection optimizes the prognosis of liver transplantation (LT) for hepatocellular carcinoma (HCC). Alpha-fetoprotein (AFP) is the most commonly used biomarker for diagnosis and prognosis of HCC in the clinical context. As a crucial molecule in methionine cycle, homocysteine (Hcy) level has been proved to be related to HCC progression and metastasis.

OBJECTIVE:

We aimed to explore the prognostic capacity of pre-transplant serum Hcy level in LT for HCC.

METHODS:

This study retrospectively enrolled 161 HCC patients who had underwent LT from donation after cardiac death (DCD) in the First Affiliated Hospital of Zhejiang University from 2015.01.01 to 2018.09.01. Pre-transplant serum Hcy level was incorporated into statistical analysis together with other clinical parameters and pathological features.

RESULTS:

From an overall perspective, significant difference was observed in Hcy level between recurrence ( $n=$ 61) and non-recurrence group ( $n=$ 100) though subsequent analysis showed unsatisfactory predicting performance. In the whole cohort, multivariate analysis showed that lnAFP ( $p=$ 0.010) and Milan criteria (MC, $p<$ 0.001) were independent risk factors of HCC recurrence after LT. MA score based on MC and lnAFP performed well in predicting post-LT tumor recurrence with the AUROC at 0.836 ( $p<$ 0.001) and 3-year recurrence-free survival rate at 96.8% ( $p<$ 0.001) in the low risk group ( $n=$ 69). According to the clinical practice, serum concentration lower than 20 $\mu$ g/L is considered as normal range of AFP. Elevated pre-transplant serum AFP ( $>$ 20 $\mu$ g/L) predicts high HCC recurrence after LT. We further divided the 161 recipients into AFP ${}^{-}$ group ( $n=$ 77, AFP $⩽$ 20 $\mu$ g/L) and AFP ${}^{+}$ group ( $n=$ 84, AFP $>$ 20 $\mu$ g/L). MA score was still well presented in the AFP ${}^{+}$ group and the AUROC for tumor recurrence was 0.823 ( $p<$ 0.001), whereas the predicting accuracy was reduced in AFP ${}^{-}$ group (AUROC: 0.754, $P<$ 0.001). After subsequent analysis, we found that elevated pre-transplant Hcy level ( $>$ 12.75 $\mu$ mol/L) predicted increased tumor recurrence risk in AFP ${}^{-}$ group. The 3-year recurrence-free survival rates were 92.0% and 53.7% ( $p<$ 0.001) in low Hcy subgroup ( $n=$ 40) and high Hcy subgroup ( $n=$ 37) respectively. Multivariate analysis showed that Hcy ( $p=$ 0.040) and Milan criteria ( $p=$ 0.003) were independent risk factors for post-transplant tumor recurrence in AFP ${}^{-}$ group. Further combination of Hcy level and Milan criteria identified a subgroup of AFP ${}^{-}$ recipients with acceptable outcomes even though beyond Milan criteria (3-year recurrence-free survival rate: 77.7%, $p<$ 0.001).

CONCLUSION:

As a classic predictor in HCC prognosis, AFP performed well in our study cohort when combined with Milan criteria. Homocysteine was an effective prognostic biomarker in LT for AFP ${}^{-}$ hepatocellular carcinoma. In recipients exceeding Milan criteria, acceptable post-transplant outcome could be seen in those with low Hcy and AFP level.

1. Introduction

Hepatocellular carcinoma (HCC) [1] is the most common liver cancer worldwide and accounts for majority of cancer-related death. LT is a decent curative option for HCC patients who have no indication for surgical resection though with limited tumor burden. Due to donor organ shortage, technical difficulties, complicated perioperative management and high tumor recurrence rate, the candidate selection for HCC-related LT should be carefully reviewed and optimized. Clinical indexes such as neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, systemic immune-inflammation index, protein induced by vitamin K absence or antagonist II, as well as AFP showed certain predictive value for tumor recurrence after LT [2, 3, 4, 5, 6, 7, 8]. Thus, we decided to explore other clinical molecules to improve recipient selection before LT for HCC.

