ABSTRACT

Long-read DNA and RNA sequencing facilitate genome assembly, haplotyping, complex variant calling, and gene isoform identification. Structural variant calling is integral to the molecular characterization of tumors; thus, long-read nanopore DNA sequencing technology is becoming routinely used in cancer research. As a standard practice, high molecular weight (HMW) DNA is extracted from both tumor and matched normal samples from blood or buffy coat to redact germline variants from somatic mutations. However, we found that buffy coat DNA consistently underperformed compared to DNA extracted from tumor tissue. Furthermore, this observation was unique to DNA extracted from buffy coat cells collected in Streck, but not ethylenediaminetetraacetic acid (EDTA), tubes. We therefore investigated whether the released formaldehyde in Streck tubes resulted in DNA crosslinking, which would explain the low data throughput. Indeed, a decrosslinking step during Streck DNA extraction significantly improved data yield and fragment length without compromising data quality. We therefore recommend a tailored DNA extraction protocol of Streck-derived buffy coat samples for nanopore sequencing.

Address correspondence to: Neeman Mohibullah, Integrated Genomics Operation, Memorial Sloan Kettering Cancer Center, 1250 1st Ave., New York, NY 10065, USA (Tel: 212-639-2853; Fax: 646-888-3407; Email: [email protected]).

Keywords: buffy coat, crosslinking, DNA, extraction, nanopore sequencing, Streck

Conflicts of Interest: The authors declare no conflicts of interest.

Note: Anonymized DNA samples were used from three healthy volunteers under an approved protocol, IRB 06-107, by the Institutional Review Board of Memorial Sloan Kettering Cancer Center.

INTRODUCTION

The study aim was to develop a protocol that allowed the use of DNA derived from buffy coat samples collected in Streck Cell-Free DNA Blood Collection Tubes (cfDNA BCTs). While processing more than 600 samples in our core, we noticed that Streck BCT DNA consistently underperformed in nanopore long-read sequencing, failing to reach target coverage of 30X for human genome, approximately equivalent to data yield of 90 Gb (Fig. 1). However, both buffy coat DNA from Ethylenediaminetetraacetic acid (EDTA) BCTs and the optimal cutting temperature (OCT) compound-derived matched tumor tissue reliably yielded >100 Gb per PromethION flow cell. Streck tubes are coated with imidazolidinyl urea, a formaldehyde-releasing agent, which prevents release of cellular DNA, facilitating analysis of cfDNA.[1]^,[2] A previous report showed that Streck tubes induce peripheral mononuclear cell protein formaldehyde crosslinking, particularly in cytoskeletal proteins and histones.[3] We hypothesized that Streck buffy coat DNA is also crosslinked by the released formaldehyde, impeding nanopore sequencing. Due to their frequent use in the clinic for liquid biopsy testing, it is important to enable compatibility between Streck-derived genomic DNA (gDNA) and nanopore long-read sequencing, a powerful technology for analyzing cancer genomes.

Here we describe a simple alteration in high molecular weight (HMW) DNA extraction protocol for Streck-derived buffy coat samples, which is necessary to successfully prepare and sequence nanopore libraries.

MATERIALS & METHODS

Sample collection and processing

One 5 mL Streck cfDNA BCT (S) (230470, Streck, La Vista, NE, USA) and two 5 mL BD Vacutainer EDTA tubes (367863, Becton Dickinson, Franklin Lakes, NJ, USA) (untreated: E; treated with final concentration of 0.1% formaldehyde (47608, Sigma, St. Louis, MO, USA) immediately after blood draw: EF) were collected from three healthy volunteers. After a 20-hr incubation, DNA was extracted with (+) and without (–) decrosslinking (Fig. 2).

Buffy coat DNA extraction

Whole blood samples were centrifuged at 800 g for 10 min at room temperature. The plasma phase was discarded, and the buffy coat was split into two 1.5 mL Eppendorf DNA LoBind tubes (13698791; Thermo Fisher Scientific, Waltham, MA, USA). Genomic DNA (gDNA) samples were extracted using the Monarch HMW DNA Extraction Kit for Cells & Blood (T3050L; New England Biolabs, Ipswich, MA, USA) following the manufacturer’s protocol except for an initial 1:1 dilution of buffy coat in 1xPBS (10010072; Gibco, Grand Island, NY, USA) to bring the buffy coat volume to 500 µL followed by erythrocyte lysis. For decrosslinked samples only: the addition of Nuclei Lysis Solution (Proteinase K and Nuclei Lysis Buffer) was followed by 56°C incubation for 10 min at 2000 rpm (ThermoMixer C; Eppendorf, Hamburg, Germany) and overnight incubation at 65°C, before continuing with leukocyte lysis. Extracted DNA was quantified by Quant-iT 1X dsDNA High Sensitivity Assay Kit (Q33232; Thermo Fisher Scientific), according to manufacturer’s instructions. Quality and size were analyzed by NanoDrop (ND1000; Thermo Fisher Scientific) and gDNA TapeStation 4200 (Agilent, Santa Clara, CA, USA), respectively.

