Ethics statement

The study protocol was approved by the Riverside Research Ethics Committee (Imperial College Healthcare NHS Trust; REC reference number: 07/Q0401/20) and the Leicestershire, Northamptonshire & Rutland Research Ethics Committees (REC reference number: 05/Q2502/28; 11/EM0023). Blood sample collection was conducted in accordance with the Declaration of Helsinki. All patients gave written informed consent prior to participation.

Blood processing

20 ml venous blood samples were drawn in the morning into EDTA-containing tubes (BD Biosciences) and centrifuged at 1000g (Jouan CR422; Jouan) for 10 min at 4 °C. Plasma was carefully removed, leaving 3-5 mm above the buffy coat, added to a 15 ml Falcon tube (BD Biosciences) and re-centrifuged at 1000g for 10 min at 4 °C. Plasma was carefully removed and stored at -80 °C in 1 ml aliquots in 1.5 ml tubes (Starlab) for cfDNA, and in 1.5 ml RNase-free tubes (Starlab) for miRNA. We also varied the speed of the second centrifugation step: 1000g, 2000g (Jouan CR422; Jouan) or 10000g (Microcentaur; MSE) each for 10 min at 4 °C, as well as increasing the time between venepuncture and the first centrifugation step (at 2 hr and at 6 hr). On thawing a third and final spin was performed at 1000g for 5 min at room temperature and plasma carefully removed into a separate tube away from any residual debris before extraction of CNAs.

Extraction of circulating nucleic acids, spike-in templates and QC checks

We purchased 3 plasma samples from healthy females (age range 35 - 39 years) (Seralab) and pooled an equal volume to generate a reference healthy plasma sample. For cfDNA we compared isolation from 1 ml starting plasma volume, with each sample/kit performed in triplicate, using the QIAamp^® DNA Blood Mini Kit, QIAamp^® Circulating Nucleic Acid Kit, NucleoSpin^® Plasma XS Kit (Macherey-Nagel) and FitAmp^® Plasma/Serum DNA Isolation kit (Epigentek). Each extraction was performed exactly according to manufacturers’ instructions and as described previously [12,13]. With each sample/kit the elution volume was 1/10th of the starting plasma volume. As additional validation we spiked λ/HindIII DNA (Ambion) into 1 ml of the pooled healthy plasma at 4 fixed concentrations (50 ng, 5 ng, 0.5 ng and 0.05 ng) and assessed recovery of two fragments by real-time qPCR.

Circulating miRNAs were extracted from 1 ml, 500 µl or 200 µl of plasma. As miRNA was not recovered from the pooled healthy plasma (Seralab) we isolated miRNA from pooled plasmas from 3 patients with either primary or metastatic breast cancer. Each extraction was performed in triplicate, using the miRNeasy Serum/Plasma Kit (Qiagen), miRvana microRNA Isolation kit (Ambion) or QIAamp^® Circulating Nucleic Acid Kit (Qiagen), according to the manufacturers’ instructions. 30 pg or 30 ng of the synthetic Arabidopsis thaliana miR-159a (ath-miR-159a; 5’-UUUGGAUUGAAGGGAGCUCUA-3’) (Integrated DNA Technologies) was spiked-in to each sample after the initial lysis step to test extraction efficiency. All samples were eluted in 50 µl of RNase-free water.

Initial QC checks of eluted cfDNA and RNA (1 µl each) were performed using the Qubit® 2.0 dsDNA high sensitivity assay (Life Technologies) and Agilent Bioanalyser 2100 analysis with RNA 6000 Pico reagents and chips (Agilent), respectively, according to manufacturers instructions..

Reverse transcription and analysis of miRNA profiles

MicroRNAs were reverse transcribed from 3 µl of sample using the TaqMan Megaplex RT system (Applied Biosystems), and pre-amplified from 2.5 µl of cDNA using the TaqMan Megaplex Pre-Amp system (Applied Biosystems) according to manufacturer’s instructions on a Veriti thermal cycler (Applied Biosystems). Pre-amplified cDNA was diluted 1:25 prior to use and compared with unamplified cDNA, diluted 1:5 in 1xTE buffer. A RT-negative control was included in each experiment to assess genomic DNA contamination, all of which showed no amplification upon analysis by qPCR for hsa-miR-21. For validation of the TLDA cards, pre-amplified cDNA was diluted 1:100.

