Autophagy-related gene LC3 expression in tumor and liver microenvironments significantly predicts recurrence of hepatocellular carcinoma after surgical resection

Background The role of autophagy-related markers as the prognostic factor of post-operative hepatocellular carcinoma (HCC) recurrence remained controversial. Methods Overall, 535 consecutive HCC patients undergoing curative resection from 2010 to 2014 were followed and classified with early (ER, <2 years) or late recurrence (LR). Autophagy-related markers, LC3, Beclin-1, and p62 expression was immunohistochemically assessed in HCC and adjacent non-tumor (ANT) tissues. Results HCC recurred in 245 patients: 116 with ER and 129 with LR. The cumulative incidence of recurrence at 1, 3, 5, and 7 years was 9.7%, 33.9%, 53.3%, and 66.3%, respectively. In multivariate analysis, HCC recurrence was significantly associated with low LC3 expression in tumor and ANT tissues, HCC tissues only and ANT tissues only (hazard ratio/95% confidence interval: 6.12/2.473–17.53, 4.18/1.285–13.61, and 1.89/1.299–2.757) and macrovascular invasion (1.63/1.043–2.492) and cirrhosis (1.59/1.088–2.326). ER was significantly associated with low LC3 expression in tumor and ANT tissues, HCC tissues only and ANT tissues only (6.54/2.934–15.81, 3.26/1.034–10.27, and 2.09/1.313–3.321) and macrovascular and microvascular invasion (2.65/1.306–5.343 and 2.55/1.177–5.504). LR was significantly associated with low LC3 expression in tumor and ANT tissues, HCC tissues only and ANT tissues only (5.02/1.372–18.83, 3.19/1.13–12.09, and 1.66/1.051–2.620) and cirrhosis (1.66/1.049–2.631). Patients with low and high LC3 expression in tumor and ANT tissues showed a 5-year cumulative recurrence of 94.3% and 41.7%, respectively (p < 0.001). Conclusions The high LC3 expression in the tumor and liver microenvironments is significantly associated with lower HCC recurrence. Furthermore, tumor characteristics and liver microenvironment were also significantly associated with ER and LR, respectively. Translational impact The analysis for LC3 expression in both the HCC and ANT tissues could identify patients at risk of HCC recurrence.


Introduction
Hepatocellular carcinoma (HCC) is currently the fifth most common type of cancer and the third leading cause of cancer-related mortality worldwide [1][2][3] . In Taiwan, HCC is highly associated with viral-and alcoholicassociated cirrhosis 4 and ranks as the second leading cause of cancer-related death 5 . Even after curative resection for HCC, the 5-year recurrence rate and survival rate remain as high as 60 and 50%, respectively 6,7 . Identifying the factors associated with HCC recurrence after surgical resection could provide a promising strategy to improve the prognosis of HCC patients undergoing curative hepatectomy.
Autophagy is involved in the physiology and pathogenesis of human disease 8,9 . Enhancement or inhibition of autophagy-related proteins has been reported to have therapeutic efficacy in cancer patients 10,11 . However, studies of the involvement of autophagy in tumor recurrence have yielded controversial results [12][13][14][15] . In addition, the prognostic significance of autophagy-related markers such as LC3 and Beclin-1 in predicting the clinical outcome of HCC patients has been reported in previous studies [13][14][15][16][17][18] , but results have been conflicting due to the relatively limited case numbers 19 .
Recently, the molecular and histological changes that occur in the tumor microenvironment have become a main area of focus. The non-tumor liver microenvironment plays an important role in hepatocarcinogenesis [20][21][22][23] . The presence of autophagy in the non-tumor microenvironment was also found to promote tumor growth via the provision of nutrients 24 . These findings highlight the importance of studying both the tumor and non-tumor microenvironments to comprehensively understand the impact of autophagy in HCC development and progression. Hence, we conducted the current large-scale study to explore the impact of autophagy-related markers in both the tumor and adjacent non-tumor (ANT) microenvironments on HCC recurrence after surgical resection. Our results suggest that the low LC3 expression in both tumor and ANT microenvironments strongly predicts HCC recurrence in patients who have undergone curative resection.

