Specific cell populations were captured based on primary Gleason pattern, and their gene expression profiles were analyzed. In general, unsupervised clustering grouped cases in accordance with Gleason pattern. However, the grouping was not perfect in this regard since two of seven Gleason pattern 3 cases clustered with Gleason pattern 5 cases and lymph node metastases. The molecular profile may provide more detailed information than the Gleason pattern, indicating that the Gleason 3 cases are more aggressive, or have the potential to be more aggressive, than expected based on histology alone.
In agreement with previous results [28–35] we found that the type II transmembrane serine protease hepsin is approximately 5-fold overexpressed in prostate cancer cells compared to benign prostatic epithelial cells. Furthermore, metastatic cells and Gleason pattern 5 cells, which grouped together in unsupervised clustering, had hepsin expression levels intermediate between Gleason pattern 3 and benign cells. Hepsin immunohistochemistry has indicated that the staining intensity of hormone-refractory metastatic cancers is intermediate between localized prostate tumors (Gleason score 6–8) and benign prostate . The lymph node metastasis cases we analyzed were not hormone-refractory, indicating that hepsin expression may not be regulated by hormones. Also in agreement with previously results [29, 31, 35–37] we observed up-regulation of AMACR expression in prostate cancer relative to benign cells (about 5-fold). The expression pattern of AMACR was similar to that of hepsin in that AMACR expression in high grade, Gleason pattern 5 and metastatic cancer, cells was intermediate between low grade, Gleason pattern 3, and benign cells. Interestingly, AMACR expression has been reported to be lower in hormone-refractory metastatic prostate cancer than hormone-naïve-localized prostate cancer , suggesting that AMACR protein expression might be regulated by androgens. A subsequent report  suggests, however, that AMACR expression is not hormone-dependent, but that it may be a marker of tumor differentiation. This is consistent with the data presented here showing decreased AMACR expression in Gleason pattern 5 cells and non-hormone-refractory lymph node metastases relative to Gleason pattern 3 cells. Interestingly, in a recently published gene expression study of laser microbeam microdissected populations of prostate cancer, prostatic intraepithelial neoplasia (PIN), and normal prostatic epithelial cells  hepsin was not found to be differentially expressed between normal epithelium and PINs and prostate tumors, but AMACR was found to be up-regulated.
Recommendations of the Working Group  for fold amplification during the first and second rounds of linear amplification may not be based on experiments where minimum amounts of input RNA were amplified (less than 200 ng). Therefore, the relevance of these recommendations to such experiments is unclear. The ability to successfully amplify small amounts of total RNA is becoming increasingly important. LCM is labor-intensive, especially given the need for significant numbers of biological replicates when analyzing human specimens. Minimizing the requirements for input RNA can have a significant impact on study feasibility. In addition, interest in analyzing individual cells continues to grow .
Standards for the size of biotinylated cRNA obtained from two cycle amplification protocols are needed. For one cycle experiments, the Working Group recommends that the biotinylated cRNA should be 500–3,000 base pairs in size, and that samples that do not meet these criteria should be discarded . However, two cycle amplification protocols are biased to the 3' ends of transcripts.
We developed qPCR assays to assess RNA quality and quantity. Our qPCR assay for RNA quality is analogous to the 3'/5' and 3'/M ratios for GAPDH and β-actin that are quality control parameters for Affymetrix microarrays . Arcturus has also provided a qPCR protocol to assess RNA quality in formalin fixed paraffin embedded tissues . Quantitative PCR assays are fairly reproducible [see Additional file 11] and extremely sensitive, requiring significantly less sample than the NanoDrop® ND-1000 spectrophotometer and the Agilent 2100 bioanalyzer for analysis of RNA quality and quantity. In addition, qPCR is a more direct, functional assay for mRNA quality than analysis of ribosomal RNA as a surrogate for mRNA using the Agilent bioanalyzer. It is also noteworthy that measurements of absorbance at 260 nm can be misleading because degraded RNA will also contribute to the absorbance, which could give an erroneously high estimate of intact RNA concentration.
Using the U133 Plus 2.0 arrays, the reliability of the signal was observed to decrease dramatically with decreased intensity [see Additional file 6]. It may be possible to increase the sensitivity by hybridizing larger amounts of labeled targets to the arrays. As expected, correlations between hybridization replicates were significantly better than correlations between replicates of the entire protocol performed on different days. The correlation between biological replicates would be expected to be even less.
We compared the IVT and ENZO kits using signal correlation plots. However, it is difficult to determine which kit is superior in the absence of a comprehensive set of known differences in gene expression. A better comparison could be achieved by performing spike-in experiments [25, 26] or comprehensive qPCR confirmation of differentially expressed genes.