The Effect of Preservative and Temperature on the Analysis of Circulating Tumor DNA

The clinical applicability of liquid biopsies is becoming increasingly evident with the emergence of novel targeted therapies (1–3) combined with next-generation technologies for assessment of tumor-specific sequence and structural alterations in cell-free DNA (cfDNA) derived from blood (4–12). These noninvasive approaches provide a method to evaluate tumor-specific alterations present in the tumor sample without the need for a tissue biopsy through direct identification of circulating tumor DNA (ctDNA) similar to noninvasive prenatal screening tests (NIPT) that scan a maternal blood sample for genetic conditions in the fetus. On average, the fetal fraction of DNA in maternal plasma ranges from 10% to 30% (13–15). In contrast, ctDNA is often present at significantly lower levels and can be considerably affected by less robust sample collection techniques as there may be only a few mutant DNA fragments per milliliter (mL) of plasma (6, 16). Therefore, ctDNA levels may depend upon how the cell and cell-free components were isolated into plasma or serum, especially as cfDNA has a very short half-life (6, 17, 18). Once cfDNA is released into circulation from apoptotic or necrotic cells, the half-life is estimated to be about 2 hours (6, 18). Although the rapid turnover of cfDNA allows for sophisticated monitoring of tumor dynamics, blood collection procedures should be carefully considered if the end goal is to detect ctDNA. To date, plasma has been the primary cell-free blood component used for ctDNA analyses, but studies have been conducted using serum (5). To isolate plasma, whole blood is anticoagulated to prevent clotting of the cellular component, and following centrifugation, the cell-free supernatant is removed. When three commonly used anticoagulants [di-potassium ethylenediaminetetraacetic acid (K₂EDTA), lithium heparin, and sodium citrate] were evaluated in their ability to stabilize cfDNA levels after venipuncture, it was found that total plasma DNA concentrations were stable with all 3 anticoagulants if whole blood was processed into plasma within 6 hours. However, K₂EDTA was the anticoagulant with the smallest change in cfDNA levels after 6 hours (19). To better understand the influence of preanalytic whole blood processing on cfDNA levels, we evaluated several conditions to find the optimal approach.

All biologic specimens utilized in this study were obtained from patients enrolled in a study approved by the Johns Hopkins University Institutional Review Board and provided informed consent. Patients whose tumors had been sequenced as part of standard clinical care and had “hot spot” mutations in KRAS, EGFR, or BRAF were recruited into this study.

Serum was prepared after collection of the whole blood into tubes containing no anticoagulant. Whole blood was allowed to clot by leaving it undisturbed at room temperature for 15 to 30 minutes. The clot was removed by centrifuging at 1,000 to 2,000 × g for 10 minutes in a refrigerated centrifuge. The resulting supernatant is designated serum and was carefully aspirated from the tube without disturbing the clot (cells) and transferred in 1-mL aliquots into 1.5-mL Eppendorf tubes and then centrifuged a second time at 18,400 × g for 10 minutes to remove cellular debris. Serum aliquots were then transferred to a new 1.5-mL Eppendorf tube and frozen at −80°C.

Plasma was prepared from whole blood collected in K₂EDTA or BCT tubes. Cells were removed from the plasma by centrifugation at 375 × g for 10 minutes at room temperature. The resulting supernatant is designated plasma and was carefully aspirated from the tube without disturbing the buffy coat layer (cells) and transferred in 1-mL aliquots into 1.5-mL Eppendorf tubes and then centrifuged a second time at 18,400 × g for 10 minutes to remove cellular debris. Plasma aliquots were then transferred to a new 1.5-mL Eppendorf tube and frozen at −80°C.

Ten tubes of peripheral blood (∼10 mL per tube) were collected from each patient with tumors that harbored KRAS mutations. Of these 10 tubes, 5 were K₂EDTA tubes and the 5 tubes were BCT tubes. After venipuncture, the tubes were inverted 10 times immediately after the blood draw. Whole blood was kept at room temperature and processed into plasma at time points (0, 24, 72, 120, 168 hours), except for the baseline sample (time 0) that was processed into plasma (as described above) within 4 hours of venipuncture.

When the influence of temperature on ctDNA levels was assessed, 12 tubes of peripheral blood (∼10 mL per tube) were collected from 6 patients with tumors that harbored KRAS mutations. Of these 12 tubes, 6 were K₂EDTA tubes and the 6 tubes were BCT tubes. After venipuncture, the tubes were inverted several times. K₂EDTA tubes and BCT tubes were equally split into two sets: one set was kept at 4°C whereas the other set was stored at room temperature and processed into plasma at time points 0, 24, and 72 hours, except for the baseline sample (time 0) that was processed into plasma (as described above) within 4 hours of venipuncture.

