Background

Plasma cell-free DNA (cfDNA) is widely used in clinical testing(1), including non-invasive prenatal testing (NIPT)(2–5), cancer testing(6–8), infectious disease testing(9,10), and organ transplantation monitoring(11,12). In cancer testing, cfDNA is actively studied for the detection of residual cancer disease(13) and early detection (14,15).

Reliable cfDNA analysis typically depends on specialized preservative tubes (e.g., Streck) or rapid processing of EDTA tubes(16). However, both methods limit accessibility by requiring specific tubes or on-call personnel, which adds costs and restricts cfDNA testing in resource-limited settings.

In contrast, heparin separators are standardly used in clinical settings, ubiquitous in everyday phlebotomy trays at many institutions for routine chemistry testing, such as the basic metabolic panel. For example, at our institution alone (Stanford Health Care), over 900,000 of these tubes are processed annually. These tubes contain a gel barrier that separates plasma from cellular components following routine centrifugation. They are typically handled through high-throughput, automated, and CLIA-compliant workflows that incorporate prompt centrifugation and refrigerated storage.

This study explores the potential utilization of millions of leftover plasma samples and the associated healthcare infrastructure as an untapped resource for cfDNA research and clinical testing. To assess the feasibility of using heparin separators for cfDNA testing, we conducted whole-genome sequencing (WGS) and the FLEXseq methylation assay(17) on paired samples collected from the same venipuncture in both heparin separators and gold standard EDTA (and/or Streck) tubes. Our evaluation spanned multiple benchmarks relevant to a broad assortment of cfDNA tests, including metagenomics(9,18), methylation profiling (17,19), copy number alterations (CNAs)(20), epigenetic tissue-of-origin mapping(19), and cfDNA fragmentomics(21).

Results

Heparin separators preserve cfDNA integrity under ideal conditions

To evaluate the potential of repurposing leftover plasma from heparin separators, we conducted a controlled comparison in the Healthy Cohort (n = 5) using matched samples collected from the same blood draw in heparin separators and in gold-standard EDTA and Streck tubes (Fig. 1a, Table S1).

Figure 1.

Fragment size distributions were highly similar across all tube types, with a modal peak at ∼166 bp, consistent with known cfDNA profiles (Fig. 1b)(4,24). End motif analysis of 256 4-mers showed comparable rankings (Fig. 1c), suggesting that heparin separators do not perturb fragmentation patterns when processed immediately.

To assess the preservation of epigenomic features, we calculated methylation levels at individual CpG sites using the FLEXseq methylation assay’s output. CpG sites with ≥30× coverage showed strong correlations between heparin separator and EDTA (Pearson’s r = 0.93, Fig. 1d, left), and between heparin separator and Streck (r = 0.93, Fig. 1d, right). Tissue deconvolution revealed highly concordant cell type proportions across all collection tube types (Fig. 1e, P = 0.86, PERMANOVA), with consistent results across all five cases (Heparin separator versus EDTA: Pearson’s r = 0.92-0.93; heparin separator versus Streck: Pearson’s r = 0.92-0.93; Supplemental Fig. 1). These findings show that heparin separators preserve the biological status of methylation under ideal processing conditions.

To mimic clinical storage, we tested the impact of delayed processing. Blood from a healthy volunteer was collected into EDTA, heparin separator, and non-gel heparin tubes, then divided into two sets (Supplementary Fig. 2a). One set was processed immediately, while the other was refrigerated (4°C) for one hour before the first spin and an additional four days before the second spin. Fragment size profiles were identical (Supplementary Fig. 2b), confirming that refrigerated storage does not compromise cfDNA integrity.

Hospital Cohort: Residual plasma from Heparin separators after clinical usage

To evaluate the feasibility of heparin separator tubes for cfDNA analysis, we identified 34 matched pairs of viral PCR–positive plasma samples (viral load >1000 IU/mL) collected in both heparin separator and EDTA tubes from the same blood draw. These cases were retrieved by querying the electronic medical record systems at Stanford (n=28) and UCSF (n=6) for samples that tested positive by the Clinical Virology Laboratories. All samples underwent cfDNA extraction, followed by whole-genome sequencing (WGS) to evaluate viral load, copy number alterations (CNAs), and fragmentation profiles, as well as methylation profiling via FLEXseq.