Methionine is a metabolic dependency of tumor-initiating cells. It has been reported that methionine cycle had an impact on the epigenetic state of cancer cells and ignited tumor initiation [9]. A previous study showed that switching gene expression of two crucial methionine adenosyltransferases, MAT1A to MAT2A (M1-M2 switch) promoted HCC invasion and metastasis [10]. Homocysteine (Hcy), a critical molecule in methionine cycle, folate metabolism and transsulfuration pathway, is routinely reviewed to evaluate the cardiovascular risk for patients. Liver is the key metabolic center for Hcy biosynthesis and metabolism. In fact, elevated Hcy level has been demonstrated to be related to HCC progression and metastasis [11, 12, 13]. However, the relationship between Hcy level and the outcomes of LT patients remains relatively vague. An in-depth analysis is therefore necessary to explicate the role of Hcy in the prognosis of LT patients.

In present study, we detected the Hcy level in the pre-transplant serum of LT recipients with HCC and carried out multiple factor analysis combining other important clinical features. The ultimate purpose is to disclose the predictive value of Hcy in HCC-related LT recipients and optimize the candidate selection criteria, as well as to provide potential treatment guidelines for them.

2. Material and methods

2.1 Patients and demographics

One hundred and sixty-one HCC patients who underwent LT from donation after cardiac death (DCD) in the First Affiliated Hospital of Zhejiang University from 2015.01.01 to 2018.09.01 were enrolled in this study. HCC was confirmed exclusively based on post-operative result. Pre-transplant factors including demographics, preoperative complications, preoperative serum level of alpha-fetoprotein (AFP, $\mu$ g/L), cystatin (CYC, mg/L), creatinine (Cr, $\mu$ mol/L), homocysteine (Hcy, $\mu$ mol/L), blood urea nitrogen (BUN, mmol/L), total protein (TP, g/L), albumin (Alb, g/L), globulin (g/L), and glomerular filtration rate (GFR, ml/min) were selected and presented in the catalogued data. Patient’s details about morphological features, etiology, tumor differentiation grades, inflammatory status, post-transplant recurrence and patient survival were also collected.

Endpoint of follow-up was on 2019.09.30, the median follow-up time was 2.75 years (95% CI: [2.26–3.23]). In the 161 patients enrolled, 147 were male and 14 were female. Their ages ranged from 42 to 60 years old, with an average of 51 years old. The average length of survival was 1.83 years. One hundred and fifty-four cases of HCC were derived from HBV infection, 1 from HCV infection, 2 from alcoholic hepatitis and 4 from unknown reason (Table 1).

Table 1

Demographic and clinical characteristics of enrolled patients with/without tumor recurrence