Circulating cfDNA was extracted from plasma using the QIAamp Circulating Nucleic Acid Kit (Qiagen, Catalog #55114). The purified DNA concentration was determined using Qubit dsDNA HS Assay Kit (Life Technologies, Catalog #Q32854). All extracted DNA was analyzed with the appropriate droplet digital PrimePCR assay (BioRad, Catalog #100-31246 and #100-31249) on the BioRad QX200 droplet digital PCR system to determine the number of wild-type and mutant genomic equivalents. The mutant fractional abundance was calculated as a percentage of the total DNA present at each time point. The fragment size of cfDNA derived from plasma and serum was measured using the High Sensitivity DNA Kit on the 2100 Bioanalyzer (Agilent, Catalog #5067-4627).

Circulating cfDNA from healthy controls was extracted and quantified as described above. Genomic libraries were created from the total yield of extracted DNA using NEB enzyme mixes for End Repair (cat# E6050), A-tail (cat# E6053), and ligation (cat# E6056, incorporating 8 pooled adapter oligonucleotides generated by IDT) followed by 12 cycles of PCR and purified as described previously using Agencourt AMPure XP beads (Beckman Coulter; ref. 20). A 58-gene panel (Supplementary Table S4) interrogating 80,930 nucleotides was employed to detect sequence mutations by next-generation sequencing, using the Agilent SureSelect in-solution targeted enrichment system (Agilent; ref. 21).

Targeted regions were captured in-solution using a custom SureSelect kit according to the manufacturer's instructions. The captured DNA library was immobilized from the beads using 200 μL of a 95% Formamide and 10 mmol/L EDTA solution at 65°C for 5 minutes. Once separated from the beads, the DNA was purified using Agencourt AMPure XP beads (Beckman Coulter) in a ratio of 1.0× to 1.8× ratio of captured DNA library to beads and washed using 70% ethanol according to manufacturer's instructions. The captured DNA library was amplified and purified as described previously (20). The captured genomic library underwent 100-bp paired-end sequencing on a HiSeq 2500 instrument (Illumina) to obtain an average of 26,574× total coverage in the targeted region of interest (Supplementary Table S5), with variant identification performed as previously described (21, 22).

The statistical difference between BCT and K₂EDTA tubes at various time points as well as the difference in total genomic equivalents or mutant allele fraction (MAF) between serum and plasma was calculated using a paired, 2-tailed Student t test where P < 0.05 was considered significant (GraphPad Prism v5.00). To study sample size, a power analysis was performed in R (v3.2.4). Six samples were determined to provide an 80% power for the detection of a 10% difference between blood collection tube types (P < 0.05, SD = 7). One sample set was removed from the analysis as the baseline MAF was <1% and had the potential to confound these findings. For a reduced study size of 5, the study was determined to provide 67% power for the detection of a 10% difference between blood collection tube types.

To evaluate the yield of cfDNA between serum and plasma, the quantity of total genome equivalents and the MAF were compared between serum and plasma collected from patients with stage II and III pancreatic cancer (n = 46 and n = 51, respectively). All cases were known to harbor mutations in KRAS at p.G12C, p.G12D, p.G12I, p.G12R, p.G12S, p.G12V, or p.G13D positions. DNA extracted from serum was analyzed for mutations known to be present in the matched tumor using standard droplet digital PCR assays. We compared our results in serum with a similar dataset we had previously analyzed for mutations in plasma from 51 stage-matched patients with pancreatic cancer from a study by Sausen and colleagues (20). Total genome equivalents were significantly greater in the DNA obtained from serum specimens (P < 0.001, Fig. 1A), whereas the observed MAF was significantly higher in the DNA obtained from plasma than from serum specimens (P < 0.001, Fig. 1B). While distinct peaks of cfDNA were observed in plasma with the largest fraction measuring 155 base pairs on average, the cfDNA fragments in the serum varied considerably in size ranging from 150 to 2,000 base pairs (Fig. 1C).

