Prognostic value of autophagy related proteins ULK1, Beclin 1, ATG3, ATG5, ATG7, ATG9, ATG10, ATG12, LC3B and p62/SQSTM1 in gastric cancer

Abstract

Autophagy-related (ATG) genes contributed to tumorigenesis and cancer progression. This study aims to investigate the expression of ATG proteins and their clinicopathological significance in gastric cancer. Nine well-known ATG proteins, (ULK1, Beclin 1, ATG3, ATG5, ATG7, ATG9, ATG10, ATG12 and LC3B) and p62/SQSTM1, which represented key regulators that participated in whole autophagosomes stepwise processes, were detected in a large cohort of 352 primary gastric cancer patients. Among these 352 patients, 117 cases were randomly assigned to the training set to detect the clinicopathological value of ATG proteins, and another 235 patients were used as the testing set for further validation. Except for Beclin 1, ATG9 and ATG10, another six ATG proteins and p62/SQSTM1 were closely correlated with histological types for gastric cancer. Moreover, low expression of ULK1, Beclin 1 and ATG10 were associated with lymph node metastasis. In addition, down-regulation of ULK1, Beclin 1, ATG7 and ATG10, up-regulation of ATG12 correlated with advanced TNM stage. Importantly, multivariate cox analysis identified ULK1, Beclin 1, ATG3 and ATG10 as favorable independent prognostic factors for overall survival. Combination analysis of ULK1, Beclin 1, ATG3, ATG10 revealed the improved prognostic accuracy for gastric cancer. Our study showed that ATG proteins might serve as novel prognostic biomarkers in gastric cancer, and supply a new valuable insight into cancer treatment targeting autophagy for patients.

Introduction

Gastric cancer, a chronic Helicobacter pylori infection related malignancy, remains the second leading cause of cancer mortality worldwide [1]. Although the fluoropyrimidine and platinum-based combination therapies have partially benefited the early-stage gastric cancer patients, local recurrence and distant metastasis are still the major reasons for the poor survival of advanced subgroups [2]. Supported by accumulated clinical and prognostic biomarkers, such as peritoneal metastases, plasma alkaline phosphatase, ERCC1, HER2 and HER3 [3,4], the risk classification of patient outcome was defined more accurately. HER2 and HER3 amplification, for example, were detected in 12-64% of patients with gastric cancer and predicted a poor overall survival (OS) [3,4]. Importantly, Trastuzumab, a recombinant monoclonal antibody against HER2, had greatly improved progression free survival and overall survival for advanced gastric cancer in the phase III ToGA trial [5]. Therefore, it will be of great clinical value to identify more novel HER2-alike molecular markers, that not only predicting the prognosis individually but also providing promising therapeutic molecular targets for gastric cancer.

The modulation of autophagy is considered as one of the hallmarks of cancer cells. There are three primary types of autophagy: macroautophagy (commonly referred to as autophagy), microautophagy and chaperone-mediated autophagy. As described in detail previously [6-9], the most distinguishing feature of macroautophagy is the formation of the double-membrane bound phagophore and autophagosome. Autophagosomes undergo a stepwise maturation process, including initiation, nucleation, elongation, maturation and degradation. Simultaneously, a number of key signaling events that orchestrate the sequestration process have been identified, which has led to the discovery of more than 30 autophagy-related (ATG) genes. The key autophagy regulatory ATG genes are listed in Figure S1. At the autophagosomes initiation step, the dephosphorylation of mTOR-dependent sites on the ULK1/Atg13/FIP200 complex releases ULK1’s activity, and auto-phosphorylation of the ULK1/ATG13/FIP200 complex by localization of ATG9 to phagophore [10-12]. This leads to the formation of autophagy specific Beclin 1/ATG14L/VPS34/VPS15 complex and coats a cup-shaped isolation membrane, which serves as a recruitment signal for the isolation membrane elongation [13]. Two protein conjugation systems, which are respectively triggered by ubiquitin-like molecules ATG12 and LC3, are required to autophagosome elongation and maturation. Firstly, ATG12 is activated by ATG7 (E1) and transferred to ATG10 (E2) following by covalent linking to ATG5 [14,15]. Secondly, LC3 is cleaved by ATG4, and formed a cleaved LC3-I with a C-terminal glycine residue. ATG7 (E1) further activates LC3-I and transfers it to ATG3 (E2), subsequent conversion of LC3-I into LC3-II [16-18]. With the assistance of ATG5/ATG12 conjugates, p62/SQSTM1 bound LC3-II is then conjugated to phosphatidylethanolamine and coated the outer surfaces of the autophagosome [19-21]. The autophagosome ultimately fuses with a lysosome that facilitates the turnover of engulfed material by lysosomal/vacuolar acid hydrolases.

