An optimized toolkit for high-molecular-weight DNA extraction and ultra-long-read nanopore sequencing using glass beads and hexamminecobalt(III) chloride

Previously developed protocols

Both the standard and UL sequencing workflows consist of three main steps: DNA extraction, library preparation, and library cleanup prior to flow cell loading (Fig. 1A). UL sequencing necessitates minimal DNA shearing throughout the library preparation process and during the extraction methods to maintain UHMW DNA. Early experiments pioneering UL sequencing used the transposase-based rapid kit (ONT, e.g., RAD004) to obtain reads with an N50 > 100 kb (Jain et al. 2018a). This method eliminated the need for DNA shearing before library preparation. However, the yield from rapid runs was typically low, which might be a consequence of the protocol itself. The lack of purification or cleanup of rapid libraries prior to flow cell loading likely contributed to reduced yields compared to ligation-based sequencing approaches. To better understand the complex interplay of DNA quality, library preparation, read length, and yield, we tested a range of UL protocols aimed at maximizing both read length and yield (Fig. 1B).

Figure 1.

Summary of the FindingNemo toolkit. (A) Similarities and differences between the standard and UL protocols that center on the way UHMW DNA is handled. Created with BioRender (https://www.biorender.com). (B) The route of FindingNemo toolkit consists of five main steps, with options at the extraction, library preparation, and cleanup steps. Options with asterisks (*) are our current laboratory workflow for the UL protocol.

Open in new tabDownload PowerPointLink to figure

The goal in obtaining UL reads with the rapid kit is to ensure that each DNA molecule is cut only once by a transposase complex, maximizing read length. Initially, Quick (2018) achieved this by saturating the transposase reaction with UHMW DNA (Table 1A). An alternative approach taken by Logsdon et al. (2021) was to reduce the amount of transposase for a fixed amount of DNA (Table 1B). However, the sequencing output of this method, when applied to the CHM13 cell line (Logsdon et al. 2021), was lower overall compared to the publicly available nanopore data for GM24385 and GM24631 ligation-based libraries (Supplemental Fig. S1A). We hypothesized that this was most likely due to suboptimal pore occupancy (i.e., the number of available pores sequencing at any time), resulting from the reduced number of adapted DNA ends available for sequencing because of the lower amount of transposase. In testing the Quick protocol, we observed two different read length distributions (see Supplemental Fig. S1B,C). While both had N50s of ∼90 kb, one was dominated by shorter reads. This suggests that the DNA was not cut uniformly. Moreover, although these protocols generated UL reads, they suffered from high variability in output N50, occupancy, and yield (Supplemental Fig. S1B,C). Typically, as N50 increases, yield and occupancy drop (Supplemental Table S1A,B).

The Rocky Mountain protocol (Table 1C; Tyson 2019) was developed to use the ligation kit (ONT, e.g., LSK109) to address the yield issues experienced with the rapid kit. This protocol utilized varying concentrations of salts and polyethylene glycol (PEG) to precipitate longer reads after the library preparation step. The ligation kit requires light shearing of the input DNA, resulting in non-UL N50s albeit with much improved yields (Tyson 2019). Therefore, this protocol was further modified to use the rapid kit followed by precipitation of the fragmented DNA using a filter paper disc or star as a form of SPRI matrix (Krazy StarFish [KSF]) (Table 1D). The KSF protocol worked consistently in producing UL N50s, but occupancies and yields were still suboptimal, perhaps a consequence of having DNA cleanup prior to sequencing adapter ligation, consequently producing many free adapters (Methods; Supplemental Fig. S1D,E; Supplemental Table S1C,D).

New and developing protocols

ONT released a rapid kit for UL sequencing, SQK-ULK001 (hereafter ULK001), in which the transposase reaction is performed in a large volume (Table 1E), keeping the DNA concentration at ∼50 ng/µL. We hypothesized that larger reaction volumes facilitated even diffusion and mixing of DNA and transposase complexes in a more homogeneous state in solution, a feature that we included in our Nemo protocols (Table 1F,G). Efficient transposase reactions result in consistent UL sequencing output as shown by the high rates of occupancy (Fig. 2; Supplemental Figs. S1F–H, S2; Supplemental Table S1). Occupancy as a parameter is specifically defined by the percentage of sequencing pores (i.e., pores in “adapter” and “strand” states) compared to all pores in available states (i.e., “adapter,” “strand,” and “pore” states).

Figure 2.

Validation of the FindingNemo protocol. Sequencing outputs between rapid kit-based library preparation protocols (ULK001-Nemo and RAD004-Nemo protocol) show comparable performance when compared to the control ULK001 protocol. (A) Workflow of the library preparation protocols, (B) total yields subcategorized by read lengths, (C) time lapse of yields, (D) read length distributions (dashed black vertical lines denote N50s), and (E) time lapse of relative pore occupancies. Each library was loaded three times (i.e., total DNA equivalent to 3 million GM12878 cells) on a MinION R.9.4.1 flow cell following a nuclease flush protocol. All data are after 65 h of sequencing on the GridION platform.

