Normalization was performed using Fragments per Kilobase per Million, and

isoform expression values were generated using Cufflinks with Ensembl version 69 as the reference transcriptome [37]. Cufflinks calculates isoform expression levels using a statistical model in which the probability of observing a given fragment is a linear function of the transcript abundance. Gene level AZD0530 expression is the sum of transcript level expression, as each read is assigned to a single transcript. Tophat was chosen because it is the standard sequence aligner used by Cufflinks [38]. Correlation coefficients were generated using Spearman’s correlation. Hierarchical clustering was performed on the covariance matrices to generate heat maps. Expression levels of the isoforms and at the gene level were compared across clinical and pathologic groups such as cancer versus normal, tumor stage, histology, hormone receptor status, and PAM50 cluster [39]. Means selleck screening library between groups were compared using analysis of variance. Expression was divided into high versus low expression using the median expression value. Kaplan-Meier curves were generated for the high and low expression groups and compared using the log-rank test for metastasis-free survival (MFS), recurrence-free survival (RFS), and overall survival

(OS). Hazard ratios (HRs) were generated using univariate Cox regression. Multi-gene analysis was performed using Cox regression with expression of each gene/isoform as a covariate. Comparison of expression between metastatic versus non-metastatic cell lines was performed using Student’s t-test. Statistics

and plots were generated using the R statistical computing software and GraphPad Prism. Studies of isoforms of CXCL12 in cancer and other diseases have been limited by the lack of isoform-specific probes on microarrays and antibodies for IHC. As a result, studies have focused predominantly on only the α and β isoforms of CXCL12. To overcome limitations of microarrays and antibodies, we investigated expression levels FAD of all isoforms of CXCL12 and receptors CXCR4 and CXCR7 in breast cancer using the TCGA RNA sequencing data set. The clinical and pathologic characteristics of the tumor samples and patients in this data set are shown in Table 1. The Cufflinks analysis program assigns each read to individual isoforms such that the sum of expression levels for a specific isoform is equal to the gene level of expression. On the basis of this analysis, we determined that the most common isoform of CXCL12 in breast cancer is α (65%), followed by β (27%) > γ (5%) > δ (2%). We detected only very low levels of expression for CXCL12-ε (0.1%) and -φ (0.2%) and therefore refrained from statistical inference using these isoforms.