Whole blood from 9 patients with colorectal cancer, where the tumor genotype was known to harbor a mutation in KRAS, EGFR, or BRAF, was collected and monitored to assess the stability of DNA obtained from K₂EDTA tubes and BCT tubes at room temperature over a period of 7 days (Supplementary Fig. S1 and Supplementary Table S1). During this time, the change in overall genome equivalents observed in DNA obtained from BCT tubes remained stable compared with K₂EDTA tubes stored at room temperature (P < 0.005 between 0 and 168 hours, Fig. 2A, Supplementary Table S2). Within 24 hours, a significant increase in the number of total genome equivalents was observed in the DNA obtained from K₂EDTA tubes, which continued to increase over time. After 1 week, there was a 13-fold increase in total genome equivalents per milliliter in K₂EDTA tubes, whereas the percentage increase of total genome equivalents per milliliter obtained from BCT tubes was 6%. For cases where the MAF was >1% in both tube types, the observed MAF was stable in blood collected and stored in BCT tubes but decreased considerably in blood collected and stored in K₂EDTA tubes (Fig. 2B, one time point in P1 was unable to be evaluated and therefore excluded from analysis, Supplementary Table S2).

To determine whether storage of K₂EDTA tubes at 4°C could increase the allowable time interval between blood collection and fractionation, blood was collected in both K₂EDTA and BCT tubes from 6 KRAS- or EGFR mutant–positive patients with cancer (Supplementary Fig. S1 and Supplementary Table S1) and held at either room temperature (Streck BCT) or at 4°C (K₂EDTA) for up to 72 hours. When K₂EDTA tubes were stored at 4°C, there was no significant increase in yield of total genomic equivalents over the 3-day period (Fig. 3A, Supplementary Table S3) in contrast to the observed increase in total genomic equivalents when whole blood was kept at room temperature and stored in K₂EDTA tubes. We also analyzed the change in total mutant DNA fragments per milliliter and the change in MAF over the same 3-day period. The percentage change in the number of mutant DNA fragments initially increased at 24 hours in both tubes but decreased by 4% (Streck BCT) or 21% (K₂EDTA) at 72 hours, with no significant difference observed between K₂EDTA and BCT tubes (P > 0.05 at each time point, Fig. 3B, Supplementary Table S3). The MAF decreased in plasma isolated from K₂EDTA tubes and BCT at 24 and 72 hours with a statistically significant difference observed between tube types (P < 0.05, Fig. 3C, Supplementary Table S3) at 72 hours. There was an observable difference in gDNA levels between plasma DNA extracted at baseline (within 4 hours of blood draw) and plasma DNA extracted at 72 hours in K₂EDTA tubes stored at 4°C or room temperature (Fig. 3D), whereas BCT tubes show minimal differences in total cfDNA levels regardless of storage temperature or time (Fig. 3E). In all cases, the trends in cfDNA and ctDNA levels were consistent between K₂EDTA tubes and BCT in each series tested.

In addition to total genomic equivalents, we evaluated the volume of plasma obtained from fractionated blood at room temperature over the 7-day period. We found that the volume of plasma obtained decreased in both tube types over time using the same fractionation conditions (Fig. 4A). By day 7 (168 hours), less than 2 mL of plasma was fractionated from each 10 mL Vacutainer tube (Streck or K₂EDTA) containing 8 to 10 mL of whole blood.

The number of mutations between BCT and K₂EDTA tubes was analyzed using a hybrid-capture, next-generation sequencing approach. Three BCT and 3 K₂EDTA tubes of whole blood (∼10 mL per tube) from 3 healthy individuals were analyzed for sequence mutations over a 72-hour period where BCT tubes were stored at RT while K₂EDTA tubes at 4°C (Supplementary Table S5). Using a targeted 58-gene panel designed to cover 80,930 nucleotides and detect sequence mutations in plasma, the observed per base error rate, calculated as the number of bases mutated divided by the number of targeted bases, was similar between plasma isolated from K₂EDTA and BCT tubes (0.00041% vs. 0.00014%, P = 0.28, Student t test). Consistent with previous findings, the yield of cfDNA in healthy individuals was significantly lower than the yield in diseased individuals (P < 0.02, Student t test, Supplementary Table S6; refs. 4, 23, 24).