Importantly, the aberrant ATG genes were crucial for tumorigenesis and cancer progression. As the central player indispensable for the first phases of autophagy, Beclin 1 is monoallelically deleted in human ovarian, breast and prostate cancers, and might be a prognostic marker in a variety of solid tumors [22]. Ectopic ATG10 and ULK1 were significantly correlated with lymph node metastasis and a poor OS for colorectal and breast cancers [23,24]. Nevertheless, the clinicopathological values of ATG proteins in cancer remain debatable. Beclin 1 high expression, for example, a favorable biomarker to predict outcome for gastric cancer, hepatocellular carcinoma and high-grade gliomas [25-29] and also as an inferior prognostic biomarker for nasopharyngeal carcinoma [30]. For gastric cancer, the clinicopathological significance of ATG proteins, except Beclin 1 [27-29] had not yet been characterized.

Collected data suggesting that key autophagy regulatory ATG proteins contribute to tumorigenesis and tumor progression, and may predict prognosis in human cancers. In order to find out the relationship between ATG proteins and clinicopathological parameters, especially, the prognostic value in gastric cancer, we detected the clinicopathological value of 9 known ATG proteins (ULK1, Beclin 1, ATG3, ATG5, ATG7, ATG9, ATG10, ATG12 and LC3B) and p62/SQSTM1, which represented key regulators that participated in whole autophagosomes stepwise processes in a large cohort of 352 gastric cancers.

Materials and methods

This study was approved by the Human Ethics Committee of the Sixth Affiliated Hospital, Sun Yat-sen University. A written informed consent was obtained from all the patients at the time of admission, with which the tissue, blood and other samples might be used for scientific research but did not relate to patient’s privacy.

Patients

A total of 352 primary gastric cancer patients who underwent initial surgical resection were recruited in the present study. The archived formalin-fixed, paraffin-embedded tissues were collected in both the First Affiliated Hospital and the Sixth Affiliated Hospital of Sun Yat-sen University from January 2002 to October 2006. Of these patients, 117 cases were randomly assigned to the training set to assess the clinicopathological value of these 10 proteins, and another independent cohort of 235 patients were used as the testing set for further validation. Patients were selected by the following inclusion criteria: pathologically confirmed as gastric adenocarcinoma; without oncological surgery, chemotherapy, and radiotherapy history; completed follow-up information and paraffin-embedded specimens; and received postsurgical chemotherapy depending on the severity of the disease and according to the National Comprehensive Cancer Network (NCCN) guidelines. Moreover, patient would be excluded for any of the following reasons: previously received any anticancer therapy; prior malignancy; and pregnancy. The patient stage was redefined according to 2010 AJCC staging system for gastric cancer [31]. The pathological differentiation of intestinal subtype denoted to papillary and/or tubular adenocarcinomas, and diffuse subtype included of poorly differentiated, signet-ring cell, and/or mucinous adenocarcinomas [32].

Tissue microarrays (TMAs) construction

TMAs were constructed in accordance with a previously described method [30]. For each case, two cores taken from the selected tumor area and additional one core from normal adjacent mucosa (within 2 cm distance from the tumor margin) were used to construct the TMAs. Briefly, a hollow needle was utilized to punch and remove bipartite cylinders tissue cores (1.0 mm in diameter) from selected donor tissues regions. Further, the punched tissue cores were inserted into a recipient paraffin block with a precisely spaced, array pattern, using an automatic tissue arraying instrument (Beecher Instruments, Silver Spring, Maryland, USA).