Open in new tabDownload PowerPointLink to figure

We next tested the performance of RAD004 and ULK001 rapid kit-based protocols, modifying parameters such as the number of input cells, the RAD004 dilution buffer, and the cleanup steps while maintaining the large volume ratio, and compared these to the control ULK001 protocol (Fig. 2A). Regardless of cleanup method (Nanobind disc or glass beads), ULK001-based protocol outputs were comparable in terms of N50s and occupancies (Fig. 2B–E). Meanwhile, the output of the library prepared with the RAD004 protocol showed slightly lower performance in general (Fig. 2B–E; Supplemental Table S1G).

The total yield of an UL sequencing is not a direct measure of the library quality as yield is also flow cell dependent. The control ULK001 library produced the highest total yield (Fig. 2B,C) likely because of the highest pore count of the flow cell (Supplemental Fig. S3A; Supplemental Table S1E). Normalizing yields between flow cells by taking a ratio of yield to the number of pores utilized during a period of data collection (i.e., yield per pore) allows comparison of run performance. When normalized, the yields per pore of the two ULK001 libraries were comparable (Supplemental Table S1E,F). In all three libraries, the distribution of read lengths against read quality was similar (Supplemental Fig. S3B), as well as sequencing speed (Supplemental Fig. S3C). These similarities imply that either of the rapid kit-based protocols could be used to successfully produce UL sequencing outputs. Considering all sequencing output parameters, the ULK001 was the optimal protocol for UL sequencing compared to the RAD004 (Supplemental Table S2).

DNA precipitation and recovery: maintaining quality and homogeneity

Prior to the launch of the ULK001 kit, we developed a protocol to clean the library after adapter ligation using DNA precipitation chemistry compatible with the motor protein–DNA complex. The filter paper from the KSF protocol was replaced with glass beads, adapting the glass bead's effectiveness to act as a DNA-binding substrate as in the Monarch extraction kit protocol (NEB, T3050). We also surveyed compounds with well-studied DNA precipitating (Deng and Bloomfield 1999; Ouameur and Tajmir-Riahi 2004) and selected spermine and CoHex for testing DNA precipitation and recovery profiles using glass beads (Supplemental Fig. S4A). Optimum final concentrations of the compounds were tested empirically based on previous studies (Pelta et al. 1996).

The ULK001 protocol is carried out in a dilute condition of ∼50 ng/µL UHMW DNA. We tested the impact of DNA concentration (10, 20, and 50 ng/µL) on precipitation and recovery rate. CoHex is the best precipitant for both recovery and homogenization of UHMW DNA at a concentration range of 20 ng/µL to <50 ng/µL (Supplemental Fig. S4B–D). The optimal DNA recovery of ∼75% was obtained using CoHex at a DNA concentration of 20 ng/µL. Spermine shows a significantly higher recovery rate at 50 ng/µL DNA but show greater heterogeneity and sample-to-sample variation (Supplemental Fig. S4B–D).

The early ULK001 kit version used a spermine-based precipitation buffer (PPT buffer). We tested this alongside our FindingNemo and spermine cleanup protocols, using a DNA input of 40 ng/µL, and split the adapted library into three cleanup preparations. These runs yielded similar UL sequencing metrics (Fig. 3A). Albeit with a slightly longer N50, ONT's PPT protocol showed the lowest yield per pore, presumably from the lack of any washing step after the library precipitation that might cause the library to have a higher rate of blocking than the others (Fig. 3B).

Figure 3.

Choosing the best precipitating agent that ensures fast homogenization. (A) Read length distribution of libraries precipitated with different buffers; dashed black vertical lines denote N50s. (B) Sequencing metrics after 6 h of run (excluding the first 10 min). All libraries were extracted from GM12878 cells using Monarch direct lysis protocol with an input DNA concentration of ∼40 ng/µL split into three cleanup preparations, loaded on MinION R.9.4.1 flow cells and sequenced on GridION platform.