For miRNA PCR arrays, TLDA v2.0 cards were used according to manufacturer’s instructions on an Applied Biosystems 7900 thermal cycler as follows: 50 °C for 2 min, 94.5 °C for 10 min, followed by 40 cycles of 97 °C for 30 sec and 59.7 °C for 1 min. The cycle threshold (C_T) values were determined using a fixed threshold of 0.1 for all qPCR experiments. Exiqon miRCURY LNA Serum/Plasma focus panels were analysed on a Roche Lightcycler 480 thermal cycler as follows: 95 °C for 10 min, followed by 45 cycles of 95 °C for 10 sec and 60 °C for 1 min with a ramp rate of 1.6 °C/sec.

MiRNA samples were analysed in triplicate using 3 µl of diluted template and TaqMan microRNA assays (hsa-miR-191 (assay code 000490), hsa-miR-21 (000397), ath-miR-159a (000338), hsa-miR-15b (000390), hsa-miR-16 (000391), hsa-miR-24 (000402), hsa-miR-155 (002623) or hsa-miR-484 (001821) (Applied Biosystems)) according to manufacturer’s instructions. Cycling was performed using a 10 µl reaction on a Step 1 Plus thermal cycler: for cfDNA: 95 °C for 20 sec, followed by 50 cycles of 95 °C for 1 sec and 60 °C for 20 sec; for miRNA assays: 95 °C for 20 sec, followed by 40 cycles of 95 °C for 1 sec and 60 °C for 20 sec. Results were analysed with Step 1 v2.2 software and Microsoft® Excel.

Quantitation of cfDNA and Targeted Deep Sequencing

Each sample (3.6 µl of eluted cfDNA) was quantified in triplicate relative to a standard curve, generated from a serially-diluted template (Human Genomic DNA; Roche) of known concentration, and converted to ng/mL using the following equation: (Total quantity of cfDNA/3.6) x 1000, and an in-house locus-specific TaqMan assay targeting GAPDH [11], and two inventoried assays for ACTBL2 (Hs001101944_s1, targeted amplicon 67 bp) and HPRT1 (Hs03929096_g1, targeted amplicon 64 bp) (Applied Biosystems). For the λ/HindIII DNA spike-in experiments, two fragments (23kb and 564 bp) were quantified using custom primers (Sigma) and FAM-MGB probes (Applied Biosystems). The spike-in was quantified in triplicate relative to a standard curve generated from serially-diluted λ/HindIII DNA. Sequences were: 23 kb fragment: forward, 5’-CGCACAGGAACTGAAGAATG-3’; reverse, 5’-CCGTCGAGAATACTGGCAAT-3’; FAM-MGB probe, 5’-TGTACTTTCGTGCTGTCGCGGATCG-3’ (targeted amplicon 108 bp); 564 bp fragment: forward, 5’-TGGGATCATTGGGTACTGTGG-3’; reverse, 5’-CAGACTTGGGGGTGATGAGTTT-3’; FAM-MGB probe, 5’-CCGCTATCCCTGATCAG-3’ (targeted amplicon 97 bp).

For DNA sequencing, cfDNA was isolated from 1 ml plasma of four patients with metastatic breast cancer, using two kits (QIAamp^® DNA Blood Mini Kit and the QIAamp^® Circulating Nucleic Acid Kit) to generate 8 templates. 10 ng of each template was sequenced in pairs on two 316 chips on the Ion PGM™ Sequencer using the Ion AmpliSeq™Cancer Hotspot Panel v2 according to manufacturer’s instructions. DNA sequencing data was accessed through the Torrent Suite v.3.4.2. Reads were aligned against the human genome (hg19) using Alignment v 3.4.48996 and variants called using the coverageAnalysis v 3.4.51836 and variantCaller v 3.4.51874, respectively. All raw BAM files have been deposited at the European Nucleotide Archive (ENA - https://www.ebi.ac.uk/ena) under accession number PRJEB4510.

Statistical analysis

All statistical analysis was performed using GraphPad Prism 6 for Windows. An unpaired Mann-Whitney test was used for comparison of RNA extractions and a two-tailed Spearman’s correlation analysis was used for correlation analysis of the TLDA and miRCURY cards. Where indicated, miRNA ΔC_T values were normalised to the average C_T values obtained on the miRCURY PCR array and TLDA card A before cycle 35 [15]. A matched-paired Wilcoxon test was used to compare the variants detected in the cfDNA sequenced on the Ion PGM™ Sequencer using the Ion AmpliSeq™Cancer Hotspot Panel v2.

