Blood samples were collected from 6 healthy volunteers and from 28 patients with CRC who underwent surgical resection between April 2006 and March 2009 at Tsukuba University Hospital (Tsukuba, Japan). Blood sampling was periodically conducted up until 12 weeks after surgical resection. None of the patients had received radiotherapy or chemotherapy prior to surgery. The primary clinical characteristics for each patient with CRC involved in the current study are presented in Table I. Informed consent was obtained from all patients for the collection of blood samples and the ethics committee of Tsukuba University Hospital approved the study protocol.

Blood samples were collected from the patients using PAXgene Vacutainer tubes (BD Biosciences, Franklin Lakes, NJ, USA) and subjected to RNA isolation and extraction using a PAXgene Blood RNA Isolation kit (Qiagen, Inc., Valencia, CA, USA), following the manufacturer's instructions. The quantity of total RNA obtained was determined using a NanoDrop Spectrophotometer (NanoDrop Technologies; Thermo Fisher Scientific, Inc., Waltham, MA, USA) according to the 280/260 nm absorbance ratio, and RNA integrity was evaluated using an RNA 6000 Nano LabChip kit (Agilent Technologies, Inc., Santa Clara, CA, USA) and an Agilent 2100 Bioanalyzer (Agilent Technologies, Inc.).

Cyanine 3 (Cy3)-labeled cDNA was synthesized from 10 µg total RNA extracted from the blood samples using a LabelStar Array kit (Qiagen, Inc.) with Cy3-dUTP (GE Healthcare Bio-Sciences, Pittsburgh, PA, USA) and random nonamer primers. The labeled cDNA was hybridized with probe sequences on an Agilent 44 Kx4 human sense and antisense custom microarray slide (Agilent Technologies, Inc.) (16) in a hybridization solution prepared with an In Situ Hybridization Plus kit (Agilent Technologies, Inc.), according to the manufacturer's instructions. The Cy3 fluorescence signals were imaged using an Agilent C DNA microarray scanner (Agilent Technologies, Inc.) and processed using the Feature Extraction version 8.1 software (Agilent Technologies, Inc.).

The microarray data were processed using the GeneSpring GX version 12 software (Agilent Technologies, Inc.) to perform the log transformations and the normalization of all values to the 75th percentile of the respective microarray, followed by normalization to the respective median expression level for all samples. Additionally, the normalized gene expression data were filtered on flags following the protocol of the manufacturer GeneSpring GX (http://www.chem.agilent.com/cag/bsp/products/gsgx/manuals/GeneSpring-manual.pdf). Only those genes classified as flag-present or flag-marginal in >70% of all the array samples, were allowed to pass the filter. The expression profiles of the RNA samples were analysed, using GeneSpring GX version 12 software, with unpaired t-tests, with Benjamini-Hochberg false discovery rate correction (17) for unequal variances, as described in the results section. P<0.05 was considered to indicate a statistically significant difference.

Two-dimensional hierarchical clustering based on Euclidean distance measures was performed using Ward's method (18). The data were visualized using heat maps and dendrograms, as described previously (14). Sample trees were drawn horizontally and gene trees were drawn vertically. Principal component analysis (PCA) was used to identify and characterize trends in multigene expression profiles.

The quality of the RNAs obtained from the blood samples of the healthy volunteers and patients with CRC were initially examined using a NanoDrop Spectrophotometer at an absorbance ratio of 280/260 nm, revealing that the ratios were between 1.8 and 2.0. These results indicated that the total RNAs prepared were usable for labeling with Cy3-dUTP and subsequent microarray analysis. The integrity of the RNAs was then examined using an Agilent 2100 Bioanalyzer, revealing that the Rin values of the RNA samples ranged from 5.9 to 9.2. As the random priming method was utilized for the synthesis of Cy3-labeled cDNAs from the RNA samples, the lower Rin values were considered not to affect the quality of the cDNAs for analysis.

Cy3-labeled cDNAs were synthesized from the extracted RNA and subjected to microarray analysis. A total of 40 transcripts were identified as differentially expressed with a magnitude of >2-fold (P<0.05) between the CRC blood cells and non-cancerous blood cells; of these transcripts, 20 were sense sequences and 20 were antisense sequences (Table II). According to the cluster analysis, CRC and non-cancerous blood samples were revealed to form separate clusters for the antisense transcripts, but not to form separate clusters for the sense transcripts (Fig. 1; the clustering data for the sense transcripts are not presented).

As presented in Fig. 2, PCA analysis revealed that the CRC and non-cancerous blood samples were well separated with the first principal component; its contribution rate was ~43.5%. The contribution rates of the second and third principal components were calculated to be ~7.7% and ~6.4%, respectively. These results indicate that the CRC and non-cancerous blood samples were effectively separated with only the first component.

The results of the microarray with RNA samples from patients with stage I and II CRC and controls revealed a total of 72 transcripts that were significantly differentially expressed with a magnitude of >2-fold (P<0.01) between the blood cells of healthy volunteers and of patients with CRC. A total of 33 transcripts were sense sequences and 39 were antisense sequences (Table III). When the blood samples from patients with stage I or II CRC and healthy volunteers were subjected to cluster analysis for the aforementioned sense and antisense transcripts, the patients were observed to form respective clusters only with antisense transcripts (Fig. 3; the clustering data for the sense transcripts are not presented).

The 39 antisense transcripts were further examined using PCA analysis. As presented in Fig. 4, CRC and non-cancerous RNA samples were revealed to be well separated with regard to the first principal component; its contribution rate was ~43.5%. The contribution rates of the second and third principal components were calculated to be ~7.9% and ~7.0%, respectively. These results indicate that the patients with stage I or II CRC were effectively separated from healthy volunteers with the first component.

The antisense transcript of haloacid dehalogenase-like hydrolase 1 domain-containing (HDHD1), which was the highest and third highest differentially expressed transcript in the earlier analyses (Tables II and III, respectively), was selected for evaluation at various time points following the surgical resection of tumor tissues from patients with CRC. The expression levels of the HDHD1 antisense transcript at 1 week and at 3 months post-surgery were decreased approximately to the expression levels observed in healthy volunteers (Fig. 5). The results suggest that certain antisense transcripts, including HDHD1, may serve as potential biomarkers of CRC diagnosis and prognosis.