Characteristics	Total	Non-recurrence	Recurrence	$p$ -value
Patients ( $n$ )	161	100	61
Etiology ( $n$ [%])			0.534
HBV	154 (95.7)	94 (94.0)	60 (98.4)
HCV	1 (0.6)	1 (1.0)	0 (0.0)
ASH	2 (1.2)	2 (2.0)	0 (0.0)
Unknown	4 (2.5)	3 (3.0)	1 (1.6)
CVD ( $n$ [%])				0.302
No	156 (96.9)	98 (98.0)	58 (95.1)
Yes	5 (3.1)	2 (2.0)	3 (4.9)
HP ( $n$ [%])				0.842
No	141 (87.6)	88 (88.0)	53 (86.9)
Yes	20 (12.4)	12 (12.0)	8 (13.1)
DM ( $n$ [%])				0.344
No	143 (88.9)	87 (87.0)	56 (91.8)
Yes	18 (11.1)	13 (13.0)	5 (8.2)
Age (year)	51.29 $\pm$ 9.31	51.27 $\pm$ 9.04	51.34 $\pm$ 9.81	0.948
Gender ( $n$ [%])				0.688
Male	147 (91.3)	92 (92.0)	55 (90.2)
Female	14 (8.7)	8 (8.0)	6 (9.8)
BMI (kg/m ${}^2$ )	22.92 $\pm$ 3.15	23.15 $\pm$ 3.27	22.54 $\pm$ 2.93	0.230
MELD score	11 (6–46)	12 (6–40)	11 (6–46)	0.527
Child-pugh classification ( $n$ [%])			0.741
Grade A	70 (43.5)	42 (42.0)	25 (41.7)
Grade B	58 (36.0)	34 (34.0)	24 (40.0)
Grade C	33 (20.5)	22 (22.0)	11 (18.3)
MC ( $n$ [%])				0.000
Within	69 (42.9)	66 (66.0)	3 (4.9)
Beyond	92 (57.1)	34 (34.0)	58 (95.1)
Tumor number ( $n$ [%])				0.001
Single	78 (48.4)	59 (59.0)	19 (31.1)
Multifocal	83 (51.6)	41 (41.0)	42 (68.9)
Max diameter of tumor ( $n$ [%])		0.001
$⩽$ 5 cm	117 (72.7)	82 (82.0)	35 (57.4)
$>$ 5 cm	44 (27.3)	18 (18.0)	26 (42.6)
Total diameter ( $n$ [%])				0.000
$⩽$ 8 cm	101 (62.7)	76 (76.0)	25 (41.0)
$>$ 8 cm	60 (37.3)	24 (24.0)	36 (59.0)
Pathological differentiation ( $n$ [%])			0.708
High/moderate	121 (75.2)	74 (74.0)	47 (77.0)
Low	40 (24.8)	26 (26.0)	14 (23.0)
AFP ( $\mu$ g/L)	27.7 (0.8–80000)	13.6 (0.8–80000.0)	145.7 (0.9–80000.0)	0.000
CYC (mg/L)	0.94 $\pm$ 0.33	0.98 $\pm$ 0.33	0.87 $\pm$ 0.32	0.046
Hcy ( $\mu$ mol/L)	12.96 $\pm$ 4.67	12.50 $\pm$ 4.74	13.71 $\pm$ 4.47	0.011
Cr ( $\mu$ mol/L)	69.41 $\pm$ 15.51	71.20 $\pm$ 16.64	66.51 $\pm$ 13.07	0.034
BUN (mmol/L)	5.10 $\pm$ 1.62	5.20 $\pm$ 1.73	4.94 $\pm$ 1.40	0.196
Total protein (g/L)	66.03 $\pm$ 7.58	65.53 $\pm$ 7.54	66.86 $\pm$ 7.64	0.298
Albumin (g/L)	37.45 $\pm$ 5.92	37.12 $\pm$ 6.18	37.99 $\pm$ 5.45	0.274
Globulin (g/L)	28.59 $\pm$ 6.23	28.41 $\pm$ 6.32	28.87 $\pm$ 6.12	0.819
AG	1.38 $\pm$ 0.38	1.38 $\pm$ 0.39	1.38 $\pm$ 0.38	0.729
GFR (ml/min)	107.95 $\pm$ 69.55	109.72 $\pm$ 87.43	105.01 $\pm$ 14.56	0.606
NLR	2.86 (0.65–30)	2.74 (0.72–11.20)	3.14 (0.65–30)	0.318
SII	241.67 (24.57–3253.33)	198.67 (24.57–2274.23)	279.60 (38.5–3253.33)	0.013
ST (year)	1.83 (0.11–4.74)	2.08 (0.11–4.74)	1.45 (0.16–4.74)	0.003
RFST (year)	1.32 (0.02–4.74)	2.08 (0.11–4.74)	0.43 (0.02–3.12)	0.000
MFT (year)	2.74 (2.52–2.95)

