One advantage of the BioMark arrays is the capability to analyse a large number of genes in a single sample. In order to facilitate this, up to ~ 2 μL of sample is loaded into each sample inlet of the chip and further distributed in the channels of the microfluidic chip as 48 separate 9 nL reactions using the integrated fluidic circuit (IFC). Thus the original sample is diluted more than 200-fold prior to the PCR reaction. In order to ensure that there are sufficient copies of target molecules in each reaction, Fluidigm^® recommends using either RNA samples that do not have a concentration lower than 250 ng total RNA/μL or that a pre-amplification stage is included, whereby the cDNA sample undergoes 14-18 cycles of amplification with a mix of up to 100 different primer pairs (Fluidigm Advanced Development Protocols 3, 5 and 8). In order to further investigate the requirement for pre-amplification, RNA standards were spiked into human total RNA at different concentrations (for sample composition, see Additional File 2) with the aim of mimicking a range of physiological abundances, from highly abundant mRNA transcripts (10⁶ copies/ng total RNA; equivalent to 10⁴ copies per cell) to transcripts only expressed in a sub-population of cells (1 copy/ng total RNA; equivalent to 0.01 copies per cell), based on the RNA content of a cell estimated as 26 pg 12 . A single RT reaction was performed for each RNA sample followed by 3 independent qPCR runs, with replicate assay measurements for each ERCC standard.

Figure 1 compares the results of real-time PCR with cDNA samples (equivalent to 1 ng total RNA) or pre-amplified cDNA on the BioMark arrays with non-amplified cDNA using the ABI 7900HT system. The linear range in terms of transcript copy numbers for the pre-amplified cDNA samples on the BioMark arrays was similar to the non-amplified cDNA samples on the ABI 7900HT system, covering five orders of magnitude between 10 and 10⁶ copies/ng total RNA. Although the linear range of both platforms was similar, the Ct values from BioMark were over 10 units lower than those observed for ABI 7900HT system. This could be due to the higher concentration of the template in the 9 nL nanofluidic reaction chambers of BioMark arrays compared to the standard 20 μL volume used on the ABI 7900HT system, such that the fluorescence output of the PCR reaction exceeds the threshold level at an earlier cycle (personal communication: A. Meliss, Fluidigm, September 2009). For the RT-PCRs performed with non-amplified cDNA samples on BioMark arrays, the linear detection range covered only two orders of magnitude between the transcript numbers of 10⁴ and 10⁶ copies/ng.

Unlike hybridization-based technologies such as microarrays, the LOD cannot be ascribed for RT-qPCR using a baseline for sample blanks, as a Ct value is not obtained for zero control samples. Therefore, the incidence of failed PCR reactions (undetermined Ct value) across the range of transcript abundances was also compared for conventional and nanofluidic PCR platforms with pre- or non-amplified cDNA as the template (Figure 2). As in Figure 1, a high percentage of PCR failures was observed at 1 copy/ng total RNA for both pre-amplified cDNA on the BioMark arrays and cDNA using the ABI 7900HT system. The rate of failures was slightly lower on the BioMark arrays as the projected template concentration per 9 nL reaction was 6 copies (assuming 100% efficiency of RT and pre-amplification), whilst it was only 1 RNA copy per ABI 7900HT reaction. For non-amplified cDNA, high reaction failure rates with BioMark arrays were observed below 10³ copies/ng, which equates to 2 copies per 9 nL reaction.

The accuracy and precision of qPCR detection across the above noted linear range was assessed by linear regression, comparing the slope and R² of the data from three independent dynamic arrays (Table 1). Pre-amplification of cDNA samples resulted in a significant improvement in the slope of the linear regression of copy number against Ct value (p < 0.05), with the mean slope within 6% of the ideal slope of 1. The pre-amplified cDNA samples demonstrated greater precision of the instrument over the linear detection range (Figure 1) than cDNA (p < 0.05) (Table 1). The accuracy and precision of pre-amplified cDNA detection on the BioMark arrays were comparable to those of the ABI 7900HT where non-amplified cDNA was used as the template (Table 1).

Precision of qPCR detection as a function of transcript copy number for the two platforms is shown in Figure 3. For concentrations of RNA standards above 10⁴ copies/ng (using cDNA with the ABI 7900HT system or pre-amplified cDNA for the BioMark arrays), and above 10⁶ copies/ng (with non-amplified cDNA for the BioMark arrays), Ct standard deviation values are below 0.1 units, corresponding to less than 7% variation 13 . As RNA copy numbers decrease, variation between replicate qPCR measurements increases, with maximum average standard deviation values corresponding to 46% and 66% variation at 10 copies/ng for the BioMark arrays (pre-amplified cDNA) and ABI 7900HT system respectively (Figure 3).

The accuracy of nanofluidic dynamic PCR for detection of multiple genetic biomarkers was further tested using two panels of RNA standards. With the aim of mimicking 'normal' and 'disease' states where some biomarkers are differentially expressed whilst others remain unchanged in their expression, standards were spiked at different ratios (1.0, 1.5, 2.0, 5.0, 10.0 and 20-fold differences) over a range of transcript copy numbers (Table 2). Three independent RT experiments, each containing two replicate RT reactions, were performed in order to investigate how technical noise associated with the whole RT-qPCR process impacts on the detection of differential or non-differential transcript expression levels. The resulting cDNA was quantified on the ABI 7900HT system or pre-amplified and measured on the BioMark. Fold change values were calculated using ΔCt values and the results of pair-wise comparison of the expression levels of each ERCC standard in the two panels displayed in Figure 4.

