Human gastric cancer tissue samples were obtained from 564 patients (405 males and 159 females; age range, 29–82 years) with primary gastric tumors who underwent D1 or D2 radical gastrectomy surgery at the First Affiliated Hospital of Dalian Medical University (Dalian, China) between 2011 and 2013. The gastric tissues outside the cancer loci were selected as the tumor-adjacent tissue samples; 335 tumor-adjacent tissue samples from these patients were collected as controls. The diagnosis of gastric cancer was confirmed by pathological staining. Clinicopathological data including patient age, gender, tumor location, tumor size, histological differentiation, invasion depth, Helicobacter pylori (H. pylori) infection, ascites, lymphatic invasion, lymph node metastasis, distant metastasis and tumor-node-metastasis (TNM) stage (26) were retrospectively retrieved from the medical records. The patients did not undergo radiotherapy or chemotherapy prior to surgery. Outcomes of interest included the overall survival (OS) and disease-free survival (DFS) rates. OS was defined as the duration from diagnosis to the last follow-up, or to mortality. DFS was calculated as the time from the initial diagnosis to local recurrence, or distant metastasis.

The patients consented to have tissue samples collected at the time of admission for surgery, according to protocols authorized by the Regional Human Ethics Committee of Dalian Medical University. The Medical Ethics Committee of Dalian Medical University approved the present study. Due to the retrospective nature of the present study, the Ethics Committees waived the requirement for written informed consent from the patients.

Tissue cores were extracted from formalin-fixed and paraffin-embedded tissue blocks containing the tumor tissue samples and the tumor-adjacent tissue samples and arrayed into a paraffin recipient block. Sections (4 µm thick) were obtained from the TMA blocks using a microtome, mounted on poly-L-lysine-coated glass slides and used for immunohistochemistry.

TMA sections were washed in xylene, rehydrated in a graded ethanol series and washed in tap water. The tissue sections were then heated in 100 ml 10 mM sodium citrate buffer (pH 6.0) in a microwave oven (high power, 700 W for 3 min; medium power, 400 W for 3 min; low power, 100 W for 3 min; a total of 9 min) to retrieve antigen. The sections were then incubated at 37°C for 30 min in 3% H₂O₂/methanol to block endogenous peroxidase activity. Nonspecific protein binding sites were blocked by 10% normal goat serum (Boster Biological Technology, Ltd., Wuhan, China) at 37°C for 30 min. The sections were incubated in a primary polyclonal rabbit anti-human antibody against PARP-1 (cat. ab6079; dilution 1:200; Abcam, Cambridge, UK) overnight at 4°C, and subsequently incubated with a biotinylated goat anti-rabbit secondary antibody (cat. no. SAEP031; dilution, 1:150; Wuhan Elabscience Biotechnology Co., Ltd., Wuhan, China) for 30 min and streptavidin horseradish peroxidase (LSAB kit; Dako, Glostrup, Denmark) for an additional 30 min. Sections were stained with 3,3-diaminobenzidine at room temperature for 50–60 sec, counterstained with hematoxylin, dehydrated with graded ethanol and mounted with neutral resin. For the negative controls, the primary antibodies were replaced with phosphate-buffered saline.

Two pathologists blinded to the experimental conditions examined the final effective immunostaining under a light microscope (Eclipse 50i; Nikon Corporation, Tokyo, Japan). The intensity of immunoreactivity was graded on a scale of 0–3, as follows: 0) For no visible staining, i) for low staining, ii) for moderate staining and iii) for high staining. The percentage of the stained nuclei in gastric cancer gland cells and normal gland cells was assigned using 5% increments. Five random, non-overlapping fields were defined and 40 cells for each field (a total of 200 cells) were selected in order to calculate the percentage of stained cells present in each sample. The immunoreactivity score (IRS) was determined by multiplying the intensity score and the percentage of stained nuclei, giving a minimum IRS score of 0 and a maximum of 300%. Receiver operating characteristic (ROC) curve analysis was performed to determine an optimal cutoff IRS for PARP-1 expression. Plotting the sensitivity and specificity for each outcome under study generated the ROC curves.

The Pearson's χ² test and the Fisher's exact probability test were used to determine significant differences between the categorical data. The Mann-Whitney U test was used to detect differences in the IRS or lymph node metastasis number between various groups. The Wilcoxon rank-sum test was used to compare the IRS of gastric cancer tissues with tumor-adjacent tissues, for the paired tissue samples. Kaplan-Meier survival plots were generated and comparisons between the survival curves were determined with a log-rank test. Cox's proportional hazards regression model was used to evaluate the association between the potential confounding variables and the prognosis (OS or DFS). Only those cases with complete data on all the variables (n=153) were included in the multivariate analyses. The data were processed using SPSS 16.0 software (SPSS, Inc., Chicago, IL, USA). P<0.05 was considered to indicate statistical significance.