Note. HBV, hepatitis B virus; HCV, hepatitis C virus; ASH, alcoholic hepatitis; PHB, primary biliary cirrhosis; CVD, cardiovascular diseases; HP, hypertension; DM, diabetes; BMI, Body Mass Index; MELD, Model for end-stage liver disease; MC, Milan criteria; Total diameter, total tumor diameter; AFP, alpha fetal protein; CYC, cystatin; Cr, creatinine; Hcy, homocysteine; BUN, blood urea nitrogen; AG, albumin to globulin ratio; GFR, glomerular filtration rate; NLR, lymphocyte ratio; SII, systemic immune-inflammation index; ST, survival time; RFST, recurrence free survival time; MFT, median follow-up time. ${}^{*}P$ -values written in bold indicate a statistical significance.

Figure 1.

The tumor recurrence-predicting capacity of AFP and Hcy in LT for HCC. A. Recipients with low pre-transplant serum AFP level ( $n=$ 100) were associated with reduced recurrence risk compared with recipients who had high AFP level ( $n=$ 61); B. Recipients with low pre-transplant serum Hcy level ( $n=$ 116) were associated with reduced recurrence risk.

Figure 2.

ROC curve and Kaplan-Meier survival analysis of MA score in predicting tumor recurrence after LT. A. ROC curve for HCC recurrence of the MA score, the AUROC $=$ 0.836 ( $p<$ 0.001). B. Kaplan-Meier estimates of cumulative risk of HCC recurrence according to the MA score. Recipients in MAT ${}^{\text{high}}$ group were associated significantly higher recurrence risk with 3-year RFS rate at 32.6%, whereas the 3-year RFS rate was 96.8% in MAT ${}^{\text{low}}$ recipients ( $p<$ 0.001).

The procedures of this study were in accordance with the Helsinki Declaration. All livers were acquired from deceased donors strictly according to the guidelines of the China donation after citizen’s death. No donor organs were obtained from executed prisoners or other institutionalized persons. The study was approved by China Liver Transplantation Registry (CLTR), the only national registry in Mainland China, according to the Regulations on Human Organ Transplant and national legal requirements. Patient consents were obtained. The study protocol was also approved by the Human Ethics Committee of First Affiliated Hospital, Zhejiang University School of Medicine.

2.2 Statistical analysis

The results were analyzed by SPSS (Version 20.0, inc., Chicago, IL, USA). AUROC curve and Youden index were calculated for optimal cut-off value. Kaplan-Meier method was used for univariate analysis. Student’s $t$ -test or Mann-Whitney U test was used, when available, for comparisons between groups. To improve the statistical rationality, repeated univariate analysis by Cox regression was carried out before multivariate analysis to reassess the correlation between the independent variable and the outcome variable. Cox regression method was used for multivariate analysis and construction of prognostic model. A $p$ -value lower than 0.05 was considered statistically significant.

3. Results

3.1 Novel pre-operative tumor recurrence predicting score based on Milan criteria and serum AFP level

All characteristics shown in Table 1 were enrolled to evaluate their correlation with post-transplant tumor recurrence. Among them, pre-operative serum AFP level ( $p<$ 0.001), Cr level ( $p=$ 0.034), Hcy level ( $p=$ 0.011), SII ( $p=$ 0.013), MC ( $p<$ 0.001), tumor number ( $p=$ 0.001), max diameter of tumor ( $p=$ 0.001), and total tumor diameter ( $p<$ 0.001) were significantly different in the recurrence group ( $n=$ 61) and non-recurrence group ( $n=$ 100) according to the primary univariate analysis. As a classic biomarker for HCC prognosis, AFP was well presented in our data set ( $p<$ 0.001, Fig. 1A). However, in the whole cohort, the pre- dicting ability of Hcy for post-LT tumor recurrence risk was unsatisfactory ( $p=$ 0.011, Fig. 1B).