Open in new tabDownload PowerPointLink to figure

Homogeneity is measured by calculating the coefficient of variation percentage (%CV) of the DNA concentration. We take at least three concentration measurements from the top, middle, and bottom parts of the solution using a NanoDrop spectrophotometer device. The smaller the %CV, the more homogeneous and vice versa. Recovered DNA became more homogeneous after being in solution for a longer period, except when precipitated using CoHex at 10 ng/µL DNA concentration (Supplemental Fig. S4E). Previous studies showed that a charge neutralization of 88%–90% is required for DNA condensation (Deng and Bloomfield 1999). We hypothesize that at 10 ng/µL DNA, excess CoHex cations affected the water hydration at the DNA–cations interface (Deng and Bloomfield 1999) resulting in a more condensed, difficult-to-dissolve DNA as observed during pipetting. Moreover, NanoDrop is more sensitive when measuring absorbance of the differentially condensed DNA in the solution, which explains the larger variation in DNA concentration compared to Qubit measurements (Supplemental Fig. S4C,D). In contrast, spermine-precipitated DNA shows the least homogeneity in solution (Supplemental Fig. S4E). Spermine is also less stable than CoHex; as a solution, it is readily oxidized at room temperature (Sigma-Aldrich). CoHex was chosen to precipitate DNA at a concentration range of 20–40 ng/µL UHMW DNA.

The importance of DNA homogeneity and optimum concentration during UL library preparation was shown in the use of a concentrated and nonhomogeneous UHMW DNA in sequencing, i.e., more than 60 ng/µL DNA in reaction (Supplemental Fig. S5). The sequencing output of this library typifies the output of an undercut library where N50 reaches above 100 kb, but accumulated yield and occupancy are low, as well as a rapid decrease of sequencing pore number (Supplemental Fig. S5D). In essence, ensuring the homogeneity of extracted UHMW DNA and setting the reaction concentration below 50 ng/µL provide the optimum conditions for UL sequencing. Longer elution of the library (e.g., overnight) at room temperature with gentle rotation may also help untangle long DNA molecules, facilitating further homogenization and eventually improving UL sequencing output.

Quantification of UHMW DNA

One of the complications in UL sequencing is in accurately quantitating the UHMW DNA. There is often large variation between the two most used quantification methods: the spectrophotometric- and fluorometric-based methods, represented by NanoDrop and Qubit 3, respectively (Supplemental Fig. S4F). The viscous nature of UHMW DNA also makes it difficult to withdraw a homogeneous small volume to be measured, even when using a positive displacement pipette. Therefore, we followed a novel Qubit concentration measurement of UHMW DNA to complement the standard NanoDrop method. This method uses Jurkat genomic DNA as the baseline standard for UHMW DNA and a glass bead to homogenize samples before each measurement (‘Giron’ Koetsier and Cantor 2021). We modified it by combining small volumes from three to four different parts of the DNA solution to average out concentration differences within the solution (Cahyani et al. 2021). An accurate concentration measurement requires an ideal ratio of NanoDrop to Qubit DNA values between 1 and 1.5 (Simbolo et al. 2013).

As expected, sheared DNA homogenized faster and better than UHMW DNA as shown by the smaller %CV of the sample a few hours after elution (Supplemental Fig. S6A). This fits with the fact that shorter DNA molecules will diffuse faster than longer ones. Longer DNA molecules may also have more complex structure and behavior in solution (Robertson and Smith 2007). The %CV of both types of samples were all below 50% indicating that the FindingNemo method could homogenize DNA solution relatively quickly (Supplemental Fig. S6A). It also filtered out contaminants in the input DNA as shown by the increase of the NanoDrop absorbance values after purification (Supplemental Fig. S6B; Koetsier and Cantor 2019). Last, the FindingNemo protocol did not interfere with the DNA quality as shown by their size distribution on the pulsed-field electrophoresis gel before and after DNA cleanup (Supplemental Fig. S6C).

The FindingNemo protocol versus undercut UL libraries

Combining the FindingNemo approach with either the Quick's or KSF protocols improved sequencing yield but did not significantly improve occupancy or N50 (Supplemental Table S3). We hypothesize that the transposase reaction upstream of the library cleanup is a more critical step in producing optimal UL sequencing outputs. Suboptimal transposase fragmentation may result in long DNA molecules without adapters in the library, because of an insufficient number of transposome complexes compared to DNA molecules, a situation we refer to as an “undercut.”

An undercut library may also occur if the transposome complex-to-DNA ratio is balanced, but the DNA solution is not homogeneous. Viscous, nonhomogeneous UHMW DNA is difficult to dilute and can have varying DNA densities at different points in the solution. Consequently, the cutting frequency of the transposome complex will be higher where DNA is more accessible and will be lower where long DNA molecules remain in a compacted, less-accessible mass. This scenario counterintuitively results in a library with a short N50 despite the HMW nature of the input DNA (Supplemental Fig. S8—0 rpm sample).

It is also important to consider the interaction between UHMW DNA molecules and the sequencing pores, especially just before a mux scan occurs. During this period, only pores with actively threading (sequencing) DNA molecules will continue sequencing; pores without DNA are inactivated. This process terminates at a fixed time, which is 10 min by default. After this period, all pores are inactivated. Molecules still traversing a pore at that point will be ejected, but UHMW DNA increases the risk of stalling or blocking nanopores. At a sequencing speed of 400 bases per sec, any molecule longer than 240 kb is at risk of blocking a pore and negatively impacting yield. Therefore, in the FindingNemo protocol, we modified this script parameter (wind_down) from 600 to 1800 sec (Cahyani et al. 2024).