Recovery of cfDNA and λ/HindIII spike-in

We compared 4 commercial kits for isolation of cfDNA, using 1 ml of pooled healthy plasma. Each extraction was performed in triplicate and each cfDNA sample was quantified in triplicate by qPCR targeting GAPDH [13], ACTBL2 and HPRT1. The 3 assays showed equivalent results, with the QIAamp^® DNA Blood Mini and QIAamp^® CNA kits giving the highest quantities of cfDNA/ml of plasma (mean 264.28 ng/ml and 239.85 ng/ml, respectively), 10 fold higher than the other two kits (Figure 1A). As validation, we also performed Qubit® 2.0 analysis, which confirmed this trend but with a slightly higher recovery of cfDNA (Figure 1A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Comparison of cfDNA recovery using different commercially available kits.

A. Average cfDNA yield per 1 ml of plasma with 4 extraction kits, quantified using qPCR for GAPDH, ACTBL2 and HPRT and Qubit® 2.0 High Sensitivity dsDNA Assay. Error bar = ±SEM, n = 9 from 3 independent experiments each performed in triplicate. ND = Not detectable. B. Percent recovery of serial dilutions of λ/HindIII DNA, quantified using assays targeting the 23 kb and 564 bp fragments. Error bar = ±SEM, n = 9 from 3 independent experiments performed in triplicate. CNA = Circulating Nucleic Acid kit.

https://doi.org/10.1371/journal.pone.0077963.g001

To determine the efficiency of cfDNA recovery, we assessed recovery of spiked-in λ/HindIII DNA using the 3 kits that gave the highest yields of cfDNA (QIAamp^® DNA Blood Mini Kit, QIAamp^® CNA Kit and NucleoSpin^® Plasma XS Kit). The QIAamp^® DNA Blood Mini kit and the QIAamp^® CNA Kit showed similar but variable recovery of the 23kb fragment (range 20-70%). The QIAamp^® CNA Kit gave the highest yield of the smaller (564 bp) fragment, recovering between 90 - 100% at each dilution tested; whereas the NucleoSpin^® Plasma XS Kit showed a lower recovery of both amplicons (Figure 1B).

cfDNA extraction methods generate equivalent targeted deep sequencing data

To determine whether the two kits giving the best recovery of cfDNA (QIAamp^® DNA Blood Mini kit and the QIAamp^® CNA Kit) showed any variance in cfDNA profiles, paired cfDNA from 4 metastatic patients were sequenced using the Ion AmpliSeq™Cancer Hotspot Panel v2 on the Ion PGM™ Sequencer. A total of 412M bases were generated across the eight cfDNA samples, with an average of 52 ± 5 M bases per sample (Table S1). This was equivalent to 503 ± 46K reads, > 96% of which mapped to the target region providing an average coverage of 2000x. High similarity was observed in the coverage profile for each cfDNA pair across the 207 amplicons (average size 154 bp, range 111bp -187 bp) (Figure 2). Of note, one sample (P4) showed a peak in the number of reads (>30,000 reads) at ERBB2 suggesting gene amplification. The total number of reads at each amplicon showed high correlation, (ρ > 0.95, p<0.001) for 3 of the 4 cfDNA pairs (Table S1). In the one sample (P3) with lower but significant correlation (ρ = 0.87) the QIAamp^® CNA method gave a more uniform coverage that the QIAamp^® DNA Blood Mini method.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Comparison of number of reads between two different extraction kits using the Ion AmpliSeq™Cancer Hotspot Panel v2.

Analysis of 4 plasma samples extracted using both the QIAamp^® DNA Blood Mini kit and the QIAamp^® CNA Kit. In each trace, the number of reads is shown on the Y-axis. Red lines indicate the QIAamp^® CNA Kit and the blue lines represent the QIAamp^® DNA Blood Mini kit. For in depth analysis see Tables S1 and S2.

https://doi.org/10.1371/journal.pone.0077963.g002

A total of 167 germline and somatic variants were called with an average of 21 variants per sample (Table S2). 136 (81%) of the total variants were matched in paired samples in terms of variant position and frequency with no significant difference in percentage variant call (p>0.05, p<0.001). Thirty-one (19%) variants were unique, but the majority of these were in P3 and detected at <10% depth.