Oxford Nanopore Technologies (ONT) library preparation and sequencing

3 μg gDNA was normalized to 49 µL of nuclease-free water and transferred to a g-TUBE (520079; Covaris, Woburn, MA, USA). gDNA was sheared at 4600-6000 rpm for 2-4 min following the manufacturer’s guidelines. DNA size was re-analyzed, to ensure ~25 kb fragments, by TapeStation, and DNA purity was re-assessed by NanoDrop, before library preparation.

Libraries were prepared with the Ligation Sequencing Kit (SQK-LSK114; Oxford Nanopore Technologies (ONT), Oxford, UK), according to manufacturer’s guidelines. Libraries were quantified by Quant-iT and 50 fmol were loaded onto a FLO-PRO114M (ONT) flow cell. Run options in the MinKNOW UI included 1 kb minimum read length and 90-hr run limit. The high accuracy (HAC) basecalling model was selected. Flow cells were washed using Flow Cell Wash Kit (EXP-WSH004; ONT) once the active pores reached ~25-30%, and 50 fmol of the library was reloaded according to manufacturer’s guidelines.

Data Analysis

Estimated coverage was calculated using the equation $C = LN/G$ where L = average of N50 and median read length, and N = read number, all of which were reported by MinKNOW 23.11.7, and G = haploid genome size. The sequencing reads were aligned to human genome reference GRCh38 using minimap2 with default Oxford nanopore reads setting. OxoG Metrics stats were collected using CollectOxoGMetrics from Picard tools (https://broadinstitute.github.io/picard/).

RESULTS

In an early study of five tumor and matched normal pairs processed in the core, Streck-derived buffy coat ONT DNA libraries consistently underperformed, defined as data yield below 90 Gb, unlike matched tumor and other EDTA-derived libraries, which had an average yield of 116 Gb, and an average coverage of 47X (Fig. 1A). Moreover, most Streck-derived samples resisted gTube fragmentation, maintaining a >60 kb size as assessed by TapeStation, even after multiple attempts to shear to 25-30 kb (Table 1). While very long DNA can lead to lower data yield and pore clogging, the size alone could not explain the significant difference in data yield and the rapid pore depletion (Fig. 1B), as some of the tumor samples also appeared >55-60 kb. Importantly, even though the Streck buffy coat samples seemingly had the longest fragments after extraction, the post-sequencing average N50 value was the lowest compared to tumor and EDTA buffy coat N50s (Fig. 1C). We therefore questioned whether the Streck preservative, potentially releasing formaldehyde, was responsible for these outcomes. A formaldehyde test strip confirmed that Streck tubes have a formaldehyde concentration of >100 mg/l (0.01%), potentially crosslinking gDNA over prolonged incubation time (Fig. 3). Crosslinked DNA could explain the rapid pore depletion and low data yield. To address our hypothesis and test whether a decrosslinking step could mitigate the outcome, blood samples were collected from each of three healthy volunteers and processed under the conditions shown in the schematic (EDTA, E; EDTA + Formaldehyde, EF; Streck, S) (Fig. 2). All extracted DNA samples were of optimal quality. It is noteworthy that all EF samples resisted shearing, maintaining >50 kb size, and had the lowest library yield (Table 2). Comparison between the samples extracted without decrosslinking showed that S and EF samples behaved similarly with lower library yield, rapid pore depletion, and lower N50s versus E samples, corroborating our previous observations (Fig. 4). Decrosslinking rescued sequencing output for both S and EF samples, resulting in data yield of >100 Gb (Fig. 4A-B). E sample performances were comparable with or without decrosslinking (Fig. 4A-C). It has been previously shown that extended heat exposure causes DNA damage, with 8-oxoguanine (oxoG) being the most frequent base lesion.[4] Analysis of the oxidative DNA lesion oxoG showed that the higher temperature for the decrosslinking step did not induce oxidative DNA damage. Additionally, decrosslinking resulted in improved oxidation error rates in S and EF samples, versus no decrosslinking, probably due to the resolution of crosslinks reducing the number of alternate base calls (Fig. 4D).[5]

In this study, we hypothesize that the released formaldehyde in Streck tubes crosslinks cellular DNA impeding nanopore sequencing. Streck DNA nanopore performance was comparable to formaldehyde treated DNA, suggesting that the released formaldehyde in Streck tubes was responsible for the poor outcome. As such, decrosslinking at 65^oC overnight during the HMW DNA extraction process of Streck-derived buffy coat DNA samples is essential for a high-throughput nanopore sequencing run.

ACKNOWLEDGMENTS

This work was funded by the MSK Cancer Center Support Grant P30 CA008748; PI: Selwyn Vickers. We are grateful to Ms. Tina M. Alano and Dr. Joseph M. Scandura for the blood collection of volunteers, and Ms. Emily Stockfisch for blood sample information. We thank all members of the Integrated Genomics Operation for helpful insights and support.