Table 2

Cox regression analysis of the variables of recurrence-free survival in 161 patients with HCCs after LT

Characteristics	Univariate analysis			Multivariate analysis
	HR	95% CI	$p$ -value	HR	95% CI	$p$ -value
lnAFP	3.16	1.89–5.27	$<$ 0.001	1.69	1.00–2.85	0.050
MC	22.16	6.93–70.84	$<$ 0.001	17.42	5.36–56.60	$<$ 0.001
Tumor number	2.53	1.47–4.36	0.001	1.02	0.56–1.86	0.995
Total diameter	3.10	1.86–5.19	$<$ 0.001	0.64	0.37–1.12	0.115
Hcy	1.05	1.00–1.10	0.058	1.02	0.96–1.09	0.560
Cr	0.98	0.96–1.00	0.015	0.99	0.97–1.01	0.328
NLR	1.051	1.00–1.10	0.045	1.01	0.96–1.07	0.603
SII	1.001	1.00–1.00	0.002	1.00	1.00	0.972

Note. HR, hazard ratio. ${}^{*}P$ -values written in bold indicate a statistical significance.

Table 3

Cox regression analysis of the variables of recurrence-free survival in 77 patients with AFP ${}^{-}$ HCCs after LT

Characteristics	Univariate analysis			Multivariate analysis
	HR	95% CI	$p$ -value	HR	95% CI	$p$ -value
Age	1.01	0.96–1.07	0.700
Gender	2.91	0.96–8.78	0.059	2.84	0.91–8.81	0.071
BMI	0.95	0.83–1.08	0.404
MELD	1.01	0.97–1.06	0.572
Child	1.21	0.67–2.19	0.524
MC	9.29	2.70–31.98	$<$ 0.001	6.88	1.91–24.82	0.003
Number	1.58	0.64–2.90	0.317
Total diameter	3.67	1.47–9.13	0.005	0.95	0.30–2.55	0.804
Pathology	1.82	0.71–4.63	0.210
CYC	1.07	0.26–4.49	0.926
Hcy	1.18	1.08–1.29	$<$ 0.001	1.11	1.00–1.22	0.040
Cr	0.98	0.95–1.02	0.322
BUN	0.83	0.61–1.14	0.243
Total protein	1.01	0.95–1.08	0.702
AG	1.60	0.51–5.02	0.422
GFR	1.01	0.98–1.04	0.634
NLR	1.07	0.87–1.32	0.500
SII	1.00	1.00	0.092	1.00	1.00	0.94

Note. HR, hazard ratio. ${}^{*}P$ -values written in bold indicate a statistical significance.

Figure 3.

Hcy performed well in predicting post-transplant tumor recurrence of AFP ${}^{-}$ HCC patients. A. Kaplan-Meier estimates of cumulative risk of HCC recurrence based on AFP level. All 161 patients were divided into AFP ${}^{-}$ ( $n=$ 77) and AFP ${}^{+}$ ( $n=$ 84) group according to AFP level at the cut-off value of 20 $\mu$ g/L. The 3-year RFS rates were 73.8% and 45.4% in these two groups respectively ( $p<$ 0.001); B. ROC curve for HCC recurrence of the MA score in AFP ${}^{+}$ group. The AUROC was 0.823 ( $p<$ 0.001). C. ROC curve for HCC recurrence of the MA score in AFP ${}^{-}$ group. The AUROC was 0.754 ( $p<$ 0.001). D. Kaplan-Meier estimates of cumulative risk of HCC recurrence according to Hcy level in AFP ${}^{-}$ group. 77 recipients in AFP ${}^{-}$ group were further divided into high Hcy subgroup ( $n=$ 40) and low Hcy subgroup ( $n=$ 37) by the optimal cut-off value of 12.75 $\mu$ mol/L. The 3-year RFS rates were 92.0% and 53.7% respectively ( $p<$ 0.001).