Recovery of circulating miRNA

We compared 3 commercial kits for extraction of circulating miRNA. Unfortunately, none were successful at recovering miRNA from 1 ml of the pooled control plasma, as two miRNAs (hsa-miR-21 or hsa-miR-191) were undetected by qPCR. Therefore, we trialled the 3 kits using pooled plasma from 3 primary breast cancer patients recruited to our studies. There were marked differences in recovery of miRNA between the three kits. Bioanalyser analysis showed minor differences in miRNA yields for the QIAamp^® CNA or miRNeasy Plasma/Serum kit; which were significantly higher than the miRvana microRNA Isolation kit (p <0.001) (Figure 3). Considering individual miRNAs, qRT-PCR revealed lower C_T values (4 - 10 cycles earlier) for both hsa-miR-21 and hsa-miR-191 with the miRNeasy Serum/Plasma kit compared to the QIAamp^® CNA and miRvana kits when analysing both cDNA and pre-amplified cDNA (Figure 3A, B). The C_T values for the synthetic ath-miR-159a spike-in were also consistently lower using the miRNeasy Serum/Plasma kit compared to the miRvana kit, but not to the QIAamp^® CNA kit (Figure 4A, B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Bioanalyzer analysis of circulating miRNA iolated with different kits.

A-C. Representative Bioanalyser traces of cell-free RNA obtained from 1 ml of plasma using 3 commercial kits. The kits used and RNA integrity (RIN) values are highlighted above the trace and representative gels are shown below each trace. A = extractions using the miRNeasy Serum/Plasma kit, B = mirVana microRNA Isolation kit and C = Circulating Nucleic Acid kit. Quantities obtained are shown in D. Error bar = ±SEM, n = 3 from 3 independent experiments.

https://doi.org/10.1371/journal.pone.0077963.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. qRT-PCR analysis of miRNA isolated with different kits.

A. Average C_T values obtained for hsa-miR-191, hsa-miR-21 and ath-miR-159a on cDNA and pre-amplified cDNA (Pre-Amp) from 1 ml of plasma. Error bar = ±SEM, n = 9 from 3 independent experiments performed in triplicate. S/P = Serum/Plasma kit, CNA = Circulating Nucleic Acid kit. B. Average ΔC_T values comparing the miRNeasy Serum/Plasma and CNA kit to the miRvana Isolation kit. ΔC_T values calculated using the equation: Average miRNA C_T using the miRvana kit - Average miRNA C_T using another kit (i.e miRNeasy Serum/Plasma or CNA kit). Error bar = ±SEM, n = 9 from 3 independent experiments performed in triplicate. S/P = Serum/Plasma kit, CNA = Circulating Nucleic Acid kit. C. Two-tailed Spearman’s correlation analysis of ΔC_T values obtained from TLDA A cards and samples extracted from 1 ml of plasma using the miRvana Isolation kit or 200 µl of plasma using the miRNeasy Serum/Plasma kit. Dotted lines represent the 95% confidence interval. Rho (ρ) and p-values are indicated on the graph. D. Average C_T values obtained analysing hsa-miR-191, hsa-miR-21 and ath-miR-159a in cDNA produced from 1 ml, 500 µl or 200 µl of plasma using the miRNeasy Serum/Plasma kit. Error bar = ±SEM, n = 9 from 3 independent experiments performed in triplicate.

https://doi.org/10.1371/journal.pone.0077963.g004

Of note, although the miRvana microRNA Isolation kit gave the lowest yield of miRNAs, even from 1 ml of pooled plasma, the profile of miRNAs detected by TLDA cards showed significant correlation with the profile generated following extraction of miRNA from 200 µl of plasma using the miRNeasy Serum/Plasma kit (p <0.0001) (Figure 4C). Based on these results, further optimisation of miRNA isolation was only performed using the miRNeasy Serum/Plasma kit.

The yield of miRNA is not dependent on plasma volume

Next, we compared miRNA yields from 200 µl, 500 µl and 1 ml starting plasma volume. There was no significant increase in miRNA yield with increasing plasma volume by either Bioanalyser analysis (Figure S1) or qRT-PCR (hsa-miR-21 and hsa-miR-191) (Figure 4D), suggesting the quantity of miRNA recovered was not volume dependent. As part of this experiment, we also spiked-in ath-miR-159a at a concentration of 30 ng (1000-fold higher than in the previous kit comparisons) to test whether this resulted in saturation of either the column or the qPCR. Although the 1000-fold increase in concentration of the spike-in was reflected by an appropriate C_T value with 200 µl plasma volume, there was an increase in C_T values (and therefore decrease in detected concentration of the spike-in) with increasing plasma volume, suggesting less efficient recovery.