Methods of whole blood processing, preservation, and storage are important considerations for the analysis of ctDNA. A common question is whether serum or plasma are equivalent for cfDNA analyses. It is clear that the abundance and nature of the cfDNA is very different in these 2 fractions of whole blood. Serum contains high-molecular-weight DNA, likely attributed to how serum is produced by clotting whole blood into a solid and liquid phase. This relatively violent process results in the release of cellular DNA into the cell-free liquid component by lysis and apoptosis (Fig. 1). In contrast, plasma is produced using an anticoagulant (potassium EDTA, sodium citrate, lithium heparin) and the cellular (solid) component gently separates from the plasma (liquid) component resulting in cfDNA that is uniform in size (median size of 180 bp). This is likely the reason that K₂EDTA tubes are recommended by the International Council for Standardization in Hematology (ICSH) and the Clinical and Laboratory Standards Institute (CLSI) as the ideal tube for molecular testing (such as routine PCR-based molecular diagnostics and HIV viral load testing) because the anticoagulant dipotassium salts preserve cellular components (25, 26) and prevent contamination and dilution with cellular DNA that may interfere with analyses of cfDNA. In addition, EDTA salts have been shown to inhibit DNAse activity and thereby help preserve cfDNA levels ex vivo during storage, transport, and processing (27). Di- and tri-potassium (K2 and K3) salts of EDTA are standard anticoagulants that have been used in studies with cfDNA. K₂EDTA is spray-dried on the walls of the tube rather than added in liquid form, as is the case for K₃EDTA. The spray-dried coating has been shown to prevent both cell shrinkage and other osmotic effects on blood cells more effectively than in liquid form (28, 29). Several studies have evaluated cfDNA and ctDNA levels when whole blood was stored and preserved with K₃EDTA. These analyses show varying results, many with findings similar to ours when evaluating cfDNA and ctDNA levels in stored whole blood (27, 28, 30–33).

In our study, we compared K₂EDTA tubes with BCT tubes, which are intended to preserve cfDNA through proprietary components that are free of formaldehyde. cfDNA levels remained stable in BCT tubes over a 7-day period even when whole blood was stored at room temperature and likewise cfDNA levels were stable when whole blood was stored at 4°C in K₂EDTA tubes for 24 hours. Not surprisingly, temperature has been shown to have a strong influence on peripheral blood storage in K₂EDTA-coated tubes. Prior studies have shown that storage of whole blood in K₂EDTA tubes at 4°C results in the stability of leukocytes and plasma proteins for an extended period (29). Likewise, storage of whole blood at 4°C preserves both total cfDNA and ctDNA at levels equivalent to the BCT tubes. In contrast, when blood was stored at room temperature in K₂EDTA-coated tubes, there was a massive release of cfDNA presumably by cells undergoing unorganized death by lysis or necrosis as demonstrated by the high-molecular-weight DNA fragments found in the plasma after 24 hours (Figs. 2A and 3D). At room temperature, there was also a concomitant decrease in the fraction of ctDNA (Fig. 2B) likely from the dilution of the mutant DNA fragments by wild-type DNA released by the lysed lymphocytes. We also noted that at room temperature, the volume of plasma recovered significantly decreased regardless of storage tube used and should be considered when storage times exceed 1 day, especially when samples are idle. Furthermore, it appears that using either of these preservatives have no observable effects on the quality of next-generation sequencing analyses of cfDNA.

As the number of ctDNA analyses increase for the determination of mutation profiles in patients with cancer for research and clinical care, the methods used for whole blood preservation and storage before processing into plasma and DNA extraction will influence the quality of analysis. It is clear that minimizing cellular lysis during whole blood storage is of high importance and this can be accomplished using BCT tubes at room temperature or in EDTA tubes stored at 4°C.

V. Adleff is a consultant/advisory board member for Personal Genome Diagnostics. V.E. Velculescu holds ownership interest (including patents) in and is a consultant/advisory board member for Personal Genome Diagnostics. L.A. Diaz Jr is an employee of Personal Genome Diagnostics; reports receiving speakers' bureau honoraria from Merck; and holds ownership interest (including patents) in PapGene and Personal Genome Diagnostics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Parpart-Li, D. Laheru, V.E. Velculescu, M. Sausen, L.A. Diaz Jr

Development of methodology: S. Parpart-Li, B. Bartlett, V. Adleff, V.E. Velculescu, M. Sausen, L.A. Diaz Jr

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Bartlett, V. Adleff, L. Tucker, J. Phallen, J.R. Brahmer, N. Azad, I. Browner, L.A. Diaz Jr

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Parpart-Li, R. Steinberg, A. Georgiadis, J.R. Brahmer, V.E. Velculescu, M. Sausen, L.A. Diaz Jr

Writing, review, and/or revision of the manuscript: S. Parpart-Li, B. Bartlett, V. Adleff, A. Georgiadis, J.R. Brahmer, I. Browner, D. Laheru, V.E. Velculescu, M. Sausen, L.A. Diaz Jr

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B. Bartlett, M. Popoli, A. Georgiadis, L.A. Diaz Jr

Study supervision: V.E. Velculescu, M. Sausen, L.A. Diaz Jr

We would like to thank S. Bonerigo and A. Ryan (Johns Hopkins) for their clinical coordination.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.