Leftover plasma from the heparin separators enables viral load detection

We performed metagenomic sequencing to compare viral load measurements and assess concordance between plasma collected in EDTA and heparin separators (Fig. 2). WGS achieved a median coverage of 4.52× (IQR: 1.76-5.08×). Viral read counts normalized to the number of reads aligned to the human genome were highly correlated between plasma collection tube types (Pearson’s r = 0.99, P < 0.001), with no significant difference observed (Wilcoxon signed-rank test, P = 0.27). This high level of concordance was consistently observed across institutions and among distinct viral types.

Figure 2.

Leftover plasma from the heparin separator enables copy number profiling

CNAs are commonly observed in cancer and can be detected in cfDNA as a surrogate marker for tumor burden. To evaluate whether heparin separators preserve CNA signals, we compared the genome-wide copy number profiles of samples from heparin separators and EDTA tubes. CNA profiles were generated using a 500 kbp bin size, with a median of 97.68 million reads per sample (range: 0.73–360.73 million). Representative genome-wide CNA plots from a matched pair (P171: heparin separator; P187: EDTA) revealed similar patterns of chromosomal gains and losses (Fig. 3a). Genome-wide log₂ copy ratio deviations were concordant between tube types (Pearson’s r = 0.97; Fig. 3b). Similar patterns and concordance were observed in other CNA-positive samples (n = 5, Pearson’s r = 0.86–0.99; Supplementary Fig. 3). Tumor fraction estimates based on copy number also correlated strongly between matched samples (Pearson’s r = 0.97; Fig. 3c). These findings demonstrate that plasma from heparin separators is suitable for CNA detection and tumor fraction estimation.

Figure 3.

Correlated Methylation Profiling from Heparin Separator Plasma

To assess the reliability of methylation profiling across blood collection tube types, we analyzed the available residual matched plasma samples in the Hospital Cohort (n=12). Using FLEXseq(17), a high-resolution methylation assay, we observed strong genome-wide concordance of CpG methylation levels between heparin separator and EDTA tubes (P168: heparin separator, P185: EDTA tube; Pearson’s r = 0.90; Fig. 4a; other cases in Supplementary Fig. 4). To evaluate the biological relevance of the cfDNA methylation profiles, we performed cell type deconvolution on the paired plasma(19). The results showed comparable tissue-of-origin proportions between plasma collected in heparin separator and EDTA tubes (Fig. 4b). These findings demonstrate that cfDNA methylation signatures are well preserved in heparin separator plasma and are reliable for epigenomic applications such as cell-of-origin estimation.

Figure 4.

Comparable cfDNA yield deduced from the sequencing result

We quantified cfDNA concentrations (normalized to plasma volume) in the 25 matched pairs from the Healthy Cohort and the Hospital Cohort with sufficient leftover DNA. Qubit measurements revealed variability across samples (0.14–43.19 ng/mL; Fig. 5a & Fig. 5b) and poor correlation between EDTA and heparin separator data. However, Qubit measurements are related to both addressable (intact short fragments) and non-addressable (genomic length or broken) DNA, whereas NGS output is related to only addressable DNA.

Figure 5.

To enable precise measurements of the addressable DNA, we spiked in lambda DNA at a constant quantity across all samples at the beginning of the library preparation and normalized human-aligned read counts to this internal control. This strategy produces a measure (Human/Lambda ratio) proportional to the addressable DNA content of the sample. This normalized ratio yielded strong concordance between EDTA and heparin separator samples (R² = 0.998; Fig. 5c), indicating that library preparation effectively mitigated preanalytical variability.