Overall, fold change estimation was found to be accurate for both ABI 7900HT system and BioMark arrays (Figures 4A and 4B), with the observed fold change values overlapping with the expected fold change measurements for all standards. For ERCC standards mimicking low abundance transcripts, the technical noise associated with the resulting fold change measurement was considerably greater than higher abundance RNA species. For example, a 20-fold increase in expression level at 5 copies/ng (ERCC-51) is associated with 16-fold and 10-fold difference in the minimum and maximum fold changes detected by the BioMark arrays and ABI 7900HT system respectively. For non-differential expression at low copy numbers (fold change = 1.0; ERCC-113), fold change measurements spanned a range of over 50% and 150% of the expected value on the BioMark and ABI 7900HT system respectively. At levels of abundance exceeding 100 copies/ng, mean fold change measurements were accurate to within 10% of expected values.

In vitro transcribed ERCC RNA standards were produced from ERCC plasmid DNA (courtesy of Dr. Marc Salit, NIST, USA). Plasmid DNA from standards ERCC-13, 25, 42, 51, 81, 84, 95, 99, 113 and 171 was cleaved into a single linear molecule using BamHI restriction endonuclease (New England Biolabs, UK). 500 ng of plasmid DNA was used for each sample and digested by adding 40 U of BamHI enzyme in NEB3 buffer provided by the manufacturer. The digestion mixture was incubated at 37°C for 2 hours followed by purification using QiaQuick PCR purification kit with an elution volume of 32 μl. In vitro transcription was carried out with 8 μl digested plasmid DNA using MEGAscript^® T7 Kit (Applied Biosystems/Ambion, UK) followed by DNase treatment and clean-up using RNeasy columns (Qiagen, UK). RNA concentration and insert sizes were estimated using the Nanodrop 1000 spectrophotometer (Thermo Scientific, UK) and 2100 Bioanalyzer (Agilent Technologies, USA) respectively. RNA standards were diluted in nuclease free-water and spiked into Universal Human Reference RNA (Stratagene, UK) (final concentration 100 ng/μl). For experiments investigating the linear range of platform detection, standards were spiked at 10-fold intervals between 1 and 10⁶ copies/ng total RNA (Additional File 2). For the simulated 'normal' and 'disease' panels, standards were spiked at various copy numbers and ratios (Table 2).

RNA samples were reverse-transcribed using the TaqMan^® Reverse Transcription Reagents kit (Applied Biosystems, UK) in 40 μL reactions containing 400 ng total RNA and oligo(dT) primers according to manufacturer's instructions. cDNA samples were diluted to a concentration of 0.5 ng/μL (total RNA equivalent) with nuclease-free water. For experiments investigating the linear range of platform detection (Figures 1, 2, 3), a single RT reaction was performed for each RNA sample whilst for the simulated 'normal' and 'disease' panels (Figure 4), 6 replicate RT reactions were performed. A single aliquot of each cDNA sample, equivalent to 12.5 ng RNA, was pre-amplified with assays corresponding to all 10 standards in a 25 μL volume reaction using TaqMan^® PreAmp Mastermix (Applied Biosystems, UK) according to manufacturer's protocol. Following pre-amplification, the samples were diluted 1:5 (v/v) in TE buffer, pH 8.0.

Further information on sample preparation and real-time PCR validation complying with the Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines 22 is available in Additional Files 2 and 3 (MIQE Additional Information and Checklist).

Custom-designed primers and TaqMan^® FAM-TAMRA probes for each ERCC standard (Additional File 1) were supplied by Applied Biosystems and a 20 × assay mix was prepared containing 18 μM primer and 5 μM probe (final concentration 900 nM primer and 250 nM probe). qPCR assays were tested initially using a serial dilution of ERCC cDNA and PCR efficiencies calculated (see Additional Data Table 2: MIQE Additional information). All 10 assays were found to have PCR efficiencies of greater than 86%.

BioMark arrays were prepared according to the manufacturer's instructions. TaqMan^® assays were diluted 1:1 (v/v) with DA Assay Loading Reagent (Fluidigm^®) and 5 μL was added to each assay inlet of the array. Also, 5 μL reaction mix was prepared by mixing 2 × TaqMan^® Universal Mastermix (Applied Biosystems), DA Sample Loading Reagent and nuclease-free water containing 2 μL of cDNA or pre-amplified cDNA. The samples were loaded into each sample inlet as per manufacturer's recommendations. Following loading of the assays and samples into the chip by the IFC controller, PCR was performed with the following reactions conditions: 50°C for 2 minutes, 95°C for 10 minutes, followed by 45 cycles of 95°C for 15 seconds and 60°C for 60 seconds. Data was processed by automatic global threshold setting with the same threshold value for all assays and linear baseline correction using BioMark Real-time PCR Analysis software (version 2.1.1). The quality threshold was set at the default setting of 0.65. For experiments investigating the linear range of platform detection (Figures 1, 2, 3), 8 qPCR reactions consisting of 4 assay inlet and 2 sample inlet replicates were performed for each cDNA or pre-amplified cDNA sample. For the simulated 'normal' and 'disease' panels (Figure 4), 12 qPCR reactions consisting of 4 assay inlet and 3 sample inlet replicates were performed for each cDNA sample.

Conventional real-time PCR was performed using ABI 7900HT system in 20 μL reaction volumes containing TaqMan^® Universal PCR Master Mix and 2 μL of respective cDNA in optical 96-well plates (Applied Biosystems). Cycling conditions were as those used for the BioMark arrays. Triplicate qPCR reactions were performed for each cDNA sample for all experiments. The threshold fluorescence level was set manually for each plate using SDS software version 2.3 (Applied Biosystems). Following export of Cycle threshold (Ct) data, further data analysis for both platforms was performed in Microsoft^® Excel 2003. Comparison of slope and R² values between pre-amplified and non-amplified cDNA, as a template on the BioMark arrays, was performed as paired t-test in Microsoft^® Excel 2003.