Extractions without kits

We next compared the effect of DNA extraction methods on UL sequencing. The original approaches by Quick and Logsdon et al. utilized phenol–chloroform isolation and alcohol-based precipitation with some differences (Table 2A,B). We succeeded in scaling down these protocols in combination with glass beads to use less cell input and shorten hands-on time to at least half (Table 2C). To reduce the toxicity from phenol use, we also developed phenol-free glass beads–based extraction protocols utilizing either SDS, CTAB, or CTAB with CoHex as the lysis buffer (Table 2D,E). These extraction protocols produced (U)HMW DNA, characterized by the distinctly narrow vertical lines/smears and DNA in the wells of the pulsed-field gel electrophoresis results (Supplemental Fig. S7).

We also successfully produced UL runs using these DNA samples with varying degrees of occupancies and yields (Fig. 4A–J). Only the extraction protocol using a combination of CTAB and CoHex in the lysis buffer could not produce UL N50 (Fig. 4F). The inclusion of CoHex during lysis was to test whether stable DNA toroid aggregates could be induced (Deng and Bloomfield 1999; Ouameur and Tajmir-Riahi 2004), so that DNA length could be kept intact during extraction and eventually improve sequencing output. The CoHex precipitated DNA was then washed with ethanol, and it rendered DNA elution from the beads more difficult. Ethanol is required in the wash buffer as it rinses salt, contaminants, and other impurities from the cell extract while simultaneously affecting DNA condensation (Arscott et al. 1995; Oda et al. 2016). More optimization is needed to obtain a consistent rate of elution and recovery in the CTAB–CoHex lysis protocol. Nevertheless, the option of using one of two oppositely charged surfactants as lysis agents, i.e., cationic CTAB (Arseneau et al. 2017) and anionic SDS (Xia et al. 2019), provides flexibility in extracting UHMW DNA from diverse sample types and context.

Figure 4.

Combinatorial effects of extraction and library cleanup methods, either with or without the use of kits, on the UL sequencing outputs. (A–I) Read length distributions of the libraries: Graph titles denote the used extraction kit/method and the cleanup protocol; dashed black vertical lines denote N50s. (J) Sequencing metrics after 6 h of run (excluding the first 10 min). DNA was extracted from GM12878 cells. Each library was loaded on a MinION R9.4.1 flow cell and run on the GridION platform.

Open in new tabDownload PowerPointLink to figure

Extractions using kits

We tested a range of (U)HMW DNA extraction kits during our method development and found that Nanobind CBB Big DNA (Circulomics) and Monarch (NEB) were best suited for our purposes (Table 2F–G). The Nanobind CBB Big DNA Circulomics kit could be used to extract HMW or UHMW DNA, depending on the chosen protocol. DNA extracted with the UHMW protocol showed higher size distribution on a pulsed-field electrophoresis gel compared to the HMW protocol (Supplemental Fig. S7A vs. S7B). Libraries prepared from UHMW DNA samples extracted using the Circulomics kit produced UL reads and good sequencing metrics (Fig. 4I,J).

The Monarch kit protocol initially involved a two-step DNA extraction: first extracting nuclei from cells and then isolating DNA from the nuclei (Nuclei Prep). It was later refined to a “Direct Lysis” method, which bypassed the “Nuclei Prep” step by lysing cells and nuclei simultaneously to release DNA. This method yielded high-quality UHMW DNA from GM12878 cells, as shown in pulsed-field gel samples (Supplemental Fig. S7A,B). Libraries prepared from these DNA samples indicated that the Direct Lysis approach performed slightly better than the Nuclei Prep method (Supplemental Fig. S8).

We also found that the shaking speed during lysis in the Monarch protocol influenced DNA homogeneity, which affected sequencing N50 and occupancy (Fig. 4G–H; Supplemental Fig. S8; Supplemental Table S4). Lower lysis speeds yielded longer DNA fragments by keeping DNA more intact. However, no shaking during lysis could result in a nonhomogeneous UHMW DNA solution, negatively affecting the transposase reaction during library preparation (Supplemental Fig. S8—0 rpm). This resulted in a library with non-UL read N50, the lowest occupancy, and yield compared to other libraries (Supplemental Fig. S8). The best UL output was achieved by preparing libraries from DNA lysed at a medium speed of 600–900 rpm (Fig. 4G–H; Supplemental Fig. S8; Supplemental Table S4).