Leftover plasma from the heparin separator alters fragmentation profiles

Fragmentation features of cfDNA, such as size distribution and end motifs, have been linked to nuclease activity and are known to reflect host physiology and pathology(25–27). To evaluate whether heparin separators preserve these fragmentation features, we examined two well-characterized fragmentomic features: size distribution and end-motif. Both EDTA and heparin separator plasma samples exhibited expected peak sizes near 70 bp and 170 bp (Supplementary Fig. 5). However, heparin separator plasma showed a modest increase in short fragments and a slightly diminished ∼166 bp peak, indicating a shorter size distribution. Analysis of 4-mer motifs at the 5′ ends of cfDNA fragments revealed significant differences in four of the six most frequent motifs (i.e., CCCA, CCAG, CCTG, and CAAA; P < 0.05, paired Student’s t-test), suggesting disrupted cleavage preference. These altered fragmentation profiles in the Hospital Cohort suggest that plasma samples collected in heparin separators may have influenced cfDNA fragmentation signatures. Given that the fragmentation profiles were concordant in the Healthy Cohort, the differences seen here may be due to degradation during delays in clinical lab processing after blood collection and prior to refrigerated storage.

Discussion

Our study explores the untapped potential of repurposing residual plasma from heparin separators as a practical and underutilized resource for circulating cfDNA-based research and diagnostic applications. Using matched samples from a Healthy Cohort and a real-world Hospital Cohort, we demonstrate that the leftover plasma is suitable for metagenomic viral detection, copy number profiling, and genome-wide methylation analysis, with performance comparable to gold-standard collection methods.

Although heparin separators have been previously used in COVID-19 cfDNA studies(28,29), their performance has not been systematically evaluated against standard collection methods. Historically, concerns about heparin-induced PCR inhibition and genomic DNA leakage from cells have limited the use of heparin tubes, although the available evidence remains inconclusive. For instance, Lo et al. reported no significant difference in DNA concentration between EDTA and heparin tubes when plasma was processed promptly within 2 hours after blood collection, suggesting minimal gDNA release(30). In contrast, Gerber et al. reported elevated cfDNA levels in heparin tubes compared to specialized tubes such as PAXgene, Roche cfDNA, and Streck, even at time point 0(31). The results in our Healthy Cohort suggest no significant differences in samples immediately processed or refrigerated at 4°C. Unlike earlier studies that relied on PCR-based assays and used standard heparin tubes, our study utilized heparin separators with a gel barrier to physically separate plasma from cells. Moreover, modern cfDNA sequencing workflows include multiple purification steps that could effectively reduce heparin interference on PCR and preferentially retain short cfDNA fragments, minimizing the impact of gDNA contamination on downstream analysis.

Fragmentation profiles were consistent across tube types in the Healthy Cohort but variable in the Hospital Cohort, likely reflecting preanalytical differences before laboratory processing. Prior research has shown that incubating blood with heparin at room temperature for 6 hours alters cfDNA fragmentation profiles, possibly due to heparin-induced nucleosomal disruption and enhanced DNase I activity(32). In contrast, prompt refrigeration in the Healthy Cohort preserved cfDNA integrity, potentially by inhibiting nuclease activity.

Our study demonstrates the feasibility of repurposing residual plasma from heparin separators for sequencing-based cfDNA analysis, offering a collection method that integrates seamlessly into existing clinical workflows. By utilizing samples already collected during routine care, this approach eliminates the need for additional venipuncture and enables retrospective access to clinically important cases, including those where informed consent could not be obtained at the time of collection. While promising, the approach is constrained by the limited residual plasma volume (0.5–1.5 mL), which may not meet the input requirements of applications such as minimal residual disease detection. Moreover, variability in transport and storage conditions prior to processing, as observed in the Hospital Cohort, can affect cfDNA integrity. The observed alteration in fragmentation profiles highlights the need for cold-chain transportation protocols to minimize variability and maintain cfDNA quality. Future studies should also validate this approach in larger, clinically diverse populations and assess its integration into routine preanalytical workflows.

In summary, residual plasma from heparin separators can serve as a viable source for cfDNA analysis, with stable results for copy number variation, viral load, and methylation profiling despite fragmentomic sensitivity to preanalytical variability. Repurposing residual plasma from routine blood tests could greatly simplify and expand cfDNA accessibility across a diverse range of patients.