Pre-analytical processing alters the yield of miRNA but not cfDNA

We obtained 15 ml blood samples from 5 healthy female volunteers and processed 7.5 ml for plasma at 2 hours post-venepuncture, while the remaining 7.5 ml was left for 6 hours at room temperature before processing. Plasma was separated using the 2 step 1000g, 1000g protocol (see methods). Both cfDNA and miRNA were recovered in triplicate from 1 ml and 200 µl of plasma using the QIAamp^® DNA Blood Mini kit and miRNeasy Serum/Plasma kit, respectively and analysed by qPCR. Yields of cfDNA were not altered by blood processing time (Figure 5A); however, 6 candidate miRNAs analysed were significantly different, in particular, hsa-miR-15b and hsa-miR-191 (p<0.001) (Figure 5B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Recovery of cfDNA and miRNA from plasma: effects of increasing centrifugation speed.

A. Quantitation of cfDNA using qPCR for GAPDH. Blood was processed within 2 hr or 6 hr post-venepuncture and cfDNA isolated from 1 ml of plasma. Whiskers of the box-plot show minimum and maximum values. B. Comparison of miRNA profiles from plasma processed 6 hr post-venepuncture compared to samples processed at 2 hr. Results presented as -ΔC_T (change from 2 hours. Error bar = ±Std Dev, n = 5. *p<0.05; **p<0.01; ***p<0.001. C. Two-tailed Spearman’s correlation analysis of ΔC_T values obtained from TLDA A cards for duplicate miRNA samples following centrifugation at 1000g and 2000g. Dotted lines represent the 95% confidence interval. D. Two-tailed Spearman’s correlation analysis of ΔC_T values obtained from TLDA A cards for duplicate miRNA samples following centrifugation at 1000g and 10000g. Dotted lines represent the 95% confidence interval. E. Two-tailed Spearman’s correlation analysis of ΔC_T values obtained from TLDA A cards for duplicate miRNA samples following centrifugation at 2000g and 10000g. Dotted lines represent the 95% confidence interval. For C - E, Rho (ρ) and p-values are indicated on each graph.

https://doi.org/10.1371/journal.pone.0077963.g005

We also compared recovery of cfDNA and miRNA from 1 ml and 200 µl of pooled plasma, respectively following centrifugation at 1000g then either 1000g, 2000g or 10000g. Yields of cfDNA were equivalent, as assessed by GAPDH quantitation (data not shown); however, miRNA yield was notably lower after centrifugation at 10000g. Following profiling with TLDA card A, there were a total of 195 miRNAs detected after 1000g and 187 miRNAs at 2000g, but only 138 miRNAs detected at 10000g (Table S3). However, comparison of ΔC_T values showed significant correlation between these (Figure 5C-E), particularly between samples centrifuged at 1000g and 2000g. As expected, platelet-associated miRNAs including hsa-miR-24, hsa-miR-191, hsa-miR-197 and hsa-miR-223 [16] were notably reduced after centrifugation at 10000g (Table S3).

MiRNA profiles show good reproducibility between replicate samples

To test reproducibility of the miRNA PCR arrays, we analysed pooled samples from patients with primary breast cancer in duplicate on both TLDA cards and the Exiqon miRCURY LNA microRNA PCR array. Both platforms showed high reproducibility between replicate samples (Figure S2A, B). There was also a lower but significant correlation between miRNA profiles on the two platforms (ρ = 0.6, p<0.0001 - Figure S2C). Furthermore, as melt-curve analysis showed that a number of positive probes with high C_T values (>35) on the miRCURY LNA platform had multiple melting temperatures (Figure S2C); further experiments were performed using the TLDA cards only.

Recovery of miRNA from stored plasma

Our final aim was to determine whether miRNA was recoverable from stored plasma, as we have shown previously that cfDNA can be isolated and analysed from plasma samples stored for > 12 years [14]. To achieve this we compared 20 plasma samples from patients with metastatic breast cancer as two pools: 10 extracted from freshly frozen plasma and 10 from plasma stored at -80 °C for >12 years. A total of 177 miRNAs were detected in the stored pool, compared to 202 miRNAs in the fresh pool (TLDA Pool A v2.0 Cards, Table S4), suggesting some loss of miRNAs on storage. We assessed whether C_T values were lower in fresh compared to stored samples and found no significant difference (p = 0.77). However, there was a significant correlation of both C_T and ΔC_T values between miRNAs detected in the two pools (ρ = 0.84, p<0.0001, and ρ = 0.82, p<0.0001, respectively – Figure S2D, E), suggesting similar yield of miRNA and confirming successful recovery of miRNA from long term stored plasma.