The automated isolation of ccfDNA is crucial to applications within large-scale projects. Automated extraction of ccfDNA could help to minimize human error, enable high-throughput downstream assays and implement standardized procedures for diagnostic tests. Both the EZ2 (QIAGEN) and the Maxwell (Promega) instruments reduce hands-on time for ccfDNA isolation compared to manual kits^29,51. For the QIAamp MinElute ccfDNA kit, which is a manual-only protocol, a total experiment time of 50 min was reported for the processing of 16 samples²⁹. Contrastingly, the hands-on times of the Maxwell kit and the EZ2 kit were 10 min and 30 min respectively. Notwithstanding the relatively long hands-on time observed with the EZ2 kit, the total runtime was lower (66 min) than that of the Maxwell instrument (80 min). The reason for the EZ2 instrument’s hands-on time being 20 min longer than that of the Maxwell relates to the necessity of handling multiple tubes, each of which requires placing in the appropriate positions in the EZ2 tip rack and unscrewing separately by hand. However, the EZ2 kit and EZ2 instrument were a field test set; the equipment is still in the development process. It may therefore be the case that changes in the setting will be taken into account during finalization of the instrument for commercial use. Unlike the Maxwell instrument, the EZ2 instrument offers the option of processing samples in the tubes they were stored in. However, these tubes need to match the instrument and to fit into the EZ2 rack. Not all biobanks may have tubes in use that match this system; 2D matrix tubes, for example, are in common use, but are not compatible with the EZ2 system. Alongside a rapid and reproducible extraction process, sufficient ccfDNA output is an important factor in enabling in-depth analysis of phenomena such as alterations within ctDNA. The low copy numbers of mutant alleles behind an extensive background of wild-type DNA present a technical challenge to downstream analytical platforms⁵². When mutation type and position are known, it is possible to perform very sensitive methods such as droplet digital PCR (ddPCR); otherwise, the screening of whole DNA fragments requires next-generation sequencing (NGS) methods⁵². For such downstream applications, a ccfDNA input of between 1 and 25 ng is necessary⁵³. The EZ2 kit achieved a median output of 8.6 ng from 1 ml plasma used, while the Maxwell kit yielded approximately 4.6 ng. Both of these yields are within the range of necessary input; however, for in-depth analysis using, for example, NGS methods, input of more than 1 ml of plasma would be necessary, particularly when using the Maxwell kit. qPCR-based quantification returned a median ccfDNA yield of 17.0 ng/ml plasma for EZ2 kit isolates and 13.0 ng/ml plasma for Maxwell eluates. These ccfDNA plasma concentrations were in the lower range of ccfDNA concentrations reported by others for early breast cancer (12–52 ng/ml)^54,55,56. In these studies, qPCR-based quantification was a frequent method of choice⁵⁷. More than 50% of the cohort of patients included in our comparison of the EZ2 and Maxwell kits had a tumor stage of T1; an early tumor stage is generally associated with lower levels of ccfDNA. Further, 83% of the patients were lymph node-negative (Table 1). Research has found that the concentration of ccfDNA in plasma correlates with the size of the tumor and that lymph node-positive patients show higher levels of ccfDNA^9,55. Several research groups have also reported comparably low ccfDNA plasma levels^22,56,58. In a non-triple negative cohort consisting mainly of stage II breast cancer patients, the median ccfDNA concentration was 13.22 ng/ml^22,58, which is in line with the yield achieved by EZ2 and Maxwell isolation. Interestingly, we detected lower ccfDNA concentrations when measured with fluorescence-based quantification compared to qPCR quantification. It was shown earlier that fluorescence-based DNA values were dependent on salt concentrations and decrease non-proportionally to the dilution ratio⁵⁹. In addition, fluorescence dyes, such as PicoGreen, specifically bind to double-stranded DNA (dsDNA)⁵⁹. CcfDNA, however, consists of both dsDNA and single-stranded DNA (ssDNA) with a higher proportion of ssDNA⁶⁰. Thus, quantification of ccfDNA using fluorescence dyes—in particular such kits designed for dsDNA (e.g. QuantiFluor dsDNA System kit)—might underestimate the ccfDNA concentration.

The EZ2 and Maxwell kits were both bead-based isolation methods. This technique enables a high standard of automation and is thus attractive for a variety of clinical applications. The high level of automation in bead-based methods does, however, come with the drawback of lower plasma ccfDNA yields than are attainable with manual silica-based methods. There is controversy in the literature in this area, with some researchers arguing that bead-based methods can keep up with silica membrane-based methods, while other publications assert that silica membrane-based methods are superior^20,21,30,46. Nevertheless, research groups using the QIAamp Circulating Nucleic Acid kit (QIAGEN), a silica-based-method, consistently obtained ccfDNA yields higher than those attained via bead-based methods (for a detailed overview of the literature, see Supplemental Table S1)^20,61. We conclude from these findings and our results that the choice of method for ccfDNA extraction (bead-based vs. silica membrane, automated vs. manual) should be based on the anticipated output, the level of throughput required, and the necessity or otherwise of automation. Fully automated, bead-based techniques will be crucial to high-throughput applications in routine clinical settings.