Methods

Study design and subject selection

Healthy Cohort: Five healthy adult volunteers (28–40 years old) donated blood with informed consent as approved under Stanford IRB (71230). Whole blood was collected from each donor via a single venipuncture into three tube types: EDTA, Streck, and heparin separators.

Hospital Cohort: Clinical samples were from two institutions, Stanford and UCSF. At Stanford, 28 matched pairs of remnant samples were obtained (collected between 2022 and 2025) under a no-patient-contact protocol (IRB 58461). Patients gave written consent for the use of residual clinical samples. Patients positive for Epstein-Barr virus (EBV), human herpesvirus 6 (HHV6), BK virus, adenovirus, or Torque Teno Virus were identified based on routine clinical viral PCR testing performed by the Stanford or UCSF Clinical Virology Laboratories. These viral tests were typically conducted in the context of post-transplant surveillance or infectious disease diagnostics. Residual EDTA plasma samples were processed by the clinical laboratory within four hours of collection. Matched heparin separator plasma specimens drawn at the same time were retrieved from the Stanford Clinical Chemistry Laboratory. Only matched plasma pairs with volumes exceeding 0.5 mL were included in the study. Additional sample pairs from UCSF (n = 6) were obtained (collected between 2017 and 2020) under a similar no-patient-contact IRB protocol (CA-0161925).

Sample collection and processing

Hospital Cohort: Upon arrival, plasma samples in EDTA tubes and heparin separators were immediately centrifuged at 1300 g for 10 minutes using a Cobas 8100 automated workflow. Following completion of routine clinical testing, the residual plasma is stored and transported at approximately 4°C for up to 8 days before further processing. EDTA plasma was separated into a different container, and heparin separator plasma was kept in the same original collection tube, but physically isolated above the gel separator. These procedures were conducted by the hospital staff as part of routine clinical care in accordance with a CLIA-compliant standard operating procedure. The plasma supernatant was identified by a research team, deidentified, transferred to our research lab at refrigeration, and subjected to a second centrifugation at 16,000 × g for 10 minutes. Plasma aliquots from both EDTA tubes and heparin separators were stored at –80°C until downstream analysis. UCSF cases were processed as previously described(18).

DNA extraction

Cell-free DNA (cfDNA) was extracted from 0.5 to 1 mL of plasma using the Maxwell RSC 48 instrument (Promega) with the Maxwell RSC ccfDNA Plasma Kit (AS1480), following the manufacturer’s instructions. Samples were processed in parallel in batches. cfDNA concentrations were quantified using the Varioskan spectrophotometer (Thermo Fisher Scientific). UCSF cases were extracted as previously described(18).

Whole genome sequencing DNA library preparation

A median of 18.65 ng of extracted cfDNA (range: 3.00-51.70 ng) and 2.5 ng of fragmented Lambda DNA (NEB #N3013) as internal control was used for whole-genome sequencing (WGS) library preparation with the NEBNext Ultra II DNA Library Prep Kit (New England Biolabs, #E7645L), following the manufacturer’s protocol. PCR primers were NEBNext Multiplex Oligos for Illumina (NEB #E6444L). Adapter-ligated DNA was amplified for 20 cycles using the Q5 master mix, following the manufacturer’s instructions. UCSF cases were processed as previously described(18).

FLEXseq DNA library preparation

FLEXseq DNA libraries were prepared as previously described. Briefly, extracted cfDNA molecules were adapter-ligated, dephosphorylated, digested with MspI (NEB, #R0106T), adapter-ligated a second time with custom adapters, subjected to NEBNext Enzymatic Methyl-seq Conversion Module (New England Biolabs, #E7125L), and PCR amplified.

Sequencing and alignment

Paired-end sequencing was performed on the NovaSeq X platform (Illumina) at the University of Chicago Genomics Core Facility using the 25B Reagent Kit (300 cycles). For WGS data, sequencing adapters were trimmed using Cutadapt v3.5 with the following parameters: -a AGATCGGAAGAGC -A AGATCGGAAGAGC --minimum-length 25. Trimmed reads were aligned to the hg38 human reference genome, as well as EBV, HHV6, and other viral genomes using BWA v0.7.17. PCR duplicates were removed using Samtools.