Maximum yield is obtained at a read N50 between 90 and 110 kb

On a MinION, yields in excess of 3 Mb per pore after 6 h of sequencing could be obtained at 93% occupancy, reinforcing that high occupancy is essential for high yield (Fig. 4J). However, UL libraries with many short fragments had reduced occupancy as short reads pass quickly through the pores, lowering total yields (Fig. 4B,D,E,J). This underscores the importance of homogeneously cut UHMW DNA in UL library preparation. Additionally, targeting an N50 between 90 and 110 kb is ideal for optimal UL sequencing output, as N50 is inversely proportional to yield (Fig. 4J; Supplemental Table S4)

Scalable library loading amount with the FindingNemo protocol

We also tested the impact of the amount of input DNA and library loaded per flow cell. When the input amount of UHMW DNA was <5 µg, sequencing output parameters were markedly reduced (Supplemental Table S4, Monarch-01 vs. Monarch-05). To produce the optimum N50, occupancy, and yield, DNA equivalent to at least 1 million GM12878 cells (or at least 5 µg) should be used for library preparation, and a minimum of 2 µg of this library was loaded onto a MinION flow cell. At least 2 million cells and 4 µg of library should be used on a PromethION flow cell (Supplemental Table S4, Monarch-01 vs. Monarch-02).

FindingNemo in 1 day: from cells to UL sequencing in 1 working day

The relatively fast homogenization rate of the FindingNemo cleanup method, combined with the fast DNA extraction when using the Monarch kit, can support the preparation of UL libraries from fresh or frozen cells to sequencing in just a working day (Supplemental Fig. S9A; Cahyani 2021). We can obtain UL N50s using this protocol; however, yields are markedly reduced (Supplemental Fig. S9B–D), likely as a result of the heterogeneity of DNA samples. Yield reductions were exacerbated by the low occupancy level and the abundance of short reads and would be interpreted as undercut libraries (Supplemental Fig. S9B–D, Runs 2 and 3). Moreover, the FindingNemo cleanup protocol might not significantly remove these highly abundant short reads because of their high concentration in the libraries.

The FindingNemo protocol improved N50, occupancy, and yield with ULK114 kit

In 2022, ONT released a new pore and chemistry running on the R10.4.1 flow cells and a new UL kit, ULK114. This included new sequencing scripts and a sample rate shift from 4 to 5 kHz. We tested the ULK114 kit according to the manufacturer's instructions and with the FindingNemo protocol on three human cell lines (Fig. 5). Libraries prepared with the FindingNemo protocol showed longer N50s, optimal yield, and occupancy (Fig. 5A,B). This protocol consistently produced higher yields, regardless of kit chemistry and sequencing programs. Additionally, the 5 kHz script increased the overall mapping rate of reads compared to the 4 kHz program (Fig. 5B).

Figure 5.

UL N50s were obtained using the new ULK114 kit from three cell lines. (A) Read length distribution of GM12878 library run with the earlier ULK114 script version (left panel), while the other two libraries (middle and right panels) were run with the most current 5 kHz version of the script. Vertical dashed black lines indicate N50s. (B) Run metrics of the three libraries showed that the Nemo protocol performed better than ONT's.

Open in new tabDownload PowerPointLink to figure

Ratio of transposase to DNA

The ratio of transposase to DNA is crucial for optimizing UL sequencing. The ULK001 protocol specifies 1 µL of transposase fragmentation mix (FRA) per 1 million human cells, corresponding to 5–6 µg of UHMW DNA for a 3.2 Gb genome (Table 3). We used this ratio for all GM12878 (Figs. 2Figure 3.Figure 4.–5) and HEK293 (Fig. 6) runs. Additionally, we tested different FRA to DNA ratios in Mongolian gerbil, tick Amblyomma variegatum, and yeast Saccharomyces cerevisiae (Table 3). These ratios produced UL sequencing outputs for the gerbil and tick samples (Fig. 6A,C). Although the yeast run did not yield UL read N50s (Fig. 6B,C), it showed a significant increase in read length compared to existing nanopore-sequenced data (Giordano et al. 2017; Salazar et al. 2017).

Figure 6.

UL sequencing output and metrics of non-GM12878 cells and different sequencing platforms: gerbil blood, tick A. variegatum, yeast S. cerevisiae, and human HEK293 cells using either the ULK001 or the ULK001-Nemo protocol (as labeled). (A) Read length distributions of libraries run on PromethION R9.4.1 flow cells; dashed black vertical lines denote N50s. (B) Read length distribution of the HEK293 and yeast S. cerevisiae libraries run on MinION R9.4.1 flow cells. (C) Sequencing metrics of the libraries. Each metric was shown after 6 h of run (excluding the first 10 min).