Immunohistochemical (IHC) staining

TMAs containing 352 gastric cancers were detected by IHC. The sections (4 μm thick) were dewaxed in xylene and rehydrated in gradient ethanol solutions. Endogenous peroxidase activity and nonspecific binding of antibodies were blocked by hydrogen peroxide and goat serum albumin, respectively. Antigenic retrieval was performed in EDTA antigenic retrieval buffer (pH 8.0) or citrate buffer (pH 6.0) with a microwave. Sections were then incubated with primary antibodies at 4°C overnight. A total of 10 proteins representing the key players in autophagosomes formation processes, consisting of initiation and nucleation: Beclin 1 (Cell Signaling, #3738, 1:100 dilution), ULK1 (Sigma-Aldrich, #A7481, 1:200 dilution) and ATG9 (Abcam, #ab108338, 1:100 dilution); elongation: ATG12 (Sigma-Aldrich, #A8731, 1:200 dilution), ATG7 (Sigma-Aldrich, #A2856, 1:200 dilution), ATG10 (MBL International, #M151-3, 1:300 dilution), ATG5 (Sigma-Aldrich, #A0731, 1:400 dilution); formation: ATG3 (Sigma-Aldrich, #A3606, 1:300 dilution), p62/SQSTM1 (MBL International, #PM045, 1:200 dilution) and LC3B (Sigma-Aldrich, #L7543, 1:300 dilution), were detected in this study. Then, the sections were washed in PBS buffer for 5 min 3 times, treated with 100 μl secondary antibody for 30 min at room temperature and stained with DAB (Dako, Envision System/DAB-chromogen, K5007, DK-2600 Glostrup, Denmark) according to the manufacturer’s instructions. The sections were washed in PBS buffer for 10 min 3 times and counterstained with hematoxylin for 2 min at room temperature. Finally, the sections were washed in dH₂0 for 5 min 3 times and mounted the coverslips. A negative control was utilized by changing the specific primary antibody with non-immune serum immunoglobulins at the 1:200 dilutions.

Semiquantitative assessment of IHC staining

The expression level of each protein was evaluated by combined assessment of staining intensity and extent as we previously described [33]. We scored the staining intensity as following: negative (score 0), bordering (score 1), weak (score 2), moderate (score 3) and strong (score 4). Staining extent was graded into five parts according to the percentage of elevated staining cells in the field: negative (score 0), 0-25% (score 1), 26-50% (score 2), 51-75% (score 3) and 76-100% (score 4). The multiplied overall score was subjected to further clinicopathological analysis. Immunohistochemical staining was evaluated and scored by two independent pathologists (Fan XJ and Cao QH) blinded to clinical follow-up data. In case completely different results occurred, they would work together to confirm the score.

Selection of cutoff score for each biomarker

The receiver operating characteristic (ROC) curve analysis was used to selection of cutoff score in the training set as we previously reported [33]. Briefly, the sensitivity and specificity for patient outcome at each score were plotted to generate a ROC curve. The score localized closest to the point at both maximum sensitivity and specificity, ie., the point (0.0, 1.0) on the curve, was identified as the cutoff score that could be correctly classified patient outcome as death or alive.

Western blot analysis

Gastric cancer tissue samples preserved in liquid nitrogen were subsequently ground and lysed with the RIPA buffer (Sigma-Aldrich, R0278) for Western blot analysis. The protein concentration was tested by the Bradford method with BSA (Sigma-Aldrich, A4503) as the standard protein. Equal amounts of tissue extract were electrophilic run in SDS-polyacrylamide gel and transferred to nitrocellulose membrane (BioRad Laboratories, 162-0094) for antibody blotting. The membrane was further blocked and incubated with according primary antibodies (ULK1, Sigma-Aldrich, #A7481, 1:1000 dilution; Beclin 1, Cell Signaling, #3738, 1:1000 dilution; ATG3, Sigma-Aldrich, #A3606, 1:3000 dilution; ATG5, Sigma-Aldrich, #A0731, 1:1000 dilution; ATG7, Sigma-Aldrich, #A2856, 1:2000 dilution; ATG9, Abcam, #ab108338, 1:2000 dilution; ATG10, MBL International, #M151-3, 1:500 dilution; ATG12, Sigma-Aldrich, #A8731, 1:500 dilution; p62/SQSTM1, MBL International, #PM045, 1:1000 dilution and LC3B, Sigma-Aldrich, #L7543, 1:1000 dilution). The HRP labeled secondary antibody was diluted in 1% BBB buffer (1:15000). The membrane was then incubated in diluted secondary antibody for 1 hour at room temperature on an orbital shaker. The membrane was washed in the 1X PBST for 5 min 3 times. Finally, the membrane was incubated in 10 ml SignalFire^TM ECL Reagent (CST, #6883) with gentle agitation for 1 min at room temperature, drained off excess developing solution, wrapped in plastic wrap and exposed to x-ray film.

