Lonergan, Peter E.

Objectives To prospectively evaluate the performance of a pre‐specified statistical model based on four kallikrein markers in blood (total prostate‐specific antigen [PSA], free PSA, intact PSA, and human kallikrein‐related peptidase 2), commercially available as the 4Kscore, in predicting Gleason Grade Group (GG) ≥2 prostate cancer at biopsy in an international multicentre study at three academic medical centres, and whether microseminoprotein‐β (MSP) adds predictive value. Patients and Methods A total of 984 men were prospectively enrolled at three academic centres. The primary outcome was GG ≥2 on prostate biopsy. Three pre‐specified statistical models were used: a base model including PSA, age, digital rectal examination and prior negative biopsy; a model that added free PSA to the base model; and the 4Kscore. Results A total of 947 men were included in the final analysis and 273 (29%) had GG ≥2 on prostate biopsy. The base model area under the receiver operating characteristic curve of 0.775 increased to 0.802 with the addition of free PSA, and to 0.824 for the 4Kscore. Adding MSP to the 4Kscore model yielded an increase (0.014–0.019) in discrimination. In decision‐curve analysis of clinical utility, the 4Kscore showed a benefit starting at a 7.5% threshold. Conclusion A prospective multicentre evaluation of a pre‐specified model based on four kallikrein markers (4Kscore) with the addition of MSP improves the predictive discrimination for GG ≥2 prostate cancer on biopsy and could be used to inform biopsy decision‐making.

Abstract Recent studies suggest that progression to castrate-resistant prostate cancer (CRPC) is due to generation of androgen receptor (AR) splice variants, lacking the ligand binding domain. These AR variants originally identified in cell-based models of CRPC, have also been detected in benign and malignant human prostate tissue, with the highest levels observed in late state CRPC. The diversity and complexity of AR variants is well appreciated but not fully understood. Using RT-PCR we assessed the expression of a number of AR variants in cell-based models of both castrate-resistant and androgen-dependent prostate cancer. Several AR variants were expressed in both 22Rv1 and VCaP cells, but not in LNCaP cells. Coding sequences for AR variants were cloned into expression vectors and constitutive transcriptional activity was observed using a luciferase reporter assay. Augmented transcriptional activity of AR variants was observed in cells that express full-length AR, when full-length AR was silenced using shRNA targeting exons 7/8. This observation suggests the constitutive activity of AR variants may be independent of full-length AR. We profiled the AR variant transcriptome in 22Rv1 cells, using the Illumina HumanHT12 v4 whole-genome expression array platform. shRNA targeting exon 1 was used to silence all AR isoforms and shRNA targeting exons 7/8 was used to silence full-length AR. 5,256 genes were modulated by knocking down all AR isoforms and 5,086 genes were modulated by knocking down full-length AR. A total of 3,714 genes were common to both groups, leaving 1542 genes modulated by AR variants and 1372 modulated by full-length AR (p<0.05). The represents a subset of genes differentially regulated by AR variants independently of full length AR. Aggregate function of multiple AR variants may confer a CRPC phenotype independent of full-length AR. Pathway analysis of these genes differentially regulated by AR variants may reveal potential targets for therapeutic intervention in patients who progress despite androgen deprivation or anti-androgen therapy. (Supported by NIH grants CA125747, CA091956 and CA121277, T.J. Martell Foundation and the Mayo Clinic Prostate Cancer SPORE). Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 2924. doi:1538-7445.AM2012-2924