mRNA Expression of Candidate Genes

The NanoString assay allows for gene expression assays on their nCounter system and provides reliable, sensitive, and highly multiplexed detection of mRNA targets. They are designed to measure up to 800 targets in a single reaction, and the platform processes 12 samples simultaneously. ICGI has established this high-throughput system to complement our propriety systems and solutions with basic molecular data.

Although isolation of mRNA is straightforward and follows established protocols, and the use of the nCounter system is uncomplicated, there are a number of important considerations regarding sample inclusion and normalisation of data. A sample to be analysed has to fulfil a set of biological criteria (e.g., the proportion of malignant epithelial cells above threshold), a set of technical criteria related to the experiment (e.g., binding density) and a set of criteria related to the resulting estimates of mRNA molecule counts (e.g., that there must be a signal level significantly above background), briefly described below. The normalisation procedure is carried out on the final set of samples that fulfil the inclusion criteria.

Biological properties of sample:

• Malignant tumour must be present in the sample to be analysed (HE1).
• The proportion of malignant tumour cells must be sufficiently high for the sample to be included in the analysis. The proportion of malignant tumour cells in the RNA sample can be estimated based on the proportion of malignant tumour cells in the annotated region for analysis in HE1, together with the corresponding region in HE2/HE3 (the mean proportion in the two corresponding regions represents an estimate for the RNA sample tumour cell content). A reasonable threshold for malignant tumour cell content may be 50%.
• Establish an optimal amount of RNA to include in the analysis, such that results are robust, and we do not use more material than necessary. The amount of RNA required may be defined as a fixed value of RNA mass (e.g., 100ng) or based on the scaled mass of RNA after Bioanalyzer analysis that compensates for the level of RNA degradation. Samples that have RNA below the required levels can be included if there is sufficient additional material available and/or the sample can be up-concentrated within the loading limits of the NanoString cassette.

Technical properties:

Samples that do not meet quality requirements regarding technical properties may be remeasured if the material and budget allow. The criteria include:

• Field of view (FOV) registration less than 75% (default in nSolver)
• Binding density is outside of 0.1-2.25 range (default in nSolver)
• Linearity of positive controls – R2 value less than 0.95 - (default in nSolver)
• Positive control limit of detection – 0.5fM positive control is < 2 standard deviations above the mean of the negative controls (default in nSolver)
• Scaling of positive control - positive control normalisation factor <0.3 or >3

Sample analysis result properties:

Exclude samples where the Housekeeping-gene to negative control count test fails, i.e., where the mean of q-values for the comparison between Housekeeping genes and background is below a threshold (e.g., 0.01 or 0.001). The q-value compares the count of a Housekeeping gene to the counts of all negative controls; a small q-value indicates that the Housekeeping gene count is significantly larger than the negative control counts.

Normalise mRNA counts:

• Background subtraction – background thresholding: Use mean of all negative control samples ± 2 SD.
• Housekeeping genes: Calculate the geometric mean of all included Housekeeping genes for each sample, calculate the mean of geometric means across all samples and calculate a scaling factor for each sample such that the geometric mean for the sample multiplied by the scaling factor equals the mean of sample-wise geometric means. Multiply each of the genes in the sample by this scaling factor.

The procedure for analysing the results in univariate and multivariate analyses by survival is as follows:

For each patient, mRNA measurement is done for different tumour areas, and these are aggregated by using the lowest, average, and highest value. Each aggregated gene expression value is analysed as 1) continuous value on a linear scale, 2) continuous value on a logarithmic scale, 3) categorised into four groups after quartiles, 4) categorised into two groups, first quartile against the rest, 5) categorised into two groups, first half against last half and 6) categorised into two groups, the three first quartiles against the last.

Candidate genes in Prostate Cancer

We have expanded our list of 71 candidate genes selected for protein expression to 255 genes that are either common in prostate cancer, related to genomic instability, gained or lost during cancer development, involved in the mitotic checkpoint or known as prostate stem cell markers. We have isolated mRNA from three different tumour samples (tissue blocks) from each of the 255 prostate cancer patients who have undergone surgery (radical prostatectomy) at Oslo University Hospital between 1987 and 2005. The patients have been followed for at least ten years after surgery or until death. Using the NanoString technology, we will quantify the mRNA expression of the 255 candidate genes in the 765 tumour samples. This project aims to gain knowledge about the prostate tumour heterogeneity at the mRNA level, identify genes of particular importance in prostate cancer and contribute to new independent prognostic markers. The data will also be critical in our project on clinical decision support (see Clinical > Decision support on page 65). Each gene will be analysed for its ability to assess time to recurrence in univariate, and in multivariate analyses together with the following variables:

• Age at surgery 1-year increment
• Preoperative PSA (< 6 ng/ml, > 6 ng/ml and < 10 ng/ml, > 10 ng/ml and < 20 ng/ml, > 20 ng/ml)
• Gleason grade (<7, 3+4, 4+3, >7)
• Surgical margins
• Extracapsular extension
• Seminal vesicle invasion
• Pathologic node (N) stage

Preliminary results indicate that mRNA measurements of 23 are significant in both uni- and multivariate analysis, and these 23 genes will be further evaluated as potential biomarkers of prognosis. We hope to get a further understanding of the mechanism behind each gene’s relation to prostate cancer by studying their interactions and functions, and the data will be used to assess tumour heterogeneity. We will validate all findings of independent and combined prognostic markers in a more recent cohort of 284 patients (852 tumour tissue samples) with prostate cancer, which have undergone surgery at Oslo University Hospital between 2000 and 2006.

Candidate genes in Colorectal carcinomas

Having the same goals and using the same principles and procedures as described above for prostate cancer, we will analyse the mRNA expression of 146 candidate genes in 1220 tumour samples from 256 patients with stage II colon cancer treated in the Gloucester colorectal cancer study during the period of 1988-1996. All findings of independent and combined prognostic markers will be validated using the QUASAR2 trial.