Statistical analysis

In the training set, ROC analysis was employed to generate the IHC cutoff score for each ATG protein (SPSS-Analyze-ROC Curve). For validation, the clinicopathological value of each ATG proteins were evaluated in the testing set and overall patients. The chi-square test or Fisher’s exact test was used to evaluate the relationship between each ATG proteins and clinicopathological variables (SPSS-Analyze-Descriptive Statistics--Crosstabs). The multivariate Cox proportional hazards model was utilized to estimate the hazard ratio (HR) and 95% confidence interval (CI) (SPSS-Survival-Cox Regression). The OS difference between patients subsets was analyzed by Kaplan-Meier analysis and log-rank tests (SPSS-Analyze-Kaplan-Meier). Statistically significant difference was considered if the P value from a two-tailed test was less than 0.05. Statistical analysis was performed using SPSS v. 17.0 (SPSS, Inc., Chicago, IL).

Results

Patient characteristics

Of these 352 patients, the median duration of overall survival (OS) for training set and testing set was 33.34 ± 2.75 months and 31.82 ± 1.74 months, respectively (P = 0.613). The clinicopathological features of these two subgroups, including gender, age, tumor location, tumor size, histological type, TNM stage, and lymph node metastases status, were all comparable (Table 1 and Table S1). Moreover, the Kaplan-Meier survival analysis showed that the training set and testing set had the similar overall survival probability (data not shown), indicating a balanced clinicopathological features of both cohorts for further analysis.

Table 1.

Expression status of autophagy related proteins in relation to patient characteristics in 352 gastric cancers

	ULK1		Beclin 1		ATG3		ATG5		ATG7		ATG9		ATG10		ATG12		P62/SQSTM1		LC3B
	Low	High	Low	High	Low	High	Low	High	Low	High	Low	High	Low	High	Low	High	Low	High	Low	High
Gender
Male	105	130	136	99	86	149	47	188	63	172	89	146	152	83	44	191	112	123	85	150
Female	65	52	72	45	38	79	29	88	34	83	56	61	68	49	18	99	57	60	38	79
P valuea	0.070		0.565		0.479		0.336		0.704		0.085		0.244		0.462		0.910		0.553
Age
≥ 58b	79	95	90	84	59	115	42	132	43	131	54	120	106	68	27	147	75	99	49	125
< 58	91	87	118	60	65	123	34	144	54	124	91	87	114	64	35	143	94	84	74	104
P value	0.288		0.007		0.912		0.300		0.283		< 0.001		0.583		0.330		0.071		0.010
Location
Upper third	54	46	59	41	27	73	21	79	20	80	32	68	68	32	17	83	44	56	41	59
Middle third	34	62	59	37	34	62	14	82	20	76	38	58	53	43	15	81	48	48	36	60
Lower third	72	69	78	63	57	84	35	106	51	90	67	74	86	55	28	113	65	76	40	101
Whole	10	5	12	3	6	9	6	9	6	9	8	7	13	2	2	13	12	3	6	9
P value	0.017		0.288		0.188		0.083		0.010		0.077		0.062		0.809		0.067		0.192
Size
≥ 5 cm	111	101	131	81	72	140	50	162	66	146	94	118	138	74	43	169	105	107	82	130
< 5 cm	59	81	77	63	53	87	26	114	31	109	51	89	82	58	19	121	64	76	41	99
P value	0.065		0.224		0.495		0.291		0.069		0.151		0.219		0.117		0.514		0.087
Histological type
Intestinal	126	152	162	116	81	197	50	228	63	215	108	170	170	108	38	240	119	159	85	193
Diffuse	44	30	46	28	43	31	26	48	34	40	37	37	50	24	24	50	50	24	38	36
P value	0.036		0.596		< 0.001		0.002		< 0.001		0.086		0.346		< 0.001		< 0.001		0.001
Lymph node metastasis
With	143	102	161	84	94	151	58	187	73	172	106	139	176	69	49	196	125	120	91	154
Without	27	80	47	60	30	77	18	89	24	83	39	68	44	63	13	94	44	63	32	75
P value	< 0.001		< 0.001		0.069		0.162		0.194		0.242		< 0.001		0.094		0.104		0.224
Tumor-Node-Metastasis stage
I	9	32	11	30	23	28	6	39	8	37	17	28	23	22	7	38	18	27	12	33
II	21	63	40	44	26	58	14	66	17	63	27	53	31	49	7	73	31	49	23	57
III	112	77	124	65	67	122	44	145	56	133	84	105	140	29	39	150	98	91	73	116
IV	28	10	33	5	18	20	12	26	16	22	17	21	26	12	9	29	22	16	15	23
P value	< 0.001		< 0.001		0.195		0.160		0.042		0.381		< 0.001		0.040		0.089		0.243