Then we carried out univariate analysis to reassess the correlation between post-LT tumor recurrence and potential risk variables by Cox regression. Eight variables including lnAFP, MC, tumor number, total tumor diameter, Hcy, Cr, NLR and SII were subsequently enrolled in multivariate Cox regression analysis, the results showed that lnAFP ( $p=$ 0.010) and MC ( $p<$ 0.001) were independent risk factors of tumor recurrence after LT (Table 2). We further established a predicting model based on above two variables and named it as MA score $=$ 2.858 $\times$ MC (0, within MC; 1, beyond MC) $+$ 0.523 $\times$ lnAFP. The AUROC $=$ 0.837 ( $p<$ 0.001, Fig. 2A). The optimal cut-off of MA score calculated according to Youden index is 1.88. The whole cohort was divided into two groups: MA ${}^{\text{low}}$ group ( $n=$ 69) and MA ${}^{\text{high}}$ group ( $n=$ 92). The Kaplan-Meier analysis showed that recipients in MA ${}^{\text{high}}$ group had reduced recurrence-free survival (RFS) time with 3-year RFS rate at 32.6%, whereas the 3-year RFS rate was 96.8% in MA ${}^{\text{low}}$ recipients ( $p<$ 0.001, Fig. 2B.)

3.2 Hcy level precisely predicted post-LT tumor recurrence for AFP ${}^{-}$ HCC recipients

As we mentioned above, AFP is the most commonly used clinical biomarker which has strong correlation with the prognosis of HCC patients, we further divided all 161 patients into AFP ${}^{-}$ ( $n=$ 77) and AFP ${}^{+}$ ( $n=$ 84) group according to AFP level at the cut-off value of 20 $\mu$ g/L. The 3-year RFS rates were 73.8% and 45.4% in these two groups respectively ( $p<$ 0.001, Fig. 3A). Then we verified the predicting capacity of MA score in above two groups, the ROC curve showed that MA score still performed well in AFP ${}^{+}$ group with AUROC at 0.823 ( $p<$ 0.001, Fig. 3B), though reduced accuracy was observed in the AFP ${}^{-}$ group with AUROC at 0.754 ( $p=$ 0.001, Fig. 3C). It’s obvious that even with negative AFP expression, some recipients were still at high recurrence risk. Therefore additional effective pre-operative serum predictors should be developed in the context of clinical practice to refine the subtyping of AFP ${}^{-}$ recipients. We further evaluated the recurrence predicting ability of clinical and pathological features by univariate analysis in AFP ${}^{-}$ group ( $n=$ 77), five variables including gender, MC, total tumor diameter, Hcy and SII were subsequently enrolled in the multivariate Cox regression analysis through backward wald method (Table 3). The results showed that Hcy and MC outperformed other parameters in these recipients. The AUROC of Hcy in predicting tumor recurrence was 0.726 ( $p=$ 0.003). Recipients with negative pre-transplant serum AFP level were further divided into Hcy ${}^{\text{high}}$ group ( $n=$ 40) and Hcy ${}^{\text{low}}$ group ( $n=$ 37) by the optimal cut-off value of 12.75 $\mu$ mol/L. The 3-year RFS rates were 92.0% and 53.7% respectively according to survival analysis ( $p<$ 0.001, Fig. 3D).

