Up-regulation of cell cycle arrest protein BTG2 correlates with increased overall survival in breast cancer, as detected by immunohistochemistry using tissue microarray

The breast cancer samples were obtained from 144 patients undergoing surgical resection at Malmö University Hospital, Malmö, Sweden, between 2001 and 2002. One patient lacked five years follow-up time resulting in the exclusion of this sample from the 5-year survival analysis, although not from the multivariate analysis. The 5-year survival analysis was performed based on overall survival, including 111 samples from alive and 32 samples from dead patients. Further clinical information is compiled in Table 1. Tissue microarrays (TMAs) containing duplicate 1.00 mm cores from each tumour were constructed as previously described [11]. The utilization of the tumour material for research purposes was approved by regional ethical committees in Lund, Sweden.

The expression of ADIPOR1, ADORA1, BTG2 and CD46 proteins was investigated using IHC. Prior to hybridisation to the tissue microarrays, antibodies corresponding to the selected genes were optimised on paraffin-embedded sections of breast tumours. After deparaffinisation in Xylene, the tissue microarrays were autoclaved for at least one hour in buffer S1699 or S2367 (Dako Norden A/S, Denmark) or Borgs Decloaker pH9 buffer solution (Biocare Medical, CA, USA) (Table 2). The immunohistochemical staining was performed in an automated immunostainer (TechMate 228 500 Plus; Dako Norden A/S, Denmark). The TMA sections were incubated with the different antibodies at a dilution of 1:300 for ADIPOR1 (Phoenix Pharmaceuticals, Inc, CA, USA), 1:500 for ADORA1 (Genway Biotech, Inc, CA USA), 1:1000 for BTG2 (Genway Biotech, Inc, CA, USA), and 1:40 for CD46 (BD Biosciences, New Jersey, USA); (Table 2). The antibodies were visualised with the EnVision (K5007, Dako Norden A/S, Denmark) or LSAB (K5001, Dako Norden A/S, Denmark) visualization system according to the manufacturer's instructions (Table 2). EnVision uses a secondary antibody against both rabbit and mouse that is directly labelled with HRP (horseradish peroxidase) reacting with DAB, whereas LSAB uses a secondary antibody against rabbit and mouse, labelled with biotin, then streptavidin-HRP is added and the staining is done with DAB. The TMA sections were then washed in water, dehydrated in an alcohol gradient followed by Xylen treatment, and mounted.

The immunostained tissue microarray sections were analysed by a pathologist (AK). CD46 protein is a cell membrane protein, while ADIPOR1 and ADORA1 are cytoplasm proteins. The subcellular location of BTG2 has varied in previous publications, and in this study, the cytoplasm and cell membrane were stained. The cell membrane staining intensities were graded as no expression (0), low expression (+), moderate expression (++), and high expression (+++). The expression of the cytoplasm proteins was also graded from 0-3 (ranging from no expression to high expression). The proportion of tumour cells expressing membrane and/or cytoplasm protein was determined. In the evaluation of ADIPOR1, ADORA1 and CD46, any level of sample staining was considered positive. For the BTG2 protein, moderate to strong staining of at least 50% of the cells was required for the sample to be scored as positive.

The difference in expression of the four proteins between tumours from alive and dead patients was tested using two-tailed Fisher's exact test. Kaplan-Meier survival curves were produced using the SPSS version 16 software to demonstrate the difference in overall survival between overall BTG2-positive and overall BTG2-negative samples, as well as BTG2 membrane-positive and BTG2 membrane-negative samples. Significant differences between the curves were compared using the Breslow-Wilcoxon test [12]. Additionally, a multivariate analysis (Cox-regression) was performed to evaluate the clinical significance of using current prognostic markers (tumour size, nodal status, HER2 expression, hormone receptor status, histological grade, and patient age) versus using a model containing the proposed markers HER2 expression, patient age, and increasingly used marker Ki67 in combination with BTG2 expression. To cope with the reduced validity of the scientific inference due to misrepresentation we used Multiple Imputation by Chained Equations. The effect of the markers on survival probability was modelled by Proportional Hazards Model. Variable selection was based on a combination bootstrap and information theory approach [13]. As an external validating measure we used time-dependent AUC and Concordance index (C-index) [14]. The time-dependent AUC characterises the temporal changes in predictive accuracy. Concordance index offers an easy way to interpret global accuracy measure that varies between 0.5 and 1. A concordance index of 1 means that with 100% precision we can rank the patient's survival time given the recorded marker information. If the concordance index converges to 0.5 the ranking of survival times becomes more and more driven by chance, and becomes completely random at 0.5. To quantify the impact of a single marker on the predictive accuracy we removed one marker at the time from the final model and refitted a Proportional Hazards Model and re-estimated the C-index. An estimation of the correlation between expression of the different proteins were performed using Pearson correlation.