Interestingly, the correlation of the ccfDNA yields from the EZ2 and Maxwell extraction methods differed depending on the quantification platforms used. On all quantification platforms used in our research, the angle of the correlation line was lower than the bisector, confirming generally higher ccfDNA yields from use of the EZ2 instrument. The correlation curve of the qPCR data was closer to the bisector line than the correlation curves of the fluorescence quantification and the electrophoresis data. Therefore, while there may be higher concentrations of ccfDNA in EZ2 eluates, the effect is smaller when looking at qPCR data, a method that measures the amplifiability of DNA templates. A low ratio of long to short fragments can reduce the amplifiability of a template, as the proportion of amplifiable fragments decreases when the fragments are too short⁶². Applied to our data, this could point toward differences in fragment size distribution between the Maxwell and the EZ2 eluates, resulting in differences in amplifiability.

We also compared the two extraction instruments for differences in fragment length distribution within the extracts. This is a well-documented quality control procedure that is crucial to cancer diagnosis and the predicting progression and prediction of prognosis^7,22,63. Prior work has shown that the method of extraction used may affect fragment sizes within the samples^29,44. Quantification of the copy numbers of ALU60 and ALU247 templates revealed differences in the distribution of fragment lengths between the isolates. We found that the numbers of ALU60 copies, representing all fragments longer than 60 bp, were significantly higher for EZ2 kit extracts than for those obtained with the Maxwell kit. Conversely, higher copy numbers of large fragments (ALU247) were extracted with the Maxwell instrument. Looking at the ratio of longer to-total fragment size, we note that ccfDNA integrity of plasma samples isolated by the Maxwell kit was significantly higher than that of EZ2 samples. In summary, it appears that the EZ2 instrument is superior in extracting total ccfDNA and small DNA fragments in particular. Research has found that high yields of short, fragmented DNA are more likely to contain higher percentages of ctDNA than lower yields and can therefore serve as a measure of quality⁶⁴. Further, fragment size distribution is a characteristic relevant to the choice of a downstream application. For example, sequencing methods relying on double-stranded library preparation were shown to be insensitive on ultra-short fragments and should be rejected in favor of single-stranded library preparation to avoid depletion of short DNA molecules⁶⁵.

Our results regarding integrity are in line with other studies showing integrities of between 0.5 and 0.8^7,8,9,18. Notwithstanding differences in target design, most of the groups that conducted these studies determined integrity using an ALU-specific qPCR. However, the field remains without a standardized procedure for sample handling, ccfDNA isolation and determination of integrity⁶⁶. These factors contribute to the high divergences in integrity values observed among different studies.

Differences between the isolation methods evaluated were particularly apparent after quantification of mtDNA. Median hmito copy numbers were 15.9-fold higher in the EZ2 samples than in the Maxwell probes. The literature observes that cell-free mtDNA of the plasma is highly enriched in short sizes (30–60 bp)⁶⁷. This supports the assumption that the EZ2 kit is better able to extract short fragments than the Maxwell kit. While ALU-specific qPCR enables quantification of the major part of nuclear ccfDNA, it does not include the representative amount of circulating mtDNA levels. Many current studies are limited to nuclear ccfDNA quantification^20,30,68. To our knowledge, this is the first report comparing different extraction methods with regard to specific yields of mtDNA. Current work has discussed circulating mtDNA as a non-invasive biomarker in several cancers^69,70. Analysis of hmito copy numbers in ccfDNA enables specific quantification of mtDNA plasma levels^50,71. High mtDNA yield is crucial to downstream analysis such as mutational screening; however, we are yet to completely understand the complex role of changes in mtDNA levels themselves and in mitochondria-specific alterations during cancer evolution and progression. Further research is therefore vital if we are to improve the use of mtDNA as a biomarker. Currently, the major challenge of working with mtDNA is its low concentration in plasma samples, which indicates the importance of techniques that enable the extraction of as much mtDNA as possible⁶⁷.

The present study has some limitations. First, due to the small size of the study cohort, we did not compare values obtained from samples of breast cancer patients with those from healthy individuals. We are therefore unable to draw conclusions regarding the diagnostic potential of ccfDNA extracts. A larger cohort would be required in order to establish whether the method used for ccfDNA isolation impacts the sensitivity and specificity of ccfDNA yield and integrity for the purpose of distinguishing between healthy individuals and patients with breast cancer. Future studies might usefully investigate this question with the aim of providing more information on the diagnostic application of ccfDNA^9,27. Second, the study focused solely on ccfDNA quantity and quality, but not on ctDNA detection. We did not perform any downstream analysis identifying mutant alleles and were therefore unable to estimate ctDNA yield. Further, we used slightly different plasma input conditions for the two isolation platforms; for the Maxwell kit, 1 ml of undiluted plasma was used, while for the EZ2 kit, 1 ml plasma was diluted with 1 ml 1xPBS. This adaptation was necessary and recommended by the manufacturer for the purpose of increasing the input volume to the recommenended 2 ml of plasma. However, such a dilution has the potential to influence the ccfDNA extraction process due to dilution of inhibitory substances present in the plasma sample.

Overall, this study presents novel data on the quantitative and time-related efficiency of two fully automated ccfDNA extraction instruments, EZ2 and Maxwell. Of these, the EZ2 instrument provided better results with respect to ccfDNA and mtDNA yield, while the Maxwell instrument appeared superior in isolating long fragments. We also confirmed that the method of isolation used has a direct effect on ccfDNA integrity, a factor which should be taken into account when comparing studies. This finding further emphasizes the need for standardized procedures and quality controls where ccfDNA analysis is used in a diagnostic or prognostic clinical setting. Supplementary quality control should additionally take place before downstream analysis; this should include evaluation of fragment size distribution, checking for contamination with DNA from white blood cells leading to the presence of clonal hematopoietic variants, and assessment of the purity of extracted ccfDNA (e.g. presence of PCR inhibitors such as heparin)^64,72.