3.3 Combination of Hcy level and Milan criteria optimized selection of recipients with negative pre-transplant serum AFP level

As shown in Table 3, both Hcy level ( $p=$ 0.040) and Milan criteria ( $p=$ 0.003) were independent risk factors in predicting tumor recurrence after LT for AFP ${}^{-}$ recipients. The Kaplan-Meier analysis proved that the 3-year RFS rate was significantly higher in recipients within Milan criteria ( $n=$ 45) than those beyond Milan ( $n=$ 32) criteria (95.3% vs 47.1%, $p<$ 0.001, Fig. 4A). Also, the combination of Hcy and Milan criteria effectively distinguished AFP ${}^{-}$ recipients who might obtain better recurrence-free survival even though beyond Milan criteria. Survival analysis showed that 9 out of the 32 recipients beyond Milan criteria showed a better 3-year RFS rate (77.7%) which was a rather acceptable prognosis after LT ( $p<$ 0.001, Fig. 4B). Since MA score presented satisfactory post-LT tumor recurrence predicting capacity in AFP ${}^{+}$ group, no further analysis was conducted in this group.

Figure 4.

Combination of Hcy level and Milan criteria optimized selection of AFP ${}^{-}$ recipients. A. Kaplan-Meier estimates of cumulative risk of HCC recurrence according to Milan criteria. Three-year RFS rate was significantly higher in recipients within Milan criteria ( $n=$ 45) than those beyond Milan ( $n=$ 32) criteria (95.3% vs 47.1%, $p<$ 0.001); B. Kaplan-Meier estimates of cumulative risk of HCC recurrence according to Milan criteria and Hcy level. Nine out of the 32 recipients beyond Milan criteria showed a better 3-year recurrence-free survival rate (77.7%) which was a rather acceptable prognosis after LT ( $p<$ 0.001).

4. Discussion

HCC accounts for the majority of liver cancer worldwide and the fourth-leading cause of cancer-related death [1, 14]. LT is an effective curative option for HCC patients with limited tumor burden and optimized recipient selection is of great significance for HCC patients to obtain favorable prognosis. Pre-transplant serum AFP level has been recognized to effectively predict tumor recurrence after LT for HCC patients though HCC recurrence were still observed in some AFP negative recipients [15]. In fact, single biomarker is far from enough to meet with the command of precise selection of LT recipients while multi-biomarker models can bridge the gap by serving as a more effective predictor for post-operative prognosis. Our results showed that pre-transplant serum Hcy level can be applied to distinguish AFP ${}^{-}$ recipients with different recurrence risk and refine existing recipient selection criteria.

In the past decade, several prognostic models were established to evaluated the post-transplant HCC recurrence risk for recipients with intent to optimize MC [16]. AFP is a surrogate marker of both tumor differentiation and vascular invasion. Accordingly, high AFP levels have been reported to be associated with high recurrence rates [17]. Thus, AFP was applied to construct predicting models in several researches [18]. In our research, the multivariate analysis showed that MC and lnAFP were independent risk factors for tumor recurrence. We further established a novel prognostic score name MA score which effectively predicting post-transplant HCC recurrence. The MA score is convenient and might be useful in clinical practice after multi-center validation and optimization.

The most significant finding in current article is the tumor-recurrence predicting capacity of homocysteine in certain HCC-related LT recipient group. Homocysteine is an important intermediate product of one-carbon metabolism, which encompasses folic acid and methionine cycle [19]. One-carbon units are essential substances for the proliferation and regeneration of cancer cells by influencing nucleotide synthesis and methylation pathways [20]. HCC is a highly malignant cancer and the needs for one-carbon units is extremely integrant. Previous study of our research team reported that genetic variations in one-carbon metabolism pathway genes of donor liver was associated with increased risk of HCC frecurrence [21]. Butler et al. reported that serum cystathionine, methionine, and S-adenosylmethionine but not HCY were associated with increased HCC risk ( $p<$ 0.05) [22]. While in our study, correlation was seen between Hcy level and overall HCC recurrence rate after LT. One possible reason for this is that LT recipients were generally more vulnerable to metabolic disorders which contribute to the recurrence of HCC.