Open in new tabDownload PowerPointLink to figure

Genome ploidy

Genome ploidy anomalies must be considered when preparing UL libraries. For instance, about 4.2% of HEK293 cells are hypotriploid (Synthego HEK293). This means the DNA mass extracted from the same number of cells is higher than from the GM12878 cell line, altering the FRA to DNA ratio (Table 3) and potentially affecting sequencing output by increasing read length and decreasing yield. Sequencing HEK293 DNA with the Monarch kit confirmed this, producing N50s over 190 kb but with lower yields (Fig. 6B,C). To maximize yield and N50, it is better to use the FRA volume to DNA mass ratio rather than the absolute cell number when genome ploidy information is available.

Sample types: to count or to weigh?

The UL protocol requires the same number of human cells per microliter of FRA, whether from a cell line or nucleated blood. However, counting cells in blood samples can be impractical, as seen in the gerbil UL run (Table 3). In such cases, the FRA to DNA mass ratio can be used, provided the DNA is extracted to maintain native chromosome length and properly quantified (Methods). Using this ratio, UL libraries were produced from gerbil blood following the original ULK001 protocol (Fig. 6B,C).

The FindingNemo protocol works on different nanopore platforms

We ran pairwise sequencing of identical libraries on both the MinION and PromethION platforms, showing that PromethION produced more yield per pore (Supplemental Table S5). With nearly six times as many pores as a MinION, PromethION is expected to generate more data. The best PromethION results had an N50 of 126 kb and a yield of 3.85 Mb per pore at 88% occupancy (Fig. 6C—tick cells ULK001-Nemo). On the MinION platform, the best results ranged from 106 to 147 kb N50, with yields of 3.19–3.89 Mb per pore and at least 95% occupancy (Fig. 4J). These metrics were all obtained after 6 h of sequencing.

Relative performance of extraction kits in tick samples

We compared two extraction kits, Circulomics and Monarch, for UL sequencing of a tick cell line with a ∼6 Gb genome (Fig. 6A,C). Although flow cell occupancies were similar, the Monarch-extracted DNA had longer N50s, higher yields, and a greater proportion of UL reads (Fig. 6A right panel, C). The Circulomics library had more short reads, likely because of undercutting (Fig. 6C, middle panel). The Circulomics-extracted DNA concentration was ∼50 ng/µL, while Monarch-extracted DNA concentration was ∼26 ng/µL, reflecting differences in sample homogeneity. Moreover, Monarch extractions yielded DNA with a longer size distribution, as shown by pulsed-field electrophoresis gel (Supplemental Fig. S10). The Monarch protocol is also more flexible, allowing modifications to target different read length distributions by adjusting lysis speed (Methods).

Cell sources

GM12878 cells were grown by Darren Crowley (University of Nottingham) and additionally purchased from Coriell Institute for Medical Research, Camden, NJ, USA. Yeast S. cerevisiae cells were given by Stephen Gray (University of Nottingham). Tick A. variegatum (AVL/CTVM17) cells were obtained from Alistair Darby's group (University of Liverpool). The HG00233 and HG04054 cells were obtained from Danny Miller's laboratory (University of Washington, Seattle, WA, USA). Last, HEK293 cells were obtained from New England Biolabs (NEB) in Ipswich, MA, USA.

Kit-based extraction

Kits used in the extraction of UHMW DNA were Monarch HMW DNA Extraction Kit for Cells & Blood (NEB T3050) and Nanobind CBB Big DNA Kit (Circulomics SKU NB-900-001-01) plus the auxiliary Nanobind UL Library Prep Kit (Circulomics SKU NB-900-601-01) according to the manufacturer's protocols with modifications as previously described for the Monarch protocol (Cahyani et al. 2021). The Monarch nuclei prep approach followed the original protocol (NEB T3050), while the direct lysis approach combined the prep and lysis buffers into one step (NEB Direct Lysis). Lysis speed at 600 up to 900 rpm resulted in an optimum UL sequencing output and a minimum of overnight incubation of DNA in solution helped homogenization.

