Approval for the present study was obtained from the Medical Ethics Committee of China Medical University (Liaoning, China). The need for written informed consents by the patients was waived due to the retrospective nature of the present study. A total of 91 osteosarcoma specimens were collected from patients with osteosarcoma who underwent curative tumor resection at the First Affiliated Hospital of China Medical University (Liaoning, China) between January 2003 and January 2008. During surgery, a total of 47 adjacent non-tumorous bone tissues were collected as controls. No patients underwent chemotherapy or radiotherapy prior to surgery. The clinical stage of the patients was classified according to the 6th edition of the Tumor Node Metastases Classification of Malignant Tumors, International Union against Cancer (25). The clinicopathological information for all patients is presented in Table I. All 91 patients with osteosarcoma were monitored during follow-up appointments, which lasted between 72 and 132 months. Mortality occurred in 9 patients during the follow-up period. The median overall survival (OS) and disease-free survival (DFS) of patients was 68 and 55 months, respectively.

miR-335 expression and subcellular localization in osteosarcoma tissues and matched noncancerous bone tissues were measured using in situ hybridization. Briefly, following the manufacturer's protocol, the tissue slides were mixed with 5′-digoxigenin LNA-modified-miR-335 (Exiqon A/S, Vedbaek, Denmark) using the IsHyb in situ Hybridization kit (BioChain Institute Inc., Eureka Drive, Newark, CA, USA).

Rock1 protein expression and subcellular localization in osteosarcoma tissues and matched noncancerous bone tissues was detected using immunohistochemical staining. Firstly, tissue sections (4 µm thick) were incubated with the rabbit monoclonal Rock1 antibody (dilution, 1:100; cat. no. ab134181; Abcam, Cambridge, UK) at 4°C overnight and subsequently incubated with biotinylated secondary antibodies (dilution, 1:1,000; cat. no. E043201; Dako; Agilent Technologies, Inc., Santa Clara, CA, USA) at 37°C for 30 min. Samples were subsequently incubated with streptavidin horseradish peroxidase for another 30 min (LSAB kit; Dako; Agilent Technologies, Inc.) and stained with 3,3-diaminobenzidine. Finally, the slides were counterstained with hematoxylin, dehydrated in a graded ethanol series (absolute ethyl alcohol for 3 min, 95% ethanol for 3 min and 85% ethanol for 3 min), then mounted. Negative control sections were performed under the same conditions, except without primary antibodies.

The immunoreactivity intensity was scored according to four values: 0) Negative staining; i) weak positive staining; ii) moderate positive staining; and iii) strong positive staining. The proportion of stained cells for that intensity over the total number of tumor cells on the slide was recorded in 5% increments from a range of 0–100. The final immunoreactive score was obtained by multiplying the intensity score with the percentage of positively stained cells, and ranged between 0 and 300% and assigned with 5% increments (0, 5, 10, …300%). These scores were used to determine the cutoff value in receiver operating characteristic (ROC) curves to discriminate between tumors of high miR-335 or Rock1 expression and those of low expression. The sensitivity and specificity was plotted on ROC curves to investigate the survival status of patients with osteosarcoma.

All data were analyzed using SPSS 17.0 (SPSS, Inc., Chicago, IL, USA). The Student's t-test was used to compare the difference for data with normal distribution, while Wilcoxon rank-sum tests or Kruskal-Wallis tests were used to compare the difference for data with unequal variance. The association between miR-335 and Rock1 expression and clinicopathological characteristics in patients with osteosarcoma was evaluated using the Pearson's χ² test or Fisher's exact probability test. The association between miR-335 and Rock1 expression was assessed by Spearman's rank correlation analysis. The Kaplan-Meier curves were plotted to demonstrate the survival difference, and the survival probabilities were assessed using a log-rank test. Univariate and multivariate Cox's proportional hazards regression models were used to assess the association of potential confounding variables with the OS or DFS. P≤0.05 was considered to indicate a statistically significant difference.

In situ hybridization and immunohistochemistry analysis were used to confirm the subcellular localizations and the expression patterns of miR-335 and Rock1 protein in 91 osteosarcoma samples and 47 noncancerous bone samples respectively. As demonstrated in Fig. 1, miR-335 and Rock1 immunopositive staining were mainly localized in the cytoplasm. It was also observed that compared to noncancerous bone tissues, the miR-335 expression levels were considerably decreased in osteosarcoma tissues (P<0.001), while Rock1 expression levels increased significantly in osteosarcoma tissues (P<0.001).