Patients and follow-up
This retrospective study included 535 consecutive, histologically proven HCC patients who underwent curative surgical resection between 2010 and 2014 at E-Da Hospital, I-Shou University, Kaohsiung, Southern Taiwan (n = 318) and Changhua Christian Hospital, Changhua, Central Taiwan (n = 217). All patients received regular follow-up every 3 months after surgery. The follow-up period was defined as the duration from the date of operation to the date of either death or the last follow-up. The last follow-up was on December 2016. Time to recurrence (TTR) was defined as the duration from the date of operation to the date of recurrence. Recurrent HCC was defined based on histological confirmation or highly elevated serum alpha-fetoprotein (AFP) in addition to diagnosis via at least two imaging methods according to the recommendations of the American Association for the Study of Liver Disease (AASLD) 25 . The patients were divided into four groups according to their TTR: patients experiencing recurrence within 2 years after operation (early recurrence group, ER, n = 116); patients experiencing recurrence 2-7 years after operation (late recurrence group, LR, n = 129); an all-patient recurrence group (AR, n = 245), consisting of both the ER and LR groups; and patients with no recurrence during the follow-up period after the first hepatectomy (non-recurrence group, NR, n = 290).
For the remaining materials and methods, please see Supporting information.

Baseline demographic data
The demographic and clinicopathological factors of the 535 patients (73.1% male, mean age of 63 years) are shown in Table 1. Regarding the etiology of HCC, 46.7% of the patients had HBV, 28.4% had HCV, 3.9% had HBV/HCV co-infection, and 20.9% were not infected with HBV/ HCV. One-third of the patients had liver cirrhosis, of which 22.8 and 9.5%, respectively, had a Child-Pugh score of A and B. One-tenth of the patients had an Edmondson-Steiner Grade of I-II, and approximately one-fifth of the patients had multiple tumors. Macrovascular and microvascular tumor invasion were observed in 20.7 and 46.0% of the patients, respectively. Regarding tumor stage, 16.4 and 36.1% of the patients were TNM stage III-IV and BCLC stage B-C, respectively. Regarding the expression of autophagy-related markers, 91.6% of the HCC tissues and 59.8% of the ANT tissues were high for LC3; 86.7% of the HCC tissues and 34.8% of the ANT tissues were high for Beclin-1; and 81.1% of the HCC tissues and 8.4% of the ANT tissues were high for p62.

Factors related to HCC recurrence in patients who underwent hepatectomy
In univariate analysis, the presence of liver cirrhosis and macrovascular invasion and the low LC3 expression in HCC tissues or ANT tissues were significantly associated with HCC recurrence (  Table 3).
In multivariate analysis, the Cox proportional hazard analysis identified that patients with low LC3 expression in both HCC and ANT tissues had the highest risk of HCC recurrence (−/−; HR: 6 Table 3).
The 1-and 2-year cumulative incidence of HCC recurrence was 16.9% and 36.2%, respectively, in those with macrovascular invasion (Fig. 2a) and 12.1% and 30.1%, respectively, in those with microvascular invasion (Fig. 2b), which were significantly higher than those in their counterparts. To evaluate whether LC3 expression in HCC and ANT tissues has an effect on ER, LC expression patterns were analyzed in both tissue types. The results revealed a 1-and 2-year cumulative incidence of HCC recurrence of 8.0% and 15.7%, respectively, for patients with high LC3 expression in both tumor and ANT tissues (+/+). Compared to this group, patients with low LC3 in both tissues (−/−; 18.5% and 55.6%, respectively), those low LC3 in HCC tissues only (−/+; 11.7% and 30.5%, respectively) and those low LC3 in ANT tissues only (+/−; 6.7% and 33.3%, respectively) were significantly more prone to early recurrence (Fig. 2c).