The clinicopathological characteristics of 564 patients with gastric cancer are summarized in Table I. For these 564 patients, the features of tumor location of 498 patients (88.3%), tumor size of 522 patients (92.6%), histological differentiation of 555 patients (98.4%), depth of tumor invasion of 529 patients (93.8%), status of H. pylori infection of 297 patients (52.7%), presence of ascites of 479 patients (84.9%), lymphatic invasion of 540 patients (95.7%), lymph node metastasis of 553 patients (98.0%), distant metastasis of 434 patients (77.0%) and TNM stage of 481 patients (85.3%) were recorded. The average age (mean ± standard deviation) of gastric cancer patients in the present study was 60.1±10.4 years (range, 29–82 years). The histological differentiation of the cancers was determined in 555 patients as follows: 14.6% high differentiation (n=81), 21.4% moderate differentiation, (n=119) and 64.0% low differentiation (n=355). The depth of tumor invasion was evaluated in 529 patients as follows: 10.8% T1 (n=57, tumor invades the mucosa or submucosa), 15.9% T2 (n=84, tumor invades the muscularis propria), 64.1% T3 (n=339, tumor invasion of the serosa), and 9.3% T4 (n=49, tumor invades the adjacent organs and structures). The status of H. pylori infection was defined by a ¹³C-urea breath test. Of the total group, 208 patients (70.0%) were diagnosed with H. pylori infection while 89 patients (30.0%) were negative for H. pylori infection. The stage of the cancer was evaluated in 481 patients according to the TNM staging system as follows: 13.7% Stage I (n=66), 24.9% stage II (n=120), 35.3% stage III (n=170) and 26.0% stage IV (n=125). Ascites, lymphatic invasion, lymph node metastasis and distant metastasis occurred in 176 (36.7%) of 479 patients, 203 (37.6%) of 540 patients, 316 (57.1%) of 553 patients and 148 (34.1%) of 434 patients, respectively.

Follow-up information was available for 523 patients with gastric cancer. During the follow-up period of 0–95 months, relapses occurred in 246 cases and mortality occurred in 208 cases. The 5-year survival rate was determined to be 46.5%. The mean OS time was 65.8 months (95% confidence interval (CI); 62.6–68.9 months) and the mean DFS time was 60.5 months (95% CI; 57.3–63.7 months) respectively.

Using immunohistochemistry, the expression levels of PARP-1 in 564 gastric cancer tissue samples and 335 tumor-adjacent control tissue samples were studied. Positive PARP-1 staining in gastric tissues appeared as brown particles that were primarily nuclear (Fig. 1). The gastric cancer tissue samples exhibited significantly more intense staining for PARP-1 expression, as compared with tumor-adjacent tissue samples (P<0.001; Fig. 2A). There was also a significant increase in the IRS of PARP-1 expression among paired samples (P<0.001; Fig. 2B).

The ROC analysis was performed to determine an optimal cutoff score for PARP-1 expression; the ROC curves for the clinicopathological features with statistical significance are indicated in Fig. 3. The area under the curve for the lymph node metastasis status had the largest relative area (Fig. 3A). Based on this outcome, a cutoff score of 175% was selected for PARP-1 expression levels: The low and high expression levels of PARP-1 were defined as IRS≤175% and IRS>175%, respectively. A total of 298 (52.8%) tumors exhibited low expression levels of PARP-1 and 266 (47.2%) tumors exhibited high expression levels of PARP-1.

The association between PARP-1 expression levels and the clinicopathological characteristics of gastric cancer were investigated (Table II). Statistical analysis indicated that PARP-1 expression levels were increased with the following: Decreased differentiation (P<0.001), increased depth of invasion (P=0.037), presence of lymphatic invasion (P<0.001), presence of lymph node metastasis (P<0.001) and advanced TNM stage (P=0.015). In addition, high PARP-1 expression levels were increased among patients with H. pylori infection (P=0.032). The expression levels of PARP-1 were not identified to be significantly associated with age, gender, tumor location, tumor size, distal metastasis or presence of ascites in the patients (P>0.05).

The association between PARP-1 expression levels and lymph node metastasis in gastric cancer patients was also investigated (Fig. 4). The numbers of lymph nodes positive for metastasis were recorded from 553 gastric cancer patients, and ranged from 0 to 33 lymph nodes among the 553 patients. Compared with PARP-1 low expression levels, PARP-1 high expression levels were associated with a larger number of positive lymph nodes (P<0.001; Fig. 4A). Furthermore, lymph node metastasis-positive tumors exhibited a higher IRS of PARP-1 expression (P<0.001; Fig. 4B).

Kaplan-Meier analysis and a log-rank test were used to evaluate the association between the expression levels of PARP-1 protein and patient mortality. High expression levels of PARP-1 were associated with significantly shorter OS time (P<0.001; Fig. 5A) and DFS time (P=0.001; Fig. 5B).

Univariate analysis of the potential prognostic impact of the clinicopathological parameters identified age, gender, tumor size, histological differentiation, invasion depth, presence of ascites, lymph node metastasis, TNM stage and PARP-1 expression as significantly associated with OS and DFS in gastric cancer patients (Table III). Subsequently, multivariate Cox regression models (using the same clinicopathological features) revealed that only histological differentiation, presence of ascites, lymph node metastasis and TNM stage remained as independent prognostic factors (Table IV).

The present study used Kaplan-Meier analysis to investigate the association of PARP-1 expression levels with OS and DFS in subgroups of gastric cancer patients, which were categorized according to clinicopathological parameters. The results of the current study indicated that high expression levels of PARP-1 were associated with shorter OS (P<0.001; Fig. 6A) and DFS (P<0.001; Fig. 6B) in the subset of patients with H. pylori infection. However, in those patients without H. pylori infection, the expression levels of PARP-1 were not significantly associated with OS or DFS (OS, P=0338; DFS, P=0.999; Fig. 6C and D). In addition, high expression levels of PARP-1 were associated with shorter OS (P=0.001; Fig. 7A) and DFS (P=0.006; Fig. 7B) in the subset of patients with an advanced TNM stage, but not with an early TNM stage (OS, P=0.834; DFS, P=0.955; Fig. 7C and D). The impact of PARP-1 expression levels on the prognosis was not significantly different between the subgroups of gastric cancer patients categorized according to age, gender, tumor size, tumor location, invasion depth, histological differentiation, presence of ascites, distant metastasis, lymphatic invasion or lymph node metastasis.