Since the introduction of gene expression microarrays in the mid-1990s, genome-wide expression profiling has been widely utilized in cancer research 
. Gene expression profiling experiments have led to significant advances in our understanding of a wide range of human malignancies, but clinical research efforts have been frustrated by lack of specimens. A major hindrance to the translation of gene expression profiling to the clinic is the fact that gene expression microarrays are best performed on fresh frozen tissue, and few samples are stored as fresh frozen. In contrast, essentially all tumor specimens are stored as FFPET 
. This fixation and storage technique results in extensive RNA fragmentation 
. Several groups have attempted to use FFPET for gene expression profiling by microarrays 
with mixed results. Difficulties of gene expression profiling by microarray on FFPET include the inability to obtain adequate RNA for gene microarray profiling from most archival samples 
and a lack of sensitivity for identifying genes known to be expressed from frozen tissue in matched FFPET samples 
Due to the difficulty of gene expression profiling from FFPET using microarrays, several groups have utilized RT-PCR to quantify the expression of targeted sets of genes 
. Recently, Hoshida et al.
performed DASL for gene expression profiling from FFPET and identified a gene expression pattern that correlates with survival in hepatocellular carcinoma 
; however, this study does not demonstrate the ability for full unbiased genome-wide expression profiling from FFPET, as the analysis was limited to only 6,100 genes targeted by the probe set.
3SEQ is sequencing-based and overcomes the major limitations of hybridization-based expression arrays, allowing more sensitive and precise quantitative measurements for genome-wide expression profiling. The primary feature that differentiates the 3SEQ protocol from standard RNA-Seq protocols is that while standard RNA-Seq typically generates a non-directional sequencing library comprised of fragments of RNA that span the transcript length, the 3SEQ protocol generates a directional sequencing library comprised predominantly of ~200 bp cDNA fragments with a poly-A tail, in which sequencing will proceed directionally toward the poly-A tail. This novel design is essential for performing accurate transcriptional profiling from severely degraded RNA, because it ensures that one read per transcript molecule is produced, regardless of degradation and regardless of transcript length.
Compared to other existing gene expression profiling methods, the 3SEQ method facilitates gene expression profiling of degraded RNA from FFPET. In the current study, we identified similar numbers of differentially expressed genes between DTF and SFT on frozen tissue (~9.6K) and FFPET (~8.1K) by 3SEQ. In contrast, we identified fewer differentially expressed genes on frozen tissue (~4.6K) and far fewer genes on FFPET (only 69 genes) by HEEBO microarray. These data clearly indicate that, in contrast to HEEBO microarray, 3SEQ is effective for genome-wide gene expression profiling on both frozen tissue and FFPET, with similar performance on the two tissue types.
Functional gene set analysis of 3SEQ data from frozen tissue and FFPET revealed two key sets of pathways (Wnt signalling-related pathways, and extracellular matrix-related pathways) to be enriched in DTF. HEEBO analysis identified the extracellular matrix-related pathways, but failed to identify significant enrichment in the Wnt signalling pathways, which are known to play a critical role in DTF 
. In addition, functional gene set analysis of genes with relatively increased expression in SFT by 3SEQ identified insulin signalling as a significantly enriched pathway and recent studies suggest that insulin receptor activation is frequently seen in SFT 
. Analysis of 3SEQ data from both frozen tissue and FFPET revealed several related cancer-associated pathways (prostate cancer, VEGF signalling, acute myeloid leukemia) containing a number of genes known to be important in carcinogenesis (AKT2, IGF1, BAD, PIK3R1, CCND1, PML, RARA), whose concerted role in SFT has not previously been characterized. There were no pathways identified by HEEBO-FFPET as enriched in SFT. These findings further suggest that 3SEQ is superior to HEEBO microarray for gene expression profiling from FFPET.
An additional advantage of 3SEQ data is that it can be utilized to identify genes expressed in at least low or moderate levels in one of the tumor types with virtually no expression in the other tumor type. This type of quantitative analysis, which is very difficult to perform on microarray data due to background noise and hybridization artefacts, may facilitate a deeper understanding of differential pathways of tumor pathogenesis and the identification of highly specific diagnostic markers.
These findings have significant implications for translational cancer research. A major goal of translational cancer research is to identify diagnostic markers, prognostic markers, and markers to predict response to treatment. Each of these goals requires the acquisition of large numbers of well-annotated clinical specimens with long term follow-up 
. The lack of adequate samples with detailed clinical information (such as drug response) has been a major impediment to the translation of gene expression profiling findings to the clinic 
. We believe 3SEQ could revolutionize the field of translational cancer genomics, by allowing investigators to perform gene expression profiling on large numbers of well-annotated archival tumor specimens with long term follow-up. Experiments could then be designed to identify gene expression signatures to predict specific clinically important phenotypes (such as drug response, progression/recurrence risk, and survival) and to gain a deeper understanding of cancer biology.