Factors related to late HCC recurrence in patients who underwent hepatectomy
To identify factors associated with late HCC recurrence, clinicopathological factors were compared between the LR (n = 129) and the NR groups (n = 290) ( Table 2). In univariate analysis, the presence of liver cirrhosis and microvascular invasion, early BCLC stage, and the low LC3 expression in HCC tissues or ANT tissues were significantly associated with late HCC recurrence.

LC3 staining of tumor and non-tumor tissues in reoperated patients
Having established the association between LC3 expression and HCC recurrence in patients who underwent first curative HCC hepatectomy, we next evaluated LC3 expression patterns in patients who underwent second (n = 32) and third (n = 5) operable HCC hepatectomy due to HCC recurrence ( Table 4). The low LC3 expression in both tumor and ANT tissues was significantly associated with repeated HCC recurrence. Four and 3 patients showed a change from high to low LC3 staining in tumor tissues at the second and third surgical resection, respectively. Moreover, 5 and 4 patients showed a change from high to low LC3 staining in ANT tissues at the second and third surgical resection, respectively. In total, 100% of the patients showing a lossof-LC3 staining in tumor and ANT tissues experienced  ANT adjacent non-tumor, BCLCstage Barcelona clinic liver cancer, + high, − low The significance of bold enteries in tables 1, 2 and 3 is p-value < 0.005. Fig. 2 Cumulative incidence of early recurrence with respect to various clinicopathological factors. Patients with macrovascular invasion (a) and microvascular invasion (b) were significantly more likely to develop early recurrence of HCC. Patients low LC3 expression in HCC tissues (−/+), adjacent non-tumor (ANT) tissues (+/−), or both (−/−) had a significantly higher incidence of early recurrence than patients with LC3 expression in both HCC and ANT tissues (+/+; c). HCC hepatocellular carcinoma, ANT adjacent non-tumor, HR hazard ratio, CI confidence intervals, + high, − low HCC recurrence. This result further demonstrates that loss-of-LC3 expression in both HCC and ANT tissues is associated with a high risk of HCC recurrence, highlighting that LC3 expression significantly predicts HCC recurrence.