^aChi-square test or Fisher’s exact test.

^bmedian age.

ATG proteins expression status in gastric cancer tumor and normal adjacent tissue

As shown in Figure 1 and Figure S2, Beclin 1, ATG10 and p62/SQSTM1 were respectively weakly expressed in gastric cancer tissues, whereas strongly expressed in normal adjacent tissues. Conversely, the other 7 markers, including ULK1, ATG3, ATG5, ATG7, ATG9, ATG12 and LC3B, were highly expressed in the gastric cancer tissues, and lowly or moderately expressed in adjacent non-tumor tissues. In addition, the western blot analysis showed that the LC3-II expression level was up-regulated in the tumor tissues than adjacent normal tissues. Interestingly, we also observed an increased LC3-II to LC3-I ratio in the tumor tissues, indicating an activation of autophagy (Figure 1).

Figure 1.

ATG proteins expression in human gastric cancer and non-tumor tissues. Western blot analysis of ATG expression in four pairs of matched gastric tumor (T) and normal adjacent epithelia (N). Equal loading of ATG proteins was determined by β-actin.

Correlations between ATG proteins expression and clinicopathological features

In the training set, ROC analysis showed that the IHC cutoff scores of ULK1, Beclin 1, ATG3, ATG5, ATG7, ATG9, ATG10, ATG12, LC3B and p62/SQSTM1 were 8, 6, 6, 8, 8, 6, 6, 6, 8 and 8, respectively. Dichotomized these proteins according to their own cutoff scores, we found that ULK1, Beclin 1, ATG3, ATG5, ATG7, ATG9, ATG10, ATG12, LC3B and p62/SQSTM1 were respectively highly expressed in 51.7%, 40.9%, 66.8%, 78.4%, 72.4%, 58.8%, 37.5%, 82.4%, 65.1% and 52.0% of overall patients (Figure 2; Table 1). Moreover, we also detected the similar expression status of these 10 markers in the training set and testing set (Table S1).

Figure 2.

The expression of ATG proteins in gastric cancer. A. The expression of representative ATG proteins at initiation step, containing ULK1, Beclin 1 and ATG9, In left panel a, immunohistochemistry (IHC) revealed high expression of ULK1, ATG9 and low expression of Beclin 1 in the case 36. In right panel b, IHC demonstrated low expression of ULK1, ATG9 and high expression of Beclin 1 in the case 157. B. The expression of representative ATG proteins at elongation step, including ATG5, ATG7, ATG10, ATG12. In left panel a, IHC showed high expression of ATG5, ATG7 ATG12 and low expression of ATG10 in the case 239. In right panel b, high expression of ATG10 and low expression of ATG5, ATG7, ATG12 were displayed in the same case 12. C. The expression of representative ATG proteins at maturation step containing ATG3, LC3B, p62/SQSTM1. In left panel a, IHC showed that ATG3 and LC3B were highly expressed, while p62/SQSTM1 was lowly expressed in the same case 65. In right panel b, the low expression of ATG3 and LC3B and high expression of p62/SQSTM1 in case 112. In panel a and b, the right figures displayed representative ATGs proteins expression in selected tumor zone with enlarged view.

As shown in Table 1, Beclin 1, ATG9 and LC3B were closely correlated with age in overall patients (P = 0.007, P < 0.001 and P = 0.010, respectively). In addition, ULK1 and ATG7 expression levels were significantly varied among different tumor locations (P = 0.017 and P = 0.010, respectively). In addition, comparing intestinal with diffuse gastric cancer subtypes, significantly different ULK1, ATG3, ATG5, ATG7, ATG12, p62/SQSTM1 and LC3B expression levels were detected (P = 0.036, P < 0.001, P = 0.002, P < 0.001, P < 0.001, P < 0.001 and P = 0.001, respectively). Importantly, low expression of ULK1, Beclin 1 and ATG10 were associated with lymph node metastasis (P < 0.001, P < 0.001 and P < 0.001, respectively) and advanced TNM stage (P < 0.001, P < 0.001 and P < 0.001, respectively). Moreover, low expression of ATG7 and ATG10 correlated with high TNM stage (P = 0.042 and P = 0.040, respectively).