We further demonstrated that in a special HCC-related LT recipient group, which was the AFP ${}^{-}$ patients, elevated pre-transplant serum Hcy level ( $>$ 12.75 $\mu$ mol/L) indicated a higher tumor recurrence risk after LT (Fig. 2B). When combined with Milan criteria, Hcy level distinguished AFP ${}^{-}$ recipients with better prognosis in which the 3-year recurrence-free survival rate was as high as 77.7% even though they were beyond the criteria ( $p<$ 0.001, Fig. 3B). There is robust evidence to support our research. One of which reported that high Hcy level in HCC not only resulted from but also promoted hepatocarcinogenesis via CYP450-EET metabolism by crosstalk of DNA demethylation of CYP2J2 and ERK1/2 signaling [23]. Although the underlying reason of which Hcy portrays remarkable predicting ability in AFP ${}^{-}$ HCC group was uncertain, it is undoubtedly an effective post-transplant tumor recurrence predicting biomarker which can be routinely examined in clinical interviewing to optimize existing recipient selection criteria.

We assumed that as a valid biomarker for SAM metabolism, abnormal elevation of Hcy level reflects increment in Hcy remethylation activity which results in stimulated methionine synthesis and reduced SAM level. This in turn increasing the HCC risk by triggering several tumor-promoting pathways. Previous study demonstrated that the production of SAM, which is the biological methyl donor essential to vast reactions within the cell, relies greatly on methionine adenosyltransferases (MATs) regulation [24]. Impaired biosynthesis of SAM in patients with HCC indicates aberrant M1-M2 switch which subsequently induces genomic instability [25], mitochondrial dysfunction and dysregulated metabolic pathways including RAS/ERK [26], IKK/NF-kB [27], PI3K/AKT [28] and LKB1/AMPK [29] and eventually leads to HCC development. A recent research reported that simultaneous downregulation of MAT1A enhanced osteopontin (OPN) expression which triggered ERK signaling and thereby promoted metastasis [10]. Thus, these findings suggested a promising anticancer treatment since the restoration of SAM level can inhibit stimulation of cancer-promoting signaling pathways and potentially reduce the tumor recurrence risk after LT for HCC recipients.

Molecules which play crucial roles in SAM and HCY regulation are considerable therapeutic targets. At present, there are two ongoing SAM-related clinical trials in HCC treatment. One of which (NCT03178929) is designed to investigate SAM treatment in HCC patients after radical treatment, where tumor recurrence is the endpoint while another (NCT02586285) aims to study the SAM treatment in HCC patients after curative treatment. Inspired by above researches, we believed that Hcy level might be a practical biomarker for peri-transplant management. It’s rational to proactively administer SAM to recipients with high Hcy level in order to decrease the post-transplant tumor recurrence risk. Our research team have also carried out animal experiments to validate our hypothesis.

In a nutshell, we revealed the predicting capacity of homocysteine in post-transplant tumor recurrence risk for HCC recipients and suggested potential novel therapeutic targets for those at high risk. Despite the encouraging results of our study, there are a few shortcomings which need to be improved: 1) limited sample size; 2) no multi-center validation as pre-transplant serum Hcy level is not routinely examined in other medical centers, we plan to generalize the detection of serum Hcy level test in the peri-operative examination and carry out multi-center validation in future; 3) this study focused on clinical phenomenon and further experimental evidence is required to support our hypothesis.

5. Conclusion

We innovatively came up with that homocysteine was a promising prognostic biomarker in LT for AFP ${}^{-}$ HCC. In recipients exceeding Milan criteria, those with low Hcy and AFP level also have acceptable outcomes. Restoration of SAM level can be integrated to individualized management for HCC patients with high Hcy and AFP level to reduce post-LT tumor recurrence risk. Further exploration of the underlying molecular mechanisms is needed to meet with the clinical demands of treatment-oriented phenotyping. We also established a prognostic model based on Milan criteria and lnAFP named MA score which performed well in predicting tumor recurrence in the whole research cohort. To resolve the sample size limitation and improve the predicting capacity of MA as well as Hcy, multi-center validation will be carried out in future.