Discussion
Our study demonstrated that the high LC3 expression in both the tumor and non-tumor liver microenvironments is significantly associated with lower recurrence, regardless of early or late recurrence. This suggests that the measurement of LC3 expression in both tissues may serve as a predictor of HCC recurrence. The findings that the majority of the patients who underwent second and third operable hepatectomy also had a low LC3 expression in both tumor and ANT tissues and that the loss-of-LC3 expression in both HCC and ANT tissues led to a high risk of HCC recurrence further support the use of LC3 as a prognostic factor for HCC recurrence. In addition to LC3, vascular invasion was significantly associated with ER, and the presence of liver cirrhosis was significantly associated with LR. These results suggest that tumor characteristics have a higher impact on ER, whereas the liver microenvironment has a higher impact on LR.
Tumor risk factors such as tumor invasion into the portal vein and intrahepatic metastasis have been reported to be associated with ER 26 . In the current study, patients with vascular tumor invasion were at a significantly higher risk of developing ER. We postulate that oncogenes and tumor suppressor genes may have undergone genetic alterations during tumor progression 27 , hence allowing tumor cells to acquire invasive and metastatic potential. Intrahepatic metastasis might have occurred prior to hepatectomy or during tumor manipulation, contributing to ER 26,27 . Cirrhosis is associated with carcinogenic potential and is also a predisposing factor for HCC recurrence 28 . Patients with pre-existing cirrhosis have been reported to have lower rates of recurrence-free survival at 3 years or later, suggesting that underlying liver status has an effect on LR 29 . This finding supports our observation that patients with liver cirrhosis, which . HCC hepatocellular carcinoma, ANT adjacent non-tumor, HR hazard ratio, CI confidence intervals, + high, − low denotes poor liver function, are more susceptible to LR. According to our results, the presence of vascular tumor invasion and liver cirrhosis may serve as prognostic factors for ER and LR, respectively.
Using a Drosophila melanogaster malignant tumor model, Katheder et al. 24 recently demonstrated that dormant autophagy-deficient and growth-impaired tumors are capable of reactivating tumor growth when transplanted into an autophagy-efficient host, suggesting that autophagy in the microenvironment impacts tumor growth. Although many studies have revealed that LC3 expression in HCC tissues is higher than in non-tumor tissues and have associated LC3 expression in HCC tissues with tumor development and prognosis 14,18,30 , the relationship between LC3 expression in the non-tumor microenvironment and HCC progression has not been discussed in the literature. Here, we showed that the high LC3 expression in the tumor and ANT microenvironments have additional protective effects against HCC recurrence. Our results clearly demonstrate the importance and potential role of LC3 expression in the tumor and non-tumor liver microenvironments in the prognosis of HCC recurrence. This study is the first to demonstrate that LC3 expression in the non-tumor liver microenvironment is significantly associated with HCC recurrence and that the low LC3 expression in both tumor and ANT tissues significantly increases the risk of HCC recurrence.
Our findings indicate that patients with low LC3 expression in both tumor and ANT tissues are significantly more likely to experience first and second HCC recurrence. All the patients with a loss-of-LC3 staining in tumor and ANT tissues experienced recurrent HCC. The loss-of-LC3 in tumor and ANT tissues was also associated with a high risk of HCC recurrence. Regardless of any other HCC tumor characteristics or the status of ANT tissues in the liver, a low LC3 expression in tumor and ANT tissues at the time of surgery was associated with a significantly increased risk of HCC recurrence. This finding suggests that autophagy-related marker LC3 predicts HCC recurrence.
Our results revealed that the high LC3 in both the tumor and liver microenvironments provides patients who undergo curative hepatectomy with a survival advantage against HCC recurrence. However, there are some limitations to the current study. First, this study used a retrospective design, which could have resulted in unintended bias. Second, the findings of this study must still be validated in Western populations with different ethnicities. Third, the involvement of LC3 expression in the tumor and non-tumor microenvironment in limiting tumorigenesis related to HCC recurrence and its underlying mechanism related to HCC need further investigation in vivo and in vitro.
In summary, the low LC3 expression in both the tumor and non-tumor liver microenvironments was significantly associated with a very high risk of HCC recurrence in patients who underwent curative hepatectomy for HCC. Different factors were associated with early and late recurrence: while the presence of vascular tumor invasion was associated with ER, the liver microenvironment was associated with LR. This study is the first to demonstrate that, in addition to the tumor microenvironment, assessment of autophagy-related markers in the nontumor liver microenvironment is very important for predicting HCC recurrence. The analysis of LC3 expression in tumor and ANT tissues, in conjunction with an assessment of the presence of vascular tumor invasion and liver cirrhosis, could identify patients at risk of HCC recurrence after curative resection. Our results indicated that autophagy-related marker LC3 is significantly associated with HCC recurrence and that LC3 may serve as a potential biomarker for predicting HCC recurrence.

Study highlights
What is current knowledge Autophagy is involved in the physiology and pathogenesis of human disease, including HCC. Autophagy-related markers, such as LC3 and Beclin-1 are used as prognostic factors of HCC. The impact of autophagy-related markers on postoperative HCC recurrence is not documented.
What is new here LC3 expression in the tumor and liver microenvironments is associated with risk of post-operative HCC recurrence. The tumor characteristics and liver microenvironments have effects on early and late recurrence, respectively.

Translational impact
LC3 may serve as a potential biomarker for predicting post-operative HCC recurrence. ANT adjacent non-tumor