ATG proteins expression and survival analysis

As shown in Figure 3A, the median OS duration for the subgroup with high ULK1 expression was 38.6 months, whereas was 26.2 months for the subgroup with low ULK1 expression in testing set (P < 0.001). Moreover, the similar OS difference between high and low ULK expression subgroups was also obtained in overall patients (median OS: 39.1 VS 27.7 months, P < 0.001, Figure 4A). Similarly, the statistically significant OS differences between the highly and lowly expressed subgroups were also observed for Beclin 1 (Figures 3B and 4B), ATG3 (Figures 3C and 4C) and ATG10 (Figures 3D and 4D) in the testing set and overall patients. However, we failed to detect the prognostic values for ATG5 (Figures S3A and S4A), ATG7 (Figures S3B and S4B), ATG9 (Figures S3C and S4C), ATG12 (Figures S3D and S4D), LC3B (Figures S3E and S4E) and p62/SQSTM1 (Figures S3F and S4F) in both testing set and overall patients (all P > 0.05).

Figure 3.

Kaplan-Meier estimated of overall survival according to ATG proteins expression level in testing set. In testing set, high expression of ULK1 (A), Beclin 1 (B), ATG3 (C) and ATG10 (D) were closely correlated with favorable overall survival.

Figure 4.

Kaplan-Meier estimated of overall survival according to ATG proteins expression level in overall patients. In overall patients, high expression of ULK1 (A), Beclin 1 (B), ATG3 (C) and ATG10 (D) were closely correlated with favorable overall survival.

Importantly, Cox multivariate regression analyses demonstrated that ULK1 (HR: 0.620, P = 0.008), Beclin 1 (HR: 0.617, P = 0.006), ATG3 (HR: 0.703, P = 0.045) and ATG10 (HR: 0.642, P = 0.022) were indeed the independent indicators to predict the prognosis for overall patients with gastric cancer (Table 2). Interestingly, we found that histological type (HR: 0.501, P = 0.001), other than gender, age, tumor location and size, was also displayed a significantly predictive value for OS (Table 2).

Table 2.

Cox multivariate regression analyses of ATG proteins, p62/SQSTM1 expression and clinicophathological variables on overall survival in testing set and overall patients

Variable	Testing set			Overall patients
	HR	95% CI	P value	HR	95% CI	P value
ULK1 (low VS high)	0.626	0.405-0.966	0.034	0.620	0.435-0.884	0.008
Beclin 1 (low VS high)	0.625	0.401-0.974	0.038	0.617	0.437-0.872	0.006
ATG3 (low VS high)	0.591	0.384-0.910	0.017	0.703	0.498-0.992	0.045
ATG5 (low VS high)	0.945	0.601-1.486	0.807	1.218	0.835-1.778	0.306
ATG7 (low VS high)	0.926	0.560-1.532	0.764	0.977	0.645-1.480	0.913
ATG9 (low VS high)	1.226	0.789-1.905	0.366	1.210	0.845-1.735	0.298
ATG10 (low VS high)	0.577	0.360-0.923	0.022	0.642	0.440-0.938	0.022
ATG12 (low VS high)	1.437	0.804-2.659	0.222	1.164	0.740-1.831	0.512
P62/SQSTM1 (low VS high)	1.054	0.591-1.879	0.859	1.056	0.658-1.695	0.822
LC3B (low VS high)	1.065	0.625-1.814	0.817	1.214	0.750-1.963	0.430
Gender (male VS female)	1.012	0.659-1.554	0.957	1.022	0.721-1.450	0.902
Age (≥ 58 VS < 58)	1.366	0.894-2.088	0.149	1.288	0.915-1.811	0.146
Location (upper third & middle third VS lower third & whole)	0.909	0.603-1.372	0.650	0.867	0.622-1.210	0.403
Size (≥ 5 cm VS < 5 cm)	0.992	0.654-1.504	0.96	1.012	0.727-1.408	0.944
Histological type (intestinal VS diffuse)	0.535	0.329-0.862	0.010	0.501	0.331-0.759	0.001

Abbreviation: TNM, Tumor-Node-Metastasis stage.