Phenol–chloroform extraction with glass beads

This is a scaled-down version of Quick's original protocol (Quick 2018), and the modifications are as previously described (Cahyani et al. 2021). In summary, pellet of 5 million cells was washed with phosphate-buffered saline (PBS) (Fisher Scientific 15453819) and lysed with an SDS buffer (0.5% SDS, 100 mM NaCl, 25 mM EDTA pH 8.0, and 10 mM Tris-HCl pH 8.0) in the presence of RNase A (Qiagen 19101, 20 µg/mL) at 37°C for 5 min. Proteinase K (Qiagen 19131, 200 µg/mL) was added and incubated at 56°C for 15 min. Cell lysate was split into two 5PRIME phase-lock gels (Quanta-Bio 2302820). BioUltra TE-saturated phenol (Merck 77607) was added at 1:1 volume ratio. Sample was homogenized by vertical rotation at 20–30 rpm for 10 min and then centrifuged at 4000g for 10 min. The aqueous phase was transferred into new phase-lock gel tubes. DNA purification was repeated with a second round of phenol:chloroform:isoamyl alcohol = 25:24:1 volume ratio. In the third round, only chloroform:isoamyl alcohol = 24:1 was used. After the last centrifugation, the aqueous phase from the two tubes was combined in a 5 or 15 mL tube and added with 5 M ammonium acetate (Sigma-Aldrich A-7330) (0.4 volume), three glass beads (3 mm diameter), and absolute ethanol (2.5 volume). The tube was inverted by hand 20–30 times to precipitate DNA onto the glass beads (or placed in a tube rotator at 10 rpm for 3 min). The supernatant was removed, and bound DNA was washed twice with 60%–70% ethanol. Beads with DNA were poured into a bead retainer and quickly spun to remove excess ethanol (or absorbed with a filter paper or tissue). DNA was quickly eluted from beads by pouring them into a 2 mL LoBind (Eppendorf 0030108051) tube containing 10 mM Tris-HCl pH 9.0 and incubating at 37°C for 30 min with regular wide-bore pipette mixing. Incubation was continued overnight at room temperature. Afterward, using a bead retainer, DNA was spun down at maximum speed for 1 min.

Phenol-free, home-brew extractions

This protocol is adapted from the Monarch HMW DNA Extraction Kit for Cells & Blood as described previously (Cahyani et al. 2021). In summary, a pellet of 1–3 million cells was washed with PBS and lysed in either SDS or CTAB buffer with 200 µg Proteinase K and incubated at 56°C for 15 min (preferably with shaking at 600–700 rpm). The SDS buffer composition was the same as used in the phenol-based extraction described in the previous section. The CTAB buffer compositions were 2% CTAB, 1.4 M NaCl, 25 mM EDTA pH 8.0, and 10 mM Tris-HCl pH 8.0. After lysis, 100 µg RNase A was added and incubated at 37°C for 5 min. To precipitate DNA, 5 M ammonium acetate (0.4 volume), two to three glass beads (depending on cell number used), and isopropanol (0.9 volume) were added. Sample was mixed on a rotator at 9 rpm for 5 min (or inverted by hand 20–30 times). Liquid was removed by pipetting, and bound DNA was washed twice with 60%–70% ethanol. Beads with DNA were poured into a bead retainer and quickly spun to remove excess ethanol (or absorbed with a filter paper or tissue). DNA was quickly eluted with 10 mM Tris-HCl pH 8.0 in a 2 mL tube and incubated at 37°C for 30 min with regular wide-bore pipette mixing. Incubation was continued overnight at room temperature. Afterward, using a bead retainer, DNA was spun down at a maximum speed for 1 min.

CTAB–CoHex extraction

This protocol is similar to the home-brew CTAB method as described above with the following modifications. The CTAB–CoHex buffer compositions used were 2% CTAB, 20 mM CoHex, 1.4 M NaCl, 25 mM EDTA pH 8.0, and 10 mM Tris-HCl pH 8.0. To precipitate DNA, sample was diluted with 10 mM Tris-HCl pH 8.0, so CoHex concentration became ∼5 mM and NaCl was ∼400 mM, i.e., the ratio of CTAB–CoHex lysis buffer to Tris-HCl was 1:2.5. Then, —two to three glass beads (depending on cell number used) were added. Sample was mixed on a rotator at 9 rpm for 5 min (or inverted by hand 20–30 times). Liquid was removed by pipetting, and bound DNA was washed twice with an ethanol buffer (50% ethanol and 50% of 1 mM CoHex). Beads with DNA were poured into a bead retainer and quickly spun to remove excess ethanol (or absorbed with a filter paper or tissue). DNA was quickly eluted with 10 mM Tris-HCl pH 8.0 in a 2 mL tube and incubated at 37°C for 30 min with regular wide-bore pipette mixing. Incubation was continued overnight at room temperature. Afterward, using a bead retainer, DNA was spun down at a maximum speed for 1 min.