FLEXseq data were processed following the protocol described by Yu et al.(17). Briefly, reads were trimmed to remove adapters and filtered based on quality scores. High-quality reads were aligned to the hg38 genome using Bismark v0.23.0. Unconverted and duplicated reads were removed using Bismark. All SNP-containing regions were masked for de-identification using Bedtools.

Viral load analysis

The ratio of reads aligned to viral genomes (e.g., EBV, HHV6) relative to every one million reads aligned to the human genome (hg38) was calculated and log-transformed. Pearson correlation analysis was performed between matched tube types, with the linear regression constrained to pass through the origin (0, 0).

CNA analysis

Copy number alterations were analyzed using ichorCNA on autosomal chromosomes(22). HMMcopy’s readCounter was used to generate 500 kb bin-level Wig files from aligned BAM files, which included only reads with a mapping quality score >20. Reference files for GC content (gcWig), mappability score (mapWig), and centromere positions were obtained from the ichorCNA GitHub repository corresponding to the selected bin size. Panel-of-normal (PoN) Wig files were generated separately using the cases in the Healthy Cohort in the EDTA and heparin separator groups, respectively.

Tumor purity was estimated based on log2 ratio outputs from ichorCNA using previously reported calculations(23).

Methylation correlation analysis

Methylation calls were extracted using the bismark_methylation_extractor. For each CpG site, the methylation level was calculated as the ratio of methylated reads to total coverage depth. CpG sites with coverage ≥30× were retained for correlation analysis between matched samples.

Tissue-of-origin analysis

Tissue-of-origin deconvolution was performed using cell type-specific unmethylated markers as previously described by Loyfer et al.(19). The top 250 unmethylated CpG markers for each cell type were extracted from the reference dataset. Markers located on sex chromosomes or overlapping with common single nucleotide polymorphisms (SNPs) were excluded. Deconvolution was conducted using the uxm function based on a panel of nine reference cell types: B cells, T cells, natural killer (NK) cells, granulocytes, monocytes/macrophages, endothelial cells, erythroid progenitors, megakaryocytes, and hepatocytes.

cfDNA fragmentomics

Fragment size distributions were generated for each sample by calculating the frequency across fragment lengths between 50 and 600 bp. End motif frequencies were also calculated for each sample. The Shapiro-Wilk test was performed separately for each motif in different tubes (EDTA, Heparin separator, and Streck) to assess whether the frequency distributions followed a normal distribution. For the Healthy Cohort, differences among the three collection tubes were evaluated using the Friedman test. For the Hospital Cohort, statistical significance between tube types was assessed using the Paired Student’s t-test. To control the false discovery rate (FDR), multiple testing corrections were applied using the Benjamini-Hochberg procedure across all tested motifs. Significance was determined based on FDR-adjusted P-values.

Data Availability

Raw WGS and genome-wide enriched methylation sequencing data are available for the Healthy Cohort (n = 5) with informed consent. FASTQ files were uploaded to the Sequence Read Archive (SRA), with the Bioproject ID of PRJNA1260066. Viral reads detailed statistic files (n=28), methylation calls (n=17), copy number plots (n=6), and fragment length files (n=34) are available on Zenodo (https://doi.org/10.5281/zenodo.15367094).

Acknowledgements

This work was funded by an NIH K08 (CA230156) grant, a Burroughs-Wellcome CAMS Award to WG. We thank the members of the Stanford Clinical Molecular Genetic Pathology, Clinical Virology, and Clinical Chemistry Laboratories, and UCSF Clinical Virology and Clinical Chemistry Laboratories for saving residual specimens. We thank Debbie Chan and Dianna Ng for their critical insights at the inception of this project and Ruben Luo for facilitating Clinical Chemistry collections at Stanford. We thank the University of Chicago Genomics Facility (RRID:SCR_019196) for Novaseq sequencing.

Additional files

Supplementary table

Supplementary Figures

Supplementary Figure 1.

Supplementary Figure 2.

Supplementary Figure 3.

Supplementary Figure 4.

Supplementary Figure 5.