Combination analysis of ULK1, Beclin 1, ATG3, ATG10 expression and survival

To analyze the prognostic value of combining ULK1, Beclin 1, ATG3, ATG10 for gastric cancer, we divided the patients into five groups: without positive maker (0), with only one positive marker (1), with two (2), three (3) and four (4) positive markers. The OS curve of these five subgroups was showed respectively in Figure 5A. More importantly, patients with four positive markers displayed a superior survival compared with those with 0-3 positive makers (P = 0.008, Figure 5B). The same difference was found between patients with 0-2 positive makers and those with 3-4 positive markers (P = 0.001, Figure 5C), patients with 0-1 positive makers and those with 2-4 positive markers (P < 0.001, Figure 5D).

Figure 5.

Combination of ULK1, Beclin 1, ATG3, ATG10 expression and overall survival. A. Survival curve of patients without positive marker (0), with only one positive marker (1), with two (2), three (3) and four (4) positive markers. B. Patients with four positive markers (4) displayed a superior prognosis compared with those with zero-to-three positive makers (0-3). C. Patients with three-to-four positive markers (3-4) displayed a favorable prognosis compared with those with zero-to-two positive makers (0-2). D. Patients with two-to-four positive markers (2-4) displayed a better survival compared with those with zero-to-one positive maker (0-1).

Discussion

In this study, we assessed the expression pattern of 9 ATG proteins and p62/SQSTM1 involved the autophagosomes formation processes (ULK1, Beclin 1 and ATG9), elongation steps (ATG5, ATG7, ATG10 and ATG12), and maturation (ATG3, p62/SQSTM1 and LC3B). Our results revealed that low expression of ULK1, Beclin 1, ATG3 and ATG10 predicted a favorable prognosis in a 5-year follow-up analysis. More importantly, the four-biomarker-based combination had a significantly better prognosis compared with one to three-protein-based combination. Furthermore, multivariate analysis revealed that ULK1, Beclin 1, ATG3 and ATG10 were indeed the significant independent prognostic factors for alive, not the death.

ULK1, Beclin 1 and ATG9 play important roles at the autophagosomes formation processes [10,12,34]. In non-small cell lung cancer, activation of ULK1 preserved cancer cell cytoskeletal dynamics and released the cell motility effector FAK, leading to enhanced metastatic dissemination to bone/liver [35]. Beclin 1 is classified as a haplo-insufficient tumor suppressor gene [22,36]. Knockdown of Beclin 1 enhanced chemotherapy-induced apoptosis and reduced clone formation and impaired cell growth of hepatocellular carcinoma cells [37]. Indeed, accumulated studies demonstrated that ULK1 and Beclin 1 deficiency were closely correlated with poor clinicopathological features, including lymph node metastasis, advanced TNM stage and unfavorable outcome, in a variety of cancers [24-26,30,38,39]. Consistent with previous studies, low expressions of ULK1 and Beclin 1 were correlated with lymph node metastasis, advanced TNM stage and poor prognosis. Interestingly, ULK1 expression was up-regulated in gastric tumor compared to adjacent normal tissues, and predicted a favorable OS, suggesting ULK1 playing a tumor suppressor gene role in the gastric cancer. However, the underlying mechanism is not clear and need to be further investigated. Unfortunately, we did not find any association between ATG9 with clinicopathological variables except age. Importantly, multivariate analysis confirmed that ULK1 and Beclin 1 were the independent prognostic biomarker for gastric cancer, suggesting that autophagosomes initiation step played an essential role in regulating cancer cell metastasis and predicting patient prognosis for gastric cancer.