Yeast extraction using CTAB and glass beads

Yeast pellet (∼50 µL volume or equivalent to 200 million cells) was washed twice in cold PBS and resuspended in 480 µL SpheroBuffer (1 M sorbitol, 100 mM Na₂HPO₄, and 100 mM EDTA) (Cahyani et al. 2024). Lyticase enzyme (5 U/µL) was added (20 μL). Sample was incubated at 30°C for 30 min, and spheroplast formation was checked from time to time. Sample was lysed by adding 480 µL of 2× CTAB lysis buffer (4% CTAB, 2 M NaCl, 50 mM EDTA pH 8.0, and 20 mM Tris-HCl pH 9.0), mixed well with wide-bore pipet tip as the buffer is viscous, added with 400 µg Proteinase K (20 mg/mL), and incubated at 56°C for 15 min while shaking at 700 rpm. To remove RNA, 400 µg of RNase A (100 mg/mL) was added followed by incubation at 56°C for 10 min without shaking. Sample was cooled off at room temperature and added with 400 µL of 5 M ammonium acetate and 100 µL of 5 M NaCl (Thermo Fisher Scientific AM9759) and mixed by a vertical rotator for 1 min at 9 rpm. Sample was then centrifuged for 7 min at 16,000g, and the lysate supernatant was removed to a new 2 mL tube. Three glass beads and 0.5 mL isopropanol were added, and the sample was mixed on a vertical rotating mixer at 9 rpm for 5 min. Between 0.5 and 0.7 mL liquid was removed and replaced with the same amount of isopropanol and was rotated for an additional 3 min. When DNA had bound tightly around the beads, the liquid was discarded, and bound DNA was washed twice with 1 mL of 70% ethanol. Beads with DNA were poured into a bead retainer and quickly spun to remove excess ethanol (or absorbed with a filter paper or tissue). DNA was quickly eluted with elution buffer (EB) (ONT) or 10 mM Tris-HCl pH 9.0 in a 2 mL tube and incubated at 37°C for 30 min with regular wide-bore pipette mixing. Incubation was continued overnight at room temperature. Afterward, using a bead retainer, DNA was separated from the beads by centrifugation at a maximum speed for 1 min.

UL kit Nemo

The ONT UL kits, i.e., ULK001 and ULK114, are based on the cleaving and tagging of the DNA by the transposome complexes to insert transposase adapters, which then are ligated to the sequencing adapters (ONT ULK001). Our modifications of the ULK001 method were described in previously published protocol (Cahyani et al. 2021). To summarize, DNA extracted from 1 to 6 million cells was diluted to a final concentration of 20–40 ng/µL (in the final reaction volume). All mixing was done with wide-bore pipette tips. Meanwhile on ice, the FRA dilution buffer (FDB) was well mixed with FRA (1 µL per 1 million human cells DNA) and then combined to the DNA solution and quickly incubated at 23°C for 5–10 min (depending on the reaction volume). Enzyme was inactivated at 70°C for 5 min and cooled down to room temperature. Rapid adapter-F (RAP-F) was added (0.83 µL per 1 µL FRA used), and the sample was incubated at room temperature for at least 30 min.

The ULK114 modifications are similar to ULK001 and are described in the newly published and updated protocol also on protocols.io (Cahyani et al. 2024). In summary, library preparation to run on a PromethION flow cell needs extracted DNA from 6 million cells that was diluted to a final concentration of 20–40 ng/µL (in the final reaction volume). On ice, the dilution buffer (FDB) was well mixed with FRA (6 µL per 6 million human cells DNA) and then combined to the DNA solution and quickly incubated at 23°C for 10 min with 9 rpm rotation. Enzyme was inactivated at 75°C for 10 min and cooled down to room temperature. Sequencing adapter (RA) was added (5 µL per 6 µL FRA), and the sample was rotated at 9 rpm for 30 min.

RAD004-Nemo

The RAD004-Nemo protocol followed the same route as the ULK001-Nemo protocol described previously (Cahyani et al. 2021). However, the transposase and adapter used were the ones from the RAD004 kit (i.e., FRA and RAP, respectively). The dilution buffer used to replace FDB was the 4× MuA buffer (100 mM Tris-HCl pH 8.0, 40 mM MgCl_2, 440 mM NaCl, 0.2% Triton X-100, and 40% glycerol).

Krazy StarFish

The KSF protocol is an alternative UL sequencing that uses a filter paper instead of glass beads. It is a cheap, quick-and-dirty method to produce UL sequencing based on the RAD004 kit. This method was described previously (Cahyani et al. 2021) and summarized as follows. An amount of ∼7.5 µg DNA in 75 µL volume was mixed well with 25 µL 4× MuA buffer. Meanwhile in another tube on ice, 25 µL 4× MuA buffer was diluted with 74 µL water and added with 1 µL FRA (or up to 1.2 µL FRA if DNA was quite viscous). These two solutions were mixed well on ice, aliquoted into two PCR tubes, and treated at 30°C for 1 min, 80°C for 1 min, and then cooled to room temperature. Reactions were pooled into a single tube, and 12 µL 5 M NaCl was added and mixed gently. A KSF filter (made from Whatman filter paper no. 3 with a star hole puncher) was submerged in the solution, and 142 µL isopropanol added and mixed by inversion of 20–30 times allowing UHMW DNA to collect and condense onto the filter. Liquid was removed, and filter paper was washed twice with 60%–70% ethanol. Excess ethanol was removed by pipetting, and paper was air-dried for 2 min. After the paper was moved to a clean tube, the library was eluted by adding 120 µL of elution buffer (10 mM Tris-HCl pH 8.0). Before sequencing, 37.5 µL library was added with 0.5 µL RAP and incubated for at least 30 min at room temperature.