[2059] Validation of Prognostic Biomarkers on Prostate Cancer Tissue Microarrays

Andrew D Johnson, Anthony E Rizzardi, Lauren O Marston, Joseph S Koopmeiners, Gregory J Metzger, Colleen L Forster, Rachel I Vogel, James B McCarthy, Eva A Turley, Jessica R Tiffany, Ze'ev Ronai, Christopher A Warlick, Stephen C Schmechel. University of Minnesota, Minneapolis, MN; NCI-Designated Cancer Center, San Diego, CA; University of Western Ontario, London, ON, Canada

Background: Biomarkers that predict disease aggressiveness leading to prostate specific antigen (PSA) failure after prostatectomy may facilitate selection of adjuvant therapies. We assessed the prognostic value of 33 biomarkers using immunohistochemistry (IHC) on tissue microarray (TMA) sections representing PCa tissue from prostatectomies of 153 patients for whom clinicopathologic features and PSA follow-up after surgery were available.
Design: Potential prognostic biomarkers were assessed based on RNA expression profiling, data analysis and literature review. Using a cross-study data analysis we have selected 11 genes (ACPP, ADAM9, ALDH1A2, CASR, CCPG1, GADD45B, HOXC6, IGF1, IQCK, PAGE4 and PLIN2) that best distinguished aggressive (leading to PSA failure after prostatectomy) versus non-aggressive PCa. 4 additional genes (CHMP1A, EI24, MAP4K4, and MKI67) were selected based on published expression analyses. Additional biomarker candidates, based on literature review, included 6 associated with hyaluronan (HA) processing (HA, HAS2, HMMR, HYAL1, CD44, and CD44v6), 7 implicated in prognosis of PCa (CCND1, PTEN, SMAD4, SPP1 [PMID:21289624]; HES6, SIAH2, and SOX9 [PMID:20609350]), 3 neuroendocrine markers (CHGA, NSE, SYP), an angiogenesis marker CD34, and the tumor suppressor p53. Software algorithms (Genie Histology Pattern Recognition and Color Deconvolution, Aperio) were used on whole slide images of stained TMAs to automate tissue annotation into several classes (tumor, stroma, and empty glass) and quantify IHC staining. Hazard ratios associating increased expression with decreased time to biochemical failure were calculated.
Results: Established pathologic parameters were each separately associated with time to PSA failure. In tumor areas, IHC expression of CCND1, HMMR, IGF1, KI67, SIAH2 [nuclear], and SMAD4 were separately associated with shorter time to PSA failure. After adjustment for adverse clinicopathologic features HMMR, HOXC6, IGF1, MAP4K4, and SMAD4 were independently associated with shorter time to PSA failure. Notably, 3-gene signature (HMMR, SIAH2 [nuclear], and SMAD4) had the best n-gene performance with AUC 0.721 for 5-year PSA failure (p=0.002).
Conclusions: We identify that expression of CCND1, HMMR, HOXC6, IGF1, KI67, MAP4K4 and SIAH2 [nuclear] were associated with shorter time to PSA failure. Further, a 3-gene model (HMMR, SIAH2 [nuclear], and SMAD4) appears to serve as a useful signature associated with shorter time to PSA failure.
Category: Techniques

Tuesday, March 5, 2013 2:00 PM

Proffered Papers: Section E, Tuesday Afternoon

 

Close Window