At autophagosome elongation steps, ATG12 is activated by ATG7 (E1) and ATG10 (E2), and followed by covalent linkage to an internal lysine on ATG5 [14,15]. Functionally, inhibition of Atg5/7 by siRNA or chloroquine suppressed cancer cell epithelial-mesenchymal transition (EMT) and diminished invasiveness for hepatocellular carcinoma [40]. Serum deprivation of mesenchymal stem cells (MSCs) upregulated ATG10 and ATG12, and further activated autophagy to facilitate MCF-7 breast cancer cell survival and progression [41], indicating that autophagosome elongation step contributed to tumorigenesis and tumor progression. In the present study, the associations between the expression of ATG5, ATG7, ATG10, ATG12 and the clinicopathological parameters were investigated. We further confirmed that deficiency of ATG10 and ATG12 is associated with lymph node metastasis in gastric cancer (Table 1 and Table S1). Significantly, our multivariate analysis demonstrated that ATG10 was an independent favorable prognostic biomarker for gastric cancer. Similar to the controversial role of autophagy and Beclin 1 in tumorigenesis and predicting prognosis, our study showed an opposite outcome of the prognostic impact of ATG10 when compared with the previous study. Jo YK et al [23] reported that increased expression of ATG10 was associated with lymphovascular invasion and lymph node metastasis, and predicted an unfavorable OS for colorectal cancer. The possibility was that ATG10 may have different roles in different types of cancer, the expression of ATG10 was also depending upon the cancer cell types. However, we did not find any association of ATG5, ATG7, ATG12 expression with patients’ OS.

At the last step of autophagosome maturation, ATG7 (E1) activates LC3 and transfers it to ATG3 (E2), and subsequent converses LC3-I into LC3-II [16,17,42]. Together with ATG5/ATG12, p62/SQSTM1 bound LC3-II coats the outer surfaces of the autophagosome. In the present study, we found that high expression of ATG3 predicted a favorable OS for gastric cancer. To our best of knowledge, this is the first report about the positive prognostic value of ATG3 in human cancers. Consistent with reported results, an in vitro study found that siRNA targeting ATG3 and ATG7 evidently decreased AKT inhibitor AZD5363 treated prostate cancer cell proliferation and survival [43]. Inversely, high-expressed ATG3 gene and protein, the myelodysplastic syndrome SKM-1 cell line was inhibited cell proliferation and vitality. The malignancy of SKM-1 cell line was decreased after transfected with ATG3 [44]. Similar to ULK1, the expression of ATG3 was upregulated in tumor tissues, but played a role as putative suppressor gene. The potential reason may be complicated. Our ongoing study will focus on the mechanisms research. These data demonstrated that ATG3 may play discriminatory role in different types of cancer and/or phases of tumorigenesis. The underlying mechanism linking ATG3 and tumor progression will be further investigated. However, the statistically OS difference was not shown between the subgroups with high or low expression of LC3 and p62/SQSTM1. Nonetheless, previous studies showed that increased LC3 expression typically has poor prognosis in lung and pancreatic cancers and melanoma [45,46] while decreased LC3 expression has been correlated with tumorigenesis and an inferior prognosis in breast cancer and early stage cervical squamous cell carcinoma [47,48]. When it comes to p62/SQSTM1, the results of prognostic value were also divers depending on different cancer types. The role of p62/SQSTM1 as an inferior prognostic marker has been reported in lung adenocarcinoma and prostate cancer [49]. In contrast to this, Ellis RA et al showed that high p62 expression status predict longer disease-free survival in cutaneous malignant melanoma [50]. Combined, these results indicated that ATG proteins play different roles in different cancer types and stage, and autophagosome maturation ATG may mediate tumor progression in gastric cancer.

Moreover, we interestingly found that, histological type, other than gender, age, tumor location and size, was also an independent factor for OS. The patients with diffuse type had an inferior OS compared to those with intestinal type [51]. We founded that different tumor locations did not show any predictive value for gastric survival. Some reports revealed that the patients with upper third gastric cancer showed a poorer prognosis than those with other locations due to the stomach anatomy, complex lymphatic systems, or technical difficulties during surgery [52,53].

In addition, there are limitations in the present study. The absence of data for metastasis and progression analyses was one limitation. Furthermore, another independent validation cohort with larger sample size from multicenter was needed to evaluate the prognostic value of ATG proteins in patients with gastric cancer.

Taken together, our study demonstrated that autophagosomes formation processes related ATG proteins are associated with histological differentiation, refined the risk to lymph node metastasis, and might to be the prognostic indicators for gastric cancer. Also, these findings supply a comprehensive understanding of correlation between autophagy and gastric cancer, and new valuable insight into cancer treatment targeting autophagy for gastric cancer patients.

Disclosure of conflict of interest

None.