ROC curve analysis was applied to determine an optimal cutoff value for the expression of miR-335 and ROCK1 in osteosarcoma samples. Considering the OS status, cutoff values of 115 and 87.5% were selected for the expression of miR-335 and Rock1, respectively (Fig. 2A and B). Tumors in which the immunohistological scores were ≥115 and <115% were defined as those of high and low expression of miR-335, respectively. Tumors with immunohistological scores ≥87.5 and <87.5% were defined as those with high and low expression of Rock1, respectively.

In Table II, it can be observed how miR-335 and Rock1 were expressed in 91 osteosarcoma samples. As shown in the cutoff value used in the present study, 11/91 (12.09%) samples were of high miR-335 and high Rock1 expression, 11/91 (12.09%) were of low miR-335 and low Rock1 expression, 31/91 (34.07%) were of high miR-335 expression, but low Rock1 expression, and 38/91 (41.76%) were of low miR-335 expression but high Rock1 expression. There was a significant negative correlation between miR-335 and Rock1 expression in osteosarcoma tissues (r=−0.378, P<0.001, Table II), according to Spearman's correlation analysis.

As demonstrated in Table III, the low expression of miR-335 was significantly associated with distant metastasis (P=0.016) and grade IIB/III osteosarcoma (P=0.004). The high expression of Rock1 was also significantly associated with distant metastasis (P=0.022), grade IIB/III osteosarcoma (P=0.027) and higher tumor size (P=0.013). Table IV presents the association of combined miR-335 and Rock1 expression with distant metastasis and clinical stage. Out of all comparisons investigated in the present study, tumors with a high expression of Rock1 but low expression of miR-335 were the most strongly associated with distant metastasis (P=0.010) and a higher clinical stage (P=0.010).

Kaplan-Meier analysis and log-rank test were used to evaluate the association of different miR-335 and Rock1 expression level with the OS or DFS in osteosarcoma patients. The data demonstrates that a high expression of Rock1 was significantly associated with shorter OS (P<0.001, Fig. 3A) and DFS (P<0.001, Fig. 3B), and the low expression of miR-335 was also evidently associated with shorter OS (P<0.001, Fig. 3C) and DFS (P<0.001, Fig. 3D). Furthermore, univariate and multivariate Cox's proportional hazards regression models were used to evaluate the association of potential confounding variables with the OS or DFS.

Univariate results identified that gender (P=0.020 for OS), higher tumor size (P=0.008 for DFS; P=0.001 for OS), higher clinical stage (P<0.001 for DFS; P<0.001 for OS), distant metastasis (P=0.006 for DFS; P<0.001 for OS), anatomic location (P<0.001 for DFS), miR-335 low expression (P<0.001 for DFS; P<0.001 for OS) and high Rock1 expression (P<0.001 for DFS; P=0.001 for OS), were associated with the OS and DFS in patients with osteosarcoma (Table V). In addition, multivariate Cox regression analysis identified that higher clinical stage (P=0.002 for DFS; P=0.015 for OS), distant metastasis (P=0.024 for DFS; P=0.002 for OS) and low expression of miR-335 (P<0.001 for DFS; P=0.002 for OS) remained independent prognostic factors for OS in patients with osteosarcoma. Subsequently, the present study combined miR-335 and Rock1 expression into three groups: High expression of miR-335 and Rock1, either high expression of miR-335 or Rock1, and low expression miR-335 and Rock1 to investigate whether the combined expression of miR-335 and Rock1 exhibited an effect on the progression and prognosis of osteosarcoma.

Kaplan-Meier analysis and log-rank tests (Fig. 3E and F) revealed that the OS and DFS of patients with high expression of Rock1 and low expression of miR-335 was considerably shorter compared with those with high or low expression of miR-335 and Rock1, and those with low Rock1 expression and high miR-335 expression (P<0.001 for the two). Results of univariate and multivariate Cox proportional hazards regression analysis are presented in Table VI. The findings demonstrate that combined high Rock1 expression and low miR-335 expression was considered as an independent prognostic factor for shorter OS (P=0.050) and DFS (P=0.021) in patients with osteosarcoma.